Implicit conversion from float -139.43661971830983 to int loses precision