Implicit conversion from float -1.0067114093959901 to int loses precision