Implicit conversion from float -48.900000000000006 to int loses precision