1165158 - Reenable pointerlock test on most platforms

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[Xidorn Quan [:xidorn] UTC+11]

Currently, pointerlock tests (dom/tests/mochitest/pointerlock) are disabled on almost all platforms by different ways. They are disabled on Windows in mochitest.ini, on Linux in dom/tests/moz.build, on Mac OS X >= 10.7 in the test file. Hence, the tests are currently only run on Mac OS X 10.6. They are also disabled for e10s, Android, and b2g. Android and b2g are probably irrelevant, but Android does support mice, doesn't it?

We definintely should try to fix either side (impl or test) to enable them in most of those platforms, so that we could detect problems once we break pointerlock feature.

Comment 1

[Xidorn Quan [:xidorn] UTC+11]

An initial try push: https://treeherder.mozilla.org/#/jobs?repo=try&revision=78bfa62e2136

The tests have been running very well on Linux builds. (Windows builds were cancelled because this try push doesn't really enable Windows tests)

Comment 2

[Xidorn Quan [:xidorn] UTC+11]

Another try push for windows: https://treeherder.mozilla.org/#/jobs?repo=try&revision=7aaf3cfe27cf

Comment 3

[Xidorn Quan [:xidorn] UTC+11]

RyanVM, given the try run, do you think the tests are reliable enough to be enabled on Linux and Windows other than XP?

Comment 4

[Ryan VanderMeulen [:RyanVM]]

One hit on bug 931445. Are you still looking at the other various intermittents filed on this test?

Comment 5

[Ryan VanderMeulen [:RyanVM]]

Oh, ugh, it's a Start Screen failure:
http://mozilla-releng-blobs.s3.amazonaws.com/blobs/try/sha512/eb25576a9014fcc59d94721e5d7dfa2f2b9f42c696ec4024e3f7c80e55d9d6c162eeafba55a4a2a282d6c0d231ab1993c417aa0d5528725edd88c009656b148b

Comment 6

[Xidorn Quan [:xidorn] UTC+11]

(In reply to Ryan VanderMeulen [:RyanVM UTC-4] from comment #4)
> One hit on bug 931445. 

That is... I assume it's a different hit. Previously, timeouts mostly happen on file_childIframe and several other files because of a real bug in our code which will be fixed in bug 1174966.

The timeout in that try run happens in file_infiniteMovement. I would need more time to investigate. But it seems this intermittent is much harder to reproduce.

The bad thing of pointer lock tests is that, all of the tests are in one single test, which makes it hard to realize which one actually times out from the error signature. I've also filed bug 1175403 to split the tests, and hopefully that would help.

> Are you still looking at the other various
> intermittents filed on this test?

The intermittent in Linux build has been fixed in bug 1175390.

The bust in ASan and OS X is actually unrelated to this, and will be fixed in bug 1175485.

The intermittent in OS X, which is bug 1174323, I really have no idea how to fix that completely... see bug 1174323 comment 128.

Comment 7

[Xidorn Quan [:xidorn] UTC+11]

It seems that timeout isn't that easy to trigger: https://treeherder.mozilla.org/#/jobs?repo=try&revision=8905e35c2d91

Comment 8

[Xidorn Quan [:xidorn] UTC+11]

Attachment: enable on Linux and Windows other than XP

Comment 9

[Ryan VanderMeulen [:RyanVM]]

Comment on attachment 8624679 [details] [diff] [review]
enable on Linux and Windows other than XP

Review of attachment 8624679 [details] [diff] [review]:
-----------------------------------------------------------------

I assume the GTK3 folks are OK with this being re-enabled? Either way, better to disable in the manifest if needed rather than at the moz.build level.

Comment 10

[Xidorn Quan [:xidorn] UTC+11]

You removed the leave-open keyword... Okay, I guess that's fine, although it is not "most platforms". We can leave OS X Lion+ and Win XP to bug 1170369 anyway.

Comment 11

[Ryan VanderMeulen [:RyanVM]]

Yeah, probably easier for tracking anyway (especially since next week is Whistler and the uplift to Aurora is on the 29th).

Comment 12

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/478258a3ba5f

Comment 13

[Ryan VanderMeulen [:RyanVM]]

This made Mulet mochitests very unhappy. Given the earlier green Try pushes, I guess it depends on one of the other fixes that hasn't landed yet. Re-disabling on Mulet for now.
https://treeherder.mozilla.org/logviewer.html#?job_id=10967432&repo=mozilla-inbound

https://hg.mozilla.org/integration/mozilla-inbound/rev/76c92f57be06

Comment 14

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/478258a3ba5f
https://hg.mozilla.org/mozilla-central/rev/76c92f57be06