Implicit conversion from float -34.71910112359551 to int loses precision