We ICEd on the following testcase because even though the underlying
types were the same, theirs TREE_TYPEs were not. We can use
TYPE_MAIN_VARIANT to see through the typedefs which confused us.
Tested x86_64-pc-linux-gnu, applying to the ubsan branch.
2013-08-14 Marek Polacek <pola...@redhat.com>
* c-ubsan.c (ubsan_instrument_division): Use TYPE_MAIN_VARIANT
in the assert.
* c-c++-common/ubsan/typedef-1.c: New test.
--- gcc/c-family/c-ubsan.c.mp 2013-08-14 16:55:56.322784886 +0200
+++ gcc/c-family/c-ubsan.c 2013-08-14 16:56:03.716812775 +0200
@@ -42,8 +42,10 @@ ubsan_instrument_division (location_t lo
tree type = TREE_TYPE (op0);
/* At this point both operands should have the same type,
- because they are already converted to RESULT_TYPE. */
- gcc_assert (type == TREE_TYPE (op1));
+ because they are already converted to RESULT_TYPE.
+ Use TYPE_MAIN_VARIANT since typedefs can confuse us. */
+ gcc_assert (TYPE_MAIN_VARIANT (TREE_TYPE (op0))
+ == TYPE_MAIN_VARIANT (TREE_TYPE (op1)));
/* TODO: REAL_TYPE is not supported yet. */
if (TREE_CODE (type) != INTEGER_TYPE)
--- gcc/testsuite/c-c++-common/ubsan/typedef-1.c.mp 2013-08-14
16:58:05.767181047 +0200
+++ gcc/testsuite/c-c++-common/ubsan/typedef-1.c 2013-08-14
16:57:21.086022479 +0200
@@ -0,0 +1,12 @@
+/* { dg-do compile } */
+/* { dg-options "-fsanitize=undefined" } */
+
+typedef int V;
+int
+foo (void)
+{
+ V v = 9;
+ int a = 3;
+ v += v % a;
+ return v / 3;
+}
Marek
