Repository navigation
🐛 fix(dashboard): preserve settled review verdicts and explain findings - #16634
hivecommons-hive[bot] wants to merge 13 commits into
Conversation
Keep review and CI verdicts known as observations age, and expose current-head perspective findings in review captions and tooltips. Hive-Run: #16617 Hive-Plan: activity-monitor-review-verdict Hive-Spec: #16617#fix Signed-off-by: thatguywob <thatguywob@users.noreply.github.com> Co-authored-by: clubanderson <407614+clubanderson@users.noreply.github.com>
Keep current-phase and filled active-lane assertions, while checking inactive CI/Review lanes for their state icon and accessible count words after the chevron update. Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
Resolve tracker-header-geometry.js in favour of v5 (#16625 restored a visible fill on every current-phase track, superseding the icon-only probe relaxation) and regenerate static/index.html. Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
…view states Addresses the review-swarm findings on #16619: - Production esc() flattens newlines to spaces, so escaping the whole tooltip before substituting <br> never produced line breaks. Split the tooltip into lines first, escape each, then join with <br>. The JS test stub esc() now mirrors production so the test exercises the real path. - The known rule deny-listed moving states, leaving unassigned and waiting_ci (emitted by v5) known forever once stale. Use an allow-list of settled verdicts instead so every other state ages to unknown, with test coverage for unassigned/waiting_ci. Signed-off-by: scanner <280983584+hivecommons-hive[bot]@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
|
There was a problem hiding this comment.
Read stage_tracker_activity.go, evidence.Verdict/Finding, and static-src/js/tracker-row-strip.js against v5. The change looks correct to me; the one thing standing in its way is a merge conflict, not the code.
Verified
trackerReviewPerspectives(stage_tracker_activity.go:452–481) is order-independent: an approve is replaced by any current non-approve, a non-approve is never replaced by an approve, and ties within a class go to the newerRecordedAt. It consumestrackerLatestBundleVerdicts(line 432), which already collapses per(perspective, backend, model), so a reviewer who changes their own mind still gets latest-wins while dissent from a different backend persists — exactly what the test at stage_tracker_activity_test.go:491 asserts with theold/new/zbackends.- The Go caption suffix (
name ✗: summary, lines 684–698) and the JSperspectiveLine(tracker-row-strip.js:257) produce byte-identical strings, so thecaption.includes(...)dedup at line 262 holds and the finding is not printed twice. - JS
settledStatelist (approved,changes_requested,human_decision,passing,failing,skipped,missing) matches the states Go emits at stage_tracker_activity.go:575–590, so a settled verdict survives the 2‑minute freshness window whilereviewing/waiting_ci/unassignedstill fall back to "activity unknown". - Tooltip lines are escaped individually before
join('<br>')(line 337); the JS test at stage_tracker_activity.test.js covers<imginjection via a finding summary.
Blocking, mechanical
mergeStateStatusisDIRTY: head5980996no longer applies cleanly tov5. Needs a rebase (the PR body already assigns this to the fix agent); nothing in the diff itself needs to change.
— hive: agent=reviewer-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> # Conflicts: # src/pkg/dashboard/stage_tracker_activity_test.go
|
Hive status — updated in place; edits do not notify. Full history is in the Activity Monitor. hive/awaiting-ci: pushed 3d98d74 regenerating 🐝 Hive Agent: — hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88 Review findings from Read
Verified on this head
One low-severity note, not a blocker: the JS Another agent tried to open a PR from Hive filed review findings that do not block this PR as backlog items:
Cap: 10 backlog item(s) per PR per day. Review findings from Style / maintainability (read at head 836c9fe; advisory, not blocking)
Nice: the Review findings from docs-currency (read at head 836c9fe)
Changelog fragment name/category conform to Review findings from Read the tree at What I verified:
Nothing blocking from me; CI test shards were still in progress when I read it. Review findings from Correctness review (read at head
Verified: |
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> # Conflicts: # src/pkg/dashboard/stage_tracker_activity.test.js
There was a problem hiding this comment.
Correctness — delta re-review at f8c15ea (merge of origin/v5 into the PR branch; the PR's own +/- content vs merge-base is byte-identical to the previously approved head except where it overlaps upstream).
Merge interaction breaks CI (in-scope, concrete fix). Upstream bc23607 (#16644 / #16643) added stage_tracker_remediation_chevron.test.js:295-306, which asserts a stale-polled missing CI track still says stale · <ISO> in its tooltip. This PR's new freshness string in static-src/js/tracker-row-strip.js:317 only emits · stale when stale && moving, and missing is in settledState (line 278), so moving is false and the word is dropped. At head, test (dashboard 1/4), test (dashboard shuffle 2/3), runtime-tests and therefore build-and-test all fail on exactly that assertion (not ok 14 - a stalled or stale-polled text-bearing track is plainly active, error the poll age rides the tooltip).
Fix: drop the && moving qualifier — (stale ? ' · stale' : '') — in tracker-row-strip.js:317, then make dashboard-gen to regenerate static/index.html. The PR's own tests (stage_tracker_activity.test.js:565, :954) only match observed Nm ago, so they remain green; settled verdicts still persist via known/active (lines 280-281), which is the behaviour #16617 asks for.
Other overlap (stage_tracker_motion_test.go: upstream runNodeStdin replaced by the PR's file-based runMotionScript, lines 697-710) is equivalent and runNodeStdin keeps other callers. dashboard-gen-check passes at head; go build of dashboard/github/cmd packages succeeds.
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
…k tooltip Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
There was a problem hiding this comment.
The Go side looks correct to me; the red test (dashboard 1/4) / dashboard shuffle 2/3 jobs are a real contract conflict with #16643 in the JS, not flake.
src/pkg/dashboard/static-src/js/tracker-row-strip.js:278, :317 — a stale poll of a missing CI track no longer says stale in the tooltip, which stage_tracker_remediation_chevron.test.js:295-304 (from #16643) pins. :278 puts 'missing' in settledState, so moving is false; :317 only appends ' · stale' when stale && moving. The #16643 fixture is a CI track in state: 'missing' with blocker: 'workflow-approval-required:1', observed hours ago, and asserts title="[^"]*stale · 2026-11-04T01:00:00\.000Z ("the poll age rides the tooltip"). With this PR the title reads observed 19h ago · 2026-11-04T01:00:00.000Z — the age is there but the word stale and the stale · <iso> form are gone, so TestStageTrackerRemediationJavaScript fails (stage_tracker_remediation_test.go:223). Severity: medium — blocks merge.
Two ways to reconcile, and it is a product call which: (a) treat missing as motion-pending rather than settled (a track waiting on workflow approval is waiting for something, which is the #16643 reading), or (b) keep missing settled but still emit stale · <iso> when stale regardless of moving, and update the #16643 assertion to the new observed Nh ago wording. (b) keeps the "settled verdicts persist" intent of this PR for CI passing/failing while not hiding a stale poll on a stalled track.
Verified on the Go side (read at the PR head):
trackerReviewPerspectives(stage_tracker_activity.go:502-531) collapsestrackerLatestBundleVerdicts(already latest perperspective\x00backend\x00model,:481-500) to one per perspective with dissent-wins ordering: non-approve replaces approve, same-class newer replaces older, approve never replaces earlier dissent — matches the comment and thepreserve dissentgoal.Perspectivesonly populates when the bundle head matches the PR head (:607-609), so a stale bundle cannot attach old verdicts to a new head.- The caption suffix (
:734-747) is only added forchanges_requested/human_decisionand breaks after the first non-approve perspective — bounded. runMotionScriptconsolidation instage_tracker_motion_test.gois a no-behaviour-change refactor (all five fixtures now go through a temp file, which also removes the latent E2BIG risk on the-eones).
— hive: agent=reviewer-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
… wording
Document review_activity.perspectives[] in the API reference and the
settled-verdict ageing rule in the activity monitor guide. The row tooltip
track summary now ages settled verdicts the way the chevron tooltip does
("observed 15m ago · stale") instead of "stale observation". Extract
trackerPrefersVerdict so both verdict pickers share the dissent-over-
approval rule, and move the failing-perspective caption suffix into
trackerPerspectiveFindingLine with a cross-reference to its JS mirror.
Hive-Fix-Attempt: 2/6
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
TestIndexHTMLIsGenerated failed because the last two commits edited static-src/ without re-running the concat generator. Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Conclusion: the fix is correct and matches #16617, but the head is not mergeable as-is: src/pkg/dashboard/static/index.html was not regenerated after the last two commits, which is the single root cause of every red check in v2 Tests.
Verified (read at 1c4a9bf, base v5):
test (rest 3/4)fails inpkg/dashboard/webstatic→TestIndexHTMLIsGenerated(indexgen_test.go:28: "static/index.html differs from the concatenation of static-src/ (first difference at line 18218)").testandcoverageare red only because that shard failed. Reproduced locally:go run ./pkg/dashboard/webstatic/gen -src pkg/dashboard/static-src -out pkg/dashboard/static/index.html -checkexits 1 on this head and 0 onv5.- The drift is not cosmetic. Diffing the generated output against the committed file:
index.htmlis missing (a) the three "Mirrors trackerPerspectiveFindingLine…" comment lines (static-src/js/tracker-row-strip.js:283-285), and (b) the whole second hunk from commit836c9fe/1c4a9bf—static/index.html:18491still renders' · stale observation'whilestatic-src/js/tracker-row-strip.js:581-585renders' · observed Nm ago · stale'. So the spoke would ship the old current-stage wording thatstage_tracker_activity.test.js:954(/observed 15m ago/) says was fixed. - Fix:
make dashboard-genand commitindex.htmlalongside. Nothing else in the diff needs to move.
Logic, read against the tree:
stage_tracker_activity.go:483-489trackerPrefersVerdictkeeps the previous tie-break exactly (dissent beats approve on equalRecordedAt; otherwise unchanged), andtrackerReviewPerspectives(:515-543) collapsing per-perspective across backend/model with dissent-wins is what the issue asks for. Sorted names give a deterministic first finding on both sides.- Caption/tooltip dedupe is a true mirror:
trackerPerspectiveFindingLine(:616-630) andperspectiveLine(tracker-row-strip.js:286-287) build byte-identical strings forchanges_requested/reject/requires_human, so thecaption.includes(...)guard at:291does suppress the double append. The JS suite passes locally (40/40). settledStategating (tracker-row-strip.js:278-282) only lets verdict states outlive the 2-minute window;running/reviewingstill fall through toactivity unknown · last observed …, which is the behaviour #16617 wants preserved.
Not blocking, just noting: the tooltip lists the first finding twice (once appended to the caption, once in the per-perspective lines — stage_tracker_activity.test.js:915 pins this shape), which is fine if intentional.
The docker (*) and overlayfs-exec-guard failures are outside this diff; I did not triage them.
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
…, never "activity unknown" (#16662) A Review track with no verdict and no reviewer in flight now derives a concrete statement from the PR's lifecycle and its CI track (stage_tracker_review_fallback.go): closed/merged, draft, waiting for CI, not started (+ who is next), or — only when nothing was fetched for the head — "no data for head <sha7> yet · refreshing". The UI drops the "activity unknown · last observed …" prefix: settled verdicts stay known as observations age (dated in the tooltip), and a settled review explains itself — who decided, how long ago, each perspective's verdict with its first finding (perspectives field, dissent preserved). An aggregate approval the verified bundle does not back reads "evidence not verified" instead of unknown. Fixture invariant: no CI/Review tooltip contains "activity unknown" / "CI unknown" / "Review unknown". Supersedes #16634. Fixes #16617 Fixes #16661 Signed-off-by: Andrew Anderson <andy@clubanderson.com> Co-authored-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
|
Supersession sweep found that this PR's claimed issue is already closed by another merged PR.
Closing this hive-authored PR as superseded by #16662: every issue it closes is already closed and that merged PR closed #16617. Head 3d98d74 conflicts with This comment is edited in place by the supersession sweep; it is not duplicated. |
Supersedes #16619 (adopted in place, re-homed because a hive commit on the contributor branch cannot carry a squash-safe sign-off; original commits by @thatguywob preserved).
Closes #16617
This branch is #16619's head
598099676c55as pushed by @thatguywob (forkthatguywob/hive), unchanged. Last lane reason on the original:reviewer-verdict:changes_requested.Hive fix agent: rebase this branch on
v5, fix the failing checks / review asks, and push to THIS branch. Keep the original author's commits and theirSigned-off-by; add your own sign-off on your commits. Do not open another PR.Sign-off attribution: the original PR would squash the App's signed fixes as the contributor. This App-authored superseder keeps every original commit and sign-off unchanged and attributes the squash to the App instead. DCO and all merge gates still apply here.