Implicit conversion from float -1.881028938906752 to int loses precision