Implicit conversion from float -17.297687861271683 to int loses precision