-- This pretends to spot two more bugs in ghc-4.02:
----------------------------------------------------------------------
module Bug where
data R = R {r :: Int}
rr = R {r=-1}
data Key = K1 | K2 | K3 deriving(Eq)
-- data Mix a = Mix [(Key, c a)]
-- f :: Mix a -> Mix a
-- f (Mix ps) = case lookup K1 ps of Just _ -> Mix ps
----------------------------------------------------------------------
ghc -c Bug.hs yields
...:4:10: Haskell 98 does not support 'punning' on records on input: "=-"
Then, replace `=-1' with `= -1', un-comment the rest of the module and
command
ghc -c -fglasgow-exts Bug.hs
It reports
panic! (the `impossible' happened):
zonkTcTypeToType: free type variable with non-* type: c{-a144-}
Also i have a question on the language extension:
is this Mix a, f a correct attempt to use existential types?
I mean putting into association list ps :: Mix Int
the items like, say, (K1,[0]), (K2,Just 0) - here [], Just are the
costructor instances for `c' from the Mix definition.
------------------
Sergey Mechveliani
[EMAIL PROTECTED]
