On Wed, Apr 14, 2010 at 11:23:26PM +0100, Alex Francis wrote:
> Hi, I have a bit of a puzzle on my hands and I'm looking for some guidance.
>
> Say you have a Moose base class Thing with various subclasses. Through
> an oversight, one of these subclasses ('BadThing') isn't declared as a
> Moose class - it just has the base class in its @ISA. (Sure, this is a
> bug, but imagine it's one you didn't know was there).
>
> BadThing doesn't define a constructor of its own. When I call
> BadThing->new, I inherit 'new' from Moose::Object via Thing.
>
> Moose::Object::new does this:
>
> my $real_class = Scalar::Util::blessed($class) || $class;
> my $self =
> Class::MOP::Class->initialize($real_class)->new_object($params);
>
> at which point Class::MOP::Class spots that BadThing has no metaclass
> and creates one for it - a Class::MOP::Class, not a
> Moose::Meta::Class.
>
> My instantiation of BadThing will probably succeed, because
> Class::MOP::Class knows about quite a lot of the features that
> Moose::Meta::Class knows about. But.. if I use a Moose-specific
> feature, such as 'trigger', nothing will happen. I'll get no warning.
> I'll just be stuck wondering why my trigger didn't fire.
>
> So my question is, should Moose::Object::new complain if it tries to
> instantiate a class whose metaclass is a non-Moose one?
The actual bug here is that Class::MOP::Class->initialize shouldn't
initialize a CMOP metaclass for a class with a Moose parent (feel free
to read up on metaclass compatibility if you're interested in the
details). I have this partially fixed in a branch
(topic/more_metaclass_compat), but there are still some details that
need to be worked out... hopefully I'll get it fixed up and released
soon.
-doy
