[semantic-sil] objc_{existential_,}metatype_to_object should be treated as owned.

Metatypes in SIL are trivial, until we conver them to objects (mainly for cases
where one converts the metatype to an object for use in objc). Thus these
conversion functions must be treated as returning an @owned value.

Author: gottesmm

lib/SIL/SILValue.cpp

@@ -188,6 +188,16 @@ CONSTANT_OWNERSHIP_INST(Owned, LoadWeak)
 CONSTANT_OWNERSHIP_INST(Owned, PartialApply)
 CONSTANT_OWNERSHIP_INST(Owned, StrongPin)
 CONSTANT_OWNERSHIP_INST(Owned, ThinToThickFunction)
+
+// One would think that these /should/ be unowned. In truth they are owned since
+// objc metatypes do not go through the retain/release fast path. In their
+// implementations of retain/release nothing happens, so this is safe.
+//
+// You could even have an optimization that just always removed retain/release
+// operations on these objects.
+CONSTANT_OWNERSHIP_INST(Owned, ObjCExistentialMetatypeToObject)
+CONSTANT_OWNERSHIP_INST(Owned, ObjCMetatypeToObject)
+
 // All addresses have trivial ownership. The values stored at the address may
 // not though.
 CONSTANT_OWNERSHIP_INST(Trivial, AddressToPointer)
@@ -214,9 +224,6 @@ CONSTANT_OWNERSHIP_INST(Trivial, MarkFunctionEscape)
 CONSTANT_OWNERSHIP_INST(Trivial, MarkUninitialized)
 CONSTANT_OWNERSHIP_INST(Trivial, MarkUninitializedBehavior)
 CONSTANT_OWNERSHIP_INST(Trivial, Metatype)
-CONSTANT_OWNERSHIP_INST(Trivial,
-                        ObjCExistentialMetatypeToObject) // Is this right?
-CONSTANT_OWNERSHIP_INST(Trivial, ObjCMetatypeToObject)   // Is this right?
 CONSTANT_OWNERSHIP_INST(Trivial, ObjCProtocol)           // Is this right?
 CONSTANT_OWNERSHIP_INST(Trivial, ObjCToThickMetatype)
 CONSTANT_OWNERSHIP_INST(Trivial, OpenExistentialAddr)