[GHC] #14101: Type synonyms can make roles too conservative

[GHC]

#14101: Type synonyms can make roles too conservative
-------------------------------------+-------------------------------------
        Reporter:  dfeuer            |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:  8.4.1
       Component:  Compiler (Type    |              Version:  8.2.1
  checker)                           |
      Resolution:                    |             Keywords:  roles
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  GHC rejects       |  Unknown/Multiple
  valid program                      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #8234             |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by goldfire):

 Sorry -- I wrote a post here but somehow it got lost. (I'm sure this is my
 fault.)

 This is by design. Any non-nominal role on a type parameter must be
 assigned to a proper type parameter to that type. When you eta-reduce,
 there's no place to put the non-nominal role.

 Happily, Ryan's patch solves the problem and looks utterly correct to me.
 So, while the synonym's argument's role is still nominal, that fact
 becomes irrelevant during role inference.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14101#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler