Implicit conversion from float -90.40540540540539 to int loses precision