On Tue, Jul 10, 2007 at 09:15:12PM +0200, ngaba@petra.hos.u-szeged.hu wrote:
Back to pacman: There is a problem with the old/new resolvedeps: this is not deterministic (multiple provision) and sometimes one choice is better than an other one: for example "choice A" induces a package conflict, and choice B would solve the problem and would install the required package (and its dependencies) but pacman is not clever enough to find "choice B"...
Hmm, the old way was already not good, but after your patch, it's worse imo, because we lose the simple hint of trying the actual package name first. I think that's what we want in most cases when we let pacman auto resolve the dependencies for us. One example : I install acidrip which depends on mplayer, I most likely want mplayer to be pulled, not mplayer-svn. If pacman has to make one blind choice, I think it should rather try the actual package first. But well, that's already what you wanted to do for speed concern : http://www.archlinux.org/pipermail/pacman-dev/2007-June/008539.html