memcmp and cross-compilation (was: Re: memcmp-tests module)

[Simon Josefsson]

Simon Josefsson <address@hidden> writes:

> Simon Josefsson <address@hidden> writes:
>
>> MinGW apparently needs gnulib's memcmp module
>
> That was rather surprising to me, so I started investigating why that is
> the case.

The reason was that Autoconf's AC_FUNC_MEMCMP uses AC_RUN_IFELSE with a
cross-compile default of no.  I prefer to assume memcmp exists on
cross-compile targets since I assume C89 or later in my projects.  That
is probably not the right choice for autoconf, but I think it is for
gnulib.  Thus it seems gnulib needs to extend autoconf's test somewhat?
I'll suggest something for review.

autoconf lib/autoconf/functions.m4:

# AC_FUNC_MEMCMP
# --------------
AC_DEFUN([AC_FUNC_MEMCMP],
[AC_CACHE_CHECK([for working memcmp], ac_cv_func_memcmp_working,
[AC_RUN_IFELSE([AC_LANG_PROGRAM([AC_INCLUDES_DEFAULT], [[
  /* Some versions of memcmp are not 8-bit clean.  */
  char c0 = '\100', c1 = '\200', c2 = '\201';
  if (memcmp(&c0, &c2, 1) >= 0 || memcmp(&c1, &c2, 1) >= 0)
    return 1;

  /* The Next x86 OpenStep bug shows up only when comparing 16 bytes
     or more and with at least one buffer not starting on a 4-byte boundary.
     William Lewis provided this test program.   */
  {
    char foo[21];
    char bar[21];
    int i;
    for (i = 0; i < 4; i++)
      {
        char *a = foo + i;
        char *b = bar + i;
        strcpy (a, "--------01111111");
        strcpy (b, "--------10000000");
        if (memcmp (a, b, 16) >= 0)
          return 1;
      }
    return 0;
  }
]])],
               [ac_cv_func_memcmp_working=yes],
               [ac_cv_func_memcmp_working=no],
               [ac_cv_func_memcmp_working=no])])
test $ac_cv_func_memcmp_working = no && AC_LIBOBJ([memcmp])
])# AC_FUNC_MEMCMP

/Simon