Reproducible Builds and sstate over CDN - Update on AGL's Infra and CIAT
summary
AGL runs reproducible builds so identical binaries come from the same sources and configuration, with Yocto 4.0 on by default and OE core reporting only about one or two out of ~3,000 packages still needing fixes. Binary sstate caches move to a CDN for artifacts under 64 MB, covering 99% of artifacts, and a hash equivalency server (HASHserv/PRServ) short-circuits rebuilds by recognizing identical binary hashes to save large amounts of rebuild time.
AI-written summary · about
chapters (14)
- 0:00Reproducible builds intro
- 1:49What are reproducible builds
- 3:17Why builds may fail
- 5:54Reproducible builds overview
- 7:58Benefits of reproducible builds
- 9:22BitBake workflow
- 10:28sstate cache usage
- 12:20Hash equivalent server
- 14:30Hash equivalency saves time
- 15:53sstate + hash equivalency
- 16:35sstate via CDN
- 17:39Scaling hash server
- 18:48Future sstate distribution
- 20:15References and closing
related talks
40mReproducible Builds: Unexpected Benefits and Problems
51mHow We Added Software Updates to AGL
42mThe Long Road to Reproducible Builds
56mReproducible Builds, the first ten years
37mAutomated Testing and Continuous Integration/Delivery: Lessons Learned
47mReproducible builds
42mSpeed-up bitbake with SSTATE, PRSERV and HashSERV
42mSecurity and Productivity Pick Two with Reproducible Builds