Automatic reply from Jenkins: Can I test this?

Jenkins: ok to test

Test PASSed.

• ALADIN: trivial patch needed
• ESMF: look OK (both mpich and mpich2 as values are known so the mpi_family.lower() should be fine), cfr. http://www.earthsystemmodeling.org/esmf_releases/public/ESMF_3_1_0rp3/ESMF_usrdoc/node7.html)
• imkl: needs patch to handle MPICH case, should use mpi_family() when possible (i.e. for non-dummy toolchains)
• NWChem: trivial patch needed
• Rmpi: seems OK, it only needs to distinguish between the high-level types OpenMPI vs MPICH, and make an exception for Intel MPI
• ScaLAPACK & SCOTCH: looks fine too, are using mpi_family but no patch needed

required changes in geimer#1

Builds with the touched easyblocks should be retested once that PR in included in here.

I've tested installing the existing Scalasca and Score-P gompi/1.4.12-no-OFED easyconfigs using the gmpolf/1.4.8 toolchain from easybuilders/easybuild-easyconfigs#1217, on top of this PR and easybuilders/easybuild-framework#1112 (which wraps around easybuilders/easybuild-framework#1101).

Scalasca v2.0 and Score-P v1.2.1 (and their dependencies) were not a problem, but Scalasca v1.4.3 didn't work, see https://gist.github.com/boegel/ae07827244011a32f87e.

I don't know how much of an issue this is since i) it's an old version, and ii) something else may be causing this problem. I haven't checked whether building Scalasca v1.4.3 with gmolf/1.4.8 worked prior to the changes we're dealing with here. @geimer: thoughts?

I'm also checking the installation of the following with gmolf/1.4.8 (based on the existing goolf/1.4.10 easyconfigs), since the respective easyblocks were touched in this PR (once geimer#1) is merged in, results are pending:

• ALADIN 36t1_op2bf1
• ESMF v5.3.0
• ESMF v6.1.1

I'm also retesting all existing imkl easyconfigs, to ensure nothing broke with the changes made in geimer#1.

Update: building ALADIN, and NWChem (v6.1.1) with the gmpolf/1.4.8 toolchain works without problems. Reinstalling all the imkl easyconfigs (on top of geimer#1) worked fine too.

Building ESMF v5.3.0 or v6.1.1 with gmpolf/1.4.8 is a problem though, and fails with:

make: mpiCC: Command not found

So, this needs a closer look, but this is not caused by any of the changes being discussed here, it simply never worked. :-)

@boegel: Regarding ESMF: Though both mpich and mpich2 are recognized, mpich may actually mean MPICH 1.x which is quite different from MPICH 2/3. I believe MPICH 1.x created an mpiCC wrapper for C++, whereas MPICH 2/3 use mpicxx or mpic++. Therefore I suggest to try again with mapping MPICH to mpich2.

@boegel: Don't waste any more time on Scalasca v1.4.3. It's a bug in our code base and not caused by these changes. It also won't compile with a recent Open MPI (>= v1.7). We are likely to release a v1.4.4 soon where these issues will be fixed (a final bugfix release for the unteachable who are not willing to switch to the better technology ;-)).

Test PASSed.

@boegel: Thanks for taking care of the missing easyblocks!

just a confirmation: the ESMF issue is fixed with the change included in geimer#1, so this should be good to go in

Good to go, thoroughly tested on top of easybuilders/easybuild-framework#1112