[Patch] : Fix configure check for have_attribute_alias
Danny Smith
dannysmith@clear.net.nz
Wed Aug 9 04:24:00 GMT 2006
Hello,
The have_attribute_alias check in libgomp/acinclude.m4 is wrong.
The aliased target does not want a __USER_LABEL_PREFIX__. If if it did,
all the testcase involving the attribute in the testsuite should fail on
mingw32 and any other target that defines a non-empty
__USER_LABEL_PREFIX__
Here is the (obvious?) fix. Is this OK for mainline or should it wait
until branching?
Danny
libgomp/ChangeLog
* acinclude.m4 (have_attribute_alias): Don't use
__USER_LABEL_PREFIX__.
* configure: Regenerate.
libgfortran/ChangeLog
* acinclude.m4 (have_attribute_alias): Don't use
__USER_LABEL_PREFIX__.
* configure: Regenerate.
This is the diff for libgomp. Same except for line numbers in
libgfortran
Index: acinclude.m4
===================================================================
--- acinclude.m4 (revision 116011)
+++ acinclude.m4 (working copy)
@@ -47,11 +47,10 @@
AC_CACHE_CHECK([whether the target supports symbol aliases],
have_attribute_alias, [
AC_TRY_LINK([
-#define ULP STR1(__USER_LABEL_PREFIX__)
#define STR1(x) STR2(x)
#define STR2(x) #x
void foo(void) { }
-extern void bar(void) __attribute__((alias(ULP "foo")));],
+extern void bar(void) __attribute__((alias("foo")));],
[bar();], have_attribute_alias=yes, have_attribute_alias=no)])
if test $have_attribute_alias = yes; then
AC_DEFINE(HAVE_ATTRIBUTE_ALIAS, 1,
More information about the Gcc-patches
mailing list