%X => %_X, %_X => %__X

src/parser.cc

Index: src/parser.cc
diff --git a/src/parser.cc b/src/parser.cc
index f6f84622e42e49386c8ad7bd8f6068ae1b1b1b3d..c444b97d3153d520bf0525a580c6f241bb84854a 100644
--- a/src/parser.cc
+++ b/src/parser.cc
@@ -4643,11 +4643,16 @@ Expression* Parser::ParseV8Intrinsic(bool* ok) {
     return NULL;
   }
 
-  // Check that the function is defined if it's an inline runtime call.
-  if (function == NULL && name->Get(0) == '_') {
-    ReportMessage("not_defined", Vector<Handle<String> >(&name, 1));
-    *ok = false;
-    return NULL;
+  if (function == NULL && name->length() > 0 && name->Get(0) == '_') {
+    // Check that the function is defined if it's an inline runtime call.
+    if (name->length() > 1 && name->Get(1) == '_') {
+      ReportMessage("not_defined", Vector<Handle<String> >(&name, 1));
+      *ok = false;
+      return NULL;
+    }
+    // Assume this is a runtime-js function called via %_name.
+    name = isolate()->factory()->NewSubString(name, 1, name->length());

[Michael Starzinger, 2013/01/31 15:04:53]

This essentially merge the "%foo" and "%_foo" namespace at the moment, which is
probably not what we want. Instead the "%foo" namespace should be checked
against a white-list of allowed API-instrinsics. That white-list is empty for
now.

Also the AST node generated at the end of this function needs to know the
distinction between API-intrinsics and runtime-intrinsics, because otherwise
that distinction would be lost by the name mangling.

[drcarney, 2013/01/31 15:20:48]

Yes, I was planning this for the patch that actually introduces the public
intrinsics.

+    name = isolate()->factory()->LookupSymbol(name);
   }
 
   // We have a valid intrinsics call or a call to a builtin.