ghc-6.7: panic! (the 'impossible' happened) -- typechecker getting confused by Data.Generics?

This bug is triggered by http://www.cse.unsw.edu.au/~dons/lambdabot/Lib/Pointful.hs. A reduced test case follows.

{-# OPTIONS_GHC -fglasgow-exts #-}
module Pointful' where

import Language.Haskell.Syntax
import Data.Generics
import Control.Monad.State

-- rename :: () -> HsExp -> State (HsName, [HsName]) HsExp
rename = \_ -> everywhereM (mkM (\e -> case e of HsWildCard -> return e))
{-
Pointful.hs:9:0:ghc-6.7: panic! (the 'impossible' happened)
  (GHC version 6.7 for i386-unknown-linux):
        dictPred
    ic{v a1nF} :: {forall a{tv a1iH} [sk].
                   (base:Data.Generics.Basics.Data{tc 29} a{tv a1iH} [sk]) =>
                   <pred>base:GHC.Base.Monad{tc 28} m{tv a1nq} [tau]}
                  Refinement {[]}
-}

-- rename _ = everywhereM (mkM (\e -> case e of  HsWildCard -> return e))
{-
Pointful.hs:30:17:
    Couldn't match expected type `Monad m' against inferred type `()'
    In the first argument of `rename', namely `()'
    In the expression: rename () undefined
    In the definition of `uncomb'': uncomb' = rename () undefined
-}

uncomb' :: State (HsName, [HsName]) HsExp
uncomb' = rename () undefined

The errors can be reproduced by uncommenting the line immediately preceding it (and commenting out the other one). In both cases the error can be fixed by uncommenting the type signature for rename.

Tested with ghc-6.7 as of 2006-11-27. ghc-6.6 and ghc-6.4.2 are both fine.

Trac metadata

Trac field	Value
Version	6.7
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system	Unknown
Architecture	Unknown