Cost-Effective Regression Testing Using Bloom Filters in Continuous Integration Development Environments

Abstract

Regression testing in continuous integration development environments must be cost-effective and should provide fast feedback on test suite failures to the developers. In order to provide faster feedback on failures to developers while using computing resources efficiently, two types of regression testing techniques have been developed: Regression Testing Selection (RTS) and Test Case Prioritization (TCP). One of the factors that reduces the effectiveness of the RTS and TCP techniques is the inclusion of test suites that fail only once over a period. We propose an approach based on Bloom filtering to exclude such test suites during the RTS process, and to assign such test suites with a lower priority during the TCP process. We experimentally evaluate our approach using a Google dataset, and demonstrate that cost-effectiveness of the proposed RTS and TCP techniques outperforms the state-of-the-art techniques.