Implicit conversion from float -9.458728010825439 to int loses precision