Re: Bug#220437: cgiwrap.allow replaced by package upgrade
Someone, it is not clear who or when, wrote:
> > > It seems the problem is a conflict between the way Debian
> > > does configuration files (if it's there it won't be touched,
> > > but if it's not there it will be put there), and logic that
> > > involves the presence or absence of a file.
On Wed, 2004-05-26 at 17:18, allomber@math.u-bordeaux.fr wrote:
> Hello, I would like to point out that this statement is false:
Bill,
As you have written it - with a colon at the end of your statement - you
appear to be saying that the following policy is false. However,
reading the rest of the note as context, I understand you to mean that
the unattributed statement above is false, and then cite the following
policy to back up this assertion. Is that correct?
> Debian policy E.1.:
>
> E.1. Automatic handling of configuration files by `dpkg'
>
> However, note that `dpkg' will _not_ replace a conffile that was
> removed by the user (or by a script). This is necessary because with
> some programs a missing file produces an effect hard or impossible to
> achieve in another way, so that a missing file needs to be kept that
> way if the user did it.
>
> So I wonder whether you purged the package and reinstalling without realizing.
I have sometimes wondered if an audit trail kept, in perpetuity, of dpkg
package database status changes would be a good idea. It would come in
handy for mysteries like this one. Unfortunately, so far as I can see,
the only such status change info kept is in /var/lib/dpkg, and that is
only for the last set of changes, not the last several. This is
probably too little, too late by the time an investigation is in full
swing.
> Alternatively it might be a bug in a old version of the package that did
> not register cgiwrap.allow as conffiles.
This is testable. Look at woody, potato, slink ... etc. until you find
a version that doesn't list cgiwrap.allow as a conffile (ignoring for
the moment the possibility that an interim version might have not listed
it as a conffile.) So far I checked woody & potato versions of cgiwrap,
and both have cgiwrap.allow in ./debian/conffiles. While not proof
positive that this wasn't what happened, it certainly makes it seem less
likely.
Ben
--
synrg at debian dot org
Reply to: