h5py seg-fault at import

[Jerome Kieffer]

Importing h5py (without MPI environment) results in a seg-fault.

Apparently h5py is linked to libhdf5_openmpi.so.103 which requires the program to be started via a mpirun. I wonder what is the rational behind this, so I suspect it is simply a bug.

```
kieffer@p9-01:~$ uname -a
Linux p9-01 5.4.0-18-generic #22-Ubuntu SMP Sat Mar 7 18:06:34 UTC 2020 ppc64le ppc64le ppc64le GNU/Linux

kieffer@p9-01:~$ lsb_release -rd
Description: Ubuntu Focal Fossa (development branch)
Release: 20.04

kieffer@p9-01:~$ dpkg -l |grep h5py
ii python3-h5py 2.10.0-2build2 ppc64el general-purpose Python interface to hdf5 (Python 3)

kieffer@p9-01:~$ python3 -c "import h5py"
*** An error occurred in MPI_Init_thread
*** on a NULL communicator
*** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,
*** and potentially your MPI job)
[p9-01:505264] Local abort before MPI_INIT completed completed successfully, but am not able to aggregate error messages, and not able to guarantee that all other processes were killed!
```

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Simon Chopin (schopin)]

I can't reproduce this in a fresh Focal container, I guess it must have been fixed by https://launchpad.net/ubuntu/+source/openmpi/4.0.3-0ubuntu1

Closing this as Fix Released, please reopen if it still happens for you.