Re: For the historians: Do we still need the Conflicts: on tetex-lib?
Hilmar Preusse <hille42@web.de> wrote:
> Is that "policy" or common practice? If yes, we could start to clean
I don't think it is mentioned in the Policy Manual, which is why I wrote
"policy", but this is definitely perfectly acceptable and accepted
practice for Debian developers. *Some* of them try to support upgrades
from 'release' to 'release'+n with n >= 2, but this is neither required
nor expected behaviour.
For many transitions, the assumption that the user will not skip a
release is quite fundamental. For instance, on sid:
% grep-available -sPre-Depends -XF Package dpkg
Pre-Depends: dselect, libc6 (>= 2.3.2.ds1-4)
Many people wonder what the point of having dpkg and dselect in
different packages in sid is since dpkg Pre-Depends on dselect. The plan
is to drop this dependency in sarge+1, because if you drop it in sarge,
you may break woody to sarge upgrades performed with dselect. Now,
assuming the user upgrades to sarge before upgrading to sarge+1 ensures
that he will have the dselect package installed in sarge and therefore
also at the moment he upgrades to sarge+1. Then, he will be able to
remove dselect because the Pre-Depends will be dropped in sarge+1.
If you cannot make this assumption, transitions in complex packages
sooner or later lie to madness.
> these long Conflicts- and Replaces-lines of tetex-bin and tetex-base.
> AFAICS they where introduced to guarantee a smooth upgrade from
> non-teTeX to teTeX, which was done between pre-hamm and hamm, i.e.
> 4(!) releases ago. I guess a smooth upgrade from bo (or rex) to sarge
> will be quite impossible.
Very unlikely to happen smoothly, yes.
--
Florent
Reply to: