Implicit conversion from float -48.80000000000001 to int loses precision