[Aptitude-devel] Bug#653284: unreasonable "Do not install"

[Harald Dunkel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Hi Manuel,

On 03/09/16 19:36, Manuel A. Fernandez Montecelo wrote:
> I think that there might be duplicates, some of them are old, but probably some are fresh enough.  I might even have stumbled over it now and then, but since the resolver gives very bad solutions so often, I skip over the bad solutions so quickly that I don't always pay attention.  So I'll take
> a look at this along to other reports at some point, although it's always daunting.
> 
> 
> One of the things that might have made it worse is enabling multi-arch, which I did a few years ago.
> 

I can imagine.

One thing you might consider on resolving package conflicts
is looking at source packages. If there is a source package S
used to build s1 and s2, then it appears to be pretty unlikely
that upgrading s1 and keeping the old version of s2 makes any
sense. Since there are quite a number of source packages
creating a large list of binary packages each, this scheme
could help to reduce the dependency graph and to get rid of
a lot of "bad solutions" (optionally, of course).

Maybe this scheme could be extended to multiarch as well.
I wouldn't like that additional 32bit packages sneak in just
to resolve a conflict in the 64bit environment.


Regards
Harri

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2

iQEcBAEBCAAGBQJW47D9AAoJEAqeKp5m04HLoIIH/j+deD4cZPKx/EfZqg4+GvTD
lmIACLr6x+WZKiX7u+RBzD45E3YV07V23cXxG+Lt2sgO8iYPDzTYCq97kOLtLgbf
BSHmZF2ABbmyAyoS/yMFhbccEjbG6H1/0IrSg43H2zWLcBEE6fjYp3T73YRNUzRh
uLhIqVRKLcGbQfpyQUBJhB1vpVkxN1dIzKYutcy22BGbZhjhxfs5z0A9SQdGoDYQ
Ns7TwsbMXdA5qIyAAVABCkNLQts0TcrCF0OEKSbN++oI2V3N0vUUPX7AQJ6AoAtG
L7UIqMn0AmAKi5wxIAXSMqhI5DPHwLD3uKPjPCAtCP7B4B188ZVCHzgvoxGVuLA=
=fQ7r
-----END PGP SIGNATURE-----