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

Issue 784527 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Last visit > 30 days ago
Closed: Jan 2018
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Android
Pri: 2
Type: Bug-Regression

Blocking:
issue 774267



Sign in to add a comment

stack_tool_for_tombstones hangs on Nexus 5 GPU.FYI bot

Project Member Reported by ynovikov@chromium.org, Nov 13 2017

Issue description

Status: Assigned (was: Untriaged)

Comment 2 by kbr@chromium.org, Nov 13 2017

Blocking: 774267

Comment 3 by hzl@chromium.org, Nov 13 2017

I just reverted the suspicious cl.

Comment 4 by hzl@chromium.org, Nov 13 2017

Waiting for the revert to land. Why that bot is all green, but have so many tombstones?
It could be because we have some tests which are expected to fail, not sure.

Comment 6 by hzl@chromium.org, Nov 13 2017

Cc: kbr@chromium.org ynovikov@chromium.org
If you take a look at webgl_conformance_tests and search for "Build fingerprint" you will see about 12 crashes. And the bot still appears green. Is that intentional?

https://logs.chromium.org/v/?s=chromium%2Fbb%2Fchromium.gpu.fyi%2FAndroid_Release__Nexus_5_%2F8044%2F%2B%2Frecipes%2Fsteps%2Fwebgl_conformance_tests%2F0%2Fstdout

These crashes created the tombstones and the stack script hangs when symbolizing the tombstones. Of course, the stack script should never hang when symbolizing the tombstones, but I am wondering whether the about 12 crashes seen in webgl_conformance_tests are expected?


Comment 7 by kbr@chromium.org, Nov 13 2017

Yes, they're expected. See for example this suppression:
https://cs.chromium.org/chromium/src/content/test/gpu/gpu_tests/webgl_conformance_expectations.py?q=webgl_conformance_expectations.py&sq=package:chromium&l=541

which refers to  Issue 733599 . Unfortunately we didn't follow up yet to triage these failures.

Yes, I think these crashes are expected. For example, WebglConformance_conformance_glsl_bugs_qualcomm_loop_with_continue_crash which generates the first tombstone I think is for a known Qualcomm bug, the test description is:
This test exercises a crash on Adreno 300 series GPUs when compiling certain loop constructs. crbug.com/527761
Project Member

Comment 9 by bugdroid1@chromium.org, Nov 17 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/0cde78d1a55e0f99e18ef7f0ab472cda6af406eb

commit 0cde78d1a55e0f99e18ef7f0ab472cda6af406eb
Author: Zhiling Huang <hzl@chromium.org>
Date: Fri Nov 17 19:21:10 2017

Close and join pool after resolving tombstones.

Bug:  784527 
Change-Id: I6f57b53db6f3f898238efa4617e5d8e0bbbfae36
Reviewed-on: https://chromium-review.googlesource.com/776014
Reviewed-by: John Budorick <jbudorick@chromium.org>
Commit-Queue: Zhiling Huang <hzl@chromium.org>
Cr-Commit-Position: refs/heads/master@{#517508}
[modify] https://crrev.com/0cde78d1a55e0f99e18ef7f0ab472cda6af406eb/build/android/tombstones.py

Comment 10 by hzl@chromium.org, Jan 20 2018

Status: Fixed (was: Assigned)
No longer hangs.

Sign in to add a comment