We segfaulted on attached testcase, because we were accessing
->typed.type field via TREE_TYPE, but IDENTIFIER_NODEs don't contain
->typed element at all. Fixed by switching the expressions in
a condition, so we always first check that we're operating on
INTEGER_CST.
Regtested/bootstrapped on x86_64-linux, ok for trunk and 4.7 branch?
2012-12-03 Marek Polacek <pola...@redhat.com>
PR c/55570
* c-common.c (check_user_alignment):
* gcc.dg/pr55570.c: New test.
--- gcc/c-family/c-common.c.mp 2012-12-03 13:07:58.233832299 +0100
+++ gcc/c-family/c-common.c 2012-12-03 13:08:55.395259841 +0100
@@ -7261,8 +7261,8 @@ check_user_alignment (const_tree align,
{
int i;
- if (!INTEGRAL_TYPE_P (TREE_TYPE (align))
- || TREE_CODE (align) != INTEGER_CST)
+ if (TREE_CODE (align) != INTEGER_CST
+ || !INTEGRAL_TYPE_P (TREE_TYPE (align)))
{
error ("requested alignment is not an integer constant");
return -1;
--- gcc/testsuite/gcc.dg/pr55570.c.mp 2012-12-03 13:06:02.298558986 +0100
+++ gcc/testsuite/gcc.dg/pr55570.c 2012-12-03 13:05:39.200504521 +0100
@@ -0,0 +1,4 @@
+/* PR c/55570 */
+/* { dg-do compile } */
+
+char array[16] __attribute__((aligned (SOME_NOT_DEFINED_MACRO))); /* {
dg-error "requested alignment is not an integer constant" } */
Marek
