Re: Bug#33430: Release-critical Bugreport for February 16, 1999
On Thursday 18 February 1999, at 11 h 49,
Adam Di Carlo <aph@debian.org> wrote:
> Well, optimally, the master2files stuff *should* in fact check
> depends. Sure, the user may skip the 'Select' step in dselect, but
> you can be sure they'll run that step at some later point.
I'm afraid there is no simple solution. Let's imagine package A and B on the
master file. They do not conflict. Fine.
Now, A recommends C and B recommends D. C and D conflicts. Even if
master2files checked conflicts it would not have catch this one, which will
manifest itself only with dselect. (It is one of the things I hate in dselect:
conflict resolution can need iteration.)
[Does anyone know if apt automatically installs recommended packages? And
gnome-apt?]
I'm afraid that, if we check conflicts in recommended packages, we would
extend the graph to a point where conflicts will be certain.
> I would suggest, for your sanity, you start using pkg-order to help
> you check dependancies/recommends? Would that work?
It would be a good idea (but the doc is a little too hairy for me) but it
would not solve the above problem (which is a simplified version of bug 33430).
--
http://www.debian.org/~bortz/
Reply to: