fix: use UTF-16 wire format for @utf8InCpp strings in codegen

The @utf8InCpp annotation only affects the C++ in-memory representation
(std::string vs android::String16). The wire format is always UTF-16
because the C++ binder backend uses writeUtf8AsUtf16/readUtf8FromUtf16.

Remove the special-case in marshalForType that mapped @utf8InCpp String
to WriteString/ReadString (UTF-8), and fix AttributionSourceState to use
WriteString16/ReadString16 accordingly.

Author: xaionaro@dx.center

android/content/attributionsourcestate.go

@@ -99,13 +99,11 @@ func MarshalForTypeWithRegistry(
 		return MarshalInfo{}
 	}
 
-	// Check for @utf8InCpp annotation on String type.
-	if ts.Name == "String" && hasAnnotation(ts.Annots, "utf8InCpp") {
-		return MarshalInfo{
-			WriteExpr: "_data.WriteString(%s)",
-			ReadExpr:  "_reply.ReadString()",
-		}
-	}
+	// The @utf8InCpp annotation only affects the in-memory representation
+	// in C++ (std::string vs android::String16). The wire format is always
+	// UTF-16 because the C++ binder backend uses writeUtf8AsUtf16 /
+	// readUtf8FromUtf16 for these fields. We therefore do NOT special-case
+	// @utf8InCpp here -- all Strings use WriteString16/ReadString16.
 
 	if info, ok := marshalPrimitiveMap[ts.Name]; ok {
 		return info

tools/pkg/codegen/marshal.go

@@ -90,13 +90,16 @@ func TestMarshalForType_Primitives(t *testing.T) {
 }
 
 func TestMarshalForType_Utf8String(t *testing.T) {
+	// @utf8InCpp only affects the C++ in-memory representation; the wire
+	// format is always UTF-16. Verify that @utf8InCpp String produces the
+	// same marshal info as a plain String.
 	ts := &parser.TypeSpecifier{
 		Name:   "String",
 		Annots: []*parser.Annotation{{Name: "utf8InCpp"}},
 	}
 	info := MarshalForType(ts)
-	assert.Equal(t, "_data.WriteString(%s)", info.WriteExpr)
-	assert.Equal(t, "_reply.ReadString()", info.ReadExpr)
+	assert.Equal(t, "_data.WriteString16(%s)", info.WriteExpr)
+	assert.Equal(t, "_reply.ReadString16()", info.ReadExpr)
 	assert.False(t, info.NeedsCast)
 }