1188698 - [gaia-automation] gaia_test.py publish method should *not* fail when succeed count is 0

Status: RESOLVED FIXED

(Release Engineering :: Applications: MozharnessCore, defect)

Description

[Aus Lacroix [:aus]]

Right now we fail when we have empty test chunks because mozharness stupidly assumes that passed count == 0 means we failed. 

It was a stop-gap put in place when we had issues with what make test-integration for gaia actually invokes. The underlying issues are fixed and so now this behavior is *not* desirable.

Comment 1

[Aus Lacroix [:aus]]

try run with 40 chunks for Gij -- https://treeherder.mozilla.org/#/jobs?repo=try&revision=715bbbf00682

Comment 2

[Ryan VanderMeulen [:RyanVM]]

Don't forget that mozharness lives in-tree too now, so you'll need to land the change there as well.

Also, why aren't we fixing the chunking logic to not allow for chunks with no tests being run? That seems wasteful to allow it.

Comment 3

[Aus Lacroix [:aus]]

(In reply to Ryan VanderMeulen [:RyanVM UTC-4] from comment #2)
> Don't forget that mozharness lives in-tree too now, so you'll need to land
> the change there as well.

True, will land there as well.

> Also, why aren't we fixing the chunking logic to not allow for chunks with
> no tests being run? That seems wasteful to allow it.

Yeah, unfortunately it has to do with when we apply the manifest for the test files that are disabled which is much later in the process. It would be nice to move that up the chain but that's going to take some major work. :/

Comment 4

[Pulsebot]

https://hg.mozilla.org/integration/b2g-inbound/rev/572d0d046629

Comment 5

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/572d0d046629

Comment 6

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

mozharness production tag moved to: https://hg.mozilla.org/build/mozharness/rev/80df8b8b2376