Implicit conversion from float -25.968586387434556 to int loses precision