Characterising Faults in Aspect-Oriented Programs: Towards Filling the Gap Between Theory and Practice

Since the proposal of Aspect-Oriented Programming, several candidate fault taxonomies for aspect-oriented (AO) software have been proposed. Such taxonomies, however, generally rely on language features, hence still requiring practical evaluation based on realistic implementation scenarios. The current lack of available AO systems for evaluation as well as historical data are the two major obstacles for this kind of study. This paper quantifies, documents and classifies faults uncovered in several releases of three AO systems, all from different application domains. Our empirical analysis naturally led us to revisit and refine a previously defined fault taxonomy. We identified particular fault types that stood out amongst the categories defined in the taxonomy. Besides this, we illustrate recurring faulty scenarios extracted from the analysed systems. We believe such scenarios should be considered for the establishment of testing strategies along the software development process.