[gmx-developers] How to find out why a regression test is failing in Jenkins

[Mark Abraham]

On Thu, Dec 19, 2013 at 2:34 AM, Szilárd Páll <pall.szilard at gmail.com>wrote:

> On Wed, Dec 18, 2013 at 11:33 AM, Carsten Kutzner <ckutzne at gwdg.de> wrote:
> > Hi,
> >
> > currently the regressiontests that I added for the comp. e-phys
> > functionality fail in Jenkins
> > (but not on my own machine). I get the following output in Jenkins (for
> each
> > of the tests):
> >
> > Failed
> >
> > complex.swap_x
> >
> > Failing for the past 1 build (Since #3533 )
> > Took 0 ms.
> > Error Message
> >
> > Erorrs in
> > Stacktrace
> >
> >
> > The raw console output tells me that a .tpr file could not be produced,
> but
> > I cannot see why.
> > Where do I find the output produced by grompp?
>
> The outputs are on the build slaves, so one needs to log in an
> manually check them; I grabbed the ones from the following config:
>
> http://jenkins.gromacs.org/job/Gromacs_Gerrit_master/3533/OPTIONS=Compiler=gcc%20CompilerVersion=4.7%20GMX_GPU=OFF%20CUDA=4.2%20GMX_DOUBLE=ON%20GMX_MPI=OFF%20NRANKS=1%20GMX_CPU_ACCELERATION=SSE4.1%20host=bs_nix1204,label=bs_nix1204/
>
> You can get the tarball from here: https://db.tt/THnW9JKa


Thanks.

Cheers,
> --
> Szilárd
>
> PS: couldn't we get the content of .out files to display in the test
> result jenkins page? Alternatively, we could copy back some of the
> output files to the jeknins server, but this could quickly turn into a
> storage challenge.


My understanding is imperfect, but the readily accessible outputs are
stdout/stderr, and those that get sent to an XML file that CTest and/or
Jenkins know to deal with (and how).

In principle, we might be able to redirect further content such as the
*.out files to fields in the XML files, so that error output can be readily
accessed from the web interface. I'm not very keen to invest more time in
the perl test harness, though.

Alternatively, Jenkins can probably be told to make certain output files
accessible, but maintaining the lists is a bit awkward. Already we have
duplication in that CTest knows about the list of regressiontests
directories from content in the source repo, and the perl harness has the
same list in the regressiontests repo (and I recently reduced some internal
duplication there). Jenkins having a magic list of files/globs to preserve
would be similarly icky to keep remembering to maintain.

Tarballing the whole workspace automatically (only in case of failure?)
could probably be done, not sure how easy maintaining the storage would be.

Mark

>
> > 11:14:43 Abnormal return value for 'gmx grompp -maxwarn 10 -n index
> >>grompp.out 2>&1' was 255
> > 11:14:43 No topol.tpr file in swap_x. grompp failed
> > 11:14:43 FAILED. Check  files in swap_x
> >
> > Thanks,
> >   Carsten
> > --
> > Gromacs Developers mailing list
> >
> > * Please search the archive at
> > http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List before
> > posting!
> >
> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >
> > * For (un)subscribe requests visit
> > https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developersor
> > send a mail to gmx-developers-request at gromacs.org.
> --
> Gromacs Developers mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developersor send a mail to
> gmx-developers-request at gromacs.org.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20131220/bb02349c/attachment.html>