Implicit conversion from float -127.17171717171718 to int loses precision