spack repo of hdf5libs and its dependencies for use by DUNE offline

[dingp]

This is an issue to keep track of the effort to make hdf5libs available in the offline software stack.

Per discussion today with Fermilab's SciSoft team and Tom Junk, DUNE DAQ will provide a spack repo containing necessary spack recipe files related to hdf5libs. SciSoft will build the hdf5libs spack module using Fermilab's spack variant, which automatically makes spack modules available as UPS products.

The following tarball contains the spack recipes for hdf5libs and its dependencies. Additionally, the package depends on nlohmann-json, highfive ~mpi, and a few other standard python modules, all of which are available from the spack built-in repo.

There is also an dependency of cetlib in the chain. I did not include it here since DUNE offline might already have that.

spack-repo-hdf5libs-dunedaq-v3.2.0.tar.gz

$ tree spack-repo-hdf5libs-dunedaq-v3.2.0
spack-repo-hdf5libs-dunedaq-v3.2.0
├── packages
│   ├── daq-cmake
│   │   └── package.py
│   ├── daqdataformats
│   │   └── package.py
│   ├── detchannelmaps
│   │   └── package.py
│   ├── detdataformats
│   │   └── package.py
│   ├── ers
│   │   └── package.py
│   ├── hdf5libs
│   │   └── package.py
│   ├── logging
│   │   └── package.py
│   ├── py-moo
│   │   └── package.py
│   └── trace
│       ├── disable_cetmodules.diff
│       ├── install-exec.diff
│       ├── install-scripts-317.diff
│       ├── install-scripts.diff
│       └── package.py
└── repo.yaml

10 directories, 14 files

[gartung]

I get this error trying to build hdf5libs with the v0.18-fermi branch of spack
==> Error: Package 'py-jsonnet' not found.

[dingp]

Sorry, I missed the package and thought it was from the built-in repo. You can find the package.py file here https://github.com/DUNE-DAQ/daq-release/blob/dingpf/gcc12/spack-repos/externals/packages/py-jsonnet/package.py

[gartung]

There is a jsonnet with a +python variant. I ended up checking out the daq-release repo and copying all of the packages from externals into the packages repo in the tarball.

[dingp]

Good to know! We will try to switch to jsonnet +python in the next update of our externals software stack.