Skip to content

Conversation

ggouaillardet
Copy link
Contributor

ompi_request_check_same_instance(): ignore requests whose communicator is ompi_mpi_comm_null. That can occur when pml/ucx is used and mca_pml_ucx_completed_request_init() was invoked.

Thanks Christian Huettig for the report and helping with the troubleshooting.

Refs. #12942

ompi_request_check_same_instance(): ignore requests whose communicator
is ompi_mpi_comm_null. That can occur when pml/ucx is used and
mca_pml_ucx_completed_request_init() was invoked.

Thanks Christian Huettig for the report and helping with the troubleshooting.

Refs. open-mpi#12942

Signed-off-by: Gilles Gouaillardet <[email protected]>
@ggouaillardet
Copy link
Contributor Author

FWIW, I think

ompi_req->req_mpi_object.comm = &ompi_mpi_comm_null.comm;
is triggering the bug.

@janjust
Copy link
Contributor

janjust commented Dec 16, 2024

@ggouaillardet is this still a draft? Can it be merged?

@ggouaillardet ggouaillardet marked this pull request as ready for review December 17, 2024 01:22
@ggouaillardet
Copy link
Contributor Author

@janjust sure, let's merge it!

@janjust janjust merged commit cfb4d38 into open-mpi:main Dec 17, 2024
15 checks passed
@janjust
Copy link
Contributor

janjust commented Dec 17, 2024

@ggouaillardet thanks! I opened the v5.0.x cherry-pick here: #12984 mind give it a quick thumbs up?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants