Re: SAT based britney
On Thu, 23 Jun 2011, Mehdi Dogguy wrote:
> Yes, but zack's answer confirms that (at least for me) we won't have any
> better solutions because testing migration problem isn't a trivial
> problem. So, personally, I remain not convinced that a new SAT solver
> would make things easier or better for us. We will have to enter "hints"
> (no matter what implementation is used) to make choices the solver can't
> infer automatically. (as julien said in the beginning of this thread).
Well, from my understanding the SAT solver would drop the need to provide
"easy" hints. "force" hints where you deliberately break dependencies
would require some input to tell the solver which constraints can be
But it's still a big win IMO.
> Having that said, what might help us is to optimize britney and fix some
> corner cases that could be handled in a better way.
Maybe start by documenting them with wishlist bug reports against
Raphaël Hertzog ◈ Debian Developer
Follow my Debian News ▶ http://RaphaelHertzog.com (English)
▶ http://RaphaelHertzog.fr (Français)