Extending C Global Surveyor

Software failure are noted for their blowing large sums of money and sometimes even human life, in particular in the area of safety critical mission software. The most well-known desaster happened in 1996 when Ariane 501 exploded shortly after launch. The least it did was to cost the European space program half a billion US$ due to an overflow in an arithmetic conversion. The Automated Software Engineering Group at the NASA Ames Research Center has developed C Global Surveyor (CGS), a static analysis tool based on abstract interpretation. It particularly concentrates on runtime errors that are hard to find during development such as out-of-bound array accesses, acesses to non-initialised variables, and de-references of null pointers. CGS proved to analyse large, pointer intensive and heavily multithreaded code (up to 280 KLoC) in a couple of hours with a constant precision of 80%. It is used to successfully analyse mission-critical flight software of NASA's "Mars Path-Finder" (MPF) and Deep Space 1 (DS1) legacy as well as software of the Mars Exploration Rover (MER) mission (650 KLoC) and other JPL-based missions. However, the abstract interpretation techniques on which CGS is based, need to be augmented by complimentary program analysis techniques in order to enhance CGS and support the developer when analysing very large systems. As a first step, we included the construction of control flow graphs that represent the programs to be analysed. It is a first step towards the application of more advanced techniques such as program slicing.