Implicit conversion from float -44.79729729729729 to int loses precision