[GHC] #12124: Ambiguous type variable: it's a red herring!

[GHC]

#12124: Ambiguous type variable: it's a red herring!
-------------------------------------+-------------------------------------
        Reporter:  drb226            |                Owner:
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.0.1
      Resolution:                    |             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:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 For the Description, what is happening is this:

 * The ill-arity'd pattern `Whoops a -> ..` makes the typechecker emit an
 error message and fail with a hard error; rightly so.

 * Usually the typing constraints arising from the failing computation
 would be discarded. But as a short-cut for a common case, GHC does not
 begin a new implication constraint for the RHS of `foo`, because it has no
 tyvars and no given constraints.

 * So the `Monad a` constraint arising from the call of `return` is
 accumulated into the top-level constraints of the whole program; so after
 recovering from the error GHC tries to solve the constraint.

 Two possible solutions

 * Remove the special case.  That looks OK, but it'd mean that we might
 solve `(Num Int)` repeatedly in different function definitions.  CSE can
 common-up later, but it's just more work.

 * In the special case, accumulate constraints in a separate variable, and
 union them into the outer context only if the typecheck succeeds.  Not
 hard to do; I think this is probably best.

 Thanks for the accurate report!

 Simon

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12124#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler