Backporting is so 1993
summary
Argues that cherry‑picking fixes and long LTS forks don’t scale, shows concrete pain: 1900 commits per release in Zephyr that are impossible to backport cleanly, and real incidents like a meltdown backport bricked many systems. Recommends staying near upstream tip, running CI that runs sanitycheck on hardware and sample apps, taking small batches of commits, using containerized isolation for components, and coordinating via kernel‑CI / upstream testing so fixes land faster and regression bisection stays tractable.
AI-written summary · about
chapters (10)
- 0:00Start
- 2:11Traditional Embedded Approach
- 3:00Traditional Product Maintenance
- 4:14Embedded Linux Development
- 5:32Meltdown/Spectre?
- 6:35Stable Maintenance Is Not Easy
- 10:15And now for something not Linux The Zephyr Project
- 12:42Latest Software
- 12:59Working with Latest Zephyr
- 16:04Working with Latest: Linux
related talks
36mNo, It's Never Too Late to Upstream Your Legacy Linux Based Platform
36mNo, It's (Still) Never Too Late to Upstream Your Legacy Linux Based Platforms
40mContributing to Zephyr vs (Linux and U-boot)
36mNo, It's (Still) Never Too Late to Upstream Your Legacy Linux Based Platforms
39mThe Life of a Kernel Bugfix
17mSustaining Yesterday’s Linux: Delivering Security for Today’s Workloads
43mLinus Torvalds in conversation with Dirk Hohndel
21mCommunities, Enterprises and All the Rest
