You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I marked this as "minor" because although technically it's a bug in MPI conformance, this bug has been un-noticed -- for years -- in both Open MPI and MPICH. So it's definitely not too important.
But OTOH, it should be simple cherry-picks from master to v4.1.x. Hopefully, it's easy to do.
The text was updated successfully, but these errors were encountered:
This is simple if we first cherry-pick 98bc7af (change the definition of type(MPI_Status) in use-mpi-f08).
I cannot remember what we decided about that
this breaks ABI, so no backport
the code is broken, so breaking ABI is a lesser evil here
If we choose not to backport the type definition change, we would have reimplement the conversion subroutine
(no more memcpy() ...). That's not rocket science, but a bit time consuming.
can you please advise on how you want to move forward?
(I should be able to find some time next week)
Excellent point -- I forgot about the ABI-breaking stuff. Thanks for the reminder.
Let's not bring this over to v4.1.x, then. We only found this bug by an internal audit, not a real MPI code -- this is pretty low priority. We'll let it all be fixed in v5.0.
Bring over #7762 and #8043 to the v4.1 branch.
I marked this as "minor" because although technically it's a bug in MPI conformance, this bug has been un-noticed -- for years -- in both Open MPI and MPICH. So it's definitely not too important.
But OTOH, it should be simple cherry-picks from master to v4.1.x. Hopefully, it's easy to do.
The text was updated successfully, but these errors were encountered: