Implicit conversion from float -26.76646706586827 to int loses precision