Really, don't use the C api - it will break at some point, since it is
not intended in any way to be a public api.

Use a FFI callback to call from C to ruby.

http://wiki.github.com/ffi/ffi/examples has a callback example, as
does http://wiki.github.com/ffi/ffi/windows-examples



2009/12/9 Simon Chiang <simon.a.chiang@gmail.com>:

> I'm building a callback to ruby from an extension and I would like to
> pass back pointers much like an attached function that returns
> a :pointer.  Basically something like this:
>
>  int RubyCallback(void *thing)
>  {
>    VALUE ffi_pointer = rbffi_Pointer_NewInstance(thing);
>    VALUE ruby_module = rb_const_get(rb_cObject, rb_intern
> ("SomeModule"));
>    VALUE ruby_result = rb_funcall(ruby_module, rb_intern("callback"),
> 1, ffi_pointer);
>
>    long result = FIX2INT(ruby_result);
>    return(result);
>  }
>
> And then in ruby:
>
>  module SomeModule
>    module_function
>
>    def callback(ptr)
>       ptr.inspect         # => shows this to be an FFI::Pointer
>    end
>  end
>
> A method like this works on OS X with ffi-0.5.0 (pre snow-lepoard,
> I'll get the version if necessary) but results in a segfault on snow
> leopard 10.6.2.  The segfault is triggered upon inspect.  I don't know
> how FFI works... I really only got this working by taking a guess at
> what rbffi_Pointer_NewInstance does.  Any ideas what's going wrong on
> 10.6.2 and/or how I can do this better?
>