Implicit conversion from float -109.21875 to int loses precision