#1061: default class methods can be given too general a type as GHC tries to 
share
them between instances
-------------------------+--------------------------------------------------
    Reporter:  igloo     |       Owner:         
        Type:  bug       |      Status:  new    
    Priority:  normal    |   Milestone:  6.8    
   Component:  Compiler  |     Version:  6.6    
    Severity:  normal    |    Keywords:         
  Difficulty:  Unknown   |    Testcase:  tc199  
Architecture:  Unknown   |          Os:  Unknown
-------------------------+--------------------------------------------------
GHC does not accept this program:

 {{{
 class Baz v x where
    foo :: x -> x
    foo y = y

 instance Baz Int Int
 }}}

 even though it does accept this one:

 {{{
 class Baz v x where
    foo :: x -> x
    foo y = y

 instance Baz Int Int where
    foo y = y
 }}}

 I can't find an explicit answer in the report, but I think both should be
 accepted.

 The bug says, in a comment,

 {{{
 This code defines a default method with a highly dubious type,
 because 'v' is not mentioned, and there are no fundeps

 However, arguably the instance declaration should be accepted,
 beause it's equivalent to
  instance Baz Int Int where { foo x = x }
 which *does* typecheck

 GHC does not actually macro-expand the instance decl.  Instead, it
 defines a default method function, thus

  $dmfoo :: Baz v x => x -> x
  $dmfoo y = y

 Notice that this is an ambiguous type: you can't call $dmfoo
 without triggering an error.  And when you write an instance decl,
 it calls the default method:

  instance Baz Int Int where foo = $dmfoo

 I'd never thought of that.  You might think that we should just
 *infer* the type of the default method (here forall a. a->a), but
 in the presence of higher rank types etc we can't necessarily do
 that.
 }}}

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/1061>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs

Reply via email to