This directory contains the recipes that are used to generate the packages on the diracgrid conda channel. This should only be used in situations where it is impractical to add packages to conda-forge directly. New packages can be added to conda-forge using staged-recipes and existing packages can be modified by making a PR to the corresponding feedstock.

If it is absolutely essential for DIRAC to have a custom build of a package, a new directory can be added here. A reference for the meta.yaml syntax can be found here and the conda-forge is a good source of examples. You can then test building packages by running:

conda build -c diracgrid -c conda-forge -m conda_build_config.yaml my-package/

The conda-recipes/conda_build_config.yaml file is a "variant config file" which is used to constrain the versions of dependencies. Currently this is only used to set the Python interpreter version. If shared library dependencies are added, new values should be added based on the main conda-forge configuration.

Once a newly built package is ready, it can be uploaded to the diracgrid conda channel using:

# Install the anaconda CLI client
conda install anaconda-client
# Login using your anaconda.org credentials
anaconda login
# Upload the binary, the file path will be printed at the end of the conda-build log
anaconda upload path/to/my/package.tar.bz2 --user diracgrid

As this is forseen as to be a rare operation, contious integration is not currently being used for building and uploading packages.

Reason for not using conda-forge: This package is distributed using DIRACGrid's channel as it is a fork of the main upstream package, making it ineligible for conda-forge.

Version history:

• 5.1.1+dirac.1-0 Initial release

Reason for not using conda-forge: This package is completely non-functional without DIRACGrid's fork of tornado.

Version history:

• 0.1.1-0 Initial release