Bug#1021658: Processed: autopkgtest regressions are RC (and this one also blocks the perl transition)
On Sun, Oct 23, 2022 at 12:35:41PM +0200, Paul Gevers wrote:
> Hi Hilmar,
>
> On 23-10-2022 12:15, Hilmar Preuße wrote:
> > Am 22.10.2022 um 10:21 teilte Debian Bug Tracking System mit:
> > the breakage is not caused by the perl upload, but due to the latest TeX
> > Live upload at the beginning of this month. Hence TL either does not
> > migrate to testing (which is good in the moment IMHO). I reported the
> > issue upstream, not sure when we get a solution.
>
> I think both Adrian and I understood that, I *assume* he raised the severity
> because failing autopkgtest (on amd64 and arm64) are RC [1].
Yes, that's what I tried to say in the Subject.
> > I could remove the few failing tests from the test suite, but we would
> > get a degraded TL / latexml combination in testing this way.
> >
> > Can we add an override / hint to make sure perl migrates to testing?
>
> We'll handle it somehow yes.
It seems someone already handled it or it disappeared by itself.
> > Further the autorm for latexml would need to be removed, else the new TL
> > will migrate to testing and will break an (already) installed latexml
> > package.
>
> You can block migration of packages by creating an RC bug against it. For
> your info, if there's progress on an RC issue, the Release Team considers it
> acceptable that you regularly ping the RC bug with some update, which will
> reset the removal date counter. It goes without saying that the Release Team
> can at any moment remove any package from testing if they deem that the
> right action to solve issues regarding testing and/or migration.
>
> > Any further ideas how to handle the situation?
>
> We'll handle perl, you can handle TL yourself by filling an RC bug.
The autopkgtest regression in latexml already prevents the migration of TL.
The only possible action on the TL side will be after latexml is fixed
to add Breaks in TL against unfixed versions of latexml to avoid users
(or test environments) ending up with incompatble combinations.
> Paul
>...
cu
Adrian
Reply to: