Implicit conversion from float -2.440191387559807 to int loses precision