David Baron :dbaron: Reporter
	Description • 15 years ago

The Mac debug everythingelse machine is now red most of the time (I think last week it was only red about a quarter of the time).

The last two failures have both been the following build step failing:

'bash ../tools/buildfarm/utils/installdmg.sh ...' failed

though there have also been other failures (which may or may not need to be split into different bugs; I'm starting with this one).

	Nick Thomas [:nthomas] (UTC+12) Assignee
	Comment 1 • 15 years ago

I see four builds with the unpack error, and it looks like they're all on bm-xserve17.

	Nick Thomas [:nthomas] (UTC+12) Assignee
	Comment 2 • 15 years ago

Ok, this definitely looks like it's slave specific. The "OS X 10.5.2 mozilla-central test debug everythingelse" builds starting at
* Wed Oct 28 13:39:41 2009
* Wed Oct 28 07:39:13 2009
* Wed Oct 28 06:31:21 2009
* Tue Oct 27 09:39:53 2009
all failed to unpack the dmg. Taking the 13:39 dmg and manually calling hdiutil attach ... works fine on bm-xserve19, which is nominally equivalent spec.

There are also a couple of failures in "OS X 10.5.2 mozilla-central test mochitests-N/5" where N = 2 and 4.

	David Baron :dbaron: Reporter
	Comment 3 • 15 years ago

Whatever the problem was, it seems to have gone away.

I wonder if we might not have been killing processes properly if the build failed because of hitting the log size limit?  (Going over the log size limit was fixed by bug 525100.)

	Nick Thomas [:nthomas] (UTC+12) Assignee
	Comment 4 • 15 years ago

Yeah, I left bm-xserve17 disconnected from the buildbot master so it couldn't re-offend. I doubt it can be anything to do with logs because the runs that were failing were packaged unit tests, and we reboot the slaves after every job.

I've dug around on bm-xserve17 and can't find anything out of the ordinary, so we'll reimage it to get back to a good state.

	Nick Thomas [:nthomas] (UTC+12) Assignee
	Comment 5 • 15 years ago

Bug 525398 for that.