On Wed, 12 Jul 2023 13:14:47 +0200 Michael Biebl <bi...@debian.org> wrote:
> Am 12.07.23 um 11:14 schrieb Josh Triplett:
> > Package: libblockdev-mdraid3
> > Version: 3.0.1-1
> > Severity: normal
> > X-Debbugs-Cc: j...@joshtriplett.org
> > 
> > Current udisks2 has a dependency on libblockdev-mdraid3, and
> > libblockdev-mdraid3 has a dependency on mdadm. This effectively makes
> > mdadm required on desktop systems.
> > 
> > As I understand it from
> > https://github.com/storaged-project/udisks/issues/1142 , udisks2 needs
> > to depend on libblockdev-mdraid3. (And that much isn't necessarily an
> > issue, it's just a library.)
> > 
> > Would it be possible for libblockdev-mdraid3 to avoid pulling in mdadm,
> 
> By avoid, I assume you mean not even demoting it to Recommends would be 
> an option for you, but drop the dependency completely (or make it a 
> Suggests)

I was hoping for a Suggests, yes. Using Recommends would still leave
this as an issue for most desktop systems, and would contribute to the
reasons why some people disable Recommends by default and just
selectively install Recommends when installing packages.

> > so that systems not using RAID don't have to have mdadm and its
> > associated configuration and boot-time logic installed?
> 
> And then I'll get bug reports like
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=989493

My understanding was that that error happens *unconditionally* whether
there's a RAID device on the system or not, and that that bug is fixed
by depending on libblockdev-mdraid3 (which seems fine). But if I'm
understanding the upstream issue correctly, libblockdev-mdraid3 doesn't
actually need the mdadm binary *unless* the user's system actually has a
RAID device on it, in which case they would more likely want to have
mdadm installed. So, I *think* removing the Depends on mdadm would mean
that the only users getting an error message are those who have a RAID
device and *don't* have mdadm installed, which seems like exactly the
set of users who *should* get an error message.

- Josh Triplett

Reply via email to