[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#330078: latex points to etex, not to tex



Anton Zinoviev <anton@lml.bas.bg> wrote:

> On Mon, Sep 26, 2005 at 08:43:38PM +0200, Frank K.ster wrote:
>> Anton Zinoviev <anton@lml.bas.bg> wrote:
>> 
>> > Do you think that this bug is due to misbehaviour of dpkg?  True,
>> > 01tetex.cnf is a configuration file, but I hadn't edited it so it had
>> > to be removed during the downgrading.
>> 
>> No, dpkg won't remove unused configuration files upon upgrading, and it
>> won't do so upon downgrading, either.  For the upgrading case, the
>> maintainer scripts has to take care - rename the file in preinst, take
>> over settings, moving out of the way or whatever.  For the downgrading
>> case, we don't provide any support.
>> 
>> dpkg will *only* remove a file when the package is purged.
>
> Yes.  My point was that the file was not changed by the user so
> keeping it is evil.  If dpkg doesn't ask me for a permission to change
> the contents of an unchanged conffile then it logical to also remove
> it when it does not exist in the upgraded/downgraded package any more.

No, it isn't.  Everything is designed from the point of view of an
upgrade, and in this case "preserving local changes" of course must
include *not* to remove an unused conffile.  It might contain important
work of the local administrator.  

In case the maintainer knows that the presence of the old conffile would
be harmful, he has to do something about it (like moving it out of the
way), but he may still not silently delete it.

In case of a downgrade, this could in principle be done.  But supporting
downgrades would increase both the work for maintainers and the
complexity of maintainer scripts by an insane amount, and is therefore
not supported.  People who try unstable or even experimental packages
are expected to be able to perform a downgrade manually.

For example, you can purge nearly any package (unless it's essential, I
think) with --force-depends if you install the working, older version
immediately afterwards.

So essentially there is no bug, and there never was one.  You just
expect too much from us.

Regards, Frank
-- 
Frank Küster
Inst. f. Biochemie der Univ. Zürich
Debian Developer




Reply to: