This repository was archived by the owner on Nov 8, 2026. It is now read-only.
Repository navigation
Cancel superseded CI runs and bound Ubuntu mirror downloads - #536
Merged
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Older commits of one MSeries PR left whole validation matrices running and queued, consuming shared GitHub runner capacity while Rupert awaited deployment. Every PR workflow now groups concurrency by workflow name and PR number and cancels superseded PR runs. Other PRs and workflows remain independent. Non-PR runs are not preempted, and the global package publishing lock is preserved.
A lightweight workflow checks all PR workflow definitions, including newly added ones. Nine regression tests cover missing/misconfigured cancellation, workflow/PR/ref isolation, case-insensitive workflow-name collisions, future workflows, and preservation of release and publisher locks. Handoff and roadmap record activation limits.
Validation:
python3 mseries/ci/workflow_policy.pypassed across all ten PR workflows.python3 -B -m unittest discover -s mseries/ci -v: 9 passed, including mutation checks on every PR workflow.git diff --checkpassed.This prevents superseded-run buildup once updated workflow revisions are in use. It does not retroactively cancel legacy runs or reserve shared capacity against legitimate current builds. No desktop installation, restart, publishing, or hardware changes.
Live verification: pushing a newer PR commit automatically cancelled older active runs 39878629706 (validation), 39878629771 (Disk Space), 39878629859 (packages), and 39878629749 (Aurora). Another PR package run 39877802029 remained in progress. A manual package control with publish=false (39879234844) continued into active jobs after a second PR update; it was then explicitly cancelled as test cleanup. No actual publishing was started. Final head f7019eb now awaits CI before merge.
Mirror hardening follow-up: Aurora dependency setup now uses signed, temporary Ubuntu ARM64 sources (HTTPS ports first, HTTPS Azure fallback), with 60-second metadata and 240-second package-download budgets. Installation uses verified downloaded packages with --no-download; runner sources and Ubuntu signing requirements stay intact. Rustup is bounded separately. All 17 CI tests pass locally, including eight mirror/timeout/installation-safety cases. The updated candidate and unified72 kernel builds and package inspection must pass before merge. Current dev documentation was merged with both sets of notes preserved to resolve a PR conflict that prevented checks from dispatching.