Implicit conversion from float -1.7694369973190192 to int loses precision