872602 - Please give :nalexander access to a build slave for testing Android Nightly language repacks

Status: RESOLVED FIXED

(Infrastructure & Operations Graveyard :: CIDuty, task)

Description

[Nick Alexander :nalexander [he/him]]

Hello releng,

This is a follow up to the #releng conversation below.  Summary: I'd like to reland Bug 863445 but avoid breaking Nightly (https://bugzilla.mozilla.org/show_bug.cgi?id=863445#c25).  Testing a language repack locally seems difficult; breaking Android N seems bad.  RyanVM suggested I could borrow a build slave to test language repacks.

Is it easy to trigger a repack on a slave?

#releng conversation:

08:45 nalexander: Is it possible to trigger a Nightly build on try?  I have updated patches for Bug 863445 that pass (https://tbpl.mozilla.org/?tree=Try&rev=a6cab6824fda) but I'd like to avoid https://bugzilla.mozilla.org/show_bug.cgi?id=863445#c25 if possible.
08:47 bhearsum: nope
08:48 bhearsum: which specific part of the nightly build are you looking to have run?
08:48 nalexander: bhearsum: I did wonder, since it would be an easy way to get malicious code signed with Nightly keys.
08:48 nalexander: bhearsum: that was my next question.  My original patch failed due to some bad naming interactions in language repacks.
08:48 nalexander: bhearsum: maybe I could just run a language repack locall?
08:48 nalexander: locally?
08:49 bhearsum: maybe, but we already run some l10n tests on try...namely, the l10n-check target
08:49 bhearsum: oh, but that doesn't consider android
08:49 nalexander: Yeah, l10n on Android is pretty funky.
08:49 bhearsum: you could try running one by hand locally - it might be a little challenging
08:49 bhearsum: i can send you a log link that has a list of commands in it, though
08:50 nalexander: bhearsum: please.  I see https://developer.mozilla.org/en/docs/Creating_a_Language_Pack but it doesn't appear to consider Android at all -- perhaps that's fine.
08:52 bhearsum: ok, looks like you want to look at https://ftp.mozilla.org/pub/mozilla.org/mobile/nightly/2013-05-15-03-10-27-mozilla-central-android/mozilla-central-android-nightly-bm65-build1-build9.txt.gz
08:52 bhearsum: starting with the step started at  2013-05-15 04:51:32.503812
08:52 bhearsum: you'll need to tweak that a fair, since you won't be running through mock_mozilla
08:55 bhearsum: you'll need to have https://hg.mozilla.org/build/mozharness/ cloned somewhere, too
08:56 nalexander: bhearsum: yeah.  Thanks for the info.  In the grand scheme of things, how bad is breaking an N build?  'cuz I'm quite confident I can avoid the presenting issue, but looking at that log and the tools required, it's going to be a great deal of effort to test this locally.
08:57 bhearsum: that's more a question for edmorley or RyanVM
08:57 RyanVM: breaking a nightly?
08:57 RyanVM: that sounds bad at first glance
08:59 nalexander: RyanVM: I'm trying to avoid breaking an N build (I.e., build or tests not completing), not shipping a horked Nightly.
09:00 nalexander: RyanVM: the issue is that it seems really hard to test Android language repacks.
09:00 RyanVM: nalexander: I'm wondering if with enough mozconfig magic, you could generate a nightly off Try
09:01 RyanVM: but my general stance is that breaking an N build is a bad thing :)
09:01 nalexander: RyanVM: presumably not, since Try doesn't have signing keys.
09:01 RyanVM: especially because recent experience tells that us that Android nightlies are different enough from dep builds that they can break where deps don't
09:01 nalexander: RyanVM: you have any idea how to make a builder, somewhere, do a language repack?
09:01 RyanVM: file a releng bug to borrow a machine?
09:01 RyanVM: catlee^ ?
09:04 nalexander: RyanVM: thanks for the suggestion, I'll move this to a ticket.
09:04 RyanVM: thanks

Comment 1

[u429623]

disabled bld-centos6-hp-013, waiting for current build to complete

Comment 2

[u429623]

credentials sent via IRC.

When done with host, unassign this bug. Thanks.

Comment 3

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

(In reply to Hal Wine [:hwine] from comment #2)
> credentials sent via IRC.
> 
> When done with host, unassign this bug. Thanks.

:nalexander, all the dep bugs here are fixed, are you still using this machine? If so, thats fine. If not, we'd like to put it back in production.

Comment 4

[Nick Alexander :nalexander [he/him]]

(In reply to John O'Duinn [:joduinn] from comment #3)
> (In reply to Hal Wine [:hwine] from comment #2)
> > credentials sent via IRC.
> > 
> > When done with host, unassign this bug. Thanks.
> 
> :nalexander, all the dep bugs here are fixed, are you still using this
> machine? If so, thats fine. If not, we'd like to put it back in production.

I believe so, fingers crossed :)  Thanks for its use!