Fixing regression bug 68139. Accessing radiobutton.value always returning the defalue value. r=harishd, sr=vidur, a=blizzard@mozilla.org/asa@mozilla.org

content/html/content/src/nsHTMLInputElement.cpp

@@ -427,19 +427,21 @@ nsHTMLInputElement::GetValue(nsAWritableString& aValue)
     }
       
     return NS_OK;
-  } else if (type == NS_FORM_INPUT_RADIO) {
-    // The value of a radio input is always "on". Why you ask? Well, I
-    // don't know but that's how it appears to work in older
-    // browsers. -- jst@netscape.com
+  }
+
+  // Treat value == defaultValue for other input elements
+  nsresult rv = nsGenericHTMLLeafFormElement::GetAttribute(kNameSpaceID_HTML,
+                                                           nsHTMLAtoms::value,
+                                                           aValue);
+
+  if (rv == NS_CONTENT_ATTR_NOT_THERE && type == NS_FORM_INPUT_RADIO) {
+    // The defauly value of a radio input is "on".
     aValue.Assign(NS_LITERAL_STRING("on"));
 
     return NS_OK;
   }
 
-  // Treat value == defaultValue for other input elements
-  return nsGenericHTMLLeafFormElement::GetAttribute(kNameSpaceID_HTML,
-                                                    nsHTMLAtoms::value,
-                                                    aValue);
+  return rv;
 }
 
 NS_IMETHODIMP