[GHC] #13295: Failure to resolve type parameter determined by type family

[GHC]

#13295: Failure to resolve type parameter determined by type family
-------------------------------------+-------------------------------------
        Reporter:  alexvieth         |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler (Type    |              Version:  8.0.1
  checker)                           |
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  GHC rejects       |  Unknown/Multiple
  valid program                      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by goldfire):

 The error is poor (ambiguity isn't really the issue), but I agree that GHC
 should reject your first function but accept your second.

 Your first function relies critically on the fact that `D` has but one
 constructor. Thus, we know that an element of `D` is constructed with
 `'D`. Without this knowledge, then `introG` really is problematic.
 Although we believe it's sound to do so, GHC does not currently do this
 kind of reasoning (i.e., eta-expansion). In `introG'`, the type signature
 tells us that `F t` evaluates to a use of `'D`, so the missing piece is
 provided.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13295#comment:1>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler