Implicit conversion from float -12.451838879159368 to int loses precision