Implicit conversion from float -34.54954954954954 to int loses precision