Fix wrong term "process group with Cartesian structure" in section 7.5.5

[cniethammer]

Problem

The text describing MPI_Cart_rank in Section 7.5.5 uses the term "process group with Cartesian structure".
This is wrong, as topology information is added to communicators as specified in 7.3.
So, the interface of MPI_Cart_rank is build around a communicator as input - and not a process group.

Proposal

Replace "process group with Cartesian structure" with "communicator with an associated Cartesian topology"

Impact on Implementations

• Behavior of implementation clarified

Impacts on Users

• Behavior of implementation clarified

References

PR: https://github.com/mpi-forum/mpi-standard/pull/280
The latest PDF file (chapter with marked changes): topol_fix-unspecified-cartesian-process-group.pdf

[tonyskjellum]

As discussed, this needs to move to and through a committee. @GuillaumeMercier --- are you OK with my putting through the Collective et al. WG? I think you said that :-) Please advise.

We should put this forward as an erratum. I will look for @cniethammer (Christoph's) other ones too.

@RolfRabenseifner, please do not create crypto-errata without WG :-) Wait---Now I see, you are the chapter author for the Topologies WG :-)

Thank you all.

[tonyskjellum]

@cniethammer , you need to do your PR through the right WG's github, not through your personal one. That way, it is fully WG-accessible and shared with the requisite WG. Once I confirm with @GuillaumeMercier , we will do so under collective or topology github presence, and you can push your branch there, and do a new PR. Personal forks are not the right process!
Thank you, Tony

[GuillaumeMercier]

@tonyskjellum: you can take over this issue/PR in your own WG if you wish.

[cniethammer]

@cniethammer , you need to do your PR through the right WG's github, not through your personal one. That way, it is fully WG-accessible and shared with the requisite WG. Once I confirm with @GuillaumeMercier , we will do so under collective or topology github presence, and you can push your branch there, and do a new PR. Personal forks are not the right process!
Thank you, Tony

Thanks.
Waiting for a go to move and make the PR at the WG repo.

[wesbland]

Is this intended for MPI 4.0 or MPI 4.1?

[cniethammer]

@RolfRabenseifner or @tonyskjellum could one of you label this for mpi-4.0

Chapter PDF with marked change is attached to the initial comment.

[wesbland]

2020-12-07 Errata vote results

Yes	No	Abstain	Missed	Result
29	0	1	0	Passed