Change CodeGenModule GlobalDeclMap to directly reference globals
instead of mapping the decl to a bitcast of the global to the correct
type.
- GetAddrOf{Function,GlobalVar} introduce the bitcast on every use now.
- This solves a problem where a dangling pointer could be introduced
by the RAUW done when replacing a forward or tentative
definition. See testcase for more details.
- Fixes <rdar://problem/6108358>
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@54211 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/test/CodeGen/2008-07-30-redef-of-bitcasted-decl.c b/test/CodeGen/2008-07-30-redef-of-bitcasted-decl.c
new file mode 100644
index 0000000..f7c2908
--- /dev/null
+++ b/test/CodeGen/2008-07-30-redef-of-bitcasted-decl.c
@@ -0,0 +1,28 @@
+// RUN: clang --emit-llvm -o - %s
+// <rdar://problem/6108358>
+
+/* For posterity, the issue here begins initial "char []" decl for
+ * s. This is a tentative definition and so a global was being
+ * emitted, however the mapping in GlobalDeclMap referred to a bitcast
+ * of this global.
+ *
+ * The problem was that later when the correct definition for s is
+ * emitted we were doing a RAUW on the old global which was destroying
+ * the bitcast in the GlobalDeclMap (since it cannot be replaced
+ * properly), leaving a dangling pointer.
+ *
+ * The purpose of bar is just to trigger a use of the old decl
+ * sometime after the dangling pointer has been introduced.
+ */
+
+char s[];
+
+static void bar(void *db) {
+ eek(s);
+}
+
+char s[5] = "hi";
+
+int foo() {
+ bar(0);
+}