[Buildbot-devel] Recovering from having wrong git installed on cygwin

[Dan Kegel]

On Sun, Dec 9, 2012 at 8:43 AM, Dan Kegel <dank at kegel.com> wrote:
> I will try making sure nothing gets installed after rebaseall
> and see if that stabilizes things.

The problem happened again last night, this time on
a pure linux setup, so it's not cygwin-specific.
A build has been showing up as in progress for 17 hours,
stuck on the 'git' step:
            factory.addStep(Git(repourl=repourl, mode='full',
method='copy', branch=sbranch))
with stdout

git --version
 in dir /home/buildbot/slave-state/ubu1004/nobuild-buildhost03-ubu1004/gst-plugins-bad-ubu1004-master/build
(timeout 1200 secs)
 watching logfiles {}
 argv: ['git', '--version']
 environment:
  CCACHE_DIR=/home/buildbot/slave-state/ubu1004/nobuild-buildhost03-ubu1004/ccache.dir
  HOME=/home/buildbot
  INSTANCE=lo
  INTERFACE=lo
  JOB=network-interface
  LOGNAME=buildbot
  MAIL=/var/mail/buildbot
  OLDPWD=/home/buildbot/ob-repobot-git
  PATH=/home/buildbot/slave-state/ubu1004/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games
  PS1=(ubu1004)$
  PWD=/home/buildbot/slave-state/ubu1004/nobuild-buildhost03-ubu1004/gst-plugins-bad-ubu1004-master/build
  SHELL=/bin/sh
  TERM=linux
  UPSTART_EVENTS=started
  UPSTART_INSTANCE=
  UPSTART_JOB=buildslave-ubu1004-nobuild
  USER=buildbot
  VIRTUAL_ENV=/home/buildbot/slave-state/ubu1004
 using PTY: False
git version 1.7.0.4
program finished with exit code 0
elapsedTime=0.034024

and interrupt

[Failure instance: Traceback (failure with no frames): <class
'twisted.internet.error.ConnectionLost'>: Connection to the other side
was lost in a non-clean fashion.
]

It's curious that the 1200 second timeout didn't clear the bad state.

Interestingly, although maxbuilds is set to 1 on all
slaves, another builder on the same slave is functioning
normally.

This is with a one-shot slave (but the windows one wasn't).

I hope that narrows down the problem a bit.  Want me to file a bug?
- Dan