541357 - [SeaMonkey 2.0 on "Parallels" (cb-seamonkey-osx-0?)] xpcshell-tests: near-perma-orange T-FAIL + timeout

Status: RESOLVED FIXED

(SeaMonkey :: Testing Infrastructure, defect)

Description

[Serge Gautherie (:sgautherie)]

http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey2.0/1264138905.1264146407.1348.gz
OS X 10.5 comm-1.9.1 unit test on 2010/01/21 21:41:45
{
TEST-PASS | /builds/slave/comm-1.9.1-macosx-unittest/build/objdir/mozilla/_tests/xpcshell/test_mailnewsglobaldb/unit/test_index_junk_imap_online.js | test passed

command timed out: 300 seconds without output, killing pid 50456
}

The next test should be test_mailnewsglobaldb/unit/test_index_junk_local.js

NB: Like bug 541111, I wonder if we have a too short timeout or if it's just our "broken" box again...

Comment 1

[Robert Kaiser]

(In reply to comment #0)
> NB: Like bug 541111, I wonder if we have a too short timeout or if it's just
> our "broken" box again...

And which "broken" box would that be? Hard to say if you don't even post box names...

Comment 2

[Serge Gautherie (:sgautherie)]

(In reply to comment #1)
> And which "broken" box would that be? Hard to say if you don't even post box
> names...

I meant (possible) "Parallels" misbehaviors...

NB: To be explicit, I filed these 2 bugs because you asked me about these failures yesterday, but otherwise I just ignore them when we have (1) perma-orange and (2) MacOSX on Parallels.

Comment 3

[Robert Kaiser]

(In reply to comment #2)
> NB: To be explicit, I filed these 2 bugs because you asked me about these
> failures yesterday, but otherwise I just ignore them when we have (1)
> perma-orange and (2) MacOSX on Parallels.

1) We have one non-Parallels box now and from all I see it also has timeouts happening.
2) The real problems we had with Parallels were on mochitest-plain and mitigated by stepping back to 1 CPU per VM - those problems started happening on the 1-CPU configuaration and didn't change with going back to 2 CPUs and mochitest-plain seems to have no problems at all, i.e. to me, Parallels looks fixed now.

Comment 4

[Andrew Sutherland [:asuth] (he/him)]

Without output from the xpcshell test, it's hard to know exactly what's going on, but the following should be noted about the test:

1) On my local box, it takes 3 seconds to run test_index_junk_local.js.
2) The test has (and all gloda tests also have) a 600 second self-destruct timer.  (This is the default value for DEFAULT_LONGEST_TEST_RUN_CONCEIVABLE_SECS in mailnews/test/resources/asyncTestUtils.js).  If this fires, because of how xpcshell works, we should actually get the output from the test.

I'd suggest increasing your tinderbox timeout so the self-destruct has a chance to give us a better idea of what is going on.  Also, I think the thunderbird tinderboxen may be the reason why we set the timeout at 10 minutes, suggesting this may not be unusual...

Comment 6

[Serge Gautherie (:sgautherie)]

(In reply to comment #3)

I had a look at the waterfall from now back to 2010/01/30 20:57:02 (only).

> 1) We have one non-Parallels box now and from all I see it also has timeouts
> happening.

Unless I've missed something,

*all cb-seamonkey-osx-0? builds fail after some (random) 'test_mailnewsglobaldb' test.
{ examples:
http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey2.0/1265351094.1265353626.31419.gz
OS X 10.5 comm-1.9.1 unit test on 2010/02/04 22:24:54
/test_mailnewsglobaldb/unit/test_index_junk_imap_online.js | test passed
+
http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey2.0/1265407602.1265411540.7979.gz
OS X 10.5 comm-1.9.1 unit test on 2010/02/05 14:06:42
/test_mailnewsglobaldb/unit/test_gloda_content_local.js | test passed
+
http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey2.0/1265439775.1265442340.1241.gz
OS X 10.5 comm-1.9.1 unit test on 2010/02/05 23:02:55
/test_mailnewsglobaldb/unit/test_index_messages_imap_online_to_offline.js | test passed
+
http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey2.0/1265649762.1265652923.26424.gz
OS X 10.5 comm-1.9.1 unit test on 2010/02/08 09:22:42
/test_mailnewsglobaldb/unit/test_noun_mimetype.js | test passed
+
http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey2.0/1265685420.1265688665.22272.gz
OS X 10.5 comm-1.9.1 unit test on 2010/02/08 19:17:00
/test_mailnewsglobaldb/unit/test_mime_emitter.js | test passed
+
http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey2.0/1265719694.1265722472.31231.gz
OS X 10.5 comm-1.9.1 unit test on 2010/02/09 04:48:14
/test_mailnewsglobaldb/unit/test_index_messages_local.js | test passed
}

*only 1 cb-sea-miniosx01 build failed, and that case ('test_dm' test) is different.
{
http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey2.0/1265153137.1265154882.31641.gz
OS X 10.5 comm-1.9.1 unit test on 2010/02/02 15:25:37
/test_dm/unit/test_bug_401582.js | test passed
}
I'd be interested if you could point me to other (related) failures on that box.

> 2) The real problems we had with Parallels were on mochitest-plain and
> mitigated by stepping back to 1 CPU per VM - those problems started happening
> on the 1-CPU configuaration and didn't change with going back to 2 CPUs and
> mochitest-plain seems to have no problems at all, i.e. to me, Parallels looks
> fixed now.

Well, unless we would find a flaw in 'test_mailnewsglobaldb', Parallels (VMs) still looks broken to me.
It would be interesting to check when the 2nd CPU was re-enabled and what was the situation before. (or to disable it again and monitor if it makes a difference.)

Comment 7

[Serge Gautherie (:sgautherie)]

KaiRo, could you do something about it? Like to restrict the tests to the mini box (and maybe 1 VM only), ftb?

Comment 8

[Robert Kaiser]

I can't limit any tests to just some random machine or collection of machines, and I won't invest any second in trying to do that. Let's hope bug 526208 will just fix all our problems.

Comment 9

[Serge Gautherie (:sgautherie)]

Not fully perma-orange actually: a few runs pass.

R.Fixed by bug 526208 which removed these VMs.