Make sure build IDs are both unique and reproducible for brew pipelines
With the variant work, we're going to get multiple builds per architecture per brew build. E.g. x86_64 rt, x86_64 regular, + both for debug. Each of these will have a separate pipeline. If we keep the current brew build KCIDB IDs, this would lead to overwriting the builds, as the current build ID schema is brew-<taskid>_<arch>
(with an optional hardcoded debug suffix). We don't want to overwrite the builds, we want to uniquely identify each build. The brew checkout and build IDs are reproducible in order to allow QE to submit results for them so whatever naming we pick needs to be something QE can recreate.
Based on that, probably including package name in the ID should be a good candidate. It would also take care of the debug suffix. E.g. brew-<taskid>_<arch>_<package_name>
AC:
-
agree on whether the above makes sense or if we should go with a different schema -
change the pipeline to create these new IDs. It should also be acceptable to use the package name for regular (non-brew) pipelines so we can change the common code for that -
inform QE that is sending us results about the change because they need to change their scripts @kamaxeon @bgoncalv