[GHC] #8631: Need ImpredicativeTypes for GeneralizedNewtypeDeriving?

[GHC]

#8631: Need ImpredicativeTypes for GeneralizedNewtypeDeriving?
-------------------------------------+------------------------------------
        Reporter:  goldfire          |            Owner:
            Type:  bug               |           Status:  new
        Priority:  normal            |        Milestone:
       Component:  Compiler          |          Version:  7.7
      Resolution:                    |         Keywords:
Operating System:  Unknown/Multiple  |     Architecture:  Unknown/Multiple
 Type of failure:  None/Unknown      |       Difficulty:  Unknown
       Test Case:                    |       Blocked By:
        Blocking:                    |  Related Tickets:
-------------------------------------+------------------------------------

Comment (by goldfire):

 Yes, absolutely, explicit type application would make all of this much
 easier. But, I believe we can still fix this bug without it, once we have
 some design decisions about the questions I posed originally.

 As for question 1: I think it's safe to turn on !ImpredicativeTypes over
 the code produced by GND. This would make using GND easier, and I don't
 want to bother users with this technicality.

 As for question 2: In general, I'm a little worried about error messages
 that mention `coerce` yet refer to code that patently does ''not'' mention
 `coerce`. If we can't get rid of the `coerce`, could we add some extra
 context describing the use of GND and perhaps suggesting `-ddump-deriv`?

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8631#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler