[macOS] Some tests do not run on macOS 14 (94% tests passed, 13 tests failed out of 205)

[barracuda156]

Several tests do not run for some reason and self-report as failed, at least on Sonoma / aarch64:

94% tests passed, 13 tests failed out of 205

Total Test time (real) = 104.95 sec

The following tests FAILED:
    121 - core/test/mpi/base/communicator (Not Run)
    122 - core/test/mpi/base/exception_helpers (Not Run)
    123 - core/test/mpi/base/bindings (Not Run)
    124 - core/test/mpi/base/polymorphic_object (Not Run)
    125 - core/test/mpi/base/rank_mapping (Not Run)
    126 - core/test/mpi/distributed/helpers (Not Run)
    127 - core/test/mpi/distributed/matrix (Not Run)
    128 - core/test/mpi/distributed/preconditioner/schwarz (Not Run)
    201 - test/mpi/matrix_reference (Not Run)
    202 - test/mpi/partition_helpers_reference (Not Run)
    203 - test/mpi/vector_reference (Not Run)
    204 - test/mpi/preconditioner/schwarz_reference (Not Run)
    205 - test/mpi/solver/solver_reference (Not Run)
Errors while running CTest

Is this an expected behavior?

[upsj]

No, this is not expected. It seems like there is something going wrong with the test launch itself - What compiler/MPI library are you using? What is the output for the individual tests?

I am running on Sonoma/aarch64, and the tests succeed here

[barracuda156]

@upsj Our port for 1.6.0: https://github.com/macports/macports-ports/blob/master/math/ginkgo/Portfile

My update now to 1.7.0: https://github.com/barracuda156/macports-ports/commit/7c30106e3ef71c81e9a3588447212d6eac9a6417

No patches from our side.

MPICH being used here is 4.1.2 which uses gfortran-13. I can try using another MPICH to see if it makes a difference, a bit later tonight.

[upsj]

What I need is the full output of ctest, because the tests are not failing, but instead can't be run. This might be caused by the mpirun executable path not being found from the FindMPI module

[barracuda156]

@upsj Here are logs:

ginkgo_sonoma_build_log.txt ginkgo_ctest_log.txt

[barracuda156]

@upsj Are logs of any help, or should I provide something else? We definitely want to fix running tests properly.

P. S. There is another issue with tests, but I have to repeat it for 1.7.0 – should be able to do it in about 10 days, when back to PowerPC hardware: https://github.com/ginkgo-project/ginkgo/issues/1260

[upsj]

Indeed, it seems like the MPIEXEC_EXECUTABLE variable is empty when setting up the tests. I see you are specifying it as part of your CMake invocation, is it then properly set in CMakeCache.txt?

EDIT: Sorry, I looked in the wrong place

Could not find executable /opt/local/bin/mpiexec

Is your mpiexec file executable? Does it exist? Are you testing in the same environment that you are building in?