Implicit conversion from float -4.580152671755741 to int loses precision