Focused Diagnosis for Failing Software Tests

Ranging from firmware to cloud services, software is an essential part of almost any imaginable system, or at the very least assists us in their design or maintenance. The sheer complexity and sophisticated concepts of today’s software products thus demand for solutions that assist us in assuring their quality. We aim to contribute in this direction by proposing a fault localization approach for failing test cases that draws on model-based diagnosis techniques from the AI community and focuses the search on dynamic executions. With this focus, we offer the scalability needed to consider also designs like service oriented architectures (SOAs). Furthermore, we opt for a flexible approach that allows a user to refine the reasoning by annotating our basic structure of a control flow graph with further information, e.g., for black box components. First experiments with standard software examples, as well as examples taken from the SOA domain show promising results.