It looks like we may need to update the pmix submodule to avoid the CI failure. @jsquyres I think this is happening because the CI is using HEAD of openpmix. someone (i wonder who) did a bunch of refactoring of pmix/prrte that removed PMIX_SESSION_PROVISION from the pmix header file(s).

i'm sorry - i may not be clear enough. We need to rebase the capstone-devel prrte on top of upstream. or we need to a patch to capstone-devel prrte to build against head of openpmix master.

Note that you got build errors across the board here in CI:

prted/pmix/pmix_server_queries.c:837:71: warning: incompatible pointer types passing 'void **' to parameter of type 'pmix_value_t **' (aka 'struct pmix_value **') [-Wincompatible-pointer-types]
                ret = PMIx_Get(&pproc, PMIX_MEM_ALLOC_KIND, &info, 1, (void**)&value);
                                                                      ^~~~~~~~~~~~~~
/Users/runner/work/_temp/pmixinstall/include/pmix.h:208:51: note: passing argument to parameter 'val' here
                                   pmix_value_t **val);
                                                  ^
prted/pmix/pmix_server_session.c:145:50: error: use of undeclared identifier 'PMIX_SESSION_PROVISION'
        } else if (PMIX_CHECK_KEY(&req->info[n], PMIX_SESSION_PROVISION) ||
                                                 ^
1 error generated.
make[2]: *** [prted/pmix/libprrte_la-pmix_server_session.lo] Error 1

You might want to try building before pushing up to the PR.

No the problem here doesn't have anything to do with the capstone work. it has to do with the fact that certain "changes" were made to the upstream openpmix that require us to either slurp up the corresponding changes in the upstream prrte master or figure out how to change the CI to use a working version of openpmix or else patch the capstone prrte devel branch so it can work with CI using head of upstream openpmix master or something. I believe btw, that open-mpi/ompi#12906 is failing owing to changes introduced in upstream prrte/openpmix that included the code changes that are now breaking CI for this project. So i'm reluctant to just patch both the Open MPI prrte and the capstone prrte with those changes in upstream prrte/openpmix.

One idea to try folks is to modify the .github/workflows yml files to build against a specific sha of openpmix.
See https://stackoverflow.com/questions/73879978/how-to-create-a-github-action-to-checkout-a-specific-commit-in-a-private-reposit

You would want to use the same openpmix sha as what we are currently using in Open MPI for main branch.
See https://github.com/open-mpi/ompi/tree/main/3rd-party. looks like a sha to try is 08e41ed

No the problem here doesn't have anything to do with the capstone work. it has to do with the fact that certain "changes" were made to the upstream openpmix that require us to either slurp up the corresponding changes in the upstream prrte master or figure out how to change the CI to use a working version of openpmix or else patch the capstone prrte devel branch so it can work with CI using head of upstream openpmix master or something.

Yes, PRRTE has always tracked with PMIx. However, I have gone through some effort to "protect" the code so that PRRTE will build with earlier versions. I'm afraid you just haven't kept up with the changes 🤷‍♂️ Current PRRTE head of master branch will build with older PMIx versions down to PMIx v4.2

okay looks like one more fixed needed for make distcheck

ERROR: files left after uninstall:
./lib/libprrte.so.4.0.0
make[1]: *** [Makefile:1026: distuninstallcheck] Error 1
make[1]: Leaving directory '/home/runner/work/prrte/prrte/prrte-gitclone/_build/sub'
make: *** [Makefile:973: distcheck] Error 1