[chiark.chat] Re: RFH: sysvinit etc. in Debian

Ian Jackson ijackson at chiark.greenend.org.uk
Wed Nov 7 22:50:17 GMT 2018


Andreas Messer writes ("Re: [chiark.chat] Re: RFH: sysvinit etc. in Debian"):
> One thing we really need to to discuss is the 
> "Provides: libpam-systemd" field. We're willing to remove this and
> prepare some kind of glue package for Devuan. But i'm not sure about
> the benefit. Because if you want to use elogind on Debian with lets 
> say Gnome or KDE, you'll still have to provide "libpam-systemd"
> somehow, or these will not fully install.

Right.

> I heard about some ongoing discussion on debian-devel about this?

This has been discussed at some length.  Some Debian folks are really
not happy about the Provides.  The medium term plan advocated there is
to change the dependencies in Debian so that this is Provides no
longer needed.

But for testing the new elogind package in Debian we will have to have
the glue package.  It will have to be obtained separately for testing
with Debian, whether totally out of band (initially) or via Debian
experimental.

If we want elogind to be identical in Debian and Devuan (which seems
quite desirable) that glue package would be needed in Devuan too.  It
is possible that in practice Devuan would retain the glue package
indefinitely.

To avoid getting elogind's entry into Debian held up it would be wise
for the glue package to be generated from a separate source package.
That source package might easily be acceptable since it could stay
in experimental indefinitely, in Debian.

Does that all seem to make sense ?

> Anyway, I have uploaded the package to "suites/unstable" branch
> and tagged it in the Devuan elogind package repository found here:
> 
> https://git.devuan.org/devuan-packages/elogind.git

Thanks.  At some point are we going to want a sponsor for a Debian
upload ?  I can be that sponsor unless someone else wants to do the
upload.  I think getting a version of elogind into Debian experimental
(well, into the NEW queue) sooner rather than later would be good.

I think the only blockers for that are:
 - the source package administriva, copyright notices, etc.,
   are in good shape;
 - it builds well and is lintian-clean;
 - it works for at least one thing, when equivs or some other
   mechanism are used to subvert the libpam-systemd dependency

Last time I looked at the Devuan elogind package the first point was
already satisfied.  I don't remember whether I ran lintian but I doubt
that will be a big problem.

The last point is probably not too hard to satisfy but a test report
from someone would be good.

Having the package in Debian, even in experimental, would let us use
the Debian BTS for bugs that affect it in Debian, as well as getting
an unknown-duration item off our critical path.

In principle it would also let us file bugs about dependencies.  That
should be done with some care to avoid making an unnecessary political
mess[1].  I can help with that.

[1] I should say that I am very pleased with the atmosphere we have
here on this list.  I like that people are not afraid to make
mistakes, and that we can change our minds and ask for things to be
done differently, without rancour.

Ian.

-- 
Ian Jackson <ijackson at chiark.greenend.org.uk>   These opinions are my own.

If I emailed you from an address @fyvzl.net or @evade.org.uk, that is
a private address which bypasses my fierce spamfilter.




More information about the Debian-init-diversity mailing list