Fixes to shmem

[hpcraink]

Dear @jladd-mlnx and @miked-mellanox ,
could You please review this PR?

The fix to oshmem/shmem/fortran/shmem_put_f.c has been in since the earliest logs of git's repo: I see df7654e

AFAICT there are no other missing MCA_SPML_CALL and MCA_ATOMIC_CALL.

[mike-dubman]

👍

[jsquyres]

@hpcraink @miked-mellanox @jladd-mlnx Can someone file a corresponding PR for v2.0.1? Thanks!

[karasevb]

@hppritcha @jsquyres
We are considering to PR this to ompi-release to be able to PR #1883 on top of it.

It seems like portion of this PR depends on open-mpi/ompi-release#1041 that is targeted for v2.1.0 while this one is set for v2.0.1 milestone.
Does that mean that we want to defer this PR and #1883 till v2.1.0 too?

[artpol84]

@hppritcha @jsquyres @jladd-mlnx @miked-mellanox Any thoughts?

[jsquyres]

Thanks for the ping; sorry for the delay in replying.

My $0.02:

• v2.x: oshmem: Align OSHMEM API with spec v1.3 (nonblocking put/get) (this PR depends on #913) ompi-release#1041 is clearly a 2.1.0 PR (because it's a new feature)
• these other two fixes (Fixes to shmem #1867 and oshmem/fortran: fix warning mesages #1883) are possible candidates for 2.0.1 (because they're bug fixes)
• however, some/all of these two bug fixes are dependent upon the new feature.
• I think it would be ok to split apart Fixes to shmem #1867 and oshmem/fortran: fix warning mesages #1883 into:
  1. what is needed for v2.0.1
  2. what is needed for v2.1.0

and PR them accordingly.