Bug#699759: apt: score computation may prefer obsolete installed packages over their successors
On 2013-05-09 18:43, David Kalnischkies wrote:
> In general, I wonder a bit why we are not just +2 all priority scores and
> have therefore always positive (unsigned) numbers.
I tried this when I looked into this priority issue ... I think I even
added 3 s.t. there would be no (interesting) packages with a score of 0
that would not be dumped.
I quickly discared that approach ... I don't remember exactly why, but
it was not helpful for debugging. Probably because it was not preserving
the relative priority ordering of the packages at all. And I only looked
at this for exactly one upgrade path. Perhaps a global rescoring would
have made things different (and better) globally.
Anyway, we still could test new scores and how they influence the
squeeze2wheezy updates ...
In the end I added some methods to manipulate the scores that do a "lazy
initialization to 0" after initializing all scores to some magic
(-9999), s.t. I could get a dump of all scores *including 0* that were
updated during the computation. Having these missing is nasty for
debugging. (There is alaready a bug filed about this.) Would you be
interested in this patch? (In its current form it's a hack and
proof-of-concept that I wouldn't propose for direct inclusion).
Andreas
Reply to: