On Thu, 08 Apr 2010 16:08:57 +0200 Patrick Lauer <patr...@gentoo.org> wrote: > > Please detail your cycle breaking algorithm that works in such a way > > that it's guaranteed not to toggle flags that will break a system, > > but that does not require any changes to be made to ebuilds etc > > telling the package manager what those flags are. > > > That would violate the Entscheidungsproblem. > > But why would you make the cycle breaking depend on an oracle? Once we > have the method in place we can find the two special cases and think > of ways to fix them.
The problem is, the special cases where it goes horribly wrong aren't rare. As soon as you start looking for cycles, you find flags like 'build', 'bootstrap' and 'acl'. > Abandoning the idea as a whole just because there may be a corner > case that isn't as easy appears quite silly to me. I'm not after abandoning the idea. I'm after doing it properly, and doing it properly starts by handling the problematic cases rather than pretending they don't exist. We've already seen repeatedly what goes wrong when you start with the assumption "it'll probably work" and then pile on special exceptions every time someone gets screwed over by something you didn't think of. Why don't we go with "we'll only do it for things where we know it'll work" instead this time? -- Ciaran McCreesh
Description: PGP signature