[GHC] #11511: Type family producing infinite type accepted as injective

[GHC]

#11511: Type family producing infinite type accepted as injective
-------------------------------------+-------------------------------------
        Reporter:  jstolarek         |                Owner:
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler (Type    |              Version:  8.1
  checker)                           |             Keywords:  TypeFamilies,
      Resolution:                    |  Injective
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 jstolarek):

 I had a talk about injective type families yesterday and it provoked me to
 rethink this issue. What we do in the pre-unification algorithm is:

   1. assume that type family applications unify with anything
   2. but we look under injective type family applications and unify the
 arguments

 Now, my question is: when we are checking whether `F` is injective why do
 we assume that `F`'s injectivity annotation is true? This is what we're
 trying to prove and I think we should not assume that as part of our
 proof. This is exactly what's happening here.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11511#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler