Build failed in Jenkins: cyrus-imapd-master #756

Greg Banks gnb at fastmail.fm
Tue Aug 7 07:17:45 EDT 2012



On Tue, Aug 7, 2012, at 08:58 PM, Bron Gondwana wrote:
> On Tue, Aug 7, 2012, at 11:56 AM, Greg Banks wrote:
> > 
> > 
> > On Tue, Aug 7, 2012, at 07:12 PM, Jenkins wrote:
> > > See <http://ci.cyrusimap.org/job/cyrus-imapd-master/756/changes>
> > > 
> > > Changes:
> > > 
> > > [gnb] Bug #3726 Add a portability stub for posix_fadvise
> > > 
> > > [gnb] Update K&R code in portability stubs.
> > > 
> > > [gnb] Bug #3726 Add a portability stub for strsep
> > > 
> > > [gnb] Bug #3726 Add a portability stub for memmem
> > > 
> > > [...]
> > > Cassandane::Test::Metronome.basic
> > >     http://ci.cyrusimap.org/job/cyrus-imapd-master/756//testReport/%28root%29/Cassandane__Test__Metronome/test_basic/
> > 
> > Cassandane::Test::Metronome.test_basic (from
> > Cassandane__Test__Metronome)
> > Failing for the past 1 build (Since Failed#756 )
> > Took 1 sec.
> > Error Message
> > 
> > Standard deviation 1.43380587135031 is too high
> > 
> > Hmm, this failure had nothing to do with the recent commits - that test
> > is not even running C code.  The test is just time sensitive.  This is
> > the second time it's happened recently; perhaps the physical hardware is
> > more heavily loaded these days. If it happens again I'll bump up the
> > threshold in the test.
> 
> For sure.  Why are we caring about relative speed of the virtual machine
> across time?

Master has a fork rate limiting feature.  It was very broken.  Now it's
fixed and has a test.  That test needs to measure the actual fork rate
that master allows when fed a rate of connection attempts which is above
the configured fork rate limit.  To generate that load, the test uses
the Metronome class, which uses the POSIX realtime API to generate a
controlled average rate of events.  The test that is failing, measures
the rate stability of the Metronome class itself.  It doesn't need to be
super accurate, but we do need to know if it's like 50% wrong.

> Spurious test failures just lead to being ignored :(

Sure.

-- 
Greg.


More information about the Cyrus-devel mailing list