Title: #531: Add the needed machinery to have automated builds for our COPR
The COPR repos should be at least usable and currently the git HEAD of our
branches might or might not be usable in some aspects -- for example AD
provider or IPA provider are not tested at all upstream. The integration tests
we have test only a small subset of the functionality and for a lot of
functionality we rely on greentea tests. Prior to the release of 1.16.1 there
was quite a few bugs (3594 for example, I remember that one because I was
working on it) which were only revealed by the greentea tests.
Until we have commit gating with all the available tests, we can't be sure that
everything works. And we only do a release at the point where our downstream
tests are also passing. After that, we tag the release and the release is the
sssd upstream saying "this bundle of commits is known to work well together".
I don't think there is any harm in having "sssd-1-13-rolling" branch and
"sssd-1-16-rolling" in addition to building the releases in COPR, but I don't
think having the git HEAD as the default without warning is a good idea until
we drastically improve our coverage.
About testing the sssd-1-13 branch before a release, we need to do that before
the release IMO. There are some downstream tests that would fail because they
might check for functionality or regressions not present in 1.13 but we should
at least run all the tests with 1.13.4 and 1.13.5 prior to 1.13 release and
compare regressions. This work is not related to COPR, but in general, I don't
think we should release upstream without testing it.
About the smaller releases, I agree. I think there we need to improve our
failure detection and stability of tests, because currently we check the
failures downstream more or less manually, but I would also like to have montly
or bi-monhly releases.
See the full comment at
sssd-devel mailing list -- email@example.com
To unsubscribe send an email to sssd-devel-le...@lists.fedorahosted.org