Implicit conversion from float -105.6359102244389 to int loses precision