Re: Wrapped loads and Module::nesting

The patch below fixes Module.nesting so that it behaves in 1.9 like it 
does in 1.8. Note, however, that I don't really understand what I'm 
doing in eval.c, but I guessed and it seems to work. Of course, there 
may be some good reason for this change in Module.nesting that I'm 
unaware of...

	David

$ diff -u eval.c~ eval.c
--- eval.c~     2007-08-29 22:59:14.000000000 -0700
+++ eval.c      2007-09-13 15:01:41.000000000 -0700
@@ -361,6 +361,13 @@
         }
         cref = cref->nd_next;
      }
+
+    if (RARRAY(ary)->len == 0) {
+      rb_thread_t *th = GET_THREAD();
+      if (th->top_wrapper)
+       rb_ary_push(ary, th->top_wrapper);
+    }
+
      return ary;
  }


David Flanagan wrote:
> When I call load with a second argument of true, the file is loaded into 
> an anonymous module.  In Ruby 1.8 I can use Module::nesting[0] to obtain 
> this anonymous module.  It no longer works in Ruby 1.9.  I think Ruby 
> 1.9 ought to work like 1.8 because then a wrapped module can pass itself 
> back to the loader with code like this:
> 
> # This code is in the file to be loaded wrapped
> if Module.nesting.size > 0       # If we're loaded into a wrapper module
>   $wrapper = Module.nesting[0]   # Pass the module back to the loader
> end
> 
> module Foo
>   class Bar
>     def b
>       "Foo::Bar.b"
>     end
>   end
> end
> 
> # This code is the file that does the wrapped load:
> # Use include to safely allow the wrapped file to define new modules
> # without modifying existing ones
> load "wrapmod.rb", true
> include $wrapper if $wrapper
> puts Foo::Bar.new.b
> 
> Doing something like this is the only way that using wrapped loads makes 
> sense to me. If we can't do this in 1.9 then wrapped loads no longer 
> seem useful.
> 
> Are there security implications I'm missing here?
> 
>     David
>