Timeline
Apr 21, 2019:
- 11:44 PM Changeset in webkit [244497] by
-
- 8 edits in trunk/Source/bmalloc
[bmalloc] Use StaticPerProcess' mutex as bmalloc::Heap does with PerProcess
https://bugs.webkit.org/show_bug.cgi?id=197135
Reviewed by Darin Adler.
This patch leverages StaticPerProcess::mutex() for per process instance's lock in various classes,
as Heap does with PerProcess::mutex().
- bmalloc/AllIsoHeaps.cpp:
(bmalloc::AllIsoHeaps::add):
(bmalloc::AllIsoHeaps::head):
- bmalloc/AllIsoHeaps.h:
- bmalloc/CryptoRandom.cpp:
(bmalloc::ARC4RandomNumberGenerator::randomValues):
- bmalloc/DebugHeap.cpp:
(bmalloc::DebugHeap::memalignLarge):
(bmalloc::DebugHeap::freeLarge):
- bmalloc/DebugHeap.h:
- bmalloc/Scavenger.cpp:
(bmalloc::Scavenger::run):
(bmalloc::Scavenger::runSoon):
(bmalloc::Scavenger::scheduleIfUnderMemoryPressure):
(bmalloc::Scavenger::schedule):
(bmalloc::Scavenger::timeSinceLastFullScavenge):
(bmalloc::Scavenger::scavenge):
(bmalloc::Scavenger::threadRunLoop):
- bmalloc/Scavenger.h:
- 11:22 PM Changeset in webkit [244496] by
-
- 3 edits in trunk/Source/WTF
[WTF] Generic memoryFootprint() implementation should use bmalloc on Linux
https://bugs.webkit.org/show_bug.cgi?id=196963
Reviewed by Don Olmstead.
Have the generic memoryFootprint() implementation use bmalloc's
memoryFootprint() API on Linux, whenever the system malloc option is
not enabled. Limitation to Linux platforms is due to the bmalloc
implementation being limited to those configurations (excluding iOS
which doesn't use MemoryFootprintGeneric.cpp).
- wtf/PlatformWPE.cmake: Switch to building MemoryFootprintGeneric.cpp.
- wtf/generic/MemoryFootprintGeneric.cpp:
(WTF::memoryFootprint):
- 5:24 PM Changeset in webkit [244495] by
-
- 4 edits in trunk/Source/WebCore
WHLSLPrepare.cpp always recompiles, even if nothing was changed
https://bugs.webkit.org/show_bug.cgi?id=197151
Reviewed by Dan Bernstein.
- DerivedSources-input.xcfilelist: Script updated this automatically after
DerivedSources.make was corrected.
- DerivedSources-output.xcfilelist: Ditto, although I had to manually remove
one bogus leftover reference to WHLSLStandardLibrary.cpp.
- DerivedSources.make: Updated the rule that builds WHSLStandardLibrary.h to
no longer refer to nonexistent WHLSLStandardLibrary.cpp. Because the dependency
was on a file that was never created, the rule to regenerate WHSLStandardLibrary.h
was running on every build, instead of only when one of the dependencies changed.
- 2:10 PM Changeset in webkit [244494] by
-
- 8 edits in trunk
Defer EditorState updates until the next layer tree flush in a few additional circumstances
https://bugs.webkit.org/show_bug.cgi?id=197145
<rdar://problem/50078170>
Reviewed by Darin Adler.
Source/WebKit:
Gets rid of sendPartialEditorStateAndSchedulePostLayoutUpdate(), in favor of always scheduling a full editor
state update after the next compositing flush.
- WebProcess/WebPage/WebPage.cpp:
(WebKit::WebPage::updateEditorStateAfterLayoutIfEditabilityChanged):
(WebKit::WebPage::setNeedsFontAttributes):
(WebKit::WebPage::didChangeOverflowScrollPosition):
(WebKit::WebPage::didChangeSelection):
(WebKit::WebPage::didChangeSelectionOrOverflowScrollPosition):
(WebKit::WebPage::sendPartialEditorStateAndSchedulePostLayoutUpdate): Deleted.
- WebProcess/WebPage/WebPage.h:
- WebProcess/WebPage/ios/WebPageIOS.mm:
(WebKit::WebPage::platformEditorState const):
(WebKit::WebPage::updateVisibleContentRects):
Tools:
Adjusts several editing API tests to wait for a presentation update following a selection change, programmatic
focus, or showing the font manager.
- TestWebKitAPI/Tests/WebKitCocoa/EditorStateTests.mm:
(TestWebKitAPI::TEST):
- TestWebKitAPI/Tests/WebKitCocoa/WKContentViewTargetForAction.mm:
(TEST):
- TestWebKitAPI/Tests/mac/FontManagerTests.mm:
(TestWebKitAPI::TEST):
- 1:14 PM Changeset in webkit [244493] by
-
- 4 edits in trunk/Source/WebKit
Regression(r243767) WebFrame::m_navigationIsContinuingInAnotherProcess flag is never reset
https://bugs.webkit.org/show_bug.cgi?id=197144
Reviewed by Darin Adler.
WebFrame::m_navigationIsContinuingInAnotherProcess flag is never reset since it was introduced in
r243767. This leads to leaking Navigation objects in the UIProcess when reusing a previously
suspended process because such process will no longer send the DidDestroyNavigation IPC.
It turns out that resetting the flags causes API tests such as ProcessSwap.QuickBackForwardNavigationWithPSON
to ASSERT. This is because when the UIProcess quickly navigate back and forth without waiting for policy
decisions, we may end up getting the policy decision for a particular navigation *after* we've sent the
DidDestroyNavigation.
As a result, this patch reverts r243767 and fixes in the assertion in http/tests/adClickAttribution/store-ad-click-attribution.html
another way. We initially assumed that the logic in WebPageProxy::didDestroyNavigation() was failing to
ignore the DidDestroyNavigation from the previous process after a swap due to a race, maybe because it was
sometimes received too late and m_provisionalPage was already cleared. However, this would not make sense
since the test is crashing consistently and the page would no longer be able to receive IPC from the
previous process *after* we've committed the provisional process/page.
The real issue was that the DidDestroyNavigation IPC was received *before* we could construct the
provisional page, which is why the logic in WebPageProxy::didDestroyNavigation() was failing to ignore
the bad IPC. In WebPageProxy::receivedNavigationPolicyDecision(), we were calling receivedPolicyDecision()
(which would send the DidReceivePolicyDecision to the previous WebProcess) and then continueNavigationInNewProcess()
in order to construct the provisional page. I personally did not expect we could receive IPC between the
calls to receivedNavigationPolicyDecision() and receivedPolicyDecision(), since we are not yielding and since
the DidReceivePolicyDecision IPC is asynchronous. However, this is exactly what was happening in the context
of this test. The reason is that the DidReceivePolicyDecision IPC was getting wrapped in a synchronous message
and sent as synchronous message due to the Connection::m_inDispatchMessageMarkedToUseFullySynchronousModeForTesting
flag which seems to get set in the test due to some EventSender IPC. I believe this is because the test uses
EventSender to do a click on a link which triggers the navigation.
To address the issue, I now call receivedNavigationPolicyDecision() *after* continueNavigationInNewProcess()
to make sure that we always start the provisional load in the new process before we tell the previous process
to stop loading. This way, there is no way we get IPC from the previous process about the current navigation
before we have a provisional page.
- UIProcess/WebPageProxy.cpp:
(WebKit::WebPageProxy::receivedNavigationPolicyDecision):
(WebKit::WebPageProxy::didDestroyNavigation):
- WebProcess/WebPage/WebFrame.cpp:
(WebKit::WebFrame::didReceivePolicyDecision):
(WebKit::WebFrame::documentLoaderDetached):
- WebProcess/WebPage/WebFrame.h:
- 12:48 PM Changeset in webkit [244492] by
-
- 2 edits in trunk/Tools
WebDriver: add hooks to make it possible to easily run WPT WebDriver web server by itself
https://bugs.webkit.org/show_bug.cgi?id=197125
<rdar://problem/50061468>
Reviewed by Darin Adler.
Sometimes I want to run wptserve independently of tests. Make it easier to script that
using this class rather than making our own ./wpt serve invocation.
- Scripts/webkitpy/webdriver_tests/webdriver_w3c_web_server.py:
(WebDriverW3CWebServer):
(WebDriverW3CWebServer.document_root): Expose this to clients.
(WebDriverW3CWebServer.wait): Add a method that blocks until the server dies.
(WebDriverW3CWebServer.enter):
(WebDriverW3CWebServer.exit):
Make it possible to use the web server with a 'with' statement.
Apr 20, 2019:
- 5:03 PM Changeset in webkit [244491] by
-
- 3 edits2 adds in trunk
REGRESSION (r243137): SVGViewElement.viewTarget should not return a new object
https://bugs.webkit.org/show_bug.cgi?id=197137
Reviewed by Darin Adler.
Source/WebCore:
All the DOM objects accessing the viewTarget of the same SVGViewElement
should hold a Ref pointer to the same SVGStringList property.
Test: svg/dom/SVGViewElement-viewTarget.html
- svg/SVGViewElement.idl:
LayoutTests:
- svg/dom/SVGViewElement-viewTarget-expected.txt: Added.
- svg/dom/SVGViewElement-viewTarget.html: Added.
- 4:46 PM Changeset in webkit [244490] by
-
- 2 edits in trunk/Source/WebCore
REGRESSION(r243958): Unnecessary deactivation of AudioSession (PLT Regression)
https://bugs.webkit.org/show_bug.cgi?id=197123
<rdar://problem/49783264>
Reviewed by Per Arne Vollan.
Only set m_becameActive if we actually activated the AudioSession before starting playback. This
avoids unnecessarily deactivating the AudioSession in processWillSuspend().
- platform/audio/PlatformMediaSessionManager.cpp:
(WebCore::PlatformMediaSessionManager::sessionWillBeginPlayback):
- 12:16 PM Changeset in webkit [244489] by
-
- 2 edits in trunk/Source/WTF
HashTable::removeIf always shrinks the hash table by half even if there is nothing left
https://bugs.webkit.org/show_bug.cgi?id=196681
<rdar://problem/49917764>
Reviewed by Darin Adler.
Address Darin's comments by removing the explicit type from std::max.
- wtf/HashTable.h:
(WTF::KeyTraits>::computeBestTableSize):
(WTF::KeyTraits>::shrinkToBestSize):
- 12:00 PM Changeset in webkit [244488] by
-
- 12 edits in trunk
[CMake][Win] Use target oriented design for WebKitLegacy
https://bugs.webkit.org/show_bug.cgi?id=197112
Reviewed by Konstantin Tokarev.
.:
Override WebKitLegacy_DERIVED_SOURCES_DIR within WinCairo in support of a target
oriented design.
- Source/cmake/OptionsWinCairo.cmake:
Source/WebKitLegacy:
Enumerate the public framework headers for WebKitLegacy and copy them. Include
directories should be PRIVATE. Migrate to using WebKitLegacy_DERIVED_SOURCES_DIR.
- CMakeLists.txt:
- PlatformMac.cmake:
- PlatformWin.cmake:
Source/WebKitLegacy/win:
Fix an include directory.
- WebKitQuartzCoreAdditions/CMakeLists.txt:
Tools:
Fix includes and libraries for DumpRenderTree after changes to WebKitLegacy targets.
- DumpRenderTree/CMakeLists.txt:
- DumpRenderTree/PlatformWin.cmake:
- DumpRenderTree/win/EventSender.cpp:
- 11:44 AM Changeset in webkit [244487] by
-
- 2 edits in trunk/Tools
Update text for old EWS submit button
https://bugs.webkit.org/show_bug.cgi?id=197140
Reviewed by Alexey Proskuryakov.
- QueueStatusServer/templates/statusbubble.html:
- 11:31 AM Changeset in webkit [244486] by
-
- 2 edits in trunk/Tools
[ews-app] New EWS should poll bugzilla more frequently
https://bugs.webkit.org/show_bug.cgi?id=197138
Reviewed by Alexey Proskuryakov.
- BuildSlaveSupport/ews-app/ews/fetcher.py:
(FetchLoop.init):
- 11:30 AM Changeset in webkit [244485] by
-
- 3 edits in trunk/Tools
[ews-app] Submit to EWS buttons are not aligned properly with status-bubbles
https://bugs.webkit.org/show_bug.cgi?id=197139
Reviewed by Alexey Proskuryakov.
- BuildSlaveSupport/ews-app/ews/templates/statusbubble.html:
- QueueStatusServer/templates/statusbubble.html:
- 9:44 AM Changeset in webkit [244484] by
-
- 5 edits in trunk
Unreviewed, fix iOS build with recent SDKs.
Source/WebKit:
- UIProcess/API/Cocoa/WKWebView.mm:
(deviceOrientation):
- UIProcess/ios/fullscreen/WKFullScreenViewController.mm:
(-[WKFullScreenViewController initWithWebView:]):
(-[WKFullScreenViewController viewWillTransitionToSize:withTransitionCoordinator:]):
(-[WKFullScreenViewController _statusBarFrameDidChange:]):
Tools:
- TestWebKitAPI/ios/DragAndDropSimulatorIOS.mm:
- 12:08 AM Changeset in webkit [244483] by
-
- 2 edits in trunk/Source/JavaScriptCore
Unreviewed, suppress warnings in non Darwin environments
- jit/ExecutableAllocator.cpp:
(JSC::dumpJITMemory):
