* coding/RefPtr.html: "smart pointer" -> "smart pointers"

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20482 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    * coding/RefPtr.html: Added. First draft of a document about RefPtr.

        Also tweaked the MIME types of a few files and removed a stray executable bit.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20481 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Amend the statistics reporting for the WebCore cache to include XSL and to report live/decoded sizes.

        * Misc/WebCache.mm:
        (+[WebCache statistics]):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20480 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by Darin.

        - test for http://bugs.webkit.org/show_bug.cgi?id=13156
          REGRESSION (r19621): Pasting breakable content where wrapped line is too long to fit in a textarea fails to draw a horizontal scrollbar

        * fast/text/midword-break-after-breakable-char-expected.checksum: Added.
        * fast/text/midword-break-after-breakable-char-expected.png: Added.
        * fast/text/midword-break-after-breakable-char-expected.txt: Added.
        * fast/text/midword-break-after-breakable-char.html: Added.

WebCore:

        Reviewed by Darin.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13156
          REGRESSION (r19621): Pasting breakable content where wrapped line is too long to fit in a textarea fails to draw a horizontal scrollbar

        Test: fast/text/midword-break-after-breakable-char.html

        Breaking in the middle of the word
        is allowed only if no breaking opportunity between words has occurred yet. The
        first position on the line should not be considered "between words" even if
        it is a breaking opportunity.

        * rendering/bidi.cpp:
        (WebCore::RenderBlock::findNextLineBreak): Changed according to the above. Also
        cleaned up a couple of lines.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20479 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Fix for bugzilla bug 13187, place a hard limit on the amount of decoded data that WebCore will keep
        around even on live pages (pages being displayed in tabs/windows or in the back/forward cache).
        Images will decode to paint and once the decoded data exceeds the cache size, the decoded data will be
        tossed.

        Refine the flushing algorithm to apply the two-pass decode/evict model to each LRU-SP queue individually, rather
        than to all the lists at once.  This allows the cache to evict large encoded resources before flushing small
        or frequently accessed decoded onces.

        Added information on live/decoded data to the cache statistics.

        Reviewed by darin

        * loader/Cache.cpp:
        (WebCore::Cache::Cache):
        (WebCore::Cache::pruneLiveResources):
        (WebCore::Cache::pruneAllResources):
        (WebCore::Cache::setMaximumSize):
        (WebCore::Cache::remove):
        (WebCore::Cache::lruListFor):
        (WebCore::Cache::removeFromLRUList):
        (WebCore::Cache::insertInLRUList):
        (WebCore::Cache::liveLRUListFor):
        (WebCore::Cache::removeFromLiveResourcesList):
        (WebCore::Cache::insertInLiveResourcesList):
        (WebCore::Cache::addToLiveResourcesSize):
        (WebCore::Cache::removeFromLiveResourcesSize):
        (WebCore::Cache::adjustSize):
        (WebCore::Cache::getStatistics):
        * loader/Cache.h:
        (WebCore::Cache::TypeStatistic::TypeStatistic):
        * loader/CachedImage.cpp:
        (WebCore::CachedImage::decodedSizeWillChange):
        (WebCore::CachedImage::decodedSizeChanged):
        * loader/CachedImage.h:
        * loader/CachedResource.cpp:
        (WebCore::CachedResource::CachedResource):
        (WebCore::CachedResource::ref):
        (WebCore::CachedResource::deref):
        (WebCore::CachedResource::setEncodedSize):
        (WebCore::CachedResource::liveResourceAccessed):
        * loader/CachedResource.h:
        (WebCore::CachedResource::liveAccessCount):
        (WebCore::CachedResource::resetLiveAccessCount):
        (WebCore::CachedResource::increaseLiveAccessCount):
        * platform/graphics/BitmapImage.cpp:
        (WebCore::BitmapImage::destroyDecodedData):
        (WebCore::BitmapImage::cacheFrame):
        * platform/graphics/ImageObserver.h:
        * rendering/RenderBox.cpp:
        (WebCore::RenderBox::paintBackgroundExtended):
        * rendering/RenderImage.cpp:
        (WebCore::RenderImage::paint):
        * rendering/RenderListMarker.cpp:
        (WebCore::RenderListMarker::paint):
        * rendering/RenderObject.cpp:
        (WebCore::RenderObject::paintBorderImage):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20478 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by Mitz.

        - test for http://bugs.webkit.org/show_bug.cgi?id=13123
          CSS selectors that use :lang, :not, and namespaces do not serialize correctly

        * fast/css/css-selector-text-expected.txt: Added.
        * fast/css/css-selector-text.html: Added.

WebCore:

        Reviewed by Mitz.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13123
          CSS selectors that use :lang, :not, and namespaces do not serialize correctly

        Test: fast/css/css-selector-text.html

        * css/CSSSelector.cpp: (WebCore::CSSSelector::selectorText): Add the namespace
        prefix for tag or attribute names, the sub-selector for :not, and the argument
        for :lang.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20477 268f45cc-cd09-0410-ab3c-d52691b4dbfc

JavaScriptCore:

        Reviewed by Adam

        Whoops, RetainPtr should be in the WTF namespace

        * wtf/RetainPtr.h:

WebKit:

        Reviewed by Adam

        RetainPtr is no longer in the WebCore namespace

        * History/WebBackForwardList.mm:
        * WebCoreSupport/WebEditorClient.h:
        * WebCoreSupport/WebFrameLoaderClient.h:
        * WebView/WebDocumentLoaderMac.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20476 268f45cc-cd09-0410-ab3c-d52691b4dbfc

JavaScriptCore:

        Reviewed by Adam

        <rdar://problem/5086210> - Move RetainPtr to WTF

        * wtf/RetainPtr.h: Added
        * JavaScriptCore.xcodeproj/project.pbxproj: Add it to the project file
        * JavaScriptCore.vcproj/WTF/WTF.vcproj: Ditto

WebCore:

        Reviewed by Adam

        <rdar://problem/5086210> - Move RetainPtr to WTF

        * ForwardingHeaders/wtf/RetainPtr.h: Added.
        * WebCore.xcodeproj/project.pbxproj: Removed WebCore/RetainPtr.h

        * history/HistoryItem.h: Changed #include to <wtf/RetainPtr.h>
        * history/PageCache.h: Ditto
        * page/mac/WebCoreFrameBridge.mm: Ditto
        * platform/ContextMenu.h: Ditto
        * platform/ContextMenuItem.h: Ditto
        * platform/DragImage.h: Ditto
        * platform/FileChooser.h: Ditto
        * platform/PlatformKeyboardEvent.h: Ditto
        * platform/PopupMenu.h: Ditto
        * platform/SharedBuffer.h: Ditto
        * platform/cf/RetainPtr.h: Removed.
        * platform/graphics/Icon.h: Changed #include
        * platform/mac/ClipboardMac.h: Ditto
        * platform/mac/PasteboardMac.mm: Ditto
        * platform/mac/WidgetMac.mm: Ditto
        * platform/network/AuthenticationChallenge.h: Ditto
        * platform/network/ResourceError.h: Ditto
        * platform/network/ResourceHandle.h: Ditto
        * platform/network/ResourceRequest.h: Ditto
        * platform/network/ResourceResponse.h: Ditto
        * rendering/RenderThemeMac.mm: Ditto

WebKit:

        Reviewed by Adam

        <rdar://problem/5086210> - Move RetainPtr to WTF

        * ForwardingHeaders/wtf/RetainPtr.h: Added.
        * History/WebBackForwardList.mm: Changed #import to <wtf/RetainPtr.h>
        * WebCoreSupport/WebEditorClient.h: Ditto
        * WebCoreSupport/WebFrameLoaderClient.h: Ditto
        * WebView/WebDocumentLoaderMac.h: Ditto

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20475 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Fix for <rdar://problem/5086797>, layoutPending is now reporting incorrect results, which will lead
        to all sorts of layout issues.  This is in bugzilla as 13179.

        Reviewed by hyatt

        - fix http://bugs.webkit.org/show_bug.cgi?id=13179
          REGRESSION (r20410): In debug builds, loading a certain page is suspended until redisplay is forced

        Moved the root->needsLayout() check from layoutPending() to needsLayout()
        to restore the behavior before r20324.

        * page/FrameView.cpp:
        (WebCore::FrameView::layoutPending):
        (WebCore::FrameView::needsLayout):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20474 268f45cc-cd09-0410-ab3c-d52691b4dbfc

WebCore:

        Reviewed by Adele.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13180
              <rdar://problem/5084478>
          Another debug build crash from ASSERTION FAILED: !needsLayout()

        No test possible because updateRendering() is always called after script
        execution.

        * page/FrameView.cpp:
        (WebCore::FrameView::layout): Get the layout root after calling recalcStyle()
        since a style recalc may result in needing to do start layout at the root.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20473 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Not reviewed. Build fix.
Fix FloatRect vs. IntRect confusion.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20472 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Adele

        - fixed <rdar://problem/5084872> Need to add flickr to spoof list in WebKit
        - only do site-specific spoofing if a preference is set

        * WebView/WebPreferenceKeysPrivate.h:
        added WebKitUseSiteSpecificSpoofingPreferenceKey

        * WebView/WebPreferences.m:
        (+[WebPreferences initialize]):
        initialize WebKitUseSiteSpecificSpoofingPreferenceKey to false
        (-[WebPreferences _useSiteSpecificSpoofing]):
        get value of WebKitUseSiteSpecificSpoofingPreferenceKey
        (-[WebPreferences _setUseSiteSpecificSpoofing:]):
        set value of WebKitUseSiteSpecificSpoofingPreferenceKey

        * WebView/WebPreferencesPrivate.h:
        declare _useSiteSpecificSpoofing and _setUseSiteSpecificSpoofing

        * WebView/WebView.mm:
        cache the value of WebKitUseSiteSpecificSpoofingPreferenceKey in a bool in _private
        (-[WebView _preferencesChangedNotification:]):
        update the cached value
        (-[WebView setPreferences:]):
        ditto
        (-[WebView WebCore::_userAgentForURL:WebCore::]):
        Only spoof here if the new site-specific spoofing preference is enabled. If it is, pass
        Safari 2.0.4's user agent string for flickr.com. We can remove this case when 5081617 is addressed.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20471 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fix incorrect radar number.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20470 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Reviewed by Oliver.

Fix all known RenderSVGImage problems.

Fixes: http://bugs.webkit.org/show_bug.cgi?id=12126 (RenderSVGImage seems to suffer from integer overflow)
Fixes: http://bugs.webkit.org/show_bug.cgi?id=12442 (raster images disappearing during script execution (SVG))
Fixes: http://bugs.webkit.org/show_bug.cgi?id=12572 (WebKit does not properly invalidate image region after image load)
Added test: svg/custom/scrolling-embedded-svg-file-image-repaint-problem.html

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20469 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2007-03-24  Mitz Pettel  <mitz@webkit.org>

        Reviewed by Dave Hyatt.

        - http://bugs.webkit.org/show_bug.cgi?id=13177
          Suppress control characters in the ATSUI code path

        * platform/mac/FontMac.mm:
        (WebCore::overrideLayoutOperation):
        (WebCore::ATSULayoutParameters::initialize):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20468 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Fix for <rdar://problem/5086400>, images scrolled offscreen continue to animate.  Safari 2 would
        halt animations when images were no longer visible.

        Reviewed by mjs, olliej

        * platform/graphics/BitmapImage.cpp:
        (WebCore::BitmapImage::advanceAnimation):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20467 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2007-03-24  Mark Rowe  <mrowe@apple.com>

        Reviewed by Dave Hyatt.

        * JavaScriptGlue.xcodeproj/project.pbxproj: Use GCC 4.0 for testjsglue.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20462 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Fix for <rdar://problem/508328>, sluggish text entry in search field.  Make sure
        our size actually changed before we decide to do a full repaint because of background/border
        complexities.

        Reviewed by aroben

        * rendering/RenderObject.cpp:
        (WebCore::RenderObject::repaintAfterLayoutIfNeeded):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20461 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fix Gmail spurious glyphs issue.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20460 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by Adele.

        - pixel test and updated results for http://bugs.webkit.org/show_bug.cgi?id=13153
          REGRESSION: Visual highlighting of pre-populated blank line in textarea is broken

        * editing/selection/doubleclick-crash-expected.checksum:
        * editing/selection/doubleclick-crash-expected.png:
        * fast/text/selection-hard-linebreak-expected.checksum: Added.
        * fast/text/selection-hard-linebreak-expected.png: Added.
        * fast/text/selection-hard-linebreak-expected.txt: Added.
        * fast/text/selection-hard-linebreak.html: Added.

WebCore:

        Reviewed by Adele.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13153
          REGRESSION: Visual highlighting of pre-populated blank line in textarea is broken

        - fix a bug where selecting across a soft line break did not highlight to
          the end of the first line if it contained skipped whitespace

        Test: fast/text/selection-hard-linebreak.html

        * rendering/InlineTextBox.cpp:
        (WebCore::InlineTextBox::selectionState): Changed to treat a selection that ends on the
        end of a hard line break as if it ends after the line break. Fixed the case of a selection
        that starts and ends in the same text object as the box but does not intersect it
        to return SelectionNone instead of selectionBoth.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20459 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by Adele.

        - test for http://bugs.webkit.org/show_bug.cgi?id=13124
          REGRESSION: Reproducible crash in Widget::getView

        * fast/frames/hover-timer-crash-expected.txt: Added.
        * fast/frames/hover-timer-crash.html: Added.

WebCore:

        Reviewed by Adele.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13124
          REGRESSION: Reproducible crash in Widget::getView

        Test: fast/frames/hover-timer-crash.html

        * page/Frame.cpp:
        (WebCore::Frame::setView): Added a call to EventHandler::clear().

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20458 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Forgot to include these nil checks in my previous checkin.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20457 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by adele

        <rdar://problem/5078739>
        Crash in in ReplaceSelectionCommand if the fragment contains only a style span

        * editing/pasteboard/5078739-expected.txt: Added.
        * editing/pasteboard/5078739.html: Added.

WebCore:

        Reviewed by adele

        <rdar://problem/5078739>
        Crash in in ReplaceSelectionCommand if the fragment contains only a style span

        * editing/ReplaceSelectionCommand.cpp:
        (WebCore::ReplaceSelectionCommand::doApply): Check
        for an empty fragment after removing a redundant
        style span instead of before.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20456 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed/landed by Adam.

        Bug 13175: Make apollo mac project files for JavaScriptCore actually
        build something
        http://bugs.webkit.org/show_bug.cgi?id=13175

        - Changing apollo mac project files for JavaScriptCore such that they actually build
        JavaScriptCore source code.

        * JavaScriptCore.apolloproj/ForwardingSources/grammar.cpp: Added.
        * JavaScriptCore.apolloproj/mac/JavaScriptCore/JavaScriptCore.xcconfig:
        * JavaScriptCore.apolloproj/mac/JavaScriptCore/JavaScriptCore.xcodeproj/project.pbxproj:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20455 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    <rdar://problem/5085063>

        Fix non-looping animation on cuteoverload.com. Just always flush
        decoded data for large animated images to avoid our cache getting
        out of sync with the image source.

        Reviewed by ggaren

        * platform/graphics/BitmapImage.cpp:
        (WebCore::BitmapImage::advanceAnimation):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20454 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Darin.

        Fix for <rdar://problem/5061024> REGRESSION: Returning false from onkeypress event on Input is ignored when key is Tab
        http://bugs.webkit.org/show_bug.cgi?id= 13020

        No known way to add a layout test for this.  We would need a way to know exactly how the event was consumed.

        * page/EventHandler.cpp: (WebCore::EventHandler::defaultKeyboardEventHandler):
          If the keypress event has its default behavior prevented, then we should consider the keydown event handled.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20453 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by darin

        Fixed (was missing a newline):
        * editing/execCommand/4917055-expected.txt:
        Tests setting a caet before the image and changing its alignment:
        * editing/execCommand/5080333-1-expected.checksum: Added.
        * editing/execCommand/5080333-1-expected.png: Added.
        * editing/execCommand/5080333-1-expected.txt: Added.
        * editing/execCommand/5080333-1.html: Added.
        Tests selecting the image and changing its alignment:
        * editing/execCommand/5080333-2-expected.checksum: Added.
        * editing/execCommand/5080333-2-expected.png: Added.
        * editing/execCommand/5080333-2-expected.txt: Added.
        * editing/execCommand/5080333-2.html: Added.

WebCore:

        Reviewed by darin

        <rdar://problem/5080333>
        REGRESSION: Selection changes when changing the alignment of an image

        Regression occurred when we started using moveParagraphs
        to move content in applyBlockStyle.  moveParagraphs
        moves by copying, deleting and reinserting content, and
        so must be accompanied by selection preservation code.
        That code uses rangeFromLocationAndLength and rangeLength,
        which use TextIterators, which don't emit anything for images
        and other replaced elements, causing this bug.

        * editing/ApplyStyleCommand.cpp:
        (WebCore::ApplyStyleCommand::applyBlockStyle): Ask rangeLength
        and rangeFromLocationAndLength to request that their
        TextIterators emit spaces for replaced elements.
        Use rangeCompliantEquivalent()s when creating a Range from
        VisiblePositions, since some VisiblePositions have illegal
        deepEquivalent()s.
        * editing/CompositeEditCommand.cpp:
        (WebCore::CompositeEditCommand::moveParagraphs): Ditto.
        * editing/TextIterator.cpp:
        (WebCore::TextIterator::TextIterator):
        (WebCore::TextIterator::handleReplacedElement): Emit
        a space if requested.
        (WebCore::TextIterator::representNodeOffsetZero): Emit
        ranges before m_node, not around m_lastTextNode.  These
        ranges should represent the part of the document associated
        with the emitted character.
        (WebCore::TextIterator::rangeLength): Take in the new bool.
        (WebCore::TextIterator::rangeFromLocationAndLength): Ditto.
        Also, don't loop an extra time after finding the end of the
        range when we're looking for zero length ranges.  This appeared
        to be a workaround for the bugs fixed in representNodeOffsetZero
        in this patch.
        * editing/TextIterator.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20449 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2007-03-24  Mark Rowe  <mrowe@apple.com>

        Rubber-stamped by Darin.

        * Configurations/JavaScriptCore.xcconfig: Remove unnecessary INFOPLIST_PREPROCESS.

2007-03-24  Mark Rowe  <mrowe@apple.com>

        Rubber-stamped by Darin.

        * Configurations/JavaScriptGlue.xcconfig: Remove unnecessary INFOPLIST_PREPROCESS.

2007-03-24  Mark Rowe  <mrowe@apple.com>

        Rubber-stamped by Darin.

        * Configurations/WebCore.xcconfig: Remove unnecessary INFOPLIST_PREPROCESS.

2007-03-24  Mark Rowe  <mrowe@apple.com>

        Rubber-stamped by Darin.

        * Configurations/WebKit.xcconfig: Remove unnecessary INFOPLIST_PREPROCESS.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20445 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    * html/HTMLBodyElement.cpp: Fix comment typo.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20443 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Patch from Adam Treat to make the Qt build work without SVG and XSLT as well as
with.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20442 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Hyatt.

        - <rdar://problem/5074439> handle margins on frame and iframe elements
          without involving FrameView, FrameLoader, etc.

        * html/HTMLBodyElement.cpp: (WebCore::HTMLBodyElement::insertedIntoDocument):
        Grab the margin from the frame directly here. There was no real benefit
        to doing this via FrameView. Later we can delete quite a bit of unneeded code
        here and in WebKit that exists only to set up the margin values in FrameView,
        which are now unused.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20441 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by Darin.

        Tests for <rdar://problem/5080428> REGRESSION: getting value for hidden input element fails after the fix for Bug 11866
        http://bugs.webkit.org/show_bug.cgi?id=13121

        * fast/dom/HTMLInputElement/input-hidden-value-expected.txt: Added.
        * fast/dom/HTMLInputElement/input-hidden-value.html: Added.
        * fast/forms/input-value-expected.txt:

WebCore:

        Reviewed by Darin.

        Fix for <rdar://problem/5080428> REGRESSION: getting value for hidden input element fails after the fix for Bug 11866
        http://bugs.webkit.org/show_bug.cgi?id=13121

        Rolled out the change for 11866 that made hidden input elements store the value separately from the value attribute.
        Added an m_originalValue field that gets set when the element is done being parsed (in closeRenderer).  In reset,
        use the m_originalValue for hidden input elements.

        * html/HTMLInputElement.cpp:
        (WebCore::HTMLInputElement::closeRenderer):
        (WebCore::HTMLInputElement::reset):
        (WebCore::HTMLInputElement::setValue):
        (WebCore::HTMLInputElement::storesValueSeparateFromAttribute):
        * html/HTMLInputElement.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20440 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Patch by Adam Treat.  Removes hardcoded path to .pro file.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20439 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Add radar reference.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20430 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2007-03-23  Mitz Pettel  <mitz@webkit.org>

        Reviewed by Dave Hyatt.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13157
          List markers positioned incorrectly and don't repaint properly on Oxford homepage

        Test: fast/lists/marker-image-error.html

        * rendering/RenderListMarker.cpp:
        (WebCore::RenderListMarker::imageChanged): Mark ourselves for relayout if
        the image failed to load, since that means we will use a bullet instead.

2007-03-23  Mitz Pettel  <mitz@webkit.org>

        Reviewed by Dave Hyatt.

        - test for http://bugs.webkit.org/show_bug.cgi?id=13157
          List markers positioned incorrectly and don't repaint properly on Oxford homepage

        * fast/lists/marker-image-error-expected.checksum: Added.
        * fast/lists/marker-image-error-expected.png: Added.
        * fast/lists/marker-image-error-expected.txt: Added.
        * fast/lists/marker-image-error.html: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20429 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fix the compile

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20428 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2007-03-23  Mark Rowe  <mrowe@apple.com>

        Build fix for when BUILDING_ON_TIGER is not defined.

        * Misc/WebTypesInternal.h:
        * WebView/WebHTMLView.mm:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20427 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Adam.

        Make script-add.html more deterministic.

        * fast/dom/resources/script3.js:
        * fast/dom/resources/script4.js:
        * fast/dom/resources/script5.js:
        * fast/dom/script-add.html:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20426 268f45cc-cd09-0410-ab3c-d52691b4dbfc

printing changes

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20425 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed/landed by Adam.

        - Adding mac project files for apollo port of JavaScriptCore. Currently project
        just builds dftables.

        * JavaScriptCore.apolloproj/mac/JavaScriptCore/JavaScriptCore.Debug.xcconfig: Added.
        * JavaScriptCore.apolloproj/mac/JavaScriptCore/JavaScriptCore.Release.xcconfig: Added.
        * JavaScriptCore.apolloproj/mac/JavaScriptCore/JavaScriptCore.xcconfig: Added.
        * JavaScriptCore.apolloproj/mac/JavaScriptCore/JavaScriptCore.xcodeproj/project.pbxproj: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20422 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Ada.

        <rdar://problem/5074974>
        CrashTracer: [USER] repro crash in Safari at com.apple.WebCore: WebCore::FrameLoader::activeDocumentLoader const + 6

        * loader/mac/NetscapePlugInStreamLoaderMac.mm:
        (WebCore::NetscapePlugInStreamLoader::didReceiveResponse):
        Don't use ResourceResponse:isHTTP here since that only looks at the protocol instead of if the response came from a
        HTTP server (and not a web archive).

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20419 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Not reviewed - gdk build fix.

        * platform/graphics/cairo/ImageSourceCairo.cpp:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20417 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fix wheeling issue on gamespot.com

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20416 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by kevin

        <rdar://problem/5081257>
        REGRESSION: New Mail signatures start out empty

        * editing/selection/5081257-1-expected.checksum: Added.
        * editing/selection/5081257-1-expected.png: Added.
        * editing/selection/5081257-1-expected.txt: Added.
        * editing/selection/5081257-1.html: Added.
        * editing/selection/5081257-2-expected.checksum: Added.
        * editing/selection/5081257-2-expected.png: Added.
        * editing/selection/5081257-2-expected.txt: Added.
        * editing/selection/5081257-2.html: Added.

WebCore:

        Reviewed by kevin

        <rdar://problem/5081257>
        REGRESSION: New Mail signatures start out empty

        * editing/VisiblePosition.cpp:
        (WebCore::VisiblePosition::canonicalPosition): Allow
        descent from positions inside the document into an
        editable body.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20413 268f45cc-cd09-0410-ab3c-d52691b4dbfc

WebKit:

        Reviewed by Darin.

        Use BUILDING_ON_TIGER from WebKitPrefix.h instead of local
        MAC_OS_X_VERSION_MAX_ALLOWED <= MAC_OS_X_VERSION_10_4 tests.

        * Misc/WebTypesInternal.h:
        * WebView/WebHTMLView.mm:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20412 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Adele and John

        <rdar://5043528> - Prepare for new API

        * platform/network/ResourceHandle.cpp:
        (WebCore::ResourceHandle::clearAuthentication): Clear CF type

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20411 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Patch by Hyatt, reviewed and committed by me.

        Does not reproduce consistently, so a layout test does not seem
        possible.

        Fix for <rdar://problem/5082421> Constantly hitting ASSERT(!
        needsLayout()) in RenderView::paint()

        * page/FrameView.cpp:
        (WebCore::FrameView::layoutPending): It is not enough to ask if the
        layout timer is active. There may be times that we don't have a
        body yet so we cannot schedule layout yet, but the root still needs
        layout.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20410 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by Geoff.

        Replaced all calls to eventSender.mouseClick with paired calls to
        mouseDown/mouseUp.

        * fast/events/click-count.html:
        * fast/events/event-sender-mouse-click-expected.txt: Removed.
        * fast/events/event-sender-mouse-click.html: Removed.
        * fast/events/frame-click-focus.html:
        * fast/events/objc-event-api.html:
        * fast/forms/focus-selection-input.html:
        * fast/forms/focus-selection-textarea.html:

WebKitTools:

        Reviewed by Geoff.

        Removed eventSender.mouseClick because it was only useful for AppKit
        controls, which we don't use anymore.

        * DumpRenderTree/EventSendingController.m:
        (+[EventSendingController isSelectorExcludedFromWebScript:]): Removed
        mouseClick.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20407 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Darin

        Followup for recent fix to 5079700, found by assertion added to Safari.

        * page/mac/FrameMac.mm:
        (WebCore::Frame::searchForLabelsBeforeElement):
        Fixed another place in this method where an empty NSString was being returned instead of nil.
        Cleaned up the style a little too.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20404 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Remove unnecessary destroyFrameAtIndex call.  Now that we throw away
        the image source, it is no longer needed.

        Reviewed by beth

        * platform/graphics/BitmapImage.cpp:
        (WebCore::BitmapImage::destroyDecodedData):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20392 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Geoff, Darin.

        <rdar://problem/5077892> In RapidWeaver 3.5.1, a crash occurs when attempting to load Safari Bookmarks in iMedia Browser

        * platform/mac/ThreadCheck.mm:
        (WebCore::_WebCoreThreadViolationCheck):
        Change default on Tiger to NSLog rather than raise an exception.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20390 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Brady.

        - use binary_search instead of hash table; slightly better performance and
          less memory use

        * platform/network/ResourceHandle.cpp: (WebCore::ResourceHandle::portAllowed):
        Use binary_search.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20389 268f45cc-cd09-0410-ab3c-d52691b4dbfc

WebCore:

        Reviewed by Adele.

        - fix <rdar://problem/5074630> detachChildren call should move from WebKit to WebCore

        * loader/FrameLoader.cpp: (WebCore::FrameLoader::setDocumentLoader):
        Add a call to detachChildren() after the call to prepareForDataSourceReplacement().
        There was no reason for this crucial loading step to be left to the client.

WebKit:

        Reviewed by Adele.

        - fix <rdar://problem/5074630> detachChildren call should move from WebKit to WebCore

        * WebCoreSupport/WebFrameLoaderClient.mm: (WebFrameLoaderClient::prepareForDataSourceReplacement):
        Remove call to detachChildren. This should be a WebCore responsibility.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20388 268f45cc-cd09-0410-ab3c-d52691b4dbfc

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20387 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by Ada.

        Added a new test to make sure that eventSender.mouseClick works
        correctly.

        * fast/events/click-count-expected.txt: Added.
        * fast/events/click-count.html: Added.

        Updated results for some tests that depended on the broken version of
        eventSender.mouseClick.

        * fast/events/objc-event-api-expected.txt:
        * fast/forms/focus-selection-input-expected.txt:
        * fast/forms/focus-selection-textarea-expected.txt:

WebKitTools:

        Reviewed by Ada.

        Make eventSender.mouseClick actually send a mouse up event.

        * DumpRenderTree/EventSendingController.m:
        (-[EventSendingController mouseClick]):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20386 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Minor refactoring and cleanup of the bridge calls that want to control layout settings on the RenderView.
        Have the bridge talk through the FrameView instead of just asking for the RenderView directly.

        Add an assert to help catch situations where the RenderView needs layout at paint time, since this is a known
        catastrophic scenario that will (much of the time) result in a crash in RenderTableSection::paint.

        Reviewed by aroben

        * page/FrameView.cpp:
        (WebCore::FrameView::adjustViewSize):
        (WebCore::FrameView::needsLayout):
        (WebCore::FrameView::setNeedsLayout):
        * page/FrameView.h:
        * page/mac/WebCoreFrameBridge.mm:
        (-[WebCoreFrameBridge needsLayout]):
        (-[WebCoreFrameBridge setNeedsLayout]):
        * rendering/RenderView.cpp:
        (WebCore::RenderView::paint):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20385 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Beth Dakin.

        Re-worked the fix in r20375. 'M' and 'm' are not the only commands that
        accept "extra" coordinates -- virtually all commands accept them.

        * ksvg2/svg/SVGParserUtilities.cpp:
        (WebCore::SVGPathParser::parseSVG):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20384 268f45cc-cd09-0410-ab3c-d52691b4dbfc

WebKitSite:

        Reviewed by NOBODY (more word tweaking).

        * building/checkout.html: By popular demand, the link to the svn 1.3 client has been removed.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20383 268f45cc-cd09-0410-ab3c-d52691b4dbfc

WebKitSite:

        Reviewed by NOBODY (tweaked wording).

        * building/checkout.html: Got rid of "click here" link and restructured sentences
        linking to svn clients.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20382 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by hyatt.

        <rdar://problem/5072460> CrashTracer: [USER] 1 crashes in Xcode at com.apple.ImageIO.framework: getBandProcPNG + 168

        An NSData object was being alloc/init'd, then returned callers who cast the pointer as a
        CFDataRef, including calling CFRelease on it. The problem is that under garbage collection, the NS
        retain count is ignored (it's always 0), but the CFRetain and CFRelease are not ignored. This
        caused the object to be over-released. The solution that works in both GC and non-GC is to "transfer"
        the initial NS retain count to the CF retain count, using HardRetainWithNSRelease.

        The creator of the NSData was SharedBuffer::createNSData. The callers were PDFDocumentImage::dataChanged()
        and ImageSource::setData(). This particular crash involved the ImageSource::setData() case.

        * platform/SharedBuffer.h:
        Declare createCFData().

        * platform/graphics/cg/ImageSourceCG.cpp:
        (WebCore::ImageSource::setData):
        Call createCFData instead of createNSData.

        * platform/graphics/cg/PDFDocumentImage.cpp:
        (WebCore::PDFDocumentImage::dataChanged):
        Call createCFData instead of createNSData.

       * platform/mac/SharedBufferMac.mm:
        (WebCore::SharedBuffer::createCFData):
        Implement createCFData(). Use HardRetainWithNSRelease for gc safety.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20381 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by Darin.  Patch by Mitz.

        - test for http://bugs.webkit.org/show_bug.cgi?id=13145
          Regression: Scrollbar not resizing after display none

        * fast/dynamic/view-overflow-expected.checksum: Added.
        * fast/dynamic/view-overflow-expected.png: Added.
        * fast/dynamic/view-overflow-expected.txt: Added.
        * fast/dynamic/view-overflow.html: Added.

WebCore:

        Reviewed by Darin.  Patch by Mitz.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13145
          Regression: Scrollbar not resizing after display none

        Test: fast/dynamic/view-overflow.html

        * rendering/RenderView.cpp:
        (WebCore::RenderView::layout): Reset overflowHeight() and overflowWidth()
        before calling docHeight() and docWidth(), since the former act as a lower
        bound for the latter.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20380 268f45cc-cd09-0410-ab3c-d52691b4dbfc

WebKitSite:

        Reviewed by Adam.

        * building/checkout.html: Add link to svn 1.4 client for Mac OS X 10.4.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20379 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Fixed bad version of results that I check in accidentally.

        * svg/dom/string-list-parser-expected.txt:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20378 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Beth Dakin.

        A little cleanup based work I've been doing on the SVG parsers.

        * ksvg2/svg/SVGParserUtilities.h:
        (WebCore::skipString):
        * ksvg2/svg/SVGPreserveAspectRatio.cpp: Moved checkString into
        SVGParserUtilities.h, so it could be with all its friends. Renamed
        "checkString" to "skipString" to match the rest of the code and to be
        clear about which functions move the buffer pointer.
        (WebCore::SVGPreserveAspectRatio::parsePreserveAspectRatio):
        * ksvg2/svg/SVGTransformable.cpp:
        (WebCore::SVGTransformable::parseTransformAttribute): Changed gotos into
        returns. Removed unnecessary magic number.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20377 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by Beth Dakin.

        Added fuzz tester for the SVG rgb color parser.

        * svg/dom/rgb-color-parser-expected.txt: Added.
        * svg/dom/rgb-color-parser.html: Added.

WebCore:

        Reviewed by Beth Dakin.

        Fixed <rdar://problem/5078471> SVG rgb color parser runs past end of
        buffer when running fuzzing test

        * ksvg2/svg/SVGColor.cpp:
        (WebCore::parseNumberOrPercent): Check for past the end condition.
        (WebCore::SVGColor::colorFromRGBColorString): Reversed "read past end
        of buffer, then check if you're past the end" logic.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20376 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by Beth Dakin.

        Added fuzz tester for the path parser.

        * svg/dom/path-parser-expected.txt: Added.
        * svg/dom/path-parser.html: Added.

WebCore:

        Reviewed by Beth Dakin.

        Fixed <rdar://problem/5079410> SVG path parser hangs, exhausts memory when
        running fuzzer test

        * ksvg2/svg/SVGParserUtilities.cpp:
        (WebCore::SVGPathParser::parseSVG): Don't assume that extra digits are
        always preceded by an 'M' or an 'm'.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20375 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Beth Dakin.

        Added fuzz tester for the string list parser.

        * svg/dom/string-list-parser-expected.txt: Added.
        * svg/dom/string-list-parser.html: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20374 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Beth Dakin.

        Added fuzz tester for the length list parser.

        * svg/dom/length-list-parser-expected.txt: Added.
        * svg/dom/length-list-parser.html: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20373 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Beth Dakin.

        Factored randomizing functions into their own file, in preparation for future
        tests that rely on them.

        * svg/dom/resources/scripted-random.js: Added.
        * svg/dom/transform-parser.html: Factored out scripted-random.js, added
        a small string that increases the normal-seemingness of the last test.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20372 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    New results after rolling out r19828 in r20352.

        * fast/overflow/scrollRevealButton-expected.checksum:
        * fast/overflow/scrollRevealButton-expected.png:
        * fast/overflow/scrollRevealButton-expected.txt:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20371 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Adam

        - fixed <rdar://problem/5079700> REGRESSION (r19702): Name/password aren't autofilled
          or saved for .Mac homepage login

        * page/mac/FrameMac.mm:
        (WebCore::Frame::searchForLabelsBeforeElement):
        don't count empty strings as valid results

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20370 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by darin

        * editing/markup.cpp:
        (WebCore::styleFromMatchedRulesAndInlineDecl):
        Release the RefPtr to avoid increasing/decreasing
        the refcount unnecessarily.
        (WebCore::propertyMissingOrEqualToNone): Added and used
        instead of converting values to strings.
        (WebCore::elementHasTextDecorationProperty):
        (WebCore::createMarkup):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20369 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Darin.

        <rdar://problem/5076599> JavaScriptCore has a weak export (vtable for KJS::JSCell)

        * JavaScriptCore.exp: Remove __ZTVN3KJS6JSCellE.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20368 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    - some specs to propose by email

        * specs/HTML_Timed_Media_Elements.html: Added.
        * specs/Timed_Media_CSS.html: Added.
        * specs/Timed_Media_Elements-Open_Issues.html: Added.
        * specs/mediaelement.css: Added.
        * specs/movie-status-states.png: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20367 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by oliver

        <rdar://problem/5075944>
        REGRESSION: Partially selected underlined content pastes as plain text

        Tests for the bug:
        * editing/pasteboard/5075944-expected.checksum: Added.
        * editing/pasteboard/5075944-expected.png: Added.
        * editing/pasteboard/5075944-expected.txt: Added.
        * editing/pasteboard/5075944.html: Added.
        Tests that the element that has the text-decoration
        and all its descendants down to the common ancestor
        are included in the copied markup (instead of the
        text-decoration property being treated as though it
        were inheritable and placed on the style span wrapper):
        * editing/pasteboard/5075944-2-expected.checksum: Added.
        * editing/pasteboard/5075944-2-expected.png: Added.
        * editing/pasteboard/5075944-2-expected.txt: Added.
        * editing/pasteboard/5075944-2.html: Added.
        Tests that moveParagraphs gets underlining right (it doesn't
        use annotation):
        * editing/pasteboard/5075944-3-expected.checksum: Added.
        * editing/pasteboard/5075944-3-expected.png: Added.
        * editing/pasteboard/5075944-3-expected.txt: Added.
        * editing/pasteboard/5075944-3.html: Added.

WebCore:

        Reviewed by oliver

        <rdar://problem/5075944>
        REGRESSION: Partially selected underlined content pastes as plain text

        We stopped adding markup for all the ancestors
        of lastClosed up to the commonAncestorBlock
        because it was adding a lot of unnecessary markup.
        This caused us to lose underlining when copying a
        partially selected underlined element (even though
        we put all styles that the copied markup inherits
        into a style span, because the text-decoration
        isn't inheritable).

        * editing/markup.cpp:
        (WebCore::styleFromMatchedRulesAndInlineDecl): Moved
        code here.
        (WebCore::elementHasTextDecorationProperty): Added.
        (WebCore::createMarkup): If the copied markup has a
        text-decoration because some common ancestor has
        a text-decoration property set, include that ancestor
        and all its descendants in the copied markup.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20366 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2007-03-21  Oliver Hunt  <oliver@apple.com>

        Reviewed by Antti.
WebCore:
        Update a few SVG attribute parsers to be more defensive against
        reading beyond the end of input.

        Fixes rdar://problem/5077218 -- SVG transform parser runs past
        end of buffer when running fuzzing test

        * ksvg2/svg/SVGParserUtilities.cpp:
        (WebCore::SVGPathParser::parseSVG):
        * ksvg2/svg/SVGParserUtilities.h:
        (WebCore::skipOptionalSpacesOrDelimiter):
        * ksvg2/svg/SVGTransformable.cpp:
        (WebCore::parseTransformParamList):
        (WebCore::SVGTransformable::parseTransformAttribute):

LayoutTests:
        Fuzz test for SVG transform parser

        * svg/dom/transform-parser-expected.txt: Added.
        * svg/dom/transform-parser.html: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20364 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Mitz.

        http://bugs.webkit.org/show_bug.cgi?id=13140
        Crash after loading SVG fragment with XMLHttpRequest (Yosemite Map Geology Layer)

        Test: svg/carto.net/frameless-svg-parse-error.html

        * ksvg2/misc/SVGDocumentExtensions.cpp:
        (WebCore::SVGDocumentExtensions::reportError): Added a null check.
        (WebCore::SVGDocumentExtensions::reportWarning): Ditto.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20363 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Geoff.

        * API/JSStringRef.cpp: (JSStringIsEqual): Added JSLock.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20362 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fix the compile when USE(MULTIPLE_THREADS) isn't defined

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20361 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Fix for <rdar://problem/5072678> Crash in
        RenderLayer::scrollRectToVisible with MallocScribble enabled

        I tried to fix this earlier today but my fix caused a regression
        scrolling through RSS pages. It turns out that sometimes scroll
        events need to propagate immediately, and sometimes they must be
        delayed or they risk deleting objects that are expected to be
        around after the event has propagated. Mitz's original fix made
        sheduleEvent() only delay events that happen during layout. This
        fix marks two other places in addition to layout where events also
        need to be delayed. These two places are places that were marked
        with FIXMEs that Mitz removed in his original patch. (There was a
        third FIXME in RenderLayer::updateScrollInfoAfterLayout() but that
        case is only called through layout and is covered by Mitz's
        original patch.)

        * page/FrameView.cpp:
        (WebCore::FrameView::~FrameView): Added assertion.
        (WebCore::FrameView::layout): Call new functions to increment and
        decrement the queued events count.
        (WebCore::FrameView::pauseScheduledEvents):
        (WebCore::FrameView::resumeScheduledEvents): Decrement queued
        events count and dispatch events if the count has zeroed.
        * page/FrameView.h:
        * rendering/RenderLayer.cpp:
        (WebCore::RenderLayer::scrollRectToVisible):
        (WebCore::Marquee::start):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20360 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2007-03-21  Mitz Pettel  <mitz@webkit.org>

        Reviewed by John Sullivan.

        - fix http://bugs.webkit.org/show_bug.cgi?id=13130
          REGRESSION: Sometimes a frame resizer keeps following the mouse after mouseup

        * rendering/RenderFrameSet.cpp:
        (WebCore::RenderFrameSet::userResize): Reordered the code to allow resizing
        to be terminated by a mouseup event regardless of whether the frameset needs
        layout.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20359 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Oliver.

        <rdar://problem/5077252>
        In NetNewsWire 2.1.1 and 3.0d62, a crash occurs after triple-clicking a link on the page

        * loader/FrameLoader.cpp:
        (WebCore::FrameLoader::load):
        Move the m_inStopAllLoaders check to the load function that's common for loading a request or substitute data.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20358 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Anders

        <rdar://problem/5073391> and http://bugs.webkit.org/show_bug.cgi?id=13137

        Crash in IconDatabase when private browsing is enabled.

        The problem was caused by http://trac.webkit.org/projects/webkit/changeset/20182
        which changed many uses of char[] and Vector<char> to SharedBuffer.  The patch
        tended to literally replace a Vector<char> with RefPtr<SharedBuffers> but forgot
        to enforce the concept that Vector<char>'s always exist, whereas RefPtr<SharedBuffers>
        can be null.  This led to derefs.

        I took the opportunity to rework the iconDB functions to live in a SharedBuffer
        world, as that didn't exist when they were originally written - now they just return
        SharedBuffers instead of taking a Vector<char>& as a parameter

        * loader/icon/IconDatabase.cpp:
        (WebCore::IconDatabase::imageDataForIconURL): Return a SharedBuffer
        (WebCore::IconDatabase::iconForPageURL): Null check the SharedBuffer before asking
          it if it's empty
        (WebCore::IconDatabase::imageDataForIconURLQuery): Return a new SharedBuffer

        * loader/icon/IconDatabase.h: Return SharedBuffer's instead of taking Vector<char>&'s

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20357 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Rubberstamped by Adele.

        Roll out r20348 because it was causing crashes.

        * page/FrameView.cpp:
        (WebCore::FrameViewPrivate::FrameViewPrivate):
        (WebCore::FrameView::layout):
        (WebCore::FrameView::scheduleEvent):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20355 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2007-03-20  Matt Lilek  <pewtermoose@gmail.com>

        Reviewed by Darin.

        Clear up the steps necessary to get a patch landed.

        http://bugs.webkit.org/show_bug.cgi?id=12877 - Bug 12877: "Contributing Code" page could be clearer
        http://bugs.webkit.org/show_bug.cgi?id=8690 - Bug 8690: Contributing code doesn't mention what to do with new files

        * coding/contributing.html:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20354 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Build fix.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20353 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Kevin Decker.

        Rolling out part of r19828 that caused a scrolling regression in Mail.

        <rdar://problem/4820817> Message jumps around when trying to select text

        * platform/mac/ScrollViewMac.mm:
        (WebCore::ScrollView::visibleContentRect):
        (WebCore::ScrollView::updateContents):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20352 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Geoff and Adam.

        - make USE(MULTIPLE_THREADS) support more portable
        http://bugs.webkit.org/show_bug.cgi?id=13069

        - fixed a threadsafety bug discovered by testing this

        - enhanced threadsafety assertions in collector

        * API/JSCallbackObject.cpp:
        (KJS::JSCallbackObject::~JSCallbackObject): This destructor can't
        DropAllLocks around the finalize callback, because it gets called
        from garbage collection and we can't let other threads collect!

        * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
        * kjs/JSLock.cpp:
        (KJS::JSLock::currentThreadIsHoldingLock): Added new function
        to allow stronger assertions than just that the lock is held
        by some thread (you can now assert that the current thread is
        holding it, given the new JSLock design).
        * kjs/JSLock.h:
        * kjs/collector.cpp: Refactored for portability plus added some
        stronger assertions.
        (KJS::Collector::allocate):
        (KJS::currentThreadStackBase):
        (KJS::Collector::registerAsMainThread):
        (KJS::onMainThread):
        (KJS::PlatformThread::PlatformThread):
        (KJS::getCurrentPlatformThread):
        (KJS::Collector::Thread::Thread):
        (KJS::destroyRegisteredThread):
        (KJS::Collector::registerThread):
        (KJS::Collector::markCurrentThreadConservatively):
        (KJS::suspendThread):
        (KJS::resumeThread):
        (KJS::getPlatformThreadRegisters):
        (KJS::otherThreadStackPointer):
        (KJS::otherThreadStackBase):
        (KJS::Collector::markOtherThreadConservatively):
        (KJS::Collector::markStackObjectsConservatively):
        (KJS::Collector::protect):
        (KJS::Collector::unprotect):
        (KJS::Collector::collectOnMainThreadOnly):
        (KJS::Collector::markMainThreadOnlyObjects):
        (KJS::Collector::collect):
        * kjs/collector.h:
        * wtf/FastMalloc.cpp:
        (WTF::fastMallocSetIsMultiThreaded):
        * wtf/FastMallocInternal.h:
        * wtf/Platform.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20351 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John

        <rdar://5071341> - Crash in IconLoader::finishLoading()

        Though the reproducibility of the crash is still elusive, there's been a crasher likely uncovered
        Hyatt's recent caching changings to the IconLoader.  This crash tends to hit on sites without favicons,
        that serve up complex 404 pages, often when the load is cancelled or redirected.  Occasionally, it *seems*
        the IconLoader is getting failed/cancelled twice.

        This patch adds some assertions and a "fail gracefully in a release build" check that will have to
        serve us until we know more about the cause.

        * loader/icon/IconLoader.cpp:
        (WebCore::IconLoader::didFail): Add an assertion to catch this case
        (WebCore::IconLoader::finishLoading): Add an assertion to catch the case, and fail
          gracefully in debug builds.  Also add an efficiency check to not commit to the DB if there is no
          IconURL

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20350 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Darin.

        - fix http://bugs.webkit.org/show_bug.cgi?id=10747
        REGRESSION: Using two fingers on a trackpad to scroll inside a text area makes the whole page move
        <rdar://problem/5045710>

        Eat smooth scroll events in WebCore if pointer is over scrollable area. Not
        really a good fix, smooth scrolling should be supported properly. It should do for now.

        * page/EventHandler.cpp:
        (WebCore::EventHandler::handleWheelEvent):
        * rendering/RenderListBox.cpp:
        (WebCore::RenderListBox::isScrollable):
        * rendering/RenderListBox.h:
        * rendering/RenderObject.cpp:
        (WebCore::RenderObject::isScrollable):
        * rendering/RenderObject.h:
        * rendering/RenderTextControl.cpp:
        (WebCore::RenderTextControl::isScrollable):
        * rendering/RenderTextControl.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20349 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Geoff.

        Covered by existing test case.

        Fix for <rdar://problem/5072678> Crash in
        RenderLayer::scrollRectToVisible with MallocScribble enabled

        This removes some code that was added with http://bugs.webkit.org/
        show_bug.cgi?id=8360. This code was not needed to fix that bug, but
        the idea behind it was that there are times when it is safe for
        scheduleEvent() to propagate the event immediately. The
        MallocScribble crash revealed that this assumption was a bit too
        broad; the cause of the crash was that we were propagating events
        immediately when it was not safe. Mitz and I discussed this online
        and we agree that the safest solution is to revert to the old
        behavior here, and always delay the propagation of the event.

        * page/FrameView.cpp:
        (WebCore::FrameViewPrivate::FrameViewPrivate):
        (WebCore::FrameView::layout):
        (WebCore::FrameView::scheduleEvent):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20348 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John Sullivan

        Add some very helpful logging to the IconLoader's SubresourceLoaderClient impl.

        * loader/icon/IconLoader.cpp:
        (WebCore::IconLoader::didReceiveResponse):
        (WebCore::IconLoader::didReceiveData):
        (WebCore::IconLoader::didFail):
        (WebCore::IconLoader::didFinishLoading):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20347 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests:

        Reviewed by darin

        <rdar://problem/5071074>
        REGRESSION: Links are pasted as plain text at certain positions

        * editing/pasteboard/5071074-2-expected.checksum: Added.
        * editing/pasteboard/5071074-2-expected.png: Added.
        * editing/pasteboard/5071074-2-expected.txt: Added.
        * editing/pasteboard/5071074-2.html: Added.

WebCore:

        Reviewed by darin

        <rdar://problem/5071074>
        REGRESSION: Links are pasted as plain text at certain positions

        moveParagraphs calls createMarkup without annotation, which leaves
        out anchor elements that aren't fully selected.

        * editing/CompositeEditCommand.cpp:
        (WebCore::CompositeEditCommand::moveParagraphs): We don't want
        createMarkup to do full annotation here.  Doing so could include
        special ancestor blocks, which would prevent the copied content
        from appearing in the destination paragraph when it's re-inserted.
        Annotation would also inline styles from style sheets,  which is
        unnecessary because the content is being moved to the same
        document it's being moved from.
        Added a new bool to createMarkup that when true allows markup for
        special ancestors as long as they are inline when annotation is off.
        * editing/markup.cpp:
        (WebCore::createMarkup): Don't set specialCommonAncestor if annotation
        is off, unless it's inline and we've been told it's ok to include such
        elements.
        * editing/markup.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20346 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Darin.

        http://bugs.webkit.org/show_bug.cgi?id=12497
        Implement XPath result ordering.

WebCore:
        XPath::NodeVector typedef is replaced with a real XPath::NodeSet class that knows how
        to sort itself, and can remember whether it has been already sorted.

        * CMakeLists.txt:
        * WebCore.pro:
        * WebCore.xcodeproj/project.pbxproj:
        * WebCoreSources.bkl:
        Added XPathNodeSet files.

        * xml/XPathExpression.cpp: Fixed includes.

        * xml/XPathFunctions.cpp:
        (WebCore::XPath::FunId::evaluate): Mark the resulting node-set as unsorted.
        (WebCore::XPath::FunLocalName::evaluate): Replacing NodeVector with NodeSet.
        (WebCore::XPath::FunNamespaceURI::evaluate): Ditto.
        (WebCore::XPath::FunName::evaluate): Ditto.
        (WebCore::XPath::FunCount::evaluate): Ditto.
        (WebCore::XPath::FunSum::evaluate): Ditto.

        * xml/XPathNodeSet.cpp: Added.
        (WebCore::XPath::parentWithDepth):
        (WebCore::XPath::sortBlock):
        (WebCore::XPath::NodeSet::sort): Sort the node-set in document order.
        (WebCore::XPath::NodeSet::reverse): Reverse the order (useful for making axes such as
        parent or ancestor sorted).
        (WebCore::XPath::NodeSet::firstNode): Returns the first node in document order; currently
        implemented via fully sorting the node-set, but this can obviously be optimized.
        (WebCore::XPath::NodeSet::anyNode): Added for symmetry with firstNode().

        * xml/XPathNodeSet.h: Added.
        (WebCore::XPath::NodeSet::NodeSet):
        (WebCore::XPath::NodeSet::operator=):
        (WebCore::XPath::NodeSet::size):
        (WebCore::XPath::NodeSet::isEmpty):
        (WebCore::XPath::NodeSet::operator[]):
        (WebCore::XPath::NodeSet::reserveCapacity):
        (WebCore::XPath::NodeSet::clear):
        (WebCore::XPath::NodeSet::swap):
        (WebCore::XPath::NodeSet::append):
        (WebCore::XPath::NodeSet::markSorted):
        (WebCore::XPath::NodeSet::isSorted):
        Most of these methods just call Vector counterparts.

        * xml/XPathParser.cpp: Updated the copyright notice.

        * xml/XPathPath.cpp:
        (WebCore::XPath::Filter::evaluate): Replacing NodeVector with NodeSet.
        (WebCore::XPath::Path::evaluate): Ditto.
        (WebCore::XPath::LocationPath::evaluate): Replacing NodeVector with NodeSet. This function
        always marks the result as unsorted, because it is hard to tell whether a step breaks node order.
        Identifying and implementing special cases when it is not necessary to do so is an important
        future optimization.

        * xml/XPathPath.h: Replacing NodeVector with NodeSet.
        * xml/XPathPredicate.cpp:
        (WebCore::XPath::EqTestOp::compare): Replacing NodeVector with NodeSet.
        (WebCore::XPath::Union::evaluate): Replacing NodeVector with NodeSet. Currently, Union just
        marks the result as unordered; we can consider using merge sort to avoid this.

        * xml/XPathResult.cpp:
        (WebCore::XPathResult::XPathResult): Replacing NodeVector with NodeSet.
        (WebCore::XPathResult::singleNodeValue): Ditto.
        (WebCore::XPathResult::snapshotLength): Ditto.
        (WebCore::XPathResult::iterateNext): Ditto.
        (WebCore::XPathResult::snapshotItem): Ditto.
        (WebCore::XPathResult::convertTo): Ditto. Sort the result when requested to.

        * xml/XPathResult.h: Replacing NodeVector with NodeSet.

        * xml/XPathStep.cpp:
        (WebCore::XPath::Step::evaluate): If the input is not sorted, mark the output as such, too.
        (WebCore::XPath::Step::nodesInAxis): Fixed a number of bugs when enumerating with an
        attribute context node.
        (WebCore::XPath::Step::nodeTestMatches): Replacing NodeVector with NodeSet.
        * xml/XPathStep.h: Ditto.

        * xml/XPathUtil.cpp:
        (WebCore::XPath::isValidContextNode): XPath data model doesn't put attribute data into child
        nodes, so passing such node as a context could cause problems.

        * xml/XPathUtil.h: Removed NodeVector typedef.

        * xml/XPathValue.cpp:
        (WebCore::XPath::Value::Value):
        (WebCore::XPath::Value::toNodeSet):
        (WebCore::XPath::Value::toBoolean):
        (WebCore::XPath::Value::toNumber):
        (WebCore::XPath::Value::toString):
        * xml/XPathValue.h:
        (WebCore::XPath::Value::):
        (WebCore::XPath::Value::isNodeSet):
        Replacing NodeVector with NodeSet.

LayoutTests:
        * fast/xpath/document-order-expected.txt: Added.
        * fast/xpath/document-order.html: Added.

        * fast/xpath/text-nodes-expected.txt:
        * fast/xpath/text-nodes.html:
        * fast/xpath/4XPath/Core/test_step-expected.txt:
        Updated results for tests that now pass.

        * fast/xpath/xpath-test-pre.js: Added.
        * fast/xpath/4XPath/Core/test.js:
        Moved checkSnapshot() to a separate file.

        * fast/xpath/4XPath/Core/test_core_functions.html:
        * fast/xpath/4XPath/Core/test_location_path.html:
        * fast/xpath/4XPath/Core/test_nodeset_expr.html:
        * fast/xpath/4XPath/Core/test_parser.html:
        * fast/xpath/4XPath/Core/test_predicate_list.html:
        * fast/xpath/4XPath/Core/test_step.html:
        Load xpath-test-pre.js for checkSnapshot().

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20345 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2007-03-21  Mark Rowe  <mrowe@apple.com>

        Build fix.

        * bindings/objc/WebScriptObject.mm:
        (-[WebScriptObject methodForSelector:]):
        (-[WebScriptObject respondsToSelector:]):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20344 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2007-03-21  Mark Rowe  <mrowe@apple.com>

        Reviewed by Tim Hatcher.

        Fix for http://bugs.webkit.org/show_bug.cgi?id=13129.
        Bug 13129: Democracy Player dies in NSException from WebScriptObject on startup.

        Some cross-language scripting bridges use the presence of -(int)count to determine
        whether to treat objects as sequence-like.  All WebScriptObject's exposed this method
        which lead to non-collection WebScriptObject's being mishandled by PyObjC.  We now
        expose -count only when the object we wrap looks like a collection from a JavaScript
        point of view (eg, it has a length property).

        * bindings/objc/WebScriptObject.mm:
        (-[WebScriptObject _shouldRespondToCount]):
        (-[WebScriptObject methodForSelector:]):
        (-[WebScriptObject respondsToSelector:]):
        (-[WebScriptObject _count]):
        * bindings/objc/WebScriptObjectPrivate.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20343 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    * page/EventHandler.cpp: (WebCore::EventHandler::handleMouseMoveEvent):
        Fix obvious typo in the change from last night -- forgot to assign to the variable.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20342 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Refactor fonts a bit for portability.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20341 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Adam.

        Stub out a couple of CFNetwork backend functions.

        * platform/network/ResourceHandle.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20340 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2007-03-19  Oliver Hunt  <oliver@apple.com>

        Reviewed by Ada.

        Protect against possibility of JS detaching a frame (and losing its view)
        midway through a mouse move.

        * page/EventHandler.cpp:
        (WebCore::EventHandler::handleMouseMoveEvent):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@20339 268f45cc-cd09-0410-ab3c-d52691b4dbfc