Consistent Feature-Model Driven Software Product Line Evolution

SPLs are an approach to manage families of closely related software systems in terms of configurable functionality. A feature model captures common and variable functionalities of an SPL on a conceptual level in terms of features. Reusable artifacts, such as code, documentation, or tests are related to features using a feature-artifact mapping. A product of an SPL can be derived by selecting features in a configuration. Over the course of time, SPLs and their artifacts are subject to change. As SPLs are particularly complex, their evolution is a challenging task. Consequently, SPL evolution must be thoroughly planned well in advance. However, plans typically do not turn out as expected and, thus, replanning is required. Feature models lean themselves for driving SPL evolution. However, replanning of feature-model evolution can lead to inconsistencies and feature-model anomalies may be introduced during evolution. Along with feature-model evolution, other SPL artifacts, especially configurations, need to consistently evolve. The work of this thesis provides remedy to the aforementioned challenges by presenting an approach for consistent evolution of SPLs. The main contributions of this thesis can be distinguished into three key areas: planning and replanning feature-model evolution, analyzing feature-model evolution, and consistent SPL artifact evolution. As a starting point for SPL evolution, we introduce Temporal Feature Models (TFMs) that allow capturing the entire evolution timeline of a feature model in one artifact, i.e., past history, present changes, and planned evolution steps. We provide an execution semantics of feature-model evolution operations that guarantees consistency of feature-model evolution timelines. To keep feature models free from anomalies, we introduce analyses to detect anomalies in feature-model evolution timelines and explain these anomalies in terms of their causing evolution operations. To enable consistent SPL artifact evolution, we generalize the concept of modeling evolution timelines in TFMs to be applicable for any modeling language. Moreover, we provide a methodology that enables involved engineers to define and use guidance for configuration evolution.