Implicit conversion from float -6.96652719665272 to int loses precision