[MEDApp] Minor Fixes

[matekelemen]

• remove the dummy definitions of MPI_Comm and MPI_Info that lead to multiple definitions. Forward declaration also fails because med.h indirectly includes mpi.h that defines these as typedef structs that are incompatible with structs. I haven't tried this on non-MPI builds.
• update testing macros
• delete a try-catch-rethrow block from a destructor (destructors are implicitly noexcept)

[philbucher]

I will review tonight, sorry for the delays

[philbucher]

We need to figure out a better solution for the MPI-issue, but thx otherwise

[philbucher]

the issue here is when you compile without MPI

I am not sure how we can go about this. Maybe ifdef? Might be the only option we have for now, can you give it a try?

Also, how did you install the med-lib? By hand or via the package manager?

[matekelemen]

Maybe ifdef?

Yep, I think we'll have to resort to the preprocessor.

Also, how did you install the med-lib? By hand or via the package manager?

I installed salome-medcoupling (should include both the serial and OpenMPI med lib) using the system package manager.

[philbucher]

Ok ifdef it is (at least for now, until someone complains :D), can you please add it?

Also, can you please add how you installed the lib to the readme? I did it already for ubuntu

[matekelemen]

I found that the created geometries were automatically assigned IDs (really high ones) that eventually lead to exceptions when trying to output the resulting ModelPart to MDPA via ModelPartIO::WriteModelPart, so I added explicit IDs for the created geometries.

The current impl obviously won't work with ModelParts that are already populated, but we can worry about those later because the Nodes are created in the same manner.

[philbucher]

I found that the created geometries were automatically assigned IDs (really high ones) that eventually lead to exceptions when trying to output the resulting ModelPart to MDPA via ModelPartIO::WriteModelPart, so I added explicit IDs for the created geometries.

The current impl obviously won't work with ModelParts that are already populated, but we can worry about those later because the Nodes are created in the same manner.

ok thanks that makes sense
Lets add a check in read mode if nodes already exist (I thought I had added that one already). Can you please take care of this one?

Ids of entities and mypy, my programming arch-enemies :D

The base branch was changed.