955447 - mozilla-central desktop and Android nightlies not triggered when there haven't been any l10n changes, since that's required to bypass last-good-rev because win64 debug is permared but still considered

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[Phil Ringnalda (:philor)]

We've had fresh pushes they could have built on, and we've gotten b2g nightlies, but desktop and Android failed to trigger on 12-29 and 12-30.

There'll be self-serve-triggered nightlies on 4b242b19b006 shortly, but the start time should make it clear they're mine, not automation's.

Comment 1

[Chris AtLee [:catlee]]

The scheduler things there's nothing appropriate for building for some reason:
2013-12-30 03:02:02-0800 [-] lastGoodRev: took 0.31 seconds to run; returned None
2013-12-30 03:02:03-0800 [-] mozilla-central nightly: No sourcestamp returned from ssfunc; not scheduling a build

Comment 2

[Chris AtLee [:catlee]]

For some reason the win64-debug build is in the list of builders to look at when trying to find a good revision to build....It's been burning for ages, so I'm not sure how this ever worked ;)

Comment 3

[Chris AtLee [:catlee]]

I could be misintepreting the data...

Looking at the scheduler on bm81 via manhole:

>>> s
<buildbotcustom.scheduler.SpecificNightly-props instance at 0x10f85368>
>>> s.name
'mozilla-central nightly'
>>> s.builderNames
['Android 2.2 mozilla-central nightly', 'Android 2.2 Armv6 mozilla-central nightly', 'Android 4.2 x86 mozilla-central nightly', 'Linux mozilla-central nightly', 'Linux x86-64 mozilla-central nightly', 'Linux x86-64 mozilla-central asan nightly', 'Linux x86-64 mozilla-central debug asan nightly', 'OS X 10.7 mozilla-central nightly', 'WINNT 5.2 mozilla-central nightly', 'WINNT 6.1 x86-64 mozilla-central nightly', 'Linux mozilla-central xulrunner nightly', 'Linux x86-64 mozilla-central xulrunner nightly', 'OS X 10.7 mozilla-central xulrunner nightly', 'WINNT 5.2 mozilla-central xulrunner nightly']
>>> s.ssFunc.func_closure[2].cell_contents
['Android 2.2 mozilla-central build', 'Android 2.2 Armv6 mozilla-central build', 'Android 2.2 Debug mozilla-central build', 'Android 4.2 x86 mozilla-central build', 'Linux mozilla-central build', 'Linux mozilla-central leak test build', 'Linux x86-64 mozilla-central build', 'Linux x86-64 mozilla-central leak test build', 'OS X 10.7 mozilla-central build', 'OS X 10.7 64-bit mozilla-central leak test build', 'WINNT 5.2 mozilla-central build', 'WINNT 5.2 mozilla-central leak test build', 'WINNT 6.1 x86-64 mozilla-central leak test build']

I'm assuming this is the builderNames parameter passed to lastGoodFunc.

Comment 4

[Phil Ringnalda (:philor)]

The last two scheduled nightlies were on pushes which had a finished and red win64 debug build before they were scheduled, so either we aren't actually checking for it, or we only started checking for it on Sunday.

Comment 5

[Phil Ringnalda (:philor)]

Mmm, http://mxr.mozilla.org/build/source/buildbotcustom/misc_scheduler.py?mark=283,302#281. Have we been doing mozilla-central nightlies on the tip revision all this time, because there has never been a good rev for as long as we've had win64 debug builds, but up until now there has always been an l10n change to trigger them anyway?

Comment 6

[Chris AtLee [:catlee]]

Yes, that seems very likely. I haven't found anything in buildbot-configs or buildbotcustom that has changed lately that would explain this otherwise.

Comment 7

[Phil Ringnalda (:philor)]

Attachment: fix

Well, better rearrange the deck chairs before we just stop doing last-good-rev on m-c, then :)

Comment 8

[Chris AtLee [:catlee]]

Per irc - we can wait until Monday to reconfig for this.

Comment 9

[Phil Ringnalda (:philor)]

Comment on attachment 8354953 [details] [diff] [review]
fix

https://hg.mozilla.org/build/buildbot-configs/rev/3370c8db5936

Comment 10

[Armen [:armenzg]]

In production (not CCed).