Building a Quality Assurance System for SDK Development: Methods, Tools, and Practices

The author, a test development engineer at Agora, shares the experience of establishing a robust quality system for SDK testing, emphasizing the need for end‑to‑end workflow tracking from requirement creation to release, data collection for retrospectives, and continuous improvement.

The methodology is divided into five stages: (1) "Small‑step rapid iteration" for fast releases with minimal test case maintenance; (2) "Steady‑state high quality" focusing on thorough testing across multiple dimensions (manual, automated, performance, lab); (3) "Documentation‑driven process" where accumulated test cases and metrics are formalized into clear, reusable documents; (4) "Diverse toolchain" showcasing eight categories of testing tools and the necessity of custom development for specific business needs; (5) "Refined development flow" highlighting code‑centric practices, second‑stage tool development (e.g., STF, Bugly, Sonic, HttpRunner) and the importance of high‑quality code.

Practical implementations include a real‑time closed‑loop handling pipeline for SDK crash data, a simplified circuit‑plug/unplug test setup that reduces manual effort dramatically, the HENGE cloud testing platform built on STF for automated test execution and reporting, and the open‑source Moonlight SDK for low‑overhead performance data collection.

The article concludes with a comprehensive quality‑system diagram covering requirement, development, code integration, testing, pre‑release, delivery, and post‑release monitoring, reinforcing that every testing activity and tool serves the ultimate goal of delivering high‑quality, reliable products.

In the Q&A section, the author explains coverage checking using the Bullseye tool within CI, the placement of CI/CD in the release pipeline for startups, and recommendations for end‑to‑end performance testing tools across Android, iOS, and Windows platforms.