Implicit conversion from float -23.400000000000006 to int loses precision