Timeline



Mar 1, 2015:

10:35 PM Changeset in webkit [180882] by Simon Fraser
  • 26 edits
    8 adds in trunk

Make clip-path work on <video>, <canvas> etc.
​https://bugs.webkit.org/show_bug.cgi?id=138684

Reviewed by Darin Adler.

Source/WebCore:

clip-path only worked in compositing layers on the painted contents of the layer,
and failed to clip children. Fix this by translating the clip path into a Path
which is set on a CA shape layer (for Mac and iOS), or painted into the
RenderLayerBacking's mask layer. There are two code paths:

  1. clip-path which is a <basic-shape> or <geometry-box>, and no mask.

Here we can use the optimal code path of converting the clip into a path
that is put onto a CAShapeLayer, which is then used as a mask. There is no
additional backing store.

  1. clip-path with an SVG reference, or clip-path combined with -webkit-mask:

Here we have to allocate backing store for the mask layer, and paint the
clip path (possibly with the mask).

We add GraphicsLayer::Type::Shape, and add a getter for the layer type.

Tests: compositing/masks/compositing-clip-path-and-mask.html

compositing/masks/compositing-clip-path-mask-change.html
compositing/masks/compositing-clip-path.html
compositing/masks/reference-clip-path-on-composited.html

  • platform/graphics/GraphicsLayer.cpp:

(WebCore::GraphicsLayer::GraphicsLayer): Store the type in the layer so the getter can return it.
(WebCore::GraphicsLayer::shapeLayerPath): Get and set the shape layer path.
(WebCore::GraphicsLayer::setShapeLayerPath): Ditto.
(WebCore::GraphicsLayer::shapeLayerWindRule): Get and set the shape layer wind rule.
(WebCore::GraphicsLayer::setShapeLayerWindRule): Ditto.

  • platform/graphics/GraphicsLayer.h:

(WebCore::GraphicsLayer::type): Expose the type.
(WebCore::GraphicsLayer::supportsLayerType): Allow the cross-platform code to use
shape layers when it knows they are available.
(WebCore::GraphicsLayer::needsClippingMaskLayer): Deleted. This was never used.

  • platform/graphics/GraphicsLayerClient.h: Align the bits (helps avoid typos). Add a

GraphicsLayerPaintClipPath phase.

  • platform/graphics/Path.h: Some exports since WK2 needs to encode Paths now.
  • platform/graphics/ca/GraphicsLayerCA.cpp:

(WebCore::GraphicsLayerCA::initialize): Make shape layers.
(WebCore::GraphicsLayerCA::setShapeLayerPath): Setter for the shape path. Sadly we
can't early return on unchanged paths yet.
(WebCore::GraphicsLayerCA::setShapeLayerWindRule):
(WebCore::GraphicsLayerCA::commitLayerChangesBeforeSublayers): Updates for shape path
and wind rule.
(WebCore::GraphicsLayerCA::updateShape):
(WebCore::GraphicsLayerCA::updateWindRule):

  • platform/graphics/ca/GraphicsLayerCA.h: Some new dirty bits for shape path and wind rule.
  • platform/graphics/ca/PlatformCALayer.h:
  • platform/graphics/ca/mac/PlatformCALayerMac.h:
  • platform/graphics/ca/mac/PlatformCALayerMac.mm: Got rid of lots of m_layer.get().

(PlatformCALayerMac::~PlatformCALayerMac):
(PlatformCALayerMac::setNeedsDisplay):
(PlatformCALayerMac::setNeedsDisplayInRect):
(PlatformCALayerMac::removeFromSuperlayer):
(PlatformCALayerMac::setSublayers):
(PlatformCALayerMac::removeAllSublayers):
(PlatformCALayerMac::appendSublayer):
(PlatformCALayerMac::insertSublayer):
(PlatformCALayerMac::replaceSublayer):
(PlatformCALayerMac::adoptSublayers):
(PlatformCALayerMac::addAnimationForKey):
(PlatformCALayerMac::removeAnimationForKey):
(PlatformCALayerMac::animationForKey):
(PlatformCALayerMac::setMask):
(PlatformCALayerMac::isOpaque):
(PlatformCALayerMac::setOpaque):
(PlatformCALayerMac::bounds):
(PlatformCALayerMac::setBounds):
(PlatformCALayerMac::position):
(PlatformCALayerMac::setPosition):
(PlatformCALayerMac::anchorPoint):
(PlatformCALayerMac::setAnchorPoint):
(PlatformCALayerMac::transform):
(PlatformCALayerMac::setTransform):
(PlatformCALayerMac::sublayerTransform):
(PlatformCALayerMac::setSublayerTransform):
(PlatformCALayerMac::setHidden):
(PlatformCALayerMac::setGeometryFlipped):
(PlatformCALayerMac::isDoubleSided):
(PlatformCALayerMac::setDoubleSided):
(PlatformCALayerMac::masksToBounds):
(PlatformCALayerMac::setMasksToBounds):
(PlatformCALayerMac::acceleratesDrawing):
(PlatformCALayerMac::setAcceleratesDrawing):
(PlatformCALayerMac::contents):
(PlatformCALayerMac::setContents):
(PlatformCALayerMac::setContentsRect):
(PlatformCALayerMac::setMinificationFilter):
(PlatformCALayerMac::setMagnificationFilter):
(PlatformCALayerMac::backgroundColor):
(PlatformCALayerMac::setBackgroundColor):
(PlatformCALayerMac::setBorderWidth):
(PlatformCALayerMac::setBorderColor):
(PlatformCALayerMac::opacity):
(PlatformCALayerMac::setOpacity):
(PlatformCALayerMac::copyFiltersFrom):
(PlatformCALayerMac::setName):
(PlatformCALayerMac::setSpeed):
(PlatformCALayerMac::setTimeOffset):
(PlatformCALayerMac::contentsScale):
(PlatformCALayerMac::setContentsScale):
(PlatformCALayerMac::cornerRadius):
(PlatformCALayerMac::setCornerRadius):
(PlatformCALayerMac::setEdgeAntialiasingMask):
(PlatformCALayerMac::shapeWindRule): New function.
(PlatformCALayerMac::setShapeWindRule): Ditto.
(PlatformCALayerMac::shapePath): Ditto.
(PlatformCALayerMac::setShapePath): Ditto.
(PlatformCALayer::isWebLayer):

  • platform/graphics/cg/PathCG.cpp:

(WebCore::Path::Path): nullptr.

  • rendering/RenderLayer.cpp:

(WebCore::RenderLayer::paintsWithClipPath): Return true if the clip path is painted.
(WebCore::RenderLayer::computeClipPath): Factor code that computes the clip path into this
function, so we can call it from RenderLayerBacking too.
(WebCore::RenderLayer::setupClipPath):
(WebCore::RenderLayer::paintLayerContents): We only want to apply the clip path
for painting when we're either painting a non-composited layer, or we're painting the
mask layer of a composited layer. We in the latter case, we just want to fill the clip
path with black, so re-use the paintChildClippingMaskForFragments() which does this.

  • rendering/RenderLayer.h: Align the bits, add PaintLayerPaintingCompositingClipPathPhase.
  • rendering/RenderLayerBacking.cpp:

(WebCore::RenderLayerBacking::~RenderLayerBacking):
(WebCore::RenderLayerBacking::updateConfiguration):
(WebCore::RenderLayerBacking::updateGeometry): Move mask updating into its own function.
(WebCore::RenderLayerBacking::updateMaskingLayerGeometry): If we're using the shape layer
code path, compute the Path and set it and the wind rule on the mask layer.
(WebCore::RenderLayerBacking::updateMaskingLayer): This is now more complex, as it has
to deal with combinations of clip-path and mask, some of which allow for the shape layer
mask, and we handle dynamic changes between these and painted masks.
(WebCore::RenderLayerBacking::paintingPhaseForPrimaryLayer): Include the GraphicsLayerPaintClipPath phase.
(WebCore::RenderLayerBacking::paintIntoLayer): Map GraphicsLayerPaintClipPath to PaintLayerPaintingCompositingClipPathPhase.
(WebCore::RenderLayerBacking::updateMaskLayer): Deleted.

  • rendering/RenderLayerBacking.h:

Source/WebKit2:

Support encode/decode for WebCore Path objects, which is done by traversing
the path.

  • Shared/WebCoreArgumentCoders.cpp:

(IPC::pathPointCountApplierFunction):
(IPC::pathEncodeApplierFunction):
(IPC::ArgumentCoder<Path>::encode):
(IPC::ArgumentCoder<Path>::decode):

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

(WebKit::applyPropertiesToLayer): Actually apply the path and wind rule to the shape layer.

  • Shared/mac/RemoteLayerTreeTransaction.h: Include path and wind rule in the layer properties.
  • Shared/mac/RemoteLayerTreeTransaction.mm:

(WebKit::RemoteLayerTreeTransaction::LayerProperties::LayerProperties):
(WebKit::RemoteLayerTreeTransaction::LayerProperties::encode): Encode shape and wind rule.
(WebKit::RemoteLayerTreeTransaction::LayerProperties::decode): Decode shape and wind rule.

  • WebProcess/WebPage/mac/PlatformCALayerRemote.cpp:

(WebKit::PlatformCALayerRemote::shapePath):
(WebKit::PlatformCALayerRemote::setShapePath):
(WebKit::PlatformCALayerRemote::shapeWindRule):
(WebKit::PlatformCALayerRemote::setShapeWindRule):

  • WebProcess/WebPage/mac/PlatformCALayerRemote.h:

LayoutTests:

Tests for various combinations of clip-path and mask, and dynamic changes
thereof.

  • compositing/masks/compositing-clip-path-and-mask-expected.html: Added.
  • compositing/masks/compositing-clip-path-and-mask.html: Added.
  • compositing/masks/compositing-clip-path-expected.html: Added.
  • compositing/masks/compositing-clip-path-mask-change-expected.html: Added.
  • compositing/masks/compositing-clip-path-mask-change.html: Added.
  • compositing/masks/compositing-clip-path.html: Added.
  • compositing/masks/reference-clip-path-on-composited-expected.html: Added.
  • compositing/masks/reference-clip-path-on-composited.html: Added.
8:33 PM Changeset in webkit [180881] by commit-queue@webkit.org
  • 2 edits in trunk/Source/WebCore

[Cairo] Implement Path::addEllipse
​https://bugs.webkit.org/show_bug.cgi?id=142144

Patch by Hunseop Jeong <Hunseop Jeong> on 2015-03-01
Reviewed by Gyuyoung Kim.

Add support for addEllipse method for platforms using cairo.

  • platform/graphics/cairo/PathCairo.cpp:

(WebCore::Path::addEllipse):

7:58 PM Changeset in webkit [180880] by bshafiei@apple.com
  • 5 edits in branches/safari-600.4.10-branch/Source

Versioning.

7:56 PM Changeset in webkit [180879] by gyuyoung.kim@samsung.com
  • 2 edits in trunk/LayoutTests

Unreviewed EFL gardening. Mark crash tests of webgl to CRASH.
WebGL isn't supported by EFL port now.

  • platform/efl/TestExpectations:
7:40 PM Changeset in webkit [180878] by bshafiei@apple.com
  • 1 copy in tags/Safari-600.4.10.6

New tag.

7:26 PM Changeset in webkit [180877] by bshafiei@apple.com
  • 30 edits in branches/safari-600.4.10-branch/Source/WebCore

Merged r180839. rdar://problem/20001723

6:59 PM Changeset in webkit [180876] by mitz@apple.com
  • 2 edits in trunk/Source/WebKit

Silence non-fatal errors about failing to create WebKitPluginHost.app and WebKitPluginAgent symlinks.

Rubber-stamped by Alexey Proskuryakov.

  • WebKit.xcodeproj/project.pbxproj: If a link already exist, don’t try to create it.
6:08 PM Changeset in webkit [180875] by fpizlo@apple.com
  • 5 edits in trunk/Source/JavaScriptCore

BytecodeGenerator shouldn't emit op_resolve_scope as a roundabout way of returning the scopeRegister
​https://bugs.webkit.org/show_bug.cgi?id=142153

Reviewed by Michael Saboff.

We don't need a op_resolve_scope if we know that it will simply return the scope register.
This changes the BytecodeGenerator to use the scope register directly in those cases where
we know statically that we would just have returned that from op_resolve_scope.

This doesn't appear to have a significant impact on performance.

  • bytecode/CodeBlock.cpp:

(JSC::CodeBlock::CodeBlock):

  • bytecompiler/BytecodeGenerator.cpp:

(JSC::BytecodeGenerator::emitResolveScope):
(JSC::BytecodeGenerator::emitReturn):
(JSC::BytecodeGenerator::emitGetOwnScope): Deleted.

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

(JSC::ResolveNode::emitBytecode):
(JSC::EvalFunctionCallNode::emitBytecode):
(JSC::FunctionCallResolveNode::emitBytecode):
(JSC::PostfixNode::emitResolve):
(JSC::DeleteResolveNode::emitBytecode):
(JSC::TypeOfResolveNode::emitBytecode):
(JSC::PrefixNode::emitResolve):
(JSC::ReadModifyResolveNode::emitBytecode):
(JSC::AssignResolveNode::emitBytecode):
(JSC::ConstDeclNode::emitCodeSingle):
(JSC::EmptyVarExpression::emitBytecode):
(JSC::ForInNode::emitLoopHeader):
(JSC::ForOfNode::emitBytecode):
(JSC::BindingNode::bindValue):

5:53 PM Changeset in webkit [180874] by gyuyoung.kim@samsung.com
  • 2 edits in trunk/LayoutTests

Unreviewed EFL gardening on 2nd March.

Mark css3 shape tests to flaky. Unskip passing tests and so on.

  • platform/efl/TestExpectations:
5:48 PM Changeset in webkit [180873] by commit-queue@webkit.org
  • 7 edits in trunk/Source/WebCore

Use std::unique_ptr instead of PassOwnPtr|OwnPtr for ScrollAnimator
​https://bugs.webkit.org/show_bug.cgi?id=142143

Patch by Joonghun Park <​jh718.park@samsung.com> on 2015-03-01
Reviewed by Darin Adler.

No new tests, no behavior changes.

  • platform/ScrollAnimator.cpp:

(WebCore::ScrollAnimator::create):

  • platform/ScrollAnimator.h:
  • platform/ScrollAnimatorNone.cpp:

(WebCore::ScrollAnimator::create):

  • platform/ScrollableArea.h:
  • platform/ios/ScrollAnimatorIOS.mm:

(WebCore::ScrollAnimator::create):

  • platform/mac/ScrollAnimatorMac.mm:

(WebCore::ScrollAnimator::create):

5:46 PM Changeset in webkit [180872] by aestes@apple.com
  • 3 edits in trunk/Source/WebCore

[Content Filtering] Move another declaration to WebFilterEvaluatorSPI.h
​https://bugs.webkit.org/show_bug.cgi?id=142066

Reviewed by Andreas Kling.

  • platform/ios/ContentFilterIOS.mm:
  • platform/spi/cocoa/WebFilterEvaluatorSPI.h:
4:08 PM Changeset in webkit [180871] by Chris Dumez
  • 21 edits
    4 adds in trunk

Make NotificationCenter / Notification suspendable
​https://bugs.webkit.org/show_bug.cgi?id=142117
<rdar://problem/19923085>

Reviewed by Andreas Kling.

Source/WebCore:

Make NotificationCenter / Notification suspendable so that pages using
them can enter the PageCache.

NotificationCenter can safely be suspended if there are no pending
permission requests. This required adding an
"hasPendingPermissionRequests()" callback to the NotificationClient.

Notification can safely be suspended if it is either idle (not showing
yet) or closed.

Tests: fast/history/page-cache-notification-non-suspendable.html

fast/history/page-cache-notification-suspendable.html

  • Modules/notifications/Notification.cpp:

(WebCore::Notification::canSuspend):

  • Modules/notifications/NotificationCenter.cpp:

(WebCore::NotificationCenter::canSuspend):

  • Modules/notifications/NotificationClient.h:

Source/WebKit/mac:

Provide implementation for NotificationClient::hasPendingPermissionRequests().
The implementation is very simplistic. it will only return false if no
request for permission for ever made. This is because there is currently no
easy way to figure out if a permission request is pending or not.

  • WebCoreSupport/WebNotificationClient.h:
  • WebCoreSupport/WebNotificationClient.mm:

(WebNotificationClient::requestPermission):
(WebNotificationClient::hasPendingPermissionRequests):

Source/WebKit/win:

Provide implementation for NotificationClient::hasPendingPermissionRequests().

  • WebCoreSupport/WebDesktopNotificationsDelegate.cpp:

(WebDesktopNotificationsDelegate::requestPermission):
(hasPendingPermissionRequests):

  • WebCoreSupport/WebDesktopNotificationsDelegate.h:

Source/WebKit2:

Provide implementation for NotificationClient::hasPendingPermissionRequests().

  • WebProcess/Notifications/NotificationPermissionRequestManager.cpp:

(WebKit::NotificationPermissionRequestManager::hasPendingPermissionRequests):

  • WebProcess/Notifications/NotificationPermissionRequestManager.h:
  • WebProcess/WebCoreSupport/WebNotificationClient.cpp:

(WebKit::WebNotificationClient::hasPendingPermissionRequests):

  • WebProcess/WebCoreSupport/WebNotificationClient.h:

LayoutTests:

Add layout tests to cover cases where notifications should prevent
entering the PageCache or not.

  • fast/history/page-cache-notification-non-suspendable-expected.txt: Added.
  • fast/history/page-cache-notification-non-suspendable.html: Added.
  • fast/history/page-cache-notification-suspendable-expected.txt: Added.
  • fast/history/page-cache-notification-suspendable.html: Added.
  • platform/efl/TestExpectations:
  • platform/gtk/TestExpectations:
  • platform/ios-simulator/TestExpectations:
  • platform/win/TestExpectations:
  • platform/wincairo/TestExpectations:
2:34 PM Changeset in webkit [180870] by rniwa@webkit.org
  • 11 edits in trunk/LayoutTests

EFL, GTK+, and Windows rebaselines after r180867.

  • platform/efl/TestExpectations:
  • platform/efl/editing/execCommand/5142012-1-expected.txt:
  • platform/efl/editing/execCommand/nsresponder-outdent-expected.txt:
  • platform/efl/editing/inserting/insert-at-end-02-expected.txt:
  • platform/gtk/editing/execCommand/5142012-1-expected.txt:
  • platform/gtk/editing/execCommand/nsresponder-outdent-expected.txt:
  • platform/gtk/editing/inserting/insert-at-end-02-expected.txt:
  • platform/gtk/editing/pasteboard/4989774-expected.txt:
  • platform/win/editing/execCommand/5142012-1-expected.txt:
  • platform/win/editing/execCommand/nsresponder-outdent-expected.txt:
  • platform/win/editing/inserting/insert-at-end-02-expected.txt:
1:56 PM Changeset in webkit [180869] by Antti Koivisto
  • 4 edits in trunk/Source/WebKit2

Enable new disk cache on iOS
​https://bugs.webkit.org/show_bug.cgi?id=142148

Reviewed by Sam Weinig.

  • NetworkProcess/cache/NetworkCacheStorageCocoa.mm:

iOS build fix.

  • UIProcess/Cocoa/WebProcessPoolCocoa.mm:

(WebKit::registerUserDefaultsIfNeeded):

Disable the efficacy logging by default for now. It has significant performance impact.

  • config.h:

Enable it.

12:43 PM Changeset in webkit [180868] by ap@apple.com
  • 2 edits in trunk/LayoutTests

media/track/track-in-band-cues-added-once.html flakily fails
​https://bugs.webkit.org/show_bug.cgi?id=142152

  • platform/mac/TestExpectations: Marked it as such.
11:52 AM Changeset in webkit [180867] by rniwa@webkit.org
  • 17 edits
    2 adds in trunk

isContentEditable shouldn't trigger synchronous style recalc in most cases
​https://bugs.webkit.org/show_bug.cgi?id=129034

Reviewed by Antti Koivisto.

Source/WebCore:

Avoid style recalc inside isContentEditable when the document doesn't contain -webkit-user-modify or
-webkit-user-select: all. Instead, compute the value from contenteditable attributes in ancestors.
However, still compute the editability from the style tree when it's up-to-date in order to avoid
repeatedly walking up the DOM tree in a hot code path inside editing.

Test: fast/dom/HTMLElement/dynamic-editability-change.html

  • css/CSSGrammar.y.in: No need to pass in "true" as we never call this function with false.
  • css/CSSParser.cpp:

(WebCore::isValidKeywordPropertyAndValue): Calls parserSetUsesStyleBasedEditability as needed.
(WebCore::parseKeywordValue): Passes around StyleSheetContents*.
(WebCore::CSSParser::parseValue): Ditto.
(WebCore::CSSParser::parseFont): Ditto.

  • css/StyleSheetContents.cpp:

(WebCore::StyleSheetContents::StyleSheetContents): Initializes and copies m_usesStyleBasedEditability.

  • css/StyleSheetContents.h:

(WebCore::StyleSheetContents::parserSetUsesRemUnits): Removed the argument since it was always true.
(WebCore::StyleSheetContents::parserSetUsesStyleBasedEditability): Added.
(WebCore::StyleSheetContents::usesStyleBasedEditability): Added.

  • dom/Document.cpp:

(WebCore::Document::recalcStyle): Added a FIXME as well as a comment explaining why we don't call
setUsesStyleBasedEditability. Since Node::computeEditability triggers style recalc only when the flag
is set to true, it's too late to update the flag here.
(WebCore::Document::updateStyleIfNeeded): Uses a newly extracted needsStyleRecalc.
(WebCore::Document::updateBaseURL): Preserves m_usesStyleBasedEditability as well as m_usesRemUnit.
(WebCore::Document::usesStyleBasedEditability): Added. Returns true when inline style declarations or
any active stylesheet uses -webkit-user-modify or -webkit-user-select: all. Flushing pending stylesheet
changes here is fine because the alternative is to trigger a full blown style recalc.

  • dom/Document.h:

(WebCore::Document::needsStyleRecalc): Added. Extracted from updateStyleIfNeeded.

  • dom/DocumentStyleSheetCollection.cpp:

(WebCore::DocumentStyleSheetCollection::DocumentStyleSheetCollection):
(WebCore::styleSheetsUseRemUnits): Deleted.
(WebCore::DocumentStyleSheetCollection::updateActiveStyleSheets): Updates m_usesStyleBasedEditability
as well as m_usesRemUnit.

  • dom/DocumentStyleSheetCollection.h:

(WebCore::DocumentStyleSheetCollection::usesStyleBasedEditability): Added.
(WebCore::DocumentStyleSheetCollection::setUsesStyleBasedEditability): Added.

  • dom/Node.cpp:

(WebCore::computeEditabilityFromComputedStyle): Extracted from computeEditability.
(WebCore::Node::computeEditability): When the style recalc is requested and the render tree is dirty,
check if the document uses any CSS property that can affect the editability of elements. If it doesn't,
compute the editability from contenteditable attributes in the anchors via matchesReadWritePseudoClass.
Continue to use the style-based computation when the render tree isn't dirty to avoid the tree walk.

  • html/HTMLElement.cpp:

(WebCore::HTMLElement::editabilityFromContentEditableAttr): Extracted from matchesReadWritePseudoClass
to be called in Node::computeEditability. Also made it return Editability instead of boolean.
(WebCore::HTMLElement::matchesReadWritePseudoClass):

  • html/HTMLElement.h:

LayoutTests:

Added a regression test to update the editability of elements dynamically. Also rebaselined
tests per style recalc timing changes.

  • fast/dom/HTMLElement/dynamic-editability-change-expected.txt: Added.
  • fast/dom/HTMLElement/dynamic-editability-change.html: Added.
  • platform/mac/editing/execCommand/5142012-1-expected.txt: anonymous render block differences.
  • platform/mac/editing/execCommand/nsresponder-outdent-expected.txt: Ditto.
  • platform/mac/editing/inserting/insert-at-end-02-expected.txt: Empty render text differences.
  • platform/mac/editing/pasteboard/4989774-expected.txt: Ditto.
11:48 AM Changeset in webkit [180866] by ddkilzer@apple.com
  • 2 edits in trunk/Tools

LayoutTestRealy: Prepend XPC_ to the key, not the value!

Follow-up fix for:

LayoutTestRelay: App environment variables not set for --guard-malloc or --leaks
<​http://webkit.org/b/142145>

  • LayoutTestRelay/LayoutTestRelay/LTRelayController.m:

(-[LTRelayController _environmentVariables]): Fix think-o.

11:12 AM Changeset in webkit [180865] by Brent Fulgham
  • 2 edits in trunk/Source/WebCore

[Win] Unreviewed build fix.

  • WebCorePrefix.h: Provide some default definitions to help build on Windows

machines with different application support libraries.

10:57 AM Changeset in webkit [180864] by ddkilzer@apple.com
  • 2 edits in trunk/Tools

LayoutTestRelay: App environment variables not set for --guard-malloc or --leaks
<​http://webkit.org/b/142145>

Reviewed by Simon Fraser.

  • LayoutTestRelay/LayoutTestRelay/LTRelayController.m:

(-[LTRelayController _environmentVariables]): Add.
(-[LTRelayController launchApp]): Use -_environmentVariables.

10:28 AM Changeset in webkit [180863] by Brent Fulgham
  • 2 edits in trunk/LayoutTests

[Win] Document some more debug assertions.

  • platform/win/TestExpectations:
1:52 AM Changeset in webkit [180862] by Carlos Garcia Campos
  • 2 edits in releases/WebKitGTK/webkit-2.8/Source/WebCore

Merge r180846 - FrameView::layoutTimerFired() should update style if needed before doing layout
​https://bugs.webkit.org/show_bug.cgi?id=141688

Reviewed by Andreas Kling.

If the style recalc timer has been scheduled to fire after the layout timer,
when the layout timer fires, we might as well just do the style recalc
too. The call to updateStyleIfNeeded() will cancel the pending style
recalc timer.

This doesn't have much impact on the number of layouts (measured via PLT)
but seems like a reasonable thing to do.

  • page/FrameView.cpp:

(WebCore::FrameView::layoutTimerFired):

1:48 AM Changeset in webkit [180861] by Carlos Garcia Campos
  • 14 edits
    1 copy in releases/WebKitGTK/webkit-2.8/Source/bmalloc

Merge r180797 - bmalloc: Pathological madvise churn on the free(malloc(x)) benchmark
​https://bugs.webkit.org/show_bug.cgi?id=142058

Reviewed by Andreas Kling.

The churn was caused by repeatedly splitting an object with physical
pages from an object without, and then merging them back together again.
The merge would conservatively forget that we had physical pages, forcing
a new call to madvise on the next allocation.

This patch more strictly segregates objects in the heap from objects in
the VM heap, with these changes:

(1) Objects in the heap are not allowed to merge with objects in the VM
heap, and vice versa -- since that would erase our precise knowledge of
which physical pages had been allocated.

(2) The VM heap is exclusively responsible for allocating and deallocating
physical pages.

(3) The heap free list must consider entries for objects that are in the
VM heap to be invalid, and vice versa. (This condition can arise
because the free list does not eagerly remove items.)

With these changes, we can know that any valid object in the heap's free
list already has physical pages, and does not need to call madvise.

Note that the VM heap -- as before -- might sometimes contain ranges
or pieces of ranges that have physical pages, since we allow splitting
of ranges at granularities smaller than the VM page size. These ranges
can eventually merge with ranges in the heap during scavenging.

  • bmalloc.xcodeproj/project.pbxproj:
  • bmalloc/BoundaryTag.h:

(bmalloc::BoundaryTag::owner):
(bmalloc::BoundaryTag::setOwner):
(bmalloc::BoundaryTag::initSentinel):
(bmalloc::BoundaryTag::hasPhysicalPages): Deleted.
(bmalloc::BoundaryTag::setHasPhysicalPages): Deleted. Replaced the concept
of "has physical pages" with a bit indicating which heap owns the large
object. This is a more precise concept, since the old bit was really a
Yes / Maybe bit.

  • bmalloc/Deallocator.cpp:
  • bmalloc/FreeList.cpp: Adopt

(bmalloc::FreeList::takeGreedy):
(bmalloc::FreeList::take):
(bmalloc::FreeList::removeInvalidAndDuplicateEntries):

  • bmalloc/FreeList.h:

(bmalloc::FreeList::push): Added API for considering the owner when
deciding if a free list entry is valid.

  • bmalloc/Heap.cpp:

(bmalloc::Heap::Heap): Adopt new API.

(bmalloc::Heap::scavengeLargeRanges): Scavenge all ranges with no minimum,
since some ranges might be able to merge with ranges in the VM heap, and
they won't be allowed to until we scavenge them.

(bmalloc::Heap::allocateSmallPage):
(bmalloc::Heap::allocateMediumPage):
(bmalloc::Heap::allocateLarge): New VM heap API makes this function
simpler, since we always get back physical pages now.

  • bmalloc/Heap.h:
  • bmalloc/LargeObject.h:

(bmalloc::LargeObject::end):
(bmalloc::LargeObject::owner):
(bmalloc::LargeObject::setOwner):
(bmalloc::LargeObject::isValidAndFree):
(bmalloc::LargeObject::merge): Do not merge objects across heaps since
that causes madvise churn.
(bmalloc::LargeObject::validateSelf):
(bmalloc::LargeObject::init):
(bmalloc::LargeObject::hasPhysicalPages): Deleted.
(bmalloc::LargeObject::setHasPhysicalPages): Deleted. Propogate the Owner API.

  • bmalloc/Owner.h: Added.
  • bmalloc/SegregatedFreeList.cpp:

(bmalloc::SegregatedFreeList::SegregatedFreeList):
(bmalloc::SegregatedFreeList::insert):
(bmalloc::SegregatedFreeList::takeGreedy):
(bmalloc::SegregatedFreeList::take):

  • bmalloc/SegregatedFreeList.h: Propogate the owner API.
  • bmalloc/VMAllocate.h:

(bmalloc::vmDeallocatePhysicalPagesSloppy):
(bmalloc::vmAllocatePhysicalPagesSloppy): Clarified these functions and
removed an edge case.

  • bmalloc/VMHeap.cpp:

(bmalloc::VMHeap::VMHeap):

  • bmalloc/VMHeap.h:

(bmalloc::VMHeap::allocateSmallPage):
(bmalloc::VMHeap::allocateMediumPage):
(bmalloc::VMHeap::allocateLargeObject):
(bmalloc::VMHeap::deallocateLargeObject): Be sure to give each object
a new chance to merge, since it might have been prohibited from merging
before by virtue of not being in the VM heap.

(bmalloc::VMHeap::allocateLargeRange): Deleted.
(bmalloc::VMHeap::deallocateLargeRange): Deleted.

1:43 AM Changeset in webkit [180860] by Carlos Garcia Campos
  • 18 edits in releases/WebKitGTK/webkit-2.8/Source/WebCore

Merge r180772 - Use NeverDestroyed for JS wrapper owners.
<​https://webkit.org/b/142090>

Reviewed by Chris Dumez.

Using NeverDestroyed puts these objects in BSS which is preferable
since that prevents them from pinning down entire malloc pages forever.

  • bindings/scripts/CodeGeneratorJS.pm:

(GenerateHeader): Use NeverDestroyed instead of DEPRECATED_DEFINE_STATIC_LOCAL.

  • bindings/scripts/test/JS/*: Rebaseline bindings tests for this change.
1:21 AM Changeset in webkit [180859] by Carlos Garcia Campos
  • 3 edits
    2 adds in releases/WebKitGTK/webkit-2.8

Merge r180767 - Use after free in WebCore::RenderNamedFlowFragment::restoreRegionObjectsOriginalStyle
​https://bugs.webkit.org/show_bug.cgi?id=138366

Reviewed by Dave Hyatt.

This patch ensures that we clean up RenderNamedFlowFragment::m_renderObjectRegionStyle when embedded flow content is getting destroyed.

In m_renderObjectRegionStyle hash map, we store style information about the named flow's descendant children.
When a child is being detached from the tree, it removes itself from this hashmap.
We do it by traversing up on the ancestor chain and call removeFlowChildInfo() on the parent flow.
However in case of embedded flows (for example multicolumn content inside a region), we need to check whether the parent flow
is inside a flow too and continue the cleanup accordingly.

Source/WebCore:

Test: fast/regions/region-with-multicolumn-embedded-crash.html

  • rendering/RenderObject.cpp:

(WebCore::RenderObject::removeFromRenderFlowThreadIncludingDescendants):

LayoutTests:

  • fast/regions/region-with-multicolumn-embedded-crash-expected.txt: Added.
  • fast/regions/region-with-multicolumn-embedded-crash.html: Added.
1:18 AM Changeset in webkit [180858] by Carlos Garcia Campos
  • 3 edits in releases/WebKitGTK/webkit-2.8/Source/JavaScriptCore

Merge r180716 - MachineThreads::Thread clean up has a use after free race condition.
<​https://webkit.org/b/141990>

Reviewed by Filip Pizlo.

MachineThreads::Thread clean up relies on the clean up mechanism
implemented in _pthread_tsd_cleanup_key(), which looks like this:

void _pthread_tsd_cleanup_key(pthread_t self, pthread_key_t key)
{

void (*destructor)(void *);
if (_pthread_key_get_destructor(key, &destructor)) {

void ptr = &self->tsd[key];
void *value = *ptr;

=== Start of window for the bug to manifest =================

At this point, this thread has cached "destructor" and "value"
 (which is a MachineThreads*). If the VM gets destructed (along
with its MachineThreads registry) by another thread, then this
 thread will have no way of knowing that the MachineThreads* is
now pointing to freed memory. Calling the destructor below will
 therefore result in a use after free scenario when it tries to
access the MachineThreads' data members.

if (value) {

*ptr = NULL;
if (destructor) {

=== End of window for the bug to manifest ==================

destructor(value);

}

}

}

}

The fix is to add each active MachineThreads to an ActiveMachineThreadsManager,
and always check if the manager still contains that MachineThreads object
before we call removeCurrentThread() on it. When MachineThreads is destructed,
it will remove itself from the manager. The add, remove, and checking
operations are all synchronized on the manager's lock, thereby ensuring that
the MachineThreads object, if found in the manager, will remain alive for the
duration of time we call removeCurrentThread() on it.

There's also possible for the MachineThreads object to already be destructed
and another one happened to have been instantiated at the same address.
Hence, we should only remove the exiting thread if it is found in the
MachineThreads object.

There is no test for this issue because this bug requires a race condition
between 2 threads where:

  1. Thread B, which had previously used the VM, exiting and getting to the bug window shown in _pthread_tsd_cleanup_key() above.
  2. Thread A destructing the VM (and its MachineThreads object) within that window of time before Thread B calls the destructor.

It is not possible to get a reliable test case without invasively
instrumenting _pthread_tsd_cleanup_key() or MachineThreads::removeCurrentThread()
to significantly increase that window of opportunity.

  • heap/MachineStackMarker.cpp:

(JSC::ActiveMachineThreadsManager::Locker::Locker):
(JSC::ActiveMachineThreadsManager::add):
(JSC::ActiveMachineThreadsManager::remove):
(JSC::ActiveMachineThreadsManager::contains):
(JSC::ActiveMachineThreadsManager::ActiveMachineThreadsManager):
(JSC::activeMachineThreadsManager):
(JSC::MachineThreads::MachineThreads):
(JSC::MachineThreads::~MachineThreads):
(JSC::MachineThreads::removeThread):
(JSC::MachineThreads::removeThreadIfFound):
(JSC::MachineThreads::removeCurrentThread): Deleted.

  • heap/MachineStackMarker.h:
12:56 AM Changeset in webkit [180857] by Carlos Garcia Campos
  • 6 edits in releases/WebKitGTK/webkit-2.8/Source/bmalloc

Merge r180701 - bmalloc: Large object free list can grow infinitely
​https://bugs.webkit.org/show_bug.cgi?id=142055

Reviewed by Andreas Kling.

By design, we don't eagerly remove large objects from the free list.
This creates two simple pathologies:

(1) If you free and then allocate the same object repeatedly, it will
duplicate itself in the free list repeatedly. Since it is never
invalid at the time of allocation, it will never be removed.

(2) If you split and then merge the same object repeatedly, it will
duplicate its split sibling in the free list repeatedly. If its
sibling is in a separate free list size class, it will never be
consulted at the time of allocation, so it will never be removed.

So, a simple "while (1) { free(malloc(x)); }" causes infinite memory
use in the free list.

The solution in this patch is a simple helper to remove garbage from the
free list if it grows too large. This pathology is not common, so the
cost is OK.

Long-term, perhaps we should rethink the laziness of these free lists.

  • bmalloc/BoundaryTag.h:

(bmalloc::BoundaryTag::isMarked):
(bmalloc::BoundaryTag::setMarked): New bit, used by free list GC.

  • bmalloc/FreeList.cpp:

(bmalloc::FreeList::removeInvalidAndDuplicateEntries): The GC algorithm.

  • bmalloc/FreeList.h:

(bmalloc::FreeList::FreeList):
(bmalloc::FreeList::push): Invoke the GC if we're getting huge.

  • bmalloc/LargeObject.h:

(bmalloc::LargeObject::isMarked):
(bmalloc::LargeObject::setMarked):
(bmalloc::LargeObject::validateSelf): Expose the new bit.

  • bmalloc/Sizes.h: New constant to control GC frequency.
12:51 AM Changeset in webkit [180856] by Carlos Garcia Campos
  • 8 edits
    1 copy
    1 move in releases/WebKitGTK/webkit-2.8/Source/bmalloc

Merge r180693 - bmalloc: Refactored SegregatedFreeList and BoundaryTag::init
​https://bugs.webkit.org/show_bug.cgi?id=142049

Reviewed by Anders Carlsson.

Split out a FreeList class from SegregatedFreeList. This will make it
easier to add behaviors on free list insertion and removal -- and it's
probably how I should have designed things at the start.

Moved BoundaryTag::init into LargeObject, since all the related logic
lives in LargeObject now too, and this allows us to remove BoundaryTagInlines.h.

  • bmalloc.xcodeproj/project.pbxproj:
  • bmalloc/BoundaryTagInlines.h: Removed.
  • bmalloc/FreeList.cpp: Copied from Source/bmalloc/bmalloc/SegregatedFreeList.cpp.

(bmalloc::FreeList::takeGreedy):
(bmalloc::FreeList::take):
(bmalloc::SegregatedFreeList::SegregatedFreeList): Deleted.
(bmalloc::SegregatedFreeList::insert): Deleted.
(bmalloc::SegregatedFreeList::takeGreedy): Deleted.
(bmalloc::SegregatedFreeList::take): Deleted.

  • bmalloc/FreeList.h: Copied from Source/bmalloc/bmalloc/SegregatedFreeList.h.

(bmalloc::FreeList::push):

  • bmalloc/LargeObject.h:

(bmalloc::LargeObject::init):

  • bmalloc/SegregatedFreeList.cpp:

(bmalloc::SegregatedFreeList::SegregatedFreeList):
(bmalloc::SegregatedFreeList::insert):
(bmalloc::SegregatedFreeList::takeGreedy):
(bmalloc::SegregatedFreeList::take):

  • bmalloc/SegregatedFreeList.h:
  • bmalloc/Sizes.h:
  • bmalloc/VMHeap.cpp:

(bmalloc::VMHeap::grow):

12:45 AM Changeset in webkit [180855] by Carlos Garcia Campos
  • 4 edits in releases/WebKitGTK/webkit-2.8/Source/bmalloc

Merge r180688 - bmalloc: free up a bit in BoundaryTag
​https://bugs.webkit.org/show_bug.cgi?id=142048

Reviewed by Brady Eidson.

We were wasting a bit by accident, and I need one now.

  • bmalloc/Algorithm.h:

(bmalloc::rightShift): Deleted. Not needed, now that I've simplified
the math.

  • bmalloc/BoundaryTag.h: Since each boundary tag bucket is 1024 bytes

long, the maximum offset into a bucket is 1023.

You need 5 bits to count up to 1024, but only 4 to count up to 1023.

Math is hard.

(bmalloc::BoundaryTag::compactBegin): Switched to division because it
is simpler, and easier to match up with our ASSERT. The compiler will
turn division by constant power of two into a shift for us.

(bmalloc::BoundaryTag::setRange): Added an ASSERT for compactBegin
because we do encode it, so we should ASSERT that encoding did not
lose information.

  • bmalloc/Sizes.h: Shifting is no longer used since we use division

instead.

12:43 AM Changeset in webkit [180854] by Carlos Garcia Campos
  • 6 edits
    6 adds in releases/WebKitGTK/webkit-2.8

Merge r180683 - Setting any of the <object> element plugin controlling attributes does not have any affect.
​https://bugs.webkit.org/show_bug.cgi?id=141936.

Patch by Said Abou-Hallawa <​sabouhallawa@apple.com> on 2015-02-26
Reviewed by Zalan Bujtas.

Source/WebCore:

When setting any of the <object> element plugin controlling attributes
dynamically we need to mark the the element to be dirty by calling
setNeedsStyleRecalc(), so it has to recreate its renderer when needed.

Test: svg/as-object/svg-in-object-dynamic-attribute-change.html

  • dom/Element.h: Delete unimplemented function.
  • html/HTMLObjectElement.cpp:

(WebCore::HTMLObjectElement::parseAttribute): Dirty the element by calling
setNeedsStyleRecalc() when one of the plugin controlling attributes gets
changed. We have to clear the m_useFallbackContent because the attribute's
new value might fix the object rendering.

  • html/HTMLObjectElement.h: Add a function to clear m_useFallbackContent.
  • html/HTMLPlugInImageElement.cpp:

(WebCore::HTMLPlugInImageElement::willRecalcStyle): We might need to
reconstruct the object renderer in the image case. This can happen if the
image was rendering fallback content and the attribute's new value fixes
the object rendering.

LayoutTests:

  • svg/as-object/resources/lime100x100.html: Added.
  • svg/as-object/resources/lime100x100.png: Added.
  • svg/as-object/resources/lime100x100.svg: Added.
  • svg/as-object/resources/red100x100.svg: Added.
  • svg/as-object/svg-in-object-dynamic-attribute-change-expected.html: Added.
  • svg/as-object/svg-in-object-dynamic-attribute-change.html: Added.

Ensure that changing the 'type' and the 'data' attributes of the <object>
element will have the expected outcome. Also make sure that the <object>
element renderer falls back correctly when setting any of the attributes
to some unexpected value.

12:36 AM WebKitGTK/2.8.x edited by Carlos Garcia Campos
(diff)
12:28 AM Changeset in webkit [180853] by Carlos Garcia Campos
  • 2 edits in releases/WebKitGTK/webkit-2.8/Source/JavaScriptCore

Merge r180667 - Add calleeSaveRegisters() implementation for ARM Traditional
​https://bugs.webkit.org/show_bug.cgi?id=141903

Reviewed by Darin Adler.

  • jit/RegisterSet.cpp:

(JSC::RegisterSet::calleeSaveRegisters):

12:24 AM Changeset in webkit [180852] by Carlos Garcia Campos
  • 7 edits
    2 adds in releases/WebKitGTK/webkit-2.8

Merge r180643 - Horizontal and vertical lines are clipped completely if clip-path is included in the tag but the referenced element is defined later.
​https://bugs.webkit.org/show_bug.cgi?id=141776.

Reviewed by Dean Jackson.
Source/WebCore:

Tests: svg/clip-path/clip-path-line-use-before-defined-expected.svg

svg/clip-path/clip-path-line-use-before-defined.svg

  • rendering/svg/RenderSVGResourceClipper.cpp:

(WebCore::RenderSVGResourceClipper::applyClippingToContext): Ensure the renderer
is added to m_clipper if it does not exist. The same renderer might have been
added to m_clipper in resourceBoundingBox().

(WebCore::RenderSVGResourceClipper::addRendererToClipper): Add the renderer to
m_clipper if it does not exist. Return the associated ClipperData.

(WebCore::RenderSVGResourceClipper::resourceBoundingBox): If the clipper is
referenced before it is defined, add the renderer to m_clipper. While doing the
layout() for the clipper, we can check if m_clipper has values or not. If it does
have, we are going to mark the clipper for client invalidation which is done by
the SVG root.

  • rendering/svg/RenderSVGResourceClipper.h:
  • rendering/svg/RenderSVGResourceContainer.h:

(WebCore::RenderSVGResourceContainer::selfNeedsClientInvalidation): Define a
new function selfNeedsClientInvalidation() which controls marking the clipper
for client invalidation. In RenderSVGResourceClipper, override it so it checks
m_clipper to force clients validation even if it the first time we do layout
for this clipper.

  • rendering/svg/RenderSVGResourceContainer.cpp:

(WebCore::RenderSVGResourceContainer::layout): Call the virtual function
selfNeedsClientInvalidation() to check whether we need to mark the clipper for
client invalidation.

  • svg/SVGElement.cpp: Delete unneeded header file.

LayoutTests:

New test cases for SVG lines which are clipped to a <clipPath>. The <clipPath>
is referenced before it is defined.

  • svg/clip-path/clip-path-line-use-before-defined-expected.svg: Added.
  • svg/clip-path/clip-path-line-use-before-defined.svg: Added.
12:21 AM Changeset in webkit [180851] by Carlos Garcia Campos
  • 6 edits
    1 add in releases/WebKitGTK/webkit-2.8

Merge r180639 - CodeBlock crashes when dumping op_push_name_scope
​https://bugs.webkit.org/show_bug.cgi?id=141953

Patch by Benjamin Poulain <​bpoulain@apple.com> on 2015-02-25
PerformanceTests/SunSpider:

Reviewed by Filip Pizlo.

  • profiler-test.yaml:

Source/JavaScriptCore:

Reviewed by Filip Pizlo and Csaba Osztrogonác.

  • bytecode/CodeBlock.cpp:

(JSC::CodeBlock::dumpBytecode):

  • tests/stress/op-push-name-scope-crashes-profiler.js: Added.

Tools:

Reviewed by Filip Pizlo.

  • Scripts/run-jsc-stress-tests:
12:17 AM Changeset in webkit [180850] by Carlos Garcia Campos
  • 2 edits in releases/WebKitGTK/webkit-2.8/Source/WebCore

Merge r180634 - REGRESSION (r180018 ): Holding a rubber-band in place can get stuck
​https://bugs.webkit.org/show_bug.cgi?id=142020
-and corresponding-
rdar://problem/19945216

Reviewed by Tom Horton.

It was a mistaken assumption that it was necessary to return false in the zero-
delta case. That is clearly conceptually wrong since false represents the DOM
doing something special with the event, which is clearly not the case if we never
even send the event to the DOM. Returning true will allow the rest of the
scrolling machinery the ability to handle the event.

  • dom/Element.cpp:

(WebCore::Element::dispatchWheelEvent):

12:15 AM Changeset in webkit [180849] by Carlos Garcia Campos
  • 16 edits in releases/WebKitGTK/webkit-2.8

Merge r180621 - AX: Implement support for ARIA 1.1 'searchbox' role
​https://bugs.webkit.org/show_bug.cgi?id=142004

Reviewed by Chris Fleizach.

Source/WebCore:

Add a new accessible SearchFieldRole to handle both the ARIA role
and the "search" input type.

No new tests. Instead, added a new test case to roles-exposed.html
for the mapping, and updated roles-computedRoleString.html because
there is now a one-to-one mapping between the "search" input type
and an ARIA role.

  • accessibility/AccessibilityNodeObject.cpp:

(WebCore::AccessibilityNodeObject::determineAccessibilityRole):
(WebCore::AccessibilityNodeObject::isSearchField):

  • accessibility/AccessibilityObject.cpp:

(WebCore::AccessibilityObject::isARIATextControl):
(WebCore::AccessibilityObject::isARIAInput):
(WebCore::initializeRoleMap):

  • accessibility/AccessibilityObject.h:
  • accessibility/AccessibilityRenderObject.cpp:

(WebCore::AccessibilityRenderObject::determineAccessibilityRole):

  • accessibility/atk/WebKitAccessibleWrapperAtk.cpp:

(atkRole):

  • accessibility/ios/WebAccessibilityObjectWrapperIOS.mm:

(-[WebAccessibilityObjectWrapper accessibilityCanFuzzyHitTest]):
(-[WebAccessibilityObjectWrapper accessibilityTraits]):
(-[WebAccessibilityObjectWrapper determineIsAccessibilityElement]):

  • accessibility/mac/WebAccessibilityObjectWrapperMac.mm:

(createAccessibilityRoleMap):

LayoutTests:

  • accessibility/roles-computedRoleString-expected.txt: Updated for new role.
  • accessibility/roles-computedRoleString.html: Updated for new role.
  • accessibility/roles-exposed.html: New test case added.
  • platform/efl/accessibility/roles-exposed-expected.txt: Updated for new test case.
  • platform/gtk/accessibility/roles-exposed-expected.txt: Updated for new test case.
  • platform/mac-mavericks/accessibility/roles-exposed-expected.txt: Updated for new test case.
  • platform/mac/accessibility/roles-exposed-expected.txt: Updated for new test case.

Feb 28, 2015:

8:36 PM Changeset in webkit [180848] by Simon Fraser
  • 6 edits in trunk/Source

Viewport units should not dirty style just before we do layout
​https://bugs.webkit.org/show_bug.cgi?id=141682

Reviewed by Zalan Bujtas.
Source/WebCore:

In documents using viewport units, we dirtied style every time layout changed
the size of the document. This is nonsensical, because viewport units depend on the
viewport size, not the document size.

Move the style dirtying from layout() into availableContentSizeChanged(). Hook
this up for WebKit1 by calling from -[WebFrameView _frameSizeChanged], and,
since that causes availableContentSizeChanged() to be called for WK1 for the first
time, protect the call to updateScrollbars() with a !platformWidget check.

Covered by existing viewport unit tests.

  • page/FrameView.cpp:

(WebCore::FrameView::layout):
(WebCore::FrameView::availableContentSizeChanged):
(WebCore::FrameView::viewportSizeForCSSViewportUnits): Add a FIXME comment. Whether
scrollbars are ignored depends on the value of the overflow property on the root element.

  • page/FrameView.h:
  • platform/ScrollView.cpp:

(WebCore::ScrollView::availableContentSizeChanged):

Source/WebKit/mac:

  • WebView/WebFrameView.mm:

(-[WebFrameView _frameSizeChanged]):

6:31 PM Changeset in webkit [180847] by akling@apple.com
  • 2 edits in trunk/Source/WebCore

[Cocoa] Purge SQLite page cache when under memory pressure.
<​https://webkit.org/b/142139>
<rdar://problem/19997739>

Reviewed by Pratik Solanki.

Call out to sqlite3 cache purging SPI on Cocoa platforms when
we need to free up some extra memory.

  • platform/cocoa/MemoryPressureHandlerCocoa.mm:

(WebCore::MemoryPressureHandler::platformReleaseMemory):

5:04 PM Changeset in webkit [180846] by Simon Fraser
  • 2 edits in trunk/Source/WebCore

FrameView::layoutTimerFired() should update style if needed before doing layout
​https://bugs.webkit.org/show_bug.cgi?id=141688

Reviewed by Andreas Kling.

If the style recalc timer has been scheduled to fire after the layout timer,
when the layout timer fires, we might as well just do the style recalc
too. The call to updateStyleIfNeeded() will cancel the pending style
recalc timer.

This doesn't have much impact on the number of layouts (measured via PLT)
but seems like a reasonable thing to do.

  • page/FrameView.cpp:

(WebCore::FrameView::layoutTimerFired):

3:06 PM Changeset in webkit [180845] by ddkilzer@apple.com
  • 2 edits in trunk/Tools

[iOS] Stop running webkit-build-directory on every layout test

This was originally fixed by David Farler for Bug 135409 in
r172602.

  • Scripts/webkitpy/port/ios.py:

(IOSSimulatorPort.relay_path): Mark as @memoized so it doesn't
run webkit-build-directory every time it's called.

2:47 PM Changeset in webkit [180844] by youenn.fablet@crf.canon.fr
  • 2 edits in trunk/Tools

W3C importer should use filesystem instead of shutil/host
​https://bugs.webkit.org/show_bug.cgi?id=142012

Reviewed by Bem Jones-Bey.

Removed direct use of python shutil and os, except for os.walk which will require its own fix.

  • Scripts/webkitpy/w3c/test_importer.py:

(main):
(TestImporter.do_import):
(TestImporter.find_importable_tests):
(TestImporter.import_tests):
(TestImporter.remove_deleted_files):
(TestImporter.write_import_log):

2:33 PM Changeset in webkit [180843] by Csaba Osztrogonác
  • 2 edits in trunk/LayoutTests

REGRESSION(r85798): Lists of crashing/timeouting/stderr tests aren't sorted
​https://bugs.webkit.org/show_bug.cgi?id=142081

Reviewed by Ryosuke Niwa.

  • fast/harness/results.html:
2:20 PM Changeset in webkit [180842] by andersca@apple.com
  • 4 edits in trunk/Source/WebKit2

WebsiteDataStore should handle fetching and deleting local storage data
​https://bugs.webkit.org/show_bug.cgi?id=142137

Reviewed by Sam Weinig.

  • UIProcess/Storage/StorageManager.cpp:

(WebKit::StorageManager::deleteEntriesForOrigins):
Add a new function that deletes entries from multiple origins.

  • UIProcess/Storage/StorageManager.h:

Add new members.

  • UIProcess/WebsiteData/WebsiteDataStore.cpp:

(WebKit::WebsiteDataStore::fetchData):
Fetch local storage data as well.

(WebKit::WebsiteDataStore::removeData):
Delete local storage data as well.

1:52 PM Changeset in webkit [180841] by andersca@apple.com
  • 3 edits in trunk/Source/WebKit2

Fetch cache origins from the network process
​https://bugs.webkit.org/show_bug.cgi?id=142135

Reviewed by Dan Bernstein.

  • NetworkProcess/NetworkProcess.cpp:

(WebKit::cfURLCacheOrigins):
Add a new helper function that returns a vector of CFURL cache origins.

(WebKit::fetchDiskCacheOrigins):
Fetch the disk cache origins and pass them along to the completion handler. Currently we don't handle the new disk cache.

(WebKit::NetworkProcess::fetchWebsiteData):
Create a callback aggregator and fetch disk cache origins if we're asked for it.

  • UIProcess/WebsiteData/WebsiteDataStore.cpp:

(WebKit::computeNetworkProcessAccessTypeForDataFetch):
New helper function that computes if we need to launch the network process in order to fetch data.

(WebKit::WebsiteDataStore::fetchData):
Fetch website data from the network process as well.

1:41 PM Changeset in webkit [180840] by youenn.fablet@crf.canon.fr
  • 3 edits in trunk/LayoutTests/imported/w3c/web-platform-tests

Unreviewed - set svn:ignore for generated web-platform-tests files - see bug 142110

12:58 PM Changeset in webkit [180839] by Simon Fraser
  • 31 edits in trunk/Source/WebCore

Fullscreen video layers are off by one sometimes
​https://bugs.webkit.org/show_bug.cgi?id=142122
rdar://problem/19878821

Reviewed by Eric Carlson.

Convert MediaPlayer::naturalSize() to return a FloatSize, since the natural size
isn't always integral (because of preserving pixel aspect ratio etc). Fix all the media
backends to use FloatSizes for natural size. Convert the video image drawing code
paths to FloatSize, since naturalSize is used on the destination rect computation,
and painting should be floating point anyway.

Give the layer created by SourceBufferPrivateAVFObjC a name in debug builds.

  • html/HTMLVideoElement.cpp:

(WebCore::HTMLVideoElement::videoWidth):
(WebCore::HTMLVideoElement::videoHeight):
(WebCore::HTMLVideoElement::paintCurrentFrameInContext):

  • html/HTMLVideoElement.h:
  • html/canvas/CanvasRenderingContext2D.cpp:

(WebCore::size):
(WebCore::CanvasRenderingContext2D::drawImage):

  • html/canvas/WebGLRenderingContextBase.cpp:

(WebCore::WebGLRenderingContextBase::videoFrameToImage):

  • platform/graphics/MediaPlayer.cpp:

(WebCore::NullMediaPlayerPrivate::naturalSize):
(WebCore::MediaPlayer::naturalSize):
(WebCore::MediaPlayer::paint):
(WebCore::MediaPlayer::paintCurrentFrameInContext):
(WebCore::NullMediaPlayerPrivate::paint): Deleted.

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

(WebCore::MediaPlayerPrivateInterface::paintCurrentFrameInContext):

  • platform/graphics/avfoundation/MediaPlayerPrivateAVFoundation.cpp:

(WebCore::MediaPlayerPrivateAVFoundation::naturalSize):
(WebCore::MediaPlayerPrivateAVFoundation::setNaturalSize):

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

(WebCore::MediaPlayerPrivateAVFoundationObjC::paintCurrentFrameInContext):
(WebCore::MediaPlayerPrivateAVFoundationObjC::paint):
(WebCore::MediaPlayerPrivateAVFoundationObjC::paintWithImageGenerator):
(WebCore::MediaPlayerPrivateAVFoundationObjC::createImageForTimeInRect):
(WebCore::MediaPlayerPrivateAVFoundationObjC::tracksChanged):
(WebCore::MediaPlayerPrivateAVFoundationObjC::sizeChanged):
(WebCore::MediaPlayerPrivateAVFoundationObjC::paintWithVideoOutput):

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

(WebCore::MediaPlayerPrivateMediaSourceAVFObjC::naturalSize):
(WebCore::MediaPlayerPrivateMediaSourceAVFObjC::paint):
(WebCore::MediaPlayerPrivateMediaSourceAVFObjC::paintCurrentFrameInContext):

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

(WebCore::MediaSourcePrivateAVFObjC::naturalSize):

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

(WebCore::SourceBufferPrivateAVFObjC::trackDidChangeEnabled):
(WebCore::SourceBufferPrivateAVFObjC::naturalSize):

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

(WebCore::VideoTrackPrivateMediaSourceAVFObjC::naturalSize):

  • platform/graphics/mac/MediaPlayerPrivateQTKit.h:
  • platform/graphics/mac/MediaPlayerPrivateQTKit.mm:

(WebCore::MediaPlayerPrivateQTKit::naturalSize):
(WebCore::MediaPlayerPrivateQTKit::paintCurrentFrameInContext):
(WebCore::MediaPlayerPrivateQTKit::paint):

  • platform/mock/mediasource/MockMediaPlayerMediaSource.cpp:

(WebCore::MockMediaPlayerMediaSource::naturalSize):
(WebCore::MockMediaPlayerMediaSource::paint):

  • platform/mock/mediasource/MockMediaPlayerMediaSource.h:
  • rendering/RenderVideo.cpp:

(WebCore::RenderVideo::calculateIntrinsicSize):
(WebCore::RenderVideo::paintReplaced):

12:20 PM Changeset in webkit [180838] by andersca@apple.com
  • 4 edits in trunk/Source/WebKit2

Simplify WebResourceCacheManagerCFNet.mm code
​https://bugs.webkit.org/show_bug.cgi?id=142134

Reviewed by Dan Bernstein.

  • ENABLE(CACHE_PARTITIONING) is always true on Mac and iOS, so remove those #ifdefs.
  • Make cfURLCacheHostNamesWithCallback and clearCFURLCacheForHostNames public so they can be used by the new WebsiteDataStore code in an upcoming patch.
  • Change cfURLCacheHostNamesWithCallback to take an std::function instead of a block.
  • WebProcess/ResourceCache/WebResourceCacheManager.cpp:

(WebKit::WebResourceCacheManager::getCacheOrigins):

  • WebProcess/ResourceCache/WebResourceCacheManager.h:
  • WebProcess/ResourceCache/cf/WebResourceCacheManagerCFNet.mm:

(WebKit::partitionName):
(WebKit::WebResourceCacheManager::cfURLCacheHostNamesWithCallback):
(WebKit::WebResourceCacheManager::clearCFURLCacheForHostNames):
(WebKit::WebResourceCacheManager::cfURLCacheHostNames): Deleted.

10:45 AM Changeset in webkit [180837] by commit-queue@webkit.org
  • 10 edits
    1 copy
    1 move
    2 adds in trunk/Source/WebKit2

Unreviewed, rolling out r180804.
​https://bugs.webkit.org/show_bug.cgi?id=142131

Broke nightlies (Requested by ap on #webkit).

Reverted changeset:

"[WK2] Drop legacy WKBundlePageDiagnosticLoggingClient API"
​https://bugs.webkit.org/show_bug.cgi?id=141176
​http://trac.webkit.org/changeset/180804

7:05 AM Changeset in webkit [180836] by gyuyoung.kim@samsung.com
  • 2 edits in trunk/LayoutTests

Unreviewed EFL gardening 1st Mar.

  • platform/efl/TestExpectations:
  • Unskip tests which have been passed since r180672.
  • Skip compositing/webgl.
5:22 AM WebKitGTK/2.8.x edited by Carlos Garcia Campos
(diff)
5:20 AM Changeset in webkit [180835] by Carlos Garcia Campos
  • 2 edits in releases/WebKitGTK/webkit-2.8/Source/WTF

Merge r180620 - Enable concurrent JIT on GTK
​https://bugs.webkit.org/show_bug.cgi?id=142007

Reviewed by Benjamin Poulain.

Seems weird that GTK keeps it off. No good reason for that as far as I can tell.

  • wtf/Platform.h:
5:20 AM Changeset in webkit [180834] by Carlos Garcia Campos
  • 2 edits
    1 add in releases/WebKitGTK/webkit-2.8/Source/JavaScriptCore

CMake build of libllvmForJSC.so should limit its export list like the Xcode build does
​https://bugs.webkit.org/show_bug.cgi?id=141989

Reviewed by Gyuyoung Kim.

  • CMakeLists.txt:
  • llvm/library/libllvmForJSC.version: Added.
3:59 AM Changeset in webkit [180833] by Carlos Garcia Campos
  • 5 edits in releases/WebKitGTK/webkit-2.8/Source/bmalloc

Merge r180604 - Rolling out ​http://trac.webkit.org/changeset/180430 as it causes the PLT to crash.
<rdar://problem/19948015>

Unreviewed.

  • bmalloc/VMHeap.cpp:

(bmalloc::VMHeap::grow):

  • bmalloc/VMHeap.h:
  • bmalloc/Zone.cpp:

(bmalloc::Zone::Zone):
(bmalloc::Zone::size): Deleted.

  • bmalloc/Zone.h:
3:55 AM Changeset in webkit [180832] by Carlos Garcia Campos
  • 15 edits
    10 adds in releases/WebKitGTK/webkit-2.8

Merge r180600 - AX: Implement support for ARIA 1.1 'switch' role
​https://bugs.webkit.org/show_bug.cgi?id=141986

Reviewed by Chris Fleizach.

Source/WebCore:

Map the role to ATK_ROLE_TOGGLE_BUTTON for Gtk and Efl; on the Mac, to
AXCheckBox with a subrole of AXSwitch. Ensure it looks and acts like a
widget to accessibility APIs (supports and emits notifications when
toggled, doesn't have children, exposes a name and description when
provided).

Tests: accessibility/aria-switch-checked.html

accessibility/aria-switch-sends-notification.html
accessibility/aria-switch-text.html

  • accessibility/AccessibilityNodeObject.cpp:

(WebCore::AccessibilityNodeObject::canHaveChildren):
(WebCore::AccessibilityNodeObject::isChecked):
(WebCore::AccessibilityNodeObject::visibleText):
(WebCore::AccessibilityNodeObject::title):

  • accessibility/AccessibilityObject.cpp:

(WebCore::AccessibilityObject::isARIAInput):
(WebCore::AccessibilityObject::actionVerb):
(WebCore::initializeRoleMap):
(WebCore::AccessibilityObject::supportsChecked):
(WebCore::AccessibilityObject::checkboxOrRadioValue):

  • accessibility/AccessibilityObject.h:

(WebCore::AccessibilityObject::isSwitch):

  • accessibility/atk/AXObjectCacheAtk.cpp:

(WebCore::AXObjectCache::postPlatformNotification):

  • accessibility/atk/WebKitAccessibleWrapperAtk.cpp:

(atkRole):

  • accessibility/ios/WebAccessibilityObjectWrapperIOS.mm:

(-[WebAccessibilityObjectWrapper accessibilityCanFuzzyHitTest]):
(-[WebAccessibilityObjectWrapper accessibilityTraits]):
(-[WebAccessibilityObjectWrapper determineIsAccessibilityElement]):

  • accessibility/mac/WebAccessibilityObjectWrapperMac.mm:

(createAccessibilityRoleMap):
(-[WebAccessibilityObjectWrapper subrole]):
(-[WebAccessibilityObjectWrapper accessibilityAttributeValue:]):

LayoutTests:

  • accessibility/aria-switch-checked-expected.txt: Added.
  • accessibility/aria-switch-checked.html: Added.
  • accessibility/aria-switch-sends-notification-expected.txt: Added.
  • accessibility/aria-switch-sends-notification.html: Added.
  • accessibility/aria-switch-text.html: Added.
  • accessibility/roles-exposed.html: Added a test case for the new role.
  • platform/efl/accessibility/aria-fallback-roles-expected.txt: Added.
  • platform/efl/accessibility/aria-switch-text-expected.txt: Added.
  • platform/efl/accessibility/roles-exposed-expected.txt: Updated for the new role.
  • platform/gtk/accessibility/aria-fallback-roles-expected.txt: Added.
  • platform/gtk/accessibility/aria-switch-text-expected.txt: Added.
  • platform/gtk/accessibility/roles-exposed-expected.txt: Updated for the new role.
  • platform/mac-mavericks/accessibility/roles-exposed-expected.txt: Updated for the new role.
  • platform/mac/TestExpectations: Skip the 'checked' notifcation as the Mac doesn't have it.
  • platform/mac/accessibility/aria-switch-text-expected.txt: Added.
  • platform/mac/accessibility/roles-exposed-expected.txt: Updated for the new role.
3:52 AM Changeset in webkit [180831] by Carlos Garcia Campos
  • 6 edits in releases/WebKitGTK/webkit-2.8/Source/JavaScriptCore

Merge r180591 - Rolling out r179753. The fix was invalid.
<​https://webkit.org/b/141990>

Not reviewed.

  • API/tests/testapi.mm:

(threadMain):
(useVMFromOtherThread): Deleted.
(useVMFromOtherThreadAndOutliveVM): Deleted.

  • heap/Heap.cpp:

(JSC::Heap::Heap):
(JSC::Heap::~Heap):
(JSC::Heap::gatherStackRoots):

  • heap/Heap.h:

(JSC::Heap::machineThreads):

  • heap/MachineStackMarker.cpp:

(JSC::MachineThreads::Thread::Thread):
(JSC::MachineThreads::MachineThreads):
(JSC::MachineThreads::~MachineThreads):
(JSC::MachineThreads::addCurrentThread):
(JSC::MachineThreads::removeThread):
(JSC::MachineThreads::removeCurrentThread):

  • heap/MachineStackMarker.h:
3:49 AM Changeset in webkit [180830] by youenn.fablet@crf.canon.fr
  • 2 edits in trunk/Tools

W3C test importer should use filesystem to read and write files
​https://bugs.webkit.org/show_bug.cgi?id=142084

Reviewed by Bem Jones-Bey.

Use of FileSystem.write_binary_file, read_text_file and write_text_file in lieu of open().

  • Scripts/webkitpy/w3c/test_importer.py:

(TestImporter.import_tests):
(TestImporter.remove_deleted_files):
(TestImporter.write_import_log):

3:31 AM Changeset in webkit [180829] by Carlos Garcia Campos
  • 13 edits
    1 add in releases/WebKitGTK/webkit-2.8/Source/bmalloc

Merge r180576 - bmalloc: Added a little more abstraction for large objects
​https://bugs.webkit.org/show_bug.cgi?id=141978

Reviewed by Sam Weinig.

Previously, each client needed to manage the boundary tags of
a large object using free functions. This patch introduces a LargeObject
class that does things a little more automatically.

  • bmalloc.xcodeproj/project.pbxproj:
  • bmalloc/Allocator.cpp:

(bmalloc::Allocator::reallocate): Use the new LargeObject class.

  • bmalloc/BeginTag.h:

(bmalloc::BeginTag::isInFreeList): Deleted. Moved this logic into the
LargeObject class.

  • bmalloc/BoundaryTag.h:

(bmalloc::BoundaryTag::isSentinel):
(bmalloc::BoundaryTag::compactBegin):
(bmalloc::BoundaryTag::setRange):
(bmalloc::BoundaryTag::initSentinel): Added an explicit API for sentinels,
which we used to create and test for implicitly.

  • bmalloc/BoundaryTagInlines.h:

(bmalloc::BoundaryTag::init):
(bmalloc::validate): Deleted.
(bmalloc::validatePrev): Deleted.
(bmalloc::validateNext): Deleted.
(bmalloc::BoundaryTag::mergeLeft): Deleted.
(bmalloc::BoundaryTag::mergeRight): Deleted.
(bmalloc::BoundaryTag::merge): Deleted.
(bmalloc::BoundaryTag::deallocate): Deleted.
(bmalloc::BoundaryTag::split): Deleted.
(bmalloc::BoundaryTag::allocate): Deleted. Moved this logic into the
LargeObject class.

  • bmalloc/EndTag.h:

(bmalloc::EndTag::init):
(bmalloc::EndTag::operator=): Deleted. Re-reading this code, I found
special behavior in the assignment operator to be a surprising API.
So, I replaced the assignment operation with an explicit initializing
function.

  • bmalloc/Heap.cpp:

(bmalloc::Heap::scavengeLargeRanges):
(bmalloc::Heap::allocateXLarge):
(bmalloc::Heap::findXLarge):
(bmalloc::Heap::deallocateXLarge):
(bmalloc::Heap::allocateLarge):
(bmalloc::Heap::deallocateLarge):

  • bmalloc/Heap.h: No behavior changes here -- just adopting the

LargeObject interface.

  • bmalloc/LargeObject.h: Added.

(bmalloc::LargeObject::operator!):
(bmalloc::LargeObject::begin):
(bmalloc::LargeObject::size):
(bmalloc::LargeObject::range):
(bmalloc::LargeObject::LargeObject):
(bmalloc::LargeObject::setFree):
(bmalloc::LargeObject::isFree):
(bmalloc::LargeObject::hasPhysicalPages):
(bmalloc::LargeObject::setHasPhysicalPages):
(bmalloc::LargeObject::isValidAndFree):
(bmalloc::LargeObject::merge):
(bmalloc::LargeObject::split):
(bmalloc::LargeObject::validateSelf):
(bmalloc::LargeObject::validate): Moved this code into a class, out of
BoundaryTag free functions.

New to the class are these features:

(1) Every reference to an object is validated upon creation and use.

(2) There's an explicit API for "This is a reference to an object
that might be stale (the DoNotValidate API)".

(3) The begin and end tags are kept in sync automatically.

  • bmalloc/SegregatedFreeList.cpp:

(bmalloc::SegregatedFreeList::insert):
(bmalloc::SegregatedFreeList::takeGreedy):
(bmalloc::SegregatedFreeList::take):

  • bmalloc/SegregatedFreeList.h: Adopt the LargeObject interface.
  • bmalloc/VMHeap.cpp:

(bmalloc::VMHeap::grow):

  • bmalloc/VMHeap.h:

(bmalloc::VMHeap::allocateLargeRange):
(bmalloc::VMHeap::deallocateLargeRange): Adopt the LargeObject interface.

3:29 AM Changeset in webkit [180828] by Carlos Garcia Campos
  • 50 edits
    2 adds in releases/WebKitGTK/webkit-2.8

Merge r180570 - REGRESSION(r179429): Can't type comments in Facebook
​https://bugs.webkit.org/show_bug.cgi?id=141859

Reviewed by Brent Fulgham.

Source/JavaScriptCore:

When window.Symbol is exposed to user-space pages,
Facebook's JavaScript use it (maybe, for immutable-js and React.js's unique key).
However, to work with Symbols completely, it also requires
1) Object.getOwnPropertySymbols (for mixin including Symbols)
2) the latest ES6 Iterator interface that uses Iterator.next and it returns { done: boolean, value: value }.
Since they are not landed yet, comments in Facebook don't work.

This patch introduces RuntimeFlags for JavaScriptCore.
Specifying SymbolEnabled flag under test runner and inspector to continue to work with Symbol.
And drop JavaScriptExperimentsEnabled flag
because it is no longer used and use case of this is duplicated to runtime flags.

(GlobalObject::javaScriptRuntimeFlags):
(GlobalObject::javaScriptExperimentsEnabled): Deleted.

  • runtime/JSGlobalObject.cpp:

(JSC::JSGlobalObject::JSGlobalObject):
(JSC::JSGlobalObject::init):

  • runtime/JSGlobalObject.h:

(JSC::JSGlobalObject::finishCreation):
(JSC::JSGlobalObject::javaScriptRuntimeFlags):
(JSC::JSGlobalObject::javaScriptExperimentsEnabled): Deleted.

  • runtime/RuntimeFlags.h: Added.

(JSC::RuntimeFlags::RuntimeFlags):
(JSC::RuntimeFlags::createAllEnabled):

Source/WebCore:

Enable SymbolEnabled runtime flag in inspector context.

  • ForwardingHeaders/runtime/RuntimeFlags.h: Added.
  • WebCore.order:
  • WebCore.vcxproj/WebCore.vcxproj:
  • WebCore.vcxproj/WebCore.vcxproj.filters:
  • bindings/js/JSDOMWindowBase.cpp:

(WebCore::JSDOMWindowBase::javaScriptRuntimeFlags):
(WebCore::JSDOMWindowBase::javaScriptExperimentsEnabled): Deleted.

  • bindings/js/JSDOMWindowBase.h:
  • bindings/js/JSWorkerGlobalScopeBase.cpp:

(WebCore::JSWorkerGlobalScopeBase::javaScriptRuntimeFlags):
(WebCore::JSWorkerGlobalScopeBase::javaScriptExperimentsEnabled): Deleted.

  • bindings/js/JSWorkerGlobalScopeBase.h:
  • inspector/InspectorFrontendClientLocal.cpp:

(WebCore::InspectorFrontendClientLocal::InspectorFrontendClientLocal):

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

Source/WebKit/mac:

Introduce SymbolEnabled and drop javaScriptExperimentsEnabled.
Private API, javaScriptExperimentsEnabled is dropped.

  • Misc/WebNSDictionaryExtras.h:
  • Misc/WebNSDictionaryExtras.m:

(-[NSMutableDictionary _webkit_setUnsignedInt:forKey:]):

  • WebKit.order:
  • WebView/WebPreferenceKeysPrivate.h:
  • WebView/WebPreferences.mm:

(+[WebPreferences initialize]):
(-[WebPreferences _setUnsignedIntValue:forKey:]):
(-[WebPreferences javaScriptRuntimeFlags]):
(-[WebPreferences setJavaScriptRuntimeFlags:]):
(-[WebPreferences setJavaScriptExperimentsEnabled:]): Deleted.
(-[WebPreferences javaScriptExperimentsEnabled]): Deleted.

  • WebView/WebPreferencesPrivate.h:
  • WebView/WebView.mm:

(-[WebView _preferencesChanged:]):

Source/WebKit/win:

Added Windows support.

  • Interfaces/IWebPreferences.idl:
  • Interfaces/IWebPreferencesPrivate.idl:
  • WebPreferenceKeysPrivate.h:
  • WebPreferences.cpp:

(WebPreferences::initializeDefaultSettings):
(WebPreferences::javaScriptRuntimeFlags):
(WebPreferences::setJavaScriptRuntimeFlags):
(WebPreferences::isWebSecurityEnabled):

  • WebPreferences.h:
  • WebView.cpp:

(WebView::notifyPreferencesChanged):

Source/WebKit2:

Enable SymbolEnabled in inspector context.

  • Shared/WebPreferencesDefinitions.h:
  • UIProcess/API/C/WKPreferences.cpp:

(WKPreferencesSetJavaScriptRuntimeFlags):
(WKPreferencesGetJavaScriptRuntimeFlags):
(WKPreferencesSetJavaScriptExperimentsEnabled): Deleted.
(WKPreferencesGetJavaScriptExperimentsEnabled): Deleted.

  • UIProcess/API/C/WKPreferencesRef.h:
  • UIProcess/API/C/WKPreferencesRefPrivate.h:
  • UIProcess/API/Cocoa/WKPreferences.mm:

(-[WKPreferences _javaScriptRuntimeFlags]):
(-[WKPreferences _setJavaScriptRuntimeFlags:]):

  • UIProcess/API/Cocoa/WKPreferencesPrivate.h:
  • UIProcess/efl/WebInspectorProxyEfl.cpp:

(WebKit::WebInspectorProxy::platformCreateInspectorPage):

  • UIProcess/gtk/WebInspectorProxyGtk.cpp:

(WebKit::WebInspectorProxy::platformCreateInspectorPage):

  • UIProcess/mac/WebInspectorProxyMac.mm:

(WebKit::WebInspectorProxy::platformCreateInspectorPage):

  • WebProcess/WebPage/WebPage.cpp:

(WebKit::WebPage::updatePreferences):

  • mac/WebKit2.order:

Tools:

Drop javaScriptExperimentsEnabled and specify JavaScriptRuntimeFlagsAllEnabled as KJavaScriptRuntimeFlags.

  • DumpRenderTree/mac/DumpRenderTree.mm:

(resetWebPreferencesToConsistentValues):

  • DumpRenderTree/win/DumpRenderTree.cpp:

(resetWebPreferencesToConsistentValues):

  • WebKitTestRunner/TestController.cpp:

(WTR::TestController::resetPreferencesToConsistentValues):

3:16 AM WebKitGTK/2.8.x edited by Carlos Garcia Campos
(diff)
3:14 AM WebKitGTK/2.8.x edited by Carlos Garcia Campos
(diff)
3:13 AM WebKitGTK/2.8.x edited by Carlos Garcia Campos
(diff)
3:05 AM Changeset in webkit [180827] by Carlos Garcia Campos
  • 4 edits in releases/WebKitGTK/webkit-2.8

Merge r180566 - [GTK] Layout Test accessibility/roles-exposed.html is failing
​https://bugs.webkit.org/show_bug.cgi?id=141960

Reviewed by Martin Robinson.

Source/WebCore:

The test was failing because Gtk now uses GtkColorChooserDialog for the
color input, making the input field a button which results in the color
chooser dialog appearing. As a side effect of this change, the input now
has an accessible role of ColorWell, which is currently mapped to
ATK_ROLE_COLOR_CHOOSER (which is by definition the dialog which results
upon activating the button input field). Changed the Gtk platform mapping
to ATK_ROLE_BUTTON, leaving the Efl mapping as-is.

No new tests. Instead, updated and unskipped failing test.

  • accessibility/atk/WebKitAccessibleWrapperAtk.cpp:

(atkRole):

LayoutTests:

  • platform/gtk/TestExpectations: Unskip the failing test.
  • platform/gtk/accessibility/roles-exposed-expected.txt: Update the expectations.
3:03 AM Changeset in webkit [180826] by Carlos Garcia Campos
  • 5 edits in releases/WebKitGTK/webkit-2.8

Merge r180565 - Crash loading local file with WebPageProxy::loadAlternateHTMLString
​https://bugs.webkit.org/show_bug.cgi?id=141867

Patch by Michael Catanzaro <Michael Catanzaro> on 2015-02-24
Reviewed by Anders Carlsson.

Source/WebKit2:

WebPageProxy::loadAlternateHTMLString needs to assume read access to unreachableURL as well
as baseURL, because unreachableURL will get added to the back/forward list, causing us to
crash later on when we notice the unexpected URL received in checkURLReceivedFromWebProcess.

  • UIProcess/WebPageProxy.cpp:

(WebKit::WebPageProxy::loadAlternateHTMLString):

Tools:

  • TestWebKitAPI/Tests/WebKit2/LoadAlternateHTMLStringWithNonDirectoryURL.cpp:

(TestWebKitAPI::loadAlternateHTMLString): Split most of this test into a function so it can
be shared with the new test.
(TestWebKitAPI::TEST): Add a cross-platform test for this crash.

  • TestWebKitAPI/Tests/WebKit2Gtk/TestLoaderClient.cpp: Add a GTK+ test for this crash.

(testLoadAlternateHTMLForLocalPage):
(beforeAll):

3:00 AM Changeset in webkit [180825] by Carlos Garcia Campos
  • 3 edits in releases/WebKitGTK/webkit-2.8/Source/WebCore

Merge r180563 - [GTK] Fonts loaded via @font-face look bad
​https://bugs.webkit.org/show_bug.cgi?id=140994

Patch by Michael Catanzaro <Michael Catanzaro> on 2015-02-24
Reviewed by Martin Robinson.

We've had several complaints that woff fonts look bad on some websites. This seems to be a
combination of multiple issues. For one, we don't look at Fontconfig settings at all when
creating a web font. This commit changes FontPlatformData::initializeWithFontFace to instead
use sane default settings from Fontconfig when loading a web font, rather than accepting the
default settings from GTK+, which are normally overridden by Fontconfig when loading system
fonts. However, we will hardcode the hinting setting for web fonts to always force use of
the light autohinter, so that we do not use a font's native hints. This avoids compatibility
issues when fonts with poor native hinting are only tested in browsers that do not use the
native hints. It also allows us to sidestep future security vulnerabilities in FreeType's
bytecode interpreter.

The net result of this is that there should be little noticable difference if you have GTK+
set to use slight hinting (which forces use of the autohinter) unless you have customized
Fontconfig configuration, but a dramatic improvement with particular fonts if you use medium
or full hinting. This should reduce complaints about our font rendering on "fancy sites."

No new tests, since the affected fonts we've found are not freely redistributable.

  • platform/graphics/freetype/FontCustomPlatformDataFreeType.cpp:

(WebCore::FontCustomPlatformData::FontCustomPlatformData): Force web fonts to be autohinted.

  • platform/graphics/freetype/FontPlatformDataFreeType.cpp:

(WebCore::getDefaultCairoFontOptions): Renamed to disambiguate.
(WebCore::getDefaultFontconfigOptions): Added.
(WebCore::FontPlatformData::initializeWithFontFace): Always call
FontPlatformData::setCairoOptionsFromFontConfigPattern. If the FontPlatformData was not
created with an FcPattern (e.g. because this is a web font), call
getDefaultFontconfigOptions to get a sane default FcPattern.
(WebCore::FontPlatformData::setOrientation): Renamed call to getDefaultCairoFontOptions.
(WebCore::getDefaultFontOptions): Deleted.

2:56 AM Changeset in webkit [180824] by Carlos Garcia Campos
  • 18 edits in releases/WebKitGTK/webkit-2.8

Merge r180558 - Always serialize :lang()'s arguments to strings
​https://bugs.webkit.org/show_bug.cgi?id=141944

Reviewed by Benjamin Poulain.

Source/WebCore:

As specified in [1] :lang()'s arguments are always serialized to strings.

[1] ​http://dev.w3.org/csswg/cssom/#serializing-selectors

Related tests are updated.

  • css/CSSGrammar.y.in:
  • css/CSSParserValues.cpp:

(WebCore::CSSParserSelector::setLangArgumentList):

  • css/CSSParserValues.h:

(WebCore::CSSParserString::init):
(WebCore::CSSParserString::clear):
(WebCore::CSSParserString::tokenType): Deleted.
(WebCore::CSSParserString::setTokenType): Deleted.

  • css/CSSSelector.cpp:

(WebCore::appendLangArgumentList):
(WebCore::CSSSelector::setLangArgumentList):

  • css/CSSSelector.h:

(WebCore::CSSSelector::langArgumentList):

  • css/SelectorCheckerTestFunctions.h:

(WebCore::matchesLangPseudoClass):

  • cssjit/SelectorCompiler.cpp:

(WebCore::SelectorCompiler::addPseudoClassType):
(WebCore::SelectorCompiler::SelectorCodeGenerator::generateElementIsInLanguage):

LayoutTests:

Some tests results are updated to reflect the always serialize
:lang()'s arguments to strings.

  • fast/css/css-lang-selector-with-string-arguments-text-expected.txt:
  • fast/css/css-lang-selector-with-string-arguments-text.html:
  • fast/css/parsing-css-lang-expected.txt:
  • fast/css/parsing-css-lang.html:
  • fast/css/css-selector-text-expected.txt:
  • fast/css/css-selector-text.html:
  • fast/css/css-set-selector-text-expected.txt:
  • fast/css/css-set-selector-text.html:
  • fast/dom/css-selectorText-expected.txt:
2:51 AM Changeset in webkit [180823] by Carlos Garcia Campos
  • 2 edits in releases/WebKitGTK/webkit-2.8/Source/WebCore

Merge r180554 - Give TemporaryChange for m_inLoadPendingImages assertion a name so it actually does something.

  • css/StyleResolver.cpp:

(WebCore::StyleResolver::loadPendingImages):

1:31 AM Changeset in webkit [180822] by Carlos Garcia Campos
  • 5 edits in releases/WebKitGTK/webkit-2.8/Source/WebCore

Merge r180548 - EventHandler references deleted Scrollbar
​https://bugs.webkit.org/show_bug.cgi?id=141931
<rdar://problem/19915210>

Reviewed by Tim Horton.

Tested by scrollbars/overflow-custom-scrollbar-crash.html

Update the EventHandler class to use a WeakPtr to reference the
last used Scrollbar, rather than retaining the Scrollbar and
artificially extending its life. This keeps the EventHandler
state in proper sync with the state of the render tree, and
avoids cases where we have destroyed a ScrollableArea (and
Scrollbar) but are still sending messages to a fake zombie
version of the element.

  • page/EventHandler.cpp:

(WebCore::EventHandler::clear):
(WebCore::EventHandler::handleMousePressEvent):
(WebCore::EventHandler::updateMouseEventTargetNode):
(WebCore::EventHandler::updateLastScrollbarUnderMouse):

  • page/EventHandler.h:
  • platform/Scrollbar.cpp:

(WebCore::Scrollbar::Scrollbar): Initialize WeakPtrFactory.

  • platform/Scrollbar.h:

(WebCore::Scrollbar::createWeakPtr): Added,

1:26 AM Changeset in webkit [180821] by Carlos Garcia Campos
  • 4 edits in releases/WebKitGTK/webkit-2.8

Merge r180535 - WTF::WeakPtr should rename 'forgot' to 'clear' and support nullptr assignment
​https://bugs.webkit.org/show_bug.cgi?id=141935

Reviewed by Myles C. Maxfield.

Source/WTF:

  • wtf/WeakPtr.h:

(WTF::WeakPtr::operator=): Added 'nullptr_t' overload.
(WTF::WeakPtr::clear): Renamed from 'forget'
(WTF::WeakPtr::forget): Deleted.

Tools:

  • TestWebKitAPI/Tests/WTF/WeakPtr.cpp:

(TestWebKitAPI::TEST): Updated for 'clear' method rename, and added a few
tests for assigning from nullptr.

1:08 AM Changeset in webkit [180820] by Carlos Garcia Campos
  • 4 edits in releases/WebKitGTK/webkit-2.8

Merge r180528 - Source/WTF:
WTF::WeakPtr should have a 'forget' method
​https://bugs.webkit.org/show_bug.cgi?id=141923

Reviewed by Myles C. Maxfield.

  • wtf/WeakPtr.h:

(WTF::WeakPtr::forget): Added.

Tools:
WTF::WeakPtr should have a 'forget' method.
​https://bugs.webkit.org/show_bug.cgi?id=141923

Reviewed by Myles C. Maxfield.

  • TestWebKitAPI/Tests/WTF/WeakPtr.cpp:

(TestWebKitAPI::TEST): Added 'Forget' tests case.

12:41 AM Changeset in webkit [180819] by Carlos Garcia Campos
  • 3 edits
    2 adds in releases/WebKitGTK/webkit-2.8

Merge r180530 - Default value of HTMLSelectElement size IDL attribute should be 0.
​https://bugs.webkit.org/show_bug.cgi?id=141795

Reviewed by Andreas Kling.

Source/WebCore:

Default value of HTMLSelectElement size IDL attribute should be 0.
As in spec: ​http://www.w3.org/html/wg/drafts/html/master/forms.html#the-select-element, also this matches the behavior of Chrome, IE and
Gecko.

Test: fast/dom/select-size.html

  • html/HTMLSelectElement.cpp:

(WebCore::HTMLSelectElement::parseAttribute):

LayoutTests:

  • fast/dom/select-size-expected.txt: Added.
  • fast/dom/select-size.html: Added.
12:38 AM Changeset in webkit [180818] by Carlos Garcia Campos
  • 2 edits
    1 add in releases/WebKitGTK/webkit-2.8/Source/JavaScriptCore

Merge r180516 - r9 is volatile on ARMv7 for iOS 3 and up.
​https://bugs.webkit.org/show_bug.cgi?id=141489
rdar://problem/19432916

Reviewed by Michael Saboff.

  • jit/RegisterSet.cpp:

(JSC::RegisterSet::calleeSaveRegisters): removed r9 from the list of ARMv7 callee save registers.

  • tests/stress/regress-141489.js: Added.

(foo):

Note: See TracTimeline for information about the timeline view.