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

Issue 689577 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Closed: Apr 2017
Cc:
EstimatedDays: ----
NextAction: ----
OS: Windows
Pri: 2
Type: Bug



Sign in to add a comment

Flaky DCHECK in Pixel_OffscreenCanvas2DResizeOnWorker on Win10 Release (Intel HD 530) GPU.FYI bot

Project Member Reported by ynovikov@chromium.org, Feb 7 2017

Issue description

Sample log:
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win10%20Release%20%28Intel%20HD%20530%29/builds/161/steps/pixel_test/logs/stdio

Pixel_ScissorTestWithPreserveDrawingBuffer (gpu_tests.pixel_integration_test.PixelIntegrationTest) ... (WARNING) 2017-02-07 03:02:51,861 pixel_integration_test.RestartBrowserIfNecessaryWithArgs:93  Restarting browser with arguments: []
[2340:5548:0207/030251.887:ERROR:texture_manager.cc(2119)] [.DisplayCompositor-0B29D760]GL ERROR :GL_INVALID_ENUM : glTexParameteri: param was GL_FALSE
[2340:5548:0207/030251.887:ERROR:texture_manager.cc(2119)] [.DisplayCompositor-0B29D760]GL ERROR :GL_INVALID_ENUM : glTexParameteri: param was GL_FALSE
[1508:1000:0207/030252.052:FATAL:surface_manager.cc(71)] Check failed: frame_sink_source_map_.size() == 0u (1 vs. 0)
Backtrace:
	base::debug::StackTrace::StackTrace [0x717CA3F7+55]
	base::debug::StackTrace::StackTrace [0x717CEFAA+10]
	cc::SurfaceManager::~SurfaceManager [0x7204ED2C+396]
	std::default_delete<cc::SurfaceManager>::operator() [0x7132A7FB+17]
	content::GpuProcessTransportFactory::~GpuProcessTransportFactory [0x7132A6E9+256]
	content::GpuProcessTransportFactory::`vector deleting destructor' [0x7132A88C+11]
	content::ImageTransportFactory::Terminate [0x7132CEA6+16]
	content::BrowserMainRunnerImpl::Shutdown [0x7107237D+640]
	content::BrowserMain [0x7106D3A2+139]
	content::RunNamedProcessTypeMain [0x7149C975+176]
	content::ContentMainRunnerImpl::Run [0x7149C894+280]
	content::ContentMain [0x7149BEE4+35]
	ChromeMain [0x70DF1869+180]
	MainDllLoader::Launch [0x001452F2+539]
	wWinMain [0x00142E93+373]
	__scrt_common_main_seh [0x00193308+246] (f:\dd\vctools\crt\vcstartup\src\startup\exe_common.inl:253)
	BaseThreadInitThunk [0x767E38F4+36]
	RtlUnicodeStringToInteger [0x774A5DE3+595]
	RtlUnicodeStringToInteger [0x774A5DAE+542]

Looks like this is going on for a while:
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win10%20Release%20%28Intel%20HD%20530%29/builds/20

fsamuel@, I see you've fixed something similar in  issue 663453  / 666481, could you please take a look?
 
OffscreenCanvasCompositorFrameSinkProviderImpl needs to be destroyed before SurfaceManager is destroyed. I'm going to look into moving it to GpuProcessTransportFactory, which owns SurfaceManager.

In a subsequent patch I'm going to move services/ui/surfaces/display_compositor.{h|cc} to components/display_compositor/display_compositor.{h|cc} which owns SurfaceManager. That will replace OffscreenCanvasCompositorFrameSinkProviderImpl. For this bug, I will simply try to fix the crasher.
Blocking: 601863
Cc: rjkroege@chromium.org sadrul@chromium.org kylec...@chromium.org enne@chromium.org danakj@chromium.org piman@chromium.org
Labels: displaycompositor
Owner: kylec...@chromium.org
I haven't updated this bug in a while. It might already be fixed as some DCHECKs have changed in SurfaceManager. Regardless, I've passed along the work to put "DisplayCompositor" in GpuProcessTransportFactory to kylechar@. That will definitely address any lingering lifetime concerns.
Status: Fixed (was: Assigned)
With this patch: https://codereview.chromium.org/2789753002/ this bug should be fixed because CompositorFrameSinkSupports belonging to offscreen canvas will be destroyed prior to destroying SurfaceManager. Marking as FIXED.
Blocking: -601863

Sign in to add a comment