A generalized, rule-based method for the detection of intermittent faults in software programs

Intermittent faults are a very common problem in the software world, and are very hard to locate and correct, due to the fact that they manifest their presence only under certain circumstances. Most of the existing approaches for intermittent fault identification assume that suitable instrumentation has been provided in the program, typically in the form of assertions that dictate which program states are considered to be erroneous, by examining variable values. This approach is, however, inefficient, since only errors for which appropriate instrumentation has been provided will be uncovered. In this paper we propose a method that can be used to detect probable sources of intermittent faults within a program. Our method proposes certain points in the code, whose data interdependencies combined with their execution interweaving indicate that they could be the root cause of intermittent faults. The approach presented in this paper extends our previous work, by examining shared variable access sequences and taking into account not only direct dependencies between shared variables, but additionally indirect ones, i.e. cases where values of local variables are computed using values of some shared variable(s), and subsequently the local variable values are used to set the value of other shared variables. The detection of suspicious access pattern, which may indicate the presence of intermittent faults, is formalized through the introduction of generalized rules; these rules are combined with model-based checking to ensure that all program execution paths are covered. The list of suspicious locations within the code is finally presented to the developer, who will decide after a thorough examination of the code, to accept or reject each of the proposals.