The unit tests are quite fast after the initial jitting. Only the demos
take time. We could just pick a subset of the demos for a quicker
regression suite.

I also have a sketch of a framework for running serial demos in parallel
with timeouts, that would help somewhat.
26. nov. 2014 10:19 skrev "Anders Logg" <[email protected]>:

> On Wed, Nov 26, 2014 at 09:50:49AM +0100, Johannes Ring wrote:
> > On Wed, Nov 26, 2014 at 9:27 AM, Garth N. Wells <[email protected]> wrote:
> > > There a lots of failures on the buildbots at the moment that I can't
> > > reproduce locally. A number seem to be timing out in parallel. What
> can we
> > > do about it? Do the buildbots enough resources to run in parallel?
> >
> > I think most of the random timeouts will be fixed by upgrading PETSc
> > to the latest version. This has been on my todo list for a while. I
> > will try to do it this week.
> >
> > > My medium term wish is to have the tests run inside Docker containers
> so
> > > that anyone can run and debug a buildbot system configuration locally.
> >
> > Can we have different versions of Ubuntu, Debian or other Linux
> > distributions inside the Docker containers? What about OS X?
>
> Another pressing issue regarding buildbots is to reduce the time it
> takes to run the tests. The current tests take about 2 hours to run,
> which does not encourage developers to run them often. Even the
> quicktest is very expensive to run.
>
> --
> Anders
> _______________________________________________
> fenics mailing list
> [email protected]
> http://fenicsproject.org/mailman/listinfo/fenics
>
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics

Reply via email to