Test native x86-linux and cross to cris-elf.
Ok to commit?
-- >8 --
In r17-6022-gbcf14085d8270f, conditional calls were added for applying
__atomic_fetch_add to a uint64_t. This broke libgfortran for 32-bit
systems, that don't have such an operation and lack libatomic.
Better change the configure test to check a 64-bit type. Use
__INT64_TYPE__ to avoid having to find and include a usable stdint.h,
(which would require extra measures for e.g. newlib systems that
configure and build the library together with gcc)
PR libfortran/128618
* acinclude.m4 (LIBGFOR_CHECK_ATOMIC_FETCH_ADD): Check that
__atomic_fetch_add works on a 64-bit type, not just int.
* configure: Regenerate.
---
libgfortran/acinclude.m4 | 2 +-
libgfortran/configure | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/libgfortran/acinclude.m4 b/libgfortran/acinclude.m4
index 890516a6aedb..044514770b33 100644
--- a/libgfortran/acinclude.m4
+++ b/libgfortran/acinclude.m4
@@ -49,7 +49,7 @@ dnl Check whether the target supports __atomic_fetch_add.
AC_DEFUN([LIBGFOR_CHECK_ATOMIC_FETCH_ADD], [
AC_CACHE_CHECK([whether the target supports __atomic_fetch_add],
libgfor_cv_have_atomic_fetch_add, [
- AC_LINK_IFELSE([AC_LANG_PROGRAM([[int foovar = 0;]], [[
+ AC_LINK_IFELSE([AC_LANG_PROGRAM([[__INT64_TYPE__ foovar = 0;]], [[
if (foovar <= 0) return __atomic_fetch_add (&foovar, 1, __ATOMIC_ACQ_REL);
if (foovar > 10) return __atomic_add_fetch (&foovar, -1,
__ATOMIC_ACQ_REL);]])],
libgfor_cv_have_atomic_fetch_add=yes,
libgfor_cv_have_atomic_fetch_add=no)])
diff --git a/libgfortran/configure b/libgfortran/configure
index ed2c497c7298..3e3e68a46b7d 100755
--- a/libgfortran/configure
+++ b/libgfortran/configure
@@ -32288,7 +32288,7 @@ else
fi
cat confdefs.h - <<_ACEOF >conftest.$ac_ext
/* end confdefs.h. */
-int foovar = 0;
+__INT64_TYPE__ foovar = 0;
int
main ()
{
--
2.43.0