Implicit conversion from float -1.3138686131386805 to int loses precision