Re: Possible bug in init_copy or rb_gc_copy_finalizer?

any word?

On Apr 15, 2004, at 21:09, Ryan Davis wrote:

> The following code:
>
> 	#!/usr/bin/ruby -w
>
> 	obj = Object.new
> 	ObjectSpace.define_finalizer(obj) { |an_id| puts "finalizing 
> #{an_id}" }
> 	obj2 = obj.clone
>
> causes the following output:
>
> 	./blah.rb:5: warning: copy_finalizer: descarding old finalizers
> 	finalizing 935212
> 	finalizing 935242
>
> Eric and I think this is the correct patch:
>
> <517> cvs diff -du object.c gc.c
> Index: object.c
> ===================================================================
> RCS file: /src/ruby/object.c,v
> retrieving revision 1.134.2.8
> diff -d -u -r1.134.2.8 object.c
> --- object.c    14 Apr 2004 04:06:52 -0000      1.134.2.8
> +++ object.c    16 Apr 2004 04:03:46 -0000
> @@ -263,7 +263,7 @@
>      }
>      clone = rb_obj_alloc(rb_obj_class(obj));
>      RBASIC(clone)->klass = rb_singleton_class_clone(obj);
> -    RBASIC(clone)->flags = (RBASIC(obj)->flags | FL_TEST(clone, 
> FL_TAINT)) & ~FL_FREEZE;
> +    RBASIC(clone)->flags = (RBASIC(obj)->flags | FL_TEST(clone, 
> FL_TAINT)) & ~(FL_FREEZE|FL_FINALIZE);
>      init_copy(clone, obj);
>      RBASIC(clone)->flags |= RBASIC(obj)->flags & FL_FREEZE;
>
> Index: gc.c
> ===================================================================
> RCS file: /src/ruby/gc.c,v
> retrieving revision 1.168.2.1
> diff -d -u -r1.168.2.1 gc.c
> --- gc.c        12 Apr 2004 10:11:34 -0000      1.168.2.1
> +++ gc.c        16 Apr 2004 04:03:47 -0000
> @@ -1719,12 +1719,10 @@
>
>      if (!finalizer_table) return;
>      if (!FL_TEST(obj, FL_FINALIZE)) return;
> -    if (FL_TEST(dest, FL_FINALIZE)) {
> -       rb_warn("copy_finalizer: discarding old finalizers");
> -    }
>      if (st_lookup(finalizer_table, obj, &table)) {
>         st_insert(finalizer_table, dest, table);
>      }
> +    RBASIC(dest)->flags |= FL_FINALIZE;
>  }
>
>  static VALUE
>
> -----
>
> 1) We added FL_FINALIZE to the flags to strip from the cloned object,
>
> 2) We think that the removed FL_TEST isn't necessary because 
> rb_gc_copy_finalizer is only called from init_copy so far and we have 
> stripped FL_FINALIZE from the flags just before calling init_copy.
>
> 3) We set the finalize flag on the cloned object after we insert the 
> cloned object into the finalizer_table.
>
> Please let us know if this is sufficient.
>
> Thank you
>
>