Random test failures...

[Dag Andersen]

Tirsdag 30 august 2011 16:20:09 skrev Cyrille Berger Skott:
> On Tuesday 30 August 2011, Dag Andersen wrote:
> > Tirsdag 30 august 2011 09:58:33 skrev Cyrille Berger Skott:
> > > Hi,
> > > 
> > > It actually seems we get a lot of random test failures this days:
> > > 
> > > * Plan-TJTester (now pass, but failed on the previous run
> > > http://my.cdash.org/viewBuildError.php?buildid=225474 )
> > > * krita-image-KisProjectionTest (now pass, but failed on the previous
> > > run) * stage-part-TestCustomSlideShows and stage-part-
> > > TestRenameCustomSlideShowCommand are randomly hanging after the test
> > > run (see for instance:
> > > http://my.cdash.org/testDetails.php?test=7239540&build=226002 )
> > > 
> > > I remember seeing others well (in plan's tests), if I see them again I
> > > will update that list.
> > 
> > Plan:
> > The build error seems to be reports. Things go on in that module so we
> > will see when it calms down.
> 
> Sorry, I paste a wrong link, I get this:
> 
> http://my.cdash.org/testDetails.php?test=7235833&build=225827
Ahh, yes. I had a crash here a few days ago in QMutex d'tor. There is no mutex 
in the actual test so must be QTest. Of course the crash disappeared when I 
started to investigate :(
I've run valgrind but can't see any obvious problems that could cause this.
> 
> > Otherwise, my relatively new scripting unit test fails afaics because
> > kross cannot find python. Could this be fixed?
> 
> Yes, going to install krosspython.
Great
> 
> > Then sometimes the CalendarTest times out (it normally takes a few
> > milliseconds) with the output below. The test itself passes, but maybe a
> > crash on termination? Btw, I've never seen this at home.
> > 
> > Totals: 8 passed, 0 failed, 0 skipped
> > ********* Finished testing of KPlato::CalendarTester *********
> > kdeinit4: preparing to launch /usr/lib/libkdeinit4_kconf_update.so
> > kdeinit4: preparing to launch /usr/lib/kde4/kio_archive.so
> > kdeinit4: preparing to launch /usr/lib/libkdeinit4_kconf_update.so
> > kdeinit4: preparing to launch /usr/lib/kde4/libexec/drkonqi
> > kdeinit4: preparing to launch /usr/bin/knotify4
> > kdeinit4: preparing to launch /usr/bin/knotify4
> > <unknown program name>(24535)/ main: Running knotify found
> > Connecting to deprecated signal
> > QDBusConnectionInterface::serviceOwnerChanged(QString,QString,QString)
> > knotify(24532) KNotify::event: 1  ref= 0
> > knotify(24532) KNotify::event: 2  ref= 0
> > ...
> > knotify(24532) KNotify::event: 272  ref= 0
> > knotify(24532) KNotify::event: 273  ref= 0
> > kdeinit4: preparing to launch /usr/lib/kde4/libexec/drkonqi
> 
> Sounds very similar to the stage-part-TestCustomSlideShows and stage-part-
> TestRenameCustomSlideShowCommand.
> 
> The weird thing is that it seems to only affect a limited subset of test.
Yes, I've only seen the calendar test for plan, noone else.
Could be a coincidence, of course, but...
> The only reason I can think of, that is different from running the tests
> "at home" is that the tests are run concurently on the server (12 at the
> same time), so maybe the load affects those tests.

-- 
Mvh.
Dag Andersen