On 28/11/2024 20:00, Drew Parsons wrote:
On 2024-11-28 20:40, Emilio Pozuelo Monfort wrote:
On 28/11/2024 20:13, Alastair McKinstry wrote:
(5) mpi4py fails to build with py3.13 (due to cffi dependency).
Builds ok with 'nocheck' allowing more of the stack to build.
Are you sure? It looks like it was already built with py3.13, see
e.g. [1]
I haven't noticed any python 3.13 problem with mpi4py, at least with
the openmpi 5 in unstable.
cffi 1.17 supports python 3.13, and has migrated to testing.
In any case, if there is a problem with mpi4py, the upstream developer
is usually fast to respond.
Ok my notes are out of date on this. It has since been fixed.
(Alastair: this is a good opportunity to ask on behalf of the upstream
mpi4py developer if we could switch on a more complete test suite for
OpenMPI to catch things like PMIX or bigendian issues. mpi4py tests
run a full suite which triggers them, but then we (including I)
misjudge the origin of the bug and file it against mpi4py, and that
taxes the mpi4py developer sorely.
e.g. https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1087800 ,
https://github.com/mpi4py/mpi4py/issues/586
Would to great if we could catch more issues before they get to
mpi4py. At the moment openmpi's debian/tests is only checking that
mpicc compiles sample code successful).
Ok I agree.
There was a package that did full MPI testing separately of
openmpi/mpich (I can't find it right now, and it wasn't well maintained)
but enabling better testing in openmpi/mpich/pmix/ucx is needed.
Drew
Alastair
--
Alastair McKinstry,
GPG: 82383CE9165B347C787081A2CBE6BB4E5D9AD3A5
e: alast...@mckinstry.ie, im: @alastair:mckinstry.ie @amckins...@mastodon.ie
Commander Vimes didn’t like the phrase “The innocent have nothing to fear,”
believing the innocent had everything to fear, mostly from the guilty but in
the longer term
even more from those who say things like “The innocent have nothing to fear.”
- T. Pratchett, Snuff
