[GHC] #9118: Can't eta-reduce representational coercions

[GHC]

#9118: Can't eta-reduce representational coercions
-------------------------------------+------------------------------------
        Reporter:  goldfire          |            Owner:
            Type:  feature request   |           Status:  new
        Priority:  normal            |        Milestone:
       Component:  Compiler          |          Version:  7.8.2
      Resolution:                    |         Keywords:
Operating System:  Unknown/Multiple  |     Architecture:  Unknown/Multiple
 Type of failure:  None/Unknown      |       Difficulty:  Unknown
       Test Case:                    |       Blocked By:
        Blocking:                    |  Related Tickets:  #9117
-------------------------------------+------------------------------------

Comment (by goldfire):

 I agree with rwbarton on this -- I think the `F1`/`F2` example doesn't
 really apply.

 But, there remains a problem case:

 {{{
 newtype Pair a = Mk (a, a)
 }}}

 We can easily derive `Coercible (Pair Int) ((,) Int Int)`, but it would be
 bad to derive `Coercible Pair ((,) Int)`. The problem is that the newtype
 definition has a repeated variable on the right, and thus the definition
 cannot be eta-reduced. Probably, a sufficiently complicated Core
 construction can detect and avoid this case, but I don't think it would be
 easy.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9118#comment:6>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler