Implicit conversion from float -65.55555555555554 to int loses precision