Build external packages with spack in a work area

Developers occasionally need to utilize a different variant, a new version, or a completely new external package before its inclusion and deployment in the DUNE DAQ external software stack on CVMFS. Here's a step-by-step guide on how to do this:

1. Create a work area using the -s flag with the dbt-create command. This will generate a .spack subdirectory within the work area, with the external stack and release stack as upstreams to it.
2. For the external package, either copy or create the Spack recipe file and place it in the directory: <workarea>/.spack/spack-repo/packages/<package_name>/package.py. Detailed instructions on creating this recipe file are provided below.
3. Run the spack install command, for example spack install <package>@version%[email protected] arch=linux-almalinux9-x86_64. Replace <package> with the package name, version with the desired version, and specify the appropriate compiler and architecture.
4. After installation, load the newly installed Spack package and run dbt-build to build your DUNE DAQ packages. If dbt-build complains about dbt-workarea-env not being run, reload the Python virtual environment as follows:

deactivate
source .venv/bin/activate

Spack recipe files for the new external packages

New external package

For a new external package, there may already be a recipe file in Spack's built-in repository. To check for available versions and their dependencies, use the following command:

spack info <package_name>

You can also open the recipe file in your default editor with:

spack edit <package_name>

If no recipe file exists, consider reaching out to Software Coordination for assistance in creating one.

Existing external package

For existing external packages, you can copy the recipe file (and related patches) from either the built-in Spack repository or daq-release/spack-repo/externals. Copy the entire directory from the source path to your work area's Spack repository path:

cp -pr <spack-repo-path>/packages/<package_name> <workarea>/.spack/spack-repo/packages/<package_name>

This ensures you copy both the recipe file and related patch files.

Adding a new version

If a recipe file exists but lacks entries for newer versions, use the following command to have Spack check for available versions and generate checksums:

spack checksum <package_name>

Copy the output for the new versions and their associated checksums into the recipe file to make these versions available. Additionally, check if the package uses patches when building. If patches are in use, confirm if existing patches are still applicable; otherwise, create new ones. Contact Software Coordination for assistance as needed.

Example: Building and Using a New HighFive version in a work area

## Creating a work area with the "-s" option of "dot-create"
source /cvmfs/dunedaq.opensciencegrid.org/setup_dunedaq.sh
setup_dbt latest
dbt-create -c -s -n NAFD23-10-02 dev-hdf5
cd dev-hdf5/

## Obtaining the Spack recipe file and placing it into the local Spack repository
git clone https://github.com/DUNE-DAQ/daq-release
cp -pr daq-release/spack-repos/externals/packages/highfive .spack/spack-repo/packages/

## Setting up Spack and building the new HighFive version/variant
dbt-workarea-env
spack install --reuse [email protected]%[email protected]+mpi arch=linux-almalinux9-x86_64

## Loading the locally installed HighFive
spack load highfive+mpi

## Reloading the Python virtual environment
## "spack load" sometimes modifies PYTHONPATH, which could cause issues with "dbt-build."
## Reloading the environment before running "dbt-build" avoids this potential issue.
deactivate
source .venv/bin/activate

## Building the DAQ package using features available only in the locally installed HighFive
cd sourcecode/
git clone https://github.com/DUNE-DAQ/hdf5libs -b leo-update-tests-apps
# Modify hdf5libs's CMakeLists.txt to set:
# option(WITH_HIGHFIVE_AS_PACKAGE "HIGHFIVE externals as a dunedaq package" ON)
# option(WITH_HDF5_AS_PACKAGE "HDF5 externals as a dunedaq package" ON)

dbt-build