Implicit conversion from float -23.765541740674948 to int loses precision