existentials containing type-class(dictionaries) require more type-signatures in 6.7

I don't understand the "rules" well enough to know if this is actually a bug, or if the type signature is supposed to be required. (Also I need to update my build of HEAD... but before I forget,)

{-# LANGUAGE ExistentialQuantification #-}
module Test where

import Data.Typeable

data D = forall a. Typeable a => D a

--type signature only required in 6.7, not 6.6.1
fromD :: Typeable result => D -> result -> Bool
fromD (D a) def =
   typeOf a == typeOf def

ghc 6.7's error message when the type signature is commented out:

ghc67tysigreg.hs:10:15:
    Could not deduce (Typeable a) from the context (Typeable a1)
      arising from a use of `typeOf' at ghc67tysigreg.hs:10:15-24
    Possible fix:
      add (Typeable a) to the context of the constructor `D'
    In the second argument of `(==)', namely `typeOf def'
    In the expression: (typeOf a) == (typeOf def)
    In the definition of `fromD':
        fromD (D a) def = (typeOf a) == (typeOf def)

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