Re: Patch to projects.html: glibc's string function macros

["Kaveh R. Ghazi" <ghazi at caip dot rutgers dot edu>]

 > From: "Joseph S. Myers" <jsm28@cam.ac.uk>
 > 
 > On Mon, 30 Oct 2000, Kaveh R. Ghazi wrote:
 > 
 > > Done and done.  Since va_end isn't broken, I didn't bother with a
 > > testcase.
 > 
 > va_end was, however, broken in 2.95.2, which is a (weak) case for adding a
 > testcase for it anyway.

Alright, here's a testcase that I'd like to install as
execute/va-arg-21.c.  I've tested that it fails on hosts where va_end
is defined to "(void)0" like solaris2.7 cc and older gcc versions.

Ok to install?

		--Kaveh



/* Copyright (C) 2000  Free Software Foundation.

   If the argument to va_end() has side effects, test whether side
   effects from that argument are honored.

   Written by Kaveh R. Ghazi, 10/31/2000.  */

#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>

#ifndef __GNUC__
#define __attribute__(x)
#endif

static void __attribute__ ((__format__ (__printf__, 1, 2)))
doit (const char *s, ...) 
{
  va_list *ap_array[3], **ap_ptr = ap_array;

  ap_array[0] = malloc (sizeof(va_list));
  ap_array[1] = NULL;
  ap_array[2] = malloc (sizeof(va_list));
  
  va_start (*ap_array[0], s);
  vprintf (s, **ap_ptr);
  /* Increment the va_list pointer once.  */
  va_end (**ap_ptr++);

  /* Increment the va_list pointer a second time.  */
  ap_ptr++;
  
  va_start (*ap_array[2], s);
  /* If we failed to increment ap_ptr twice, then the parameter passed
     in here will dereference NULL and should cause a crash.  */
  vprintf (s, **ap_ptr);
  va_end (**ap_ptr);

  /* Just in case, If *ap_ptr is NULL abort anyway.  */
  if (*ap_ptr == 0)
    abort();
}

int main()
{
  doit ("%s", "hello world\n");
  return 0;
}