Failing build: https://ci.chromium.org/buildbot/chromium.gpu.fyi/Linux%20Debug%20%28NVIDIA%29/41100

Swarming task: https://chromium-swarm.appspot.com/task?id=3a8e9b43250ad010&refresh=10&show_raw=1

Only useful thing I could find in the log:
[9654:9694:1220/131829.260945:ERROR:browser_gpu_channel_host_factory.cc(120)] Failed to launch GPU process.

This bot has been red for some time.

First failing build: https://ci.chromium.org/buildbot/chromium.gpu.fyi/Linux%20Debug%20%28NVIDIA%29/41072

This CL looked suspicious:

f920e13ac4b3b345e20c70d113d0874176a17106 Shift Telemetry and about:gpu to use GpuFeatureInfo.

Specifically this part:

"Also, this CL makes Telemetry crash if Chrome doesn't send GPU data back in 5s."

And the swarming logs confirm this:

https://chromium-swarm.appspot.com/task?id=3a899bed1100f910&refresh=10

[14050:14050:1219/140030.619147:FATAL:system_info_handler.cc(171)] Check failed: false. Gathering system GPU info took more than 5 seconds.

However, this is different from the failure above. We should fix this first, and then see if there are more failures.

BTW I got an email because of one my changes (not the culprit) was in the build, and the first line of the email says:

Automatically closing tree for "gpu_process_launch_tests on NVIDIA GPU on Linux on Ubuntu" on "Linux Debug (NVIDIA)"

But we don't close the tree for gpu.fyi failures, right?