If you eliminate the path where native code needs to access the class information for managed types then these issues can all be resolved with the `const std::type_info&` proposed. When serializing the lookup of type to classId should be happening in the managed layer anyway, so the table for type to classId in the unmanaged code should not be invoked at all. When deserializing we could hook earlier that we are actually managed code and divert to managed code for the type lookups. We do this already for PDX types.
[ Full content available at: https://github.com/apache/geode-native/pull/339 ] This message was relayed via gitbox.apache.org for [email protected]
