Implicit conversion from float -1.3446676970633717 to int loses precision