ReadCallerDataNode should only skip RubyNode not RubyBaseNode

* It used to skip RubyContextNode before the previous commit.
  Before that, it used RubyBaseNode but that was before the DSL changed,
  which means data nodes for inline caches are just Nodes and not
  subclasses of the node using the DSL like before. Also, there is an
  intermediate CallInternalMethodNode since the new dispatch chain.

Author: eregon

src/main/java/org/truffleruby/language/arguments/ReadCallerDataNode.java

@@ -24,6 +24,7 @@
 import org.truffleruby.language.FrameAndVariablesSendingNode;
 import org.truffleruby.language.NotOptimizedWarningNode;
 import org.truffleruby.language.RubyBaseNode;
+import org.truffleruby.language.RubyNode;
 import org.truffleruby.language.control.RaiseException;
 
 public abstract class ReadCallerDataNode extends RubyBaseNode {
@@ -78,7 +79,8 @@ private boolean notifyCallerToSendData() {
                     startSending((FrameAndVariablesSendingNode) parent);
                     return true;
                 }
-                if (parent instanceof RubyBaseNode) {
+                if (parent instanceof RubyNode) {
+                    // A node with source info representing Ruby code, we could not find the FrameAndVariablesSendingNode
                     return false;
                 }
                 parent = parent.getParent();