gecko-dev/xpcom/io/nsFastLoadFile.cpp

2577 строки
82 KiB
C++
Исходник Обычный вид История

/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */
/* ***** BEGIN LICENSE BLOCK *****
* Version: MPL 1.1/GPL 2.0/LGPL 2.1
*
* The contents of this file are subject to the Mozilla Public License Version
* 1.1 (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
* http://www.mozilla.org/MPL/
*
* Software distributed under the License is distributed on an "AS IS" basis,
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
* for the specific language governing rights and limitations under the
* License.
*
* The Original Code is Mozilla FastLoad code.
*
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
* The Initial Developer of the Original Code is
* Netscape Communications Corporation.
* Portions created by the Initial Developer are Copyright (C) 2001
* the Initial Developer. All Rights Reserved.
*
* Contributor(s):
* Brendan Eich <brendan@mozilla.org> (original author)
*
* Alternatively, the contents of this file may be used under the terms of
* either of the GNU General Public License Version 2 or later (the "GPL"),
* or the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
* in which case the provisions of the GPL or the LGPL are applicable instead
* of those above. If you wish to allow use of your version of this file only
* under the terms of either the GPL or the LGPL, and not to allow others to
* use your version of this file under the terms of the MPL, indicate your
* decision by deleting the provisions above and replace them with the notice
* and other provisions required by the GPL or the LGPL. If you do not delete
* the provisions above, a recipient may use your version of this file under
* the terms of any one of the MPL, the GPL or the LGPL.
*
* ***** END LICENSE BLOCK ***** */
#include <string.h>
#include "prtypes.h"
#include "nscore.h"
#include "nsDebug.h"
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
#include "nsEnumeratorUtils.h"
#include "nsMemory.h"
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
#include "nsXPIDLString.h"
#include "nsString.h"
#include "nsReadableUtils.h"
#include "nsIComponentManager.h"
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
#include "nsIFile.h"
#include "nsILocalFile.h"
#include "nsISeekableStream.h"
#include "nsISerializable.h"
#include "nsIStreamBufferAccess.h"
#include "nsBinaryStream.h"
#include "nsFastLoadFile.h"
#include "nsInt64.h"
#ifdef DEBUG_brendan
# define METERING
# define DEBUG_MUX
#endif
#ifdef METERING
# define METER(x) x
#else
# define METER(x) /* nothing */
#endif
#ifdef DEBUG_MUX
# include <stdio.h>
# include <stdarg.h>
static void trace_mux(char mode, const char *format, ...)
{
va_list ap;
static FILE *tfp;
if (!tfp) {
char tfn[16];
sprintf(tfn, "/tmp/mux.%ctrace", mode);
tfp = fopen(tfn, "w");
if (!tfp)
return;
setvbuf(tfp, NULL, _IOLBF, 0);
}
va_start(ap, format);
vfprintf(tfp, format, ap);
va_end(ap);
}
# define TRACE_MUX(args) trace_mux args
#else
# define TRACE_MUX(args) /* nothing */
#endif
/*
* Fletcher's 16-bit checksum, using 32-bit two's-complement arithmetic.
*/
#define FOLD_ONES_COMPLEMENT_CARRY(X) ((X) = ((X) & 0xffff) + ((X) >> 16))
#define ONES_COMPLEMENT_ACCUMULATE(X,Y) (X) += (Y); if ((X) & 0x80000000) \
FOLD_ONES_COMPLEMENT_CARRY(X)
#define FLETCHER_ACCUMULATE(A,B,U) ONES_COMPLEMENT_ACCUMULATE(A, U); \
ONES_COMPLEMENT_ACCUMULATE(B, A)
PR_IMPLEMENT(PRUint32)
NS_AccumulateFastLoadChecksum(PRUint32 *aChecksum,
const PRUint8* aBuffer,
PRUint32 aLength,
PRBool aLastBuffer)
{
PRUint32 C = *aChecksum;
PRUint32 A = C & 0xffff;
PRUint32 B = C >> 16;
PRUint16 U = 0;
if (aLength >= 4) {
PRBool odd = PRWord(aBuffer) & 1;
switch (PRWord(aBuffer) & 3) {
case 3:
U = (aBuffer[0] << 8) | aBuffer[1];
FLETCHER_ACCUMULATE(A, B, U);
U = aBuffer[2];
aBuffer += 3;
aLength -= 3;
break;
case 2:
U = (aBuffer[0] << 8) | aBuffer[1];
FLETCHER_ACCUMULATE(A, B, U);
U = 0;
aBuffer += 2;
aLength -= 2;
break;
case 1:
U = *aBuffer++;
aLength--;
break;
}
PRUint32 W;
if (odd) {
while (aLength > 3) {
W = *NS_REINTERPRET_CAST(const PRUint32*, aBuffer);
U <<= 8;
#ifdef IS_BIG_ENDIAN
U |= W >> 24;
FLETCHER_ACCUMULATE(A, B, U);
U = PRUint16(W >> 8);
FLETCHER_ACCUMULATE(A, B, U);
U = W & 0xff;
#else
U |= W & 0xff;
FLETCHER_ACCUMULATE(A, B, U);
U = PRUint16(W >> 8);
U = NS_SWAP16(U);
FLETCHER_ACCUMULATE(A, B, U);
U = W >> 24;
#endif
aBuffer += 4;
aLength -= 4;
}
aBuffer--; // we're odd, we didn't checksum the last byte
aLength++;
} else {
while (aLength > 3) {
W = *NS_REINTERPRET_CAST(const PRUint32*, aBuffer);
#ifdef IS_BIG_ENDIAN
U = W >> 16;
FLETCHER_ACCUMULATE(A, B, U);
U = PRUint16(W);
FLETCHER_ACCUMULATE(A, B, U);
#else
U = NS_SWAP16(W);
FLETCHER_ACCUMULATE(A, B, U);
U = W >> 16;
U = NS_SWAP16(W);
FLETCHER_ACCUMULATE(A, B, U);
#endif
aBuffer += 4;
aLength -= 4;
}
}
}
if (aLastBuffer) {
NS_ASSERTION(aLength <= 4, "aLength botch");
switch (aLength) {
case 4:
U = (aBuffer[0] << 8) | aBuffer[1];
FLETCHER_ACCUMULATE(A, B, U);
U = (aBuffer[2] << 8) | aBuffer[3];
FLETCHER_ACCUMULATE(A, B, U);
break;
case 3:
U = (aBuffer[0] << 8) | aBuffer[1];
FLETCHER_ACCUMULATE(A, B, U);
U = aBuffer[2];
FLETCHER_ACCUMULATE(A, B, U);
break;
case 2:
U = (aBuffer[0] << 8) | aBuffer[1];
FLETCHER_ACCUMULATE(A, B, U);
break;
case 1:
U = aBuffer[0];
FLETCHER_ACCUMULATE(A, B, U);
break;
}
aLength = 0;
}
while (A >> 16)
FOLD_ONES_COMPLEMENT_CARRY(A);
while (B >> 16)
FOLD_ONES_COMPLEMENT_CARRY(B);
*aChecksum = (B << 16) | A;
return aLength;
}
PR_IMPLEMENT(PRUint32)
NS_AddFastLoadChecksums(PRUint32 sum1, PRUint32 sum2, PRUint32 sum2ByteCount)
{
PRUint32 A1 = sum1 & 0xffff;
PRUint32 B1 = sum1 >> 16;
PRUint32 A2 = sum2 & 0xffff;
PRUint32 B2 = sum2 >> 16;
PRUint32 A = A1 + A2;
while (A >> 16)
FOLD_ONES_COMPLEMENT_CARRY(A);
PRUint32 B = B2;
for (PRUint32 n = (sum2ByteCount + 1) / 2; n != 0; n--)
ONES_COMPLEMENT_ACCUMULATE(B, B1);
while (B >> 16)
FOLD_ONES_COMPLEMENT_CARRY(B);
return (B << 16) | A;
}
#undef FOLD_ONES_COMPLEMENT_CARRY
#undef ONES_COMPLEMENT_ACCUMULATE
#undef FLETCHER_ACCUMULATE
static const char magic[] = MFL_FILE_MAGIC;
// -------------------------- nsFastLoadFileReader --------------------------
nsID nsFastLoadFileReader::nsFastLoadFooter::gDummyID;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsFastLoadFileReader::nsObjectMapEntry
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
nsFastLoadFileReader::nsFastLoadFooter::gDummySharpObjectEntry;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
NS_IMPL_ISUPPORTS_INHERITED5(nsFastLoadFileReader,
nsBinaryInputStream,
nsIObjectInputStream,
nsIFastLoadFileControl,
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsIFastLoadReadControl,
nsISeekableStream,
nsIFastLoadFileReader)
MOZ_DECL_CTOR_COUNTER(nsFastLoadFileReader)
nsresult
nsFastLoadFileReader::ReadHeader(nsFastLoadHeader *aHeader)
{
nsresult rv;
PRUint32 bytesRead;
rv = Read(NS_REINTERPRET_CAST(char*, aHeader), sizeof *aHeader, &bytesRead);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
if (bytesRead != sizeof *aHeader ||
memcmp(aHeader->mMagic, magic, MFL_FILE_MAGIC_SIZE)) {
return NS_ERROR_UNEXPECTED;
}
aHeader->mChecksum = NS_SWAP32(aHeader->mChecksum);
aHeader->mVersion = NS_SWAP32(aHeader->mVersion);
aHeader->mFooterOffset = NS_SWAP32(aHeader->mFooterOffset);
aHeader->mFileSize = NS_SWAP32(aHeader->mFileSize);
return NS_OK;
}
// nsIFastLoadFileControl methods:
NS_IMETHODIMP
nsFastLoadFileReader::GetChecksum(PRUint32 *aChecksum)
{
*aChecksum = mHeader.mChecksum;
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileReader::SetChecksum(PRUint32 aChecksum)
{
mHeader.mChecksum = aChecksum;
return NS_OK;
}
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
struct nsStringMapEntry : public PLDHashEntryHdr {
const char* mString; // key, must come first
nsISupports* mURI; // for SelectMuxedDocument return value
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
};
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
struct nsDocumentMapEntry : public nsStringMapEntry {
PRUint32 mInitialSegmentOffset; // offset of URI's first segment in file
};
struct nsDocumentMapReadEntry : public nsDocumentMapEntry {
PRUint32 mNextSegmentOffset; // offset of URI's next segment to read
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
PRUint32 mBytesLeft : 31, // bytes remaining in current segment
mNeedToSeek : 1; // flag to defer Seek from Select to
// Read, in case there is no Read before
// another entry is Selected (to improve
// input stream buffer utilization)
PRInt64 mSaveOffset; // in case demux schedule differs from
// mux schedule
};
PR_STATIC_CALLBACK(void)
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
strmap_ClearEntry(PLDHashTable *aTable, PLDHashEntryHdr *aHdr)
{
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsStringMapEntry* entry = NS_STATIC_CAST(nsStringMapEntry*, aHdr);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (entry->mString)
nsMemory::Free((void*) entry->mString);
NS_IF_RELEASE(entry->mURI);
PL_DHashClearEntryStub(aTable, aHdr);
}
static const PLDHashTableOps strmap_DHashTableOps = {
PL_DHashAllocTable,
PL_DHashFreeTable,
PL_DHashGetKeyStub,
PL_DHashStringKey,
PL_DHashMatchStringKey,
PL_DHashMoveEntryStub,
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
strmap_ClearEntry,
PL_DHashFinalizeStub,
NULL
};
// An nsObjectMapEntry holds a strong reference to an XPCOM object, unless the
// mObject member, when cast to NSFastLoadOID, has its MFL_OBJECT_DEF_TAG bit
// set. NB: we rely on the fact that an nsISupports* is never an odd pointer.
struct nsObjectMapEntry : public PLDHashEntryHdr {
nsISupports* mObject; // key, must come first
};
// Fast mapping from URI object pointer back to spec-indexed document info.
struct nsURIMapReadEntry : public nsObjectMapEntry {
nsDocumentMapReadEntry* mDocMapEntry;
};
PR_STATIC_CALLBACK(void)
objmap_ClearEntry(PLDHashTable *aTable, PLDHashEntryHdr *aHdr)
{
nsObjectMapEntry* entry = NS_STATIC_CAST(nsObjectMapEntry*, aHdr);
// Ignore tagged object ids stored as object pointer keys (the updater
// code does this).
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if ((NS_PTR_TO_INT32(entry->mObject) & MFL_OBJECT_DEF_TAG) == 0)
NS_IF_RELEASE(entry->mObject);
PL_DHashClearEntryStub(aTable, aHdr);
}
static const PLDHashTableOps objmap_DHashTableOps = {
PL_DHashAllocTable,
PL_DHashFreeTable,
PL_DHashGetKeyStub,
PL_DHashVoidPtrKeyStub,
PL_DHashMatchEntryStub,
PL_DHashMoveEntryStub,
objmap_ClearEntry,
PL_DHashFinalizeStub,
NULL
};
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
NS_IMETHODIMP
nsFastLoadFileReader::HasMuxedDocument(const char* aURISpec, PRBool *aResult)
{
nsDocumentMapReadEntry* docMapEntry =
NS_STATIC_CAST(nsDocumentMapReadEntry*,
PL_DHashTableOperate(&mFooter.mDocumentMap, aURISpec,
PL_DHASH_LOOKUP));
*aResult = PL_DHASH_ENTRY_IS_BUSY(docMapEntry);
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileReader::StartMuxedDocument(nsISupports* aURI, const char* aURISpec)
{
nsDocumentMapReadEntry* docMapEntry =
NS_STATIC_CAST(nsDocumentMapReadEntry*,
PL_DHashTableOperate(&mFooter.mDocumentMap, aURISpec,
PL_DHASH_LOOKUP));
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
// If the spec isn't in the map, return NS_ERROR_NOT_AVAILABLE so the
// FastLoad service can try for a file update.
if (PL_DHASH_ENTRY_IS_FREE(docMapEntry))
return NS_ERROR_NOT_AVAILABLE;
nsCOMPtr<nsISupports> key(do_QueryInterface(aURI));
nsURIMapReadEntry* uriMapEntry =
NS_STATIC_CAST(nsURIMapReadEntry*,
PL_DHashTableOperate(&mFooter.mURIMap, key,
PL_DHASH_ADD));
if (!uriMapEntry)
return NS_ERROR_OUT_OF_MEMORY;
NS_ASSERTION(uriMapEntry->mDocMapEntry == nsnull,
"URI mapped to two different specs?");
if (uriMapEntry->mDocMapEntry)
return NS_ERROR_UNEXPECTED;
docMapEntry->mURI = aURI;
NS_ADDREF(docMapEntry->mURI);
uriMapEntry->mObject = key;
NS_ADDREF(uriMapEntry->mObject);
uriMapEntry->mDocMapEntry = docMapEntry;
TRACE_MUX(('r', "start %p (%p) %s\n", aURI, key.get(), aURISpec));
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileReader::SelectMuxedDocument(nsISupports* aURI,
nsISupports** aResult)
{
nsresult rv;
// Find the given URI's entry and select it for more reading.
nsCOMPtr<nsISupports> key(do_QueryInterface(aURI));
nsURIMapReadEntry* uriMapEntry =
NS_STATIC_CAST(nsURIMapReadEntry*,
PL_DHashTableOperate(&mFooter.mURIMap, key,
PL_DHASH_LOOKUP));
// If the URI isn't in the map, return NS_ERROR_NOT_AVAILABLE so the
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
// FastLoad service can try selecting the file updater.
if (PL_DHASH_ENTRY_IS_FREE(uriMapEntry))
return NS_ERROR_NOT_AVAILABLE;
// If we're interrupting another document's segment, save its offset so
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
// we can seek back when it's reselected. If prevDocMapEntry->mNeedToSeek
// is set, that means the stream is not positioned for prevDocMapEntry, to
// avoid buffer thrashing. See below in this function for more.
nsDocumentMapReadEntry* prevDocMapEntry = mCurrentDocumentMapEntry;
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
if (prevDocMapEntry &&
prevDocMapEntry->mBytesLeft &&
!prevDocMapEntry->mNeedToSeek) {
rv = Tell(&prevDocMapEntry->mSaveOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
}
// It turns out we get a fair amount of redundant select calls, thanks to
// non-blocking hunks of data from the parser that are devoid of scripts.
// As more data gets FastLoaded, the number of these useless selects will
// decline.
nsDocumentMapReadEntry* docMapEntry = uriMapEntry->mDocMapEntry;
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
if (docMapEntry == prevDocMapEntry) {
TRACE_MUX(('r', "select prev %s same as current!\n",
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
docMapEntry->mString));
}
// Invariant: docMapEntry->mBytesLeft implies docMapEntry->mSaveOffset has
// been set non-zero by the Tell call above.
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
else if (docMapEntry->mBytesLeft) {
NS_ASSERTION(docMapEntry->mSaveOffset != 0,
"reselecting from multiplex at unsaved offset?");
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
// Defer Seek till Read, in case of "ping-pong" Selects without any
// intervening Reads, to avoid dumping the underlying mInputStream's
// input buffer for cases where alternate "pongs" fall in the same
// buffer.
docMapEntry->mNeedToSeek = PR_TRUE;
}
*aResult = prevDocMapEntry ? prevDocMapEntry->mURI : nsnull;
NS_IF_ADDREF(*aResult);
mCurrentDocumentMapEntry = docMapEntry;
#ifdef DEBUG_MUX
PRInt64 currentSegmentOffset;
Tell(&currentSegmentOffset);
trace_mux('r', "select %p (%p) offset %ld\n",
aURI, key.get(), (long) currentSegmentOffset);
#endif
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileReader::EndMuxedDocument(nsISupports* aURI)
{
nsCOMPtr<nsISupports> key(do_QueryInterface(aURI));
nsURIMapReadEntry* uriMapEntry =
NS_STATIC_CAST(nsURIMapReadEntry*,
PL_DHashTableOperate(&mFooter.mURIMap, key,
PL_DHASH_LOOKUP));
// If the URI isn't in the map, return NS_ERROR_NOT_AVAILABLE so the
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
// FastLoad service can try to end a select on its file updater.
if (PL_DHASH_ENTRY_IS_FREE(uriMapEntry))
return NS_ERROR_NOT_AVAILABLE;
// Drop our ref to the URI object that was passed to StartMuxedDocument,
// we no longer need it, and we do not want to extend its lifetime.
if (uriMapEntry->mDocMapEntry)
NS_RELEASE(uriMapEntry->mDocMapEntry->mURI);
// Shrink the table if half the entries are removed sentinels.
PRUint32 size = PL_DHASH_TABLE_SIZE(&mFooter.mURIMap);
if (mFooter.mURIMap.removedCount >= (size >> 2))
PL_DHashTableOperate(&mFooter.mURIMap, key, PL_DHASH_REMOVE);
else
PL_DHashTableRawRemove(&mFooter.mURIMap, uriMapEntry);
TRACE_MUX(('r', "end %p (%p)\n", aURI, key.get()));
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileReader::Read(char* aBuffer, PRUint32 aCount, PRUint32 *aBytesRead)
{
nsresult rv;
nsDocumentMapReadEntry* entry = mCurrentDocumentMapEntry;
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
if (entry) {
// Don't call our Seek wrapper, as it clears mCurrentDocumentMapEntry.
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
if (entry->mNeedToSeek) {
rv = mSeekableInput->Seek(nsISeekableStream::NS_SEEK_SET,
entry->mSaveOffset);
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
if (NS_FAILED(rv))
return rv;
entry->mNeedToSeek = PR_FALSE;
}
// Loop to handle empty segments, which may be generated by the
// writer, given Start A; Start B; Select A; Select B; write B data;
// multiplexing schedules, which do tend to occur given non-blocking
// i/o with LIFO scheduling. XXXbe investigate LIFO issues
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
while (entry->mBytesLeft == 0) {
// Check for unexpected end of multiplexed stream.
NS_ASSERTION(entry->mNextSegmentOffset != 0,
"document demuxed from FastLoad file more than once?");
if (entry->mNextSegmentOffset == 0)
return NS_ERROR_UNEXPECTED;
rv = mSeekableInput->Seek(nsISeekableStream::NS_SEEK_SET,
entry->mNextSegmentOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
// Clear mCurrentDocumentMapEntry temporarily to avoid recursion.
mCurrentDocumentMapEntry = nsnull;
rv = Read32(&entry->mNextSegmentOffset);
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
if (NS_SUCCEEDED(rv)) {
PRUint32 bytesLeft = 0;
rv = Read32(&bytesLeft);
entry->mBytesLeft = bytesLeft;
}
mCurrentDocumentMapEntry = entry;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
NS_ASSERTION(entry->mBytesLeft >= 8, "demux segment length botch!");
entry->mBytesLeft -= 8;
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
}
}
rv = mInputStream->Read(aBuffer, aCount, aBytesRead);
if (NS_SUCCEEDED(rv) && entry) {
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
NS_ASSERTION(entry->mBytesLeft >= *aBytesRead, "demux Read underflow!");
entry->mBytesLeft -= *aBytesRead;
#ifdef NS_DEBUG
// Invariant: !entry->mBytesLeft implies entry->mSaveOffset == 0.
if (entry->mBytesLeft == 0)
entry->mSaveOffset = 0;
#endif
}
return rv;
}
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
NS_IMETHODIMP
nsFastLoadFileReader::ReadSegments(nsWriteSegmentFun aWriter, void* aClosure,
PRUint32 aCount, PRUint32 *aResult)
{
nsDocumentMapReadEntry* entry = mCurrentDocumentMapEntry;
NS_ASSERTION(!entry || (!entry->mNeedToSeek && entry->mBytesLeft != 0),
"ReadSegments called from above nsFastLoadFileReader layer?!");
nsresult rv = nsBinaryInputStream::ReadSegments(aWriter, aClosure, aCount,
aResult);
if (NS_SUCCEEDED(rv) && entry) {
NS_ASSERTION(entry->mBytesLeft >= *aResult,
"demux ReadSegments underflow!");
entry->mBytesLeft -= *aResult;
#ifdef NS_DEBUG
// Invariant: !entry->mBytesLeft implies entry->mSaveOffset == 0.
if (entry->mBytesLeft == 0)
entry->mSaveOffset = 0;
#endif
}
return rv;
}
NS_IMETHODIMP
nsFastLoadFileReader::SetInputStream(nsIInputStream *aInputStream)
{
nsresult rv = nsBinaryInputStream::SetInputStream(aInputStream);
mSeekableInput = do_QueryInterface(aInputStream);
NS_ASSERTION(!mInputStream || mSeekableInput,
"FastLoad requires a seekable input stream");
return rv;
}
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
/**
* XXX tuneme
*/
#define MFL_CHECKSUM_BUFSIZE 8192
NS_IMETHODIMP
nsFastLoadFileReader::ComputeChecksum(PRUint32 *aResult)
{
nsCOMPtr<nsIInputStream> stream = mInputStream;
nsCOMPtr<nsISeekableStream> seekable = mSeekableInput;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
PRInt64 saveOffset;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsresult rv = seekable->Tell(&saveOffset);
if (NS_FAILED(rv))
return rv;
if (mBufferAccess) {
rv = mBufferAccess->GetUnbufferedStream(getter_AddRefs(stream));
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
seekable = do_QueryInterface(stream);
if (!seekable)
return NS_ERROR_UNEXPECTED;
}
rv = seekable->Seek(nsISeekableStream::NS_SEEK_SET, 0);
if (NS_FAILED(rv))
return rv;
char buf[MFL_CHECKSUM_BUFSIZE];
PRUint32 len, rem;
rem = offsetof(nsFastLoadHeader, mChecksum);
rv = stream->Read(buf, rem, &len);
if (NS_FAILED(rv))
return rv;
if (len != rem)
return NS_ERROR_UNEXPECTED;
rv = seekable->Seek(nsISeekableStream::NS_SEEK_CUR, 4);
if (NS_FAILED(rv))
return rv;
memset(buf + rem, 0, 4);
rem += 4;
PRUint32 checksum = 0;
while (NS_SUCCEEDED(rv = stream->Read(buf + rem, sizeof buf - rem, &len)) &&
len) {
len += rem;
rem = NS_AccumulateFastLoadChecksum(&checksum,
NS_REINTERPRET_CAST(PRUint8*, buf),
len,
PR_FALSE);
if (rem)
memcpy(buf, buf + len - rem, rem);
}
if (NS_FAILED(rv))
return rv;
if (rem) {
NS_AccumulateFastLoadChecksum(&checksum,
NS_REINTERPRET_CAST(PRUint8*, buf),
rem,
PR_TRUE);
}
rv = seekable->Seek(nsISeekableStream::NS_SEEK_SET, saveOffset);
if (NS_FAILED(rv))
return rv;
*aResult = checksum;
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileReader::GetDependencies(nsISimpleEnumerator* *aDependencies)
{
return NS_NewArrayEnumerator(aDependencies, mFooter.mDependencies);
}
nsresult
nsFastLoadFileReader::ReadFooter(nsFastLoadFooter *aFooter)
{
nsresult rv;
rv = ReadFooterPrefix(aFooter);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
aFooter->mIDMap = new nsID[aFooter->mNumIDs];
if (!aFooter->mIDMap)
return NS_ERROR_OUT_OF_MEMORY;
PRUint32 i, n;
for (i = 0, n = aFooter->mNumIDs; i < n; i++) {
rv = ReadSlowID(&aFooter->mIDMap[i]);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
}
aFooter->mObjectMap = new nsObjectMapEntry[aFooter->mNumSharpObjects];
if (!aFooter->mObjectMap)
return NS_ERROR_OUT_OF_MEMORY;
for (i = 0, n = aFooter->mNumSharpObjects; i < n; i++) {
nsObjectMapEntry* entry = &aFooter->mObjectMap[i];
rv = ReadSharpObjectInfo(entry);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
entry->mReadObject = nsnull;
entry->mSkipOffset = 0;
entry->mSaveStrongRefCnt = entry->mStrongRefCnt;
entry->mSaveWeakRefCnt = entry->mWeakRefCnt;
}
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (!PL_DHashTableInit(&aFooter->mDocumentMap, &strmap_DHashTableOps,
(void *)this, sizeof(nsDocumentMapReadEntry),
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
aFooter->mNumMuxedDocuments)) {
aFooter->mDocumentMap.ops = nsnull;
return NS_ERROR_OUT_OF_MEMORY;
}
if (!PL_DHashTableInit(&aFooter->mURIMap, &objmap_DHashTableOps,
(void *)this, sizeof(nsURIMapReadEntry),
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
aFooter->mNumMuxedDocuments)) {
aFooter->mURIMap.ops = nsnull;
return NS_ERROR_OUT_OF_MEMORY;
}
for (i = 0, n = aFooter->mNumMuxedDocuments; i < n; i++) {
nsFastLoadMuxedDocumentInfo info;
rv = ReadMuxedDocumentInfo(&info);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
nsDocumentMapReadEntry* entry =
NS_STATIC_CAST(nsDocumentMapReadEntry*,
PL_DHashTableOperate(&aFooter->mDocumentMap,
info.mURISpec,
PL_DHASH_ADD));
if (!entry) {
nsMemory::Free((void*) info.mURISpec);
return NS_ERROR_OUT_OF_MEMORY;
}
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
NS_ASSERTION(!entry->mString, "duplicate URISpec in MuxedDocumentMap");
entry->mString = info.mURISpec;
entry->mURI = nsnull;
entry->mInitialSegmentOffset = info.mInitialSegmentOffset;
entry->mNextSegmentOffset = info.mInitialSegmentOffset;
entry->mBytesLeft = 0;
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
entry->mNeedToSeek = PR_FALSE;
entry->mSaveOffset = 0;
}
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsCOMPtr<nsISupportsArray> readDeps;
rv = NS_NewISupportsArray(getter_AddRefs(readDeps));
if (NS_FAILED(rv))
return rv;
- Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Writer or Updater must close after the Reader, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-17 13:52:55 +04:00
nsCAutoString filename;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
for (i = 0, n = aFooter->mNumDependencies; i < n; i++) {
rv = ReadCString(filename);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
PRInt64 fastLoadMtime;
rv = Read64(NS_REINTERPRET_CAST(PRUint64*, &fastLoadMtime));
if (NS_FAILED(rv))
return rv;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsCOMPtr<nsILocalFile> file;
rv = NS_NewNativeLocalFile(filename, PR_TRUE, getter_AddRefs(file));
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
PRInt64 currentMtime;
rv = file->GetLastModifiedTime(&currentMtime);
if (NS_FAILED(rv))
return rv;
if (LL_NE(fastLoadMtime, currentMtime)) {
#ifdef DEBUG
nsCAutoString path;
file->GetNativePath(path);
printf("%s mtime changed, invalidating FastLoad file\n",
path.get());
#endif
return NS_ERROR_FAILURE;
}
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
rv = readDeps->AppendElement(file);
if (NS_FAILED(rv))
return rv;
}
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
aFooter->mDependencies = readDeps;
return NS_OK;
}
nsresult
nsFastLoadFileReader::ReadFooterPrefix(nsFastLoadFooterPrefix *aFooterPrefix)
{
nsresult rv;
rv = Read32(&aFooterPrefix->mNumIDs);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Read32(&aFooterPrefix->mNumSharpObjects);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Read32(&aFooterPrefix->mNumMuxedDocuments);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Read32(&aFooterPrefix->mNumDependencies);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
return NS_OK;
}
nsresult
nsFastLoadFileReader::ReadSlowID(nsID *aID)
{
nsresult rv;
rv = Read32(&aID->m0);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Read16(&aID->m1);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Read16(&aID->m2);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
PRUint32 bytesRead;
rv = Read(NS_REINTERPRET_CAST(char*, aID->m3), sizeof aID->m3, &bytesRead);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
if (bytesRead != sizeof aID->m3)
return NS_ERROR_FAILURE;
return NS_OK;
}
nsresult
nsFastLoadFileReader::ReadFastID(NSFastLoadID *aID)
{
nsresult rv = Read32(aID);
if (NS_SUCCEEDED(rv))
*aID ^= MFL_ID_XOR_KEY;
return rv;
}
nsresult
nsFastLoadFileReader::ReadSharpObjectInfo(nsFastLoadSharpObjectInfo *aInfo)
{
nsresult rv;
rv = Read32(&aInfo->mCIDOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
NS_ASSERTION(aInfo->mCIDOffset != 0,
"fastload reader: mCIDOffset cannot be zero!");
rv = Read16(&aInfo->mStrongRefCnt);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Read16(&aInfo->mWeakRefCnt);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
return NS_OK;
}
nsresult
nsFastLoadFileReader::ReadMuxedDocumentInfo(nsFastLoadMuxedDocumentInfo *aInfo)
{
nsresult rv;
nsCAutoString spec;
rv = ReadCString(spec);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Read32(&aInfo->mInitialSegmentOffset);
if (NS_FAILED(rv))
return rv;
aInfo->mURISpec = ToNewCString(spec);
return NS_OK;
}
nsresult
nsFastLoadFileReader::Open()
{
nsresult rv;
// Don't bother buffering the header, as we immediately seek to EOF.
if (mBufferAccess)
mBufferAccess->DisableBuffering();
rv = ReadHeader(&mHeader);
if (mBufferAccess)
mBufferAccess->EnableBuffering();
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
if (mHeader.mVersion != MFL_FILE_VERSION)
return NS_ERROR_UNEXPECTED;
if (mHeader.mFooterOffset == 0)
return NS_ERROR_UNEXPECTED;
rv = mSeekableInput->Seek(nsISeekableStream::NS_SEEK_END, 0);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
PRInt64 fileSize;
rv = mSeekableInput->Tell(&fileSize);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
nsInt64 fileSize64 = fileSize;
const nsInt64 maxUint32 = PR_UINT32_MAX;
NS_ASSERTION(fileSize64 <= maxUint32, "fileSize must fit in 32 bits");
if ((PRUint32) fileSize64 != mHeader.mFileSize)
return NS_ERROR_UNEXPECTED;
rv = mSeekableInput->Seek(nsISeekableStream::NS_SEEK_SET,
PRInt32(mHeader.mFooterOffset));
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = ReadFooter(&mFooter);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
return mSeekableInput->Seek(nsISeekableStream::NS_SEEK_SET,
sizeof(nsFastLoadHeader));
}
NS_IMETHODIMP
nsFastLoadFileReader::Close()
{
// Give up our strong "keepalive" references, in case not all objects that
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
// were deserialized were fully re-connected.
//
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
// This happens for sure when an nsFastLoadFileUpdater is created and wraps
// an nsFastLoadFileReader whose data was already deserialized by an earlier
// FastLoad episode. The reader is useful in the second such episode during
// a session not so much for reading objects as for its footer information,
// which primes the updater's tables so that after the update completes, the
// FastLoad file has a superset footer.
for (PRUint32 i = 0, n = mFooter.mNumSharpObjects; i < n; i++) {
nsObjectMapEntry* entry = &mFooter.mObjectMap[i];
entry->mReadObject = nsnull;
}
return mInputStream->Close();
}
nsresult
nsFastLoadFileReader::DeserializeObject(nsISupports* *aObject)
{
nsresult rv;
NSFastLoadID fastCID;
rv = ReadFastID(&fastCID);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
const nsID& slowCID = mFooter.GetID(fastCID);
nsCOMPtr<nsISupports> object(do_CreateInstance(slowCID, &rv));
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
nsCOMPtr<nsISerializable> serializable(do_QueryInterface(object));
if (!serializable)
return NS_ERROR_FAILURE;
rv = serializable->Read(this);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
*aObject = object;
NS_ADDREF(*aObject);
return NS_OK;
}
nsresult
nsFastLoadFileReader::ReadObject(PRBool aIsStrongRef, nsISupports* *aObject)
{
nsresult rv;
NSFastLoadOID oid;
rv = Read32(&oid);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
oid ^= MFL_OID_XOR_KEY;
nsCOMPtr<nsISupports> object;
if (oid == MFL_DULL_OBJECT_OID) {
// A very dull object, defined at point of single (strong) reference.
NS_ASSERTION(aIsStrongRef, "dull object read via weak ref!");
rv = DeserializeObject(getter_AddRefs(object));
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
} else {
NS_ASSERTION((oid & MFL_WEAK_REF_TAG) ==
(aIsStrongRef ? 0 : MFL_WEAK_REF_TAG),
"strong vs. weak ref deserialization mismatch!");
nsObjectMapEntry* entry = &mFooter.GetSharpObjectEntry(oid);
// Check whether we've already deserialized the object for this OID.
object = entry->mReadObject;
if (!object) {
PRInt64 saveOffset;
nsDocumentMapReadEntry* saveDocMapEntry = nsnull;
rv = mSeekableInput->Tell(&saveOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
PRUint32 saveOffset32 = saveOffset;
if (entry->mCIDOffset != saveOffset32) {
// We skipped deserialization of this object from its position
// earlier in the input stream, presumably due to the reference
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
// there being an nsFastLoadPtr, or (more likely) because the
// object was muxed in another document, and deserialization
// order does not match serialization order. So we must seek
// back and read it now.
NS_ASSERTION(entry->mCIDOffset < saveOffset32,
"out of order object?!");
// Ape our Seek wrapper by clearing mCurrentDocumentMapEntry.
// This allows for a skipped object to be referenced from two
// or more multiplexed documents in the FastLoad file.
saveDocMapEntry = mCurrentDocumentMapEntry;
mCurrentDocumentMapEntry = nsnull;
rv = mSeekableInput->Seek(nsISeekableStream::NS_SEEK_SET,
entry->mCIDOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
}
rv = DeserializeObject(getter_AddRefs(object));
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
if (entry->mCIDOffset != saveOffset32) {
// Save the "skip offset" in case we need to skip this object
// definition when reading forward, later on.
rv = mSeekableInput->Tell(&entry->mSkipOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
// Restore stream offset and mCurrentDocumentMapEntry in case
// we're still reading forward through a part of the multiplex
// to get object definitions eagerly.
rv = mSeekableInput->Seek(nsISeekableStream::NS_SEEK_SET,
saveOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
mCurrentDocumentMapEntry = saveDocMapEntry;
}
// Save object until all refs have been deserialized.
entry->mReadObject = object;
} else {
// What if we are at a definition that's already been read? This
// case arises when a sharp object's def is serialized before its
// refs, while a non-defining ref is deserialized before the def.
// We must skip over the object definition.
if (oid & MFL_OBJECT_DEF_TAG) {
NS_ASSERTION(entry->mSkipOffset != 0, "impossible! see above");
rv = mSeekableInput->Seek(nsISeekableStream::NS_SEEK_SET,
entry->mSkipOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
}
}
if (aIsStrongRef) {
NS_ASSERTION(entry->mStrongRefCnt != 0,
"mStrongRefCnt underflow!");
--entry->mStrongRefCnt;
} else {
NS_ASSERTION(MFL_GET_WEAK_REFCNT(entry) != 0,
"mWeakRefCnt underflow!");
MFL_DROP_WEAK_REFCNT(entry);
}
if (entry->mStrongRefCnt == 0 && MFL_GET_WEAK_REFCNT(entry) == 0)
entry->mReadObject = nsnull;
}
if (oid & MFL_QUERY_INTERFACE_TAG) {
NSFastLoadID iid;
rv = ReadFastID(&iid);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = object->QueryInterface(mFooter.GetID(iid),
NS_REINTERPRET_CAST(void**, aObject));
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
} else {
*aObject = object;
NS_ADDREF(*aObject);
}
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileReader::ReadID(nsID *aResult)
{
nsresult rv;
NSFastLoadID fastID;
rv = ReadFastID(&fastID);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
*aResult = mFooter.GetID(fastID);
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileReader::Seek(PRInt32 aWhence, PRInt64 aOffset)
{
mCurrentDocumentMapEntry = nsnull;
return mSeekableInput->Seek(aWhence, aOffset);
}
NS_IMETHODIMP
nsFastLoadFileReader::Tell(PRInt64 *aResult)
{
return mSeekableInput->Tell(aResult);
}
NS_IMETHODIMP
nsFastLoadFileReader::SetEOF()
{
return mSeekableInput->SetEOF();
}
NS_COM nsresult
NS_NewFastLoadFileReader(nsIObjectInputStream* *aResult,
nsIInputStream* aSrcStream)
{
nsFastLoadFileReader* reader = new nsFastLoadFileReader(aSrcStream);
if (!reader)
return NS_ERROR_OUT_OF_MEMORY;
// Stabilize reader's refcnt.
nsCOMPtr<nsIObjectInputStream> stream(reader);
nsresult rv = reader->Open();
if (NS_FAILED(rv))
return rv;
*aResult = stream;
NS_ADDREF(*aResult);
return NS_OK;
}
// -------------------------- nsFastLoadFileWriter --------------------------
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
NS_IMPL_ISUPPORTS_INHERITED4(nsFastLoadFileWriter,
nsBinaryOutputStream,
nsIObjectOutputStream,
nsIFastLoadFileControl,
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsIFastLoadWriteControl,
nsISeekableStream)
MOZ_DECL_CTOR_COUNTER(nsFastLoadFileWriter)
struct nsIDMapEntry : public PLDHashEntryHdr {
NSFastLoadID mFastID; // 1 + nsFastLoadFooter::mIDMap index
nsID mSlowID; // key, used by PLDHashTableOps below
};
PR_STATIC_CALLBACK(const void *)
idmap_GetKey(PLDHashTable *aTable, PLDHashEntryHdr *aHdr)
{
nsIDMapEntry* entry = NS_STATIC_CAST(nsIDMapEntry*, aHdr);
return &entry->mSlowID;
}
PR_STATIC_CALLBACK(PLDHashNumber)
idmap_HashKey(PLDHashTable *aTable, const void *aKey)
{
const nsID *idp = NS_REINTERPRET_CAST(const nsID*, aKey);
return idp->m0;
}
PR_STATIC_CALLBACK(PRBool)
idmap_MatchEntry(PLDHashTable *aTable,
const PLDHashEntryHdr *aHdr,
const void *aKey)
{
const nsIDMapEntry* entry = NS_STATIC_CAST(const nsIDMapEntry*, aHdr);
const nsID *idp = NS_REINTERPRET_CAST(const nsID*, aKey);
return memcmp(&entry->mSlowID, idp, sizeof(nsID)) == 0;
}
static const PLDHashTableOps idmap_DHashTableOps = {
PL_DHashAllocTable,
PL_DHashFreeTable,
idmap_GetKey,
idmap_HashKey,
idmap_MatchEntry,
PL_DHashMoveEntryStub,
PL_DHashClearEntryStub,
PL_DHashFinalizeStub,
NULL
};
nsresult
nsFastLoadFileWriter::MapID(const nsID& aSlowID, NSFastLoadID *aResult)
{
nsIDMapEntry* entry =
NS_STATIC_CAST(nsIDMapEntry*,
PL_DHashTableOperate(&mIDMap, &aSlowID, PL_DHASH_ADD));
if (!entry)
return NS_ERROR_OUT_OF_MEMORY;
if (entry->mFastID == 0) {
entry->mFastID = mIDMap.entryCount;
entry->mSlowID = aSlowID;
}
*aResult = entry->mFastID;
return NS_OK;
}
nsresult
nsFastLoadFileWriter::WriteHeader(nsFastLoadHeader *aHeader)
{
nsresult rv;
PRUint32 bytesWritten;
rv = Write(aHeader->mMagic, MFL_FILE_MAGIC_SIZE, &bytesWritten);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
if (bytesWritten != MFL_FILE_MAGIC_SIZE)
return NS_ERROR_FAILURE;
rv = Write32(aHeader->mChecksum);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write32(aHeader->mVersion);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write32(aHeader->mFooterOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write32(aHeader->mFileSize);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
return NS_OK;
}
// nsIFastLoadFileControl methods:
NS_IMETHODIMP
nsFastLoadFileWriter::GetChecksum(PRUint32 *aChecksum)
{
if (mHeader.mChecksum == 0)
return NS_ERROR_NOT_AVAILABLE;
*aChecksum = mHeader.mChecksum;
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileWriter::SetChecksum(PRUint32 aChecksum)
{
mHeader.mChecksum = aChecksum;
return NS_OK;
}
struct nsDocumentMapWriteEntry : public nsDocumentMapEntry {
PRUint32 mCurrentSegmentOffset; // last written segment's offset
};
// Fast mapping from URI object pointer back to spec-indexed document info.
// We also may need the slow mapping from mURISpec to nsDocumentMapWriteEntry,
// because the writer's mDocumentMap double hash table may grow "behind the
// back of" each mURIMap entry's mDocMapEntry member.
struct nsURIMapWriteEntry : public nsObjectMapEntry {
nsDocumentMapWriteEntry* mDocMapEntry;
PRUint32 mGeneration;
const char* mURISpec;
};
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
NS_IMETHODIMP
nsFastLoadFileWriter::HasMuxedDocument(const char* aURISpec, PRBool *aResult)
{
nsDocumentMapWriteEntry* docMapEntry =
NS_STATIC_CAST(nsDocumentMapWriteEntry*,
PL_DHashTableOperate(&mDocumentMap, aURISpec,
PL_DHASH_LOOKUP));
*aResult = PL_DHASH_ENTRY_IS_BUSY(docMapEntry);
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileWriter::StartMuxedDocument(nsISupports* aURI,
const char* aURISpec)
{
// Save mDocumentMap table generation and mCurrentDocumentMapEntry key in
// case the hash table grows during the PL_DHASH_ADD operation.
PRUint32 saveGeneration = mDocumentMap.generation;
const char* saveURISpec = mCurrentDocumentMapEntry
? mCurrentDocumentMapEntry->mString
: nsnull;
nsDocumentMapWriteEntry* docMapEntry =
NS_STATIC_CAST(nsDocumentMapWriteEntry*,
PL_DHashTableOperate(&mDocumentMap, aURISpec,
PL_DHASH_ADD));
if (!docMapEntry)
return NS_ERROR_OUT_OF_MEMORY;
// If the generation number changed, refresh mCurrentDocumentMapEntry.
if (mCurrentDocumentMapEntry && mDocumentMap.generation != saveGeneration) {
mCurrentDocumentMapEntry =
NS_STATIC_CAST(nsDocumentMapWriteEntry*,
PL_DHashTableOperate(&mDocumentMap, saveURISpec,
PL_DHASH_LOOKUP));
NS_ASSERTION(PL_DHASH_ENTRY_IS_BUSY(mCurrentDocumentMapEntry),
"mCurrentDocumentMapEntry lost during table growth?!");
// Refresh saveGeneration for use below when initializing uriMapEntry.
saveGeneration = mDocumentMap.generation;
}
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
NS_ASSERTION(docMapEntry->mString == nsnull,
"redundant multiplexed document?");
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (docMapEntry->mString)
return NS_ERROR_UNEXPECTED;
void* spec = nsMemory::Clone(aURISpec, strlen(aURISpec) + 1);
if (!spec)
return NS_ERROR_OUT_OF_MEMORY;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
docMapEntry->mString = NS_REINTERPRET_CAST(const char*, spec);
docMapEntry->mURI = aURI;
NS_ADDREF(docMapEntry->mURI);
nsCOMPtr<nsISupports> key(do_QueryInterface(aURI));
nsURIMapWriteEntry* uriMapEntry =
NS_STATIC_CAST(nsURIMapWriteEntry*,
PL_DHashTableOperate(&mURIMap, key, PL_DHASH_ADD));
if (!uriMapEntry)
return NS_ERROR_OUT_OF_MEMORY;
NS_ASSERTION(uriMapEntry->mDocMapEntry == nsnull,
"URI mapped to two different specs?");
if (uriMapEntry->mDocMapEntry)
return NS_ERROR_UNEXPECTED;
uriMapEntry->mObject = key;
NS_ADDREF(uriMapEntry->mObject);
uriMapEntry->mDocMapEntry = docMapEntry;
uriMapEntry->mGeneration = saveGeneration;
uriMapEntry->mURISpec = NS_REINTERPRET_CAST(const char*, spec);
TRACE_MUX(('w', "start %p (%p) %s\n", aURI, key.get(), aURISpec));
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileWriter::SelectMuxedDocument(nsISupports* aURI,
nsISupports** aResult)
{
// Capture the current file offset (XXXbe maintain our own via Write?)
nsresult rv;
PRInt64 currentSegmentOffset;
rv = mSeekableOutput->Tell(&currentSegmentOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
PRUint32 currentSegmentOffset32 = currentSegmentOffset;
// Look for an existing entry keyed by aURI, added by StartMuxedDocument.
nsCOMPtr<nsISupports> key(do_QueryInterface(aURI));
nsURIMapWriteEntry* uriMapEntry =
NS_STATIC_CAST(nsURIMapWriteEntry*,
PL_DHashTableOperate(&mURIMap, key, PL_DHASH_LOOKUP));
NS_ASSERTION(PL_DHASH_ENTRY_IS_BUSY(uriMapEntry),
"SelectMuxedDocument without prior StartMuxedDocument?");
if (PL_DHASH_ENTRY_IS_FREE(uriMapEntry))
return NS_ERROR_UNEXPECTED;
// Beware that uriMapEntry->mDocMapEntry may be stale, if an mDocumentMap
// addition caused that table to grow. We save the mDocumentMap generation
// in each uriMapEntry and compare it to the current generation, rehashing
// uriMapEntry->mURISpec if necessary.
nsDocumentMapWriteEntry* docMapEntry = uriMapEntry->mDocMapEntry;
if (uriMapEntry->mGeneration != mDocumentMap.generation) {
docMapEntry =
NS_STATIC_CAST(nsDocumentMapWriteEntry*,
PL_DHashTableOperate(&mDocumentMap,
uriMapEntry->mURISpec,
PL_DHASH_LOOKUP));
NS_ASSERTION(PL_DHASH_ENTRY_IS_BUSY(docMapEntry), "lost mDocMapEntry!?");
uriMapEntry->mDocMapEntry = docMapEntry;
uriMapEntry->mGeneration = mDocumentMap.generation;
}
docMapEntry = uriMapEntry->mDocMapEntry;
// If there is a muxed document segment open, close it now by setting its
// length, stored in the second PRUint32 of the segment.
nsDocumentMapWriteEntry* prevDocMapEntry = mCurrentDocumentMapEntry;
if (prevDocMapEntry) {
if (prevDocMapEntry == docMapEntry) {
TRACE_MUX(('w', "select prev %s same as current!\n",
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
prevDocMapEntry->mString));
*aResult = docMapEntry->mURI;
NS_ADDREF(*aResult);
return NS_OK;
}
PRUint32 prevSegmentOffset = prevDocMapEntry->mCurrentSegmentOffset;
TRACE_MUX(('w', "select prev %s offset %lu\n",
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
prevDocMapEntry->mString, prevSegmentOffset));
rv = mSeekableOutput->Seek(nsISeekableStream::NS_SEEK_SET,
prevSegmentOffset + 4);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
// The length counts all bytes in the segment, including the header
// that contains [nextSegmentOffset, length].
rv = Write32(currentSegmentOffset32 - prevSegmentOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
// Seek back to the current offset only if we are not going to seek
// back to *this* entry's last "current" segment offset and write its
// next segment offset at the first PRUint32 of the segment.
if (!docMapEntry->mInitialSegmentOffset) {
rv = mSeekableOutput->Seek(nsISeekableStream::NS_SEEK_SET,
currentSegmentOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
}
}
// If this entry was newly added, set its key and initial segment offset.
// Otherwise, seek back to write the next segment offset of the previous
// segment for this document in the multiplex.
if (!docMapEntry->mInitialSegmentOffset) {
docMapEntry->mInitialSegmentOffset = currentSegmentOffset32;
} else {
rv = mSeekableOutput->Seek(nsISeekableStream::NS_SEEK_SET,
docMapEntry->mCurrentSegmentOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write32(currentSegmentOffset32);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = mSeekableOutput->Seek(nsISeekableStream::NS_SEEK_SET,
currentSegmentOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
}
// Update this document's current segment offset so we can later fix its
// next segment offset (unless it is last, in which case we leave the zero
// placeholder as a terminator).
docMapEntry->mCurrentSegmentOffset = currentSegmentOffset32;
rv = Write32(0); // nextSegmentOffset placeholder
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write32(0); // length placeholder
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
*aResult = prevDocMapEntry ? prevDocMapEntry->mURI : nsnull;
NS_IF_ADDREF(*aResult);
mCurrentDocumentMapEntry = docMapEntry;
TRACE_MUX(('w', "select %p (%p) offset %lu\n",
aURI, key.get(), currentSegmentOffset));
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileWriter::EndMuxedDocument(nsISupports* aURI)
{
nsCOMPtr<nsISupports> key(do_QueryInterface(aURI));
nsURIMapWriteEntry* uriMapEntry =
NS_STATIC_CAST(nsURIMapWriteEntry*,
PL_DHashTableOperate(&mURIMap, key, PL_DHASH_LOOKUP));
// If the URI isn't in the map, nsFastLoadFileWriter::StartMuxedDocument
// must have been called with a redundant URI, *and* its caller must have
// ignored the NS_ERROR_UNEXPECTED it returned in that case.
if (PL_DHASH_ENTRY_IS_FREE(uriMapEntry)) {
TRACE_MUX(('w', "bad end %p (%p)\n", aURI, key.get()));
return NS_ERROR_UNEXPECTED;
}
// Drop our ref to the URI object that was passed to StartMuxedDocument,
// we no longer need it, and we do not want to extend its lifetime.
if (uriMapEntry->mDocMapEntry)
NS_RELEASE(uriMapEntry->mDocMapEntry->mURI);
// Shrink the table if half the entries are removed sentinels.
PRUint32 size = PL_DHASH_TABLE_SIZE(&mURIMap);
if (mURIMap.removedCount >= (size >> 2))
PL_DHashTableOperate(&mURIMap, key, PL_DHASH_REMOVE);
else
PL_DHashTableRawRemove(&mURIMap, uriMapEntry);
TRACE_MUX(('w', "end %p (%p)\n", aURI, key.get()));
return NS_OK;
}
struct nsDependencyMapEntry : public nsStringMapEntry {
PRInt64 mLastModified;
};
NS_IMETHODIMP
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsFastLoadFileWriter::AddDependency(nsIFile* aFile)
{
nsCAutoString path;
nsresult rv = aFile->GetNativePath(path);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
nsDependencyMapEntry* entry =
NS_STATIC_CAST(nsDependencyMapEntry*,
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
PL_DHashTableOperate(&mDependencyMap, path.get(),
PL_DHASH_ADD));
if (!entry)
return NS_ERROR_OUT_OF_MEMORY;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (!entry->mString) {
const char *tmp = ToNewCString(path);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (!tmp)
return NS_ERROR_OUT_OF_MEMORY;
entry->mString = tmp;
// If we can't get the last modified time from aFile, assume it does
// not exist, or is otherwise inaccessible to us (due to permissions),
// remove the dependency, and suppress the failure.
//
// Otherwise, we would end up aborting the fastload process due to a
// missing .js or .xul or other file on every startup.
rv = aFile->GetLastModifiedTime(&entry->mLastModified);
if (NS_FAILED(rv)) {
PL_DHashTableOperate(&mDependencyMap, path.get(), PL_DHASH_REMOVE);
rv = NS_OK;
}
}
return rv;
}
nsresult
nsFastLoadFileWriter::WriteFooterPrefix(const nsFastLoadFooterPrefix& aFooterPrefix)
{
nsresult rv;
rv = Write32(aFooterPrefix.mNumIDs);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write32(aFooterPrefix.mNumSharpObjects);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write32(aFooterPrefix.mNumMuxedDocuments);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write32(aFooterPrefix.mNumDependencies);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
return NS_OK;
}
nsresult
nsFastLoadFileWriter::WriteSlowID(const nsID& aID)
{
nsresult rv;
rv = Write32(aID.m0);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write16(aID.m1);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write16(aID.m2);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
PRUint32 bytesWritten;
rv = Write(NS_REINTERPRET_CAST(const char*, aID.m3), sizeof aID.m3,
&bytesWritten);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
if (bytesWritten != sizeof aID.m3)
return NS_ERROR_FAILURE;
return NS_OK;
}
nsresult
nsFastLoadFileWriter::WriteFastID(NSFastLoadID aID)
{
return Write32(aID ^ MFL_ID_XOR_KEY);
}
nsresult
nsFastLoadFileWriter::WriteSharpObjectInfo(const nsFastLoadSharpObjectInfo& aInfo)
{
nsresult rv;
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
NS_ASSERTION(aInfo.mCIDOffset != 0,
"fastload writer: mCIDOffset cannot be zero!");
rv = Write32(aInfo.mCIDOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write16(aInfo.mStrongRefCnt);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write16(aInfo.mWeakRefCnt);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
return NS_OK;
}
nsresult
nsFastLoadFileWriter::WriteMuxedDocumentInfo(const nsFastLoadMuxedDocumentInfo& aInfo)
{
nsresult rv;
rv = WriteStringZ(aInfo.mURISpec);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write32(aInfo.mInitialSegmentOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
return NS_OK;
}
PLDHashOperator PR_CALLBACK
nsFastLoadFileWriter::IDMapEnumerate(PLDHashTable *aTable,
PLDHashEntryHdr *aHdr,
PRUint32 aNumber,
void *aData)
{
nsIDMapEntry* entry = NS_STATIC_CAST(nsIDMapEntry*, aHdr);
PRUint32 index = entry->mFastID - 1;
nsID* vector = NS_REINTERPRET_CAST(nsID*, aData);
NS_ASSERTION(index < aTable->entryCount, "bad nsIDMap index!");
vector[index] = entry->mSlowID;
return PL_DHASH_NEXT;
}
struct nsSharpObjectMapEntry : public nsObjectMapEntry {
NSFastLoadOID mOID;
nsFastLoadSharpObjectInfo mInfo;
};
PLDHashOperator PR_CALLBACK
nsFastLoadFileWriter::ObjectMapEnumerate(PLDHashTable *aTable,
PLDHashEntryHdr *aHdr,
PRUint32 aNumber,
void *aData)
{
nsSharpObjectMapEntry* entry = NS_STATIC_CAST(nsSharpObjectMapEntry*, aHdr);
PRUint32 index = MFL_OID_TO_SHARP_INDEX(entry->mOID);
nsFastLoadSharpObjectInfo* vector =
NS_REINTERPRET_CAST(nsFastLoadSharpObjectInfo*, aData);
NS_ASSERTION(index < aTable->entryCount, "bad nsObjectMap index!");
vector[index] = entry->mInfo;
NS_ASSERTION(entry->mInfo.mStrongRefCnt, "no strong ref in serialization!");
// Ignore tagged object ids stored as object pointer keys (the updater
// code does this).
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if ((NS_PTR_TO_INT32(entry->mObject) & MFL_OBJECT_DEF_TAG) == 0)
NS_RELEASE(entry->mObject);
return PL_DHASH_NEXT;
}
PLDHashOperator PR_CALLBACK
nsFastLoadFileWriter::DocumentMapEnumerate(PLDHashTable *aTable,
PLDHashEntryHdr *aHdr,
PRUint32 aNumber,
void *aData)
{
nsFastLoadFileWriter* writer =
NS_REINTERPRET_CAST(nsFastLoadFileWriter*, aTable->data);
nsDocumentMapWriteEntry* entry =
NS_STATIC_CAST(nsDocumentMapWriteEntry*, aHdr);
nsresult* rvp = NS_REINTERPRET_CAST(nsresult*, aData);
nsFastLoadMuxedDocumentInfo info;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
info.mURISpec = entry->mString;
info.mInitialSegmentOffset = entry->mInitialSegmentOffset;
*rvp = writer->WriteMuxedDocumentInfo(info);
return NS_FAILED(*rvp) ? PL_DHASH_STOP : PL_DHASH_NEXT;
}
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
PLDHashOperator PR_CALLBACK
nsFastLoadFileWriter::DependencyMapEnumerate(PLDHashTable *aTable,
PLDHashEntryHdr *aHdr,
PRUint32 aNumber,
void *aData)
{
nsFastLoadFileWriter* writer =
NS_REINTERPRET_CAST(nsFastLoadFileWriter*, aTable->data);
nsDependencyMapEntry* entry = NS_STATIC_CAST(nsDependencyMapEntry*, aHdr);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsresult* rvp = NS_REINTERPRET_CAST(nsresult*, aData);
*rvp = writer->WriteStringZ(entry->mString);
if (NS_SUCCEEDED(*rvp))
*rvp = writer->Write64(entry->mLastModified);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
return NS_FAILED(*rvp) ? PL_DHASH_STOP :PL_DHASH_NEXT;
}
nsresult
nsFastLoadFileWriter::WriteFooter()
{
nsresult rv;
PRUint32 i, count;
nsFastLoadFooterPrefix footerPrefix;
footerPrefix.mNumIDs = mIDMap.entryCount;
footerPrefix.mNumSharpObjects = mObjectMap.entryCount;
footerPrefix.mNumMuxedDocuments = mDocumentMap.entryCount;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
footerPrefix.mNumDependencies = mDependencyMap.entryCount;
rv = WriteFooterPrefix(footerPrefix);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
// Enumerate mIDMap into a vector indexed by mFastID and write it.
nsID* idvec = new nsID[footerPrefix.mNumIDs];
if (!idvec)
return NS_ERROR_OUT_OF_MEMORY;
count = PL_DHashTableEnumerate(&mIDMap, IDMapEnumerate, idvec);
NS_ASSERTION(count == footerPrefix.mNumIDs, "bad mIDMap enumeration!");
for (i = 0; i < count; i++) {
rv = WriteSlowID(idvec[i]);
if (NS_FAILED(rv)) break;
}
delete[] idvec;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
// Enumerate mObjectMap into a vector indexed by mOID and write it.
nsFastLoadSharpObjectInfo* objvec =
new nsFastLoadSharpObjectInfo[footerPrefix.mNumSharpObjects];
if (!objvec)
return NS_ERROR_OUT_OF_MEMORY;
#ifdef NS_DEBUG
memset(objvec, 0, footerPrefix.mNumSharpObjects *
sizeof(nsFastLoadSharpObjectInfo));
#endif
count = PL_DHashTableEnumerate(&mObjectMap, ObjectMapEnumerate, objvec);
NS_ASSERTION(count == footerPrefix.mNumSharpObjects,
"bad mObjectMap enumeration!");
for (i = 0; i < count; i++) {
rv = WriteSharpObjectInfo(objvec[i]);
if (NS_FAILED(rv)) break;
}
delete[] objvec;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
// Enumerate mDocumentMap, writing nsFastLoadMuxedDocumentInfo records
count = PL_DHashTableEnumerate(&mDocumentMap, DocumentMapEnumerate, &rv);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
NS_ASSERTION(count == footerPrefix.mNumMuxedDocuments,
"bad mDocumentMap enumeration!");
// Write out make-like file dependencies.
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
count = PL_DHashTableEnumerate(&mDependencyMap, DependencyMapEnumerate, &rv);
if (NS_FAILED(rv))
return rv;
return NS_OK;
}
nsresult
nsFastLoadFileWriter::Init()
{
if (!PL_DHashTableInit(&mIDMap, &idmap_DHashTableOps, (void *)this,
sizeof(nsIDMapEntry), PL_DHASH_MIN_SIZE)) {
mIDMap.ops = nsnull;
return NS_ERROR_OUT_OF_MEMORY;
}
if (!PL_DHashTableInit(&mObjectMap, &objmap_DHashTableOps, (void *)this,
sizeof(nsSharpObjectMapEntry), PL_DHASH_MIN_SIZE)) {
mObjectMap.ops = nsnull;
return NS_ERROR_OUT_OF_MEMORY;
}
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (!PL_DHashTableInit(&mDocumentMap, &strmap_DHashTableOps, (void *)this,
sizeof(nsDocumentMapWriteEntry),
PL_DHASH_MIN_SIZE)) {
mDocumentMap.ops = nsnull;
return NS_ERROR_OUT_OF_MEMORY;
}
if (!PL_DHashTableInit(&mURIMap, &objmap_DHashTableOps, (void *)this,
sizeof(nsURIMapWriteEntry), PL_DHASH_MIN_SIZE)) {
mURIMap.ops = nsnull;
return NS_ERROR_OUT_OF_MEMORY;
}
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (!PL_DHashTableInit(&mDependencyMap, &strmap_DHashTableOps, (void *)this,
sizeof(nsDependencyMapEntry), PL_DHASH_MIN_SIZE)) {
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
mDependencyMap.ops = nsnull;
return NS_ERROR_OUT_OF_MEMORY;
}
return NS_OK;
}
nsresult
nsFastLoadFileWriter::Open()
{
nsresult rv;
rv = mSeekableOutput->Seek(nsISeekableStream::NS_SEEK_SET,
sizeof(nsFastLoadHeader));
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
return Init();
}
NS_IMETHODIMP
nsFastLoadFileWriter::Close()
{
nsresult rv;
memcpy(mHeader.mMagic, magic, MFL_FILE_MAGIC_SIZE);
mHeader.mChecksum = 0;
mHeader.mVersion = MFL_FILE_VERSION;
PRInt64 footerOffset;
rv = mSeekableOutput->Tell(&footerOffset);
LL_L2UI(mHeader.mFooterOffset, footerOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
// If there is a muxed document segment open, close it now by setting its
// length, stored in the second PRUint32 of the segment.
if (mCurrentDocumentMapEntry) {
PRUint32 currentSegmentOffset =
mCurrentDocumentMapEntry->mCurrentSegmentOffset;
rv = mSeekableOutput->Seek(nsISeekableStream::NS_SEEK_SET,
currentSegmentOffset + 4);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write32(mHeader.mFooterOffset - currentSegmentOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
// Seek back to the current offset to write the footer.
rv = mSeekableOutput->Seek(nsISeekableStream::NS_SEEK_SET,
mHeader.mFooterOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
mCurrentDocumentMapEntry = nsnull;
}
rv = WriteFooter();
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
PRInt64 fileSize;
rv = mSeekableOutput->Tell(&fileSize);
LL_L2UI(mHeader.mFileSize, fileSize);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = mSeekableOutput->Seek(nsISeekableStream::NS_SEEK_SET, 0);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = WriteHeader(&mHeader);
if (NS_FAILED(rv))
return rv;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
// Now compute the checksum, using mFileIO to get an input stream on the
// underlying FastLoad file.
if (mFileIO) {
// Get the unbuffered output stream, which flushes the buffered header
// so we can read and checksum it along with the rest of the file, and
// which allows us to write the checksum directly.
nsCOMPtr<nsIOutputStream> output;
rv = mBufferAccess->GetUnbufferedStream(getter_AddRefs(output));
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv) || !output)
return NS_ERROR_UNEXPECTED;
nsCOMPtr<nsIInputStream> input;
rv = mFileIO->GetInputStream(getter_AddRefs(input));
if (NS_FAILED(rv))
return rv;
// Get the unbuffered input stream, to avoid copying overhead and to
// keep our view of the file coherent with the writer -- we don't want
// to hit a stale buffer in the reader's underlying stream.
nsCOMPtr<nsIStreamBufferAccess> bufferAccess =
do_QueryInterface(input);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
rv = bufferAccess->GetUnbufferedStream(getter_AddRefs(input));
if (NS_FAILED(rv) || !input)
return NS_ERROR_UNEXPECTED;
// Seek the input stream to offset 0, in case it's a reader who has
// already been used to consume some of the FastLoad file.
nsCOMPtr<nsISeekableStream> seekable = do_QueryInterface(input);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
rv = seekable->Seek(nsISeekableStream::NS_SEEK_SET, 0);
if (NS_FAILED(rv))
return rv;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
char buf[MFL_CHECKSUM_BUFSIZE];
PRUint32 len, rem = 0;
PRUint32 checksum = 0;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
// Ok, we're finally ready to checksum the FastLoad file we just wrote!
Reduce seeks outside the FastLoad file's underlying stream buffer (195010, r=ben, sr=bryner). - Pass null scope objects into nsIScriptContext::Compile{EventHandler,Script} when precompiling for brutal sharing. The JS engine does not need a non-null object parameter for static scope. That parameter can be non-null only if it helps the compiler "pre-bind" functions to the same scope object that they'll be parented by when executing, but with brutal sharing, functions are precompiled once and executed against many different scope objects. - A problem in XUL FastLoad was that it would serialize master .xul document out-of-line scripts (those included via script src= from non-overlay, "master" .xul docs) far from the place in the FastLoad file where XUL prototype script info was serialized for the <script src=> tag itself. I fixed that so that, unless the out-of-line script was previously serialized (by a different src= reference from another .xul file), the OOL script data immediately follows the proto-script info in the FastLoad file. This required adding a SerializeOutOfLine method to nsXULPrototypeScript, which restores symmetry by matching the existing DeserializeOutOfLine(Script) (note: I dropped the redundant "Script" from the end of the latter method's name. We need SerializeOutOfLine to handle overlay OOL scripts. They are serialized by nsXULDocument::OnStreamComplete, because that code knows the difference between an overlay and a master doc. This removes all trace of FastLoad writing from nsXULPrototypeScript::Compile -- FastLoad stuff didn't belong there, not only because we now want to write master OOL scripts later, when walking the master XUL prototype doc's element tree, but also for modularity reasons. The caller knows about FastLoad, nsXULPrototypeScript::Compile does just what its name implies. There are 132 seeks with the patch, only 49 of which dump the underlying file stream's buffer, vs. 133 seeks without the patch, 87 of which dump the buffer. - Nit-picked some comments to fit in 80 columns, and made other cosmetic fixes. - Implicated the nsXULDocument::mIsWritingFastLoad flag from the useXULCache "is the XUL cache enabled?" flag, so other places that test mIsWritingFastLoad don't have to query whether the cache is enabled. - Added METERING synchronous meter-dumping to /tmp/bufstats, only ifdef DEBUG_brendan, in netwerk/base/src/nsBufferedStreams.cpp. - Added the deferred seek optimization from the first patch in bug 195010 to nsFastLoadFile.cpp. - Fixed nsFastLoadFileReader so it overrides readSegments as well as read. This catches up with the interface extension made to nsIInputStream to add readSegments. The nsFastLoadFileReader extends nsBinaryInputStream, which is a concrete class, and overrides Read in order to snoop on *all* bytes read. It does this in order to demultiplex documents interleaved when the FastLoad file was written. But since the readSegments move into nsIInputStream.idl, certain primitives in nsBinaryStream.cpp, e.g., nsBinaryInputStream::ReadCString, have used ReadSegments, not Read, to consume parts of the underlying stream (to read the C string's chars, in that example), and the FastLoad file implementation has not accounted for those bytes. - Added a new method to nsIFastLoadFileControl and its impls: hasMuxedDocument. This is needed when serializing master XUL doc OOL scripts, because we must not serialize twice, and any OOL script that other XUL docs can include via script src= could already be in the FastLoad mux. /be
2003-04-12 05:34:11 +04:00
while (NS_SUCCEEDED(rv =
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
input->Read(buf + rem, sizeof buf - rem, &len)) &&
len) {
len += rem;
rem = NS_AccumulateFastLoadChecksum(&checksum,
NS_REINTERPRET_CAST(PRUint8*,
buf),
len,
PR_FALSE);
if (rem)
memcpy(buf, buf + len - rem, rem);
}
if (NS_FAILED(rv))
return rv;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (rem) {
NS_AccumulateFastLoadChecksum(&checksum,
NS_REINTERPRET_CAST(PRUint8*, buf),
rem,
PR_TRUE);
}
// Store the checksum in the FastLoad file header and remember it via
// mHeader.mChecksum, for GetChecksum.
seekable = do_QueryInterface(output);
rv = seekable->Seek(nsISeekableStream::NS_SEEK_SET,
offsetof(nsFastLoadHeader, mChecksum));
if (NS_FAILED(rv))
return rv;
mHeader.mChecksum = checksum;
checksum = NS_SWAP32(checksum);
PRUint32 bytesWritten;
rv = output->Write(NS_REINTERPRET_CAST(char*, &checksum),
sizeof checksum,
&bytesWritten);
if (NS_FAILED(rv))
return rv;
if (bytesWritten != sizeof checksum)
return NS_ERROR_FAILURE;
}
return mOutputStream->Close();
}
// Pseudo-tag used as flag between WriteSingleRefObject and WriteObjectCommon.
#define MFL_SINGLE_REF_PSEUDO_TAG PR_BIT(MFL_OBJECT_TAG_BITS)
nsresult
nsFastLoadFileWriter::WriteObjectCommon(nsISupports* aObject,
PRBool aIsStrongRef,
PRUint32 aTags)
{
nsrefcnt rc;
nsresult rv;
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
NS_ASSERTION((NS_PTR_TO_INT32(aObject) & MFL_OBJECT_DEF_TAG) == 0,
"odd nsISupports*, oh no!");
// Here be manual refcounting dragons!
rc = aObject->AddRef();
NS_ASSERTION(rc != 0, "bad refcnt when writing aObject!");
NSFastLoadOID oid;
nsCOMPtr<nsIClassInfo> classInfo;
if (rc == 2 && (aTags & MFL_SINGLE_REF_PSEUDO_TAG)) {
// Dull object: only one strong ref and no weak refs in serialization.
// Conservative: we don't trust the caller if there are more than two
// refs (one from the AddRef above, one from the data structure that's
// being serialized).
oid = MFL_DULL_OBJECT_OID;
aObject->Release();
} else {
// Object is presumed to be multiply connected through some combo of
// strong and weak refs. Hold onto it via mObjectMap.
nsSharpObjectMapEntry* entry =
NS_STATIC_CAST(nsSharpObjectMapEntry*,
PL_DHashTableOperate(&mObjectMap, aObject,
PL_DHASH_ADD));
if (!entry) {
aObject->Release();
return NS_ERROR_OUT_OF_MEMORY;
}
if (!entry->mObject) {
// First time we've seen this object address: add it to mObjectMap
// and serialize the object at the current stream offset.
PRInt64 thisOffset;
rv = Tell(&thisOffset);
if (NS_FAILED(rv)) {
aObject->Release();
return rv;
}
// NB: aObject was already held, and mObject is a raw nsISupports*.
entry->mObject = aObject;
oid = (mObjectMap.entryCount << MFL_OBJECT_TAG_BITS);
entry->mOID = oid;
// NB: the (32-bit, fast) CID and object data follow the OID.
entry->mInfo.mCIDOffset = thisOffset + sizeof(oid);
entry->mInfo.mStrongRefCnt = aIsStrongRef ? 1 : 0;
entry->mInfo.mWeakRefCnt = aIsStrongRef ? 0 : 1;
// Record in oid the fact that we're defining this object in the
// stream, and get the object's class info here, so we can take
// note of singletons in order to avoid reserializing them when
// updating after reading.
oid |= MFL_OBJECT_DEF_TAG;
classInfo = do_QueryInterface(aObject);
if (!classInfo) {
NS_NOTREACHED("aObject must implement nsIClassInfo");
return NS_ERROR_FAILURE;
}
PRUint32 flags;
if (NS_SUCCEEDED(classInfo->GetFlags(&flags)) &&
(flags & nsIClassInfo::SINGLETON)) {
MFL_SET_SINGLETON_FLAG(&entry->mInfo);
}
} else {
// Already serialized, recover oid and update the desired refcnt.
oid = entry->mOID;
if (aIsStrongRef) {
++entry->mInfo.mStrongRefCnt;
NS_ASSERTION(entry->mInfo.mStrongRefCnt != 0,
"mStrongRefCnt overflow");
} else {
MFL_BUMP_WEAK_REFCNT(&entry->mInfo);
NS_ASSERTION(MFL_GET_WEAK_REFCNT(&entry->mInfo) != 0,
"mWeakRefCnt overflow");
}
aObject->Release();
}
}
if (!aIsStrongRef)
oid |= MFL_WEAK_REF_TAG;
oid |= (aTags & MFL_QUERY_INTERFACE_TAG);
rv = Write32(oid ^ MFL_OID_XOR_KEY);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
if (oid & MFL_OBJECT_DEF_TAG) {
nsCOMPtr<nsISerializable> serializable(do_QueryInterface(aObject));
if (!serializable) {
NS_NOTREACHED("aObject must implement nsISerializable");
return NS_ERROR_FAILURE;
}
nsCID slowCID;
rv = classInfo->GetClassIDNoAlloc(&slowCID);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
NSFastLoadID fastCID;
rv = MapID(slowCID, &fastCID);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = WriteFastID(fastCID);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = serializable->Write(this);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
}
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileWriter::WriteObject(nsISupports* aObject, PRBool aIsStrongRef)
{
#ifdef NS_DEBUG
nsCOMPtr<nsISupports> rootObject(do_QueryInterface(aObject));
NS_ASSERTION(rootObject.get() == aObject,
"bad call to WriteObject -- call WriteCompoundObject!");
#endif
return WriteObjectCommon(aObject, aIsStrongRef, 0);
}
NS_IMETHODIMP
nsFastLoadFileWriter::WriteSingleRefObject(nsISupports* aObject)
{
#ifdef NS_DEBUG
nsCOMPtr<nsISupports> rootObject(do_QueryInterface(aObject));
NS_ASSERTION(rootObject.get() == aObject,
"bad call to WriteSingleRefObject -- call WriteCompoundObject!");
#endif
return WriteObjectCommon(aObject, PR_TRUE, MFL_SINGLE_REF_PSEUDO_TAG);
}
NS_IMETHODIMP
nsFastLoadFileWriter::WriteCompoundObject(nsISupports* aObject,
const nsIID& aIID,
PRBool aIsStrongRef)
{
nsresult rv;
nsCOMPtr<nsISupports> rootObject(do_QueryInterface(aObject));
// We could assert that |rootObject != aObject|, but that would prevent
// callers who don't know whether they're dealing with the primary
// nsISupports pointer (e.g., they don't know which implementation of
// nsIURI they have) from using this function.
#ifdef NS_DEBUG
nsCOMPtr<nsISupports> roundtrip;
rootObject->QueryInterface(aIID, getter_AddRefs(roundtrip));
NS_ASSERTION(roundtrip.get() == aObject,
"bad aggregation or multiple inheritance detected by call to "
"WriteCompoundObject!");
#endif
rv = WriteObjectCommon(rootObject, aIsStrongRef, MFL_QUERY_INTERFACE_TAG);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
NSFastLoadID iid;
rv = MapID(aIID, &iid);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
return WriteFastID(iid);
}
NS_IMETHODIMP
nsFastLoadFileWriter::WriteID(const nsID& aID)
{
nsresult rv;
NSFastLoadID fastID;
rv = MapID(aID, &fastID);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
return WriteFastID(fastID);
}
NS_IMETHODIMP
nsFastLoadFileWriter::Seek(PRInt32 aWhence, PRInt64 aOffset)
{
mCurrentDocumentMapEntry = nsnull;
return mSeekableOutput->Seek(aWhence, aOffset);
}
NS_IMETHODIMP
nsFastLoadFileWriter::Tell(PRInt64 *aResult)
{
return mSeekableOutput->Tell(aResult);
}
NS_IMETHODIMP
nsFastLoadFileWriter::SetEOF()
{
return mSeekableOutput->SetEOF();
}
NS_IMETHODIMP
nsFastLoadFileWriter::SetOutputStream(nsIOutputStream *aStream)
{
nsresult rv = nsBinaryOutputStream::SetOutputStream(aStream);
mSeekableOutput = do_QueryInterface(mOutputStream);
return rv;
}
NS_COM nsresult
NS_NewFastLoadFileWriter(nsIObjectOutputStream* *aResult,
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsIOutputStream* aDestStream,
nsIFastLoadFileIO* aFileIO)
{
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsFastLoadFileWriter* writer =
new nsFastLoadFileWriter(aDestStream, aFileIO);
if (!writer)
return NS_ERROR_OUT_OF_MEMORY;
// Stabilize writer's refcnt.
nsCOMPtr<nsIObjectOutputStream> stream(writer);
nsresult rv = writer->Open();
if (NS_FAILED(rv))
return rv;
*aResult = stream;
NS_ADDREF(*aResult);
return NS_OK;
}
// -------------------------- nsFastLoadFileUpdater --------------------------
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
NS_IMPL_ISUPPORTS_INHERITED1(nsFastLoadFileUpdater,
nsFastLoadFileWriter,
nsIFastLoadFileIO)
NS_IMETHODIMP
nsFastLoadFileUpdater::GetInputStream(nsIInputStream** aResult)
{
*aResult = mInputStream;
NS_IF_ADDREF(*aResult);
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileUpdater::GetOutputStream(nsIOutputStream** aResult)
{
*aResult = nsnull;
return NS_OK;
}
PLDHashOperator PR_CALLBACK
nsFastLoadFileUpdater::CopyReadDocumentMapEntryToUpdater(PLDHashTable *aTable,
PLDHashEntryHdr *aHdr,
PRUint32 aNumber,
void *aData)
{
nsDocumentMapReadEntry* readEntry =
NS_STATIC_CAST(nsDocumentMapReadEntry*, aHdr);
nsFastLoadFileUpdater* updater =
NS_REINTERPRET_CAST(nsFastLoadFileUpdater*, aData);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
void* spec = nsMemory::Clone(readEntry->mString,
strlen(readEntry->mString) + 1);
if (!spec)
return PL_DHASH_STOP;
nsDocumentMapWriteEntry* writeEntry =
NS_STATIC_CAST(nsDocumentMapWriteEntry*,
PL_DHashTableOperate(&updater->mDocumentMap, spec,
PL_DHASH_ADD));
if (!writeEntry) {
nsMemory::Free(spec);
return PL_DHASH_STOP;
}
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
writeEntry->mString = NS_REINTERPRET_CAST(const char*, spec);
writeEntry->mURI = nsnull;
writeEntry->mInitialSegmentOffset = readEntry->mInitialSegmentOffset;
writeEntry->mCurrentSegmentOffset = 0;
return PL_DHASH_NEXT;
}
nsresult
nsFastLoadFileUpdater::Open(nsFastLoadFileReader* aReader)
{
nsresult rv;
rv = nsFastLoadFileWriter::Init();
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
PRUint32 i, n;
// Map from dense, zero-based, uint32 NSFastLoadID in reader to 16-byte
// nsID in updater.
nsID* readIDMap = aReader->mFooter.mIDMap;
for (i = 0, n = aReader->mFooter.mNumIDs; i < n; i++) {
NSFastLoadID fastID;
rv = MapID(readIDMap[i], &fastID);
NS_ASSERTION(fastID == i + 1, "huh?");
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
}
// Map from reader dense, zero-based MFL_OID_TO_SHARP_INDEX(oid) to sharp
// object offset and refcnt information in updater.
nsFastLoadFileReader::nsObjectMapEntry* readObjectMap =
aReader->mFooter.mObjectMap;
// Prepare to save aReader state in case we need to seek back and read a
// singleton object that might otherwise get written by this updater.
nsDocumentMapReadEntry* saveDocMapEntry = nsnull;
nsISeekableStream* inputSeekable = nsnull;
PRInt64 saveOffset = 0;
for (i = 0, n = aReader->mFooter.mNumSharpObjects; i < n; i++) {
nsFastLoadFileReader::nsObjectMapEntry* readEntry = &readObjectMap[i];
NS_ASSERTION(readEntry->mCIDOffset != 0,
"fastload updater: mCIDOffset cannot be zero!");
// If the reader didn't read this object but it's a singleton, we must
// "deserialize" it now, to discover its one and only root nsISupports
// address. The object already exists in memory if it was created at
// startup without resort to the FastLoad file. The canonical example
// is the system principal object held by all XUL JS scripts.
nsISupports* obj = readEntry->mReadObject;
if (!obj && MFL_GET_SINGLETON_FLAG(readEntry)) {
if (!saveDocMapEntry) {
inputSeekable = aReader->mSeekableInput;
rv = inputSeekable->Tell(&saveOffset);
if (NS_FAILED(rv))
return rv;
saveDocMapEntry = aReader->mCurrentDocumentMapEntry;
aReader->mCurrentDocumentMapEntry = nsnull;
}
rv = inputSeekable->Seek(nsISeekableStream::NS_SEEK_SET,
readEntry->mCIDOffset);
if (NS_FAILED(rv))
return rv;
rv = aReader
->DeserializeObject(getter_AddRefs(readEntry->mReadObject));
if (NS_FAILED(rv))
return rv;
obj = readEntry->mReadObject;
// Don't forget to set mSkipOffset in case someone calls the reader
// to "deserialize" (yet again) the object we just read.
//
// Say the singleton is the system principal, and the FastLoad file
// contains data for navigator.xul including scripts and functions.
// If we update the FastLoad file to contain data for messenger.xul
// in a separate session started via mozilla -mail, *and during the
// same FastLoad episode in this session* race to open a navigator
// window, we will attempt to read all objects serialized in the
// navigator.xul portion of the FastLoad file.
//
// mSkipOffset must be set in such a case so the reader can skip
// the system principal's serialized data, because the updater for
// messenger.xul being opened here has already read it.
rv = inputSeekable->Tell(&readEntry->mSkipOffset);
if (NS_FAILED(rv))
return rv;
}
NSFastLoadOID oid = MFL_SHARP_INDEX_TO_OID(i);
void* key = obj
? NS_REINTERPRET_CAST(void*, obj)
: NS_REINTERPRET_CAST(void*, (oid | MFL_OBJECT_DEF_TAG));
nsSharpObjectMapEntry* writeEntry =
NS_STATIC_CAST(nsSharpObjectMapEntry*,
PL_DHashTableOperate(&mObjectMap, key,
PL_DHASH_ADD));
if (!writeEntry)
return NS_ERROR_OUT_OF_MEMORY;
// Hold the object if there is one, so that objmap_ClearEntry can
// release the reference.
NS_IF_ADDREF(obj);
writeEntry->mObject = NS_REINTERPRET_CAST(nsISupports*, key);
writeEntry->mOID = oid;
writeEntry->mInfo.mCIDOffset = readEntry->mCIDOffset;
writeEntry->mInfo.mStrongRefCnt = readEntry->mSaveStrongRefCnt;
writeEntry->mInfo.mWeakRefCnt = readEntry->mSaveWeakRefCnt;
}
// If we had to read any singletons, restore aReader's saved state.
if (saveDocMapEntry) {
rv = inputSeekable->Seek(nsISeekableStream::NS_SEEK_SET, saveOffset);
if (NS_FAILED(rv))
return rv;
aReader->mCurrentDocumentMapEntry = saveDocMapEntry;
}
// Copy URI spec string and initial segment offset in FastLoad file from
// nsDocumentMapReadEntry in reader to nsDocumentMapWriteEntry in updater.
// If we didn't enumerate all entries, we ran out of memory.
n = PL_DHashTableEnumerate(&aReader->mFooter.mDocumentMap,
CopyReadDocumentMapEntryToUpdater,
this);
if (n != aReader->mFooter.mDocumentMap.entryCount)
return NS_ERROR_OUT_OF_MEMORY;
// Copy source filename dependencies from reader to updater.
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsISupportsArray* readDeps = aReader->mFooter.mDependencies;
rv = readDeps->Count(&n);
if (NS_FAILED(rv))
return rv;
for (i = 0; i < n; i++) {
nsCOMPtr<nsIFile> file;
rv = readDeps->GetElementAt(i, getter_AddRefs(file));
if (NS_FAILED(rv))
return rv;
rv = AddDependency(file);
if (NS_FAILED(rv))
return rv;
}
// Seek to the reader's footer offset so we overwrite the footer. First,
// update the header to have a zero mFooterOffset, which will invalidate
// the FastLoad file on next startup read attempt, should we crash before
// completing this update.
rv = mSeekableOutput->Seek(nsISeekableStream::NS_SEEK_SET,
offsetof(nsFastLoadHeader, mFooterOffset));
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = Write32(0);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
rv = mSeekableOutput->Seek(nsISeekableStream::NS_SEEK_SET,
aReader->mHeader.mFooterOffset);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
if (NS_FAILED(rv))
return rv;
// Avoid creating yet another object by implementing nsIFastLoadFileIO on
// this updater, and save aReader's input stream so it can be returned by
// GetInputStream called from nsFastLoadFileWriter::Close. This requires
// that we override Close to break the resulting zero-length cycle.
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
mFileIO = this;
mInputStream = aReader->mInputStream;
mSeekableInput = aReader->mSeekableInput;
return NS_OK;
}
NS_IMETHODIMP
nsFastLoadFileUpdater::Close()
{
// Call base-class Close implementation, which uses mFileIO.
nsresult rv = nsFastLoadFileWriter::Close();
// Break degenerate cycle from this->mFileIO to this.
mFileIO = nsnull;
return rv;
}
NS_COM nsresult
NS_NewFastLoadFileUpdater(nsIObjectOutputStream* *aResult,
nsIOutputStream* aOutputStream,
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsIObjectInputStream* aReaderAsStream)
{
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
// Make sure that aReaderAsStream is an nsFastLoadFileReader.
nsCOMPtr<nsIFastLoadFileReader> reader(do_QueryInterface(aReaderAsStream));
if (!reader)
return NS_ERROR_UNEXPECTED;
nsFastLoadFileUpdater* updater = new nsFastLoadFileUpdater(aOutputStream);
if (!updater)
return NS_ERROR_OUT_OF_MEMORY;
// Stabilize updater's refcnt.
nsCOMPtr<nsIObjectOutputStream> stream(updater);
Bit checkin for bug 68045, r/sr=waterson&shaver, second attempt. It all works for me on optimized and debug gcc2.96, rh7.1. - Better failure codes from nsXULPrototypeScript::Deserialize. - Call nsXULDocument::AbortFastLoads after nsXULPrototypeScript::Serialize failure, instead of just nulling the FastLoad service's output stream. - Expose nsXULDocument::AbortFastLoads via nsIXULPrototypeCache, for use from nsChromeProtocolHandler.cpp. AbortFastLoads flushes the XUL cache now, for good measure. - The needless "Current" adjective in nsIFastLoadService attribute and method names is no more. - Add a do_GetFastLoadService() helper, to use CID instead of contractid, and to let the compiler consolidate the static inline CID. - Add "nglayout.debug.checksum_xul_fastload_file" pref so people can do without the checksum verification step when reading a FastLoad file. - Verify the FastLoad file checksum, by default. Also, cache it in the FastLoad service so we don't recompute it when re-opening the FastLoad file as mailnews and other top-levels start up. Fill the checksum cache in EndFastLoad, when the last pseudo-concurrent top-level finishes loading. My hope to compute the checksum while writing the FastLoad file ran afoul of misordered writes. The old code to checksum the in-memory nsFastLoadHeader also was broken on little endian platforms. Now all checksumming is done via a separate read pass over the complete file, save for the header's checksum field, which is summed as if it contained zero. - Track and check FastLoad file dependencies. This required groveling with a bunch of Necko interfaces in nsChromeProtocolHandler::NewChannel -- read it and weep. Dependency checking, as well as checksum access and computation, use better-factored nsIFastLoad{File,Read,Write}Control interfaces. - nsBufferedStream::Seek wasn't flushing the buffer when seeking backward within the buffer, but it must, because mCursor bounds the amount to write if the buffer contains the end of file. - Add an unbufferedStream readonly attribute to nsIStreamBufferAccess, so we don't have to screw around with the bufferying layer when checksumming. Also implement nsIStreamBufferAccess in nsBufferedOutputStream. - nsISeekableOutputStream was bogus, based on a bad state I had put the nsBufferedOutputStream code in on its way from being completely broken when you seek backwards outside of the buffer. Removing this interface required using nsIFastLoadFileIO in nsFastLoadFileWriter, and it also required careful ordering of Close calls (the Reader must close after the Writer or Updater, so that the Reader's underlying, unbuffered input stream can be read by nsFastLoadFileWriter::Close to compute the checksum. - Miscellaneous tab/indentation, comment typo, bracing, if( => if ( style, nsnull vs. 0, useless variable elimination, tortured control flow, AutoString instead of String, and gratuitous ; after nsISupportsUtils.h macro call cleanups.
2001-08-22 00:51:34 +04:00
nsresult rv = updater->Open(NS_STATIC_CAST(nsFastLoadFileReader*,
aReaderAsStream));
if (NS_FAILED(rv))
return rv;
*aResult = stream;
NS_ADDREF(*aResult);
return NS_OK;
}