[GHC] #13537: Adding underscore (m :: _) to pattern variable makes it fail

[GHC]

#13537: Adding underscore (m :: _) to pattern variable makes it fail
-------------------------------------+-------------------------------------
        Reporter:  Iceland_jack      |                Owner:  (none)
            Type:  bug               |               Status:  closed
        Priority:  normal            |            Milestone:
       Component:  Compiler (Type    |              Version:  8.0.1
  checker)                           |
      Resolution:  invalid           |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------
Changes (by Iceland_jack):

 * status:  new => closed
 * resolution:   => invalid


Comment:

 Replying to [comment:2 simonpj]:
 > Now you want to force the type of the pattern to be `_`.  But since GHC
 only supports predicative polymorphism, `_` must be a monotype.  So GHC
 instantiates `forall a. a->a` to `alpha -> alpha` (for some new
 unification variable alpha), and proceeds.

 I suspected this was related to impredicativity. As a user I expected
 transforming the pattern `m` to `(m :: _)` would have no effect since it
 binds nothing (the compiler ''could'' special-case it, even though it
 brings no benefits).

 > Does that help?
 >
 > Is there some way we could improve the user manual?

 Yes it helps. I think the impredicative polymorphism entry is rather good,
 thanks for the explanation.

 To be clear, this is also expected? Closing the ticket

 {{{#!hs
 -- Works
 tId :: T -> T
 tId (MkT g) = MkT g

 -- Fails
 tId :: T -> T
 tId (MkT g) = MkT g' where
   g' = g
 }}}

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13537#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler