Closed Bug 451949 Opened 16 years ago Closed 15 years ago

Mochitest <geolocation/test_manyWindows.html> fails intermittently, when some previous tests failed (!?)

Categories

(Core :: DOM: Geolocation, defect)

defect
Not set
normal

Tracking

()

VERIFIED FIXED
mozilla1.9.2a1

People

(Reporter: sgautherie, Assigned: sgautherie)

References

()

Details

(Keywords: fixed1.9.1, intermittent-failure, Whiteboard: [fixed1.9.1b4] )

Attachments

(1 file, 1 obsolete file)

http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1219572436.1219576608.20489.gz
Linux mozilla-central qm-centos5-moz2-01 dep unit test on 2008/08/24 03:07:16

*** 31474 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_manyWindows.html | Error thrown during test: windows[i] is null - got 0, expected 1
are there other "open many windows" unit tests also failing intermittently?
Assignee: nobody → doug.turner
http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1220900437.1220907272.19130.gz
Linux mozilla-central qm-centos5-moz2-01 dep unit test on 2008/09/08 12:00:37
http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1221593497.1221600507.26180.gz
Linux mozilla-central qm-centos5-moz2-01 dep unit test on 2008/09/16 12:31:37
http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1221870970.1221875560.4320.gz
Linux mozilla-central qm-centos5-03 dep unit test on 2008/09/19 17:36:10
Blocks: 456184
Blocks: 456463
Component: General → Geolocation
Product: Firefox → Core
QA Contact: general → geolocation
this hasn't been seen in a while.  please reopen if you see it again.
Status: NEW → RESOLVED
Closed: 15 years ago
Resolution: --- → WORKSFORME
I've seen it at least a few times recently.

The latest being:
http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey/1234714851.1234720340.2182.gz
Win2k3 comm-central dep unit test on 2009/02/15 08:20:51
{
*** 26913 ERROR TEST-UNEXPECTED-FAIL | /tests/content/media/video/test/test_access_control.html | Test timed out.
*** 26916 INFO Error: Unable to restore focus, expect failures and timeouts.
[...]
*** 27154 INFO Error: Unable to restore focus, expect failures and timeouts.
*** 40976 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_manyWindows.html | Error thrown during test: windows[i] is null - got 0, expected 1
}

Maybe this test is just not enough bullet-proofed against previous test (focus) failure(s) ?
Status: RESOLVED → REOPENED
Resolution: WORKSFORME → ---
Depends on: 478652
serge, what did you have in mind?
I would hope that something could be improved so that test_manyWindows.html either succeeds,
or maybe "simply" checks what the situation is when it starts and reports a more specific failure.

Or maybe test_access_control.html could better "cleanup" after itself when it fails ?

In short, anything that would make test_manyWindows.html not fail when it seems it should have no reason to.
Attached patch patch v.1 (obsolete) — Splinter Review
manyWindows is suppose to test for leaks of geolocation objects when windows are opened, access geo, then close.  This patch will prevent the window from being closed if it isn't in our array.... but really this shouldn't happen.  someone before our tests case is blowing up and that should be fixed.

Let me know if this patch does work for you, but I would rather not have to apply this bandaide to fix a test case above ours in the test case run.
(In reply to comment #9)
> This patch will prevent the window from  being closed if it isn't in our array

That is to say if window.open() failed.

> but really this shouldn't happen. 

I agree (but it does :-/).

(Maybe a try+catch instead would help figure out what is actually happening ??
Maybe triggering "Error: Unable to restore focus, expect failures and timeouts." like multiple in-between tests "do" would be "enough" ??)

> someone before our tests case is blowing up and that should be fixed.

Sure, this is why I filed bug 478652 too.

> Let me know if this patch does work for you

I don't know how to reproduce, as it's intermittent.
The patch would have to be checked in (to 1.9.1) then see what happens.
Let's morph this bug to the currently noticed case.
Status: REOPENED → NEW
OS: Linux → Windows Server 2003
Summary: <geolocation/test_manyWindows.html> fails intermittently → Mochitest <geolocation/test_manyWindows.html> fails intermittently, when some previous tests failed (!?)
So uh, I was looking at the geo tests, because I couldn't get the geolocation stuff working. I noticed that this test uses a setInterval, but doesn't call SimpleTest.waitForExplicitFinish(). This test is basically broken by design. Maybe we should fix that so it tests something useful?

Excerpt from a green unittest run on Linux:
*** 40870 INFO Running /tests/dom/tests/mochitest/geolocation/test_manyWindows.html...
*** 40871 INFO TEST-PASS | /tests/dom/tests/mochitest/geolocation/test_manyWindows.html | Access Ci
*** 40872 INFO TEST-PASS | /tests/dom/tests/mochitest/geolocation/test_manyWindows.html | Access Cc
*** 40873 INFO TEST-PASS | /tests/dom/tests/mochitest/geolocation/test_manyWindows.html | Ensure that the geolocation object is present
*** 40875 INFO Running /tests/dom/tests/mochitest/geolocation/test_optional_api_params.html...
Whiteboard: [orange]
(In reply to comment #12)
> I noticed that this test uses a setInterval, but doesn't call
> SimpleTest.waitForExplicitFinish(). This test is basically broken by design.

Oh well, I just rediscovered this ... I can't believe Doug did not fix it yet :-<

> Maybe we should fix that so it tests something useful?

I'll do.

> Excerpt from a green unittest run on Linux:

Either you were lucky or this regressed:
currently, this test simply does not check anything:

{
http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1237204411.1237209379.15494.gz&fulltext=1
Linux mozilla-central unit test on 2009/03/16 04:53:31
http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1237214422.1237217596.4967.gz&fulltext=1
OS X 10.5.2 mozilla-central unit test on 2009/03/16 07:40:22
http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1237204411.1237213997.29399.gz&fulltext=1
WINNT 5.2 mozilla-central unit test on 2009/03/16 04:53:31

*** 41467 INFO Running /tests/dom/tests/mochitest/geolocation/test_manyWindows.html...
*** 41469 INFO Running /tests/dom/tests/mochitest/geolocation/test_optional_api_params.html...
}
and same on my local Windows SeaMonkey/1.9.2 build.
Assignee: doug.turner → sgautherie.bz
Blocks: 483407, 437948
Status: NEW → ASSIGNED
Flags: in-testsuite+
OS: Windows Server 2003 → All
Hardware: x86 → All
[Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9.2a1pre) Gecko/20090316 Minefield/3.2a1pre] (home, optim default) (W2Ksp4)
(http://hg.mozilla.org/mozilla-central/rev/8d186b98ff31)

Per comment 12:
this makes the check actually run.
Attachment #362669 - Attachment is obsolete: true
Attachment #367604 - Flags: review?(ted.mielczarek)
Attachment #367604 - Flags: review?(ted.mielczarek) → review+
Comment on attachment 367604 [details] [diff] [review]
(Av1) Add missing |SimpleTest.waitForExplicitFinish();|
[Checkin: Comment 15 & 17]


http://hg.mozilla.org/mozilla-central/rev/1b83dc5ed168
Attachment #367604 - Attachment description: (Av1) Add missing |SimpleTest.waitForExplicitFinish();| → (Av1) Add missing |SimpleTest.waitForExplicitFinish();| [Checkin: Comment 15]
Status: ASSIGNED → RESOLVED
Closed: 15 years ago15 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla1.9.2a1
V.Fixed:
{
http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1237392735.1237395724.30092.gz&fulltext=1
OS X 10.5.2 mozilla-central unit test on 2009/03/18 09:12:15

*** 41496 INFO Running /tests/dom/tests/mochitest/geolocation/test_manyWindows.html...
*** 41497 INFO TEST-PASS | /tests/dom/tests/mochitest/geolocation/test_manyWindows.html | Opened a bunch of windows and didn't crash.
}
Status: RESOLVED → VERIFIED
Comment on attachment 367604 [details] [diff] [review]
(Av1) Add missing |SimpleTest.waitForExplicitFinish();|
[Checkin: Comment 15 & 17]


http://hg.mozilla.org/releases/mozilla-1.9.1/rev/ccd8a5cc785a
Attachment #367604 - Attachment description: (Av1) Add missing |SimpleTest.waitForExplicitFinish();| [Checkin: Comment 15] → (Av1) Add missing |SimpleTest.waitForExplicitFinish();| [Checkin: Comment 15 & 17]
Keywords: fixed1.9.1
Whiteboard: [orange] → [fixed1.9.1b4] [orange]
(In reply to comment #17)
> http://hg.mozilla.org/releases/mozilla-1.9.1/rev/ccd8a5cc785a

First build, first failure:
{
http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey/1237419769.1237425924.25841.gz
Linux comm-central dep unit test on 2009/03/18 16:42:49

*** 41401 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_manyWindows.html | Error thrown during test: windows[i] is null - got 0, expected 1
}
But, as there were previous timeouts and focus errors, I won't reopen just yet.
serge, check out 484122.  i just fixed it.  It was the cause of the previous failure you are seeing.
Depends on: 484123
(In reply to comment #18)
> But, as there were previous timeouts and focus errors

I filed bug 484123.

(In reply to comment #19)
> serge, check out 484122.

Ah! Added dependency there ;->
Happened again: (See bug 484123 comment 3)
http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey/1237590448.1237596526.13756.gz
Linux comm-central dep unit test on 2009/03/20 16:07:28
Same log.

Doug, what about comment 19 ?
Do this currently depend on yet something else or shall I reopen this bug ?
i fixed the bug referenced in comment 19.  however, i did not update the strings that seamonkey uses which could cause a similar problem.
Bug 478652 comment 16 happened again.
test_manyWindows.html seems to still be fragile...
serge,
it is pretty weird that this is so fragile.  basically i just want to open up a few windows that access navigator.geolocation, close them, then verify at the end that we don't leak.  I am fine with disabling this test because if we are going to leak, we are most likely to leak from just running the other tests.

is there another way to go about building a test case that does what I am attempting to do?
If there's issues with the testcase in Mochitest, can we have it s flagged moved back to "?"
No longer blocks: 483407
Depends on: 483407
Whiteboard: [fixed1.9.1b4] [orange] → [fixed1.9.1b4]
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: