On Wed, Aug 18, 2021 at 05:55:09AM -0700, Andrea Bolognani wrote:
On Wed, Aug 18, 2021 at 12:43:21PM +0200, Pavel Hrdina wrote:
> On Wed, Aug 18, 2021 at 02:41:31AM -0700, Andrea Bolognani wrote:
> > On Tue, Aug 17, 2021 at 02:41:48PM +0200, Ján Tomko wrote:
> > > s390x-debian-sid-container:
> > > - extends: .container_job
> > > + extends: .container_optional_job
> > > variables:
> > > NAME: debian-sid-cross-s390x
> >
> > This essentially removes coverage for these architectures in a
> > permanent fashion by sweeping failures under the rug, which is not an
> > acceptable way to deal with a temporary failure IMO.
> >
> > I'm working on a different solution which would retain fully coverage
> > while still working around the current issues with Debian sid.
> >
> > NACK
>
> If I understand this patch correctly it will only mark building the
> container optional, the libvirt build would sill be mandatory so I don't
> see any issue here with coverage. We would simply reuse the latest
> working container to build libvirt.
That's true, but I'm concerned that in addition to working around
transient issues this will potentially also mask permanent issues
that actually require developer intervention because both would look
exactly the same.
Additionally, if one were to create a libvirt fork in GitLab right
now their pipeline would fail because there simply wouldn't be a
previous container image to fall back to.
Jano pointed out in another thread that we're already treating
Rawhide container builds as optional, and since the concerns I
mention above would also apply to that situation, dealing with them
for Debian only wouldn't make sense and in fact, when a different
solution is proposed, it will be helpful that the behavior is
consistent for all fast-changing distros.
I thus retract my NACK. Please go ahead and merge this patch.
--
Andrea Bolognani / Red Hat / Virtualization