944717 - Test failure "Successfully restarted in 32bit mode after requesting it - 'x86_64-gcc3' should equal 'x86-gcc3'" in /restartTests/testRestartChangeArchitecture/test3.js

Status: RESOLVED FIXED

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

Description

[mario garbi]

Happened today on osx machines:
http://mozmill-daily.blargon7.com/#/functional/report/b99421c0f132c68dec1548288a615a32
http://mozmill-daily.blargon7.com/#/functional/report/b99421c0f132c68dec1548288a614a74
http://mozmill-daily.blargon7.com/#/functional/report/b99421c0f132c68dec1548288a61af9c
http://mozmill-daily.blargon7.com/#/functional/report/b99421c0f132c68dec1548288a61bb8a

Comment 1

[Andrei Eftimie]

Something clearly changed in mozilla-central between the 28th and 29th of November.

mozilla-central pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=3c44c1f43a67&tochange=1e3712793e7d

We have 2 merges from fx-team and 1 from inbound. I'll narrow this down with tinderbox builds.

Interestingly this doesn't fail with mozmill 2.0

Comment 2

[Andrei Eftimie]

Pushlog from fx-team:
http://hg.mozilla.org/integration/fx-team/pushloghtml?fromchange=5a6373ab5fe6&tochange=514b65ab771d

I can't see how any of these changes could affect OSX restart flags...
I have doublechecked the builds I've tested, and this pushlog appears to be valid.

Comment 3

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Please continue your regression tests. It's not done yet and those builds should still exist.

Comment 4

[Andrei Eftimie]

The referenced pushlog in comment 2 is from _consecutive_ tinderbox builds from fx-team.

I took them from https://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/fx-team-macosx64/ and the mentioned builds are:

1385679919 good
1385687659 bad

I'm not sure why there aren't tinderbox builds for each commit here.

Comment 5

[Cosmin Malutan, [:cosmin-malutan]]

Attachment: skip.patch

This fails a lot, and it seems that most of the tests failing with this error fails to send report document, so I think we should skip the tests ASAP.

Comment 6

[Henrik Skupin [:whimboo][⌚️UTC+1]]

This was a single push as you can see only one user. So you will have to build Firefox then.

Comment 7

[Henrik Skupin [:whimboo][⌚️UTC+1]]

(In reply to Cosmin Malutan from comment #5)
> This fails a lot, and it seems that most of the tests failing with this
> error fails to send report document, so I think we should skip the tests
> ASAP.

Both have absolutely nothing to do with each other.

Comment 8

[Andreea Matei [:AndreeaMatei]]

Comment on attachment 8341065 [details] [diff] [review]
skip.patch

Review of attachment 8341065 [details] [diff] [review]:
-----------------------------------------------------------------

Skipped:
http://hg.mozilla.org/qa/mozmill-tests/rev/51fe4371ac05 (default)

Comment 9

[Andrei Eftimie]

I've build Firefox on OSX several time already, but I haven't been able to have a `universal binary`.

I'm still tweaking the build options to have a correct build that supports both 32b and 64b architecture. I've looked at both https://developer.mozilla.org/en-US/docs/Mac_OS_X_Universal_Binaries and compared the compile flags from about:buildconfig

Comment 10

[Henrik Skupin [:whimboo][⌚️UTC+1]]

We should wait for bug 960509 being fixed before continuing the investigation.

Comment 11

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Andrei, I think we can re-enable those tests, also given that those were always working with Mozmill 2.0.

Comment 12

[Andrei Eftimie]

Attachment: 1.patch

Great news.
I am unable to reproduce these failures at all with mozmill 2

Unskip patch attached.
Sample testrun:
http://mozmill-crowd.blargon7.com/#/functional/report/5c334d437fbb9f11a2de1d989b1b24ac

I'll check the other branches shortly

Comment 13

[Andrei Eftimie]

Attachment: 1.1.patch

Applies cleanly to default and mozilla-aurora
Only those branches have had these tests skipped.

Small update to enable the tests in the manifest only for OSX

Comment 14

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Comment on attachment 8367199 [details] [diff] [review]
1.1.patch

Review of attachment 8367199 [details] [diff] [review]:
-----------------------------------------------------------------

Why do we need a new patch for unskipping the test? We can simply backout. I do not see any differences in a local diff. Beside this the commit message should also contain better information which tests you re-enable.

Comment 15

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Backed-out the skip patch on default and will do the same on other branches if no failures are present today.

http://hg.mozilla.org/qa/mozmill-tests/rev/fba803cf9d26

Comment 16

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Attachment: Unskip (aurora) v1

I'm not sure why we change the disabled line in the manifest for aurora but because of that difference to default, I cannot backout the patch. Andrei, can you please review?

Comment 17

[Andrei Eftimie]

Comment on attachment 8369095 [details] [diff] [review]
Unskip (aurora) v1

Review of attachment 8369095 [details] [diff] [review]:
-----------------------------------------------------------------

Looks good.

Comment 18

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Also landed on aurora, so the test is enabled across all platforms:
http://hg.mozilla.org/qa/mozmill-tests/rev/d263f5e26e79

Comment 19

[Andrei Eftimie]

(In reply to Henrik Skupin (:whimboo) from comment #18)
> Also landed on aurora, so the test is enabled across all platforms:
> http://hg.mozilla.org/qa/mozmill-tests/rev/d263f5e26e79

This missed the correct flag to only run tests on OSX in the manifest file.

Comment 20

[Andrei Eftimie]

I've fixed this during the merge in bug 964172