[GHC] #14782: typeclass polymorphism defeats bang patterns

[GHC]

#14782: typeclass polymorphism defeats bang patterns
-------------------------------------+-------------------------------------
        Reporter:  dmwit             |                Owner:  (none)
            Type:  bug               |               Status:  closed
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.2.2
      Resolution:  invalid           |             Keywords:
                                     |  MonomorphismRestriction
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 goldfire):

 Let's look at your two examples:

 {{{#!hs
 main = seq (undefined :: Monoid a => a) (return ())
 }}}

 The type-checker looks at the first argument of `seq` and sees something
 of `forall a. Monoid a => a`. So it instantiates this argument, providing
 a unification variable `t` and generating the constraint `Monoid t`.
 Later, the constraint solver discovers that it can't solve for `Monoid t`,
 and you've lost the game.

 {{{#!hs
 main = let n :: Monoid a => a; !n = undefined in return ()
 }}}

 Here, the `undefined` is checked in an environment where we've quantified
 over `a` and can assume `Monoid a`, as is usual in the definition of a
 variable with a constrained type. No problem there. And, despite thinking
 of `!n` as desugaring to `seq n` somewhere, there is nowhere that GHC
 instantiates `n`. (Maybe you're arguing that we're not faithful to that
 desugaring... and you'd be right.) GHC just forces `n` right as it is,
 which exposes the lambda -- but not the `undefined` -- and then carries
 on.

 Does that help?

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14782#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler