Anomaly agglomeration as sign of product line instabilities

Abstract

Software Product Line (SPL) is a set of systems that share features. Since the components of a SPL should be easy to maintain, developers have to detect code anomalies that harm the SPL maintainability. We have found several means to detect single code anomalies. Previous work assume that single anomalies suffice to characterize SPL maintenance problems, though an anomaly represents only a part of the problem. Thus, previous studies have difficulties in characterizing anomalies that indicate SPL maintenance problems. In this dissertation, we observe that anomalies may be interconnected, thereby forming so-called agglomerations. Two or more anomalies form an agglomeration in SPL when they affect the same SPL structural element. We then assess to what extent (non-)agglomerated anomalies represent sources of a specific problem: instability. Our findings suggest that feature hierarchy agglomeration indicates up to 89% of sources of instability, i.e., more than non-agglomerated anomalies.