Jeff Hammond

This is the result with `printf` of the two arguments to the assert that fails... ``` ~/MPI/mpich/build$ ~/MPI/mpich-nvhpc-ch4-ucx-install/bin/mpirun -n 4 ./examples/cpi [proxy:0@oppenheimer] cache_put_flush (../../../../src/pm/hydra/proxy/pmip_pmi.c:185): assert (s) failed Process 0 of...

This is the result with `printf` of the two arguments to the assert that fails... ``` ~/MPI/mpich/build$ ~/MPI/mpich-nvhpc-ch4-ucx-install/bin/mpirun -n 4 ./examples/cpi [proxy:0@oppenheimer] cache_put_flush (../../../../src/pm/hydra/proxy/pmip_pmi.c:185): assert (s) failed Process 0 of...

I'm not really sure what I am looking at, but the only value I can reason about is `*p=-allgather-shm-1-0`, which comes from this: ```c while (p) { struct pmip_kvs *s...

MPI 4.1 Chapter 10 Page 473 Line 21.5 says > An implementation must support info objects as caches for arbitrary (key,value) pairs, regardless of whether it recognizes the key. Whoever...

I can probably pound it out in a day with Cursor the way I did the RMA stuff. Writing tests is probably the most useful thing.

we can just disable `least` and `fast` types. try https://github.com/jeffhammond/vapaa/pull/51.

regarding the "Not MPICH" warning, the consequence of this is that using noncontiguous subarrays AND user-defined datatypes at the same time won't work. this is the most ridiculous use case...

I think this was fixed by #51. If not, you can reopen it.

You tested this locally already, right?

There is no problem compiling Open MPI with Clang and you are free to combine Clang as the C and C++ compiler with `gfortran` as the Fortran compiler.