Tracking the propagation of data errors in software

When developing software that functions in a nonharmful manner in the presence of faults and errors (as defined in Laprie [27]), one requires knowledge of the behavior of the software under these exceptional conditions. In particular, one needs to know how faults and errors propagate to affect the execution of software. Knowing propagation pathways may, for instance, be of great help when deciding where to effectively place error detection and recovery mechanisms. Learning about error propagation Characteristics of a software system requires not only the ability to inject errors and monitor the effect that these have on system output but also the ability to monitor how these errors are transported through the system. Thus, high observability is required for these activities. Ideally, one should be able to observe every individual variable and data structure in the software. Chapter 15 gives a detailed presentation of PROPANE, the Propagation Analysis Environment, and is an extension of the work by Hiller et al. [20]. PROPANE enables the injection of faults and errors into software running on a desktop computer (currently for Windows NT/2000). PROPANE supports varied ways of probing a system, i.e., tracing internal variables and events during system operation, as well as injection of software faults and data errors. FI has been used for various purposes for many years [1, 7, 22]. FI-tools may inject errors in a number of ways, e.g., physical fault injection, radiation, etc. PROPANE uses a technique called SWIFI (Software-Implemented Fault Injection). Fault injections are performed by instrumenting the source code with both the correct code and the defect that is to be injected. With every fault, there is a