feat(janitor): unmergeable_pr_rebase rule — auto-rebase when no event fires #783

Merged

charles merged 1 commit from code-lead/781 into main

2026-05-02 22:36:11 +00:00

claude-desktop commented

2026-05-02 22:23:41 +00:00

Collaborator

Closes #781.

Summary

Adds janitor rule unmergeable_pr_rebase that closes the no-event gap on Forgejo: when a PR's mergeable flag flips to false outside of a pull_request closed && merged=true (push to base outside a merge, force-push, dependency landing elsewhere), nothing currently dispatches a rebase. The rule sweeps open PRs each janitor pass and dispatches via the same path as handlePostMergeRebase.

Design

New rule registered in _ALL_RULES and reconcileOnce.
Skips mergeable !== false, non-agent authors, and any PR that declares parents (those defer to handlePostMergeRebase's parent-merge cascade — avoids racing the stacked-rebase serialisation).
Reuses the existing _postMergeRebaseDispatched map. Two paths (event-driven + janitor sweep) cannot double-dispatch within the 10-minute window.
Dispatch logic factored out of handlePostMergeRebase into dispatchPrRebase so both call sites share the skill loader, B10 watchdog tagging, A3 label resolution, and dedup.
Exposed via the impl seam so tests swap the dispatcher / dependency-graph reader / agent resolver without mock.module (which is process-global in Bun).

Test plan

Unit: rule dispatches when mergeable=false + agent author + no parents
Unit: rule skips when mergeable=true
Unit: rule skips when author isn't a configured agent
Unit: rule skips when PR declares any parent (B13 / #429 routing)
Unit: dedup status from dispatchPrRebase is a silent skip (no JanitorAction emitted)
Unit: dry-run mode flags without calling the dispatcher
Existing post-merge tests (post-merge.test.ts, event-handlers.test.ts) still pass through the refactored helper — 100/100
just qa clean (3 pre-existing session JSONL pruning failures on main are unrelated)

Closes #781. ## Summary Adds janitor rule `unmergeable_pr_rebase` that closes the no-event gap on Forgejo: when a PR's `mergeable` flag flips to `false` outside of a `pull_request closed && merged=true` (push to base outside a merge, force-push, dependency landing elsewhere), nothing currently dispatches a rebase. The rule sweeps open PRs each janitor pass and dispatches via the same path as `handlePostMergeRebase`. ## Design - New rule registered in `_ALL_RULES` and `reconcileOnce`. - Skips `mergeable !== false`, non-agent authors, and any PR that declares parents (those defer to `handlePostMergeRebase`'s parent-merge cascade — avoids racing the stacked-rebase serialisation). - Reuses the existing `_postMergeRebaseDispatched` map. Two paths (event-driven + janitor sweep) cannot double-dispatch within the 10-minute window. - Dispatch logic factored out of `handlePostMergeRebase` into `dispatchPrRebase` so both call sites share the skill loader, B10 watchdog tagging, A3 label resolution, and dedup. - Exposed via the `impl` seam so tests swap the dispatcher / dependency-graph reader / agent resolver without `mock.module` (which is process-global in Bun). ## Test plan - [x] Unit: rule dispatches when `mergeable=false` + agent author + no parents - [x] Unit: rule skips when `mergeable=true` - [x] Unit: rule skips when author isn't a configured agent - [x] Unit: rule skips when PR declares any parent (B13 / #429 routing) - [x] Unit: dedup status from `dispatchPrRebase` is a silent skip (no `JanitorAction` emitted) - [x] Unit: dry-run mode flags without calling the dispatcher - [x] Existing post-merge tests (`post-merge.test.ts`, `event-handlers.test.ts`) still pass through the refactored helper — 100/100 - [x] `just qa` clean (3 pre-existing `session JSONL pruning` failures on `main` are unrelated)

claude-desktop added 1 commit

2026-05-02 22:23:41 +00:00

feat(janitor): unmergeable_pr_rebase rule (closes #781 )

qa / dockerfile (pull_request) Successful in 7s

Details

qa / qa (pull_request) Failing after 22m39s

Details

9cda43f575

Forgejo fires no event when a PR's mergeable flag flips after main moves
on under it (push outside a PR-merge, force-push, dependency landing
elsewhere). handlePostMergeRebase only covers the explicit merge case;
pr-changes-requested-graph only covers the review case. Stale unmergeable
PRs sat indefinitely.

Adds the eighth janitor rule. Each pass walks open PRs, picks
mergeable=false ones whose author is a configured agent and whose
declared parents are all merged, and dispatches a rebase via the shared
post-merge path — same dedup map keyed on repo#pr@sha so a janitor sweep
and a webhook-driven cascade can't double-dispatch within the 10-minute
window.

The dispatch helper is extracted from handlePostMergeRebase into
dispatchPrRebase so both call sites use the same skill loader, watchdog
tagging, label resolution, and dedup. Exposed via the impl seam so tests
can swap it without process-global mock.module.

Closes #781

charles merged commit 1d439a3a11 into main

2026-05-02 22:36:11 +00:00

charles referenced this pull request from a commit

2026-05-02 22:36:11 +00:00

feat(janitor): unmergeable_pr_rebase rule — auto-rebase when no event fires (#783)

charles deleted branch code-lead/781

2026-05-02 22:36:12 +00:00

claude-desktop referenced this pull request

2026-05-02 22:49:39 +00:00

Janitor rule: redispatch stale fix-ci on PRs whose CI is still red #784

charles referenced this pull request from a commit

2026-05-02 22:55:46 +00:00