Timeline
Apr 25, 2021:
- 10:51 PM Changeset in webkit [276581] by
-
- 2 edits in trunk/Tools
Add an option to not enable all experimental features in DumpRenderTree
https://bugs.webkit.org/show_bug.cgi?id=225041
Reviewed by Wenson Hsieh.
Added --no-enable-all-experimental-features to DumpRenderTree as we did
for WebKitTestRunner in r276559.
- DumpRenderTree/mac/DumpRenderTree.mm:
(setWebPreferencesForTestOptions):
(initializeGlobalsFromCommandLineOptions):
- 10:24 PM Changeset in webkit [276580] by
-
- 4 edits in trunk/LayoutTests
Unreviewesd EWS fixes.
Skip http/tests/security/referrer-policy-header-multipart.html on WebKit1
after r276526 split it out of http/tests/security/referrer-policy-header.html
- platform/mac-wk1/TestExpectations:
- platform/win/TestExpectations:
- platform/wincairo/TestExpectations:
- 7:05 PM Changeset in webkit [276579] by
-
- 2 edits in trunk/LayoutTests
[WinCairo] Unreviewed test gardening
- platform/wincairo/TestExpectations:
- 6:03 PM Changeset in webkit [276578] by
-
- 6 edits2 adds in trunk
Fix a crash caused by AVIF decoding failure
https://bugs.webkit.org/show_bug.cgi?id=223843
<rdar://problem/76205084>
Reviewed by Philippe Normand.
Source/WebCore:
A crash happens when the image count is accessed in AVIFImageDecoder::tryDecodeSize.
This is because an instance of AVIFImageReader is destroyed by setFailed()
in AVIFImageReader::parseHeader. Deciding a failure was hasty because
parsing can fail when necessary data is not received yet. To fix this,
we call setFailed in AVIFImageReader::parseHeader only when all data is received.
Test: http/tests/images/avif-partial-load-crash.html
- platform/image-decoders/avif/AVIFImageDecoder.cpp:
(WebCore::AVIFImageDecoder::tryDecodeSize):
- platform/image-decoders/avif/AVIFImageReader.cpp:
(WebCore::AVIFImageReader::parseHeader):
(WebCore::AVIFImageReader::repetitionCount const):
- platform/image-decoders/avif/AVIFImageReader.h:
LayoutTests:
Check if partial load of avif images crashes.
- TestExpectations: Skip the test for non-gtk platforms.
- http/tests/images/avif-partial-load-crash-expected.txt: Added.
- http/tests/images/avif-partial-load-crash.html: Added.
- 3:59 PM Changeset in webkit [276577] by
-
- 6 edits in trunk
Fix handling of overflow of /= and *= operators with double over Checked<uint64_t> and other 64-bit types.
https://bugs.webkit.org/show_bug.cgi?id=224835
Reviewed by Chris Dumez.
Source/WebCore:
- platform/graphics/gpu/cocoa/GPUBindGroupAllocatorMetal.mm:
(WebCore::GPUBindGroupAllocator::reallocate): Use integer math instead of floating point math to increase
the size by 25% each time through the loop, staying within the capabilities of our checked integer class.
Source/WTF:
- wtf/CheckedArithmetic.h: Since it is quite difficult to write correct implementations of the
operations where we multiply or divide a checked integer by a floating point number, deleted
those functions, which were barely used. Also corrected return types of various member functions,
which were unconventional "const Type", replacing them with the conventional "Type" for functions
that return a distinct value, "Type&" for assignment operators, and bool for relational operators.
Tools:
- TestWebKitAPI/Tests/WTF/CheckedArithmeticOperations.cpp:
Removed tests of the "/=" operator with floating point arguments, no longer supported.
- 3:53 PM Changeset in webkit [276576] by
-
- 5 edits in trunk/Source/WebCore
Remove templated operator<<(TextStream&, T*) that interferes with raw pointer formatting.
https://bugs.webkit.org/show_bug.cgi?id=224962
Reviewed by Darin Adler.
We can end up calling into this one (which unhelpfully casts its
argument to intptr_t) instead of theconst void*one in
TextStream.h.
Removing this means we can drop static_cast<const void*>(...) in
various places in debugging output code.
- rendering/RenderLayer.cpp:
(WebCore::outputPaintOrderTreeRecursive):
- rendering/RenderTreeAsText.cpp:
(WebCore::RenderTreeAsText::writeRenderObject):
- rendering/svg/SVGRenderTreeAsText.cpp:
(WebCore::writeStandardPrefix):
- rendering/svg/SVGRenderTreeAsText.h:
- 3:06 PM Changeset in webkit [276575] by
-
- 24 edits1 delete in trunk/Source/WebCore
Make more IndexedDB objects iso-heap'ed
https://bugs.webkit.org/show_bug.cgi?id=225029
Reviewed by Darin Adler.
Deployed IsoHeap on more IndexedDB objects.
- Modules/indexeddb/IDBFactory.cpp:
- Modules/indexeddb/IDBFactory.h:
- Modules/indexeddb/IDBGetAllResult.cpp:
- Modules/indexeddb/IDBGetAllResult.h:
- Modules/indexeddb/IDBGetResult.cpp:
- Modules/indexeddb/IDBGetResult.h:
- Modules/indexeddb/IDBIndex.cpp:
- Modules/indexeddb/IDBIndex.h:
- Modules/indexeddb/IDBKey.cpp:
- Modules/indexeddb/IDBKey.h:
- Modules/indexeddb/IDBObjectStore.cpp:
- Modules/indexeddb/IDBObjectStore.h:
- Modules/indexeddb/IDBRecordIdentifier.h: Removed since it's not used anywhere.
- Modules/indexeddb/IDBRequest.cpp:
- Modules/indexeddb/IDBRequest.h:
- Modules/indexeddb/client/IDBConnectionProxy.cpp:
- Modules/indexeddb/client/IDBConnectionProxy.h:
- Modules/indexeddb/client/IDBConnectionToServer.cpp:
- Modules/indexeddb/client/IDBConnectionToServer.h:
- Modules/indexeddb/client/TransactionOperation.cpp:
- Modules/indexeddb/client/TransactionOperation.h:
- Modules/indexeddb/shared/IDBDatabaseInfo.cpp:
- Modules/indexeddb/shared/IDBDatabaseInfo.h:
- WebCore.xcodeproj/project.pbxproj:
- 2:45 PM Changeset in webkit [276574] by
-
- 9 edits1 delete in trunk/Source
Add
ALWAYS_LOG_WITH_STREAMmacro for easier development/debugging
https://bugs.webkit.org/show_bug.cgi?id=224959
Reviewed by Darin Adler.
It's often very helpful to use existing
operator<<defined on objects, but annoying to
have to copypaste the boilerplate code that creates aWTF::TextStreamand uses it in a
WTFLogAlwaysor wait for a debug build to enable existingLOG_WITH_STREAM.
Source/WebCore:
- platform/Logging.h:
Move contents of
pal/LogMacros.htowtf/Assertions.h.
Source/WebCore/PAL:
- pal/LogMacros.h: Removed.
- PAL.xcodeproj/project.pbxproj:
- pal/CMakeLists.txt:
Move contents of
pal/LogMacros.htowtf/Assertions.h.
Source/WebKit:
- Platform/Logging.h:
Move contents of
pal/LogMacros.htowtf/Assertions.h.
Source/WTF:
- wtf/Assertions.h:
(LOG_WITH_STREAM): Moved from
Source/WebCore/pal/LogMacros.h.
(ALWAYS_LOG_WITH_STREAM): Added.
AddALWAYS_LOG_WITH_STREAMso that one change existingLOGwhen debugging with minimal
friction. As an example, many times I've wished I could've just changed
`
LOG_WITH_STREAM(Channel, stream << node);
`
to something like
`
ALWAYS_LOG_WITH_STREAM(stream << node);
`
and avoid having to modify the surrounding code too much.
- 2:09 PM Changeset in webkit [276573] by
-
- 2 edits in trunk/Source/WebInspectorUI
Web Inspector: Audit: REGRESSION(?): creating a new test in Edit mode causes the Results folder to appear
https://bugs.webkit.org/show_bug.cgi?id=224994
Reviewed by Darin Adler.
- UserInterface/Views/AuditNavigationSidebarPanel.js:
(WI.AuditNavigationSidebarPanel.prototype._addTest):
- 2:00 PM Changeset in webkit [276572] by
-
- 8 edits1 add in trunk
[css-flexbox] Table layout disregards overriding width
https://bugs.webkit.org/show_bug.cgi?id=224934
Patch by Felipe Erias <Felipe Erias> on 2021-04-25
Reviewed by Darin Adler.
LayoutTests/imported/w3c:
Update expectations.
- web-platform-tests/css/css-flexbox/table-with-percent-intrinsic-width-expected.txt: This test now passes.
Source/WebCore:
If a table's parent has set an overriding width on it, that value will get preference over the one from
the table's "width" property. This fixes several tests where the width assigned by the flexbox algorithm
did not match the width of the table after layout.
- rendering/AutoTableLayout.cpp:
(WebCore::AutoTableLayout::applyPreferredLogicalWidthQuirks const): Give precedence to the element's
overriding width if it has been set.
- rendering/RenderTable.cpp:
(WebCore::RenderTable::updateLogicalWidth): Give precedence to the element's overriding width if it
has been set, instead of discarding it if it is smaller than the style width.
(WebCore::RenderTable::computePreferredLogicalWidths): Take the overriding width into account when
computing the preferred min/max values for the element.
LayoutTests:
- TestExpectations: Remove three WPT tests that now pass.
- platform/glib/TestExpectations: Remove one WPT test that now passes.
- platform/gtk/imported/w3c/web-platform-tests/css/css-flexbox/flex-item-contains-strict-expected.txt: Added.
- 1:38 PM Changeset in webkit [276571] by
-
- 5 edits in trunk/Source/WebKit
[GPUP] REGRESSION: Selecting play/pause on Spotify.com causes page to reload
https://bugs.webkit.org/show_bug.cgi?id=225023
<rdar://76985997>
Reviewed by Darin Adler.
A zero ObjectIdentifier is apparently invalid, which causes a debug ASSERT on the sending
side and a SIGKILL on the receiving side. Replace the zero ObjectIdentifier message when
clearing a LegacyCDMSession with an Optional<ObjectIdentifier>, and send an explicit
WTF::nullopt when clearing.
- GPUProcess/media/RemoteMediaPlayerProxy.cpp:
(WebKit::RemoteMediaPlayerProxy::mediaPlayerCachedKeyForKeyId const):
(WebKit::RemoteMediaPlayerProxy::setLegacyCDMSession):
- GPUProcess/media/RemoteMediaPlayerProxy.h:
- GPUProcess/media/RemoteMediaPlayerProxy.messages.in:
- WebProcess/GPU/media/MediaPlayerPrivateRemote.cpp:
(WebKit::MediaPlayerPrivateRemote::setCDMSession):
- 1:32 PM Changeset in webkit [276570] by
-
- 5 edits in trunk/Source/WebKit
[GPUP][iOS] Silent video playback can interrupt system audio
https://bugs.webkit.org/show_bug.cgi?id=225031
<rdar://76652073>
Reviewed by Eric Carlson.
When the WebContent process asks the GPU Process to set the AVAudioSession audio session
category, the GPU Process as an optimization returns early if the session category being
requested is the same as has already been set. However, the default value of the category
is "None" (which translates to AVAudioSessionCategoryAmbient), and setting the category
to "None" becomes a no-op due to this default. As such, the GPUP never sets the underlying
AVAudioSession's category away from the default, which is AVAudioSessionCategorySoloAmbient,
and thus will interrupt other audio during playback.
Additionally, there's a subtle logic error where the audio session category is not changed
when a given WebContent process (and it's RemoteAudioSession & Proxy) goes away.
The fix for both of these issues is to re-calculate the correct audio session category
when a RemoteAudioSessionProxy is added or removed from RemoteAudioSessionProxyManager.
Since "None" is the default value for a RemoteAudioSessionProxy, the mere act of adding
a new RemoteAudioSessionProxy (which is created when a WebContent process is created)
will cause the audio session category to be set to AVAudioSessionCategoryAmbient.
- GPUProcess/media/RemoteAudioSessionProxy.cpp:
(WebKit::RemoteAudioSessionProxy::setCategory):
- GPUProcess/media/RemoteAudioSessionProxy.h:
- GPUProcess/media/RemoteAudioSessionProxyManager.cpp:
(WebKit::RemoteAudioSessionProxyManager::addProxy):
(WebKit::RemoteAudioSessionProxyManager::removeProxy):
(WebKit::RemoteAudioSessionProxyManager::updateCategory):
(WebKit::RemoteAudioSessionProxyManager::setCategoryForProcess): Deleted.
- GPUProcess/media/RemoteAudioSessionProxyManager.h:
- 12:17 PM Changeset in webkit [276569] by
-
- 7 edits in trunk/Source/ThirdParty/ANGLE
[Metal ANGLE] support primitive restart behavior with simple data types
https://bugs.webkit.org/show_bug.cgi?id=225011
Primitive restart behavior for simple data types is not defined in Metal. Add a pre-draw pass to
create a set of restart indices, rather than re-write the index buffer. This pass avoids
storing more copies of the index buffer: instead, we cache the location of restart indices, and
use them to build draw lists.
Patch by Kyle Piddington <Kyle Piddington> on 2021-04-25
Reviewed by Dean Jackson.
- src/libANGLE/renderer/metal/BufferMtl.h:
- src/libANGLE/renderer/metal/BufferMtl.mm:
(rx::IndexConversionBufferMtl::IndexConversionBufferMtl):
(rx::BufferMtl::markConversionBuffersDirty):
(rx::calculateRestartRanges):
(rx::BufferMtl::getRestartIndices):
- src/libANGLE/renderer/metal/ContextMtl.h:
- src/libANGLE/renderer/metal/ContextMtl.mm:
(rx::ContextMtl::drawLineLoopElements):
(rx::ContextMtl::drawElementsImpl):
(rx::ContextMtl::drawElementsSimpleTypesPrimitiveRestart): Deleted.
- src/libANGLE/renderer/metal/VertexArrayMtl.h:
- src/libANGLE/renderer/metal/VertexArrayMtl.mm:
(rx::VertexArrayMtl::getIndexBuffer):
(rx::VertexArrayMtl::getDrawIndices):
(rx::VertexArrayMtl::convertIndexBuffer):
(rx::VertexArrayMtl::streamIndexBufferFromClient):
- 12:16 PM Changeset in webkit [276568] by
-
- 2 edits in trunk/Source/ThirdParty/ANGLE
[Metal ANGLE] Temporarily remove AST validation even for debug builds
https://bugs.webkit.org/show_bug.cgi?id=224991
rdar://76299178
Patch by John Cunningham <johncunningham@apple.com> on 2021-04-25
Reviewed by Kenneth Russell.
- src/libANGLE/renderer/metal/ShaderMtl.mm:
(rx::ShaderMtl::compileImplMtl):
- 12:06 PM Changeset in webkit [276567] by
-
- 3 edits in trunk/Source/ThirdParty/ANGLE
[Metal ANGLE] Select proper IOSurface backing format for WebGL environment
https://bugs.webkit.org/show_bug.cgi?id=224948
<rdar://76284889>
Depending on the architecture, WebCore expects different
IOSurface texture targets for the main buffer. When running catalyst on
Arm64 devices, Metal-ANGLE needs to select the TEXTURE_2D texture target for our
IOSurface/Pbuffer bind point, rather than TEXTURE_RECTANGLE.
Reviewed by Dean Jackson.
Patch by Kyle Piddington <Kyle Piddington> on 2021-04-25
- src/libANGLE/renderer/metal/DisplayMtl.h:
- src/libANGLE/renderer/metal/DisplayMtl.mm:
(rx::needsEAGLOnMac):
(rx::DisplayMtl::EGLDrawingBufferTextureTarget):
(rx::DisplayMtl::generateConfigs):
- 11:35 AM Changeset in webkit [276566] by
-
- 2 edits in trunk/Source/WebKit
[WebXR] Should be enabled when WEBXR_INTERNALS is true
https://bugs.webkit.org/show_bug.cgi?id=225026
<rdar://problem/77111380>
Reviewed by Sam Weinig.
- Shared/WebPreferencesDefaultValues.cpp:
(WebKit::defaultWebXREnabled): Return true when HAVE(WEBXR_INTERNALS).
- 10:43 AM Changeset in webkit [276565] by
-
- 2 edits in trunk/Source/WebCore
css-contain category missing from CSSProperties.json
https://bugs.webkit.org/show_bug.cgi?id=224990
Patch by Tyler Wilcock <Tyler Wilcock> on 2021-04-25
Reviewed by Darin Adler.
- css/CSSProperties.json:
Add missing css-contain category to avoid style lint errors. Example
error:
ERROR: Source/WebCore/css/CSSProperties.json:0: property "contain" has category "css-contain" which is not in the set of categories. [json/syntax] [5]
- 10:39 AM Changeset in webkit [276564] by
-
- 2 edits in trunk/Tools
jsonchecker.py validate_string uses Python 2 basestring type instead of Python 2/3 str type
https://bugs.webkit.org/show_bug.cgi?id=224978
Patch by Tyler Wilcock <Tyler Wilcock> on 2021-04-25
Reviewed by Darin Adler.
validate_string uses the Python 2 basestring type instead of the Python 2/3
str type. This causes the check to exit early the first time it validates
a string with this error:
name 'basestring' is not defined
This resulted in jsonchecker not checking the file beyond the first
string validation.
I believe this caused jsonchecker to miss a bug in one of my previous
patches in which I put the "settings-flag" property in the wrong
place:
https://bugs.webkit.org/show_bug.cgi?id=224718#c20
- Scripts/webkitpy/style/checkers/jsonchecker.py:
(JSONCSSPropertiesChecker.validate_string):
validate_string with Python 2 / 3 str type instead of Python 2
basestring.
Apr 24, 2021:
- 10:59 PM Changeset in webkit [276563] by
-
- 4 edits in trunk/Source/WebCore
Deploy Ref/RefPtr in DeleteSelectionCommand
https://bugs.webkit.org/show_bug.cgi?id=225028
Reviewed by Wenson Hsieh.
Deployed smart pointers in DeleteSelectionCommand.
Also deployed ScriptDisallowedScope around the code which accesses the render tree.
No new tests since there should be no observable behavioral differences.
- editing/DeleteSelectionCommand.cpp:
(WebCore::isTableRowEmpty):
(WebCore::isSpecialHTMLElement): Moved from Editing.cpp.
(WebCore::firstInSpecialElement): Ditto.
(WebCore::lastInSpecialElement): Ditto.
(WebCore::positionBeforeContainingSpecialElement): Ditto. Now returns a pair instead of returning
Position and "returning" the special element via an out argument.
(WebCore::positionAfterContainingSpecialElement): Ditto.
(WebCore::DeleteSelectionCommand::initializeStartEnd):
(WebCore::DeleteSelectionCommand::initializePositionData):
(WebCore::DeleteSelectionCommand::handleSpecialCaseBRDelete):
(WebCore::firstEditablePositionInNode):
(WebCore::DeleteSelectionCommand::insertBlockPlaceholderForTableCellIfNeeded):
(WebCore::DeleteSelectionCommand::removeNode):
(WebCore::DeleteSelectionCommand::handleGeneralDelete):
(WebCore::DeleteSelectionCommand::mergeParagraphs):
(WebCore::DeleteSelectionCommand::removePreviouslySelectedEmptyTableRows):
(WebCore::DeleteSelectionCommand::originalStringForAutocorrectionAtBeginningOfSelection):
(WebCore::DeleteSelectionCommand::doApply):
- editing/Editing.cpp:
(WebCore::isRenderedTable): Return false when the node is not a HTMLElement for consistency.
(WebCore::isSpecialHTMLElement): Moved to DeleteSelectionCommand.cpp.
(WebCore::firstInSpecialElement): Ditto.
(WebCore::lastInSpecialElement): Ditto.
(WebCore::positionBeforeContainingSpecialElement): Ditto.
(WebCore::positionAfterContainingSpecialElement): Ditto.
- editing/Editing.h:
- 7:12 PM Changeset in webkit [276562] by
-
- 15 edits5 adds in trunk
Changing the source of a model element with clipping applied does not update the model
https://bugs.webkit.org/show_bug.cgi?id=224917
Reviewed by Simon Fraser.
Source/WebCore:
Tests: model-element/model-element-contents-layer-updates-with-clipping.html
model-element/model-element-contents-layer-updates.html
Previously, a <model> with a contents clipping layer (e.g. border-radius)
would not reparent its contents layer in the right place when setContentsToModel
was called again (because the source changed), leaving the old model
contents layer in place.
- platform/graphics/ca/GraphicsLayerCA.cpp:
(WebCore::GraphicsLayerCA::updateSublayerList):
Ensure that updateSublayerList always parents contentsLayer in one of its two homes:
under contentsClippingLayer, if it exists; otherwise, directly under the primary layer.
(WebCore::GraphicsLayerCA::setContentsToModel):
Drive-by fix a bug revealed by the tests for this patch: when swapping out the
contents layer in setContentsToModel, we also need to mark ContentsRectsChanged,
or the new contents layer will not get its bounds set during the subsequent flush.
(WebCore::GraphicsLayerCA::setContentsToPlatformLayer):
Remove special-case code that was added to fix this bug just for setContentsToPlatformLayer;
this case is now correctly handled for all contents layers by updateSublayerList.
(WebCore::GraphicsLayerCA::dumpInnerLayer const):
- platform/graphics/GraphicsLayerClient.h:
- platform/graphics/ca/PlatformCALayer.cpp:
(WebCore::PlatformCALayer::dumpAdditionalProperties):
- platform/graphics/ca/PlatformCALayer.h:
- testing/Internals.cpp:
(WebCore::toPlatformLayerTreeFlags):
- testing/Internals.h:
- testing/Internals.idl:
Add a bit to platformLayerTreeAsText() that makes PlatformCALayerRemoteModelHosting
dump the size of the model that it is hosting, which is used in the test for this bug.
Remove the IncludeOpacity bit since we can just always log opacity if it's not the default.
Source/WebKit:
- WebProcess/WebPage/RemoteLayerTree/PlatformCALayerRemoteModelHosting.h:
- WebProcess/WebPage/RemoteLayerTree/PlatformCALayerRemoteModelHosting.mm:
(WebKit::PlatformCALayerRemoteModelHosting::dumpAdditionalProperties):
Add a bit to platformLayerTreeAsText() that makes PlatformCALayerRemoteModelHosting
dump the size of the model that it is hosting, which is used in the test for this bug.
LayoutTests:
- model-element/model-element-contents-layer-updates-expected.txt: Added.
- model-element/model-element-contents-layer-updates-with-clipping-expected.txt: Added.
- model-element/model-element-contents-layer-updates-with-clipping.html: Added.
- model-element/model-element-contents-layer-updates.html: Added.
- model-element/resources/cube.usdz: Added.
- platform/ios-wk2/TestExpectations:
- platform/mac/TestExpectations:
Add tests that ensure that adding a <model> with one source, then changing
it to another, correctly updates the content layer. Test this both
with and without clipping (the without-clipping case passed before this change,
with-clipping failed).
These tests only work on Cocoa ports with UI-side compositing enabled
because they depend on the PlatformCALayer subclass holding on to the model
data (and logging its size) in order to distinguish between the two models.
- 5:50 PM Changeset in webkit [276561] by
-
- 2 edits in trunk/Tools
Rename slave to worker in webkitpy - part 2
https://bugs.webkit.org/show_bug.cgi?id=224988
Reviewed by Jonathan Bedard.
- Scripts/webkitpy/common/system/crashlogs_unittest.py:
- 5:47 PM Changeset in webkit [276560] by
-
- 1 copy in tags/Safari-611.2.5
Tag Safari-611.2.5.
- 5:39 PM Changeset in webkit [276559] by
-
- 8 edits in trunk/Tools
Add an option to not enable all experimental features in WebKitTestRunner
https://bugs.webkit.org/show_bug.cgi?id=224958
Reviewed by Tim Horton.
Added --no-enable-all-experimental-features to run-webkit-tests and WebKitTestRunner.
It causes a crash in DumpRenderTree for now. We should rectify this in the future
when we align the behaviors of WebKitTestRunner and DumpRenderTree.
- Scripts/webkitpy/layout_tests/run_webkit_tests.py:
(parse_args): Added --no-enable-all-experimental-features.
- Scripts/webkitpy/port/driver.py:
(Driver.cmd_line): Pass along the option to WebKitTestRunner.
- WebKitTestRunner/Options.cpp:
(WTR::handleOptionNoEnableAllExperimentalFeatures): Added.
(WTR::OptionsHandler::OptionsHandler): Added --no-enable-all-experimental-features.
- WebKitTestRunner/Options.h:
- WebKitTestRunner/TestController.cpp:
(WTR::TestController::initialize):
(WTR::TestController::resetPreferencesToConsistentValues):
- WebKitTestRunner/TestController.h:
- 3:31 PM Changeset in webkit [276558] by
-
- 4 edits2 adds in trunk
Crash in BreakBlockquoteCommand::doApply()
https://bugs.webkit.org/show_bug.cgi?id=224941
Patch by Julian Gonzalez <julian_a_gonzalez@apple.com> on 2021-04-24
Reviewed by Ryosuke Niwa.
Source/WebCore:
Despite assertions to the contrary, it is possible for there not to be any node
to move into the new blockquote in BreakBlockquoteCommand::doApply() as a result
of layout updates, so remove the assertions and handle this case.
Test: editing/pasteboard/paste-as-quotation-then-paste-crash.html
- editing/BreakBlockquoteCommand.cpp:
(WebCore::BreakBlockquoteCommand::doApply):
LayoutTests:
Add test for this crash, running only on Release for now.
Thanks to Tuomas Karkkainen for its basic structure.
- TestExpectations:
- editing/pasteboard/paste-as-quotation-then-paste-crash-expected.txt: Added.
- editing/pasteboard/paste-as-quotation-then-paste-crash.html: Added.
- 2:29 PM Changeset in webkit [276557] by
-
- 8 edits in branches/safari-611-branch/Source
Versioning.
WebKit-7611.2.5
- 2:21 PM Changeset in webkit [276556] by
-
- 8 edits in branches/safari-611-branch/Source/WebKit
Reland r275846 with Unreviewed crash fix. rdar://77106929
Corrects crash due to bad merge.
- 2:14 PM Changeset in webkit [276555] by
-
- 8 edits in branches/safari-611-branch/Source/WebKit
Revert "Re-land Cherry-pick r275846. rdar://problem/76727548"
This reverts commit r276398.
- 12:36 PM Changeset in webkit [276554] by
-
- 11 edits in trunk
Improve parsing and computed style of the rotate CSS property
https://bugs.webkit.org/show_bug.cgi?id=225019
Reviewed by Dean Jackson.
LayoutTests/imported/w3c:
Mark 2 WPT progressions and add some new parsing WPT tests for rotate.
- web-platform-tests/css/css-transforms/animation/rotate-composition-expected.txt: We don't support additive interpolation,
the FAIL result is most likely a test that passed without proper support for the feature which now fails due to lack of
support for this feature.
- web-platform-tests/css/css-transforms/parsing/rotate-parsing-valid-expected.txt:
- web-platform-tests/css/css-transforms/parsing/rotate-parsing-valid.html:
- web-platform-tests/web-animations/animation-model/animation-types/accumulation-per-property-002-expected.txt:
- web-platform-tests/web-animations/animation-model/animation-types/addition-per-property-002-expected.txt:
- web-platform-tests/web-animations/animation-model/animation-types/interpolation-per-property-002-expected.txt:
Source/WebCore:
There were a few issues with how we parsed the rotate CSS property and how we dealt with it
when resolving the computed style.
When parsing, we required individual vector coordinates to be 1 when others were 0 for a vector
to be considered parallel with the x, y or z axis. We now support any positive value. Then, as
we create the TransformOperation from the CSSValue, we must set the type to ROTATE_3D instead
of ROTATE_Z if the z axis was specified through the CSS value. Indeed, ROTATE_Z is equivalent
to ROTATE and without setting this type to ROTATE_3D, TransformOperation::is3DOperation() would
return false.
This is important because when figuring out the computed style, we should only return an angle
without a rotation vector if TransformOperation::is3DOperation() is false. For the case where
the operation was specified in 3D, we now use the same rule used when parsing to determine
whether the rotation vector is parallel with the x, y or z axis to use the matching keywords.
- css/CSSComputedStyleDeclaration.cpp:
(WebCore::computedRotate):
- css/TransformFunctions.cpp:
(WebCore::rotateForValue):
- css/parser/CSSPropertyParser.cpp:
(WebCore::consumeRotate):
- 12:35 PM Changeset in webkit [276553] by
-
- 6 edits in trunk
Support interpolation of the background-repeat shorthand
https://bugs.webkit.org/show_bug.cgi?id=225016
Reviewed by Dean Jackson.
LayoutTests/imported/w3c:
Mark 10 WPT progressions.
- web-platform-tests/web-animations/animation-model/animation-types/accumulation-per-property-001-expected.txt:
- web-platform-tests/web-animations/animation-model/animation-types/addition-per-property-001-expected.txt:
- web-platform-tests/web-animations/animation-model/animation-types/interpolation-per-property-001-expected.txt:
Source/WebCore:
- animation/CSSPropertyAnimation.cpp:
(WebCore::CSSPropertyAnimationWrapperMap::CSSPropertyAnimationWrapperMap):
- 12:33 PM Changeset in webkit [276552] by
-
- 7 edits in trunk
Fix interpolation of the clip CSS property
https://bugs.webkit.org/show_bug.cgi?id=225017
Reviewed by Dean Jackson.
LayoutTests/imported/w3c:
Mark 13 WPT progressions.
- web-platform-tests/web-animations/animation-model/animation-types/accumulation-per-property-001-expected.txt:
- web-platform-tests/web-animations/animation-model/animation-types/addition-per-property-001-expected.txt:
- web-platform-tests/web-animations/animation-model/animation-types/interpolation-per-property-001-expected.txt:
- web-platform-tests/web-animations/animation-model/keyframe-effects/effect-value-iteration-composite-operation-expected.txt:
Source/WebCore:
While we interpolated the clip value correctly, we wouldn't set the RenderStyle bit that
indicates that there is a non-auto value in the first place. With a new dedicated wrapper,
we can ensure we set the bit correctly as we blend, and that we use discrete interpolation
if both the from and to values aren't auto.
- animation/CSSPropertyAnimation.cpp:
(WebCore::CSSPropertyAnimationWrapperMap::CSSPropertyAnimationWrapperMap):
- 12:32 PM Changeset in webkit [276551] by
-
- 5 edits in trunk
Computed style for the translate CSS property should use px for the z value
https://bugs.webkit.org/show_bug.cgi?id=225018
Reviewed by Dean Jackson.
LayoutTests/imported/w3c:
Mark a single WPT progression.
- web-platform-tests/css/css-transforms/animation/translate-composition-expected.txt:
- web-platform-tests/web-animations/animation-model/animation-types/interpolation-per-property-002-expected.txt:
Source/WebCore:
- css/CSSComputedStyleDeclaration.cpp:
(WebCore::computedTranslate):
- 9:02 AM Changeset in webkit [276550] by
-
- 16 edits in trunk
Refactor ValueRange from enum to enum class
https://bugs.webkit.org/show_bug.cgi?id=224981
Patch by Tyler Wilcock <Tyler Wilcock> on 2021-04-24
Reviewed by Sam Weinig.
Source/WebCore:
Refactor ValueRange from:
enum ValueRange {
ValueRangeAll,
ValueRangeNonNegative
}
to:
enum ValueRange : uint8_t {
All,
NonNegative
}
making it smaller (one byte versus four) and harder to misuse (no
auto-coercion to number types).
- animation/CSSPropertyAnimation.cpp:
(WebCore::blendFunc):
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
- css/CSSCalculationValue.cpp:
(WebCore::CSSCalcValue::create):
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
- css/CSSCalculationValue.h:
(WebCore::CSSCalcValue::createCalculationValue const):
(WebCore::CSSCalcValue::setPermittedValueRange):
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
- css/MediaQueryExpression.cpp:
(WebCore::consumeFirstValue):
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
- css/parser/CSSPropertyParser.cpp:
(WebCore::CSSPropertyParser::consumeTransformOrigin):
(WebCore::consumeLetterSpacing):
(WebCore::consumeWordSpacing):
(WebCore::consumeTabSize):
(WebCore::consumeTextSizeAdjust):
(WebCore::consumeFontSize):
(WebCore::consumeLineHeight):
(WebCore::consumeSize):
(WebCore::consumeTextIndent):
(WebCore::consumeScrollPadding):
(WebCore::consumeMaxWidthOrHeight):
(WebCore::consumeWidthOrHeight):
(WebCore::consumeMarginOrOffset):
(WebCore::consumeClipComponent):
(WebCore::consumeLineClamp):
(WebCore::consumeHyphenateLimit):
(WebCore::consumeColumnWidth):
(WebCore::consumeGapLength):
(WebCore::consumeZoom):
(WebCore::consumeAnimationIterationCount):
(WebCore::consumeAnimationValue):
(WebCore::consumeLineWidth):
(WebCore::consumeTranslate3d):
(WebCore::consumeNumbers):
(WebCore::consumePerspective):
(WebCore::consumeTransformValue):
(WebCore::consumeTranslate):
(WebCore::consumeScale):
(WebCore::consumeRotate):
(WebCore::consumePositionLonghand):
(WebCore::consumeFlexBasis):
(WebCore::consumeKerning):
(WebCore::consumeStrokeDasharray):
(WebCore::consumeBaselineShift):
(WebCore::consumeRxOrRy):
(WebCore::consumeBorderRadiusCorner):
(WebCore::consumeTextUnderlineOffset):
(WebCore::consumeTextDecorationThickness):
(WebCore::consumeVerticalAlign):
(WebCore::consumeShapeRadius):
(WebCore::consumeBasicShapePolygon):
(WebCore::consumeRadii):
(WebCore::consumeBasicShapeInset):
(WebCore::consumeBorderImageSlice):
(WebCore::consumeBorderImageOutset):
(WebCore::consumeBorderImageWidth):
(WebCore::consumeReflect):
(WebCore::consumeBackgroundSize):
(WebCore::consumeFitContent):
(WebCore::consumeGridBreadth):
(WebCore::consumeInitialLetter):
(WebCore::consumeWebkitMarqueeIncrement):
(WebCore::consumeWebkitMarqueeRepetition):
(WebCore::consumeWebkitMarqueeSpeed):
(WebCore::consumeAspectRatio):
(WebCore::CSSPropertyParser::parseSingleValue):
(WebCore::CSSPropertyParser::consumeBorderSpacing):
(WebCore::CSSPropertyParser::consumeFlex):
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
- css/parser/CSSPropertyParserHelpers.cpp:
(WebCore::CSSPropertyParserHelpers::consumeNumberOrPercentDividedBy100Raw):
(WebCore::CSSPropertyParserHelpers::CalcParser::CalcParser):
(WebCore::CSSPropertyParserHelpers::consumeNumberRaw):
(WebCore::CSSPropertyParserHelpers::consumeFontWeightNumberRaw):
(WebCore::CSSPropertyParserHelpers::consumeLengthRaw):
(WebCore::CSSPropertyParserHelpers::consumePercentRaw):
(WebCore::CSSPropertyParserHelpers::consumeAngleRaw):
(WebCore::CSSPropertyParserHelpers::consumeAngleWorkerSafe):
(WebCore::CSSPropertyParserHelpers::consumeTime):
(WebCore::CSSPropertyParserHelpers::consumePositionComponent):
(WebCore::CSSPropertyParserHelpers::consumeDeprecatedGradientPoint):
(WebCore::CSSPropertyParserHelpers::consumeDeprecatedGradient):
(WebCore::CSSPropertyParserHelpers::consumeGradientColorStops):
(WebCore::CSSPropertyParserHelpers::consumeDeprecatedRadialGradient):
(WebCore::CSSPropertyParserHelpers::consumeRadialGradient):
(WebCore::CSSPropertyParserHelpers::consumeFilterFunction):
(WebCore::CSSPropertyParserHelpers::consumeSingleShadow):
(WebCore::CSSPropertyParserHelpers::consumeFontSizeRaw):
(WebCore::CSSPropertyParserHelpers::consumeLineHeightRaw):
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
- css/parser/CSSPropertyParserHelpers.h:
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
- css/parser/CSSPropertyParserWorkerSafe.cpp:
(WebCore::CSSPropertyParserHelpersWorkerSafe::consumeFontStretch):
(WebCore::CSSPropertyParserHelpersWorkerSafe::consumeFontStretchRange):
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
- page/IntersectionObserver.cpp:
(WebCore::parseRootMargin):
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
- platform/CalculationValue.h:
(WebCore::CalculationValue::CalculationValue):
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
- platform/Length.cpp:
(WebCore::convertTo100PercentMinusLength):
(WebCore::blendMixedTypes):
(WebCore::blend):
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
- platform/Length.h:
Refactor enum ValueRange to enum class ValueRange.
- style/StyleBuilderConverter.h:
(WebCore::Style::BuilderConverter::convertTo100PercentMinusLength):
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
Tools:
ValueRange has been refactored from enum to enum class : uint8_t to
make it smaller and harder to misuse.
- TestWebKitAPI/Tests/WebCore/CalculationValue.cpp:
(TestWebKitAPI::createTestValue):
Refactor ValueRangeAll and ValueRangeNonNegative to ValueRange::All
and ValueRange::NonNegative.
- 6:42 AM Changeset in webkit [276549] by
-
- 3 edits2 adds in trunk
[RenderTreeBuilder] Subtree moving should clear the floats on all the descendants
https://bugs.webkit.org/show_bug.cgi?id=224996
<rdar://76837320>
Reviewed by Antti Koivisto.
Source/WebCore:
While moving a subtree, we invalidate the floating object list so that we don't end up with incorrectly placed floats (they'll get regenerated during the subsequent layout).
A float can be "assigned" to more than one RenderBlockFlow (e.g intruding floats). It's very common that a set of descendant RenderBlockFlow
renderers "see" the same set of floats (each RenderBlockFlow has its own list of floating objects).
Now the invalidation is based on ancestor-to-descendant direction starting with finding the outer most containing block for a particular float (see outermostBlockContainingFloatingObject)
The invalidation logic also expects no gaps in the ancestor chain e.g.
RenderBlockFlow (A) -> float X
RenderBlockFlow (B) -> float X
RenderBlockFlow (C) -> float X
if float X is assigned to both A and C, then it must be assigned to B as well.
RenderBlockFlow::removeFloatingObjects() simply removes the float from the renderer. It does not invalidate the ancestor/descendant chain.
e.g. calling B.removeFloatingObjects() would just remove float X from RenderBlockFlow (B)
RenderBlockFlow (A) -> float X
RenderBlockFlow (B)
RenderBlockFlow (C) -> float X
and any subsequent invalidation attempt would fail to clear up A or C (depending on whether it is initiated on A or C).
Test: fast/multicol/floating-boxes-moved-under-multi-column.html
- rendering/updating/RenderTreeBuilder.cpp:
(WebCore::RenderTreeBuilder::moveChildren):
LayoutTests:
- fast/multicol/floating-boxes-moved-under-multi-column-expected.txt: Added.
- fast/multicol/floating-boxes-moved-under-multi-column.html: Added.
- 6:10 AM Changeset in webkit [276548] by
-
- 2 edits in trunk/Source/WebCore
[LFC] Ignore content height and width when 'contain: size' is present
https://bugs.webkit.org/show_bug.cgi?id=225013
Reviewed by Antti Koivisto.
Enable size containment for "shrink to fit width" and for "height: auto" formatting context roots.
https://www.w3.org/TR/css-contain-2/#size-containment.
- layout/FormattingContextGeometry.cpp:
(WebCore::Layout::FormattingContext::Geometry::contentHeightForFormattingContextRoot const):
(WebCore::Layout::FormattingContext::Geometry::shrinkToFitWidth):
- 12:41 AM Changeset in webkit [276547] by
-
- 4 edits2 adds in trunk
Move selectedOptions cache invalidation timing
https://bugs.webkit.org/show_bug.cgi?id=224894
Patch by Rob Buis <rbuis@igalia.com> on 2021-04-24
Reviewed by Ryosuke Niwa.
Source/WebCore:
Move selectedOptions cache invalidation timing out
of setSelectedState, having it in setSelectedState
can cause selectedOptions cache invalidation while
determining the cache.
Test: fast/dom/HTMLSelectElement/selected-options-item-crash.html
- html/HTMLOptionElement.cpp:
(WebCore::HTMLOptionElement::setSelectedState):
- html/HTMLSelectElement.cpp:
(WebCore::HTMLSelectElement::updateListBoxSelection):
(WebCore::HTMLSelectElement::selectOption):
(WebCore::HTMLSelectElement::deselectItemsWithoutValidation):
(WebCore::HTMLSelectElement::restoreFormControlState):
(WebCore::HTMLSelectElement::reset):
(WebCore::HTMLSelectElement::updateSelectedState):
LayoutTests:
Add test for this.
- fast/dom/HTMLSelectElement/selected-options-item-crash-expected.txt: Added.
- fast/dom/HTMLSelectElement/selected-options-item-crash.html: Added.
