Bug#245253: apt-get: claims it will upgrade a packages that should/will not be
On Thu, Apr 22, 2004 at 11:15:01AM -0700, Matt Zimmerman wrote:
> > You're right, it unpacked the version from the cache. But then there is
> > still a bug in apt, since unless I clean the cache, it will continuously
> > unpack it, thinking it is the right deb.
>
> apt has no way to accomodate having two distinct packages in the cache with
> the same version number, and I don't see any reason why it should.
Simple one: in this case a NM applicant makes available his packages until
his sponsor can take time to upload.
Normally the more common case of a DD making his packages available while
they are waiting in queue/new should not cause a problem, I suppose.
> (package,version) should refer unambiguously to a package, and that
> assumption is deep in the packaging toolchain.
That's understandable, but see above - the good practice would probably be
to present the package with a backport-like versionning, I suppose. Isaac,
that suggestion is for you ;)
> > What makes it think the one installed is not the available one ? Why
> > wouldn't it be possible to detect the same problem wrt the cache ?
> >
> > And above all, why if I comment out any one of the 2 source lines, would it
> > think that the installed package is good ?
>
> apt is trying to deal sanely with a broken situation. This sort of thing
> happens, for example, if the metadata in the Packages file doesn't match the
> package that is actually available. apt has no way to d
What makes apt think the situation is broken ? Correct me if I'm wrong: if
the metadata for the 2 packages is different, then at least the md5sum
fields are different, and this is really the case here, even for package
wesnoth for which nothing special happens.
First, that shows that apt has a simple way of detecting the problem. It
would be good if it would warn the user instead of resorting to a DWIM
behaviour.
Secondly, that also shows that in some way this DWIM handling apt has of this
broken situation is not consistent among broken packages. It may be worth
to keep it instead of failing, to spare the user to deal with an error that
not his own, but I do not think it should do so silently.
Probably even a "warn and ask user before proceeding" behaviour would be
better.
Regards,
--
Yann Dirson <ydirson@altern.org> | Why make M$-Bill richer & richer ?
Debian-related: <dirson@debian.org> | Support Debian GNU/Linux:
Pro: <yann.dirson@fr.alcove.com> | Freedom, Power, Stability, Gratuity
http://ydirson.free.fr/ | Check <http://www.debian.org/>
Reply to: