Implicit conversion from float -39.054054054054035 to int loses precision