search

JCSDA / spack

A flexible package manager that supports multiple versions, configurations, platforms, and compilers.
https://spack.io
Other
0 stars 14 forks source link

release/1.60: bug fix for mapl with mvapich2 #449

Closed climbfuji closed 4 months ago

climbfuji commented 4 months ago

Description

Yet another bug fix for mapl for release/1.6.0 because it wants to configure MPI providers itself. Need to add mvapich2, because Hercules uses mvapich2 for spack-stack-1.6.0 with gcc. I will also cherry-pick this and send it to spack develop.

Ignore the failing style tests!

Testing

Corresponding spack-stack PR: https://github.com/JCSDA/spack-stack/pull/1190

mathomp4 commented 4 months ago

@climbfuji Ah. Apologies. I guess I haven't used mvapich2 in a while and didn't realize there was an mvapich2 and mvapich in spack (odd...but 🤷🏼 ):

One suggestion would be:

        elif self.spec.satisfies("^mvapich2"):
            args.append(self.define("MPI_STACK", "mvapich"))

to match the ^mvapich bit below it.[^1]

[^1]: Again, if you aren't running GEOS, this whole thing is moot. Grah. I think I should just say "If unknown, just assume mpich" which is probably not a bad assumption.

climbfuji commented 4 months ago

Thanks @mathomp4 also pushed your bug fix to the spack develop PR

mathomp4 commented 4 months ago

Huh. I'll have to look at that license style issue. I mean, I think it's right...

climbfuji commented 4 months ago

Huh. I'll have to look at that license style issue. I mean, I think it's right...

That's only because it's an old release of spack-stack. As long as it works for spack develop and our spack-stack-dev fork, all good.