[GHC] #8745: GeneralizedNewtypeDeriving is still not Safe

[GHC]

#8745: GeneralizedNewtypeDeriving is still not Safe
-------------------------------------+------------------------------------
        Reporter:  goldfire          |            Owner:
            Type:  bug               |           Status:  new
        Priority:  normal            |        Milestone:
       Component:  Compiler          |          Version:  7.8.1-rc1
      Resolution:                    |         Keywords:  Safe
Operating System:  Unknown/Multiple  |     Architecture:  Unknown/Multiple
 Type of failure:  None/Unknown      |       Difficulty:  Unknown
       Test Case:                    |       Blocked By:
        Blocking:                    |  Related Tickets:
-------------------------------------+------------------------------------

Comment (by goldfire):

 I've found a weird corner case that is very relevant.

 Consider

 {{{
 class Foo a where ...
 type role Foo representational
 data PackFoo a where
   MkPF :: Foo a => PackFoo a
 instance Foo Int where ...
 newtype Age = MkAge Int
 instance Foo Age where ...
 incoherent :: PackFoo Age
 incoherent = coerce (MkPF :: PackFoo Int)
 useIncoherent = case incoherent of MkPF -> {- here, we two incompatible
 instances of Foo Age available! -}
 }}}

 The problem, as I see it, boils down to the role annotation on `Foo`,
 which allows users to create incoherence. Should we perhaps disallow role
 annotations on classes in Safe Haskell? It would be analogous to the
 restrictions on overlapping instances, I think.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8745#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler