Skip to content

🌱 Forward-merge v5 into v6 (82 commits, manual: conflicts) - #16997

Merged
clubanderson merged 84 commits into
v6from
topup/v5-to-v6-20261107
Nov 7, 2026
Merged

clubanderson merged 84 commits into
v6from
topup/v5-to-v6-20261107

Conversation

@clubanderson

Copy link
Copy Markdown
Member

Manual forward-merge of v5 into v6 (82 commits; merge base 4b4177cea, v5 tip 9a485a7f9). The automated v6-topup.yml run has refused to open a PR since 2026-11-06 because git merge v5 conflicts. 62 files conflicted; each was resolved by hand under the standing top-up rules (v5 wins unless the v6 side is intentional v6-only restructuring; port v5 behaviour into v6's structure rather than resurrecting v5 files; keep DashboardMatrixConfig; discord/ stays deleted; 7-tile grid; v6 dispatch relocation + v5 fixQueue both kept; the two auto-resolve patterns in the workflow file).

Part of #7563

Conflicts and resolutions

Area Files Resolution
gh wrapper bin/gh-wrapper.sh, bin/gh-wrapper.test.sh Took v5 (#16759 maintainer-approved gh pr list --state merged allowance) over v6 #16686 narrowing — judgement call.
cmd/hive hold_guard_mergeability_test.go, main_helpers.go, main_sweeps.go, dependency_bot_* Union; v6 dispatch relocation kept, v5 fixQueue / dependency-bot helpers kept.
automerge: update-branch automerge_update_branch*.go Union of updateBranchAndSkip with pr *gh.PullRequest.
automerge: workflow-fence approval carry automerge_contributor_workflow_approval.go (+test), automerge_review_gate.go, automerge_review_head_binding_test.go, automerge_human_changes_requested.go Unified v6 reach-based review carry (reviewReachesHead) with v5 #16858 BaseRefreshChecker comment/marker carry into one workflowApproval{Login,Typed,AnswersVerdict,CarriedFrom}; latestTrustedHumanApprovalReach returns (login, carriedFrom, ok); audit marker gets from=<sha> whenever carried by either mechanism. v5 wording ported into stagelabels/human.go, prstage/causes.go, pr_approve_command.go; decisionActions stays removed on v6 (stagelabels.DecisionActions).
automerge: fork guard automerge_stage_sharedci.go, automerge_stage_labels.go, prstage/types.go, automerge_fork_guard_test.go Both branches fixed utah-packages#862/#869 independently (v5 #16845 foldForkGuard digest check, v6 #16814 carriedForkGuard by lane reason). Kept both: a carried guard is flagged ForkGuardCarried and verified against the head's check digest in foldForkGuard; dropped when another check is red, so a real failure is still nudged/adopted. v5 test adapted to v6's forkPR fixture and the #14306 "never re-home a fork" rule.
dashboard: Let-hive-decide plan failures api_repo_item_decide_delegated_plan.go, delegated_gather_error.go (+tests), tracker-actions.js, index.html Union of v5 #16782 structured {ok,error,reason,retryable,attempts,elapsed_ms,hive,hint} + retries and v6 #16951 429/Retry-After. Rate-limit is the first classification case (429, reason: github, not retried); delegatedGatherError writer replaced by writeDelegatedPlanFailure.
dashboard: config JS config-helpers.js, index.html Both sides (v6 aria-checked + v5 merge-train stall-warn dirty tracking); index.html regenerated via webstatic/gen.
config pkg/config/config.go → governor_config.go, dashboard_config.go, github_api_budget.go config.go kept at v6 (workflow auto-resolve rule); v5 GovernorCIConfig, merge_train_stall_warn_s explicit-zero handling ported into v6's split files.
audit audit.go v6 realigned map + v5 config_governor row.
policies / kick message pkg/policies/defaults/quality*.md, src/policies/quality*.md, scheduler/kickmessage.go governorWorkSourceRule = v5 merged-history allowance + v6 --comments/hive-gh compact form; the five quality templates carry the identical sentence (otherwise addGovernorWorkSourceRule appends a second copy).
blocked sweep pkg/github/issue_blocked*.go Union; v5 TestReconcileBlockedIssuesBlockedOnIssue adapted to v6's marker-less → needs-human ask behaviour.
docs docs/api-reference/*.md (9 files), operator-reference/README.md, hive-merge.md, scale-envelope.md, agent-configuration.md, labels-and-control-signals/{reference,stage-labels}.md, dashboard-sections/activity-monitor.md, github-cli-compact.md v6 text + v5 additions (workflow-fence carry, gated-run approval sweep #16603, dependency-bot behind, blocked-on-issue #16852, plan failure body). API-reference tables: ours rows + v5-only rows, citations re-resolved with check-api-reference-citations.sh --fix.
misc kickmessage.go, issue_blocked_test.go, pr_decision_base_refresh_test.go, notify, delegatedecision, fleetreport, hub, review, crux/* Union / both-appended.

Verification (local)

  • gofmt -l clean (two pre-existing v6 files unchanged), go build ./..., go vet ./..., golangci-lint run ./... → 0 issues.
  • go test -cover on touched packages: config 90.3, delegatedecision 92.8, fleetreport 93.0, hub 89.0, notify 94.7, review 92.3, stageengine 92.3, prstage 94.7, stagelabels 91.6, cmd/hive 75.0; dashboard / github / automerge / policies / scheduler re-run after the test adaptations above (CI shards confirm).
  • bin/gh-wrapper.test.sh, bin/contributor-agent.test.sh running.

Ignore probot DCO / prow dco / native-queue-policy: they always fail on forward-merges carrying others' commits.

hivecommons-hive Bot and others added 30 commits November 5, 2026 08:19
A pack agent moved onto an inference gateway had its model cleared so the
gateway's default_model would apply. When the gateway has none (the operator
picked a model per agent on the dashboard), the agent launched with no model,
crash-looped and was paused by the watchdog. It now takes the model the
operator's agents on that gateway run, and a crash-loop pause on an agent
this repairs is resumed.

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: MikeSpreitzer <14296719+MikeSpreitzer@users.noreply.github.com>
* fix(packs): clarify gateway model reconciliation logs

Hive-Run: #16732
Hive-Plan: gateway model reconciliation review
Hive-Spec: issue-16732#gateway-model-log
Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>

* fix: document gateway model log change

Hive-Run: #16732
Hive-Plan: gateway model reconciliation logs
Hive-Spec: gateway-model-logging#changelog
Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>

---------

Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: Karibhou <Karibhou@users.noreply.github.com>
Hive-Run: #16733
Hive-Plan: gateway model fallback
Hive-Spec: gateway-model-reconciliation#override-loss

Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: Karibhou <Karibhou@users.noreply.github.com>
Hive-Run: #16737
Hive-Plan: docs-currency
Hive-Spec: repo-acmm-onboarding#readiness

Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Hive-Run: #16742
Hive-Plan: citation drift correction
Hive-Spec: repo-acmm-onboarding#hive-wide-readiness

Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: Karibhou <Karibhou@users.noreply.github.com>
…ntime error (#16749)

* 🐛 fix(ioscan): do not fleet-report an unconfigured classifier as a runtime error

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* 🐛 fix(ioscan): address review findings on unconfigured-classifier WARN

- Add isNoEndpointErr helper so fail-closed and fail-open share the
  no-endpoint special-case instead of duplicating the substring check.
- Add classifierFixHintOpen so the fail-open branches no longer recommend
  fail_mode: open, which is already in effect there.
- Update fleet-report.md: the fail-open no-endpoint case now logs
  "ioscan classifier not configured" and is not fleet-reported.

Hive-Fix-Attempt: 1/6

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

---------

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…16719)

Keep governor-supplied work discovery restrictions while describing merged-only history reads as conditional on wrapper and tier permissions. Add regression coverage for the injected header and remove its categorical proxy-block claim.

Hive-Run: #16714
Hive-Plan: kick-header-work-source-rule
Hive-Spec: issue-16714#governorWorkSourceRule

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
Co-authored-by: thatguywob <thatguywob@users.noreply.github.com>
* 🐛 fix: keep superseder decision questions parked

Preserve a sign-off-only superseder park and show the reviewer's question
instead of silently approving it through the automatic superseder rule.

Hive-Run: #16679
Hive-Plan: superseder-decision-question
Hive-Spec: issue-16679#decision-question
Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>

* 🐛 fix: preserve scope auto-decisions for hive superseders

Classify recorded reviewer questions with the existing review classifier and keep the superseder parked only for policy sign-off questions. Read every changed-file page, including rename origins, and fail the delegated run if the classification cannot read the files.

Cover scope approval, policy sign-off recommendations, hard-stop precedence, workflow and config policy edits, pagination, renames, and file-read failures.

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>

---------

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
Co-authored-by: thatguywob <thatguywob@users.noreply.github.com>
Hive-Run: #16746
Hive-Plan: correct-handlePackSetLevel-citation
Hive-Spec: #16746#finding

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
Co-authored-by: thatguywob <thatguywob@users.noreply.github.com>
* 🐛 fix(packs): retain gateway placement after override loss

Use persisted pack gateway placements as fallback votes when runtime overrides disappear and no operator backend picks remain. Carry their copied model to agents added at later ACMM levels while preserving operator choices and gateway defaults.

Fixes #16740

Hive-Run: #16740
Hive-Plan: preserve gateway placement after override loss
Hive-Spec: #16740#gateway-placement
Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>

* 🐛 fix(packs): wire named gateway in placement regression

Register the gateway backend checker in the named-gateway regression setup, matching production validation before setting overrides. Document how retained pack-owned gateway placements count after operator overrides disappear.

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>

* 🐛 fix(packs): use host interface for gateway test wiring

Access the gateway backend checker through crux.Host, since the dashboard dependency exposes the narrower crux.Runtime interface.

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>

---------

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
Co-authored-by: thatguywob <thatguywob@users.noreply.github.com>
Hive-Run: #16750
Hive-Plan: correct-ioscan-fleet-report-documentation
Hive-Spec: #16750#ioscan-classifier-unavailable

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
Co-authored-by: thatguywob <thatguywob@users.noreply.github.com>
Honor usable Retry-After headers after HTTP 429 and suppress ntfy delivery until the deadline. Without a usable header, back off from one minute to one hour and reset after successful delivery. Keep other notification channels independent.

Hive-Run: #16689
Hive-Plan: ntfy-rate-limit-recovery
Hive-Spec: issue-16689#evidence

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
Co-authored-by: thatguywob <thatguywob@users.noreply.github.com>
* ✨ feat(settings): expose merge-train controls in Auto-merge

Expose nested merge-train controls with owner-only partial API updates and persistence-aware explicit-zero stall warnings.

Hive-Run: #16677
Hive-Plan: merge-train-settings
Hive-Spec: #16677#proposed
Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
Co-authored-by: clubanderson <407614+clubanderson@users.noreply.github.com>

* 📖 docs: refresh dashboard API source citations

Resync the 201 route citations shifted by the new merge-train PATCH registrations using the repository citation fixer.

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>

* 📖 docs: publish merge-train PATCH API contracts

Document the registered auto-merge and governor PATCH routes in OpenAPI and the API reference, including merge-train validation and effective responses. Fixes the CI route parity failure.

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>

---------

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
Co-authored-by: thatguywob <thatguywob@users.noreply.github.com>
Co-authored-by: clubanderson <407614+clubanderson@users.noreply.github.com>
StateIdle is published before runSandboxKick finishes its deferred disk writes. Join the real sandbox run persister in test cleanup so temporary work-dir removal cannot race the remaining writes, including when assertions fail.

Hive-Run: #16724
Hive-Plan: stabilize-review-spool-registration
Hive-Spec: issue-16724#ci-flake

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
Co-authored-by: thatguywob <thatguywob@users.noreply.github.com>
…e loads (#16755)

* 🐛 fix: decision pills no longer flash disabled while the viewer's role loads

On a hard reload the stage tracker's first paint ran before the auth
fetch had named the viewer, so stageTrackerActionCtx read the missing
role as 'read' and every role-gated decision option rendered
aria-disabled with a sign-in reason until the next tracker poll.

- stageTrackerAuthPending: unknown is not read. While auth is pending the
  ctx is optimistic (user/canHuman true, canHold null), so the pills
  render enabled and a click that beats auth gets the server's 403.
- showGHAuthState marks auth known on both paths (and checkGHAuth when
  both endpoints fail) and repaints the tracker once when the user/role
  actually changed.
- Logged-out viewers still see the disabled pills with the sign-in
  reason once auth is known.

Fixes #16754

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

* test: selection-bar assertion checks the role-gated chips, not row state

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

---------

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
…await a verdict (#16757)

The stage engine files every base-refresh reason under 050, and the 050
caption and next-move line are composed from the stage alone, before the
review track for the head is attached. A train member with an approved
head therefore read "waiting for a verdict on the current head" next to
a Review 5/5 approved chevron. Once the review track is known, rewrite
the queued-behind 050 row to "Approved N/M · queued for base refresh
behind PR #<front>"; reviewing heads and the 060 wording are unchanged.

Fixes #16756

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
* 📖 docs: clarify permitted merged PR history reads

Keep quality work discovery scoped to the governor-supplied queue while documenting merged-only history reads conditional on wrapper and tier permissions.

Hive-Run: #16700
Hive-Plan: docs-currency-quality-prompt
Hive-Spec: #16700#merged-pr-history
Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>

* 📖 docs: align quality variants with merged history guidance

Keep every quality mode consistent with the conditional history guidance and cover both source and embedded copies in the regression.

Hive-Run: #16715
Hive-Plan: quality-policy-variant-parity
Hive-Spec: #16715#quality-policy-variants
Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>

---------

Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
Co-authored-by: thatguywob <thatguywob@users.noreply.github.com>
Co-authored-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
The capture walk's depth and entry budgets, the newest-wins ordering in
consider, and the id checks that keep a transcript-read id shell-safe
(over-length, non-hex codex uuids, unterminated goose session.id,
dot-prefixed names) had no test reaching them; neither had omp's
-tmp-<relative> and bare-root session buckets. These are the guards that
stop a large or hostile session tree from stalling the PR-opened hook or
smuggling metacharacters into a resume command, so each now has a test
that fails without it.

Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…16760)

* fix(github): post the prow/lgtm warning once with a live context

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* fix(github): keep the prow/lgtm warning on the caller's cancellation

The detached warning context ignored lifecycle cancellation, so a hung
comment bridge held the operator queue for the full warning timeout. Only
the deadline is renewed now; a withdrawn caller posts nothing. The fake
API gains the /user route the marker lookup needs, and the scale-envelope
citation is re-anchored.

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* fix(github): follow only the caller's cancellation for the prow/lgtm warning

The AfterFunc tied the warning context to any Done() of the caller, so a
caller whose deadline expired mid-warning still cancelled the lookup and
post, reproducing the deadline-exceeded error from #16753 with no marker
left behind. Cancel the warning context only on context.Canceled, and
cover a live-but-short caller deadline against a slow comment handler.

Hive-Fix-Attempt: 1/6
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* docs(scale-envelope): re-anchor the RankActionableIssues citation

The client.go edit shifted the function; the docs-citations check anchors on
AgeMinutes inside the cited range.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

---------

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The ⋯ decision Next sentences, the Approve-and-let-hive-merge consequence
and the GitHub approval note used the lane's internal word "sweep"; they
now say "next merge pass" per #15575. Adds a ratchet
test over the decision result/option/note builders.

Fixes #16767

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
…16771)

* ✨ feat: PR in progress chevron names the actor with their kind icon

A lit PR in progress chevron reads "PR in progress · @login" with the
actor's kind glyph (hive / clanker / person), the way "Waiting for @x"
names its person. The recorded holder leads, then stage_actor.login,
then the PR's author; the name rides inside the clamped chevron copy so
it truncates like every holder, the icon sits outside it, the tooltip
carries the full text, and nobody known keeps the bare word.

Fixes #16770

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

* fix: PR in progress actor never names the issue's own author

An issue drawn through a covering PR with no author fell back to the
issue's author via stageTrackerReturnedAuthor; the helper now resolves
the one PR drawn here itself and keeps the bare word when it has none.

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

---------

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
Hive-Run: #16763
Hive-Plan: community-meeting-links
Hive-Spec: issue-16763#proposed-action

Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
…16736)

* feat(github): auto-approve gated workflow runs for trusted actors

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: clubanderson <407614+clubanderson@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix(github): bind workflow-run approval admission to the approved head

The sweep trusted a PR for the automerge label alone, which is bound to
no head and names no approver. Read the stage record's head-bound
admission (dashboard Approve by a verified owner or merger, #15136)
instead; a label on its own never approves a run. Move the ledger
mutations into markDeniedCounted/claim so the once-per-head invariant
lives on the ledger type. Document that the default-on sweep needs the
App's Actions: Read and write and how it relates to /ok-to-test.

Hive-Fix-Attempt: 1/6
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* docs(github): state the App-actor trust boundary for gated-run approval

Document why trusting the App as a run's triggering actor cannot approve
contributor code: GitHub gates only fork pull_request runs and the App
cannot push to a fork, so an App-pushed same-repository branch is never
gated. Code comment on workflowApprovalTrusted and the operator reference.

The default of governor.ci.auto_approve_workflows stays as #16603
specifies (on); flipping it is the owner's call, not a code defect.

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>

* docs(github): state which approval paths reach fork heads

github-app-setup.md still said Hive never approves fork heads, which the
gated-run approval sweep contradicts: GitHub holds only fork pull_request
runs behind the gate, so the sweep (and the registry-trusted contributor
lane) are exactly the paths that approve fork heads with the App token.
Say so, keep the sentinel-alert exclusion, note the same second exception
in the operator-reference automerge bullet, and document that a trusted
author covers runs a fork collaborator pushed to the head.

Hive-Fix-Attempt: 3/6
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* docs(github): name the approval sweep as the primary cure for held runs

operator-reference "Held and never-started required checks" now leads with
the default-on gated-run approval sweep (App token, Actions: Read and write)
and describes /ok-to-test as the fallback that never reaches fork heads;
hive-merge.md no longer claims the hive never approves a human-authored
lane's held runs, since dashboard-admitted heads are approved by the sweep.

Hive-Fix-Attempt: 4/6

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* docs(github): state that a recorded approval refusal is not retried on the same head

Spell out the operational consequence of the once-per-head ledger in the
operator reference and App setup guide: granting Actions: write after a
403 was recorded does not retry that head; a new push, a restart, or
Approve and run on GitHub is needed.

Hive-Fix-Attempt: 5/6
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

* docs: align operator docs with the workflow auto-approval sweep

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

---------

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: clubanderson <407614+clubanderson@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…goes offline (#16776)

* 🐛 fix(stream): watchdog restarts ffmpeg and the pod when the channel goes offline

ffmpeg can sit at speed=1x for hours after Twitch drops the ingest session,
with the pod 1/1 Running and the channel offline (hivecommons, 2026-11-04
and 2026-11-05). Move the stall watchdog into stream-watchdog.sh and teach
it to ask Twitch whether TWITCH_CHANNEL_LOGIN is live (public GQL
Client-Id, or Helix with app creds): N misses outside a grace window
restart ffmpeg with exponential backoff, M relaunches still offline exit
the supervisor so Kubernetes restarts the container. A heartbeat touched
only while frames advance and the channel is live drives healthcheck.sh;
probe windows now cover the ~2 min key lock Twitch holds after a publisher
drops, which today's startup probe tripped on. Fleet, base and matrix get
TWITCH_CHANNEL_LOGIN and a 15 s terminationGracePeriodSeconds.

Fixes #16775

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

* 🌱 stream: silence SC2329 on the trap handler so the new shellcheck step passes

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

* 🌱 stream: resync PR head after GitHub push outage

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

* 🌱 stream: re-pin the hivecommons stream config hashes for the watchdog change

The pin guards against the fleet work touching twitch.tv/hivecommons by
accident; this PR changes it on purpose (TWITCH_CHANNEL_LOGIN, probe
windows, terminationGracePeriodSeconds, compose passthrough).

Signed-off-by: Andrew Anderson <andy@clubanderson.com>

---------

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
…of "hive unavailable" (#16782)

The preview answered every failure as 502 — a PR merged since the row
loaded, an intermittent GitHub 500 on check-runs — and the hosted
Ingress intercepts 502/503 and swaps the body for its generic outage
page, so the owner read "hive unavailable" for a hive that was up.

- retry transient GitHub failures twice with backoff before giving up
- answer not-open as 409 and unrecoverable reads as 424, never 502/503,
  with {reason, retryable, attempts, elapsed_ms, hive, hint}
- log a Warn with the real error on every failed plan
- the dialog names the hive, explains the failure (merged since the row
  loaded / GitHub failed after N attempts / no GitHub client / gateway
  502), hides Retry when it cannot help, and says Let hive do that
  stays off until a preview loads

Refs #16776

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
* docs: correct Prow LGTM error comment

Hive-Run: #16761
Hive-Plan: correct bindProwLgtm documentation
Hive-Spec: issue-16761#reporting-error
Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>

* docs: add changelog for Prow LGTM comment

Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>

---------

Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
…es transient POST failures, and distrusts fork heads on the actor path (#16783)

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
…the check digest (#16781)

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Hive-Run: #16690
Hive-Plan: phone navigation label
Hive-Spec: issue-16690#phone-navigation-label

Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
* fix: classify REST reserve shedding as rate limiting

Hive-Run: #16784
Hive-Plan: issue-activity-reserve-classification
Hive-Spec: fleet-report#unclassified-error-repeated
Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>

* chore: add changelog fragment for #16784

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

---------

Signed-off-by: Karibhou <Karibhou@users.noreply.github.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Co-authored-by: Karibhou <Karibhou@users.noreply.github.com>
Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
@kubestellar-prow kubestellar-prow Bot added the dco-signoff: yes Indicates the PR's author has signed the DCO. label Nov 7, 2026
@kubestellar-prow

Copy link
Copy Markdown
Contributor

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:
Once this PR has been reviewed and has the lgtm label, please assign clubanderson for approval. For more information see the Code Review Process.

The full list of commands accepted by this bot can be found here.

Details Needs approval from an approver in each of these files:

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

@kubestellar-prow kubestellar-prow Bot added the size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files. label Nov 7, 2026
@github-actions

github-actions Bot commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor

Open PR overlap check

Base branch: v6

Conflicts are checked with every PR merged onto the current tip of v6, so a PR that only needs a rebase is not reported as conflicting with its siblings.

PR Title Relation Files
#16993 🐛 fix(automerge): a maintainer keeps role/trusted when the hive merges main into their approved PR same files src/pkg/github/automerge/automerge_contributor_lane.go
#17001 🐛 fix(github): the overlap duplicate guard defers only to hive agents' PRs and tells the blocked issue same files src/docs/notifications.md
#17000 🐛 fix(adoption): take a review-asked fix the undo guard refused once the reviewer approves it same files src/docs/labels-and-control-signals/stage-labels.md
src/pkg/github/automerge/automerge_contributor_adopt.go
src/pkg/github/automerge/automerge_stage_labels.go
src/pkg/stagelabels/store.go
#16994 🐛 fix(review): A review a rate limit kept from reading the PR is no verdict and is asked again, not a human decision same files src/docs/review-swarm.md
#17002 🐛 fix(escalation): a red PR triaged as infrastructure is re-run once instead of sent back to fixers until it escalates same files src/cmd/hive/main_eval_cycle.go
src/cmd/hive/main_sweeps.go
src/docs/operator-reference/sweeps-and-operations.md

@hivecommons-hive

hivecommons-hive Bot commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor

Hive status — updated in place; edits do not notify. Full history is in the Activity Monitor.

clubanderson this PR is held at stage/071-needs-author and the hive will not merge it yet.

Failing checks:

  • native-queue-policy

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.

@hivecommons-hive

Copy link
Copy Markdown
Contributor

CI report for 4f8f849 (scanner lane — human-authored branch, so this is a comment, not a push):

1. relative links and anchors resolve — PR-local, two citations drifted (src/scripts/check-docs-citations.py):

DRIFT src/docs/labels-and-control-signals/reference.md:817 src/pkg/config/governor_config.go:303-314 -> lacks anchor(s): LabelsConfig
DRIFT src/docs/labels-and-control-signals/reference.md:879 src/pkg/config/governor_config.go:303-314 -> lacks anchor(s): LabelsConfig, DefaultAutoMergeLabel

On this head type LabelsConfig struct is at src/pkg/config/governor_config.go:317-336 (the DefaultAutoMergeLabel mention is inside it at line 325). Fix on both lines 817 and 879 of src/docs/labels-and-control-signals/reference.md:

-`src/pkg/config/governor_config.go:303-314`
+`src/pkg/config/governor_config.go:317-336`

or run python3 src/scripts/check-docs-citations.py --fix src/docs and commit the result.

2. native-queue-policy — shared incident, not this PR's fault: #17003. Every v6-based PR fails with isCosmeticLabelFlip is not a function because the replay step added by #16952 runs against v6's older .github/scripts/native-queue-policy.cjs. This PR's head already carries the exporting .cjs, so merging it is the fix for the incident; the check cannot go green here until it lands.

DCO / dco / tide: non-blocking on forward-merges carrying others' commits, as the body notes.


🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88

… landed

Part of #7563

Signed-off-by: Andrew Anderson <andy@clubanderson.com>
@hivecommons-hive

Copy link
Copy Markdown
Contributor
3e788b5

ℹ️ Sentinel notice — informational; author is trusted (owner), merge is not blocked

Hive flagged this PR (author @clubanderson, head 3e788b54af20) because it matches behaviors that can override security controls, escalate privileges or damage the codebase. This is a heuristic, not an accusation — a maintainer should confirm the change is intended before it merges.

  • sensitive_path — changes 31 sensitive path(s) (Touches a sensitive path (OWNERS, workflows, policies, hive config, security docs, …))
    • .github/workflows/native-queue-policy.yml
    • .github/workflows/stream-image.yml
    • bin/gh-wrapper.sh
    • bin/gh-wrapper.test.sh
    • src/Dockerfile
    • src/Dockerfile.contributor
    • src/deploy/entrypoint.sh
    • src/deploy/stream/Dockerfile
    • src/deploy/stream/README.md
    • src/deploy/stream/docker-compose.yaml
    • src/deploy/stream/entrypoint.sh
    • src/deploy/stream/fleet/configmap.yaml
    • src/deploy/stream/fleet/deployment.yaml
    • src/deploy/stream/healthcheck.sh
    • src/deploy/stream/k8s/configmap.yaml
    • src/deploy/stream/k8s/deployment.yaml
    • src/deploy/stream/overlays/hive-hosted-matrix/kustomization.yaml
    • src/deploy/stream/stream-watchdog.sh
    • src/deploy/stream/test-stream-watchdog.sh
    • src/deploy/test_entrypoint_replica_names.sh
    • src/pkg/policies/defaults/quality-advisory.md
    • src/pkg/policies/defaults/quality-full.md
    • src/pkg/policies/defaults/quality-holdgated.md
    • src/pkg/policies/defaults/quality-measured.md
    • src/pkg/policies/defaults/quality.md
    • src/pkg/policies/source_sync_test.go
    • src/policies/quality-advisory.md
    • src/policies/quality-full.md
    • src/policies/quality-holdgated.md
    • src/policies/quality-measured.md
    • src/policies/quality.md
  • remote_code_execution — adds a remote-fetch-and-execute, decoded payload, reverse shell or embedded credential (Adds curl|sh, eval of encoded payloads, reverse shells or embedded private keys)
    • src/crux/logscrub/find_secret_test.go
    • src/crux/pushbroker/pushbroker_test.go
    • src/deploy/stream/Dockerfile

Hive did not add the sentinel-alert label because this author is trusted. The finding remains recorded for audit and the dashboard Security tab; set sentinel.trusted_authors_block: true to require the blocking label for trusted authors too.

@clubanderson
clubanderson merged commit e1bad50 into v6 Nov 7, 2026
87 of 90 checks passed
@kubestellar-prow
kubestellar-prow Bot deleted the topup/v5-to-v6-20261107 branch November 7, 2026 15:40
@github-actions

github-actions Bot commented Nov 7, 2026

Copy link
Copy Markdown
Contributor

Thank you for your contribution! Your PR has been merged.

We'd love to hear how your experience was: share feedback

Danathar added a commit to Danathar/hive that referenced this pull request Nov 7, 2026
Signed-off-by: Danathar <Danathar@users.noreply.github.com>
Danathar added a commit to Danathar/hive that referenced this pull request Nov 7, 2026
…verdict

Signed-off-by: Danathar <Danathar@users.noreply.github.com>
Danathar added a commit to Danathar/hive that referenced this pull request Nov 7, 2026
…ner-edits

Signed-off-by: Danathar <Danathar@users.noreply.github.com>
Danathar added a commit to Danathar/hive that referenced this pull request Nov 7, 2026
…p-bots

Signed-off-by: Danathar <Danathar@users.noreply.github.com>
Danathar added a commit to Danathar/hive that referenced this pull request Nov 7, 2026
…doption

Signed-off-by: Danathar <Danathar@users.noreply.github.com>
Danathar added a commit to Danathar/hive that referenced this pull request Nov 7, 2026
Signed-off-by: Danathar <Danathar@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

dco-signoff: yes Indicates the PR's author has signed the DCO. size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants