Minimum cmake for trilinos

[xylar]

Checklist

• Document (in a comment titled Testing in this PR) any testing that was used to verify the changes

[xylar]

@trhille, this is the more permanent fix related to what I asked you to test in
E3SM-Project/mache#109 (comment)
Could you try this out and report back? It will need to be tested in conjunction with:

--mache_fork xylar/mache --mache_branch  update_cmake_for_albany

as before.

[trhille]

@xylar, I'm an error when I run this:

./conda/configure_compass_env.py --conda ${CONDA_BASE} --mache_fork xylar/mache --mache_branch  update_cmake_for_albany --update_spack --spack /lustre/scratch4/turquoise/.mdt1/trhille/spack_test --tmpdir ${TMPDIR} --compiler gnu --mpi mpich --recreate --with_albany

...

recreating environment: dev_compass_1_2_0-alpha_4_gnu_mpich_albany
==> Error: /lustre/scratch4/turquoise/.mdt1/trhille/test_COMPASS_PR549/minimum_cmake_for_trilinos/conda/build_chicoma-cpu_gnu_mpich_albany/dev_compass_1_2_0-alpha_4_gnu_mpich_albany.yaml:14: CommentedMap([('[email protected]', None)]) is not valid under any of the given schemas

[xylar]

Okay, better leave this to me to debug. This method is way too slow.

[xylar]

Thank you for being a good sport about a very tedious process.

[trhille]

No problem!

[xylar]

@trhille, these latest fixes are working for me on Chyrsalis. Could you give it another go?

[xylar]

It probably doesn't make sense to build new spack environments for all machines and configurations because of this change. So far, there is still no change in compass dependencies that requires new spack environments. So I suggest we make this an "in progress" PR until we need it.

[trhille]

@xylar, I'm sorry to say I'm getting an error when running this on Chicoma. The traceback is pretty heinous, but here are the bits that seem relevant:

==> Installing trilinos-for-albany-develop-g5rgby5vl5bvyuyip66ahy2cwxkh2o6k
==> No binary for trilinos-for-albany-develop-g5rgby5vl5bvyuyip66ahy2cwxkh2o6k found: installing from source
==> No patches needed for trilinos-for-albany
==> trilinos-for-albany: Executing phase: 'cmake'
==> trilinos-for-albany: Executing phase: 'build'
==> Error: ProcessError: Command exited with status 2:
    'make' '-j16'

4 errors found in build log:
...
>> 31070    /lustre/scratch4/turquoise/.mdt1/trhille/spack_test/spack_for_mache_1.13.0/opt/spack/linux-sles15-zen2/gcc-12.1.0/boost-1.80.0-6zelek6xnl45a65oyer3jppic44w3i6h/include/boost/stacktrace/fram
              e.hpp:18:10: fatal error: boost/core/explicit_operator_bool.hpp: No such file or directory
... 
>> 31074    make[2]: *** [packages/stk/stk_util/stk_util/util/CMakeFiles/stk_util_util.dir/build.make:247: packages/stk/stk_util/stk_util/util/CMakeFiles/stk_util_util.dir/ReportHandler.cpp.o] Error 1
...
>> 31089    make[1]: *** [CMakeFiles/Makefile2:12476: packages/stk/stk_util/stk_util/util/CMakeFiles/stk_util_util.dir/all] Error 2
...
  >> 31375    make: *** [Makefile:159: all] Error 
See build log for details:
  /lustre/scratch4/turquoise/.mdt1/trhille/spack_temp/spack-stage/spack-stage-trilinos-for-albany-develop-g5rgby5vl5bvyuyip66ahy2cwxkh2o6k/spack-build-out.txt

==> Warning: Skipping build of albany-develop-hjhdlza4dnb36bgfzxe357hquncv4ivo since trilinos-for-albany-develop-g5rgby5vl5bvyuyip66ahy2cwxkh2o6k failed
==> Error: albany-develop-hjhdlza4dnb36bgfzxe357hquncv4ivo: Package was not installed
...
==> Error: Installation request failed.  Refer to reported errors for failing package(s).

[xylar]

AAAArgh!

I don't think that piece of the stack trace is very helpful. I can give it a try myself...

[trhille]

😞

[xylar]

@trhille, it worked fine for me. I deleted my spack_test and spack_tmp directories and started fresh. Maybe that made a difference, not sure.

[trhille]

Okay, I can try that as well.

[trhille]

@xylar , okay I was able to run this successfully after removing my old spack_test and spack_temp directories! I'll set up and run full_integration now.

[xylar]

Thanks @trhille. I ran into (and fixed) an issue I had with building the polaris environment on Chrysalis using the new environment without system cmake. I don't yet understand why my compass test didn't reveal the same issue but maybe it was just luck in terms for what was already in the environment I was testing.

Anyway, it sounds like (assuming the full_integration testing goes well) we're done here until we have an actual need to build new spack environments. Luckily, we're not there yet.

[trhille]

All full_integration test cases passed on Chicoma.

[trhille]

I don't know if it makes sense to approve a draft PR, but I'm doing so anyway!

[xylar]

I think this is a case where approving makes sense. Thanks, @trhille!

[xylar]

Closing this because these changes are in #555