Implicit conversion from float -2.4965893587994543 to int loses precision