467485 - reftest, crashtest and BrowserChrome broken on unittest mac since Shiretoko branding

Status: RESOLVED FIXED

(Release Engineering :: General, defect, P2)

Description

[Nick Thomas [:nthomas] (UTC+12)]

unittest hard codes Minefield.app in a few places and we have Shiretoko.app now, so three of the test suites are broken there.

Trying a couple of runs with a symlink added manually, eg in
moz2-darwin9-slave05:/builds/slave/trunk_darwin-5/build/objdir/dist/, from Minefield.app to Shiretoko.app

Similar to bug 456281, but I didn't want to push all of that out because it's not clear if any testing has been done in staging.

Comment 1

[Nick Thomas [:nthomas] (UTC+12)]

Also, the it wouldn't help for the ref and crashtest.

Comment 2

[Nick Thomas [:nthomas] (UTC+12)]

Attachment: Temporary (!) kludge to keep us going

Comment 3

[Nick Thomas [:nthomas] (UTC+12)]

Comment on attachment 350913 [details] [diff] [review]
Temporary (!) kludge to keep us going

Checked in and pushed live, to get the Firefox3.1 tree green. Passed checkconfig before hand.

Comment 4

[Nick Thomas [:nthomas] (UTC+12)]

s/name/description/ but otherwise working OK.

Comment 5

[Chris Cooper [:coop] (he/him)]

Comment on attachment 350913 [details] [diff] [review]
Temporary (!) kludge to keep us going

This is fine as a temporary hack, but it will be fixed properly by bug 456281 which I'm planning to test in staging later today.

Comment 6

[:Gavin Sharp [email: [e-mail]]]

(In reply to comment #5)
> (From update of attachment 350913 [details] [diff] [review])
> This is fine as a temporary hack, but it will be fixed properly by bug 456281
> which I'm planning to test in staging later today.

That won't fix reftest and crashtest, will it? Not sure what the best approach for those is... perhaps a checked in python script similar to runtests.py that finds the right bundle? Maybe even just use runtests.py, since we'll potentially end up doing that for bug 374458 anyways.

Comment 7

[Nick Thomas [:nthomas] (UTC+12)]

Leaving open for a proper fix.

Comment 8

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

(In reply to comment #8)
> unittest hard codes Minefield.app in a few places and we have Shiretoko.app
> now, so three of the test suites are broken there.

urgh. ok, so doing symlinks can workaround the hardcoding, but still... I'd be worried that this comes up again with other pending project branches. 

Feels to me like the real fix is to change the unittest to just use the name of the project, not hardcode "Minefield". Should this bug be reassigned to Core:Testing?

Comment 9

[:Gavin Sharp [email: [e-mail]]]

The hardcoding of Minefield.app is in the buildbot config, not in the test harness.

For the tests that use runtests.py (mochi/mochichrome, browser chrome), you can fix this by omitting the --appname parameter completely, and the patch in bug 374458 does that. For the tests that invoke Firefox directly (reftest/crashtest), you need to either find the right bundle in buildbot code somehow, or we'll need to introduce a python script in the test harness that does it for you. That's what comment 6 is about.

Comment 10

[:Gavin Sharp [email: [e-mail]]]

(In reply to comment #9)
> and the patch in bug 374458 does that.

Sorry, I meant bug 456281.

Comment 11

[[e-mail] (:bhearsum)]

Coop says this will be solved as part of bug 456281.

Comment 12

[[e-mail] (:bhearsum)]

Looks like the patch in 456281 is already running, duping this over.

Comment 13

[[e-mail] (:bhearsum)]

Looks like the patch in 456281 is already running, duping this over.

Comment 14

[Nick Thomas [:nthomas] (UTC+12)]

There's no talk of reftest or crashtest over on 456281

Comment 15

[[e-mail] (:bhearsum)]

Whoops, you're right Nick.

Comment 16

[[e-mail] (:bhearsum)]

Attachment: symlink shiretoko to minefield

We hit this again when during our consolidation downtime. This is a really crappy kludge that works for now. We need to fix it for realz soon.

Comment 17

[Chris AtLee [:catlee]]

Comment on attachment 353680 [details] [diff] [review]
symlink shiretoko to minefield

This step is now failing because Minefield.app exists.

Comment 18

[[e-mail] (:bhearsum)]

Attachment: test for shiretoko and minefield

Comment 19

[Chris AtLee [:catlee]]

Attachment: Set brand name in config.py and pass along to unittest factory

Comment 20

[Chris AtLee [:catlee]]

Attachment: Use branch name to run certain unit tests on mac

Comment 21

[[e-mail] (:bhearsum)]

Comment on attachment 353809 [details] [diff] [review]
test for shiretoko and minefield

Checking in factory.py;
/cvsroot/mozilla/tools/buildbotcustom/process/factory.py,v  <--  factory.py
new revision: 1.63; previous revision: 1.62
done

Comment 22

[[e-mail] (:bhearsum)]

Comment on attachment 353813 [details] [diff] [review]
Use branch name to run certain unit tests on mac

Checking in factory.py;
/cvsroot/mozilla/tools/buildbotcustom/process/factory.py,v  <--  factory.py
new revision: 1.64; previous revision: 1.63
done

Comment 23

[[e-mail] (:bhearsum)]

Comment on attachment 353812 [details] [diff] [review]
Set brand name in config.py and pass along to unittest factory

changeset:   607:b50d6b3214e3

Comment 24

[[e-mail] (:bhearsum)]

Comment on attachment 353813 [details] [diff] [review]
Use branch name to run certain unit tests on mac

Checking in unittest/steps.py;
/cvsroot/mozilla/tools/buildbotcustom/unittest/steps.py,v  <--  steps.py
new revision: 1.18; previous revision: 1.17
done

Comment 26

[[e-mail] (:bhearsum)]

I think this is fixed for realz this time.