Race Condition Detection In Concurrent Erlang Applications Using Static Analysis

Abstract

In safety-critical and high-reliability concurrent systems, software development and maintenance require great effort. This effort could be significantly reduced if concurrency defects, among other software errors, were identified through automatic tools such as program analyzers and compile-time checkers. We therefore address the problem of finding some commonly occurring kinds of race conditions in Erlang programs using static analysis. Our analysis is completely automatic, fast and scalable, and avoids false alarms by taking language characteristics into account. We have integrated our analysis in a publicly available, commonly used tool for detecting software defects in Erlang programs and evaluate its effectiveness and performance on a suite of widely used industrial and open source programs of considerable size. The number of previously unknown race conditions that we have detected in them is significant.