The handbook names the appliance check build-and-boot and describes its old behaviour #9
Labels
No labels
accessibility
bug
documentation
duplicate
enhancement
good first issue
help wanted
invalid
question
wontfix
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
Keel-Linux/handbook#9
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Keel-Linux/.github#12 renames the reusable appliance job
build-and-boottoboot-published-layer, so the check isappliance / boot-published-layer, andmakes a layer that has never been published fail the job instead of passing it.
The ten caller repositories are updated with it. The handbook still carries the
old name and, in two places, the old behaviour:
docs/ci-cd.md, section on check names: "test-appliance.yml(job idbuild-and-boot), so its check is exactly:appliance / build-and-boot".docs/ci-cd.md, the new-repository procedure: "its check isappliance / build-and-bootand it is required only once it has passedonce, because it depends on a layer being published". The reason is now
different: it fails while the layer is unpublished, rather than passing
vacuously.
docs/ci-cd.md, the keel-nodebb note and the keel-core row of the table.docs/releases-host.md, "Publishing a layer": "appliance / build-and-bootskips with a notice and passes. Publishing a layer is therefore what turns
that check from a skip into a real run." It now fails rather than skipping.
Worth adding at the same time, because it is the substance of that pull
request and not only a rename: the check boots the layer the mirror publishes,
so on a pull request that changes a recipe it is evidence about code that is
already merged. The procedure in
docs/ci-cd.mdthat tells a maintainer torequire it should say so, and should say which way renaming a required check
fails. GitHub fails closed: a required context that is never reported leaves
the pull request blocked, waiting for a status that will never arrive, so a
rule left on the old name locks the repository rather than quietly ceasing to
apply. Measured on keel-mariadb on 2026-09-28: pull request 16, reporting
appliance / build-and-boot,clean; pull request 17, reportingappliance / boot-published-layerand identical otherwise,blocked. Renamethe rule in the same change.
unit-mariadb,unit-postgresqlandunit-rediseach mentionbuild-and-bootonce in a header comment as well; those are prose, not gates.