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

Issue 696656 link

Starred by 1 user

Issue metadata

Status: Verified
Owner:
OOO until 2019-01-24
Closed: Feb 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: All
Pri: 1
Type: Bug-Regression

Blocked on:
issue 636153



Sign in to add a comment

Switch to typ caused GPU info to disappear from stdio

Project Member Reported by kainino@chromium.org, Feb 27 2017

Issue description

The GPU info should be available in the test output, but it is not.

Bad:
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win10%20Release%20%28Intel%20HD%20530%29/builds/276/steps/webgl2_conformance_tests/logs/stdio

Good:
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win10%20Release%20%28Intel%20HD%20530%29/builds/240/steps/webgl2_conformance_tests/logs/stdio

(INFO) 2017-02-22 13:54:38,505 browser._LogBrowserInfo:123  GPU device 0: VENDOR = 0x8086 (Intel), DEVICE = 0x1912
(INFO) 2017-02-22 13:54:38,505 browser._LogBrowserInfo:125  GPU Attributes:
(...)
 

Comment 1 by kbr@chromium.org, Feb 27 2017

Cc: dpranke@chromium.org nedngu...@google.com
Labels: -Type-Bug Build-Tools-TYP Type-Bug-Regression
Status: Available (was: Untriaged)
More info: it looks like this information was displayed via logging, and the log verbosity for the typ sub-processes is set to less than it used to be. Here are the logs from the smaller context_lost tests on the builds before the typ switch:
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win10%20Release%20%28Intel%20HD%20530%29/builds/240

and after:
https://build.chromium.org/p/chromium.gpu.fyi/builders/Win10%20Release%20%28Intel%20HD%20530%29/builds/276

Dirk, Ned, how do we increase the logging level in typ's sub-processes? src/third_party/catapult/telemetry/telemetry/testing/run_browser_tests.py seems to set:
  runner.args.verbose = options.verbose

but that's not increasing the log level to "INFO".

context-lost-log-before-typ.txt
128 KB View Download
context-lost-log-after-typ.txt
97.0 KB View Download
I think maybe just change your command line args from `-v` to `-vv` should do the trick? typ has multiple levels of logging.

Though, I'm not sure if (or how) the logging handler configs get propagated down to the subprocesses, particularly on Windows, so there might be something to change there.
I think the problem is that logging isn't being configured at all in the typ-based code; it got lost during the switchover.

Looking at 

https://codereview.chromium.org/2708403004/diff/1/telemetry/telemetry/testing/browser_test_runner.py

there was a call to logging.basicConfig() in the old code, but not in the new code AFAICT (though I didn't look super-hard).

Comment 4 by kbr@chromium.org, Feb 27 2017

Owner: kbr@chromium.org
Status: Assigned (was: Available)
Thanks Dirk. That's a great catch.

Let me try fixing this.

Comment 5 by kbr@chromium.org, Feb 27 2017

Status: Started (was: Assigned)
Project Member

Comment 6 by bugdroid1@chromium.org, Feb 28 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/e3a08e0c72b10c81f8fdfe3420871f265c6a66af

commit e3a08e0c72b10c81f8fdfe3420871f265c6a66af
Author: catapult-deps-roller <catapult-deps-roller@chromium.org>
Date: Tue Feb 28 02:14:14 2017

Roll src/third_party/catapult/ 0545e3502..4f3d6da8a (1 commit).

https://chromium.googlesource.com/external/github.com/catapult-project/catapult.git/+log/0545e3502780..4f3d6da8af75

$ git log 0545e3502..4f3d6da8a --date=short --no-merges --format='%ad %ae %s'
2017-02-27 kbr Fix logging after switch to typ in browser_test_runner.

Created with:
  roll-dep src/third_party/catapult
BUG= 696656 

Documentation for the AutoRoller is here:
https://skia.googlesource.com/buildbot/+/master/autoroll/README.md

If the roll is causing failures, see:
http://www.chromium.org/developers/tree-sheriffs/sheriff-details-chromium#TOC-Failures-due-to-DEPS-rolls

CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.android:android_optional_gpu_tests_rel
TBR=catapult-sheriff@chromium.org

Review-Url: https://codereview.chromium.org/2718163003
Cr-Commit-Position: refs/heads/master@{#453456}

[modify] https://crrev.com/e3a08e0c72b10c81f8fdfe3420871f265c6a66af/DEPS

Comment 7 by kbr@chromium.org, Feb 28 2017

Status: Verified (was: Started)
The above fix is rolling out. Example build containing it:

https://build.chromium.org/p/chromium.gpu.fyi/builders/Linux%20Release%20%28NVIDIA%29/builds/47423

Excerpt from https://build.chromium.org/p/chromium.gpu.fyi/builders/Linux%20Release%20%28NVIDIA%29/builds/47423/steps/context_lost_tests%20on%20NVIDIA%20GPU%20on%20Linux%20on%20Ubuntu/logs/stdio :

(INFO) 2017-02-27 18:56:23,460 browser._LogBrowserInfo:116  OS: linux trusty
(INFO) 2017-02-27 18:56:23,504 browser._LogBrowserInfo:123  GPU device 0: VENDOR = 0x10de (Nvidia), DEVICE = 0x104a
(INFO) 2017-02-27 18:56:23,504 browser._LogBrowserInfo:125  GPU Attributes:
(INFO) 2017-02-27 18:56:23,504 browser._LogBrowserInfo:127    amd_switchable      : False
(INFO) 2017-02-27 18:56:23,504 browser._LogBrowserInfo:127    basic_info_state    : 1
(INFO) 2017-02-27 18:56:23,504 browser._LogBrowserInfo:127    context_info_state  : 1
(INFO) 2017-02-27 18:56:23,504 browser._LogBrowserInfo:127    direct_rendering    : True
(INFO) 2017-02-27 18:56:23,504 browser._LogBrowserInfo:127    driver_date         : 
(INFO) 2017-02-27 18:56:23,504 browser._LogBrowserInfo:127    driver_vendor       : NVIDIA
(INFO) 2017-02-27 18:56:23,504 browser._LogBrowserInfo:127    driver_version      : 367.27
...

Sign in to add a comment