Implicit conversion from float -137.91519434628975 to int loses precision