Re: Bug#1068189: debhelper: --link-doc checking for known packages makes linux-signed build FTBFS

[Salvatore Bonaccorso <carnil@debian.org>]

Control: reassign -1 src:linux 6.7.9-2

Hi Niels,

On Mon, Apr 01, 2024 at 05:19:43PM +0200, Niels Thykier wrote:
> Salvatore Bonaccorso:
> > Source: debhelper
> > Version: 13.15
> > Severity: serious
> > Tags: ftbfs
> > Justification: Regression for other package builds, FTBFS
> > X-Debbugs-Cc: carnil@debian.org,debian-kernel@lists.debian.org
> > Control: affects -1 + src:linux,src:linux-signed-amd64,src:linux-signed-arm64
> > 
> > Hi Niels,
> > 
> > Not fully investigated, but starting to fill a bugreport. I noticed
> > that the src:linux pipeline on salsa started to fail for the
> > jobs in th build-signed stage (in the build-signed job).
> > 
> > https://salsa.debian.org/carnil/linux/-/jobs/5527774
> > 
> > (and for saving the output):
> > 
> > [...]
> > 
> > (attached as well the raw log)
> > 
> > I'm not 100% sure yet, this might be a problem in our packaging in
> > which case we can re-eassign. But it only got triggered with the
> > change recently in debhelper:
> > 
> > https://salsa.debian.org/debian/debhelper/-/commit/dec5cfad00e2abd9ee3594f90c93f3fa42bb73ff
> > 
> > Regards,
> > Salvatore
> 
> Hi Salvatore
> 
> It was a suggestion raised (I think on IRC) to have debhelper explicitly
> check these parameters, because a lot of t64 breakage was "unnoticed" by
> debhelper. That is, when people forgot to update --link-doc parameters
> (etc.).
> 
> The code for `--link-doc` uses `${binary:Version}` for the dependency, so
> the package should really be from the same source[1]. In my view, it was
> never a case that was expected to work between source packages.
> 
> I think `linux` with `linux-signed` is doing something really special here
> (especially considering it has worked so far), and I think the question is
> whether `linux`/`linux-signed` should get a special-case or concluding that
> the `--link-doc` is not suitable for the `linux`/`linux-signed` case.
> 
> I would like to hear your case for what makes `--link-doc` sensible for the
> `linux-signed` case. I know of `linux-signed`, but I have no idea what you
> are dealing with in practice, so it is hard for me to make a judgement call
> on this (other than my biased gut feeling of wanting to minimize
> special-cases).

Thanks for your very quick reply, this is much appreicated.

I understand the reason and src:linux should not get really to be
exceptionally handled. So for now I will re-assign it to src:linux
and we can search for a solution in our package.

Thanks a lot for your work on debhelper!

Regards,
Salvatore