gecko-dev/widget/xremoteclient/XRemoteClient.cpp

743 строки
21 KiB
C++
Исходник Обычный вид История

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim:expandtab:shiftwidth=2:tabstop=8:
*/
/* vim:set ts=8 sw=2 et cindent: */
2012-05-21 15:12:37 +04:00
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#include "mozilla/ArrayUtils.h"
#include "mozilla/IntegerPrintfMacros.h"
#include "mozilla/Sprintf.h"
#include "mozilla/Unused.h"
#include "XRemoteClient.h"
#include "RemoteUtils.h"
#include "plstr.h"
#include "prsystem.h"
#include "mozilla/Logging.h"
#include "prenv.h"
#include "prdtoa.h"
#include <stdlib.h>
#include <unistd.h>
#include <string.h>
#include <strings.h>
#include <sys/time.h>
#include <sys/types.h>
#include <unistd.h>
#include <limits.h>
#include <X11/Xatom.h>
#define MOZILLA_VERSION_PROP "_MOZILLA_VERSION"
#define MOZILLA_LOCK_PROP "_MOZILLA_LOCK"
#define MOZILLA_COMMANDLINE_PROP "_MOZILLA_COMMANDLINE"
#define MOZILLA_RESPONSE_PROP "_MOZILLA_RESPONSE"
#define MOZILLA_USER_PROP "_MOZILLA_USER"
#define MOZILLA_PROFILE_PROP "_MOZILLA_PROFILE"
#define MOZILLA_PROGRAM_PROP "_MOZILLA_PROGRAM"
#ifdef IS_BIG_ENDIAN
#define TO_LITTLE_ENDIAN32(x) \
((((x) & 0xff000000) >> 24) | (((x) & 0x00ff0000) >> 8) | \
(((x) & 0x0000ff00) << 8) | (((x) & 0x000000ff) << 24))
#else
#define TO_LITTLE_ENDIAN32(x) (x)
#endif
#ifndef MAX_PATH
#ifdef PATH_MAX
#define MAX_PATH PATH_MAX
#else
#define MAX_PATH 1024
#endif
#endif
using mozilla::LogLevel;
using mozilla::Unused;
static mozilla::LazyLogModule sRemoteLm("XRemoteClient");
static int (*sOldHandler)(Display *, XErrorEvent *);
static bool sGotBadWindow;
XRemoteClient::XRemoteClient()
{
mDisplay = 0;
mInitialized = false;
mMozVersionAtom = 0;
mMozLockAtom = 0;
mMozCommandLineAtom = 0;
mMozResponseAtom = 0;
mMozWMStateAtom = 0;
mMozUserAtom = 0;
mMozProfileAtom = 0;
mMozProgramAtom = 0;
mLockData = 0;
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("XRemoteClient::XRemoteClient"));
}
XRemoteClient::~XRemoteClient()
{
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("XRemoteClient::~XRemoteClient"));
if (mInitialized)
Shutdown();
}
// Minimize the roundtrips to the X-server
static const char *XAtomNames[] = {
MOZILLA_VERSION_PROP,
MOZILLA_LOCK_PROP,
MOZILLA_RESPONSE_PROP,
"WM_STATE",
MOZILLA_USER_PROP,
MOZILLA_PROFILE_PROP,
MOZILLA_PROGRAM_PROP,
MOZILLA_COMMANDLINE_PROP
};
static Atom XAtoms[MOZ_ARRAY_LENGTH(XAtomNames)];
nsresult
XRemoteClient::Init()
{
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("XRemoteClient::Init"));
if (mInitialized)
return NS_OK;
// try to open the display
mDisplay = XOpenDisplay(0);
if (!mDisplay)
return NS_ERROR_FAILURE;
// get our atoms
XInternAtoms(mDisplay, const_cast<char**>(XAtomNames),
MOZ_ARRAY_LENGTH(XAtomNames), False, XAtoms);
int i = 0;
mMozVersionAtom = XAtoms[i++];
mMozLockAtom = XAtoms[i++];
mMozResponseAtom = XAtoms[i++];
mMozWMStateAtom = XAtoms[i++];
mMozUserAtom = XAtoms[i++];
mMozProfileAtom = XAtoms[i++];
mMozProgramAtom = XAtoms[i++];
mMozCommandLineAtom = XAtoms[i];
mInitialized = true;
return NS_OK;
}
void
XRemoteClient::Shutdown (void)
{
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("XRemoteClient::Shutdown"));
if (!mInitialized)
return;
// shut everything down
XCloseDisplay(mDisplay);
mDisplay = 0;
mInitialized = false;
if (mLockData) {
free(mLockData);
mLockData = 0;
}
}
static int
HandleBadWindow(Display *display, XErrorEvent *event)
{
if (event->error_code == BadWindow) {
sGotBadWindow = true;
return 0; // ignored
}
return (*sOldHandler)(display, event);
}
nsresult
Bug 1080319 - Remove the -remote option. r=bsmedberg The -remote option has existed essentially forever, but its usefulness is questionable: - It requires a running instance to be any useful, so any script actually using it should first do -remote 'ping()' and handle the response properly. - It is not cross-application. The remote service dispatches the -remote commands to the command line handler, and, for example, desktop b2g builds don't have handlers for -remote (although thunderbird and seamonkey do). - It is not a cross-platform option, which leads to the following point: - There are other command line ways to do the same thing (at least in Firefox), without having to jump through hoops with -remote 'ping()', because there are command line options to do those same things on non-X11 platforms. For the latter, in Firefox case: - -remote 'openURL(url)' can be replaced with firefox url - -remote 'openURL(url,new-tab)' can be replaced with firefox -new-tab url - -remote 'openURL(url,new-window)' can be replaced with firefox -new-window url - -remote 'openfile(file,...)' is the same as -remote 'openurl(file,...) so, can be replaced as above - -remote 'xfedocommand(openbrowser)' is inherited from the mozilla suite and doesn't make much sense, but can be replaced with firefox -new-window The interesting part is that without changing nsBrowserContentHandler.js, -remote still works, meaning that if people really feel strongly about -remote, they'll still be able to write an addon to bring it back. This also means this patch actually doesn't remove -remote for applications other than Firefox that do support it, although -remote 'ping()' doesn't work as expected. However, other -remote commands will now work even without a running instance.
2014-10-14 02:19:52 +04:00
XRemoteClient::SendCommandLine (const char *aProgram, const char *aUsername,
const char *aProfile,
int32_t argc, char **argv,
const char* aDesktopStartupID,
char **aResponse, bool *aWindowFound)
{
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("XRemoteClient::SendCommandLine"));
Bug 1080319 - Remove the -remote option. r=bsmedberg The -remote option has existed essentially forever, but its usefulness is questionable: - It requires a running instance to be any useful, so any script actually using it should first do -remote 'ping()' and handle the response properly. - It is not cross-application. The remote service dispatches the -remote commands to the command line handler, and, for example, desktop b2g builds don't have handlers for -remote (although thunderbird and seamonkey do). - It is not a cross-platform option, which leads to the following point: - There are other command line ways to do the same thing (at least in Firefox), without having to jump through hoops with -remote 'ping()', because there are command line options to do those same things on non-X11 platforms. For the latter, in Firefox case: - -remote 'openURL(url)' can be replaced with firefox url - -remote 'openURL(url,new-tab)' can be replaced with firefox -new-tab url - -remote 'openURL(url,new-window)' can be replaced with firefox -new-window url - -remote 'openfile(file,...)' is the same as -remote 'openurl(file,...) so, can be replaced as above - -remote 'xfedocommand(openbrowser)' is inherited from the mozilla suite and doesn't make much sense, but can be replaced with firefox -new-window The interesting part is that without changing nsBrowserContentHandler.js, -remote still works, meaning that if people really feel strongly about -remote, they'll still be able to write an addon to bring it back. This also means this patch actually doesn't remove -remote for applications other than Firefox that do support it, although -remote 'ping()' doesn't work as expected. However, other -remote commands will now work even without a running instance.
2014-10-14 02:19:52 +04:00
*aWindowFound = false;
// FindBestWindow() iterates down the window hierarchy, so catch X errors
// when windows get destroyed before being accessed.
sOldHandler = XSetErrorHandler(HandleBadWindow);
Bug 1080319 - Remove the -remote option. r=bsmedberg The -remote option has existed essentially forever, but its usefulness is questionable: - It requires a running instance to be any useful, so any script actually using it should first do -remote 'ping()' and handle the response properly. - It is not cross-application. The remote service dispatches the -remote commands to the command line handler, and, for example, desktop b2g builds don't have handlers for -remote (although thunderbird and seamonkey do). - It is not a cross-platform option, which leads to the following point: - There are other command line ways to do the same thing (at least in Firefox), without having to jump through hoops with -remote 'ping()', because there are command line options to do those same things on non-X11 platforms. For the latter, in Firefox case: - -remote 'openURL(url)' can be replaced with firefox url - -remote 'openURL(url,new-tab)' can be replaced with firefox -new-tab url - -remote 'openURL(url,new-window)' can be replaced with firefox -new-window url - -remote 'openfile(file,...)' is the same as -remote 'openurl(file,...) so, can be replaced as above - -remote 'xfedocommand(openbrowser)' is inherited from the mozilla suite and doesn't make much sense, but can be replaced with firefox -new-window The interesting part is that without changing nsBrowserContentHandler.js, -remote still works, meaning that if people really feel strongly about -remote, they'll still be able to write an addon to bring it back. This also means this patch actually doesn't remove -remote for applications other than Firefox that do support it, although -remote 'ping()' doesn't work as expected. However, other -remote commands will now work even without a running instance.
2014-10-14 02:19:52 +04:00
Window w = FindBestWindow(aProgram, aUsername, aProfile);
nsresult rv = NS_OK;
if (w) {
// ok, let the caller know that we at least found a window.
*aWindowFound = true;
// Ignore BadWindow errors up to this point. The last request from
// FindBestWindow() was a synchronous XGetWindowProperty(), so no need to
// Sync. Leave the error handler installed to detect if w gets destroyed.
sGotBadWindow = false;
// make sure we get the right events on that window
XSelectInput(mDisplay, w,
(PropertyChangeMask|StructureNotifyMask));
bool destroyed = false;
// get the lock on the window
rv = GetLock(w, &destroyed);
if (NS_SUCCEEDED(rv)) {
// send our command
Bug 1080319 - Remove the -remote option. r=bsmedberg The -remote option has existed essentially forever, but its usefulness is questionable: - It requires a running instance to be any useful, so any script actually using it should first do -remote 'ping()' and handle the response properly. - It is not cross-application. The remote service dispatches the -remote commands to the command line handler, and, for example, desktop b2g builds don't have handlers for -remote (although thunderbird and seamonkey do). - It is not a cross-platform option, which leads to the following point: - There are other command line ways to do the same thing (at least in Firefox), without having to jump through hoops with -remote 'ping()', because there are command line options to do those same things on non-X11 platforms. For the latter, in Firefox case: - -remote 'openURL(url)' can be replaced with firefox url - -remote 'openURL(url,new-tab)' can be replaced with firefox -new-tab url - -remote 'openURL(url,new-window)' can be replaced with firefox -new-window url - -remote 'openfile(file,...)' is the same as -remote 'openurl(file,...) so, can be replaced as above - -remote 'xfedocommand(openbrowser)' is inherited from the mozilla suite and doesn't make much sense, but can be replaced with firefox -new-window The interesting part is that without changing nsBrowserContentHandler.js, -remote still works, meaning that if people really feel strongly about -remote, they'll still be able to write an addon to bring it back. This also means this patch actually doesn't remove -remote for applications other than Firefox that do support it, although -remote 'ping()' doesn't work as expected. However, other -remote commands will now work even without a running instance.
2014-10-14 02:19:52 +04:00
rv = DoSendCommandLine(w, argc, argv, aDesktopStartupID, aResponse,
&destroyed);
// if the window was destroyed, don't bother trying to free the
// lock.
if (!destroyed)
FreeLock(w); // doesn't really matter what this returns
}
}
XSetErrorHandler(sOldHandler);
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("SendCommandInternal returning 0x%" PRIx32 "\n",
static_cast<uint32_t>(rv)));
return rv;
}
Window
XRemoteClient::CheckWindow(Window aWindow)
{
Atom type = None;
int format;
unsigned long nitems, bytesafter;
unsigned char *data;
Window innerWindow;
XGetWindowProperty(mDisplay, aWindow, mMozWMStateAtom,
0, 0, False, AnyPropertyType,
&type, &format, &nitems, &bytesafter, &data);
if (type) {
XFree(data);
return aWindow;
}
// didn't find it here so check the children of this window
innerWindow = CheckChildren(aWindow);
if (innerWindow)
return innerWindow;
return aWindow;
}
Window
XRemoteClient::CheckChildren(Window aWindow)
{
Window root, parent;
Window *children;
unsigned int nchildren;
unsigned int i;
Atom type = None;
int format;
unsigned long nitems, after;
unsigned char *data;
Window retval = None;
if (!XQueryTree(mDisplay, aWindow, &root, &parent, &children,
&nchildren))
return None;
// scan the list first before recursing into the list of windows
// which can get quite deep.
for (i=0; !retval && (i < nchildren); i++) {
XGetWindowProperty(mDisplay, children[i], mMozWMStateAtom,
0, 0, False, AnyPropertyType, &type, &format,
&nitems, &after, &data);
if (type) {
XFree(data);
retval = children[i];
}
}
// otherwise recurse into the list
for (i=0; !retval && (i < nchildren); i++) {
retval = CheckChildren(children[i]);
}
if (children)
XFree((char *)children);
return retval;
}
nsresult
XRemoteClient::GetLock(Window aWindow, bool *aDestroyed)
{
bool locked = false;
bool waited = false;
*aDestroyed = false;
nsresult rv = NS_OK;
if (!mLockData) {
char pidstr[32];
char sysinfobuf[SYS_INFO_BUFFER_LENGTH];
SprintfLiteral(pidstr, "pid%d@", getpid());
PRStatus status;
status = PR_GetSystemInfo(PR_SI_HOSTNAME, sysinfobuf,
SYS_INFO_BUFFER_LENGTH);
if (status != PR_SUCCESS) {
return NS_ERROR_FAILURE;
}
// allocate enough space for the string plus the terminating
// char
mLockData = (char *)malloc(strlen(pidstr) + strlen(sysinfobuf) + 1);
if (!mLockData)
return NS_ERROR_OUT_OF_MEMORY;
strcpy(mLockData, pidstr);
if (!strcat(mLockData, sysinfobuf))
return NS_ERROR_FAILURE;
}
do {
int result;
Atom actual_type;
int actual_format;
unsigned long nitems, bytes_after;
unsigned char *data = 0;
XGrabServer(mDisplay);
result = XGetWindowProperty (mDisplay, aWindow, mMozLockAtom,
0, (65536 / sizeof (long)),
False, /* don't delete */
XA_STRING,
&actual_type, &actual_format,
&nitems, &bytes_after,
&data);
// aWindow may have been destroyed before XSelectInput was processed, in
// which case there may not be any DestroyNotify event in the queue to
// tell us. XGetWindowProperty() was synchronous so error responses have
// now been processed, setting sGotBadWindow.
if (sGotBadWindow) {
*aDestroyed = true;
rv = NS_ERROR_FAILURE;
}
else if (result != Success || actual_type == None) {
/* It's not now locked - lock it. */
XChangeProperty (mDisplay, aWindow, mMozLockAtom, XA_STRING, 8,
PropModeReplace,
(unsigned char *)mLockData,
strlen(mLockData));
locked = True;
}
XUngrabServer(mDisplay);
XFlush(mDisplay); // ungrab now!
if (!locked && !NS_FAILED(rv)) {
/* We tried to grab the lock this time, and failed because someone
else is holding it already. So, wait for a PropertyDelete event
to come in, and try again. */
MOZ_LOG(sRemoteLm, LogLevel::Debug,
("window 0x%x is locked by %s; waiting...\n",
(unsigned int) aWindow, data));
waited = True;
while (true) {
XEvent event;
int select_retval;
fd_set select_set;
struct timeval delay;
delay.tv_sec = 10;
delay.tv_usec = 0;
FD_ZERO(&select_set);
// add the x event queue to the select set
FD_SET(ConnectionNumber(mDisplay), &select_set);
select_retval = select(ConnectionNumber(mDisplay) + 1,
&select_set, nullptr, nullptr, &delay);
// did we time out?
if (select_retval == 0) {
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("timed out waiting for window\n"));
rv = NS_ERROR_FAILURE;
break;
}
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("xevent...\n"));
XNextEvent (mDisplay, &event);
if (event.xany.type == DestroyNotify &&
event.xdestroywindow.window == aWindow) {
*aDestroyed = true;
rv = NS_ERROR_FAILURE;
break;
}
if (event.xany.type == PropertyNotify &&
event.xproperty.state == PropertyDelete &&
event.xproperty.window == aWindow &&
event.xproperty.atom == mMozLockAtom) {
/* Ok! Someone deleted their lock, so now we can try
again. */
MOZ_LOG(sRemoteLm, LogLevel::Debug,
("(0x%x unlocked, trying again...)\n",
(unsigned int) aWindow));
break;
}
}
}
if (data)
XFree(data);
} while (!locked && !NS_FAILED(rv));
if (waited && locked) {
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("obtained lock.\n"));
} else if (*aDestroyed) {
MOZ_LOG(sRemoteLm, LogLevel::Debug,
("window 0x%x unexpectedly destroyed.\n",
(unsigned int) aWindow));
}
return rv;
}
Window
XRemoteClient::FindBestWindow(const char *aProgram, const char *aUsername,
Bug 1080319 - Remove the -remote option. r=bsmedberg The -remote option has existed essentially forever, but its usefulness is questionable: - It requires a running instance to be any useful, so any script actually using it should first do -remote 'ping()' and handle the response properly. - It is not cross-application. The remote service dispatches the -remote commands to the command line handler, and, for example, desktop b2g builds don't have handlers for -remote (although thunderbird and seamonkey do). - It is not a cross-platform option, which leads to the following point: - There are other command line ways to do the same thing (at least in Firefox), without having to jump through hoops with -remote 'ping()', because there are command line options to do those same things on non-X11 platforms. For the latter, in Firefox case: - -remote 'openURL(url)' can be replaced with firefox url - -remote 'openURL(url,new-tab)' can be replaced with firefox -new-tab url - -remote 'openURL(url,new-window)' can be replaced with firefox -new-window url - -remote 'openfile(file,...)' is the same as -remote 'openurl(file,...) so, can be replaced as above - -remote 'xfedocommand(openbrowser)' is inherited from the mozilla suite and doesn't make much sense, but can be replaced with firefox -new-window The interesting part is that without changing nsBrowserContentHandler.js, -remote still works, meaning that if people really feel strongly about -remote, they'll still be able to write an addon to bring it back. This also means this patch actually doesn't remove -remote for applications other than Firefox that do support it, although -remote 'ping()' doesn't work as expected. However, other -remote commands will now work even without a running instance.
2014-10-14 02:19:52 +04:00
const char *aProfile)
{
Window root = RootWindowOfScreen(DefaultScreenOfDisplay(mDisplay));
Window bestWindow = 0;
Window root2, parent, *kids;
unsigned int nkids;
// Get a list of the children of the root window, walk the list
// looking for the best window that fits the criteria.
if (!XQueryTree(mDisplay, root, &root2, &parent, &kids, &nkids)) {
MOZ_LOG(sRemoteLm, LogLevel::Debug,
("XQueryTree failed in XRemoteClient::FindBestWindow"));
return 0;
}
if (!(kids && nkids)) {
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("root window has no children"));
return 0;
}
// We'll walk the list of windows looking for a window that best
// fits the criteria here.
for (unsigned int i = 0; i < nkids; i++) {
Atom type;
int format;
unsigned long nitems, bytesafter;
unsigned char *data_return = 0;
Window w;
w = kids[i];
// find the inner window with WM_STATE on it
w = CheckWindow(w);
int status = XGetWindowProperty(mDisplay, w, mMozVersionAtom,
0, (65536 / sizeof (long)),
False, XA_STRING,
&type, &format, &nitems, &bytesafter,
&data_return);
if (!data_return)
continue;
double version = PR_strtod((char*) data_return, nullptr);
XFree(data_return);
Bug 1080319 - Remove the -remote option. r=bsmedberg The -remote option has existed essentially forever, but its usefulness is questionable: - It requires a running instance to be any useful, so any script actually using it should first do -remote 'ping()' and handle the response properly. - It is not cross-application. The remote service dispatches the -remote commands to the command line handler, and, for example, desktop b2g builds don't have handlers for -remote (although thunderbird and seamonkey do). - It is not a cross-platform option, which leads to the following point: - There are other command line ways to do the same thing (at least in Firefox), without having to jump through hoops with -remote 'ping()', because there are command line options to do those same things on non-X11 platforms. For the latter, in Firefox case: - -remote 'openURL(url)' can be replaced with firefox url - -remote 'openURL(url,new-tab)' can be replaced with firefox -new-tab url - -remote 'openURL(url,new-window)' can be replaced with firefox -new-window url - -remote 'openfile(file,...)' is the same as -remote 'openurl(file,...) so, can be replaced as above - -remote 'xfedocommand(openbrowser)' is inherited from the mozilla suite and doesn't make much sense, but can be replaced with firefox -new-window The interesting part is that without changing nsBrowserContentHandler.js, -remote still works, meaning that if people really feel strongly about -remote, they'll still be able to write an addon to bring it back. This also means this patch actually doesn't remove -remote for applications other than Firefox that do support it, although -remote 'ping()' doesn't work as expected. However, other -remote commands will now work even without a running instance.
2014-10-14 02:19:52 +04:00
if (!(version >= 5.1 && version < 6))
continue;
data_return = 0;
if (status != Success || type == None)
continue;
// If someone passed in a program name, check it against this one
// unless it's "any" in which case, we don't care. If someone did
// pass in a program name and this window doesn't support that
// protocol, we don't include it in our list.
if (aProgram && strcmp(aProgram, "any")) {
Unused << XGetWindowProperty(mDisplay, w, mMozProgramAtom,
0, (65536 / sizeof(long)),
False, XA_STRING,
&type, &format, &nitems, &bytesafter,
&data_return);
// If the return name is not the same as what someone passed in,
// we don't want this window.
if (data_return) {
if (strcmp(aProgram, (const char *)data_return)) {
XFree(data_return);
continue;
}
// This is actually the success condition.
XFree(data_return);
}
else {
// Doesn't support the protocol, even though the user
// requested it. So we're not going to use this window.
continue;
}
}
// Check to see if it has the user atom on that window. If there
// is then we need to make sure that it matches what we have.
const char *username;
if (aUsername) {
username = aUsername;
}
else {
username = PR_GetEnv("LOGNAME");
}
if (username) {
Unused << XGetWindowProperty(mDisplay, w, mMozUserAtom,
0, (65536 / sizeof(long)),
False, XA_STRING,
&type, &format, &nitems, &bytesafter,
&data_return);
// if there's a username compare it with what we have
if (data_return) {
// If the IDs aren't equal, we don't want this window.
if (strcmp(username, (const char *)data_return)) {
XFree(data_return);
continue;
}
XFree(data_return);
}
}
// Check to see if there's a profile name on this window. If
// there is, then we need to make sure it matches what someone
// passed in.
if (aProfile) {
Unused << XGetWindowProperty(mDisplay, w, mMozProfileAtom,
0, (65536 / sizeof(long)),
False, XA_STRING,
&type, &format, &nitems, &bytesafter,
&data_return);
// If there's a profile compare it with what we have
if (data_return) {
// If the profiles aren't equal, we don't want this window.
if (strcmp(aProfile, (const char *)data_return)) {
XFree(data_return);
continue;
}
XFree(data_return);
}
}
// Check to see if the window supports the new command-line passing
// protocol, if that is requested.
// If we got this far, this is the best window. It passed
// all the tests.
bestWindow = w;
break;
}
if (kids)
XFree((char *) kids);
return bestWindow;
}
nsresult
XRemoteClient::FreeLock(Window aWindow)
{
int result;
Atom actual_type;
int actual_format;
unsigned long nitems, bytes_after;
unsigned char *data = 0;
result = XGetWindowProperty(mDisplay, aWindow, mMozLockAtom,
0, (65536 / sizeof(long)),
True, /* atomic delete after */
XA_STRING,
&actual_type, &actual_format,
&nitems, &bytes_after,
&data);
if (result != Success) {
MOZ_LOG(sRemoteLm, LogLevel::Debug,
("unable to read and delete " MOZILLA_LOCK_PROP
" property\n"));
return NS_ERROR_FAILURE;
}
if (!data || !*data){
MOZ_LOG(sRemoteLm, LogLevel::Debug,
("invalid data on " MOZILLA_LOCK_PROP
" of window 0x%x.\n",
(unsigned int) aWindow));
return NS_ERROR_FAILURE;
}
else if (strcmp((char *)data, mLockData)) {
MOZ_LOG(sRemoteLm, LogLevel::Debug,
(MOZILLA_LOCK_PROP " was stolen! Expected \"%s\", saw \"%s\"!\n",
mLockData, data));
return NS_ERROR_FAILURE;
}
if (data)
XFree(data);
return NS_OK;
}
nsresult
XRemoteClient::DoSendCommandLine(Window aWindow, int32_t argc, char **argv,
const char* aDesktopStartupID,
char **aResponse, bool *aDestroyed)
{
*aDestroyed = false;
int commandLineLength;
char* commandLine = ConstructCommandLine(argc, argv, aDesktopStartupID,
&commandLineLength);
XChangeProperty (mDisplay, aWindow, mMozCommandLineAtom, XA_STRING, 8,
PropModeReplace, (unsigned char *) commandLine,
commandLineLength);
free(commandLine);
if (!WaitForResponse(aWindow, aResponse, aDestroyed, mMozCommandLineAtom))
return NS_ERROR_FAILURE;
return NS_OK;
}
bool
XRemoteClient::WaitForResponse(Window aWindow, char **aResponse,
bool *aDestroyed, Atom aCommandAtom)
{
bool done = false;
bool accepted = false;
while (!done) {
XEvent event;
XNextEvent (mDisplay, &event);
if (event.xany.type == DestroyNotify &&
event.xdestroywindow.window == aWindow) {
/* Print to warn user...*/
MOZ_LOG(sRemoteLm, LogLevel::Debug,
("window 0x%x was destroyed.\n",
(unsigned int) aWindow));
*aResponse = strdup("Window was destroyed while reading response.");
*aDestroyed = true;
return false;
}
if (event.xany.type == PropertyNotify &&
event.xproperty.state == PropertyNewValue &&
event.xproperty.window == aWindow &&
event.xproperty.atom == mMozResponseAtom) {
Atom actual_type;
int actual_format;
unsigned long nitems, bytes_after;
unsigned char *data = 0;
Bool result;
result = XGetWindowProperty (mDisplay, aWindow, mMozResponseAtom,
0, (65536 / sizeof (long)),
True, /* atomic delete after */
XA_STRING,
&actual_type, &actual_format,
&nitems, &bytes_after,
&data);
if (result != Success) {
MOZ_LOG(sRemoteLm, LogLevel::Debug,
("failed reading " MOZILLA_RESPONSE_PROP
" from window 0x%0x.\n",
(unsigned int) aWindow));
*aResponse = strdup("Internal error reading response from window.");
done = true;
}
else if (!data || strlen((char *) data) < 5) {
MOZ_LOG(sRemoteLm, LogLevel::Debug,
("invalid data on " MOZILLA_RESPONSE_PROP
" property of window 0x%0x.\n",
(unsigned int) aWindow));
*aResponse = strdup("Server returned invalid data in response.");
done = true;
}
else if (*data == '1') { /* positive preliminary reply */
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("%s\n", data + 4));
/* keep going */
done = false;
}
else if (!strncmp ((char *)data, "200", 3)) { /* positive completion */
*aResponse = strdup((char *)data);
accepted = true;
done = true;
}
else if (*data == '2') { /* positive completion */
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("%s\n", data + 4));
*aResponse = strdup((char *)data);
accepted = true;
done = true;
}
else if (*data == '3') { /* positive intermediate reply */
MOZ_LOG(sRemoteLm, LogLevel::Debug,
("internal error: "
"server wants more information? (%s)\n",
data));
*aResponse = strdup((char *)data);
done = true;
}
else if (*data == '4' || /* transient negative completion */
*data == '5') { /* permanent negative completion */
MOZ_LOG(sRemoteLm, LogLevel::Debug, ("%s\n", data + 4));
*aResponse = strdup((char *)data);
done = true;
}
else {
MOZ_LOG(sRemoteLm, LogLevel::Debug,
("unrecognised " MOZILLA_RESPONSE_PROP
" from window 0x%x: %s\n",
(unsigned int) aWindow, data));
*aResponse = strdup((char *)data);
done = true;
}
if (data)
XFree(data);
}
else if (event.xany.type == PropertyNotify &&
event.xproperty.window == aWindow &&
event.xproperty.state == PropertyDelete &&
event.xproperty.atom == aCommandAtom) {
MOZ_LOG(sRemoteLm, LogLevel::Debug,
("(server 0x%x has accepted "
Bug 1080319 - Remove the -remote option. r=bsmedberg The -remote option has existed essentially forever, but its usefulness is questionable: - It requires a running instance to be any useful, so any script actually using it should first do -remote 'ping()' and handle the response properly. - It is not cross-application. The remote service dispatches the -remote commands to the command line handler, and, for example, desktop b2g builds don't have handlers for -remote (although thunderbird and seamonkey do). - It is not a cross-platform option, which leads to the following point: - There are other command line ways to do the same thing (at least in Firefox), without having to jump through hoops with -remote 'ping()', because there are command line options to do those same things on non-X11 platforms. For the latter, in Firefox case: - -remote 'openURL(url)' can be replaced with firefox url - -remote 'openURL(url,new-tab)' can be replaced with firefox -new-tab url - -remote 'openURL(url,new-window)' can be replaced with firefox -new-window url - -remote 'openfile(file,...)' is the same as -remote 'openurl(file,...) so, can be replaced as above - -remote 'xfedocommand(openbrowser)' is inherited from the mozilla suite and doesn't make much sense, but can be replaced with firefox -new-window The interesting part is that without changing nsBrowserContentHandler.js, -remote still works, meaning that if people really feel strongly about -remote, they'll still be able to write an addon to bring it back. This also means this patch actually doesn't remove -remote for applications other than Firefox that do support it, although -remote 'ping()' doesn't work as expected. However, other -remote commands will now work even without a running instance.
2014-10-14 02:19:52 +04:00
MOZILLA_COMMANDLINE_PROP ".)\n",
(unsigned int) aWindow));
}
}
return accepted;
}