The ability of object-oriented metrics to predict change-proneness: a meta-analysis