13 years ago Get the size from the shared buffer.
andersca [Wed, 14 Mar 2007 17:38:51 +0000 (17:38 +0000)]
    Get the size from the shared buffer.

        * platform/graphics/cg/ImageSourceCG.cpp:
        * platform/graphics/cg/PDFDocumentImage.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20189 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
weinig [Wed, 14 Mar 2007 17:25:04 +0000 (17:25 +0000)]

        Reviewed by Hyatt, thumbs up by Darin.

        - test for http://bugs.webkit.org/show_bug.cgi?id=12782
          Reproducible crash in BidiContext::deref

        * fast/dynamic/anonymous-block-orphaned-lines-expected.checksum: Added.
        * fast/dynamic/anonymous-block-orphaned-lines-expected.png: Added.
        * fast/dynamic/anonymous-block-orphaned-lines-expected.txt: Added.
        * fast/dynamic/anonymous-block-orphaned-lines.html: Added.


        Reviewed by Hyatt, thumbs up by Darin.

        - fix http://bugs.webkit.org/show_bug.cgi?id=12782
          Reproducible crash in BidiContext::deref

        Test: fast/dynamic/anonymous-block-orphaned-lines.html

        * rendering/InlineBox.cpp:
        (WebCore::InlineBox::root): Added an assertion that we return a root inline box.
        * rendering/RenderBlock.cpp:
        (WebCore::RenderBlock::removeChild): Added code to adopt the line boxes of
        anonymous blocks being destroyed instead of leaving them orphaned, which
        is what caused this crash. The boxes will be deleted on the next layout, but
        this ensures consistency in the mean time.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20188 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Antti.
lars [Wed, 14 Mar 2007 13:38:49 +0000 (13:38 +0000)]
    Reviewed by Antti.

        Fix the Qt build.
        Add a getter to SharedBuffer that returns a reference to the internal
        Vector to avoid an extra copy of the data.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20187 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
antti [Wed, 14 Mar 2007 11:37:21 +0000 (11:37 +0000)]

        Reviewed by Mitz.

        Test case for <rdar://problem/5058774>
        REGRESSION: In Mail, caret appears oversized when typing in a To Do note

        * editing/style/highlight-expected.checksum: Added.
        * editing/style/highlight-expected.png: Added.
        * editing/style/highlight-expected.txt: Added.
        * editing/style/highlight.html: Added.


        Reviewed by Mitz.

        Fix <rdar://problem/5058774>
        REGRESSION: In Mail, caret appears oversized when typing in a To Do note

        Horizontal and vertical were switched. Was regression from

        * rendering/RootInlineBox.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20186 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Tweak the data() functions of stylesheets and scripts to be internally consis...
hyatt [Wed, 14 Mar 2007 11:02:56 +0000 (11:02 +0000)]
    Tweak the data() functions of stylesheets and scripts to be internally consistent.

        * loader/CachedCSSStyleSheet.cpp:
        * loader/CachedScript.cpp:
        * loader/CachedXSLStyleSheet.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20185 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Fix Radar 5050688.
hyatt [Wed, 14 Mar 2007 09:20:47 +0000 (09:20 +0000)]
    Fix Radar 5050688.

        For large animated GIFs, destroy and recreate the source for every animation frame.  This keeps
        the memory consumption down while giant images are animating.

        Reviewed by andersca

        * platform/graphics/BitmapImage.cpp:
        * platform/graphics/BitmapImage.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20184 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Make sure to use CFDataCreateWithBytesNoCopy where we can.
hyatt [Wed, 14 Mar 2007 08:49:55 +0000 (08:49 +0000)]
    Make sure to use CFDataCreateWithBytesNoCopy where we can.

        Reviewed by andersca

        * platform/graphics/cg/ImageSourceCG.cpp:
        * platform/graphics/cg/PDFDocumentImage.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20183 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoWebCore:
hyatt [Wed, 14 Mar 2007 08:07:59 +0000 (08:07 +0000)]

        Fix for bugzilla bug 13050 and also radar p1 5050645.

        This patch reworks resource loading to avoid having redundant buffers in the icon database and in cached
        images in the WebCore cache.  It also avoids overcopying in top-level image documents and in the icon

        There is now only one SharedBuffer for a resource and everybody observes that buffer now instead of ever
        making their own.  Even ImageIO uses the SharedBuffer while decoding.

        The page in 13050 dropped from 145mb down to 45mb of memory use with this change for a stunning savings
        of 100mb.

        Reviewed by olliej, mjs

        * WebCore.exp:
        * loader/CachedCSSStyleSheet.cpp:
        * loader/CachedCSSStyleSheet.h:
        * loader/CachedImage.cpp:
        * loader/CachedImage.h:
        * loader/CachedResource.cpp:
        * loader/CachedResource.h:
        * loader/CachedScript.cpp:
        * loader/CachedScript.h:
        * loader/CachedXSLStyleSheet.cpp:
        * loader/CachedXSLStyleSheet.h:
        * loader/DocLoader.cpp:
        * loader/ImageDocument.cpp:
        * loader/icon/IconDataCache.cpp:
        * loader/icon/IconDataCache.h:
        * loader/icon/IconDatabase.cpp:
        * loader/icon/IconDatabase.h:
        * loader/icon/IconLoader.cpp:
        * loader/icon/IconLoader.h:
        * loader/icon/SQLStatement.cpp:
        * loader/icon/SQLStatement.h:
        * loader/loader.cpp:
        * page/mac/WebCoreFrameBridge.mm:
        (-[WebCoreFrameBridge getData:andResponse:forURL:]):
        (-[WebCoreFrameBridge getAllResourceDatas:andResponses:]):
        * platform/SharedBuffer.h:
        * platform/graphics/BitmapImage.cpp:
        * platform/graphics/BitmapImage.h:
        * platform/graphics/Image.cpp:
        * platform/graphics/Image.h:
        * platform/graphics/ImageSource.h:
        * platform/graphics/cg/ImageSourceCG.cpp:
        * platform/graphics/cg/PDFDocumentImage.cpp:
        * platform/graphics/cg/PDFDocumentImage.h:
        * platform/graphics/mac/ImageMac.mm:
        * platform/graphics/svg/SVGImage.cpp:
        * platform/mac/PasteboardMac.mm:


        Fixes to ensure that the resource loader's shared buffer can always be used.

        Reviewed by olliej, mjs

        * Misc/WebIconDatabase.mm:
        (-[WebIconDatabase _convertToWebCoreFormat]):
        * WebCoreSupport/WebFrameLoaderClient.mm:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20182 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
justing [Wed, 14 Mar 2007 05:37:27 +0000 (05:37 +0000)]

        Reviewed by darin

        Gmail Editor: Applying alignment to selected text in message also applies alignment to signature

        * editing/style/5046875-1-expected.checksum: Added.
        * editing/style/5046875-1-expected.png: Added.
        * editing/style/5046875-1-expected.txt: Added.
        * editing/style/5046875-1.html: Added.
        * editing/style/5046875-2-expected.checksum: Added.
        * editing/style/5046875-2-expected.png: Added.
        * editing/style/5046875-2-expected.txt: Added.
        * editing/style/5046875-2.html: Added.
        moveParagraphs doesn't move unnecessary brs:
        * editing/style/create-block-for-style-002-expected.txt:
        * editing/style/create-block-for-style-004-expected.txt:
        * editing/style/create-block-for-style-001-expected.txt:
        * editing/style/create-block-for-style-009-expected.txt:
        * editing/style/create-block-for-style-010-expected.txt:
        moveParagraphs clears out unrendered text during the move:
        * editing/style/create-block-for-style-011-expected.txt:
        * editing/style/create-block-for-style-007-expected.txt:
        * editing/style/create-block-for-style-012-expected.txt:


        Reviewed by darin

        Gmail Editor: Applying alignment to selected text in message also applies alignment to signature

        * editing/ApplyStyleCommand.cpp:
        (WebCore::ApplyStyleCommand::doApply): Don't call applyBlockStyle unless
        there is a block style to apply.
        (WebCore::ApplyStyleCommand::applyBlockStyle): Don't do the remove step.
        It was unnecessary and removed properties from blocks that could contain
        content outside the range being operated on (added a testcase).
        (WebCore::ApplyStyleCommand::addBlockStyleIfNeeded): Used an early return
        instead of if-nesting.
        * editing/ApplyStyleCommand.h:
        * editing/CompositeEditCommand.cpp:
        Return the new block, if one was created.  Use moveParagraphs to move
        paragraphs into the new block, instead of moving nodes.  The old code moved
        too much (added a testcase).
        * editing/CompositeEditCommand.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20181 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago2007-03-13 Oliver Hunt <oliver@apple.com>
oliver [Wed, 14 Mar 2007 05:00:07 +0000 (05:00 +0000)]
2007-03-13  Oliver Hunt  <oliver@apple.com>

        Reviewed by Brady.
        To fix <rdar://problem/5044366> we now pass a NSString
        representation of the URL extracted with _web_originalDataAsString
        instead of relying on [NSURL absoluteString] in the bridge

        * page/mac/WebCoreFrameBridge.h:
        * page/mac/WebCoreFrameBridge.mm:
        (-[WebCoreFrameBridge getData:andResponse:forURL:]):
        Modify subresourceForURL to take NSString argument
        so we can avoid [NSURL absoluteString]

        * WebView/WebDataSource.mm:
        (-[WebDataSource subresourceForURL:]):
        Test for <rdar://problem/5044366> make sure we can create
        a WebArchive from a page with a partially decoded url.

        * webarchive/archive-with-unencoded-url-expected.txt: Added.
        * webarchive/archive-with-unencoded-url.html: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20180 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Rubberstamped by Alice
beidson [Wed, 14 Mar 2007 04:44:17 +0000 (04:44 +0000)]
    Rubberstamped by Alice

        Meant to be part of my previous checkin... pruning unused code from WebKit

        * WebView/WebFrame.mm: Removed _canCachePage
        * WebView/WebFrameInternal.h: Ditto

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20179 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Anders
beidson [Wed, 14 Mar 2007 03:59:08 +0000 (03:59 +0000)]
    Reviewed by Anders

        <rdar://problem/5048818> - REGRESSION: Incompletely loaded resources being saved to the object cache

        Due to a subtle change in loader behavior back in 10904, we would stop all loaders before calling
        didFail() on them in the Cache loader.  As a result, we basically cleared all of the Subresource Loaders
        out of the Cache loader before more properly failing them as errored out.  The result?  Partially loaded
        resources being cached.

        Since Loader::didFail() both calls error() on the object *and* removes the loader, the solution is to call
        didFail() for all cancelled loaders instead of *only* removing them from the set of active loaders.

        In addition, pages that didn't completely load were being saved to the back/forward cache.  To fix that,
        I added a null check on the DocumentLoader's error to see if the page ended in an error, or did indeed
        completely load.

        Note that the layout test for this - if possible - will require other enhancements including possibly adding
        support for window.stop().  That task is documented in <rdar://problem/5061826>

        * loader/FrameLoader.cpp:
        (WebCore::FrameLoader::provisionalLoadStarted): Fixed a few bugs relating to my original BFCache rewrite to
          more perfectly restore the original behavior - including only caching HTML documents via the
          m_client->canCachePage() call
        (WebCore::FrameLoader::canCachePage): Don't make the call to m_client->canCachePage() as that serves a different
          - Check the mainDocumentError to see if the load ended in error as a further criteria in determining the
          cachability of a page

        * loader/loader.cpp:
        (WebCore::Loader::cancelRequests): Call didFail(cancelledError()) instead of just removing the loaders from the
          loaders-in-progress set.  This adds the effect of properly cleaning up the cached object.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20178 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoWebCore:
bdakin [Wed, 14 Mar 2007 03:56:43 +0000 (03:56 +0000)]
        Rendering part reviewed by Hyatt. Editing part consulted with and
        rubber stamped by Justin and Harrison.

        Fix for <rdar://problem/5025925> A hang occurs in Safari when
        attempting to print page at http://www.pcadvisor.co.uk

        * rendering/RenderBlock.cpp:
        RenderBlock::makeChildrenNonInline() takes a block's inline
        children and turns them into block children. If the children had
        line boxes, those boxes were being leaked. In the layout test I
        added with the change (and at pcadvisor.co.uk during printing)
        children were being made non-inline, and then they were being made
        inline again. This meant that some of the children ended up
        pointing to totally stale line boxes that are normally just leaked.
        This caused an infinite loop in RenderFlow::destroy(). This patch
        simply deletes everyone's line boxes in

        * editing/InsertParagraphSeparatorCommand.cpp:
        (WebCore::InsertParagraphSeparatorCommand::doApply): The other part
        of this fix is that I added a call to updateLayout in
        InsertParagraphSeparatorCommand::doApply(). One layout test
        (editing/spelling/spelling.html) was changed by my patch to
        RenderBlock. doApply() inserts a node into the render tree. In at
        least one case in spelling.html, that caused some line boxes to be
        deleted. Back in doApply() this meant that the RenderTree was out-
        of-date, and we mistakenly thought we were at the end of the
        paragraph. This caused us to insert a RenderBR() at the end of the
        tree instead of an empty RenderText(). No one seems to know exactly
        why we insert either, or if the change is necessarily a problem. It
        is clear, though, that the RenderTree in doApply() is out-of-date
        after inserting the node and deleting some line boxes, so it seems
        prudent to call into updateLayout().

        Reviewed by Hyatt.

        Test for <rdar://problem/5025925> A hang occurs in Safari when
        attempting to print page at http://www.pcadvisor.co.uk

        * fast/block/float/nestedAnonymousBlocks-expected.checksum: Added.
        * fast/block/float/nestedAnonymousBlocks-expected.png: Added.
        * fast/block/float/nestedAnonymousBlocks-expected.txt: Added.
        * fast/block/float/nestedAnonymousBlocks.html: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20177 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Anders.
aroben [Wed, 14 Mar 2007 02:37:57 +0000 (02:37 +0000)]
    Reviewed by Anders.

        * platform/FontData.h: Added m_isSystemFont parameter to match NSFont.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20176 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago <rdar://problem/5031181> cntl-k at end of paragraph adds nothing to the kill...
harrison [Wed, 14 Mar 2007 00:39:53 +0000 (00:39 +0000)]
    <rdar://problem/5031181> cntl-k at end of paragraph adds nothing to the kill ring
        <rdar://problem/5031189> REGRESSION: cntl-y yanks only the most recently killed content

        Updated results.

        * editing/deleting/delete-to-end-of-paragraph-expected.checksum:
        * editing/deleting/delete-to-end-of-paragraph-expected.txt:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20175 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agotypo fix. thanks aroben!
bdakin [Tue, 13 Mar 2007 23:55:32 +0000 (23:55 +0000)]
typo fix. thanks aroben!

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20174 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoWebCore:
bdakin [Tue, 13 Mar 2007 23:51:25 +0000 (23:51 +0000)]
        Reviewed by Maciej.

        Export DocumentLoader::setFrame(). Part for fix for <rdar://
        problem/4277074> 8F32: Help Viewer crashed on clicking link -
        KHTMLView::viewportMouseReleaseEvent (12647)

        * WebCore.exp:

        Reviewed by Maciej.

        Fix for <rdar://problem/4277074> 8F32: Help Viewer crashed on
        clicking link - KHTMLView::viewportMouseReleaseEvent (12647)

        Re-set the DocumentLoader's frame when loading it from the page
        cache before setting the document view.

        * WebCoreSupport/WebFrameLoaderClient.mm:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20173 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoof -> be
darin [Tue, 13 Mar 2007 23:25:28 +0000 (23:25 +0000)]
of -> be

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20171 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
darin [Tue, 13 Mar 2007 23:18:23 +0000 (23:18 +0000)]

        Reviewed by Maciej.

        - tests for http://bugs.webkit.org/show_bug.cgi?id=12794
          <rdar://problem/5028154> REGRESSION: TripTik planner at aaa.com never
          finishes loading due to unclosed canvas tag (12794)

        * fast/canvas/canvas-hides-fallback-expected.txt: Added.
        * fast/canvas/canvas-hides-fallback.html: Added.
        * fast/canvas/script-inside-canvas-fallback-expected.txt: Added.
        * fast/canvas/script-inside-canvas-fallback.html: Added.
        * fast/canvas/unclosed-canvas-1-expected.txt: Added.
        * fast/canvas/unclosed-canvas-1.html: Added.
        * fast/canvas/unclosed-canvas-2-expected.txt: Added.
        * fast/canvas/unclosed-canvas-2.html: Added.
        * fast/canvas/unclosed-canvas-3-expected.txt: Added.
        * fast/canvas/unclosed-canvas-3.html: Added.
        * fast/canvas/unclosed-canvas-4-expected.txt: Added.
        * fast/canvas/unclosed-canvas-4.html: Added.


        Reviewed by Maciej.

        - fix http://bugs.webkit.org/show_bug.cgi?id=12794
          <rdar://problem/5028154> REGRESSION: TripTik planner at aaa.com never
          finishes loading due to unclosed canvas tag (12794)

        Change <canvas> elements so that their contents are parsed normally,
        but not rendered. This change fixes the bug, because normal parsing
        rules close the <canvas> element in that case. The special parser
        stuff was just getting in the way.

        Also do some basic cleanup to the HTML parser. This was motivated by
        an earlier version of this patch that made even more changes to the
        parser, but the cleanup is still worth landing.

        Test: fast/canvas/canvas-hides-fallback.html
        Test: fast/canvas/script-inside-canvas-fallback.html
        Test: fast/canvas/unclosed-canvas-1.html
        Test: fast/canvas/unclosed-canvas-2.html
        Test: fast/canvas/unclosed-canvas-3.html
        Test: fast/canvas/unclosed-canvas-4.html

        * html/HTMLCanvasElement.h: Added a data member to keep track of whether the
        renderer is a RenderHTMLCanvas or not.
        * html/HTMLCanvasElement.cpp:
        (WebCore::HTMLCanvasElement::createRenderer): If JavaScript is enabled, create
        a RenderHTMLCanvas. If it's not, let the default code create the default type
        of renderer, which will result in fallback content being visible. The
        RenderHTMLCanvas class already hides all of its children. Set the m_rendererIsCanvas
        boolean accordingly. Since the actual storage for the canvas is allocated lazily
        when you actually get a drawing context, we don't need to do anything special
        to prevent it when JavaScript is disabled; the relevant functions won't be called.
        (WebCore::HTMLCanvasElement::reset): Protect the code that manipulates the
        RenderHTMLCanvas with a check of m_rendererIsCanvas. This is the only code inside
        the DOM element that relies on the renderer type.

        * html/HTMLParser.h: Removed unneeded includes. Marked HTMLParser as
        Noncopyable. Changed the Document parameter to the constructor to instead
        of HTMLDocument. Renamed discard_until to m_skipModeTag for clarity.
        Removed unused noSpaces function and unneeded public doc() function.
        Moved data members all down to the end so you can see them together in order.
        Renamed map to m_currentMapElement and isindex to m_isindexElement.
        Removed unused end and headLoaded data members.  Renamed m_fragment to
        m_isParsingFragment to make it clearer that it's a boolean, not a fragment.

        * html/HTMLParser.cpp:
        (WebCore::HTMLParser::HTMLParser): Changed to use member construction
        syntax instead of calling reset(). This is especially helpful in the
        fragment case, where calling reset() later on is illegal, so not using
        it in the constructor lets us assert.
        (WebCore::HTMLParser::~HTMLParser): Did an explicit deref instead of
        calling setCurrent for its side effect.
        (WebCore::HTMLParser::reset): Updated for member name changes and removal
        and to use document instead of doc().
        (WebCore::HTMLParser::setCurrent): Use document instead of doc().
        (WebCore::HTMLParser::setSkipMode): Added. No longer inline. Now sets the
        m_inCanvasBeforeFirstOpenTag data member to false.
        (WebCore::HTMLParser::parseToken): Tightened up the skip mode logic at the
        top of the function, and added a FIXME about the strange case there where
        we don't skip yet stay in skip mode. Updated for renaming and doc().
        (WebCore::HTMLParser::insertNode): Updated for renaming and doc().
        (WebCore::HTMLParser::handleError): Ditto.
        (WebCore::HTMLParser::framesetCreateErrorCheck): Ditto.
        (WebCore::HTMLParser::isindexCreateErrorCheck): Changed to use RefPtr.
        (WebCore::HTMLParser::noscriptCreateErrorCheck): Updated for renaming and doc().
        (WebCore::HTMLParser::mapCreateErrorCheck): Ditto.
        (WebCore::HTMLParser::getNode): Removed the special case for canvas here.
        Canvas fallback is now handled in the DOM, not the parser. Updated for
        renaming and doc().
        (WebCore::HTMLParser::allowNestedRedundantTag): Changed a #define into a C++
        (WebCore::HTMLParser::processCloseTag): Updated for renaming and doc().
        (WebCore::HTMLParser::isInline): Ditto.
        (WebCore::HTMLParser::tagIsOnStack): Added. Used by new canvas logic.
        (WebCore::HTMLParser::popBlock): Updated for renaming and doc(). Also renamed
        the local variable Elem to elem.
        (WebCore::HTMLParser::createHead): Ditto.
        (WebCore::HTMLParser::handleIsindex): Changed to use RefPtr.
        (WebCore::HTMLParser::startBody): Updated for renaming and doc().
        (WebCore::HTMLParser::finished): Ditto.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20170 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Two more cleanup fixes to the cache. Don't call destroyDecodedData in the...
hyatt [Tue, 13 Mar 2007 23:07:55 +0000 (23:07 +0000)]
    Two more cleanup fixes to the cache.  Don't call destroyDecodedData in the BitmapImage destructor, since
        clearing the image source and calling setData on it again causes it to do an extra copy of the encoded
        data.  Since we're about to be destroyed this is just wasteful.

        When the cache prunes, don't allow it to destroy the decoded data of an image that is still actively loading,
        since we've established that ImageIO can actually crash if you yank the rug out from under it like that.

        Reviewed by ggaren

        * loader/Cache.cpp:
        * platform/graphics/BitmapImage.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20169 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Try fixing the Qt build.
andersca [Tue, 13 Mar 2007 22:46:01 +0000 (22:46 +0000)]
    Try fixing the Qt build.

        * editing/Editor.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20168 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Geoff.
thatcher [Tue, 13 Mar 2007 22:40:16 +0000 (22:40 +0000)]
    Reviewed by Geoff.

        <rdar://problem/5057117> Spoof user agent on Yahoo.com with Safari and WebKit as version 4xx

        * Plugins/WebBaseNetscapePluginView.mm:
        (-[WebBaseNetscapePluginView userAgent]): Stop using the deprecated lossyCString method.
        * WebCoreSupport/WebFrameBridge.mm: Removed dead code, userAgentForURL: wasn't used.
        * WebCoreSupport/WebFrameLoaderClient.mm:
        (WebFrameLoaderClient::userAgent): Call WebView's _userAgentForURL:.
        * WebView/WebView.mm:
        (-[WebView _cachedResponseForURL:]): Call userAgentForURL: instead of _userAgent.
        (-[WebView userAgentForURL:]): Call _userAgentForURL:.
        (-[WebView _userAgentWithApplicationName:andWebKitVersion:]): New method to construct a UA.
        (-[WebView _computeUserAgent]): Ractored out into _userAgentWithApplicationName:andWebKitVersion:
        (-[WebView _userAgentForURL:]): Tail compare for Yahoo.com, and return a UA with an older WebKit version.
        * WebView/WebViewInternal.h: Declare _userAgentForURL:.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20167 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
harrison [Tue, 13 Mar 2007 22:31:43 +0000 (22:31 +0000)]

        Reviewed by Justin.

        <rdar://problem/5031181> cntl-k at end of paragraph adds nothing to the kill ring
        <rdar://problem/5031189> REGRESSION: cntl-y yanks only the most recently killed content

        * editing/pasteboard/emacs-ctrl-k-y-001-expected.checksum: Added.
        * editing/pasteboard/emacs-ctrl-k-y-001-expected.png: Added.
        * editing/pasteboard/emacs-ctrl-k-y-001-expected.txt: Added.
        * editing/pasteboard/emacs-ctrl-k-y-001.html: Added.


        Reviewed by Justin.

        <rdar://problem/5031181> cntl-k at end of paragraph adds nothing to the kill ring
        <rdar://problem/5031189> REGRESSION: cntl-y yanks only the most recently killed content

        For rdar://5031181, properly extend the selection before the killring handling, and
        make sure plainText of that selection returns a linefeed.

        For rdar://5031189, restore Editor::deleteRange() code that continued current killring,
        even though the range deletion implicitly stopped it via changing the selection.

        A byproduct of this change is the elimination of RUNDFINDER vs CONTENT TextIterator. The
        only difference between the two was whether to emit a newline when the range started
        with a blockflow element. No callers actually need that any more.

        Tests added:
        * editing/pasteboard/emacs-ctrl-k-y-001-expected.checksum: Added.
        * editing/pasteboard/emacs-ctrl-k-y-001-expected.png: Added.
        * editing/pasteboard/emacs-ctrl-k-y-001-expected.txt: Added.
        * editing/pasteboard/emacs-ctrl-k-y-001.html: Added.

        * editing/Editor.cpp:
        Clear the "start new kill ring sequence" setting, because it was set to true
        when the selection was updated by deleting the range.

        If extending the selection to the end of paragraph resulted in a caret selection,
        extend by character, to handle the case when the selection started as a caret at
        the end of paragraph.

        * editing/TextIterator.cpp:
        Initialize new member variables for tracking handling of the beginning of the range.

        Call representNodeOffsetZero on the m_endContainer.
        Move visibility checks into handleTextNode and handleReplacedElement.

        Call emitText.

        Moved visibility check into here.


        New. Emits proper sequence when encountering offset 0 of a node, including the
        m_endContainer. Started with code from handleNonTextNode.

        Call representNodeOffsetZero.

        Similar to shouldRepresentNodeOffsetZero, do not emit the newline if the node
        was collapsed, and before any other emitted content.


        New. Consolidates code used by handleText and handleTextBox.

        Removed RUNFINDER.

        Removed RUNFINDER.



        * editing/TextIterator.h:
        Added member variables for tracking handling of the beginning of the range.
        Eliminated concept of RUNDFINDER vs CONTENT TextIterator.

        * editing/visible_units.cpp:
        Eliminated concept of RUNDFINDER vs CONTENT TextIterator.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20166 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Clean up the null image case in CachedImage::data to make sure the size total...
hyatt [Tue, 13 Mar 2007 22:11:09 +0000 (22:11 +0000)]
    Clean up the null image case in CachedImage::data to make sure the size totals will stay accurate.
        I'm not convinced this case can even be hit, but I'm cleaning it up just in case.

        * loader/CachedImage.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20165 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Add a test that currently crashes the Qt build.
lars [Tue, 13 Mar 2007 22:10:54 +0000 (22:10 +0000)]
    Add a test that currently crashes the Qt build.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20164 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reorder the call to allReferencesRemoved, since otherwise the live object...
hyatt [Tue, 13 Mar 2007 22:03:36 +0000 (22:03 +0000)]
    Reorder the call to allReferencesRemoved, since otherwise the live object size will become

        * loader/CachedResource.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20163 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by George.
lars [Tue, 13 Mar 2007 22:00:09 +0000 (22:00 +0000)]
    Reviewed by George.

        Don't try to load <object> tags with an invalid url.
        Fixes LayoutTests/fast/dom/object-plugin-hides-properties.html
        which hit an assertion in the frameloader.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20162 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
antti [Tue, 13 Mar 2007 21:41:49 +0000 (21:41 +0000)]

        Reviewed by Darin.

        REGRESSION: Typing tab key fails to insert a tab character in Google Docs editable area

        * editing/inserting/typing-tab-designmode-expected.txt: Added.
        * editing/inserting/typing-tab-designmode-forms-expected.checksum: Added.
        * editing/inserting/typing-tab-designmode-forms-expected.png: Added.
        * editing/inserting/typing-tab-designmode-forms-expected.txt: Added.
        * editing/inserting/typing-tab-designmode-forms.html: Added.
        * editing/inserting/typing-tab-designmode.html: Added.


        Reviewed by Darin.

        Fix http://bugs.webkit.org/show_bug.cgi?id=11083
        REGRESSION: Typing tab key fails to insert a tab character in Google Docs editable area

        Allow inserting tabs in designMode. Backtab behaves like before (matches FF).

        * page/EventHandler.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20159 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoFixes builds using buildit.
thatcher [Tue, 13 Mar 2007 21:31:18 +0000 (21:31 +0000)]
Fixes builds using buildit.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20158 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by .
kmccullo [Tue, 13 Mar 2007 20:25:24 +0000 (20:25 +0000)]
    Reviewed by .

        - Adding expected failures until the are truly fixed.
        - rdar://problem/5060302

        * tests/mozilla/expected.html:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20157 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Fix two counting errors in the decoded size of objects in the WebCore cache...
hyatt [Tue, 13 Mar 2007 20:18:51 +0000 (20:18 +0000)]
    Fix two counting errors in the decoded size of objects in the WebCore cache.  We need to explicitly
        destroy decoded data when clearing out a partially loaded image that had an error, since the destructor
        (although it does destroy the data) disconnects the observer so that the notification doesn't happen.

        Some CachedImages aren't in the cache (like image documents).  When the decoded size of such an image
        changes we should not notify the cache, since - newsflash - we aren't actually in it.

        Reviewed by ggaren

        * loader/CachedImage.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20156 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoSilly wrapping change for testing a post-commit hook change.
thatcher [Tue, 13 Mar 2007 19:33:00 +0000 (19:33 +0000)]
Silly wrapping change for testing a post-commit hook change.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20154 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoFix 99% of the crashes in the loader I've been seeing lately.
lars [Tue, 13 Mar 2007 19:20:34 +0000 (19:20 +0000)]
Fix 99% of the crashes in the loader I've been seeing lately.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20153 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Tim Hatcher and John Sullivan.
darin [Tue, 13 Mar 2007 18:22:43 +0000 (18:22 +0000)]
    Reviewed by Tim Hatcher and John Sullivan.

        - fix <rdar://problem/4915303> CrashTracer: 36 crashes in Safari at
          com.apple.AppKit: -[NSView getRectsBeingDrawn:count:] + 502

        * page/mac/FrameMac.mm: (WebCore::Frame::imageFromRect):
        This was calling drawRect: directly, but NSView's getRectsBeingDrawn:count:
        method was never really safe to call unless it was AppKit that called your
        drawRect: method. Changed it to call drawSingleRect: instead. A little ugly,
        but seems to work and will almost certainly fix the bug.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20152 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Maciej.
weinig [Tue, 13 Mar 2007 16:53:26 +0000 (16:53 +0000)]
    Reviewed by Maciej.

        Fixes: http://bugs.webkit.org/show_bug.cgi?id=12974

        Call forgetGenericContext in JSSVGPathSeg destructor, otherwhise
        we'll hit an ASSERT in a debug build, when running svg/custom/js-update-path-changes.svg
        a few dozen times in a single WebKit instance. The ASSERT is good and just warns that
        there was already a generic context pointer registered, and the new "to be registered"
        object already exists, but pointing to a different object. That's because garbage collection
        calls the JSSVGPathSeg destructor, but that didn't cleanup the generic context map.

        Only JSSVGPathSeg is hit by this problem, as it's the only non SVGAnimated* type
        using the generic context system while using a custom JSSVGPathSegList implementation.

        Also cleanup JSSVGPathSegListCustom code to call the static forgetGenericContext
        method instead of doing the same using custom code.

        * bindings/js/JSSVGPathSegListCustom.cpp:
        * bindings/scripts/CodeGeneratorJS.pm:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20151 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago * fast/js/kde/RegExp-expected.txt: Fix the one other test affected by
darin [Tue, 13 Mar 2007 14:47:11 +0000 (14:47 +0000)]
    * fast/js/kde/RegExp-expected.txt: Fix the one other test affected by
        the js-test-pre.js change.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20149 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
darin [Tue, 13 Mar 2007 14:19:32 +0000 (14:19 +0000)]

        Reviewed by Adele.

        - test for http://bugs.webkit.org/show_bug.cgi?id=12595
          <rdar://problem/4722863> REGRESSION: Can't add item to cart at lnt.com
          (JS type error) (12595)

        * fast/forms/old-names-expected.txt: Added.
        * fast/forms/old-names.html: Added.

        - check in files that were somehow left out in the last check-in

        * fast/js/resources/js-test-pre.js: Improve the format of the failure message when
        both the real results and expected results are strings.
        * fast/dom/wrapper-classes-expected.txt: Changed back to use the new format.


        Reviewed by Adele.

        - fix http://bugs.webkit.org/show_bug.cgi?id=12595
          <rdar://problem/4722863> REGRESSION: Can't add item to cart at lnt.com
          (JS type error) (12595)

        Test: fast/forms/old-names.html

        * bindings/js/JSHTMLFormElementCustom.cpp:
        (WebCore::JSHTMLFormElement::canGetItemsForName): If the form collection has
        nothing for a given name, try the form's oldNamedElement function.
        (WebCore::JSHTMLFormElement::nameGetter): Ditto.

        * bindings/js/kjs_dom.h: Removed the DOMNamedNodesCollection. Instead we will use
        a class derived from NodeList.
        * bindings/js/kjs_dom.cpp: Ditto.

        * bindings/js/kjs_html.cpp:
        (KJS::VectorNodeList::VectorNodeList): Added. Constructor for a new class derived
        from NodeList to be used for the named items result from a collection -- uses a
        vector of node pointers.
        (KJS::VectorNodeList::length): Added.
        (KJS::VectorNodeList::item): Added.
        (KJS::JSHTMLCollection::getNamedItems): Use VectorNodeList and the existing wrapper
        for NodeList rather than a custom JavaScript class, DOMNamedNodesCollection.

        * dom/ChildNodeList.h:
        * dom/ChildNodeList.cpp:
        (WebCore::ChildNodeList::ChildNodeList): Updated to derive from TreeNodeList,
        since NodeList is now a simpler class.
        (WebCore::ChildNodeList::elementMatches): Updated for name and parameter change.

        * dom/NameNodeList.h:
        * dom/NameNodeList.cpp:
        (WebCore::NameNodeList::NameNodeList): Updated to derive from TreeNodeList,
        since NodeList is now a simpler class.
        (WebCore::NameNodeList::rootNodeAttributeChanged): Updated for name and
        parameter change.

        * dom/Node.h: Change register/unregister functions to take TreeNodeList.
        * dom/Node.cpp:
        (WebCore::TagNodeList::TagNodeList): Updated to derive from TreeNodeList,
        since NodeList is now a simpler abstract class.
        (WebCore::TagNodeList::elementMatches): Updated for name and parameter change.
        (WebCore::Node::registerNodeList): Changed type from NodeList to TreeNodeList.
        (WebCore::Node::unregisterNodeList): Ditto.

        * dom/NodeList.h: Broke NodeList into a simpler base class and a derived class
        with the machinery for iterating a tree, called TreeNodeList.
        * dom/NodeList.cpp:
        (WebCore::NodeList::~NodeList): Added.
        (WebCore::NodeList::itemWithName): Factored out of the old itemWithName.
        (WebCore::TreeNodeList::TreeNodeList): Renamed from NodeList.
        (WebCore::TreeNodeList::~TreeNodeList): Ditto.
        (WebCore::TreeNodeList::recursiveLength): Ditto.
        (WebCore::TreeNodeList::itemForwardsFromCurrent): Ditto.
        (WebCore::TreeNodeList::itemBackwardsFromCurrent): Ditto.
        (WebCore::TreeNodeList::recursiveItem): Ditto.
        (WebCore::TreeNodeList::itemWithName): Factored half of this into this function,
        the other half in NodeList::itemWithName.
        (WebCore::TreeNodeList::rootNodeAttributeChanged): Added. No longer inline.
        (WebCore::TreeNodeList::rootNodeChildrenChanged): Renamed from NodeList.

        * html/HTMLFormElement.h: Added formElementNameChanged and oldNamedElement
        fucntions, and a map called m_oldNames. Also removed m_boundary, which I
        thought I had already done.
        * html/HTMLFormElement.cpp:
        (WebCore::HTMLFormElement::HTMLFormElement): Initialize m_oldNames to 0.
        Switched the rest of the members to initialization syntax.
        (WebCore::HTMLFormElement::~HTMLFormElement): Delete m_oldNames.
        (WebCore::HTMLFormElement::formElementNameChanged): Added. Stores a reference
        to one element under each of its old names.
        (WebCore::HTMLFormElement::oldNamedElement): Added. Returns the old element
        that once had a given name.

        * html/HTMLGenericFormElement.h:
        * html/HTMLGenericFormElement.cpp:
        (WebCore::HTMLGenericFormElement::parseMappedAttribute): When the name
        attribute changes, tell the form about the old name.
        (WebCore::HTMLGenericFormElement::insertedIntoTree): When telling a form
        about an element, also store away the old name so that we can use it
        when the name changes later.

        * html/HTMLInputElement.cpp:
        (WebCore::HTMLInputElement::parseMappedAttribute): Added a call to the
        base class in the nameAttr case, so the code in HTMLGenericFormElement
        above will get called in the input element case.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20148 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Lars.
weinig [Tue, 13 Mar 2007 14:17:32 +0000 (14:17 +0000)]
    Reviewed by Lars.

        Update text in tests to match expected output.

        * fast/dom/gc-6-expected.txt:
        * fast/dom/gc-6.html:
        * fast/dom/gc-7-expected.txt:
        * fast/dom/gc-7.html:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20147 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
antti [Tue, 13 Mar 2007 13:36:57 +0000 (13:36 +0000)]

        Reviewed by Alexey.

        Test for DOS using composed character.

        * fast/text/large-text-composed-char-dos-expected.txt: Added.
        * fast/text/large-text-composed-char-dos.html: Added.


        Reviewed by Alexey.

        Alexey spotted a DOS by using string of 64k unbreakable character in
        fix for http://bugs.webkit.org/show_bug.cgi?id=12833

        * dom/Text.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20146 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Tim H.
thatcher [Tue, 13 Mar 2007 12:54:03 +0000 (12:54 +0000)]
    Reviewed by Tim H.

        - fix test results for http://bugs.webkit.org/show_bug.cgi?id=13054
          Layout tests results require updates after JavaScript wrapper class name fixes

        * fast/dom/gc-6-expected.txt: Renamed DOMHTMLElement to HTMLElement.
        * fast/dom/gc-7-expected.txt: Ditto.
        * fast/dom/prototype-chain-expected.txt: Renamed DOMHTMLElementPrototype to HTMLElementPrototype.
        * fast/dom/undetectable-document-all-expected.txt: Renamed Collection to HTMLCollection.
        * fast/dom/wrapper-classes-expected.txt: Updated for change to output for string comparison failure
        in fast/js/resources/js-test-pre.js.
        * fast/events/mouseclick-target-and-positioning-expected.txt: Renamed DOMHTMLElement to HTMLElement.
        * fast/tokenizer/external-script-document-write-expected.checksum: Ditto.
        * fast/tokenizer/external-script-document-write-expected.png: Ditto.
        * fast/tokenizer/external-script-document-write-expected.txt: Ditto.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20145 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoTweaked again.
ddkilzer [Tue, 13 Mar 2007 10:13:49 +0000 (10:13 +0000)]
Tweaked again.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20144 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoAdded bug references and a period.
ddkilzer [Tue, 13 Mar 2007 10:11:34 +0000 (10:11 +0000)]
Added bug references and a period.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20143 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoFix the Qt build.
lars [Tue, 13 Mar 2007 10:03:04 +0000 (10:03 +0000)]
Fix the Qt build.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20142 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago2007-03-13 Mark Rowe <mrowe@apple.com>
bdash [Tue, 13 Mar 2007 09:21:41 +0000 (09:21 +0000)]
2007-03-13  Mark Rowe  <mrowe@apple.com>

        Reviewed by Maciej.

        Have the nightly launcher prefer Safari found in /Applications or ~/Applications
        if present, otherwise fall back on using LaunchServices to locate it elsewhere
        on the system.  The motivation for this is to make the behaviour of the nightly
        builds more predictable on machines with multiple copies of Safari present.

        * WebKitLauncher/main.m:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20141 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Darin.
rwlbuis [Tue, 13 Mar 2007 08:00:29 +0000 (08:00 +0000)]
    Reviewed by Darin.

        WebKit does not support xlink:show attributes

        Make an exception for non-empty targets, these should be opened
        in a new window, unless the value is _self.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20140 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Double the encoded size of images for now. We do this to account for a bug...
hyatt [Tue, 13 Mar 2007 07:32:07 +0000 (07:32 +0000)]
    Double the encoded size of images for now.  We do this to account for a bug in ImageIO where they hold
        a separate copy of image data.

        Reviewed by aroben

        * loader/CachedImage.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20139 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
harrison [Tue, 13 Mar 2007 04:24:28 +0000 (04:24 +0000)]

        Reviewed by Darin.

        <rdar://problem/4743256> ctrl-y key binding (yank) should do nothing when kill ring is empty

        * editing/pasteboard/emacs-cntl-y-001-expected.checksum:
        * editing/pasteboard/emacs-cntl-y-001-expected.png:
        * editing/pasteboard/emacs-cntl-y-001-expected.txt:
        * editing/pasteboard/emacs-cntl-y-001.html:


        Reviewed by Darin.

        <rdar://problem/4743256> ctrl-y key binding (yank) should do nothing when kill ring is empty

        Test updated:
        * editing/pasteboard/emacs-cntl-y-001.html:

        * WebView/WebHTMLView.mm:
        (-[WebHTMLView yank:]):
        (-[WebHTMLView yankAndSelect:]):
        Do nothing if the killring is empty.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20138 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Not reviewed - gdk build fix.
kjk [Tue, 13 Mar 2007 04:05:41 +0000 (04:05 +0000)]
    Not reviewed - gdk build fix.

        * platform/gdk/TemporaryLinkStubs.cpp:
        (SearchPopupMenu::enabled): add missing stub.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20137 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoWebCore:
ddkilzer [Tue, 13 Mar 2007 03:52:09 +0000 (03:52 +0000)]

        Reviewed by Darin.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13045
          REGRESSION: Blackboard CourseWare Error with Nightlies after Mar 8

        * html/HTMLFormElement.cpp:
        (WebCore::getUniqueBoundaryString): Speculative fix: remove dashes from
        the boundary prefix that appear after non-dash characters.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20136 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
ddkilzer [Tue, 13 Mar 2007 03:48:49 +0000 (03:48 +0000)]

        Reviewed by Anders.

        - test for http://bugs.webkit.org/show_bug.cgi?id=13015
          REGRESSION (r17233-r17241): Repro crash when leaving a page whose unload handler submits a form

        * fast/loader/onunload-form-submit-crash-2-expected.txt: Added.
        * fast/loader/onunload-form-submit-crash-2.html: Added.


        Reviewed by Anders.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13015
          REGRESSION (r17233-r17241): Repro crash when leaving a page whose unload handler submits a form

        Test: fast/loader/onunload-form-submit-crash-2.html

        * loader/DocumentLoader.cpp:
        (WebCore::DocumentLoader::finishedLoading): Added null check.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20135 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoRemove conflict marker.
andersca [Tue, 13 Mar 2007 02:13:50 +0000 (02:13 +0000)]
Remove conflict marker.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20134 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Add expected results.
andersca [Tue, 13 Mar 2007 02:11:39 +0000 (02:11 +0000)]
    Add expected results.

        * fast/frames/frame-navigation-expected.txt: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20133 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
andersca [Tue, 13 Mar 2007 01:09:53 +0000 (01:09 +0000)]

        Reviewed by Darin.

        ASSERTION: Navigating 'back' in frameset: !_private->previousItem (6454)

        * fast/frames/frame-navigation.html: Added.
        * fast/frames/resources/frame-navigation-child.html: Added.


        Reviewed by Darin.

        ASSERTION: Navigating 'back' in frameset: !_private->previousItem (6454)

        Change back the behavior of checkLoadComplete to traverse the entire frame tree instead of
        just the parent frames of the current frame. This is needed in order to reset the previous history item
        for all frames when doing a frame navigation. (This was changed in revision 11819)

        * loader/FrameLoader.cpp:
        * loader/FrameLoader.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20132 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by darin
justing [Tue, 13 Mar 2007 00:53:38 +0000 (00:53 +0000)]
    Reviewed by darin

        REGRESSION: Gmail Editor: Dragging text into Reply (textarea) field results in a crash at WebCore::InsertNodeBeforeCommand::doApply()

        * editing/SelectionController.cpp:
        (WebCore::removingNodeRemovesPosition): Added.
        (WebCore::SelectionController::nodeWillBeRemoved): Clear the
        selection if it's inside a shadow tree.
        * page/DragController.cpp:
        (WebCore::setSelectionToDragCaret): Return false to signal to
        clients that a drop shouldn't be performed if the second attempt
        to set a selection ends up in non-editable content.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20131 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
darin [Tue, 13 Mar 2007 00:49:02 +0000 (00:49 +0000)]

        Reviewed by Tim Hatcher.

        - added a test that checks the classes of wrappers created by both
          the JavaScript binding and the Objective-C binding

        * fast/dom/wrapper-classes-expected.txt: Added.
        * fast/dom/wrapper-classes.html: Added.
        * fast/js/resources/js-test-pre.js: Improve the format of the failure message when
        both the real results and expected results are strings.


        Reviewed by Tim Hatcher.

        - fixed JavaScript wrapper classes to be correct for a variety of cases
          that a new test uncovered: was broken for at least 5 classes

        - fixed Objective-C wrapper classes to be correct for a variety of cases
          that a test case uncovered: was broken for ast least 50 classes

        - added missing DOM API for creating OverflowEvent and WheelEvent instances

        Test: fast/dom/wrapper-classes.html

        * DerivedSources.make: Added missing bindings: HTMLCanvasElement for ObjC,
        CDATASection, Comment, and EntityReference for JavaScript.
        * WebCore.xcodeproj/project.pbxproj: Added those new generated files.

        * bindings/js/JSHTMLElementWrapperFactory.cpp:
        (WebCore::createJSHTMLWrapper): Corrected the wrapper classes for <keygen>,
        which needs an HTMLSelectElement wrapper, and <xmp>, which needs an
        HTMLPreElement wrapper.

        * bindings/objc/DOMInternal.h: Updated for new naming scheme.
        Also moved createDOMWrapper from the KJS namespace to the WebCore namespace.
        * bindings/objc/DOMUtility.mm:
        (KJS::createDOMWrapper): Broke the core function into a separate one, and
        left it in the KJS namespace because Objective-C++ rules make it impossible
        for it to work in the WebCore namespace. Used a macro-based implementation
        to cut down on repeated code, and added missing cases for Counter,
        HTMLOptionsCollection, Range, XPathExpression, XPathResult, Event, RGBColor,
        Rect, Window, DOMImplementation, NodeIterator, TreeWalker, and HTMLCollection.
        (WebCore::createDOMWrapper): The other half of the function.

        * bindings/objc/DOM.mm:
        (WebCore::createElementClassMap): Corrected the wrapper classes for
        <canvas>, which needs a DOMHTMLCanvasElement wrapper, <del>, which needs
        a DOMHTMLModElement wrapper, <embed>, which needs a DOMHTMLEmbedElement
        wrapper, <ins>, which needs a DOMHTMLModElement wrapper, <th>, which needs
        a DOMHTMLTableCellElement wrapper, and <xmp>, which needs an
        DOMHTMLPreElement wrapper.
        (+[DOMNode _wrapNode:]): Updated for new naming scheme.
        (+[DOMNode _wrapEventTarget:]): Ditto.
        (+[DOMNodeFilter _wrapNodeFilter:]): Ditto.
        (ObjCNodeFilterCondition::acceptNode): Ditto.
        (-[DOMDocument createNodeIterator:whatToShow:filter:expandEntityReferences:]): Ditto.
        (-[DOMDocument createTreeWalker:whatToShow:filter:expandEntityReferences:]): Ditto.
        (WebCore::ObjCEventListener::handleEvent): Ditto.

        * dom/Document.cpp: (WebCore::Document::createEvent):
        Added cases for OverflowEvent and WheelEvent.

        * dom/OverflowEvent.h: Added empty constructor and initOverflowEvent.
        * dom/OverflowEvent.cpp:
        (WebCore::OverflowEvent::OverflowEvent): Added.
        (WebCore::OverflowEvent::initOverflowEvent): Added.
        * dom/OverflowEvent.idl: Added initOverflowEvent.

        * bindings/objc/PublicDOMInterfaces.h: Added initOverflowEvent.

        * bindings/objc/DOMCSS.mm:
        (+[DOMStyleSheet _wrapStyleSheet:]): Updated for new naming scheme.
        (+[DOMCSSRule _wrapCSSRule:]): Corrected wrapper for CSSUnknownRule.
        (+[DOMCSSValue _wrapCSSValue:]): Updated for new naming scheme.

        * bindings/js/kjs_css.h:
        (KJS::DOMRGBColor::impl): Added. Used when making an ObjC wrapper.
        (KJS::DOMRect::impl): Ditto.

        * bindings/js/kjs_dom.cpp: (KJS::toJS): Corrected the wrapper classes for
        CDATASection, Comment, and EntityReference.

        * bindings/js/kjs_html.cpp: Corrected the class name for HTMLElement
        (was "DOMHTMLElement") and HTMLCollection (was "Collection").

        * bindings/objc/DOMImplementationFront.h:
        * bindings/objc/DOMImplementationFront.cpp:
        (WebCore::implementationFront): Added new overload that returns a front given
        a JavaScript wrapper. Needed by the code that makes the Objective-C wrapper.

        * bindings/objc/WebScriptObject.mm:
        (+[WebScriptObject _convertValueToObjcValue:originRootObject:rootObject:]):
        Updated since createDOMWrapper is now in the WebCore namespace.

        * bindings/scripts/CodeGeneratorObjC.pm: Update the naming scheme for the
        wrapper creation functions to use _wrapElement: rather then _elementWith:
        style. Removed now-unneeded special cases for things that needed to stay
        upper-case, since we don't have to do the whole lcfirst thing.

        * html/HTMLCanvasElement.idl: Added #if so we can successfully generate the
        Objective-C wrapper for this class, even though we can't yet handle the
        DOMObject return type.

        * html/HTMLOptionsCollection.idl: Added GenerateNativeConverter so we get
        an appropriate toJS function.

        * bindings/objc/DOMEvents.mm:
        (+[DOMEvent _wrapEvent:]):
        * bindings/objc/DOMHTML.mm:
        (-[DOMHTMLDocument createDocumentFragmentWithMarkupString:baseURL:]):
        (-[DOMHTMLDocument createDocumentFragmentWithText:]):
        * bindings/objc/DOMObject.mm:
        (-[DOMObject sheet]):
        * bindings/objc/DOMRGBColor.mm:
        (-[DOMRGBColor red]):
        (-[DOMRGBColor green]):
        (-[DOMRGBColor blue]):
        (-[DOMRGBColor alpha]):
        (+[DOMRGBColor _wrapRGBColor:]):
        * bindings/objc/DOMSVGPathSegInternal.mm:
        (+[DOMSVGPathSeg _wrapSVGPathSeg:]):
        * bindings/objc/DOMXPath.mm:
        (+[DOMNativeXPathNSResolver _wrapXPathNSResolver:]):
        * page/mac/WebCoreFrameBridge.mm:
        (-[WebCoreFrameBridge nodesFromList:]):
        (-[WebCoreFrameBridge elementWithName:inForm:]):
        (-[WebCoreFrameBridge formForElement:]):
        (-[WebCoreFrameBridge currentForm]):
        (-[WebCoreFrameBridge controlsInForm:]):
        (-[WebCoreFrameBridge rangeByAlteringCurrentSelection:SelectionController::direction:SelectionController::granularity:]):
        (-[WebCoreFrameBridge convertNSRangeToDOMRange:]):
        (-[WebCoreFrameBridge markDOMRange]):
        (-[WebCoreFrameBridge markedTextDOMRange]):
        (-[WebCoreFrameBridge smartDeleteRangeForProposedRange:]):
        (-[WebCoreFrameBridge documentFragmentWithMarkupString:baseURLString:]):
        (-[WebCoreFrameBridge documentFragmentWithText:inContext:]):
        (-[WebCoreFrameBridge documentFragmentWithNodesAsParagraphs:]):
        (-[WebCoreFrameBridge replaceSelectionWithNode:selectReplacement:smartReplace:matchStyle:]):
        (-[WebCoreFrameBridge replaceSelectionWithText:selectReplacement:smartReplace:]):
        (-[WebCoreFrameBridge increaseSelectionListLevel]):
        (-[WebCoreFrameBridge increaseSelectionListLevelOrdered]):
        (-[WebCoreFrameBridge increaseSelectionListLevelUnordered]):
        (-[WebCoreFrameBridge dragCaretDOMRange]):
        (-[WebCoreFrameBridge editableDOMRangeForPoint:]):
        (-[WebCoreFrameBridge characterRangeAtPoint:]):
        (-[WebCoreFrameBridge typingStyle]):
        (-[WebCoreFrameBridge rangeOfCharactersAroundCaret]):
        * platform/mac/ClipboardMac.mm:
        * platform/mac/PasteboardMac.mm:
        Updated for new naming scheme.


        Reviewed by Tim Hatcher.

        - update for the new naming scheme for the Objective-C wrapper-creation
          functions: _wrapElement: instead of _elementWith:, etc.

        * WebCoreSupport/WebEditorClient.mm:
        * WebCoreSupport/WebFrameLoaderClient.mm:
        * WebView/WebFrame.mm: (kit):
        Use the _wrapElement-style functions.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20130 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Darin.
thatcher [Tue, 13 Mar 2007 00:34:17 +0000 (00:34 +0000)]
    Reviewed by Darin.

        <rdar://problem/4990691> REGRESSION: Selecting text in Adium's Messages field causes horizontal shift

        Corrected the recursive point conversion and scrolling done in scrollPointRecursively to use the
        document view of each NSClipView.

        * platform/mac/ScrollViewMac.mm:
        (WebCore::ScrollView::scrollPointRecursively): Call convertPoint:fromView: on the document views,
        not the clip views as we encounter them. Then call constrainScrollPoint: on the converted point to
        constrain to the document view bounds. And finally call scrollPoint: on the document view, not the clip view.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20129 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Tim Hatcher.
darin [Tue, 13 Mar 2007 00:07:04 +0000 (00:07 +0000)]
    Reviewed by Tim Hatcher.

        - add a DumpRenderTree feature where you can get the Objective-C
          class name of a JavaScript object

        * DumpRenderTree/DumpRenderTree.m:
        (+[LayoutTestController isSelectorExcludedFromWebScript:]): Add the
        objCClassNameOf: selector.
        (+[LayoutTestController webScriptNameForSelector:]): Add the
        objCClassNameOf: selector, with the name "objCClassName".
        (-[LayoutTestController objCClassNameOf:]): Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20128 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
antti [Mon, 12 Mar 2007 23:57:56 +0000 (23:57 +0000)]

        Reviewed by Alexey

        Test that this works correctly with composed characters
        REGRESSION: Selecting text in 6.6MB txt file is sluggish as of the Feb 19th nightly

        * fast/text/large-text-composed-char-expected.checksum: Added.
        * fast/text/large-text-composed-char-expected.png: Added.
        * fast/text/large-text-composed-char-expected.txt: Added.
        * fast/text/large-text-composed-char.html: Added.


        Reviewed by Alexey.

        Fix http://bugs.webkit.org/show_bug.cgi?id=12833
        REGRESSION: Selecting text in 6.6MB txt file is sluggish as of the Feb 19th nightly

        Divide large text blocks (>64kB) over multiple text nodes. This limits linebox searches to
        a manageable subset.

        * dom/Text.cpp:
        * dom/Text.h:
        * html/HTMLParser.cpp:
        * loader/TextDocument.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20127 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Fix a regression in printing. Printer fonts need to be part of the
hyatt [Mon, 12 Mar 2007 23:48:55 +0000 (23:48 +0000)]
    Fix a regression in printing.  Printer fonts need to be part of the
        font cache key, since printer fonts have different glyph widths.
        (Integer antialiasing is used for screen fonts but not for printer fonts.)

        Reviewed by aroben

        * platform/FontCache.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20126 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Maciej.
andersca [Mon, 12 Mar 2007 23:07:51 +0000 (23:07 +0000)]
    Reviewed by Maciej.

        REGRESSION: Repro Crash in FrameLoader::frame loading about:blank in PLT

        Always get the web view from the current web frame, since the document loader's frame can have been zeroed out
        (for example when detaching the document loader).

        * WebCoreSupport/WebFrameLoaderClient.mm:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20125 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago add missing symbol to fix the Qt build again.
lars [Mon, 12 Mar 2007 21:32:03 +0000 (21:32 +0000)]
    add missing symbol to fix the Qt build again.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20124 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Oliver.
adele [Mon, 12 Mar 2007 20:43:14 +0000 (20:43 +0000)]
    Reviewed by Oliver.

        Adding the ability to enable or disable a SearchPopupMenu.

        * platform/SearchPopupMenu.h:
        * platform/mac/SearchPopupMenuMac.mm: (WebCore::SearchPopupMenu::enabled):
        * rendering/RenderTextControl.cpp: (WebCore::RenderTextControl::showPopup):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20123 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Nikolas Zimmermann.
rwlbuis [Mon, 12 Mar 2007 20:17:49 +0000 (20:17 +0000)]
    Reviewed by Nikolas Zimmermann.

        SVG fails to correctly handle all link style selectors
        <text> elements ignore <a> children

        Allow <a> inside svg text and handle xlink:show.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20122 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by .
kmccullo [Mon, 12 Mar 2007 19:05:31 +0000 (19:05 +0000)]
    Reviewed by .

        - Actually update tests for new DST rules.

        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/Date/
        * tests/mozilla/ecma/jsref.js:
        * tests/mozilla/ecma_2/jsref.js:
        * tests/mozilla/ecma_3/Date/shell.js:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20121 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Oliver.
adele [Mon, 12 Mar 2007 18:49:33 +0000 (18:49 +0000)]
    Reviewed by Oliver.

        Add a missing parameter to the constructor.

        * platform/PlatformKeyboardEvent.h:
        * platform/win/KeyEventWin.cpp: (WebCore::PlatformKeyboardEvent::PlatformKeyboardEvent):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20120 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by .
kmccullo [Mon, 12 Mar 2007 17:44:39 +0000 (17:44 +0000)]
    Reviewed by .

        - Update tests for new DST rules.

        * tests/mozilla/ecma/shell.js:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20118 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoJavaScriptCore:
ggaren [Mon, 12 Mar 2007 15:09:37 +0000 (15:09 +0000)]

        Reviewed by Oliver Hunt.

        Fixed <rdar://problem/4681051> Installer crashes in KJS::Collector::
        markOtherThreadConservatively(KJS::Collector::Thread*) trying to install
        iLife 06 using Rosetta on an Intel Machine

        The problem was that our thread-specific data destructor would modify the
        list of active JavaScript threads without holding the JSLock, corrupting
        the list. Corruption was especially likely if one JavaScript thread exited
        while another was starting up.

        * JavaScriptCore.exp:
        * kjs/JSLock.cpp: Don't conflate locking the JSLock with registering a
        thread, since the thread-specific data destructor needs to lock
        without registering a thread. Instead, treat thread registration as a
        part of the convenience of the JSLock object, and whittle down JSLock::lock()
        to just the bits that actually do the locking.
        * kjs/JSLock.h: Updated comments to mention the new behavior above, and
        other recent changes.
        * kjs/collector.cpp:
        (KJS::destroyRegisteredThread): Lock here.
        (KJS::Collector::registerThread): To match, assert that we're locked here.


        Reviewed by Oliver Hunt.

        Updated in light of fix for <rdar://problem/4681051> Installer crashes
        in KJS::Collector::markOtherThreadConservatively(KJS::Collector::Thread*)
        trying to install iLife 06 using Rosetta on an Intel Machine

        * JavaScriptGlue.cpp:
        (JSLockInterpreter): Ensure backwards compatibility by calling
        registerThread() when explicitly taking the JSLock. (This doesn't happen
        automatically anymore.) I doubt this actally matters, but in JavaScriptGlue
        territory, that kind of thinking will get you killed.


        Reviewed by Oliver Hunt.

        Beefed up --threaded mode in light of <rdar://problem/4681051> Installer
        crashes in KJS::Collector::markOtherThreadConservatively(KJS::Collector::Thread*)
        trying to install iLife 06 using Rosetta on an Intel Machine

        --threaded mode now runs a bunch of different JavaScript threads, randomly
        killing and respawning them. This was sufficient for reproducing the
        bug on my MacBook Pro.

        * DumpRenderTree/DumpRenderTree.m:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20115 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Fix the Qt build. I still get crashes though :/
lars [Mon, 12 Mar 2007 09:18:26 +0000 (09:18 +0000)]
    Fix the Qt build. I still get crashes though :/

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20112 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Not reviewed - gdk build fixes.
kjk [Mon, 12 Mar 2007 05:49:55 +0000 (05:49 +0000)]
    Not reviewed - gdk build fixes.

        * platform/gdk/EditorClientGdk.cpp:
        (WebCore::EditorClientGdk::respondToChangedSelection): add stub
        * platform/gdk/EditorClientGdk.h:
        * platform/gdk/FrameGdk.cpp:
        * platform/gdk/TemporaryLinkStubs.cpp:
        (Editor::markMisspellings): add stub

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20111 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Adele.
ap [Mon, 12 Mar 2007 05:39:34 +0000 (05:39 +0000)]
    Reviewed by Adele.

        W3C XPath test Text_Nodes.svg fails

        * xml/XPathStep.cpp:
        (WebCore::XPath::Step::nodeTestMatches): Revert the fix, as the behavior doesn't appear
        all that desirable as it did at first glance.

        * dom/svg/level3/xpath/Text_Nodes-expected.txt: This fails again.

        * fast/xpath/4XPath/Borrowed/kd_20010423-expected.txt:
        * fast/xpath/4XPath/Borrowed/kd_20010423.html:
        Brought the test back to its original form (in which Firefox passes).

        * fast/xpath/text-nodes-expected.txt: Added.
        * fast/xpath/text-nodes.html: Added.
        A new test for text nodes, verifying some more arcane cases.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20110 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago2007-03-11 Oliver Hunt <oliver@apple.com>
oliver [Mon, 12 Mar 2007 05:34:38 +0000 (05:34 +0000)]
2007-03-11  Oliver Hunt  <oliver@apple.com>

        Reviewed by hyatt.

        Fix for <rdar://problem/5055690> ASSERTION failure on drop into
        editable element with content changed on drop

        After setting the selection for a drop into an editable region
        we make sure the we succeeded.  If we didn't we assume a focus handler
        or similar altered the element contents and try again, if the second
        attempt fails we bail out.

        * page/DragController.cpp:

        <rdar://problem/5055690> ASSERTION failure on drop into editable element with content changed on drop

        * fast/events/content-changed-during-drop-expected.txt: Added.
        * fast/events/content-changed-during-drop.html: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20109 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago2007-03-11 Oliver Hunt <oliver@apple.com>
oliver [Mon, 12 Mar 2007 04:41:34 +0000 (04:41 +0000)]
2007-03-11  Oliver Hunt  <oliver@apple.com>

        Reviewed by Adele.
        Moving Frame{Mac}::respondToChangedSelection to Frame.cpp
        Added new EditorClient method to handle old bridge function

        * bridge/EditorClient.h:
           Added respondToChangedSelection to replace old bridge function
        * editing/Editor.cpp:
           Add client call to replace old bridge call from Frame::respondToChangedSelection
        * editing/SelectionController.cpp:
           No longer directly call Editor as Frame::respondToChangedSelection
           makes the call
        * page/Frame.cpp:
           Moved from FrameMac, replaced bridge call with call to Editor
        * page/mac/FrameMac.mm:
           Moved respondToChangedSelection to Frame.cpp
        * page/mac/WebCoreFrameBridge.h:
           Removed respondToChangedSelection from bridge
        * page/qt/FrameQt.cpp:
           Remove stub method for respondToChangedSelection
        Moved respondToChangedSelection from FrameBridge to EditorClient

        * WebCoreSupport/WebEditorClient.h:
        * WebCoreSupport/WebEditorClient.mm:
        * WebCoreSupport/WebFrameBridge.mm:
          Removed respondToChangedSelection from bridge
        Stub for EditorClientQt::respondToChangedSelection

        * WebCoreSupport/EditorClientQt.cpp:
        * WebCoreSupport/EditorClientQt.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20108 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Adele.
darin [Mon, 12 Mar 2007 03:24:45 +0000 (03:24 +0000)]
    Reviewed by Adele.

        - fix http://bugs.webkit.org/show_bug.cgi?id=12964
          <rdar://problem/5045717> REGRESSION: crash in -[WebBaseNetscapePluginStream _deliverData]
          at simpsonsmovie.com (12964)

        * Plugins/WebBaseNetscapePluginStream.m:
        (-[WebBaseNetscapePluginStream startStreamResponseURL:expectedContentLength:lastModifiedDate:MIMEType:]):
        At every place we call the plug-in, since it could destroy the stream, get pluginView into a local
        variable; it will be set to nil if the stream is destroyed.
        (-[WebBaseNetscapePluginStream _destroyStream]): Added calls to retain/release to handle the case where
        one of the calls to the plug-in destroys the stream. Added a call to cancelPreviousPerformRequestsWithTarget
        in case _deliverData has been scheduled but not yet delivered. Also get pluginView into a local variable
        as mentioned above, and check at strategic points and exit if the stream was already destroyed to avoid
        multiple calls to NPP_DestroyStream or NPP_URLNotify.
        (-[WebBaseNetscapePluginStream _deliverData]): Ditto.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20107 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoFixed my name in the 2 ChangeLogs kjk just comitted
beidson [Mon, 12 Mar 2007 01:32:07 +0000 (01:32 +0000)]
Fixed my name in the 2 ChangeLogs kjk just comitted

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20106 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago WebCore:
kjk [Mon, 12 Mar 2007 01:09:17 +0000 (01:09 +0000)]

        Reviewed by Brady Edison.

        Linux/gdk fix.

        * loader/gdk/FrameLoaderClientGdk.cpp:
        (WebCore::FrameLoaderClientGdk::shouldGoToHistoryItem): return true so
        that FrameLoader:goBackOrForwards() works.


        Reviewed by Brady Edison.

        Add forward/backward/quit menus for easier testing.

        * GdkLauncher/main.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20105 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoJavaScriptCore:
ggaren [Sun, 11 Mar 2007 23:57:11 +0000 (23:57 +0000)]

        Reviewed by Darin Adler.

        Fixed <rdar://problem/4587763> PAC file: lock inversion between QT and
        JSCore causes a hang @ www.panoramas.dk

        With a PAC file, run-webkit-tests --threaded passes, the reported site
        works, and all the Quicktime/JavaScript and Flash/JavaScript examples
        I found through Google work, too.

        Any time JavaScript causes arbitrary non-JavaScript code to execute, it
        risks deadlock, because that code may block, trying to acquire a lock
        owned by a thread that is waiting to execute JavaScript. In this case,
        the thread was a networking thread that was waiting to interpret a PAC file.

        Because non-JavaScript code may execute in response to, well, anything,
        a perfect solution to this problem is impossible. I've implemented an
        optimistic solution, instead: JavaScript will drop its lock whenever it
        makes a direct call to non-JavaScript code through a bridging/plug-in API,
        but will blissfully ignore the indirect ways it may cause non-JavaScript
        code to run (resizing a window, for example).

        Unfortunately, this solution introduces significant locking overhead in
        the bridging APIs. I don't see a way around that.

        This patch includes some distinct bug fixes I saw along the way:

        * bindings/objc/objc_instance.mm: Fixed a bug where a nested begin() call
        would leak its autorelease pool, because it would NULL out _pool without
        draining it.

        * bindings/runtime_object.cpp:
        (RuntimeObjectImp::methodGetter): Don't copy an Identifier to ASCII only
        to turn around and make an Identifier from the ASCII. In an earlier
        version of this patch, the copy caused an assertion failure. Now it's
        just unnecessary work.
        (RuntimeObjectImp::getOwnPropertySlot): ditto

        * bindings/objc/objc_instance.h: Removed overrides of setVAlueOfField and
        getValueOfField, because they did exactly what the base class versions did.
        Removed overrides of Noncopyable declarations for the same reason.

        * bindings/runtime.h: Inherit from Noncopyable instead of rolling our own.
        * bindings/c/c_instance.h: ditto

        And the actual patch:

        * API/JSCallbackConstructor.cpp: Drop all locks when calling out to C.
        * API/JSCallbackFunction.cpp: ditto
        * API/JSCallbackObject.cpp: ditto

        * bindings/c/c_instance.cpp: Drop all locks when calling out to C.
        * bindings/c/c_runtime.cpp: Drop all locks when calling out to C.
        * bindings/jni/jni_objc.mm:
        (KJS::Bindings::dispatchJNICall): Drop all locks when calling out to Java.

        * bindings/objc/objc_instance.mm: The changes here are to accomodate the
        fact that C++ unwinding of DropAllLocks goes crazy when you put it inside
        a @try block. I moved all JavaScript stuff outside of the @try blocks, and
        then prefixed the whole blocks with DropAllLocks objects. This required some
        supporting changes in other functions, which now acquire the JSLock for
        themselves, intead of relying on their callers to do so.
        * bindings/objc/objc_runtime.mm: Same as above, except I didn't want to
        change throwError to acquire the JSLock for itself.
        * bindings/objc/objc_utility.mm: Supporting changes mentioned above.

        * kjs/JSLock.cpp:
        (1) Fixed DropAllLocks to behave as advertised, and drop the JSLock only
        if the current thread actually acquired it in the first place. This is
        important because WebKit needs to ensure that the JSLock has been
        dropped before it makes a plug-in call, even though it doesn't know if
        the current thread actually acquired the JSLock. (We don't want WebKit
        to accidentally drop a lock belonging to *another thread*.)
        (2) Used the new per-thread code written for (1) to make recursive calls
        to JSLock very cheap. JSLock now knows to call pthread_mutext_lock/
        pthread_mutext_unlock only at nesting level 0.
        * kjs/JSLock.h: Don't duplicate Noncopyable.

        * wtf/Assertions.h: Blind attempt at helping the Windows build.


        Reviewed by Darin Adler.

        Fixed <rdar://problem/4587763> PAC file: lock inversion between QT and
        JSCore causes a hang @ www.panoramas.dk

        See JavaScriptCore ChangeLog for details.

        * bindings/objc/WebScriptObject.mm:
        (_didExecute): Added helpful ASSERT.
        (+[WebScriptObject throwException:]): Added missing JSLock.


        Reviewed by Darin Adler.

        Fixed <rdar://problem/4587763> PAC file: lock inversion between QT and
        JSCore causes a hang @ www.panoramas.dk

        See JavaScriptCore ChangeLog for details.

        Drop the JSLock before making calls through the plug-in API from functions
        that may have been called by JavaScript.

        * Plugins/WebBaseNetscapePluginView.mm:
        (-[WebBaseNetscapePluginView sendEvent:]):
        (-[WebBaseNetscapePluginView setWindowIfNecessary]):
        (-[WebBaseNetscapePluginView initWithFrame:pluginPackage:URL:baseURL:MIMEType:attributeKeys:attributeValues:loadManually:DOMElement:]):
        (-[WebBaseNetscapePluginView createPluginScriptableObject]):
        (-[WebBaseNetscapePluginView evaluateJavaScriptPluginRequest:]):
        (-[WebBaseNetscapePluginView webFrame:didFinishLoadWithReason:]):
        (-[WebBaseNetscapePluginView loadPluginRequest:]):
        (-[WebBaseNetscapePluginView _printedPluginBitmap]):
        * Plugins/WebPluginController.mm:
        (+[WebPluginController plugInViewWithArguments:fromPluginPackage:]):
        (-[WebPluginController startAllPlugins]):
        (-[WebPluginController stopAllPlugins]):
        (-[WebPluginController addPlugin:]):
        (-[WebPluginController destroyPlugin:]):
        (-[WebPluginController destroyAllPlugins]):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20104 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Hyatt.
antti [Sun, 11 Mar 2007 11:14:34 +0000 (11:14 +0000)]
    Reviewed by Hyatt.

        Optimize linebox memory consumption:
        - move all bitfields to baseclass compacting them
        - make InlineTextBox::m_truncation unsigned short and make it relative to m_start
        - remove extremely rarely used EllipsisBox pointer from RootInlineBox and instead
          use a global hashmap to store it if needed
        - use minimum required number of bits to store BidiStatus enum variables in RootInlineBox
        - move overflow variables in RootInlineBox to a separate struct that is instantiated
          only if any of the variables is set to a value that can't trivially be derived from
          box x, y, width and height

        As a result line box objects shrink:
            InlineBox: 44 -> 44 bytes
            InlineTextBox: 68 -> 60 bytes
            InlineFlowBox: 68 -> 64 bytes
            RootInlineBox: 128 -> 88 bytes

        The optimizations possiblity was noticed when debugging http://bugs.webkit.org/show_bug.cgi?id=12833
        Bug 12833: REGRESSION: Selecting text in 6.6MB txt file is sluggish as of the Feb 19th nightly

        On that page the patch saves 11.5MB or some 21% of linebox memory consumption. It also
        actually improves selection performance somewhat by improving memory locality.

        * rendering/InlineBox.h:
        * rendering/InlineFlowBox.h:
        * rendering/InlineTextBox.cpp:
        * rendering/InlineTextBox.h:
        * rendering/RootInlineBox.cpp:
        (WebCore::RootInlineBox::Overflow::operator delete):
        * rendering/RootInlineBox.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20103 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Darin.
ap [Sun, 11 Mar 2007 08:21:13 +0000 (08:21 +0000)]
    Reviewed by Darin.

        A partial fix for http://bugs.webkit.org/show_bug.cgi?id=13021
        XPath can be very slow

        * xml/XPathExpression.cpp:
        (WebCore::XPathExpression::evaluate): Cache evaluationContext in a local variable.

        * xml/XPathExpressionNode.cpp:
        * xml/XPathExpressionNode.h:
        * xml/XPathFunctions.cpp:
        * xml/XPathFunctions.h:
        Made one-liners critical for performance inline.

        * xml/XPathGrammar.y: Fully parse NodeTests, so that strings are no longer passed for what is
        essentially an enum. Use LocationPath accessors to add steps, instead of directly manipulating
        internal data members.

        * xml/XPathParser.cpp:
        * xml/XPathParser.h:
        Added support methods for changes in XPathGrammar.y.

        * xml/XPathPath.cpp:
        (WebCore::XPath::Filter::evaluate): Cache evaluationContext in a local variable. Use swap() to avoid
        performing vector assignments.
        (WebCore::XPath::LocationPath::evaluate): Use swap() to avoid performing vector assignments.
        (WebCore::XPath::LocationPath::optimizeStepPair): This new method is called during LocationPath construction,
        to simplify the path as it's being built. Currently, the only optimized case is "//*" - it is a basis for
        important operations that cannot be efficiently written in XPath 1.0, but can be optimized with a little bit
        of XPath 2.0.
        (WebCore::XPath::LocationPath::appendStep): A new accessor that modifies m_steps and calls optimizeStepPair().
        (WebCore::XPath::LocationPath::insertFirstStep): Ditto.
        * xml/XPathPath.h:
        (WebCore::XPath::LocationPath::setAbsolute): A new accessor.

        * xml/XPathStep.h:
        Step::NodeTest is a new sub-class that represents a fully parsed NodeTest.
        New accessors that let optimizeStepPair() manipulate Step data.

        * xml/XPathStep.cpp:
        (WebCore::XPath::Step::Step): Use the new NodeTest class.
        (WebCore::XPath::Step::evaluate): Cache evaluationContext in a local variable. Use swap() to avoid
        performing unneeded vector assignments.
        (WebCore::XPath::Step::nodesInAxis): Cosmetic changes.
        (WebCore::XPath::Step::nodeTestMatches): Use NodeTest instead of parsing the test from string each time.
        Added a partial implementation of XPath 2.0 element() node test.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20102 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Darin.
ap [Sun, 11 Mar 2007 08:10:38 +0000 (08:10 +0000)]
    Reviewed by Darin.

        FCKeditor: <hr>, <ul> and <ol> have id="undefined"

        This fixes the attached reduction, but not the original issue.

        Test: editing/execCommand/default-parameters.html

        * dom/Document.h:
        * dom/Document.idl:
        Make second and third execCommand() parameters optional.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20101 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Maciej.
adele [Sat, 10 Mar 2007 23:37:15 +0000 (23:37 +0000)]
    Reviewed by Maciej.

        Fix for http://bugs.webkit.org/show_bug.cgi?id=13028
        REGRESSION: textField:doCommandBySelector:inFrame: not being called properly

        * html/HTMLInputElement.cpp: (WebCore::HTMLInputElement::defaultEventHandler):
          Before calling the base class defaultEventHandler, which will call handleKeypress, call doTextFieldCommandFromEvent.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20100 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
ddkilzer [Sat, 10 Mar 2007 22:46:31 +0000 (22:46 +0000)]

        Reviewed by Darin.

        - test for http://bugs.webkit.org/show_bug.cgi?id=13013
          REGRESSION: Selection box does not scroll to where the focus jumps when pressing an alphanumeric key

        * fast/forms/listbox-typeahead-scroll-expected.txt: Added.
        * fast/forms/listbox-typeahead-scroll.html: Added.


        Reviewed by Darin.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13013
          REGRESSION: Selection box does not scroll to where the focus jumps when pressing an alphanumeric key

        Test: fast/forms/listbox-typeahead-scroll.html

        * html/HTMLSelectElement.cpp:
        (WebCore::HTMLSelectElement::setSelectedIndex): Reordered to set the active selection's
        anchor and end before selecting the option, since the active selection is used to
        decide where to scroll when the selection is made.
        (WebCore::HTMLSelectElement::defaultEventHandler): Removed redundant check.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20099 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
ddkilzer [Sat, 10 Mar 2007 22:43:14 +0000 (22:43 +0000)]

        Reviewed by Darin.

        - test for http://bugs.webkit.org/show_bug.cgi?id=12973
          REGRESSION: Reproducible assert while loading this test file if css is already in the cache

        * fast/dom/css-cached-import-rule-expected.checksum: Added.
        * fast/dom/css-cached-import-rule-expected.png: Added.
        * fast/dom/css-cached-import-rule-expected.txt: Added.
        * fast/dom/css-cached-import-rule.html: Added.


        Reviewed by Darin.

        - fix http://bugs.webkit.org/show_bug.cgi?id=12973
          REGRESSION: Reproducible assert while loading this test file if css is already in the cache

        Test: fast/dom/css-cached-import-rule.html

        Replaced some direct calls to document->stylesheetLoaded() with calls to
        the sheet's checkLoaded(). The latter calls back to the element's sheetLoaded() --
        which notifies the document of the load -- and then updates the sheet's
        loadCompleted() flag, ensuring that it stays in sync with whether the stylesheet
        is still considered pending by the document.

        * dom/ProcessingInstruction.cpp:
        * dom/StyleElement.cpp:
        * html/HTMLLinkElement.cpp:
        * ksvg2/svg/SVGStyleElement.cpp:
        * ksvg2/svg/SVGStyleElement.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20098 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoWebCore:
ddkilzer [Sat, 10 Mar 2007 22:38:07 +0000 (22:38 +0000)]

        Reviewed by Darin.

        - fix http://bugs.webkit.org/show_bug.cgi?id=9609
          REGRESSION: Missing image icon needs to be moved back to WebKit

        * WebCore.exp: Export WebCore::Image::loadPlatformResource(const char*) for use in
        [WebHTMLView _startDraggingImage:at:operation:event:sourceIsDHTML:DHTMLWroteData:].


        Reviewed by Darin.

        - fix http://bugs.webkit.org/show_bug.cgi?id=9609
          REGRESSION: Missing image icon needs to be moved back to WebKit

        * WebView/WebHTMLView.mm:
        (-[WebHTMLView _startDraggingImage:at:operation:event:sourceIsDHTML:DHTMLWroteData:]):
        Use WebCore::Image::loadPlatformResource(const char*) to load the missingImage image.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20097 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago2007-03-10 Mark Rowe <mrowe@apple.com>
bdash [Sat, 10 Mar 2007 08:08:36 +0000 (08:08 +0000)]
2007-03-10  Mark Rowe  <mrowe@apple.com>

        Reviewed by John.

        <rdar://problem/5051827> HIWebView handling of kEventControlGetData is broken in 64-bit

        On Leopard the kEventParamControlDataBufferSize event parameter is of type typeByteCount.
        The 32-bit implementation of GetEventParameter will coerce between integer types and
        typeByteCount while the 64-bit version will return a failure. As typeByteCount is new
        in Leopard we must continue using typeSInt32 when building for Tiger.

        * Carbon/HIWebView.m:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20096 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoWebKitTools:
ddkilzer [Sat, 10 Mar 2007 00:41:48 +0000 (00:41 +0000)]

        Reviewed by Darin.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13007
          svn-create-patch doesn't handle UTF files with BOMs as text

        Force diff to treat files that svn-create-patch thinks are text as text.

        * Scripts/svn-create-patch: Added -a switch to diff command.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20095 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Reviewed by Darin.
thatcher [Fri, 9 Mar 2007 22:58:02 +0000 (22:58 +0000)]
    Reviewed by Darin.

        <rdar://problem/4976254> Please get off _NSSoftLinkingGetFrameworkFuncPtr

        Use dlopen and dlsym to access the DCSShowDictionaryServiceWindow function.

        * WebView/WebHTMLView.mm:
        (-[WebHTMLView _lookUpInDictionaryFromMenu:]):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20094 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago * blog/wp-content/themes/webkit/style.css: Removed rule for code since
darin [Fri, 9 Mar 2007 19:45:36 +0000 (19:45 +0000)]
    * blog/wp-content/themes/webkit/style.css: Removed rule for code since
        we want it to match pre, and the default font looks fine -- no need to
        ask for "Courier New" explicitly.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20093 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
darin [Fri, 9 Mar 2007 18:54:29 +0000 (18:54 +0000)]

        Reviewed by Justin.

        - test for http://bugs.webkit.org/show_bug.cgi?id=8928
          <rdar://problem/5045708> REPRODUCIBLE ASSERT: Cannot paste HTML into a
          contenteditable region in an XHTML document (8928)

        * editing/pasteboard/paste-xml-expected.checksum: Added.
        * editing/pasteboard/paste-xml-expected.png: Added.
        * editing/pasteboard/paste-xml-expected.txt: Added.
        * editing/pasteboard/paste-xml.xhtml: Added.


        Reviewed by Justin.

        - fix http://bugs.webkit.org/show_bug.cgi?id=8928
          <rdar://problem/5045708> REPRODUCIBLE ASSERT: Cannot paste HTML into a
          contenteditable region in an XHTML document (8928)

        Test: editing/pasteboard/paste-xml.xhtml

        * editing/markup.cpp: (WebCore::createFragmentFromMarkup): Added a check for 0
        here, since createContextualFragment can return 0 for XML documents that fail
        to parse. In my testing, callers all seem equipped to handle 0.


        Reviewed by Justin.

        - fix http://bugs.webkit.org/show_bug.cgi?id=8928
          <rdar://problem/5045708> REPRODUCIBLE ASSERT: Cannot paste HTML into a
          contenteditable region in an XHTML document (8928)

        * WebView/WebHTMLView.mm:
        (-[WebHTMLView _hasHTMLDocument]): Added.
        (-[WebHTMLView _documentFragmentFromPasteboard:forType:inContext:subresources:]):
        Don't call AppKit's conversion from the DOM to an attributed string if the document
        is not an HTML document, to work around an AppKit limitation (Radar 5052390).

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20092 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago * English.lproj/StringsNotToBeLocalized.txt: Updated for recent changes.
darin [Fri, 9 Mar 2007 16:19:18 +0000 (16:19 +0000)]
    * English.lproj/StringsNotToBeLocalized.txt: Updated for recent changes.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20091 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoLayoutTests:
weinig [Fri, 9 Mar 2007 16:05:37 +0000 (16:05 +0000)]

        Reviewed by Darin.

        - test for http://bugs.webkit.org/show_bug.cgi?id=9929
          REGRESSION: crash on logging in on mijnpostbank.nl

        * http/tests/misc/onload-remove-iframe-crash-2-expected.txt: Added.
        * http/tests/misc/onload-remove-iframe-crash-2.html: Added.
        * http/tests/misc/resources/onload-remove-iframe-crash-2.html: Added.
        * http/tests/misc/resources/onload-remove-iframe.html: Added.


        Reviewed by Darin.

        - fix http://bugs.webkit.org/show_bug.cgi?id=9929
          REGRESSION: crash on logging in on mijnpostbank.nl

        Test: http/tests/misc/onload-remove-iframe-crash-2.html

        The resulted from an iframe's load event handler removing the iframe
        from the document.

        * dom/Document.cpp:
        (WebCore::Document::implicitClose): Bail out early if an event handler
        removed the frame.
        * loader/FrameLoader.cpp:
        (WebCore::FrameLoader::checkCompleted): Protect the frame from deletion
        by event handlers.
        (WebCore::FrameLoader::detachFromParent): Schedule a completion check
        on the parent (in case the child is what has been keeping it from completing).
        * loader/FrameLoader.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20090 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoI forgot to commit my ChangeLog as part of r20028, sorry about that.
rwlbuis [Fri, 9 Mar 2007 12:35:46 +0000 (12:35 +0000)]
I forgot to commit my ChangeLog as part of r20028, sorry about that.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20089 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoRoll out r20078 as it's causing a crash during layout tests.
bdash [Fri, 9 Mar 2007 10:00:15 +0000 (10:00 +0000)]
Roll out r20078 as it's causing a crash during layout tests.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20088 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years agoWebCore:
ddkilzer [Fri, 9 Mar 2007 06:05:00 +0000 (06:05 +0000)]

        Reviewed by Beth.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13019
          REGRESSION (r20074): Forms don't submit on a variety of websites

        No tests added since LayoutTests/fast/forms/document-write.html was timing out
        and causing a layout test failure.

        * html/HTMLFormElement.cpp:
        (WebCore::HTMLFormElement::submit): Removed stray code.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20084 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago Not reviewed - gdk build fix.
kjk [Fri, 9 Mar 2007 04:55:44 +0000 (04:55 +0000)]
    Not reviewed - gdk build fix.

        * platform/graphics/cairo/ImageSourceCairo.cpp:
        (WebCore::ImageSource::frameIsCompleteAtIndex): add empty stub

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20083 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago2007-03-08 Mark Rowe <mrowe@apple.com>
bdash [Fri, 9 Mar 2007 04:23:41 +0000 (04:23 +0000)]
2007-03-08  Mark Rowe  <mrowe@apple.com>

        Reviewed by Tim Hatcher.

        <rdar://problem/5051616> Mouse clicks and movement are ignored in HIWebView on 64-bit

        Mouse events are not being handled correctly as GetControlKind is returning an error
        on 64-bit.  The more modern HIObjectIsOfClass behaves correctly for this use.

        * Carbon/HIWebView.m:
        (WindowHandler): Use HIObjectIsOfClass in place of GetControlKind.
        (HIWebViewEventHandler): Don't leak the NSEvent.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20082 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago2007-03-09 Mark Rowe <mrowe@apple.com>
bdash [Fri, 9 Mar 2007 04:06:10 +0000 (04:06 +0000)]
2007-03-09  Mark Rowe  <mrowe@apple.com>

        Unreviewed.  Use the new SVN URL on the buildbot.

        * BuildSlaveSupport/build.webkit.org-config/webkit/steps.py:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20081 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago2007-03-09 Mark Rowe <mrowe@apple.com>
bdash [Fri, 9 Mar 2007 03:57:17 +0000 (03:57 +0000)]
2007-03-09  Mark Rowe  <mrowe@apple.com>

        Rubber-stamped by Tim Hatcher.

        Sexy blog permalinks!

        * blog/.htaccess: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20080 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago2007-03-08 MorganL <morganl.webkit@yahoo.com>
bdash [Fri, 9 Mar 2007 03:11:46 +0000 (03:11 +0000)]
2007-03-08  MorganL  <morganl.webkit@yahoo.com>

        Reviewed by Darin.

        Bug 13018: allow embedders to override the definition of CRASH.

        * wtf/Assertions.h: make it possible to override CRASH.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20079 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago2007-03-08 Mitz Pettel <mitz@webkit.org>
bdash [Fri, 9 Mar 2007 03:08:35 +0000 (03:08 +0000)]
2007-03-08  Mitz Pettel  <mitz@webkit.org>

        Reviewed by Brady.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13015
          REGRESSION (r17233-r17241): Repro crash when leaving a page whose unload handler submits a form

        Test: fast/loader/onunload-form-submit-crash-2.html

        * loader/DocumentLoader.cpp:
        (WebCore::DocumentLoader::finishedLoading): Added null check.

2007-03-08  Mitz Pettel  <mitz@webkit.org>

        Reviewed by Brady.

        - test for http://bugs.webkit.org/show_bug.cgi?id=13015
          REGRESSION (r17233-r17241): Repro crash when leaving a page whose unload handler submits a form

        * fast/loader/onunload-form-submit-crash-2-expected.txt: Added.
        * fast/loader/onunload-form-submit-crash-2.html: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20078 268f45cc-cd09-0410-ab3c-d52691b4dbfc

13 years ago2007-03-08 Bruce Q Hammond <bruceq@apple.om>
bdash [Fri, 9 Mar 2007 03:02:08 +0000 (03:02 +0000)]
2007-03-08  Bruce Q Hammond  <bruceq@apple.om>

        Reviewed by Darin.

        Fix for http://bugs.webkit.org/show_bug.cgi?id=13009
        Console spews "CGContextGetType: invalid context" non-stop on web site

        * Plugins/WebBaseNetscapePluginView.mm:
        (-[WebBaseNetscapePluginView saveAndSetNewPortStateForUpdate:]):
        (-[WebBaseNetscapePluginView restorePortState:]):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20077 268f45cc-cd09-0410-ab3c-d52691b4dbfc