watchlist: Add a way to detect a net increase or decrease of a pattern (in a file).
https://bugs.webkit.org/show_bug.cgi?id=69031

Reviewed by Adam Barth.

* Scripts/webkitpy/common/config/watchlist: Added usage of the new pattern,
a comment to explain something important about the email addresses, and
a comment to let emacs know to treat the file as a python file.
* Scripts/webkitpy/common/watchlist/amountchangedpattern.py: Added.
Detects increases or decreases in a pattern.
* Scripts/webkitpy/common/watchlist/amountchangedpattern_unittest.py: Added.
* Scripts/webkitpy/common/watchlist/watchlist_unittest.py:
Added unit tests for watchlist which has "more" or "less".
* Scripts/webkitpy/common/watchlist/watchlistparser.py:
Added support for "more" or "less".

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96441 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fix assertion failure in XSS Auditor
https://bugs.webkit.org/show_bug.cgi?id=69050
https://code.google.com/p/chromium/issues/detail?id=97346

Reviewed by Daniel Bates.

Source/WebCore:

Test: fast/forms/xss-auditor-doesnt-crash.html

* html/parser/XSSAuditor.cpp:
(WebCore::XSSAuditor::init): Don't try to decode the HTML body if it is empty.

LayoutTests:

* fast/forms/xss-auditor-doesnt-crash-expected.txt: Added.
* fast/forms/xss-auditor-doesnt-crash.html: Added.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96440 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Add support for eventSender.mouseScrollBy in WTR
https://bugs.webkit.org/show_bug.cgi?id=69189

Source/WebKit2:

Reviewed by Simon Fraser.

* UIProcess/WebPageProxy.cpp:
(WebKit::WebPageProxy::handleWheelEvent):
* WebProcess/WebPage/WebPage.cpp:
(WebKit::WebPage::wheelEventSyncForTesting):
* WebProcess/WebPage/WebPage.h:
* WebProcess/WebPage/WebPage.messages.in:
Add sync version of wheel event handling for use in testing.

Tools:

Add implementation of eventSender.mouseScrollBy for the mac.

Reviewed by Simon Fraser.

* WebKitTestRunner/EventSenderProxy.h:
* WebKitTestRunner/InjectedBundle/Bindings/EventSendingController.idl:
* WebKitTestRunner/InjectedBundle/EventSendingController.cpp:
(WTR::EventSendingController::keyDown):
(WTR::EventSendingController::mouseScrollBy):
* WebKitTestRunner/InjectedBundle/EventSendingController.h:
* WebKitTestRunner/TestController.cpp:
(WTR::TestController::didReceiveSynchronousMessageFromInjectedBundle):
* WebKitTestRunner/mac/EventSenderProxy.mm:
(WTR::EventSenderProxy::mouseScrollBy):

LayoutTests:

Reviewed by Simon Fraser.

* platform/mac-wk2/Skipped:
Removes tests that used eventSender.mouseScrollBy.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96439 268f45cc-cd09-0410-ab3c-d52691b4dbfc

StringRecursionChecker should not work in terms of EncodedJSValue
https://bugs.webkit.org/show_bug.cgi?id=69188

Reviewed by Oliver Hunt.

0 is not the empty value on 32_64.
Code that casts literals to EncodedJSValues may be unsafe if we change our internal representation.

* runtime/ArrayPrototype.cpp:
(JSC::arrayProtoFuncToString):
(JSC::arrayProtoFuncToLocaleString):
(JSC::arrayProtoFuncJoin):
* runtime/ErrorPrototype.cpp:
(JSC::errorProtoFuncToString):
* runtime/RegExpPrototype.cpp:
(JSC::regExpProtoFuncToString):
* runtime/StringRecursionChecker.cpp:
(JSC::StringRecursionChecker::throwStackOverflowError):
(JSC::StringRecursionChecker::emptyString):
* runtime/StringRecursionChecker.h:
(JSC::StringRecursionChecker::performCheck):
(JSC::StringRecursionChecker::earlyReturnValue):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96438 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Loading page on potterybankids causes reproducible assertion failure in debug builds
https://bugs.webkit.org/show_bug.cgi?id=69185
<rdar://problem/10218241>

Reviewed by Simon Fraser.

* rendering/RenderLayer.cpp:
(WebCore::RenderLayer::computeRepaintRects): Remove assert condition that isn't guaranteed
to be true.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96437 268f45cc-cd09-0410-ab3c-d52691b4dbfc

DFG JIT, Branch on integer can always be a 32-bit compare.
https://bugs.webkit.org/show_bug.cgi?id=69174

Reviewed by Sam Weinig.

if (shouldSpeculateInteger(node.child1()) && !isStrictInt32(node.child1())),
the JSVALUE64 JIT will currently compare all 64bits in the register, but in
these cases the DataFormat is always a JS boxed integer. In these cases we
can just compare the low 32bits anyway - no need to check the tag.
This allows the code to be unified with the JSVALUE32_64 JIT.

* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96436 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Qt] Add missing include for QMimeData
https://bugs.webkit.org/show_bug.cgi?id=69178

Reviewed by Tor Arne Vestbø.

This was implicitly included by other headers, but due to changes in
Qt5 we need to include explicitly.

* UIProcess/qt/QtWebPageProxy.cpp:
* WebProcess/WebPage/WebPage.cpp:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96435 268f45cc-cd09-0410-ab3c-d52691b4dbfc

fast/forms/text-control-intrinsic-widths.html fails on Mac if Word/Office fonts are installed
https://bugs.webkit.org/show_bug.cgi?id=35273

Reviewed by Dan Bernstein.

There are two types of font issue that can cause layout test failure.
First, the user may have non-standard fonts installed that affect the layout
of some tests. For example, some tests use the MS Gothic font if installed.

Secondly, the user may have duplicate fonts installed in ~/Library/Fonts,
which override the system fonts (e.g. a copy of Arial from MS Office). For
some tests, this can affect font metrics or fallback behavior.

This change addresses the first problem only. It limits the set of fonts
available through NSFontManager, by swizzling some methods to return
a list of fonts or font families which only include fonts from a hardcoded
whitelist.

* DumpRenderTree/mac/DumpRenderTree.mm:
(allowedFontFamilySet):
(drt_NSFontManager_availableFontFamilies):
(drt_NSFontManager_availableFonts):
(swizzleNSFontManagerMethods):
(activateTestingFonts):
(adjustFonts):
(createWebViewAndOffscreenWindow):
(prepareConsistentTestingEnvironment):
* WebKitTestRunner/InjectedBundle/mac/InjectedBundleMac.mm:
(WTR::allowedFontFamilySet):
(WTR::drt_NSFontManager_availableFontFamilies):
(WTR::drt_NSFontManager_availableFonts):
(WTR::swizzleNSFontManagerMethods):
(WTR::InjectedBundle::platformInitialize):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96434 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Source/WebCore: Support -webkit-tap-highlight-color when enabling touch events support.
https://bugs.webkit.org/show_bug.cgi?id=48544

Reviewed by Kenneth Rohde Christiansen.

Original code from the iOS WebCore code dump, extracted and modified by tijiang@rim and jnd@chromium.
Supports the new CSS property -webkit-tap-highlight-color on platforms which support touch events.
Platform can override WebCore::RenderTheme::platformTapHighlightColor to use its own platform
specific default tap highlight color. Otherwise RenderTheme::defaultTapHighlightColor will be
used as default tap highlight color.

Test: fast/events/touch/tap-highlight-color.html

* css/CSSComputedStyleDeclaration.cpp:
(WebCore::CSSComputedStyleDeclaration::getPropertyCSSValue):
* css/CSSParser.cpp:
(WebCore::CSSParser::parseValue):
* css/CSSPropertyNames.in:
* css/CSSStyleSelector.cpp:
(WebCore::CSSStyleSelector::applyProperty):
* rendering/RenderTheme.cpp:
(WebCore::RenderTheme::tapHighlightColor):
* rendering/RenderTheme.h:
(WebCore::RenderTheme::platformTapHighlightColor):
* rendering/style/RenderStyle.cpp:
(WebCore::RenderStyle::initialTapHighlightColor):
* rendering/style/RenderStyle.h:
(WebCore::InheritedFlags::tapHighlightColor):
(WebCore::InheritedFlags::setTapHighlightColor):
* rendering/style/StyleRareInheritedData.cpp:
(WebCore::StyleRareInheritedData::StyleRareInheritedData):
(WebCore::StyleRareInheritedData::operator==):
* rendering/style/StyleRareInheritedData.h:

LayoutTests: Support -webkit-tap-highlight-color in WebCore when enabling touch events support.
https://bugs.webkit.org/show_bug.cgi?id=48544.

Reviewed by Kenneth Rohde Christiansen.

Add a layout test to get customized tap highlight color set by page author.

* fast/css/getComputedStyle/computed-style-without-renderer.html:
* fast/css/getComputedStyle/computed-style.html:
* fast/events/touch/script-tests/tap-highlight-color.js: Added.
(onTouchStart):
(touchTargets):
* fast/events/touch/tap-highlight-color-expected.txt: Added.
* fast/events/touch/tap-highlight-color.html: Added.
* svg/css/getComputedStyle-basic.xhtml:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96433 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2011-09-30  Oliver Hunt  <oliver@apple.com>

        Need a sensible GGC policy

        Reviewed by Geoff Garen.

        This replaces the existing random collection policy
        with a deterministic policy based on nursery size.

        * heap/AllocationSpace.cpp:
        (JSC::AllocationSpace::allocateSlowCase):
        * heap/Heap.cpp:
        (JSC::Heap::Heap):
        (JSC::Heap::markRoots):
        (JSC::Heap::collect):
        * heap/Heap.h:
        * heap/MarkedSpace.cpp:
        (JSC::MarkedSpace::MarkedSpace):
        (JSC::MarkedSpace::resetAllocator):
        * heap/MarkedSpace.h:
        (JSC::MarkedSpace::nurseryWaterMark):
        (JSC::MarkedSpace::allocate):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96432 268f45cc-cd09-0410-ab3c-d52691b4dbfc

https://bugs.webkit.org/show_bug.cgi?id=69173

RTL/LTR mixtures still not correct in regions. Simplify the loop once I realized it's only
your directionality that matters and not your containing block's.

Reviewed by Dan Bernstein.

Source/WebCore:

Added new test in fast/regions.

* rendering/RenderBox.cpp:
(WebCore::RenderBox::borderBoxRectInRegion):

LayoutTests:

* fast/regions/multiple-directionality-changes-in-variable-width-regions.html: Added.
* platform/mac/fast/regions/multiple-directionality-changes-in-variable-width-regions-expected.png: Added.
* platform/mac/fast/regions/multiple-directionality-changes-in-variable-width-regions-expected.txt: Added.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96431 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[chromium] Pass a processed bool in WebCompositorClient::didHandleInputEvent to indicate if the event was processed
https://bugs.webkit.org/show_bug.cgi?id=69169

Patch by James Robinson <jamesr@chromium.org> on 2011-09-30
Reviewed by Darin Fisher.

If the compositor does not process the input event the embedder might want to do something with it - for
example, if the input event type is part of a CTRL+T keyboard accelerator sequence the browser might want to
open a new tab. This passes that state through the didHandleInputEvent callback.

* public/WebCompositorClient.h:
* src/WebCompositorImpl.cpp:
(WebKit::WebCompositorImpl::handleInputEvent):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96430 268f45cc-cd09-0410-ab3c-d52691b4dbfc

DFG 32-bit support for op_call and op_construct causes
run-javascriptcore-tests to fail
https://bugs.webkit.org/show_bug.cgi?id=69171

Reviewed by Gavin Barraclough.

This fixes one obvious bug that was causing test failures (no
support for dummy slow case for op_add in 32_64), and disables
op_call and op_construct by default.

* dfg/DFGCapabilities.h:
(JSC::DFG::canCompileOpcode):
* jit/JITArithmetic32_64.cpp:
(JSC::JIT::emit_op_add):
(JSC::JIT::emitSlow_op_add):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96429 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Need to skip webkitpy.common.checkout.baselineoptimizer_unittest.BaselineOptimizerTest on Windows..
https://bugs.webkit.org/show_bug.cgi?id=69039

Reviewed by Adam Barth.

* Scripts/webkitpy/common/checkout/baselineoptimizer_unittest.py:
Actually skip only one specific test that is failing on Windows.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96428 268f45cc-cd09-0410-ab3c-d52691b4dbfc

<rdar://problem/10208291> first-letter in generated before content with display: table is not updated

Reviewed by Sam Weinig.

Source/WebCore:

Test: fast/css-generated-content/first-letter-in-nested-before.html

* rendering/RenderObjectChildList.cpp:
(WebCore::RenderObjectChildList::updateBeforeAfterContent): Preserve the style type of anonymous
containers around the generated content when updating their style. This preserves the first-letter-ness
of the inline wrapping the first letter, which allows updateFirstLetter() to identify it later.

LayoutTests:

* fast/css-generated-content/first-letter-in-nested-before-expected.png: Added.
* fast/css-generated-content/first-letter-in-nested-before-expected.txt: Added.
* fast/css-generated-content/first-letter-in-nested-before.html: Added.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96427 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Inset focus ring (Skia on Mac)
https://bugs.webkit.org/show_bug.cgi?id=69166
http://code.google.com/p/chromium/issues/detail?id=97956

This focus ring was drawn too large and was left tracks
in subsequent invalidations. Make it more closely match
the CG version.

Reviewed by Stephen White.

No new tests. This platform is not enabled.

* platform/graphics/skia/GraphicsContextSkia.cpp:
(WebCore::getFocusRingOutset):
(WebCore::GraphicsContext::drawFocusRing):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96426 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Unreviewed, rolling out r96422.
http://trac.webkit.org/changeset/96422
https://bugs.webkit.org/show_bug.cgi?id=69170

Broke Windows build and kling wanted to roll it out as well
(Requested by rniwa on #webkit).

Patch by Sheriff Bot <webkit.review.bot@gmail.com> on 2011-09-30

* rendering/InlineBox.cpp:
(WebCore::InlineBox::prevOnLineExists):
* rendering/InlineBox.h:
(WebCore::InlineBox::InlineBox):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96425 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Crash due to out of bounds read/write in MarkedSpace
https://bugs.webkit.org/show_bug.cgi?id=69148

This was a case of being surprised by a poorly aritulcated cell size limit,
plus an incorrect ASSERT guarding the cell size limit.

Reviewed by Oliver Hunt.

* heap/MarkedSpace.h:
(JSC::MarkedSpace::sizeClassFor): Changed heap size ranges to be inclusive,
since it makes the ranges easier to understand.

Bumped up the max cell size to support the use case in this bug. Since the
atomSize is much bigger than it used to be, there isn't much accounting
cost to handling more size classes.

Switched to FixedArray, to help catch SizeClass indexing bugs in the future.

* heap/MarkedSpace.cpp:
(JSC::MarkedSpace::MarkedSpace):
(JSC::MarkedSpace::resetAllocator):
(JSC::MarkedSpace::canonicalizeCellLivenessData): Updated for size ranges
being inclusive.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96424 268f45cc-cd09-0410-ab3c-d52691b4dbfc

https://bugs.webkit.org/show_bug.cgi?id=69167

[CSS3 Regions] Fix some bugs in the computation of the logical left and width offsets
for blocks in variable width regions. Specifically, this patch addresses a minor bug
with overflow:hidden/scroll/auto block placement and with RTL directionality.

Reviewed by Sam Weinig.

Source/WebCore:

Added five new tests in fast/regions.

* rendering/RenderBlock.cpp:
(WebCore::RenderBlock::logicalRightOffsetForContent):
Avoid a double fetch of the border box in a given region by not calling
logicalLeftOffsetForContent. No correctness change here. It's just cleaner.

* rendering/RenderBox.cpp:
(WebCore::RenderBox::borderBoxRectInRegion):
Change the shifted optimization to be cached for "self" rather than for "containing block".
Rework the adjustment loop to handle changes in directionality and to fix correctness
issues with RTL.

(WebCore::RenderBox::containingBlockLogicalWidthForContentInRegion):
Don't double adjust if an object uses line width as the containing block width. In that
case we already computed the line width in the region, so we don't need to apply any
additional fixup.

(WebCore::RenderBox::renderBoxRegionInfo):
* rendering/RenderBoxRegionInfo.h:
(WebCore::RenderBoxRegionInfo::RenderBoxRegionInfo):
(WebCore::RenderBoxRegionInfo::isShifted):
Changed the shifted code to be on self rather than on containing block.

* rendering/RenderFlowThread.cpp:
(WebCore::RenderFlowThread::layout):
Refactor the layout code for RenderFlowThread to properly set up region rects assuming
right alignment in RTL code. This change forces us to compute our maximum logical width
first before looping through the regions again to set their content rects.

LayoutTests:

* fast/regions/overflow-in-variable-width-regions.html: Added.
* fast/regions/overflow-rtl-in-variable-width-regions.html: Added.
* fast/regions/percentage-margins-mixed-ltr-dominant-regions.html: Added.
* fast/regions/percentage-margins-mixed-rtl-dominant-regions.html: Added.
* fast/regions/percentage-margins-rtl-variable-width-regions.html: Added.
* platform/mac/fast/regions/overflow-in-variable-width-regions-expected.png: Added.
* platform/mac/fast/regions/overflow-in-variable-width-regions-expected.txt: Added.
* platform/mac/fast/regions/overflow-rtl-in-variable-width-regions-expected.png: Added.
* platform/mac/fast/regions/overflow-rtl-in-variable-width-regions-expected.txt: Added.
* platform/mac/fast/regions/percentage-margins-mixed-ltr-dominant-regions-expected.png: Added.
* platform/mac/fast/regions/percentage-margins-mixed-ltr-dominant-regions-expected.txt: Added.
* platform/mac/fast/regions/percentage-margins-mixed-rtl-dominant-regions-expected.png: Added.
* platform/mac/fast/regions/percentage-margins-mixed-rtl-dominant-regions-expected.txt: Added.
* platform/mac/fast/regions/percentage-margins-rtl-variable-width-regions-expected.png: Added.
* platform/mac/fast/regions/percentage-margins-rtl-variable-width-regions-expected.txt: Added.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96423 268f45cc-cd09-0410-ab3c-d52691b4dbfc

REGRESSION(r82611) InlineBox has 33 bits of bitset, causing alignment issues and extra memory use.
https://bugs.webkit.org/show_bug.cgi?id=64914

Patch by Andreas Kling <kling@webkit.org> on 2011-09-30
Reviewed by Antti Koivisto.

Remove InlineBox::prevOnLineExists() and its two accompanying bitfields
since nobody is using them anymore. nextOnLineExists() is still used by
GTK+ accessibility code.

Also added a compile-time assertion to guard against future bloating of
the InlineBox class.

* rendering/InlineBox.cpp:
(WebCore::SameSizeAsInlineBox::~SameSizeAsInlineBox):
* rendering/InlineBox.h:
(WebCore::InlineBox::InlineBox):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96422 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Qt] Build fix: Qt::escape is deprecated in Qt5
https://bugs.webkit.org/show_bug.cgi?id=69162

Use QString::toHtmlEscaped in the Qt5 case.

Patch by Pierre Rossi <pierre.rossi@gmail.com> on 2011-09-30
Reviewed by Andreas Kling.

Source/JavaScriptCore:

* JavaScriptCore.pri:
* wtf/qt/UtilsQt.h: Added.
(escapeHtml):
* wtf/wtf.pri:

Source/WebCore:

No new tests needed.

* WebCore.pro: adjust the include path accordingly
in the v8 case.

Source/WebKit/qt:

* Api/qwebpage.cpp:
(QWebPage::javaScriptAlert):
(QWebPage::javaScriptConfirm):
(QWebPage::javaScriptPrompt):
* WebCoreSupport/ChromeClientQt.cpp:
(WebCore::ChromeClientQt::setToolTip):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96421 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Update baselines after trac.webkit.org/changeset/96404.

* platform/chromium-cg-mac/fast/text/atsui-multiple-renderers-expected.png: Added.
* platform/chromium-cg-mac/fast/text/complex-synthetic-bold-space-width-expected.png: Added.
* platform/chromium-linux/fast/text/complex-synthetic-bold-space-width-expected.png: Added.
* platform/chromium-mac/fast/text/atsui-multiple-renderers-expected.png:
* platform/chromium-mac/fast/text/complex-synthetic-bold-space-width-expected.png: Added.
* platform/chromium-win/fast/text/complex-synthetic-bold-space-width-expected.png: Added.
* platform/chromium-win/fast/text/complex-synthetic-bold-space-width-expected.txt: Added.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96420 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fix bug in getHostCallReturnValue of DFG JIT on X86
https://bugs.webkit.org/show_bug.cgi?id=69133

Patch by Yuqiang Xian <yuqiang.xian@intel.com> on 2011-09-30
Reviewed by Gavin Barraclough.

We need to insert the additional argument in the stack slot before
return address instead of simply pushing it afterwards.
Also getHostCallReturnValue* should be attributed as stdcall
to make the stack cleaned up by the callee.

* dfg/DFGOperations.cpp:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96419 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Qt] wtf header files are unknown to Qt Creator
https://bugs.webkit.org/show_bug.cgi?id=69158

Adding the HEADERS variable in wtf.pri so that
the header files can be accessed easily.

Patch by Pierre Rossi <pierre.rossi@gmail.com> on 2011-09-30
Reviewed by Andreas Kling.

* wtf/wtf.pri:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96418 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Qt] Prevent qDebug() output from DRT and WTR unless --verbose

For DRT we didn't install the message handler early enough to
catch output while constructing the QApplication. For WTR we
didn't even install a message handler.

Since the UI process will forward any output from the web process
we set an environment variable QT_WEBKIT_SUPPRESS_WEB_PROCESS_OUTPUT
in WTR before the web process is started. This is picked up by the
web process which installs its own message handler.

The environment variable can be overriden on the command line if you
want to see output from the web process, or you can pass --verbose to
WTR to see output from both processes.

https://bugs.webkit.org/show_bug.cgi?id=69132

Reviewed by Andreas Kling.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96417 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Qt] Make sure WTR sizes the window and item correctly

Revision 96345 changed the logic for how the view and
window was created, but missed a vital part, setting
the size.

We now use a QSGView for the window, that has a simple
item as its root object that is always resized to fit
within the window. The webview is then parented to the
root object and set to anchors.fill: parent. That way
any window geometry changes will propagate to the web
view.

https://bugs.webkit.org/show_bug.cgi?id=69134

Reviewed by Andreas Kling.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96416 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Merge some more of DFGSpeculativeJIT 32_64/64
https://bugs.webkit.org/show_bug.cgi?id=69164

Reviewed by Oliver Hunt.

* dfg/DFGJITCodeGenerator.h:
* dfg/DFGJITCodeGenerator32_64.cpp:
* dfg/DFGJITCodeGenerator64.cpp:
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compilePeepHoleBranch):
* dfg/DFGSpeculativeJIT.h:
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::fillSpeculateIntInternal):
(JSC::DFG::SpeculativeJIT::compare):
(JSC::DFG::SpeculativeJIT::compileValueAdd):
(JSC::DFG::SpeculativeJIT::compileLogicalNot):
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compare):
(JSC::DFG::SpeculativeJIT::compileValueAdd):
(JSC::DFG::SpeculativeJIT::compileLogicalNot):
(JSC::DFG::SpeculativeJIT::compile):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96415 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[EFL] Only save the current viewport in PixelDumpSupportEfl.
https://bugs.webkit.org/show_bug.cgi?id=68450

Patch by Raphael Kubo da Costa <kubo@profusion.mobi> on 2011-09-30
Reviewed by Antonio Gomes.

The previous implementation saved the whole page contents in the final
PNG, however the baselines and expectations in the tree only want the
current viewport (with a scrollbar, if needed), so we now only save the
currently visible contents.

* DumpRenderTree/efl/PixelDumpSupportEfl.cpp:
(createBitmapContextFromWebView): Use
ewk_frame_visible_content_geometry_get instead of
ewk_frame_contents_size_get.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96414 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Remove didCompleteRubberBandForMainFrame UIClient callback
https://bugs.webkit.org/show_bug.cgi?id=69161

Reviewed by Adam Roben.

* UIProcess/API/C/WKPage.h:
* UIProcess/WebPageProxy.cpp:
* UIProcess/WebPageProxy.h:
* UIProcess/WebPageProxy.messages.in:
* UIProcess/WebUIClient.cpp:
* UIProcess/WebUIClient.h:
* WebProcess/WebCoreSupport/WebChromeClient.cpp:
(WebKit::WebChromeClient::didCompleteRubberBandForFrame):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96413 268f45cc-cd09-0410-ab3c-d52691b4dbfc

webkit->chromium DEPS roll 102910->103471
https://bugs.webkit.org/show_bug.cgi?id=69160

Patch by Elliot Poger <epoger@google.com> on 2011-09-30
Reviewed by Adam Barth.

* DEPS:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96412 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[EFL] Support and keep track of extra windows (views).
https://bugs.webkit.org/show_bug.cgi?id=68454

Patch by Raphael Kubo da Costa <kubo@profusion.mobi> on 2011-09-30
Reviewed by Antonio Gomes.

So far, window.open and related JavaScript code would not work, as
DRTView did not implementation
Ewk_View_Smart_Class::window_{open,close}. The default implementation
in ewk just returns the current window/view, which broke any
layouttest which relied on window.open.

We now properly create and remove these new views. Unfortunately, the
current way ewk works (and due to the fact that we cannot pass class
methods as function pointers to the EFL functions), DRTView needs to
know that DRTChrome exists and call `browser' directly in some points.

* DumpRenderTree/efl/DumpRenderTreeView.cpp:
(onWindowCreate):
(onWindowCloseDelayed):
(onWindowClose):
(drtViewTiledAdd):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96411 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Remove unused ReplaceSelectionCommand::copyStyleToChildren
https://bugs.webkit.org/show_bug.cgi?id=69153

Reviewed by Antonio Gomes.

Removed the function because it's no called anywhere.

* editing/ReplaceSelectionCommand.cpp:
* editing/ReplaceSelectionCommand.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96410 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fix EnabledAtRuntime support for constants.
https://bugs.webkit.org/show_bug.cgi?id=67311

Patch by Aaron Colwell <acolwell@chromium.org> on 2011-09-30
Reviewed by Adam Barth.

* bindings/scripts/CodeGeneratorV8.pm:
(GenerateImplementation):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96409 268f45cc-cd09-0410-ab3c-d52691b4dbfc

https://bugs.webkit.org/show_bug.cgi?id=69043

[CSS3 Regions] Blocks split across regions with variable width need to size
differently in each region.

This patch adds the capability for blocks to have variable width and positioning
when split across regions. It is very rudimentary and so far only operates on basic
normal flow RenderBlocks. Future patches will expand the support to cover other layout
constructs like flexible boxes and tables.

Reviewed by Sam Weinig.

Added new tests in fast/regions and updated existing results.

Source/WebCore:

* WebCore.xcodeproj/project.pbxproj:
* rendering/HitTestResult.cpp:
(WebCore::HitTestResult::HitTestResult):
(WebCore::HitTestResult::operator=):
* rendering/HitTestResult.h:
(WebCore::HitTestResult::region):
(WebCore::HitTestResult::setRegion):
The hit test result now holds the active RenderRegion so that hit testing can adjust
block widths and positions when hit testing their border boxes.

* rendering/PaintInfo.h:
(WebCore::PaintInfo::PaintInfo):
The paint info struct now holds the active RenderRegion so that painting can adjust
block widths and positions when painting borders and backgrounds and shadows.

* rendering/RenderBlock.cpp:
(WebCore::RenderBlock::layoutBlock):
If our width ever changes, we invalidate all of our cached RenderBoxRegionInfo in all
regions that we span.

(WebCore::RenderBlock::clearRenderBoxRegionInfo):
Called to clear out our cached region-specific information in all regions that we span.

(WebCore::RenderBlock::borderBoxRectInRegionAtPosition):
Returns the borderBoxRect for the region at the specified vertical offset. This rect
can be both shifted horizontally and have a different width from our original border rect.

(WebCore::RenderBlock::logicalLeftOffsetForContent):
(WebCore::RenderBlock::logicalRightOffsetForContent):
Modified to call borderBoxRectInRegionAtPosition so that lines will fit inside
the content rect of the border box rect for the specific region.

* rendering/RenderBlock.h:
(WebCore::RenderBlock::logicalRightOffsetForContent):
Modified to call logicalLeftOffsetForContent since it's cleaner.

* rendering/RenderBox.cpp:
(WebCore::RenderBox::borderBoxRectInRegion):
Called to compute the border box rect in a specific region. The result is cached in a HashMap
in the region itself so that subsequent lookups are fast.

(WebCore::RenderBox::nodeAtPoint):
Modified to use the border box rect in the active region when hit testing the backgrounds of boxes.

(WebCore::RenderBox::paintBoxDecorations):
Modified to use the border box rect in the active region when painting the backgrounds of boxes.

(WebCore::RenderBox::containingBlockLogicalWidthForContent):
(WebCore::RenderBox::containingBlockLogicalWidthForContentInRegion):
(WebCore::RenderBox::computeLogicalWidth):
(WebCore::RenderBox::computeLogicalWidthInRegion):
Helpers for computing logical widths and margins in a specific region. The result is then cached
in a HashMap in the region.

(WebCore::RenderBox::renderBoxRegionInfo):
The function for obtaining the region-specific information for a given box.

* rendering/RenderBox.h:
(WebCore::RenderBox::borderBoxRectInRegion):
New function for returning the border box rect in a given region.

* rendering/RenderBoxModelObject.cpp:
(WebCore::RenderBoxModelObject::paintFillLayerExtended):
Make sure mask painting honors the active region.

* rendering/RenderBoxRegionInfo.h: Added.
(WebCore::RenderBoxRegionInfo::RenderBoxRegionInfo):
(WebCore::RenderBoxRegionInfo::logicalLeft):
(WebCore::RenderBoxRegionInfo::logicalWidth):
(WebCore::RenderBoxRegionInfo::containingBlockChainIsShifted):
New class held by RenderRegions that caches box-specific info for a given region. This will eventually
expand to include custom styles as well, but for now it is limited to a new logical left and a new
logical width, along with a bit for optimizing accumulated shifting when painting/hit testing to avoid
too much groveling up the containing block chain.

* rendering/RenderFlowThread.cpp:
(WebCore::RenderFlowThread::RenderFlowThread):
Removed the region fitting optimization, since eventually everyone is going to care, and there's no reason
to limit it with the results now being cached.

(WebCore::RenderFlowThread::layout):
All box-specific region information is always cleared whenever the regions are invalidated.

(WebCore::RenderFlowThread::computeLogicalWidth):
Modified to set up RenderBoxRegionInfo for the flow thread in all regions.

(WebCore::RenderFlowThread::paintIntoRegion):
(WebCore::RenderFlowThread::hitTestRegion):
Modified to take the region instead of the region's rectangle so that painting and hit testing of the flow
thread layer tree can properly receive the active region.

(WebCore::RenderFlowThread::removeRenderBoxRegionInfo):
Helper for removing a box's information from all regions. For now it grovels through every region, so eventually
we may want to have a cache of the start/end regions for a given box somewhere.

* rendering/RenderFlowThread.h:
Removal of the region fitting stuff.

* rendering/RenderLayer.cpp:
(WebCore::RenderLayer::paint):
(WebCore::RenderLayer::paintOverlayScrollbars):
(WebCore::RenderLayer::paintLayer):
(WebCore::RenderLayer::paintList):
(WebCore::RenderLayer::paintPaginatedChildLayer):
(WebCore::RenderLayer::paintChildLayerIntoColumns):
* rendering/RenderLayer.h:
* rendering/RenderLayerBacking.cpp:
(WebCore::RenderLayerBacking::paintIntoLayer):
Modified to pass the current region down through painting functions so that it is known at paint time.

* rendering/RenderObject.cpp:
(WebCore::RenderObject::RenderObject):
Added a new bit to RenderObjects, inRenderFlowThread(), so that it is quick to determine whether or not an
object needs flow thread special casing.

(WebCore::RenderObject::enclosingRenderFlowThread):
Modified enclosingRenderFlowThread to be able to quickly return 0 if the object is not in a flow thread.

(WebCore::RenderObject::containerForRepaint):
Same.

* rendering/RenderObject.h:
(WebCore::RenderObject::setParent):
setParent now updates inRenderFlowThread() state.

(WebCore::RenderObject::inRenderFlowThread):
(WebCore::RenderObject::setInRenderFlowThread):
Adding the new bit.

* rendering/RenderObjectChildList.cpp:
(WebCore::RenderObjectChildList::removeChildNode):
Make sure when an object is removed from a RenderFlowThread that it deletes its box-specific information
in all regions.

* rendering/RenderRegion.cpp:
(WebCore::RenderRegion::~RenderRegion):
(WebCore::RenderRegion::paintReplaced):
(WebCore::RenderRegion::nodeAtPoint):
(WebCore::RenderRegion::renderBoxRegionInfo):
(WebCore::RenderRegion::setRenderBoxRegionInfo):
(WebCore::RenderRegion::removeRenderBoxRegionInfo):
(WebCore::RenderRegion::deleteAllRenderBoxRegionInfo):
(WebCore::RenderRegion::matchesRenderFlowThreadLogicalWidth):
* rendering/RenderRegion.h:
Added the new HashMap for holding box-specific region information. Also added an additional optimization to
check if a specific region matches the overall width of the RenderFlowThread. If it does, we don't need to
cache box-specific information for that region.

* rendering/RenderReplica.cpp:
(WebCore::RenderReplica::paint):
* rendering/RenderScrollbarPart.cpp:
(WebCore::RenderScrollbarPart::paintIntoRect):
* rendering/svg/SVGImageBufferTools.cpp:
(WebCore::SVGImageBufferTools::renderSubtreeToImageBuffer):
Modified the PaintInfo construction to include the region argument. It's not optional so that callers will
have to consider it (since masks illustrated this is necessary).

LayoutTests:

* fast/regions/percentage-margins-variable-width-regions.html: Added.
* platform/mac/fast/regions/percentage-margins-variable-width-regions-expected.png: Added.
* platform/mac/fast/regions/percentage-margins-variable-width-regions-expected.txt: Added.
* platform/mac/fast/regions/webkit-flow-double-pagination-float-push-expected.png:
* platform/mac/fast/regions/webkit-flow-float-pushed-to-last-region-expected.png:
* platform/mac/fast/regions/webkit-flow-float-unable-to-push-expected.png:
* platform/mac/fast/regions/webkit-flow-floats-inside-regions-bounds-expected.png:
* platform/mac/fast/regions/webkit-flow-inlines-dynamic-expected.png:
* platform/mac/fast/regions/webkit-flow-inlines-inside-regions-bounds-expected.png:
* platform/mac/fast/regions/webkit-flow-inlines-inside-regions-bounds-vertical-expected.png:
* platform/mac/fast/regions/webkit-flow-inlines-inside-regions-bounds-vertical-rl-expected.png:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96408 268f45cc-cd09-0410-ab3c-d52691b4dbfc

feBlend uses a table of function pointers which reduces inlineability inside the main loop
https://bugs.webkit.org/show_bug.cgi?id=69154
<rdar://problem/10215221>

Reviewed by Darin Adler.

Don't use a table of function pointers inside the feBlend inner loop, instead
use switch and inline functions, bringing a 20% performance gain across the
board to feBlend.

No new tests, minor performance improvement.

* platform/graphics/filters/FEBlend.cpp:
(WebCore::normal):
(WebCore::multiply):
(WebCore::screen):
(WebCore::darken):
(WebCore::lighten):
(WebCore::FEBlend::apply):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96407 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Add getCallData to MethodTable in ClassInfo
https://bugs.webkit.org/show_bug.cgi?id=69024

Reviewed by Sam Weinig.

Source/JavaScriptCore:

* JavaScriptCore.exp:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:

Added the getCallData to the MethodTable in the ClassInfo struct.
* runtime/ClassInfo.h:

Source/WebCore:

No new tests.

* WebCore.exp.in:

Changed getCallData from private to protected to allow subclasses who
don't override getCallData themselves to reference it in their own
method tables when calling the CREATE_METHOD_TABLE macro.
* bridge/runtime_method.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96406 268f45cc-cd09-0410-ab3c-d52691b4dbfc

WebKitTestRunner needs layoutTestController.setAppCacheMaximumSize
https://bugs.webkit.org/show_bug.cgi?id=42698

Reviewed by Andreas Kling.

Source/WebKit2:

Implemented layoutTestController.setAppCacheMaximumSize and layoutTestController.clearApplicationCache,
and call clearApplicationCache when layoutTest start up.

* WebProcess/ApplicationCache/WebApplicationCacheManager.cpp:
(WebKit::WebApplicationCacheManager::setAppCacheMaximumSize):
* WebProcess/ApplicationCache/WebApplicationCacheManager.h:
* WebProcess/InjectedBundle/API/c/WKBundle.cpp:
(WKBundleClearApplicationCache):
(WKBundleSetAppCacheMaximumSize):
* WebProcess/InjectedBundle/API/c/WKBundlePrivate.h:
* WebProcess/InjectedBundle/InjectedBundle.cpp:
(WebKit::InjectedBundle::clearApplicationCache):
(WebKit::InjectedBundle::setAppCacheMaximumSize):
* WebProcess/InjectedBundle/InjectedBundle.h:

Tools:

Implemented layoutTestController.setAppCacheMaximumSize and layoutTestController.clearApplicationCache,
and call clearApplicationCache when layoutTest start up.

* WebKitTestRunner/InjectedBundle/Bindings/LayoutTestController.idl:
* WebKitTestRunner/InjectedBundle/InjectedBundle.cpp:
(WTR::InjectedBundle::beginTesting):
* WebKitTestRunner/InjectedBundle/LayoutTestController.cpp:
(WTR::LayoutTestController::clearAllApplicationCaches):
(WTR::LayoutTestController::setAppCacheMaximumSize):
* WebKitTestRunner/InjectedBundle/LayoutTestController.h:

LayoutTests:

Removed http/tests/appcache/max-size.html from skipped list, and also the "failed" expected file for http/tests/appcache/interrupted-update-expected.html

* platform/mac-wk2/http/tests/appcache: Removed.
* platform/mac-wk2/http/tests/appcache/interrupted-update-expected.txt: Removed.
* platform/wk2/Skipped:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96405 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Complex spaces with synthetic bold are too wide
https://bugs.webkit.org/show_bug.cgi?id=69033

Patch by Ned Holbrook <nholbrook@apple.com> on 2011-09-30
Reviewed by Dan Bernstein.

Source/WebCore:

Test: fast/text/complex-synthetic-bold-space-width.html

* platform/graphics/mac/ComplexTextController.cpp:
(WebCore::ComplexTextController::adjustGlyphsAndAdvances): Subtract synthetic bold offset from spaceWidth.
* platform/graphics/win/UniscribeController.cpp:
(WebCore::UniscribeController::shapeAndPlaceItem): Ditto.

LayoutTests:

* fast/text/complex-synthetic-bold-space-width.html: Added.
* platform/mac/fast/text/atsui-multiple-renderers-expected.png:
* platform/mac/fast/text/atsui-multiple-renderers-expected.txt:
* platform/mac/fast/text/complex-synthetic-bold-space-width-expected.png: Added.
* platform/mac/fast/text/complex-synthetic-bold-space-width-expected.txt: Added.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96404 268f45cc-cd09-0410-ab3c-d52691b4dbfc

LayoutTests: Rewrite fast/dom/HTMLInputElement/input-size-attribute.html
https://bugs.webkit.org/show_bug.cgi?id=69130

Patch by Antaryami Pandia <antaryami.pandia@motorola.com> on 2011-09-30
Reviewed by Alexey Proskuryakov.

* fast/dom/HTMLInputElement/input-size-attribute-expected.txt:
* fast/dom/HTMLInputElement/input-size-attribute.html:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96403 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Qt][WK2] Change WebKitTestRunner to use QMouseEvents instead of QGraphicsSceneMouseEvents
https://bugs.webkit.org/show_bug.cgi?id=69145

This is a follow up to https://bugs.webkit.org/show_bug.cgi?id=69105
Due to refactorings in Qt5, QMouseEvent now supports floating point coordinates
and is perfectly suitable for SceneGraph as well. GraphicsScene events
will not be available in QtGui anymore in future. Therefore we also have to change
QtWebKit to use QMouseEvents instead of QGraphicsSceneMouseEvents.

Patch by Zeno Albisser <zeno.albisser@nokia.com> on 2011-09-30
Reviewed by Chang Shu.

* WebKitTestRunner/EventSenderProxy.h:
* WebKitTestRunner/qt/EventSenderProxyQt.cpp:
(WTR::EventSenderProxy::createMouseEvent):
(WTR::EventSenderProxy::mouseDown):
(WTR::EventSenderProxy::mouseUp):
(WTR::EventSenderProxy::mouseMoveTo):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96402 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Replace textInputController.doCommand with eventSender.keyDown in password-echo tests
https://bugs.webkit.org/show_bug.cgi?id=69143

textInputController.doCommand is not supported very well across the platforms. Replace it
with eventSender.keyDown would make the tests pass on more platforms and potentially WK2.

Patch by Chang Shu <cshu@webkit.org> on 2011-09-30
Reviewed by Antonio Gomes.

* editing/input/resources/password-echo.js:
(init):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96401 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[UNIX] Add a method to get information for an already loaded plugin module
https://bugs.webkit.org/show_bug.cgi?id=69139

Reviewed by Martin Robinson.

* Shared/Plugins/Netscape/NetscapePluginModule.h: Add pluginInfo().
* Shared/Plugins/Netscape/x11/NetscapePluginModuleX11.cpp:
(WebKit::NetscapePluginModule::pluginInfo): Private method to get
the plugin information from the module, it must be called when the
module has been already initialized.
(WebKit::NetscapePluginModule::getPluginInfo): Use pluginInfo() to
get information of the plugin module for the given path.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96400 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Qt][WK2] Event delivery in QDesktopWebView is broken after merge of Qt5 refactor
https://bugs.webkit.org/show_bug.cgi?id=69105

This changes QDesktopWebView and QDesktopWebPageProxy to use QMouseEvent instead
of QGraphicsSceneMouseEvent.
Further it removes an endless loop introduced by the QSGItem::event() function
delivering inputMethodEvents but no other events.

Patch by Zeno Albisser <zeno.albisser@nokia.com> on 2011-09-30
Reviewed by Andreas Kling.

* Shared/NativeWebMouseEvent.h:
(WebKit::NativeWebMouseEvent::nativeEvent):
* Shared/NativeWebWheelEvent.h:
(WebKit::NativeWebWheelEvent::nativeEvent):
* Shared/qt/NativeWebMouseEventQt.cpp:
(WebKit::NativeWebMouseEvent::NativeWebMouseEvent):
* Shared/qt/NativeWebWheelEventQt.cpp:
(WebKit::NativeWebWheelEvent::NativeWebWheelEvent):
* Shared/qt/WebEventFactoryQt.cpp:
(WebKit::mouseButtonForEvent):
(WebKit::webEventTypeForEvent):
(WebKit::WebEventFactory::createWebMouseEvent):
(WebKit::WebEventFactory::createWebWheelEvent):
* Shared/qt/WebEventFactoryQt.h:
* UIProcess/API/qt/qdesktopwebview.cpp:
(QDesktopWebView::mousePressEvent):
(QDesktopWebView::mouseMoveEvent):
(QDesktopWebView::mouseReleaseEvent):
(QDesktopWebView::mouseDoubleClickEvent):
(QDesktopWebView::wheelEvent):
(QDesktopWebView::event):
* UIProcess/API/qt/qdesktopwebview.h:
* UIProcess/qt/qdesktopwebpageproxy.cpp:
(QDesktopWebPageProxy::handleEvent):
(QDesktopWebPageProxy::handleMouseMoveEvent):
(QDesktopWebPageProxy::handleMousePressEvent):
(QDesktopWebPageProxy::handleMouseReleaseEvent):
(QDesktopWebPageProxy::handleMouseDoubleClickEvent):
(QDesktopWebPageProxy::handleWheelEvent):
(QDesktopWebPageProxy::handleHoverMoveEvent):
* UIProcess/qt/qdesktopwebpageproxy.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96399 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Web Inspector: [chromium] expose inspector protocol version to the embedder.
https://bugs.webkit.org/show_bug.cgi?id=69092

Source/WebCore:

This change also introduces major/minor inspector version components.

Reviewed by Yury Semikhatsky.

* CMakeLists.txt:
* CodeGenerators.pri:
* DerivedSources.make:
* GNUmakefile.am:
* GNUmakefile.list.am:
* WebCore.gyp/WebCore.gyp:
* WebCore.gypi:
* inspector/Inspector-0.1.json: Renamed from Source/WebCore/inspector/Inspector.draft-01.json.
* inspector/Inspector.json:
* inspector/generate-inspector-protocol-version: Renamed from Source/WebCore/inspector/validate-protocol-compatibility.

Source/WebKit/chromium:

Reviewed by Yury Semikhatsky.

* public/WebDevToolsAgent.h:
* src/WebDevToolsAgentImpl.cpp:
(WebKit::WebDevToolsAgent::inspectorProtocolVersion):
(WebKit::WebDevToolsAgent::supportsInspectorProtocolVersion):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96398 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Chromium] Layout test failures under Snow Leopard
https://bugs.webkit.org/show_bug.cgi?id=54322

Unreviewed test_expectations.txt update. acid3 seems to be IMAGE+TEXT, not IMAGE

* platform/chromium/test_expectations.txt:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96397 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Chromium] Layout Test fast/frames/frame-set-rotation-hit.html and fast/frames/frame-set-scaling-hit.html are failing
https://bugs.webkit.org/show_bug.cgi?id=69136

Unreviewed test_expectations.txt update

* platform/chromium/test_expectations.txt:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96396 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Versioning.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96395 268f45cc-cd09-0410-ab3c-d52691b4dbfc

https://bugs.webkit.org/show_bug.cgi?id=69106
Universal attribute selectors disable style sharing

Reviewed by Dave Hyatt.

Selectors of type [foo="bar"] ended up marking every element style with the affectedByAttributeSelectors bit
rendering style sharing inoperative. This happens on http://www.whatwg.org/specs/web-apps/current-work/ for example.

Instead we now mark style with affectedByUncommonAttributeSelectors bit only if an attribute selector actually
matches the element. Before sharing, we also check the current element against collected attribute rules.
We can share the style if neither element was affected.

This speeds up style matching and applying ~15% on full HTML5 spec (=~0.7s). Sharing percentage goes from 0% to ~30%.
Increased sharing should also save a substantial amount of memory.

* css/CSSSelector.h:
(WebCore::CSSSelector::isAttributeSelector):
* css/CSSStyleSelector.cpp:
(WebCore::RuleData::containsUncommonAttributeSelector):
(WebCore::collectSpecialRulesInDefaultStyle):
(WebCore::assertNoSiblingRulesInDefaultStyle):
(WebCore::CSSStyleSelector::CSSStyleSelector):
(WebCore::CSSStyleSelector::matchRules):
(WebCore::CSSStyleSelector::matchesRuleSet):
(WebCore::CSSStyleSelector::canShareStyleWithElement):
(WebCore::CSSStyleSelector::locateSharedStyle):
(WebCore::CSSStyleSelector::styleForElement):
(WebCore::selectorListContainsUncommonAttributeSelector):
(WebCore::isCommonAttributeSelectorAttribute):
(WebCore::containsUncommonAttributeSelector):
(WebCore::RuleData::RuleData):
(WebCore::collectFeaturesFromSelector):
(WebCore::collectFeaturesFromList):
* css/CSSStyleSelector.h:
* css/SelectorChecker.cpp:
(WebCore::SelectorChecker::checkOneSelector):
* rendering/style/RenderStyle.cpp:
(WebCore::RenderStyle::RenderStyle):
* rendering/style/RenderStyle.h:
(WebCore::InheritedFlags::affectedByUncommonAttributeSelectors):
(WebCore::InheritedFlags::setAffectedByUncommonAttributeSelectors):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96393 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[chromium] Add WebKit API for sending input events to the compositor thread
https://bugs.webkit.org/show_bug.cgi?id=69117

Reviewed by Darin Fisher.

Source/WebCore:

Convert to using an externally-provided thread instead of one managed by CCThreadProxy internally. Most changes
in these files are mechanically changing references from ccThread to s_ccThread to reflect the class-static
nature of this pointer, the rest are removing the old thread creation codepath.

These changes are covered by the CC* unit tests and by the gpu layout tests when run with the
use_threaded_compositing=1 gyp define.

* platform/graphics/chromium/cc/CCLayerTreeHost.cpp:
* platform/graphics/chromium/cc/CCLayerTreeHost.h:
* platform/graphics/chromium/cc/CCThreadProxy.cpp:
(WebCore::CCThreadProxy::setThread):
(WebCore::CCThreadProxy::CCThreadProxy):
(WebCore::CCThreadProxy::~CCThreadProxy):
(WebCore::CCThreadProxy::compositeAndReadback):
(WebCore::CCThreadProxy::finishAllRendering):
(WebCore::CCThreadProxy::initializeLayerRenderer):
(WebCore::CCThreadProxy::setNeedsCommit):
(WebCore::CCThreadProxy::setNeedsCommitAndRedraw):
(WebCore::CCThreadProxy::setNeedsRedraw):
(WebCore::CCThreadProxy::start):
(WebCore::CCThreadProxy::stop):
(WebCore::CCThreadProxy::beginFrameAndCommit):
(WebCore::CCThreadProxy::scheduleDrawTaskOnCCThread):
* platform/graphics/chromium/cc/CCThreadProxy.h:

Source/WebKit/chromium:

Adds a new WebKit interface to allow the embedder to pass input events directly to the compositor thread.
Exposes the new WebCompositor interface and associated client. These interfaces are slightly different in that
they can only be interacted with from the compositor thread.

Changes the initialization path for the compositor thread to using an externally allocated and managed thread
instead of one internal to WebCore so that the embedder can interact with the thread more easily. This also
simplifies the startup and shutdown paths.  The contract is that WebCompositor::setThread() must be called
before any compositor can be initialized and the passed-in thread must outlive all compositor instances.

Each WebCompositor instance is associated with a WebWidget.  There isn't any way to get a WebCompositor for a
given WebWidget in this patch, that will be done in a follow-up.

* WebKit.gyp:
* public/WebCompositor.h:
(WebKit::WebCompositor::~WebCompositor):
* public/WebCompositorClient.h:
(WebKit::WebCompositorClient::~WebCompositorClient):
* src/CCThreadImpl.cpp:
(WebKit::CCThreadImpl::create):
(WebKit::CCThreadImpl::CCThreadImpl):
* src/CCThreadImpl.h:
* src/WebCompositorImpl.cpp: Added.
(WebKit::WebCompositor::setThread):
(WebKit::WebCompositorImpl::WebCompositorImpl):
(WebKit::WebCompositorImpl::~WebCompositorImpl):
(WebKit::WebCompositorImpl::setClient):
(WebKit::WebCompositorImpl::handleInputEvent):
* src/WebCompositorImpl.h:
(WebKit::WebCompositorImpl::create):
* src/WebViewImpl.cpp:
* src/WebViewImpl.h:
* tests/CCLayerTreeHostTest.cpp:
(WTF::CCLayerTreeHostTest::CCLayerTreeHostTest):
* tests/CCThreadTest.cpp:
(WebKit::TEST):

Tools:

Adds support for creating and setting a compositor thread in DRT.

* DumpRenderTree/chromium/WebViewHost.cpp:
(WebViewHost::WebViewHost):
* DumpRenderTree/chromium/WebViewHost.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96392 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Add op_call/op_constructor support to JSVALUE32_64 DFG JIT
https://bugs.webkit.org/show_bug.cgi?id=69120

Patch by Yuqiang Xian <yuqiang.xian@intel.com> on 2011-09-29
Reviewed by Gavin Barraclough.

Improve the coverage of JSVALUE32_64 DFG JIT.

* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::parseBlock):
* dfg/DFGCapabilities.h:
(JSC::DFG::canCompileOpcode):
* dfg/DFGJITCodeGenerator.h:
(JSC::DFG::tagOfCallData):
(JSC::DFG::payloadOfCallData):
* dfg/DFGJITCodeGenerator32_64.cpp:
(JSC::DFG::JITCodeGenerator::emitCall):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96391 268f45cc-cd09-0410-ab3c-d52691b4dbfc

watchlist: Add cross-checks for WatchList once it is filled.
https://bugs.webkit.org/show_bug.cgi?id=68975

Reviewed by Eric Seidel.

* Scripts/webkitpy/common/watchlist/watchlist.py: Made the data members public
instead of having trivial getter and setters.
* Scripts/webkitpy/common/watchlist/watchlist_unittest.py: Fix the unit tests to
pass the validation checks.
* Scripts/webkitpy/common/watchlist/watchlistparser.py: Add validation checks
and fix a few style nits.
* Scripts/webkitpy/common/watchlist/watchlistparser_unittest.py: Add tests for the
validation checks.
* Scripts/webkitpy/common/watchlist/watchlistrule.py: Make definitions_to_match public.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96390 268f45cc-cd09-0410-ab3c-d52691b4dbfc

DFG JIT - register not unlocked after usage in ArithDiv
https://bugs.webkit.org/show_bug.cgi?id=69122

Patch by Yuqiang Xian <yuqiang.xian@intel.com> on 2011-09-29
Reviewed by Geoffrey Garen.

Some allocated register is not unlocked after the usage in ArithDiv.
Also there's a typo in "ENBALE_DFG_CONSISTENTCY_CHECK".

* dfg/DFGNode.h:
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96389 268f45cc-cd09-0410-ab3c-d52691b4dbfc

run-api-tests fails on chromium-win bot
https://bugs.webkit.org/show_bug.cgi?id=69121

On Windows running a perl script needs the interpreter.

Patch by Xianzhu Wang <wangxianzhu@chromium.org> on 2011-09-29
Reviewed by Adam Barth.

* Scripts/run-api-tests:
(buildTestTool):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96388 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Chromium] Layout Test platform/chromium/compositing/3d-corners.html is failing
https://bugs.webkit.org/show_bug.cgi?id=69124

Unreviewed test_expectations.txt update

* platform/chromium/test_expectations.txt:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96387 268f45cc-cd09-0410-ab3c-d52691b4dbfc

fast/text/line-initial-and-final-swashes.html fails on Chromium
https://bugs.webkit.org/show_bug.cgi?id=69103

Unreviewed test_expectations.txt update

* platform/chromium/test_expectations.txt:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96386 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fixed Hit testing on Framesets when Frameset is transformed
https://bugs.webkit.org/show_bug.cgi?id=67740

Reviewed by Simon Fraser.

Source/WebCore:

Tests: fast/frames/frame-set-rotation-hit.html
       fast/frames/frame-set-scaling-hit.html

* rendering/RenderFrameSet.cpp:
(WebCore::RenderFrameSet::userResize):

LayoutTests:

* fast/frames/frame-set-rotation-hit-expected.png: Added.
* fast/frames/frame-set-rotation-hit-expected.txt: Added.
* fast/frames/frame-set-rotation-hit.html: Added.
* fast/frames/frame-set-scaling-hit-expected.png: Added.
* fast/frames/frame-set-scaling-hit-expected.txt: Added.
* fast/frames/frame-set-scaling-hit.html: Added.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96385 268f45cc-cd09-0410-ab3c-d52691b4dbfc

fix gyp warnings in chromium for renamed files
https://bugs.webkit.org/show_bug.cgi?id=69116

Reviewed by Adam Barth.

* WebCore.gypi:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96384 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Initial implementation of WebInspector for WebKit2 GTK port.
https://bugs.webkit.org/show_bug.cgi?id=68235

Patch by Ravi Phaneendra Kasibhatla <ravi.kasibhatla@motorola.com> on 2011-09-29
Reviewed by Martin Robinson.

* UIProcess/WebInspectorProxy.cpp:
(WebKit::WebInspectorProxy::WebInspectorProxy): Initialize the native window handles of GTK.
* UIProcess/WebInspectorProxy.h: Add native window handle for inspector window.
* UIProcess/gtk/WebInspectorGtk.cpp:
(WebKit::inspectorFilesBasePath): Base folder for inspector resources.
(WebKit::inspectorWindowDestroyed): Invoked when inspector window is closed by user.
(WebKit::WebInspectorProxy::windowDestroyed): Resets the native handles to 0 when user closes the window.
(WebKit::WebInspectorProxy::platformCreateInspectorPage): Create the WebView for inspector.
(WebKit::WebInspectorProxy::platformOpen): Create the window for inspector.
(WebKit::WebInspectorProxy::platformDidClose): Close both WebView & window of inspector.
(WebKit::WebInspectorProxy::platformInspectedURLChanged): Update the title of inspector window.
(WebKit::WebInspectorProxy::inspectorPageURL): Return the absolute URL of inspector.html.
(WebKit::WebInspectorProxy::inspectorBaseURL): Return the location where inspector resources are present.
WebProcess is given read access to this folder only for loading file URLs by sandboxed extension.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96383 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[EFL] Use C++ type cast instead of C style type cast
https://bugs.webkit.org/show_bug.cgi?id=68321

Reviewed by Martin Robinson.

This is a second step in order to be more compliant with WebKit coding style.
The second step is to change C-type casting with C++-type casting in .cpp files.

In addition, remained efl style '*' pointer placements are moved to the right place as well.

* ewk/ewk_auth_soup.cpp:
* ewk/ewk_contextmenu.cpp:
* ewk/ewk_frame.cpp:
* ewk/ewk_history.cpp:
* ewk/ewk_view.cpp:
* ewk/ewk_view_single.cpp:
* ewk/ewk_view_tiled.cpp:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96382 268f45cc-cd09-0410-ab3c-d52691b4dbfc

De-virtualize JSCell::toObject
https://bugs.webkit.org/show_bug.cgi?id=68937

Reviewed by Darin Adler.

* JavaScriptCore.exp:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:

De-virtualized JSCell::toObject and changed its implementation to manually check the
cases for JSString and JSObject rather than leaving it up to the virtual method call.
* runtime/JSCell.cpp:
(JSC::JSCell::toObject):
* runtime/JSCell.h:

Removed JSNotAnObject::toObject because the case for JSObject works for it.
Also removed JSObject::toObject because it was essentially the identity function,
which is not necessary since toObject is no longer virtual.
* runtime/JSNotAnObject.cpp:
* runtime/JSNotAnObject.h:
* runtime/JSObject.cpp:
* runtime/JSObject.h:

De-virtualized JSObject::toObject and JSString::toObject.
* runtime/JSString.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96381 268f45cc-cd09-0410-ab3c-d52691b4dbfc

These tests can also time out.

* platform/chromium/test_expectations.txt:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96380 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Start refactoring DFGSpeculativeJIT
https://bugs.webkit.org/show_bug.cgi?id=69112

Reviewed by Oliver Hunt.

Again, move JSVALUE64 code into a DFJSpeculativeJIT64.cpp

* JavaScriptCore.xcodeproj/project.pbxproj:
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::ValueSource::dump):
(JSC::DFG::ValueRecovery::dump):
(JSC::DFG::OSRExit::OSRExit):
(JSC::DFG::OSRExit::dump):
(JSC::DFG::SpeculativeJIT::compilePeepHoleDoubleBranch):
(JSC::DFG::SpeculativeJIT::compilePeepHoleObjectEquality):
(JSC::DFG::SpeculativeJIT::compilePeepHoleIntegerBranch):
(JSC::DFG::SpeculativeJIT::compile):
(JSC::DFG::SpeculativeJIT::checkArgumentTypes):
(JSC::DFG::SpeculativeJIT::computeValueRecoveryFor):
* dfg/DFGSpeculativeJIT.h:
(JSC::DFG::SpeculativeJIT::emitAllocateJSFinalObject):
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compare):
* dfg/DFGSpeculativeJIT64.cpp: Copied from Source/JavaScriptCore/dfg/DFGSpeculativeJIT.cpp.
(JSC::DFG::SpeculativeJIT::fillSpeculateIntInternal):
(JSC::DFG::SpeculativeJIT::fillSpeculateCell):
(JSC::DFG::SpeculativeJIT::compile):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96379 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Freetype] Some text in Planet GNOME renders in the wrong place
https://bugs.webkit.org/show_bug.cgi?id=69099

Reviewed by Dirk Schulze.

Source/WebCore:

Test: platform/gtk/fonts/synthetic-oblique-positioning.html

Fold the oblique transform into the font matrix itself rather than transforming
the CTM of the context at render time. Not only does this fix the issue, it
prevents unnecessary work on every paint.

* platform/graphics/cairo/FontCairo.cpp:
No longer set the synthetic oblique transformation matrix on the context
when rendering the text. Instead it is now folded into the TM of the font
itself.
* platform/graphics/freetype/FontPlatformDataFreeType.cpp:
(WebCore::FontPlatformData::initializeWithFontFace): Fold the oblique transform into the scaled font.
* platform/graphics/win/FontPlatformDataCairoWin.cpp:
(WebCore::FontPlatformData::FontPlatformData): Ditto.

LayoutTests:

Add a test which exercises this issue. It seems the problem grows worse
as the y component of the text position increases, so position the text
about halfway down the page.

* platform/gtk/fonts/synthetic-oblique-positioning-expected.png: Added.
* platform/gtk/fonts/synthetic-oblique-positioning-expected.txt: Added.
* platform/gtk/fonts/synthetic-oblique-positioning.html: Added.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96378 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Refactor out trivially duplicated code in DFGJITCodeGenerator.
https://bugs.webkit.org/show_bug.cgi?id=69109

Reviewed by Oliver Hunt.

Some code is trivially redundant between DFGJITCodeGenerator.cpp & DFGJITCodeGenerator32_64.cpp

Basically move a JSVALUE64 specific code into a new DFGJITCodeGenerator64.cpp, leave common code
in DFGJITCodeGenerator.cpp, and remove copies from DFGJITCodeGenerator32_64.cpp.

For some function differences are trivial & make more sense to ifdef individually, and some
Operand methods make more sense left in DFGJITCodeGenerator.cpp alongside similar constructors.

* JavaScriptCore.xcodeproj/project.pbxproj:
* dfg/DFGJITCodeGenerator.cpp:
(JSC::DFG::JITCodeGenerator::isKnownNotNumber):
(JSC::DFG::JITCodeGenerator::isKnownBoolean):
(JSC::DFG::JITCodeGenerator::writeBarrier):
(JSC::DFG::JITCodeGenerator::dump):
(JSC::DFG::JITCodeGenerator::checkConsistency):
(JSC::DFG::GPRTemporary::GPRTemporary):
(JSC::DFG::FPRTemporary::FPRTemporary):
* dfg/DFGJITCodeGenerator32_64.cpp:
* dfg/DFGJITCodeGenerator64.cpp: Copied from Source/JavaScriptCore/dfg/DFGJITCodeGenerator.cpp.
* dfg/DFGJITCompiler.h:
(JSC::DFG::JITCompiler::branchIfNotCell):
* dfg/DFGJITCompilerInlineMethods.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96377 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Update image baselines after trac.webkit.org/changeset/96366.  These are progressions.

* platform/chromium-mac/fast/css/font-smoothing-expected.png:
* platform/chromium-mac/fast/repaint/text-emphasis-v-expected.png:
* platform/chromium-mac/fast/text/emphasis-avoid-ruby-expected.png:
* platform/chromium-mac/fast/text/emphasis-overlap-expected.png:
* platform/chromium-mac/fast/text/soft-hyphen-4-expected.png:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96376 268f45cc-cd09-0410-ab3c-d52691b4dbfc

DFG JIT should infer which uses of a variable are not aliased
https://bugs.webkit.org/show_bug.cgi?id=68593

Reviewed by Oliver Hunt.

This separates how a variable is stored (i.e. its virtual register)
from how it's predicted. Each variable now takes a
VariableAccessData as its operand, instead of the virtual register.
The VariableAccessData stores the operand and the prediction. If
multiple uses of a variable are aliased, their VariableAccessDatas
are unified.

This also adds tracking of which argument values are used. It
correctly observes that an argument value is not used, if the
argument is assigned to inside the function before being used.

This also adds tracking of which variables are live at the head of
a basic block, and separates that from a variable being live at the
tail.

Finally, this communicates to both OSR entry and OSR exit code how
a variable is predicted at a particular point in the code, rather
than just communicating how it was predicted in the entire code
block (since with this patch there is no longer the notion of a
variable having just one prediction for a code block).

* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.vcproj/WTF/WTF.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* bytecode/ActionablePrediction.h: Added.
(JSC::actionablePredictionFromPredictedType):
(JSC::valueObeysPrediction):
(JSC::actionablePredictionToString):
(JSC::ActionablePredictions::ActionablePredictions):
(JSC::ActionablePredictions::setArgument):
(JSC::ActionablePredictions::argument):
(JSC::ActionablePredictions::setVariable):
(JSC::ActionablePredictions::variable):
(JSC::ActionablePredictions::argumentUpperBound):
(JSC::ActionablePredictions::variableUpperBound):
(JSC::ActionablePredictions::pack):
(JSC::ActionablePredictions::packVector):
* bytecode/CodeBlock.h:
* bytecode/PredictionTracker.h:
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::newVariableAccessData):
(JSC::DFG::ByteCodeParser::getLocal):
(JSC::DFG::ByteCodeParser::setLocal):
(JSC::DFG::ByteCodeParser::getArgument):
(JSC::DFG::ByteCodeParser::setArgument):
(JSC::DFG::ByteCodeParser::parseBlock):
(JSC::DFG::ByteCodeParser::processPhiStack):
(JSC::DFG::ByteCodeParser::parse):
* dfg/DFGDriver.cpp:
(JSC::DFG::compile):
* dfg/DFGGraph.cpp:
(JSC::DFG::Graph::nameOfVariableAccessData):
(JSC::DFG::Graph::dump):
(JSC::DFG::Graph::predictArgumentTypes):
* dfg/DFGGraph.h:
(JSC::DFG::operandIsArgument):
(JSC::DFG::VariableRecord::setFirstTime):
(JSC::DFG::BasicBlock::BasicBlock):
(JSC::DFG::Graph::predict):
(JSC::DFG::Graph::getPrediction):
* dfg/DFGJITCompiler.h:
(JSC::DFG::JITCompiler::noticeOSREntry):
* dfg/DFGNode.h:
(JSC::DFG::Node::hasVariableAccessData):
(JSC::DFG::Node::hasLocal):
(JSC::DFG::Node::variableAccessData):
(JSC::DFG::Node::local):
* dfg/DFGOSREntry.cpp:
(JSC::DFG::prepareOSREntry):
* dfg/DFGOSREntry.h:
* dfg/DFGPropagator.cpp:
(JSC::DFG::Propagator::propagateNodePredictions):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::ValueSource::dump):
(JSC::DFG::OSRExit::OSRExit):
(JSC::DFG::SpeculativeJIT::compile):
(JSC::DFG::SpeculativeJIT::checkArgumentTypes):
(JSC::DFG::SpeculativeJIT::computeValueRecoveryFor):
* dfg/DFGSpeculativeJIT.h:
(JSC::DFG::ValueSource::ValueSource):
(JSC::DFG::ValueSource::forPrediction):
(JSC::DFG::ValueSource::isSet):
(JSC::DFG::ValueSource::kind):
(JSC::DFG::ValueSource::nodeIndex):
(JSC::DFG::ValueSource::nodeIndexFromKind):
(JSC::DFG::ValueSource::kindFromNodeIndex):
(JSC::DFG::SpeculativeJIT::isKnownArray):
(JSC::DFG::SpeculativeJIT::computeValueRecoveryFor):
(JSC::DFG::SpeculativeJIT::SpeculativeJIT):
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::OSRExit::OSRExit):
(JSC::DFG::SpeculativeJIT::compile):
(JSC::DFG::SpeculativeJIT::checkArgumentTypes):
(JSC::DFG::SpeculativeJIT::computeValueRecoveryFor):
* wtf/PackedIntVector.h: Added.
(WTF::PackedIntVector::PackedIntVector):
(WTF::PackedIntVector::operator=):
(WTF::PackedIntVector::size):
(WTF::PackedIntVector::ensureSize):
(WTF::PackedIntVector::resize):
(WTF::PackedIntVector::clearAll):
(WTF::PackedIntVector::get):
(WTF::PackedIntVector::set):
(WTF::PackedIntVector::mask):
* wtf/Platform.h:
* wtf/UnionFind.h: Added.
(WTF::UnionFind::UnionFind):
(WTF::UnionFind::find):
(WTF::UnionFind::unify):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96375 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Follow-up fix for <rdar://problem/10191243> Glyph variants (line final swashes) appear where they should not

Reviewed by Simon Fraser.

Source/WebCore:

* platform/graphics/cocoa/FontPlatformDataCocoa.mm:
(WebCore::FontPlatformData::ctFont): Check for Hoefler Text Black Italic as well.

LayoutTests:

Added Hoefler Text Black Italic to the test case.

* fast/text/line-initial-and-final-swashes.html:
* platform/mac/fast/text/line-initial-and-final-swashes-expected.png:
* platform/mac/fast/text/line-initial-and-final-swashes-expected.txt:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96374 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Build fix.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96373 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Add logic to collect dirty objects as roots
https://bugs.webkit.org/show_bug.cgi?id=69100

Reviewed by Geoff Garen.

This gives us the ability to walk all the MarkedBlocks in an
AllocationSpace and collect the dirty objects, and then use
them as GC roots.

* dfg/DFGJITCodeGenerator.cpp:
(JSC::DFG::JITCodeGenerator::markCellCard):
* dfg/DFGJITCodeGenerator32_64.cpp:
(JSC::DFG::JITCodeGenerator::markCellCard):
* heap/AllocationSpace.cpp:
   Tidy up the write barrier logic a bit
(JSC::MarkedBlock::gatherDirtyObjects):
(JSC::TakeIfDirty::returnValue):
(JSC::TakeIfDirty::TakeIfDirty):
(JSC::TakeIfDirty::operator()):
(JSC::AllocationSpace::gatherDirtyObjects):
* heap/AllocationSpace.h:
* heap/CardSet.h:
(JSC::::isCardMarked):
(JSC::::clearCard):
* heap/Heap.cpp:
(JSC::Heap::markRoots):
* heap/Heap.h:
(JSC::Heap::writeBarrier):
* heap/MarkStack.cpp:
(JSC::SlotVisitor::visitChildren):
* heap/MarkedBlock.h:
(JSC::MarkedBlock::setDirtyObject):
(JSC::MarkedBlock::addressOfCardFor):
* heap/SlotVisitor.h:
* jit/JITPropertyAccess.cpp:
(JSC::JIT::emitWriteBarrier):
   Tidy the write barrier a bit

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96372 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Attempt to make this tool work on Windows.

* Scripts/run-api-tests:
(buildTestTool):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96367 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Enable LCD text in Skia on Mac
https://bugs.webkit.org/show_bug.cgi?id=68734

Reviewed by Stephen White.

No new tests. Existing layout tests are generated
with LCD text disabled for pixel comparisons.

Duplicate the logic in FontMac.mm to pass settings
for antialiasing and smoothing. Also disable smoothing
for DumpRenderTree.

* platform/graphics/skia/FontSkia.cpp:
(WebCore::setupPaint):
(WebCore::Font::drawGlyphs):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96366 268f45cc-cd09-0410-ab3c-d52691b4dbfc

 Implement flick gesture in Chromium Gesture Recognizer
 https://bugs.webkit.org/show_bug.cgi?id=67930

 Reviewed by Dimitri Glazkov.

 Test: Source/WebKit/chromium/tests/InnerGestureRecognizerTest.cpp

 * platform/chromium/GestureRecognizerChromium.cpp:
 (WebCore::GestureRecognizerChromium::isOverMinFlickSpeed):
 (WebCore::GestureRecognizerChromium::appendScrollGestureEnd):
 (WebCore::GestureRecognizerChromium::updateValues):
 (WebCore::GestureRecognizerChromium::scrollEnd):
 * platform/chromium/GestureRecognizerChromium.h:

Patch by Varun Jain <varunjain@google.com> on 2011-09-29

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96365 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[chromium] canvas.toDataURL("image/jpeg"): use libjpeg-turbo data swizzle
https://bugs.webkit.org/show_bug.cgi?id=67402

Reviewed by Kenneth Russell.

libjpeg-turbo can read directly from the input data for the premultiplied BRGX
(SkBitmap) encoding case.

No change in behavior, covered by existing canvas 2d/3d tests
canvas/philip/tests/toDataURL.jpeg.alpha.html
fast/canvas/webgl/premultiplyalpha-test.html

* platform/image-encoders/skia/JPEGImageEncoder.cpp:
(WebCore::encodePixels): If JCS_EXTENSIONS is defined (libjpeg-turbo), swizzle
the input BRGX pixels directly within libjpeg-turbo.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96364 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Unreviewed windows build fix.

* assembler/MacroAssemblerCodeRef.h:
* dfg/DFGOperations.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96363 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[EFL] API to get plain text of the frame
https://bugs.webkit.org/show_bug.cgi?id=67114

Patch by Raphael Kubo da Costa <kubo@profusion.mobi> on 2011-09-29
Reviewed by Ryosuke Niwa.

Add a function to return a frame's contents converted to plain text,
as already exposed by other ports and needed by DRT when dumping a
frame's contents.

* ewk/ewk_frame.cpp:
(ewk_frame_plain_text_get):
* ewk/ewk_frame.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96362 268f45cc-cd09-0410-ab3c-d52691b4dbfc

This test can also time out.

* platform/chromium/test_expectations.txt:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96361 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Qt] Try to fix the minial build

* QtTestBrowser/launcherwindow.cpp:
* QtTestBrowser/webpage.cpp:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96360 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Delete bogus expectation.

* platform/chromium-cg-mac/svg/text/selection-background-color-expected.txt: Removed.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96359 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Attempt to update these image results again.

* platform/chromium-cg-mac-leopard/http/tests/misc: Added.
* platform/chromium-cg-mac-leopard/http/tests/misc/acid3-expected.png: Added.
* platform/chromium-cg-mac-leopard/http/tests/misc/acid3-expected.txt: Added.
* platform/chromium-cg-mac/http/tests/misc/acid3-expected.png:
* platform/chromium-cg-mac/http/tests/misc/acid3-expected.txt:
* platform/chromium-mac/http/tests/misc/acid3-expected.png:
* platform/chromium-mac/http/tests/misc/acid3-expected.txt: Removed.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96358 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fixenate the build.

* bindings/v8/V8DOMWindowShell.cpp:
(WebCore::V8DOMWindowShell::namedItemAdded):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96357 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Mark test as expected to fail.

* platform/chromium/test_expectations.txt:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96356 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Qt][WK2] Skip all failing tests (10605!) after refactored Qt5.

Rubber-stamped by Tor Arne Vestbø.

* platform/qt-wk2/Skipped:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96355 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Structure transitions involving many (> 64) properties sometimes cause structure corruption
https://bugs.webkit.org/show_bug.cgi?id=69102

Reviewed by Darin Adler.

Made m_offset an int instead of a signed char. Changed the code to ensure that transitions
don't lead to the dictionary kind being forgotten.

* runtime/Structure.cpp:
(JSC::Structure::Structure):
* runtime/Structure.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96354 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Remove direct reads to m_firstNodeInserted and m_lastLeafInserted in ReplaceSelectionCommand
https://bugs.webkit.org/show_bug.cgi?id=69023

Reviewed by Enrica Casucci.

Converted to removeUnrenderedTextNodesAtEnds to use InsertedNodes instead of m_firstNodeInserted
and m_lastLeafInserted. Extracted the logic to update nodes as InsertedNodes::willRemoveNode
(old one was renamed to InsertedNodes::willRemoveNodePreservingChildren).

Also extracted shouldPerformSmartReplace and addSpacesForSmartReplace out of doApply,
and rewrote the logic to obtain endNode and startNode using startOfInsertedContent and
endOfInsertedContent instead of m_firstNodeInserted and m_lastLeafInserted.

Finally, replaced the nullity checks of m_firstNodeInserted and m_lastLeafInserted in
completeHTMLReplacement by nullity checks of start and end positions.

* dom/Node.cpp:
(WebCore::Node::traversePreviousSibling): Added.
* dom/Node.h:
* editing/ReplaceSelectionCommand.cpp:
(WebCore::ReplaceSelectionCommand::InsertedNodes::willRemoveNodePreservingChildren): Renamed from
willRemoveNode.
(WebCore::ReplaceSelectionCommand::InsertedNodes::willRemoveNode): Extracted from
removeUnrenderedTextNodesAtEnds.
(WebCore::ReplaceSelectionCommand::removeRedundantStylesAndKeepStyleSpanInline):
(WebCore::ReplaceSelectionCommand::removeUnrenderedTextNodesAtEnds):
(WebCore::ReplaceSelectionCommand::positionAtEndOfInsertedContent): Made const.
(WebCore::ReplaceSelectionCommand::positionAtStartOfInsertedContent): Made const.
(WebCore::ReplaceSelectionCommand::handleStyleSpans): Takes firstNodeInserted instead of directly
accessing m_firstNodeInserted.
(WebCore::ReplaceSelectionCommand::doApply):
(WebCore::ReplaceSelectionCommand::shouldPerformSmartReplace):
(WebCore::ReplaceSelectionCommand::addSpacesForSmartReplace):
(WebCore::ReplaceSelectionCommand::completeHTMLReplacement):
* editing/ReplaceSelectionCommand.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96353 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Shrink HTMLAnchorElement on 32-bit.
https://bugs.webkit.org/show_bug.cgi?id=69094

Patch by Andreas Kling <kling@webkit.org> on 2011-09-29
Reviewed by Antti Koivisto.

* html/HTMLAnchorElement.h: Pack members into a bitfield.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96352 268f45cc-cd09-0410-ab3c-d52691b4dbfc

<rdar://problem/10191243> Glyph variants (line final swashes) appear where they should not

Reviewed by Simon Fraser.

Source/WebCore:

Test: fast/text/line-initial-and-final-swashes.html

Hoefler Text Italic enables line-initial and -final swashes by default, so disable them. This
change targets only this known-bad font rather than all fonts, because at least one font (Khmer MN)
incorrectly claims to have the line-initial feature enabled, but disabling it actually does
something different and bad (breaking some combining marks).

* platform/graphics/cocoa/FontPlatformDataCocoa.mm:
(WebCore::createFeatureSettingDictionary): Added this helper function.
(WebCore::cascadeToLastResortFontDescriptor): Deployed WTF_ARRAY_LENGTH().
(WebCore::cascadeToLastResortAndDisableSwashesFontDescriptor): Added. Returns a font descriptor
that, in addition to having a cascade list consisting of the last resort font, also has feature
settings to disable line-initial and line-final swashes.
(WebCore::FontPlatformData::ctFont): For Hoefler Text Italic, use cascadeToLastResortAndDisableSwashesFontDescriptor().

LayoutTests:

* fast/text/line-initial-and-final-swashes.html: Added.
* platform/mac/fast/text/line-initial-and-final-swashes-expected.png: Added.
* platform/mac/fast/text/line-initial-and-final-swashes-expected.txt: Added.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96351 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Unreviewed: resetting baseline for code generator bindings

Unreviewed build fix

No new tests.

* bindings/scripts/test/JS/JSTestObj.cpp:
* bindings/scripts/test/JS/JSTestObj.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96350 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[V8, chromium] More logging to determine cause of a null
v8::Context in V8DOMWindowShell::namedItemAdded().

https://bugs.webkit.org/show_bug.cgi?id=68099

Reviewed by Adam Barth.

* bindings/v8/V8DOMWindowShell.cpp:
(WebCore::V8DOMWindowShell::namedItemAdded):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96349 268f45cc-cd09-0410-ab3c-d52691b4dbfc

https://bugs.webkit.org/show_bug.cgi?id=69040
ScrollbarThemeComposite requires a ScrollView to draw scroll corner

Reviewed by Simon Fraser.

No new tests. This will be needed later.

* platform/ScrollbarThemeComposite.cpp:
(WebCore::pageForScrollView):
(WebCore::ScrollbarThemeComposite::paintScrollCorner):
Use pageForScrollView() function which already existed in this file, used in another similar
location.

* platform/chromium/FramelessScrollView.cpp:
* platform/chromium/FramelessScrollView.h:
* platform/gtk/ScrollbarThemeGtk.cpp:
* platform/gtk/ScrollbarThemeGtk.h:
* platform/wx/ScrollbarThemeWx.cpp:
* platform/wx/ScrollbarThemeWx.h:
All these overrides are no longer needed, ScrollbarThemeComposite will do the right thing.

* platform/qt/ScrollbarThemeQt.cpp: (WebCore::ScrollbarThemeQt::paintScrollCorner):
Removed a special case for updatingControlTints phase. The same case is present in cross-platform
code now, and Qt doesn't have any custom subclasses of ScrollableArea or ScrollView to need
special handling.
This was added in r37377 without a bug or much ChangeLog explanation.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96348 268f45cc-cd09-0410-ab3c-d52691b4dbfc

DFG operation calls should be stdcall in Linux JSVALUE32_64 DFG JIT
https://bugs.webkit.org/show_bug.cgi?id=69058

Patch by Yuqiang Xian <yuqiang.xian@intel.com> on 2011-09-29
Reviewed by Gavin Barraclough.

Also Fixed the stdcall FunctionPtr constructors to make them compiled correctly on Linux

* assembler/MacroAssemblerCodeRef.h:
(JSC::FunctionPtr::FunctionPtr):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96347 268f45cc-cd09-0410-ab3c-d52691b4dbfc

De-virtualize JSCell::visitChildrenVirtual and remove all other visitChildrenVirtual methods
https://bugs.webkit.org/show_bug.cgi?id=68839

Reviewed by Geoffrey Garen.

Source/JavaScriptCore:

Removed the remaining visitChildrenVirtual methods.  This patch completes the process of
de-virtualizing visitChildren.

* API/JSCallbackObject.h:
* JavaScriptCore.exp:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
* debugger/DebuggerActivation.cpp:
* debugger/DebuggerActivation.h:
* runtime/Arguments.cpp:
* runtime/Arguments.h:
* runtime/Executable.cpp:
* runtime/Executable.h:
* runtime/GetterSetter.cpp:
* runtime/GetterSetter.h:
* runtime/JSActivation.cpp:
* runtime/JSActivation.h:
* runtime/JSArray.cpp:
* runtime/JSArray.h:
* runtime/JSFunction.cpp:
* runtime/JSFunction.h:
* runtime/JSGlobalObject.cpp:
* runtime/JSGlobalObject.h:
* runtime/JSObject.cpp:
* runtime/JSPropertyNameIterator.cpp:
* runtime/JSPropertyNameIterator.h:
* runtime/JSStaticScopeObject.cpp:
* runtime/JSStaticScopeObject.h:
* runtime/JSValue.h:
* runtime/NativeErrorConstructor.cpp:
* runtime/NativeErrorConstructor.h:
* runtime/RegExpObject.cpp:
* runtime/RegExpObject.h:
* runtime/Structure.cpp:
* runtime/Structure.h:
* runtime/StructureChain.cpp:
* runtime/StructureChain.h:

Inlined the method table access and call to the visitChildren function (the only call sites
to visitChildren are here).
* heap/MarkStack.cpp:
(JSC::SlotVisitor::visitChildren):

Changed the field name for the visitChildren function pointer to visitChildren (from
visitChildrenFunctionPtr) to make call sites less verbose.
* runtime/ClassInfo.h:

Discovered JSBoundFunction doesn't have its own ClassInfo (it used JSFunction's ClassInfo) but
overrides visitChildren, so it needs to have its own ClassInfo.
* runtime/JSBoundFunction.cpp:
* runtime/JSBoundFunction.h:

Had to move className up to make sure that the virtual destructor in JSObject wasn't
the first non-inline virtual method in JSObject (as per the comment in the file).
Also moved JSCell::visitChildrenVirtual into JSObject.h in order for it be inline-able
to mitigate the cost of an extra method call.

Also added a convenience accessor function methodTable() to JSCell to return the MethodTable to make
call sites more concise.  Implementation is inline in JSObject.h.
* runtime/JSObject.h:
(JSC::JSCell::methodTable):
* runtime/JSCell.h:

Added an out of line virtual destructor to JSWrapperObject and ScopeChainNode to
appease the vtable gods.  It refused to compile if there were no virtual methods in
both of these classes due to the presence of a weak vtable pointer.
* runtime/JSWrapperObject.cpp:
(JSC::JSWrapperObject::~JSWrapperObject):
* runtime/JSWrapperObject.h:
* runtime/ScopeChain.cpp:
(JSC::ScopeChainNode::~ScopeChainNode):
* runtime/ScopeChain.h:

Source/JavaScriptGlue:

Removed the remaining visitChildrenVirtual methods.  This patch completes the process of
de-virtualizing visitChildren.

* UserObjectImp.cpp:
* UserObjectImp.h:

Source/WebCore:

No new tests.

Removed the remaining visitChildrenVirtual methods.  This patch completes the process of
de-virtualizing visitChildren.

* WebCore.exp.in:
* bindings/js/JSAttrCustom.cpp:
* bindings/js/JSAudioContextCustom.cpp:
* bindings/js/JSCSSRuleCustom.cpp:
* bindings/js/JSCSSStyleDeclarationCustom.cpp:
* bindings/js/JSCanvasRenderingContextCustom.cpp:
* bindings/js/JSDOMGlobalObject.cpp:
(WebCore::JSDOMGlobalObject::~JSDOMGlobalObject):
(WebCore::JSDOMGlobalObject::finishCreation):
* bindings/js/JSDOMGlobalObject.h:
* bindings/js/JSDOMWindowCustom.cpp:
* bindings/js/JSDOMWindowShell.cpp:
* bindings/js/JSDOMWindowShell.h:
* bindings/js/JSJavaScriptAudioNodeCustom.cpp:
* bindings/js/JSMessageChannelCustom.cpp:
* bindings/js/JSMessagePortCustom.cpp:
* bindings/js/JSNamedNodeMapCustom.cpp:
* bindings/js/JSNodeCustom.cpp:
* bindings/js/JSNodeFilterCustom.cpp:
* bindings/js/JSNodeIteratorCustom.cpp:
* bindings/js/JSSVGElementInstanceCustom.cpp:
* bindings/js/JSSharedWorkerCustom.cpp:
* bindings/js/JSStyleSheetCustom.cpp:
* bindings/js/JSTreeWalkerCustom.cpp:
* bindings/js/JSWebGLRenderingContextCustom.cpp:
* bindings/js/JSWorkerContextCustom.cpp:
* bindings/js/JSXMLHttpRequestCustom.cpp:
* bindings/js/JSXPathResultCustom.cpp:
* bindings/scripts/CodeGeneratorJS.pm:
(GenerateHeader):
(GenerateImplementation):
* bridge/qt/qt_instance.cpp:
* bridge/qt/qt_runtime.cpp:
* bridge/qt/qt_runtime.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96346 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[Qt] Fix build against Qt5 after refactor of widgets out of QtGUi

QWidget and friends now live in the QtWidgets library. We update
includes in implementation files and private headers to us the
non-module-prefixed path, and leave the lookup for the include
path. For public headers we have to ifdef the includes as the
user might now have the modules we need in his QT config.

Finally, QSGCanvas is no longer a QWidget but a QWindow, so we
have to update our code and use windowHandle() for setting the
parent relationships.

https://bugs.webkit.org/show_bug.cgi?id=68687

Reviewed by Andreas Kling.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96345 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Bug fixes for CreateThis, NewObject and GetByOffset in JSVALUE32_64 DFG JIT
https://bugs.webkit.org/show_bug.cgi?id=69075

Patch by Yuqiang Xian <yuqiang.xian@intel.com> on 2011-09-29
Reviewed by Gavin Barraclough.

* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96344 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Unreviewed, rolling out r96340.
http://trac.webkit.org/changeset/96340
https://bugs.webkit.org/show_bug.cgi?id=69098

Caused 10 tests to crash in Debug (Requested by
abarth|gardener on #webkit).

Patch by Sheriff Bot <webkit.review.bot@gmail.com> on 2011-09-29

Source/WebCore:

* accessibility/AXObjectCache.cpp:
(WebCore::AXObjectCache::AXObjectCache):
(WebCore::AXObjectCache::~AXObjectCache):
(WebCore::AXObjectCache::remove):
(WebCore::AXObjectCache::childrenChanged):
* accessibility/AXObjectCache.h:
* accessibility/AccessibilityMenuList.cpp:
(WebCore::AccessibilityMenuList::childrenChanged):
* accessibility/AccessibilityMenuList.h:
* accessibility/AccessibilityMenuListPopup.cpp:
(WebCore::AccessibilityMenuListPopup::childrenChanged):
* accessibility/AccessibilityMenuListPopup.h:
* accessibility/AccessibilityObject.h:
(WebCore::AccessibilityObject::childrenChanged):
* accessibility/AccessibilityRenderObject.cpp:
(WebCore::startOfContinuations):
(WebCore::AccessibilityRenderObject::updateAccessibilityRole):
(WebCore::AccessibilityRenderObject::childrenChanged):
* accessibility/AccessibilityRenderObject.h:
* rendering/RenderObject.cpp:
(WebCore::RenderObject::willBeDestroyed):

LayoutTests:

* platform/mac/accessibility/aria-liveregion-without-element-access-expected.txt: Removed.
* platform/mac/accessibility/aria-liveregion-without-element-access.html: Removed.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96343 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Shrink FontFallbackList.
https://bugs.webkit.org/show_bug.cgi?id=69093

Patch by Andreas Kling <kling@webkit.org> on 2011-09-29
Reviewed by Antti Koivisto.

Reduce the size of FontFallbackList by one CPU word, decreasing memory
consumption by 300 kB (on 64-bit) when loading the full HTML5 spec.

* platform/graphics/FontCache.h:
* platform/graphics/FontCache.cpp:
(WebCore::FontCache::generation):

    Store the FontCache generation as an ushort rather than uint.

* platform/graphics/FontFallbackList.cpp:
(WebCore::FontFallbackList::FontFallbackList):
* platform/graphics/FontFallbackList.h:

    Pack enum and bool members in a bitfield.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96342 268f45cc-cd09-0410-ab3c-d52691b4dbfc

We should ignore the return value of GetRealNamedProperty
https://bugs.webkit.org/show_bug.cgi?id=68840

Reviewed by Nate Chapin.

Source/WebCore:

Instead of skipping the lookup process by using the result of real
named property, we should just indiciate that it wasn't handled by the
interceptor.

Test: http/tests/security/window-named-valueOf.html

* bindings/v8/custom/V8DOMWindowCustom.cpp:
(WebCore::V8DOMWindow::namedPropertyGetter):

LayoutTests:

* http/tests/security/window-named-valueOf-expected.txt: Added.
* http/tests/security/window-named-valueOf.html: Added.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96341 268f45cc-cd09-0410-ab3c-d52691b4dbfc

ARIA live regions don't trigger notifications for elements that aren't in the AX tree
https://bugs.webkit.org/show_bug.cgi?id=62289

Source/WebCore:

If an ARIA Live region udpates an element that is not in the AX object cache, then the Live region
notification is not sent. To fix this, the childrenChanged() method needs to actually create
the appropriate objects, but since that method gets called during a render tree update, we've learned
that it's generally not safe to create objects.

Instead a one shot timer can be fired that will update and create the necessary objects so that the
correct notification can be sent.

Reviewed by Darin Adler.

Test: platform/mac/accessibility/aria-liveregion-without-element-access.html

* accessibility/AXObjectCache.cpp:
(WebCore::AXObjectCache::AXObjectCache):
(WebCore::AXObjectCache::~AXObjectCache):
(WebCore::AXObjectCache::remove):
(WebCore::AXObjectCache::childrenUpdateTimerFired):
(WebCore::AXObjectCache::childrenChanged):
* accessibility/AXObjectCache.h:
* accessibility/AccessibilityMenuList.cpp:
(WebCore::AccessibilityMenuList::childrenChanged):
* accessibility/AccessibilityMenuList.h:
* accessibility/AccessibilityMenuListPopup.cpp:
(WebCore::AccessibilityMenuListPopup::childrenChanged):
* accessibility/AccessibilityMenuListPopup.h:
* accessibility/AccessibilityObject.h:
(WebCore::AccessibilityObject::childrenChanged):
* accessibility/AccessibilityRenderObject.cpp:
(WebCore::startOfContinuations):
(WebCore::AccessibilityRenderObject::updateAccessibilityRole):
(WebCore::AccessibilityRenderObject::childrenChanged):
* accessibility/AccessibilityRenderObject.h:
* rendering/RenderObject.cpp:
(WebCore::RenderObject::willBeDestroyed):

LayoutTests:

Reviewed by Darin Adler.

* platform/mac/accessibility/aria-liveregion-without-element-access-expected.txt: Added.
* platform/mac/accessibility/aria-liveregion-without-element-access.html: Added.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96340 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[GTK] Dragging a selection does not produce a drag image
https://bugs.webkit.org/show_bug.cgi?id=69064

Patch by Martin Robinson <mrobinson@igalia.com> on 2011-09-29
Reviewed by Andreas Kling.

Moved Chromium's generic implementation of Frame::nodeImage and
Frame::dragImageForSeletion to Frame.cpp and simply excluded via
#ifdefs platforms that do not have their own implementation. Removed
all empty implementations of these functions and Qt's which was
functionally identical.

No new tests. Drag-and-drop drag image functionality changes
are incredibly difficult to test in a non-manual way.

* CMakeListsEfl.txt: Delist FrameEfl.cpp.
* GNUmakefile.list.am: Delist FrameGtk.cpp.
* WebCore.gypi: Delist removed Frame*.cpp.
* WebCore.pro: Delist FrameQt.cpp.
* page/Frame.cpp: Moved FrameChromium implementation here.
(WebCore::ScopedFramePaintingState::ScopedFramePaintingState):
(WebCore::ScopedFramePaintingState::~ScopedFramePaintingState):
(WebCore::Frame::nodeImage):
(WebCore::Frame::dragImageForSelection):
* page/chromium/FrameChromium.cpp: Removed.
* page/efl/FrameEfl.cpp: Removed.
* page/gtk/FrameGtk.cpp: Removed.
* page/qt/FrameQt.cpp: Removed.
* page/wx/FrameWx.cpp: Removed.
* platform/gtk/ClipboardGtk.cpp:
(WebCore::ClipboardGtk::createDragImage): Call nodeImage to get
the drag image for a node.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96339 268f45cc-cd09-0410-ab3c-d52691b4dbfc

[chromium skia] PNGImageEncoder: hoist constants out of the encoding loop
https://bugs.webkit.org/show_bug.cgi?id=68988

Reviewed by Kenneth Russell.

No new tests.  Covered by existing canvas 2d and 3d tests.
canvas/philip/tests/toDataURL.png.*.html
fast/canvas/toDataURL-alpha.html
fast/canvas/webgl/premultiplyalpha-test.html

* platform/image-encoders/skia/PNGImageEncoder.cpp:
(WebCore::encodePixels):  Move constant out of the encoding loop.
(WebCore::PNGImageEncoder::encode):  Consistency & style: call encodePixels()
just like we do in the JPEG encoder.

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96338 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Add unit test for CCLayerSorter
https://bugs.webkit.org/show_bug.cgi?id=68622

Source/WebCore:

Minor refactoring for testability:
- Made pointInTriangle public.
- Added LayerShape to decouple LayerIntersector and GraphNode.
- Added a public wrapper function for LayerIntersector.

Patch by Iain Merrick <husky@google.com> on 2011-09-29
Reviewed by James Robinson.

* platform/graphics/chromium/cc/CCLayerSorter.cpp:
(WebCore::CCLayerSorter::pointInTriangle):
(WebCore::CCLayerSorter::calculateZDiff):
(WebCore::CCLayerSorter::LayerIntersector::LayerIntersector):
(WebCore::CCLayerSorter::LayerIntersector::go):
(WebCore::CCLayerSorter::LayerIntersector::checkZDiff):
(WebCore::CCLayerSorter::LayerIntersector::layerZFromProjectedPoint):
(WebCore::CCLayerSorter::checkOverlap):
(WebCore::CCLayerSorter::LayerShape::LayerShape):
(WebCore::CCLayerSorter::createGraphNodes):
* platform/graphics/chromium/cc/CCLayerSorter.h:
(WebCore::CCLayerSorter::LayerShape::LayerShape):
(WebCore::CCLayerSorter::GraphNode::GraphNode):

Source/WebKit/chromium:

Patch by Iain Merrick <husky@google.com> on 2011-09-29
Reviewed by James Robinson.

* WebKit.gypi:
* tests/CCLayerSorterTest.cpp: Added.
(WebCore::TEST):

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@96337 268f45cc-cd09-0410-ab3c-d52691b4dbfc