A framework of an assertion-based algorithmic debugging for distributed programs

In this paper, we show a framework and an assertion-based prototype system of an algorithmic debugging for distributed programs written in procedural languages. First, an abstraction of previous related works is presented and we argue that an algorithmic debugging is a pruning process of a directed graph. Second, we present how to concrete the abstract method for distributed procedural languages. The method is divided into two phases. The purpose of the first phase is to reduce a range of erroneous program codes that cause a global state anomaly. We use only a global state in this phase. In the second phase, we locate a fault only using local stales of processes. Third, we describe an overview of our prototype system based on the framework. The system is based on replaying technique. We also show an automatic method for the first phase by using an assertion technique. Finally, the limitation of this approach and the future work is presented. The major contribution of this paper is as follows. First, separating the treatment of global and focal anomaly by dividing debugging processes into two phase enables to apply algorithmic debugging to distributed programs. Second, we can use appearable execution points instead of really appeared execution points to check whether the execution is correct or not. And third, we show an automatic operation of the first phase to avoid troublesome global state checking and to aim the range of codes which may contains fault.