Le sam. 25 juil. 2026 à 11:10, Otto Kekäläinen <[email protected]> a écrit :

> Hi,
>
> > +1, I'm also not a fan of monorepos or repositories which only contain
> > debian/.
>
> Monorepos could make sense for Debian perhaps in the future if the
> overall packaging workflow and packaging formats converge clearly on
> something. Until that happens, jumping directly to a monorepo would
> alienate many people who would essentially be forced to redo all their
> workflows immediately in one go. Hence I would not recommend it for
> now.
>
> > But concerning the initial question (changing multiple packages at
> > once), isn't that exactly what the Debian Janitor[1] excels at? I guess
> > we could talk to Jelmer about adding go-specific tasks (or maybe come up
> > with patches ourselves, I haven't look at the code).
> >
> > IIRC, the Janitor can even be configured to automatically commit the
> > changes to the repository, so there would be no need to manually merge
> > all changes.
>
> Janitor is a Salsa bot that runs lintian-brush on packages, and files
> the results as Merge Requests. You can just run lintian-brush yourself
> every time you touch a package to have lots of changes fixed
> automatically. My `debcraft improve` runs lintian-brush and several
> similar tools under the hood.
>
>
> In general, the way wide-scale changes are done in Debian is either by
>
> a) change in a build tool (e.g. dh-golang) which once uploaded will
> make all builds/packages behave in the new way
>

Improving dh-golang is surely a good idea.
dh-nodejs helped us a lot at js-team. It made things a more bearable.


> or
>
> b) change in policy, communicating that to all maintainers and
> updating tools such as lintian and lintian-brush to automatically
> detect and fix the issues in the packages that need to be updated
>
> Currently there is no fast/easy technical way to do changes to
> multiple packages (e.g. via monorepo), nor do we have any
> social/organisational way to do it, as reaching any decision takes
> endless amount of discussion and vague consensus, and there is no
> clear way to propose changes and have them voted on in a binding way
> (apart from the GR as described in the Debian constitution, but that
> is overkill in most day-to-day changes we want to drive). This is e.g.
> why Stapelberg stopped being active in the Go team as described in
> https://michael.stapelberg.ch/posts/2019-03-10-debian-winding-down/.
>
> I have a feeling we should solve the social/organisational way first
> on how to make decisions that affect all or most packages before we
> are ready to have automation such as monorepos or Flight Deck to do
> changes on a lot of packages in parallel.
>

Doing large-scale changes like that surely require some level of consensus.
Otherwise, improving dh-golang and uscan to e.g. support versions in go.mod,
would help a lot (while probably break things for a while) and improve
automation.

Reply via email to