bosilca

I can't get your branch to work, it fails with ``` $ python3.11 test/test_ulfm.py Traceback (most recent call last): File "test_ulfm.py", line 1, in from mpi4py import MPI ImportError: dlopen(Python/3.11/lib/python/site-packages/mpi4py/MPI.cpython-311-darwin.so,...

Right, I was using the 5.0-rc without this patch. It would be nice if instead of that error the python wrapper would have raise a non-implemented exception (like you have...

Let me take a loook at this, han should never be involved in any collective related to a flavor of MPI_COMM_SELF.

My take: For the long term only one of these two APIs is blessed by the standard, and that one shall survive. The other will be deprecated once the MPI...

> I am running in non-singleton case for that test. The han_comm_create_new procedure has no error checking, which means that when the SPLIT_TYPE fails line 121 (there are dead procs)...

@abouteiller and @dalcinl I am unable to replicate your issues with the current version (and I merged two pending PRs related to ULFM last Friday), neither on my laptop (OSX)...

I'm still not able to reproduce but I think I know why. The python ULFM tester skips 4 tests, all related to the new functionality. It does use the correct...

@dalcinl I found a typo in your commit that prevented things from working properly on my side. I tried to push to your branch but I'm not allowed, so here...

@dalcinl The requested defines are in. @abouteiller the test seems to pass for me, feel free to merge this PR once all the github actions passed.

This seems to have the same root cause as https://github.com/open-mpi/ompi/pull/11475, the underlying MCA param does not deal gracefully with the re-registration of events, aka multiple calls to opal_utils_init/_fini.