Implicit conversion from float -45.400000000000006 to int loses precision