[Buildbot-devel] "but we weren't running, quitting silently" error message

[Damian Krzeminski]

My buildslaves seem to get stuck from time to time with "SlaveBuilder.commandComplete <>  but we 
weren't running, quitting silently". Usually it happens when executing "svnversion" command of 
standard SVN step.
If that happens master thinks that build is in progress and waterfall refuses to display the logfile 
associated with the step that failed.
Other builds on the same slave are OK and I can ping the slave. But restarting the slave seem to be 
the only way to unstuck the build.

I was trying to find some references to this on the list and in sourceforge tracker without much 
luck. Is this a known problem? If someone has a better workaround than restarting the slave please 
let me know.
Thanks,
Damian

Fedora Core 5 2.6.16-1.2133_FC5
Python 2.4.3
Buildbot version: 0.7.4
Twisted version: 1.3.0rc1

twisted.log snippet from the stuck slave:

2006/09/11 15:24 EDT [-] ShellCommand._startCommand
2006/09/11 15:24 EDT [-]  /usr/bin/svnversion .
2006/09/11 15:24 EDT [-]   in dir /home/buildbot/sipx-slave/sipx-fc5-fast/build
2006/09/11 15:24 EDT [-]   watching logfiles {}
2006/09/11 15:24 EDT [-]   argv: ['/usr/bin/svnversion', '.']
2006/09/11 15:24 EDT [-]   environment: {'LC_ALL': 'C', 'SHELL': '/bin/sh', 'SHLVL': '2', 
'PYTHONPATH': ':/usr/local/lib/python2.4/site-packages::/usr/local/lib/python2.4/site-packages', 
'PWD': '/home/buildbot', 'LOGNAME': 'buildbot', 'USER': 'buildbot', 'PATH': '/usr/bin:/bin', 'HOME': 
'/home/buildbot', '_': '/usr/local/bin/buildbot'}
2006/09/11 15:24 EDT [-] sending app-level keepalive
2006/09/11 15:24 EDT [-] command finished with signal None, exit code 0
2006/09/11 15:24 EDT [-] SlaveBuilder.commandComplete <buildbot.slave.commands.SVN instance at 
0xb7d267cc>
2006/09/11 15:24 EDT [-]  but we weren't running, quitting silently