[Debichem-devel] Bug#816590: Bug#816590: elpa: still FTBFS on mipsel/powerpc

[Michael Banck]

Hi,

On Thu, Mar 03, 2016 at 10:10:41AM +0100, Emilio Pozuelo Monfort wrote:
> Source: elpa
> Version: 2015.05.001-6
> Severity: serious
> 
> Your package failed to build on mipsel/powerpc:
> 
> Build killed with signal TERM after 150 minutes of inactivity
> 
> Logs at https://buildd.debian.org/status/logs.php?pkg=elpa&ver=2015.05.001-6

Well, yeah.

I've extensively debugged the powerpc issue on a porter box the other
day, and it only hangs if I use more than one MPI process (otherwise it
segfaults unless we use a very small testsuite size). In other news, I
think it's an OpenMPI issue, but the mpi-testsuite package I've uploaded
is still in NEW.

So I've done the following now for ELPA:

 * I've (hopefully) forced mpiexec -n 1 (i.e. one process only) on
   powerpc.

 * I've further reduced the testsuite matrix size from 2000 to 500 on
   mipsel.

Let's see how -7 goes, I've just uploaded it.


Michael