588746 - Fake l10n repack builder as part of regular depend builds

Status: RESOLVED WONTFIX

(Release Engineering :: General, defect, P3)

Description

[Axel Hecht [:Pike]]

Given that we're busting l10n nightlies for the second time in a row, we should have a test.

Something like a triggered build like the tests, with a fake partial locale and an l10n-merge build reporting to the regular tinderbox would be useful, obviously.

Not sure if we can test release code paths, too, or if nightly and l10n dep need separate tests, depends a good deal on how shared their factories are these days.

Comment 1

[[e-mail] (:bhearsum)]

I don't think testing release paths make sense here -- we don't test any other release paths for on-change builds. Testing an l10n build makes a lot of sense though. If we set it up to trigger x-testing only, would that be good enough?

Comment 2

[Axel Hecht [:Pike]]

I'd prefer us to be independent of x-testing, as that's good to have as an independent playground.

Re release code path, I'm never sure what variable we exactly pass in, thus I'd prefer to have a test for it.

Comment 3

[[e-mail] (:bhearsum)]

OK, but I don't think we should be publishing real updates to any locale for this. Could we create a new fake locale for it?

Comment 5

[Lukas Blakk [:lsblakk] use ?needinfo]

We'll be getting our summer intern Nick to start looking into adding this to Try (and then to other branches as needed).

Comment 6

[Lukas Blakk [:lsblakk] use ?needinfo]

(In reply to comment #5)
> We'll be getting our summer intern Nick to start looking into adding this to
> Try (and then to other branches as needed).

Sorry I never got back round to this bug until now to update. We didn't end up getting that intern, so this project did not start and at present I can't give a good estimate of when it might be tackled.  It will stay attached to try_enhancements tracking so that at least it won't drop completely off my radar. That's the best I can say for now.

Comment 7

[[e-mail] (:bhearsum)]

Is this useful in the face of l10n-check?

Comment 8

[Axel Hecht [:Pike]]

I think so. Maybe more so if we're doing this with a pseudo localization similar to what we doing for gaia in bug 900182.

CCing glandium and gps, as they're the folks suffering from lack of testability of l10n automation mostly these days, I think.

Comment 9

[Justin Wood (:Callek)]

:Pike if this bug was solved with 'real' locales (eg. en-GB, ru, ja-jp[-mac], zh-CN, etc) would there still be value for a "fake" locale, if so who/what controls said "fake" locale, would it be in tree/generated from in tree or something?

Comment 10

[Axel Hecht [:Pike]]

I think this is good to dupe on the real locales at this point.