On 2020-08-11 09:46, Drew Parsons wrote:
Thanks Jed. I think you're replying here in response to discussion raised elsewhere. Could you provide a url link if its publicly accessible so we can catch the rest of the context?
Oh wait, I'm wrong aren't I? Your reply is to Christian Engwer's question in Debian Bug#791676 (https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=791676)
The glitch with mpifort a.f90 and gfortran-10/gfortran-9 might be a temporary problem. We're still in the transition from gcc-9 to gcc-10. "mpifort a.f90" succeeds in debian unstable. The problem here is that the gfortran package is 1.185.1 (defaulting to gcc-9) in debian testing, but 1.188 (gcc-10) in debian unstable. The transition is held up by seqan, which is failing tests (see https://tracker.debian.org/pkg/gcc-defaults)
Since our default MPI is OpenMPI, we didn't catch the inconsistency in mpich soon enough. mpich 3.4~a2+really3.3.2-2 was recently built against gcc-10, but its gfortran dependency was not strict and so it migrated into debian testing ahead of gfortran (with the gfortran-10 default).
Drew
