Implicit conversion from float -10.000000000000014 to int loses precision