Re: autoconf-2.59 and AC_C_LONG_DOUBLE: comments and a proposal

[Paul Eggert]

"Nelson H. F. Beebe" <address@hidden> writes:

> This then results in run-time failures if [s]printf floating-point
> format items with the L type modifier, or long double library
> routines, like powl(), are used.

I'd like to distinguish between failures of the compiler and failures
of the run-time library.  How about if we have AC_C_LONG_DOUBLE stand
only for the compiler issues, and add new macros for the
runtime-library issues?  After all, the C compiler need not be a
hosted implementation.

I installed the following patch to autoconf that I hope resolves all
the issues that you mentioned for the compiler proper, plus a few
other issues that I thought of while I was at it.  The library issues
remain.

2004-06-02  Paul Eggert  <address@hidden>

        * lib/autoconf/c.m4 (AC_LONG_DOUBLE): Don't check LDBL_MAX
        and LDBL_EPSILON, as the resulting expression isn't an
        integer constant expression and violates the C standard.
        Problem reported by Nelson H. F. Beebe.  Also, check
        for "L" suffix, and check that long double doesn't have
        worse range or precision than double, that mixed-mode
        arithmetic doesn't generate a diagnostic, that double
        constants fit in long double.

--- c.m4.~1.187.~       2004-05-03 13:15:44 -0700
+++ c.m4        2004-06-02 16:38:23 -0700
@@ -948,11 +948,25 @@ AC_DEFUN([AC_C_LONG_DOUBLE],
    [ac_cv_c_long_double],
    [AC_COMPILE_IFELSE(
       [AC_LANG_BOOL_COMPILE_TRY(
-        [#include <float.h>
-         long double foo = 0.0;],
-        [/* Using '|' rather than '||' catches a GCC 2.95.2 x86 bug.  */
-         (DBL_MAX < LDBL_MAX) | (LDBL_EPSILON < DBL_EPSILON)
-         | (DBL_MAX_EXP < LDBL_MAX_EXP) | (DBL_MANT_DIG < LDBL_MANT_DIG)])],
+        [[#include <float.h>
+          long double const a[] =
+            {
+               0.0L, DBL_MIN, DBL_MAX, DBL_EPSILON,
+               LDBL_MIN, LDBL_MAX, LDBL_EPSILON
+            };
+          long double
+          f (long double x)
+          {
+             return ((x + (unsigned long int) 10) * (-1 / x) + a[0]
+                      + (x ? f (x) : 'c'));
+          }
+        ]],
+        [[(0 < ((DBL_MAX_EXP < LDBL_MAX_EXP)
+                 + (DBL_MANT_DIG < LDBL_MANT_DIG)
+                 - (LDBL_MAX_EXP < DBL_MAX_EXP)
+                 - (LDBL_MANT_DIG < DBL_MANT_DIG)))
+          && (int) LDBL_EPSILON == 0
+         ]])],
       ac_cv_c_long_double=yes,
       ac_cv_c_long_double=no)])
 if test $ac_cv_c_long_double = yes; then