Timeline

Jul 26, 2021:

9:42 PM Changeset in webkit [280337] by Said Abou-Hallawa

• 20 edits
  1 copy
  1 add in trunk/Source

Regulate the WebPage RenderingUpdates from the WebProcess to the GPUProcess
​https://bugs.webkit.org/show_bug.cgi?id=227791
<rdar://78430639>

Reviewed by Simon Fraser.

Source/WebCore:

Make ChromeClient decide whether a RenderingUpdate can be triggered or
should be rescheduled. Keep track of how many times RenderingUpdate was
rescheduled to ensure RenderingUpdate will be triggered anyway after a
certain number of reschedules.

• page/ChromeClient.h:

(WebCore::ChromeClient::canTriggerRenderingUpdate const):

• page/RenderingUpdateScheduler.cpp:

(WebCore::RenderingUpdateScheduler::displayRefreshFired):

• page/RenderingUpdateScheduler.h:

Source/WebKit:

This is the workflow to regulate the WebPage RenderingUpdates:

-- WebPage::finalizeRenderingUpdate() calls RemoteRenderingBackendProxy::

finalizeRenderingUpdate().

-- RemoteRenderingBackendProxy::finalizeRenderingUpdate() sends its

renderingUpdateID to GPUPProcess and asks it to process all the
pending DisplayList items.

-- RemoteRenderingBackend::finalizeRenderingUpdate) processes all the

pending items and sends the received renderingUpdateID back to the
WebProcess.

-- RemoteRenderingBackendProxy::didFinalizeRenderingUpdate() stores the

received didRenderingUpdateID.

-- RemoteRenderingBackendProxy::delayedRenderingUpdateCount() returns

'renderingUpdateID - didRenderingUpdateID' which is the number of
un-processed finalizeRenderingUpdate by GPUP.

-- WebPage::canTriggerRenderingUpdate() uses delayedRenderingUpdateCount()

to decides whether the current RenderingUpdate can be triggered or
should be rescheduled.

-- RenderingUpdateScheduler::displayRefreshFired() calls

WebPage::canTriggerRenderingUpdate() through the page chrome client.

• GPUProcess/graphics/RemoteRenderingBackend.cpp:

(WebKit::RemoteRenderingBackend::submit):
(WebKit::RemoteRenderingBackend::finalizeRenderingUpdate):

• GPUProcess/graphics/RemoteRenderingBackend.h:
• GPUProcess/graphics/RemoteRenderingBackend.messages.in:
• Scripts/webkit/messages.py:

(types_that_cannot_be_forward_declared):

• Shared/MonotonicObjectIdentifier.h: Copied from Source/WebKit/Shared/TransactionID.h.

(WebKit::MonotonicObjectIdentifier::MonotonicObjectIdentifier):
(WebKit::MonotonicObjectIdentifier::isHashTableDeletedValue const):
(WebKit::MonotonicObjectIdentifier::encode const):
(WebKit::MonotonicObjectIdentifier::decode):
(WebKit::MonotonicObjectIdentifier::operator== const):
(WebKit::MonotonicObjectIdentifier::operator> const):
(WebKit::MonotonicObjectIdentifier::operator>= const):
(WebKit::MonotonicObjectIdentifier::operator< const):
(WebKit::MonotonicObjectIdentifier::operator<= const):
(WebKit::MonotonicObjectIdentifier::operator!= const):
(WebKit::MonotonicObjectIdentifier::increment):
(WebKit::MonotonicObjectIdentifier::next const):
(WebKit::MonotonicObjectIdentifier::toUInt64 const):
(WebKit::MonotonicObjectIdentifier::operator bool const):
(WebKit::MonotonicObjectIdentifier::loggingString const):
(WebKit::MonotonicObjectIdentifier::hashTableDeletedValue):
(WebKit::MonotonicObjectIdentifier::isValidIdentifier):
(WebKit::operator<<):

• Shared/RenderingUpdateID.h: Added.
• Shared/TransactionID.h:

(WebKit::MonotonicObjectIdentifier::MonotonicObjectIdentifier): Deleted.
(WebKit::MonotonicObjectIdentifier::isHashTableDeletedValue const): Deleted.
(WebKit::MonotonicObjectIdentifier::encode const): Deleted.
(WebKit::MonotonicObjectIdentifier::decode): Deleted.
(WebKit::MonotonicObjectIdentifier::operator== const): Deleted.
(WebKit::MonotonicObjectIdentifier::operator> const): Deleted.
(WebKit::MonotonicObjectIdentifier::operator>= const): Deleted.
(WebKit::MonotonicObjectIdentifier::operator< const): Deleted.
(WebKit::MonotonicObjectIdentifier::operator<= const): Deleted.
(WebKit::MonotonicObjectIdentifier::operator!= const): Deleted.
(WebKit::MonotonicObjectIdentifier::increment): Deleted.
(WebKit::MonotonicObjectIdentifier::next const): Deleted.
(WebKit::MonotonicObjectIdentifier::toUInt64 const): Deleted.
(WebKit::MonotonicObjectIdentifier::operator bool const): Deleted.
(WebKit::MonotonicObjectIdentifier::loggingString const): Deleted.
(WebKit::MonotonicObjectIdentifier::hashTableDeletedValue): Deleted.
(WebKit::MonotonicObjectIdentifier::isValidIdentifier): Deleted.
(WebKit::operator<<): Deleted.

• WebKit.xcodeproj/project.pbxproj:
• WebProcess/GPU/graphics/RemoteRenderingBackendProxy.cpp:

(WebKit::RemoteRenderingBackendProxy::gpuProcessConnectionDidClose):
(WebKit::RemoteRenderingBackendProxy::finalizeRenderingUpdate):
(WebKit::RemoteRenderingBackendProxy::didFinalizeRenderingUpdate):

• WebProcess/GPU/graphics/RemoteRenderingBackendProxy.h:

(WebKit::RemoteRenderingBackendProxy::renderingUpdateID const):
(WebKit::RemoteRenderingBackendProxy::delayedRenderingUpdateCount const):

• WebProcess/GPU/graphics/RemoteRenderingBackendProxy.messages.in:
• WebProcess/GPU/graphics/RemoteResourceCacheProxy.cpp:

(WebKit::RemoteResourceCacheProxy::cacheFont):
(WebKit::RemoteResourceCacheProxy::prepareForNextRenderingUpdate):
(WebKit::RemoteResourceCacheProxy::clearFontMap):
(WebKit::RemoteResourceCacheProxy::finalizeRenderingUpdateForFonts):
(WebKit::RemoteResourceCacheProxy::finalizeRenderingUpdate):
(WebKit::RemoteResourceCacheProxy::didFinalizeRenderingUpdate): Deleted.

• WebProcess/GPU/graphics/RemoteResourceCacheProxy.h:
• WebProcess/WebCoreSupport/WebChromeClient.cpp:

(WebKit::WebChromeClient::canTriggerRenderingUpdate const):

• WebProcess/WebCoreSupport/WebChromeClient.h:
• WebProcess/WebPage/WebPage.cpp:

(WebKit::WebPage::canTriggerRenderingUpdate const):
(WebKit::WebPage::finalizeRenderingUpdate):

• WebProcess/WebPage/WebPage.h:

7:21 PM Changeset in webkit [280336] by Wenson Hsieh

• 10 edits
  1 add in trunk

[macOS Monterey] Pressing Fn+E should present the emoji picker when editing
​https://bugs.webkit.org/show_bug.cgi?id=228300
rdar://77558672

Reviewed by Tim Horton.

Source/WebCore/PAL:

Add a new SPI declaration on NSMenu. See WebKit and WebKitLegacy changes for more information.

• pal/spi/mac/NSMenuSPI.h:

Source/WebKit:

In macOS Monterey, the system-wide Fn+E key command presents the Emoji & Symbols UI as a popover. However, this
new key command currently does not work in editable contexts in both WebKit1 and WebKit2 views on macOS, where
we end up just inserting the character "e".

This is because, on both ports, we attempt to interpret NSEvents by calling back into AppKit to handle the
event, recording the selectors that *would've* been performed, and then using this information to build up a
list of WebCore::KeyCommand for the event (which are later converted into editing commands). In WebKit2, we
call into -[NSTextInputContext handleEventByKeyboardLayout:] to handle the event, and capture the resulting
calls back into WKWebView by storing entries in m_collectedKeypressCommands (something similar happens in
WebKit1). In the case of these new Fn commands, they are handled by the system before entering
NSTextInputContext, so NSTextInputContext ends up handling them as regular text insertion.

As an aside, the reason the meta (⌘) key doesn't encounter this problem is because meta, option, control and
shift are considered valid key bindings in NSKeyBindingManager, whereas the function key mask is specifically
excluded from this list for compability reasons. NSTextInputContext consults NSKeyBindingManager in order to
check whether the given NSEvent is a known key binding in -handleEventByKeyboardLayout:, so it fails to
acknowledge Fn+E as a key binding.

To work around this, we adopt new AppKit SPI added in rdar://81123724 to determine whether an NSEvent maps to
any main menu items and avoid handling "Fn-*" key events as text insertion if they already map to items in the
main menu.

Test: KeyboardEventTests.FunctionKeyCommand

• UIProcess/Cocoa/WebViewImpl.mm:

(WebKit::WebViewImpl::collectKeyboardLayoutCommandsForEvent):

Source/WebKitLegacy/mac:

Apply a similar fix, this time for WebKit1 on macOS.

• WebView/WebHTMLView.mm:

(-[WebHTMLView insertText:]):

Only append "insertText:" for a corresponding NSEvent with the Function key held down if it doesn't also map
to a item in the main menu.

Tools:

Add a new API test that simulates a keypress for "Fn+E" with a main menu that contains a system menu item for
that key binding; the test then verifies that text was not inserted into the page as a result.

• TestWebKitAPI/TestWebKitAPI.xcodeproj/project.pbxproj:
• TestWebKitAPI/Tests/mac/KeyboardEventTests.mm: Added.

(-[KeyboardTestMenu _containsItemMatchingEvent:includingDisabledItems:]):
(TestWebKitAPI::TEST):

• TestWebKitAPI/cocoa/TestWKWebView.h:
• TestWebKitAPI/cocoa/TestWKWebView.mm:

(-[TestWKWebView typeCharacter:]):
(-[TestWKWebView typeCharacter:modifiers:]):

6:43 PM Changeset in webkit [280335] by ysuzuki@apple.com

• 17 edits in trunk

Unreviewed, speculative revert of r280193
​https://bugs.webkit.org/show_bug.cgi?id=228142

Source/WebCore:

• html/parser/HTMLConstructionSite.cpp:

(WebCore::HTMLConstructionSite::insertTextNode):

• html/parser/HTMLConstructionSite.h:
• html/parser/HTMLDocumentParser.cpp:

(WebCore::HTMLDocumentParser::pumpTokenizerLoop):

• html/parser/HTMLMetaCharsetParser.cpp:

(WebCore::HTMLMetaCharsetParser::checkForMetaCharset):

• html/parser/HTMLPreloadScanner.cpp:

(WebCore::HTMLPreloadScanner::scan):

• html/parser/HTMLToken.h:

(WebCore::HTMLToken::clear):
(WebCore::HTMLToken::beginStartTag):
(WebCore::HTMLToken::beginEndTag):
(WebCore::HTMLToken::shrinkToBestFit): Deleted.

• html/parser/HTMLTokenizer.h:

(WebCore::HTMLTokenizer::shrinkToBestFit): Deleted.

• html/parser/HTMLTreeBuilder.cpp:

(WebCore::HTMLTreeBuilder::ExternalCharacterTokenBuffer::characterPredicate):
(WebCore::HTMLTreeBuilder::processFakeCharacters):
(WebCore::HTMLTreeBuilder::insertPhoneNumberLink):
(WebCore::HTMLTreeBuilder::linkifyPhoneNumbers):
(WebCore::HTMLTreeBuilder::processCharacterBuffer):
(WebCore::HTMLTreeBuilder::processCharacterBufferForInBody):
(WebCore::HTMLTreeBuilder::defaultForInTableText):
(WebCore::HTMLTreeBuilder::processTokenInForeignContent):

• html/parser/HTMLTreeBuilder.h:

Source/WTF:

• wtf/Vector.h:

(WTF::Malloc>::shrinkCapacity):
(WTF::Malloc>::shrinkToBestFit): Deleted.

• wtf/text/AtomStringImpl.cpp:

(WTF::UCharBufferTranslator::equal):
(WTF::LCharBufferTranslator::equal):
(WTF::BufferFromStaticDataTranslator::equal):

• wtf/text/StringView.h:

(WTF::StringView::stripLeadingMatchedCharacters): Deleted.

Tools:

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

6:40 PM Changeset in webkit [280334] by mmaxfield@apple.com

• 12 edits in trunk/Source

[GPU Process] Add resource use counter infrastructure to RemoteResourceCache
​https://bugs.webkit.org/show_bug.cgi?id=228222

Reviewed by Said Abou-Hallawa.

Source/WebCore:

This is part 2 of ​https://bugs.webkit.org/show_bug.cgi?id=228216.

No new tests because there is no behavior change yet. This patch is just adding infrastructure. For now, all these
use counts sent by the web process are 0 - I'll implement that part in 3 follow-up patches.

• platform/graphics/displaylists/DisplayListReplayer.h:

(WebCore::DisplayList::Replayer::Delegate::recordResourceUse):

Source/WebKit:

This patch adds the infrastructure for associating a counter with each resource in the RemoteResourceCache.
As the web process uses a resource during display list recording, it increments a counter, and as the GPU
process uses a resource, it increments a counter in the RemoteResourceCache. When the web process is done
with a resource, it sends a release message to the GPU process which crucially includes the web process's
counter. The GPU process has to then wait to actually delete the resource until its own use counter matches
the one it received from the web process. This patch implements this deferred deletion infrastructure
in the GPU process.

• GPUProcess/graphics/RemoteRenderingBackend.cpp:

(WebKit::RemoteRenderingBackend::ReplayerDelegate::recordResourceUse):
(WebKit::RemoteRenderingBackend::releaseRemoteResource): Pass the use count to the RemoteResourceCache.

• GPUProcess/graphics/RemoteRenderingBackend.h: Add the use count field.
• GPUProcess/graphics/RemoteRenderingBackend.messages.in: Ditto.
• GPUProcess/graphics/RemoteResourceCache.cpp:

(WebKit::RemoteResourceCache::cacheImageBuffer): It's actually okay if there's already an resource cached.
That just means that the web process started using a resource again before the GPU process consumed all
commands that used it the first time. All we need to do is increment the open count.
(WebKit::RemoteResourceCache::cacheNativeImage): Ditto.
(WebKit::RemoteResourceCache::cacheFont): Ditto.
(WebKit::RemoteResourceCache::ensureResourceUseCounter):
(WebKit::RemoteResourceCache::maybeRemoveResource): This is called whenever it's possible for us to be in
situation where we should be removing a resource (aka whenever the use count increments). It checks to see
if we can remove the resource, and if it can, does so.
(WebKit::RemoteResourceCache::recordResourceUse): The callback that runs for each resource use during
replaying. It can't actually delete any resources, so instead it just records which resources had their
use counts incremented, and waits for prune() to actually call maybeRemoveResource().
(WebKit::RemoteResourceCache::prune): After the replay is complete, call maybeRemoveResource() on all the
resources which had their use counters touched.
(WebKit::RemoteResourceCache::releaseRemoteResource): Put the relevant resource into the state where we
are waiting for the use counters to match before we can delete the resource. Also, call maybeRemoveResource()
so it can be deleted immediately if the counters already match.

• GPUProcess/graphics/RemoteResourceCache.h: Add a new data structure to hold the use counters. There's

a big comment in here that describes what these fields mean.

• WebProcess/GPU/graphics/RemoteImageBufferProxy.h:

(WebKit::RemoteImageBufferProxy::~RemoteImageBufferProxy): Send a dummy use count for now. This will be
implemented in a follow-up patch.

• WebProcess/GPU/graphics/RemoteRenderingBackendProxy.cpp:

(WebKit::RemoteRenderingBackendProxy::releaseRemoteResource): Add the useCount parameter.

• WebProcess/GPU/graphics/RemoteRenderingBackendProxy.h: Ditto.
• WebProcess/GPU/graphics/RemoteResourceCacheProxy.cpp:

(WebKit::RemoteResourceCacheProxy::releaseNativeImage): Send a dummy use count for now. This will be
implemented in a follow-up patch.
(WebKit::RemoteResourceCacheProxy::clearFontMap): Ditto.
(WebKit::RemoteResourceCacheProxy::finalizeRenderingUpdateForFonts): Ditto.

6:07 PM Changeset in webkit [280333] by Chris Dumez

• 13 edits
  2 adds in trunk

Location.href setter uses wrong window to resolve relative URLs
​https://bugs.webkit.org/show_bug.cgi?id=228287

Reviewed by Darin Adler.

LayoutTests/imported/w3c:

Rebaseline a few WPT tests now that more checks are passing.

• web-platform-tests/html/semantics/scripting-1/the-script-element/execution-timing/083-expected.txt:
• web-platform-tests/html/semantics/scripting-1/the-script-element/execution-timing/084-expected.txt:
• web-platform-tests/html/webappapis/scripting/processing-model-2/window-onerror-with-cross-frame-event-listeners-1-expected.txt:
• web-platform-tests/html/webappapis/scripting/processing-model-2/window-onerror-with-cross-frame-event-listeners-2-expected.txt:

Source/WebCore:

Location::setLocation() is using "firstWindow" to resolve any input relative URL.
We rely on JSC::VM::deprecatedVMEntryGlobalObject() to get the first/entry window.
However, the test I wrote shows that we were using the wrong window as "first window"
in some cases (since this test would pass in other browsers but not in WebKit).

It turns out that JSEventListener::handleEvent() had a VMEntryScope internally that
would impact the global object being returned by JSC::VM::deprecatedVMEntryGlobalObject().
We were incorrectly using the scriptExecutionContext passed as parameter when constructing
the VMEntryScope instead of using the global object of the function we're about to call.
This bug addresses this.

Credits to Geoff Garen for identifying the issue in JSEventListener::handleEvent().

Test: http/tests/dom/window-location-set-href-relative-url.html

• bindings/js/JSEventListener.cpp:

(WebCore::JSEventListener::handleEvent):

• page/Location.cpp:

(WebCore::Location::setLocation):

LayoutTests:

Add layout test coverage. This test is passing in Chome and Firefox but was failing with
WebKit.

• http/tests/dom/window-location-set-href-relative-url-expected.txt: Added.
• http/tests/dom/window-location-set-href-relative-url.html: Added.

5:37 PM Changeset in webkit [280332] by commit-queue@webkit.org

• 4 edits in trunk/Source/WebCore

This is responding to feedback on r280310.
​https://bugs.webkit.org/show_bug.cgi?id=227718

Patch by Johnson Zhou <​qiaosong_zhou@apple.com> on 2021-07-26
Reviewed by Alex Christensen.

• dom/FormDataEvent.h:

(WebCore::FormDataEvent::formData const):

• html/DOMFormData.cpp:

(WebCore::DOMFormData::create):
(WebCore::DOMFormData::clone const):
(WebCore::DOMFormData::clone): Deleted.

• html/DOMFormData.h:

5:34 PM Changeset in webkit [280331] by Eric Hutchison

• 2 edits in trunk/LayoutTests

imported/w3c/web-platform-tests/css/css-scoping/css-scoping-shadow-dynamic-remove-style-detached.html is a flaky failure.
​https://bugs.webkit.org/show_bug.cgi?id=228311.

Unreviewed test gardening.

• platform/mac/TestExpectations:

5:25 PM Changeset in webkit [280330] by Jean-Yves Avenard

• 4 edits
  2 adds in trunk

Video pauses after scrubbing with Touch Bar
​https://bugs.webkit.org/show_bug.cgi?id=228277
rdar://80606886

Reviewed by Jer Noble.

Source/WebCore:

In ​https://trac.webkit.org/r206487 ; in order to ensure that the playback state
was properly reflected following a seek using the touch bar, the element was paused.
It's unclear if that workaround is still required, but for now we will record if the
element was playing before the seek and if so, resume playback once the seek completes.
Now that the touch bar and Now Playing are hooked to the Media Session action handlers
the behaviour change will occur for all those components.

Test: media/media-session/play-after-seek.html

• html/HTMLMediaElement.cpp:

(WebCore::HTMLMediaElement::HTMLMediaElement): Initialize new member in constructor.
(WebCore::HTMLMediaElement::clearSeeking):
(WebCore::HTMLMediaElement::finishSeek): Call play() once seek completes if the element
was playing before.
(WebCore::HTMLMediaElement::pause): Ensure that if pause() is called before the seek
completes, the element stays paused.
(WebCore::HTMLMediaElement::handleSeekToPlaybackPosition): Record playing state before
pausing the element.

• html/HTMLMediaElement.h: Add new boolean member.

LayoutTests:

• media/media-session/play-after-seek-expected.txt: Added.
• media/media-session/play-after-seek.html: Added.

5:15 PM Changeset in webkit [280329] by Eric Hutchison

• 2 edits in trunk/LayoutTests

Update test expectations for inspector/canvas/recording-bitmaprenderer-memoryLimit.html.
​https://bugs.webkit.org/show_bug.cgi?id=227881.

Unreviewed test gardening.

• platform/mac-wk2/TestExpectations:

5:06 PM Changeset in webkit [280328] by jer.noble@apple.com

• 2 edits in trunk/Source/WebKit

REGRESSION: [iOS] ASSERTION FAILED: !m_messageReceiverMapCount under WebKit::RemoteAudioHardwareListener::~RemoteAudioHardwareListener()
​https://bugs.webkit.org/show_bug.cgi?id=228038
<rdar://problem/80705471>

Reviewed by Chris Dumez.

Remove RemoteAudioHardwareListener from the GPUProcessConnection messageMap when the connection closes.

• WebProcess/GPU/media/RemoteAudioHardwareListener.cpp:

(WebKit::RemoteAudioHardwareListener::gpuProcessConnectionDidClose):

5:06 PM Changeset in webkit [280327] by jer.noble@apple.com

• 6 edits in trunk/Source

[Cocoa] WebKit is making GroupActivities API calls for all WebKit clients
​https://bugs.webkit.org/show_bug.cgi?id=228299
<rdar://80802982>

Reviewed by Eric Carlson.

Source/WebKit:

Tie registering for GroupActivities notifications to the MediaSessionCoordinatorEnabled
WebPreference, and make the default value for that preference depend on having the
"com.apple.developer.group-session.urlactivity" entitlement. All WebKit clients with
that entitlement will have the preference enabled by default; all other clients will
have that preference disabled by default.

• Shared/WebPreferencesDefaultValues.cpp:

(WebKit::defaultMediaSessionCoordinatorEnabled):

• Shared/WebPreferencesDefaultValues.h:
• UIProcess/WebPageProxy.cpp:

(WebKit::m_limitsNavigationsToAppBoundDomains):
(WebKit::WebPageProxy::~WebPageProxy):
(WebKit::WebPageProxy::didCommitLoadForFrame):

Source/WTF:

• Scripts/Preferences/WebPreferencesExperimental.yaml:

5:02 PM Changeset in webkit [280326] by jer.noble@apple.com

• 2 edits in trunk/Source/WebKit

REGRESSION: [iOS] ASSERTION FAILED: !m_useCount in WebKit::SandboxExtension::~SandboxExtension()
​https://bugs.webkit.org/show_bug.cgi?id=228171
<rdar://problem/80926509>

Reviewed by Eric Carlson.

SandboxExtension expects to be revoked before it is destroyed, something that is done by
RemoteMediaPlayerProxy::invalidate(). However, invalidate() won't be called if the
RemoteMediaPlayerManagerProxy is destroyed before all its proxies are invalidated.

Ensure every RemoteMediaPlayerProxy contained by the RemoteMediaPlayerManagerProxy is
invalidated in RemoteMediaPlayerManagerProxy's destructor.

• GPUProcess/media/RemoteMediaPlayerManagerProxy.cpp:

(WebKit::RemoteMediaPlayerManagerProxy::~RemoteMediaPlayerManagerProxy):

• GPUProcess/media/RemoteMediaPlayerManagerProxy.h:

4:58 PM Changeset in webkit [280325] by commit-queue@webkit.org

• 4 edits in trunk

WKUserContentController.removeAllScriptMessageHandlers() doesn't release the message handlers
​https://bugs.webkit.org/show_bug.cgi?id=228271

Patch by Alex Christensen <​achristensen@webkit.org> on 2021-07-26
Reviewed by Sihui Liu.

Source/WebKit:

Covered by an API test.

• UIProcess/UserContent/WebUserContentControllerProxy.cpp:

(WebKit::WebUserContentControllerProxy::removeAllUserMessageHandlers):

Tools:

• TestWebKitAPI/Tests/WebKitCocoa/UserContentController.mm:

(TEST):

4:55 PM Changeset in webkit [280324] by Ayumi Kojima

• 2 edits in trunk/LayoutTests

[ iOS Debug ] http/tests/xmlhttprequest/access-control-preflight-credential-sync.html is a flaky crash.
​https://bugs.webkit.org/show_bug.cgi?id=228305

Unreviewed test gardening.

• platform/ios-wk2/TestExpectations:

4:54 PM Changeset in webkit [280323] by rniwa@webkit.org

• 4 edits in trunk/Source/WebCore

Deploy smart pointers in ApplyBlockElementCommand, IndentOutdentCommand and InsertListCommand
​https://bugs.webkit.org/show_bug.cgi?id=228304

Reviewed by Chris Dumez.

Use RefPtr in more places instead of raw pointers.

• editing/ApplyBlockElementCommand.cpp:

(WebCore::ApplyBlockElementCommand::formatSelection):
(WebCore::isNewLineAtPosition):
(WebCore::ApplyBlockElementCommand::rangeForParagraphSplittingTextNodesIfNeeded):
(WebCore::ApplyBlockElementCommand::endOfNextParagraphSplittingTextNodesIfNeeded):

• editing/IndentOutdentCommand.cpp:

(WebCore::IndentOutdentCommand::tryIndentingAsListItem):
(WebCore::IndentOutdentCommand::indentIntoBlockquote):
(WebCore::IndentOutdentCommand::outdentParagraph):

• editing/InsertListCommand.cpp:

(WebCore::InsertListCommand::mergeWithNeighboringLists):
(WebCore::InsertListCommand::unlistifyParagraph):
(WebCore::adjacentEnclosingList):
(WebCore::InsertListCommand::listifyParagraph):

4:51 PM Changeset in webkit [280322] by Alan Coon

• 1 copy in tags/Safari-611.3.10.1.6

Tag Safari-611.3.10.1.6.

4:37 PM Changeset in webkit [280321] by Eric Hutchison

• 2 edits in trunk/LayoutTests

[ Catalina+ Debug wk2 ] inspector/model/remote-object/iterator-large.html is a flaky timeout.
​https://bugs.webkit.org/show_bug.cgi?id=228210.

Unreviewed test gardening.

• platform/mac-wk2/TestExpectations:

4:27 PM Changeset in webkit [280320] by Eric Hutchison

• 2 edits in trunk/LayoutTests

Updated test expectations for fullscreen/full-screen-remove-children.html.
​https://bugs.webkit.org/show_bug.cgi?id=227874.

Unreviewed test gardening.

• platform/mac-wk2/TestExpectations:

3:09 PM Changeset in webkit [280319] by Kate Cheney

• 5 edits in trunk

ServiceWorkerRegistration.unregister method fails in WKWebView
​https://bugs.webkit.org/show_bug.cgi?id=227524
<rdar://problem/80264108>

Reviewed by Youenn Fablet.

Source/WebCore:

We should only check the app-bound domain list when registering a new
worker. Unregistering or updating an existing registration is fine because
we know that every registration has already gone through this check.

• workers/service/server/SWServer.cpp:

(WebCore::SWServer::addRegistrationFromStore):
(WebCore::SWServer::validateRegistrationDomain):
(WebCore::SWServer::scheduleJob):

• workers/service/server/SWServer.h:

Tools:

API test coverage. Drive-by fix to remove unnecessary WKWebView
configuration flags from tests that don't need them.

Note this will also fix ​https://bugs.webkit.org/show_bug.cgi?id=227531
where updating an existing worker fails after registering 3 domains,
but writing a test for that would require registering service workers
on 3 domains that are not localhost or a loopback IP address, which
we currently do not have infrastructure for.

Testing unregister was only possible because the script URL passed in
during the unregister job is null, so it does not get caught in the
localhost/loopback IP address check, unlike the update case.

• TestWebKitAPI/Tests/WebKitCocoa/InAppBrowserPrivacy.mm:

(TEST):
(-[SWInAppBrowserPrivacyMessageHandler userContentController:didReceiveScriptMessage:]):

3:07 PM Changeset in webkit [280318] by pvollan@apple.com

• 6 edits in trunk

The layout test fast/images/heic-as-background-image.html is a constant failure
​https://bugs.webkit.org/show_bug.cgi?id=228195
<rdar://80334724>

Reviewed by Said Abou-Hallawa.

Source/WebKit:

Call correct system function for required initialization.

No new tests, covered by existing test.

• WebProcess/cocoa/WebProcessCocoa.mm:

(WebKit::WebProcess::platformInitializeWebProcess):

Source/WTF:

Add HAVE macro for the system function CMPhotoIsTileDecoderAvailable.

• wtf/PlatformHave.h:

LayoutTests:

Update test expectations.

• platform/mac/TestExpectations:

2:59 PM Changeset in webkit [280317] by Kate Cheney

• 2 edits in trunk/Source/WebKit

Attribution context causes some performance regressions
​https://bugs.webkit.org/show_bug.cgi?id=228294

Unreviewed followup to address Darin's comment regarding unnecessary
semi-colons.

• NetworkProcess/cocoa/NetworkDataTaskCocoa.mm:

(processPCMRequest):
(overrideAttributionContext):

2:39 PM Changeset in webkit [280316] by Cameron McCormack

• 4 edits in trunk/Source/WebCore

Don't allow descriptors to be set to CSS-wide keywords
​https://bugs.webkit.org/show_bug.cgi?id=228218
<rdar://80675715>

Reviewed by Simon Fraser.

And store the parent rule type directly on
StyleRuleCSSStyleDeclaration so that we can still check it after
m_parentRule has been cleared.

• css/PropertySetCSSStyleDeclaration.cpp:

(WebCore::StyleRuleCSSStyleDeclaration::StyleRuleCSSStyleDeclaration):
(WebCore::StyleRuleCSSStyleDeclaration::cssParserContext const):

• css/PropertySetCSSStyleDeclaration.h:
• css/parser/CSSParserFastPaths.cpp:

(WebCore::parseKeywordValue):

2:26 PM Changeset in webkit [280315] by Russell Epstein

• 1 copy in tags/Safari-612.1.25

Tag Safari-612.1.25.

2:26 PM Changeset in webkit [280314] by Russell Epstein

• 1 delete in tags/Safari-612.1.25

Delete tag.

2:19 PM Changeset in webkit [280313] by Chris Dumez

• 3 edits in trunk/Source/WebCore

Rename activeWindow to incumbentWindow in the Location class
​https://bugs.webkit.org/show_bug.cgi?id=228295

Reviewed by Darin Adler.

Rename activeWindow to incumbentWindow in the Location class, since this is the naming used
both in the HTML specification and Location.idl.

• page/Location.cpp:

(WebCore::Location::setHref):
(WebCore::Location::setProtocol):
(WebCore::Location::setHost):
(WebCore::Location::setHostname):
(WebCore::Location::setPort):
(WebCore::Location::setPathname):
(WebCore::Location::setSearch):
(WebCore::Location::setHash):
(WebCore::Location::assign):
(WebCore::Location::replace):
(WebCore::Location::reload):
(WebCore::Location::setLocation):

• page/Location.h:

1:47 PM Changeset in webkit [280312] by commit-queue@webkit.org

• 3 edits
  2 adds in trunk

Crash in InsertParagraphSeparatorCommand::doApply
​https://bugs.webkit.org/show_bug.cgi?id=224977

Patch by Frédéric Wang <​fwang@igalia.com> on 2021-07-26
Reviewed by Ryosuke Niwa.

Source/WebCore:

Because <html> elements are handled specially in Position::isCandidate() (a) and
PositionIterator::isCandidate() (b), the function InsertParagraphSeparatorCommand::doApply()
may end up in a edge case where the startBlock is a sibling of the visible position per (a)
but isFirstInBlock,isLastInBlock is true,false per (b). This leads to hitting the debug
assertion ASSERT(startBlock->firstChild()) and dereferencing a nullptr pointer in release.
This patch fixes that by exiting early if the visible position is not a descendant of the
start block.

Test: editing/inserting/insert-paragraph-separator-with-html-elements-crash.html

• editing/InsertParagraphSeparatorCommand.cpp:

(WebCore::InsertParagraphSeparatorCommand::doApply):

LayoutTests:

Add regression test.

• editing/inserting/insert-paragraph-separator-with-html-elements-crash-expected.txt: Added.
• editing/inserting/insert-paragraph-separator-with-html-elements-crash.html: Added.

1:40 PM Changeset in webkit [280311] by Kate Cheney

• 2 edits in trunk/Source/WebKit

Attribution context causes some performance regressions
​https://bugs.webkit.org/show_bug.cgi?id=228294
<rdar://problem/76663528>

Reviewed by Per Arne Vollan.

Attribution context is used by the networking stack for a variety of
purposes, but is not always needed and can negatively affect
performance. For these reasons, we should override it in some cases.

• NetworkProcess/cocoa/NetworkDataTaskCocoa.mm:

(overrideAttributionContext):
(WebKit::NetworkDataTaskCocoa::NetworkDataTaskCocoa):

1:15 PM Changeset in webkit [280310] by commit-queue@webkit.org

• 37 edits
  3 adds in trunk

LayoutTests/imported/w3c:
Added FormDataEvent support.
​https://bugs.webkit.org/show_bug.cgi?id=227718

Patch by Johnson Zhou <​qiaosong_zhou@apple.com> on 2021-07-26
Reviewed by Chris Dumez.

• web-platform-tests/html/semantics/forms/form-submission-0/FormDataEvent.window-expected.txt:
• web-platform-tests/html/semantics/forms/form-submission-0/constructing-form-data-set-expected.txt:
• web-platform-tests/html/semantics/forms/form-submission-0/text-plain.window-expected.txt:
• web-platform-tests/html/semantics/forms/form-submission-0/urlencoded2.window-expected.txt:
• web-platform-tests/html/webappapis/scripting/events/event-handler-all-global-events-expected.txt:
• web-platform-tests/html/webappapis/scripting/events/event-handler-attributes-body-window-expected.txt:
• web-platform-tests/html/webappapis/scripting/events/event-handler-attributes-windowless-body-expected.txt:
• web-platform-tests/xhr/formdata-expected.txt:

Source/WebCore:
FormDataEvent added, and dispatched upon creation of DOMFormData or submission of HTMLFormElement.
​https://bugs.webkit.org/show_bug.cgi?id=227718

Patch by Johnson Zhou <​qiaosong_zhou@apple.com> on 2021-07-26
Reviewed by Chris Dumez.

• CMakeLists.txt:
• DerivedSources-input.xcfilelist:
• DerivedSources-output.xcfilelist:
• DerivedSources.make:
• Sources.txt:
• WebCore.xcodeproj/project.pbxproj:
• dom/EventNames.h:
• dom/EventNames.in:
• dom/FormDataEvent.cpp: Added.

(WebCore::FormDataEvent::create):
(WebCore::FormDataEvent::FormDataEvent):
(WebCore::FormDataEvent::eventInterface const):

• dom/FormDataEvent.h: Added.

(WebCore::FormDataEvent::formData const):

• dom/FormDataEvent.idl: Added.
• dom/GlobalEventHandlers.idl:
• html/DOMFormData.cpp:

(WebCore::DOMFormData::DOMFormData):
(WebCore::DOMFormData::create):
(WebCore::DOMFormData::clone):

• html/DOMFormData.h:

(WebCore::DOMFormData::create): Deleted.

• html/HTMLAttributeNames.in:
• html/HTMLElement.cpp:

(WebCore::HTMLElement::createEventHandlerNameMap):

• html/HTMLFormElement.cpp:

(WebCore::HTMLFormElement::submitIfPossible):
(WebCore::HTMLFormElement::submit):
(WebCore::HTMLFormElement::constructEntryList):

• html/HTMLFormElement.h:
• loader/FormSubmission.cpp:

(WebCore::FormSubmission::create):

• loader/FormSubmission.h:
• platform/network/FormData.cpp:

(WebCore::FormData::appendNonMultiPartKeyValuePairItems):

Source/WebInspectorUI:
FormDataEvent added.
​https://bugs.webkit.org/show_bug.cgi?id=227718

Patch by Qiaosong Zhou <​qiaosong_zhou@apple.com> on 2021-07-26
Reviewed by Chris Dumez.

• UserInterface/Models/ScriptTimelineRecord.js:

(WI.ScriptTimelineRecord.EventType.displayName):

LayoutTests:
Added support for FormDataEvent. Rebaselined.
​https://bugs.webkit.org/show_bug.cgi?id=227718

Patch by Johnson Zhou <​qiaosong_zhou@apple.com> on 2021-07-26
Reviewed by Chris Dumez.

• platform/ios-wk2/imported/w3c/web-platform-tests/html/dom/idlharness.https-expected.txt:
• platform/ios-wk2/imported/w3c/web-platform-tests/mathml/relations/html5-tree/math-global-event-handlers.tentative-expected.txt:
• platform/mac-wk1/imported/w3c/web-platform-tests/html/dom/idlharness.https-expected.txt:
• platform/mac-wk1/imported/w3c/web-platform-tests/mathml/relations/html5-tree/math-global-event-handlers.tentative-expected.txt:
• platform/mac-wk2/imported/w3c/web-platform-tests/html/dom/idlharness.https-expected.txt:
• platform/mac-wk2/imported/w3c/web-platform-tests/mathml/relations/html5-tree/math-global-event-handlers.tentative-expected.txt:

12:39 PM Changeset in webkit [280309] by ysuzuki@apple.com

• 2 edits in trunk/Source/JavaScriptCore

Add Temporal to features.json
​https://bugs.webkit.org/show_bug.cgi?id=228292

Reviewed by Myles C. Maxfield.

• features.json:

12:36 PM Changeset in webkit [280308] by jer.noble@apple.com

• 16 edits
  2 adds in trunk

[Cocoa] Playback stalls on bilibili.com
​https://bugs.webkit.org/show_bug.cgi?id=228239
<rdar://80419477>

Reviewed by Eric Carlson.

Source/WebCore:

Test: media/media-source/media-source-effectiveplaybackrate.html

bilibili.com will attempt to detect playback stalls by detecting when video.currentTime
returns the same value while video.paused is false. When the GPU process is enabled, we use
a heuristic to provide the answer to currentTime to avoid making synchronous calls to the
GPU process. This heuristic uses the reported playback rate of the MediaPlayerPrivate to
derive the currentTime. However, MediaPlayerPrivateAVFoundationObjC does not override the
default implementation of rate(), and so always returns a reported rate of zero.

To make the call less ambiguous, add a MediaPlayer and MediaPlayerPrivate effectiveRate()
method, which by default will just return rate(), as some ports only report the
requested rate and not the effective rate. Then add overrides for both rate() and
effectiveRate() on both MediaPlayerPrivateAVFoundationObjC and
MediaPlayerPrivateMediaSourceAVFObjC.

• html/HTMLMediaElement.cpp:

(WebCore::HTMLMediaElement::updatePlaybackRate):
(WebCore::HTMLMediaElement::mediaPlayerRateChanged):

• platform/graphics/MediaPlayer.cpp:

(WebCore::MediaPlayer::effectiveRate const):

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

(WebCore::MediaPlayerPrivateInterface::effectiveRate const):

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

(WebCore::MediaPlayerPrivateAVFoundationObjC::rate const):
(WebCore::MediaPlayerPrivateAVFoundationObjC::effectiveRate const):

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

(WebCore::MediaPlayerPrivateMediaSourceAVFObjC::rate const):
(WebCore::MediaPlayerPrivateMediaSourceAVFObjC::effectiveRate const):

• testing/Internals.cpp:

(WebCore::Internals::isMediaElementHidden):
(WebCore::Internals::elementEffectivePlaybackRate):

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

Source/WebKit:

Replace rate() -> effectiveRate().

• GPUProcess/media/RemoteMediaPlayerProxy.cpp:

(WebKit::RemoteMediaPlayerProxy::mediaPlayerRateChanged):

LayoutTests:

• media/media-source/media-source-effectiveplaybackrate-expected.txt: Added.
• media/media-source/media-source-effectiveplaybackrate.html: Added.

12:22 PM Changeset in webkit [280307] by commit-queue@webkit.org

• 2 edits in trunk/Source/WebCore

[WebXR] WebXRSession::m_visibilityState is uninitialized
​https://bugs.webkit.org/show_bug.cgi?id=228286

Patch by Ada Chan <​ada.chan@apple.com> on 2021-07-26
Reviewed by Tim Horton.

WebXRSession::m_visibilityState should be initialized to XRVisibilityState::Visible.

• Modules/webxr/WebXRSession.h:

12:18 PM Changeset in webkit [280306] by Aditya Keerthi

• 7 edits
  2 adds in trunk

[iOS] Page background color does not update after UIUserInterfaceLevel change
​https://bugs.webkit.org/show_bug.cgi?id=228282
rdar://80490391

Reviewed by Wenson Hsieh.

Source/WebCore:

The default background color of the root element is a semantic color
that adapts to changes in user interface style (light/dark mode) and
user interface level (base/elevated).

Currently, the default background color is correctly recalculated
after a change to the user interface style. However, the existing
logic does not update the color unless there is a change to the user
interface style. This behavior is incorrect, since a change to the
user interface level, without a change to the user interface style,
is ignored.

A common scenario in which a user interface level change is not
accompanied by a user interface style change, is when a WKWebView is
created and then presented as a page sheet, form sheet, or popover.
In this scenario, the default background color is currently incorrect.

To fix, ensure that the background color is recalculated if any of the
traits that affect semantic colors is changed.

Test: fast/css/ios/update-user-interface-level.html

• page/FrameView.cpp:

(WebCore::FrameView::recalculateBaseBackgroundColor):

Recalculate the background color if any of the traits that affect
semantic colors is changed, not just a change in user interface
style (light/dark mode).

• page/FrameView.h:
• testing/InternalSettings.cpp:

Added a testing hook to change the current user interface level.

(WebCore::InternalSettings::resetToConsistentState):
(WebCore::InternalSettings::setUseDarkAppearance):
(WebCore::InternalSettings::setUseElevatedUserInterfaceLevel):

• testing/InternalSettings.h:
• testing/InternalSettings.idl:

LayoutTests:

Added a test to verify that a change in user interface level correctly
updates the page's background color.

• fast/css/ios/update-user-interface-level-expected.txt: Added.
• fast/css/ios/update-user-interface-level.html: Added.

12:01 PM Changeset in webkit [280305] by Ayumi Kojima

• 2 edits in trunk/LayoutTests

[ iOS Debug] 3 editing/pasteboard/smart-paste-paragraph tests are flaky failing.
​https://bugs.webkit.org/show_bug.cgi?id=228285

Unreviewed test gardening.

• platform/ipad/TestExpectations:

11:54 AM Changeset in webkit [280304] by Russell Epstein

• 1 copy in tags/Safari-612.1.24.11.5

Tag Safari-612.1.24.11.5.

11:52 AM Changeset in webkit [280303] by Russell Epstein

• 8 edits in branches/safari-612.1.24.11-branch/Source

Versioning.

WebKit-7612.1.24.11.5

11:45 AM Changeset in webkit [280302] by Eric Hutchison

• 3 edits in trunk/LayoutTests

REGRESSION (r279427): [ Mac wk1 and iOS ] imported/w3c/web-platform-tests/html/interaction/focus/the-autofocus-attribute/document-with-fragment-valid.html is a flaky failure.
​https://bugs.webkit.org/show_bug.cgi?id=227762.

Unreviewed test gardening.

• platform/ios-wk2/TestExpectations:
• platform/mac-wk1/TestExpectations:

11:43 AM Changeset in webkit [280301] by Russell Epstein

• 8 edits in branches/safari-611.3.10.1-branch/Source

Versioning.

WebKit-7611.3.10.1.6

11:42 AM Changeset in webkit [280300] by Chris Dumez

• 3 edits
  2 adds in trunk

XML documents end up with a unique origin in WebKit only
​https://bugs.webkit.org/show_bug.cgi?id=228254

Reviewed by Darin Adler.

Source/WebCore:

XML documents end up with a unique origin in WebKit only. They have a regular origin in Blink and Firefox.
This patch is aligning our behavior with other browsers.

Test: http/tests/misc/xml-document-origin.html

• xml/XMLTreeViewer.cpp:

(WebCore::XMLTreeViewer::transformDocumentToTreeView):

LayoutTests:

Add layout test coverage.

• http/tests/misc/xml-document-origin-expected.txt: Added.
• http/tests/misc/xml-document-origin.html: Added.

10:53 AM Changeset in webkit [280299] by Chris Dumez

• 2 edits in trunk/Source/WebKit

Make sure the ProcessAssertion gets destroyed on the main thread
​https://bugs.webkit.org/show_bug.cgi?id=228281
<rdar://81048308>

Reviewed by Geoffrey Garen.

Follow-up to r279877 to make sure that the ProcessAssertion gets destroyed on the main thread in
ProcessAssertion::acquireAsync(), even if the completionHandler is null.

• UIProcess/ios/ProcessAssertionIOS.mm:

(WebKit::ProcessAssertion::acquireAsync):

10:51 AM Changeset in webkit [280298] by jer.noble@apple.com

• 12 edits
  2 adds in trunk

[iOS] All home screen web apps resume when any home screen web app is foregrounded
​https://bugs.webkit.org/show_bug.cgi?id=228246
<rdar://72949281>

Reviewed by Eric Carlson.

Source/WebCore:

Test: media/video-page-visibility-restriction.html

On iOS, home screen web apps all run from the same UIProcess, SafariViewService. So when
one Web App is foregrounded, the SafariViewService itself is foregrounded, and all WKWebViews
(one for each Web App) are foregrounded as well, allowing all Web Apps to resume audio
playback. This is not ideal; ideally, all Web Apps will be allowed to continue to play
audio in the background. But until we can fix that bug, the current behavior of pausing
audio from Web App A when A is backgrounded, and resuming audio from A when Web App B is
foregrounded feels super broken.

Add a new WKPreference/WebPreference/Setting and matching MediaElementSession restriction
that will block playback of audible media elements when the media element's page is not
visible. When adopted by SafariViewService, this would keep multiple Web Apps (and indeed
SafariViewController pages) from starting playback when any other is foregrounded.

• html/HTMLMediaElement.cpp:

(WebCore::HTMLMediaElement::initializeMediaSession):
(WebCore::HTMLMediaElement::visibilityStateChanged):

• html/MediaElementSession.cpp:

(WebCore::MediaElementSession::visibilityChanged):
(WebCore::MediaElementSession::playbackStateChangePermitted const):

• html/MediaElementSession.h:
• platform/audio/PlatformMediaSession.h:
• testing/Internals.cpp:

(WebCore::Internals::setMediaElementRestrictions):

Source/WebKit:

Add a private WKPreference for setting the new WebPreference.

• UIProcess/API/Cocoa/WKPreferences.mm:

(-[WKPreferences _requiresPageVisibilityToPlayAudio]):
(-[WKPreferences _setRequiresPageVisibilityToPlayAudio:]):

• UIProcess/API/Cocoa/WKPreferencesPrivate.h:

Source/WTF:

• Scripts/Preferences/WebPreferences.yaml:

LayoutTests:

• media/video-page-visibility-restriction-expected.txt: Added.
• media/video-page-visibility-restriction.html: Added.

10:24 AM Changeset in webkit [280297] by Kocsen Chung

• 1 copy in tags/Safari-612.1.24.11.4

Tag Safari-612.1.24.11.4.

10:15 AM Changeset in webkit [280296] by Russell Epstein

• 8 edits
  2 adds in branches/safari-612.1.25-branch

Cherry-pick r280271. rdar://problem/81117003

REGRESSION (r279751): WebContent process often crashes when hovering over content on apple.com
​https://bugs.webkit.org/show_bug.cgi?id=228247
rdar://81010093

Reviewed by Tim Horton.

Source/WebCore:

Add an internal testing hook that can be used to trigger text recognition for the given element. While we should
eventually combine this with another testing hook to simulate VisionKit text recognition results, the new test
using this internal hook shouldn't make its way into VisionKit anyways, so this isn't necessary for now.

See WebKit ChangeLog for more details.

Test: fast/images/text-recognition/text-recognition-in-transparent-video.html

• testing/Internals.cpp: (WebCore::Internals::requestTextRecognition):
• testing/Internals.h:
• testing/Internals.idl:

Source/WebKit:

After r279751, the snapshot fallback codepath I added in createShareableBitmap to handle the edge case of
fully transparent images causes us to now take snapshots when hovering over fully transparent video elements,
and attempt to recognize text in them. This is because RenderVideo is a RenderImage subclass without a cached
image, so we'll end up going down the transparent renderer codepath instead of bailing with a null bitmap.

However, since CachedImages are null for video elements, before we even get to VisionKit, we end up crashing
with a nullptr-deref inside WebPage::requestTextRecognition, which assumes that RenderImage::cachedImage()
is non-null.

To address this, we make two minor adjustments (see below).

• WebProcess/WebCoreSupport/ShareableBitmapUtilities.cpp: (WebKit::createShareableBitmap):

Limit the snapshotting fallback to non-media images (i.e. non-RenderMedia).

• WebProcess/WebPage/WebPage.cpp: (WebKit::WebPage::requestTextRecognition):

Make this robust in the case where CachedImage is null, to avoid the possibility for similar crashes in the
future.

LayoutTests:

• fast/images/text-recognition/text-recognition-in-transparent-video-expected.txt: Added.
• fast/images/text-recognition/text-recognition-in-transparent-video.html: Added.

git-svn-id: ​https://svn.webkit.org/repository/webkit/trunk@280271 268f45cc-cd09-0410-ab3c-d52691b4dbfc

10:15 AM Changeset in webkit [280295] by Kocsen Chung

• 8 edits in branches/safari-612.1.24.11-branch/Source

Versioning.

WebKit-7612.1.24.11.4

5:10 AM Changeset in webkit [280294] by Adrian Perez de Castro

• 1 copy in releases/WPE WebKit/webkit-2.32.3

WPE WebKit 2.32.3

5:10 AM Changeset in webkit [280293] by Adrian Perez de Castro

• 4 edits in releases/WebKitGTK/webkit-2.32

Unreviewed. Update OptionsWPE.cmake and NEWS for the 2.32.3 release

.:

• Source/cmake/OptionsWPE.cmake: Bump version numbers.

Source/WebKit:

• wpe/NEWS: Add release notes for 2.32.3

2:50 AM Changeset in webkit [280292] by commit-queue@webkit.org

• 3 edits in trunk/Source/WebCore

[Pipewire] Muting the display capture closes the Portal session
​https://bugs.webkit.org/show_bug.cgi?id=228265

Patch by Philippe Normand <​pnormand@igalia.com> on 2021-07-26
Reviewed by Xabier Rodriguez-Calvar.

The session should not be closed in that case, because setting the capture state to active
again would have no effect. This patch also removes useless CRLFs from WTFLogAlways() calls
and increases the session Close call timeout to 100ms, 10ms was too low, at least for my
setup.

• platform/mediastream/gstreamer/GStreamerDisplayCaptureDeviceManager.cpp:

(WebCore::GStreamerDisplayCaptureDeviceManager::createDisplayCaptureSource):
(WebCore::GStreamerDisplayCaptureDeviceManager::stopSource):

• platform/mediastream/gstreamer/GStreamerVideoCaptureSource.cpp:

(WebCore::GStreamerVideoCaptureSource::~GStreamerVideoCaptureSource):
(WebCore::GStreamerVideoCaptureSource::stopProducingData):

12:48 AM Changeset in webkit [280291] by commit-queue@webkit.org

• 4 edits in trunk/Source/WebKit

[GTK] MiniBrowser crashes when closed while capturing desktop
​https://bugs.webkit.org/show_bug.cgi?id=228232

Patch by Philippe Normand <​pnormand@igalia.com> on 2021-07-26
Reviewed by Adrian Perez de Castro.

Disconnect the WebView from the WebPageProxy just before the final GObject dispose call, in
order to avoid UIClient notifications on the being-disposed WebView.

• UIProcess/API/glib/WebKitUIClient.cpp:

(detachUIClientFromView):

• UIProcess/API/glib/WebKitUIClient.h:
• UIProcess/API/glib/WebKitWebView.cpp:

(webkitWebViewDispose):

12:42 AM Changeset in webkit [280290] by Ziran Sun

• 4 edits in trunk

[css-grid] svg image as grid items should use the overriding logical width/height when defined to compute the logical height/width
​https://bugs.webkit.org/show_bug.cgi?id=228105

Source/WebCore:

As discussed at ​https://github.com/w3c/csswg-drafts/issues/6286#issuecomment-866986544, degenerate
aspect ratios derived from SVG width/height attributes fall back to viewbox aspect ratio
(whether due to negative values or zero values).

When computing the logical height/width using an intrinsic aspect ratio, RenderReplaced uses the
overridingLogicalWidth/overridingLogicalHeight whenever defined as long as the flex or
grid item has an intrinsic size. For an SVG graphic though, it's common to have an intrinsic aspect
ratio but not to have an intrinsic width or height. For this special case, we still should use
overridingLogicalWidth/overridingLogicalHeight for logical height/width calculations.

Reviewed by Javier Fernandez.

• rendering/RenderReplaced.cpp:

(WebCore::RenderReplaced::computeReplacedLogicalWidth const):
(WebCore::RenderReplaced::computeReplacedLogicalHeight const):

LayoutTests:

Reviewed by Javier Fernandez.

Update TestExpectations as 4 failed grid WPT tests are now passing.

• TestExpectations:

Jul 25, 2021:

3:32 PM Changeset in webkit [280289] by Alexey Shvayka

• 36 edits
  2 adds in trunk

Partly implement Function.prototype.{caller,arguments} reflection proposal
​https://bugs.webkit.org/show_bug.cgi?id=158116

Reviewed by Yusuke Suzuki.

JSTests:

• ChakraCore/test/strict/19.function.baseline:
• ChakraCore/test/strict/22.callerCalleeArguments.baseline-jsc:
• microbenchmarks/function-prototype-get.js: Added.
• microbenchmarks/reflect-own-keys-function.js: Added.
• stress/for-in-shadow-non-enumerable.js:
• stress/function-hidden-as-caller.js:
• stress/has-own-property-arguments.js:
• stress/object-assign-fast-path.js:
• stress/put-to-proto-chain-overrides-put.js:
• stress/reflect-set.js:
• test262/config.yaml: Skip 3 test cases that are now incorrect.
• test262/expectations.yaml: Mark 2 test cases as passing.

Source/JavaScriptCore:

To ensure web-compatibility, only the safe subset of Function.prototype.{caller,arguments}
reflection proposal [1] is implemented, which is currently shipped in SpiderMonkey.

Complete list of differences from the proposed spec:

1. Cross-realm receiver function is allowed instead of throwing a TypeError.

Throwing is likely safe to ship, but #225997 needs to be fixed first for
custom properties to receive correct global object.

2. Cross-realm caller function is returned instead of null.

Hiding cross-realm caller may break things: we currently have a test for
the opposite behavior.

3. Defines "caller" and "arguments" setters that throw for disallowed receivers, instead failing silently in sloppy mode.

This is actually more restrictive than the spec, which is preferable,
and aligns with V8 and SM.

Most importantly, this patch removes own "caller" and "arguments" properties from
sloppy mode ES5 functions. They were non-configurable, making it harder to use
their holder as a ProxyTarget. They were also non-writable, with a constantly
changing Value, which violated the invariants of internal methods [2].

As a result, JSFunction methods are greatly simplified, especially defineOwnProperty()
and getOwnSpecialPropertyNames(). The latter is now 2.1x faster according to the
provided microbenchmark. Also, removes double "prototype" lookup from Get,
which is a 10% progression.

[1]: ​https://github.com/claudepache/es-legacy-function-reflection
[2]: ​https://tc39.es/ecma262/#sec-invariants-of-the-essential-internal-methods

• runtime/ClonedArguments.cpp:

(JSC::ClonedArguments::getOwnPropertySlot):
(JSC::ClonedArguments::materializeSpecials):

• runtime/FunctionExecutable.h:
• runtime/FunctionPrototype.cpp:

(JSC::FunctionPrototype::addFunctionProperties):
(JSC::isAllowedReceiverFunctionForCallerAndArguments):
(JSC::RetrieveArgumentsFunctor::RetrieveArgumentsFunctor):
(JSC::RetrieveArgumentsFunctor::result const):
(JSC::RetrieveArgumentsFunctor::operator() const):
(JSC::retrieveArguments):
(JSC::JSC_DEFINE_CUSTOM_GETTER):
(JSC::RetrieveCallerFunctionFunctor::RetrieveCallerFunctionFunctor):
(JSC::RetrieveCallerFunctionFunctor::result const):
(JSC::RetrieveCallerFunctionFunctor::operator() const):
(JSC::retrieveCallerFunction):
(JSC::JSC_DEFINE_CUSTOM_SETTER):
(JSC::FunctionPrototype::initRestrictedProperties): Deleted.

• runtime/FunctionPrototype.h:
• runtime/JSFunction.cpp:

(JSC::JSFunction::getOwnPropertySlot):
(JSC::JSFunction::getOwnSpecialPropertyNames):
(JSC::JSFunction::put):
(JSC::JSFunction::deleteProperty):
(JSC::JSFunction::defineOwnProperty):
(JSC::RetrieveArgumentsFunctor::RetrieveArgumentsFunctor): Deleted.
(JSC::RetrieveArgumentsFunctor::result const): Deleted.
(JSC::RetrieveArgumentsFunctor::operator() const): Deleted.
(JSC::retrieveArguments): Deleted.
(JSC::JSC_DEFINE_CUSTOM_GETTER): Deleted.
(JSC::RetrieveCallerFunctionFunctor::RetrieveCallerFunctionFunctor): Deleted.
(JSC::RetrieveCallerFunctionFunctor::result const): Deleted.
(JSC::RetrieveCallerFunctionFunctor::operator() const): Deleted.
(JSC::retrieveCallerFunction): Deleted.

• runtime/JSGlobalObject.cpp:

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

• runtime/JSGlobalObject.h:

Remove unused m_throwTypeErrorGetterSetter and make ThrowTypeError lazily-created.

• runtime/JSGlobalObjectFunctions.cpp:

(JSC::JSC_DEFINE_HOST_FUNCTION):

• runtime/JSGlobalObjectFunctions.h:
• runtime/JSObject.cpp:

(JSC::JSObject::putDirectCustomGetterSetterWithoutTransition):

• runtime/JSObject.h:

LayoutTests:

• inspector/model/remote-object-get-properties-expected.txt:
• inspector/runtime/getDisplayableProperties-expected.txt:
• inspector/runtime/getProperties-expected.txt:
• js/Object-getOwnPropertyNames-expected.txt:
• js/basic-strict-mode-expected.txt:
• js/kde/function_arguments-expected.txt:
• js/kde/script-tests/function_arguments.js:
• js/non-strict-function-properties-expected.txt:
• js/script-tests/Object-getOwnPropertyNames.js:
• js/script-tests/basic-strict-mode.js:
• js/script-tests/non-strict-function-properties.js:
• js/script-tests/throw-type-error-is-unique.js:

12:40 PM Changeset in webkit [280288] by Wenson Hsieh

• 20 edits
  2 adds in trunk

[iOS] Unified field is unselected after focusing URL bar if text was selected in a fixed position container
​https://bugs.webkit.org/show_bug.cgi?id=228269
rdar://80556392

Reviewed by Tim Horton.

Source/WebKit:

In Safari on iOS 15, if the selection (either ranged or caret) is inside a fixed position container when the
user taps on the unified field, we'll immediately clear the text selection inside the unified field upon
bringing up the keyboard. This happens because the tab pill in iOS 15 is lowered as the URL bar is focused,
which causes the web view to scroll slightly. This, in turn, induces a brief unstable scrolling tree state,
which then causes us to temporarily hide and show selection views while scrolling in unstable state (see
r209931) by calling -deactivateSelection and then -activateSelection on the text interaction assistant.
Calling -[UIWKTextInteractionAssistant activateSelection] then causes UIKit to dispatch a
UITextSelectionViewActivatedNotification; In the unified field, which is a UITextField subclass, UIKit code then
listens for this notification and responds to it by clearing the selection if the newly activated selection's
host view (WKContentView) is different than itself, thereby causing the bug.

To fix this, we simply make two (minor) adjustments to the logic for temporarily hiding and showing the
selection while performing an unstable scroll. See below for more details.

Test: editing/selection/ios/scrolling-with-fixed-selection-does-not-unselect-native-text-field.html

• UIProcess/ios/WKContentViewInteraction.mm:

(-[WKContentView shouldHideSelectionWhenScrolling]):

Only hide and (later) restore the selection in non-editable text if the selection is ranged. This is because
caret selections in non-editable content are not user-visible anyways, so there's no need to temporarily
suppress the selection.

(-[WKContentView _updateChangedSelection:]):

Only attempt to show the selection views again if doing so doesn't cause us to steal first responder status away
from the existing first responder; otherwise, we'll wait until we -becomeFirstResponder to -activateSelection.

(-[WKContentView selectionInteractionAssistant]): Deleted.

Tools:

We already have some very basic support for installing and removing native text fields in the view hierarchy,
through TestRunner::(add|remove)ChromeInputField(). In order to support the new layout test, we additionally
implement the ability to:

• Set text inside the native chrome input field that was installed using addChromeInputField().
• Select all text inside the chrome input field.
• Query the chrome input field for the currently selected text.

We only support iOS for the time being, with stubs on other platforms, since the new test that uses this
functionality is iOS-specific; if needed in the future for a similar test on other platforms, we can implement
the stubbed testing hooks on TestRunner and PlatformWebView as well.

• DumpRenderTree/TestRunner.h:
• DumpRenderTree/mac/TestRunnerMac.mm:

(TestRunner::setTextInChromeInputField):
(TestRunner::selectChromeInputField):
(TestRunner::getSelectedTextInChromeInputField):

• DumpRenderTree/win/TestRunnerWin.cpp:

(TestRunner::setTextInChromeInputField):
(TestRunner::selectChromeInputField):
(TestRunner::getSelectedTextInChromeInputField):

• WebKitTestRunner/InjectedBundle/Bindings/TestRunner.idl:
• WebKitTestRunner/InjectedBundle/InjectedBundle.cpp:

(WTR::InjectedBundle::didReceiveMessageToPage):
(WTR::InjectedBundle::postSetTextInChromeInputField):
(WTR::InjectedBundle::postSelectChromeInputField):
(WTR::InjectedBundle::postGetSelectedTextInChromeInputField):

• WebKitTestRunner/InjectedBundle/InjectedBundle.h:
• WebKitTestRunner/InjectedBundle/TestRunner.cpp:

(WTR::TestRunner::setTextInChromeInputField):
(WTR::TestRunner::selectChromeInputField):
(WTR::TestRunner::getSelectedTextInChromeInputField):
(WTR::TestRunner::callSetTextInChromeInputFieldCallback):
(WTR::TestRunner::callSelectChromeInputFieldCallback):
(WTR::TestRunner::callGetSelectedTextInChromeInputFieldCallback):

• WebKitTestRunner/InjectedBundle/TestRunner.h:
• WebKitTestRunner/PlatformWebView.h:
• WebKitTestRunner/TestInvocation.cpp:

(WTR::TestInvocation::didReceiveMessageFromInjectedBundle):

• WebKitTestRunner/gtk/PlatformWebViewGtk.cpp:

(WTR::PlatformWebView::setTextInChromeInputField):
(WTR::PlatformWebView::selectChromeInputField):
(WTR::PlatformWebView::getSelectedTextInChromeInputField):

• WebKitTestRunner/ios/PlatformWebViewIOS.mm:

(WTR::chromeInputField):

Additionally do some light refactoring by pulling out logic for grabbing the chrome input field (i.e. a view
with a tag of 1 under the window) out into a separate helper method. Use this helper in a few places below.

(WTR::PlatformWebView::addChromeInputField):
(WTR::PlatformWebView::setTextInChromeInputField):
(WTR::PlatformWebView::selectChromeInputField):
(WTR::PlatformWebView::getSelectedTextInChromeInputField):
(WTR::PlatformWebView::removeChromeInputField):

• WebKitTestRunner/mac/PlatformWebViewMac.mm:

(WTR::PlatformWebView::setTextInChromeInputField):
(WTR::PlatformWebView::selectChromeInputField):
(WTR::PlatformWebView::getSelectedTextInChromeInputField):

• WebKitTestRunner/win/PlatformWebViewWin.cpp:

(WTR::PlatformWebView::setTextInChromeInputField):
(WTR::PlatformWebView::selectChromeInputField):
(WTR::PlatformWebView::getSelectedTextInChromeInputField):

• WebKitTestRunner/wpe/PlatformWebViewWPE.cpp:

(WTR::PlatformWebView::setTextInChromeInputField):
(WTR::PlatformWebView::selectChromeInputField):
(WTR::PlatformWebView::getSelectedTextInChromeInputField):

LayoutTests:

Add a new layout test that installs, focuses, and selects text inside a native UITextField (simulating Safari's
URL field) while the DOM selection is inside a fixed position container, and then scrolls the web view a bit to
temporarily induce an unstable scrolling tree state. After this, we verify that the text selection inside the
native text field has not been cleared.

See Tools/ changes for more details.

• editing/selection/ios/scrolling-with-fixed-selection-does-not-unselect-native-text-field-expected.txt: Added.
• editing/selection/ios/scrolling-with-fixed-selection-does-not-unselect-native-text-field.html: Added.
• resources/ui-helper.js:

(window.UIHelper.addChromeInputField):
(window.UIHelper.removeChromeInputField):
(window.UIHelper.setTextInChromeInputField):
(window.UIHelper.selectChromeInputField):
(window.UIHelper.getSelectedTextInChromeInputField):
(window.UIHelper):

11:55 AM WebKitGTK/2.32.x edited by Adrian Perez de Castro

(diff)

11:29 AM Changeset in webkit [280287] by commit-queue@webkit.org

• 2 edits in trunk

[GTK] USE_OPENGL_OR_ES should not be an automagic feature
​https://bugs.webkit.org/show_bug.cgi?id=228266

Patch by Michael Catanzaro <Michael Catanzaro> on 2021-07-25
Reviewed by Fujii Hironori.

The USE_OPENGL_OR_ES feature flag added in r271220 is currently automagic: that is, if
neither OpenGL nor OpenGL ES is available at build time, it gets silently disabled. This is
not OK because it makes it easy for distributors to accidentally fail to enable OpenGL
support. We should require manually disabling the feature with -DUSE_OPENGL_OR_ES=OFF in
order to build with OpenGL disabled.

• Source/cmake/OptionsGTK.cmake:

6:00 AM Changeset in webkit [280286] by cathiechen

• 1 edit
  1 add in trunk/PerformanceTests

[Performance test][css-contain] Add test to contain: size layout
​https://bugs.webkit.org/show_bug.cgi?id=227948

Reviewed by Ryosuke Niwa.

This test emulates the scenario that a small part of the page is changed. If it's applied contain: size layout,
the performance should be improved. This test is a transform of the test [1] in blink.

[1] third_party/blink/perf_tests/layout/css-contain-change-text.html

• Layout/css-contain-change-size.html: Added.