[GHC] #9131: Experiment with a dedicated solver for Coercible

[GHC]

#9131: Experiment with a dedicated solver for Coercible
-------------------------------------+------------------------------------
        Reporter:  nomeata           |            Owner:
            Type:  task              |           Status:  new
        Priority:  low               |        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:
-------------------------------------+------------------------------------

Comment (by goldfire):

 Hmmm... this makes sense. Solving for a `Coercible` constraint involves
 "looking through" newtypes, which may be recursive. This is analogous to
 the situation with type families, where we require `UndecidableInstances`
 to do this sort of thing. For good reasons, we can't require
 `UndecidableInstances` for recursive newtypes. I'm not sure what the best
 solution here is, though. A part of me would like to define some subset of
 this problem for which we guarantee completeness (perhaps this subset is
 "non-recursive newtypes"), not unlike with type families.

 Good idea to pursue this line.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9131#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler