Implicit conversion from float -4.054054054054049 to int loses precision