Implicit conversion from float -6.916426512968286 to int loses precision