Implicit conversion from float -4.579646017699119 to int loses precision