Implicit conversion from float -12.305246422893475 to int loses precision