Implicit conversion from float -4.444444444444443 to int loses precision