From cd420aa73974e60fdd46cf73e30ebd0afc0f9c60 Mon Sep 17 00:00:00 2001 From: Danathar Date: Wed, 28 Oct 2026 09:33:56 -0400 Subject: [PATCH 1/2] =?UTF-8?q?=F0=9F=93=96=20docs(v6):=20behind=20fork=20?= =?UTF-8?q?leaves=20the=20lane=20front=20at=20once;=20serialized=20lane=20?= =?UTF-8?q?readiness=20row=20from=20the=20dd406b3=20re-run?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit operator-reference.md said a behind fork pull request leaves the front once the front timeout runs out; forkUpdateOrLeave leaves at once and posts one comment, which the live exercise on #7563 observed. The readiness row now records the full re-run at dd406b3: steps 1-3 and 5-7 pass, step 4 (hold at the front) failed and is tracked in #14934. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_0113KwWU4TkJ8wfy3XjusyGo Signed-off-by: Danathar --- src/docs/operator-reference.md | 18 ++++++++++++------ src/docs/v6-readiness.md | 2 +- 2 files changed, 13 insertions(+), 7 deletions(-) diff --git a/src/docs/operator-reference.md b/src/docs/operator-reference.md index 4a031d9e1b..dbdbcc8665 100644 --- a/src/docs/operator-reference.md +++ b/src/docs/operator-reference.md @@ -768,12 +768,18 @@ into its branch, pinned to the evaluated head; it never rebases or force-pushes, and never rewrites a contributor's commits. A fork pull request that already contains the current tip with passing required checks needs no update and merges normally. Fork CI waiting for maintainer approval -(`action_required`) counts as not passed. When Hive cannot update a fork's -branch (GitHub refuses the update, for example because "Allow edits by -maintainers" is off), the pull request does not merge from the front; it -leaves the front with the recorded reason once the front timeout runs out, -the next pull request starts, and it stays eligible to merge at a later turn -once its author has brought it up to date. +(`action_required`) counts as not passed. The lane updates a behind fork +branch only when GitHub reports that Hive can update it +(`viewerCanUpdateBranch`, which needs "Allow edits by maintainers") **and** +`review.contributor_prs.base_sync` is on. Otherwise the pull request leaves +the front at once, without waiting for the front timeout, with the reason +`fork pull request is behind the target branch and Hive is not allowed to +update it: …` naming which condition failed. Hive posts one comment asking the +author to merge the target tip, and the next pull request starts. Once the +author has merged the tip and the required checks pass, the pull request +rejoins the lane and merges at a later turn. A queued pull request whose new +head its queue approval no longer covers (for example, the update also changed +files) must be queued again. ### Branches with GitHub's merge queue diff --git a/src/docs/v6-readiness.md b/src/docs/v6-readiness.md index ecd0a1f693..1f084ed38b 100644 --- a/src/docs/v6-readiness.md +++ b/src/docs/v6-readiness.md @@ -119,7 +119,7 @@ not a unit test: | Runs via chat | One live run reaches a human gate and is approved from the chat spine; the next runs snapshot shows it advancing. | ✅ Live exercise passed with `just runs-e2e-v6` against `clubanderson/hive-runs-e2e#11`; evidence on [#8466](https://github.com/hivecommons/hive/issues/8466#issuecomment-5805851875) and tracker row checked in the [#7563 v6 readiness bar section](https://github.com/hivecommons/hive/issues/7563). | | Email | One HUMAN DECISION NEEDED escalation delivered; one allowlisted inbound reply acted on (or reply-to-act explicitly deferred here). | ⬜ | | Push / on-call | One `requires_human` verdict pages a real device via at least one provider. | ⬜ | -| Serialized merge lane (`merge_strategy: hive-serialized`, [#10884](https://github.com/hivecommons/hive/issues/10884)) | On a personal-account repo with no GitHub merge queue, following the [runbook](v6-readiness-runbook.md#evidence-template) evidence template: (1) strategy switch and ACMM Eval credit ("satisfied by Hive serialized merge lane"); (2) a single merge; (3) two PRs at once (ordering, and the behind PR is updated); (4) a hold refused at the front; (5) a failed required check at the front; (6) the front timeout; (7) a fork PR. | 🟨 Partial. A self-hosted v6 hive at `f55b5fe66` merged four PRs through the lane, which produced [#10984](https://github.com/hivecommons/hive/issues/10984) (fixed by [#10985](https://github.com/hivecommons/hive/pull/10985)) and [#11023](https://github.com/hivecommons/hive/issues/11023). **Done:** single merge; update-then-merge on the App sweep path; front timeout. **Not yet done:** hold at the front; red-at-front leave (#11023 fixed by [#11031](https://github.com/hivecommons/hive/pull/11031)); fork PR. Row stays unchecked until every step has linked evidence. | +| Serialized merge lane (`merge_strategy: hive-serialized`, [#10884](https://github.com/hivecommons/hive/issues/10884)) | On a personal-account repo with no GitHub merge queue, following the [runbook](v6-readiness-runbook.md#evidence-template) evidence template: (1) strategy switch and ACMM Eval credit ("satisfied by Hive serialized merge lane"); (2) a single merge; (3) two PRs at once (ordering, and the behind PR is updated); (4) a hold refused at the front; (5) a failed required check at the front; (6) the front timeout; (7) a fork PR. | 🟨 Partial. A self-hosted v6 hive at `dd406b3` re-ran all seven steps against one build, which includes #11031 ([evidence](https://github.com/hivecommons/hive/issues/7563#issuecomment-6470375751)). **Done:** (1) ACMM credit; (2) single merge; (3) two at once, with the behind PR updated by a merge of the tip; (5) red at the front left on the first sweep after the check failed, and the next PR was promoted in the same transaction; (6) front timeout; (7) up-to-date fork merges, and a behind fork with `base_sync` off leaves the front at once with one comment, then rejoins and merges after the author merges the tip. **Failed:** (4) a `hold` on the front PR never released the front, which waited out the 60m timeout ([#14934](https://github.com/hivecommons/hive/issues/14934)). **Also found:** contributor-lane merges bypass the lane ([#14935](https://github.com/hivecommons/hive/issues/14935)); a 4xx on the lane's merge call wedges the front until the timeout ([#14936](https://github.com/hivecommons/hive/issues/14936)). **Not yet done:** `front-extend` at a short `merge_front_timeout`; the `base_sync`-on fork leg. Row stays unchecked until step 4 passes. An earlier partial run at `f55b5fe66` produced [#10984](https://github.com/hivecommons/hive/issues/10984) (fixed by [#10985](https://github.com/hivecommons/hive/pull/10985)) and [#11023](https://github.com/hivecommons/hive/issues/11023) (fixed by [#11031](https://github.com/hivecommons/hive/pull/11031)). | ## Admin MCP readiness evidence From e2b74a6046a799cd6bbf7abc6476973f1e45e0d6 Mon Sep 17 00:00:00 2001 From: Danathar Date: Wed, 28 Oct 2026 14:02:27 -0400 Subject: [PATCH 2/2] =?UTF-8?q?=F0=9F=93=96=20docs(v6):=20queue=20approval?= =?UTF-8?q?=20covers=20exactly=20the=20approved=20commit?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit A /hive queue approval records the reviewed commit SHA (latestHiveQueueApproval) and both queueApprovalStillHolds and the queued sweep compare it by exact SHA, so any new head, including a plain merge of the tip, needs a re-queue. Drop the parenthetical that implied a tip-only merge keeps the approval. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01BQYnPGFJx81c8JXR3o1sg1 Signed-off-by: Danathar --- src/docs/operator-reference.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/src/docs/operator-reference.md b/src/docs/operator-reference.md index dbdbcc8665..5c73e6be4b 100644 --- a/src/docs/operator-reference.md +++ b/src/docs/operator-reference.md @@ -777,9 +777,9 @@ the front at once, without waiting for the front timeout, with the reason update it: …` naming which condition failed. Hive posts one comment asking the author to merge the target tip, and the next pull request starts. Once the author has merged the tip and the required checks pass, the pull request -rejoins the lane and merges at a later turn. A queued pull request whose new -head its queue approval no longer covers (for example, the update also changed -files) must be queued again. +rejoins the lane and merges at a later turn. A pull request merged through a +`/hive queue` approval must be queued again after any new head, including a +merge of the tip: the approval covers exactly the approved commit. ### Branches with GitHub's merge queue