Fault localization in ABAP Programs

In this paper we present an automated technique for localizing faults in data centric programs. Data-centric programs typically interact with databases to get collections of content, process each entry in the collection and output another collection or write entries back to database. Many of the production faults in data centric programs are manifested because of presence of certain data values or data patterns in the database that were not checked for or handled correctly during development of the initial code. Our technique collects the execution trace of the faulty program, uses a novel precise slicing algorithm to break the trace into slices mapping to each entry in the output collection and finally does a semantic difference between the slices corresponding to correct output entries versus incorrect ones to suggest potentially faulty statements. We implemented our approach for ABAP programs. ABAP is the language used to code in SAP systems and interacts heavily with databases. It also contains embedded SQL like commands to work on collections of data. We applied our technique to a suite of 13 ABAP programs with faults from the field and it was able to identify the precise cause in 12 cases.