Testcase for something that already works. Clang agrees with Comeau,

and this makes sense.  likely a gcc bug that it doesn't diagnose the bad
restrict.

llvm-svn: 60923

clang/test/Sema/types.c

@@ -0,0 +1,10 @@
+// RUN: clang %s -pedantic -verify
+
+// rdar://6097662
+typedef int (*T)[2];
+restrict T x;
+
+typedef int *S[2];
+restrict S y; // expected-error {{restrict requires a pointer or reference ('S' is invalid)}}
+
+