Implicit conversion from float -3.673469387755091 to int loses precision