[Debian policy: For reference, this is bug #603680. On Wed, Nov 17, 2010 at 12:59:00PM +0100, David Kalnischkies wrote: > On Wed, Nov 17, 2010 at 10:34, Bill Allombert > <bill.allomb...@math.u-bordeaux1.fr> wrote: > > On Tue, Nov 16, 2010 at 09:35:55PM +0100, David Kalnischkies wrote: > [… snip …] > >> apt-squeeze recently (see #591882) got a third option: > >> c) try installing another or-group member > >> > >> Note that while c) seems to be the "captain obvious" solution it introduces > >> a big problem: a) and b) reduce the number of broken packages, > >> but c) can add a lot more which could (real-world will tell if really) > >> work against the current resolver determinism… > > > > Probably option c) should be removed. This makes upgrade process much less > > predictable. > > An secondary issue with option c) is that this can lead apt to upgrade free > > packages with non-free packages, if non-free packages are listed as > > alternative. > > Only if the free package is uninstallable in squeeze, but in this case a > new installation has the same result.
Not necessarily: during an upgrade there are much more constraint to satisfy that during an installation, as this is the case here. > APT will try to fix the free package before it tries to fix the packages > depending on the free one, so the non-free option is still only the fallback. I do not think it is correct to ever upgrade a free package to a non-free one. Now, apt is not at fault, the problem rather lie in a strange interpretation of policy 2.1.2 by some developers. But we cannot ignore the issue either. Cheers, -- Bill. <ballo...@debian.org> Imagine a large red swirl here. -- To UNSUBSCRIBE, email to debian-policy-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20101117135400.gb22...@yellowpig
