v2.x: fix and revamp MPI_Pack and friends #3070
Conversation
|
This is an alternative to #3060 |
|
@ggouaillardet #3060 was merged. Does that create a logical conflict with this one? |
|
i will rebase tomorrow and fix the conflict if any |
There was a problem hiding this comment.
Thanks @ggouaillardet. You manage to extract the minimally invasive code from the trunk, that fixes the problem in 2.x.
While going through the code I noticed that ompi_datatype_dfd is assigned but never used in the 2.x. I think it deserves to disappear (as it did in the trunk).
|
@hppritcha Need to examine this one a little more carefully because #3060 was merged before this one (this one might be ok -- and @ggouaillardet might just rebase it / make it ok depending on the timing if we want to get this in before tonight's MTT). |
…type so it can be directly used by test/datatype/external32 (back-ported from commit open-mpi/ompi@5932287) Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
this regression was previously introduced in open-mpi/ompi@221e6e2 (back-ported from commit open-mpi/ompi@f2e33c7) Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
this fixes a regression from open-mpi/ompi@f2e33c7 (cherry picked from commit open-mpi/ompi@7b803ac)
Thanks Alex and the folks at Mellanox for the help. Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp> (back-ported from commit open-mpi/ompi@880f2d5)
ggouaillardet
force-pushed
the
topic/v2.x/mpi_pack_unpack
branch
from
570806c to
0c69572
Compare
|
@ggouaillardet Thank you! |
|
@hppritcha Good to go when CI finishes. Would be good to get in the next ~80 minutes so that it's part of tonight's nightly snapshot build / tonight's MTT. |
|
Travis is having an outage tonight and builds are currently paused. I'm merging so that this fix gets in for tonight's MTT runs. |
3 participants
Fixes #3065