Implicit conversion from float -120.00000000000003 to int loses precision