Analysis of Code Review Blind Spots in Open-Source Repositories

This report presents findings from a study analyzing code review blind spots across 90 major open-source repositories. The objective was to measure how frequently bugs introduced in pull requests are later fixed by subsequent pull requests. The study identified 1,258 proven bugs, with a median time to fix of 16 days and a mean time of 46 days. The analysis included repositories in various programming languages, such as Python, JavaScript, and TypeScript, and highlighted that bugs were present even in projects utilizing AI-powered code review tools. The findings categorize the bugs into correctness, reliability, observability, performance, and security issues, with 61% classified as critical. The report also discusses the structural gaps in code review processes, revealing that over half of the studied repositories scored an F in code health. The results indicate a significant need for improved code review practices to prevent the introduction of bugs in production code.