OS: Linux

cast_shell_browsertests frequently fails with the crash shown below:

https://luci-logdog.appspot.com/v/?s=chromium%2Fbb%2Fchromium.linux%2FCast_Linux%2F43843%2F%2B%2Frecipes%2Fsteps%2Fcast_shell_browsertests%2F0%2Fstdout

[31638:31638:INFO:render_media_log.cc(33)] MediaEvent: PAUSE {}
[31638:31638:INFO:HTMLMediaElement.cpp(1868)] addPlayedRange(0x3cf1809e3620, 0.9621, 1.069)
[31638:31638:INFO:HTMLMediaElement.cpp(3597)] mediaControlsDidBecomeVisible(0x3cf1809e3620)
[31613:31613:WARNING:pref_notifier_impl.cc(23)] Pref observer found at shutdown.
BrowserTestBase received signal: Segmentation fault. Backtrace:
#0 0x0000015447d7 base::debug::StackTrace::StackTrace()
#1 0x0000033a3621 content::(anonymous namespace)::DumpStackTraceSignalHandler()
#2 0x7f6e84a49d40 <unknown>
#3 0x00000084a2b6 chromecast::media::MediaPipelineImpl::SetPlaybackRate()
#4 0x000000782afc media::mojom::RendererStubDispatch::Accept()
#5 0x0000015e05e9 mojo::internal::MultiplexRouter::ProcessIncomingMessage()
#6 0x0000015dff30 mojo::internal::MultiplexRouter::Accept()
#7 0x0000015da295 mojo::Connector::ReadSingleMessage()
#8 0x0000015da992 mojo::Connector::ReadAllAvailableMessages()
#9 0x0000007d9d40 mojo::SimpleWatcher::DiscardReadyState()
#10 0x0000015e864c mojo::SimpleWatcher::OnHandleReady()
#11 0x000001544f36 base::debug::TaskAnnotator::RunTask()
#12 0x000001559c74 base::MessageLoop::RunTask()
#13 0x00000155a2f1 base::MessageLoop::DoWork()
#14 0x00000155b4aa base::MessagePumpDefault::Run()
#15 0x000001572820 base::RunLoop::Run()
#16 0x000001591cba base::Thread::ThreadMain()
#17 0x00000158cac3 base::(anonymous namespace)::ThreadFunc()
#18 0x7f6e87759182 start_thread
#19 0x7f6e84b0d47d clone
[19/19] CastNavigationBrowserTest.VideoPlaybackMp4 (2355 ms)
1 test failed:
    CastNavigationBrowserTest.VideoPlaybackMp4 (../../chromecast/browser/test/cast_navigation_browsertest.cc:84)

Even some of the runs which claim to have passed have actually failed with this error which is somehow not detected (but not always). The failure rate is around 5% - see https://uberchromegw.corp.google.com/i/chromium.linux/builders/Cast%20Linux?numbuilds=100.

For instance, 43860 was reported as failing. 43859 was reported as succeeding but contained the same failure crash. Most of the runs that claim to have succeeded actually did.

Presumably there is some sort of race condition? The test should be made more reliable or should be disabled.