remove h5py from spack and from tests

[Yurlungur]

PR Summary

As requested by @mauneyc-LANL @dholladay00 and motivated by issues @annapiegra is having with oneapi, this PR does the following:

• Stores an hdf5 "gold" file as a release
• Downloads the file in cmake if needed
• Removes py-h5py from spack and cmake dependencies.

PR Checklist

• Adds a test for any bugs fixed. Adds tests for new features.
• Format your changes by using the make format command after configuring with cmake.
• Document any new features, update documentation for changes made.
• Make sure the copyright notice on any files you modified is up to date.
• After creating a pull request, note it in the CHANGELOG.md file

[annapiegra]

Singularity-eos builds with a fresh spack checkout %oneApi w/ the following changes to package.py:

--- a/var/spack/repos/builtin/packages/singularity-eos/package.py
+++ b/var/spack/repos/builtin/packages/singularity-eos/package.py
@@ -19,6 +19,7 @@ class SingularityEos(CMakePackage, CudaPackage):
     maintainers = ["rbberger"]
 
     version("main", branch="main")
+    version("fixed", branch="jmm/goldfiles")    
     version("1.6.2", sha256="9c85fca679139a40cc9c72fcaeeca78a407cc1ca184734785236042de364b942")
     version("1.6.1", sha256="c6d92dfecf9689ffe2df615791c039f7e527e9f47799a862e26fa4e3420fe5d7")
 
@@ -62,7 +63,6 @@ class SingularityEos(CMakePackage, CudaPackage):
     depends_on("[email protected]", when="+tests")
     depends_on("python@3:", when="+python")
     depends_on("[email protected]:", when="+python")
-    #    depends_on("py-h5py", when="+tests build_extra=stellarcollapse")
     depends_on("py-sphinx", when="+doc")
     depends_on("[email protected]", when="+doc")
     depends_on("py-sphinx-multiversion", when="+doc")
@@ -116,9 +116,6 @@ class SingularityEos(CMakePackage, CudaPackage):
     # common MPI dependence
     for _flag in ("~mpi", "+mpi"):
         depends_on("hdf5~cxx+hl" + _flag, when=_flag)
-        depends_on("py-h5py" + _flag, when="+tests build_extra=stellarcollapse " + _flag)
-        #        depends_on("hdf5+hl" + _flag, when=_flag)
-        depends_on("py-h5py" + _flag, when=_flag)
         depends_on("kokkos-nvcc-wrapper" + _flag, when="+cuda+kokkos" + _flag)
 

[Yurlungur]

Thanks, @annapiegra ! @dholladay00 @mauneyc-LANL does this satisfy?

[dholladay00]

Does not including h5py fix any of the issues we see with hdf5 and mpi @Yurlungur @mauneyc-LANL ? I'm curious if we could further simplify the spackage?

[Yurlungur]

I don't think this simplifies hdf5 proper, which needs to be compiled as either with or without MPI.

[Yurlungur]

@mauneyc-LANL I switched to FetchContent. Let me know what you think now.

[mauneyc-LANL]

With cmake@>=3.24, you'll get this warning blurb.

[cmake] CMake Warning (dev) at /usr/share/cmake/Modules/FetchContent.cmake:1267 (message):
[cmake]   The DOWNLOAD_EXTRACT_TIMESTAMP option was not given and policy CMP0135 is
[cmake]   not set.  The policy's OLD behavior will be used.  When using a URL
[cmake]   download, the timestamps of extracted files should preferably be that of
[cmake]   the time of extraction, otherwise code that depends on the extracted
[cmake]   contents might not be rebuilt if the URL changes.  The OLD behavior
[cmake]   preserves the timestamps from the archive instead, but this is usually not
[cmake]   what you want.  Update your project to the NEW behavior or specify the
[cmake]   DOWNLOAD_EXTRACT_TIMESTAMP option with a value of true to avoid this
[cmake]   robustness issue.

See https://cmake.org/cmake/help/latest/policy/CMP0135.html

Not breaking, and for this case I think is over-doing it. But if you want to avoid terminal bloat, you could add

# Avoid warning about DOWNLOAD_EXTRACT_TIMESTAMP in CMake 3.24:
if (CMAKE_VERSION VERSION_GREATER_EQUAL "3.24.0")
  cmake_policy(SET CMP0135 NEW)
endif()