Implicit conversion from float -45.80000000000001 to int loses precision