Arne Morten Kvarving
Arne Morten Kvarving
did not look at details but sounds like a fixture initializing mpi et al.
not really, what i mean is if you have a test fixture which calls MPI_Init / MPI_Finalize (directly or indirectly), you are in trouble because those can only be called...
you can see an example of the typical way of solving this in https://github.com/OPM/opm-simulators/blob/master/tests/test_broadcast.cpp in particular note the use of BOOST_TEST_NO_MAIN and implementation of a main method.
ah that old fun. there's a PHG_RANDOMIZE_INPUT parameter that might help. it could also be LB_APPROACH not set to PARTITION.
there is no "corresponding" commit, the different histories are entirely independent. best you have to go by is date.
In principle it sounds like a good idea, but I would do a internet review to see if there isn't an existing system for this already before brewing our own...
jenkins build this opm-common=4545 please
jenkins build this opm-common=4545 please
jenkins build this opm-common=4545 please
jenkins build this please