Mar 15, 2015:

11:51 PM Changeset in webkit [181525] by benjamin@webkit.org
  • 14 edits
    6 adds in trunk

CSS: fix the case-insensitive matching of the attribute selectors Begin, End and Hyphen

Reviewed by Brent Fulgham.


Fix attribute matching with:
-Begin: [a=b].
-End: [a$=b].
-Hyphen: [a|=b].

Tests: fast/selectors/attribute-endswith-value-matching-is-ascii-case-insensitive.html


  • css/SelectorChecker.cpp:

I forgot to change CSSSelector::Exact in my last patch.
The tests could not catch that since we use the CSS JIT almost everywhere.

  • cssjit/SelectorCompiler.cpp:



Add the necessary infrastructure to test startsWith() and endsWith() with
ASCII case-insentive comparisons.

  • wtf/text/AtomicString.h:


  • wtf/text/StringCommon.h:

I moved the low level equal() code from StringImpl to StringCommon
since it is used by both StringImpl and StringView.

Ideally we should drop the "Common" part of the name but StringView
wants this inline for some reason. I prefered keeping the current behavior
since I don't know how StringView's matching performance was evaluated.

Make all that code shared between StringView and Stringimpl.

  • wtf/text/StringImpl.cpp:

(WTF::stringImplContentEqual): Deleted.

  • wtf/text/StringImpl.h:

(WTF::loadUnaligned): Deleted.
(WTF::equal): Deleted.

  • wtf/text/StringView.cpp:


  • wtf/text/StringView.h:

Since those are new, we can safely make them out-of-line and
evaluate the inlining impact as needed.

  • wtf/text/WTFString.h:



  • TestWebKitAPI/Tests/WTF/StringImpl.cpp:



  • fast/selectors/attribute-endswith-value-matching-is-ascii-case-insensitive-expected.txt: Added.
  • fast/selectors/attribute-endswith-value-matching-is-ascii-case-insensitive.html: Added.
  • fast/selectors/attribute-hyphen-value-matching-is-ascii-case-insensitive-expected.txt: Added.
  • fast/selectors/attribute-hyphen-value-matching-is-ascii-case-insensitive.html: Added.
  • fast/selectors/attribute-startswith-value-matching-is-ascii-case-insensitive-expected.txt: Added.
  • fast/selectors/attribute-startswith-value-matching-is-ascii-case-insensitive.html: Added.
11:28 PM Changeset in webkit [181524] by mitz@apple.com
  • 2 edits in trunk/Source/WebCore

Fixed the iOS build after r181522.

  • page/FrameView.cpp:


10:44 PM Changeset in webkit [181523] by aestes@apple.com
  • 5 edits in trunk/Source/WebCore

[Content Filtering] Adopt new NEFilterSource SPI

Reviewed by Dan Bernstein.

Teach NetworkExtensionContentFilter to use a new, alternate NEFilterSource SPI on platforms where it is available.

  • platform/ContentFilter.cpp:

(WebCore::ContentFilter::types): Renamed HAVE(NE_FILTER_SOURCE) to HAVE(NETWORK_EXTENSION).

  • platform/cocoa/NetworkExtensionContentFilter.h: Renamed member variables to remove redundancy, forward-declared NEFilterSourceStatus,

added a dispatch_semaphore member variable to avoid creating and destroying multiple semaphores, and made m_originalData a SharedBuffer.

  • platform/cocoa/NetworkExtensionContentFilter.mm:

(decisionInfoReplacementData): Returned the replacement data from a decision handler info dictionary.
(WebCore::createNEFilterSource): Created either an old-style or new-style NEFilterSource object.
(WebCore::NetworkExtensionContentFilter::NetworkExtensionContentFilter): Called receivedResponse:decisionHandler: when using the new SPI.
(WebCore::NetworkExtensionContentFilter::~NetworkExtensionContentFilter): Released the dispatch_semaphore.
(WebCore::NetworkExtensionContentFilter::addData): Appended the copied NSData to m_originalData, avoiding an additional copy previously
being made by NSMutableData. Used the new receivedData:decisionHandler: SPI when appropriate.
(WebCore::NetworkExtensionContentFilter::finishedAddingData): Used the new finishedLoadingWithDecisionHandler: SPI when appropriate.
(WebCore::NetworkExtensionContentFilter::needsMoreData): Changed m_neFilterSourceStatus to m_status.
(WebCore::NetworkExtensionContentFilter::didBlockData): Ditto.
(WebCore::NetworkExtensionContentFilter::getReplacementData): Returned the replacement data from NEFilterSource if the load was blocked.
Otherwise, returned the original data.
(WebCore::NetworkExtensionContentFilter::handleDecision): Added a helper to set m_status and m_replacementData, and to signal m_semaphore.

  • platform/spi/cocoa/NEFilterSourceSPI.h: Declared the new NEFilterSource SPI on platforms that support it.
10:34 PM Changeset in webkit [181522] by Brent Fulgham
  • 3 edits
    4 adds in trunk

Scroll snap points are not supported on iframe content


Tested by css3/scroll-snap/scroll-snap-iframe.html

Reviewed by Simon Fraser.

The scroll snap points were not being applied to the iframe contents because the code
that sets up the scroll snap point content is not called for iframes.

To correct this, we need to make sure the snap offsets are set during post-frame layout
for iframes. We also need to make sure (on Mac) that the scroll animator and timers are updated.

  • page/FrameView.cpp:

(WebCore::FrameView::performPostLayoutTasks): Call 'updateSnapOffsets' if the frame is not a
MainFrame. Also notify scroll animators they need to update their snap point settings.


Reviewed by Simon Fraser.

  • css3/scroll-snap/resources: Added.
  • css3/scroll-snap/resources/iframe-content.html: Added.
  • css3/scroll-snap/scroll-snap-iframe-expected.txt: Added.
  • css3/scroll-snap/scroll-snap-iframe.html: Added.
10:31 PM Changeset in webkit [181521] by Simon Fraser
  • 3 edits in trunk/Source/WebCore

Add the same is<RenderBox> test to KeyframeAnimation::computeExtentOfTransformAnimation()
that ImplicitAnimation::computeExtentOfTransformAnimation() has, and change the latter
to the more canonical is<RenderBox>() form.

Fixes an assertion in animations/animation-on-inline-crash.html

  • page/animation/ImplicitAnimation.cpp:


  • page/animation/KeyframeAnimation.cpp:


7:26 PM Changeset in webkit [181520] by Simon Fraser
  • 2 edits in trunk/Source/WebCore

And another fix. Thanks to Hunseop Jeong for the fix.

  • html/HTMLInputElement.cpp:


7:24 PM Changeset in webkit [181519] by Simon Fraser
  • 2 edits in trunk/Source/WebCore

Fix the touch-event build.

  • dom/Document.cpp:


5:40 PM Changeset in webkit [181518] by Nikita Vasilyev
  • 2 edits in trunk/Source/WebInspectorUI

Web Inspector: Console Scrolls Unexpectedly when Clicking inside un-selected Expanding Object

Don't scroll when selection change is triggered by clicking, e.g.
only scroll on arrow up and down key press events.

Reviewed by Joseph Pecoraro.

  • UserInterface/Views/LogContentView.js:


5:19 PM Changeset in webkit [181517] by Nikita Vasilyev
  • 10 edits
    2 moves in trunk/Source/WebInspectorUI

Web Inspector: Rename ConsoleMessage and ConsoleMessageImpl to LegacyConsoleMessage and LegacyConsoleMessageImpl respectively

As a first step of ConsoleMessage refactoring (https://bugs.webkit.org/show_bug.cgi?id=142599):

  • Rename WebInspector.ConsoleMessage class to WebInspector.LegacyConsoleMessage
  • Rename WebInspector.ConsoleMessageImpl class to WebInspector.LegacyConsoleMessageImpl
  • Rename ConsoleMessage.js file to LegacyConsoleMessage.js
  • Rename ConsoleMessageImpl file to LegacyConsoleMessageImpl.js

Reviewed by Joseph Pecoraro.

  • UserInterface/Controllers/JavaScriptLogViewController.js:


  • UserInterface/Controllers/LogManager.js:


  • UserInterface/Main.html:
  • UserInterface/Models/DefaultDashboard.js:


  • UserInterface/Views/ConsoleCommandResult.js:


  • UserInterface/Views/ConsoleGroup.js:


  • UserInterface/Views/LegacyConsoleMessage.js: Renamed from Source/WebInspectorUI/UserInterface/Views/ConsoleMessage.js.


  • UserInterface/Views/LegacyConsoleMessageImpl.js: Renamed from Source/WebInspectorUI/UserInterface/Views/ConsoleMessageImpl.js.

(WebInspector.LegacyConsoleMessageImpl.prototype.get message):
(WebInspector.LegacyConsoleMessageImpl.prototype.get formattedMessage):
(WebInspector.LegacyConsoleMessageImpl.prototype.get text):
(WebInspector.LegacyConsoleMessageImpl.prototype.get stackTrace):
(WebInspector.LegacyConsoleMessageImpl.prototype.get levelString):
(WebInspector.LegacyConsoleMessageImpl.prototype.get clipboardPrefixString):

  • UserInterface/Views/LogContentView.js:


  • WebInspectorUI.vcxproj/WebInspectorUI.vcxproj:
  • WebInspectorUI.vcxproj/WebInspectorUI.vcxproj.filters:
5:16 PM Changeset in webkit [181516] by commit-queue@webkit.org
  • 5 edits in trunk

Web Inspector: Better handle displaying -0

Patch by Joseph Pecoraro <Joseph Pecoraro> on 2015-03-15
Reviewed by Timothy Hatcher.


Modeled after a blink change:

Patch by <aandrey@chromium.org>
DevTools: DevTools: Show -0 for negative zero in console

  • inspector/InjectedScriptSource.js:

When creating a description string, or preview value string
for -0, be sure the string is "-0" and not "0".


Add tests for -0 in different places.
I don't understand yet why the test says "0" for
the property previews of "-0". Everything behaves
correctly, but I can't see to make the test show
the right value appear in the test. That is worth
investigating separately though.

  • inspector/model/remote-object-expected.txt:
  • inspector/model/remote-object.html:
4:07 PM Changeset in webkit [181515] by Simon Fraser
  • 33 edits
    20 adds in trunk

Reduce the side-effects of animations turning off overlap testing

Reviewed by Dean Jackson.


When a layer is running a transition or animation of the transform property,
we would simply disable overlap testing for later layers, which had the side-effect
of promoting lots of unrelated elements into layers temporarily.

Fix by maintaining overlap, but computing an overlap extent that takes the animation
into account.

Rotations are currently treated as full rotations. If an extent for the overlap is
hard to compute (e.g. 3d transforms, or matrix animations with a rotation component),
then we fall back to the current behavior.

Tests: compositing/layer-creation/mismatched-rotated-transform-animation-overlap.html


  • page/animation/AnimationBase.cpp:

(WebCore::AnimationBase::computeTransformedExtentViaTransformList): When we have matched
transform lists, we can map a rectangle through the various operations. Transform-origin
is used to shift the origin of the box first, and then unshift after. If we encounter
a rotation, for now assume it's a full rotation (a future patch could tighten this up).
(WebCore::AnimationBase::computeTransformedExtentViaMatrix): If we're using matrix
interpolation, we have to decompose the matrix to see if there's any rotation component,
and, if there is, fall back to current behavior.

  • page/animation/AnimationBase.h:
  • page/animation/AnimationController.cpp:


  • page/animation/AnimationController.h:
  • page/animation/AnimationControllerPrivate.h:
  • page/animation/CompositeAnimation.cpp:

(WebCore::CompositeAnimation::computeExtentOfTransformAnimation): Ask active keyframe
animations and transitions to compute the bounds extent.

  • page/animation/CompositeAnimation.h:
  • page/animation/ImplicitAnimation.cpp:

(WebCore::ImplicitAnimation::computeExtentOfTransformAnimation): Compute the extent
of the start and end transforms, and union them.

  • page/animation/ImplicitAnimation.h:
  • page/animation/KeyframeAnimation.cpp:

(WebCore::KeyframeAnimation::getAnimatedStyle): Some nullptr goodness.
(WebCore::KeyframeAnimation::computeExtentOfTransformAnimation): Compute an extent
for each keyframe, and take their union.

  • page/animation/KeyframeAnimation.h:
  • platform/graphics/GeometryUtilities.cpp:

(WebCore::euclidianDistance): Use Pythagoras to compute a distance.
(WebCore::boundsOfRotatingRect): Given a rect whose location is relative
to the rotation origin, compute a bounds for the rotated rect by computing
the furthest corner from the origin, and sweeping out a circle.

  • platform/graphics/GeometryUtilities.h:
  • platform/graphics/transforms/Matrix3DTransformOperation.h:
  • platform/graphics/transforms/MatrixTransformOperation.h:
  • platform/graphics/transforms/PerspectiveTransformOperation.h:
  • platform/graphics/transforms/RotateTransformOperation.h:
  • platform/graphics/transforms/ScaleTransformOperation.h:
  • platform/graphics/transforms/SkewTransformOperation.h:
  • platform/graphics/transforms/TransformOperation.h:


  • platform/graphics/transforms/TransformOperations.cpp:

(WebCore::TransformOperations::affectedByTransformOrigin): Ask all the operations if
they are affected by transform-origin.
(WebCore::TransformOperations::blendByMatchingOperations): nullptr.

  • platform/graphics/transforms/TransformOperations.h:
  • rendering/RenderBox.cpp:

(WebCore::RenderBox::pushMappingToContainer): Comment fix. Only take transforms into account
if the geometry map says so (which is most of the time).

  • rendering/RenderGeometryMap.cpp:

(WebCore::RenderGeometryMap::mapToContainer): RenderLayerCompositor is now using the
geometry map in a way that is incompatible with this assertion; it deliberately ignores
transforms sometimes, so we can't easily verify that the mapping matches mapping through
(WebCore::RenderGeometryMap::pushMappingsToAncestor): Save and restore the UseTransforms

  • rendering/RenderGeometryMap.h:
  • rendering/RenderLayer.cpp:

(WebCore::RenderLayer::boundingBox): Whitespace.
(WebCore::RenderLayer::getOverlapBoundsIncludingChildrenAccountingForTransformAnimations): Helper
function to get the bounds of a layer, including descendants, when a transform animation is running.

  • rendering/RenderLayer.h:
  • rendering/RenderLayerCompositor.cpp:

(WebCore::RenderLayerCompositor::CompositingState::CompositingState): Add a ancestorHasTransformAnimation
flag to detect nested animated transforms.
(WebCore::RenderLayerCompositor::OverlapExtent::knownToBeHaveExtentUncertainty): This returns true when
the layer is animating transform, and the transition/animation is such that we can't easily compute the
bounds of the animation.
(WebCore::RenderLayerCompositor::computeExtent): const RenderLayer&.
Compute the animated bounds if there's a transform animation running.
(WebCore::RenderLayerCompositor::addToOverlapMap): const RenderLayer&
(WebCore::RenderLayerCompositor::addToOverlapMapRecursive): const RenderLayer&
Delay the call to pushMappingsToAncestor() until knowing if there's a transform animation running, and
if there is, push the mapping while ignoring transforms (since the transform is implicitly taken into account
for overlap via the computed animated bounds).
If this layer is running a transform animation, set the childState.ancestorHasTransformAnimation flag so
that descendants will know (nested transform animations fall back to current behavior).
The if (.... && isRunningAcceleratedTransformAnimation()) is what previously caused us to turn off overlap
testing in the face of animations. That now only happens if we were unable to easily compute the animation bounds.
(WebCore::RenderLayerCompositor::isRunningTransformAnimation): This previously tested whether an accelerated animation
was running, but that's timing sensitive; AnimationController can start the transform animation, but it's not yet
considered accelerated until we get an async callback from GraphicsLayer, yet this code needed to know if the
animation was running.
Since transform animations are always accelerated, we can just test for a running transform animation.
(WebCore::RenderLayerCompositor::isRunningAcceleratedTransformAnimation): Deleted.

  • rendering/RenderLayerCompositor.h:
  • rendering/style/RenderStyle.cpp:

(WebCore::requireTransformOrigin): Some FIXME comments.


These test overlay an animated element with a grid of position: relative squares.
These reveal the overlap area by selectively getting composited.

  • compositing/layer-creation/animation-overlap-with-children-expected.txt:
  • compositing/layer-creation/mismatched-rotated-transform-animation-overlap-expected.txt: Added.
  • compositing/layer-creation/mismatched-rotated-transform-animation-overlap.html: Added.
  • compositing/layer-creation/mismatched-rotated-transform-transition-overlap-expected.txt: Added.
  • compositing/layer-creation/mismatched-rotated-transform-transition-overlap.html: Added.
  • compositing/layer-creation/mismatched-transform-transition-overlap-expected.txt: Added.
  • compositing/layer-creation/mismatched-transform-transition-overlap.html: Added.
  • compositing/layer-creation/multiple-keyframes-animation-overlap-expected.txt: Added.
  • compositing/layer-creation/multiple-keyframes-animation-overlap.html: Added.
  • compositing/layer-creation/scale-rotation-animation-overlap-expected.txt: Added.
  • compositing/layer-creation/scale-rotation-animation-overlap.html: Added.
  • compositing/layer-creation/scale-rotation-transition-overlap-expected.txt: Added.
  • compositing/layer-creation/scale-rotation-transition-overlap.html: Added.
  • compositing/layer-creation/translate-animation-overlap-expected.txt: Added.
  • compositing/layer-creation/translate-animation-overlap.html: Added.
  • compositing/layer-creation/translate-scale-animation-overlap-expected.txt: Added.
  • compositing/layer-creation/translate-scale-animation-overlap.html: Added.
  • compositing/layer-creation/translate-scale-transition-overlap-expected.txt: Added.
  • compositing/layer-creation/translate-scale-transition-overlap.html: Added.
  • compositing/layer-creation/translate-transition-overlap-expected.txt: Added.
  • compositing/layer-creation/translate-transition-overlap.html: Added.
3:47 PM Changeset in webkit [181514] by Simon Fraser
  • 11 edits in trunk/Source

Clean up related to wheelEvent names

Reviewed by Anders Carlsson.

Add EventNames::isWheelEventType() and use it in places that test for the
two wheel event names.


  • dom/Document.cpp:


  • dom/Document.h: No need for exports (I grepped). Pass the Node*, which

will be used in a later patch.

  • dom/EventNames.h:


  • dom/Node.cpp:


  • html/shadow/MediaControlsApple.cpp:


  • page/DOMWindow.cpp:


  • page/ios/FrameIOS.mm:

(WebCore::ancestorRespondingToScrollWheelEvents): Remove dead code.


  • WebProcess/Plugins/PluginView.cpp:


1:42 PM Changeset in webkit [181513] by Simon Fraser
  • 2 edits in trunk/Source/WebCore

Remove a redundant repaint when a layer becomes composited

Reviewed by Anders Carlsson.

RenderLayerCompositor::computeCompositingRequirements() doesn't need to call
repaintOnCompositingChange() when a layer is going to become composited,
because updateBacking() does exactly the same thing. I used an assertion
and ran the tests to ensure this wasn't a behavior change.

  • rendering/RenderLayerCompositor.cpp:


12:43 PM Changeset in webkit [181512] by benjamin@webkit.org
  • 19 edits
    3 adds in trunk

Change the exact attribute matching to be ASCII case-insensitive

Patch by Benjamin Poulain <bpoulain@apple.com> on 2015-03-15
Reviewed by Darin Adler.


In CSS, testing attribute values should be ASCII case-insensitive,
previously we were using full unicode case conversion.

Test: fast/selectors/attribute-exact-value-match-is-ascii-case-insensitive.html

  • css/CSSParser.cpp:

The CSS parser has its own fast version for ASCII case insensitive.
This code was using the general equalIgnoringASCIICase() which was causing name conflicts,
change that to the normal CSS parser version.

  • css/SelectorCheckerTestFunctions.h:

(WebCore::equalIgnoringASCIICase): Deleted.

  • cssjit/SelectorCompiler.cpp:



Add support for ASCII case insensitive comparisons to all the string

The new file StringCommon.h has the common algorithm to avoid repeating
the same code with different types.

  • WTF.vcxproj/WTF.vcxproj:
  • WTF.vcxproj/WTF.vcxproj.filters:
  • wtf/ASCIICType.h:


  • wtf/CMakeLists.txt:
  • wtf/text/AtomicString.h:


  • wtf/text/StringCommon.h: Added.


  • wtf/text/StringImpl.cpp:

(WTF::StringImpl::defaultWritingDirection): Deleted.
(WTF::StringImpl::adopt): Deleted.
(WTF::StringImpl::sizeInBytes): Deleted.
(WTF::putUTF8Triple): Deleted.
(WTF::StringImpl::utf8): Deleted.

  • wtf/text/StringImpl.h:

(WTF::StringImpl::isSubString): Deleted.
(WTF::find): Deleted.


  • TestWebKitAPI/Tests/WTF/StringImpl.cpp:
  • TestWebKitAPI/Tests/WTF/StringView.cpp:


  • fast/selectors/attribute-exact-value-match-is-ascii-case-insensitive-expected.txt: Added.
  • fast/selectors/attribute-exact-value-match-is-ascii-case-insensitive.html: Added.
12:30 PM Changeset in webkit [181511] by mitz@apple.com
  • 6 edits in trunk/Source/WebKit2

[iOS] Presenting a modal sheet on top of a WKWebView causes it to lose focused, active state

Reviewed by Anders Carlsson.

  • UIProcess/API/Cocoa/WKWebView.mm:

(-[WKWebView becomeFirstResponder]): Override to delegate first responder status to the
WKContentView if possible.
(-[WKWebView _retainActiveFocusedState]): New SPI that prevents view hierarchy changes from
affecting the active and focused state of the view. Increments a counter ivar and returns a
completion block (which callers must call when they’re done) that decrements it back.

  • UIProcess/API/Cocoa/WKWebViewInternal.h: Declared new _activeFocusedStateRetainCount ivar

with @package access.

  • UIProcess/API/Cocoa/WKWebViewPrivate.h: Declared new method.
  • UIProcess/ios/PageClientImplIOS.mm:

(WebKit::PageClientImpl::isViewWindowActive): Return true if active/focused state is to be
maintained despite not being visible.
(WebKit::PageClientImpl::isViewFocused): Return true if active/focused state is to be
maintained despite not being active.

  • UIProcess/ios/WKContentViewInteraction.mm:

(-[WKContentView resignFirstResponder]): Don’t blur the assisted node if active/focused
state is to be maintained.

12:13 PM Changeset in webkit [181510] by Brent Fulgham
  • 5 edits in trunk/Source/WebCore

scroll snap points do not properly account for zoomed pages

Reviewed by Anders Carlsson.

When a WebView is zoomed (such that it has a non-unity pageScaleFactor), we need to account for this
scaling value when selecting our correct scroll snap point target, as well as when specifying the
pixel location for our animation to target.

  • page/scrolling/mac/ScrollingTreeFrameScrollingNodeMac.h:
  • page/scrolling/mac/ScrollingTreeFrameScrollingNodeMac.mm:

(WebCore::ScrollingTreeFrameScrollingNodeMac::pageScaleFactor): Added new delegate method.

  • platform/cocoa/ScrollController.h:

(WebCore::ScrollControllerClient::pageScaleFactor): Added new default delegate.

  • platform/cocoa/ScrollController.mm:

(WebCore::ScrollController::beginScrollSnapAnimation): Calculate the correct scroll target
based on the page scale factor.

11:16 AM WebKitGTK/2.6.x edited by berto@igalia.com
10:52 AM Changeset in webkit [181509] by Brent Fulgham
  • 2 edits in trunk/LayoutTests

[Win] Document more debug assertions for later review.

  • platform/win/TestExpectations:
10:45 AM Changeset in webkit [181508] by Csaba Osztrogonác
  • 41 edits in trunk/Source/WebCore

Fix run-bindings-tests on the WinCairo bot

Reviewed by Alex Christensen.

  • bindings/scripts/test/JS/JSFloat64Array.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSFloat64Array.h: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestActiveDOMObject.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestActiveDOMObject.h: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestCallback.cpp: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestCallback.h: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestCustomNamedGetter.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestCustomNamedGetter.h: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestEventConstructor.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestEventConstructor.h: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestEventTarget.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestEventTarget.h: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestException.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestException.h: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestGenerateIsReachable.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestGenerateIsReachable.h: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestImplements.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestImplements.h: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestInterface.cpp: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestInterface.h: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestMediaQueryListListener.cpp: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestMediaQueryListListener.h: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestNamedConstructor.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestNamedConstructor.h: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestNondeterministic.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestNondeterministic.h: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestObj.cpp: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestObj.h: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestOverloadedConstructors.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestOverloadedConstructors.h: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestSerializedScriptValueInterface.cpp: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestSerializedScriptValueInterface.h: Modified property svn:eol-style.
  • bindings/scripts/test/JS/JSTestSupplemental.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestSupplemental.h: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestTypedefs.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSTestTypedefs.h: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSattribute.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSattribute.h: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSreadonly.cpp: Added property svn:eol-style.
  • bindings/scripts/test/JS/JSreadonly.h: Added property svn:eol-style.
10:00 AM Changeset in webkit [181507] by Darin Adler
  • 74 edits in trunk

More event handler improvements

Reviewed by Anders Carlsson.


These are the improvements:

  • Use EventHandler rather than EventListener as the the type for event handler attributes. This matches the HTML specification, and also makes sense, since EventListener means something else (and we use it to mean that!). Also renamed JSWindowEventListener to WindowEventHandler. Even though this only affects the JavaScript code generated, it's not really a JavaScript-specific flag.
  • Tweak formatting on addEventListener/removeEventListener/dispatchEvent in all the IDL files where the appear. This includes changing the spelling from "evt" to "event". Some day we should fix this so these functions only need to appear in EventTarget.idl.
  • Tweak formatting a bit on the IDL files we had to modify anyway.
  • Use [Conditional] more often and #if less often in IDL files.
  • Added a new [DocumentEventHandler] attribute for the selectionchange event. This involved adding new event handler attribute functions to JSEventListener.h for use by the JavaScript bindings.
  • Removed a little unused code from the JavaScript code bindings generator.
  • Improved the mechanism used by HTMLElement and SVGElement to share the list of content attributes that are event handlers so there is only one map rather than two. Made a similar mechanism so that HTMLBodyElement and HTMLFrameSetElement can share the list of window event handlers.
  • Followed the HTML specification by putting all the event handler support in the HTMLElement class rather than having event handlers apply only a the particular element that uses those events. We already did this for most of our event handlers, but we are now doing it for all of them.
  • Modules/battery/BatteryManager.idl: Use EventHandler instead of EventListener

as appropriate. Also reformatted addEventListener/removeEventListener/dispatchEvent.

  • Modules/encryptedmedia/MediaKeySession.idl: Ditto.
  • Modules/indexeddb/IDBDatabase.idl: Ditto.
  • Modules/indexeddb/IDBOpenDBRequest.idl: Ditto.
  • Modules/indexeddb/IDBRequest.idl: Ditto.
  • Modules/indexeddb/IDBTransaction.idl: Ditto.
  • Modules/mediastream/MediaStream.idl: Ditto.
  • Modules/mediastream/MediaStreamTrack.idl: Ditto.
  • Modules/mediastream/RTCDTMFSender.idl: Ditto.
  • Modules/mediastream/RTCDataChannel.idl: Ditto.
  • Modules/mediastream/RTCPeerConnection.idl: Ditto.
  • Modules/speech/SpeechSynthesisUtterance.idl: Ditto.
  • Modules/webaudio/AudioBufferSourceNode.idl: Ditto.
  • Modules/webaudio/AudioContext.idl: Ditto.
  • Modules/webaudio/OscillatorNode.idl: Ditto.
  • Modules/webaudio/ScriptProcessorNode.idl: Ditto.
  • Modules/websockets/WebSocket.idl: Ditto.
  • css/FontLoader.idl: Ditto.
  • dom/EventListener.idl: Ditto.
  • dom/EventTarget.idl: Ditto.
  • dom/MessagePort.idl: Ditto.
  • dom/Node.idl: Ditto.
  • dom/WebKitNamedFlow.idl: Ditto.
  • fileapi/FileReader.idl: Ditto.
  • html/MediaController.idl: Ditto.
  • html/track/AudioTrackList.idl: Ditto.
  • html/track/TextTrackCue.idl: Ditto.
  • html/track/TextTrackList.idl: Ditto.
  • html/track/VideoTrackList.idl: Ditto.
  • loader/appcache/DOMApplicationCache.idl: Ditto.
  • page/EventSource.idl: Ditto.
  • page/Performance.idl: Ditto.
  • workers/AbstractWorker.idl: Ditto.
  • workers/DedicatedWorkerGlobalScope.idl: Ditto.
  • workers/Worker.idl: Ditto.
  • workers/WorkerGlobalScope.idl: Ditto.
  • xml/XMLHttpRequest.idl: Ditto.
  • xml/XMLHttpRequestUpload.idl: Ditto.
  • Modules/notifications/Notification.idl: Did the above, but also used

[Conditional] instead of #if throughout.

  • html/track/TextTrack.idl: Ditto.
  • Modules/webaudio/AudioNode.idl: Tweaked paragraphing of this file.
  • bindings/js/JSEventListener.cpp:

(WebCore::windowEventHandlerAttribute): Renamed to take the word "forwarded"
out of this. More closely matches the terminology used in IDL files and the
HTML specification.
(WebCore::setWindowEventHandlerAttribute): Ditto.
(WebCore::documentEventHandlerAttribute): Added.
(WebCore::setDocumentEventHandlerAttribute): Added.

  • bindings/js/JSEventListener.h: Updated for above changes.
  • bindings/objc/PublicDOMInterfaces.h: Renamed argument from "evt" to "event".
  • bindings/scripts/CodeGeneratorGObject.pm:

(SkipAttribute): Skip attributes of type "EventHandler" rather than attributes
of type "EventListener".

  • bindings/scripts/CodeGeneratorObjC.pm:

(SkipAttribute): Ditto.

  • bindings/scripts/CodeGeneratorJS.pm:

(GenerateImplementation): Look for the type EventHandler instead of the type
EventListener for event handler attributes. Also added code to handle the new
DocumentEventHandler, and use the name WindowEventHandler instead of the name
JSWindowEventListener. Removed unneeded preflight check to see if we have
writable attributes; it was not doing us any good. (That caused a lot of code
to be un-indented and makes the diff hard to read.)

  • bindings/scripts/IDLAttributes.txt: Removed JSWindowEventListener, and added

WindowEventHandler and DocumentEventHandler.

  • bindings/scripts/test/TestObj.idl: Use the type EventHandler instead of the

type EventListener. The test output is unchanged, though.

  • dom/Document.idl: Got rid of the conditionals and merged all the event handler

attributes into a single sorted, unconditional list. Added some that were missing,
as detected by the event-handler-attributes.html test.

  • dom/Element.idl: Ditto. This includes moving attributes here that were formerly

only in certain element classes. Note that the script event handler attribute
support is still here, even though it should be in HTMLElement and SVGElement
instead. There's a FIXME about that, but no real urgency in fixing it.

  • html/HTMLAttributeNames.in: Added onmessage. Previously, the support for this

was from the script attribute only, not the content attribute.

  • html/HTMLBodyElement.cpp:

(WebCore::HTMLBodyElement::createWindowEventHandlerNameMap): Added. This
contains the list of all the window event handlers that can be set on a body
or frameset element and which are triggered by events on the window.
(WebCore::HTMLBodyElement::eventNameForWindowEventHandlerAttribute): Added.
This is the function to call to use the map above. Given an attribute, it
returns either null if it is not a window event handler attribute, or the
event type if it is one.
(WebCore::HTMLBodyElement::parseAttribute): Updated to use the new
functions above. Handling of these attributes is now unconditional, but
also we don't have to keep the nested if statements here up to date, just
the list of event handler attributes names in the map create function above.

  • html/HTMLBodyElement.h: Added public eventNameForWindowEventHandlerAttribute

and private createWindowEventHandlerNameMap functions.

  • html/HTMLBodyElement.idl: Updated to use WindowEventHandler, DocumentEventHandler,

and EventHandler. Also made everything unconditional. Also filled out the list here
to match the list handled as content attributes. All covered by the test.

  • html/HTMLElement.cpp:

(WebCore::HTMLElement::createEventHandlerNameMap): Added. Replaces the old
populate function. Changed the list of event handler attributes to be a bit more
complete, and to be entirely unconditional. Also refactored this function to use
a new populateEventHandlerNameMap helper, shared with HTMLBodyElement.
(WebCore::HTMLElement::populateEventHandlerNameMap): Added. Factors out the code
that both this class and HTMLBodyElement use to build event handler name maps.
(WebCore::HTMLElement::eventNameForEventHandlerAttribute): Added. This is the
function that call to use the map above. Given an attribute it returns either
null if it is not an event handler attribute, or the event type if it is one.
This is actually two functions. One is a protected function for use by both this
class and HTMLBodyElement so they can share things like the optimization to look
for the "on" prefix. The other is the public function that we actually use in
HTMLElement and SVGElement.
(WebCore::HTMLElement::editabilityFromContentEditableAttr): Tweaked and refactored
to use lineageOfType. Would have been even simpler if this took an element instead
of a node. Unrelated to the event handler changes.
(WebCore::HTMLElement::parseAttribute): Removed long-obsolete code that decided
whether to call through to the base class. The base class function is empty and
never needs to be called, and in any case there is no value in doing work to
decide whether to call through to an empty function. Changed the style of the
function to use early return instead of else. Worth considering whether we want
to return early or call through to base class in this family of functions. It's
more efficient to return early, but doesn't work well if both the derived class
and base class want to respond to changes to the same attribute. The new logic
for event handler attributes is more straightforward than the old, since the
eventNameForEventHandlerAttribute has the logic.
(WebCore::HTMLElement::textToFragment): Tweaked and refactored a bit, and also
changed to return a Ref since this never fails and needs to return null.

  • html/HTMLElement.h: Updated for above changes. This includes a template version

of populateEventHandlerNameMap that extracts the array size at compile time and
passes it to the non-template function that does the work.

  • html/HTMLFrameElementBase.cpp:

(WebCore::HTMLFrameElementBase::parseAttribute): Removed unneeded code to handle
event handler attributes handled by HTMLElement.

  • html/HTMLImageElement.cpp:

(WebCore::HTMLImageElement::parseAttribute): Ditto.

  • html/HTMLLinkElement.cpp:

(WebCore::HTMLLinkElement::parseAttribute): Ditto.

  • html/HTMLObjectElement.cpp:

(WebCore::HTMLObjectElement::parseAttribute): Ditto.

  • html/HTMLScriptElement.cpp:

(WebCore::HTMLScriptElement::parseAttribute): Ditto.

  • html/HTMLFrameSetElement.cpp:

(WebCore::HTMLFrameSetElement::parseAttribute): Changed function to early return
style, and added FIXMEs about the many problems in the attribute handling code.
Replaced all the code to handle window event handlers with a new bit of code that
calls HTMLBodyElement::eventNameForWindowEventHandlerAttribute.

  • html/HTMLFrameSetElement.idl: Changed to match the list of window event handlers

in HTMLBodyElement.idl, although I did not add the document event handler here.
As in the various other cases, having some extra event handlers does not seem to
do harm and this is covered by the event-handler-attributes.html test.

  • html/HTMLMarqueeElement.idl: Renamed EventListener to EventHandler in comment.
  • page/DOMWindow.idl: As with Element and Document, removed conditionals, and

filled out the list of event handlers so all the tests in
event-handler-attributes.html will pass.

  • svg/SVGElement.cpp:

(WebCore::SVGElement::parseAttribute): Changed code to the early return style,
and replaced the event handler attribute code with new much simpler code that
uses the new HTMLElement::eventNameForEventHandlerAttribute. Also changed the
way we call through to base classes. Just call through to the
SVGLangSpace::parseAttribute function unconditionally, and don't try to use
early return style to arbitrate among base classes. We should make this
simplification throughout the SVG code; there's no need for the complexity
that was there before just to cut down slightly on calls through to base
class parseAttribute functions.

  • svg/SVGSVGElement.cpp:

(WebCore::SVGSVGElement::parseAttribute): Changed some of this code to the
early return style and corrected some comments about the window event handler
attributes here. These could use some further testing and might later need to be
properly supported when the attributes are set in script, not just in content.

  • svg/SVGScriptElement.cpp:

(WebCore::SVGScriptElement::isSupportedAttribute): Deleted.
(WebCore::SVGScriptElement::parseAttribute): Changed this function to use the
early return style and also to call through to all three base classes. This is
a pattern we should follow elsewhere in SVG to simplify the code. There is no
need for the supportedAttributes sets like the one in this calass, and the code
is unnecessarily complex, perhaps in an attempt to optimize performance. I'm
pretty sure the old code was slower than this new code will be. No need for the
extra hash table lookup every time. Also removed handling of event handler
attribute which is taken care of by SVGElement now.
(WebCore::SVGScriptElement::svgAttributeChanged): Made similar changes for
the same reason as in parseAttribute. This function really needs a new name:
It's the same as parseAttribute, except it's also used when implementing
changes due to SVG animation.

  • svg/SVGScriptElement.h: Removed isSupportedAttribute.


  • fast/dom/event-handler-attributes-expected.txt: Updated to expect more passing tests,

and to accomodate some changes to what's tested and expected.

  • fast/dom/event-handler-attributes.html: Update test to cover more events since we

no longer compile the event handlers conditionally; it's OK to have some handlers for
events even if the features in question aren't turned on.

8:05 AM Changeset in webkit [181506] by ddkilzer@apple.com
  • 1 edit
    1 add in trunk/LayoutTests

[iOS] REGRESSION (r181191): Add results for css3/blending/svg-blend-plus-lighter.html

Regressed with:

Support "plus-lighter" in mix-blend mode

  • platform/ios-simulator/css3/blending/svg-blend-plus-lighter-expected.txt: Add.

Mar 14, 2015:

10:32 PM Changeset in webkit [181505] by Simon Fraser
  • 28 edits in trunk/Source

Clean up use of flags in localToContainer-type functions

Reviewed by Alexey Proskuryakov.


RenderObject::localToContainerQuad() had the questionable behavior of always
enforcing UseTransforms in the flags. However, a future patch will need to call
localToContainerQuad() without this flag.

Fix by requiring callers of localToAbsoluteQuad(), localToContainerQuad(), localToContainerPoint(),
and absoluteToLocalQuad() to pass the UseTransforms flag in, providing it as a default argument.

The default value of the MapCoordinatesFlags parameter to mapLocalToContainer(),
a lower-level function, is removed.

Sprinkle a few more nullptrs around.

No behavior change.

  • rendering/RenderBlock.cpp:

(WebCore::RenderBlock::selectionGapRectsForRepaint): call the higher-level localToContainerPoint()
instead of mapLocalToContainer().
(WebCore::RenderBlock::absoluteQuads): Pass UseTransforms.

  • rendering/RenderBox.cpp:

(WebCore::RenderBox::absoluteQuads): Ditto.

  • rendering/RenderBox.h:
  • rendering/RenderFlowThread.h:
  • rendering/RenderGeometryMap.h:


  • rendering/RenderImage.cpp:

(WebCore::RenderImage::collectSelectionRects): This function was erroneously passing
'false' as flags. Pass UseTransforms instead (but no behavior change since
UseTransforms was forced on lower down).

  • rendering/RenderInline.h:
  • rendering/RenderLayer.cpp:

(WebCore::RenderLayer::scrollRectToVisible): UseTransforms is the parameter default, remove it.

  • rendering/RenderLineBreak.cpp:

(WebCore::RenderLineBreak::collectSelectionRects): Another erroneous 'false'.

  • rendering/RenderNamedFlowFragment.cpp:


  • rendering/RenderObject.cpp:

(WebCore::RenderObject::localToContainerQuad): Here's where we no longer force
the UseTransforms bit.
(WebCore::RenderObject::localToContainerPoint): Also here.

  • rendering/RenderObject.h: I prefer bit flags lined up. Makes it easier to spot errors.


  • rendering/RenderText.cpp:

(WebCore::RenderText::collectSelectionRects): Another bad 'false'.

  • rendering/RenderTextLineBoxes.cpp:


  • rendering/RenderView.h:
  • rendering/SimpleLineLayoutFunctions.cpp:


  • rendering/svg/RenderSVGForeignObject.h:
  • rendering/svg/RenderSVGInline.cpp:

(WebCore::RenderSVGInline::absoluteQuads): Another bad 'false'.

  • rendering/svg/RenderSVGInline.h:
  • rendering/svg/RenderSVGModelObject.cpp:


  • rendering/svg/RenderSVGModelObject.h:
  • rendering/svg/RenderSVGRoot.h:
  • rendering/svg/RenderSVGText.cpp:


  • rendering/svg/RenderSVGText.h:
  • rendering/svg/SVGRenderSupport.h:


Pass UseTransforms explicitly to localToContainerPoint()

  • WebProcess/WebPage/ios/WebPageIOS.mm:


10:11 PM Changeset in webkit [181504] by Brent Fulgham
  • 5 edits in trunk/Source

[iOS] scroll snap points are animating to the wrong positions...

Reviewed by Simon Fraser.

Avoid adding an extra '0' snap point to our set. We always start with one zero; this
extra append just forces us to do more steps in our search for nearest snap point.

  • page/scrolling/AxisScrollSnapOffsets.cpp:

(WebCore::updateFromStyle): Remove extra '0' appended to offsets.

[iOS] scroll snap points are animating to the wrong positions.

Reviewed by Simon Fraser.

Scroll snapping was landing in the wrong place on iOS because of two problems:
(1) It was searching for the closest snap offset point using unscaled 'screen' pixels,
which caused it to always choose one of the earliest snap point options.
(2) It was then selecting a scaled snap point coordinate and passing it back to UIKit
to animate the snap. This caused it to select a target point beyond the 'screen' pixel
we want to hit.

The solution to both problems are to scale the scroll destination UIKit suggests so that
we search among the scaled points with a valid value. Then, we need to scale the returned
value back to screen units before handing it back to UIKit to process.

  • UIProcess/API/Cocoa/WKWebView.mm:

(-[WKWebView scrollViewWillBeginDragging:]): Drive-by fix. Get rid of extra ';' at
the end of the line.

  • UIProcess/ios/RemoteScrollingCoordinatorProxyIOS.mm:


9:58 PM Changeset in webkit [181503] by rniwa@webkit.org
  • 4 edits
    3 adds in trunk

parseClass should popScope after pushScope

Reviewed by Benjamin Poulain.


Pop the parser scope as needed.

  • parser/Parser.cpp:



Added a regression test for the test case that caught this bug.

  • TestExpectations: Skipped the test since ES6 class syntax isn't enabled by default.
  • js/class-syntax-scoping-expected.txt: Added.
  • js/class-syntax-scoping.html: Added.
  • js/script-tests/class-syntax-scoping.js: Added.


3:54 PM Changeset in webkit [181502] by dino@apple.com
  • 10 edits in trunk/Source

Feature flag for Animations Level 2

Reviewed by Brent Fulgham.

Add ENABLE_CSS_ANIMATIONS_LEVEL_2 and a runtime flag animationTriggersEnabled.

  • Configurations/FeatureDefines.xcconfig:


  • bindings/generic/RuntimeEnabledFeatures.cpp:


  • bindings/generic/RuntimeEnabledFeatures.h:


2:42 PM Changeset in webkit [181501] by msaboff@apple.com
  • 2 edits in trunk/Source/WTF

Disable Yarr JIT for ARMv7k

Rubber stamped by Geoffrey Garen.

Changed the setting of ENABLE_YARR_JIT to be predicated on ENABLED_JIT.

  • wtf/Platform.h:
1:49 PM Changeset in webkit [181500] by cavalcantii@gmail.com
  • 3 edits in trunk/Source/WebCore

RenderBlock::imageChange() calling const methods on exit

Reviewed by Brent Fulgham.

No new tests, no change on behavior.

  • rendering/RenderBlock.cpp:

(WebCore::RenderBlock::imageChanged): Deleted.

  • rendering/RenderBlock.h:
11:25 AM Changeset in webkit [181499] by vjaquez@igalia.com
  • 9 edits in trunk

[GStreamer] share GL context in pipeline

Reviewed by Philippe Normand.


Add search of gstreamer-gl library in the GStreamer installation. If
it is found, WTF_USE_GSTREAMER_GL macro is defined.

  • Source/cmake/FindGStreamer.cmake:
  • Source/cmake/OptionsGTK.cmake:


GstGL elements in a pipeline need to be aware of the application's
display and its GL context. This information is shared through context
messages between the pipeline and the browser.

This patch shares this context through a GStreamer's synchronous
message, using the GL information held in the web process.

This patch is based on the work of Philippe Normand for Bug 138562.

No new tests because this is platform specific and it depends in the
run-time availability and configurations of GstGL elements.

  • PlatformGTK.cmake: appends the GstGL header files in the include

directories. Also its library directory is appended.

  • platform/graphics/gstreamer/MediaPlayerPrivateGStreamer.cpp:

(WebCore::mediaPlayerPrivateSyncMessageCallback): New callback function.
Initialize the new class attributes.
(WebCore::MediaPlayerPrivateGStreamer::handleSyncMessage): New method
for handling synchronous messages from the pipeline. This method
currently only handles the GL context sharing.
(WebCore::MediaPlayerPrivateGStreamer::createGSTPlayBin): Configures
the pipeline's bus to handle the synchronous messages.

  • platform/graphics/gstreamer/MediaPlayerPrivateGStreamer.h: Add new

class methods and attributes.


  • WebProcess/gtk/WebProcessMainGtk.cpp: Enable XInitThreads() if

GSTREAMER_GL is used, since GstGL elements use another thread for
queuing GL operations.

10:45 AM Changeset in webkit [181498] by commit-queue@webkit.org
  • 2 edits in trunk/Source/JavaScriptCore

Unreviewed, rolling out r181487.

Caused Speedometer/Full.html to fail (Requested by smfr on

Reverted changeset:

"DFG::PutStackSinkingPhase should eliminate GetStacks that
have an obviously known source"

9:29 AM Changeset in webkit [181497] by msaboff@apple.com
  • 8 edits
    6 adds in trunk

ES6: Add binary and octal literal support

Reviewed by Ryosuke Niwa.


Added a binary literal parser function, parseBinary(), to Lexer patterned after the octal parser.
Refactored the parseBinary, parseOctal and parseDecimal to use a constant size for the number of
characters to try and handle directly. Factored out the shifting past any prefix to be handled by
the caller. Added binary and octal parsing to toDouble() via helper functions.

  • parser/Lexer.cpp:


  • parser/Lexer.h:
  • parser/ParserTokens.h:
  • runtime/JSGlobalObjectFunctions.cpp:



  • wtf/ASCIICType.h:

(WTF::isASCIIBinaryDigit): New support function.
(WTF::isASCIIOctalDigit): Updated to use logical and (&&) instead of binary and (&).


New tests.

  • js/binary-literals-expected.txt: Added.
  • js/binary-literals.html: Added.
  • js/octal-literals-expected.txt: Added.
  • js/octal-literals.html: Added.
  • js/script-tests/binary-literals.js: Added.
  • js/script-tests/octal-literals.js: Added.

Mar 13, 2015:

7:57 PM Changeset in webkit [181496] by achristensen@apple.com
  • 11 edits
    4 adds in trunk

Progress towards CMake on Mac.

Reviewed by Gyuyoung Kim.


  • CMakeLists.txt:
  • Source/PlatformMac.cmake: Added stub.
  • Source/cmake/OptionsMac.cmake:

Change defines to get CMake working.


  • CMakeLists.txt:
  • PlatformMac.cmake:

Added Zone.cpp to Mac CMake builds.


  • PlatformMac.cmake:

Generate TracingDtrace.h based on project.pbxproj.


  • PlatformMac.cmake:

Added new include directories and removed old source files.


  • CMakeLists.txt: Added.


  • PlatformMac.cmake: Added.
7:50 PM Changeset in webkit [181495] by fpizlo@apple.com
  • 2 edits in trunk/Source/JavaScriptCore

Object allocation sinking phase shouldn't re-decorate previously sunken allocations on each fixpoint operation

Reviewed by Oliver Hunt.

Just because promoteHeapAccess() notifies us of an effect to a heap location in a node doesn't
mean that we should handle it as if it was for one of our sinking candidates. Instead we should
prune based on m_sinkCandidates.

This fixes a benign bug where we would generate a lot of repeated IR for some pathological

  • dfg/DFGObjectAllocationSinkingPhase.cpp:


7:50 PM Changeset in webkit [181494] by commit-queue@webkit.org
  • 44 edits in trunk/Source

Unreviewed, rolling out r181483.

Caused use-after-free on many tests (Requested by ap on

Reverted changeset:

"Allow clients to selectively disable plug-ins"

6:44 PM Changeset in webkit [181493] by bshafiei@apple.com
  • 8 edits in tags/Safari-601.1.22.1/Source

Merged r181491. rdar://problem/20153362

6:10 PM Changeset in webkit [181492] by Antti Koivisto
  • 12 edits in trunk/Source/WebCore

Cache glyph widths to GlyphPages

Reviewed by Andreas Kling.

Currently we have a separate cache in Font for glyph widths. In practice we always need
the widths so we can just cache them in GlyphPages. This simplifies the code and removes
a per-character hash lookup from WidthIterator.

  • platform/graphics/Font.cpp:


Rename to make it clear this doesn't cache.


Initialize the width.
This could go to GlyphPage.cpp if we had one.

  • platform/graphics/Font.h:

(WebCore::Font::zeroGlyph): Deleted.
(WebCore::Font::setZeroGlyph): Deleted.
(WebCore::Font::widthForGlyph): Deleted.

  • platform/graphics/FontCascade.cpp:


  • platform/graphics/FontCascadeFonts.cpp:


  • platform/graphics/GlyphPage.h:


Return width too as part of GlyphData.


  • platform/graphics/WidthIterator.cpp:


No need to lookup width separately now.

  • platform/graphics/mac/ComplexTextController.cpp:


  • platform/graphics/mac/ComplexTextControllerCoreText.mm:


  • rendering/mathml/RenderMathMLOperator.cpp:


  • rendering/svg/SVGTextRunRenderingContext.cpp:


  • svg/SVGFontData.cpp:


5:52 PM Changeset in webkit [181491] by eric.carlson@apple.com
  • 8 edits in trunk/Source


Reviewed by Darin Adler.

  • Configurations/FeatureDefines.xcconfig:
4:01 PM Changeset in webkit [181490] by rniwa@webkit.org
  • 14 edits
    3 adds in trunk

Class constructor should throw TypeError when "called"

Reviewed by Michael Saboff.


Added ConstructorKind::None to denote code that doesn't belong to an ES6 class.
This allows BytecodeGenerator to emit code to throw TypeError when generating code block
to call ES6 class constructors.

Most of changes are about increasing the number of bits to store ConstructorKind from one
bit to two bits.

  • bytecode/UnlinkedCodeBlock.cpp:


  • bytecode/UnlinkedCodeBlock.h:

(JSC::UnlinkedFunctionExecutable::constructorKindIsDerived): Deleted.
(JSC::UnlinkedCodeBlock::constructorKindIsDerived): Deleted.

  • bytecompiler/BytecodeGenerator.cpp:

(JSC::BytecodeGenerator::generate): Don't emit bytecode when we had already emitted code
to throw TypeError.
(JSC::BytecodeGenerator::BytecodeGenerator): Emit code to throw TypeError when generating
code to call.

  • bytecompiler/BytecodeGenerator.h:

(JSC::BytecodeGenerator::constructorKindIsDerived): Deleted.

  • bytecompiler/NodesCodegen.cpp:


  • parser/Nodes.cpp:


  • parser/Nodes.h:
  • parser/Parser.cpp:

(JSC::Parser<LexerType>::parseFunctionInfo): Renamed the incoming function argument to
ownerClassKind. Set constructorKind to Base or Derived only if we're parsing a constructor.
(JSC::Parser<LexerType>::parseClass): Don't parse static methods using MethodMode since that
would result in BytecodeGenerator erroneously treating static method named "constructor" as
a class constructor.

  • parser/Parser.h:
  • parser/ParserModes.h:
  • runtime/Executable.h:



Added tests for calling class constructors.

  • TestExpectations: Skipped the test since ES6 class syntax isn't enabled by default.
  • js/class-syntax-call-expected.txt: Added.
  • js/class-syntax-call.html: Added.
  • js/script-tests/class-syntax-call.js: Added.
3:34 PM March 2015 Meeting edited by achristensen@apple.com
3:33 PM March 2015 Meeting edited by achristensen@apple.com
3:32 PM CMake edited by achristensen@apple.com
3:30 PM Changeset in webkit [181489] by commit-queue@webkit.org
  • 2 edits in trunk/Source/WebCore

Fix typo restoreUserInterfaceForOptimizedFullscreeStopWithCompletionHandler.

Patch by Jeremy Jones <jeremyj@apple.com> on 2015-03-13
Reviewed by Eric Carlson.

Add the missing 'n' in "fullscreen".

  • platform/spi/ios/AVKitSPI.h:
3:28 PM CMake created by achristensen@apple.com
2:40 PM WebKitDotOrgUpdates created by bfulgham@webkit.org
2:07 PM Changeset in webkit [181488] by timothy_horton@apple.com
  • 5 edits in trunk/Source

Sites that use a device-width viewport but don't have enough height to fill the view are scaled up

Reviewed by Benjamin Poulain.

  • page/ViewportConfiguration.cpp:

Split shouldIgnoreScalingConstraints into one for each dimension.

Don't force the initial and minimum scales to cover the whole view if the
page claims to want to lay out to device width but then lays out too big.
This will allow pages that misbehave in this way to scale down further
than they previously could, but will result in a region of empty background
color being exposed at the initial/minimum scale.

Update the logging to show each dimension separately.

  • page/ViewportConfiguration.h:
  • UIProcess/ios/WKScrollView.mm:

(-[WKScrollView _rubberBandOffsetForOffset:maxOffset:minOffset:range:outside:]):
Now that the WKContentView can (without pinching) be smaller than the unobscured
region of the WKWebView, we need to take that into account when deciding where
to retarget scrolling.

1:57 PM March 2015 Meeting edited by bfulgham@webkit.org
1:56 PM March 2015 Meeting edited by bfulgham@webkit.org
1:55 PM March 2015 Meeting edited by bfulgham@webkit.org
1:18 PM Changeset in webkit [181487] by fpizlo@apple.com
  • 2 edits in trunk/Source/JavaScriptCore

DFG::PutStackSinkingPhase should eliminate GetStacks that have an obviously known source

Reviewed by Oliver Hunt.

This was an obvious omission from the original PutStackSinkingPhase. Previously, we would treat
GetStacks conservatively and assume that the stack slot escaped. That's pretty dumb, since a
GetStack is a local load of the stack. This change makes GetStack a no-op from the standpoint of
this phase's deferral analysis. At the end we either keep the GetStack (if there was no concrete
deferral) or we replace it with an identity over the value that would have been stored by the
deferred PutStack. Note that this might be a Phi that the phase creates, so this is strictly
stronger than what GCSE could do.

This is probably not a speed-up now, but it will be very useful for the varargs simplification
done in bug 141174.

  • dfg/DFGPutStackSinkingPhase.cpp:
1:14 PM Changeset in webkit [181486] by ggaren@apple.com
  • 3 edits in trunk/Source/JavaScriptCore

Prohibit GC while sweeping

Reviewed by Andreas Kling.

I noticed in https://bugs.webkit.org/show_bug.cgi?id=142636 that a GC
could trigger a sweep which could trigger another GC. Yo Dawg.

I tried to figure out whether this could cause problems or not and it
made me cross-eyed.

(Some clients like to report extra memory cost during deallocation as a
way to indicate that the GC now owns something exclusively. It's
arguably a bug to communicate with the GC in this way, but we shouldn't
do crazy when this happens.)

This patch makes explicit the fact that we don't allow GC while sweeping.

Usually, sweeping implicitly defers GC by virtue of happening during
allocation. But not always.

  • heap/Heap.cpp:

(JSC::Heap::collectAllGarbage): Defer GC while sweeping due to an
explicit GC request.

(JSC::Heap::didFinishCollection): Make sure that zombifying sweep
defers GC by not returning to the non-GC state until we're all done.

  • heap/IncrementalSweeper.cpp:

(JSC::IncrementalSweeper::sweepNextBlock): Defer GC while sweeping due
to a timer.

1:04 PM Changeset in webkit [181485] by mark.lam@apple.com
  • 28 edits
    1 add
    1 delete in trunk/Source

Replace TCSpinLock with a new WTF::SpinLock based on WTF::Atomic.

Reviewed by Filip Pizlo.


  • API/JSValue.mm:


  • API/JSWrapperMap.mm:
  • dfg/DFGCommon.cpp:


  • Changed to use a StaticSpinLock since that's what this code was trying to do anyway.
  • heap/CopiedBlock.h:


  • heap/CopiedSpace.cpp:


  • heap/CopiedSpace.h:
  • heap/GCThreadSharedData.cpp:


  • heap/GCThreadSharedData.h:
  • heap/ListableHandler.h:


  • parser/SourceProvider.cpp:
  • profiler/ProfilerDatabase.cpp:



No new tests because there is no new functionality. This is a refactoring effort.

  • bindings/objc/WebScriptObject.mm:
  • platform/ios/wak/WAKWindow.mm:

(-[WAKWindow initWithLayer:]):
(-[WAKWindow initWithFrame:]):


  • WebProcess/WebPage/EventDispatcher.cpp:


  • WebProcess/WebPage/EventDispatcher.h:
  • WebProcess/WebPage/ViewUpdateDispatcher.cpp:


  • WebProcess/WebPage/ViewUpdateDispatcher.h:


We no longer use TCMalloc in our code, and we now have C++11. This replaces the
TCMalloc_SpinLock with a new WTF::SpinLock based on WTF::Atomic (which is a
wrapper around std::atomic).

Note that there is a StaticSpinLock and a SpinLock:

  1. StaticSpinLock is an alias for SpinLockBase, and its initialization relies on static / global std:atomic being automatically initialized to 0 at compile time. Hence, StaticSpinLock should only be used for statics / globals (including static members of classes / structs).
  1. SpinLock is derived from SpinLockBase, and adds a default constructor to initialize its internal atomic value to 0. Because SpinLock has a constructor, it cannot be used for statics / globals. Objects that want to embed a spin lock as an instance field should use SpinLock so that it is initialized properly. std::atomic will not automatically initialize instance fields to 0. Hence, StaticSpinLock will not work correctly as instance fields of a class / struct.
  • WTF.vcxproj/WTF.vcxproj:
  • WTF.vcxproj/WTF.vcxproj.filters:
  • WTF.xcodeproj/project.pbxproj:
  • wtf/CMakeLists.txt:
  • wtf/MetaAllocator.cpp:


  • wtf/MetaAllocator.h:
  • wtf/SpinLock.h: Added.


  • wtf/TCSpinLock.h: Removed.
  • wtf/text/AtomicString.cpp:
12:24 PM Changeset in webkit [181484] by commit-queue@webkit.org
  • 11 edits
    2 adds in trunk

AX: Provide API for assistive tech to ignore DOM key event handlers

Patch by Doug Russell <d_russell@apple.com> on 2015-03-13
Reviewed by Beth Dakin.

Assistive technology applications on the desktop are heavily dependent on keyboard navigation being reliable. This is greatly hindered by sites that handle key events without updating keyboard selection and then consume the event. It is important for assistive technology apps to allow users to decide to ignore these handlers that are incorrect for their purposes.

This can be fixed by exposing, via a new accessibility attribute, a way to decide, for a given WebCore::Frame, to pre-empt DOM dispatch and instead let accessibility caret browsing take place.


Test: platform/mac/accessibility/prevent-keyboard-event-dispatch.html

  • accessibility/AccessibilityObject.cpp:


  • accessibility/AccessibilityObject.h:
  • accessibility/mac/WebAccessibilityObjectWrapperMac.mm:

(-[WebAccessibilityObjectWrapper accessibilityAttributeNames]):
(-[WebAccessibilityObjectWrapper accessibilityAttributeValue:]):
(-[WebAccessibilityObjectWrapper accessibilityIsAttributeSettable:]):
(-[WebAccessibilityObjectWrapper _accessibilitySetValue:forAttribute:]):

  • dom/Element.cpp:


  • page/EventHandler.cpp:


  • page/EventHandler.h:
  • page/Settings.in:


  • accessibility/parent-delete-expected.txt:
  • platform/mac/accessibility/document-attributes-expected.txt:
  • platform/mac/accessibility/prevent-keyboard-event-dispatch-expected.txt: Added.
  • platform/mac/accessibility/prevent-keyboard-event-dispatch.html: Added.
11:58 AM Changeset in webkit [181483] by Conrad Shultz
  • 44 edits in trunk/Source

Allow clients to selectively disable plug-ins

Reviewed by Anders Carlsson.


Add new functionality allow clients to declaratively disable individual plug-ins (in a manner that conceals them
from the page). As part of this:

1) Introduce the concept of web-visible plug-ins and related concepts, as distinct from the real underlying

set of plug-ins.

2) Where applicable, plumb additional information about plug-ins (specifically, bundle identification) deeper

down the stack.

3) Add generic functionality to PluginStrategy to support filtering plug-ins for visibility and introduce

a concrete implementation thereof in WebPlatformStrategies in WebKit2.

4) Add messaging infrastructure to allow clients to set and clear plug-in policies.

While currently only used in a very limited manner, the new declarative plug-in policies are written generically
so that they could be easily used in the future to reduce synchronous messaging to the client when loading plug-ins.

  • dom/DOMImplementation.cpp:

Update to reflect function rename.

  • loader/SubframeLoader.cpp:

Adopt getWebVisibleMimesAndPluginIndices().
Update to reflect function rename.

  • platform/PlatformStrategies.h:

Export platformStrategies(), since it is now used in WebProcess.cpp.

  • plugins/DOMMimeType.cpp:

Adopt getWebVisibleMimesAndPluginIndices().

  • plugins/DOMMimeType.h:

Don't return references in a few places where it is no longer safe to do so.
(WebCore::DOMMimeType::mimeClassInfo): Deleted.

  • plugins/DOMMimeTypeArray.cpp:

Adopt getWebVisibleMimesAndPluginIndices().

  • plugins/DOMPlugin.cpp:

Adopt getWebVisiblePlugins().
Adopt getWebVisibleMimesAndPluginIndices().

  • plugins/DOMPlugin.h:

(WebCore::DOMPlugin::pluginInfo): Deleted.

  • plugins/DOMPluginArray.cpp:

Adopt getWebVisiblePlugins().

  • plugins/PluginData.cpp:

Stash the passed-in Page and call initPlugins().
New member function; call through to PluginStrategy::getWebVisiblePluginInfo().
New member function; build up the mimes and mimePluginIndices vectors in the same manner as before, but
limited to the web-visible plug-ins.
Renamed from supportsMimeType(); update to work in terms of web-visible plug-ins.
Renamed from pluginInfoForMimeType(); ditto.
Renamed from pluginNameForMimeType(); ditto.
Renamed from pluginFileForMimeType(); ditto.
(WebCore::PluginData::supportsMimeType): Deleted.
(WebCore::PluginData::pluginInfoForMimeType): Deleted.
(WebCore::PluginData::pluginNameForMimeType): Deleted.
(WebCore::PluginData::pluginFileForMimeType): Deleted.

  • plugins/PluginData.h:

Add a member variable for the associate Page; declare the PluginLoadClientPolicy enumeration; add
new members to PluginInfo for the clientLoadPolicy and bundle information.
Replace some member functions with new ones that will hide plug-ins upon request from the client;
(WebCore::PluginData::mimes): Deleted.
(WebCore::PluginData::mimePluginIndices): Deleted.

  • plugins/PluginStrategy.h:

Declare new member functions for retrieving web-visible plug-ins and setting/clearing plug-in policies.

  • replay/SerializationMethods.cpp:

Remove now-obsolete code for handling MIME types and add a FIXME.
Handle the new members in PluginInfo.

  • replay/WebInputs.json:

Teach Replay about PluginLoadClientPolicy.


Add new functionality allow clients to declaratively disable individual plug-ins (in a manner that conceals them
from the page). As part of this:

1) Introduce the concept of web-visible plug-ins and related concepts, as distinct from the real underlying

set of plug-ins.

2) Where applicable, plumb additional information about plug-ins (specifically, bundle identification) deeper

down the stack.

3) Add generic functionality to PluginStrategy to support filtering plug-ins for visibility and introduce

a concrete implementation thereof in WebPlatformStrategies in WebKit2.

4) Add messaging infrastructure to allow clients to set and clear plug-in policies.

While currently only used in a very limited manner, the new declarative plug-in policies are written generically
so that they could be easily used in the future to reduce synchronous messaging to the client when loading plug-ins.

  • Plugins/WebBasePluginPackage.mm:

(-[WebBasePluginPackage getPluginInfoFromPLists]):
Set new members in PluginInfo.

  • Plugins/WebNetscapePluginPackage.mm:

(-[WebNetscapePluginPackage getPluginInfoFromResources]):
Set new members in PluginInfo.

  • WebCoreSupport/WebPlatformStrategies.h:
  • WebCoreSupport/WebPlatformStrategies.mm:

Override to wrap getPluginInfo().
Override to do nothing.

  • WebView/WebFrame.mm:

(-[WebFrame _canProvideDocumentSource]):
Update to reflect function rename.


  • WebCoreSupport/WebPlatformStrategies.h:
  • WebCoreSupport/WebPlatformStrategies.cpp:

Wrap getPluginInfo();

  • WebView.cpp:

Update to reflect function rename.


Add new functionality allow clients to declaratively disable individual plug-ins (in a manner that conceals them
from the page). As part of this:

1) Introduce the concept of web-visible plug-ins and related concepts, as distinct from the real underlying

set of plug-ins.

2) Where applicable, plumb additional information about plug-ins (specifically, bundle identification) deeper

down the stack.

3) Add generic functionality to PluginStrategy to support filtering plug-ins for visibility and introduce

a concrete implementation thereof in WebPlatformStrategies in WebKit2.

4) Add messaging infrastructure to allow clients to set and clear plug-in policies.

While currently only used in a very limited manner, the new declarative plug-in policies are written generically
so that they could be easily used in the future to reduce synchronous messaging to the client when loading plug-ins.

  • Shared/Plugins/Netscape/mac/NetscapePluginModuleMac.mm:

Set new PluginInfo members.

  • Shared/WebCoreArgumentCoders.cpp:

Handle new PluginInfo members.

  • Shared/WebProcessCreationParameters.h:

Declare a member to store client plug-in policies.

  • Shared/WebProcessCreationParameters.cpp:

Encode the policies.
Decode the policies.

  • UIProcess/API/C/WKAPICast.h:

Teach how to cast between WebCore and WebKit2 policy definitions.

  • UIProcess/API/C/WKPluginLoadPolicy.h:

Declare WKPluginLoadClientPolicy, matching PluginLoadClientPolicy in WebCore.

  • UIProcess/API/C/mac/WKContextPrivateMac.h:
  • UIProcess/API/C/mac/WKContextPrivateMac.mm:

New function; wrap WebProcessPool::setPluginLoadClientPolicy().
New function; wrap WebProcessPool::clearPluginClientPolicies().

  • UIProcess/WebProcessPool.h:
  • UIProcess/WebProcessPool.cpp:

Copy any client plug-in policies to the new process' creation parameters.
Send SetPluginLoadClientPolicy to all processes and stash the passed-in policy info.
Send ClearPluginClientPolicies to all processes and clear any stashed policy info.

  • WebProcess/Plugins/PDF/PDFPlugin.mm:

Set new PluginInfo member.

  • WebProcess/WebCoreSupport/WebFrameLoaderClient.cpp:

Adopt getWebVisibleMimesAndPluginIndices().
Update to reflect reflect function rename.

  • WebProcess/WebCoreSupport/WebPlatformStrategies.h:
  • WebProcess/WebCoreSupport/WebPlatformStrategies.cpp:

Pass the Page to populatePluginCache().
New member function; wrap getPluginInfo(); if appropriate, filter out plug-ins the client has asked
to block on the current page.
New member function; map empty inputs to a wildcard character and insert the new policy into the nested map.
New member function; empty m_hostsToPluginIdentifierData.
New member function; look up the most specific policy for the given plug-in and host.
If appropriate, set the client load policy on the plug-ins; do this irrespective of whether the cache has
been populated since the Page, and therefore host, may change.

  • WebProcess/WebPage/WebPage.cpp:

Update to reflect function rename.

  • WebProcess/WebProcess.h:
  • WebProcess/WebProcess.cpp:

Enumerate and process the client plug-in policies in the creation parameters.
New member function; wrap PluginStrategy::setPluginLoadClientPolicy.
New member function; wrap PluginStrategy::clearPluginClientPolicies().

  • WebProcess/WebProcess.messages.in:

Add messages for setting and clearing policies.

11:19 AM Changeset in webkit [181482] by rniwa@webkit.org
  • 3 edits in trunk/Source/JavaScriptCore

BytecodeGenerator needs to be re-entrant to support miranda functions

Reviewed by Filip Pizlo.

Made CodeCache::getGlobalCodeBlock and CodeCache::getFunctionExecutableFromGlobalCode re-entrant
by not keeping AddResult while invoking BytecodeGenerator::generate.

This is needed to support Miranda functions since they need to be lazily initialized.

  • runtime/CodeCache.cpp:


  • runtime/CodeCache.h:

(JSC::CodeCacheMap::findCacheAndUpdateAge): Extracted from add.
(JSC::CodeCacheMap::addCache): Extracted from add.
(JSC::CodeCacheMap::add): Deleted.

11:02 AM Changeset in webkit [181481] by mark.lam@apple.com
  • 7 edits in trunk/Source

Introduce WTF::Atomic to wrap std::atomic for a friendlier CAS.

Reviewed by Filip Pizlo.


Changed CodeBlock, and the DFG's crashLock to use WTF::Atomic instead of

  • bytecode/CodeBlock.cpp:


  • bytecode/CodeBlock.h:
  • dfg/DFGCommon.cpp:



The CAS functions provided by std::atomic takes a reference to the expected
value and modifies it if the CAS fails. However, in a lot of our CAS usage,
we don't want the expected value to change. The solution to this is to
provide a WTF::Atomic struct that wraps std::atomic, and provide CAS
methods that won't alter the expected value if the CAS fails.

The method names in WTF::Atomic are chosen to be identical to those
in std::atomic so that WTF::Atomic can be a simple drop in replacement
for std::atomic.

Also changed ByteSpinLock to use WTF::Atomic instead of std::atomic.

  • wtf/Atomics.h:


  • wtf/ByteSpinLock.h:


10:58 AM Changeset in webkit [181480] by Chris Dumez
  • 12 edits
    3 adds in trunk

XMLHttpRequests should not prevent a page from entering PageCache

Reviewed by Alexey Proskuryakov.


Make XMLHttpRequest ActiveDOMObjects suspendable in most cases to
drastically improve the likelihood of pages using them to enter
PageCache. XMLHttpRequest used to be only suspendable when not
loading. After this patch, if the XMLHttpRequest is loading when
navigating away from the page, it will be aborted and the page
will enter the PageCache. Upon restoring the page from PageCache,
the XMLHttpRequests' error handlers will be executed to give them
a chance to reload if they want to.

Test: http/tests/navigation/page-cache-xhr.html

  • history/PageCache.cpp:

Do not prevent a page to enter the page cache ff the main document has
an error that is a cancellation and all remaining subresource loaders
are for XHR. We extend the pre-existing mechanism used on iOS, which
allowed PageCaching if the remaining resource loads are for images.

  • loader/DocumentLoader.cpp:

Mark XHR loaders as PageCache acceptable.

  • loader/DocumentThreadableLoader.cpp:


  • loader/DocumentThreadableLoader.h:
  • loader/ThreadableLoader.h:
  • loader/cache/CachedResource.cpp:


  • loader/cache/CachedResource.h:
  • loader/cache/CachedResourceClient.h:


  • xml/XMLHttpRequest.cpp:

Report that we can suspend XMLHttpRequests as long as the window load
event has already fired. If the window load event has not fired yet,
it would be unsafe to cancel the load in suspend() as it would
potentially cause arbitrary JS execution while suspending.

If suspending for PageCache and the request is currently loading, abort
the load and mark that we should fire the error event upon restoring
from PageCache.

Upon resuming, fire the error event in a timer if the load was aborted
for suspending. We need to do this in a timer because we are not allowed
to execute arbitrary JS inside resume().

Add a assertion to make sure we are not firing event inside stop() as
this would potentially cause arbitrary JS execution and it would be
unsafe. It seems to me that our code is currently unsafe but the
assertion does not seem to be hit by our current layout tests. I am
adding the assertion as it would make it clear we have a bug and we
need to fix it.

  • xml/XMLHttpRequest.h:


Add a tests to make sure that loading XMLHttpRequests do not prevent a
page from entering PageCache.

  • http/tests/navigation/page-cache-xhr-expected.txt: Added.
  • http/tests/navigation/page-cache-xhr.html: Added.
  • http/tests/navigation/resources/page-cache-helper.html: Added.
9:00 AM Changeset in webkit [181479] by youenn.fablet@crf.canon.fr
  • 3 edits
    3 adds in trunk

WebKit test infrastructure should automate the process of cloning W3C test suite and importing tests from it

Reviewed by Ryosuke Niwa.


TestRepositories defines the git repositories used to import tests.
It defines for each repository the infrastructure paths to always import, the paths to skip and the revision to use.
ImportExpectartions defines the tests to import.

  • resources/ImportExpectations: Added.
  • resources/TestRepositories: Added.


TestDownloader is introduced to download W3C tests from the official
W3C tests repositories (https://github.com/w3c/web-platform-tests.git
and https://github.com/w3c/csswg-test.git).
TestDownloader clones the repositories and checkouts revisions of the tests
defined in LayoutTests/imported/w3c/resources/TestRepositories.

Similarly to https://codereview.chromium.org/148173016,
TestDownloader copies part of the files based on the rules
defined in LayoutTests/imported/w3c/resources/ImportExpectations.

TestDownloader is used by TestImporter when no source directory is provided.

  • Scripts/webkitpy/w3c/test_downloader.py: Added.


  • Scripts/webkitpy/w3c/test_importer.py:


8:45 AM WebKitGTK/Gardening/Calendar edited by clopez@igalia.com
6:50 AM Changeset in webkit [181478] by commit-queue@webkit.org
  • 2 edits
    1 add in trunk/LayoutTests

Unreviewed Gardening 13th March.


Patch by Marcos Chavarría Teijeiro <chavarria1991@gmail.com> on 2015-03-13

  • platform/gtk/TestExpectations:
  • platform/gtk/fast/borders/mixed-border-style2-expected.txt: Added. The result was missing.
6:47 AM Changeset in webkit [181477] by commit-queue@webkit.org
  • 3 edits in trunk/LayoutTests

Rebaseline accessibility/color-well.html

Patch by Marcos Chavarría Teijeiro <chavarria1991@gmail.com> on 2015-03-13
Reviewed by Chris Fleizach.

The revision (r179922) which makes this test fail, implements a color
picker for GTK+ platform. So now we have different accessibility
information for a color input and we could update the expected file.

TestExpectations has also been updated.

  • platform/gtk/TestExpectations:
  • platform/gtk/accessibility/color-well-expected.txt:
6:00 AM WebKitGTK/Gardening/Calendar edited by chavarria1991@gmail.com
1:27 AM WebKitGTK/Gardening/Calendar edited by chavarria1991@gmail.com
1:08 AM Changeset in webkit [181476] by jhoneycutt@apple.com
  • 3 edits in trunk/Source/WebKit2

Use UIDocumentMenuViewController to allow selection of files from
iCloud Drive in WKFileUploadPanel

Part of <https://bugs.webkit.org/show_bug.cgi?id=142596> WebKit on
iOS should support file upload from iCloud Drive

Reviewed by Andy Estes.

  • Platform/spi/ios/UIKitSPI.h:

Include UIDocumentMenuViewController_Private.h if it exists. If not,
declare this new SPI.

  • UIProcess/ios/forms/WKFileUploadPanel.mm:

Create a UIDocumentInteractionController for the file, and return a
thumbnail image from its smallest icon.

If we can get a UIImage from the file, create a thumbnail from it.
Otherwise, call fallbackIconForFile().

Code moved from -[_WKVideoFileUploadItem displayImage].

Get the file's extension, and get the extension's preferred UTI. If the
UTI is an image type, call iconForImageFile(). If it's a video type,
call iconForVideoFile(). Otherwise, return the fallback icon.

(-[_WKFileUploadItem initWithFileURL:]):
Added an initializer that takes a file URL.

(-[_WKFileUploadItem fileURL]):
(-[_WKImageFileUploadItem initWithFileURL:originalImage:]):
(-[_WKVideoFileUploadItem displayImage]):
Inherit from UIDocumentPickerDelegate and UIDocumentMenuDelegate. Add
an ivar for the document menu controller.

(-[WKFileUploadPanel dealloc]):
(-[WKFileUploadPanel presentWithParameters:resultListener:]):
If instances of UIDocumentMenuViewController respond to
-_setIgnoreApplicationEntitlementForImport:, meaning we can disable the
assertion that the current app has the iCloud Documents entitlement,
then show the document picker menu. Otherwise, show the current source
selection UI. The runtime check will be removed when a newer UIKit is

Return the UTIs for the <input> element's "accept" attribute's MIME
type list.

(-[WKFileUploadPanel _mediaTypesForPickerSourceType:]):
Call UTIsForMIMETypes(). Fall back to available source types for this
picker type.

(-[WKFileUploadPanel _documentPickerMenuMediaTypes]):
Call UTIsForMIMETypes(). Fall back to the all-encompassing
"public.item" UTI.

(-[WKFileUploadPanel _photoLibraryButtonLabel]):
(-[WKFileUploadPanel _cameraButtonLabel]):
Factored out of -_showMediaSourceSelectionSheet. Return the
appropriate label for the camera button, based on whether it can take a
photo or video, or return nil if the camera is not available.

(-[WKFileUploadPanel _showMediaSourceSelectionSheet]):
Changed to use -_cameraButtonLabel and

(-[WKFileUploadPanel _showDocumentPickerMenu]):
Create a UIDocumentMenuViewController, and populate it with options to
open the photo library or take a picture or video. If iCloud Drive or
other document sources are available, they will be added automatically.

(-[WKFileUploadPanel _presentForCurrentInterfaceIdiom:]):
(-[WKFileUploadPanel documentMenu:didPickDocumentPicker:]):
(-[WKFileUploadPanel documentMenuWasCancelled:]):
(-[WKFileUploadPanel documentPicker:didPickDocumentAtURL:]):
(-[WKFileUploadPanel documentPickerWasCancelled:]):
(-[WKFileUploadPanel _uploadItemFromMediaInfo:successBlock:failureBlock:]):

(-[_WKImageFileUploadItem initWithFilePath:originalImage:]): Deleted.
(-[_WKImageFileUploadItem fileURL]): Deleted.
(-[_WKVideoFileUploadItem initWithFilePath:mediaURL:]): Deleted.
(-[_WKVideoFileUploadItem fileURL]): Deleted.

12:52 AM Changeset in webkit [181475] by commit-queue@webkit.org
  • 3 edits in trunk/Source/WebCore

Fix Debug build error 'comparison is always true due to limited range of data type [-Werror=type-limits]'

Patch by Joonghun Park <jh718.park@samsung.com> on 2015-03-13
Reviewed by Csaba Osztrogonác.

No new tests, no behavior changes.

Now CSSPropertyID type is uint16_t, so propertyID >= 0 check is needed no more.

  • css/CSSPrimitiveValue.cpp:


  • css/makeprop.pl:
12:18 AM Changeset in webkit [181474] by commit-queue@webkit.org
  • 1 edit
    3 adds in trunk/LayoutTests

Unreviewed GTK+ Gardening

Rebaseline the missing test result

Patch by Hunseop Jeong <Hunseop Jeong> on 2015-03-13

  • platform/gtk/fast/borders/mixed-border-style2-expected.png: Added.
  • platform/gtk/fast/css/css2-system-fonts-expected.txt: Added.
  • platform/gtk/fast/text/tatechuyoko-expected.txt: Added.

Mar 12, 2015:

11:26 PM Changeset in webkit [181473] by bshafiei@apple.com
  • 5 edits in tags/Safari-601.1.22.1/Source


11:23 PM Changeset in webkit [181472] by bshafiei@apple.com
  • 1 copy in tags/Safari-601.1.22.1

New tag.

11:17 PM Changeset in webkit [181471] by bshafiei@apple.com
  • 5 edits in branches/safari-600.5-branch/Source


11:17 PM Changeset in webkit [181470] by bshafiei@apple.com
  • 5 edits in branches/safari-600.1.4.15-branch/Source


8:02 PM Changeset in webkit [181469] by mark.lam@apple.com
  • 2 edits in trunk/Source/JavaScriptCore

Change the DFG crashLock to use std::atomic.

Reviewed by Filip Pizlo.

  • dfg/DFGCommon.cpp:


7:01 PM Changeset in webkit [181468] by commit-queue@webkit.org
  • 14 edits
    4 deletes in trunk/Source/WebCore

Remove DrawingBuffer

Patch by Zan Dobersek <zdobersek@igalia.com> on 2015-03-12
Reviewed by Darin Adler.

Remove the DrawingBuffer class. Objects of this type were only held in the
WebGLRenderingContext (later renamed to WebGLRenderingContextBase) on the
Chromium port, with the relevant code removed in r147888. Since then, the
m_drawingBuffer member variable has always been null.

  • CMakeLists.txt:
  • PlatformEfl.cmake:
  • PlatformGTK.cmake:
  • WebCore.vcxproj/WebCore.vcxproj:
  • WebCore.vcxproj/WebCore.vcxproj.filters:
  • WebCore.xcodeproj/project.pbxproj:
  • html/canvas/WebGL2RenderingContext.cpp:


  • html/canvas/WebGLRenderingContext.cpp:


  • html/canvas/WebGLRenderingContextBase.cpp:


  • html/canvas/WebGLRenderingContextBase.h:

(WebCore::ScopedDrawingBufferBinder::ScopedDrawingBufferBinder): Deleted.
(WebCore::ScopedDrawingBufferBinder::~ScopedDrawingBufferBinder): Deleted.

  • platform/graphics/GraphicsContext.h:
  • platform/graphics/GraphicsContext3D.h:
  • platform/graphics/cairo/DrawingBufferCairo.cpp: Removed.
  • platform/graphics/gpu/DrawingBuffer.cpp: Removed.
  • platform/graphics/gpu/DrawingBuffer.h: Removed.
  • platform/graphics/gpu/mac/DrawingBufferMac.mm: Removed.
  • platform/graphics/opengl/GraphicsContext3DOpenGLCommon.cpp:


6:57 PM Changeset in webkit [181467] by fpizlo@apple.com
  • 6 edits in trunk/Source

Bytecode liveness analysis should have more lambdas and fewer sets

Reviewed by Mark Lam.


In bug 141174 I'll need to identify all of the bytecode kill sites. This requires hooking into
the bytecode analysis' stepOverFunction method, except in such a way that we observe uses that
are not in outs. This refactors stepOverFunction so that you can pass it use/def functors that
can either be used to propagate outs (as we do right now) or to additionally detect kills or
whatever else.

In order to achieve this, the liveness analysis was moved off of maintaining uses/defs
bitvectors. This wasn't helping the abstraction and was probably inefficient. The new code
should be a bit faster since we don't have to clear uses/defs bitvectors on each instruction. On
the other hand, being able to intercept each use means that our code for exception handlers is
no longer a bitwise-merge; it requires finding set bits. Fortunately, this code only kicks in
for instructions inside a try, and its performance is O(live at catch), so that's probably not

  • bytecode/BytecodeLivenessAnalysis.cpp:

(JSC::setForOperand): Deleted.

  • bytecode/BytecodeUseDef.h:


  • bytecode/CodeBlock.cpp:


Add a method for iterating each set bit in a FastBitVector. Uses a functor as a callback since
this allows for a more efficient algorithm.

  • wtf/FastBitVector.h:


6:11 PM Changeset in webkit [181466] by rniwa@webkit.org
  • 31 edits
    10 adds in trunk/Source/JavaScriptCore

"this" should be in TDZ until super is called in the constructor of a derived class

Reviewed by Mark Hahnenberg.

DFG and FTL implementations co-authored by Filip Pizlo.

In ES6 class syntax, "this" register must be in the "temporal dead zone" (TDZ) and throw ReferenceError until
super() is called inside the constructor of a derived class.

Added op_check_tdz, a new OP code, which throws a reference error when the first operand is an empty value
to all tiers of JIT and LLint. The op code throws in the slow path on the basis that a TDZ error should be
a programming error and not a part of the programs' normal control flow. In DFG, this op code is represented
by a no-op must-generate node CheckNotEmpty modeled after CheckCell.

Also made the constructor of a derived class assign the empty value to "this" register rather than undefined
so that ThisNode can emit the op_check_tdz to check the initialized-ness of "this" in such a constructor.

  • bytecode/BytecodeList.json: Added op_check_tdz.
  • bytecode/BytecodeUseDef.h:

(JSC::computeUsesForBytecodeOffset): Ditto.
(JSC::computeDefsForBytecodeOffset): Ditto.

  • bytecode/CodeBlock.cpp:

(JSC::CodeBlock::dumpBytecode): Ditto.

  • bytecode/ExitKind.cpp:

(JSC::exitKindToString): Added TDZFailure.

  • bytecode/ExitKind.h: Ditto.
  • bytecompiler/BytecodeGenerator.cpp:

(JSC::BytecodeGenerator::BytecodeGenerator): Assign the empty value to "this" register to indicate it's in TDZ.
(JSC::BytecodeGenerator::emitTDZCheck): Added.
(JSC::BytecodeGenerator::emitReturn): Emit the TDZ check since "this" can still be in TDZ if super() was never
called. e.g. class B extends A { constructor() { } }

  • bytecompiler/BytecodeGenerator.h:
  • bytecompiler/NodesCodegen.cpp:

(JSC::ThisNode::emitBytecode): Always emit the TDZ check if we're inside the constructor of a derived class.
We can't omit this check even if the result was ignored per spec.

  • dfg/DFGAbstractInterpreterInlines.h:

(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects): Previously, empty value could never appear
in a local variable. This is no longer true so generalize this code. Also added the support for CheckNotEmpty.
Like CheckCell, we phantomize this DFG node in the constant folding phase if the type of the operand is already
found to be not empty. Otherwise filter out SpecEmpty.

  • dfg/DFGByteCodeParser.cpp:

(JSC::DFG::ByteCodeParser::parseBlock): Added op_check_tdz.

  • dfg/DFGCapabilities.cpp:

(JSC::DFG::capabilityLevel): op_check_tdz can be compiled and inlined.

  • dfg/DFGClobberize.h:

(JSC::DFG::clobberize): CheckNotEmpty doesn't read or write values.

  • dfg/DFGConstantFoldingPhase.cpp:

(JSC::DFG::ConstantFoldingPhase::foldConstants): Convert CheckNotEmpty to a phantom if non-emptiness had already
been proven for the operand prior to this node.

  • dfg/DFGDoesGC.cpp:

(JSC::DFG::doesGC): CheckNotEmpty does not trigger GC.

  • dfg/DFGFixupPhase.cpp:

(JSC::DFG::FixupPhase::fixupNode): CheckNotEmpty is a no-op in the fixup phase.

  • dfg/DFGNodeType.h: CheckNotEmpty cannot be removed even if the result was ignored. See ThisNode::emitBytecode.
  • dfg/DFGPredictionPropagationPhase.cpp:

(JSC::DFG::PredictionPropagationPhase::propagate): CheckNotEmpty doesn't return any value.

  • dfg/DFGSafeToExecute.h:

(JSC::DFG::safeToExecute): CheckNotEmpty doesn't load from heap so it's safe.

  • dfg/DFGSpeculativeJIT32_64.cpp:

(JSC::DFG::SpeculativeJIT::compile): Speculative the operand to be not empty. OSR exit if the speculation fails.

  • dfg/DFGSpeculativeJIT64.cpp:

(JSC::DFG::SpeculativeJIT::compile): Ditto.

  • ftl/FTLCapabilities.cpp:

(JSC::FTL::canCompile): CheckNotEmpty can be compiled in FTL.

  • ftl/FTLLowerDFGToLLVM.cpp:

(JSC::FTL::LowerDFGToLLVM::compileNode): Calls compileCheckNotEmpty for CheckNotEmpty.
(JSC::FTL::LowerDFGToLLVM::compileCheckNotEmpty): OSR exit with "TDZFailure" if the operand is not empty.

  • jit/JIT.cpp:

(JSC::JIT::privateCompileMainPass): Added op_check_tdz.
(JSC::JIT::privateCompileSlowCases): Ditto.

  • jit/JIT.h:
  • jit/JITOpcodes.cpp:

(JSC::JIT::emit_op_check_tdz): Implements op_check_tdz in Baseline JIT.
(JSC::JIT::emitSlow_op_check_tdz): Ditto.

  • jit/JITOpcodes32_64.cpp:

(JSC::JIT::emit_op_check_tdz): Ditto.
(JSC::JIT::emitSlow_op_check_tdz): Ditto.

  • llint/LowLevelInterpreter32_64.asm: Implements op_check_tdz in LLint.
  • llint/LowLevelInterpreter64.asm: Ditto.
  • runtime/CommonSlowPaths.cpp:

(JSC::SLOW_PATH_DECL): Throws a reference error for op_check_tdz. Shared by LLint and Baseline JIT.

  • runtime/CommonSlowPaths.h:
  • tests/stress/class-syntax-no-loop-tdz.js: Added.
  • tests/stress/class-syntax-no-tdz-in-catch.js: Added.
  • tests/stress/class-syntax-no-tdz-in-conditional.js: Added.
  • tests/stress/class-syntax-no-tdz-in-loop-no-inline-super.js: Added.
  • tests/stress/class-syntax-no-tdz-in-loop.js: Added.
  • tests/stress/class-syntax-no-tdz.js: Added.
  • tests/stress/class-syntax-tdz-in-catch.js: Added.
  • tests/stress/class-syntax-tdz-in-conditional.js: Added.
  • tests/stress/class-syntax-tdz-in-loop.js: Added.
  • tests/stress/class-syntax-tdz.js: Added.
5:57 PM Changeset in webkit [181465] by rniwa@webkit.org
  • 3 edits
    2 adds in trunk

REGRESSION(r180726): Removing an empty line at the end of textarea clears the entire texture

Reviewed by Darin Adler.


The bug was caused by TypingCommand::deleteKeyPressed erroneously determining the editable root to be empty because
Position::atStartOfTree returns true when it's anchored at a BR that is immediately below the root editable element.

Fixed the bug by replacing the use of the deprecated atFirstEditingPositionForNode by a code that understands modern
position types such as PositionIsBeforeAnchor in atStartOfTree and atEndOfTree. These two functions will no longer
return true when anchored before or after BR after this patch.

Test: editing/deleting/delete-empty-line-breaks-at-end-of-textarea.html

  • dom/Position.cpp:



Added a regression test for deleting empty lines at the end of a textarea element.

  • editing/deleting/delete-empty-line-breaks-at-end-of-textarea-expected.txt: Added.
  • editing/deleting/delete-empty-line-breaks-at-end-of-textarea.html: Added.
5:55 PM Changeset in webkit [181464] by bshafiei@apple.com
  • 1 copy in tags/Safari-600.

New tag.

5:53 PM Changeset in webkit [181463] by bshafiei@apple.com
  • 1 copy in tags/Safari-600.5.13

New tag.

5:40 PM Changeset in webkit [181462] by msaboff@apple.com
  • 2 edits in trunk/Source/WTF

Disable Yarr JIT for ARMv7k

Reviewed by Oliver Hunt.

Make the setting of ENABLE_YARR_JIT match ENABLE_JIT for ARMv7k.

  • wtf/Platform.h:
5:20 PM Changeset in webkit [181461] by mark.lam@apple.com
  • 2 edits in trunk/Source/WTF

Change WTF::ByteSpinLock to use std::atomic.

Reviewed by Filip Pizlo.

  • wtf/ByteSpinLock.h:


4:54 PM Changeset in webkit [181460] by Brent Fulgham
  • 2 edits in trunk/Source/ThirdParty/ANGLE

[Win] Unreviewed build fix (part 2).

  • ANGLE.vcxproj/libGLESv2Common.props: Missed a directory separator.
4:47 PM Changeset in webkit [181459] by Brent Fulgham
  • 2 edits in trunk/Source/ThirdParty/ANGLE

[Win] Unreviewed build fix for some DirectX SDK Installs

  • ANGLE.vcxproj/libGLESv2Common.props: Correct path for Lib and

Include paths based on problems others have had building WebKit
under different Windows and DirectX revisions.

3:57 PM Changeset in webkit [181458] by Yusuke Suzuki
  • 23 edits
    1 move
    1 add in trunk

Integrate MapData into JSMap and JSSet

Reviewed by Filip Pizlo.


This patch integrates MapData into JSMap and JSSet.
This removes 2 object allocation per one JSMap / JSSet.

MapDataImpl is specialized into MapData and SetData.
In the case of SetData, it does not have the dummy values
previously stored in the MapDataImpl. So the storage size of SetData
becomes the half of the previous implementation.

And now MapData and SetData are completely integrated into JSMap and JSSet,
these structures are not exposed to the other code even in WebCore world.

And at the same time, this patch fixes missing destroy functions
in JSMapIterator and JSSetIterator.
They are needed because MapData::const_iterator is a non-trivial destructor.

(JSC::JSMap::finishCreation): Deleted.

  • runtime/JSMap.h:

(JSC::JSMap::mapData): Deleted.

  • runtime/JSMapIterator.cpp:


  • runtime/JSMapIterator.h:


  • runtime/JSSet.cpp:

(JSC::JSSet::finishCreation): Deleted.

  • runtime/JSSet.h:

(JSC::JSSet::mapData): Deleted.

  • runtime/JSSetIterator.cpp:


  • runtime/JSSetIterator.h:


  • runtime/MapConstructor.cpp:


  • runtime/MapData.h:

(JSC::MapData::const_iterator::key): Deleted.
(JSC::MapData::const_iterator::value): Deleted.
(JSC::MapData::create): Deleted.
(JSC::MapData::createStructure): Deleted.
(JSC::MapData::size): Deleted.
(JSC::MapData::clear): Deleted.
(JSC::MapData::KeyType::KeyType): Deleted.
(JSC::MapData::const_iterator::internalIncrement): Deleted.
(JSC::MapData::const_iterator::ensureSlot): Deleted.
(JSC::MapData::const_iterator::const_iterator): Deleted.
(JSC::MapData::const_iterator::~const_iterator): Deleted.
(JSC::MapData::const_iterator::operator*): Deleted.
(JSC::MapData::const_iterator::end): Deleted.
(JSC::MapData::const_iterator::operator!=): Deleted.
(JSC::MapData::const_iterator::operator==): Deleted.

  • runtime/MapDataInlines.h: Renamed from Source/JavaScriptCore/runtime/MapData.cpp.


  • runtime/MapPrototype.cpp:

(JSC::getMapData): Deleted.

  • runtime/SetPrototype.cpp:

(JSC::getMapData): Deleted.

  • runtime/VM.cpp:


  • runtime/VM.h:


Now Set has SetData and it's different from MapData.
And MapData/SetData are completely integrated into JSSet and JSMap.
Structured-cloning algorithm need to be aware of these changes.
And in the case of JSSet, since JSSet doesn't need dummy value for construction,
Structured-cloning only serialize the keys in JSSet.

  • ForwardingHeaders/runtime/MapDataInlines.h: Added.
  • bindings/js/SerializedScriptValue.cpp:

(WebCore::CloneDeserializer::consumeMapDataTerminationIfPossible): Deleted.


Test structure-clone implementation that becomes aware of the difference between Map and Set.

  • fast/storage/serialized-script-value.html:
3:50 PM Changeset in webkit [181457] by ggaren@apple.com
  • 2 edits in trunk/Source/bmalloc

Assertion failure in bmalloc::LargeObject::validateSelf on Mavericks Debug layout test bot

Reviewed by Michael Saboff.

The typical backtrace to this crash shows the main thread trying to
realloc a large string while a DFG compiler thread tries to
free a large vector buffer.

I believe that this is a race condition -- at least in debug builds --
since the main thread will try to validate its object's neighbors
without holding a lock, even though those neighbors might be in the
midst of changing.

In general, there may be sneaky times when it is valid to look at an
object's metadata without holding the heap lock, but it is best not to
do so unless we have a really really good reason to.

  • bmalloc/Allocator.cpp:

(bmalloc::Allocator::reallocate): Take a lock before reading the metadata
for this object, since we generally require any access to shared heap
metadata to take a lock.

3:49 PM March 2015 Meeting edited by Simon Fraser
3:14 PM Changeset in webkit [181456] by mark.lam@apple.com
  • 4 edits in trunk/Source/JavaScriptCore

Use std::atomic for CodeBlock::m_visitAggregateHasBeenCalled.

Reviewed by Mark Hahnenberg.

We used to spin our own compare and swap on a uint8_t. Now that we can
use C++11, let's use std::atomic instead.

  • bytecode/CodeBlock.cpp:


  • The CAS here needs std::memory_order_acquire ordering because it requires lock acquisition semantics to visit the CodeBlock.
  • bytecode/CodeBlock.h:


  • heap/CodeBlockSet.cpp:


  • These can go with relaxed ordering because they are all done before the GC starts parallel marking.
3:04 PM FiveYearPlan edited by bfulgham@webkit.org
2:54 PM FiveYearPlan edited by bfulgham@webkit.org
2:33 PM FiveYearPlan edited by bfulgham@webkit.org
2:27 PM FiveYearPlan edited by bfulgham@webkit.org
2:15 PM FiveYearPlan created by bfulgham@webkit.org
2:04 PM WikiStart edited by bfulgham@webkit.org
2:01 PM Changeset in webkit [181455] by mitz@apple.com
  • 3 edits in trunk/Source/WebCore

Finish up <rdar://problem/20086546> [Cocoa] Add an option to treat certificate chains with SHA1-signed certificates as insecure

Added back OS X bits that I couldn’t land initially in r181317 or had to remove in r181327.

  • platform/network/mac/CertificateInfoMac.mm:


  • platform/spi/cocoa/SecuritySPI.h:
1:59 PM WebKitGTK/2.8.x edited by clopez@igalia.com
1:49 PM Changeset in webkit [181454] by bshafiei@apple.com
  • 2 edits in branches/safari-600.1.4.15-branch/Source/WebKit2

Merged r181448. rdar://problem/20130417

1:37 PM Changeset in webkit [181453] by ggaren@apple.com
  • 3 edits in trunk/Source/WebCore

REGRESSION: Crash under Heap::reportExtraMemoryAllocatedSlowCase for media element

Reviewed by Mark Hahnenberg.

This was a pre-existing bug that I made a lot worse in

  • html/HTMLMediaElement.cpp:

(WebCore::HTMLMediaElement::parseAttribute): Compare size before
subtracting rather than subtracting and then comparing to zero. The
latter technique is not valid for unsigned integers, which will happily
underflow into giant numbers.

  • Modules/mediasource/SourceBuffer.cpp:

(WebCore::SourceBuffer::reportExtraMemoryAllocated): This code was

technically correct, but I took the opportunity to clean it up a bit.
There's no need to do two checks here, and it smells bad to check for
a negative unsigned integer.

1:32 PM Changeset in webkit [181452] by jonowells@apple.com
  • 6 edits in trunk/Source/WebInspectorUI

Web Inspector: Debugger sidebar should group global breakpoints together

Reviewed by Timothy Hatcher.

Update the DebuggerSidebarPanel class to hold global breakpoints such as "All Exceptions" in one container. This
will be the place future such breakpoints are added.

  • Localizations/en.lproj/localizedStrings.js: Added string.
  • UserInterface/Main.html: Small rearrangement.
  • UserInterface/Views/DebuggerSidebarPanel.js: Change how exception breaking options are displayed.


  • UserInterface/Views/FolderTreeElement.js: Support additional classes for icons.
  • UserInterface/Views/ResourceSidebarPanel.js: Change call to FolderTreeElement constructor.
1:24 PM Changeset in webkit [181451] by ap@apple.com
  • 2 edits in trunk/Tools

[Mavericks] Test snapshots are 1600x1200 on Retina devices

Reviewed by Tim Horton.

  • WebKitTestRunner/mac/PlatformWebViewMac.mm: (WTR::PlatformWebView::windowSnapshotImage):

Removed an ifdef that disabled the fix on Mavericks.

1:04 PM schedule.png attached to March 2015 Meeting by Simon Fraser
12:44 PM Changeset in webkit [181450] by Simon Fraser
  • 2 edits in trunk/LayoutTests

These hidpi filter tests pass sometimes.

  • platform/mac/TestExpectations:
12:31 PM Changeset in webkit [181449] by commit-queue@webkit.org
  • 8 edits in trunk/Source/WebCore

Stop using single-include headers that are only available since GStreamer >= 1.2.


Patch by Sebastian Dröge <sebastian@centricular.com> on 2015-03-12
Reviewed by Philippe Normand.

  • platform/audio/gstreamer/AudioDestinationGStreamer.cpp:
  • platform/audio/gstreamer/AudioFileReaderGStreamer.cpp:
  • platform/audio/gstreamer/AudioSourceProviderGStreamer.cpp:
  • platform/audio/gstreamer/WebKitWebAudioSourceGStreamer.cpp:
  • platform/graphics/gstreamer/GStreamerUtilities.cpp:
  • platform/graphics/gstreamer/GStreamerUtilities.h:
  • platform/graphics/gstreamer/ImageGStreamer.h:

Instead of using single-include headers for the GStreamer libraries,
directly include the headers we need. The single-include headers were
only added in 1.2, and this would be the only reason why we would
depend on 1.2.

12:26 PM Changeset in webkit [181448] by enrica@apple.com
  • 2 edits in trunk/Source/WebKit2

Adopt new API for keyboard event handling.

Reviewed by Sam Weinig.

Adopts the new API.

  • UIProcess/ios/WKContentViewInteraction.mm:

(-[WKContentView _didHandleKeyEvent:]):

12:13 PM Changeset in webkit [181447] by Csaba Osztrogonác
  • 2 edits in trunk/Tools

Remove isWK2 cruft from webkitdirs.pm after r170426

Reviewed by Daniel Bates.

  • Scripts/webkitdirs.pm:

(isWK2): Deleted.

11:38 AM Changeset in webkit [181446] by ap@apple.com
  • 14 edits in trunk

testRunner.setAlwaysAcceptCookies does not work with NetworkProcess


Reviewed by Carlos Garcia Campos.

  • UIProcess/WebCookieManagerProxy.cpp:

(WebKit::WebCookieManagerProxy::setHTTPCookieAcceptPolicy): Don't change system-wide
cookie accept policy when in testing mode. Also, update the policy in WebProcesses too,
as we still do some networking there.

  • UIProcess/WebProcessPool.h:
  • WebProcess/InjectedBundle/API/c/WKBundle.cpp:

(WKBundleSetAlwaysAcceptCookies): Deleted.

  • WebProcess/InjectedBundle/API/c/WKBundlePrivate.h:
  • WebProcess/InjectedBundle/InjectedBundle.cpp:

(WebKit::InjectedBundle::setAlwaysAcceptCookies): Deleted.

  • WebProcess/InjectedBundle/InjectedBundle.h:

Remove WKBundleSetAlwaysAcceptCookies SPI, as WKTR now uses an API in UI process.


Reviewed by Carlos Garcia Campos.

Always change cookie policy from UI process, as changing it from WebProcess doesn't
affect other processes.

The implementation is imperfect, because the change happens asynchronously, but I
don't expect it to be a problem in practice.

  • WebKitTestRunner/InjectedBundle/InjectedBundle.cpp:


  • WebKitTestRunner/InjectedBundle/TestRunner.cpp:


  • WebKitTestRunner/TestInvocation.cpp:



Unreviewed (only code changes were reviewed).

  • platform/gtk/TestExpectations:
  • platform/mac-wk2/TestExpectations:

Unskip tests that used to be affected.

11:19 AM March 2015 Meeting edited by Simon Fraser
10:46 AM Changeset in webkit [181445] by andersca@apple.com
  • 2 edits in trunk/Source/WebKit2

WKErrorDomain should be declared nonnull

Reviewed by Antti Koivisto.

Add WK_ASSUME_NONNULL_BEGIN and WK_ASSUME_NONNULL_END so everything will be implicitly nonnull.

  • UIProcess/API/Cocoa/WKError.h:
10:41 AM Changeset in webkit [181444] by bshafiei@apple.com
  • 16 edits
    2 copies in tags/Safari-601.1.22/Source

Merged r181442. rdar://problem/11903887

10:27 AM Changeset in webkit [181443] by commit-queue@webkit.org
  • 2 edits in trunk/Source/WebKit2

PDFs don't snapshot properly in iOS Safari

Patch by Ian Henderson <ian@ianhenderson.org> on 2015-03-12
Reviewed by Tim Horton.

  • UIProcess/API/Cocoa/WKWebView.mm:

(-[WKWebView _snapshotRect:intoImageOfWidth:completionHandler:]):
If we have a _customContentView, use UIView snapshotting instead of
trying to snapshot the web page.

10:18 AM Changeset in webkit [181442] by eric.carlson@apple.com
  • 16 edits
    2 adds in trunk/Source

[Mac] Update AirPlay handling


Unreviewed, respond to post-review comments.

  • dom/Document.cpp:


  • page/ChromeClient.h:
  • page/Page.cpp:



Unreviewed, respond to post-review comments and add previously reviewed changes
missed by a bungled merge.

  • Shared/mac/WebCoreArgumentCodersMac.mm:
  • UIProcess/WebMediaPlaybackTargetPickerProxy.cpp:


  • UIProcess/mac/PageClientImpl.h:
  • UIProcess/mac/PageClientImpl.mm:


  • UIProcess/mac/WebMediaPlaybackTargetPickerProxyMac.h: Added.
  • UIProcess/mac/WebMediaPlaybackTargetPickerProxyMac.mm: Added.
  • WebKit2.xcodeproj/project.pbxproj:
  • WebProcess/WebCoreSupport/WebChromeClient.cpp:
  • WebProcess/WebCoreSupport/WebChromeClient.h:
  • WebProcess/WebPage/WebPage.h:
  • WebProcess/WebPage/WebPage.messages.in:
  • WebProcess/WebPage/mac/WebPageMac.mm:
  • WebProcess/com.apple.WebProcess.sb.in:
10:02 AM Changeset in webkit [181441] by Simon Fraser
  • 3 edits in trunk/LayoutTests

Clean up after JoePeck.

  • http/tests/security/contentSecurityPolicy/injected-inline-style-blocked-expected.txt:
  • http/tests/security/contentSecurityPolicy/inline-style-allowed-while-cloning-objects-expected.txt:
9:55 AM Changeset in webkit [181440] by Csaba Osztrogonác
  • 2 edits in trunk/Tools

[EFL] Enable building EFL 1.13 with NEON support

Reviewed by Gyuyoung Kim.

  • efl/jhbuild.modules:
9:40 AM Changeset in webkit [181439] by Csaba Osztrogonác
  • 2 edits in trunk/Source/WebCore


Reviewed by Darin Adler.

  • html/parser/HTMLPreloadScanner.cpp:


9:10 AM Changeset in webkit [181438] by commit-queue@webkit.org
  • 2 edits
    5 adds in trunk/LayoutTests

Unreviewed EFL gardening on 13th Mar.

Rebaseline the missing test result.

Patch by Hunseop Jeong <Hunseop Jeong> on 2015-03-12

  • platform/efl/TestExpectations:
  • platform/efl/editing/inserting/before-after-input-element-expected.png: Added.
  • platform/efl/editing/inserting/before-after-input-element-expected.txt: Added.
  • platform/efl/fast/css/css2-system-fonts-expected.txt: Added.
  • platform/efl/fast/text/tatechuyoko-expected.png: Added.
  • platform/efl/fast/text/tatechuyoko-expected.txt: Added.
9:10 AM Changeset in webkit [181437] by commit-queue@webkit.org
  • 2 edits in trunk/Source/WebInspectorUI

Web Inspector: Console Errors during provisional document loads get lost with "Clear Log on Reload"

Patch by Joseph Pecoraro <Joseph Pecoraro> on 2015-03-12
Reviewed by Timothy Hatcher.

  • UserInterface/Views/LogContentView.js:

Detect a provisional load has started to start save messages that come in at this time.

Reappend provisional load messages if we auto-cleared.

Helpers for dealing with the provisional loading state.

8:35 AM March 2015 Meeting edited by Simon Fraser
7:53 AM March 2015 Meeting edited by zandobersek@gmail.com
7:50 AM Changeset in webkit [181436] by Csaba Osztrogonác
  • 2 edits in trunk/Source/WTF

[ARM][Linux] GC sometimes stuck in an infinite loop if parallel GC is enabled

Reviewed by Carlos Garcia Campos.

  • wtf/Platform.h: Enable parallel GC after r181319.
7:13 AM March 2015 Meeting edited by adam.bergkvist@ericsson.com
3:27 AM Changeset in webkit [181435] by Csaba Osztrogonác
  • 2 edits in trunk/Source/JavaScriptCore

[cmake] Fix the incremental build issue revealed by r181419

Reviewed by Carlos Garcia Campos.

  • CMakeLists.txt:
1:21 AM March 2015 Meeting edited by benjamin@webkit.org
1:16 AM March 2015 Meeting edited by benjamin@webkit.org

Mar 11, 2015:

11:54 PM March 2015 Meeting edited by Chris Dumez
11:37 PM Changeset in webkit [181434] by Joseph Pecoraro
  • 2 edits in trunk/Source/WebCore

Unreviewed follow-up fix to r181426. Initialize TextPosition with zeros in case it gets used.

Address ASSERT in LayoutTests/printing/page-format-data.html.

  • dom/InlineStyleSheetOwner.cpp:

In case the TextPosition gets used because it wasn't created by a parser,
zeros are more realistic values then beforeFirst.

11:33 PM March 2015 Meeting edited by Simon Fraser
Organizing the topics (diff)
11:31 PM March 2015 Meeting edited by Simon Fraser
11:22 PM Changeset in webkit [181433] by bshafiei@apple.com
  • 2 edits in branches/safari-600.1.4.15-branch/Source/WebCore

Merged patch for rdar://problem/20128911.

11:18 PM Changeset in webkit [181432] by bshafiei@apple.com
  • 2 edits in branches/safari-600.5-branch/Source/WebCore

Merged patch for rdar://problem/20128856.

11:05 PM Changeset in webkit [181431] by bshafiei@apple.com
  • 31 edits
    5 copies in tags/Safari-601.1.22/Source

Merged r181423. rdar://problem/11903887

10:57 PM Changeset in webkit [181430] by bshafiei@apple.com
  • 5 edits in branches/safari-600.1.4.15-branch/Source


10:57 PM Changeset in webkit [181429] by bshafiei@apple.com
  • 5 edits in branches/safari-600.5-branch/Source


10:00 PM Changeset in webkit [181428] by roger_fong@apple.com
  • 2 edits in trunk/Source/WebCore

Media element time no longer updates while scrubbing following r181279.

Reviewed by Eric Carlson.

  • Modules/mediacontrols/mediaControlsApple.js:

Update time as a result of a mouse move if we are scrubbing.

9:53 PM Changeset in webkit [181427] by Brent Fulgham
  • 2 edits
    1 add in trunk/LayoutTests

[Win] Update test expectations after r181408.

Also update with more debug assertions.

  • platform/win/TestExpectations:
  • platform/win/fast/forms/input-auto-fill-button-expected.txt: Added.
9:49 PM Changeset in webkit [181426] by Joseph Pecoraro
  • 11 edits
    4 adds in trunk

Web Inspector: CSS parser errors in the console should include column numbers

Reviewed by Darin Adler.


Test: inspector-protocol/console/warnings-errors.html

  • css/CSSParser.h:

Get the current character offset depending on the source type.
Add instance variables to track column position and start
line / column for inline stylesheets.

  • css/CSSParser.cpp:

Initialize new instance variables.

Update to include column information for the token. Also, if we are on the
first line we may need to take into account a start column offset as well.

Set the token's start column.
When bumping the line number, reset the column offset for the next
line with the next character.

Include column information.

  • css/StyleSheetContents.cpp:

Include column information.

  • css/StyleSheetContents.h:
  • dom/InlineStyleSheetOwner.cpp:

Save and use column information later on.

  • dom/InlineStyleSheetOwner.h:
  • inspector/InspectorStyleSheet.cpp:

Updated parser signature needs starting column and no longer has optional parameters.


  • UserInterface/Views/ConsoleMessageImpl.js:

Column numbers in console messages are also 1 based and should be adjusted.


Test errors in both external and inline CSS and Scripts to ensure they have
expected line:column information.

  • inspector-protocol/console/warnings-errors-expected.txt: Added.
  • inspector-protocol/console/warnings-errors.html: Added.
  • inspector-protocol/resources/errors.css: Added.
  • inspector-protocol/resources/errors.js: Added.
8:27 PM Changeset in webkit [181425] by Simon Fraser
  • 1 edit
    1 move in trunk/LayoutTests

Rename the Mavericks result for input-auto-fill-button.html to use the
correct name.

  • platform/mac-mavericks/fast/forms/input-auto-fill-button-expected.txt: Renamed from LayoutTests/platform/mac-mavericks/fast/forms/input-auto-fill-button.txt.
7:20 PM March 2015 Meeting edited by Chris Dumez
7:16 PM March 2015 Meeting edited by Chris Dumez
Add WebKit Network Cache topic (diff)
7:01 PM Changeset in webkit [181424] by Antti Koivisto
  • 3 edits in trunk/Source/WebKit2

Cache traversal missing large entries

Reviewed by Andreas Kling.

When traversing cache entries we only read the header data from disk. This didn't work
correctly for entries larger than the read size.

  • NetworkProcess/cache/NetworkCacheIOChannelCocoa.mm:


Partial reads only receive one callback with 'done' set to true.
Ensure we actually return data in this case.

  • NetworkProcess/cache/NetworkCacheStorage.cpp:


Add some logging.

6:54 PM Changeset in webkit [181423] by eric.carlson@apple.com
  • 31 edits
    5 adds in trunk/Source

[Mac] Update AirPlay handling

Reviewed by Sam Weinig.


  • WebCore.xcodeproj/project.pbxproj:
  • dom/Document.cpp:


  • dom/Document.h:
  • html/HTMLMediaElement.cpp:


  • html/HTMLMediaElement.h:
  • html/HTMLMediaSession.cpp:


  • html/HTMLMediaSession.h:
  • page/ChromeClient.h:
  • page/Page.cpp:


  • page/Page.h:


  • platform/audio/MediaSession.cpp:

(WebCore::MediaSession::wirelessRoutesAvailableDidChange): Deleted.

  • platform/audio/MediaSession.h:


  • platform/audio/MediaSessionManager.cpp:

(WebCore::MediaSessionManager::wirelessRoutesAvailableChanged): Deleted.

  • platform/audio/MediaSessionManager.h:


  • platform/audio/ios/MediaSessionManagerIOS.h:
  • platform/audio/ios/MediaSessionManagerIOS.mm:

(-[WebMediaSessionHelper wirelessRoutesAvailableDidChange:]):

  • platform/graphics/AVPlaybackTarget.h: Added.


  • platform/graphics/AVPlaybackTargetPickerClient.h: Added.


  • platform/graphics/MediaPlayer.cpp:


  • platform/graphics/MediaPlayer.h:
  • platform/graphics/MediaPlayerPrivate.h:


  • platform/graphics/avfoundation/AVPlaybackTargetMac.mm: Added.


  • platform/graphics/avfoundation/objc/MediaPlayerPrivateAVFoundationObjC.h:
  • platform/graphics/avfoundation/objc/MediaPlayerPrivateAVFoundationObjC.mm:

(-[WebCoreAVFMovieObserver observeValueForKeyPath:ofObject:change:context:]):


  • Shared/WebCoreArgumentCoders.h:
  • Shared/mac/WebCoreArgumentCodersMac.mm:


  • UIProcess/PageClient.h:
  • UIProcess/WebAVPlaybackTargetPickerProxy.cpp: Added.


  • UIProcess/WebAVPlaybackTargetPickerProxy.h: Added.


  • UIProcess/WebPageProxy.cpp:


  • UIProcess/WebPageProxy.h:
  • UIProcess/WebPageProxy.messages.in:
  • UIProcess/mac/PageClientImpl.h:
  • UIProcess/mac/PageClientImpl.mm:


  • UIProcess/mac/WebAVPlaybackTargetPickerProxyMac.h: Added.
  • UIProcess/mac/WebAVPlaybackTargetPickerProxyMac.mm: Added.

(-[WebAVOutputDevicePickerMenuControllerHelper initWithCallback:]):
(-[WebAVOutputDevicePickerMenuControllerHelper clearCallback]):
(-[WebAVOutputDevicePickerMenuControllerHelper observeValueForKeyPath:ofObject:change:context:]):

  • WebKit2.xcodeproj/project.pbxproj:
  • WebProcess/WebCoreSupport/WebAVPlaybackTargetPicker.cpp: Added.


  • WebProcess/WebCoreSupport/WebChromeClient.cpp:


  • WebProcess/WebCoreSupport/WebChromeClient.h:
  • WebProcess/WebPage/WebPage.h:
  • WebProcess/WebPage/WebPage.messages.in:
  • WebProcess/WebPage/mac/WebPageMac.mm:


  • WebProcess/com.apple.WebProcess.sb.in:
6:10 PM Changeset in webkit [181422] by Simon Fraser
  • 2 edits in trunk/LayoutTests

Mark compositing/backgrounds/background-image-with-negative-zindex.html as
an image failure, since r179871 was rolled out.

  • platform/mac/TestExpectations:
6:03 PM Changeset in webkit [181421] by commit-queue@webkit.org
  • 17 edits
    1 copy
    1 add in trunk

[Content Extensions] Add resource type and load type triggers.

Patch by Alex Christensen <achristensen@webkit.org> on 2015-03-11
Reviewed by Benjamin Poulain.


New API tests.

  • WebCore.xcodeproj/project.pbxproj:
  • contentextensions/ContentExtensionCompiler.cpp:


  • contentextensions/ContentExtensionParser.cpp:


  • contentextensions/ContentExtensionRule.h:
  • contentextensions/ContentExtensionsBackend.cpp:

(WebCore::ContentExtensions::ContentExtensionsBackend::actionsForURL): Deleted.

  • contentextensions/ContentExtensionsBackend.h:
  • contentextensions/DFABytecode.h:


  • contentextensions/DFABytecodeCompiler.cpp:


  • contentextensions/DFABytecodeCompiler.h:
  • contentextensions/DFABytecodeInterpreter.cpp:


  • contentextensions/DFABytecodeInterpreter.h:
  • loader/ResourceLoadInfo.cpp: Added.


  • loader/ResourceLoadInfo.h: Added.
  • loader/cache/CachedResourceLoader.cpp:


  • page/UserContentController.cpp:

(WebCore::UserContentController::actionsForURL): Deleted.

  • page/UserContentController.h:


  • TestWebKitAPI/Tests/WebCore/ContentExtensions.cpp:

(TestWebKitAPI::testURL): Deleted.

5:38 PM Changeset in webkit [181420] by timothy_horton@apple.com
  • 2 edits in trunk/Source/WebCore

Fix the build.

  • page/EventHandler.cpp:


5:35 PM Changeset in webkit [181419] by rniwa@webkit.org
  • 4 edits in trunk/Source/JavaScriptCore

"static" should not be a reserved keyword in non-strict mode even when ES6 class is enabled

Reviewed by Mark Lam.

Make "static" RESERVED_IF_STRICT and manually detect it in parseClass.

No new tests. This is already checked by js/reserved-words.html and js/keywords-and-reserved_words.html

  • parser/Keywords.table:
  • parser/Parser.cpp:


  • parser/ParserTokens.h:
5:33 PM Changeset in webkit [181418] by timothy_horton@apple.com
  • 6 edits
    8 adds in trunk

<attachment> shouldn't use "user-select: all"

Reviewed by Darin Adler.

It turns out that "user-select: all" is rife with bugs; in lieu of fixing them
all (at least for now), let's not use "user-select: all" in the default stylesheet
for <attachment>. It's really overkill anyway, since <attachment> can't have children.
The only "user-select: all" behavior we actually want is select-on-click.
So, we'll implement that in a slightly different way.

Tests: fast/attachment/attachment-select-on-click-inside-user-select-all.html


  • css/html.css:

No more "user-select: all".

(attachment:focus): Deleted.
We stopped using attachment focus a while back and forgot to remove this.

  • dom/Node.h:

Add a virtual function that Node subclasses can override to indicate they
should be selected on mouse down.

  • html/HTMLAttachmentElement.h:

Override the aforementioned virtual function; <attachment> should always
be selected on mouse down.

  • page/EventHandler.cpp:

Determine which node should be selected when a mousedown hits the given node.
If there's any "user-select: all", we go with the outermost "user-select: all".
Otherwise, we give the node a chance to say that it wants to be selected itself.

Rename this function, it's not just about "user-select: all" anymore.
Make use of nodeToSelectOnMouseDownForNode.

(WebCore::expandSelectionToRespectUserSelectAll): Deleted.
Adjust to the new names.

  • fast/attachment/attachment-select-on-click-inside-user-select-all.html: Added.
  • fast/attachment/attachment-select-on-click.html: Added.
  • platform/mac/fast/attachment/attachment-select-on-click-expected.png: Added.
  • platform/mac/fast/attachment/attachment-select-on-click-expected.txt: Added.
  • platform/mac/fast/attachment/attachment-select-on-click-inside-user-select-all-expected.png: Added.
  • platform/mac/fast/attachment/attachment-select-on-click-inside-user-select-all-expected.txt: Added.
  • platform/mac-mavericks/fast/attachment/attachment-select-on-click-inside-user-select-all-expected.txt: Added.
  • platform/mac-mavericks/fast/attachment/attachment-select-on-click-expected.txt: Added.

Add two tests. One, for the basic functionality of clicking on an
<attachment> to select it. The second, to test that clicking on an
<attachment> inside a larger "user-select: all" element still selects
the whole "user-select: all" element.

5:29 PM Changeset in webkit [181417] by bshafiei@apple.com
  • 1 copy in tags/Safari-600.

New tag.

5:28 PM Changeset in webkit [181416] by bshafiei@apple.com
  • 1 copy in tags/Safari-600.5.12

New tag.

5:20 PM March 2015 Meeting edited by abaldeva@ea.com
5:15 PM Changeset in webkit [181415] by ggaren@apple.com
  • 4 edits in trunk/Source/WebCore

Users of Heap::deprecatedReportExtraMemory should switch to reportExtraMemoryAllocated+reportExtraMemoryVisited

Reviewed by Andreas Kling.

Fixed this bug for canvas.

  • html/HTMLCanvasElement.cpp:

(WebCore::HTMLCanvasElement::memoryCost): Factored out the helper function
required by our IDL generator.

(WebCore::HTMLCanvasElement::createImageBuffer): Use

  • html/HTMLCanvasElement.h:
  • html/HTMLCanvasElement.idl: Adopt the IDL for reporting cost in the

right way during GC. This will match our reportExtraMemoryAllocated
with a reportExtraMemoryVisited during GC.

4:52 PM Changeset in webkit [181414] by mmirman@apple.com
  • 2 edits in trunk/LayoutTests

Update windows test results

Unreviewed followup to r181353

  • platform/win/fast/events/event-attribute-expected.txt: update windows test results.
4:28 PM Changeset in webkit [181413] by roger_fong@apple.com
  • 3 edits in trunk/Source/WebCore

A number of minor edits to the media controls on OSX.
Reviewed by Darin Adler.
This covers a slew of minor edits to the new media controls. They are as follows.
Small vertical placements adjustments to inline control elements.
Make sure buttons have no focus outlines.
Expand height of mute box that triggers the volume panel appearing.
Turn all button colors into an slightly transparent white.
Center status display text in fullscreen mode.
Lower position of captions container in fullscreen mode.
Show the controls on when done loading of the video an status display is hidden.

  • Modules/mediacontrols/mediaControlsApple.css:

(audio::-webkit-media-controls-panel button:focus):
(audio::-webkit-media-controls-panel .mute-box):
(audio::-webkit-media-controls-panel .volume-box):
(audio::-webkit-media-controls-closed-captions-container li.selected:hover::before):
(video:-webkit-full-screen::-webkit-media-controls-panel .volume-box):
(audio::-webkit-media-controls-panel button:active): Deleted.

  • Modules/mediacontrols/mediaControlsApple.js:
3:52 PM Changeset in webkit [181412] by commit-queue@webkit.org
  • 11 edits in trunk/Source/WebCore

Unreviewed, rolling out r179340 and r179344.

Caused images to stay alive forever when navigating away from
the page before they finish loading. (Requested by kling on

Reverted changesets:

"CachedImage: ensure clients overrides imageChanged instead of

"HTMLImageLoader: fix build failure on assert condition after

3:32 PM Changeset in webkit [181411] by ggaren@apple.com
  • 13 edits in trunk/Source

Many users of Heap::reportExtraMemory* are wrong, causing lots of memory growth

Reviewed by Andreas Kling.

Adopt deprecatedReportExtraMemory as a short-term fix for runaway
memory growth in these cases where we have not adopted

Long-term, we should use reportExtraMemoryAllocated+reportExtraMemoryVisited.
That's tracked by https://bugs.webkit.org/show_bug.cgi?id=142595.


  • API/JSBase.cpp:


  • runtime/SparseArrayValueMap.cpp:



Using IOSDebug, I can see that the canvas stress test @ http://jsfiddle.net/fvyw4ba0/,
which used to keep > 1000 1MB NonVolatile GPU allocations live, now keeps about 10 live.

  • Modules/mediasource/SourceBuffer.cpp:


  • bindings/js/JSDocumentCustom.cpp:


  • bindings/js/JSImageDataCustom.cpp:


  • bindings/js/JSNodeListCustom.cpp:


  • dom/CollectionIndexCache.cpp:


  • html/HTMLCanvasElement.cpp:


  • html/HTMLImageLoader.cpp:


  • html/HTMLMediaElement.cpp:


  • xml/XMLHttpRequest.cpp:


3:23 PM Changeset in webkit [181410] by mmaxfield@apple.com
  • 2 edits in trunk/Source/WebKit2

Fix the build


  • Shared/Downloads/mac/DownloadMac.mm:


3:02 PM Changeset in webkit [181409] by aestes@apple.com
  • 4 edits in trunk

REGRESSION (r180985): contentfiltering/block-after-add-data.html crashes with GuardMalloc

Reviewed by Darin Adler.


  • loader/DocumentLoader.cpp:

(WebCore::DocumentLoader::dataReceived): Don't delete m_contentFilter until after we're done using its replacement data.


3:01 PM Changeset in webkit [181408] by weinig@apple.com
  • 44 edits
    5 adds in trunk

Allow adding a button in input elements for auto fill related functionality

Reviewed by Anders Carlsson.


Test: fast/forms/input-auto-fill-button.html

  • Adds a new button that can be shown in <input> elements - AutoFillButtonElement.
  • Makes the spelling of AutoFill consistent throughout WebCore and WebKit (except where not feasible due to exported API/SPI).
  • CMakeLists.txt:
  • WebCore.vcxproj/WebCore.vcxproj:
  • WebCore.vcxproj/WebCore.vcxproj.filters:
  • WebCore.xcodeproj/project.pbxproj:

Add new files.

  • css/html.css:

Add default style rules for the AutoFill button based on the ones
used for caps lock indicator.

  • html/HTMLInputElement.cpp:

(WebCore::HTMLInputElement::setAutofilled): Deleted.

  • html/HTMLInputElement.h:

(WebCore::HTMLInputElement::isAutofilled): Deleted.
Add new bit to represent whether the AutoFill button should be shown
or not. By default it is not shown, and the client must enable it.

  • html/InputType.cpp:


  • html/InputType.h:

Add new virtual methods for updating and accessing the AutoFill button.

  • html/TextFieldInputType.cpp:


  • html/TextFieldInputType.h:

Add support for adding the AutoFill to the shadow DOM of textfields. The implementation
is slightly different than for the caps lock indicator, because I didn't want to force
the creation of a container for all <input> elements just in case an AutoFill button was
added. Instead, if an AutoFill button is added, the container is created on the fly and
the existing DOM is updated to move into it. Once a container is created, it is never

  • html/shadow/AutoFillButtonElement.cpp: Added.


  • html/shadow/AutoFillButtonElement.h: Added.

Add div subclass that swallows the click event and forwards it to the ChromeClient.

  • page/ChromeClient.h:

Add handleAutoFillButtonClick client function to inform WebKit that the AutoFill
button was clicked.

  • testing/Internals.cpp:


  • testing/Internals.h:
  • testing/Internals.idl:

Expose a new internals.setShowAutoFillButton() function to allow testing
of the AutoFill button from layout tests.

  • accessibility/AccessibilityObject.cpp:


  • css/SelectorCheckerTestFunctions.h:


  • css/StyleResolver.cpp:

Update for new spelling of AutoFill.


  • DOM/WebDOMOperations.mm:

(-[DOMHTMLInputElement _setAutofilled:]):
Update for consistent spelling of AutoFill.


  • Adds new SPI to respond to clicks on the AutoFill button (both bundle and UIProcess).
  • Adds new SPI to enable/disable the display of the AutoFill button on an element.
  • Adds new SPI to get the bounds of the AutoFill button.
  • Updates spelling of AutoFill to be consistent.
  • UIProcess/API/APIUIClient.h:


  • UIProcess/API/C/WKPage.cpp:


  • UIProcess/API/C/WKPageUIClient.h:
  • UIProcess/WebPageProxy.cpp:


  • UIProcess/WebPageProxy.h:
  • UIProcess/WebPageProxy.messages.in:
  • WebProcess/InjectedBundle/API/APIInjectedBundlePageUIClient.h:


  • WebProcess/InjectedBundle/InjectedBundlePageUIClient.cpp:


  • WebProcess/InjectedBundle/InjectedBundlePageUIClient.h:
  • WebProcess/WebCoreSupport/WebChromeClient.cpp:


  • WebProcess/WebCoreSupport/WebChromeClient.h:
  • WebProcess/InjectedBundle/API/c/WKBundlePageUIClient.h:

Pipe the didClickAutoFillButton to the SPI layer.

  • WebProcess/InjectedBundle/API/Cocoa/WKWebProcessPlugInNodeHandle.mm:

(-[WKWebProcessPlugInNodeHandle HTMLInputElementIsAutoFilled]):
(-[WKWebProcessPlugInNodeHandle setHTMLInputElementIsAutoFilled:]):
Update for new spelling of AutoFill.

  • WebProcess/InjectedBundle/API/c/WKBundleNodeHandlePrivate.h:
  • WebProcess/InjectedBundle/API/c/WKBundleNodeHandle.cpp:

Added. Replaces WKBundleNodeHandleGetHTMLInputElementAutofilled/WKBundleNodeHandleSetHTMLInputElementAutofilled
which are now deprecated due to inconsistent spelling of AutoFill.



  • WebProcess/InjectedBundle/DOM/InjectedBundleNodeHandle.cpp:

(WebKit::InjectedBundleNodeHandle::isHTMLInputElementAutofilled): Deleted.
(WebKit::InjectedBundleNodeHandle::setHTMLInputElementAutofilled): Deleted.

  • WebProcess/InjectedBundle/DOM/InjectedBundleNodeHandle.h:

Fix-up InjectedBundleNodeHandle to store its underlying Node in a Ref, rather than a RefPtr.
Add helpers to implement the new SPI for WKBundleNodeHandleRef.


  • WebKitTestRunner/TestController.cpp:

Update for new callback.


  • fast/forms/input-auto-fill-button-expected.txt: Added.
  • fast/forms/input-auto-fill-button.html: Added.
2:43 PM March 2015 Meeting edited by achristensen@apple.com
2:29 PM Changeset in webkit [181407] by ggaren@apple.com
  • 29 edits in trunk/Source

Refactored the JSC::Heap extra cost API for clarity and to make some known bugs more obvious

Reviewed by Andreas Kling.


  • API/JSBase.cpp:

(JSReportExtraMemoryCost): Added a FIXME to annotate a known bug.

  • bytecode/CodeBlock.cpp:


  • bytecode/CodeBlock.h:

(JSC::CodeBlock::setJITCode): Updated for rename.

  • heap/Heap.cpp:

(JSC::Heap::deprecatedReportExtraMemorySlowCase): Renamed our reporting
APIs to clarify their relationship to each other: One must report extra
memory at the time of allocation, and at the time the GC visits it.

(JSC::Heap::willStartCollection): Updated for renames. Added explicit
API for deprecated users who can't use our best API.

(JSC::Heap::reportExtraMemoryCostSlowCase): Deleted.
(JSC::Heap::extraSize): Deleted.

  • heap/Heap.h:
  • heap/HeapInlines.h:

(JSC::Heap::reportExtraMemoryCost): Deleted. Ditto.

  • heap/SlotVisitor.h:
  • heap/SlotVisitorInlines.h:

(JSC::SlotVisitor::reportExtraMemoryUsage): Deleted. Moved this
functionality into the Heap since it's pretty detailed in its access
to the heap.

  • runtime/JSArrayBufferView.cpp:


  • runtime/JSGenericTypedArrayViewInlines.h:

(JSC::JSGenericTypedArrayView<Adaptor>::visitChildren): Updated for

  • runtime/JSString.cpp:


  • runtime/JSString.h:

(JSC::JSString::finishCreation): Updated for renames.

  • runtime/SparseArrayValueMap.cpp:

(JSC::SparseArrayValueMap::add): Added FIXME.

  • runtime/WeakMapData.cpp:

(JSC::WeakMapData::visitChildren): Updated for rename.


Updated for renames to JSC extra cost APIs.

Added FIXMEs to our 10 use cases that are currently wrong, including
canvas, which is the cause of https://bugs.webkit.org/show_bug.cgi?id=142457.

  • Modules/mediasource/SourceBuffer.cpp:

(WebCore::SourceBuffer::reportExtraMemoryCost): Deleted.

  • Modules/mediasource/SourceBuffer.h:
  • bindings/js/JSDocumentCustom.cpp:


  • bindings/js/JSImageDataCustom.cpp:


  • bindings/js/JSNodeListCustom.cpp:


  • bindings/scripts/CodeGeneratorJS.pm:


  • dom/CollectionIndexCache.cpp:

(WebCore::reportExtraMemoryCostForCollectionIndexCache): Deleted.

  • dom/CollectionIndexCache.h:


  • html/HTMLCanvasElement.cpp:


  • html/HTMLCollection.h:


  • html/HTMLImageLoader.cpp:


  • html/HTMLMediaElement.cpp:


  • xml/XMLHttpRequest.cpp:


2:14 PM Changeset in webkit [181406] by bshafiei@apple.com
  • 4 edits
    2 copies in branches/safari-600.1.4.15-branch

Merged r181351. rdar://problem/20123049

2:08 PM Changeset in webkit [181405] by benjamin@webkit.org
  • 10 edits in trunk

Add basic support for BOL and EOL assertions to the URL Filter parser

Patch by Benjamin Poulain <bpoulain@apple.com> on 2015-03-11
Reviewed by Alex Christensen.


This patch adds heavily restricted support for BOL and EOL to the URL filter parser.

Both assertions must be the first/last term of their pattern. Any advanced combination
results in a parsing error.

The BOL assertion is easy to represent: currently, any pattern starts at the beginning
of a line and the NFA are generated accordingly.

I had two options to represent the EOL assertion:
1) Add a new special transition on EOL.
2) Add a new vector of actions to the states, conditional to the EOL input.

I picked the first option to avoid growing every state by a vector
that would be empty in the vast majority of cases.

On the matching side, the interpreter was modified to support transitions on '\0'.
DFABytecodeInstruction::CheckValue now stops when running on a character after
the end of the string.

DFABytecodeInstruction::Jump gets two fixes: First we now account for the index
to avoid going past the end of the input. Second, stop on '\0' too... the reason
is that the unconditional jump is only used for fallback edges of the DFA, fallback
edge are not supposed to accept '\0'.

  • contentextensions/DFA.cpp:


  • contentextensions/DFABytecodeInterpreter.cpp:


  • contentextensions/DFANode.h:
  • contentextensions/NFA.cpp:


  • contentextensions/NFANode.h:
  • contentextensions/NFAToDFA.cpp:


  • contentextensions/URLFilterParser.cpp:



  • TestWebKitAPI/Tests/WebCore/ContentExtensions.cpp:


2:03 PM Changeset in webkit [181404] by rniwa@webkit.org
  • 6 edits in trunk

Calling super() in a base class results in a crash

Reviewed by Filip Pizlo.


The bug was caused by BytecodeGenerator trying to generate "super" expression inside the constructor of a base class.
Disallow that by keeping track of whether "super" has been used in the current scope or not (needsSuperBinding flag)
and then throwing a syntax error in parseFunctionInfo if it was used and the current scope wasn't the constructor of
a derived class.

  • parser/Parser.cpp:

(JSC::Parser<LexerType>::parseFunctionInfo): Don't allow super() or super.foo outside the constructor of a derived class.
(JSC::Parser<LexerType>::parseClass): Pass in the constructor kind to parseGetterSetter.
(JSC::Parser<LexerType>::parseGetterSetter): Ditto to parseFunctionInfo.
(JSC::Parser<LexerType>::parseMemberExpression): Set needsSuperBinding flag true on the containing scope.

  • parser/Parser.h:

(JSC::Scope::needsSuperBinding): Added.
(JSC::Scope::setNeedsSuperBinding): Added.


Added more test cases to an existing test.

  • js/class-syntax-super-expected.txt:
  • js/script-tests/class-syntax-super.js:
2:01 PM March 2015 Meeting edited by Simon Fraser
2:00 PM March 2015 Meeting edited by Simon Fraser
1:27 PM Changeset in webkit [181403] by jer.noble@apple.com
  • 5 edits in trunk/Source/WebCore

[Mac] Update fullscreen placeholder UI to use Vibrancy.

Reviewed by Eric Carlson.

Update the fullscreen placeholder with a translucent vibrant appearance
using NSVisualEffectView. Since NSVisuaEffectView is only available for
OS X 10.10 and above, wrap the new implementation in a version check and
retain the old implementation.

Drive-by: Update the strings for the placeholder view with new HI guidance
as well.

  • English.lproj/Localizable.strings:
  • platform/LocalizedStrings.cpp:


  • platform/mac/WebCoreFullScreenPlaceholderView.h:
  • platform/mac/WebCoreFullScreenPlaceholderView.mm:

(-[WebCoreFullScreenPlaceholderView setExitWarningVisible:]):

1:18 PM Changeset in webkit [181402] by bshafiei@apple.com
  • 4 edits
    2 copies in branches/safari-600.5-branch

Merged r181351. <rdar://problem/20123022>

1:08 PM Changeset in webkit [181401] by Said Abou-Hallawa
  • 3 edits in trunk/LayoutTests

svg/animations/smil-leak-*.svg tests are flaky.


Some svg tests were flaky on mac only but they became flaky on all ports
after r181345. They pass locally and on EWS but not on bots. Mark them to
be flaky on all ports till we figure out the problem with these tests.

12:16 PM Changeset in webkit [181400] by timothy_horton@apple.com
  • 6 edits in trunk/Source

Make it possible to zoom on pages that claim to lay out to device size and then fail to do so

Reviewed by Simon Fraser.

  • page/ViewportConfiguration.cpp:

Rename m_ignoreScalingConstraints to m_canIgnoreScalingConstraints,
because it being true does not guarantee that we will ignore scaling constraints,
but it being false does guarantee that we won't.

Ignore scaling constraints if the page:

a) claimed to want to lay out to device-width and then laid out too wide
b) claimed to want to lay out to device-height and then laid out too tall
c) claimed to want to lay out with initialScale=1 and then laid out too wide

Call shouldIgnoreScalingConstraints() instead of looking at the local, so we can
have some more logic here (as above).

Dump whether we're ignoring scaling constraints.

Use WTFLogAlways so that the output goes to various other logging mechanisms
instead of just stderr.

  • page/ViewportConfiguration.h:

(WebCore::ViewportConfiguration::setIgnoreScalingConstraints): Deleted.

  • Shared/WebPreferencesDefinitions.h:
  • WebProcess/WebPage/WebPage.cpp:

Flip the pref on by default.

12:14 PM Changeset in webkit [181399] by Carlos Garcia Campos
  • 2 edits in trunk/Tools

[GTK] Add support for handling TLS errors to MiniBrowser

Reviewed by Sergio Villar Senin.

It's useful for testing TLS errors handling itself, but also to
allow ignoring TLS errors when testing.

  • MiniBrowser/gtk/BrowserWindow.c:


11:26 AM March 2015 Meeting edited by Simon Fraser
11:25 AM March 2015 Meeting edited by Simon Fraser
11:22 AM Changeset in webkit [181398] by mmaxfield@apple.com
  • 35 edits in trunk

Use out-of-band messaging for RenderBox::firstLineBaseline() and RenderBox::inlineBlockBaseline()

Reviewed by David Hyatt.


Currently, RenderBox::firstLineBaseline() and RenderBox::inlineBlockBaseline() return -1 to mean
that its baseline should be skipped. Instead of using this sentinel value, this patch changes the
return type from int to Optional<int>.

No new tests because there is no behavior change.

  • rendering/RenderBlock.cpp:


  • rendering/RenderBlock.h:
  • rendering/RenderBlockFlow.cpp:


  • rendering/RenderBlockFlow.h:
  • rendering/RenderBox.h:


  • rendering/RenderDeprecatedFlexibleBox.cpp:


  • rendering/RenderFlexibleBox.cpp:


  • rendering/RenderFlexibleBox.h:
  • rendering/RenderMenuList.h:
  • rendering/RenderTable.cpp:


  • rendering/RenderTable.h:
  • rendering/RenderTableCell.cpp:


  • rendering/RenderTableSection.cpp:


  • rendering/RenderTableSection.h:
  • rendering/RenderTextControl.h:
  • rendering/mathml/RenderMathMLBlock.cpp:


  • rendering/mathml/RenderMathMLBlock.h:
  • rendering/mathml/RenderMathMLFraction.cpp:


  • rendering/mathml/RenderMathMLFraction.h:
  • rendering/mathml/RenderMathMLOperator.cpp:


  • rendering/mathml/RenderMathMLOperator.h:
  • rendering/mathml/RenderMathMLRoot.cpp:


  • rendering/mathml/RenderMathMLRoot.h:
  • rendering/mathml/RenderMathMLRow.cpp:


  • rendering/mathml/RenderMathMLScripts.cpp:


  • rendering/mathml/RenderMathMLScripts.h:
  • rendering/mathml/RenderMathMLSpace.cpp:


  • rendering/mathml/RenderMathMLSpace.h:
  • rendering/mathml/RenderMathMLUnderOver.cpp:


  • rendering/mathml/RenderMathMLUnderOver.h:


Provide a callback to Optional::valueOrCompute() which is evaluated only if necessary.

  • wtf/Optional.h:



Test Optional::valueOrCompute().

  • TestWebKitAPI/Tests/WTF/Optional.cpp:


11:21 AM Changeset in webkit [181397] by timothy_horton@apple.com
  • 5 edits in trunk

<attachment>s should be created when dropping files onto contentEditable areas

Reviewed by Anders Carlsson.

Covered by existing tests.

  • editing/mac/EditorMac.mm:

Instead of inserting the dropped URLs as strings, make an <attachment>
for each.

  • editing/pasteboard/drag-files-to-editable-element-expected.txt:
  • editing/pasteboard/drag-files-to-editable-element.html:

Update the test to expect <attachment>s instead of filenames.

11:13 AM Changeset in webkit [181396] by hyatt@apple.com
  • 6 edits in trunk

Optimize offsetWidth and offsetHeight to avoid doing layouts.

Reviewed by Beth Dakin.


  • dom/Document.cpp:


  • dom/Document.h:

Added a new method that only updates layout if it determines that the desired dimensions are out
of date.

  • dom/Element.cpp:

Patch offsetWidth and offsetHeight to call the new method rather than updateLayoutIgnorePendingStylesheets.


  • fast/images/repaint-subrect-grid.html:

Patch this test to use the window.internals update layout method rather than
relying on document.body.offsetWidth.

11:08 AM Changeset in webkit [181395] by Carlos Garcia Campos
  • 6 edits in trunk

[GTK] Add an option to enable MiniBrowser for non developer builds and always install it

Reviewed by Gustavo Noronha Silva.


Add ENABLE_MINIBROWSER option, enabled by default for development
builds and disabled for production builds unless explicilty enabled.

  • Source/cmake/OptionsGTK.cmake:


  • CMakeLists.txt: Build testing tools only for developer builds,

but MiniBrowser when ENABLE_MINIBROWSER option is ON.

  • MiniBrowser/gtk/CMakeLists.txt: Only add

-DWEBKIT_INJECTED_BUNDLE_PATH to the build for developer builds,
and add a rule to install the MiniBrowser.

  • MiniBrowser/gtk/main.c:

(main): Only set WEBKIT_INJECTED_BUNDLE_PATH env var for developer

10:57 AM Changeset in webkit [181394] by bshafiei@apple.com
  • 11 edits
    4 deletes in tags/Safari-601.1.22

Merged r181389.

10:49 AM Changeset in webkit [181393] by Brent Fulgham
  • 5 edits
    1 add in trunk/Tools

Add some DateMath tests to TestWebKitAPI

Reviewed by Alexey Proskuryakov.

  • TestWebKitAPI/CMakeLists.txt: Add new DateMath.cpp file.
  • TestWebKitAPI/TestWebKitAPI.vcxproj/TestWebKitAPI.vcxproj: Ditto.
  • TestWebKitAPI/TestWebKitAPI.vcxproj/TestWebKitAPI.vcxproj.filters: Ditto.
  • TestWebKitAPI/TestWebKitAPI.xcodeproj/project.pbxproj: Ditto."
  • TestWebKitAPI/Tests/WTF/DateMath.cpp: Added.
10:47 AM Changeset in webkit [181392] by Carlos Garcia Campos
  • 5 edits in trunk

[GTK] Do not look for child processes in the UI process binary path

Reviewed by Gustavo Noronha Silva.


  • Source/cmake/OptionsGTK.cmake: Add -DDEVELOPMENT_BUILD=1 to the

build for development builds.


It's only useful for internal tools and tests, but never when
installed, since we don't install the processes in the bin dir but
in the libexec dir.

  • Shared/gtk/ProcessExecutablePathGtk.cpp:

(WebKit::findWebKitProcess): Only look or the executables in the
UI process binary path or WEBKIT_EXEC_PATH for development builds.

  • UIProcess/API/gtk/WebKitWebContext.cpp:

(injectedBundleDirectory): Only check
WEBKIT_INJECTED_BUNDLE_PATH env var for development builds.

9:38 AM Changeset in webkit [181391] by mmaxfield@apple.com
  • 1 edit
    7 adds in trunk/LayoutTests

[Win] Test gardening after r181260


  • platform/win/svg/text/select-textLength-spacingAndGlyphs-squeeze-1-expected.txt: Added.
  • platform/win/svg/text/select-textLength-spacingAndGlyphs-squeeze-2-expected.txt: Added.
  • platform/win/svg/text/select-textLength-spacingAndGlyphs-squeeze-3-expected.txt: Added.
  • platform/win/svg/text/select-textLength-spacingAndGlyphs-squeeze-4-expected.txt: Added.
  • platform/win/svg/text/select-textLength-spacingAndGlyphs-stretch-1-expected.txt: Added.
  • platform/win/svg/text/select-textLength-spacingAndGlyphs-stretch-2-expected.txt: Added.
  • platform/win/svg/text/select-textLength-spacingAndGlyphs-stretch-3-expected.txt: Added.
9:31 AM Changeset in webkit [181390] by mmaxfield@apple.com
  • 1 edit
    3 copies
    2 adds
    1 delete in trunk/LayoutTests

Test gardening after r181278


Until Windows adopts the SVG -> OTF Converter, we should mark this test as Mac-only.

  • platform/mac/svg/fonts/resources/svg-font-general.svg: Renamed from LayoutTests/svg/fonts/resources/svg-font-general.svg.
  • platform/mac/svg/fonts/svg-font-general-expected.html: Renamed from LayoutTests/svg/fonts/svg-font-general-expected.html.
  • platform/mac/svg/fonts/svg-font-general.html: Renamed from LayoutTests/svg/fonts/svg-font-general.html.
8:58 AM Changeset in webkit [181389] by commit-queue@webkit.org
  • 11 edits
    4 deletes in trunk

Unreviewed, rolling out r181367.

Caused crashes on the debug bots (Requested by cdumez on

Reverted changeset:

"Web Inspector: CSS parser errors in the console should
include column numbers"

7:52 AM Changeset in webkit [181388] by commit-queue@webkit.org
  • 2 edits in trunk/Source/WebInspectorUI

Web Inspector: Reload after Global Search results in empty Resources Sidebar.

Patch by Joseph Pecoraro <Joseph Pecoraro> on 2015-03-11
Reviewed by Timothy Hatcher.

If you reload when the search content tree outline is showing, break out
to show the resource tree outline.

  • UserInterface/Views/ResourceSidebarPanel.js:


7:41 AM Changeset in webkit [181387] by mmaxfield@apple.com
  • 3 edits
    2 adds in trunk

Inline block children do not have correct baselines if their children are also block elements

Patch by Myles C. Maxfield <mmaxfield@apple.com> on 2015-03-11
Reviewed by Darin Adler.


Perform the same computation on child block elements as child inline elements.

Test: fast/text/baseline-inline-block-block-children.html

  • rendering/RenderBlockFlow.cpp:



  • fast/text/baseline-inline-block-block-children-expected.html: Added.
  • fast/text/baseline-inline-block-block-children.html: Added.
7:09 AM Changeset in webkit [181386] by commit-queue@webkit.org
  • 3 edits in trunk/Source/WebInspectorUI

Web Inspector: SearchResultTreeElement.representedObject is missing a saveIdentityToCookie implementation

Patch by Joseph Pecoraro <Joseph Pecoraro> on 2015-03-11
Reviewed by Timothy Hatcher.

This will restore selection of a global search tree element if you
close and reopen the inspector in such a case.

  • UserInterface/Models/DOMSearchMatchObject.js:

(WebInspector.DOMSearchMatchObject.prototype.get resource):
Cookie has the resource URL, DOM Node title, and text range.

  • UserInterface/Models/SourceCodeSearchMatchObject.js:

(WebInspector.SourceCodeSearchMatchObject.prototype.get sourceCodeTextRange):
Cookie has the source code URL and text range.

7:00 AM Changeset in webkit [181385] by clopez@igalia.com
  • 3 edits in trunk/Source/WebCore

[CMake][GStreamer] Building EFL or GTK with ENABLE_VIDEO and without ENABLE_WEB_AUDIO is broken.

Reviewed by Carlos Garcia Campos.

No new tests, this is a build fix.

  • PlatformEfl.cmake: Include GSTREAMER_AUDIO_LIBRARIES on the link step both for ENABLE_VIDEO and ENABLE_WEB_AUDIO.
  • PlatformGTK.cmake: Idem.
4:03 AM Changeset in webkit [181384] by Carlos Garcia Campos
  • 5 edits in releases/WebKitGTK/webkit-2.8/Source/JavaScriptCore

Merge r181343 - Use std::numeric_limits<unsigned>::max() instead of (unsigned)-1.

Reviewed by Benjamin Poulain.

  • jit/JIT.cpp:


  • jit/JITInlines.h:


  • jit/SlowPathCall.h:


  • yarr/Yarr.h:
3:49 AM Changeset in webkit [181383] by Carlos Garcia Campos
  • 5 edits in releases/WebKitGTK/webkit-2.8/Source

Merge r181326 - GCC: CRASH() should be annotated with NORETURN

Patch by Michael Catanzaro <Michael Catanzaro> on 2015-03-10
Reviewed by Anders Carlsson.


Don't return from a NORETURN function. This used to avoid a warning from GCC, but now it
causes one.

  • jsc.cpp:


Add COMPILER(GCC) to #ifdefs that already exist for Clang.

  • wtf/Assertions.cpp:
  • wtf/Assertions.h:
3:23 AM Changeset in webkit [181382] by Carlos Garcia Campos
  • 3 edits in releases/WebKitGTK/webkit-2.8/Source/WebKit2

Merge r181324 - [GTK] Contents not shown when entering AC mode unless the window is resized

Reviewed by Žan Doberšek.

The problem is once again that we are now creating the redirected
X window in realize method. When entering AC mode we resize the
redirected window to the drawing area size. Since the size hasn't
changed from the drawing area point of view, the web process is
not notified. The WebProcess always uses the window size, instead
of the root layer size, to make sure it's in sync, see the comment
in LayerTreeHostGtk::compositeLayersToContext(). So, we need to
enforce a resize when we change the size of the redirected window
when entering AC mode.

  • UIProcess/API/gtk/WebKitWebViewBase.cpp:


  • UIProcess/DrawingAreaProxyImpl.h:


3:22 AM Changeset in webkit [181381] by Carlos Garcia Campos
  • 2 edits in releases/WebKitGTK/webkit-2.8/Source/WebCore

Merge r181323 - [GTK] GL_MAX_VARYING_FLOATS is not defined in OpenGL ES 2

Reviewed by Žan Doberšek.

Do not use GL_MAX_VARYING_FLOATS when platform is GTK+ and using
OpenGL ES 2.

  • platform/graphics/opengl/GraphicsContext3DOpenGLCommon.cpp:


3:21 AM Changeset in webkit [181380] by Carlos Garcia Campos
  • 2 edits in releases/WebKitGTK/webkit-2.8/Source/WebCore

Merge r181322 - Unreviewed. Fix GTK+ build with OpenGL ES 2 enabled.

Remove USE(OPENGL) ifdef from GLContext.cpp, since there's nothing
specific to OpenGL in that file, and everything depending on
configure options is already protected by USE(GLX) and USE(EGL)
ifdefs. This is causing linking issues when building with OpenGL
ES 2, because the header doesn't have the ifdef, and the cpp file
is not compiled leaving all the interface undefined.

Patch by José Dapena Paz <jdapena@igalia.com> on 2015-03-10
Rubber-stamped by Carlos Garcia Campos.

  • platform/graphics/GLContext.cpp:
3:20 AM Changeset in webkit [181379] by Carlos Garcia Campos
  • 4 edits in releases/WebKitGTK/webkit-2.8

Merge r181321 - [GTK] Add a configure option to build with OpenGL ES 2

Patch by Carlos Garcia Campos <cgarcia@igalia.com> and José Dapena Paz <jdapena@igalia.com> on 2015-03-10
Reviewed by Martin Robinson.


Add ENABLE_GLES2 option. It's disabled by default, but if passed
GLES2 is required and OpenGL is not even searched. Otherwise we
search for OpenGL as usual, using it only if present.

  • Source/cmake/OptionsGTK.cmake:


Build GLES or GL specific files depending on the build options.

  • PlatformGTK.cmake:
3:19 AM Changeset in webkit [181378] by Carlos Garcia Campos
  • 3 edits in releases/WebKitGTK/webkit-2.8/Source/WebCore

Merge r181320 - Shrink the CSSPropertyID enum type

Reviewed by Sam Weinig.

Specify uint16_t as the base type for the CSSPropertyID enum.
This is enough to cover all of the CSS properties (429 at this moment,
with static_assert covering future changes). It halves the enum type size,
from 4 bytes to 2, reducing the size of various CSSPropertyID containers.

No new tests -- no change in behavior.

  • css/CSSPrimitiveValue.cpp:

(WebCore::propertyName): Remove the unnecessary propertyID < 0 check.

  • css/makeprop.pl:
3:06 AM Changeset in webkit [181377] by Carlos Garcia Campos
  • 12 edits in releases/WebKitGTK/webkit-2.8/Source

Merge r181329 - bmalloc: tryFastMalloc shouldn't crash

Reviewed by Darin Adler.


Added support for tryMalloc.

We assume that non-x-large allocations always succeed, and we crash
otherwise, since normal allocation failure will just cause the next
non-try allocation or internal metadata allocation to fail, and it's
hard and not really useful to keep limping along after that. But
extra-large allocations can meaningfully fail, and we can recover.

  • bmalloc/Heap.cpp:


  • bmalloc/Heap.h: Added support for non-crashy x-large allocation.
  • bmalloc/VMAllocate.h:

(bmalloc::vmAllocate): Added support for non-crashy VM allocation.

  • bmalloc/bmalloc.h:

(bmalloc::api::free): Tried to clarify our behavior with some comments.
Unfortunately, calling what we do "malloc" is still not quite right, since
malloc returns null on failure and we don't.


  • wtf/FastMalloc.cpp:

(WTF::fastAlignedMalloc): Don't check for null. bmalloc automatically
crashes on allocation failure, and we'd rather not pay for an extra check.

(WTF::tryFastMalloc): Added an opt-out API to return null rather than
crashing, since some clients need this.

(WTF::tryFastRealloc): Deleted. Unused.

  • wtf/FastMalloc.h:
2:38 AM Changeset in webkit [181376] by Carlos Garcia Campos
  • 6 edits
    1 delete in releases/WebKitGTK/webkit-2.8/Source/WTF

Merge r180814 - Remove unused PossiblyNull

Reviewed by Andreas Kling.

  • WTF.vcxproj/WTF.vcxproj: Removed the file.
  • WTF.vcxproj/WTF.vcxproj.filters: Ditto.
  • WTF.xcodeproj/project.pbxproj: Ditto.
  • wtf/CMakeLists.txt: Ditto.
  • wtf/PossiblyNull.h: Removed.
  • wtf/FastMalloc.h: Moved everything to the left.

Moved member functions out of the TryMallocReturnValue class definition.
(WTF::TryMallocReturnValue::operator PossiblyNull<T>): Deleted.
(WTF::TryMallocReturnValue::getValue): Marked inline, changed to work
only with pointer types, not arbitrary non-pointer types.

2:11 AM Changeset in webkit [181375] by Carlos Garcia Campos
  • 7 edits
    1 add in releases/WebKitGTK/webkit-2.8/Source

Merge r181305 - 8-bit version of weakCompareAndSwap() can cause an infinite loop.

Reviewed by Filip Pizlo.


Added a test that exercises the 8-bit CAS from multiple threads. The threads
will contend to set bits in a large array of bytes using the CAS function.

  • API/tests/CompareAndSwapTest.cpp: Added.


  • API/tests/testapi.c:



Presently, Bitmap::concurrentTestAndSet() uses the 8-bit version of
weakCompareAndSwap() (which compares and swaps an uint8_t value).
Bitmap::concurrentTestAndSet() has a loop that checks if a bit in the
byte of interest has been set. If not, it will call the 8-bit CAS
function to set the bit.

Under the covers, for ARM, the 8-bit CAS function actually works with a
32-bit CAS. The 8-bit CAS will first fetch the 32-bit value in memory
that should contain the 8-bit value, and check if it contains the
expected byte. If the value in memory doesn't have the expected byte,
it will return early to its caller. The expectation is that the caller
will reload the byte from memory and call the 8-bit CAS again.

Unfortunately, this code path that returns early does not have a
compiler fence. Without a compiler fence, the C++ compiler can
optimize away the reloading of the expected byte value, leaving it
unchanged. As a result, we'll have a infinite loop here that checks a
value that will never change, and the loop will not terminate until the
value changes.

The fix is to eliminate the early return check in the 8-bit CAS, and
have it always call down to the 32-bit CAS. The 32-bit CAS has a
compiler fence which will prevent this issue.

  • wtf/Atomics.h:


1:45 AM Changeset in webkit [181374] by Carlos Garcia Campos
  • 16 edits
    1 copy
    1 add in releases/WebKitGTK/webkit-2.8/Source

Merge r181297 - Stale entries in WeakGCMaps are keeping tons of WeakBlocks alive unnecessarily.

Reviewed by Geoffrey Garen.

Prune stale entries from WeakGCMaps as part of every full garbage collection.
This frees up tons of previously-stuck WeakBlocks that were only sitting around
with finalized handles waiting to die.

Note that WeakGCMaps register/unregister themselves with the GC heap in their
ctor/dtor, so creating one now requires passing the VM.

Average time spent in the PruningStaleEntriesFromWeakGCMaps GC phase appears
to be between 0.01ms and 0.3ms, though I've seen a few longer ones at ~1.2ms.
It seems somewhat excessive to do this on every Eden collection, so it's only
doing work in full collections for now.

Because the GC may now mutate WeakGCMap below object allocation, I've made it
so that the classic HashMap::add() optimization can't be used with WeakGCMap.
This caused intermittent test failures when originally landed due to having
an invalid iterator on the stack after add() inserted a new entry and we
proceeded to allocate the new object, triggering GC.

  • API/JSWeakObjectMapRefInternal.h:


  • API/JSWeakObjectMapRefPrivate.cpp:
  • API/JSWrapperMap.mm:

(-[JSWrapperMap initWithContext:]):
(-[JSWrapperMap jsWrapperForObject:]): Pass VM to WeakGCMap constructor.

  • JavaScriptCore.xcodeproj/project.pbxproj: Add WeakGCMapInlines.h and make

it project-private so WebCore clients can access it.

  • heap/Heap.cpp:

(JSC::Heap::pruneStaleEntriesFromWeakGCMaps): Added a new GC phase for pruning
stale entries from WeakGCMaps. This is only executed during full collections.

  • heap/Heap.h:
  • heap/HeapInlines.h:

(JSC::Heap::unregisterWeakGCMap): Added a mechanism for WeakGCMaps to register
themselves with the Heap and provide a pruning callback.

  • runtime/PrototypeMap.h:


  • runtime/Structure.cpp:

(JSC::StructureTransitionTable::add): Pass VM to WeakGCMap constructor.

  • runtime/JSCInlines.h: Add "WeakGCMapInlines.h"
  • runtime/JSGlobalObject.cpp: Include "WeakGCMapInlines.h" so this builds.
  • runtime/JSString.cpp:


  • runtime/PrototypeMap.cpp:

(JSC::PrototypeMap::emptyObjectStructureForPrototype): Remove HashMap add()
optimization since it's not safe in the GC-managed WeakGCMap world.

  • runtime/VM.cpp:

(JSC::VM::VM): Pass VM to WeakGCMap constructor.

  • runtime/WeakGCMap.h:

(JSC::WeakGCMap::WeakGCMap): Deleted.
(JSC::WeakGCMap::gcMap): Deleted.
(JSC::WeakGCMap::gcMapIfNeeded): Deleted.

  • runtime/WeakGCMapInlines.h: Added.

(JSC::WeakGCMap::pruneStaleEntries): Moved ctor, dtor and pruning callback
to WeakGCMapInlines.h to fix interdependent header issues. Removed code that
prunes WeakGCMap at certain growth milestones and instead rely on the GC
callback for housekeeping.

1:44 AM Changeset in webkit [181373] by bshafiei@apple.com
  • 5 edits in trunk/Source


1:41 AM Changeset in webkit [181372] by bshafiei@apple.com
  • 1 copy in tags/Safari-601.1.22

New tag.

1:10 AM Changeset in webkit [181371] by Philippe Normand
  • 13 edits
    6 moves in trunk/Source/WebCore

Rename MediaStreamCenter to RealtimeMediaSourceCenter

Reviewed by Eric Carlson.

As per bug title, rename MediaStreamCenter to
RealtimeMediaSourceCenter, this class manages
RealtimeMediaSources, not MediaStreams. Some un-needed includes of
the header were also removed.

12:45 AM Changeset in webkit [181370] by Carlos Garcia Campos
  • 8 edits
    4 adds in releases/WebKitGTK/webkit-2.8

Merge r181292 - REGRESSION(r176978): Inline-blocks with overflowing contents have ascents that are too large

Reviewed by David Hyatt.


When we have an inline-block element, and we want to find its baseline (to lay out other
elements on the same line) we loop through the element's children and ask them what their
baselines are. The children use the location of the top of their last line to compute this
value. However, if the child has overflow-y, this might not be the correct calculation.

This behavior is in the spec: "The baseline of an 'inline-block' is the baseline of its last
line box in the normal flow, unless it has either no in-flow line boxes or if its 'overflow'
property has a computed value other than 'visible', in which case the baseline is the bottom
margin edge."

-- http://www.w3.org/TR/CSS21/visudet.html#leading

However, we believe that a better policy is, when overflow is not "visible," to place the
baseline at the bottom of the block if the contents overflowed in the Y direction, and to place
it at the bottom of the last line if the contents did not overflow in the Y direction. This is
partially consistent with previous behavior, and isn't too far from the spec to cause too many

Test: fast/css/inline-block-tricky-baselines.html


  • rendering/RenderBlockFlow.cpp:



Update expected results.

  • css3/flexbox/child-overflow-expected.html:
  • css3/flexbox/child-overflow.html:
  • fast/css/inline-block-tricky-baselines-expected.html: Added.
  • fast/css/inline-block-tricky-baselines.html: Added.
  • fast/forms/textfield-overflow-by-value-update-expected.txt:
  • fast/text/baseline-inline-block-expected.html: Added.
  • fast/text/baseline-inline-block.html: Added.
  • platform/mac/fast/forms/search-vertical-alignment-expected.txt:
12:41 AM Changeset in webkit [181369] by Carlos Garcia Campos
  • 4 edits in releases/WebKitGTK/webkit-2.8/Source/WebKit2

Merge r181272 - bmalloc: tryFastMalloc shouldn't crash

Reviewed by Anders Carlsson.

Part 1: Stop using tryFastRealloc.

  • Shared/ShareableBitmap.cpp:

(WebKit::ShareableBitmap::resize): Deleted.

  • Shared/ShareableBitmap.h: Removed the resize function because it has

no clients.

  • WebProcess/Plugins/PluginProxy.cpp:

(WebKit::PluginProxy::updateBackingStore): Changed to allocate a new
backing store instead of resizing the old one. This has three advantages:

(1) Might be more memory-efficient, since you don't have to keep the old
one around while allocating the new one.

(2) Avoids the overhead of realloc() copying the contents of the old
backing store even though we only want uninitialized memory.

(3) Makes resize failure consistent with initial allocation failure.
Previously, while initial allocation failure would set the backing store
to null, resize failure would keep the old wrong backing store and then
tell it not to paint. Now, resize failure also sets the backing store to

12:10 AM Changeset in webkit [181368] by gyuyoung.kim@samsung.com
  • 12 edits in trunk/Source/WebCore

Remove unnecessary create() factory functions

Reviewed by Darin Adler.

create() function which just returns new instance can be replaced with std::make_unique<>.

No new tests, no behavior changes.

  • bindings/js/JSCryptoKeySerializationJWK.cpp:


  • crypto/keys/CryptoKeyAES.cpp:


  • crypto/keys/CryptoKeyDataOctetSequence.h:
  • crypto/keys/CryptoKeyHMAC.cpp:


  • crypto/keys/CryptoKeySerializationRaw.cpp:


  • platform/audio/mac/CARingBuffer.cpp:

(WebCore::CARingBuffer::create): Deleted.

  • platform/audio/mac/CARingBuffer.h:
  • platform/graphics/avfoundation/AudioSourceProviderAVFObjC.mm:


  • testing/Internals.cpp:


  • testing/MicroTaskTest.cpp:

(WebCore::MicroTaskTest::create): Deleted.

  • testing/MicroTaskTest.h:


Note: See TracTimeline for information about the timeline view.