Re: [Update] Port match to new dup, clone framework

Hi,

At Wed, 28 Aug 2002 23:42:53 +0900,
Michal Rokos wrote:
> 	This should fix it. :-) (I said should, but I know that nobu
> 	will find at least 1 bug in it... :-))

You mean this? ;-)

  /(.*)/=~"foo"; $~.become(Regexp.allocate)

will leak memory.

> @@ -528,40 +529,17 @@ match_alloc()
>  }
>  
>  static VALUE
> -match_clone(match)
> -    VALUE match;
> +match_become(obj, orig)
> +    VALUE obj, orig;
>  {
> -    NEWOBJ(clone, struct RMatch);
> -    CLONESETUP(clone, match);
> -
> -    clone->str = RMATCH(match)->str;
> -    clone->regs = 0;
> -
> -    clone->regs = ALLOC(struct re_registers);
> -    clone->regs->allocated = 0;
> -    re_copy_registers(clone->regs, RMATCH(match)->regs);
  +    if (obj == orig) return obj;
> +    RMATCH(obj)->str = RMATCH(orig)->str;
  +    re_free_registers(RMATCH(obj)->regs);
> +    RMATCH(obj)->regs->allocated = 0;
> +    re_copy_registers(RMATCH(obj)->regs, RMATCH(orig)->regs);

Or just leave allocated as is, without setting to 0.

-- 
Nobu Nakada