872522 - B2G device image builds broken by unknown non-gecko check-in

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Gaia, defect)

Description

[Ed Morley [:emorley]]

Since https://tbpl.mozilla.org/?tree=Gaia-Master doesn't work properly and bug 868597 / bug 868602 are not yet complete - we are still in the frustrating position where sheriffs have to play "guess the gaia changeset that broke the B2G device image builds". And then even when we work it out, we don't have access to the gaia repo to back it out.

Latest instance:
https://tbpl.mozilla.org/php/getParsedLog.php?id=22981299&tree=Mozilla-Central#error0

All trunk trees closed until someone on the B2G team figures out what broke and backs it out. (The alternative is to hide the device image builds on trees; if people would prefer that, just ask...).

Comment 1

[Ryan VanderMeulen [:RyanVM]]

Julien is looking into this.

Comment 2

[Julien Wajsberg [:julienw]]

offending commit is : https://github.com/mozilla-b2g/gonk-misc/commit/49469de4063a797776230194c2da019f6a4bcdbf

Comment 3

[Julien Wajsberg [:julienw]]

I reverted it in https://github.com/mozilla-b2g/gonk-misc/commit/e6beb3cf80ed83f12e1713799d9fcccca6f3a996

Comment 4

[Julien Wajsberg [:julienw]]

fixes the bustage for me.

How do you force a new build on tbpl when the commit is on an external repository ?

Comment 5

[Ed Morley [:emorley]]

You can use the blue plus sign retrigger button in the bottom panel (visible after selecting a job).

Comment 6

[Ed Morley [:emorley]]

(I'll do it now)

Comment 7

[Julien Wajsberg [:julienw]]

(note: that was Bug 867031, I commented there)

Comment 8

[Ryan VanderMeulen [:RyanVM]]

Jonathan, so how does gonk bustage fit into your plan for in-tree Gaia pointers?

Comment 9

[Masatoshi Kimura [:emk]]

(In reply to Ed Morley [:edmorley UTC+1] from comment #0)
> Since https://tbpl.mozilla.org/?tree=Gaia-Master doesn't work properly and
> bug 868597 / bug 868602 are not yet complete - we are still in the
> frustrating position where sheriffs have to play "guess the gaia changeset
> that broke the B2G device image builds". And then even when we work it out,
> we don't have access to the gaia repo to back it out.

It also frustrates non-Gaia developers.

> All trunk trees closed until someone on the B2G team figures out what broke
> and backs it out.

Why the B2G team has a privilege to obstruct others' job?

> (The alternative is to hide the device image builds on
> trees; if people would prefer that, just ask...).

Please do (at least until bug 868597 and bug 868602 are fixed).

Comment 10

[Julien Wajsberg [:julienw]]

I _thought_ that Ryan had push access to B2G repositories.

Of course, that means he could find the offending commit, which is admittedly not always easy for now.

I'd prefer to not hide the device image builds. Otherwise the gecko teams will have a privilege to obstruct B2G team's job ;) I admit this is not an ideal solution, but this bug is not the place to discuss this.

Comment 11

[Jonathan Griffin (:jgriffin)]

(In reply to Ryan VanderMeulen [:RyanVM] from comment #8)
> Jonathan, so how does gonk bustage fit into your plan for in-tree Gaia
> pointers?

It doesn't, really.  We had a plan in which we could trigger builds per-commit to any of the B2G repos, but that's been back-burnered a bit while we stand up the gaia-gecko coupling.

I think we'll eventually have to get to that point, but I think I can pretty safely say it won't be this quarter.  In order to support that, a lot of rel-eng work will be needed, as well as the new TBPLv2.