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

Issue 884783 link

Starred by 2 users
Status: Fixed
Owner: ----
Closed: Sep 24
Cc:
atotic@chromium.org
pdr@chromium.org
tkent@chromium.org
vmp...@chromium.org
█ nednguyen@chromium.org
dpranke@chromium.org
martiniss@chromium.org
robertma@chromium.org
khushals...@chromium.org
chrishtr@chromium.org
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 1
Type: Bug-Regression



Sign in to add a comment

Result merge/upload is broken in webkit_layout_tests

Project Member Reported by robertma@chromium.org, Sep 17 
Today I start to notice that the webkit_layout_tests is uploading almost empty results.

Example: https://ci.chromium.org/p/chromium/builders/luci.chromium.try/linux_chromium_rel_ng/188279

In the "webkit_layout_tests (with patch)" step, clearly tests were executed and failures were found, but the archived zip is almost empty (with only a couple of tests) and the result viewer is therefore broken.

https://test-results.appspot.com/data/layout_results/linux_chromium_rel_ng/188279/layout-test-results.zip

Interestingly, the "site_per_process_webkit_layout_tests (with patch)" doesn't seem to be affected.

This breaks a lot of tools that rely on the test results, including rebaseline-cl and wpt-importer.

Comment 1 by robertma@chromium.org, Sep 17 

I'm not 100% sure that archive_webkit_tests_results is the culprit; it's just my best guess based on the fact that all shards (and their isolated outputs) seem fine but the merged result isn't.

I don't find anything obvious in the log though:

https://logs.chromium.org/logs/chromium/buildbucket/cr-buildbucket.appspot.com/8935127583314989040/+/steps/archive_webkit_tests_results/0/stdout

Stephen, did you work on the merge/upload step recently? Can you take a look?

Comment 2 by martiniss@chromium.org, Sep 17 

I did touch it recently, in https://crrev.com/c/1217594. It shouldn't have affected this though. It does make the log less verbose, which could make it harder to debug this problem though.

Comment 3 by martiniss@chromium.org, Sep 17 

It looks like maybe the merge script isn't getting all the paths it should? From https://logs.chromium.org/logs/chromium/buildbucket/cr-buildbucket.appspot.com/8935127583314989040/+/steps/webkit_layout_tests__with_patch_/0/stdout

2018-09-17 08:07:52,183 - blinkpy.web_tests.merge_results: [DEBUG] Merging following paths:
2018-09-17 08:07:52,183 - blinkpy.web_tests.merge_results: [DEBUG] ['/b/swarming/w/ir/tmp/t/tmpu1nHvX/0',
 '/b/swarming/w/ir/tmp/t/tmpu1nHvX/1',
 '/b/swarming/w/ir/tmp/t/tmpu1nHvX/2',
 '/b/swarming/w/ir/tmp/t/tmpu1nHvX/3',
 '/b/swarming/w/ir/tmp/t/tmpu1nHvX/4',
 '/b/swarming/w/ir/tmp/t/tmpu1nHvX/5']

There are 10 total shards for that build.

Comment 4 by martiniss@chromium.org, Sep 17 

Oh wait, my bad. I was looking at site per process, the number is correct.

Comment 5 by martiniss@chromium.org, Sep 17 

Ah, this is https://crrev.com/c/1222176. I'm reverting this change, things should get better.

Comment 6 by pdr@chromium.org, Sep 17

Cc: chrishtr@chromium.org robertma@chromium.org dpranke@chromium.org pdr@chromium.org tkent@chromium.org vmp...@chromium.org
 Issue 884860  has been merged into this issue.

Comment 7 by pdr@chromium.org, Sep 17

Cc: khushals...@chromium.org atotic@chromium.org
 Issue 884891  has been merged into this issue.

Comment 9 by robertma@chromium.org, Sep 18 

Re #8: did the problem just go away itself? Do we know what happened?

Comment 10 by martiniss@chromium.org, Sep 24

Status: Fixed (was: Untriaged)
I reverted the CL in #5, which fixed it.
►

Sign in to add a comment