Bug 333703 Non-ascii directory name is garbled in directory index. patch by Masatoshi Kimura (:emk) <VYV03354@nifty.ne.jp> r=jshin, sr=darin

netwerk/base/src/nsDirectoryIndexStream.cpp

@@ -67,14 +67,14 @@ static PRLogModuleInfo* gLog;
 #include "nsURLHelper.h"
 #include "nsNetUtil.h"
 #include "nsCRT.h"
+#include "nsNativeCharsetUtils.h"
 
 static NS_DEFINE_CID(kCollationFactoryCID, NS_COLLATIONFACTORY_CID);
 
 // NOTE: This runs on the _file transport_ thread.
 // The problem is that now that we're actually doing something with the data,
 // we want to do stuff like i18n sorting. However, none of the collation stuff
-// is threadsafe, and stuff like GetUnicodeLeafName spews warnings on a debug
+// is threadsafe.
-// build, because the singletons were initialised on the UI thread.
 // So THIS CODE IS ASCII ONLY!!!!!!!! This is no worse than the current
 // behaviour, though. See bug 99382.
 // When this is fixed, #define THREADSAFE_I18N to get this code working
@@ -97,43 +97,32 @@ static int PR_CALLBACK compare(nsIFile* aElement1,
                                nsIFile* aElement2,
                                void* aData)
 {
-    // Not that this #ifdef makes much of a difference... We need it
+    if (!NS_IsNativeUTF8()) {
-    // to work out which version of GetLeafName to use, though
+        // don't check for errors, because we can't report them anyway
-#ifdef THREADSAFE_I18N
+        nsAutoString name1, name2;
-    // don't check for errors, because we can't report them anyway
+        aElement1->GetLeafName(name1);
-    nsXPIDLString name1, name2;
+        aElement2->GetLeafName(name2);
-    aElement1->GetUnicodeLeafName(getter_Copies(name1));
-    aElement2->GetUnicodeLeafName(getter_Copies(name2));
 
-    // Note - we should be the collation to do sorting. Why don't we?
+        // Note - we should do the collation to do sorting. Why don't we?
-    // Because that is _slow_. Using TestProtocols to list file:///dev/
+        // Because that is _slow_. Using TestProtocols to list file:///dev/
-    // goes from 3 seconds to 22. (This may be why nsXULSortService is
+        // goes from 3 seconds to 22. (This may be why nsXULSortService is
-    // so slow as well).
+        // so slow as well).
-    // Does this have bad effects? Probably, but since nsXULTree appears
+        // Does this have bad effects? Probably, but since nsXULTree appears
-    // to use the raw RDF literal value as the sort key (which ammounts to an
+        // to use the raw RDF literal value as the sort key (which ammounts to an
-    // strcmp), it won't be any worse, I think.
+        // strcmp), it won't be any worse, I think.
-    // This could be made faster, by creating the keys once,
+        // This could be made faster, by creating the keys once,
-    // but CompareString could still be smarter - see bug 99383 - bbaetz
+        // but CompareString could still be smarter - see bug 99383 - bbaetz
-    // NB - 99393 has been WONTFIXed. So if the I18N code is ever made
+        // NB - 99393 has been WONTFIXed. So if the I18N code is ever made
-    // threadsafe so that this matters, we'd have to pass through a
+        // threadsafe so that this matters, we'd have to pass through a
-    // struct { nsIFile*, PRUint8* } with the pre-calculated key.
+        // struct { nsIFile*, PRUint8* } with the pre-calculated key.
-    return Compare(name1, name2);
+        return Compare(name1, name2);
+    }
 
-    /*PRInt32 res;
-    // CompareString takes an nsString...
-    nsString str1(name1);
-    nsString str2(name2);
-
-    nsICollation* coll = (nsICollation*)aData;
-    coll->CompareString(nsICollation::kCollationStrengthDefault, str1, str2, &res);
-    return res;*/
-#else
     nsCAutoString name1, name2;
     aElement1->GetNativeLeafName(name1);
     aElement2->GetNativeLeafName(name2);
 
     return Compare(name1, name2);
-#endif
 }
 
 nsresult
@@ -157,13 +146,6 @@ nsDirectoryIndexStream::Init(nsIFile* aDir)
     }
 #endif
 
-#ifdef THREADSAFE_I18N
-    if (!mTextToSubURI) {
-        mTextToSubURI = do_GetService(NS_ITEXTTOSUBURI_CONTRACTID, &rv);
-        if (NS_FAILED(rv)) return rv;
-    }
-#endif
-
     // Sigh. We have to allocate on the heap because there are no
     // assignment operators defined.
     nsCOMPtr<nsISimpleEnumerator> iter;
@@ -216,24 +198,6 @@ nsDirectoryIndexStream::Init(nsIFile* aDir)
 
     mBuf.AppendLiteral("200: filename content-length last-modified file-type\n");
 
-    if (mFSCharset.IsEmpty()) {
-        // OK, set up the charset
-#ifdef THREADSAFE_I18N
-        nsCOMPtr<nsIPlatformCharset> pc = do_GetService(NS_PLATFORMCHARSET_CONTRACTID, &rv);
-        if (NS_FAILED(rv)) return rv;
-        nsString tmp;
-        rv = pc->GetCharset(kPlatformCharsetSel_FileName, tmp);
-        if (NS_FAILED(rv)) return rv;
-        LossyCopyUTF16toASCII(tmp, mFSCharset);
-#endif   
-    }
-    
-    if (!mFSCharset.IsEmpty()) {
-        mBuf.AppendLiteral("301: ");
-        mBuf.Append(mFSCharset);
-        mBuf.Append('\n');
-    }
-    
     return NS_OK;
 }
 
@@ -334,62 +298,49 @@ nsDirectoryIndexStream::Read(char* aBuf, PRUint32 aCount, PRUint32* aReadCount)
             }
 #endif
 
-        // rjc: don't return hidden files/directories!
+            // rjc: don't return hidden files/directories!
-        // bbaetz: why not?
+            // bbaetz: why not?
-        nsresult rv;
+            nsresult rv;
 #ifndef XP_UNIX
-        PRBool hidden = PR_FALSE;
+            PRBool hidden = PR_FALSE;
-        current->IsHidden(&hidden);
+            current->IsHidden(&hidden);
-        if (hidden) {
+            if (hidden) {
-            PR_LOG(gLog, PR_LOG_DEBUG,
+                PR_LOG(gLog, PR_LOG_DEBUG,
-                   ("nsDirectoryIndexStream[%p]: skipping hidden file/directory",
+                       ("nsDirectoryIndexStream[%p]: skipping hidden file/directory",
-                    this));
+                        this));
-            continue;
+                continue;
-        }
+            }
 #endif
 
-        PRInt64 fileSize = LL_Zero();
+            PRInt64 fileSize = 0;
-        current->GetFileSize( &fileSize );
+            current->GetFileSize( &fileSize );
 
-        PRInt64 tmpTime = LL_Zero();
+            PRInt64 fileInfoModifyTime = 0;
-        PRInt64 fileInfoModifyTime = LL_Zero();
+            current->GetLastModifiedTime( &fileInfoModifyTime );
-        current->GetLastModifiedTime( &tmpTime );
+            fileInfoModifyTime *= PR_USEC_PER_MSEC;
-        // Why does nsIFile give this back in milliseconds?
-        LL_MUL(fileInfoModifyTime, tmpTime, PR_USEC_PER_MSEC);
 
-        mBuf.AppendLiteral("201: ");
+            mBuf.AppendLiteral("201: ");
 
-        // The "filename" field
+            // The "filename" field
-        {
+            char* escaped = nsnull;
-#ifdef THREADSAFE_I18N
+            if (!NS_IsNativeUTF8()) {
-            nsXPIDLString leafname;
+                nsAutoString leafname;
-            rv = current->GetUnicodeLeafName(getter_Copies(leafname));
+                rv = current->GetLeafName(leafname);
-            if (NS_FAILED(rv)) return rv;
-            if (!leafname.IsEmpty()) {
-                // XXX - this won't work with directories with spaces
-                // see bug 99478 - bbaetz
-                nsXPIDLCString escaped;
-                rv = mTextToSubURI->ConvertAndEscape(mFSCharset.get(),
-                                                     leafname.get(),
-                                                     getter_Copies(escaped));
                 if (NS_FAILED(rv)) return rv;
-                mBuf.Append(escaped);
+                if (!leafname.IsEmpty())
+                    escaped = nsEscape(NS_ConvertUTF16toUTF8(leafname).get(), url_Path);
+            } else {
+                nsCAutoString leafname;
+                rv = current->GetNativeLeafName(leafname);
+                if (NS_FAILED(rv)) return rv;
+                if (!leafname.IsEmpty())
+                    escaped = nsEscape(leafname.get(), url_Path);
+            }
+            if (escaped) {
+                mBuf += escaped;
                 mBuf.Append(' ');
+                nsMemory::Free(escaped);
             }
-#else
-            nsCAutoString leafname;
-            rv = current->GetNativeLeafName(leafname);
-            if (NS_FAILED(rv)) return rv;
-            if (!leafname.IsEmpty()) {
-                char* escaped = nsEscape(leafname.get(), url_Path);
-                if (escaped) {
-                    mBuf += escaped;
-                    mBuf.Append(' ');
-                    nsCRT::free(escaped);
-                }
-            }
-#endif
-        }
 
             // The "content-length" field
             mBuf.AppendInt(fileSize, 10);

netwerk/base/src/nsDirectoryIndexStream.h

@@ -55,11 +55,6 @@ private:
     PRInt32             mPos;   // position within mArray
     nsCOMArray<nsIFile> mArray; // file objects within the directory
 
-    // XXX - we could make these statics, and share them. But we're only going to have one
-    // of directoryindexstream at once, so there's no point
-    nsXPIDLCString mFSCharset;
-    nsCOMPtr<nsITextToSubURI> mTextToSubURI;
-
     nsDirectoryIndexStream();
     /**
      * aDir will only be used on the calling thread.

netwerk/streamconv/converters/nsIndexedToHTML.cpp

@@ -166,6 +166,7 @@ nsIndexedToHTML::OnStartRequest(nsIRequest* request, nsISupports *aContext) {
     // - bbaetz
 
     PRBool isScheme = PR_FALSE;
+    PRBool isSchemeFile = PR_FALSE;
     if (NS_SUCCEEDED(uri->SchemeIs("ftp", &isScheme)) && isScheme) {
 
         // ftp urls don't always end in a /
@@ -203,7 +204,7 @@ nsIndexedToHTML::OnStartRequest(nsIRequest* request, nsISupports *aContext) {
             rv = uri->Resolve(NS_LITERAL_CSTRING(".."),parentStr);
             if (NS_FAILED(rv)) return rv;
         }
-    } else if (NS_SUCCEEDED(uri->SchemeIs("file", &isScheme)) && isScheme) {
+    } else if (NS_SUCCEEDED(uri->SchemeIs("file", &isSchemeFile)) && isSchemeFile) {
         nsCOMPtr<nsIFileURL> fileUrl = do_QueryInterface(uri);
         nsCOMPtr<nsIFile> file;
         rv = fileUrl->GetFile(getter_AddRefs(file));
@@ -226,13 +227,8 @@ nsIndexedToHTML::OnStartRequest(nsIRequest* request, nsISupports *aContext) {
             parentStr.Assign(url);
         }
 
-        // reset parser's charset to platform's default if this is file url
+        // Directory index will be always encoded in UTF-8 if this is file url
-        nsCOMPtr<nsIPlatformCharset> platformCharset(do_GetService(NS_PLATFORMCHARSET_CONTRACTID, &rv));
+        rv = mParser->SetEncoding("UTF-8");
-        NS_ENSURE_SUCCESS(rv, rv);
-        nsCAutoString charset;
-        rv = platformCharset->GetCharset(kPlatformCharsetSel_FileName, charset);
-        NS_ENSURE_SUCCESS(rv, rv);
-        rv = mParser->SetEncoding(charset.get());
         NS_ENSURE_SUCCESS(rv, rv);
 
     } else if (NS_SUCCEEDED(uri->SchemeIs("gopher", &isScheme)) && isScheme) {
@@ -304,6 +300,20 @@ nsIndexedToHTML::OnStartRequest(nsIRequest* request, nsISupports *aContext) {
     nsXPIDLString unEscapeSpec;
     rv = mTextToSubURI->UnEscapeAndConvert(encoding, titleUri.get(),
                                            getter_Copies(unEscapeSpec));
+    // unescape may fail because
+    // 1. file URL may be encoded in platform charset for backward compatibility
+    // 2. query part may not be encoded in UTF-8 (see bug 261929)
+    // so try the platform's default if this is file url
+    if (NS_FAILED(rv) && isSchemeFile) {
+        nsCOMPtr<nsIPlatformCharset> platformCharset(do_GetService(NS_PLATFORMCHARSET_CONTRACTID, &rv));
+        NS_ENSURE_SUCCESS(rv, rv);
+        nsCAutoString charset;
+        rv = platformCharset->GetCharset(kPlatformCharsetSel_FileName, charset);
+        NS_ENSURE_SUCCESS(rv, rv);
+
+        rv = mTextToSubURI->UnEscapeAndConvert(charset.get(), titleUri.get(),
+                                               getter_Copies(unEscapeSpec));
+    }
     if (NS_FAILED(rv)) return rv;
 
     nsXPIDLString htmlEscSpec;