On Wed, Sep 18, 2024 at 06:42:59PM GMT, Daniel P. Berrangé wrote:
On Wed, Sep 18, 2024 at 10:34:31PM +0900, Andrea Bolognani wrote:
> On Wed, Sep 18, 2024 at 03:02:58PM GMT, Daniel P. Berrangé wrote:
> > On Wed, Sep 18, 2024 at 09:57:32PM +0900, Andrea Bolognani wrote:
> > > I'd wait for them to be made GA.
> > >
> > > New macOS releases have historically introduced changes that require
> > > libvirt to adapt, sometimes with significant effort, and I wouldn't
> > > want anyone to spend time investigating a build failure that can
> > > potentially affect the Beta version of the OS but not the GA one.
> >
> > Actually I'm wrong. Looking more closely the sequoia image there
> > is an RC, not a Beta. The beta images was referring to 15.1, not
> > 15.0, and they are named separately as sequoia-beta. Using an RC
> > is reliable enough IMHO
>
> I still think there's no reason to rush before the image is GA'd, but
> if the build job succeeds with the RC image and there is no post-GA
> churn involved I won't stand in the way of a libvirt-ci MR taking
> that approach instead of mine[1].
By that rationale we also won't test against rawhide, or sid, etc.
IMHO the earlier we test against a platform the better, so if an
image is available we should use it.
Rawhide and sid are unstable targets *by design*, which we
acknowledge by allowing the respective jobs to fail without
considering the entire pipeline as failed.
This is more akin to adding e.g. Fedora 41 as a target when GA is not
out yet but RC is. Which is not something that, as far as I know,
we've ever done.
Anyway, I've approved the libvirt-ci MR fixing the image name.
Michal, if you squash in the obvious diff and confirm that the
pipeline still passes, you can have my
Reviewed-by: Andrea Bolognani <abologna(a)redhat.com>
and push this.
--
Andrea Bolognani / Red Hat / Virtualization