Issue metadata
Sign in to add a comment
|
WebglConformance_conformance_textures_video_tex_2d_rgb_rgb_unsigned_byte flaky on Android Release (Nexus 5) |
||||||||||||||||||||||||
Issue descriptionFlakiness dashboard shows nothing for this test. This DCHECK fails: F/chromium(28798): [FATAL:frame_sink_manager_impl.cc(228)] Check failed: registered_sources_.count(source) == 0u (1 vs. 0) fsamuel@ assigning to you since you're most familiar with this. (Stack traces aren't symbolized.) But there are lots of flakes in the last 200 builds: https://build.chromium.org/p/chromium.gpu.fyi/builders/Android%20Release%20%28Nexus%205%29/builds/6684 https://build.chromium.org/p/chromium.gpu.fyi/builders/Android%20Release%20%28Nexus%205%29/builds/6662 https://build.chromium.org/p/chromium.gpu.fyi/builders/Android%20Release%20%28Nexus%205%29/builds/6658 https://build.chromium.org/p/chromium.gpu.fyi/builders/Android%20Release%20%28Nexus%205%29/builds/6590 https://build.chromium.org/p/chromium.gpu.fyi/builders/Android%20Release%20%28Nexus%205%29/builds/6582 https://build.chromium.org/p/chromium.gpu.fyi/builders/Android%20Release%20%28Nexus%205%29/builds/6548 https://build.chromium.org/p/chromium.gpu.fyi/builders/Android%20Release%20%28Nexus%205%29/builds/6543 https://build.chromium.org/p/chromium.gpu.fyi/builders/Android%20Release%20%28Nexus%205%29/builds/6538 https://build.chromium.org/p/chromium.gpu.fyi/builders/Android%20Release%20%28Nexus%205%29/builds/6535 https://build.chromium.org/p/chromium.gpu.fyi/builders/Android%20Release%20%28Nexus%205%29/builds/6505 https://build.chromium.org/p/chromium.gpu.fyi/builders/Android%20Release%20%28Nexus%205%29/builds/6502 https://build.chromium.org/p/chromium.gpu.fyi/builders/Android%20Release%20%28Nexus%205%29/builds/6498
,
Aug 8 2017
I am receiving similar error with the same DCHECK failure today with ToT on Nexus 6P device. I compiled and installed monochrome_apk and every time I open a webpage it will crash. By a brief look at the code, the call stack CompositorImpl::RequestNewLayerTreeFrameSink() -> CompositorImpl::HandlePendingLayerTreeFrameSinkRequest() -> CompositorImpl::OnGpuChannelEstablished() -> CompositorImpl::InitializeDisplay() -> FrameSinkManagerImpl::RegisterFrameSinkManagerClient() got called twice without UnRegisterFrameSinkManagerClient in between, so the DCHECK failed. But I am lost to find call-site of RequestNewLayerTreeFrameSink(). Not sure why this is considered duplicate with 752291 though, the stack looks different to me.
,
Aug 8 2017
As I mentioned in issue 752291, it is hard to correlate between logcat and test running. I think "registered_sources_.count(source) == 0u" DCHECK is unrelated to the test failing, as I also saw it in build where the test passed. Thus, I've decided that "!release_time_.is_null()" is the DCHECK that is causing the test to fail. The only way to be sure it to have the test failing locally, then logcat can be properly correlated. |
|||||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||||
Comment 1 by kbr@chromium.org
, Aug 7 2017Mergedinto: 752291
Status: Duplicate (was: Assigned)