Fix PR7040: report duplicate methods in a class, also for virtual methods

[alainfrisch]

See http://caml.inria.fr/mantis/view.php?id=7040

There is already an error that reports duplicated methods in a class:

# class c = object method x = 1 method x = 2 end;;
Characters 30-42:
  class c = object method x = 1 method x = 2 end;;
                                ^^^^^^^^^^^^
Error: The method `x' has multiple definitions in this object

This PR triggers the same error in case of duplication between two virtual methods or one virtual and one non-virtual method.

[gasche]

When you add a Changes entry, it should marked as "may break programs", because it is more than conceivable that people would have used virtual methods as prototypes to pre-declare their methods for readability purposes.

[alainfrisch]

because it is more than conceivable that people would have used virtual methods as prototypes to pre-declare their methods for readability purposes.

Good point. Would you be in favor of not adding this stricter check, then, if this use case is considered valid?

[gasche]

Indeed, I would think that this stricter check is not worth it: it breaks reasonable programs while it doesn't prevent any kind of dangerous situation, or help users in another way. The virtual-virtual case, on the other hand, seems completely useless and could deserve at least a warning, but you may also just not bother.

[garrigue]

Even two virtual methods could have some meaning: this forces unification between the two types.
So I agree that it is better no to try to guess what the user is doing.

[alainfrisch]

I'm convinced by these arguments. Let's do nothing here.