http/tests/media/modern-media-controls/time-control/10-hours-or-more.html Is flakey timing out and flakey crashing in Big Sur and Catalina on Wk1 and Wk2 in Debug and Release. HISTORY URL: https://results.webkit.org/?suite=layout-tests&test=http%2Ftests%2Fmedia%2Fmodern-media-controls%2Ftime-control%2F10-hours-or-more.html TIMEOUT DIFF URL: https://build.webkit.org/results/Apple-BigSur-Release-WK2-Tests/r272657%20(85)/http/tests/media/modern-media-controls/time-control/10-hours-or-more-pretty-diff.html TIMEOUT DIFF TEXT: --- /Volumes/Data/slave/bigsur-release-tests-wk2/build/layout-test-results/http/tests/media/modern-media-controls/time-control/10-hours-or-more-expected.txt +++ /Volumes/Data/slave/bigsur-release-tests-wk2/build/layout-test-results/http/tests/media/modern-media-controls/time-control/10-hours-or-more-actual.txt @@ -1,11 +1,5 @@ -Testing the display of time labels with a video that is over 10 hours. +#PID UNRESPONSIVE - WebKitTestRunner (pid 23914) +FAIL: Timed out waiting for notifyDone to be called -On success, you will see a series of "PASS" messages, followed by "TEST COMPLETE". - - -PASS mediaController.controls.timeControl.elapsedTimeLabel.element.textContent became "00:00:05" - -PASS successfullyParsed is true - -TEST COMPLETE - +#EOF +#EOF CRASH LOG URL: https://build.webkit.org/results/Apple-BigSur-Debug-WK1-Tests/r272647%20(65)/http/tests/media/modern-media-controls/time-control/10-hours-or-more-crash-log.txt CRASH LOG TEXT: No crash log found for DumpRenderTree:9549. stdout: stderr: ERROR: Invalid Content-Type string '' ./platform/network/ParsedContentType.cpp(236) : bool WebCore::ParsedContentType::parseContentType(WebCore::Mode) ASSERTION FAILED: _dataTasks.contains(task) /Volumes/Data/slave/bigsur-debug/build/Source/WebCore/platform/network/cocoa/WebCoreNSURLSession.mm(270) : -[WebCoreNSURLSession taskCompleted:] 1 0x110786b69 WTFCrash 2 0x12d6323cb WTFCrashWithInfo(int, char const*, char const*, int) 3 0x132b126fc -[WebCoreNSURLSession taskCompleted:] 4 0x132b2fdaa -[WebCoreNSURLSessionDataTask _resource:loadFinishedWithError:metrics:]::$_17::operator()() const::'lambda'()::operator()() const 5 0x132b2fc8e WTF::Detail::CallableWrapper<-[WebCoreNSURLSessionDataTask _resource:loadFinishedWithError:metrics:]::$_17::operator()() const::'lambda'(), void>::call() 6 0x1107b2392 WTF::Function<void ()>::operator()() const 7 0x110834365 WTF::RunLoop::performWork() 8 0x110838c01 WTF::RunLoop::performWork(void*) 9 0x7fff20430a0c __CFRUNLOOP_IS_CALLING_OUT_TO_A_SOURCE0_PERFORM_FUNCTION__ 10 0x7fff20430974 __CFRunLoopDoSource0 11 0x7fff204306ef __CFRunLoopDoSources0 12 0x7fff2042f121 __CFRunLoopRun 13 0x7fff2042e6ce CFRunLoopRunSpecific 14 0x10d1bfada runTest(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&) 15 0x10d1bed4a runTestingServerLoop() 16 0x10d1be578 dumpRenderTree(int, char const**) 17 0x10d1c0472 DumpRenderTreeMain(int, char const**) 18 0x10d2a03e2 main 19 0x7fff20353621 start 20 0x2
<rdar://problem/74196227>
History shows crashing started in r272065(1/29/21) on Big Sur Release Wk2. Was able to duplicate a timeout in Big Sur Debug wk1 on r272638 using the following test: run-webkit-test http/tests/media/modern-media-controls/time-control/10-hours-or-more.html --iterations 1000 -f -1 --debug-rwt-logging RESULTS DIFF TEXT: --- /Volumes/Data/Spades/testbuild-272638/layout-test-results/http/tests/media/modern-media-controls/time-control/10-hours-or-more-expected.txt +++ /Volumes/Data/Spades/testbuild-272638/layout-test-results/http/tests/media/modern-media-controls/time-control/10-hours-or-more-actual.txt @@ -1,11 +1,8 @@ +FAIL: Timed out waiting for notifyDone to be called Testing the display of time labels with a video that is over 10 hours. On success, you will see a series of "PASS" messages, followed by "TEST COMPLETE". -PASS mediaController.controls.timeControl.elapsedTimeLabel.element.textContent became "00:00:05" -PASS successfullyParsed is true -TEST COMPLETE
r272764
This ended up having two causes: a race condition between the state change in WebCoreNSURLSession's cancel method allowing a double removal of a task, and the state machine in TaskData not having enough states to reflect 3 states of having generated a response: before a response is generated, after a response is generated but before the WebCoreNSURLSession calls the CompletionHandler, and after a response is received. I fixed both in the latest iteration, and before it could rarely survive 100 iterations on this test and now it can survive 1000.