This is the mail archive of the
newlib@sourceware.org
mailing list for the newlib project.
Re: [patch] Fix array aliasing conftest
On Friday 17 April 2009 18:21:44 Jeff Johnston wrote:
> The AC_TRY_COMPILE in libm/configure.in has been translated to
> AC_TRY_COMMAND.
I just noticed the spaces at the acnewlib_cv_type_long_double which lead to a
failing conftest. Your change fixed that for libm - thanks.
The libc/configure.in has the same test that should be changed in the same
manner. Is there a way to prevent that redundant long double check for libc
and libm? The tiny patch is attached.
> -- Jeff J.
Ken
newlib/ChangeLog:
2009-04-09 Ken Werner <ken.werner@de.ibm.com>
* libc/configure.in: Change AC_TRY_COMPILE to AC_TRY_COMMAND to
avoid executable test.
* libc/configure: Regenerate.
Index: src/newlib/libc/configure.in
===================================================================
--- src.orig/newlib/libc/configure.in
+++ src/newlib/libc/configure.in
@@ -144,10 +144,20 @@ dnl Autoconf 2.59 doesn't support the AC
dnl AC_TYPE_LONG_DOUBLE
dnl AM_CONDITIONAL(HAVE_LONG_DOUBLE, test x"$ac_cv_type_long_double" = x"yes")
dnl we specify our own long double test.
-AC_CACHE_CHECK([Checking long double support], [acnewlib_cv_type_long_double],
- [AC_TRY_COMPILE([], [long double foo = 0.0L;],
- [acnewlib_cv_type_long_double = yes;],
- [acnewlib_cv_type_long_double = no;])])
+AC_CACHE_CHECK([Checking long double support], [acnewlib_cv_type_long_double],[dnl
+cat > conftest.c <<EOF
+int main() {
+long double x = 0.0L;
+return 0;
+}
+EOF
+if AC_TRY_COMMAND([${CC} $CFLAGS $CPPFLAGS -c conftest.c 1>&AS_MESSAGE_LOG_FD])
+then
+ acnewlib_cv_type_long_double=yes
+else
+ acnewlib_cv_type_long_double=no
+fi
+rm -f conftest*])
AM_CONDITIONAL(HAVE_LONG_DOUBLE, test x"$acnewlib_cv_type_long_double" = x"yes")
if test -n "${machine_dir}"; then