Implicit conversion from float -32.009569377990445 to int loses precision