[GHC] #8767: Add rules involving `coerce` to the libraries

[GHC]

#8767: Add rules involving `coerce` to the libraries
-------------------------------------------------+-------------------------
        Reporter:  nomeata                       |            Owner:
            Type:  task                          |           Status:  new
        Priority:  normal                        |        Milestone:  7.8.1
       Component:  libraries/base                |          Version:
      Resolution:                                |  7.8.1-rc2
Operating System:  Unknown/Multiple              |         Keywords:
 Type of failure:  None/Unknown                  |     Architecture:
       Test Case:                                |  Unknown/Multiple
  tests/simplCore/should_run/T2110.hs            |       Difficulty:
        Blocking:                                |  Unknown
                                                 |       Blocked By:  8718
                                                 |  Related Tickets:  #2110
-------------------------------------------------+-------------------------

Comment (by ekmett):

 The `fmapCoerce` solution turned out not to be powerful enough to handle
 things like `Compose f g` while `g` remains polymorphic or many monad
 transformers. We vacillated between a few other solutions that let you add
 a different member to Functor that could lift coercions by the end of
 ICFP, but with GHC 7.8 happening "any day now" at the time we agreed to
 just punt on it until we could give it due deliberation.

 My main point was that if the code is still polymorphic in the functor and
 complicated enough not to inline into a place where that rule can fire the
 rule doesn't help. We encourage people to write very polymorphic code.
 e.g. It solves many of the bugaboos people trot out about monad
 transformers and the like if you avoid picking a concrete instance until
 the last second. I'd be really sad to see that part of our culture die to
 get a few zero-cost newtype applications.

 Anyways, not sure what the right solution is yet, but I did want to
 interject that whatever it is, `fmapCoerce` isn't it. ;)

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8767#comment:13>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler