931116 - [Tracking] b2g desktop mochitest failures

Status: RESOLVED WONTFIX

(Testing :: Mochitest, defect)

Description

[Andrew Halberstadt [:ahal]]

There are ~100 failing mochitest failures specific to b2g desktop when I run them locally. I haven't had much luck getting people to help investigate so I'm going to disable them so we can at least get the rest rolled out asap.

Comment 1

[Andrew Halberstadt [:ahal]]

Attachment: local_triage

Actually turned out to only be 17 tests after multiple failures from single tests were taken into account.

Comment 2

[Andrew Halberstadt [:ahal]]

We'll see how it does on Cedar: https://tbpl.mozilla.org/?tree=Cedar&rev=edb9156ddb0a

Comment 3

[Andrew Halberstadt [:ahal]]

Morphing this into a tracking bug. Looks like there were many more failures than 100, they were just hidden. Also lots of intermittents that look like they aren't related to any particular test.

Comment 4

[Andrew Halberstadt [:ahal]]

Attachment: Patch 1.0 - disable initial failing b2g desktop mochitests

This patch green's up Cedar (about 24 in a row now without failures). This patch is a little overzealous since some of the tests were disabled due to bug 924681 which is now fixed. I don't know which tests these were, so I'll file a follow up bug to look into re-enabling them.

Comment 5

[Jonathan Griffin (:jgriffin)]

Comment on attachment 826748 [details] [diff] [review]
Patch 1.0 - disable initial failing b2g desktop mochitests

Review of attachment 826748 [details] [diff] [review]:
-----------------------------------------------------------------

\o/

Comment 6

[Andrew Halberstadt [:ahal]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/f75f5df9bdc1

Comment 7

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/f75f5df9bdc1

Comment 8

[Andrew Halberstadt [:ahal]]

Attachment: Add b2g desktop mochitest manifest to b2g26

This is the same set of tests disabled as the last patch, except based off mozilla-b2g26, so carrying r+ forward.

Comment 9

[Andrew Halberstadt [:ahal]]

https://hg.mozilla.org/releases/mozilla-b2g26_v1_2/rev/e5b576a8bb3b

Comment 10

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

I kind of ran into a problem that might affect a lot of tests running in b2g desktop. I thought I'd post it here in case it helps anyone.

Lots of tests do synthesizeKeyEvent(key). When that code runs in a content process, the key event is handled normally. When everything runs in one process, the key event usually gets thrown away here:

http://mxr.mozilla.org/mozilla-central/source/b2g/chrome/content/shell.js?rev=8db31fe79a2d#433

B2g desktop uses one process for everything by default, so I suspect that a lot of tests are failing because of this problem.

I ran into this accidentally while fixing some e10s stuff and I haven't looked into it much on the b2g side, so I might have misdiagnosed the issue. But if someone is interested in enabling these tests, it's worth investigating.

Comment 11

[Ben Turner (not reading bugmail, use the needinfo flag!)]

All of the indexedDB tests are failing because the test page runs with a principal of "mochi.test" but then the parent window has a principal of the testing app. This looks like a third-party window and indexedDB refuses to load in third-party windows.