Implicit conversion from float -2.7777777777777715 to int loses precision