Repository navigation
π Advisory issues: skip empty dedupe checklists, show as advisory in tracker - #15418
Conversation
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here. DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
|
stage 080-awaiting-merge-admission β 050-review-pending: required-check-pending:build-and-test |
Open PR overlap checkBase branch: Conflicts are checked with every PR merged onto the current tip of
|
β¦tracker The dedupe lane filed #15381 with zero actionable findings; the Activity Monitor rendered it as 'Open Β· since 3h', indistinguishable from stalled work. - Prompt: only action-required rows (near-duplicate pair, resolved candidate, missing PR<->issue link, consolidation close) justify a new checklist; merge-order reminders and cross-links are kick output. - Relay guard: the issue-request watcher refuses issue-create for a dedupe-checklist body whose verdict is 'no' with all-zero counts and no task lines; audit agent_issue_empty_advisory_refused, bead keeps the summary. Parser unit-tested on #15381's real body. - Tracker: triage/advisory label, dedupe marker or 'Decision required: no' classifies the row as advisory. Chevron says 'Advisory Β· read & close' or 'Advisory Β· nothing found Β· safe to close' instead of 'Open Β· since Nh'; not counted as New/stalled/attention/needs-human; row menu leads with 'Close (read)' via the existing items decision endpoint (new verb close-read, closes as completed). Signed-off-by: Andrew Anderson <andy@clubanderson.com>
1095e7e to
5a6045d
Compare
5a6045d
βΉοΈ Sentinel notice β informational; author is trusted (owner), merge is not blockedHive flagged this PR (author @clubanderson, head
Hive did not add the |
There was a problem hiding this comment.
Decision needed: @clubanderson should decide whether #15417 (the earlier partial fix for #15415, Refs #15415) is closed as superseded by this PR, or whether this PR adopts #15417's stricter policy wording that forbids advisory-only checklists entirely (this PR still permits them for in-place edits).
Overlaps and conflicts with #15417 β both fix #15415, both labelled conflicts-with-open-pr, and they cannot both merge as-is.
I read both diffs in full. Same semantic change at the same hunk sites in four files:
src/pkg/github/attribution.go:48β both addAuditActionAgentIssueEmptyAdvisoryRefused = "agent_issue_empty_advisory_refused"(identical constant, different comment text β textual conflict).src/pkg/github/issue_request_watcher.go:555β both insert the empty-dedupe-checklist refusal immediately before the prior-art verdict, with the same terminal sequence (recordWriteAuditβwriteIssueResultβos.RemoveβissueClearRetry). #15417 usesemptyDedupeAdvisory(req.Body); this PR usesParseDedupeChecklistFindings(body).Actionable().src/pkg/policies/defaults/dedupe-issues.md:205-245andsrc/policies/dedupe-issues.mdβ same "no action rows β file nothing" rule.src/pkg/policies/dedupe_policy_test.go:58-67β different asserted strings.
Where they differ in a way that matters: #15417 deletes the --label triage/advisory filing path and asserts "Do not create an advisory-only checklist"; this PR keeps that path for a checklist edited in place whose action rows have all closed (dedupe-issues.md:248-250). Those are two different policies, so this is a conflict, not a pure duplicate.
This PR is the broader one: it closes #15415 (the tracker Advisory chevron, close-read decision verb, and attention/needs-human accounting that #15417 explicitly leaves "to be implemented"). I would keep this one and close #15417 as superseded β but #15417 is a fork contributor's PR, so that is a maintainer's call. I have not closed anything.
Correctness read (tree at eb61710): ParseDedupeChecklistFindings reuses the existing parseDedupeChecklist/htmlCommentPattern from dedupe_advisory_close.go:36,55; NeedsHuman/NeedsAttention are computed in the post-pass (stage_tracker.go:2365, :2414) after applyStageTrackerAdvisory runs, so the it.Advisory != nil short-circuits take effect. The count regexes match the policy's own Summary template (dedupe-issues.md:124-125) and the #15381 testdata body. Nothing defective found; the build (linux/arm64) check shows no conclusion yet and tide is pending.
β hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
There was a problem hiding this comment.
Decision needed: @clubanderson should decide whether #15418 supersedes #15417 (close #15417) or #15417 lands first and #15418 rebases onto it β they conflict in src/pkg/github/attribution.go and issue_request_watcher.go and disagree on whether advisory-only checklists may exist at all.
Conclusion: the relay half of this PR overlaps #15417 (@Karibhou) byte-for-byte in intent and position, and the two cannot both merge as-is. The code here builds and the parser logic checks out; the open question is which PR survives.
Verified overlap (read both diffs in full):
- Both add the constant
AuditActionAgentIssueEmptyAdvisoryRefused = "agent_issue_empty_advisory_refused"at the same spot insrc/pkg/github/attribution.go(afterAuditActionAgentIssuePriorArtRefused, ~line 47). Whichever lands second gets a textual conflict there, and a hand-resolved union would be a duplicate const. - Both insert a terminal "empty dedupe advisory" refusal at the top of the
kind == "issue"branch inhandleOneIssueRequest(src/pkg/github/issue_request_watcher.go:~555), writing the sameIssueResponse{OK:false}/os.Remove/issueClearRetrysequence. #15417 does it viaemptyDedupeAdvisory(req.Body)(regex counts); this PR viaParseDedupeChecklistFindings(body).Actionable(), which reuses the existingparseDedupeChecklistmarker/verdict parse indedupe_advisory_close.go:55β the better anchor, since the marker check is already the hive's definition of "is a checklist". - Both rewrite the same two paragraphs of
src/pkg/policies/defaults/dedupe-issues.md(and thesrc/policies/mirror).
Where they are not duplicates β this is the decision:
- #15417 deletes the "advisory-only checklist β file with
triage/advisory" section and thegh issue create --label triage/advisorysnippet outright ("Do not create an advisory-only checklist"). - This PR keeps advisory-only as a legitimate state ("only ever the result of editing an existing checklist in place whose action rows have all since closed") and adds the dashboard side:
RepoWorkIssue.Advisory(client.go:~4636),IssueAdvisoryFromBody,stage_tracker_advisory.go, and the exportedDedupeAdvisoryLabelName(standing_issues.go:~23) so the tracker chevron reads Advisory instead of Open. #15417 has none of that.
So #15418 is a superset on the relay + UI, but the two PRs state opposite policy on whether triage/advisory issues exist. If this one is the intended direction, #15417 should close rather than be rebased; if #15417's "never file advisory-only" is the rule, this PR's advisory chevron is drawing a state the policy says must not occur.
Correctness notes on this PR (read against the v5 tree):
withParentMarker(decomposed_parents.go:90) appends the parent marker, so the dedupe marker stays on line one andparseDedupeChecklist'sHasPrefixcheck still fires on the post-markerbody. Good.Actionable()treatsDecision required: yesas actionable even with zero counts, and the delta phrasing the policy recommends ("+2 near-duplicates") still matches\b(\d+)\s+near-duplicateβ errors land on the "file it" side, which is the right default.go build ./pkg/github/ ./pkg/dashboard/ ./pkg/policies/at5a6045dsucceeds; CI test/coverage/Dashboard Lint are green.
I have not closed or changed anything; flagging for the maintainer to pick a survivor.
β hive: agent=reviewer-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
|
@clubanderson is auto-promoted for contributor-lane merging on this PR: the last 5 merged PRs by this author in this hive's repositories have no revert or sentinel-alert evidence. CI, reviewer verdict, diff fences, soak, holds and branch protection still apply. |
|
stage 080-awaiting-merge-admission β 050-review-pending: review-gate:no-review-bundle |
|
Thank you for your contribution! Your PR has been merged. We'd love to hear how your experience was: share feedback |
Fixes #15415
Owner: "why is this sitting open and not actionable for 3h?" β about #15381, a dedupe checklist with 0 near-duplicate pairs, 0 missing links, one merge-order reminder already declared in the PR body and one cross-link note. The Activity Monitor showed it as
Open Β· since 3h.Owner's rule, verbatim: "if something is advisory then the chevron should say advisory instead of open."
1. Don't file empty advisories
dedupe-issues.md(defaults +src/policiescopy): a new checklist needs at least one action-required row (near-duplicate pair, resolved candidate, missing PRβissue link, consolidation close β each a- [ ]task line). Merge-order reminders, cross-links and "checked, not duplicates" are notes β kick output only.pkg/github/issue_request_watcher.go, before the prior-art verdict, inside the existing gate β no new direct writer):issue-createis refused for ahive:dedupe-checklistbody whose verdict isno, all counts are zero and there are no task lines. Result file explains, audit actionagent_issue_empty_advisory_refused, the summary is kept as a knowledge bead.ParseDedupeChecklistFindingsunit-tested on triage: duplicate and overlapping open issues/PRs to consolidateΒ #15381's real body (testdata/dedupe_checklist_15381.md):Actionable() == false.2. Advisory issues look like advisories in the tracker
stage_tracker_advisory.go:triage/advisorylabel, the dedupe marker, or aDecision required: nobody line βitem.advisory{kind:"advisory", source, empty, chip, caption}.Advisory Β· read & close, orAdvisory Β· nothing found Β· safe to closewhen the body reports zero findings β neverOpen Β· since Nh. No stale tint. Caption "Advisory for maintainers Β· read, then close" / "β¦ nothing found Β· safe to close".tracker-filter.js).POST /api/repos/{o}/{r}/items/{n}/decision, new verbclose-read(closes as completed with an acknowledgement comment; 400 for PRs).TestStageTrackerAdvisory{For,Row,ChipAndAction}(node harness asserts the chevron starts withAdvisoryand never saysOpen Β· since),TestRepoItemDecisionCloseReadClosesAsCompleted, operator-action mapping.src/docs/dashboard-sections.md"Advisory row";changelog.d/fixed-advisory-issue-noise.md.SPA regenerated with
go generate ./pkg/dashboard/...;static/index.htmlcommitted.