Building a sustainable codebase: 7 years and counting
summary
Seven years of work yields concrete lessons: the codebase ships 292 releases, the team grows to ~40, and risky cleverness causes bugs and slowdowns — automatic runtime scanning broke with a customer’s 50,000 classes so it is replaced by a simple register class call; an ARC conversion bug takes a week to debug; an attempted TIFF renderer costs two months and gets dropped; tests and CI (Jenkins, unit/UI tests, nightly builds) become mandatory; tooling wins when a clang-format 29-line patch fixes formatting. Boring is good, refactor constantly.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 0:10Speaker background
- 1:18Lessons from early days
- 9:03Refactoring strategies
- 12:00Technical debt
- 13:44Handling hacks
- 16:54Code reviews
- 17:55Tooling and consistency
- 21:07CI and testing
- 21:48Managing dependencies
- 25:02Conclusion and Q&A
related talks
47mAdventures with C++ Legacy Codebases: Tales of Incremental Improvement
27mMobiconf 2018: Carola Nitz, VideoLabs
41mHow to write maintainable code
22mThrough the (Open Source) Looking Glass
35mQuickly and Safely Refactoring Legacy Code
30mOptimizing for Developer Delight
01h 01mGuide to Destroying Codebases the Demise of Clever Code
43mCode quality as an engine of your IT business