[PATCH] D20985: [CUDA] Add implicit conversion of __launch_bounds__ arguments to rvalue.

[Justin Lebar via cfe-commits]

jlebar added inline comments.

================
Comment at: lib/Sema/SemaDeclAttr.cpp:4079
@@ +4078,3 @@
+  if (ValArg.isInvalid())
+    return nullptr;
+
----------------
OK, so then we want an assert, not an if?


http://reviews.llvm.org/D20985