NodeRock: uma abordagem para seleção de testes suscetíveis a event races em projetos Node.js

Abstract

The Node.js platform is widely used for JavaScript applications, employing an asynchronous execution model that is essential for its performance. However, the non-determinism of this model can lead to event races, which are subtle and hard-to-reproduce concurrency bugs that compromise software reliability. Current approaches to detect event races often require exhaustive analyses on the entire test suite, resulting in high computational costs and a lack of effective mechanisms to select and prioritize automated tests that are more susceptible to event races. This work proposes and evaluates an approach called NodeRock, which uses dynamic analysis to identify and select tests susceptible to event races in Node.js projects. NodeRock collects detailed execution traces, extracts a set of 15 dynamic features that characterize asynchronous behavior, and uses machine learning to select the tests. The results provide evidence that the approach identifies tests susceptible to event races with 75% accuracy and recall exceeding 84%, suggesting that the analysis of dynamic metrics is a promising strategy for prioritizing the investigation of these tests.