Implicit conversion from float -1.7406143344709903 to int loses precision