Re: [UPLOADED] RFC: The Future of Solving Dependency Problems in APT (SAT, CUDF)
On Di, 2010-12-28 at 16:17 +0100, Pietro Abate wrote:
>
> Here there is a description of the encoding used by other solvers :
> http://mancoosi.org/reports/d4.2.pdf
>
> It might be an interesting read.
I read it, and tried the cudfsolver, but was a bit disappointed that it
* takes 50 seconds to solve a simple problem[1] (lpsolve/glpk)
* wants to remove the whole system when using lpsolve, even in
paranoid
In contrast to this, aspcud-paranoid-1.3 takes 1.3 seconds, and seems to
produce a better result. I also played a bit with liblpsolve, but even
adding self-conflicts seems to take 1 second. The result: Solutions
using lpsolve and glpk are disqualified for a standard solver (glpk in
general, as it's license is incompatible).
The problem chosen was installing nautilus from experimental which
should be equal to an upgrade to experimental. I have put it as a
gz-compressed document at my website[1].
[1] http://jak-linux.org/tmp/example.cudf.gz
--
Julian Andres Klode - Debian Developer, Ubuntu Member
See http://wiki.debian.org/JulianAndresKlode and http://jak-linux.org/.
Reply to: