Implicit conversion from float -4.212034383954148 to int loses precision