Repository navigation
🐛 contribute: never hand umbrella or owner-only issues to a contributor relay - #19081
Conversation
…or relay A shared classifier (hub.contribute_umbrella, per-repo overrides) marks an umbrella by label, body marker, title prefix, issue-referencing task list or sub-issues, and an owner-only issue by label or marker. Contributor selection and the ready-work queue refuse them, /api/contribute/status gains umbrella_items, human_only_items and candidate_items_raw while actionable_items and candidate_items count only workable work, the contributor page shows N workable plus an excluded line, and the good-first-issue sweep withdraws help wanted and good first issue from an umbrella. They stay actionable for the owner and stay closable, unlike hive/tracking. Fixes #19078 Signed-off-by: Andy Anderson <andy@clubanderson.com>
925a444 to
7e419ae
Compare
Open PR overlap checkBase branch: Conflicts are checked with every PR merged onto the current tip of
|
|
@clubanderson this PR is back in your court ( Failing checks:
Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path. |
UI render evidenceRender evidence captured. Built from
These are fixture renders, not live data: the page was built from the PR head and filled with synthetic data from |
Signed-off-by: Andy Anderson <andy@clubanderson.com>
…-rules count test Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
[APPROVALNOTIFIER] This PR is APPROVED Approval requirements bypassed by manually added approval. This pull-request has been approved by: The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
Fixes #19078
What changes
Umbrella, epic, tracker and decomposed-parent issues, and issues that wait on a person, are never handed to a contributor relay, never counted as workable, and never advertised with
help wanted/good first issue. They stay actionable for the owner and the hive, keep their stage, and stay closable.Recognition (one classifier,
pkg/gfi/umbrella.go; confighub.contribute_umbrellawith aper_repooverride map, every field optional and inheriting):labels(defaultumbrella,tracker,epic,meta-tracker; case-insensitive, last segment sokind/epic), the<!-- hive:umbrella -->body marker (ignored inside code spans), a title prefix (default[Tracking],[Epic],[Umbrella],[Tracker]), a task list with at leasttask_list_min(default 2) issue or PR references, sub-issues, or the legacyis_trackerdetector.needs-human,human-only,human-only:*(human_only_labels), or<!-- hive:human-only -->.needs-humankeeps its existinghard-suppressname in the good-first-issue picker.hive/partially-addressedis not a reliable umbrella signal.Where it applies
/contributeselection and the ready-work queue (shared ladder, withheld reasonsumbrellaandhuman_only)./api/contribute/status: newumbrella_items,human_only_items,candidate_items_raw(the old raw count, kept for compatibility);actionable_itemsandcandidate_itemscount only workable work. Alsoumbrella_total/human_only_totalon/api/contribute/queueandumbrella_count/human_only_counton the fleet payload. No in-repo consumer readscandidate_items;actionable_items(relayneedieststrategy, federation heartbeat) now means workable, which is what they want.help wantedandgood first issueon the next sweep (logged, audited asgfi-withdraw,withdraw_advertisement: falseper hive or per repo opts out). It runs before the claim gates, so a hive-claimed umbrella loses the invitation too. There is no "recently labelled by a person" guard in the existing withdrawal code, so none was added.umbrella/owner_onlyon the stage tracker item).Stage decision. No
stage/0NN-umbrellalabel: a stage lives in the engine table, transition table, dashboard columns and filter vocabulary and every stage-aware test harness (the018change touched 53 files and cannot be validated without a local build), and umbrella is orthogonal to pipeline progress (a parent is010while waiting on children and040while a PR covers it). It is a derived pill over the shared rules instead, documented in stage-labels.md. Follow-up: a dashboard filter chip for the pill.Auto-close untouched. The markers are not
hive/tracking:CloseIssuedoes not refuse them andparent_autoclosedoes not exempt them (anepiclabel keeps its existingepics_toorule).umbrella_close_test.gopins an umbrella closable and ahive/trackingissue refused. Checked #19073 (Settings control forparent_autoclose): shared files only by adjacent edits (openapi.json, settings_parity_allowlist.txt, stage-labels.md, regenerated static/index.html; regenerate after either merges); per-repo overrides live underhub.contribute_umbrella.per_repo, notrepos:.Relay. #19053 is open and untouched. Follow-up for the relay: name
umbrella_items/human_only_itemsin its empty-queue message.Dry run: open issues the new rules would classify (nothing was labelled or edited)
Epichelp wantedhelp wantedhelp wantedNo open issue currently matches owner-only: none carries
needs-human(those are filtered already),human-onlyor the marker. #18660, #18661, #18524, #18746 and the prose-only trackers #9846 and #2364 match no rule today; label themhuman-only/umbrella(or add the marker) to take them off the relay, which the operator can decide after reviewing this list.Tests
Table-driven: each recognition rule, per-repo override and disable, counts and compatibility fields, shared selectTask/queue/withheld ladder cases, picker exclusion and withdrawal (including
help wantedalone andwithdraw_advertisement: false), dashboard text and pill, close-path eligibility. Two existing gfi tests that pinned hand-labelled epics as left alone were updated to the new decision.make dashboard-genwas run for the pill. Not run locally (CI is the gate).