Quoting Christian Hesse (2025-12-06 16:45:29)
> Lukas Fleischer <[email protected]> on Fri, 2025/12/05 18:00:
> > Quoting Christian Hesse (2025-12-05 07:47:23)
> > > Did that cover all your questions?  
> > 
> > Thanks! For the most part, yes, I think so. Two follow-ups:
> > 
> > 1. Are we going to have [extra-unstable-staging] for rebuilds, or do we
> >    plan to have a different mechanism to handle soname rebuilds across
> >    the new repos?
> 
> Oh, good question... I have not yet had that case. We would hit it when the
> pre-release bumps a soname, right? So something like systemd pushing a
> pre-release where libudev is bumped from libudev.so.1 to libudev.so.2...
> 
> At least for the packages that I care about here these cases should be
> pretty rare... :-p

Based on your earlier replies, if I understand correctly, it'd also be
required whenever there's a soname bump (for any package, not
necessarily in -unstable) while any dependent package is in -unstable.

> [...]
> >    We may also want to build some of those into devtools as checks. Does
> >    that make sense?
> 
> Hmm, not sure. Depending on package (and impact of rebuild when required)
> this may vary from over-complicating to required.

FWIW, I don't think those are extremely rare corner cases once we start
using -unstable more. Specifically, I'm thinking of cases where we push
a new package to -testing or stable repos and forget to drop the (now
older) package from -unstable.

That said, I think it's reasonable to monitor and quickly fix those
issues when they appear; as long as we have a good understanding of what
potential issues may appear and as long as they break -unstable only.

> 
> After all this has much greater impact than expected when we want to make it
> complete and correctly.

Reply via email to