Add MPI_SESSION_NULL to fortran

[boi4]

Hello,

I wanted to write some Fortran code that uses MPI Sessions and realized that MPI_SESSION_NULL is not available in Fortran. As a consequence, I added support myself by adding the constant to the Fortran header generator + implementing comparison functions for type(MPI_Session) and by making MPI_Session_finalize return an MPI_SESSION_NULL handle on return.

Here are is a code pieces that wouldn't compile (because MPI_SESSION_NULL was not defined), but works now:

program main
  use mpi
  implicit none

  integer :: ierror
  integer :: session

  call mpi_session_init(MPI_INFO_NULL, MPI_ERRORS_ARE_FATAL, session, ierror)
  if (ierror /=0) stop 'mpi session init fail'

  call mpi_session_finalize(session, ierror)
  if (ierror /=0) stop 'mpi session finalize fail'

  if (session /= MPI_SESSION_NULL) then
    print *, "Session handle was not set to MPI_SESSION_NULL!"
  end if

end program main

[ompiteam-bot]

Can one of the admins verify this patch?

[github-actions]

Hello! The Git Commit Checker CI bot found a few problems with this PR:

c3bbab7: Add MPI_SESSION_NULL to fortran

• check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

[boi4]

Just added the missing "signed off by" via force push

[hppritcha]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[hppritcha]

looks like something's borked up with mellanox ci system:

[1679155353.295017] [swx-jupiter04:24518:0]     ucp_context.c:1176 UCX  WARN  network device 'mlx5_0:1' is not available, please use one or more of: 'eth0'(tcp), 'lo'(tcp)

[yosefe]

looks like something's borked up with mellanox ci system:

[1679155353.295017] [swx-jupiter04:24518:0]     ucp_context.c:1176 UCX  WARN  network device 'mlx5_0:1' is not available, please use one or more of: 'eth0'(tcp), 'lo'(tcp)

@artemry-nv

[hppritcha]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[artemry-nv]

Still investigating the Mellanox CI issue, looks like a CI setup misconfiguration issue after recent lab shutdown.

[artemry-nv]

We run Mellanox CI for main (failed) and v5.0.x (passed) - looks there may be a regression on main. Potentially this one may relate #11499 (CI passed)

[hppritcha]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[boi4]

Should I rebase my commits onto main once the regression is fixed?

[hppritcha]

yes please do.

[jsquyres]

@boi4 The regression was fixed a few days ago; can you rebase this PR? Thanks!

[jsquyres]

Can you fix the indenting here? We use 4-space tabs in Open MPI. Thanks!

[boi4]

Yes, I don't know how that happened. It's fixed now, thanks for noticing!

[github-actions]

Hello! The Git Commit Checker CI bot found a few problems with this PR:

8f82e84: Fix invalid indentation in session_finalize_f.c

• check_signed_off: does not contain a valid Signed-off-by line

Please fix these problems and, if necessary, force-push new commits back up to the PR branch. Thanks!

[jsquyres]

Somehow you ended up with a bunch of extra commits on this PR. Can you reduce it back down to just the one commit?

[boi4]

Sorry, I am not sure how to do this. I just moved the two commits (feature+whitespace fix) to the end of my branch, hoping that github will detect that the other commits align with the main branch.
I will try to find a fix for this.

[boi4]

Ok. I believe I have fixed it now. Sorry for the confusion.

[jsquyres]

@boi4 Thank you! Sorry to be picky, but could you squash those 2 commits down into 1 commit?

[boi4]

@boi4 Thank you! Sorry to be picky, but could you squash those 2 commits down into 1 commit?

Done

[jsquyres]

Done

Awesome; thanks so much.

[jsquyres]

@boi4 Could you cherry pick this to the v5.0.x branch?

[boi4]

Sure, I have opened a new PR: #11533