On Sun, Dec 02, 2007 at 01:41:37PM +0100, "Steinar H. Gunderson" <[EMAIL
PROTECTED]> was heard to say:
> When trying to install bzrtools I get:
[snip]
> The following actions will resolve these dependencies:
>
> Keep the following packages at their current version:
> bzrtools [Not Installed]
>
> Downgrade the following packages:
> bzr [1.0~rc1-2 (unstable, now) -> 0.11-1.1 (stable)]
>
> Score is -9851
>
> Accept this solution? [Y/n/q/?] n
> Resolving dependencies...
> The following actions will resolve these dependencies:
>
> Keep the following packages at their current version:
> bzrtools [Not Installed]
>
> Score is -9881
>
> Accept this solution? [Y/n/q/?]
>
> How can doing a downgrade be _better_ than not doing anything at all?
> Clearly the second solution is better than the first one. :-)
Yes, this is a funny side effect of how aptitude scores solutions.
I can't be 100% sure, but I *think* there are two effects here:
(a) A penalty of 40 points for downgrading the package (installing
a non-default version)
(b) A bonus of 70 points for advancing the resolver one step.
(b) is obviously what's causing the problem, and it's a symptom of how
aptitude uses the score of a solution to force the search to
pursue individual branches of the search space rather than jumping
back and forth. I'm hesitant to adjust this weight, since this is the
first example I've seen where it does a severely wrong thing, and several
examples came up in the past where smaller values of StepScore would
cause the resolver to flail around for long periods of time trying to
find a solution.
Daniel
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
