Open Bug 874423 Opened 7 years ago Updated 9 months ago

Intermittent docshell/test/navigation/test_bug430723.html | Test timed out.

Categories

(Core :: DOM: UI Events & Focus Handling, defect, P3)

x86
Windows 7
defect

Tracking

()

Tracking Status
firefox51 --- wontfix
firefox52 --- disabled
firefox53 --- disabled
firefox54 --- disabled

People

(Reporter: RyanVM, Unassigned)

References

Details

(Keywords: intermittent-failure, Whiteboard: [test-disabled][stockwell disabled])

Attachments

(1 file)

Attached image screenshot
https://tbpl.mozilla.org/php/getParsedLog.php?id=23172962&tree=Fx-Team

Windows 7 32-bit fx-team opt test mochitest-2 on 2013-05-20 14:58:22 PDT for push 4ac6c72b06c8
slave: t-w732-ix-033

15:04:14     INFO -  45 INFO TEST-START | /tests/docshell/test/navigation/test_bug430723.html
15:04:14     INFO -  46 INFO TEST-PASS | /tests/docshell/test/navigation/test_bug430723.html | Ensure red page loaded.
15:04:14     INFO -  47 INFO TEST-PASS | /tests/docshell/test/navigation/test_bug430723.html | Page1: Ensure the scrollpane is at the top before we start scrolling.
15:09:37     INFO -  48 ERROR TEST-UNEXPECTED-FAIL | /tests/docshell/test/navigation/test_bug430723.html | Test timed out.
15:09:37     INFO -  args: ['C:\\slave\\test\\build\\tests\\bin\\screenshot.exe', 'c:\\users\\cltbld\\appdata\\local\\temp\\mozilla-test-fail_q7ir9m']
15:09:37     INFO -  SCREENSHOT: <see attached>
15:09:38     INFO -  49 INFO TEST-END | /tests/docshell/test/navigation/test_bug430723.html | finished in 323518ms
Disabled on B2G debug for being nearly perma-fail.

https://hg.mozilla.org/integration/b2g-inbound/rev/e1e593b61fc5
Whiteboard: [test disabled on B2G Debug][leave open]
Bulk assigning P3 to all open intermittent bugs without a priority set in Firefox components per bug 1298978.
Priority: -- → P3
Nicolas, this is currently failing frequently on OSX and Windows (62 failures in the last week). Retriggers on inbound are pointing strongly at bug 1298085 as the culprit. Can you please take a look? Thanks!

https://treeherder.mozilla.org/#/jobs?repo=mozilla-inbound&noautoclassify&filter-searchStr=M-e10s(1)&fromchange=a22abb02c078b9a5b2c2aa98f8cb9e513c90dd2e&tochange=533af0114689f1ee9cd34711b2d0f5e92fa20652
Blocks: 1298085
Flags: needinfo?(nical.bugzilla)
Whiteboard: [test disabled on B2G Debug][leave open]
(In reply to Ryan VanderMeulen [:RyanVM] from comment #44)

Bas, maybe you can help find an owner for this?
Flags: needinfo?(bas)
I can believe these are related. Nical, can you spend an hour or two seeing if that sounds feasible?
Assignee: nobody → nical.bugzilla
Flags: needinfo?(bas)
as a note, I ran this specific test with a few other tests in a 'quarantine' job as an experiment and it was the first test to run and failed 100% of the time:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=aa40ec3c1268917871b1457d7ac3309a919e5b0b&selectedJob=29488556&filter-searchStr=quarantine

possibly we are failing because this test expects something in the browser to be ready, but it might be delayed in initialization intentionally or unintentionally.


here is what we see in automation as the failure:
 04:23:54     INFO -  132 INFO TEST-START | docshell/test/navigation/test_bug430723.html
 04:24:11     INFO -  JavaScript error: resource://gre/components/nsUrlClassifierListManager.js, line 481: uncaught exception: 2147500033
 04:29:11     INFO -  TEST-INFO | started process screencapture
 04:29:11     INFO -  TEST-INFO | screencapture: exit 0
 04:29:11     INFO -  133 INFO TEST-PASS | docshell/test/navigation/test_bug430723.html | Ensure red page loaded.
 04:29:11     INFO -  134 INFO TEST-PASS | docshell/test/navigation/test_bug430723.html | Page1: Ensure the scrollpane is at the top before we start scrolling.
 04:29:11     INFO -  135 INFO TEST-PASS | docshell/test/navigation/test_bug430723.html | Page1: Ensure we can scroll down.
 04:29:11     INFO -  136 INFO TEST-PASS | docshell/test/navigation/test_bug430723.html | Page1: Ensure we can scroll up, back to the top.
 04:29:11     INFO -  137 INFO TEST-PASS | docshell/test/navigation/test_bug430723.html | Ensure blue page loaded.
 04:29:11     INFO -  138 INFO TEST-PASS | docshell/test/navigation/test_bug430723.html | Page2: Ensure the scrollpane is at the top before we start scrolling.
 04:29:11     INFO -  139 INFO TEST-PASS | docshell/test/navigation/test_bug430723.html | Page2: Ensure we could scroll.
 04:29:11     INFO -  140 INFO TEST-PASS | docshell/test/navigation/test_bug430723.html | Ensure red page restored from history.
 04:29:11     INFO -  141 INFO TEST-PASS | docshell/test/navigation/test_bug430723.html | Page1Again: Ensure scroll pane at top before we scroll.
 04:29:11     INFO -  142 INFO TEST-UNEXPECTED-FAIL | docshell/test/navigation/test_bug430723.html | Test timed out.
04:29:11 INFO - reportError@SimpleTest/TestRunner.js:114:7


that is pretty consistent.  when running the job by itself on linux, I see this:
[task 2016-10-19T19:39:59.402231Z] 19:39:59     INFO -  0 INFO SimpleTest START
[task 2016-10-19T19:39:59.405783Z] 19:39:59     INFO -  1 INFO TEST-START | docshell/test/navigation/test_bug430723.html
[task 2016-10-19T19:44:59.414982Z] 19:44:59     INFO -  TEST-INFO | started process screentopng
[task 2016-10-19T19:44:59.707120Z] 19:44:59     INFO -  TEST-INFO | screentopng: exit 0
[task 2016-10-19T19:44:59.707480Z] 19:44:59     INFO -  2 INFO TEST-PASS | docshell/test/navigation/test_bug430723.html | Ensure red page loaded.
[task 2016-10-19T19:44:59.710239Z] 19:44:59     INFO -  3 INFO TEST-PASS | docshell/test/navigation/test_bug430723.html | Page1: Ensure the scrollpane is at the top before we start scrolling.
[task 2016-10-19T19:44:59.710316Z] 19:44:59     INFO -  4 INFO TEST-UNEXPECTED-FAIL | docshell/test/navigation/test_bug430723.html | Test timed out.


that is a different failure, but hopefully this information can help shed light on the real underlying problem!  Thanks for looking at this bug Nical :)
as a note, I saw this while running tests by themselves (fresh profile, process with no tests before/after):
https://treeherder.mozilla.org/#/jobs?repo=try&revision=72375369a12a65e716cdc7afa3a81200a4857dfb&selectedJob=30144866
I did spend a lot of time on this issue, although I kept quiet on the bug since I haven't found good leads on the issue (or rather, the ones I found didn't work out so far).

I am probably not going to work on this in the short term.

As far as I understand, we appear to be "leaking" some compositor notifications (as in, we composite, we notify the content side that the composition happens and the content side looses that information somewhere. This means that until something else (any kind of interaction) causes us to composite again and receive a new compositor notification we don't tick the refresh driver.

Unfortunately we can't just tick the refresh driver because all of the buffering logic in graphics relies on content not being able to get more than one (or is it two?) frames ahead of the compositor which is actually a good thing because if for some reason the gpu freezes for a long time, i gecko kept allocating textures to send to the compositor we'd OOM pretty quickly. Anyway that makes the whole thing a bit hard to track down, because any interaction with the browser will trigger new notifications and hide the bug.
Flags: needinfo?(nical.bugzilla)
can we disable this on windows/osx opt?  it seems to run ok on debug and on linux.
Flags: needinfo?(nical.bugzilla)
Pushed by gbrown@mozilla.com:
https://hg.mozilla.org/integration/mozilla-inbound/rev/56c545ab328a
Skip test_bug430723.html on OSX/Win opt. r=me,a=test-only
Flags: needinfo?(nical.bugzilla)
Keywords: leave-open
Whiteboard: [test disabled]
Keywords: leave-open
Whiteboard: [test disabled] → [test-disabled][stockwell disabled]
Assignee: nical.bugzilla → nobody
Component: Keyboard: Navigation → User events and focus handling
You need to log in before you can comment on or make changes to this bug.