[GHC] #16146: Trivial partial type signature kills type inference in the presence of GADTs

[GHC]

#16146: Trivial partial type signature kills type inference in the presence of
GADTs
-------------------------------------+-------------------------------------
        Reporter:  goldfire          |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.6.3
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 '''Forall-or-nothing behaviour'''

 >  Another interesting example has cropped up. Re-consider

 Partial type signatures have always had the semantics that you can use a
 `forall` to bind the type variables you need to mention but the final type
 signature may have more.  E.g.
 {{{
 f :: forall a. _ -> a -> Bool
 f x y = True
 }}}
 Here the final sig is `f :: forall a b. b -> a -> Bool`.

 It would be possible to say that
 * If you use a `forall` there must be no more quantifiers (except,
 `Inferred` ones?), or
 * There can be extra quantifiers, but they are `Inferred`, so we'd get `f
 :: forall {b}. forall a. b -> a -> Bool`.  (I quite like this.), or
 * The current behaviour.

 I don't feel strongly.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/16146#comment:11>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler