[GHC] #15589: Always promoting metavariables during type inference may be wrong

[GHC]

#15589: Always promoting metavariables during type inference may be wrong
-------------------------------------+-------------------------------------
        Reporter:  goldfire          |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:  8.6.1
       Component:  Compiler          |              Version:  8.4.3
      Resolution:                    |             Keywords:  TypeInType
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):

 Here's another idea: if a type signature has any free, lexically-scoped
 type variables, then treat it like a ''partial'' type signature.   That
 is, used to guide and constrain type inference, but NOT used as the single
 canonical specification of the type of `x`, nor used to support
 polymorphic recursion.

 That's still annoying: we might want polymorphic recursion, and if so how
 would we get it.  But the point about having free unification variables is
 that we don't really know the type at all, yet.   And that's what partial
 type signatures are all about.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/15589#comment:13>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler