Implicit conversion from float -111.23456790123456 to int loses precision