ABSTRACT
In this paper we report on the results from the practical application of a method that was first presented at ESREL 2007 (Sarshar et al. 2007). The presented method identified failure modes that could cause error propagation through the usage of the system call interface of Linux. For each failure mode, its characteristics in code were determined so it could be detected when analyzing a given source code. Existing static analysis tools mainly detect language specific failures (Koenig 1988), buffer overflows, race conditions, security vulnerabilities and resource leaks, and they make use of a variety of analysis methods such as control and data flow analysis. An examination of some of the available tools showed that no existing tool can detect all the identified failure modes.
