Re: possible bug: stack dump with <<-String, #{...} and large loops

Hi,

At Mon, 27 May 2002 19:38:22 +0900,
ts wrote:
> n>   begin "#{a=1; raise}"; rescue; end
> n>   p "#{a}"
> 
>      begin "#{a = 1:}"; rescue SyntaxError; end
>      eval "a = 12"

Oops, RVarmap's in compilation time should be swept away at a
compile error.  Also the following eval's should err either.

  begin "#{a = b = 1:}"; rescue SyntaxError; end
  eval "p a"
  eval "p b"

> n> [ruby-core:00047] disposes only a dvar header
> 
>  `a' is a dynamic variable, no ?

I meant RVarmap whose id is 0 by "a dvar header", it's placed
at the top of ruby_dyna_vars list.  Other dvars are inserted
after it by dvar_asgn_curr().


Index: eval.c
===================================================================
RCS file: /cvs/ruby/src/ruby/eval.c,v
retrieving revision 1.295
diff -u -2 -p -r1.295 eval.c
--- eval.c	2002/05/21 05:39:18	1.295
+++ eval.c	2002/05/27 11:24:44
@@ -3061,6 +3061,8 @@ rb_eval(self, n)
 	    VALUE str, str2;
 	    NODE *list = node->nd_next;
+	    struct RVarmap *vars = ruby_dyna_vars;
 
 	    str = rb_str_new3(node->nd_lit);
+	    if (!ruby_dyna_vars) rb_dvar_push(0, 0);
 	    while (list) {
 		if (list->nd_head) {
@@ -3080,4 +3082,11 @@ rb_eval(self, n)
 			ruby_in_eval--;
 			if (ruby_nerrs > 0) {
+			    struct RVarmap *vp = ruby_dyna_vars;
+			    ruby_dyna_vars = vars;
+			    while (vp && vp != vars) {
+				struct RVarmap *tmp = vp;
+				vp = vp->next;
+				rb_gc_force_recycle((VALUE)tmp);
+			    }
 			    compile_error("string expansion");
 			}


-- 
Nobu Nakada