The Power of Backports : Business Impact and Automation Strategies
summary
Backporting currently runs mostly manual and slow: 80% take ~20 days, about 50% take >46 days and some take over a year, leaving old releases exposed. Automation using Fixmorph parses patches to AST, aligns divergent trees across streams and applies transformed patches, yielding a 75% success rate on kernel patches in tests. Real verification still requires build and QA (reproducers, regression suites). An in‑house tool PatchPal is under development to extend coverage.
AI-written summary · about
chapters (8)
- 0:00Backporting introduction
- 1:47Why backporting is required
- 3:33Manual backporting challenges
- 5:20Types of backporting
- 6:25Automating backporting
- 8:12AI and semantic tools
- 9:36Backporting strategies summary
- 10:40Q&A
related talks
26mImprove Upstream Code Quality by Bringing Testing to Patch Level
23mPrecise and Accurate Patch Presence Test for Binaries
30mSoftware Patching Needn't Be a Can of Worms
28mHyperscale Kernel Live Patching at Meta
36mStuck in 2009 - How I Survived
28mQuality in embargoed patches
32mReusing the BPF CI
33mLinux Kernel Live Patching