[Pkg-openmpi-maintainers] Bug#846635: openmpi: broken on kfreebsd since 2.x

[Steven Chamberlain]

Package: openmpi
Version: 2.0.1-7
Severity: important
Tags: upstream
User: debian-bsd at lists.debian.org
Usertags: kfreebsd

Hi,

(Greetings from the BSP at TU-Dresden, 2016!)

Updating mpirun-bin from 1.10.2-14 to 2.0.1-7 breaks the testsuite
of dune-common (error below), but also this simple testcase too:

| #include <mpi.h>
| 
| int main(int argc, char *argv[]) {
|         MPI_Init(&argc, &argv);
|         MPI_Finalize();
|         return 0;
| }

fails with:

> $ OMPI_MCA_plm_rsh_agent=/bin/false ./testcase
> 
> [hostname.example.com:96346] PMIX ERROR: UNREACHABLE in file
> src/client/pmix_client.c at line 983
> [hostname.example.com:96347] PMIX ERROR: NOT-SUPPORTED in file
> src/server/pmix_server_listener.c at line 540
> [hostname.example.com:96346] PMIX ERROR: UNREACHABLE in file
> src/client/pmix_client.c at line 199
> --------------------------------------------------------------------------
> It looks like orte_init failed for some reason; your parallel process is
> likely to abort.  There are many reasons that a parallel process can
> fail during orte_init; some of which are due to configuration or
> environment problems.  This failure appears to be an internal failure;
> here's some additional information (which may only be relevant to an
> Open MPI developer):
> 
>   init pmix failed
>   --> Returned value Unreachable (-12) instead of ORTE_SUCCESS

It is the same error in dune-common:

> FAIL: mpicollectivecommunication
> ================================
> 
> [falla:96461] PMIX ERROR: UNREACHABLE in file src/client/pmix_client.c
> at line 983
> [falla:96461] PMIX ERROR: UNREACHABLE in file src/client/pmix_client.c
> at line 199
> [falla:96489] PMIX ERROR: NOT-SUPPORTED in file
> src/server/pmix_server_listener.c at line 540
> --------------------------------------------------------------------------
> It looks like orte_init failed for some reason; your parallel process is
> likely to abort.  There are many reasons that a parallel process can
> fail during orte_init; some of which are due to configuration or
> environment problems.  This failure appears to be an internal failure;
> here's some additional information (which may only be relevant to an
> Open MPI developer):
> 
>   init pmix failed
>   --> Returned value Unreachable (-12) instead of ORTE_SUCCESS

What I understand so far is that, since MPI version 2.x, pmix_native.c
tries to check peer credentials on a socket.  But FreeBSD lacks
SO_PEERCRED and there isn't yet code to implement this any other way.

FreeBSD ports hasn't packed MPI version 2.x yet so, probably this
is just unimplemented/unported yet.

-- System Information:
Debian Release: stretch/sid
  APT prefers unstable
  APT policy: (500, 'unstable'), (1, 'experimental')
Architecture: kfreebsd-amd64 (x86_64)

Kernel: kFreeBSD 10.1-0-amd64
Locale: LANG=en_GB.UTF-8, LC_CTYPE=en_GB.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: sysvinit (via /sbin/init)