Implicit conversion from float -7.283531409168077 to int loses precision