Implicit conversion from float -34.68354430379746 to int loses precision