[GHC] #15918: GHC panic from QuantifiedConstraints(?)

[GHC]

#15918: GHC panic from QuantifiedConstraints(?)
-------------------------------------+-------------------------------------
        Reporter:  Iceland_jack      |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.6.2
      Resolution:                    |             Keywords:
                                     |  QuantifiedConstraints
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 > also the NB comment right after the call to isReflexiveCo.

 To be explicit, you mean
 {{{
 mkCastTy :: Type -> Coercion -> Type
 mkCastTy ty co | isReflexiveCo co = ty  -- (EQ2) from the Note
 -- NB: Do the slow check here. This is important to keep the splitXXX
 -- functions working properly. Otherwise, we may end up with something
 -- like (((->) |> something_reflexive_but_not_obviously_so) biz baz)
 -- fails under splitFunTy_maybe. This happened with the cheaper check
 -- in test dependent/should_compile/dynamic-paper.
 }}}

 > Suppose we have tyco :: Type ~ Coercion

 It's worth noting that the type checker will produce an error message for
 any such coercion.  It's like `Int ~ Bool`.  I feel somewhat relaxed about
 the impact on Core of not upholding an equality that will be rejected by
 the typechecker anyway.  (Long term I think we probably want to make Type
 and Coercion distinct everywhere.)

 > mkCastTy discards reflexive coercions that aren't Type-vs-Coercion, and
 coreView discards the rest

 Hm. You mean something like
 {{{
 tcView (ty |> co)
   | ty1 == coercionKind, ty2 == typeKind = Just ty
   | ty1 = typeKind, ty2 == coercionKind  = Just ty
   where
     (ty1,ty2) = coercionKind co
 }}}
 I suppose that'd be possible.  But I agree it's smelly.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/15918#comment:9>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler