[GHC] #9918: GHC chooses an instance between two overlapping, but cannot resolve a clause within the similar closed type family

[GHC]

#9918: GHC chooses an instance between two overlapping, but cannot resolve a
clause within the similar closed type family
-------------------------------------+-------------------------------------
        Reporter:  qnikst            |                   Owner:
            Type:  bug               |                  Status:  new
        Priority:  normal            |               Milestone:
       Component:  Compiler          |                 Version:  7.8.3
      Resolution:                    |                Keywords:
Operating System:  Unknown/Multiple  |            Architecture:
 Type of failure:  None/Unknown      |  Unknown/Multiple
      Blocked By:                    |               Test Case:
 Related Tickets:                    |                Blocking:
                                     |  Differential Revisions:
-------------------------------------+-------------------------------------

Comment (by qnikst):

 Seems I understood Simon differently. As far as I understood he advices
 make check for classes more strict, and less strict. So

 1. For classes instances:

 {{{
 instance F a (M a) where ...
 instance F a a where ...
 }}}

 Will no longer be accepted due to surely-apart check. And Equals
 typefamily will have the same rules as it has now.

 2. As less programs will be accepted, and some really relies on the
 current behaviour, it's possible to introduce `{-# UNDECIDABLE #-}` pragma
 for type class that will remove new "surely-apart" check and instances
 will be accepted

 3. (Simon didn't say it). In order to unify type families and type class
 bevaiour it's possible to allow `{-# UNDECIDANBLE #-}` pragma to be
 applied to type family. Yes, if it will be applied to a typefamily that
 should have surely-apart check (e.g. `Equals`) it will be broken. However
 if programmer takes responsibility and guarantee that this check is not
 needed (2 examples above) then
 he can use this pragma in order to make compiler accept this program.

 So as far as I understood this suggestion doesn't allow any new ill-typed
 program to be accepted, more over some programs that are accepted now will
 not be accepted then, but proposal provides a backdoor so programmer can
 have old behaviour.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9918#comment:14>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler