On 11/15/23 17:00, Dumitru Ceara wrote: > All of the above were changed to track the latest available releases. > Initially that seemed like a good idea but in practice, a new release would > potentially (silently) cause CI runs that used to pass on given stable > versions to unexpectedly start failing. > > To address that this series pins all versions we can control allowing us > to use different values for different branches. > > NOTE: the last commit of this series will look slightly different on > stable branches, e.g., on branches <= 23.06 we need Python <= 3.11 and > Fedora 37. > > The first commit of the series bumps the OVS submodule to include the > latest fixes to errors reported by recent versions of flake8. > > Changes in v2: > - added first patch to bump OVS submodule > - addressed Ales' review comments: > - moved the logic to determine which image to use out of ci.sh; > it's now in the workflow itself. > - moved setting of OVN_IMAGE in the same block with all related > env vars > - added a note for maintainers in the release process documentation > to mention that the new workflow will likely have to be triggered > manually (at least once) when branching for a new release. GitHub > doesn't allow periodic jobs on non-default branches.
IMHO, that sounds horrible to maintain. Can we just build things per run? How much value these containers actually have? Best regards, Ilya Maximets. _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
