Repository navigation
🐛 fix: explain human requirements and optional tracker actions - #16011
Conversation
Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
[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 |
Open PR overlap checkBase branch: Conflicts are checked with every PR merged onto the current tip of
|
Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
Source review of the additional approval/changes-requested example found that a recorded operator approval could project a still-current rework row onto merge admission and clear its visible reason. Head 55c9b68 prevents that: current review changes requested keep their actual stage/reason and show the author/contributor action plus a separate operator-approval receipt, never review approval or merge readiness. An earlier-head approval cannot authorize the displayed head. Added backend coverage across rework/decision stages and frontend actual covered-issue rendering. Also suppress stale head/identity actor prose and generic person-only chevron wording. No runtime gate, deployment, preparation/native train, sorting or tree changes. CI is the only test/build verdict and is still running. |
Signed-off-by: Andy Anderson <andy@clubanderson.com>
…ctor prose Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
Hive status — updated in place; edits do not notify. Full history is in the Activity Monitor. @clubanderson this PR is held at Failing checks:
Next: the hive does not push to your branch. Push a new head (or reply here); CI and the review swarm re-run automatically and the PR returns to the normal path. Evidence bundle `#16011813a1c1` sha256:20098e27a815c2225a7234cdf8db3569ba7f3f15750fa0d9b545924a24fcef8b — download from the dashboard or `hivectl review evidence #16011`. |
Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
Owner-requested same-row head synchronization is included in e19e55b. Source root cause: rework could infer pushed from old NudgeHead != displayed Head while push still required the separate push timestamp/SHA ledger, retaining waiting-to-push presence and author cues. Backend now reconciles both paths using one head-scoped predicate before row/linked-PR projection. A fresh current-head author hold is preserved; a satisfied prior-head request yields pending re-review, Hive actor and no stale human requirement. Unknown push time is explicit by omission of age, not PR updated_at guessed as a commit timestamp. The existing stage number/train style remain unchanged. Added backend actual covered-row payload regressions and shipped pill/chevron rendering cases; CI only, no deployment/config change. |
Signed-off-by: Andy Anderson <andy@clubanderson.com>
…g actor Signed-off-by: Andy Anderson <andy@clubanderson.com>
Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
Latest requested fidelity additions are at exact HEAD fc1ddc6: ownership policy is not PR rejection; human availability stays unknown despite any agent heartbeat, with explicit adoption/provenance unchanged. Current-head observed CI retry is a distinct checking state, not a fake push; actual failures, review findings, conflict/sign-off and explicit holds remain actionable. Full payload regression uses #16012 dfaadfa -> 78c30d9 and #16016 exact 94ed43c. Only gofmt and required SPA generation ran locally; fresh CI is authoritative. No rollout/configuration changes. |
There was a problem hiding this comment.
Correctness review (head 748a84b)
1. needs_human now diverges between backend and SPA for pushed human-authored 071/072 rows — medium
src/pkg/dashboard/stage_tracker_build.go:577-578 adds !(it.Push != nil && it.Push.State == pushStatePushed) to stageTrackerNeedsHuman, which the comment at :560-563 declares the ONE definition that stageTrackerItemNeedsHuman in the SPA mirrors. The mirror at src/pkg/dashboard/static-src/js/tracker-filter.js:119-126 was not updated and still returns true for any human-authored 071/072 row. The SPA does not read the payload's needs_human; the 🙋 chip count and filter call the local mirror (tracker-filter.js:189,321). Result: a human PR whose new head satisfied the old nudge gets no human_attention, the person glyph is hidden (tracker-row-strip.js:29-31), and the backend header count excludes it — but the 🙋 Needs human filter still lists it and the chip count disagrees with the payload. Reproduces with the reconciled row in stage_tracker_push_test.go (push {state:'pushed'} on a 072 human row).
Fix:
return STAGE_TRACKER_NEEDS_HUMAN_AUTHOR_STAGES.indexOf(n) >= 0 && (item.author_kind || '') === 'human' &&
!(item.push && item.push.state === 'pushed');
and add an assertion in stage_tracker_ui_test.go next to :339-341 that a 071 human row with push: { state: 'pushed' } is false.
2. Approval head compared with strict != — low
src/pkg/dashboard/stage_tracker_decided.go:175 uses a.HeadSHA != it.Head. a.HeadSHA is whatever the delegate response's head field returned (operator_actions.go:405,541), and the same package compares operator-action heads with review.SameHeadSHA (stage_tracker_decision.go:142) to tolerate case/abbreviation. A short or differently-cased recorded head would make every approval render as "Earlier approval recorded". Suggest !review.SameHeadSHA(a.HeadSHA, it.Head).
— hive: agent=adjudicator-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
… wording Signed-off-by: Andy Anderson <andy@clubanderson.com>
Signed-off-by: Andy Anderson <andy@clubanderson.com>
There was a problem hiding this comment.
Correctness — delta re-review at 790cdef (previous head 748a84b). The delta's PR-authored work (fc1ddc6, f65c659: same-head CI retry → pushStateChecking, neutral ownership wording, human presence never from agent heartbeat) reads correct and is well covered by stage_tracker_ci_retry_test.go and the new presence tests. Both prior findings are still present, and the first one is widened by this delta:
-
SPA Needs-human mirror still diverges from the server —
src/pkg/dashboard/static-src/js/tracker-filter.js:119-126.stageTrackerItemNeedsHumanreturns true for any 071/072 human-authored row, while the server'sstageTrackerNeedsHuman(src/pkg/dashboard/stage_tracker_build.go:579-581) now (a) treats anadoption.blocked_by == 'author-is-person'owner as human, and (b) excludes rows whosepush.stateispushedorchecking(the new state in this delta). Result: the "Needs human" filter in the SPA keeps showing rows the header count and server flag exclude (a same-head CI retry row still filters as needing a human). The comment atstage_tracker_build.go:562-564says the SPA mirrors this ONE definition; it doesn't.return STAGE_TRACKER_NEEDS_HUMAN_AUTHOR_STAGES.indexOf(n) >= 0 && ((item.author_kind || '') === 'human' || (item.adoption && item.adoption.blocked_by === 'author-is-person')) && !(item.push && (item.push.state === 'pushed' || item.push.state === 'checking'));(or simply trust the payload's
needs_humanflag when present and fall back to the mirror). -
Earlier-approval check uses strict
!=—src/pkg/dashboard/stage_tracker_decided.go:175:a.HeadSHA != it.Head.a.HeadSHAcomes from the endpoint'sheadresponse field or the client-suppliedreq.Head(src/pkg/dashboard/operator_actions.go:536-541), so an abbreviated or differently-cased head makes a current-head approval render as "Earlier approval recorded … previous head". The rest of the dashboard compares heads withreview.SameHeadSHA(src/pkg/dashboard/stage_tracker_decision.go:142,src/pkg/review/head_sha.go:21-37).if approval && a.HeadSHA != "" && !review.SameHeadSHA(a.HeadSHA, it.Head) {(add
"github.com/hivecommons/hive/pkg/review"to the imports.)
— hive: agent=adjudicator backend=copilot model=claude-fable-5.1 copilot=1.0.88
Signed-off-by: Andy Anderson <andy@clubanderson.com>
…uirements Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
Updated exact head: 813a1c1. Preserves landed Copy link/Close issue actions and #16022 durable author-response evidence. Independent code review: no significant findings in human/head semantics or the added pending-head column fix; the separate rail review found a fixture-only missing reconciler key, fixed before its ordering commit. Root cause of the #16014 column repro: a premature merge-call refusal for expected PR-head checks was treated as live merge-queue state at 050/060. It now remains CI/Review with the earlier refusal retained as history; actual queue/live merging and separate synthetic candidate observations are unchanged. The prior 790cdef CI failed only an old explicit-human heartbeat assertion; updated it to require unknown human availability and no heartbeat lookup. Fresh exact-head CI is running, not yet a verdict. No deployment or live configuration changes. |
|
merged by Hive trusted-author auto-merge: author @clubanderson holds role owner and GitHub permission admin on repo; CI green at 813a1c1. |
|
Thank you for your contribution! Your PR has been merged. We'd love to hear how your experience was: share feedback |
|
FINAL exact-head CI receipt: 813a1c1 has all81 CheckRun contexts terminal SUCCESS/SKIPPED and DCO SUCCESS. The final linux/arm64 image job124532466226 in run41199521160 completed SUCCESS at2026-10-31T12:28:11Z. Hive had already merged this source at12:18:34Z as95eafcdfa61ad235655987156e65406723895e97; this receipt distinguishes later final validation from the earlier merge. Legacy Tide remains PENDING after merge and is not an unfinished Actions job. No manual merge/deployment by me. |
…route "Needs human" notes through the hive-note tuck (#16584) Fix 1: when a CI or Review stage reaches a settled verdict the chevron renders only its icon (green pass / red fail, existing status tokens); the words move into the tooltip and aria-label. In-progress and awaiting-human states keep today's text. Chevron geometry unchanged. Fix 2: PR #16011 (95eafcd) drew "Needs human" context as a static .ham-human-context row with animation: none so the requirement stayed readable. Middle ground: those lines now ride the shared hive-note tuck machinery (open 30 s, slide-left into the ⬡ icon, click to re-open, keyed by row+text so re-renders never re-open them), while the row's Let hive decide / Sentinel alert controls keep the decision state. Signed-off-by: Andrew Anderson <andy@clubanderson.com> Co-authored-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Summary
Validation