1445356 - test_prompt.html fails on windows 10 with updated taskcluster worker

Status: RESOLVED WORKSFORME

(Toolkit :: Password Manager, defect)

Description

[Joel Maher ( :jmaher ) (UTC -8)]

updating our infrastructure/tooling to fix bugs/features we have a new worker which is responsible for accepting the tests to run and launching the mozharness/test script to run the tests.

The main difference in the new worker is that we login as an actual user instead of simulating it with runas and mapping directories.

While doing this we find that toolkit/components/passwordmgr/test/mochitest/test_prompt.html is failing for us consistently:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=85b4ef4fa06f4a75d9b50f8a3de2a3ecab3f7afd&filter-searchStr=10-64%20mochitest-e10s-5&selectedJob=166562571


here is what I see in a log [1]:
19:50:14     INFO -  1733 INFO TEST-START | toolkit/components/passwordmgr/test/mochitest/test_prompt.html
19:56:24     INFO -  Buffered messages logged at 19:50:14
19:56:24     INFO -  1734 INFO TEST-PASS | toolkit/components/passwordmgr/test/mochitest/test_prompt.html | Access Ci
19:56:24     INFO -  1735 INFO TEST-PASS | toolkit/components/passwordmgr/test/mochitest/test_prompt.html | Access Cc
19:56:24     INFO -  1736 INFO SpawnTask.js | Entering test setup
19:56:24     INFO -  1737 INFO SpawnTask.js | Leaving test setup
19:56:24     INFO -  1738 INFO SpawnTask.js | Entering test test_prompt_accept
19:56:24     INFO -  Buffered messages finished
19:56:24  WARNING -  TEST-UNEXPECTED-TIMEOUT | toolkit/components/passwordmgr/test/mochitest/test_prompt.html | application timed out after 370 seconds with no output
19:56:24     INFO -  1739 ERROR Force-terminating active process(es).
19:56:24     INFO -  Determining child pids from psutil...


this fails early on in the test, not sure if it is broken 100%.  When I disable this test and push to try again, I see the next test test_prompt_http.html fail with a timeout:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=8c9e190b6376aed6b6b3167f03e36da8ce61d6c0&selectedJob=167715473&filter-searchStr=mochitest-e10s-5



[1] https://taskcluster-artifacts.net/XVvQIqxzSSK-_yABwrpuUQ/0/public/logs/live_backing.log

Comment 1

[Joel Maher ( :jmaher ) (UTC -8)]

:mattn, I don't think we can realistically disable our way to success.  Do you have ideas on how to debug this or what might be going wrong?

Comment 2

[Matthew N. [:MattN]]

Do I understand correctly from bug 1399401 comment 85 that this was fixed? I wonder if that also fixed all the alert service tests that got disabled: https://dxr.mozilla.org/mozilla-central/rev/00bdc9451be6557ccce1492b9b966d4435615380/browser/base/content/test/alerts/browser.ini#7,9,11,13,15,17

Comment 3

[Joel Maher ( :jmaher ) (UTC -8)]

I am testing this on try right now:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=4de84f1ef84238364822c668778e0ea7ae3f45a7

Comment 4

[Matthew N. [:MattN]]

Since bug 1399401 isn't resolved yet, don't we have to wait for it to be deployed? I don't see anything in your series that forces the new workers to be used?

Comment 5

[Joel Maher ( :jmaher ) (UTC -8)]

it looks like the new worker did not solve this- so keeping the tests disabled for now

Comment 6

[Matthew N. [:MattN]]

OK, I see… but this bug and the other similar ones you filed can be closed, right?

Comment 7

[Joel Maher ( :jmaher ) (UTC -8)]

yes, let me do that now-

closing this bug as a fix to the generic worker resolved issues- any remaining issues with the tests are not related to the worker upgrade.