Implicit conversion from float -32.37623762376238 to int loses precision