Implicit conversion from float -34.19354838709678 to int loses precision