further fix for floating point primitives