Hi Simon,
> The code reads:
>
> x = 1.73205L;
> y = ldexpl (x, 0); ASSERT (y == x);
>
> Printing the values yield:
>
> x 1.732050 y -0.000000
Interesting...
> configure:63967: checking whether ldexpl works
> configure:64002: gcc -std=gnu99 -o conftest -g -O2 -I/opt/local/include
> conftest.c >&5
> configure:64006: $? = 0
> configure:64012: ./conftest
> configure:64016: $? = 0
> configure:64034: result: yes
This patch should make the test fail. Committed.
Bruno
--- m4/ldexpl.m4.orig 2008-12-09 12:56:13.000000000 +0100
+++ m4/ldexpl.m4 2008-12-09 12:55:03.000000000 +0100
@@ -1,5 +1,5 @@
-# ldexpl.m4 serial 2
-dnl Copyright (C) 2007 Free Software Foundation, Inc.
+# ldexpl.m4 serial 3
+dnl Copyright (C) 2007-2008 Free Software Foundation, Inc.
dnl This file is free software; the Free Software Foundation
dnl gives unlimited permission to copy and/or distribute it,
dnl with or without modifications, as long as this notice is preserved.
@@ -60,7 +60,8 @@
AC_SUBST([LDEXPL_LIBM])
])
-dnl Test whether ldexpl() works on finite numbers (this fails on AIX 5.1).
+dnl Test whether ldexpl() works on finite numbers (this fails on AIX 5.1
+dnl and MacOS X 10.4/PowerPC).
AC_DEFUN([gl_FUNC_LDEXPL_WORKS],
[
AC_REQUIRE([AC_PROG_CC])
@@ -72,9 +73,11 @@
extern long double ldexpl (long double, int);
int main()
{
- volatile long double x = 1.0;
- volatile long double y = ldexpl (x, -1);
- return (y != 0.5L);
+ volatile long double x1 = 1.0;
+ volatile long double y1 = ldexpl (x1, -1);
+ volatile long double x2 = 1.73205L;
+ volatile long double y2 = ldexpl (x2, 0);
+ return (y1 != 0.5L) || (y2 != x2);
}], [gl_cv_func_ldexpl_works=yes], [gl_cv_func_ldexpl_works=no],
[case "$host_os" in
aix*) gl_cv_func_ldexpl_works="guessing no";;
2008-12-09 Bruno Haible <[EMAIL PROTECTED]>
* m4/ldexpl.m4 (gl_FUNC_LDEXPL_WORKS): Add another check, that fails on
MacOS X 10.4/PowerPC.
Reported by Simon Josefsson.
