[mpich-discuss] Segfault in MPI_Comm_split on OS X (MPICH 3.2)

[Dmitriy Morozov]

Indeed, that's exactly the problem. Just tested it, and with the
alignment fix, the segfault goes away.

Thanks.
Dmitriy

On Fri, Aug 19, 2016 at 12:48 PM, Eric A. Borisch <eborisch at gmail.com> wrote:
> This might be the alignment bug.
> http://lists.mpich.org/pipermail/discuss/2016-May/004764.html
>
> You can try the trivial patch yourself to find out. It was already
> addressed in head; there were comments that a patched version might be
> cut...
>
> - Eric
>
> On Fri, Aug 19, 2016 at 11:43 AM, Dmitriy Morozov <dmitriy at mrzv.org> wrote:
>> When I compile and run the following code on my mac (running OS X
>> 10.11.6, El Capitan), with MPICH 3.2 built from source (via Homebrew),
>> I get a segfault in MPI_Comm_split. If I do the same with a recent
>> clone from the git repository (specifically, d8bb1df), everything
>> works fine. I'm wondering if this is a known problem in 3.2.
>>
>> Thanks.
>> Dmitriy
>>
>> // test-mpi.c
>>
>> #include <mpi.h>
>>
>> int main(int argc, char** argv)
>> {
>>     MPI_Init(&argc, &argv);
>>
>>     MPI_Comm world = MPI_COMM_WORLD;
>>
>>     int rank;
>>     MPI_Comm_rank(world, &rank);
>>
>>     MPI_Comm local;
>>     MPI_Comm_split(world, rank, 0, &local);
>>
>>     MPI_Finalize();
>> }
>>
>> mpicc test-mpi.c -o test-mpi
>> mpirun -n 2 ./test-mpi
>> _______________________________________________
>> discuss mailing list     discuss at mpich.org
>> To manage subscription options or unsubscribe:
>> https://lists.mpich.org/mailman/listinfo/discuss
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss
_______________________________________________
discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss