[GHC] #11416: GHC mistakenly believes datatype with type synonym in its type can't be eta-reduced

[GHC]

#11416: GHC mistakenly believes datatype with type synonym in its type can't be
eta-reduced
-------------------------------------+-------------------------------------
        Reporter:  RyanGlScott       |                Owner:
            Type:  bug               |               Status:  closed
        Priority:  normal            |            Milestone:  8.0.1
       Component:  Compiler (Type    |              Version:  8.0.1-rc1
  checker)                           |
      Resolution:  fixed             |             Keywords:  TypeInType
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  GHC rejects       |  Unknown/Multiple
  valid program                      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):  Phab:D1772
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by RyanGlScott):

 Hm, there's another tricky case that I hadn't previously considered:

 {{{#!hs
 type ConstT a b = a
 data family Fam a b
 data instance Fam (ConstT a b) b = Fam b deriving Functor
 }}}

 Currently, the eta-reduction check rejects this. Should it?

 On on hand, the `b` in `ConstT a b` will go away if you expand the
 `ConstT` type synonym, so one could argue this is no different from `data
 instance Fam a b = Fam b deriving Functor` (which is accepted).

 On the other hand, if you consider what the derived instance would look
 like:

 {{{#!hs
 instance Functor (Fam (ConstT a b)) where ...
 }}}

 This feels a bit off, because now the instance head is mentioning an eta-
 reduced type variable! But it does typecheck if you type it in manually...

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11416#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler