[GHC] #8565: New GeneralisedNewtypeDeriving needs help with higher rank types

[GHC]

#8565: New GeneralisedNewtypeDeriving needs help with higher rank types
-------------------------------------------------+-------------------------
        Reporter:  simonpj                       |            Owner:
            Type:  bug                           |           Status:  new
        Priority:  normal                        |        Milestone:
       Component:  Compiler                      |          Version:  7.6.3
      Resolution:                                |         Keywords:
Operating System:  Unknown/Multiple              |     Architecture:
 Type of failure:  None/Unknown                  |  Unknown/Multiple
       Test Case:                                |       Difficulty:
  typecheck/should_compile/T8565                 |  Unknown
        Blocking:                                |       Blocked By:
                                                 |  Related Tickets:
-------------------------------------------------+-------------------------

Comment (by nomeata):

 Hmm, after more reading of the type class inference code I have doubts
 that this can be implemented without a larger refactoring of the the the
 typechecker and desugarer for instances.

 One of the reasons is that when `instance (forall a. (Coercible t1 t2)) =>
 Coercible (forall a. t1) (forall a. t2)` were used when solving a
 constraint, the type variable will get turned into a parameter to the
 dictionary for the superclass, so is local to that. But if that constraint
 requires further instances, their dictionaries will require the same
 parameter, so this would require, in a way, a nested structure for the
 type class solver.

 The only work-around I can think of is asking the user to write the
 instance for `C` (or in the case of conduit, `MFunctor`) by hand. Which is
 unsatisfying.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8565#comment:8>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler