Implicit conversion from float -108.86877828054298 to int loses precision