deluser on purge (was: Piuparts testing status update)

[Russ Allbery <rra@debian.org>]

Steve Langasek <vorlon@debian.org> writes:
> On Tue, Nov 14, 2006 at 09:35:12PM +0100, Frank Lichtenheld wrote:

>> Hmm, I would read policy in a way that since a package can not rely on
>> its dependencies being present during purge, their pure absence alone
>> should not be a valid reason to fail. If this on the other hand is a
>> valid excuse to leave cruft behind is not really clear to me. I would
>> certainly prefer the package just printing a warning and exiting normal
>> rather than failing but I can't really point out a specific policy
>> reference right now to make this more than a personal preference.

> In the case of adduser, there is a strong case for not doing deluser at
> *all* on purge, because it's impossible to ensure that there are no
> off-line or remote resources referencing the uid/gid.

This is something that I'd really like to see us sort out in policy, since
I think we should be able to describe consistent behavior with regard to
system users and package purging to our users.  Right now, every
maintainer is making their own decision in this area, and I think that
provides an inferior experience for users.

-- 
Russ Allbery (rra@debian.org)               <http://www.eyrie.org/~eagle/>