Re: Patch to avoid unnecessary uses of AC_TRY_RUN

[Olly Betts]

In message <[EMAIL PROTECTED]>, Earnie Boyd writes:
>--- Steve Robbins <[EMAIL PROTECTED]> wrote:
>> Has it always been the case that `0<5' is `1'?
>
>I think I remember cases (VAX?) of 0<5 produces -1.

Dear me.  Actually, BCPL's truth value is -1, so I can believe early
versions of C might behave the same way.

We can fix this by using Tom's negative array size trick here too.
Re-revised patch attached.

I'm unsure how `int a[0];' will be handled by various compilers.  The
ANSI standard (3.5.4.2 first paragraph) says:

  The expression delimited by [ and ] (which specifies the size of an
  array) shall be an integral constant expression that has a value
  greater than zero.

So an ANSI C compiler can reasonably reject it; gcc 2.95.2 happily
compiles it (arguably a gcc bug I guess - I can't see it being argued
to be a useful extension).  Anyway, to sidestep this issue I double
and add one.

Cheers,
Olly

===

diff -u -r1.222 acspecific.m4
--- acspecific.m4       2000/02/10 13:14:58     1.222
+++ acspecific.m4       2000/02/22 18:46:20
@@ -2395,17 +2395,8 @@
 #endif
 ], ac_cv_c_char_unsigned=yes, ac_cv_c_char_unsigned=no)
 else
-AC_TRY_RUN(
-[/* volatile prevents gcc2 from optimizing the test away on sparcs.  */
-#if !defined(__STDC__) || __STDC__ != 1
-# define volatile
-#endif
-int
-main()
-{
-  volatile char c = 255;
-  exit(c < 0);
-}], ac_cv_c_char_unsigned=yes, ac_cv_c_char_unsigned=no)
+AC_TRY_COMPILE(, [int a[(char)-1];],
+               ac_cv_c_char_unsigned=yes, ac_cv_c_char_unsigned=no)
 fi])
 if test $ac_cv_c_char_unsigned = yes && test "$GCC" != yes; then
   AC_DEFINE(__CHAR_UNSIGNED__)
@@ -2420,16 +2411,16 @@
 [if test "$GCC" = yes; then
   ac_cv_c_long_double=yes
 else
-AC_TRY_RUN(
-[int
-main()
-{
-  /* The Stardent Vistra knows sizeof(long double), but does not
+AC_TRY_COMPILE(,
+[/* The Stardent Vistra knows sizeof(long double), but does not
      support it.  */
   long double foo = 0.0;
-  /* On Ultrix 4.3 cc, long double is 4 and double is 8.  */
-  exit(sizeof(long double) < sizeof(double));
-}],
+  /* On Ultrix 4.3 cc, long double is 4 and double is 8.
+     We use `* 2 + 1' to preserve the sign of small positive and negative
+     values, while avoiding the ambiguous `int a[0];'
+     (gcc 2.95.2 allows it; ANSI C specifies size "greater than zero") */
+  int a[((int)sizeof(long double) - (int)sizeof(double)) * 2 + 1];],
+],
 ac_cv_c_long_double=yes, ac_cv_c_long_double=no)
 fi])
 if test $ac_cv_c_long_double = yes; then