Build Optimization: A Systematic Literature Review

In modern software organizations, Continuous Integration (CI) consists of an automated build process triggered by change submissions and involving compilation, testing, and packaging to enable the continuous deployment of new software versions to end-users. While CI offers various advantages regarding software quality and delivery speed, it introduces challenges addressed by a large body of research. To better understand this literature, so as to help practitioners find solutions for their problems and guide future research, we conduct a systematic review of 97 studies published between 2006 and 2024, summarizing their goals, methodologies, datasets, and metrics. These studies target two main challenges: (1) long build durations and (2) build failures. To address the first, researchers have proposed techniques such as predicting build outcomes and durations, selective build execution, and build acceleration through caching or performance smell repair. On the other hand, build failure root causes have been studied, leading to techniques for predicting build script maintenance needs and automating repairs. Recent work also focuses on flaky build failures caused by environmental issues. Most techniques use machine learning and rely on build metrics, which we classify into five categories. Finally, we identify eight publicly available datasets to support future research on build optimization.