Jul 8, 2019:

10:32 PM Changeset in webkit [247254] by sbarati@apple.com
  • 8 edits
    8 adds in trunk

[WHLSL Import 23 new JS reference spec tests

Reviewed by Myles C. Maxfield.


This patch imports a bunch of JS reference spec tests on our way to
completing https://bugs.webkit.org/show_bug.cgi?id=199595

It also fixes the recursion checker phase. That phase had two bugs:

  1. We'd assert after visiting the function declaration that it was

still in the set. However, it will not be in the set when we actually
detect recursion.

  1. We would not visit the arguments to a call, so if they contained other

calls which were recursive, we would not detect such recursive calls.

Tests: webgpu/whlsl-int-literal-compare.html


  • Modules/webgpu/WHLSL/WHLSLRecursionChecker.cpp:


This moves some stuff down into whlsl-test-harness which are needed by
all tests. This also adds a new checkFail that ensures the program never
runs (e.g, it has a compile error).

  • webgpu/js/whlsl-test-harness.js:


  • webgpu/whlsl-bitwise-bool-ops-expected.txt:
  • webgpu/whlsl-bitwise-bool-ops.html:
  • webgpu/whlsl-int-literal-compare-expected.txt: Added.
  • webgpu/whlsl-int-literal-compare.html: Added.
  • webgpu/whlsl-simple-tests-expected.txt: Added.
  • webgpu/whlsl-simple-tests.html: Added.
  • webgpu/whlsl-type-mismatch-expected.txt: Added.
  • webgpu/whlsl-type-mismatch.html: Added.
  • webgpu/whlsl-uint-bitwise-expected.txt: Added.
  • webgpu/whlsl-uint-bitwise.html: Added.
8:30 PM Changeset in webkit [247253] by Devin Rousso
  • 3 edits in trunk/Source/WebInspectorUI

Web Inspector: Debugger: special breakpoints and event targets should be sorted into separate "areas"

Reviewed by Joseph Pecoraro.

Move the special All Requests breakpoint down to be right above all other URL breakpoints.
Move the window object tree element below other DOM node tree elements.

  • UserInterface/Views/DebuggerSidebarPanel.js:


  • UserInterface/Views/SourcesNavigationSidebarPanel.js:


8:16 PM Changeset in webkit [247252] by Devin Rousso
  • 3 edits
    4 adds in trunk/Source/WebInspectorUI

Web Inspector: DOM Debugger: there should be different icons for each type of DOM breakpoint

Reviewed by Joseph Pecoraro.

[S] for Subtree Modified
[A] for Attribute Modified
[R] for Node Removed

  • UserInterface/Views/DOMBreakpointTreeElement.js:


  • UserInterface/Views/DOMBreakpointTreeElement.css: Added.

(.breakpoint.dom.breakpoint-for-subtree-modified:not(.breakpoint-paused-icon) .icon):
(.breakpoint.dom.breakpoint-for-attribute-modified:not(.breakpoint-paused-icon) .icon):
(.breakpoint.dom.breakpoint-for-node-removed:not(.breakpoint-paused-icon) .icon):

  • UserInterface/Main.html:
  • UserInterface/Images/DOMBreakpointAttributeModified.svg: Added.
  • UserInterface/Images/DOMBreakpointNodeRemoved.svg: Added.
  • UserInterface/Images/DOMBreakpointSubtreeModified.svg: Added.
8:10 PM Changeset in webkit [247251] by Wenson Hsieh
  • 2 edits in trunk/LayoutTests

[iOS 13] paste-does-not-fire-promises-while-sanitizing-web-content.html times out when run with multiple iterations

Reviewed by Tim Horton.

Make this test more robust to work around bugs in iOS 13 that currently prevent some single tap gestures from
being recognized, due to conflicts with double tap gesture recognizers.

  • Make both the subframe and editable element taller, so that they're easier to hit.
  • Tap twice when attempting to trigger a paste, in hopes that at least one of the taps will be recognized as a

single click.

  • Make several event listeners one-shot, to prevent the double tap gesture from triggering paste twice.
  • editing/pasteboard/paste-does-not-fire-promises-while-sanitizing-web-content.html:
8:06 PM Changeset in webkit [247250] by Chris Dumez
  • 11 edits in trunk/Source/WebKit

Speculative fix for crashes under LocalStorageDatabaseTracker::databasePath()

Reviewed by Ryosuke Niwa.

Speculative fix for crashes under LocalStorageDatabaseTracker::databasePath():

  • Add new localStorageDirectory() getter to LocalStorageDatabaseTracker which calls isolatedCopy() on m_localStorageDirectory before returning it. Use it everywhere instead of m_localStorageDirectory since it is not safe to use the same String from various threads like it was done.
  • Move localStorageDirectory when constructing the LocalStorageDatabaseTracker instead of copying it.
  • Make sure that LocalStorageDatabaseTracker and StorageManager are both constructed and destroyed on the main thread.
  • NetworkProcess/NetworkSession.cpp:


  • NetworkProcess/NetworkSession.h:
  • NetworkProcess/WebStorage/LocalStorageDatabaseTracker.cpp:

(WebKit::LocalStorageDatabaseTracker::localStorageDirectory const):
(WebKit::LocalStorageDatabaseTracker::origins const):
(WebKit::LocalStorageDatabaseTracker::databasePath const):

  • NetworkProcess/WebStorage/LocalStorageDatabaseTracker.h:
  • NetworkProcess/WebStorage/StorageManager.cpp:


  • NetworkProcess/WebStorage/StorageManager.h:
  • NetworkProcess/WebStorage/ios/LocalStorageDatabaseTrackerIOS.mm:

(WebKit::LocalStorageDatabaseTracker::platformMaybeExcludeFromBackup const):

  • NetworkProcess/cocoa/NetworkSessionCocoa.mm:


  • NetworkProcess/curl/NetworkSessionCurl.cpp:


  • NetworkProcess/soup/NetworkSessionSoup.cpp:


7:22 PM Changeset in webkit [247249] by Devin Rousso
  • 2 edits in trunk/Source/WebInspectorUI

Uncaught Exception: Unexpected enum value: CPU

Reviewed by Joseph Pecoraro.

  • UserInterface/Controllers/TimelineManager.js:

Filter the list of auto-capture instruments based on what's actually supported, instead of
directly using the saved list from the WI.Setting.

7:19 PM Changeset in webkit [247248] by Devin Rousso
  • 2 edits in trunk/Source/WebInspectorUI

Enabling the Layers tab hides the paint flashing button

Reviewed by Joseph Pecoraro.

Still show the Compositing Borders and Paint Flashing navigation items in the Elements tab
even when the Layers tab is enabled (just like the Layers details sidebar).

  • UserInterface/Views/DOMTreeContentView.js:

(WI.DOMTreeContentView.prototype.get navigationItems):

7:18 PM Changeset in webkit [247247] by keith_miller@apple.com
  • 10 edits in trunk

Enable Intl.PluralRules and Intl.NumberFormatToParts by default

Reviewed by Yusuke Suzuki.


These features have been around for a while. We should turn them on by default.

  • runtime/IntlNumberFormatPrototype.cpp:


  • runtime/IntlObject.cpp:

(JSC::IntlObject::finishCreation): Deleted.

  • runtime/IntlObject.h:
  • runtime/Options.h:


Remove runtime flag from testing.

  • js/intl-numberformat-format-to-parts.html:
  • js/intl-pluralrules.html:
  • js/script-tests/intl-numberformat-format-to-parts.js:
  • js/script-tests/intl-pluralrules.js:
6:24 PM Changeset in webkit [247246] by Simon Fraser
  • 4 edits in trunk/Source/WebCore

YouTube embedding iframes in WebView sometimes go blank when the video starts playing

Reviewed by Tim Horton.

With some combinations of nested iframes that are being resized from empty, and toggling into
compositing mode, we'd fail to update compositing in the iframe's enclosing document, so never
host the iframes's layers.

Fix by moving some widget-resize-related code into RenderLayerCompositor::widgetDidChangeSize(),
and adding code to schedule a compositing update.

I was unable to come up with a layout test for this.

  • rendering/RenderLayerCompositor.cpp:


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


6:07 PM Changeset in webkit [247245] by Fujii Hironori
  • 3 edits in trunk/Source/WebKitLegacy/win

[WinCairo] ASSERTION FAILED: info.bmBitsPixel == 32 in createCairoContextWithHDC

Reviewed by Per Arne Vollan.

WebView::paint binds m_backingStoreBitmap to a DC by using
SelectObject. WebView::paint can be called recursively, but
m_backingStoreBitmap can't be bound to multiple DCs at the same
time. Then, SelectObject was failing in such case.

Call WebCore::Page::updateRendering before binding
m_backingStoreBitmap instead of after it.

Reverted r202744 change which won't be needed since this change.

  • WebView.cpp:

(WebView::scrollBackingStore): Removed r202744's change.
(WebView::updateBackingStore): Removed m_page->updateRendering().
(WebView::paint): Do m_page->updateRendering() before binding m_backingStoreBitmap.

  • WebView.h: Removed unused WebView::isPainting.
6:03 PM Changeset in webkit [247244] by Fujii Hironori
  • 2 edits in trunk/Tools

JSTestGlobalObject.cpp of bindings-generation-tests is failing for Windows Python

Reviewed by Ross Kirsling.

In Windows Python, preprocessor.pm is using cl.exe. cl.exe was
failing to open testglobalscope_constructors_file which was
created by using tempfile.mkstemp() because it keeps the file
open. Use tempfile.mkdtemp() to create temporary files in the
temporary directory instead of tempfile.mkstemp().

  • Scripts/webkitpy/bindings/main.py:

(BindingsTests.close_and_remove): Deleted.

5:53 PM Changeset in webkit [247243] by Chris Dumez
  • 5 edits in trunk/Source

Use WeakHashSet for WebUserContentControllerProxy::m_processes

Reviewed by Youenn Fablet.


Use WeakHashSet for WebUserContentControllerProxy::m_processses for safety. In theory, a WebProcessProxy could
stay in the map if we failed to call WebProcessProxy::shutDown() before destroying it.

  • UIProcess/UserContent/WebUserContentControllerProxy.cpp:


  • UIProcess/UserContent/WebUserContentControllerProxy.h:



Update WeakHashSet::add() to return an AddResult type, similarly to our other containers.

  • wtf/WeakHashSet.h:


5:35 PM Changeset in webkit [247242] by aakash_jain@apple.com
  • 2 edits in trunk/Tools

[ews-build] Add KillOldProcesses step before running API or Layout tests

Reviewed by Jonathan Bedard.

  • BuildSlaveSupport/ews-build/factories.py:
5:27 PM Changeset in webkit [247241] by Devin Rousso
  • 2 edits in trunk/Source/WebInspectorUI

Web Inspector: REGRESSION: Elements: pasting in the Styles sidebar adds a text node to the DOM tree

Reviewed by Joseph Pecoraro.

  • UserInterface/Base/Main.js:

Bail if event.defaultPrevented, as that means that something else has handled (and likely
overrode) this event.

5:03 PM Changeset in webkit [247240] by Devin Rousso
  • 4 edits in trunk/Source/WebInspectorUI

Web Inspector: the "x" in the Exception icon overlaps the border

Reviewed by Joseph Pecoraro.

  • UserInterface/Images/Exception.svg:
  • UserInterface/Images/Function.svg:
  • UserInterface/Images/TailDeletedFunction.svg:
4:56 PM Changeset in webkit [247239] by Chris Dumez
  • 2 edits in trunk/Source/WebCore

Make Document::postTask() safe to call from a background thread

Reviewed by Alex Christensen.

Make Document::postTask() safe to call from a background thread by not calling makeWeakPtr() on the Document.
Calling makeWeakPtr() on a document from a background thread is not safe since Document is a main thread
object. Instead, capture Document::identifier() in the lambda and lookup the document from its identifier
once we're on the main thread.

  • dom/Document.cpp:


4:51 PM Changeset in webkit [247238] by youenn@apple.com
  • 3 edits in trunk/Source/WebCore

Hop explicitly to the main thread after generating a frame in ScreenDisplayCaptureSourceMac

Reviewed by Eric Carlson.

Instead of locking and setting the current frame from a background thread, hop to the main thread.
This also makes sure the weakThis check is done in the main thread.
Manually tested.

  • platform/mediastream/mac/ScreenDisplayCaptureSourceMac.h:


  • platform/mediastream/mac/ScreenDisplayCaptureSourceMac.mm:

(WebCore::ScreenDisplayCaptureSourceMac::frameAvailable): Deleted.

4:49 PM Changeset in webkit [247237] by Devin Rousso
  • 5 edits
    1 add in trunk/Source/WebInspectorUI

Web Inspector: Debugger: there should be a different icon for URL breakpoints

Reviewed by Joseph Pecoraro.

Use the same (down|up) arrow icon for the Network Tab and the Network timeline.

  • UserInterface/Views/URLBreakpointTreeElement.js:


  • UserInterface/Views/URLBreakpointTreeElement.css:

(.breakpoint.url:not(.breakpoint-paused-icon) .icon): Added.

  • UserInterface/Views/DebuggerSidebarPanel.js:


  • UserInterface/Views/SourcesNavigationSidebarPanel.js:


  • UserInterface/Images/URLBreakpoint.svg: Added.
4:35 PM Changeset in webkit [247236] by Alan Coon
  • 1 copy in tags/Safari-607.3.9

Tag Safari-607.3.9.

4:34 PM Changeset in webkit [247235] by Devin Rousso
  • 3 edits in trunk/Source/WebInspectorUI

Web Inspector: Timelines: CPU: the Duration string isn't localized

Reviewed by Joseph Pecoraro.

  • UserInterface/Views/CPUTimelineView.js:


  • Localizations/en.lproj/localizedStrings.js:
4:31 PM Changeset in webkit [247234] by Alan Coon
  • 1 copy in tags/Safari-608.

Tag Safari-608.

4:08 PM Changeset in webkit [247233] by dbates@webkit.org
  • 6 edits
    2 adds in trunk

[iOS] Support select all in non-editable element

Reviewed by Wenson Hsieh.


Unless we are populating the callout bar always advertise that WebKit can perform Select All
when building with USE(UIKIT_KEYBOARD_ADDITIONS). This way you can press Command + A to select
all text even in a non-editable element just like you can on Mac.

  • UIProcess/ios/WKContentViewInteraction.mm:

(-[WKContentView canPerformActionForWebView:withSender:]):


Add a test to ensure that we can perform Select All even when a non-editable element is focused.

Also disallow testing of Command + A until we fix <https://bugs.webkit.org/show_bug.cgi?id=199475>.
Otherwise, the test fast/events/ios/key-events-comprehensive/key-events-meta.html fails since
Command + A now triggers page selection even in a non-editable (the purpose of this change)
and hence Command + C also triggers the copy command. The triggering of these commands causes
the test to fail the Command + A and Command + C sub-tests since no keypress event is dispatched
for the key that triggered these key commands. This is expected, but Command + A should not have
triggered selection as key-tester.js calls preventDefault() from a keypress handler. That's
the bug. See <https://bugs.webkit.org/show_bug.cgi?id=199475> for more details.

  • editing/selection/ios/select-all-non-editable-text-using-keyboard-expected.txt: Added.
  • editing/selection/ios/select-all-non-editable-text-using-keyboard.html: Added.
  • fast/events/ios/key-events-comprehensive/key-events-meta-expected.txt: Updated result.
  • fast/events/ios/resources/key-tester.js: Disallow Command + A until we fix <https://bugs.webkit.org/show_bug.cgi?id=199475>.
  • platform/ios/TestExpectations: Skip the test until the fix for <rdar://problem/48322899>

has shipped.

4:07 PM Changeset in webkit [247232] by dbates@webkit.org
  • 4 edits in trunk

Command + . generates Escape with key identifier Period, should be Escape

Reviewed by Wenson Hsieh.


Remap the key code for Command + . before we compute the Windows virtual key code.
Otherwise, the Windows virtual key code reveals the pre-mapped key code.

  • platform/ios/WebEvent.mm:

(-[WebEvent initWithKeyEventType:timeStamp:characters:charactersIgnoringModifiers:modifiers:isRepeating:withFlags:withInputManagerHint:keyCode:isTabKey:]):


Update test result.

  • fast/events/ios/key-events-comprehensive/key-events-meta-expected.txt:
3:59 PM Changeset in webkit [247231] by achristensen@apple.com
  • 2 edits in trunk/Source/WebCore

Prevent null dereferencing in SubresourceLoader::init's lambda

Reviewed by Ryosuke Niwa.

Crash logs indicate that sometimes m_documentLoader is null.
This is similar to https://bugs.webkit.org/show_bug.cgi?id=187360

  • loader/SubresourceLoader.cpp:


3:28 PM Changeset in webkit [247230] by Chris Dumez
  • 2 edits in trunk/Source/WebKit

Cleanup uses of NetworkProcess::m_sessionByConnection

Reviewed by Alex Christensen.

Avoid double HashMap lookups and call add() instead of ensure() when appropriate.

  • NetworkProcess/NetworkProcess.cpp:

(WebKit::NetworkProcess::networkSessionByConnection const):

1:51 PM Changeset in webkit [247229] by cturner@igalia.com
  • 33 edits in trunk/LayoutTests

[GTK] Some media fragment tests are flaky

Reviewed by Eric Carlson.

These tests look like they're flakey due to a small stop time. I'm
surmising that based on the bot's load, we step over the 0.5 mark
often. When I run this locally, despite running the fragment tests
100 times, I saw no timeouts.

I have seen crashes in the flakiness dashboard, but I would prefer
for these to have separate bugs so that we can squash them rather
than allowing them to linger. Having them expected to crash means
we have no crash logs to investigate.

Updated the baselines in the LayoutTests/ directory to reflect the
new stop time delta.

  • media/media-fragments/media-fragments.js:

(pause): Increase stop delta, on simulator builds the stddev
crossed the 0.5s mark, and I suspect on GTK builds, our bots being
heavily loaded cause the majority of timeouts.

  • platform/gtk/TestExpectations: Remove expectations.
12:57 PM Changeset in webkit [247228] by Jonathan Bedard
  • 3 edits in trunk/Tools

test-webkitpy: Explicitly use Python 2.7 lldb library

Reviewed by Aakash Jain.

  • Scripts/webkitpy/common/system/systemhost.py:

(SystemHost.path_to_lldb_python_directory): Explicitly use Python 2.7 lldb libary.

  • Scripts/webkitpy/test/main.py:

(_supports_building_and_running_lldb_tests): Run lldb tests on Catalina again.

12:54 PM Changeset in webkit [247227] by timothy_horton@apple.com
  • 2 edits in trunk/Source/WebKit

WKWebView fails to render when another view uses CoreImage filters

Reviewed by Dean Jackson.

  • UIProcess/WebPageProxy.cpp:

Update m_layerHostingMode when it changes, even if we can't message
the Web Content process. This ensures that the next time we launch
a Web Content process, WebPageCreationParameters will have the correct
layer hosting mode, and is the usual pattern for such things.

12:12 PM Changeset in webkit [247226] by Ryan Haddad
  • 23 edits
    18 copies
    12 moves
    120 adds in trunk

Add test expectations and baselines for Catalina

Reviewed by Jonathan Bedard.


  • Scripts/webkitpy/port/mac.py:

(MacPort): Increment current version to 10.15 so that baseline search paths are correctly constructed for High Sierra.


  • platfrom/mac/*: Adding Catalina test expectations and baselines.
  • platform/mac-mojave/*: Migrating Mojave baselines from platform/mac directory.
12:09 PM Changeset in webkit [247225] by chris.reid@sony.com
  • 5 edits in trunk

Implement MappedFileData for Windows

Reviewed by Darin Adler.


Original patch by Fujii Hironori.

Add Windows implementations for MappedFileData constructor and destructor.

  • wtf/FileSystem.cpp:
  • wtf/FileSystem.h:
  • wtf/win/FileSystemWin.cpp:


  • TestWebKitAPI/PlatformWin.cmake:
12:08 PM Changeset in webkit [247224] by Alan Coon
  • 1 edit in branches/safari-607-branch/Source/JavaScriptCore/runtime/RegExpInlines.h

Unreviewed build fix. rdar://problem/51349045

12:03 PM Changeset in webkit [247223] by Wenson Hsieh
  • 7 edits
    2 adds in trunk

[iPadOS] Viewport pops while loading sohu.com, xinhuanet.com, and various other websites

Reviewed by Simon Fraser.


Currently, the shrink-to-fit-content heuristic added to scale desktop sites fit within the viewport on iPadOS
only runs during two intervals: after document load, and after page load. On very script-heavy websites, this
may cause a visible jump, as rendering may commence well before the document is finished parsing.

To mitigate this, we move the first opportunity for the shrink-to-fit heuristic from after document, to right
after the page transition has completed (before we unfreeze the layer tree).

  • WebProcess/WebCoreSupport/WebFrameLoaderClient.cpp:


  • WebProcess/WebPage/WebPage.cpp:

(WebKit::WebPage::didFinishDocumentLoad): Deleted.

Remove this didFinishDocumentLoad hook, which was only used to schedule the shrink-to-fit timer.

  • WebProcess/WebPage/WebPage.h:
  • WebProcess/WebPage/ios/WebPageIOS.mm:


m_mainFrame might be null after page transition completes, so we need to ensure immediatelyShrinkToFitContent is
robust in this scenario.


Adjusts an existing layout test, and introduces a new test to verify that the shrink-to-fit heuristic is given a
chance to run before document load. See WebKit ChangeLog for more details.

  • fast/viewport/ios/shrink-to-fit-content-before-document-load-expected.txt: Added.
  • fast/viewport/ios/shrink-to-fit-content-before-document-load.html: Added.
  • fast/viewport/ios/shrink-to-fit-content-temporary-overflow.html:

Tweak this layout test to add the temporarily overflowing element after document load.

11:59 AM Changeset in webkit [247222] by Wenson Hsieh
  • 5 edits
    2 adds in trunk

Unable to paste from Notes into Excel 365 spreadsheet

Reviewed by Chris Dumez.


When pasting into Microsoft Excel 365, the copied data is all inserted into a single cell, even when copying a
table. To understand why this happens, we first need to understand how Excel's logic for handling paste works.
When tapping on the "Paste" button, Excel performs and expects the following:

  1. Before triggering programmatic paste, move focus into a hidden contenteditable area specifically intended to

capture pasted content.

  1. Run a promise that resolves immediately; the promise callback restores focus to the originally focused

element prior to (1).

  1. Invoke programmatic paste using document.execCommand("Paste").
  2. The callback scheduled in step (2) then runs, restoring focus to the main editable element representing a

table cell.

However, what ends up happening is this:

Steps (1)-(3): same as before.

  1. We (WebKit) create a temporary Page for the purposes of sanitizing copied web content before exposing it to

the paste handler. This involves creating and loading a document; when this is finished, we call into
Document::finishedParsing which flushes the microtask queue.

  1. This causes us to immediately run the microtask enqueued in step (2), which restores focus to the previously

focused element (importantly, this is not the element that was focused in step (1)).

  1. The paste commences, and inserts the sanitized fragment into the originally focused element rather than the

content editable area intended to capture pasted content.

Excel's script then gets confused, and does not end up using their special paste logic to handle the paste. The
pasted content is instead just inserted as plain text in a cell. To address this, we simply prevent document
load in the Page for web content sanitization from triggering a microtask checkpoint; this allows any scheduled
main thread microtasks to be deferred until the next turn of the runloop.

Test: editing/pasteboard/paste-does-not-fire-promises-while-sanitizing-web-content.html

  • dom/Document.cpp:


Don't immediately dispatch microtasks when we finish document parsing, in the case where the page is intended
only for web content sanitization, since this may end up executing script in the original document. As explained
above, this causes compatibility issues when pasting in Excel.

  • editing/markup.cpp:


When creating a page for sanitizing web content, mark it as such.

  • page/Page.h:

Add a new flag to indicate that a Page is only intended for sanitizing web content.

(WebCore::Page::isForSanitizingWebContent const):


Add a test to verify that promises scheduled right before a programmatic paste resolve in the middle of the
paste, while creating a document for web content sanitization. See WebCore ChangeLog for more details.

  • editing/pasteboard/paste-does-not-fire-promises-while-sanitizing-web-content-expected.txt: Added.
  • editing/pasteboard/paste-does-not-fire-promises-while-sanitizing-web-content.html: Added.
11:55 AM Changeset in webkit [247221] by Alan Coon
  • 4 edits in branches/safari-607-branch/Source/WebCore

Cherry-pick r246808. rdar://problem/52505041

Add didBecomePrototype() calls to global context prototypes

Reviewed by Mark Lam.

This fixes some crashes related to asserting that all prototypes
have been marked as such in JSC from
https://trac.webkit.org/changeset/246801. It's ok to call
didBecomePrototype here as we setting up the world state right now
so we won't be having a bad time.

We don't automatically call didBecomePrototype() for
setPrototypeWithoutTransition because existing objects may already
have this structure so it seems more reasonable to be explicit

  • bindings/js/JSWindowProxy.cpp: (WebCore::JSWindowProxy::setWindow):
  • bindings/js/WorkerScriptController.cpp: (WebCore::WorkerScriptController::initScript):
  • worklets/WorkletScriptController.cpp: (WebCore::WorkletScriptController::initScriptWithSubclass):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@246808 268f45cc-cd09-0410-ab3c-d52691b4dbfc

11:44 AM Changeset in webkit [247220] by Konstantin Tokarev
  • 2 edits in trunk/Source/WebCore

Remove unused #include "ImageBufferData.h"

Reviewed by Michael Catanzaro.

  • html/canvas/ImageBitmapRenderingContext.h:
11:38 AM Changeset in webkit [247219] by Chris Dumez
  • 11 edits in trunk/Source/WebCore

Fix thread safety issue in Database::scheduleTransactionCallback()

Reviewed by Alex Christensen.

I am working on adding threading assertions to WeakPtr and found a potentially
unsafe call to makeWeakPtr() on a Document from Database::scheduleTransactionCallback()
via Document::postTask(), on a background database thread. Document is a main thread
object and we should therefore not be interacting with it from a background thread.

For clarity, this patch also switches the webdatabase code to use Document instead
of ScriptExecution as type since it is only exposed to Window contexts, not workers.

  • Modules/webdatabase/Database.cpp:


  • Modules/webdatabase/Database.h:


  • Modules/webdatabase/DatabaseContext.cpp:


  • Modules/webdatabase/DatabaseContext.h:
  • Modules/webdatabase/DatabaseManager.cpp:


  • Modules/webdatabase/DatabaseManager.h:
  • Modules/webdatabase/SQLStatement.cpp:


  • Modules/webdatabase/SQLTransaction.cpp:


  • inspector/InspectorInstrumentation.h:


  • inspector/agents/InspectorDatabaseAgent.cpp:


11:26 AM Changeset in webkit [247218] by Chris Dumez
  • 15 edits in trunk/Source

Add threading assertion to WTF::CompletionHandler

Reviewed by Alex Christensen.


Update some MessagePort-related code to use WTF::Function instead of WTF::CompletionHandler
since the callback is always called on the main thread, even when it was created on a
worker thread. Ideally, this code would be refactored so that the callback gets called on
the worker thread directly.

  • dom/messageports/MessagePortChannel.cpp:


  • dom/messageports/MessagePortChannel.h:
  • dom/messageports/MessagePortChannelProvider.h:
  • dom/messageports/MessagePortChannelProviderImpl.cpp:


  • dom/messageports/MessagePortChannelProviderImpl.h:
  • dom/messageports/MessagePortChannelRegistry.cpp:


  • dom/messageports/MessagePortChannelRegistry.h:


Update some MessagePort-related code to use WTF::Function instead of WTF::CompletionHandler
since the callback is always called on the main thread, even when it was created on a
worker thread. Ideally, this code would be refactored so that the callback gets called on
the worker thread directly.

  • UIProcess/UIMessagePortChannelProvider.cpp:


  • UIProcess/UIMessagePortChannelProvider.h:
  • WebProcess/WebCoreSupport/WebMessagePortChannelProvider.cpp:


  • WebProcess/WebCoreSupport/WebMessagePortChannelProvider.h:


Add threading assertion to WTF::CompletionHandler to try and catch common cases
of unsafe usage (completion handler constructed on one thread but called on

  • wtf/CompletionHandler.h:


11:25 AM WebKitGTK/2.24.x edited by Michael Catanzaro
11:19 AM Changeset in webkit [247217] by Alan Coon
  • 4 edits in branches/safari-608.1.34.0-branch/Source/WebKit

Cherry-pick r247192. rdar://problem/52563665

AX: CrashTracer: com.apple.WebKit.WebContent at WebKit: WebKit::WebSpeechSynthesisClient::speak

Reviewed by Ryosuke Niwa.

Ensure we don't access null observers in speech callbacks.

  • WebProcess/WebCoreSupport/WebSpeechSynthesisClient.cpp: (WebKit::WebSpeechSynthesisClient::voiceList): (WebKit::WebSpeechSynthesisClient::speak): (WebKit::WebSpeechSynthesisClient::pause): (WebKit::WebSpeechSynthesisClient::resume):
  • WebProcess/WebPage/WebPage.cpp: (WebKit::WebPage::speakingErrorOccurred): (WebKit::WebPage::boundaryEventOccurred): (WebKit::WebPage::voicesDidChange):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@247192 268f45cc-cd09-0410-ab3c-d52691b4dbfc

11:17 AM Changeset in webkit [247216] by Alan Coon
  • 7 edits in branches/safari-608.1.34.0-branch/Source


11:16 AM Changeset in webkit [247215] by cturner@igalia.com
  • 4 edits
    1 add in trunk

REGRESSION(r243197): [GStreamer] Web process hangs when scrolling twitter timeline which contains HLS videos

Reviewed by Xabier Rodriguez-Calvar.


Not covered, I have a test locally that would probably trigger the
deadlock if the network requests took a realistic amount of time,
but from a local webserver the window of time to hit this deadlock
is too narrow.

  • platform/graphics/gstreamer/WebKitWebSourceGStreamer.cpp:

(webkit_web_src_init): Make the websrc start asynchronously, this
allows the main thread to be free to complete resource loader
(webKitWebSrcCreate): Calling start() from the create() vfunc is a
recipe for deadlock, since BaseSrc holds the streaming lock during
seeks, and then calls create(). In these cases, we do not want to
notify async-completion, since we've already completed from the
necessarily preceeding start() vfunc, and calling it again would
require the stream-lock and deadlock us.
(webKitWebSrcStart): Refactor to use webKitWebSrcMakeRequest, but
ensuring that we do perform an async-complete notification.
(webKitWebSrcMakeRequest): What Start() used to be, but now can be
toggled when to notify of async-completion. Start() no longer
blocks, since the return value of initiating a resource loader is
of no interest to the callers.
(webKitWebSrcCloseSession): Similarly to Start(), we do not need
to wait for the completion of cancelled net requests.


On shutdown we can easily deadlock the web process if we don't
ensure all network operations are completed before comitting state
changes. In HLS, make sure the network operations are cancelled,
and also prevent hlsdemux's retry logic from scuppering our

  • gstreamer/jhbuild.modules: Include the patch.
  • gstreamer/patches/gst-plugins-bad-do-not-retry-downloads-during-shutdown.patch: Added.
11:14 AM Changeset in webkit [247214] by Alan Coon
  • 1 copy in branches/safari-608.1.34.0-branch

New branch.

11:14 AM Changeset in webkit [247213] by Chris Dumez
  • 5 edits in trunk

Unable to play videos on xfinity.com/stream on macOS Catalina

Reviewed by Alex Christensen.


Enable invalidate Fetch signal quirk for all sites as the blacklist is getting too large and
we are finding too much breakage too late in the release cycle.

  • Modules/fetch/FetchRequest.cpp:


  • page/Quirks.cpp:

(WebCore::Quirks::shouldIgnoreInvalidSignal const):


Rebaseline existing test.

  • http/wpt/fetch/request-abort-expected.txt:
11:00 AM Changeset in webkit [247212] by graouts@webkit.org
  • 6 edits
    4 adds in trunk

[Pointer Events] "touch-action: none" does not prevent double-tap-to-zoom

Reviewed by Wenson Hsieh.


We add a new WKTouchActionGestureRecognizerDelegate method to check whether a gesture recognizer may lead to
zooming the page as a result of double-tapping, which can be caused by two different gesture recognizers
managed by WKContentViewInteraction. If that delegate method returns true and we have "touch-action: none"
set for this touch, we cause those gesture recognizers to fail and prevent double-tap-to-zoom behavior.

  • UIProcess/ios/WKContentViewInteraction.mm:

(-[WKContentView gestureRecognizerMayDoubleTapToZoomWebView:]):

  • UIProcess/ios/WKTouchActionGestureRecognizer.h:
  • UIProcess/ios/WKTouchActionGestureRecognizer.mm:

(-[WKTouchActionGestureRecognizer canPreventGestureRecognizer:]):


Add two new tests that check that setting "touch-action: none" on an element does not allow double-tap-to-zoom
and that "touch-action: manipulation" allows it.

  • pointerevents/ios/touch-action-manipulation-double-tap-to-zoom-expected.txt: Added.
  • pointerevents/ios/touch-action-manipulation-double-tap-to-zoom.html: Added.
  • pointerevents/ios/touch-action-none-double-tap-to-zoom-expected.txt: Added.
  • pointerevents/ios/touch-action-none-double-tap-to-zoom.html: Added.
  • pointerevents/utils.js:


10:46 AM Changeset in webkit [247211] by youenn@apple.com
  • 8 edits in trunk/Source/WebCore

MediaStreamTrackPrivate should always call readyStateChanged on the main thread

Reviewed by Eric Carlson.

MediaStreamTrackPrivate is sometimes calling readyStateChanged in a
background thread inside its audioSamplesAvailable method.
Instead of doing that, we hop to the main thread to call readyStateChanged.
Once the call is made in the main thread, MediaStreamTrackPrivate will
send the audio samples to its observers.

To make mock audio source closer to real capture audio sources,
audioSamplesAvailable is called on a background thread.
RealtimeMediaSource is updated to always be destroyed in the main
run loop since it is WebKit2 only.

Covered by existing tests and making sure the mock audio source calls
the audioSamplesAvailable method on a background thread.

  • platform/mediastream/MediaStreamTrackPrivate.cpp:


  • platform/mediastream/MediaStreamTrackPrivate.h:
  • platform/mediastream/RealtimeMediaSource.cpp:

scheduleDeferredTask may be called from a background thread.
It is thus safer to ref the source instead of creating a weak pointer.

  • platform/mediastream/RealtimeMediaSource.h:
  • platform/mediastream/mac/MockRealtimeAudioSourceMac.mm:


  • platform/mock/MockRealtimeAudioSource.cpp:


  • platform/mock/MockRealtimeAudioSource.h:
9:51 AM Changeset in webkit [247210] by pvollan@apple.com
  • 2 edits in trunk/Source/WebKit

[iOS] WebContent processes should not globally allow mach lookup to the AX server.

Reviewed by Chris Fleizach.

Instead, a mach extension for this service should be issued to the WebContent process. This has already been implemented.

  • Resources/SandboxProfiles/ios/com.apple.WebKit.WebContent.sb:
9:40 AM Changeset in webkit [247209] by Claudio Saavedra
  • 2 edits in trunk/Source/WebKit

REGRESSION(r246671): [WPE][GTK] Crash in NetworkProcess since the DNS cache landed

Reviewed by Michael Catanzaro.

The GError coming from the wrapped resolver shouldn't be freed,
but passed onto the caller.

  • NetworkProcess/glib/WebKitCachedResolver.cpp:


9:25 AM Changeset in webkit [247208] by youenn@apple.com
  • 3 edits in trunk/Source/WebCore

Register a MediaStreamTrack as media producer only if it is a capture track

Reviewed by Eric Carlson.

MediaStreamTrack registration as a media producer is only useful for capture tracks.
Otherwise, the audio/video playing state is computed through HTMLMediaElement.
Do not register MediaStreamTrack as a media producer if it is not a capture track.

Set the muted state of the track before registering it as an audio
producer. Otherwise, it might create unnecessary small change of playing state.
Covered by existing tests.

  • Modules/mediastream/MediaStreamTrack.cpp:


  • Modules/mediastream/MediaStreamTrack.h:
4:47 AM Changeset in webkit [247207] by cturner@igalia.com
  • 3 edits in trunk/LayoutTests

[GStreamer] media/video-volume.html broken after switching from cubic to linear scaling

Reviewed by Xabier Rodriguez-Calvar.

PulseAudio has a conversion process from volume's in
double-precision to uint32_t volumes. Depending on the environment
can introduce rounding errors. Be more lenient in our comparison

  • media/video-volume-expected.txt: Update baseline
  • media/video-volume.html: Compare volume values within a

reasonable tolerance.

4:26 AM Changeset in webkit [247206] by graouts@webkit.org
  • 3 edits in trunk/Source/WebKit

[Pointer Events] touch-action should affect the behavior of pinch-to-zoom to show tabs in Safari

Reviewed by Dean Jackson.

There are other UIPinchGestureRecognizer objects that may lead to pinch-to-zoom behavior, for instance pinching
out to show open tabs in Safari on iOS. We add a new WKUIDelegatePrivate method that will allow Safari to indicate
that a UIPinchGestureRecognizer considered for prevention would lead to pinch-to-zoom behavior.

  • UIProcess/API/Cocoa/WKUIDelegatePrivate.h:
  • UIProcess/ios/WKContentViewInteraction.mm:

(-[WKContentView gestureRecognizerMayPinchToZoomWebView:]):

4:22 AM Changeset in webkit [247205] by graouts@webkit.org
  • 20 edits
    1 copy in trunk

[Pointer Events] Enable only on the most recent version of the supported iOS family

Reviewed by Dean Jackson.


  • Configurations/FeatureDefines.xcconfig:


  • Configurations/FeatureDefines.xcconfig:


  • Configurations/FeatureDefines.xcconfig:


  • Configurations/FeatureDefines.xcconfig:
  • UIProcess/ios/WKContentViewInteraction.mm:

(-[WKContentView cancelPointersForGestureRecognizer:]):
(-[WKContentView activeTouchIdentifierForGestureRecognizer:]):
(-[WKContentView touchActionActiveTouches]):

  • UIProcess/ios/WKSyntheticTapGestureRecognizer.m:

(-[WKSyntheticTapGestureRecognizer touchesEnded:withEvent:]):

  • UIProcess/ios/WKTouchActionGestureRecognizer.h:
  • UIProcess/ios/WKTouchActionGestureRecognizer.mm:

(-[WKTouchActionGestureRecognizer canPreventGestureRecognizer:]):


  • Configurations/FeatureDefines.xcconfig:


that was added in iOS 13 that it's checking for is really required for all parts of the implementation of Pointer
Events, including dispatch of the "pointercancel" event and support for the "touch-action" CSS property.

  • wtf/Platform.h:


  • TestWebKitAPI/Configurations/FeatureDefines.xcconfig:


  • js/dom/navigator-maxtouchpoints-expected.txt: Since ENABLE_POINTER_EVENTS is now disabled by default, navigator.maxTouchPoints

is "undefined" in thed default case.

  • platform/mac/js/dom/navigator-maxtouchpoints-expected.txt: But 0 on macOS.
3:08 AM Changeset in webkit [247204] by Philippe Normand
  • 2 edits in trunk/Source/WebCore

[GStreamer] The CREATE_TRACK macro is messed up

Reviewed by Xabier Rodriguez-Calvar.

  • platform/graphics/gstreamer/MediaPlayerPrivateGStreamer.cpp:

(WebCore::MediaPlayerPrivateGStreamer::updateTracks): Fix the
CREATE_TRACK macro for !VIDEO_TRACK builds.

1:17 AM Changeset in webkit [247203] by Michael Catanzaro
  • 2 edits in trunk/Source/WebKit

[GTK] ViewGestureController constructor leaves m_direction uninitialized

Reviewed by Carlos Garcia Campos.

This is harmless because it should never be read until after it's initialized, but still not
robust. Fix it by assuming Back as the default value until the first swipe begins.

  • UIProcess/ViewGestureController.h:

Jul 7, 2019:

10:07 PM Changeset in webkit [247202] by bshafiei@apple.com
  • 5 edits in branches/safari-607-branch/Source/JavaScriptCore

Cherry-pick r246801. rdar://problem/52505041

Structure::create should call didBecomePrototype()

Reviewed by Filip Pizlo.

Structure::create should also assert that the indexing type makes sense
for the prototype being used.

  • runtime/JSObject.h:
  • runtime/Structure.cpp: (JSC::Structure::isValidPrototype): (JSC::Structure::changePrototypeTransition):
  • runtime/Structure.h: (JSC::Structure::create): Deleted.
  • runtime/StructureInlines.h: (JSC::Structure::create): (JSC::Structure::setPrototypeWithoutTransition):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@246801 268f45cc-cd09-0410-ab3c-d52691b4dbfc

10:03 PM Changeset in webkit [247201] by bshafiei@apple.com
  • 7 edits in branches/safari-607-branch/Source


7:52 AM Changeset in webkit [247200] by Alan Bujtas
  • 7 edits
    4 adds in trunk

[ContentChangeObserver] Difficult to control videos on iqiyi.com as the actions are mouse hover

Reviewed by Simon Fraser.


Decouple isVisuallyHidden and isConsideredVisible. Just because an element is not visually hidden (1px wide content)
it is not necessarily qualified to be visible in the context of hover heuristic (e.g. iqiyi.com brings up a 1px wide
clickable element when hovering over the scrubber. This element is clearly not designed to be actionable.)

Tests: fast/events/touch/ios/content-observation/tap-on-1px-height-content.html


  • dom/Node.cpp:


  • page/ios/ContentChangeObserver.cpp:

(WebCore::ContentChangeObserver::isConsideredHidden): Deleted.

  • page/ios/ContentChangeObserver.h:


  • WebProcess/WebPage/ios/WebPageIOS.mm:



  • fast/events/touch/ios/content-observation/tap-on-1px-height-content-expected.txt: Added.
  • fast/events/touch/ios/content-observation/tap-on-1px-height-content.html: Added.
  • fast/events/touch/ios/content-observation/tap-on-1px-width-content-expected.txt: Added.
  • fast/events/touch/ios/content-observation/tap-on-1px-width-content.html: Added.
6:49 AM Changeset in webkit [247199] by Alan Bujtas
  • 3 edits in trunk/Source/WebCore

[LFC][IFC] Introduce splitPosition to LineLayout

Reviewed by Antti Koivisto.

This is in preparation for breaking runs at line end.

  • layout/inlineformatting/InlineFormattingContext.h:
  • layout/inlineformatting/InlineFormattingContextLineLayout.cpp:

(WebCore::Layout::InlineFormattingContext::LineLayout::placeInlineItems const):
(WebCore::Layout::InlineFormattingContext::LineLayout::layout const):
(WebCore::Layout::InlineFormattingContext::LineLayout::computedIntrinsicWidth const):

6:47 AM Changeset in webkit [247198] by Alan Bujtas
  • 6 edits in trunk/Source/WebCore

[LFC] Fix formatting context root for inflow positioned inline containers

Reviewed by Antti Koivisto.

Relatively positioned (inflow) inline container lives in the formatting context where its parent lives unless
the parent establishes a formatting context. This is slightly different from the usual behavior which is containing block driven.

div id=outer style="position: absolute">><div id=inner><span style="position: relative">content</span></div></div>

While the relatively positioned inline container (span) is placed relative to its containing block "outer", it lives in the inline formatting context established by "inner".

  • layout/FormattingContext.cpp:

(WebCore::Layout::FormattingContext::layoutOutOfFlowDescendants const):

  • layout/layouttree/LayoutBox.cpp:

(WebCore::Layout::Box::formattingContextRoot const):

  • layout/layouttree/LayoutBox.h:
  • layout/layouttree/LayoutInlineContainer.cpp:

(WebCore::Layout::InlineContainer::formattingContextRoot const):

  • layout/layouttree/LayoutInlineContainer.h:
1:12 AM Changeset in webkit [247197] by graouts@webkit.org
  • 5 edits
    2 adds in trunk/Source/WebKit

[Pointer Events] Use a gesture recognizer to prevent pinch-to-zoom behavior

Reviewed by Dean Jackson.

We used to set the "enabled" property on the main UIScrollView's UIPinchGestureRecognizer to disable pinch-to-zoom
behavior based on the "touch-action" CSS property. This was sub-optimal since this gesture recognizer can be publicly
accessible by third-party developers, but also because it set state instead of a run-time solution.

We now introduce a new WKTouchActionGestureRecognizer dedicated to preventing clearly identified gesture recognizers
from being recognized to disable some built-in behavior. As a new touch starts, we record the computed touch-action
property for the given touch identifier on the WKTouchActionGestureRecognizer, then WKTouchActionGestureRecognizer
determines, by implementing -[canPreventGestureRecognizer:], whether the possibly-prevented gesture recognizer
is known to lead to a pinch-to-zoom behavior and whether it contains a touch that should prevent it.

To support the WKTouchActionGestureRecognizer, WKContentViewInteraction implements the new WKTouchActionGestureRecognizerDelegate
protocol to indicate whether the possibly-prevented gesture recognizer is the main UIScrollView's UIPinchGestureRecognizer.

No new test since this changes the implementation of existing behavior.

  • SourcesCocoa.txt:
  • UIProcess/ios/WKContentViewInteraction.h:
  • UIProcess/ios/WKContentViewInteraction.mm:

(-[WKContentView setupInteraction]):
(-[WKContentView cleanupInteraction]):
(-[WKContentView _removeDefaultGestureRecognizers]):
(-[WKContentView _addDefaultGestureRecognizers]):
(-[WKContentView _handleTouchActionsForTouchEvent:]):
(-[WKContentView gestureRecognizerMayPinchToZoomWebView:]):
(-[WKContentView touchActionActiveTouches]):

  • UIProcess/ios/WKTouchActionGestureRecognizer.h: Added.
  • UIProcess/ios/WKTouchActionGestureRecognizer.mm: Added.

(-[WKTouchActionGestureRecognizer initWithTouchActionDelegate:]):
(-[WKTouchActionGestureRecognizer setTouchActions:forTouchIdentifier:]):
(-[WKTouchActionGestureRecognizer clearTouchActionsForTouchIdentifier:]):
(-[WKTouchActionGestureRecognizer touchesBegan:withEvent:]):
(-[WKTouchActionGestureRecognizer touchesMoved:withEvent:]):
(-[WKTouchActionGestureRecognizer touchesEnded:withEvent:]):
(-[WKTouchActionGestureRecognizer touchesCancelled:withEvent:]):
(-[WKTouchActionGestureRecognizer _updateState]):
(-[WKTouchActionGestureRecognizer canBePreventedByGestureRecognizer:]):
(-[WKTouchActionGestureRecognizer canPreventGestureRecognizer:]):

  • WebKit.xcodeproj/project.pbxproj:

Jul 6, 2019:

7:09 PM Changeset in webkit [247196] by Nikita Vasilyev
  • 2 edits in trunk/Source/WebInspectorUI

Web Inspector: Styles: unbalanced quotes and parenthesis aren't displayed as property closed after editing values

Reviewed by Devin Rousso.

} gets added by WI.tokenizeCSSValue (called by SpreadsheetStyleProperty.prototype._renderValue)
when it encounters unbalanced quotes. Fix unbalanced quotes by re-rendering the value from the model,
not the DOM content.

  • UserInterface/Views/SpreadsheetStyleProperty.js:

For consistency, render property name from the model as well.

5:20 PM Changeset in webkit [247195] by Simon Fraser
  • 5 edits in trunk

Long hang when loading a cnn.com page on iOS

Reviewed by Zalan Bujtas.

Loading https://edition.cnn.com/travel/article/brussels-airlines-flight-to-nowhere/index.html in the iOS 13 sim
results in a long hang under OverlapMapContainer::append(). We were creating pathological clipping scopes with
thousands of entries, because OverlapMapContainer::mergeClippingScopesRecursive() had a logic error where
it added 'sourceScope' to the child instead of 'sourceChildScope'. Add a new assertion to detect that case.

I wasn't able to create a testcase that caused a hang, but a number of existing tests would have
hit the assertion.

  • rendering/LayerOverlapMap.cpp:


  • rendering/RenderLayerCompositor.cpp:

(WebCore::RenderLayerCompositor::flushPendingLayerChanges): Drive-by fix: m_overflowControlsHostLayer is null on iOS, so use rootGraphicsLayer().


Add code to load a page by default in MobileMiniBrowser so it's easy to hack it
to load a test page of your choice.

  • MobileMiniBrowser/MobileMiniBrowserFramework/WebViewController.m:

(-[WebViewController viewDidLoad]):

6:34 AM Changeset in webkit [247194] by msaboff@apple.com
  • 5 edits
    1 add in trunk

switch(String) needs to check for exceptions when resolving the string

Reviewed by Mark Lam.


New tests.

  • stress/switch-string-oom.js: Added.



Added exception checks for resolved Strings in switch processing for all tiers.

  • dfg/DFGOperations.cpp:
  • jit/JITOperations.cpp:
  • llint/LLIntSlowPaths.cpp:


4:03 AM Changeset in webkit [247193] by commit-queue@webkit.org
  • 78 edits
    64 adds in trunk/LayoutTests

Import css/cssom-view testcases from WPT.

Patch by Cathie Chen <cathiechen> on 2019-07-06
Reviewed by Frédéric Wang.


  • resources/import-expectations.json:
  • web-platform-tests/css/cssom-view/CaretPosition-001.html:
  • web-platform-tests/css/cssom-view/DOMRectList.html:
  • web-platform-tests/css/cssom-view/GetBoundingRect.html:
  • web-platform-tests/css/cssom-view/HTMLBody-ScrollArea_quirksmode.html:
  • web-platform-tests/css/cssom-view/MediaQueryList-001.html:
  • web-platform-tests/css/cssom-view/MediaQueryList-with-empty-string.html:
  • web-platform-tests/css/cssom-view/Screen-pixelDepth-Screen-colorDepth001.html:
  • web-platform-tests/css/cssom-view/cssom-getBoundingClientRect-001.html:
  • web-platform-tests/css/cssom-view/cssom-getBoundingClientRect-002.html:
  • web-platform-tests/css/cssom-view/cssom-getBoxQuads-001.html:
  • web-platform-tests/css/cssom-view/cssom-getClientRects-002.html:
  • web-platform-tests/css/cssom-view/cssom-getClientRects.html:
  • web-platform-tests/css/cssom-view/cssom-view-img-attributes-001.html:
  • web-platform-tests/css/cssom-view/cssom-view-window-screen-interface.html:
  • web-platform-tests/css/cssom-view/dom-element-scroll.html:
  • web-platform-tests/css/cssom-view/elementFromPoint-001.html:
  • web-platform-tests/css/cssom-view/elementFromPoint-002.html:
  • web-platform-tests/css/cssom-view/elementFromPoint-003.html:
  • web-platform-tests/css/cssom-view/elementFromPoint-dynamic-anon-box.html:
  • web-platform-tests/css/cssom-view/elementFromPoint-mixed-font-sizes.html:
  • web-platform-tests/css/cssom-view/elementFromPoint-parameters.html:
  • web-platform-tests/css/cssom-view/elementFromPoint-subpixel-expected.txt: Added.
  • web-platform-tests/css/cssom-view/elementFromPoint-subpixel.html: Added.
  • web-platform-tests/css/cssom-view/elementFromPoint.html:
  • web-platform-tests/css/cssom-view/elementFromPosition.html:
  • web-platform-tests/css/cssom-view/elementScroll-002.html:
  • web-platform-tests/css/cssom-view/elementScroll.html:
  • web-platform-tests/css/cssom-view/elementsFromPoint-iframes.html:
  • web-platform-tests/css/cssom-view/elementsFromPoint-inline-htb-ltr-expected.txt: Added.
  • web-platform-tests/css/cssom-view/elementsFromPoint-inline-htb-ltr.html: Added.
  • web-platform-tests/css/cssom-view/elementsFromPoint-inline-htb-rtl-expected.txt: Added.
  • web-platform-tests/css/cssom-view/elementsFromPoint-inline-htb-rtl.html: Added.
  • web-platform-tests/css/cssom-view/elementsFromPoint-inline-vlr-ltr-expected.txt: Added.
  • web-platform-tests/css/cssom-view/elementsFromPoint-inline-vlr-ltr.html: Added.
  • web-platform-tests/css/cssom-view/elementsFromPoint-inline-vlr-rtl-expected.txt: Added.
  • web-platform-tests/css/cssom-view/elementsFromPoint-inline-vlr-rtl.html: Added.
  • web-platform-tests/css/cssom-view/elementsFromPoint-inline-vrl-ltr-expected.txt: Added.
  • web-platform-tests/css/cssom-view/elementsFromPoint-inline-vrl-ltr.html: Added.
  • web-platform-tests/css/cssom-view/elementsFromPoint-inline-vrl-rtl-expected.txt: Added.
  • web-platform-tests/css/cssom-view/elementsFromPoint-inline-vrl-rtl.html: Added.
  • web-platform-tests/css/cssom-view/elementsFromPoint-invalid-cases.html:
  • web-platform-tests/css/cssom-view/elementsFromPoint-shadowroot.html:
  • web-platform-tests/css/cssom-view/elementsFromPoint-simple.html:
  • web-platform-tests/css/cssom-view/elementsFromPoint-svg-text.html:
  • web-platform-tests/css/cssom-view/elementsFromPoint-svg.html:
  • web-platform-tests/css/cssom-view/elementsFromPoint-table.html:
  • web-platform-tests/css/cssom-view/elementsFromPoint.html:
  • web-platform-tests/css/cssom-view/getClientRects-br-htb-ltr-expected.txt: Added.
  • web-platform-tests/css/cssom-view/getClientRects-br-htb-ltr.html: Added.
  • web-platform-tests/css/cssom-view/getClientRects-br-htb-rtl-expected.txt: Added.
  • web-platform-tests/css/cssom-view/getClientRects-br-htb-rtl.html: Added.
  • web-platform-tests/css/cssom-view/getClientRects-br-vlr-ltr-expected.txt: Added.
  • web-platform-tests/css/cssom-view/getClientRects-br-vlr-ltr.html: Added.
  • web-platform-tests/css/cssom-view/getClientRects-br-vlr-rtl-expected.txt: Added.
  • web-platform-tests/css/cssom-view/getClientRects-br-vlr-rtl.html: Added.
  • web-platform-tests/css/cssom-view/getClientRects-br-vrl-ltr-expected.txt: Added.
  • web-platform-tests/css/cssom-view/getClientRects-br-vrl-ltr.html: Added.
  • web-platform-tests/css/cssom-view/getClientRects-br-vrl-rtl-expected.txt: Added.
  • web-platform-tests/css/cssom-view/getClientRects-br-vrl-rtl.html: Added.
  • web-platform-tests/css/cssom-view/getClientRects-inline-expected.html: Added.
  • web-platform-tests/css/cssom-view/getClientRects-inline.html: Added.
  • web-platform-tests/css/cssom-view/historical.html:
  • web-platform-tests/css/cssom-view/htmlelement-offset-width-001.html:
  • web-platform-tests/css/cssom-view/inheritance-expected.txt: Added.
  • web-platform-tests/css/cssom-view/inheritance.html: Added.
  • web-platform-tests/css/cssom-view/interfaces.html:
  • web-platform-tests/css/cssom-view/long_scroll_composited-expected.html: Added.
  • web-platform-tests/css/cssom-view/long_scroll_composited.html: Added.
  • web-platform-tests/css/cssom-view/matchMedia-display-none-iframe-expected.txt: Added.
  • web-platform-tests/css/cssom-view/matchMedia-display-none-iframe.html: Added.
  • web-platform-tests/css/cssom-view/matchMedia.xht:
  • web-platform-tests/css/cssom-view/matchMediaAddListener.html:
  • web-platform-tests/css/cssom-view/mouseEvent.html:
  • web-platform-tests/css/cssom-view/negativeMargins.html:
  • web-platform-tests/css/cssom-view/offsetParent_element_test.html:
  • web-platform-tests/css/cssom-view/offsetTopLeft-empty-inline-expected.txt: Added.
  • web-platform-tests/css/cssom-view/offsetTopLeft-empty-inline-offset-expected.txt: Added.
  • web-platform-tests/css/cssom-view/offsetTopLeft-empty-inline-offset.html: Added.
  • web-platform-tests/css/cssom-view/offsetTopLeft-empty-inline.html: Added.
  • web-platform-tests/css/cssom-view/offsetTopLeft-inline-expected.html: Added.
  • web-platform-tests/css/cssom-view/offsetTopLeft-inline.html: Added.
  • web-platform-tests/css/cssom-view/offsetTopLeft-leading-space-inline-expected.txt: Added.
  • web-platform-tests/css/cssom-view/offsetTopLeft-leading-space-inline.html: Added.
  • web-platform-tests/css/cssom-view/offsetTopLeft-trailing-space-inline-expected.txt: Added.
  • web-platform-tests/css/cssom-view/offsetTopLeft-trailing-space-inline.html: Added.
  • web-platform-tests/css/cssom-view/offsetTopLeftInScrollableParent.html:
  • web-platform-tests/css/cssom-view/outer-svg-expected.txt: Added.
  • web-platform-tests/css/cssom-view/outer-svg.html: Added.
  • web-platform-tests/css/cssom-view/resize-event-on-initial-layout-expected.txt: Added.
  • web-platform-tests/css/cssom-view/resize-event-on-initial-layout.html: Added.
  • web-platform-tests/css/cssom-view/screenLeftTop-expected.txt: Added.
  • web-platform-tests/css/cssom-view/screenLeftTop.html: Added.
  • web-platform-tests/css/cssom-view/scroll-behavior-default-css.html:
  • web-platform-tests/css/cssom-view/scroll-behavior-element.html:
  • web-platform-tests/css/cssom-view/scroll-behavior-main-frame-root.html:
  • web-platform-tests/css/cssom-view/scroll-behavior-main-frame-window.html:
  • web-platform-tests/css/cssom-view/scroll-behavior-scrollintoview-nested.html:
  • web-platform-tests/css/cssom-view/scroll-behavior-smooth-positions.html:
  • web-platform-tests/css/cssom-view/scroll-behavior-smooth.html:
  • web-platform-tests/css/cssom-view/scroll-behavior-subframe-root.html:
  • web-platform-tests/css/cssom-view/scroll-behavior-subframe-window.html:
  • web-platform-tests/css/cssom-view/scroll-no-layout-box.html:
  • web-platform-tests/css/cssom-view/scrollIntoView-horizontal-tb-writing-mode-and-rtl-direction-expected.txt: Added.
  • web-platform-tests/css/cssom-view/scrollIntoView-horizontal-tb-writing-mode-and-rtl-direction.html: Added.
  • web-platform-tests/css/cssom-view/scrollIntoView-horizontal-tb-writing-mode-expected.txt: Added.
  • web-platform-tests/css/cssom-view/scrollIntoView-horizontal-tb-writing-mode.html: Added.
  • web-platform-tests/css/cssom-view/scrollIntoView-scrollMargin.html:
  • web-platform-tests/css/cssom-view/scrollIntoView-scrollPadding.html:
  • web-platform-tests/css/cssom-view/scrollIntoView-shadow.html:
  • web-platform-tests/css/cssom-view/scrollIntoView-smooth.html:
  • web-platform-tests/css/cssom-view/scrollIntoView-svg-shape-expected.txt: Added.
  • web-platform-tests/css/cssom-view/scrollIntoView-svg-shape.html: Added.
  • web-platform-tests/css/cssom-view/scrollIntoView-vertical-lr-writing-mode-and-rtl-direction-expected.txt: Added.
  • web-platform-tests/css/cssom-view/scrollIntoView-vertical-lr-writing-mode-and-rtl-direction.html: Added.
  • web-platform-tests/css/cssom-view/scrollIntoView-vertical-lr-writing-mode-expected.txt: Added.
  • web-platform-tests/css/cssom-view/scrollIntoView-vertical-lr-writing-mode.html: Added.
  • web-platform-tests/css/cssom-view/scrollIntoView-vertical-rl-writing-mode-expected.txt:
  • web-platform-tests/css/cssom-view/scrollIntoView-vertical-rl-writing-mode.html:
  • web-platform-tests/css/cssom-view/scrollWidthHeight.xht:
  • web-platform-tests/css/cssom-view/scrollWidthHeightWhenNotScrollable.xht:
  • web-platform-tests/css/cssom-view/scrolling-no-browsing-context.html:
  • web-platform-tests/css/cssom-view/scrolling-quirks-vs-nonquirks.html:
  • web-platform-tests/css/cssom-view/scrollingElement.html:
  • web-platform-tests/css/cssom-view/scrollintoview.html:
  • web-platform-tests/css/cssom-view/table-client-props.html:
  • web-platform-tests/css/cssom-view/table-offset-props.html:
  • web-platform-tests/css/cssom-view/table-scroll-props.html:
  • web-platform-tests/css/cssom-view/ttwf-js-cssomview-getclientrects-length.html:
  • web-platform-tests/css/cssom-view/w3c-import.log:
  • web-platform-tests/css/cssom-view/window-screen-height-immutable.html:
  • web-platform-tests/css/cssom-view/window-screen-height.html:
  • web-platform-tests/css/cssom-view/window-screen-width-immutable.html:
  • web-platform-tests/css/cssom-view/window-screen-width.html:


  • TestExpectations:
  • platform/ios-simulator-wk2/imported/w3c/web-platform-tests/css/cssom-view/offsetTopLeft-empty-inline-expected.txt: Added.
  • platform/ios-simulator-wk2/imported/w3c/web-platform-tests/css/cssom-view/offsetTopLeft-leading-space-inline-expected.txt: Added.
  • platform/ios-simulator-wk2/imported/w3c/web-platform-tests/css/cssom-view/scrollIntoView-horizontal-tb-writing-mode-and-rtl-direction-expected.txt: Added.
  • tests-options.json:
1:14 AM Changeset in webkit [247192] by Chris Fleizach
  • 4 edits in trunk/Source/WebKit

AX: CrashTracer: com.apple.WebKit.WebContent at WebKit: WebKit::WebSpeechSynthesisClient::speak

Reviewed by Ryosuke Niwa.

Ensure we don't access null observers in speech callbacks.

  • WebProcess/WebCoreSupport/WebSpeechSynthesisClient.cpp:


  • WebProcess/WebPage/WebPage.cpp:


Jul 5, 2019:

10:37 PM Changeset in webkit [247191] by youenn@apple.com
  • 2 edits in trunk/Source/WebCore

Carvana.com needs the fetch AbortSignal quirk

Reviewed by Chris Dumez.

Manually tested.

  • page/Quirks.cpp:

(WebCore::Quirks::shouldIgnoreInvalidSignal const):

5:33 PM Changeset in webkit [247190] by Chris Dumez
  • 2 edits in trunk/Source/WebKit

Fix thread safety bug in WebResourceLoadStatisticsTelemetry::calculateAndSubmit()

Reviewed by Youenn Fablet.

The resourceLoadStatisticsStore is an object that is created / used / destroyed on the background
queue. It is therefore not safe to create a WeakPtr for it and use that WeakPtr on the main thread.

  • NetworkProcess/Classifier/WebResourceLoadStatisticsTelemetry.cpp:


5:33 PM Changeset in webkit [247189] by rmorisset@apple.com
  • 2 edits in trunk/Source/WebCore

[WHLSL] The checker does not need to keep a separate m_typeAnnotations map

Reviewed by Myles C. Maxfield.

Nothing is ever deleted from it, and all of its contents are eventually copied into the expressions.
We might as well put the type annotations directly into the expressions in the first place.
It is about a 5% speedup of the checker.

No new test as there is no inteded functional change.

  • Modules/webgpu/WHLSL/WHLSLChecker.cpp:


4:46 PM Changeset in webkit [247188] by Michael Catanzaro
  • 2 edits in trunk/Source/WebKit

Unreviewed, fix -Wswitch warning

We don't support HTTPCookieAcceptPolicy::ExclusivelyFromMainDocumentDomain.

Fun fact, soup's OnlyFromMainDocumentDomain functions the same as Apple's
ExclusivelyFromMainDocumentDomain. These should be renamed....

  • NetworkProcess/Cookies/soup/WebCookieManagerSoup.cpp:


4:36 PM Changeset in webkit [247187] by Simon Fraser
  • 3 edits
    2 adds in trunk

Trigger a compositing update when video element is changing

Reviewed by Simon Fraser.


Test: compositing/video/video-update-rendering.html

  • rendering/RenderLayerBacking.cpp:



  • compositing/video/video-update-rendering-expected.txt: Added.
  • compositing/video/video-update-rendering.html: Added.
4:35 PM Changeset in webkit [247186] by Michael Catanzaro
  • 2 edits in trunk/Source/WebCore

Unreviewed, fix -Wmisleading-indentation warning introduced in r246764

  • contentextensions/ContentExtensionsBackend.cpp:


4:34 PM Changeset in webkit [247185] by dino@apple.com
  • 2 edits in trunk/Source/WebKit

[iOS] Context Menu - Add to Reading List does nothing

Reviewed by Anders Carlsson.

We were holding weak references to the _WKElementAction in
the handler provided to UIAction. By the time the handler was
called, the object had been deallocated.

  • UIProcess/API/Cocoa/_WKElementAction.mm:

(-[_WKElementAction uiActionForElementInfo:]): Use strong references in handler.

4:19 PM Changeset in webkit [247184] by timothy@apple.com
  • 6 edits in trunk/Source/WebCore

Mail's use of punchOutWhiteBackgroundsInDarkMode not working on iOS.

Reviewed by Simon Fraser.

Tests: css3/color-filters/punch-out-white-backgrounds.html

  • dom/Document.cpp:

(WebCore::Document::compositeOperatorForBackgroundColor const): Added.
Use CompositeDestinationOut for transparent frames, otherwise CompositeDestinationIn.

  • dom/Document.h:
  • rendering/InlineFlowBox.cpp:

(WebCore::InlineFlowBox::paintBoxDecorations): Use compositeOperatorForBackgroundColor.

  • rendering/RenderBox.cpp:

(WebCore::RenderBox::paintRootBoxFillLayers): Ditto.
(WebCore::RenderBox::paintBackground): Ditto.

  • rendering/RenderTableCell.cpp:

(WebCore::RenderTableCell::paintBackgroundsBehindCell): Ditto.

4:05 PM Changeset in webkit [247183] by mark.lam@apple.com
  • 3 edits
    1 add in trunk

ArgumentsEliminationPhase::eliminateCandidatesThatInterfere() should not decrement nodeIndex pass zero.

Reviewed by Filip Pizlo.


  • stress/ArgumentsEliminationPhase-eliminateCandidatesThatEscape-should-not-decrement-nodeIndex-pass-zero.js: Added.


  • dfg/DFGArgumentsEliminationPhase.cpp:
3:24 PM Changeset in webkit [247182] by jer.noble@apple.com
  • 2 edits in trunk/Source/WebCore

Revert change to block playback when process is ostensibly "suspended".

Reviewed by Eric Carlson.

In r243958, a drive-by-fix was added that blocked playback when the WebProcess thought it
was susposed to be suspended. The intent was to keep the AVAudioSession for the process
from being activated just before the process was suspended, and thus avoid receiving an
"interruption" and having the AVAudioSession deactivated out from under us upon resuming.

Unfortunately, this caused problems when another process resumes the WebContent process
in order to start playback; namely the Now Playing UI's play button. Because we may receive
the play command long before whe notice that we've been resumed, the WebContent process
may refuse to honor the play command because it thinks its supposed to be suspended.

  • platform/audio/PlatformMediaSessionManager.cpp:


3:13 PM Changeset in webkit [247181] by bshafiei@apple.com
  • 3 edits
    2 adds in tags/Safari-608.1.34

Cherry-pick r247167. rdar://problem/52547473

[ContentChangeObserver] REGRESSION (r247015): facebook photo/video upload button is unresponsive to user interaction.

Reviewed by Simon Fraser.


Apparently it's a common practice to put transparent elements over visible click targets (button like divs) and use the invisible
elements to catch the user input (e.g. Facebook's Photo/Video button).
This patch modifies the original "do not trigger click on invisible targets" heuristic to a more restrictive "do not trigger
click if the click target was previously hidden and became visible through touch start".
If this still breaks some use cases, we could turn it into YouTube quirk.

  • WebProcess/WebPage/ios/WebPageIOS.mm: (WebKit::WebPage::handleSyntheticClick):


  • fast/events/touch/ios/content-observation/opacity-change-happens-on-touchstart-with-transition3-expected.txt: Added.
  • fast/events/touch/ios/content-observation/opacity-change-happens-on-touchstart-with-transition3.html: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@247167 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2:56 PM Changeset in webkit [247180] by rniwa@webkit.org
  • 14 edits
    2 adds in trunk

[iOS] Crash in WebKit::WebPage::positionInformation via Range::startPosition

Reviewed by Wenson Hsieh.


  • editing/Editor.cpp:

(WebCore::Editor::compositionRange const): Added a FIXME.


The crash was caused because focusedElementPositionInformation asssumes Editor::compositionRange is not null
whenever Editor::hasComposition returns true, which is not necessary the case when Editor::m_compositionNode
contains no text (data is of length 0).

Fixed the crash by adding an early return for when Editor::compositionRange returns nullptr.

  • WebProcess/WebPage/ios/WebPageIOS.mm:



Added UIScriptController.ensurePositionInformationIsUpToDateAt using the existing WKWebView SPI:

  • DumpRenderTree/ios/UIScriptControllerIOS.mm:


  • DumpRenderTree/mac/UIScriptControllerMac.mm:


  • TestRunnerShared/UIScriptContext/Bindings/UIScriptController.idl:
  • TestRunnerShared/UIScriptContext/UIScriptController.cpp:


  • TestRunnerShared/UIScriptContext/UIScriptController.h:
  • WebKitTestRunner/ios/UIScriptControllerIOS.mm:


  • WebKitTestRunner/ios/UIScriptControllerMac.mm:



Added a regression test for the crash.

  • editing/input/delete-text-in-composition-expected.txt: Added.
  • editing/input/delete-text-in-composition.html: Added.
  • resources/ui-helper.js:

(window.UIHelper.ensurePositionInformationUpdateForElement): Added.

2:48 PM Changeset in webkit [247179] by Devin Rousso
  • 2 edits in trunk/Source/WebInspectorUI

Web Inspector: current call frame indicator not visible in dark mode

Reviewed by Timothy Hatcher.

  • UserInterface/Views/CallFrameTreeElement.css:

(.tree-outline .item.call-frame .status > .status-image):
(.tree-outline .item.call-frame.selected .status > .status-image):
Replace the CSS variables used for fill with their actual color value (in light mode),
as the semantic "name" of the variable doesn't really match its usage. As an example, the
current call frame shouldn't rely on the value of --border-color, as it's not a "border".

2:47 PM Changeset in webkit [247178] by bshafiei@apple.com
  • 8 edits in tags/Safari-608.1.33.2/Source

Cherry-pick r247095. rdar://problem/52695729

Unreviewed, rolling out r246616.

Caused http/tests/inspector/network/har/har-page.html to fail
on Catalina.

Reverted changeset:

"Web Inspector: Network: replace CFNetwork SPI with new API
where able"

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@247095 268f45cc-cd09-0410-ab3c-d52691b4dbfc

2:45 PM Changeset in webkit [247177] by bshafiei@apple.com
  • 7 edits in tags/Safari-608.1.33.2/Source


2:42 PM Changeset in webkit [247176] by bshafiei@apple.com
  • 1 copy in tags/Safari-608.1.33.2

Tag Safari-608.1.33.2.

2:30 PM Changeset in webkit [247175] by ysuzuki@apple.com
  • 2 edits in trunk/Source/JavaScriptCore

Unreviewed, fix build failure on ARM64_32

Implicit narrowing from uint64_t to uint32_t happens. We should explicitly narrow it because we already checked
the length is <= UINT32_MAX.

  • runtime/ArrayPrototype.cpp:


2:14 PM Changeset in webkit [247174] by mmaxfield@apple.com
  • 21 edits
    3 copies
    3 adds in trunk

[WHLSL] Standard library is too big to directly include in WebCore

Reviewed by Saam Barati.


This patch adds almost the entire remainder of the standard library. There are a few
pieces missing:

There were two problems with adding so many standard library functions:

  • We didn't want to increase the WebCore binary size that much
  • Compiling all the functions at runtime took 7 seconds, which is much too long

This patch addresses the first problem by gzipping the standard library before including it in the binary.
At runtime, we use libcompression to unzip it.

To address the second problem, we did some analysis and found that 14% of that 7 seconds was simply
destroying all the AST nodes. Even if we eliminated all processing of the AST, simply having the AST
of the entire standard library built and destroyed would still be too slow. Therefore, this patch limits
which parts of the standard library get parsed in the first place. All the functions in the standard library
file are sorted by name, and each group of functions with the same name are preceeded by a comment of the
form /* Functions named xyz */. At build time, a Python script looks for all these comments, and builds a
map from function name to character offset inside the file where those functions begin. At run time, we
parse the user program first, look for all function calls within it, and look up those function call names
in the map to see which part of the standard library holds those functions. We then parse just that part.
Because the standard library can call other functions in the standard library, we do this in a loop until
we have exhausted all the functions.

Covered by existing tests.

  • DerivedSources-input.xcfilelist:
  • DerivedSources-output.xcfilelist:
  • DerivedSources.make: gzip the standard library, and add a build step to generate the offset map.
  • Modules/webgpu/WHLSL/AST/WHLSLCallExpression.h:

(WebCore::WHLSL::AST::CallExpression::function): Every caller of this ASSERT()s the result. Might as well
move the ASSERT() into the function.

  • Modules/webgpu/WHLSL/Cocoa/WHLSLStandardLibraryUtilities.cpp: Added.

(WebCore::WHLSL::decompressStandardLibrary): Use libcompression. This is why this file is in a Cocoa/
subfolder, and is listed in SourcesCocoa.txt instead of Sources.txt.
(WebCore::WHLSL::includeStandardLibrary): Include only the bits of the standard library which are relevant,
as described above.

  • Modules/webgpu/WHLSL/Metal/WHLSLFunctionWriter.cpp: Only emit MSL code for functions which are actually

reached. The MSL compiler is a significant amount of our compile time, so reducing the size of the emitted
program can significantly improve compile times.

  • Modules/webgpu/WHLSL/Metal/WHLSLNativeFunctionWriter.cpp:


  • Change how load() and store() are written. We need explicit functions because we have explicit atomic types, which HLSL doesn't have. load() and store() aren't present in HLSL.
  • Delete f16tof32 because they're probably not important and are not obvious how to implement. We can re-add them again later if necessary.
  • Various fixes to make us generate the correct MSL code for each standard library function.
  • Modules/webgpu/WHLSL/WHLSLBuildStandardLibraryFunctionMap.py: Added. Build the function map as described


  • Modules/webgpu/WHLSL/WHLSLChecker.cpp:

(WebCore::WHLSL::resolveByInstantiation): Previously, the standard library included an operator== for two
pointers. However, that function should be generated by the compiler instead. This fixes the bug in the
compiler which allows the compiler to correctly generate the right function. This also prompted me to file

  • Modules/webgpu/WHLSL/WHLSLFunctionStageChecker.cpp: Update to the new function() signature.
  • Modules/webgpu/WHLSL/WHLSLLexer.h: Add a new position() method to the lexer. This isn't actually used

in this patch, but it's useful when doing some analysis during parsing.
(WebCore::WHLSL::Lexer::peek const):
(WebCore::WHLSL::Lexer::peekFurther const):
(WebCore::WHLSL::Lexer::position const):
(WebCore::WHLSL::Lexer::state const):
(WebCore::WHLSL::Lexer::isFullyConsumed const): Fixes a bug where isFullyConsumed() might return true even
when there is a token in the ring buffer.
(WebCore::WHLSL::Lexer::peek): Deleted.
(WebCore::WHLSL::Lexer::peekFurther): Deleted.

  • Modules/webgpu/WHLSL/WHLSLNameResolver.cpp:

(WebCore::WHLSL::NameResolver::visit): Drive-by partial fix of https://bugs.webkit.org/show_bug.cgi?id=199347

  • Modules/webgpu/WHLSL/WHLSLPrepare.cpp: Call includeStandardLibrary().


  • Modules/webgpu/WHLSL/WHLSLPreserveVariableLifetimes.cpp:
  • Modules/webgpu/WHLSL/WHLSLRecursionChecker.cpp:
  • Modules/webgpu/WHLSL/WHLSLStandardLibrary.txt:
  • Modules/webgpu/WHLSL/WHLSLStandardLibraryFunctionMap.h: Copied from Source/WebCore/Modules/webgpu/WHLSL/WHLSLRecursionChecker.cpp.
  • Modules/webgpu/WHLSL/WHLSLStandardLibraryUtilities.h: Copied from Source/WebCore/Modules/webgpu/WHLSL/WHLSLRecursionChecker.cpp.
  • SourcesCocoa.txt:
  • WebCore.xcodeproj/project.pbxproj:


The standard library doesn't include any matrix constructors that broadcast a single value
to every field in a matrix. https://bugs.webkit.org/show_bug.cgi?id=199333

  • webgpu/whlsl-matrix.html:
1:26 PM Changeset in webkit [247173] by commit-queue@webkit.org
  • 13 edits in trunk

[JSC] Clean up ArraySpeciesCreate

Patch by Alexey Shvayka <Alexey Shvayka> on 2019-07-05
Reviewed by Yusuke Suzuki.


Adjusts error message expectations in stress tests.

  • stress/array-flatmap.js:
  • stress/array-flatten.js:
  • stress/array-species-create-should-handle-masquerader.js:
  • test262/expectations.yaml: Mark 4 test cases as passing.


We have duplicate code in arraySpeciesCreate, filter, map, concatSlowPath of ArrayPrototype.js
and speciesConstructArray of ArrayPrototype.cpp. This patch fixes cross-realm Array constructor
detection in native speciesConstructArray, upgrades length type to correctly handle large integers,
and exposes it as @arraySpeciesCreate. Also removes now unused @isArrayConstructor private function.
Native speciesConstructArray is preferred because it has fast path via speciesWatchpointIsValid.

Thoroughly benchmarked: this change progresses ARES-6 by 0-1%.

  • builtins/ArrayPrototype.js:

(globalPrivate.arraySpeciesCreate): Deleted.

  • builtins/BuiltinNames.h:
  • runtime/ArrayConstructor.cpp:

(JSC::arrayConstructorPrivateFuncIsArrayConstructor): Deleted.

  • runtime/ArrayConstructor.h:
  • runtime/ArrayPrototype.cpp:


  • runtime/ArrayPrototype.h:
  • runtime/JSGlobalObject.cpp:


1:13 PM Changeset in webkit [247172] by youenn@apple.com
  • 2 edits in trunk/Source/WebCore

Add fetch quirk for www.bnz.co.nz

Reviewed by Chris Dumez.

Use fetch quirk to make bnz main page load properly.
Manually tested.

  • page/Quirks.cpp:

(WebCore::Quirks::shouldIgnoreInvalidSignal const):

1:10 PM Changeset in webkit [247171] by rmorisset@apple.com
  • 7 edits in trunk/Source/WebCore

[WHLSL] Optimize the lexer

Reviewed by Myles Maxfield.

This patch includes a number of small optimizations to the lexer:

  • By creating two new types of Token (Invalid and EOF), we can avoid having to manipulate Optional<Token> everywhere
  • By combining the lexing of identifiers, keywords and operator names, we can avoid re-reading them several times
  • By storing mere offsets in the tokens instead of string views, we save space in the common case where the string view is unused (e.g. for keywords)
  • By storing mere offsets in the tokens instead of line numbers, we can simplify the whitespace and comment lexing; and we can easily recompute the line number on demand
  • By using templates we can avoid constructing and destructing vectors at runtime every time we call tryTypes, consumeTypes or peekTypes; as well as avoiding a call to find().
  • I also hoisted the length check out of the loop in Lexer::string()

All of this combined brought the time spent in the parser when running LayoutTests/webgpu/whlsl-test-harness-test.html 5 times on my machine from 612ms to 462ms.

I also removed the 'anyCharacter' method since it was dead code.
I also removed Lexer::position(), but it is directly equivalent now to peek().startOffset.

No new tests as there is no intended functional change.

  • Modules/webgpu/WHLSL/WHLSLLexer.cpp:

(WebCore::WHLSL::Lexer::digitStar const):
(WebCore::WHLSL::Lexer::completeOperatorName const):

  • Modules/webgpu/WHLSL/WHLSLLexer.h:

(WebCore::WHLSL::Lexer::peek const):
(WebCore::WHLSL::Lexer::peekFurther const):
(WebCore::WHLSL::Lexer::state const):
(WebCore::WHLSL::Lexer::isFullyConsumed const):
(WebCore::WHLSL::Lexer::string const):

  • Modules/webgpu/WHLSL/WHLSLNameResolver.h:
  • Modules/webgpu/WHLSL/WHLSLParser.cpp:


  • Modules/webgpu/WHLSL/WHLSLParser.h:
  • Modules/webgpu/WHLSL/WHLSLPreserveVariableLifetimes.cpp:


1:09 PM Changeset in webkit [247170] by rmorisset@apple.com
  • 9 edits in trunk/Source/WebCore

[WHLSL] Remove the phase resolveCallsInFunctions

Reviewed by Myles Maxfield.

This pass only stores into each property access and call expression vectors of all the functions it might be calling, for use by the Checker afterwards.
But the checker is perfectly able to compute a pointer to these vectors by itself.
So by removing this pass, we gain the following:

  • One less pass over the AST
  • No need to copy these vectors (which can be large for heavily overloaded functions, of which there are quite a few in the stdlib)
  • No need to have these vectors in the expressions, saving 24 bytes per CallExpression and 72 bytes per PropertyAccessExpression
  • No need to allocate and then destroy these vectors.

No new tests as there is no intended functional change.

  • Modules/webgpu/WHLSL/AST/WHLSLCallExpression.h:


  • Modules/webgpu/WHLSL/AST/WHLSLPropertyAccessExpression.h:
  • Modules/webgpu/WHLSL/WHLSLAutoInitializeVariables.cpp:


  • Modules/webgpu/WHLSL/WHLSLChecker.cpp:


  • Modules/webgpu/WHLSL/WHLSLNameResolver.cpp:


  • Modules/webgpu/WHLSL/WHLSLNameResolver.h:
  • Modules/webgpu/WHLSL/WHLSLPrepare.cpp:


1:06 PM Changeset in webkit [247169] by Matt Baker
  • 2 edits in trunk/Source/WebInspectorUI

Web Inspector: Elements: use a CSS transition when showing/hiding DOM breakpoint gutter

Reviewed by Devin Rousso.

  • UserInterface/Views/DOMTreeContentView.css:

(.content-view.dom-tree .tree-outline.dom):

12:19 PM Changeset in webkit [247168] by Dewei Zhu
  • 2 edits in trunk/Websites/perf.webkit.org

Test freshness page should expose revision information for latest build on tooltip.

Reviewed by Ryosuke Niwa.

Added latest build revision information on test freshness page tooltip.

  • public/v3/pages/test-freshness-page.js: Add build revision information on tooltip.


12:12 PM Changeset in webkit [247167] by Alan Bujtas
  • 3 edits
    2 adds in trunk

[ContentChangeObserver] REGRESSION (r247015): facebook photo/video upload button is unresponsive to user interaction.

Reviewed by Simon Fraser.


Apparently it's a common practice to put transparent elements over visible click targets (button like divs) and use the invisible
elements to catch the user input (e.g. Facebook's Photo/Video button).
This patch modifies the original "do not trigger click on invisible targets" heuristic to a more restrictive "do not trigger
click if the click target was previously hidden and became visible through touch start".
If this still breaks some use cases, we could turn it into YouTube quirk.

  • WebProcess/WebPage/ios/WebPageIOS.mm:



  • fast/events/touch/ios/content-observation/opacity-change-happens-on-touchstart-with-transition3-expected.txt: Added.
  • fast/events/touch/ios/content-observation/opacity-change-happens-on-touchstart-with-transition3.html: Added.
11:57 AM Changeset in webkit [247166] by Tadeu Zagallo
  • 2 edits in trunk/Source/JavaScriptCore

Unreviewed, change the value used to scribble Heap::m_worldState

Follow-up after r247160. The value used to scribble should have the
conn bit set.

  • heap/Heap.cpp:


11:54 AM Changeset in webkit [247165] by youenn@apple.com
  • 9 edits in trunk

[iOS] Local capture MediaStreamTrack does not render in portrait mode

Reviewed by Eric Carlson.


RealtimeVideoSource was badly computing its size in case of rotation.
Whenever its underlying source is notifying of settings change,
compute the size and transpose it only in left/right case.

Update mock video source to cover that case.
Covered by updated test.

  • platform/mediastream/RealtimeVideoSource.cpp:


  • platform/mediastream/mac/MockRealtimeVideoSourceMac.h:
  • platform/mediastream/mac/MockRealtimeVideoSourceMac.mm:


  • platform/mock/MockRealtimeVideoSource.cpp:


  • platform/mock/MockRealtimeVideoSource.h:


  • webrtc/video-rotation-expected.txt:
  • webrtc/video-rotation.html:
11:05 AM Changeset in webkit [247164] by Ryan Haddad
  • 27 edits
    6 deletes in trunk

Unreviewed, rolling out r247115.

Breaks lldbWebKitTester (and by extension, test-webkitpy)

Reverted changeset:

"[WHLSL] Standard library is too big to directly include in

11:00 AM Changeset in webkit [247163] by Ryan Haddad
  • 6 edits in trunk/Source/WebCore

Unreviewed, rolling out r247159.

IndexedDB layout tests are failing the new assert.

Reverted changeset:

"ASSERT that a sessionID is valid when encoding it"

9:59 AM Changeset in webkit [247162] by Ryan Haddad
  • 12 edits in trunk

Unreviewed, rolling out r247123.

Caused TestWebKitAPI.Challenge.BasicProposedCredential to

Reverted changeset:

"Only allow fetching and removing session credentials from

9:58 AM Changeset in webkit [247161] by Wenson Hsieh
  • 8 edits
    3 adds in trunk

Click events on outer page are not being dispatched correctly after touch-zooming within an iframe

Reviewed by Simon Fraser.


Mitigations introduced in r227759 prevent a touch inside a document of origin A to result in a click event being
dispatched on an element inside a document of origin B. It accomplishes this by keeping track of the security
origin of the last touch via m_potentialTapSecurityOrigin on WebPage. However, there exists a corner case in
which m_potentialTapSecurityOrigin, set after touching a document in a subframe, may persist even after the user
has finished interacting, causing taps in subsequent documents to not result in synthetic click events due to
mismatched potential tap origins.

This may happen if the first user gesture happens inside an element in a subframe with touch event handlers, but
no click event handler (and the touch is additionally not over some clickable element). In this case,
m_potentialTapNode is set to null in WebPage::potentialTapAtPosition, and when we consult it in
WebPage::commitPotentialTap, we just end up calling commitPotentialTapFailed(); and return early. This means
that m_potentialTapNode, m_potentialTapLocation, and (importantly) m_potentialTapSecurityOrigin are not reset.
This causes subsequent taps in the top-level document to never dispatch click events, if the security origin
does not match with that of the subframe.

To fix this, we add a new async IPC message from the UI process to the web process to ensure that our current
security origin is reset before attempting to handle a tap.

Test: http/tests/events/touch/ios/click-after-handling-touch-in-cross-origin-frame.https.html

  • UIProcess/WebPageProxy.cpp:


Reset any stale potential tap security origin if needed.

  • UIProcess/WebPageProxy.h:
  • UIProcess/ios/WKContentViewInteraction.mm:

(-[WKContentView _webTouchEventsRecognized:]):

Send the new IPC message when starting a touch.

  • WebProcess/WebPage/WebPage.cpp:


  • WebProcess/WebPage/WebPage.h:
  • WebProcess/WebPage/WebPage.messages.in:


Add a new layout test to verify that after interacting with an element with touch event handlers in a cross-
origin subframe, the user is still able to click on elements on the top level document.

  • http/tests/events/touch/ios/click-after-handling-touch-in-cross-origin-frame.https-expected.txt: Added.
  • http/tests/events/touch/ios/click-after-handling-touch-in-cross-origin-frame.https.html: Added.
  • http/tests/events/touch/ios/resources/touch-target.html: Added.
9:38 AM Changeset in webkit [247160] by Tadeu Zagallo
  • 2 edits in trunk/Source/JavaScriptCore

Scribble Heap::m_worldState on destructor

Reviewed by Sam Weinig.

The worldState is dumped when we crash due to a failed checkConn, and
this will make it clear if the heap has already been destroyed.

  • heap/Heap.cpp:


9:24 AM Changeset in webkit [247159] by youenn@apple.com
  • 6 edits in trunk/Source/WebCore

ASSERT that a sessionID is valid when encoding it

Reviewed by Darin Adler.


For IDBValue, instead of encoding an invalid session ID, encode a boolean that tells there is no sessionID.
For IDBRequestData, keep track of whether there is an IDBDatabaseIdentifier
and encode/decode accordingly to not encode an invalid sessionID.
No observable change of behavior.

  • Modules/indexeddb/IDBValue.h:

(WebCore::IDBValue::sessionID const):
(WebCore::IDBValue::encode const):

  • Modules/indexeddb/shared/IDBRequestData.cpp:


  • Modules/indexeddb/shared/IDBRequestData.h:

(WebCore::IDBRequestData::databaseIdentifier const):


ASSERT that a sessionID is valid at encoding/decoding time.

  • pal/SessionID.h:

(PAL::SessionID::encode const):

9:22 AM Changeset in webkit [247158] by Wenson Hsieh
  • 14 edits
    2 adds in trunk

Touching media controls sometimes shows software keyboard

Reviewed by Eric Carlson.


In r243044, we added a compatibility hack for Google Slides (and other G-suite properties) to allow the on-
screen keyboard to show up after a prevented touch event in the case where an element was already focused, even
if the touch event handler doesn't explicitly refocus the element. However, this means that if a regular text
field (or other form control) has been programmatically focused, then interacting with any other element that
prevents default on touchstart will cause us to show the keyboard for that focused element.

To mitigate this, only fall down this refocusing codepath in the case where the focused element is a hidden
editable element (in the style of many Google productivity web apps). For non-hidden editable elements that are
already focused, this refocusing logic is not necessary, since the user should be able to interact with the
control to show the keyboard anyways; for hidden editable areas, this compatibility hack is actually needed,
since there is typically no other way for a user to focus these elements and show an on-screen keyboard.

Tests: fast/events/touch/ios/show-keyboard-after-preventing-touchstart.html


  • WebProcess/WebPage/WebPage.cpp:


  • WebProcess/WebPage/WebPage.h:
  • WebProcess/WebPage/ios/WebPageIOS.mm:

(WebKit::WebPage::isTransparentOrFullyClipped const):

Renamed from enclosingLayerIsTransparentOrFullyClipped, and pulled out into a private helper method.

(WebKit::WebPage::platformEditorState const):
(WebKit::enclosingLayerIsTransparentOrFullyClipped): Deleted.


Adds plumbing for a new testing hook to check whether or not there is an active input session. See other
ChangeLog entries for more detail.

  • DumpRenderTree/ios/UIScriptControllerIOS.mm:

(WTR::UIScriptController::hasInputSession const):

  • TestRunnerShared/UIScriptContext/Bindings/UIScriptController.idl:
  • TestRunnerShared/UIScriptContext/UIScriptController.cpp:

(WTR::UIScriptController::hasInputSession const):

  • TestRunnerShared/UIScriptContext/UIScriptController.h:
  • WebKitTestRunner/ios/UIScriptControllerIOS.mm:

(WTR::UIScriptController::hasInputSession const):


Adds a new layout test to verify that the keyboard only appears after a handled touch event if the focused
element is inside a hidden editable area; otherwise, the keyboard should not be present.

  • fast/events/touch/ios/do-not-show-keyboard-after-preventing-touchstart-expected.txt: Added.
  • fast/events/touch/ios/do-not-show-keyboard-after-preventing-touchstart.html: Added.

This test passes as long as we didn't begin showing the keyboard after tapping.

  • fast/events/touch/ios/show-keyboard-after-preventing-touchstart-expected.txt:
  • fast/events/touch/ios/show-keyboard-after-preventing-touchstart.html:

Adjust this existing test to make the focused textarea hidden.

  • resources/ui-helper.js:


Add a new testing hook to check whether there is an active input session.

8:29 AM Changeset in webkit [247157] by bshafiei@apple.com
  • 24 edits
    6 deletes in tags/Safari-608.1.34

Revert r247115. rdar://problem/51288898

8:29 AM Changeset in webkit [247156] by bshafiei@apple.com
  • 7 edits
    1 copy in tags/Safari-608.1.34/Source/WebCore

Revert r247124. rdar://problem/52687835

8:15 AM Changeset in webkit [247155] by bshafiei@apple.com
  • 7 edits in trunk/Source


8:07 AM Changeset in webkit [247154] by bshafiei@apple.com
  • 1 copy in tags/Safari-608.1.34

Tag Safari-608.1.34.

7:24 AM Changeset in webkit [247153] by Alan Bujtas
  • 2 edits in trunk/Source/WebCore

[LFC][IFC] Inline runs should be relative to the line's top/left initially

Reviewed by Antti Koivisto.

Let's construct the runs relative to the line's top/left and convert them relative to the formatting root's border box top/left
when the line is being closed. When the line moves (floats etc) we don't have to move each runs individually.

  • layout/inlineformatting/InlineLine.cpp:


7:18 AM Changeset in webkit [247152] by Antti Koivisto
  • 5 edits in trunk/Source/WebCore

REGRESSION(r244218): desmos.com: Cannot scroll formulas region when region is scrollable

Reviewed by Zalan Bujtas.

The page invokes preventDefault on simulated mouse events which stops scrolling.
To fix, add a quirk that turns simulated mouse events non-cancelable.

  • dom/MouseEvent.h:
  • dom/ios/MouseEventIOS.cpp:


  • page/Quirks.cpp:

(WebCore::Quirks::simulatedMouseEventTypeForTarget const):
(WebCore::Quirks::shouldDispatchSimulatedMouseEventsOnTarget const): Deleted.

To avoid adding more similar functions, and for future flexibility, make this one return the type too.

  • page/Quirks.h:
7:02 AM Changeset in webkit [247151] by Alan Bujtas
  • 3 edits in trunk/Source/WebCore

[LFC][IFC] Empty text node should generate empty InlineTextItem

Reviewed by Antti Koivisto.

Empty text nodes generate empty RenderText renderers but they don't generate inline boxes.
This patch aligns LFC with the current rendering code.

  • layout/inlineformatting/InlineLine.cpp:


  • layout/inlineformatting/InlineTextItem.cpp:


6:50 AM BuildingCairoOnWindows edited by Fujii Hironori
6:36 AM WebKitGTK/2.24.x edited by Adrian Perez de Castro
4:27 AM Changeset in webkit [247150] by graouts@webkit.org
  • 1 edit
    3 deletes in trunk/LayoutTests/imported/w3c

Remove Wincorrect WPT test pointerevents/pointerevent_mouse_pointercapture_inactivate_pointer.html

Unreviewed, just removing a test that was removed from the WPT repository (https://github.com/web-platform-tests/wpt/pull/17614).

  • web-platform-tests/pointerevents/pointerevent_mouse_pointercapture_inactivate_pointer-expected.txt: Removed.
  • web-platform-tests/pointerevents/pointerevent_mouse_pointercapture_inactivate_pointer.html: Removed.
  • web-platform-tests/pointerevents/resources/pointerevent_mouse_pointercapture_inactivate_pointer-iframe.html: Removed.
4:23 AM Changeset in webkit [247149] by Adrian Perez de Castro
  • 2 edits in trunk/Source/WebCore

[ATK] Do not use C linkage for functions using C++ features

Reviewed by Carlos Garcia Campos.

No new tests needed.

  • accessibility/atk/WebKitAccessible.h: Move function and type declarations involving

C++ features outside of the block delimited by the G_BEGIN_DECLS and G_END_DECLS macros,
which were setting the C linkage.

3:25 AM Changeset in webkit [247148] by graouts@webkit.org
  • 15 edits in trunk

[Pointer Events] Respect pointer capture when dispatching mouse boundary events and updating :hover

Reviewed by Dean Jackson.


Mark the progressions in 3 WPT tests.

  • web-platform-tests/pointerevents/pointerevent_boundary_events_at_implicit_release_hoverable_pointers-expected.txt:
  • web-platform-tests/pointerevents/pointerevent_mouse_capture_change_hover-expected.txt:
  • web-platform-tests/pointerevents/pointerevent_setpointercapture_relatedtarget-expected.txt:


Up until now, we would not account for pointer capture (see ​https://w3c.github.io/pointerevents/#pointer-capture) when dispatching
mouse boundary events (mouseover, mouseout, mouseenter, mouseleave) and their counterpart pointer events. We would also not account
for it when updating :hover styles.

Now, when pointer capture changes for an element, we call setCapturingMouseEventsElement() on the EventHandler such that the element
that would naturally hit-test is overridden by the pointer capture element when identifying which target to use for the dispatch of
boundary mouse events. Additionally, when calling Document::prepareMouseEvent(), we also use the pointer capture element to
pass down to Document::updateHoverActiveState() such that :hover styles are applied to the correct element.

  • dom/Document.cpp:

(WebCore::Document::prepareMouseEvent): When a new event is going to be dispatched, we must run the Process Pending Capture Element
steps as mandated by the Pointer Events spec. Calling this will dispatch the appropriate pointer capture change events and also
required boundary events since EventHandler::setCapturingMouseEventsElement() calls into EventHandler::updateMouseEventTargetNode().
Since this may update the capturing mouse events element, we ensure that we call updateHoverActiveState() with a flag that indicates that.
Finally, we use the capturing mouse events element instead of the hit-testing element to pass to updateHoverActiveState() to ensure
that is has :hover styles applied.
(WebCore::Document::updateHoverActiveState): Account for the new CaptureChange flag to force the invalidation of the :hover and :active
elements chain at all times when the capturing mouse events element changed.

  • dom/Document.h:
  • dom/PointerEvent.h: Update PointerEvent::createForPointerCapture() to take specific parameters rather than a single PointerEvent to

set the pointerId, isPrimary and pointerType properties of the generated event. This is required to call processPendingPointerCapture()
outside of PointerEvent dispatch logic since we now call it from Document::prepareMouseEvent() where we haven't yet generated such an

  • page/EventHandler.cpp:

(WebCore::EventHandler::pointerCaptureElementDidChange): When a new pointer capture element is set, call updateMouseEventTargetNode()
to ensure that boundary events are fired to indicate the pointer capture state change.
(WebCore::EventHandler::prepareMouseEvent): Keep track of the last PlatformMouseEvent used to prepare a mouse event so that we can use
it when setCapturingMouseEventsElement() is called.

  • page/EventHandler.h:
  • page/PointerCaptureController.cpp:

(WebCore::PointerCaptureController::pointerCaptureElement): Since Document::prepareMouseEvent() needs to know the current pointer capture
element, add a new public method that indicates the pointer capture element if that element is contained in the provided document. We need
to provide the document since PointerCaptureController is owned by the Page and may manage several documents.
(WebCore::PointerCaptureController::dispatchEvent): Only run the Process Pending Capture Element steps when dealing with a touch or pen
event since those steps are already ran for mouse events in Document::prepareMouseEvent(). Additionally, since the element target is already
set to be the pointer capture element with the changes made to processPendingPointerCapture(), and because on iOS pointer capture is always
active, we can remove the code that would retarget the event to the pointer capture element.
(WebCore::PointerCaptureController::processPendingPointerCapture): We now call into EventHandler::setCapturingMouseEventsElement() when the
capture target element changes. We must be careful to call this method prior to dispatching the "gotpointercapture" event and after dispatching
the "lostpointercapture" event so that boundary events are fired at the right time.

  • page/PointerCaptureController.h:


Update some WK1-specific expectations.

  • platform/mac-wk1/imported/w3c/web-platform-tests/pointerevents/pointerevent_boundary_events_at_implicit_release_hoverable_pointers-expected.txt:
  • platform/mac-wk1/imported/w3c/web-platform-tests/pointerevents/pointerevent_setpointercapture_relatedtarget-expected.txt:

Jul 4, 2019:

10:44 PM Changeset in webkit [247147] by Alan Bujtas
  • 4 edits
    2 adds in trunk

[ContentChangeObserver] didFinishTransition triggers a nested style recalc via isConsideredClickable

Reviewed by Ryosuke Niwa.


isConsideredClickable() may trigger style update through Node::computeEditability. Let's adjust the observer state in the next runloop.

Test: fast/events/touch/ios/content-observation/animation-end-with-visiblity-change-and-recursive-update-style.html

  • page/ios/ContentChangeObserver.cpp:


  • page/ios/ContentChangeObserver.h:


  • fast/events/touch/ios/content-observation/animation-end-with-visiblity-change-and-recursive-update-style-expected.txt: Added.
  • fast/events/touch/ios/content-observation/animation-end-with-visiblity-change-and-recursive-update-style.html: Added.
4:53 PM Changeset in webkit [247146] by Chris Dumez
  • 12 edits
    4 deletes in trunk/Source/WebKit

Simplify logic that handles registering WebProcessProxy objects with their WebsiteDataStore

Reviewed by Ryosuke Niwa.

Simplify logic that handles registering WebProcessProxy objects with their WebsiteDataStore to make
it more maintainable and less error-prone (avoid bugs like <rdar://problem/51899751>).

The following changes were made:

  1. The WebProcess now registers / unregisters itself directly with its WebsiteDataStore when needed, instead of having the WebPageProxy/ProvisionalPageProxy be in charge of calling the right WebProcessLifetimeTracker / WebProcessLifetimeObserver abstractions.
  2. The WebProcessLifetimeTracker / WebProcessLifetimeObserver abstractions were dropped. The WebsiteDataStore was the only observer.
  3. The WebProcessProxy is now registered with its WebsiteDataStore as soon as it has a page (provisional or not) instead of waiting until the process has finished launching. This simplifies the logic a lot and waiting until the process has launched is not needed by the WebsiteDataStore.
  • Sources.txt:
  • UIProcess/ProvisionalPageProxy.cpp:


  • UIProcess/ProvisionalPageProxy.h:
  • UIProcess/WebPageProxy.cpp:


  • UIProcess/WebPageProxy.h:
  • UIProcess/WebProcessCache.cpp:


  • UIProcess/WebProcessLifetimeObserver.cpp: Removed.
  • UIProcess/WebProcessLifetimeObserver.h: Removed.
  • UIProcess/WebProcessLifetimeTracker.cpp: Removed.
  • UIProcess/WebProcessLifetimeTracker.h: Removed.
  • UIProcess/WebProcessProxy.cpp:


  • UIProcess/WebProcessProxy.h:
  • UIProcess/WebsiteData/WebsiteDataStore.cpp:


Drop change that was added in r246097 to work around the fact that WebProcessProxy objects
were only registering themselves with their data store after their process had finished

(WebKit::WebsiteDataStore::processPools const):

  • UIProcess/WebsiteData/WebsiteDataStore.h:

(WebKit::WebsiteDataStore::processes const):

  • WebKit.xcodeproj/project.pbxproj:
3:32 PM Changeset in webkit [247145] by Alan Bujtas
  • 11 edits
    1 copy
    1 add in trunk

[ContentChangeObserver] Limit mouseOut dispatching after synthetic click to YouTube.com

Reviewed by Wenson Hsieh.


r242798 introduced synthetic mouseOut dispatching soon after sythentic click to dismiss certain type of content (e.g. video control bar on YouTube).
However on many sites, it resulted in dismissing useful content, like menu panes and other, non-clickable but informative content.
This patch limits the mouseOut dispatching to YouTube.com.

Test: fast/events/touch/ios/content-observation/mouse-out-event-should-not-fire-on-click.html

  • page/Quirks.cpp:

(WebCore::Quirks::needsYouTubeMouseOutQuirk const):

  • page/Quirks.h:
  • page/Settings.yaml:
  • testing/InternalSettings.cpp:


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


  • WebProcess/WebPage/ios/WebPageIOS.mm:



  • fast/events/touch/ios/content-observation/mouse-out-event-should-fire-on-click.html:
  • fast/events/touch/ios/content-observation/mouse-out-event-should-not-fire-on-click-expected.txt: Added.
  • fast/events/touch/ios/content-observation/mouse-out-event-should-not-fire-on-click.html: Copied from LayoutTests/fast/events/touch/ios/content-observation/mouse-out-event-should-fire-on-click.html.
11:53 AM Changeset in webkit [247144] by Ryan Haddad
  • 8 edits in trunk/Source/WebCore

Unreviewed, rolling out r247127.

Broke the watchOS build.

Reverted changeset:

"[WHLSL] Remove the phase resolveCallsInFunctions"

10:30 AM Changeset in webkit [247143] by aakash_jain@apple.com
  • 3 edits in trunk/Tools

[ews-build] Remove GTK and WPE queue from old EWS and dashboard

Reviewed by Michael Catanzaro.

  • BuildSlaveSupport/build.webkit.org-config/public_html/dashboard/Scripts/BubbleQueueServer.js:
  • QueueStatusServer/config/queues.py:
9:30 AM WebKitGTK/GCCRequirement edited by Michael Catanzaro
9:30 AM WebKitGTK/GCCRequirement edited by Michael Catanzaro
9:16 AM WebKitGTK/GCCRequirement edited by Michael Catanzaro
9:12 AM WebKitGTK/GCCRequirement edited by Michael Catanzaro
9:10 AM WebKitGTK/GCCRequirement edited by Michael Catanzaro
9:10 AM WebKitGTK/GCCRequirement edited by Michael Catanzaro
8:39 AM WebKitGTK/GCCRequirement edited by Michael Catanzaro
7:53 AM Changeset in webkit [247142] by Simon Fraser
  • 3 edits
    2 adds in trunk

Some layers inside overflow:scroll don't move back to correct positions when the contents are shrunk

Reviewed by Sam Weinig.

When an overflow scroller's content shrink and it becomes non-scrollable, the 'boundsOrigin' in
the ancestor clipping stacks of related layers wouldn't get reset to zero, leaving layers in
the wrong place, so always set boundsOrigin to zero on the clipping layers of non-overflow entries
in the ancestor clipping stack.

Test: compositing/scrolling/async-overflow-scrolling/become-non-scrollable-with-child.html

  • rendering/RenderLayerBacking.cpp:



  • compositing/scrolling/async-overflow-scrolling/become-non-scrollable-with-child-expected.html: Added.
  • compositing/scrolling/async-overflow-scrolling/become-non-scrollable-with-child.html: Added.
6:54 AM Changeset in webkit [247141] by Konstantin Tokarev
  • 2 edits in trunk/Source/WebCore

HyphenationLibHyphen: Include GLib-related headers only for PLATFORM(GTK)

Reviewed by Michael Catanzaro.

Also added compile guard for a couple of functions used only in GTK port.

  • platform/text/hyphen/HyphenationLibHyphen.cpp:
6:53 AM Changeset in webkit [247140] by clopez@igalia.com
  • 2 edits in trunk/Tools

[GTK] Fix browserperfdash script after r239522

Reviewed by Michael Catanzaro.

  • Scripts/webkitpy/browserperfdash/browserperfdash_runner.py:

(BrowserPerfDashRunner.run): Pass None for the new required BenchmarkRunner constructor.

6:28 AM Changeset in webkit [247139] by clopez@igalia.com
  • 2 edits in trunk/Tools

[webkitpy] test-webkitpy is broken on Linux since r246662

Reviewed by Michael Catanzaro.

  • Scripts/webkitpy/test/main.py:

(_supports_building_and_running_lldb_tests): Add a check for None before calling startswith().

3:52 AM Changeset in webkit [247138] by clopez@igalia.com
  • 5 edits in trunk/Tools

[GTK][WPE] Not longer needed to build woff2 on the JHBuild

Reviewed by Michael Catanzaro.

Install the Woff2 libraries and headers from the system package
and stop building them on the JHBuild.

  • gtk/install-dependencies:
  • gtk/jhbuild.modules:
  • wpe/install-dependencies:
  • wpe/jhbuild.modules:
3:50 AM Changeset in webkit [247137] by clopez@igalia.com
  • 2 edits in trunk/Tools

[GTK][WPE][CMake] The CMake cache should be clean when the JHBuild is rebuilt

Reviewed by Michael Catanzaro.

  • Scripts/webkitdirs.pm:

(shouldRemoveCMakeCache): Clean the CMake cache if we detect that the JHBuild has been rebuilt.

3:41 AM Changeset in webkit [247136] by clopez@igalia.com
  • 1 edit in trunk/Tools/Scripts/webkitpy/common/config/contributors.json

Update my status in contributors.json

Due to inactivity I lost my previous status as reviewer in r244705
but I still have svn access, so update my status temporally to
reflect the reality (checked with Alexey)

12:30 AM Changeset in webkit [247135] by Carlos Garcia Campos
  • 2 edits in trunk/Tools

Unreviewed, GTK a11y tests fix after r247071

Patch by Philippe Normand <pnormand@igalia.com> on 2019-07-01

  • WebKitTestRunner/InjectedBundle/atk/AccessibilityUIElementAtk.cpp:

(WTR::AccessibilityUIElement::popupValue const):

Note: See TracTimeline for information about the timeline view.