Marcin 'Qrczak' Kowalczyk wrote:
> Requires overlapping instances in a case where in fact something is
> X and the way it should be Z is different than that induced by X.
> And overlapping instances themselves can be lost because of context
> reduction, where a polymorphic function will use the generic instance
> of something else instead of the specific one - I can think about an
> example if it's unclear.
If context reduction choses a more generic instance when a more specific one exists,
then I consider that a bug. This was fixed in Hugs last September, and just a couple
of days ago in GHC. Check the glasgow-haskell-bugs list archive for details. The
thread is titled `bug in ghc-4.06 ?' and was reported by ... Sergey!
--Jeff
