669530 - clobberer isn't actually clobbering

Status: RESOLVED FIXED

(Release Engineering :: General, defect, P2)

Description

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

win opt bustage that should have been fixed by a clobber:
http://tinderbox.mozilla.org/showlog.cgi?log=Mozilla-Aurora/1309906114.1309908891.3005.gz

clobberer says:
WINNT 5.2 mozilla-aurora build
w32-ix-slave42
was last clobbered at:
2011-07-05 14:14:07 PDT by [e-mail]

which was before this build started (15:48), yet the build shows evidence of not being clobbered.


same problem on android:
http://tinderbox.mozilla.org/showlog.cgi?log=Mozilla-Aurora/1309908782.1309909085.3706.gz

clobberer says:
Android mozilla-aurora build
linux-ix-slave14
was last clobbered at:
2011-07-05 14:14:07 PDT by [e-mail]

yet this build started at 16:33 and also had bustage that should have been fixed by clobbering.

Comment 1

[Justin Wood (:Callek)]

dustin, anychance theres a smoking gun here [you would notice] since we recently upgraded clobberer to use MySQL?

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

(In reply to comment #0)
> win opt bustage that should have been fixed by a clobber:
> http://tinderbox.mozilla.org/showlog.cgi?log=Mozilla-Aurora/1309906114.
> 1309908891.3005.gz
> 
> clobberer says:
> WINNT 5.2 mozilla-aurora build
> w32-ix-slave42
> was last clobbered at:
> 2011-07-05 14:14:07 PDT by [e-mail]
> 
> which was before this build started (15:48), yet the build shows evidence of
> not being clobbered.

In the "checking clobber times" section of this log, you can observe that what I did in the clobberer UI clobbered the "aurora-w32" tree, but we're actually using the "m-aurora-w32" tree which was not clobbered.


> same problem on android:
> http://tinderbox.mozilla.org/showlog.cgi?log=Mozilla-Aurora/1309908782.
> 1309909085.3706.gz
> 
> clobberer says:
> Android mozilla-aurora build
> linux-ix-slave14
> was last clobbered at:
> 2011-07-05 14:14:07 PDT by [e-mail]
> 
> yet this build started at 16:33 and also had bustage that should have been
> fixed by clobbering.

Same here:  the builder is using builddir m-aurora-lnx-andrd-dbg, but we clobbered aurora-lnx-andrd-dbg.

Comment 3

[Chris AtLee [:catlee]]

I think http://hg.mozilla.org/build/buildbotcustom/rev/d046c7611110 changed the build directories used, and triggered a re-occurrence of an old clobberer bug I can't find right now.

Comment 4

[Chris AtLee [:catlee]]

I think I fixed up the clobberer db for those builders:

update clobber_times set builddir='m-aurora-w32' where builddir='aurora-w32';
update clobber_times set builddir='m-aurora-lnx-andrd' where builddir='aurora-lnx-andrd';
update clobber_times set builddir='m-aurora-lnx-andrd-dbg' where builddir='aurora-lnx-andrd-dbg';

Comment 5

[Chris AtLee [:catlee]]

Attachment: Update builds table based on builddir, not buildername

I'm getting deja vu with this, but thought I'd give it a shot.

We hit a problem when the builddir on the slave changes, but the buildername remains the same. By updating the old row in the DB, clobberer has no idea about the new build directory, and doesn't present it in the list of possible things to clobber.

Comment 6

[[e-mail] (:bhearsum)]

Comment on attachment 544118 [details] [diff] [review]
Update builds table based on builddir, not buildername

This seems like it should work to me..

Comment 7

[Chris AtLee [:catlee]]

Comment on attachment 544118 [details] [diff] [review]
Update builds table based on builddir, not buildername

landed on clobberer, and stage-clobberer.