923572 - Shut off gaia-unit on mozilla-aurora for gecko26

Status: RESOLVED WONTFIX

(Release Engineering :: General, defect)

Description

[Phil Ringnalda (:philor)]

+++ This bug was initially created as a clone of Bug #923570 +++

Unlike gaia-ui-tests where there's some hope, gaia-unit is permaorange on aurora, permaorange on trunk, and always has been. Someone who wants it running on gecko25 can jolly well green it up on the trunk first, and then ask to have it turned back on for wherever gecko25 is by then.

Comment 1

[Phil Ringnalda (:philor)]

gecko25, gecko26, the keys are right next to each other.

Comment 2

[Ed Morley [:emorley]]

John, bringing this to your attention, per blog posts :-)

Comment 3

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

hi Ed;

Fully agree no point in burning compute cycles on permaorange tests, so thanks for flagging. 

1) Before we turn these off, we need to notify B2G dev managers and see what they think is best course of action. B2G test coverage is sorely lacking, so if these are important test suites, it's possible they may prefer to prioritize fixing these suites instead of just powering them off and further reducing b2g test coverage. Obviously, if they feel these tests are not worth fixing, for whatever reason, then sure, we can disable across the board. But lets get the info together in a bug and ask them to make that call first. 

2) Is there a bug# is tracking fixing these permaorange gaia-unit testsuites, so they are green? 

If you give me (or link to this bug) the bug#s which need to be fixed to green up gaia-unit tests, we can make an informed, agreed-across-all-groups decision... either here, or in the Tuesday platform and b2g meetings.

Comment 4

[Phil Ringnalda (:philor)]

With this comment, I will be ccing every single person who was associated with the decision to run this permaorange suite on aurora for gecko26. His name is Armen, and he got zero response to his question in bug 891973 comment 6 (though had it been asked as "is this the someday when it should ride the trains?" I would have answered with a resounding "hell no").

Comment 5

[Phil Ringnalda (:philor)]

With this comment I will be ccing every single person who has, based on the evidence of bug 833666, made an effort over the last month and a half to fix up this permaorange test suite on the trunk. His name is Jonathan Griffin.

I would also cc every single person who has, since the merge to aurora, worked on fixing up the permaorange on that branch, except no such person exists. Odds are reasonably good that I am the only person who actually knows it even runs there.

Comment 6

[Phil Ringnalda (:philor)]

If B2G managers want to make managerial decisions, bug 923880 would be a good place for it: they could decide on which single trunk branch the suite will be greened up, and who will do it. This? This is permaorange nobody asked for, on a release branch.

Comment 7

[Ed Morley [:emorley]]

(In reply to John O'Duinn [:joduinn] from comment #3)
> hi Ed;
> 
> Fully agree no point in burning compute cycles on permaorange tests, so
> thanks for flagging. 
> 
> 1) Before we turn these off, we need to notify B2G dev managers and see what
> they think is best course of action. B2G test coverage is sorely lacking, so
> if these are important test suites, it's possible they may prefer to
> prioritize fixing these suites instead of just powering them off and further
> reducing b2g test coverage. Obviously, if they feel these tests are not
> worth fixing, for whatever reason, then sure, we can disable across the
> board. But lets get the info together in a bug and ask them to make that
> call first. 

This is for aurora only, and believed to be an oversight (ie: these should have never been enabled as the trains moved on).

> 2) Is there a bug# is tracking fixing these permaorange gaia-unit
> testsuites, so they are green? 
> 
> If you give me (or link to this bug) the bug#s which need to be fixed to
> green up gaia-unit tests, we can make an informed, agreed-across-all-groups
> decision... either here, or in the Tuesday platform and b2g meetings.

As philor said, bug 923880 will likely be the meta for that.

Comment 8

[Armen [:armenzg]]

I didn't touch anything.

This got changed in here:
http://hg.mozilla.org/build/buildbot-configs/rev/cface8156a58

Comment 9

[Phil Ringnalda (:philor)]

Now it's permared and takes 30 minutes, instead of permaorange and 10 minutes.

Comment 10

[Phil Ringnalda (:philor)]

Maybe in six weeks something will have changed.