Data Race Detection Based on Dependence Analysis

Concurrency issues have been a serious problem in the whole software industry. They very often lead to hard-to-reproduce bugs and thus are very difficult to locate and solve. In this paper, we introduce a novel method for detection of one major category of concurrency problems, data race. This method extracts a Variable Dependence Tree from source code and finds data race by analysis on the Variable Dependence Tree. A prototype is built in Java according to the proposed method and tested on production-level source code (C language) with more than 400k LLOC. As a result, the prototype tool successfully detected all known data races including one relying on redundant structure of software and proved the effectiveness and feasibility of our proposed method.