Implicit conversion from float -110.15625 to int loses precision