This directory must be migrated periodically with the upstrem updates of arrow-cpp-feedstock, parquet-cpp-feedstock. conda-forge repositories because of multiple vendored files.
The recipes here are tested on nightly basis, so they follow the development versions of arrow instead of the upstream recipes, which are suitable for the latest releases.
In most of the cases these recipes are more accurate, then the upstream feedstocks. Although the upstream feedstocks regularly receive automatic updates by the conda-forge team so we need to backport those changes to the crossbow recipes. Most of these updates are touching the version pinning files (under .ci_support
) and other CI related configuration files.
Because all three recipes must be built in the same continuous integration job prefer porting from the arrpw-cpp feedstock.
Copy the configuration files from arrow-cpp-feedstock/.ci_support
to the .ci_support
folder.
The .azure-pipelines/azure-pipelines-[linux|osx|win].yml
should be ported to the local counterparts under .azure-pipelines
with keeping the crossbow related parts (the cloning of arrow and the jinja templated variables) and moving the matrix definitions like this to the crossbow [tasks.yml][../tasks.yml] config file.
Theoretically these recipes should be up to date with the actual version of Arrow, so during the release procedure the content of these recipes should be copied to the upstream feedstocks.