Here is one sample run:

https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Release%20%28New%20Intel%29/builds/832

Is this related with the context lost issue on Win Intel?

Not sure whether these are the same issue. I'm going to close the other one as fixed and block this one on the earlier bug since the OUT_OF_MEMORY issue doesn't seem to be visibly happening any more. It looks like the device is being lost partway through the conformance suite run.

Looking into some of the failures like this one:
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Release%20%28New%20Intel%29/builds/831/steps/webgl2_conformance_tests/logs/stdio

there are errors inside ANGLE like:
err: rx::TextureStorage11_2DArray::createSRV(3288): 	! Assert failed in rx::TextureStorage11_2DArray::createSRV(3288): result == E_OUTOFMEMORY || SUCCEEDED(result)

and some like:
(INFO) 2016-06-04 21:34:29,266 cache_temperature.EnsurePageCacheTemperature:55  PageCacheTemperature: any
err: rx::Renderer11::testDeviceLost(2637): The D3D11 device was removed: 0x887A0006
[1168:2744:0604/213432:ERROR:gles2_cmd_decoder.cc(3802)]   GLES2DecoderImpl: Context lost during MakeCurrent.
[1168:2744:0604/213432:ERROR:gpu_command_buffer_stub.cc(384)] Context lost because MakeCurrent failed.

It looks like the D3D11 device is being lost during the test run. Is it possible that Chrome' leaking graphics resources and this is manifesting as errors inside of ANGLE? ANGLE team, would you have any suggestions on Windows-specific tools that could confirm this hypothesis?

Ken, those tests fail on ANGLE, they should be marked as Skip for now probably if they're giving ASSERTs. Looking in ANGLE dEQP test expectations the native version of the test you referenced is marked as fail always.

Thanks for pointing that out Jamie. Digging more deeply into https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20Release%20%28New%20Intel%29/builds/831/steps/webgl2_conformance_tests/logs/stdio (attached below), while the test:
deqp_functional_gles3_texturespecification_teximage3d_depth_pbo

did fail, it's the sub-test:
specification.teximage3d_depth_pbo.depth24_stencil8_2d_array

which seems to have triggered the assertion failure.

That one doesn't seem to be marked in https://cs.chromium.org/chromium/src/third_party/angle/src/tests/deqp_support/deqp_gles3_test_expectations.txt?q=deqp_gles3_test_expectations.txt&sq=package:chromium&dr .

Deleted: stdout.txt 3.9 MB

stdout.txt 3.9 MB View Download

Something fishy is definitely up with this test - I ran it locally and began to see some crazy visual corruption. It's possible the expected failures in the first four or five sub tests (which are listed in the file you referenced in comment 5) fail in such a way that they screw up the system and later trigger ASSERTs in other subtests. For now I think we should mark it as skip and the ANGLE folks can investigate down the line (though any help of course would be appreciated).

Thanks Jamie for your analysis and the suggestion. Skipping the two affected tests in https://codereview.chromium.org/2048453003 .

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/23be85976761014afa6a877026383171c7b1c499

commit 23be85976761014afa6a877026383171c7b1c499
Author: kbr <kbr@chromium.org>
Date: Tue Jun 07 08:04:10 2016

Skip two WebGL 2.0 conformance tests on Win Intel.

The following two tests are causing graphical corruption and other
strange behaviors:

  WebglConformance.conformance2_textures_misc_copy_texture_image
  WebglConformance.deqp_functional_gles3_texturespecification_teximage3d_depth_pbo

BUG= 617449 
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_optional_gpu_tests_rel;tryserver.chromium.mac:mac_optional_gpu_tests_rel;tryserver.chromium.win:win_optional_gpu_tests_rel
TBR=jmadill@chromium.org, zmo@chromium.org

Review-Url: https://codereview.chromium.org/2048453003
Cr-Commit-Position: refs/heads/master@{#398241}

[modify] https://crrev.com/23be85976761014afa6a877026383171c7b1c499/content/test/gpu/gpu_tests/webgl2_conformance_expectations.py

Ping. Should this be Pri-1? If so please assign a milestone and find an owner.

The above expectations
  WebglConformance.conformance2_textures_misc_copy_texture_image
  WebglConformance.deqp_functional_gles3_texturespecification_teximage3d_depth_pbo

are NOT failing for me on Win 10 Intel HD 620, so perhaps it's been fixed. (I think it still fails on the waterfall on Win10 Release (New Intel).)

I realized these are Skip, not Fail, so they wouldn't show up on the waterfall, sorry.

However, I also ran these as part of a full conformance run and I didn't see any weirdness from them.

Need to check if this is still happening.

+sugoi as pixel wrangler: please note I'm landing an experimental removal of these Skip entries. If they start failing/flaking please just revert.
https://chromium-review.googlesource.com/c/chromium/src/+/815946

(specifially these)
WebglConformance.conformance2_textures_misc_copy_texture_image
WebglConformance.deqp_functional_gles3_texturespecification_teximage3d_depth_pbo

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/939f1d088a55400b66d928371383f6ac76357c72

commit 939f1d088a55400b66d928371383f6ac76357c72
Author: Kai Ninomiya <kainino@chromium.org>
Date: Fri Dec 08 05:27:29 2017

Try removing old Win/Intel expectations

TBR=kbr@chromium.org

Bug:  617449 
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
Change-Id: I800d6a9ec495ccdf771e8c00752b0eeb1d39e61a
Reviewed-on: https://chromium-review.googlesource.com/815946
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Cr-Commit-Position: refs/heads/master@{#522717}
[modify] https://crrev.com/939f1d088a55400b66d928371383f6ac76357c72/content/test/gpu/gpu_tests/webgl2_conformance_expectations.py

Seems to be working.