On Sun, 10 Aug 2008 15:38:56 +0200 Francesco Poli wrote:

> On Sun, 10 Aug 2008 13:51:43 +1000 (EST) Tim Connors wrote:
> 
> [...]
> > PS.  Workaround for the guys who reported this bug is to remove
> > /etc/logrotate.d/acpid.  I decided to purge whatever else might have been
> > left behind accidentally by running (no gaurantees this won't screw your
> > system :)  Does aptitude/dpkg come with a cleaner way to purge obsolete
> > conffiles?  This seems to be affecting me more and more recently.  I
> > wonder if policy needs a change or whether packagers aren't following
> > policy related to conffiles anymore.
> > 
> > > dpkg --force-depends --purge acpid
> > > aptitude install acpid
> 
> Thanks for spotting that /etc/logrotate.d/acpid is no longer shipped
> with package acpid!
> I've just done the following:
> 
> 
[...]
> On next Sunday, I will verify if this worked out as intended...

It seems it worked great.
Three Sundays are gone without any acpid death!

As a consequence, this bug apparently won't affect any new users, but
only old users upgrading from some previous acpid version to a recent
one...

At this point, I think the bug boils down to the following flaw:
/etc/logrotate.d/acpid used to be a conffile for old versions of the
package, but now is no longer shipped or created and should no longer
exist; this change is currently not handled in any way by the postinst
script, hence old users upgrading from old versions will be hit by "the
mysterious case of weekly acpid deaths"...

I don't know whether there's a canonical way to cope with this kind of
situations.
Maybe the postinst script should check for the existence
of /etc/logrotate.d/acpid, and, when it finds it, alert (via debconf)
the user that his/her system has a configuration file probably left
over by a previous version of the package.  The debconf dialog could
even offer to remove the old file, to ignore the problem, or to never
complain again about the problem...


-- 
 http://frx.netsons.org/doc/index.html#nanodocs
 The nano-document series is here!
..................................................... Francesco Poli .
 GnuPG key fpr == C979 F34B 27CE 5CD8 DC12  31B5 78F4 279B DD6D FCF4

Attachment: pgpKrGr20IblK.pgp
Description: PGP signature

Reply via email to