Implicit conversion from float -28.400000000000006 to int loses precision