CppCon 2018

Preventing defects to reach consumers is one of the most important task of any application developer. Using machine learning and pattern matching we attempt to automatically prevent defects from reaching the central code repository by intercepting them at commit-time. In this talk, we present an approach to detect, and resolve, defects in large C++ projects. Our approach, when evaluated on past contributions, was able to detect ~65% of the commits containing defects. In addition, this project allowed us to improve our C++ practices by discovering new guidelines that are applicable for our application domain, in our case AAA game development.