Implicit conversion from float -8.457223001402525 to int loses precision