Towards a Better Understanding of Feature Dependencies in Preprocessor-based Systems

Preprocessors are a common mechanism to implement features. In this context, features usually share elements, like variables and methods, which leads to code level feature dependencies like when a feature declares a variable used by another feature. Thus, developers are susceptible to maintain a feature and cause problems to another. To better understand feature dependencies, in this paper we complement previous work by providing an empirical study about them where we investigate if there are correlations with software complexity. In addition, we investigate where they occur, i.e., in which kinds of preprocessor directives we can find them. In particular, we answer the following research questions: How does program size influence feature dependencies? How does the number of features influences feature dependencies? How does the number of different kinds of preprocessor directives influence feature dependencies? How often do feature dependencies occur for each kind of preprocessor directive? To answer them, we analyze 45 preprocessor-based systems, software families, and product lines of different domains, size, and languages. The data we collect complement previous work and show that (i) there are correlations with complexity and (ii) we can find feature dependencies in many kinds of preprocessor directives.