[GHC] #11608: Possible type-checker regression in GHC 8.0 when compiling `microlens`

[GHC]

#11608: Possible type-checker regression in GHC 8.0 when compiling `microlens`
-------------------------------------+-------------------------------------
        Reporter:  hvr               |                Owner:  goldfire
            Type:  bug               |               Status:  new
        Priority:  highest           |            Milestone:  8.0.1
       Component:  Compiler (Type    |              Version:  8.0.1-rc2
  checker)                           |
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  GHC rejects       |  Unknown/Multiple
  valid program                      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 Aha.  In a class decl we use a `TyVar` (not a skolem `TcTyVar`) for the
 class variable(s).  (There is a good reason for this: in the final `Class`
 value we don't want any `TcTyVar`s.)

 Then in `TcFlatten.flatten_tyvar` we see
 {{{
 flatten_tyvar tv
   | not (isTcTyVar tv)             -- Happens when flatten under a (forall
 a. ty)
   = flatten_tyvar3 tv
 }}}
 This case catches that class variable.  In the ambiguity check for the
 type sig for `each` we have a Given equality for the class variable.
 Then, because of the above test we fail to take advantage of the Given.
 (And the error message too is misleading.)

 Solution: delete the above equation, so that we treat `TyVar` and skolem
 `TcTyVar` the same.  I'm validating now.

 The `TyVar`/`TcTyVar` story is not great.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11608#comment:8>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler