[Pkg-utopia-maintainers] Bug#1099086: might be less verbose in calling adduser

[Simon McVittie]

On Sun, 09 Mar 2025 at 09:57:57 +0100, Marc Haber wrote:
>On Fri, Feb 28, 2025 at 10:41:26AM +0000, Simon McVittie wrote:
>>On Fri, 28 Feb 2025 at 11:21:01 +0100, Marc Haber wrote:
>>>I am afraid that our efforts in making adduser easier to use in 
>>>maintainer script have mainly occurred during the trixie release, 
>>>so if you want your maintainer script to also work in backports it 
>>>is probably better to keep it the way it is.
>>
>>I don't care about backports for this particular package (backporting
>>something as central and system'y as dbus seems like a lot of regression
>>risk for little reward). My concern is that if dbus is installed during
>>an upgrade from bookworm to trixie, its postinst needs to be guaranteed
>>to work correctly, even if adduser hasn't necessarily been upgraded to
>>its trixie version yet.
>
>You totaly have a point here that I wasnt aware of. Thanks for 
>pointing this out. Adduser will probably have to pass one release 
>cycle for all those deprecations.

You probably saw that I closed #1099086 with a new upload that now does:

     if command -v systemd-sysusers >/dev/null; then
         systemd-sysusers ${DPKG_ROOT:+--root="$DPKG_ROOT"} dbus.conf
     else
         adduser --system --quiet --group "$MESSAGEUSER"
     fi

with a versioned dependency on adduser (>= 3.130) | systemd-tmpfiles. 
This required overriding some Lintian tags: one I think is genuinely a 
Lintian bug (see #1099842), and one for 
maintainer-script-needs-depends-on-adduser because Lintian doesn't take 
into account that the adduser call is conditional (probably not feasible 
to fix in Lintian, shell script is not an easy thing to parse).

With a higher versioned dependency on adduser (>= 3.138), I believe I 
would also be able to drop the --quiet option, but I'd prefer not to 
overconstrain the order of bookworm-to-trixie upgrades: with adduser and 
dbus both being widely-installed and rather low-level, I think it's 
probably best if we can allow apt to upgrade dbus either before or after 
adduser. However, during the forky cycle I think it would make sense to 
bump the versioned dependency to adduser (>= 3.138), allowing --quiet to 
be dropped. Does that make sense to you?

     smcv