[Haskell-cafe] type refinement to a typeclass constrained type

[Simon Peyton-Jones]

| Interesting thing though, it's the same on 6.4.1.
|
| Wasn't this brokenness introduced in 6.6, but rather has lingered in
| ghc for a while, or since the introduction of GADTs, at all?

No it's always been broken.  I finally made GADTs and type classes work together in the HEAD a few months ago, using implication constraints for the first time.