Bug#849312: apt: dist-upgrade removes pkgs instead of upgrading them
2016-12-25 13:12 GMT+01:00 David Kalnischkies <david@kalnischkies.de>:
> I do think this specific instance is a rather temporary instance through
> as mariadb was only recently repeatedly uploaded, so two versions are
> still "fighting" over their spot in the archive.
Maybe, but such situations continue to keep up in testing / unstable
in the future right?
> In so far the /var/lib/dpkg/status file of the situation before the
> dist-upgrade would be interesting – you might still have it in
> /var/backups – as these solver things tend to be highly dependent on
> mirror state and the specific packages installed on a machine.
I dropped the VM, sorry.. how does this state affect the proposed
solution though?
>
>> Can't it detect mariadb-server-10.1 being a proper upgrade of
>> mariadb-server-10.0 and hence scoring this as neutral or positive?
>
> In general no, because that isn't a positive upgrade: It involves the
10.1 declares replaces on 10.0 doesn't it?
Doesn't apt use this information?
> removal of packages, which means features and services are removed other
> programs (packaged or not) might use and/or the user does. The very way
Any package upgrade might remove some behaviour / feature a user uses..
> this "upgrade" is packaged (= in another package forcibly removing the
> old) indicates that this upgrade isn't free: Something will have to
> adapt to changes – and changes are bad. Everyone hates changes: Ask all
> users of $DESKTOP_ENVIRONMENT being upgraded to $DEKTOP_ENVORNMENT+1.
> (The last bit is a bit exaggerated of course).
>
> It is what it is: A few new packages and a few packages removed – that
> you identify that as an upgrade is based on your knowledge of mariadb
> and how it (incompatibly!) changes in each new version: For apt it looks
> like the package 'default-texteditor' changed its depends from 'vim' to
> 'emacs' [actually worse than that as those two are co-installable] and
> that isn't an upgrade! ;)
It's not? :p
Does vim also declare replaces: emacs?
--
Olaf
Reply to: