GitLab

Bug 25428: 5 crashes in DumpRenderTree at com.apple.WebCore • -[AccessibilityObjectWrapper accessibilityAttributeValue:] + 830
https://bugs.webkit.org/show_bug.cgi?id=25428



git-svn-id: svn://svn.chromium.org/blink/trunk@42900 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Adam Roben.

        Update the example plug-in for new changes to the Cocoa event model.
        
        * NetscapeCocoaPlugin/main.m:
        (handleDraw):
        (handleMouseEvent):
        (NPP_HandleEvent):



git-svn-id: svn://svn.chromium.org/blink/trunk@42899 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        
        * Plugins/WebNetscapePluginView.mm: The base class of WebNetscapePluginView.mm, WebBaseNetscapePluginView, already implemented 
        the two methods below. But the Tiger compiler didn't know that. 
        (-[WebNetscapePluginView webView]): 
        (-[WebNetscapePluginView webFrame]):



git-svn-id: svn://svn.chromium.org/blink/trunk@42898 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        <rdar://problem/6352982>
        
        * Plugins/WebBaseNetscapePluginView.mm: Removed checkIfAllowedToLoadURL:frame:callbackFunc, cancelCheckIfAllowedToLoadURL, and 
        _webPluginContainerCancelCheckIfAllowedToLoadRequest from the base class. These methods now exist in the subclass WebNetscapePluginView.
        
        Added WebNetscapeContainerCheckContextInfo, which is used as a "contextInfo" object in 
        -[WebNetscapePluginView checkIfAllowedToLoadURL:frame:callbackFunc:]
        * Plugins/WebNetscapeContainerCheckContextInfo.h: Added.
        * Plugins/WebNetscapeContainerCheckContextInfo.mm: Added.
        (-[WebNetscapeContainerCheckContextInfo initWithCheckRequestID:callbackFunc:]): Added desiginated initializer.
        (-[WebNetscapeContainerCheckContextInfo checkRequestID]): Added. Returns the checkRequestID.
        (-[WebNetscapeContainerCheckContextInfo callback]): Added. Returns the callback.
        
        * Plugins/WebNetscapePluginView.h: Added two new ivars: _containerChecksInProgress and _currentContainerCheckRequestID.
        * Plugins/WebNetscapePluginView.mm: #import both WebPluginContainerCheck.h and WebNetscapeContainerCheckContextInfo.h
        
        (-[WebNetscapePluginView checkIfAllowedToLoadURL:frame:callbackFunc:]): Added. This is the implementation of WKN_CheckIfAllowedToLoadURL.
        Here, we increment the request ID and start the container check.
        
        (-[WebNetscapePluginView _containerCheckResult:contextInfo:]): Added. This is a callback method for WebPluginContainerCheck.
        It's where we actually call into the plug-in and provide the allow-or-deny result.
    
        (-[WebNetscapePluginView cancelCheckIfAllowedToLoadURL:]): Added. This is the implementation of WKN_CancelCheckIfAllowedToLoadURL. 
        Here we lookup the check, cancel it, and remove it from _containerChecksInProgress.
        
        (-[WebNetscapePluginView _webPluginContainerCancelCheckIfAllowedToLoadRequest:]): Added. WebPluginContainerCheck automatically calls
        this method after invoking our _containerCheckResult: selector. It works this way because calling -[WebPluginContainerCheck cancel] allows
        it to do it's teardown process.
        
        (-[WebNetscapePluginView fini]): Release _containerChecksInProgress ivar.
        
        * Plugins/WebPluginContainerCheck.h: Removed initWithRequest: method from header; no client was using this method directly.
        * Plugins/WebPluginContainerCheck.mm: 
        (+[WebPluginContainerCheck checkWithRequest:target:resultObject:selector:controller:contextInfo:]): Added optional contextInfo parameter.
        (-[WebPluginContainerCheck _continueWithPolicy:]): If there's a contextInfo object, pass it as a parameter to resultSelector.
        (-[WebPluginContainerCheck cancel]): Release _contextInfo ivar.
        (-[WebPluginContainerCheck contextInfo]): Added new method. Returns the contextInfo object, if one so exists.
        * Plugins/WebPluginController.mm:
        (-[WebPluginController _webPluginContainerCheckIfAllowedToLoadRequest:inFrame:resultObject:selector:]): Pass nil for contextInfo. WebKit-style
        plug-ins do not need additional context information.



git-svn-id: svn://svn.chromium.org/blink/trunk@42897 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        https://bugs.webkit.org/show_bug.cgi?id=25399
        <rdar://problem/6633943> REGRESSION: Many crashes reported accessing Lexis/Nexis database,
        beneath WebCore::Cache::evict

        The crash happened because a cached resource handle was removed from a document's cached
        resources map twice recursively, so a destructor was called for a value in a deleted bucket.
        The first call was from Cache::evict, and when destroying CachedResourceHandle destroyed
        CachedResource, DocLoader::removeCachedResource() was called again, with HashMap being in
        an inconsistent state.

        I couldn't fully reconstruct the loading sequence to make a test.

        * loader/Cache.cpp:
        (WebCore::Cache::revalidateResource): Assert that the resource being revalidated is in cache
        (it makes no sense to revalidate one that isn't).
        (WebCore::Cache::evict): Don't remove the resource from document's map. Removing a resource
        from the cache in no way implies that documents no longer use the old version. This fixes the
        crash, and also fixes many cases of resource content being unavailable in Web Inspector.

        * loader/CachedResource.h:
        (WebCore::CachedResource::setInCache): When bringing a revalidated resource back to cache,
        reset m_isBeingRevalidated to maintain the invariant of resources being revalidated never
        being in cache. This fixes another assertion I saw on LexisNexis search: in rare cases,
        switchClientsToRevalidatedResource() results in the same resource being requested again,
        but we were only enforcing CachedResource invariants after calling this function.
        (WebCore::CachedResource::unregisterHandle): Assert that the counter doesn't underflow.

        * loader/DocLoader.cpp: (WebCore::DocLoader::removeCachedResource): Assert that the passed
        resource is removed, not some other resource that happens to have the same URL (this used to
        fail on LexisNexis search before this patch).

        * loader/ImageDocument.cpp: (WebCore::ImageTokenizer::write): Replaced ASSERT_NOT_REACHED
        with notImplemented(). This method can be legally called via document.write(), and should
        work. LexisNexis takes this code path, but apparently has a fallback for Safari, so it
        doesn't affect site functionality.

        * loader/CachedResource.cpp:
        (WebCore::CachedResource::clearResourceToRevalidate): Don't assert that m_resourceToRevalidate
        is being revalidated - this may no longer be true, because we now reset this member in
        CachedResource::setInCache().



git-svn-id: svn://svn.chromium.org/blink/trunk@42896 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        - make BidiContext a RefCounted class, starting with a refcount of 1,
          and share the root BidiContexts.

        * platform/graphics/GraphicsContext.cpp:
        (WebCore::GraphicsContext::drawBidiText): Use BidiContext::create().
        * platform/text/BidiContext.cpp:
        (WebCore::BidiContext::create): Added. For the four "root" contexts,
        returns a reference to a shared static BidiContext.
        * platform/text/BidiContext.h:
        (WebCore::BidiContext::BidiContext):
        * platform/text/BidiResolver.h:
        (WebCore::::commitExplicitEmbedding): Use BidiContext::create().
        * rendering/bidi.cpp:
        (WebCore::RenderBlock::determineStartPosition): Ditto.



git-svn-id: svn://svn.chromium.org/blink/trunk@42895 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Rubber-stamped by Gustavo Noronha.

        Build fix for r42893: change GDK_exclaim to GDK_exclam

        * platform/chromium/KeyCodeConversionGtk.cpp:
        (WebCore::windowsKeyCodeForKeyEvent):
        * platform/gtk/KeyEventGtk.cpp:
        (WebCore::windowsKeyCodeForKeyEvent):

git-svn-id: svn://svn.chromium.org/blink/trunk@42894 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Xan Lopez.

        Fix wrong key code conversion.

        Missing conversion for GDK_exlaim ('!').
        Parens are opposite: GDK_parenleft is open paren '('
        and GDK_parenright is close paren ')'.
        https://bugs.webkit.org/show_bug.cgi?id=25367

        * platform/chromium/KeyCodeConversionGtk.cpp:
        (WebCore::windowsKeyCodeForKeyEvent):
        * platform/gtk/KeyEventGtk.cpp:
        (WebCore::windowsKeyCodeForKeyEvent):

git-svn-id: svn://svn.chromium.org/blink/trunk@42893 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Gustavo Noronha.

        [Gtk] gtkdrawing update
        https://bugs.webkit.org/show_bug.cgi?id=25408

        Update to the 25 Apr 2009 version of gtk2drawing.c and
        gtkdrawing.h from Mozilla (http://hg.mozilla.org/mozilla-central)

        * platform/gtk/gtk2drawing.c:
        (setup_widget_prototype):
        (moz_gtk_get_combo_box_entry_arrow):
        (ensure_toolbar_widget):
        (ensure_tree_header_cell_widget):
        (moz_gtk_init):
        (moz_gtk_button_get_inner_border):
        (moz_gtk_toggle_paint):
        (moz_gtk_caret_paint):
        (moz_gtk_entry_paint):
        (moz_gtk_tree_header_cell_paint):
        (moz_gtk_combo_box_paint):
        (moz_gtk_toolbar_separator_paint):
        (moz_gtk_menu_separator_paint):
        (moz_gtk_get_widget_border):
        (moz_gtk_get_tab_scroll_arrow_size):
        (moz_gtk_get_toolbar_separator_width):
        (moz_gtk_get_menu_separator_height):
        (moz_gtk_widget_paint):
        (moz_gtk_shutdown):
        * platform/gtk/gtkdrawing.h:

git-svn-id: svn://svn.chromium.org/blink/trunk@42892 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Gustavo Noronha.

        [GTK] Make layout tests / jsc tests usable without build-webkit
        https://bugs.webkit.org/show_bug.cgi?id=21784

        This requires WEBKITOUTPUTDIR to be set to the build directory.
        Set the location of the product dir to the location pointed to by
        WEBKITOUTPUTDIR.

        * Scripts/webkitdirs.pm:

git-svn-id: svn://svn.chromium.org/blink/trunk@42891 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Adam Roben.

        Use libsoup's facilities to resolve relative URLs, instead, since
        it is a more robust method.

        * DumpRenderTree/gtk/LayoutTestControllerGtk.cpp:
        (LayoutTestController::queueLoad):

git-svn-id: svn://svn.chromium.org/blink/trunk@42890 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Xan Lopez.

        https://bugs.webkit.org/show_bug.cgi?id=25409
        [GTK] incorrectly fallsback on cancellation

        Check whether we should fallback also on didFailLoad, so that we
        don't get error pages showing up in parts of the content of some
        sites because of, for instance, location changes.

        * WebCoreSupport/FrameLoaderClientGtk.cpp:
        (WebKit::FrameLoaderClient::dispatchDidFailLoad):

git-svn-id: svn://svn.chromium.org/blink/trunk@42889 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Xan Lopez.

        https://bugs.webkit.org/show_bug.cgi?id=25409
        [GTK] incorrectly fallsback on cancellation

        Correctly handle cancellation and frame loading interruption
        caused by policy change.  This also fixes
        fast/frames/location-change.html.

        * WebCoreSupport/FrameLoaderClientGtk.cpp:
        (WebKit::FrameLoaderClient::shouldFallBack):

git-svn-id: svn://svn.chromium.org/blink/trunk@42888 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Tor Arne Vestbø.

        https://bugs.webkit.org/show_bug.cgi?id=18475

        [Qt] Widget painting should follow the layout direction (LTR, RTL)
        of the element style, not the application layout direction.

        * platform/qt/RenderThemeQt.cpp:
        (WebCore::RenderThemeQt::applyTheme):

git-svn-id: svn://svn.chromium.org/blink/trunk@42887 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Not sure why the buildbots weren't affected by this problem.

        * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj: Let VS
        re-order the file list, and added JavaScriptCore[_debug].def to the
        project. This was not necessary for the fix, but made making the fix
        easier.

        * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore_debug.def:
        Removed a function that no longer exists.

git-svn-id: svn://svn.chromium.org/blink/trunk@42886 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Mark Rowe.

        Plug memory leak.

        * WebCoreSupport/FrameLoaderClientGtk.cpp:
        (WebKit::FrameLoaderClient::dispatchDidFailLoad):

git-svn-id: svn://svn.chromium.org/blink/trunk@42885 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Weinig Sam.

        Fix for https://bugs.webkit.org/show_bug.cgi?id=25416
        "Cached prototype accesses unsafely hoist property storage load above structure checks."
        
        Do not hoist the load of the pointer to the property storage array.

        No performance impact.

        * jit/JITPropertyAccess.cpp:
        (JSC::JIT::privateCompileGetByIdProto):
        (JSC::JIT::privateCompileGetByIdProtoList):



git-svn-id: svn://svn.chromium.org/blink/trunk@42884 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Geoffrey "Gaffe or energy?" Garen.

        Randomize address requested by ExecutableAllocatorFixedVMPool.

        * jit/ExecutableAllocatorFixedVMPool.cpp:
        (JSC::FixedVMPoolAllocator::FixedVMPoolAllocator):



git-svn-id: svn://svn.chromium.org/blink/trunk@42883 bbb929c8-8fbe-4397-9dbb-9b2b20218538

2009-04-26  Sam Weinig  <sam@webkit.org>

        Reviewed by Dan Bernstein.

        Fix for https://bugs.webkit.org/show_bug.cgi?id=25412
        The Window object should be an EventTarget

        Test: fast/dom/Window/dispatchEvent.html

        * bindings/js/JSEventTarget.cpp:
        (WebCore::toJS): Add DOMWindow case.
        (WebCore::toEventTarget): Ditto.

        * dom/EventTarget.cpp:
        (WebCore::EventTarget::toDOMWindow): Added.
        * dom/EventTarget.h:

        * page/DOMWindow.cpp:
        (WebCore::DOMWindow::scriptExecutionContext): Added. Returns the document.
        (WebCore::DOMWindow::dispatchEvent): Added.
        * page/DOMWindow.h:
        (WebCore::DOMWindow::toDOMWindow):
        (WebCore::DOMWindow::refEventTarget):
        (WebCore::DOMWindow::derefEventTarget):
        * page/DOMWindow.idl:

LayoutTests:

2009-04-26  Sam Weinig  <sam@webkit.org>

        Reviewed by Dan Bernstein.

        Tests for https://bugs.webkit.org/show_bug.cgi?id=25412
        The Window object should be an EventTarget

        * fast/dom/Window/dispatchEvent-expected.txt: Added.
        * fast/dom/Window/dispatchEvent.html: Added.
        * fast/dom/Window/resources/dispatchEvent.js: Added.
        * fast/dom/Window/window-properties-expected.txt
        * platform/mac-leopard/fast/dom/Window/window-properties-expected.txt:



git-svn-id: svn://svn.chromium.org/blink/trunk@42882 bbb929c8-8fbe-4397-9dbb-9b2b20218538

2009-04-26  Sam Weinig  <sam@webkit.org>

        Reviewed by Eric Seidel.

        Remove scons-based build system.

        * JavaScriptCore.scons: Removed.

WebCore:

2009-04-26  Sam Weinig  <sam@webkit.org>

        Reviewed by Eric Seidel.

        Remove scons-based build system.

        * WebCore.scons: Removed.



git-svn-id: svn://svn.chromium.org/blink/trunk@42881 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Xan Lopez.

        Resolve relative URLs when queueing loads. This fixes
        post-goback-same-url.html timing out.

        * DumpRenderTree/gtk/LayoutTestControllerGtk.cpp:
        (LayoutTestController::queueLoad):

git-svn-id: svn://svn.chromium.org/blink/trunk@42880 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Gustavo Noronha.

        https://bugs.webkit.org/show_bug.cgi?id=25391
        G_SIGNAL_ACTION flag in WebKitWebView signals

        Most of the signals in WebKitGTK+ have the G_SIGNAL_ACTION flag,
        but I don't think this makes sense. The flag is meant for those
        signals that can be emitted by the user to trigger some action in
        the view, as if it were a normal API, which is not the case for
        most of them.

        * webkit/webkitdownload.cpp:
        (_WebKitDownloadPrivate::webkit_download_class_init):
        * webkit/webkitwebframe.cpp:
        * webkit/webkitwebinspector.cpp:
        (_WebKitWebInspectorPrivate::webkit_web_inspector_class_init):
        * webkit/webkitwebview.cpp:

git-svn-id: svn://svn.chromium.org/blink/trunk@42879 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Unreviewed, actually apply the right patch.

        * webkit/webkitwebview.cpp:
        (webkit_web_view_move_cursor):

git-svn-id: svn://svn.chromium.org/blink/trunk@42878 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Gustavo Noronha.

        Pass a variable to hold the return value of the 'move-cursor'
        signal, otherwise a crash can happen.

        * webkit/webkitwebview.cpp:
        (webkit_web_view_move_cursor):

git-svn-id: svn://svn.chromium.org/blink/trunk@42877 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Suggested by Darin Fisher.

        Improve the fix for bug 25355: Crash when Creating New Tab or New Window when set to open Same Page
        https://bugs.webkit.org/show_bug.cgi?id=25355
        rdar://problem/6823543

        * loader/FrameLoader.cpp:
        (WebCore::FrameLoader::loadItem): Allow shouldScroll to be true even if m_currentHistoryItem is 0.
        Not sure if when this case can really arise in practice, but it's good to match the original
        logic more closely.



git-svn-id: svn://svn.chromium.org/blink/trunk@42876 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Oliver Hunt.

        A width or height of zero for the destination or source rect of an
        image causes a not invertible pattern matrix.
        This crahes newer versions of Cairo and give some graphic gliches in
        Canvas.
        With this patch we check if there is something to draw and return if not.

        * platform/graphics/cairo/ImageCairo.cpp:
        (WebCore::BitmapImage::draw):


git-svn-id: svn://svn.chromium.org/blink/trunk@42875 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Go back to disabling accessibility tests as a whole, since we have
        no real way of checking them, and no accessibilityController
        implementation.

        * platform/gtk/Skipped:

git-svn-id: svn://svn.chromium.org/blink/trunk@42874 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Disable tests that need eventSender, not yet implemented in our
        DumpRenderTree.

        * platform/gtk/Skipped:

git-svn-id: svn://svn.chromium.org/blink/trunk@42873 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Darin Adler

        https://bugs.webkit.org/show_bug.cgi?id=25390

        Fix point mapping and hit testing through transforms and perspective
        with ENABLE_3D_RENDERING. Previously the code did the more efficient
        move(), rather than getTransformFromContainer(), when the object had
        no transform. However, this skipped the application of perspective
        when the immediate child of the perspective element had no transform
        itself.

        Test: transforms/3d/point-mapping/3d-point-mapping-3.html

        * rendering/RenderBox.cpp:
        (WebCore::RenderBox::mapLocalToContainer):
        (WebCore::RenderBox::mapAbsoluteToLocalPoint):
        * rendering/RenderLayer.cpp:
        (WebCore::RenderLayer::createLocalTransformState):
        Call shouldUseTransformFromContainer() to determine if we need to
        go through the more expensive getTransformFromContainer() code
        path.

        * rendering/RenderObject.h:
        * rendering/RenderObject.cpp:
        (WebCore::RenderObject::shouldUseTransformFromContainer):
        New method that indicates whether we need use getTransformFromContainer()
        when mapping points through renderers, and hit testing.

git-svn-id: svn://svn.chromium.org/blink/trunk@42872 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Sam Weinig

        Some *obvious* style cleanup in my last patch.

        * History/WebBackForwardList.mm:
        (bumperCarBackForwardHackNeeded):



git-svn-id: svn://svn.chromium.org/blink/trunk@42871 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Oliver Hunt

        <rdar://problem/6817607> BumperCar 2.2 crashes going back (invalid WebHistoryItem)

        BumperCar was holding a pointer to a WebHistoryItem they never retain, then later
        tried to go to it.  In some cases it would be dealloc'ed first.
        When WebHistoryItems were pure Objective-C they probably got away with this more often.
        With the WebCore/Obj-C mixed WebHistoryItems it's more likely to crash.

        * History/WebBackForwardList.mm:
        (bumperCarBackForwardHackNeeded):
        (-[WebBackForwardList backListWithLimit:]):  If this is BumperCar, hang on to the
          NSArray of WebHistoryItems until the next time this method is called.
        (-[WebBackForwardList forwardListWithLimit:]):  Ditto.

        * Misc/WebKitVersionChecks.h: Added WEBKIT_FIRST_VERSION_WITHOUT_BUMPERCAR_BACK_FORWARD_QUIRK.



git-svn-id: svn://svn.chromium.org/blink/trunk@42870 bbb929c8-8fbe-4397-9dbb-9b2b20218538

2009-04-25  Adele Peterson  <adele@apple.com>

        Reviewed by Oliver Hunt.

        Fix for <rdar://problem/6712771> REGRESSION(34681): Text is no longer underlined after delete
        https://bugs.webkit.org/show_bug.cgi?id=25396

        Test: editing/deleting/delete-br-013.html

        * editing/CompositeEditCommand.cpp: (WebCore::CompositeEditCommand::moveParagraphs):
         Only preserve an empty paragraph's style when moving paragraphs around if the selection is still 
         in an empty paragraph after the move occurs.  This was causing the empty paragraph's style to overwrite
         the previous paragraph's style during a delete of an empty paragraph.

LayoutTests:

2009-04-25  Adele Peterson  <adele@apple.com>

        Reviewed by Oliver Hunt.

        Test for <rdar://problem/6712771> REGRESSION(34681): Text is no longer underlined after delete
        https://bugs.webkit.org/show_bug.cgi?id=25396

        * editing/deleting/delete-br-013.html: Added.
        * platform/mac/editing/deleting/delete-br-013-expected.checksum: Added.
        * platform/mac/editing/deleting/delete-br-013-expected.png: Added.
        * platform/mac/editing/deleting/delete-br-013-expected.txt: Added.



git-svn-id: svn://svn.chromium.org/blink/trunk@42869 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Darin Adler.

        <rdar://problem/6808171> REGRESSION (3-4): Standalone media documents don't 
        properly display non-linear media

        If a media engine claims it can support the MIME type, WebCore now creates a <video>
        element for document mode media files instead of an <embed> element. Because WebCore's 
        QuickTime backed media players do not support every kind of media the QuickTime plug-in
        supports, and because it is not always possible to tell what type of media is in a file 
        without opening and parsing it, some types of media that used to be displayed properly
        by a plug-in are no longer supported properly. To fix this, if the media engine
        sees that it can not completely support a media file it now informs the media 
        document, which replaces the <video> element with an <embed>.
        
        r42301 landed support for OSX. This patch modifies those changes slighly and adds support
        for Windows.

        * loader/MediaDocument.cpp:
        (WebCore::MediaDocument::MediaDocument): Initialize m_replaceMediaElementTimer.
        (WebCore::MediaDocument::mediaElementSawUnsupportedTracks): Don't replace the <video> 
        element immediately.
        (WebCore::MediaDocument::replaceMediaElementTimerFired): Renamed from replaceVideoWithEmbed.
        Set body margin-width and margin-height to 0 as that is what a PluginDocument uses.
        * loader/MediaDocument.h:

        * platform/graphics/mac/MediaPlayerPrivateQTKit.mm:
        (WebCore::MediaPlayerPrivate::loadStateChanged): Don't do anything if m_hasUnsupportedTracks.
        (WebCore::MediaPlayerPrivate::rateChanged): Ditto.
        (WebCore::MediaPlayerPrivate::timeChanged): Ditto.
        (WebCore::MediaPlayerPrivate::didEnd): Ditto.
        (WebCore::MediaPlayerPrivate::repaint): Ditto.
        (WebCore::MediaPlayerPrivate::paint): Ditto.
        (WebCore::MediaPlayerPrivate::sawUnsupportedTracks): Set m_hasUnsupportedTracks before callback.

        * platform/graphics/win/MediaPlayerPrivateQuickTimeWin.cpp:
        (WebCore::MediaPlayerPrivate::MediaPlayerPrivate): Initialize m_totalTrackCount and m_hasUnsupportedTracks.
        (WebCore::MediaPlayerPrivate::updateStates): Call sawUnsupportedTracks if the movie has unsupported
        media type(s) or if it fails completely.
        (WebCore::MediaPlayerPrivate::sawUnsupportedTracks): New, disable the movie object and tell the
        media player client we won't play this movie correctly.
        (WebCore::MediaPlayerPrivate::didEnd): Don't do anything if m_hasUnsupportedTracks.
        (WebCore::MediaPlayerPrivate::setSize): Ditto.
        (WebCore::MediaPlayerPrivate::setVisible): Ditto.
        (WebCore::MediaPlayerPrivate::paint): Ditto.
        (WebCore::MediaPlayerPrivate::movieEnded): Ditto.
        (WebCore::MediaPlayerPrivate::movieLoadStateChanged): Ditto.
        (WebCore::MediaPlayerPrivate::movieTimeChanged): Ditto.
        (WebCore::MediaPlayerPrivate::movieNewImageAvailable): Ditto.
        * platform/graphics/win/MediaPlayerPrivateQuickTimeWin.h:

        * platform/graphics/win/QTMovieWin.cpp:
        (QTMovieWinPrivate::task): Stop the task timer if we were disabled during the load state
        change callback.
        (QTMovieWinPrivate::drawingComplete): Don't do anything if disabled.
        (QTMovieWin::QTMovieWin): Initialize m_disabled.
        (QTMovieWin::disableUnsupportedTracks): Return total number of tracks.
        (QTMovieWin::setDisabled): New, set m_disabled flag.
        * platform/graphics/win/QTMovieWin.h:



git-svn-id: svn://svn.chromium.org/blink/trunk@42868 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Xan Lopez.

        https://bugs.webkit.org/show_bug.cgi?id=24786
        WebKitDownload sometimes suggests peculiar filenames

        Refactor the emission of the download-requested signal so that we
        have less code duplication.

        * WebCoreSupport/ContextMenuClientGtk.cpp:
        (WebKit::ContextMenuClient::downloadURL):
        * WebCoreSupport/FrameLoaderClientGtk.cpp:
        (WebKit::FrameLoaderClient::startDownload):
        * webkit/webkitprivate.h:
        * webkit/webkitwebview.cpp:

git-svn-id: svn://svn.chromium.org/blink/trunk@42867 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Xan Lopez.

        [GTK] Error reporting
        https://bugs.webkit.org/show_bug.cgi?id=18344

        Fix the SOUP resource handle to report SOUP_HTTP_ERROR for Soup
        errors and G_IO_ERROR for gio errors.

        * platform/network/soup/ResourceHandleSoup.cpp:
        (WebCore::finishedCallback):
        (WebCore::ResourceHandle::startHttp):
        (WebCore::ResourceHandle::start):
        (WebCore::readCallback):
        (WebCore::openCallback):
        (WebCore::queryInfoCallback):
        (WebCore::ResourceHandle::startGio):

2009-04-25  Jan Michael Alonzo  <jmalonzo@webkit.org>

        Reviewed by Xan Lopez.

        [GTK] Error reporting
        https://bugs.webkit.org/show_bug.cgi?id=18344

        Update FrameLoad errors to use WebKitErrors.

        * WebCoreSupport/FrameLoaderClientGtk.cpp:
        (WebKit::FrameLoaderClient::cancelledError):
        (WebKit::FrameLoaderClient::blockedError):
        (WebKit::FrameLoaderClient::cannotShowURLError):
        (WebKit::FrameLoaderClient::interruptForPolicyChangeError):
        (WebKit::FrameLoaderClient::cannotShowMIMETypeError):
        (WebKit::FrameLoaderClient::fileDoesNotExistError):
        (WebKit::FrameLoaderClient::pluginWillHandleLoadError):
        (WebKit::FrameLoaderClient::shouldFallBack):
        * webkit/webkiterror.h: Added.
        * webkit/webkiterror.cpp: Added.

2009-04-25  Jan Michael Alonzo  <jmalonzo@webkit.org>

        Reviewed by Xan Lopez.

        [GTK] Error reporting
        https://bugs.webkit.org/show_bug.cgi?id=18344

        Add webkiterror to the build.

        * GNUmakefile.am:

git-svn-id: svn://svn.chromium.org/blink/trunk@42866 bbb929c8-8fbe-4397-9dbb-9b2b20218538

         Reviewed by Xan Lopez.

         [GTK] Error reporting
         https://bugs.webkit.org/show_bug.cgi?id=18344

         Add the default error page for installation.

         * GNUmakefile.am:

2009-04-25  Jan Michael Alonzo  <jmalonzo@webkit.org>

         Reviewed by Xan Lopez.

         Display a default error page for load errors.

         * WebCoreSupport/FrameLoaderClientGtk.cpp:
         (WebKit::FrameLoaderClient::dispatchDidFailLoad):
         * resources/error.html: Added.

git-svn-id: svn://svn.chromium.org/blink/trunk@42865 bbb929c8-8fbe-4397-9dbb-9b2b20218538

         Reviewed by Xan Lopez.

         [GTK] Error reporting
         https://bugs.webkit.org/show_bug.cgi?id=18344

         Add a WebKitWebFrame API to load alternate content for unreachable URLs.
         Also add a new signal "load-error" for handling load errors.

         * WebCoreSupport/FrameLoaderClientGtk.cpp:
         (WebKit::FrameLoaderClient::postProgressFinishedNotification):
         (WebKit::FrameLoaderClient::dispatchDidFailLoading):
         (WebKit::FrameLoaderClient::dispatchDidFailProvisionalLoad):
         (WebKit::FrameLoaderClient::dispatchDidFailLoad):
         * webkit/webkitwebframe.cpp:
         * webkit/webkitwebframe.h:
         * webkit/webkitwebview.cpp:

git-svn-id: svn://svn.chromium.org/blink/trunk@42864 bbb929c8-8fbe-4397-9dbb-9b2b20218538

git-svn-id: svn://svn.chromium.org/blink/trunk@42863 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Oliver Hunt.

        Gtk build fix - check if we have MADV_FREE before using it.

        * interpreter/RegisterFile.cpp:
        (JSC::RegisterFile::releaseExcessCapacity):
        * wtf/Platform.h:

git-svn-id: svn://svn.chromium.org/blink/trunk@42862 bbb929c8-8fbe-4397-9dbb-9b2b20218538

        Reviewed by Mark Rowe.

        Only set the new URL once for the request. Doing it a second time
        after the call to willSendRequest was causing crashes when
        redirected requests got cancelled.

        * platform/network/soup/ResourceHandleSoup.cpp:
        (WebCore::restartedCallback):

git-svn-id: svn://svn.chromium.org/blink/trunk@42861 bbb929c8-8fbe-4397-9dbb-9b2b20218538