1272646 - Investigate why test from bug 449358 went perma-orange on android after the merge (but not before) - TEST-UNEXPECTED-FAIL | toolkit/content/tests/widgets/test_videocontrols_jsdisabled.html | checking video play state - got true, expected false

Status: NEW

(Toolkit :: Video/Audio Controls, defect)

Description

[:Gijs (he/him)]

It was green when it landed, and on the next run, and after that inbound merged in and it went orange. Likewise, it went orange on inbound when it was merged there from fx-team. :-(

https://treeherder.mozilla.org/logviewer.html#?job_id=27835484&repo=mozilla-inbound#L1327

Comment 1

[Ryan VanderMeulen [:RyanVM]]

I'd bet on chunking boundary changes for inexplicable Android mochitest permafails. Android still does the old-fashioned "split the tests evenly across the chunks" routine, which can introduce implicit dependencies on state from previous tests that's subject to change on a whim.

If you want to confirm that theory, look at the logs before and after and pay attention to whether the first test in the chunk changed or not.

Comment 2

[Jared Wein [:jaws] (please needinfo? me)]

So here are the weird things with this failure:

The failure condition is that video.paused is returning true, when we expect it to return false. The video has the autoplay attribute, so it should begin playing once it loads enough data. In fact, we *know* it is playing because the failure is inside of a timeupdate event listener.

Further, the test does not timeout. The next step in the test after the failure is to click on the video, which should toggle between playing and pausing. The next event we receive is indeed a pause event, so we could assume that the video *was* previously playing.

This failure seems to have nothing to do really with this test, it is just revealing some underlying issue most likely with the media APIs and their state when the video is just loading. https://bugzilla.mozilla.org/show_bug.cgi?id=1270738 looks to be the exact same problem and it started a week ago.

:alwu is assigned to that bug, so this intermittent will likely be fixed when bug 1270738 is fixed. I'll mark "see also" for that bug.

Comment 3

[Alastor Wu [:alwu]]

Hi, Jaren,
The bug 1270738 might be not related with this bug, it seems like a timing issue about missing the gecko event, and I already landed a possible patch in bug1269672.

Comment 4

[Desirae Deshazer]

Attachment: same problem

Thanks from coaplanindrak1977.