Fix TBAA information for reference accesses

[kosarev]

Summary

This patch fixes clang to decorate reference accesses as pointers and not as "omnipotent chars".

[rjmccall]

Could you flesh out this test case a bit? Please include at least:

• the initialization in B::B(S&)
• a load of the reference value (not the referent)

[kosarev]

Added a check for the initialization of the reference member in B::B(S&). (That in fact was my original intent; thanks for catching it.)

As to reference loads, we generate them with CodeGenFunction::EmitLoadOfReference() that, in contrast to CodeGenFunction::EmitLoadOfScalar(), is not TBAA-ready yet and looks to be a subject to a separate patch. So, what if we land D38126 first and then eventually re-consider the issue? Thanks again.

[kosarev]

Refined the TODO wording.

[rjmccall]

Sure, that's fine. LGTM.

lib/CodeGen/CodeGenTBAA.cpp

CodeGenTBAA::getTBAAInfo(QualType QTy) {
   }
 
   // C++1z [basic.lval]p10: "If a program attempts to access the stored value of
   // an object through a glvalue of other than one of the following types the
   // behavior is undefined: [...] a char, unsigned char, or std::byte type."
   if (Ty->isStdByteType())
     return MetadataCache[Ty] = getChar();
 
-  // Handle pointers.
+  // Handle pointers and references.
   // TODO: Implement C++'s type "similarity" and consider dis-"similar"
   // pointers distinct.
-  if (Ty->isPointerType())
+  if (Ty->isPointerType() || Ty->isReferenceType())
     return MetadataCache[Ty] = createTBAAScalarType("any pointer",
                                                     getChar());
 
   // Enum types are distinct types. In C++ they have "underlying types",
   // however they aren't related for TBAA.
   if (const EnumType *ETy = dyn_cast<EnumType>(Ty)) {
     // In C++ mode, types have linkage, so we can rely on the ODR and
     // on their mangled names, if they're external.

test/CodeGen/tbaa-reference.cpp

+// RUN: %clang_cc1 -triple x86_64-linux -O1 -disable-llvm-passes %s -emit-llvm -o - | FileCheck %s
+//
+// Check that we generate correct TBAA information for reference accesses.
+
+struct S;
+
+struct B {
+  S &s;
+  B(S &s) : s(s) {}
+  void bar();
+};
+
+void foo(S &s) {
+  B b(s);
+  b.bar();
+}
+
+// CHECK-LABEL: _Z3fooR1S
+// Check initialization of the reference parameter in foo().
+// CHECK: store %struct.S* {{.*}}, %struct.S** {{.*}}, !tbaa [[TAG_pointer:!.*]]
+//
+// CHECK-LABEL: _ZN1BC2ER1S
+// TODO: Check loading of the reference parameter in B::B(S&).
+// Check initialization of B::s in B::B(S&).
+// CHECK: store %struct.S* {{.*}}, %struct.S** {{.*}}, !tbaa [[TAG_pointer]]
+//
+// CHECK-DAG: [[TAG_pointer]] = !{[[TYPE_pointer:!.*]], [[TYPE_pointer]], i64 0}
+// CHECK-DAG: [[TYPE_pointer]] = !{!"any pointer", [[TYPE_char:!.*]], i64 0}
+// CHECK-DAG: [[TYPE_char]] = !{!"omnipotent char", {{!.*}}, i64 0}