New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 829306 link

Starred by 5 users

Issue metadata

Status: Fixed
Owner:
Closed: Apr 2018
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 2
Type: Bug



Sign in to add a comment

FATAL:host_frame_sink_manager.cc(191): Check failed: parent_data.IsFrameSinkRegistered()

Project Member Reported by kbr@chromium.org, Apr 5 2018

Issue description

Flake seen here:
https://ci.chromium.org/buildbot/tryserver.chromium.chromiumos/linux-chromeos-rel/93480

MediaEngagementAutoplayBrowserTest.UsePreloadedData_Denied/0


[6475:6475:0404/205110.076953:FATAL:host_frame_sink_manager.cc(191)] Check failed: parent_data.IsFrameSinkRegistered().
#0 0x0000043a974c base::debug::StackTrace::StackTrace()
#1 0x0000043c457b logging::LogMessage::~LogMessage()
#2 0x0000062d1719 viz::HostFrameSinkManager::RegisterFrameSinkHierarchy()
#3 0x000002e5c662 content::OffscreenCanvasSurfaceImpl::CreateCompositorFrameSink()
#4 0x000002e5b450 content::OffscreenCanvasProviderImpl::CreateCompositorFrameSink()
#5 0x0000030714d0 blink::mojom::OffscreenCanvasProviderStubDispatch::Accept()
#6 0x00000545fd86 mojo::InterfaceEndpointClient::HandleValidatedMessage()
#7 0x00000545f976 mojo::FilterChain::Accept()
#8 0x0000054610c5 mojo::InterfaceEndpointClient::HandleIncomingMessage()
#9 0x000005466dac mojo::internal::MultiplexRouter::ProcessIncomingMessage()
#10 0x0000054661b0 mojo::internal::MultiplexRouter::Accept()
#11 0x00000545f976 mojo::FilterChain::Accept()
#12 0x00000545db43 mojo::Connector::ReadSingleMessage()
#13 0x00000545e461 mojo::Connector::ReadAllAvailableMessages()
#14 0x00000545e309 mojo::Connector::OnHandleReadyInternal()
#15 0x0000028f36c7 mojo::SimpleWatcher::DiscardReadyState()
#16 0x000004a6d036 mojo::SimpleWatcher::OnHandleReady()
#17 0x000004a6d541 _ZN4base8internal7InvokerINS0_9BindStateIMN4mojo13SimpleWatcherEFvijRKNS3_18HandleSignalsStateEEJNS_7WeakPtrIS4_EEijS5_EEEFvvEE7RunImplIRKS9_RKNSt3__15tupleIJSB_ijS5_EEEJLm0ELm1ELm2ELm3EEEEvOT_OT0_NSI_16integer_sequenceImJXspT1_EEEE
#18 0x0000043a9f24 base::debug::TaskAnnotator::RunTask()
#19 0x0000044634c9 base::internal::IncomingTaskQueue::RunTask()
...

Caused failure of browser_tests on linux_chromeos_rel trybot.

If this is also related to the viz_browser_tests flakes on Windows ( Issue 829300 ) – doubtful – then this is P1. There are also viz_browser_tests flakes seen on linux_chromeos_rel :

https://ci.chromium.org/buildbot/tryserver.chromium.chromiumos/linux-chromeos-rel/?limit=200

 

Comment 2 by dmu...@chromium.org, Apr 13 2018

Owner: beccahughes@chromium.org
I see flakes on linux_chromeos_rel.

Assigning to beccahughes@, author of the test.

Change that added the test:
https://chromium.googlesource.com/chromium/src/+/eb5a05ff17880a983768affbf3d16906d796ef7e

Comment 3 by dmu...@chromium.org, Apr 13 2018

Status: Assigned (was: Untriaged)

Comment 4 by kbr@chromium.org, Apr 13 2018

Cc: fsam...@chromium.org kylec...@chromium.org
I think the Viz team ultimately needs to fix this assert.

Comment 5 by noel@chromium.org, Apr 17 2018

Just fixing flaky FileMangerBrowserTests on  issue 831074 , noticed this failure in browser_tests (with patch) linux_chromeos_rel, example 

https://ci.chromium.org/buildbot/tryserver.chromium.chromiumos/linux-chromeos-rel/103809 see browser_tests (with patch) stdout (attached).

Delete/FileManagerBrowserTest.Test/1 test internally PASSED (Yay!) but hit this check on exit, sadly.



bug-829306-browser-tests-with-patch.txt.log
930 KB View Download
Project Member

Comment 6 by bugdroid1@chromium.org, Apr 17 2018

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/88055afd28f7401e8d10654f7fea850cf714b3e0

commit 88055afd28f7401e8d10654f7fea850cf714b3e0
Author: kylechar <kylechar@chromium.org>
Date: Tue Apr 17 21:52:28 2018

Check if parent FrameSinkId is registered.

Add a check that the parent FrameSinkId is registered before creating an
OffscreeCanvasSurfaceImpl, which is really a generic embedded surface
and not just used for offscreen canvas. The IPC channel that controls
the destruction of the parent CompositorFrameSink, and subsequent
invalidation of parent FrameSinkId, is different than the IPC channel
used to create embedded surfaces. We can't rely on ordering of messages
between them. It's possible the parent FrameSinkId is invalidated before
the browser gets the request to create an embedded surface.

If this state is detected then just drop the request for the embedded
surface. This will avoid triggering the DCHECK that the parent
FrameSinkId is registered and avoid doing work for something that is
about to be destroyed.

The experiment to enable UseSurfaceLayerForVideo, which uses
OffscreenCanvasSurfaceImpl, had tests running into this problem and
hitting a DCHECK in HostFrameSinkManager::RegisterFrameSinkHierarchy().

Bug:  830003 ,  829306 
Cq-Include-Trybots: luci.chromium.try:android_optional_gpu_tests_rel
Change-Id: I4e8cb26548444ecbf941a87a984b21aa4c846580
Reviewed-on: https://chromium-review.googlesource.com/1014330
Commit-Queue: kylechar <kylechar@chromium.org>
Reviewed-by: Fady Samuel <fsamuel@chromium.org>
Cr-Commit-Position: refs/heads/master@{#551489}
[modify] https://crrev.com/88055afd28f7401e8d10654f7fea850cf714b3e0/components/viz/host/host_frame_sink_manager.cc
[modify] https://crrev.com/88055afd28f7401e8d10654f7fea850cf714b3e0/components/viz/host/host_frame_sink_manager.h
[modify] https://crrev.com/88055afd28f7401e8d10654f7fea850cf714b3e0/content/browser/renderer_host/offscreen_canvas_provider_impl_unittest.cc
[modify] https://crrev.com/88055afd28f7401e8d10654f7fea850cf714b3e0/content/browser/renderer_host/offscreen_canvas_surface_impl.cc

Cc: beccahughes@chromium.org
 Issue 833945  has been merged into this issue.
Owner: kylec...@chromium.org
Status: Fixed (was: Assigned)
It looks like my CL fixed this. There are no flakes in the past week.

Comment 9 by kbr@chromium.org, Apr 27 2018

Thanks for fixing the cause of this flakiness!

Sign in to add a comment