Re: Question about gawk unofficial patch

[Aharon Robbins]

> Hello, My name is Kimura Koichi. Gawk user in Japan.

Nice to meet you.

> In patch,
>
> @@ -279,6 +287,15 @@
>                 emalloc(r->stptr, char *, r->stlen + 2, "dupnode");
>                 memcpy(r->stptr, n->stptr, r->stlen);
>                 r->stptr[r->stlen] = '\0';
> +#if defined MBS_SUPPORT
> +               if ((n->flags & WSTRCUR) != 0) {
> +                       r->wstlen = n->wstlen;
> +                       emalloc(r->wstptr, wchar_t *, sizeof(wchar_t) * 
> (r->wstlen + 2), "dupnode");
> +                       memcpy(r->wstptr, n->wstptr, r->wstlen);
> +                       r->wstptr[r->wstlen] = L'\0';
> +                       r->flags |= WSTRCUR;
> +               }
> +#endif /* defined MBS_SUPPORT */
>
> malloced memory sizeof(wchar_t) * (r->wstlen + 2) bytes,
> but just only copied r->wstlen bytes. I guess this is not enough.

We need r->wstlen + 1 elements, note the assigment of a L'\0'
after the memcpy.  The + 2 provides a little extra breathing room;
it's been a convention in gawk for ages, before I took over full
time maintanance even.  I've decided to continue using this convention
even though in theory it's enough to just use length + 1 everywhere.

> And in node.c,
>
>       getnode(r);
>       *r = *n;
>       r->flags &= ~(PERM|TEMP|FIELD);
>       r->flags |= MALLOC;
>         r->wstptr = NULL; //###
>
> If ### line is nothing, some script cause corrupted print out.
> (If you could use Japanese environment, I can show you data for
> reproduce)

Good point.  I've fixed this, although surrounded with the
appropriate ifdef.

Thanks,

Arnold