Implicit conversion from float -139.78723404255317 to int loses precision