Can't be related; handle_closer_agent.cc is Windows-only.

For anyone investigating, it's this bot:

https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Debug%20%28New%20Intel%29?numbuilds=200

Here's the full stack trace.

[1916:4036:0916/170146:FATAL:handle_closer_agent.cc(89)] Check failed: dup_dummy == closed_handle. 
Backtrace:
	base::debug::StackTrace::StackTrace [0x1007B9A7+23]
	logging::LogMessage::~LogMessage [0x100CB92B+59]
	sandbox::HandleCloserAgent::AttemptToStuffHandleSlot [0x0052A164+612]
	sandbox::HandleCloserAgent::CloseHandles [0x0052A4B5+629]
	sandbox::CountedBuffer::Buffer [0x005196BE+78]
	sandbox::TargetServicesBase::LowerToken [0x00519C69+185]
	content::RendererMainPlatformDelegate::EnableSandbox [0x131C5297+55]
	content::RendererMain [0x131C4CD4+580]
	content::RunNamedProcessTypeMain [0x135B4457+135]
	content::ContentMainRunnerImpl::Run [0x135B4318+488]
	content::ContentMain [0x135B2184+100]
	ChromeMain [0x04AC2016+198]
	MainDllLoader::Launch [0x0042E0D4+916]
	wWinMain [0x00428EDD+541]
	invoke_main [0x0056931E+30] (f:\dd\vctools\crt\vcstartup\src\startup\exe_common.inl:118)
	__scrt_common_main_seh [0x00569180+336] (f:\dd\vctools\crt\vcstartup\src\startup\exe_common.inl:253)
	__scrt_common_main [0x0056901D+13] (f:\dd\vctools\crt\vcstartup\src\startup\exe_common.inl:296)
	wWinMainCRTStartup [0x00569338+8] (f:\dd\vctools\crt\vcstartup\src\startup\exe_wwinmain.cpp:17)
	BaseThreadInitThunk [0x74CE38F4+36]
	RtlUnicodeStringToInteger [0x77A45DE3+595]
	RtlUnicodeStringToInteger [0x77A45DAE+542]

Will, may I assign this to you to triage? I'm downgrading this to P2 because this is an FYI-only debug bot -- though we'd like it to be reliably green.

Actually, I'm upgrading this to P1 again so it gets triaged quickly.  Issue 467809  was the previous incidence of this problem. Thanks Yuly for raising the problem.

This test has indeed been flaky for a while, on all OS.

wfh@: could you please help us with the handle_closer_agent.cc problem here? That will address flakiness of these tests on Windows.

Still happens
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Release%20%28New%20Intel%29/builds/1700 - GpuRasterization.ConcavePaths
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Release%20%28New%20Intel%29/builds/1698 - trace_test.OffscreenCanvasWebGLGreenBox
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Release%20%28New%20Intel%29/builds/1691 - GpuProcess.has_transparent_visuals_gpu_process
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Release%20%28New%20Intel%29/builds/1686 - trace_test.WebGLGreenTriangle.AA.NoAlpha
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Release%20%28New%20Intel%29/builds/1683 - GpuProcess.equal_bug_workarounds_in_browser_and_gpu_process

I think the best option for the short term would be to remove the DCHECK - it should not be happening but it depends on the state of the handle table while the process is launching, so something that warrants deeper investigation (and might not be achievable in the short term)

If this is blocking you, I would lgtm a CL to remove the DCHECK with a TODO(wfh) to investigate why exactly this is failing on certain platform configurations.

Yuly: would it be possible for you to put together that CL? We really should get this bot reliably green.

Looking more at the failures, I suspect now that "Check failed: dup_dummy == closed_handle" happens later, when the test tries to recover after a previous problem.

In https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Debug%20%28New%20Intel%29/builds/1492
I see
err: rx::Renderer11::initializeD3DDevice(748): Failed creating Debug D3D11 device - falling back to release runtime.

In
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Debug%20%28New%20Intel%29/builds/1509
there is
[5300:5920:1026/205536:ERROR:scoped_com_initializer.h(58)] Multiple CoInitialize() calls for thread 5920

In
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Debug%20%28New%20Intel%29/builds/1510
there is
DevtoolsTargetCrashException: Devtools target crashed

I imagine what happens is that browser gets stuck in unkillable state, new browser gets launched, and then DCHECK fires.
Probably the same thing happens on all "New Intel" Windows and Linux bots, just on Win Debug we get more messages.

Ultimately, I suspect the source of the problem is what was mentioned in  issue 648369  - hardware.
Still, it would be nice if we were able to recover cleanly from problems even on slow hardware.

Thanks for your help diagnosing this Yuly.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/da7006ebeb66503c5988482e3631d43d44f5c74c

commit da7006ebeb66503c5988482e3631d43d44f5c74c
Author: jbauman <jbauman@chromium.org>
Date: Wed Nov 09 22:12:14 2016

Remove DCHECK when handle stuffing fails.

This seems to be causing a lot of flakes on the Win7 Intel GPU FYI bot.

BUG=649904
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.win:win10_chromium_x64_rel_ng

Review-Url: https://codereview.chromium.org/2484813005
Cr-Commit-Position: refs/heads/master@{#431049}

[modify] https://crrev.com/da7006ebeb66503c5988482e3631d43d44f5c74c/sandbox/win/src/handle_closer_agent.cc