[GHC] #14331: Overzealous free-floating kind check causes deriving clause to be rejected

[GHC]

#14331: Overzealous free-floating kind check causes deriving clause to be rejected
-------------------------------------+-------------------------------------
        Reporter:  RyanGlScott       |                Owner:  (none)
            Type:  bug               |               Status:  merge
        Priority:  normal            |            Milestone:  8.2.2
       Component:  Compiler (Type    |              Version:  8.2.1
  checker)                           |
      Resolution:                    |             Keywords:  deriving
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  GHC rejects       |            Test Case:
  valid program                      |  deriving/should_compile/T14331
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by RyanGlScott):

 It will lead to an accepting definition, yes, but I don't think it's the
 definition you'd want (or at least, it seems to differ from the proposal I
 put forth in comment:14). If I understand your argument correctly, you
 want all type variables in a `deriving` clause's type to be scoped
 differently from the data type, and expect unification to save you in the
 end? That is, this:

 {{{#!hs
 data D a = D deriving (C (a :: k))
 }}}

 Would be treated like this?

 {{{#!hs
 data D {k1} (a1 :: k1) = D deriving (forall {j} k (a :: k). C @k @j a)
 }}}

 If so, we have a problem—the //only// kinds we'd unify are the `j` in `C a
 :: j -> Constraint` and `Type` (the kind of `D a`), so the instance we'd
 get in the end is:

 {{{#!hs
 instance forall {k1} k {a1 :: k1} (a :: k). C @k @Type a (D @k1 a1) where
 ...
 }}}

 Which is not what I'd expect, since `a` and `a1` are distinct. The only
 way to ensure that they're the same is to interpret the scope like this:

 {{{#!hs
 data D {k} (a :: k) = D deriving (forall {j}. C @k @j a)
 }}}

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14331#comment:19>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler