[Haskell-cafe] Possible bug in GHC 7.0.3

[Daniel Fischer]

On Wednesday 29 June 2011, 10:22:20, Yitzchak Gale wrote:
> Ryan Ingram wrote:
> > So this is definitely a GHC bug, but I think the problem is probably
> > triggered by this line:
> > 
> > instance  Serializable a b => IResource a
> > 
> > I don't think this is a valid instance declaration without a
> > functional dependency on Serializable, as it's impossible to know
> > which type 'b' to use in the methods of IResource.
> 
> That's not exactly the reason why that line is a problem.
> The methods of IResource don't use b. Presumably, some of
> the methods of Serializable don't mention b either, and those
> could be used in the instance.
> 
> But what exactly does this instance mean? Where does the
> implicit "forall b" go? For this to make any sense, it would
> need to be:
> 
> -- Illegal syntax for instance
> instance (forall b. Serializable a b => IResource a)
> 
> I don't think it has ever been specified whether it is legal
> to have free type variables as parameters of a multi-parameter
> class in the superclass context of an instance declaration,
> and if so, how to interpret that.
> 
> Haskell 98/2010 does seem to allow it for single-parameter
> classes, but there the obvious meaning would be just to
> ignore those parts of the context.
> 
> In any case, GHC does seem to be getting very confused.
> Here is a smaller test case. While it does not trigger the
> crash, it does cause GHC to give a seemingly nonsensical
> error message:
> 
> module ClassContextBug where
> class A a
> class B b
> instance A a => B b
> 
> ClassContextBug.hs:5:17:
>     Illegal instance declaration for `B b'
>       (All instance types must be of the form (T a1 ... an)
>        where a1 ... an are *distinct type variables*,
>        and each type variable appears at most once in the instance head.
>        Use -XFlexibleInstances if you want to disable this.)
>     In the instance declaration for `B b'
> 
> I believe this is valid Haskell 98, so GHC should accept it.

No, the instance head is just a type variable, not a type constructor 
applied to type variables, so it's not allowed by H98 (nor H2010, I think, 
haven't looked it up, but since GHC defaults to 2010 now, I'm pretty sure).
So GHC rightly complains about the malformed instance head. If you enable 
FlexibleInstances, it gives

ClassContextBug.hs:5:10:
    Ambiguous constraint `A a'
      At least one of the forall'd type variables mentioned by the 
constraint
      must be reachable from the type after the '=>'
    In the instance declaration for `B b'

ClassContextBug.hs:5:10:
    Variable occurs more often in a constraint than in the instance head
      in the constraint: A a
    (Use -XUndecidableInstances to permit this)
    In the instance declaration for `B b'

which is exactly what is wrong with the instance declaration.

> But even if not, what GHC is claiming in the error message
> is simply not true.
> 
> I would like to add this to the bug report.
> Alberto, have you filed the bug report yet? Can you please post a link?
> 
> Thanks,
> Yitz