(I noticed this with my local SeaMonkey/1.9.2, but did not have time to look into it before 1.9.1 landing :-() (I can't get the tinderbox logs atm :-/) Here is what I started to copy locally: { *** 41079 INFO Running /tests/dom/tests/mochitest/geolocation/test_allowCurrent.html... *** 41080 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_allowCurrent.html | Clicked "Tell them" button *** 41081 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_allowCurrent.html | Test timed out. *** 41083 INFO Running /tests/dom/tests/mochitest/geolocation/test_allowWatch.html... *** 41084 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_allowWatch.html | Clicked "Tell them" button } I guess this could be bug 451949 comment 22. PS: Actually, I haven't seen the boxes logs yet, but the regression timeframe fits with this: http://hg.mozilla.org/releases/mozilla-1.9.1/pushloghtml?startdate=2009-03-24+17%3A22%3A33&enddate=2009-03-24+19%3A37%3A00

Boxes logs are: { *** 41271 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_allowCurrent.html | Clicked "Tell them" button *** 41272 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_allowCurrent.html | Test timed out. *** 41275 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_allowWatch.html | Clicked "Tell them" button *** 41276 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_allowWatch.html | Test timed out. *** 41279 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_cancelCurrent.html | Clicked "Don't tell them" button *** 41280 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_cancelCurrent.html | Test timed out. *** 41283 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_cancelWatch.html | Clicked "Don't tell them" button *** 41284 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_cancelWatch.html | Test timed out. *** 41285 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/geolocation/test_cancelWatch.html | Too many test timeouts, giving up. }

we changed the string names: http://hg.mozilla.org/mozilla-central/rev/c0eea14622a6 The mochitest compare against the notification bar's actual text. (pretty brittle!)

I assume such .properties cannot be shared from core/toolkit, right ? :-/ At least, why don't http://mxr.mozilla.org/mozilla-central/source/dom/tests/mochitest/geolocation/geolocation_common.js retrieve the property values too ? Should help to figure out what is broken and to support locales, shouldn't it ?

This fixes the timeouts :-) (Now, I get a JS error, I'm looking into...)

The new error is: { Error: aBar is null Source File: http://localhost:8888/tests/dom/tests/mochitest/geolocation/geolocation_common.js Line: 73 } getNotificationBox() returns a XULElement; but getNotificationBox().currentNotification is null...

Comment on attachment 369312 [details] [diff] [review] (Av1) Update notification.properties You need to fix the actual notification bar, not just its strings.

(In reply to comment #5) > getNotificationBox() returns a XULElement; An actual element or a wrapper? What's its localName?

(In reply to comment #7) > An actual element or a wrapper? What's its localName? [object XULElement], 'notificationbox'

Av1, with comment 6 suggestion(s). Looks like I missed to find nsSuiteGlue.js the first time :-/ Thanks for noticing. This fixes comment 5 error too ;-> ***** Though (now) test_timeoutWatch.html appears to (not execute and) hang if test_optional_api_params.html is ran before ... Looking...

Comment on attachment 369398 [details] [diff] [review] (Av2) Port bug 483845 to SM [Checkin: Comment 11] Looks good by code inspection, but I have no time to test right now and am probably unable to do so before Monday, but we probably want this before then. Neil, can you take a look at this?

Comment on attachment 369398 [details] [diff] [review] (Av2) Port bug 483845 to SM [Checkin: Comment 11] http://hg.mozilla.org/comm-central/rev/56cccf751e70

(In reply to comment #11) > http://hg.mozilla.org/comm-central/rev/56cccf751e70 Obviously from tinderboxes, this wasn't enough to make a difference :-/ [Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9.2a1pre) Gecko/20090326 SeaMonkey/2.0b1pre] (experimental/_m-c_, home, optim default) (W2Ksp4) (http://hg.mozilla.org/mozilla-central/rev/58c11ba7a31d +http://hg.mozilla.org/comm-central/rev/de7070b99471) Locally, I can "reproduce" two issues: the one already noted in comment 9 when running some of the geolocation tests only (which I have started to investigate); and one I discovered now when running the full test suite: test_allowCurrent.html hangs (and would probably time out). First impression is that this is caused by the "missing plugin" bar (triggered (and leftover) by some previous test) which "hides" the "geolocation" bar... Could someone else reproduce locally or check what's happening on tinderboxes ? If you confirm my observations, then I'd suggest to disable these tests until they can pass.

(In reply to comment #12) > First impression is that this is caused by the "missing plugin" bar (triggered > (and leftover) by some previous test) which "hides" the "geolocation" bar... Mossop, any idea which test this notification may be from? Thanks.

(In reply to comment #13) > (In reply to comment #12) > > First impression is that this is caused by the "missing plugin" bar (triggered > > (and leftover) by some previous test) which "hides" the "geolocation" bar... > Mossop, any idea which test this notification may be from? Thanks. I'd guess at one of http://mxr.mozilla.org/mozilla-central/source/modules/plugin/test/mochitest/ but I wouldn't know for sure.

The left-over notification is from content/base/test/file_bug391728_2.html

* Switch to clicking buttons by index, rather than label (which makes the previous patch unnecessary, I guess!) * Specifically request the geolocation notification

Oh, and it passes plain mochitests on try server. (It doesn't fully pass locally, but then all sorts of tests aren't passing locally for me...)

Comment on attachment 370198 [details] [diff] [review] Make geolocation tests more robust in getNotificationBox, don't you need the UniversalXPConnect permissions? Do you what to add constants for the 0 and 1? Might be more clear to the casual reader. Neil, are you up for fixing the others: http://mxr.mozilla.org/mozilla-central/search?string=clickNotificationButton Otherwise, i love this. if this works, r=me

Only tested locally this time.

Pushed changeset 02e2a5fb86aa to mozilla-central. Pushed changeset 23446831581c to releases/mozilla-1.9.1

SeaMonkey isn't so eager to clear the notification bar, so the test needs to.

(In reply to comment #16) > (which makes the previous patch unnecessary, I guess!) I think SeaMonkey wanted bug 483845 as a feature update anyway, no ? (Yet I agree these tests don't care for it anymore.) (In reply to comment #21) > SeaMonkey isn't so eager to clear the notification bar, so the test needs to. Oh, that's why then! I had figured out that the checks which were causing the "hang" were those which were checking for |!exception| ... but I hadn't looked into what the cause/fix could be yet. Thanks :-)

Comment on attachment 370434 [details] [diff] [review] Make geolocation tests even more robust [Checkin: Comment 24] ah. nice find. I think serge is going to be happy about this one.

Pushed changeset 5742f34d1fc8 to mozilla-central. Pushed changeset 05485c934644 to releases/mozilla-1.9.1

http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey/1238612564.1238617767.25763.gz&fulltext=1 Linux comm-central dep unit test on 2009/04/01 12:02:44 Geolocation tests pass :-)

[Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9.2a1pre) Gecko/20090402 SeaMonkey/2.0b1pre] (experimental/_m-c_, home, optim default) (W2Ksp4) (http://hg.mozilla.org/mozilla-central/rev/c4f6bb1db611 +http://hg.mozilla.org/comm-central/rev/cdfcaddd6832) V.Fixed

This is an additional fix I had noticed while investigating. Tell me if it should rather be |ok(true| than |todo(false|. I guess resume_geolocationProvider() should be called in all cases, right ?

Comment on attachment 371088 [details] [diff] [review] (Dv1) Fix test_timeoutWatch.html errorCallback() [Checkin: See comment 30 & 32] hmm. actually, it hink we might want to fail if we get position unavailable since we are now installing a geolocation provider.

(In reply to comment #28) > i think we might want to fail if we get position unavailable > since we are now installing a geolocation provider. I'll do. Do you want to keep the separate POSITION_UNAVAILABLE check, or would the TIMEOUT check be enough ?

Comment on attachment 371088 [details] [diff] [review] (Dv1) Fix test_timeoutWatch.html errorCallback() [Checkin: See comment 30 & 32] http://hg.mozilla.org/mozilla-central/rev/3678ce7c2361

(In reply to comment #30) > http://hg.mozilla.org/mozilla-central/rev/3678ce7c2361 Dv1, with comment 28 suggestion(s).

Comment on attachment 371088 [details] [diff] [review] (Dv1) Fix test_timeoutWatch.html errorCallback() [Checkin: See comment 30 & 32] http://hg.mozilla.org/releases/mozilla-1.9.1/rev/e71647c0f91d

(In reply to comment #15) > The left-over notification is from content/base/test/file_bug391728_2.html I filed bug 487911. (In reply to comment #18) > Neil, are you up for fixing the others: > http://mxr.mozilla.org/mozilla-central/search?string=clickNotificationButton I filed bug 487797.