Locating causes of inconsistency of variability models for software product line development

Abstract

The main difference between Software Product Line Development (SPLD) and traditional Software Product Development (SPD) is that development artifacts of SPLD contain variation points to deal with variabilities of a product family. By applying the SPLD approach, we can develop a product family more efficiently than by applying the SPD approach to each product of the product family. It is because the SPLD approach allows us to reuse the common parts obtained from analyzing the commonalities and variabilities of a product family for product development. Therefore, both (1) variability modeling for a product family and (2) inconsistency analysis of the resulting variability model that ensures it has been modeled correctly are crucial activities for the success of SPLD. Existing SPLD studies proposed various methods for analysis of various aspects of correctness of a variability model. In particular, analyzing whether a variability model is inconsistent or not is considered the most important analysis perspective since it is impossible to configure products from such a model. However, existing SPLD methods that analyze whether there is an inconsistency in a variability model cannot identify the exact locations of the variability model that cause the inconsistency. To overcome these limitations of the existing SPLD studies, this thesis proposes an Inconsistency Cause Locating (ICL) method. This method analyzes whether there is inconsistency in a variability model and, if there is inconsistency, it locates all the causes of the inconsistency in the model. In particular, the proposed method locates the exact locations of the nodes and links of the variability model that cause the inconsistency. As a consequence, developers using the method can easily find and fix inconsistency regardless of the size of a variability model. The key idea behind the ICL method is that, when transforming the given variability model to a CNF formula, by assigning to each literal of the formula a unique index that corresponds to each node of the variability model, the nodes that cause inconsistency, if the model is inconsistent, can be traced back once conflicting literal clauses are detected in the CNF formula. In this thesis, the ICL method is presented in the context of the variability models created by the Compositional Variability Modeling (CVM) method. This is because CVM method is a variability modeling method that can be covered throughout the entire SPLD process by solving and improving various problems of the traditional feature models. Specifically, the CVM method can systematically handle variability of a product family that occurs throughout the entire SPLD process by introducing notions of variation point and variant and clearly distinguishing between features and variability in a variability model. In addition, the CVM method can be used throughout the entire SPLD process because it is possible to construct a traceability relationship between a variability model and a development artifact and to provide compositional modeling mechanism to handle complexity and scalability of variability models. ICL method can also be applied to various variability models, including traditional feature models. To evaluate the ICL method, the research develops an ICL tool that automatically performs all the steps of the method and uses it to conduct experiments with 49 models, including real-world variability models. As a result, the ICL tool accurately identifies all models with an inconsistency and locates all causes of inconsistency in them.