Implicit conversion from float -32.128378378378386 to int loses precision