Repository navigation
π± sync: forward-merge v5 (2d67cd5) into v6 - #14815
Conversation
β¦recorded (#14706) * fix(review): wake the governor when a dispatched review verdict is recorded Review dispatch was planned only on a governor eval cycle, and recording a head's verdict did not start one, so on a quiet repo every reviewer sat idle for up to governor.eval_interval_webhook_s (900s) between kicks. The relay now calls a ReviewVerdictHook (Client.SetReviewVerdictHook) once per request whose verdicts bind a review dispatch; advisory and refused verdicts do not call it. The hive wires it to a verdict waker whose batches runLoop turns into a cached-snapshot eval cycle. The first verdict wakes at once; later ones inside governor.webhook_wake_min_gap_s fold into one wake at the end of the gap. Hive-Run: #14705 Hive-Plan: #14705 Proposed fix Hive-Spec: review-verdict-wake#14705 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HRH6kngWRcKiKJefZzZeqz Signed-off-by: Karibhou <Karibhou@users.noreply.github.com> * π docs: re-anchor RankActionableIssues citation in scale-envelope.md The verdict-hook addition to client.go shifted RankActionableIssues down 10 lines, so the citation guard reported drift. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Signed-off-by: Karibhou <Karibhou@users.noreply.github.com> * π docs: say the review verdict wake ignores webhook_wake_debounce_s Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HRH6kn4BsouThkrpz1TTCU Signed-off-by: Karibhou <Karibhou@users.noreply.github.com> Co-authored-by: gregoryhunt <31053269+gregoryhunt@users.noreply.github.com>
β¦ a runtime error Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Generated by changelog-fragment-guard from the PR title (#5675). Edit the wording freely, or replace this file with the no-changelog label if the change is not user-visible. Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Signed-off-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
β¦14731) * fix(deps): bump golang.org/x/net to v0.62.0 for GO-2026-6860 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> * fix(deps): regenerate NOTICE for golang.org/x/net v0.62.0 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>
β¦nfigured (#14618) * β¨ feature(crux): export OTel spans and kick metrics over OTLP when configured The Crux runtime opened spans through the OpenTelemetry API, but the crux binary never registered an SDK provider, so every span was discarded and there were no metrics at all (hivecommons/crux#19). - crux/telemetry: install an OTLP/HTTP TracerProvider and MeterProvider at startup only when OTEL_EXPORTER_OTLP_ENDPOINT (or a per-signal endpoint) is set; honour OTEL_SDK_DISABLED, OTEL_*_EXPORTER=none and the resource env vars. Unset keeps the no-op providers. A non-http/protobuf protocol disables export with a warning instead of failing the command. - crux/agent: crux.agent.kick.duration (histogram, s) and crux.agent.kick.outcomes (counter), labelled only by agent.role (a configured agent name, else "unknown") and outcome (ok|cancelled|error). - Both crux mains call telemetry.Init and flush on exit (5s bound). - No /metrics endpoint and no Prometheus dependency. - Document the env vars and instruments; changelog fragment. Hive-Run: hivecommons/crux#19 Hive-Plan: maintainer decision on hivecommons/crux#19 (OTLP env-gated providers, kick metrics, no /metrics) Hive-Spec: hivecommons/crux#19#recommendation Signed-off-by: Karibhou <Karibhou@users.noreply.github.com> * π fix(crux): register OTel globals only after both pipelines build Also test the OTLP export path against a fake collector and document the bounded five-second exit flush when the collector is unreachable. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HRH6kixWVJF9x9b1JF77JQ Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> * π fix(crux): bind attribute sets before calling pointer method Equivalent in kick metrics test Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> * test(crux/telemetry): cover the resource and exporter error branches The coverage gate holds crux/telemetry at 90%; the error paths Init takes when the resource or an OTLP exporter cannot be built were untested. Each is reached through the standard environment alone: an OTEL_RESOURCE_ATTRIBUTES pair with no value, and a CA certificate configured against an insecure endpoint for either signal. The metric-side case also checks that the already-built trace provider is torn down and no global is replaced. 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: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* π fix(proxy): canonicalize hostnames before host-class lookups DNS names are case-insensitive and may end in a dot; normalize inside the host predicates. Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> * docs: refresh rules.go citation shifted by host canonicalisation 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>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
π fix(forward-merge): treat the squash-refusal guard as expected, not a runtime error
β¦udge (#14732) Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
* π fix(crux): keep runbooks/ when publishing to hivecommons/crux Signed-off-by: operations <280983584+hivecommons-hive[bot]@users.noreply.github.com> * π fix(deps): bump golang.org/x/net to v0.62.0 Signed-off-by: operations <280983584+hivecommons-hive[bot]@users.noreply.github.com> --------- Signed-off-by: operations <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: operations <280983584+hivecommons-hive[bot]@users.noreply.github.com>
β¦14722) * π fix(ioscan): name classifier error and fix in blocked-kick warning Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> * fix(ioscan): clear stale classifier error once the classifier recovers lastClassifierErr was only ever written, so after one classifier error every later blocked kick, including one withheld for a genuine injection detection with a healthy classifier, carried a hint to disable the classifier. Clear it on the success path. The fail-open branch of classifierErrored is left alone: the classifier did error there, so the value is not stale, and fail-open does not withhold kicks. 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>
β¦14720) Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
* refactor(proxy): split github_proxy.go by concern (move-only) Move-only split of src/crux/proxy/github_proxy.go into per-concern files inside package proxy. No signature, behaviour or exported-name changes. Source-scanning tests now read all non-test github_proxy*.go files. Refs hivecommons/crux#23 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> * changelog: fragment for #14733 Generated by changelog-fragment-guard from the PR title (#5675). Edit the wording freely, or replace this file with the no-changelog label if the change is not user-visible. Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Signed-off-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com> * docs: re-home proxy citations to the split github_proxy_*.go files Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> --------- Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> Signed-off-by: hivecommons-hive[bot] <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: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
β¦s; add self-test (#14730) * π fix(scripts): action-pins gate fails silently on a tree with no pins; add self-test check-action-pins.sh only ever ran against this repository's own, correct workflows, so none of its failure branches had executed anywhere. Writing a hermetic self-test for it (src/scripts/test-check-action-pins.sh, fake `gh`, synthetic workflow directories) exposed two defects: - a directory with no SHA-pinned `uses:` made grep exit 1, and pipefail fed that straight into `fail=1`: "action pin check FAILED" with no finding printed above it; - NOT-PINNED printed the line number where the file name belongs (`grep -rhn` drops the name), so the message never said which workflow. The resolver now collects pins first and iterates, reporting every missing SHA instead of stopping at the first; the shape check uses `grep -rHn` so the message names the file. The self-test covers NOT-PINNED (tag, branch, short and upper-case hex), MISSING-SHA, reusable-workflow owner/repo extraction, pin de-duplication, the offline (no gh) path, EMPTY-EXPRESSION, BRACKETED-EMAIL with the numeric-noreply exclusion, combined failures and the default directory. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> * π§ chore(deps): bump golang.org/x/net to v0.62.0 (GO-2026-6860) Aligns go.mod/go.sum with v5 so govulncheck passes on this branch. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: quality <280983584+hivecommons-hive[bot]@users.noreply.github.com> --------- 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>
Signed-off-by: Andrew Anderson <andy@clubanderson.com>
β¦passes (#14756) Signed-off-by: Andy Anderson <andy@clubanderson.com>
β¦9-900px (#14656) The #13173 fix pinned .oc-topbar-right to max-content only from 901px, so at 769-900px the cluster shrank to 0, its nowrap flex-end chips overflowed leftwards and the feedback/theme buttons painted over the project title. Extend the guard down to 769px (the phone block takes over below) and drop the Settings/Health/Guide labels in that band so the drawer toggle and title keep room. Fixes #13135 Hive-Run: #13135 Hive-Plan: dashboard-header-overlap Hive-Spec: dashboard-navbar#13135 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HRH6kizfUBgiWsEGGur6Rr Signed-off-by: Karibhou <Karibhou@users.noreply.github.com> Co-authored-by: MikeSpreitzer <14296719+MikeSpreitzer@users.noreply.github.com>
β¦quest (#14745) Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
#14740) Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Captured live from `codex app-server` model/list against the 0.165.0 binary (same exchange cli_models_codex.go performs). Diff against the 0.159.0 snapshot: adds gpt-6.1-sol (visible, isDefault, present since at least 0.160.0) and gpt-6.1-terra (visible, new in 0.165.0). Hidden ids unchanged. index.html mirror and the #9804 parity test updated. Signed-off-by: Andrew Anderson <andy@clubanderson.com>
The scanner's newer issue templates write the release line as
'**Target branch:** `v6`' or inline after a sentence ('Target: v6 (open
the PR with --base v6)'). targetBranchDeclarationPattern only matched a
line beginning 'target: vN', so every relay dispatch of those issues
defaulted to the hub branch and came back wrong_base, leaving the issue
at 010 with no visible reason.
Fixes #14764
Signed-off-by: Andy Anderson <andy@clubanderson.com>
β¦competitor report (#14757) Signed-off-by: Andrew Anderson <andy@clubanderson.com> Co-authored-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Automated pin bump (#8419). Digests recomputed from the downloaded release artifacts by src/scripts/cli-pin-bump.sh: VERSION=2.1.312 Signed-off-by: hive-release[bot] <hive-release@hive.kubestellar.io> Signed-off-by: clubanderson <407614+clubanderson@users.noreply.github.com> Co-authored-by: hive-release[bot] <hive-release@hive.kubestellar.io>
β¦rows (#14772) An issue the scheduler withholds for an open "blocked by" dependency rendered as plain Actionable ("waiting for the scanner"), so sequenced RFC sub-issues read as intake starvation. The row now carries the blocked modifier, blocked_on keys, and a "waiting on #N" caption; the chip names the blocker. Signed-off-by: Andy Anderson <andy@clubanderson.com>
β¦14771) The relay's task prompt only asks the contributor model for `Closes #N`. #14701 shipped #14633 with an empty body; the PR-claim ledger derives claims from PR text and branch names only, so the issue stayed at stage/010-actionable for hours while the PR waited in review as "no linked issue" and every lane was free to re-claim it. On a verified task_complete for a GitHub-issue task the hub now appends `Fixes #N` to the contributor's own open PR (unless the body already closes the issue or says it stays open) and ensures the GitHub sidebar link so release-line PRs show in the issue list too. Detached and best-effort like the attribution reconcile. Fixes #14763 Signed-off-by: Andy Anderson <andy@clubanderson.com>
Automated pin bump (#8419). Digests recomputed from the downloaded release artifacts by src/scripts/cli-pin-bump.sh: VERSION=0.165.0 Signed-off-by: hive-release[bot] <hive-release@hive.kubestellar.io> Signed-off-by: clubanderson <407614+clubanderson@users.noreply.github.com> Co-authored-by: hive-release[bot] <hive-release@hive.kubestellar.io>
β¦MIT (#14682) * π± runtime: loopback pprof, periodic memstats and cgroup-aware GOMEMLIMIT Adds pkg/runtimeobs: a loopback-only net/http/pprof listener (debug.pprof_addr, default 127.0.0.1:6060, 'off' disables), a 60 s 'mem: runtime snapshot' INFO line (heap in-use/live/objects, next GC, goroutines, OS threads, GOMEMLIMIT) read via runtime/metrics, the same snapshot under 'runtime' in GET /api/health, and a GOMEMLIMIT default of 85% of the cgroup memory limit when the env var is unset. Refs #14681 Signed-off-by: Andy Anderson <andy@clubanderson.com> * runtime: satisfy CI ratchets for pprof/memstats - health body test decodes the new runtime object - settings parity allowlist: debug.pprof_addr is an operator knob - dashboard import ratchet: allow pkg/runtimeobs - agent runtime golden: Config gained the debug block - runtimeobs logger test waits on the record instead of sleeping - api-reference citations re-pinned after the server.go import shift Signed-off-by: Andy Anderson <andy@clubanderson.com> * changelog: fragment for #14682 Generated by changelog-fragment-guard from the PR title (#5675). Edit the wording freely, or replace this file with the no-changelog label if the change is not user-visible. Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Signed-off-by: clubanderson <407614+clubanderson@users.noreply.github.com> * runtimeobs: cover the default logger, disabled/occupied pprof and no-/proc paths Signed-off-by: Andy Anderson <andy@clubanderson.com> * changelog: single bulleted fragment for the runtime diagnostics Signed-off-by: Andy Anderson <andy@clubanderson.com> * docs: re-pin api-reference citations after the server.go import shift Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com> Signed-off-by: clubanderson <407614+clubanderson@users.noreply.github.com> Co-authored-by: clubanderson <407614+clubanderson@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
* fix(claims): retry transient GitHub side-effect failures on issue claims The claimed-stage and claimed-by label writes had one 15s attempt; a startup burst serialised behind the per-repo label writer timed them out and left issues an agent was already working at stage/010-actionable. Transient failures now retry up to four times with a doubling pause, inside the per-issue effect queue so ordering is preserved. Comments are still posted once. Signed-off-by: Andy Anderson <andy@clubanderson.com> * test(claims): instant retry pauses in TestClaimHooks_SubmitClaimFacts The unreachable test host now yields a retryable net error; without an instant pause the queued release waited behind 14s of backoff. Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
β¦verdict (no more 070β085 flap) (#14759) * fix(stage): treat a recorded /hive decision as the answer to the head's requires_human verdict The stage-label reconciler re-derived 085-needs-decision from the persisted decision-requested verdict on every tick because nothing told the verdict route that the owner's /hive decision had answered it; whoever moved the PR to 070 (unpark reply, fix planner, a maintainer by hand) was undone ~90s later (hive#14628 flapped 070<->085 six times). The recorded decision (review-human-change-requests.json, head-scoped, actor, timestamp, kind) is now folded into the reviewer verdict for that head as DecisionAnswered/DecisionAnsweredBy/DecisionAnsweredHead/ DecisionDirective, both in the lanes' verdict lookup and in the stage pass before classification, so a lane that skipped early can no longer leave the question looking open. Instructions derive 070 with reason contributor-lane:decision-answered; a bare approve clears the hold like a trusted Approve review; a new head's fresh requires_human verdict carries no answer and parks again. ReviewFacts round-trips the answer for the derive path and the verdict-route log names decision_answered/by/directive. Refs #14679 Signed-off-by: Andy Anderson <andy@clubanderson.com> * test(automerge): rename helper clashing with containsLabel Signed-off-by: Andy Anderson <andy@clubanderson.com> * test(prstage): cover ReviewVerdict.WithDecisionAnswer to restore the coverage floor Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
β¦4651) * π± dashboard: defer merge-gate pills until the PR reaches the gate Split tracker pills into actionable-now (hold, draft, stalled, regressed, unauthorized on held workflow runs) and merge-gate (sentinel, needs-human, needs-decision, merge-time unauthorized). Merge-gate pills are held back in deferred_modifiers while the PR is before stage/080 and neither render nor count under Needs attention; they render as before once the PR reaches the gate. Labels are untouched. Fixes #14650 Signed-off-by: Andy Anderson <andy@clubanderson.com> * π± dashboard: update sentinel lane wiring assertion for the deferred tooltip line Signed-off-by: Andy Anderson <andy@clubanderson.com> * π± dashboard: load stageTrackerStaleLabelsText in the deferred-pills JS harness Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
β¦ declaration (#14784) Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
* π scheduler: classify held-PR blocked-kick ioscan warning Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> * changelog: fragment for #14783 Generated by changelog-fragment-guard from the PR title (#5675). Edit the wording freely, or replace this file with the no-changelog label if the change is not user-visible. Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Signed-off-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com> * fleetreport: lower-case held-PR ioscan keyword so it matches LogFedClass/MatchesExplicitClass lower-case the message before strings.Contains, so the keyword "ioscan fail-closed blocked held-PR coordination kick" (uppercase PR) could never match. Lower-case it. Also fold the two changelog fragments into a single bulleted fixed-14783.md as check-changelog-fragment.sh requires. 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> Signed-off-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
β¦ditional (#14802) Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com> Co-authored-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
β¦14800) 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: clubanderson <407614+clubanderson@users.noreply.github.com>
* π fix: pack agents follow the hive's inference gateway An ACMM level that adds an agent to an existing hive created it on the pack's CLI backend (copilot) with a copilot model id, even when the operator had moved every other agent onto one inference gateway. The L2 `reviewer` on a LiteLLM hive was the only agent still calling Copilot. ApplyPack now reads the gateway the hive's enabled agents share (one inference backend or configured gateway name, unanimous among the gateway-routed agents) and places a new pack agent on it, clearing the pack model so the gateway's default_model applies (#9945). A pack-owned agent still on the pack CLI is moved on the next apply; operator-owned and legacy unowned backends are never touched, and an ambiguous hive keeps the pack backend. Fixes #14678 Hive-Run: #14678 Hive-Plan: issue-14678 Hive-Spec: issue-14678#reviewer-inherits-inference-config Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HRH6kiXX5Gytj12BZzAxe9 Signed-off-by: Karibhou <Karibhou@users.noreply.github.com> * π docs: re-anchor scale-envelope RankActionableIssues citation The v5 tip already drifted: RankActionableIssues now sits at src/pkg/github/client.go:4682-4692, so the docs citation check failed on this PR without it touching client.go. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HRH6kji18fCJVBCmQ8MBWb Signed-off-by: Karibhou <Karibhou@users.noreply.github.com> * π packs: follow the hive's gateway only when every other agent is on it hiveGatewayBackend counted only gateway-routed agents, so one agent moved onto LiteLLM made it "the hive's gateway" and the next restart moved every pack-owned copilot agent there. Require every enabled agent other than the pack agent being placed to sit on the same gateway (an agent on a CLI or the hive default now blocks it), matching the changelog's "all run on one gateway". Add the mixed CLI+gateway cases to the tests. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HRH6kn4BsouThkrpz1TTCU Signed-off-by: Karibhou <Karibhou@users.noreply.github.com> * π packs: decide the hive's gateway from operator-picked backends by majority hiveGatewayBackend returned "" as soon as any other enabled agent sat on a CLI, so the real L2 roster β supervisor and brainstorm pre-existing as pack-owned copilot agents β vetoed the move for the #14678 reviewer, and two late-added pack CLI agents vetoed each other on the repair path. Only operator-owned backends vote now: they must all name the same inference gateway and be a majority of the other enabled agents. Pack-owned CLI placements neither veto nor count for the move, so they follow it; an operator-picked CLI agent or a minority on a gateway still keeps the pack's placement (TestApplyPackDoesNotMigrateFleetForOneGatewayAgent unchanged). Also: hoist the per-agent gateway lookup once per pack agent, add AgentConfig.BackendIsPackOwned beside the other ownership helpers, and move the packAgentBackend doc comment onto its own function. Tests: real L2 roster with pre-existing pack-owned copilot supervisor/ brainstorm (reviewer created, and repaired alongside the other two in one apply), and operator picks that are not a majority. 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> * π packs: count agents already on the gateway toward the hive's gateway majority hiveGatewayBackend counted every other enabled agent in the majority bar but only operator-owned backends as votes. The pack agents L2 moved onto the gateway are pack-owned, so at L3 the three operator picks were three votes among six others and the new ci-maintainer landed on copilot β #14678 one level up. Operator picks still choose the gateway (any operator-picked CLI, hive default or second gateway vetoes it); the majority is now every other agent already on that gateway, whoever placed it there. Doc comment and changelog describe the implemented rule. 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> * π packs: pack-owned agents follow a later gateway change; document the rule Review follow-ups for #14777 (#14678): - A pack-owned agent an earlier apply moved onto gateway A now follows the hive again when the operator moves every operator-picked agent to gateway B; the move was gated on the agent still sitting on the pack CLI, so "follow the hive's gateway" was one hop only. Covered by TestApplyPackFollowsGatewayChange. - Rewrite the hiveGatewayBackend doc comment into separate sentences for the candidate set, the operator-owned unanimity condition and the majority bar, and reword the hiveAgentBackends snapshot comment so it no longer claims the gateway is judged on the operator's agents alone. - Document the gateway-follow placement rule in agent-configuration.md beside the backends allow/deny rule and note the exception to "pack values fill blanks" in the ACMM section. - Split the changelog fragment into readable sentences. 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> * π§ chore: DCO repair β sign off adopted commits Re-signs 38cd38c, 591ce65 and a28696b (now 66d7679, defdf80, 4a55934) with Signed-off-by; no code changes, tree unchanged. 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> Co-authored-by: MikeSpreitzer <14296719+MikeSpreitzer@users.noreply.github.com>
Forward-merge origin/v5 (2d67cd5) into v6 by hand after the v6 Top-up workflow (run 40621227070) stopped on conflicts. Both sides' intent is kept: v6-only code survives and every v5 fix lands. Conflicted files and resolutions are listed in the PR body. Signed-off-by: Andy Anderson <andy@clubanderson.com>
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 No conflicting open PRs or same-file overlaps were found. |
|
@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. |
caa1f38
βΉοΈ Sentinel notice β informational; author is trusted (owner), merge is not blockedHive flagged this PR (author @clubanderson, head
Hive did not add the |
There was a problem hiding this comment.
Verified this forward-merge mechanically rather than by reading 166 files, and it holds up. Looks correct to me.
What I checked at caa1f38
- Recomputed the merge:
git merge-tree --write-tree 6590c08 2d67cd5(the exactv6andv5heads the PR names) and diffed that tree against the PR head. Only 17 files differ: the 14 conflicted files the body lists (whose mechanical result carries conflict markers, so a difference is expected) plus three deliberate ports βsrc/pkg/config/debug_config.go,debug_config_test.goandsrc/crux/proxy/github_proxy_inference.go. Every other file is byte-identical to a clean merge, so no non-conflicting hunk was dropped or hand-edited. src/pkg/config/config.go: v5 = 8422 lines, v6 = 412, PR = 415. The v6 split layout is kept, and the one v5-only hunk to that file (ebaf1c0, theDebug DebugConfigblock +EffectivePprofAddr) is ported intodebug_config.gowith the same field and method names.src/crux/proxy/github_proxy.go: v5 split it into 11github_proxy_*.gofiles; v6's only commits on the monolith since the merge-base7a506f7are earlier forward-merges, so there is no v6-only proxy change to carry. The PR takes the v5 split (total proxy bytes 157611 vs v5's 157657).go build ./...insrc/passes at the head.
Checks
DCO= action_required comes from2d67cd5(a v5 merge commit by another author with no sign-off) β the usual forward-merge noise, not something this branch can fix.Sign-off survives the squashisCANCELLED;build (linux/arm64β¦)was still in progress when I looked. I could not run the test suite here, so the v2-ci per-package coverage floors (which forward-merge unions have undercut before, e.g. #8854/#8886) remain the one thing CI has to confirm.
β hive: agent=reviewer-2 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 caa1f38. |
|
Thank you for your contribution! Your PR has been merged. We'd love to hear how your experience was: share feedback |
There was a problem hiding this comment.
Checked the merge independently of the body, then compared. The hand resolution looks correct to me.
Method: rebuilt the automatic merge of 60a2aa2 (v6) + 2d67cd5 (v5) with git merge-tree --write-tree and diffed it against 793c13b; the only differences are the 14 conflicted files (plus the new debug_config.go/_test.go). The follow-up merge of 6590c08 into caa1f38 is identical to its automatic merge. No conflict markers remain in src/ (the two ===== hits are banner lines in scripts/generate-notice.sh:185 and test-validate-generated-notice.sh:19).
Resolutions I verified against both parents (for each: lines v5 has that the result lacks, and lines v6 has that the result lacks, are only the pre-conflict versions of the other side's hunk):
src/pkg/config/config.goβ v6's 412-line split layout kept; v5's single addition since merge-base (theDebug DebugConfigfield,config.go:86-88) is present, and the 36 lines ofDefaultPprofAddr/PprofAddrOff/DebugConfig/EffectivePprofAddrare byte-identical indebug_config.go.src/crux/proxy/github_proxy.goβ v5's split taken; v6's only change since merge-base (twoopenAIChatCompletionsURL(route.Endpoint)sites) is atgithub_proxy_inference.go:436and:652, with the helper atinference_forward.go:82.src/pkg/fleetreport/runtime.goβ result carries v5's"ioscan fail-closed blocked held-pr coordination kick"keyword and v6's"kick undeliverable"+ClassAgentGitHubCapReachedentry.src/pkg/dashboard/stage_tracker.goβDeferredModifiers(v5) andFinisher*(v6) both present;modUnauthorizedretained.src/pkg/dashboard/contribute_ws.goβlinkReportedPRToIssue(v5) andStandbyConnectionState(v6) both present.src/crux/agent/manager_kick.goβ named(err error)return atmanager_kick.go:114is what lets the v5recordKickdefer observe the outcome; that is the right way to reconcile the two.
go build ./... on caa1f38 is clean, and src/scripts/check-api-reference-citations.sh passes (api-reference citations: all pkg/<file>:<line> references verified), so the api-reference.md resolution did not leave stale line citations.
Open point, not a finding: I cannot run go test from here, so I have not independently confirmed the per-package coverage floors the v2-ci gate enforces on forward-merges beyond the pkg/config figure the body states; the in-progress CI run will settle that. probot DCO = action_required is the usual state on a forward-merge carrying other authors' commits and should not block an admin merge.
β hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
Summary
Hand-resolved forward-merge of
origin/v5(HEAD2d67cd5, 41 commits) intov6, replacing the failing automatedv6 Top-uprun 40621227070 ("v5 -> v6 forward-merge conflicts and needs a human", precedent #7199). Both sides' intent is kept: v6-only code survives and every v5 fix lands.origin/v6(6590c08) was merged back in afterwards so CI runs against current v6.14 files conflicted.
Conflict resolutions
src/crux/agent/manager_kick.goGovernorCadenceKickSource+SendKickβSendKickWithSourcesplit; gaveSendKickWithSourcea named(err error)return so v5'srecordKick(role, start, err)defer metrics (kick_metrics.go) observe the result. v5'sexemptCIPollLanelanded cleanly.src/crux/proxy/github_proxy.gogithub_proxy_{ca,connect,copilot,identity,inference,mitm,policy,setters,tunnel,violations}.go; re-applied v6'sopenAIChatCompletionsURL(route.Endpoint)change at both sites ingithub_proxy_inference.go.src/go.sumgolang.org/x/oauth2 v0.36.0lines (still required by v6go.mod).src/pkg/config/config.goDebug DebugConfigfield afterTracing.src/pkg/config/debug_config.go(new)DefaultPprofAddr,PprofAddrOff,DebugConfig,EffectivePprofAddr()(#14681) moved into their own file per v6's per-concern split. Addeddebug_config_test.gosopkg/configkeeps its 90% floor (90.2%).src/pkg/config/testdata/agent_runtime_golden.json"debug": {}after"Tracing": {}.src/pkg/fleetreport/runtime.goClassIOScanClassifierUnavailkeywords AND v6'sClassAgentGitHubCapReachedentry.src/pkg/dashboard/contribute_ws.gorecordVerifiedStandbyPR, standby outcomes); inserted v5'slinkReportedPRToIssuegoroutine (#14763) afterreconcilePRAttribution.src/pkg/dashboard/stage_tracker.goFinisher/FinisherVerb/FinisherReasonfields + v5'sDeferredModifiersrefactor (stageTrackerModifierOrder,stageTrackerClassifyModifiers,stageTrackerModifiersreturning(shown, deferred)); bothstageTrackerSetFinisherandstageTrackerBlockedBycalls kept.src/pkg/dashboard/static/index.html<select>(incl.finisher); v6's finisher fields /stageTrackerBaseRefBadgewith v5'sdeferred_modifierspassthrough.src/docs/agent-configuration.mdsrc/docs/dashboard-sections.mdsrc/docs/design/github-mention-triggers.mdsrc/docs/scale-envelope.mdsrc/docs/api-reference.mdcheck-docs-citations.py --fixre-anchored 16 citations,Conversecitation pointed atsrc/crux/proxy/rules.go. 1441 citations verified, 0 drift.Validation
go build ./...andgo build ./crux/...clean;gofmt -lclean on all touched packages;go vetclean.go test -short -coveron every Go package in the merge diff vs the floors inv2-tests.yml: cmd/crux 96.5, cmd/hive 75.6, crux/cmd/crux 77.8, crux/config 92.3, crux/proxy 91.5, crux/telemetry 98.0, pkg/config 90.2, pkg/dashboard 87.7, pkg/fleetreport 92.9, pkg/fleetreport/slogfeed 94.6, pkg/github 88.7, pkg/github/automerge 90.1, pkg/review 93.7, pkg/runtimeobs 95.6, pkg/scheduler 92.7, pkg/stageengine 93.0, pkg/stageengine/prstage 100.0 β all at or above floor. crux/agent measured 84.8% locally on both this branch and the v6 baseline (macOS tmux-socket tests skipped identically), i.e. no regression; CI on Linux is authoritative for its 88 floor.python3 src/scripts/check-docs-citations.py src/docs: 1441 verified, 0 drift.Note: the probot
DCO/dcochecks fail on forward-merges carrying other authors' commits; the merge commits themselves are signed off.