PR 10392

	From Marcus Comstedt / Dan Kegel:
	* gcc.dg/pr10392-1.c: New test.


git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@75985 138bc75d-0d04-0410-961f-82ee72b054a4

1 files changed, 62 insertions, 0 deletions

diff --git a/gcc/testsuite/gcc.dg/pr10392-1.c b/gcc/testsuite/gcc.dg/pr10392-1.c
new file mode 100644
index 00000000000..c71ec4d34b2
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr10392-1.c
@@ -0,0 +1,62 @@
+/* PR optimization/10392
+ * Reporter: marcus@mc.pp.se
+ * Summary: [3.3/3.4 regression] [SH] optimizer generates faulty array indexing
+ * Description:
+ * The address calculation of an index operation on an array on the stack 
+ * can _under some conditions_ get messed up completely
+ *
+ * Testcase tweaked by dank@kegel.com
+ * Problem only happens with -O2 -m4, so it should only happen on sh4,
+ * but what the heck, let's test other architectures, too.
+ * Not marked as xfail since it's a regression.
+*/
+/* { dg-do run } */
+/* { dg-options "-O2" } */
+/* { dg-options "-O2 -m4" { target sh4-*-* } } */
+const char *dont_optimize_function_away;
+
+const char *use(const char *str)
+{
+	dont_optimize_function_away = str;
+	if (str[0] != 'v')
+		abort();
+	if (str[1] < '1' || str[1] > '6')
+		abort();
+	if (str[2])
+		abort();
+	return str[2] ? "notused" : "v6";
+}
+
+const char *func(char *a, char *b)
+{
+	char buf[128];
+	unsigned char i;
+	const char *result;
+
+	char *item[] = {
+		"v1",
+		"v2",
+	};
+
+	buf[0] = 'v';
+	buf[1] = '3';
+	buf[2] = 0;
+
+	for (i = 0; i < 2; i++) {
+		/* bug is: following line passes wild pointer to use() on sh4 -O2 */
+		result = use(item[i]);
+
+		use(buf);
+		use(a);
+		use(b);
+		result = use(result);
+	}
+	return result;
+}
+
+int main()
+{
+	func("v4", "v5");
+	return 0;
+}
+