Repository navigation
🐛 fix: stage tracker gives immediate Let-hive-decide feedback and says why a human is wanted - #15871
Conversation
…ys why a human is wanted Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: clubanderson <407614+clubanderson@users.noreply.github.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
|
…d test fixtures Seven dashboard test fixtures lift stageTrackerActorTip by name without stageTrackerHumanWhy, so the shuffle job failed with ReferenceError. Follow the file's existing guarded-helper pattern and regenerate index.html. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
|
stage 080-awaiting-merge-admission → 050-review-pending: required-check-pending:Sign-off survives the squash |
There was a problem hiding this comment.
Read the diff against the v5 merge-base (85e9ec7) plus the callers; this looks correct to me.
What I verified:
stageTrackerPRListedOwnRowkeys off_stageTrackerRows, whichtracker-filter-render.js:99sets before the page's row HTML is built (:152-158);stageTrackerAdoptOrphanPRs(tracker-row-layout.js:276) only rewritesdrawn_underand adds no top-level rows, so the "PR has its own row" check can't miss a drawn row.- The preview pend (
tracker-actions.jsstageTrackerDelegatePreview) is taggedbusy='decide:delegate'and cleared infinally, andstageTrackerLeadBusySynconly touches a lead whosedata-ham-actmatches, restoringaria-disabled/titlefrom the saved dataset — a lead rendered disabled stays disabled afterwards.stageTrackerPendingKeys,stageTrackerClearPendingandstageTrackerActingCaptionall exist at the signatures used. stage_tracker_actor.go:204reachesstageTrackerDecisionWhyonly for 085/045;ReadyForHumanMergekeeps its non-empty "a maintainer merges it", so the client fallback tohuman_merge_reason_textinstageTrackerActorTipnever fires there and #15394's "say only who moves" is preserved.stage_tracker_decide_feedback_test.gocovers the CSS rule, pend tagging, busy/restore, the covered-issue chip and both why fallbacks.
No findings. Not a request for changes.
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
There was a problem hiding this comment.
Intent alignment (read at head 8c3280e): the busy button, pend tagging and the human-why tooltip fallback do what the body says, and the body honestly leaves B / skeleton / scroll / the no-👤-test open on #15773. One in-scope gap:
The covered-issue chip does not fire in the case #15773 actually reports. The owner's example is #15503, a secondary issue "under #15502" covered by PR #15759. The server nests a PR under its lowest listed issue (src/pkg/dashboard/stage_tracker_links.go:359-360), and outside the Pull requests view stageTrackerVisible drops a nested PR from the rows (static-src/js/tracker-filter.js:219), so _stageTrackerRows never contains #15759 while #15502 is listed. stageTrackerPRListedOwnRow (static-src/js/tracker-actions.js:505-509) therefore returns false for #15503, and stageTrackerParkedNestedPRs/stageTrackerDecisionTarget (tracker-actions.js:197-212) still resolve #15503's lead to PR #15759 regardless of drawn_under. #15503 draws no sub-row either (tracker-row-layout.js:270 filters linked PRs to stageTrackerLinkedDrawnHere), so the subRows branch at tracker-actions.js:513 never applies. Net: #15503 and #15502 both keep a live Let hive decide button — the exact duplicate the issue describes — and the new test even pins this ("the issue keeps its button when no sub-row is drawn (one covering PR)", stage_tracker_decide_feedback_test.go). The chip only appears in the narrow case where the covering PR is a top-level row beside the issue, which is not the reported layout.
Requested change: in stageTrackerLeadActionOnRow, also return the passive chip when the target PR is in item.linked_prs with a drawn_under naming a different issue that is among ctx.rows (i.e. !stageTrackerLinkedDrawnHere(lp, item) and that issue is drawn) — the decision then lives on #15502's row and #15503 reads "decision pending on #15759". Add that fixture to TestStageTrackerCoveredIssueDropsPRChipNode and soften the changelog/body claim if the general single-covering-PR case is deliberately left for later.
— hive: agent=ci-maintainer backend=copilot model=claude-fable-5.1 copilot=1.0.88
…ys the decision reason once - stageTrackerLeadActionOnRow: a secondary covered issue (hive#15503 under #15502) whose PR is drawn_under another listed issue renders the passive "decision pending on #N" chip instead of a second Let hive decide button; the chip names the row that draws the PR. Fixture added to TestStageTrackerCoveredIssueDropsPRChipNode. - stageTrackerTooltip: when the 035 next-move why repeats the caption, the Next move line keeps only who moves; the chevron hover keeps the whole why. - stageTrackerLeadBusySync stashes label/title/disabled in data-ham-* attributes like the rest of tracker-*.js; the node DOM stub drops dataset. - activity-monitor.md documents the server-side 035 why (lane reason, else decision_needed.question) and reads "Hive deciding…" like the code. Hive-Fix-Attempt: 5/6 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: ci-maintainer <280983584+hivecommons-hive[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Read the tree at d3aece9, callers included. Looks correct to me; three things I checked specifically:
- Render ordering.
tracker-filter-render.js:99assigns_stageTrackerRowsfrompage.rowsbefore the row loop at:158callsstageTrackerRowHtml, soctx.rows(tracker-row-layout.js:203) sees this render's rows, not the previous one's. Group headers ({group}withoutitem,tracker-row-layout.js:422) are dropped by the.filter(Boolean). The only otherstageTrackerRowHtmlcaller is the sound-preview sample (tracker-sound.js:728), where a stale/nullrows list just means no passive chip — harmless. - Busy button is not re-clickable.
stageTrackerLeadBusySyncsetsaria-disabled="true"and the lead click handler attracker-actions.js:809returns on exactly that attribute, so the double-click the body describes is closed; the saveddata-ham-was-disabledrestores a legitimately disabled lead afterwards. The pend is tagged'decide:delegate'and the sync only busies a button whosedata-ham-actmatches, so other row actions are untouched. - Go side.
joinClause(stage_tracker_caption.go:750) andhumanizeStageReason(:304) already exist;stageTrackerDecisionWhyfalls back toDecisionNeeded.Questionand then to the bare phrase, matching the activity-monitor.md text.
Body ↔ diff agree, including the explicit "Part of #15773" remainder. Note for whoever merges: the conflicts-with-open-pr label is real — #15969/#15970 also edit tracker-filter-render.js and static/index.html, so whichever lands second will need a rebase.
— hive: agent=reviewer-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
There was a problem hiding this comment.
Read the diff against the v5 tree. Looks correct to me; nothing blocking.
What I checked that the change depends on:
stageTrackerClearPending(tracker-actions.js:1987-1990at base) already re-runsstageTrackerPendingSync, so when the preview's tagged pend clears,stageTrackerLeadBusySyncseeson=false, isBusy=trueand restores the saved label /aria-disabled/title. ThefinallyinstageTrackerDelegatePreviewguarantees that clear even when a plan fetch throws.- The lead button is rendered with
data-ham-actinside.ham-row-head(tracker-actions.js:514,tracker-row-layout.js:230/238/540), so the new:scope > .ham-row-head .ham-lead-act[data-ham-act]selector finds it on both rows and sub-rows. _stageTrackerRowsis assigned intracker-filter-render.jsbeforestageTrackerRowHtmlruns for the page (line ~155), soctx.rowsis populated whenstageTrackerLeadActionOnRowreads it;linked_prs[].drawn_underis a real server field (stage_tracker.go:755, set atstage_tracker_links.go:394).- The tooltip dedupe compares against
lines[1], which isitem.caption || item.reasonin the non-held branch (tracker-row-strip.js:211); in the held branch an exact match against the Author line is impossible, so the worst case is no dedupe. - Go side:
joinClausealready inserts—, so the why readsa human decision is owed — workflow pathexactly as the doc now describes.
(I cannot approve; this is a comment.)
— hive: agent=reviewer-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
There was a problem hiding this comment.
Read the diff against v5 at merge-base fd50502 and the callers; looks correct to me. CI is fully green at 4055a6c (only tide pending), make's static check (webstatic/gen … -check) passes, and go build ./pkg/dashboard/ is clean.
What I verified:
- Busy button.
stageTrackerDelegatePreview(tracker-actions.js:1620-1629) opens a pend taggedbusy: 'decide:delegate'and clears it infinally, so a failed/thrownfetchinstageTrackerFetchDelegatePlanstill releases the button.stageTrackerActionBase(tracker-words.js:378) only rewrites reroute ids, sostageTrackerActingCaption('decide:delegate:preview')really resolves to "Asking the hive…" while the confirmed action keeps "Hive deciding…". The lead click handler (tracker-actions.js:809) already bails onaria-disabled="true", so the busy state blocks a double-fire without new wiring.stageTrackerPendingSyncis also called after every render (tracker-filter-render.js:173), so a re-rendered button gets its busy state back;:scope > .ham-row-head …stops a parent row from busying a sub-row's button. - Passive chip.
stageTrackerPRDrawnUnderOtherreadslinked_prs[].drawn_underwhichstage_tracker_links.godoes populate;_stageTrackerRowsis set from the drawn page before rows render, so an off-page PR correctly leaves the button on the issue. - Go side.
stageTrackerDecisionWhy→humanizeStageReason/DecisionNeeded.QuestionviajoinClause(stage_tracker_caption.go:750) — no behaviour change when both are empty.
One low-severity observation, not a blocker: the client fallback stageTrackerHumanWhy (tracker-row.js:206-208) appends raw item.reason (e.g. a slug like workflow-path) rather than a humanized clause, so on an old payload without stage_actor the tooltip may read Next move: human — workflow-path. The strip dedup at tracker-row-strip.js:221 then won't match the caption either. Only reachable on stale payloads, and the Go actor now sets why for 035/045, so fine to leave.
Intent vs #15773: the body is explicit that this is the client half and lists B (plan cache/prefetch) and the skeleton preview as remaining. Note that D's visible rendering after the chevron word ("PR in progress · #15759 needs a decision") is also not in this diff — only the tooltip half of D lands — so that should stay on the issue's open list too.
— hive: agent=reviewer-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
|
Thank you for your contribution! Your PR has been merged. We'd love to hear how your experience was: share feedback |
Let hive decide gathers the PR live for seconds with no sign anything happened, so operators click again or assume it is broken; covered issue rows repeat the PR's decision button; and the person marker on a chevron never says what the person is wanted for. The button and row now show the tracker's own working vocabulary until the preview opens, a covered issue defers to its PR row with a passive chip, and the human next-move tooltip falls back to the row's reason.
Part of #15773 — the client-side half (busy button, covered-issue chip, human-why tooltip); the plan cache/prefetch, skeleton preview, scroll-to-PR caption and the no-👤-without-a-reason test stay open on the issue.
— hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88 requested_by=@clubanderson