pá 10. 6. 2022 v 17:18 odesílatel Chris Evich <cevich(a)redhat.com> napsal:
On 6/10/22 08:49, Miloslav Trmac wrote:
> These are real: k8s.gcr.io/coredns/coredns
got some Cosign signatures, which
> are OCI artifacts, and c/image / Skopeo can’t currently copy those.
> I suppose there’s no point in delaying adding OCI artifact support, if
> only just enough for (skopeo sync) to work without having to understand
> the signatures. I’ll start working on that now.
> OTOH at least at a first guess I’m not much a fan of backporting that to
> the old stable branches — I’d rather just update the test to use a
> different repo.
Thank you so much for noticing this and letting us all know. I'm sure
this is also affecting all PRs and probably c/image testing as well, ya?
Yes, thanks for pointing that out. Currently I’m thinking that getting the
OCI artifact support done might be more useful than working around this,
assuming that that work ends up being small and quick enough — but I’m
quite open to just changing the repo for now, especially if the tests block
anyone else working on the two repos.
A third option for the release branches is to disable the failing tests.
We really only need those for backport CI. So there's room to
backport-risk with CI effectiveness. Also backports can rely more on
manual testing/verification, so 100% CI may be less relevant.
For the backports, I have already committed updates to the repo.