1282017 - it appears that we are scheduling android builds on try when not requested

Status: RESOLVED WORKSFORME

(Taskcluster :: General, defect)

Description

[Joel Maher ( :jmaher ) (UTC -8)]

odd, I don't see a lot of other cases of this, but I do see 2 cases:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=0e64234605c5c39d08e47d53a087cc2258711d40

and my push with the same commit message:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=24826d91438581abb36b250c0c0aeee41ffd2c6f

the try syntax is:
try: -b do -p win32,win64 -u none -t none


I am not sure why that would trigger android builds, especially since they are on linux.  Possibly something in the base tree is fiddling with the taskgraph and running android all the time.

Comment 1

[Geoff Brown [:gbrown]]

Actually, there's no Android *build* in those pushes, just Android lint and similar tests. Those jobs are triggered by changes to certain files: see http://hg.mozilla.org/mozilla-central/annotate/e44957e76603/taskcluster/ci/legacy/tasks/branches/base_jobs.yml#l614 for instance. There has been discussion about unexpectedly triggering these tasks with pushes that don't intentionally change those files (sorry - can't find the bug), but I think this is generally working as expected.

Comment 2

[Dustin J. Mitchell [:dustin] (he/him)]

FIXED/TRYISWEIRD