Timeline
Jun 1, 2014:
- 11:52 PM Changeset in webkit [169524] by
-
- 2 edits in trunk/LayoutTests
Unreviewed EFL gardening.
gradient-background-leakage-hidpi.html and canvas-as-image-hidpi.html pass after r169445.
- platform/efl-wk2/TestExpectations:
- 10:16 PM Changeset in webkit [169523] by
-
- 19 edits2 adds in trunk
[MSE] Appends of overlapping sample data do not clear existing samples properly.
https://bugs.webkit.org/show_bug.cgi?id=133435
Reviewed by Darin Adler.
Source/WebCore:
Test: media/media-source/media-source-overlapping-append.html
The MSE spec has a spec error in how it removes overlapping frames. The intention seems to be
to check for overlapping samples only when the incoming sample increases the highest presentation time
field. But due to a mismatch in the range of a sample, defined as [start, end), and the check, which
treats the end as inclusive, the overlapping check is almost never run.
Add the sample comparison logic to match [start, end) ranges, and rename the existing check to differentiate
it from the new one:
- Modules/mediasource/SampleMap.cpp:
(WebCore::SamplePresentationTimeIsInsideRangeComparator::operator()):
(WebCore::SamplePresentationTimeIsWithinRangeComparator::operator()):
(WebCore::SampleMap::findSamplesBetweenPresentationTimes):
(WebCore::SampleMap::findSamplesWithinPresentationRange):
- Modules/mediasource/SampleMap.h:
Update the overlapping check to occur whenever the highest presentation time increases, and update the
logic within that check to catch all overlapping frames.
- Modules/mediasource/SourceBuffer.cpp:
(WebCore::SourceBuffer::sourceBufferPrivateDidReceiveInitializationSegment):
(WebCore::SourceBuffer::sourceBufferPrivateDidReceiveSample):
(WebCore::SourceBuffer::bufferedSamplesForTrackID):
- Modules/mediasource/SourceBuffer.h:
- WebCore.exp.in:
Add a dump method to MediaSample so that samples can be easily converted to a string for testing.
- platform/MediaSample.h:
- platform/graphics/avfoundation/objc/SourceBufferPrivateAVFObjC.mm:
(WebCore::MediaSampleAVFObjC::presentationSize):
(WebCore::MediaSampleAVFObjC::dump):
Update the Mock MSE implementation to pass along a "generation" field, to aid in testing.
- platform/mock/mediasource/MockBox.cpp:
(WebCore::MockSampleBox::MockSampleBox):
- platform/mock/mediasource/MockBox.h:
- platform/mock/mediasource/MockSourceBufferPrivate.cpp:
(WebCore::MockMediaSample::dump):
Add a method to internals to dump the buffered samples to string.
- Modules/mediasource/SourceBuffer.cpp:
(WebCore::SourceBuffer::bufferedSamplesForTrackID):
- Modules/mediasource/SourceBuffer.h:
- testing/Internals.cpp:
(WebCore::Internals::bufferedSamplesForTrackID):
- testing/Internals.h:
- testing/Internals.idl:
Source/WTF:
Add a dump method to MediaTime, so that MediaTimes can be easily converted to strings (for logging purposes).
- wtf/MediaTime.cpp:
(WTF::MediaTime::dump):
- wtf/MediaTime.h:
LayoutTests:
- media/media-source/media-source-overlapping-append-expected.txt: Added.
- media/media-source/media-source-overlapping-append.html: Added.
- media/media-source/mock-media-source.js:
- 10:01 PM Changeset in webkit [169522] by
-
- 14 edits6 adds in trunk
getBBox() returns (0,0) when width or height is zero.
https://bugs.webkit.org/show_bug.cgi?id=93290
Patch by Nikos Andronikos <nikos.andronikos-webkit@cisra.canon.com.au> on 2014-06-01
Reviewed by Philip Rogers.
Source/WebCore:
The SVG spec states that a width or height of zero for rects, circles and
ellipses is valid but disables rendering of that shape.
In this case, getBBox() should return the actual bounding box of the element.
Previously, WebKit always returned (0,0,0,0) rather than the
correct bounding box. This patch fixes that.
Rects and ellipses both now follow the same behaviour where
the fall-back code path is only used when the shape will render (e.g. not
when a dimension <= 0).
This necessitates calling calculateRadiiAndCenter for ellipses even
when the calculated values will be discarded in the fall back code path.
But calculateRadiiAndCenter is cheap and this avoids making changes
elsewhere to stop CG drawing a zero width (or height) ellipse.
Tests: svg/custom/getBBox-js-circle-zerodimension.html
svg/custom/getBBox-js-ellipse-zerodimension.html
svg/custom/getBBox-js-rect-zerodimension.html
- rendering/svg/RenderSVGEllipse.cpp:
(WebCore::RenderSVGEllipse::updateShapeFromElement):
Only follow fall-back code path when shape renders.
width height==0 is not an error case so calculate bounding box. - rendering/svg/RenderSVGRect.cpp:
(WebCore::RenderSVGRect::updateShapeFromElement):
Only follow fall-back code path when shape renders.
rx ry==0 is not an error case so calculate bounding box. LayoutTests:
- platform/mac/svg/W3C-SVG-1.1/shapes-circle-02-t-expected.txt:
- platform/mac/svg/W3C-SVG-1.1/shapes-ellipse-02-t-expected.txt:
- platform/mac/svg/W3C-SVG-1.1/shapes-intro-01-t-expected.txt:
- platform/mac/svg/W3C-SVG-1.1/shapes-rect-02-t-expected.txt:
- svg/W3C-SVG-1.1/animate-elem-32-t-expected.txt:
- svg/W3C-SVG-1.1/shapes-circle-02-t-expected.txt:
- svg/W3C-SVG-1.1/shapes-ellipse-02-t-expected.txt:
- svg/W3C-SVG-1.1/shapes-intro-01-t-expected.txt:
- svg/W3C-SVG-1.1/shapes-rect-02-t-expected.txt:
- svg/custom/getBBox-js-circle-zerodimension-expected.txt: Added.
- svg/custom/getBBox-js-circle-zerodimension.html: Added.
- svg/custom/getBBox-js-ellipse-zerodimension-expected.txt: Added.
- svg/custom/getBBox-js-ellipse-zerodimension.html: Added.
- svg/custom/getBBox-js-rect-zerodimension-expected.txt: Added.
- svg/custom/getBBox-js-rect-zerodimension.html: Added.
- 6:53 PM Changeset in webkit [169521] by
-
- 2 edits in trunk/LayoutTests
EFL gardening. Unskip some iframe tests which have passed since buildbot OS bump up.
- platform/efl-wk1/TestExpectations:
- 5:33 PM Changeset in webkit [169520] by
-
- 3 edits in trunk/Source/WebKit2
[Cocoa] Add SPI to get a WebArchive of the WKWebView
<rdar://problem/16748120>
https://bugs.webkit.org/show_bug.cgi?id=133439
Reviewed by Anders Carlsson.
- UIProcess/API/Cocoa/WKWebView.mm:
(-[WKWebView _getWebArchiveDataWithCompletionHandler:]):
- UIProcess/API/Cocoa/WKWebViewPrivate.h:
- 4:07 PM Changeset in webkit [169519] by
-
- 2 edits in trunk/Source/WebCore
Don't scroll to non-integral scroll offsets
https://bugs.webkit.org/show_bug.cgi?id=133436
<rdar://problem/17088336>
Reviewed by Darin Adler.
With the gradual migration of scroll offsets to floating point, we ended
up with a fully floating point code path been event scroll deltas and
setting scroll positions in the scrolling tree.
With some input devices (e.g. Logitech scrollwheel mice), the scroll deltas
come in with non-integral values, and that caused the scroll offset to be non-integral,
resulting in fuzzy layer contents. So round the scroll position.
- page/scrolling/mac/ScrollingTreeFrameScrollingNodeMac.mm:
(WebCore::ScrollingTreeFrameScrollingNodeMac::setScrollPosition):
- 3:31 PM Changeset in webkit [169518] by
-
- 20 edits in trunk/Source
Add a LazyNeverDestroyed class template and use it
https://bugs.webkit.org/show_bug.cgi?id=133425
Reviewed by Darin Adler.
Source/JavaScriptCore:
- dfg/DFGFunctionWhitelist.cpp:
(JSC::DFG::FunctionWhitelist::ensureGlobalWhitelist):
- dfg/DFGFunctionWhitelist.h:
Source/WebCore:
- Modules/webdatabase/DatabaseBackendBase.cpp:
(WebCore::guidMutex):
- crypto/CryptoAlgorithmRegistry.cpp:
(WebCore::registryMutex):
- inspector/WorkerDebuggerAgent.cpp:
- loader/CrossOriginAccessControl.cpp:
(WebCore::isOnAccessControlResponseHeaderWhitelist):
- platform/network/NetworkStateNotifier.cpp:
(WebCore::networkStateNotifier):
- workers/DefaultSharedWorkerRepository.cpp:
(WebCore::DefaultSharedWorkerRepository::instance):
- workers/DefaultSharedWorkerRepository.h:
- workers/WorkerThread.cpp:
(WebCore::threadSetMutex):
- xml/XMLHttpRequest.cpp:
(WebCore::staticData):
Source/WebKit2:
- Shared/mac/SecItemShim.cpp:
(WebKit::responseMap):
Source/WTF:
LazyNeverDestroyed is similar to NeverDestroyed, except it's lazily constructed
by calling construct(). This makes it useful for using inside std::call_once functions.
- wtf/Forward.h:
- wtf/NeverDestroyed.h:
(WTF::LazyNeverDestroyed::construct):
(WTF::LazyNeverDestroyed::operator T&):
(WTF::LazyNeverDestroyed::get):
(WTF::LazyNeverDestroyed::asPtr):
(WTF::LazyNeverDestroyed::MaybeRelax::MaybeRelax):
- wtf/mac/DeprecatedSymbolsUsedBySafari.mm:
(WTF::atomicallyInitializedStaticMutex):
- wtf/unicode/icu/CollatorICU.cpp:
(WTF::cachedCollatorMutex):
- 2:12 PM Changeset in webkit [169517] by
-
- 3 edits in trunk/PerformanceTests
DYEBench: Move test states into benchmarkClient and remove the closure
https://bugs.webkit.org/show_bug.cgi?id=133438
Reviewed by Benjamin Poulain.
Moved all local variables in the closure wrapping benchmarkClient onto the object itself
and removed the closure to improve the readability of the code.
- DoYouEvenBench/Full.html:
- DoYouEvenBench/resources/main.js:
(window.benchmarkClient.willRunTest):
(window.benchmarkClient.didRunTest):
(window.benchmarkClient.didRunSuites):
(window.benchmarkClient.willStartFirstIteration):
(window.benchmarkClient.didFinishLastIteration):
(window.benchmarkClient._addResult): Moved.
(.): Deleted.
- 1:38 PM Changeset in webkit [169516] by
-
- 2 edits2 adds in trunk/PerformanceTests
DYEBench: Split stylesheets and scripts in Full.html into separate files
https://bugs.webkit.org/show_bug.cgi?id=133437
Reviewed by Benjamin Poulain.
Extracted main.js and main.css.
Also fixed a bug in startBenchmark that disabled suites were counted towards the total number of tests.
- DoYouEvenBench/Full.html:
- DoYouEvenBench/resources/main.css: Added.
- DoYouEvenBench/resources/main.js: Added.
(.addResult):
(window.benchmarkClient):
(startBenchmark): Renamed from startTest.
- 9:30 AM Changeset in webkit [169515] by
-
- 4 edits4 adds in trunk
Subpixel rendering: Selection gap produces a pixel line overlap on inline boxes.
https://bugs.webkit.org/show_bug.cgi?id=133429
Reviewed by Simon Fraser.
Do not floor logical right/left values when selection gap is calculated for RenderBlocks.
It produces overlapping line/gap because the floored value is intersecting/too far from the
non-floored content selection rect.
Source/WebCore:
Tests: fast/inline/hidpi-selection-gap-and-inline-selection-have-gap-rtl.html
fast/inline/hidpi-selection-gap-overlaps-inline-selection.html
- rendering/RenderBlock.cpp:
(WebCore::RenderBlock::logicalLeftSelectionGap):
(WebCore::RenderBlock::logicalRightSelectionGap):
LayoutTests:
- platform/mac/TestExpectations: add ML failure due to font sizing. (Ahem can't be used here)
- fast/inline/hidpi-selection-gap-and-inline-selection-have-gap-rtl-expected.html: Added.
- fast/inline/hidpi-selection-gap-and-inline-selection-have-gap-rtl.html: Added.
- fast/inline/hidpi-selection-gap-overlaps-inline-selection-expected.html: Added.
- fast/inline/hidpi-selection-gap-overlaps-inline-selection.html: Added.
- 6:44 AM Changeset in webkit [169514] by
-
- 9 edits in trunk/PerformanceTests
DYEBench: CSS parser warning at line 106 of base.css
https://bugs.webkit.org/show_bug.cgi?id=133433
Reviewed by Maciej Stachowiak.
Removed the old filter CSS property for Internet Explorer. The latest Internet Explorer supports -ms-linear-gradient
on background CSS property anyways.
It's interesting that some of the tests only have -webkit- prefixes. Perhaps we should update those subtests.
- DoYouEvenBench/Full.html:
- DoYouEvenBench/resources/todomvc/architecture-examples/angularjs/bower_components/todomvc-common/base.css:
(#header:before):
- DoYouEvenBench/resources/todomvc/architecture-examples/backbone/bower_components/todomvc-common/base.css:
(#header:before):
- DoYouEvenBench/resources/todomvc/architecture-examples/emberjs/bower_components/todomvc-common/base.css:
(#header:before):
- DoYouEvenBench/resources/todomvc/architecture-examples/jquery/bower_components/todomvc-common/base.css:
(#header:before):
- DoYouEvenBench/resources/todomvc/dependency-examples/flight/bower_components/todomvc-common/base.css:
(#header:before):
- DoYouEvenBench/resources/todomvc/labs/architecture-examples/react/bower_components/todomvc-common/base.css:
(#header:before):
- DoYouEvenBench/resources/todomvc/vanilla-examples/vanillajs/bower_components/todomvc-common/base.css:
(#header:before):
- 12:01 AM Changeset in webkit [169513] by
-
- 3 edits in trunk/PerformanceTests
DYEBench: Ember.js assertion hit at line 20593
https://bugs.webkit.org/show_bug.cgi?id=133431
Reviewed by Darin Adler.
The assertion was hit because ToDoMVC includes jQuery 2.1 and Ember.js 1.3.1 only recognizes jQuery 2.0.
Port the assertion from Ember.js 1.5.1 to suppress the assertion.
We should update the entire Ember.js at some point but this would do the job for now.
- DoYouEvenBench/Full.html:
- DoYouEvenBench/resources/todomvc/architecture-examples/emberjs/bower_components/ember/ember.js:
May 31, 2014:
- 11:26 PM Changeset in webkit [169512] by
-
- 3 edits in trunk/Source/WebCore
REGRESSION (r169412): Garbage pixels when vertical rubber-banding on http://yvoschaap.com/pmarca/
https://bugs.webkit.org/show_bug.cgi?id=133418
<rdar://problem/17065494>
Reviewed by Sam Weinig.
- WebCore.exp.in:
- platform/graphics/FloatPoint.h:
(WebCore::toFloatPoint):
- platform/graphics/Image.cpp:
(WebCore::Image::drawTiled):
The manual tiling code previously adjusted the initial
tiling source rect so that if the pattern was out of phase,
the first row and column would paint only the necessary part
of the tile. However, it did not also adjust the destination rect,
so the tile was painted in entirely the wrong place.
Additional testing (by enabling manual tiling mode for all patterned
images and running all of the layout tests) exposed another bug
where the last row and column would paint too much, which
was resolved by clipping to the outer image's destination rect.
- 11:26 PM Changeset in webkit [169511] by
-
- 2 edits in trunk/LayoutTests
One more missed piece of that last check-in.
- TestExpectations: Removed the flakiness expectation from the
css3/viewport-percentage-lengths/viewport-percentage-lengths-relative-font-size.html
test since that now works reliably.
- 11:25 PM Changeset in webkit [169510] by
-
- 2 edits in trunk/Source/WebKit2
View snapshots are sometimes taken at the wrong scale
https://bugs.webkit.org/show_bug.cgi?id=133419
<rdar://problem/17087497>
Reviewed by Sam Weinig.
- UIProcess/API/mac/WKView.mm:
(-[WKView _takeViewSnapshot]):
Work around <rdar://problem/17084993> by re-requesting the snapshot at
kCGWindowImageNominalResolution if it was captured at the wrong scale.
- 11:14 PM Changeset in webkit [169509] by
-
- 3 edits in trunk/LayoutTests
Try to fix another failure from the previous check-in seen on some bots.
- platform/mac-wk2/TestExpectations: Removed a couple obsolete lines.
- platform/wk2/TestExpectations: Added expectation for filter-hidden-content.svg, which
seems to be affected by the resizing for tests before it.
- 8:48 PM Changeset in webkit [169508] by
-
- 3 edits in trunk/LayoutTests
Fix test failure from the previous check-in.
- fast/dynamic/window-resize-scrollbars-test.html: Add resize handler so this test works
with asynchronous resizing.
- platform/wk2/TestExpectations: Expect failure for the test above; it's showing scrollbars,
and it should not. Before, the test was not running correctly so it seemed to pass.
- 8:04 PM Changeset in webkit [169507] by
-
- 3 edits in trunk/PerformanceTests
DYEBench: Prevent frame flattening on iOS
https://bugs.webkit.org/show_bug.cgi?id=133428
Reviewed by Andreas Kling.
Prevent frame flattening on iOS by setting scrolling=no.
- DoYouEvenBench/Full.html:
- DoYouEvenBench/resources/benchmark-runner.js:
(BenchmarkRunner.prototype._appendFrame):
- 6:34 PM Changeset in webkit [169506] by
-
- 2 edits in trunk/Source/WebCore
CSS JIT cleanup: move two utility functions out of the properties-matcher section
https://bugs.webkit.org/show_bug.cgi?id=133427
Reviewed by Andreas Kling.
Just move the code, no other modification.
- cssjit/SelectorCompiler.cpp:
(WebCore::SelectorCompiler::SelectorCodeGenerator::jumpIfNoPreviousAdjacentElement):
(WebCore::SelectorCompiler::SelectorCodeGenerator::jumpIfNoNextAdjacentElement):
- 5:01 PM Changeset in webkit [169505] by
-
- 26 edits in trunk
Viewport percentage tests that resize the viewport are flaky
https://bugs.webkit.org/show_bug.cgi?id=133351
Reviewed by Anders Carlsson.
Source/WebCore:
- page/FrameView.cpp:
(WebCore::FrameView::setFrameRect): Removed code that explicitly sends a resize event here.
The resize event will be sent as a side effect of layout instead.
(WebCore::FrameView::sendResizeEventIfNeeded): Don't send a resize event until layout is done,
since this is what propagates the new geometry in so the JavaScript code can detect it. Also
simplified the logic and streamlined a little, and added a FIXME about the bad idea here where
the code enqueues an event for an unknown time in the future. Removed the unneeded code that
passes in the DOM window as the target to dispatchWindowEvent, since that gets filled in as the
target by default.
Tools:
- DumpRenderTree/mac/DumpRenderTree.mm:
(sizeWebViewForCurrentTest): Added a call to reset the window origin, which
happens to be stored in the UI delegate object. This is part of resetting the
window position.
- DumpRenderTree/mac/UIDelegate.h: Store the window origin rather than storing
a frame rectangle. The size is stored in the size of the web view itself.
- DumpRenderTree/mac/UIDelegate.mm:
(-[UIDelegate resetWindowOrigin]): Added.
(-[UIDelegate webView:setFrame:]): Changed to set the size of the web view. The
old version would store the frame, but not resize the web view!
(-[UIDelegate webViewFrame:]): Get the size from the web view.
- WebKitTestRunner/efl/PlatformWebViewEfl.cpp:
(WTR::PlatformWebView::setWindowFrame): Added code to resize the web view.
- WebKitTestRunner/gtk/PlatformWebViewGtk.cpp:
(WTR::PlatformWebView::resizeTo): Added a comment about missing code to resize
the window.
- WebKitTestRunner/ios/PlatformWebViewIOS.mm:
(WTR::PlatformWebView::resizeTo): Changed this to call setWindowFrame so the
window gets resized too, not just the web view.
(WTR::PlatformWebView::setWindowFrame): Added code to set the size of the web view.
- WebKitTestRunner/mac/PlatformWebViewMac.mm:
(WTR::PlatformWebView::resizeTo): Changed this to call setWindowFrame so the
window gets resized too, not just the web view.
(WTR::PlatformWebView::setWindowFrame): Added code to set the size of the web view.
LayoutTests:
- css3/viewport-percentage-lengths/resources/resize-test.js:
Added code to compensate for the difference between innerWidth/Height and outerWidth/Height
so this works well in a browser window. Added code to "warm up" so the first resize triggers
a resize event. Tweaked style of the code a bit. Use resize events to avoid flakiness due
to race conditions. One thing I did not fix is the "out of order" problem where all the
test results get written out after the words TEST COMPLETE.
- css3/viewport-percentage-lengths/viewport-percentage-lengths-anonymous-block-expected.txt:
Updated due to changes to the underlying test machinery, and also to show actual expected success.
The old results said PASS with various obviously wrong results.
- css3/viewport-percentage-lengths/viewport-percentage-lengths-anonymous-block.html:
Changed this test to use the shared resize-test.js.
- css3/viewport-percentage-lengths/viewport-percentage-lengths-percent-size-child-expected.txt:
Updated due to changes to the underlying test machinery, and also to show actual expected success.
The old results said PASS with various obviously wrong results.
- css3/viewport-percentage-lengths/viewport-percentage-lengths-percent-size-child.html:
Streamlined this test a little.
- css3/viewport-percentage-lengths/viewport-percentage-lengths-relative-font-size-expected.txt:
Updated due to changes to the underlying test machinery, and also to show actual expected success.
The old results said PASS with various obviously wrong results.
- css3/viewport-percentage-lengths/viewport-percentage-lengths-relative-font-size.html:
Streamlined this test a little.
- css3/viewport-percentage-lengths/viewport-percentage-lengths-resize-expected.txt:
Updated due to changes to the underlying test machinery, and also to show both expected success,
and some expected failures. These failures were going unnoticed before since the test didn't
successfully do any resizing of the web view.
- css3/viewport-percentage-lengths/viewport-percentage-lengths-resize.html:
Streamlined this test a little.
- fast/dom/Window/window-resize-contents-expected.txt: Updated to reflect the test actually
resizing. The old test results expected no resizing to occur.
- fast/dom/Window/window-resize-contents.html: Updated to use a resize event handler so the
test can successfully test asynchronous resizes like the ones done by WebKit2.
- fast/dom/rtl-scroll-to-leftmost-and-resize.html: Updated to use a resize event handler so
this can test asynchronous resizes. Also fixed code that was treating the arguments to resizeTo
as if they were inner coordinates rather than outer. However, the bug this tests for has not
been fixed, so need to expect failures.
- platform/mac/TestExpectations: Added a failure expection for the test above.
- platform/mac/fast/dynamic/window-resize-scrollbars-test-expected.png: Updated to expect a result
that successfully resized.
- platform/mac/fast/dynamic/window-resize-scrollbars-test-expected.txt: Ditto.
- 1:29 PM Changeset in webkit [169504] by
-
- 2 edits in trunk/Source/WebKit2
[iOS] WebKit2.framework is unused
https://bugs.webkit.org/show_bug.cgi?id=133424
Reviewed by Anders Carlsson.
- Configurations/WebKit2.xcconfig: Set SKIP_INSTALL to YES when targeting iOS.
- 10:41 AM Changeset in webkit [169503] by
-
- 2 edits in trunk/Source/WebCore
Don't use std::mutex::try_lock to check if we own a mutex
https://bugs.webkit.org/show_bug.cgi?id=133421
Reviewed by Dan Bernstein.
According the C++ standard, calling try_lock on a mutex that's already owned by the current thread
is undefined behavior, so don't do that.
- Modules/webdatabase/DatabaseBackendBase.cpp:
(WebCore::guidToVersionMap):
(WebCore::updateGuidVersionMap):
(WebCore::guidToDatabaseMap):
(WebCore::guidForOriginAndName):
