On Thu, 6 Feb 2025 at 16:56, Alexander Kanavin <alex.kana...@gmail.com> wrote: > 4. How is sstate availability guaranteed? > > There would be an autobuilder job, part of a-full, that would use > bitbake-setup to set up a build matching the branch being tested, and > use that setup to run builds that populate sstate. > > Then, bitbake-setup would be used again to set up the same build, but > instead it would run a selftest inside that setup that checks that > sstate mirror has been properly populated. We already have such a > test, and such a job, but they work directly in the bitbake build set > up by the autobuilder. > > Note that master-next (or somerelease-next) doesn't always match what > then gets added to master (or somerelease); for testing exactly master > there are nightly a-full jobs. There could be a short window of not > yet provided sstate until that nightly job runs, but I'm not sure how > to address that, short of asking maintainers to run the sstate > population job from staging branch before they push that staging > branch to master/stable branch.
Ok, doing this in a-full is both excessive and unnecessary. Rather instead of the above let's run a separate job, triggered by new commits in master (or stable branches) that would populate sstate (via bitbake-setup plus autobuilder sstate tweak). Existing tests already cover for that sstate propagating to CDN. I need to check if buildbot and autobuilder code can support jobs triggered by new commits, otherwise it has to be scheduled like nightly or auh, which I'd like to avoid. Alex
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#210977): https://lists.openembedded.org/g/openembedded-core/message/210977 Mute This Topic: https://lists.openembedded.org/mt/111034286/21656 Group Owner: openembedded-core+ow...@lists.openembedded.org Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
