Sunny: as GPU pixel wrangler could you please work with the ANGLE team to repro this locally on Windows so we can more quickly track down which change caused this test to start timing out?

Changes that look like they could plausibly be related:

https://chromium.googlesource.com/chromium/src/+/9e24bd35f58fff1562b0784be8ab2e612ece6408
https://chromium.googlesource.com/chromium/src/+/b547858514737d59af3c55b2b06b4e511624c304

My patch is unlikely to be related, unless there is telemetry involved in failing tests.

Telemetry is used in the failing tests.

By the way, I relanded the V8DOMWrapper change in https://codereview.chromium.org/1764693002/ .

I'm having trouble reproducing this failure. I checked out 539cb3b2 but running the tests individually doesn't seem to produce a failure. The only thing I saw that seemed odd was sometimes I got a large popup dialog saying that a task was blocked because it wasn't allowed to be created to prevent jank (or something along those lines). A popup would match the failure mode for this test (which times out).

I'll try running the tests from start to finish. Any help here would be very appreciated, as this is blocking the ANGLE CQ.

Should we try restarting the bot?

jmadill@ Can you please paste the exact text of the popup?

If there are console errors, they should show up on stdio right? If my patch is causing the failure, there will likely be some sort of security error in the console like "Blocked a frame with origin "x.y.z" from accessing a cross-origin frame." I don't see anything like that in the logs though.

The test's hanging, for unknown reasons. Possibly because of that dialog. I've never seen it. Can you post a screenshot?

This bot runs its tasks on Swarming so the tasks are run on a different machine every time. It's definitely not the case that a particular machine is stuck in some way.

By the way, you can tell the tasks run on Swarming because of all of the [trigger] steps.

jmadill@ is working on bisecting this so reassigning.

I'm reverting https://crrev.com/b547858514737d59af3c55b2b06b4e511624c304 due to suspicious performance regression. Maybe revert will fix this as well.

Thanks dgozman@. Could you watch https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20x64%20Debug%20%28NVIDIA%29 as your revert lands and see if it clears up the WebglConformance.conformance_misc_expando_loss failures? (Those are the only ones on this bot now, so it should go green.)

My revert definitely helped. Will investigate.

Thank you Dmitry. Linked the bugs together.