Probably a flaky test uncovered by the driver upgrade. We should suppress and likely report it upstream to NV.

I see

[5316:864:0108/185301.281:INFO:CONSOLE(117)] "WebGL: INVALID_OPERATION: bindVertexArray: attempt to use a deleted object", source: http://127.0.0.1:49267/third_party/webgl/src/sdk/tests/conformance2/vertex_arrays/vertex-array-object.html?webglVersion=2 (117)
[5316:864:0108/185301.406:INFO:CONSOLE(197)] "should be 0,255,255,255
at (0, 0) expected: 0,255,255,255 was 0,0,0,255", source:  (197)
[5316:864:0108/185301.406:INFO:CONSOLE(197)] "FAIL should be 0,255,255,255
at (0, 0) expected: 0,255,255,255 was 0,0,0,255", source:  (197)
[5316:864:0108/185301.412:INFO:CONSOLE(588)] "WebGL: INVALID_OPERATION: drawElements: no buffer is bound to enabled attribute", source: http://127.0.0.1:49267/third_party/webgl/src/sdk/tests/conformance2/vertex_arrays/vertex-array-object.html?webglVersion=2 (588)

Ken is the "attempt to use a deleted object" an indication this issue is related to  issue 739604  ?

There was also an issue on Linux with another VAO test: bug 920033

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/dc969ea6f0d44803e555091f1e9ab430707b56a6

commit dc969ea6f0d44803e555091f1e9ab430707b56a6
Author: Yuly Novikov <ynovikov@chromium.org>
Date: Wed Jan 09 19:08:51 2019

Suppress two WebGL2 tests

conformance2/uniforms/incompatible-texture-type-for-sampler.html
Skip on Linux AMD
conformance2/vertex_arrays/vertex-array-object.html
Flaky on Win Nvidia GL passthrough

TBR=kbr@chromium.org

Bug:  809237 , 920265
Change-Id: I27e65f987284ae05741617e3a7ac62ec745caa7a
Reviewed-on: https://chromium-review.googlesource.com/c/1403322
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Commit-Queue: Yuly Novikov <ynovikov@chromium.org>
Cr-Commit-Position: refs/heads/master@{#621251}
[modify] https://crrev.com/dc969ea6f0d44803e555091f1e9ab430707b56a6/content/test/gpu/gpu_tests/webgl2_conformance_expectations.py

Yes, this is probably the same problem as Issue 920033. I thought it was strange that it would affect only ANGLE's implementation of the OES_vertex_array_object extension on ES 2.0 contexts.

Need to see if this is reproducible locally. When testing locally on Linux it didn't reproduce easily.

Issue 920033 reproduces fairly easily locally on both Windows and Linux just by reloading the page, but this failure doesn't. We should debug Issue 920033; it will probably fix this bug too.