Implicit conversion from float -34.10714285714285 to int loses precision