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

Issue 818727 link

Starred by 1 user

Issue metadata

Status: Duplicate
Merged: issue 818388
Owner: ----
Closed: Mar 2018
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Windows
Pri: 1
Type: Bug



Sign in to add a comment

GPU Process Startup Fails on chromium.gpu/Win10 Debug (NVIDIA)

Project Member Reported by kylec...@chromium.org, Mar 5 2018

Issue description

A number of test targets have started failing on Win10 Debug (NVIDIA) over the weekend. It looks like GpuDataManagerImpl::GpuAccessAllowed() is returning false and the GPU process isn't being started.

Failing Tests:
context_lost_tests
depth_capture_tests
gpu_process_launch_tests
hardware_accelerated_feature_tests
info_collection_tests
maps_pixel_test
pixel_test
screenshot_sync_tests
trace_test
webgl_conformance_tests
vr_browser_tests

The first failing build is here: https://ci.chromium.org/buildbot/chromium.gpu/Win10%20Debug%20%28NVIDIA%29/719

In all of the test logs I've looked it has the following log statement which will stop the GPU process from launching.

[700:8020:0302/144605.806:ERROR:gpu_process_host.cc(465)] !GpuDataManagerImpl::GpuAccessAllowed()

None of the CLs in the revision list look like the cause.
 

Comment 1 by kbr@chromium.org, Mar 5 2018

Cc: zmo@chromium.org geoffl...@chromium.org jmad...@chromium.org
Jamie, Geoff, Mo: any ideas why this bot would have suddenly started to fail to launch the GPU process?

Comment 2 by kbr@chromium.org, Mar 5 2018

Cc: iannucci@chromium.org
Currently comparing these runs:

https://ci.chromium.org/buildbot/chromium.gpu/Win10%20Debug%20%28NVIDIA%29/718
and this working shard:
https://chromium-swarm.appspot.com/task?id=3c0193806f7e2a10&refresh=10&show_raw=1

to this failing run:

https://ci.chromium.org/buildbot/chromium.gpu/Win10%20Debug%20%28NVIDIA%29/719
and this failing shard:
https://chromium-swarm.appspot.com/task?id=3c01b4c777793410&refresh=10&show_raw=1

This doesn't seem to be caused by recent Infra changes (the vpython invocation looks similar in both runs) and the blamelists in the jobs around this one don't look suspicious yet.

(Those both look like they have all the latest vpython changes, FWIW)

Comment 4 by zmo@chromium.org, Mar 5 2018

Mergedinto: 818388
Status: Duplicate (was: Untriaged)

Sign in to add a comment