Implicit conversion from float -1.743725231175688 to int loses precision