Repository navigation
π dashboard: Needs human chip counts only human-gated rows; Refs-only PRs address, not cover, an issue - #15013
Conversation
β¦Rs address, not cover, an issue Bug 1: the Activity Monitor's Needs human chip was label-based (needs_human_labels), so needs-signal / hold rows β a 040 'PR in progress' issue, held auto-merge PRs β counted. One predicate (stageTrackerNeedsHuman Go, stageTrackerItemNeedsHuman JS) now feeds the chip count, the filter and new needs_human / on_hold payload fields; held rows get an On hold chip. Bug 2: #14976 (shared-CI incident) showed 040 'open PR covers it' because the relay covered the declared issue of #14990 despite its explicit non-closing 'Refs'. Reference-only claims are carried as ClaimFact.Partial, recorded via FactSink.AddressedBy, and read 'addresses part of it'. Bare cross-references (incident deferral comments) never create coverage. Fixes #15008 Signed-off-by: Andrew 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
|
|
@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. |
1 similar comment
|
@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. |
|
@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. |
β¦n predicate Refs #15008 Signed-off-by: Andrew Anderson <andy@clubanderson.com>
There was a problem hiding this comment.
Read the core Go changes against v5 and the callers; I did not find a correctness defect. Noting what I verified so a maintainer can skip it.
Bug 2 (partial vs. covered)
prCoveredIssues(src/pkg/github/issue_inprogress.go) now splits declared issues: a declaredNthatParseReferencedIssuessees andclosingIssueRefsdoes not βpartial; anything closing βcovered. Both parsers already restrict to the same repo (prclaims.go:545-580keys ondefaultRepo;pr_duplicate_guard.go:54filters onowner/repo), so "Refs #N" and "Closes #N" in the same body resolve to covered, which is the right precedence.pr_request_watcher.go:680-682wires both halves.FactSinkgainsAddressedBy; the only two implementers in the tree (stageengine/facts_sink.go:158,stageenginetest/fake.go:207) are both updated in this PR, and there are no test-local sink implementations that would stop compiling.issueStage(derive_issue.go) keeps 040 placement and only changes the reason string; nothing else inpkg/keys on the literal"open PR covers it"besideshumanizeStageReason, which this PR extends withReasonIsPartialCoverbefore the existing switch.enrichCoveringPRonly markspartialwhen the ledger claim'sPRNumberequals the remembered PR β so a stale ledger entry for a different PR cannot flip the flag.
Bug 1 (Needs human chip)
stageTrackerNeedsHumanis hold-first, then mods, then stages, then 071/072 gated onauthorKindHumanβ matches the doc and the JS fixture cases inlabel_facets.test.js. All referenced mod/author constants exist atstage_tracker.go:481-516.
The PR body's narrative matches the diff; I found no claimed behaviour the code does not implement.
Not a defect, just flagging for whoever merges: the PR carries conflicts-with-open-pr and stage/071-needs-author, and most checks are still pending on head 65262a2, so this is waiting on a rebase/CI rather than on review.
β hive: agent=reviewer-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
There was a problem hiding this comment.
Read the tree at head d333c3e (not just the diff). Looks correct to me; nothing I could verify as a defect.
What I checked
FactSinkgainsAddressedBy(src/pkg/stageengine/facts_sink.go:21). The only production implementer is*factSink(:174), and the test double instageenginetest/fake.gois updated too βgo build ./...on the head is clean, so no third implementer was missed.prCoveredIssues(src/pkg/github/issue_inprogress.go:213-243): a declared issue goes topartialonly whenParseReferencedIssuessaw it andclosingIssueRefsdid not; a bareRefs #Nin the body that was never declared lands in neither list, which matches the doc comment ("a bare mention β¦ is not in the body's claims at all") and the earlierParseReferencedIssuesdefault-repo behaviour atprclaims.go:567-570. Closing refs still always win intocovered.enrichCoveringPR(stage_facts_sink.go:193-203) only markspartialwhen the ledger claim'sPRNumberequals the PR the store remembers, so a stale ledger entry for a different PR cannot flip the caption.BackfillCoveredByLedgeruses the sameclaim.Reference && !claim.Closingtest, consistent with theIssueClaimfield semantics atprclaims.go:195-217.issueStage(derive_issue.go:153-166):claimisis.Claimfrom:134, soclaim.Partialis nil-safe; the 040 placement is unchanged and only the reason string differs, as the body says.- Go
stageTrackerNeedsHuman(stage_tracker.go:1736-1746) and JSstageTrackerItemNeedsHuman(static/index.html:15194-15201) agree: hold short-circuits first, then modsneeds-human/needs-decision/unauthorized, then stages{15,45,85,90}, then{71,72}gated onauthor_kind == human. The numeric constants matchstagelabels.go:263-277.
Notes, not findings
build-and-testwas red on the previous run and is re-running as I write; I did not chase the earlier failure.- The PR carries
conflicts-with-open-pr;static/index.htmlis also touched by the author's #15014/#15016. GitHub still reports it mergeable, so this is a sequencing note for whoever merges, not an objection.
β hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
|
merged by Hive trusted-author auto-merge: author @clubanderson holds role owner and GitHub permission admin on repo; CI green at 65262a2. |
|
Thank you for your contribution! Your PR has been merged. We'd love to hear how your experience was: share feedback |
Fixes #15008
Seen on hive's own hosted spoke (2026-10-28 16:20Z) with the Needs human 18 chip active.
Bug 1 β "Needs human" chip counted rows nobody waits on
The chip was label-based (
needs_human_labels= the Projects preset set), soneeds-signal(#14976 at 040 "PR in progress") andhold(#7563/#14939 ready-for-auto-merge, #2364/#9699/#9701/#9735) rows matched.stageTrackerNeedsHumanβ JSstageTrackerItemNeedsHuman. Needs human = a person's action is the only thing that advances the row: stages 015, 045, 085, 090; 071/072 whenauthor_kind == human; or aneeds-human/needs-decision/unauthorizedmodifier.holdwins β never Needs human. Not: stalled, blocked, handled regression, 040 covered, auto-mergeable.needs_human/on_holdtotals and per-itemneeds_human/on_holdflags, from the same predicate.onHold); one-line tooltips on both chips; the dialog's covered-label list is gone.needs_human_labelsstays in the payload for the Projects preset.Note: 090 (ready-for-human-merge) and 072 (returned to a human contributor) are included beyond the owner's list β only a human can advance either.
Bug 2 β #14976 (shared-CI incident) at 040 "open PR covers it"
Root cause differs from the thesis. Live tracker history: 030 β 040 at 12:16:38 ET, linked PR #14990 (scanner fix, body
Refs #14976 (needs-human: β¦), no closing keyword). The relaypr_request_watcher β clearInProgressForPR(prCoveredIssues(...))covered the declared issue regardless of the explicit non-closingRefs; the sweep'sReferenceclaim then stated the same reason. Deferral comments on #14939/#14974 were not the cause: no code path derives coverage from timeline cross-references, andParseClaimedIssues/ParseReferencedIssuesignore "Deferred to shared-CI incident #N" (regression test).ClaimFact.Partial;FactSink.AddressedBy;prCoveredIssuesβ(covered, partial); relay callsaddressInProgressForPRfor partial;enrichCoveringPR/ledgerCoveringPR/BackfillCoveredByLedgerthread partial.Follow-up (not here):
[shared-ci]incident issues are detected only by title prefix; no label/marker exists for a dedicated "incident" presentation.Tests
go test ./pkg/dashboard ./pkg/github/... ./pkg/stageengine/... ./pkg/stagelabels/...green (node renderer tests included). New:TestStageTrackerNeedsHuman,TestPRCoveredIssues_DeclaredRefsIsPartialNotCovered,TestPRRequestWatcher_DeclaredRefsIssueIsAddressedNotCovered, partial cases in deriver/labels/backfill tests, rewritten needs-human node fixture.Docs:
dashboard-sections.md(filter chips),labels-and-control-signals.md(040 row),changelog.d/fixed-15008-*.md.