Re: yet another cpplib bug fix

[Dave Brolley <brolley at cygnus dot com>]

Pending your answer to my question about the @ and white space markers on your
previous patch, this looks OK to me.

Dave

Zack Weinberg wrote:

> This patch fixes the following bug:
>
> #define P(a,b) P1(a,b)
> #define P1(a,b) a##b
>
> #define ONCE(x, y) (x ?: x = y())
> #define PREFIX
>
> ONCE(memo, P(PREFIX, init))
>
> preprocesses to
>
> (memo ?: memo = @nit())
>
> The cause is an off by one error in macroexpand.  (N.B.: This patch
> must be applied over the other two cpplib bugfixes I've posted.)
>
> See egcs-bugs for a full testcase.
>
> zw
>
> 1998-09-27 16:31 -0400  Zack Weinberg  <zack@rabi.phys.columbia.edu>
>
>         * cpplib.c (macroexpand): Correct off-by-one in reverse
>         scanning of @ escapes.
>
> --- cpplib.c.bug3       Sun Sep 27 14:19:34 1998
> +++ cpplib.c    Sun Sep 27 16:26:07 1998
> @@ -2888,8 +2888,8 @@
>                           /* If whitespace is preceded by an odd number
>                              of `@' signs, the last `@' was a whitespace
>                              marker; drop it too. */
> -                         while (p2 != p1 && p2[-1] == '@') p2--;
> -                         if ((l1 - 1 - p2) & 1)
> +                         while (p2 != p1 && p2[0] == '@') p2--;
> +                         if ((l1 - p2) & 1)
>                             l1--;
>                           break;
>                         }
> @@ -2899,8 +2899,8 @@
>                           /* If a `-' is preceded by an odd number of
>                              `@' signs then it and the last `@' are
>                              a no-reexpansion marker.  */
> -                         while (p2 != p1 && p2[-1] == '@') p2--;
> -                         if ((l1 - 1 - p2) & 1)
> +                         while (p2 != p1 && p2[0] == '@') p2--;
> +                         if ((l1 - p2) & 1)
>                             l1 -= 2;
>                           else
>                             break;