On Mon, 23 Sep 2019 at 06:53:44 -0700, Felix Lechner wrote:
> On Mon, Sep 23, 2019 at 1:27 AM Simon McVittie <s...@debian.org> wrote:
> >
> > To accommodate the 0+deb10u1 convention, I think there should be an
> > exception to this tag, preventing it from being emitted for revision
> > numbers that "are based on" revision 0.
>
> I struggled with something similar last week. There was some
> discussion on IRC, but no resolution. The relevant code is here:
>
>
> https://salsa.debian.org/lintian/lintian/blob/master/checks/debian/changelog.pm#L141-149
>
> Is your value of 'release' in d/changelog set to buster?
It was for this particular run, but for the majority of my builds it's
'UNRELEASED'.
> Would it be
> better to exempt direct uploads to past releases ($release ne
> 'unstable' && $release ne 'experimental')?
In my opinion: no, because the 0+ convention is equally applicable if
(for example) you have already uploaded foo_1.2.3-1 to experimental
containing changes that are not ready for unstable (perhaps during a
freeze or a transition), and now you want to release foo_1.2.3-0+deb11u1
or foo_1.2.3-0+sid because you have realised that the upstream 1.2.3
release contains an important bug fix that needs to be fast-tracked.
In that scenario, it would be misleading to use foo_1.2.3-1~deb11u1 or
foo_1.2.3-1~sid1, because that would imply that the packaging is based
on foo_1.2.3-1, rather than being branched from foo_1.2.2-7 or whatever
the current version in unstable is.
smcv
