[Buildbot-devel] Stochastic failure of rm -rf in CVS steps

[Jean-Paul Calderone]

On Tue, 23 Sep 2008 09:55:40 -0500, Charles Bacon <bacon at mcs.anl.gov> wrote:
>I moved my buildbot from one machine to another, and I am now getting
>intermittent failures of the CVS update/checkout steps.  I'm really
>baffled, because the failing step appears to be the rm -rf: http://buildbot.globus.org/builders/kempt-complete/builds/1031/steps/cvs/logs/stdio
>
>The thing is, I can't come up with a reason why rm -rf would return a
>-1.  It never does when I run the commands by hand, and it usually
>works when buildbot tries it.  The failure rate looks to be about 10%
>or lower.  The twistd.log has this extra info:

You're probably seeing a known issue described in several tickets in the
issue tracker (eg <http://buildbot.net/trac/ticket/255>).

Jean-Paul