895996 - [ku] testAddMozSearchProvider.js fails with "Disconnect Error: Application unexpectedly closed" because the print dialog gets opened

Status: RESOLVED FIXED

(Mozilla QA Graveyard :: Mozmill Tests, defect, P1)

Description

[u279076]

Firefox 23.0 (23.0, ku, 20130718163513) on Windows NT 6.2.9200 (x86_64):
http://mozmill-release.blargon7.com/#/functional/report/180cf2548ef2865af3ae441d611e4314

Comment 1

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Mario, can you please run a quick check for the ku locale if it reproduces constantly?

Comment 2

[mario garbi]

Attachment: Screenshot from 2013-07-22 13:02:35.png

We can reproduce it locally everytime, the Firefox freezes and is not responding, it might be a regression.

http://mozmill-crowd.blargon7.com/#/functional/report/180cf2548ef2865af3ae441d616498b0

Comment 3

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Not sure what you mean with a freeze here. Given that I don't see anything related in the screenshot you gave here. Can you please explain that a bit more? When does it happen? Which version of Windows? Is it native or a VM? Is it specific to the ku locale?

Comment 4

[mario garbi]

I was running a VM - Win8 (x86). I am able to reproduce this every run with Ku but not even once with other builds so I am inclined to believe it ku locale specific. By freeze I refer to the firefox process becoming non-responsive as can be seen in the screenshot next to the page title (Not Reponding).

The reports for different locales:
http://mozmill-crowd.blargon7.com/#/functional/reports?branch=23.0&platform=Windows%20NT&from=2013-07-16&to=2013-07-22

Comment 5

[u279076]

Firefox 23.0 (23.0, ku, 20130722172257) - Windows NT 6.0.6002 Service Pack 2 (x86):
> http://mozmill-release.blargon7.com/#/functional/report/180cf2548ef2865af3ae441d61a20b51

Firefox 23.0 (23.0, ku, 20130722172257) - Windows NT 6.2.9200 (x86_64):
> http://mozmill-release.blargon7.com/#/functional/report/180cf2548ef2865af3ae441d61a1ed5d

Comment 6

[Henrik Skupin [:whimboo][⌚️UTC+1]]

We have to get this investigated and fixed soonish given that it affects our tests for each beta. Who can work on this?

Comment 7

[mario garbi]

I will work on this.

Comment 8

[mario garbi]

I noticed that we only ran tests from KU with Windows platform so I tested this on Linux and I've noticed that almost all Search tests are failing not just testAddMozSearchProvider due to an incorrect dialogue being opened (Print).
I will check if this is a localization problem or something on our part since we haven't tested KU before.

http://mozmill-crowd.blargon7.com/#/functional/report/180cf2548ef2865af3ae441d61beb1c8

Comment 9

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Might be a locale issue or that we do not use a DTD entity correctly for a shortcut.

Comment 10

[mario garbi]

Indeed, after going through the code and dtds I found this line that changes Select All shortkey from "a" to "P". This opens in our tests the print dialogue on KU and I suspect this was a mistake on their part.

DTDS:
http://hg.mozilla.org/releases/l10n/mozilla-aurora/ku/rev/baf78810a202#l6.266

Mozmill-tests line that fails:
http://hg.mozilla.org/qa/mozmill-tests/file/963408b08af3/lib/search.js#l576

Comment 11

[Henrik Skupin [:whimboo][⌚️UTC+1]]

(In reply to mario garbi from comment #10)
> Indeed, after going through the code and dtds I found this line that changes
> Select All shortkey from "a" to "P". This opens in our tests the print
> dialogue on KU and I suspect this was a mistake on their part.
> 
> DTDS:
> http://hg.mozilla.org/releases/l10n/mozilla-aurora/ku/rev/baf78810a202#l6.266

That's it! Please file a bug under the localization component and ku. 'P' should really not be used given that we don't differentiate between small and capital letters, and 'p' is assigned for printing.

Comment 12

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Anthony, as long as the problem persist, we might want to drop 'ku' from the locales to test.

Comment 13

[Andreea Matei [:AndreeaMatei]]

Mario, I see the dependency bug was fixed. Please check this test, thanks.

Comment 14

[mario garbi]

The dependency bug 897467 was only landed on Aurora and so the problem persists on Beta.

Comment 15

[mario garbi]

I have tested this again today with Mozmill 2.0.3 and we only hit this issue with ESR24 and not on the other branches.

ESR24 ku:
http://mozmill-crowd.blargon7.com/#/functional/report/40742e0746fd767e6d2fd586598cdeab

Aurora ku:
http://mozmill-crowd.blargon7.com/#/functional/report/40742e0746fd767e6d2fd586598d2307

On Aurora ku we have a similar issue as in Bug 951208 for which I will log a separate bug

Comment 16

[Andreea Matei [:AndreeaMatei]]

I don't think this will be fixed on ESR24 soon, most likely will get merged in time. I suggest we skip it on this branch, if Anthony would still want to run tests against 'ku' builds.
Would be great if we could implement soon the specific-locale skip :)

Comment 17

[Henrik Skupin [:whimboo][⌚️UTC+1]]

(In reply to Andreea Matei [:AndreeaMatei] from comment #16)
> I don't think this will be fixed on ESR24 soon, most likely will get merged
> in time. I suggest we skip it on this branch, if Anthony would still want to

I don't think that this will ever be merged into a ESR branch. But Axel could give better feedback.

Comment 18

[Axel Hecht [:Pike]]

Yeah, we don't update localizations for ESR.

I'm a bit surprised and worried that Firefox would hang for a bad command key. I'd expect the wrong dialog to open or the wrong action to take place, but not a hang. Could it be that the testsuite is hanging waiting for the print dialog?

Independent of that, I'll reduce the severity of this bug to normal, just because the affected population is small.

Comment 19

[Henrik Skupin [:whimboo][⌚️UTC+1]]

The hang is part of our Mozmill test which doesn't expect this dialog to appear, so we cannot close it. We could figure out what we could do to get it handled at least for the case when a different dialog gets opened. Andreea would you mind to file a bug for that in the Mozmill component?

Comment 20

[Andreea Matei [:AndreeaMatei]]

Filed bug 957049 for that. 
Mario, please make a skip patch for esr24. I know at some point we mentioned to Anthony that 'ku' is broken, but I'm not sure he stopped running tests on it.

Comment 21

[Henrik Skupin [:whimboo][⌚️UTC+1]]

I don't see why we should skip the test on esr24. Lets make sure that we do not run tests for the ku locale instead!

Comment 22

[Henrik Skupin [:whimboo][⌚️UTC+1]]

The dependency bug has been fixed for Firefox 26, so please re-check if we can re-enable the test for this locale.

Comment 23

[Cosmin Malutan, [:cosmin-malutan]]

This has been fixed, I checked on all os platforms, for linux I rechecked locally.
http://mozmill-crowd.blargon7.com/#/functional/report/12a8568e34c97b929089c7a61f27b3d1

Comment 24

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Was it the one and only failure which caused us to not run tests for ku? I ask because I don't see it in the list for locales to test on beta and release at https://github.com/mozilla/mozmill-ci/issues/270.

Comment 25

[Cosmin Malutan, [:cosmin-malutan]]

Yes, but as I mentioned in the last Monday meeting, not all locales are covered by the eight batches we have in issue 270. We need to add another one with the ones we missed, and test those too.

Comment 26

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Ok, so please comment on the issue, and list the locales we have to get added. Thanks.