Repository navigation
π fix: keep the preparation train connector continuous between adjacent members - #16027
Conversation
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
|
Signed-off-by: Andy Anderson <andy@clubanderson.com>
Signed-off-by: Andy Anderson <andy@clubanderson.com>
Signed-off-by: Andy Anderson <andy@clubanderson.com>
There was a problem hiding this comment.
Decision needed: @clubanderson should confirm that running sudo env "PATH=$PATH" npx --yes @puppeteer/browsers@2 install β¦ --install-deps on the self-hosted fleet in .github/workflows/dashboard-mobile-shots.yml is acceptable (root apt-get driven by a fetched npm package), or whether libnspr4 should be baked into the runner image instead.
Dashboard change looks correct to me; the workflow change needs an owner call.
Rail logic (read at a065e4c):
tracker-preparation.js:109-111βstageTrackerPreparationAdjacentnow keys ongroupIndexonly. The header emission at:157and:219and the row class at:172-177all go through this one function, so header,ham-train-lastand the connector agree on what "adjacent" means; no path still applies the oldordered/sequence-ascending rule.:172-174β gap detection walksprev.last ++ info.memberswith|Ξsequence| > 1, so reverse-FIFO neighbours (9 β 2 in the fixture) no longer getham-preparation-gap, while a true hole still does.info.orderedis still consumed bystageTrackerPreparationNodeHtml(:187) for the "linked row order differs from FIFO" tooltip, so the FIFO signal isn't lost.static/index.htmlmirrorsstatic-src1:1 for the JS and CSS hunks.- The new test asserts the pre-last members carry neither
ham-train-lastnorham-preparation-gapand that/pr#80(unrelated child) stays ahead of/pr#9without membership β matches the doc paragraph added inactivity-monitor.md.
Workflow (.github/workflows/dashboard-mobile-shots.yml:41): the Chrome install step now runs under sudo with --install-deps, which has the npm package drive apt-get as root on the self-hosted runner. It stays continue-on-error: true, and moving CHROME_BIN export after --version is a sensible tightening. I'm not calling it a defect β only that root package installs from npx --yes on shared runners is a trust-surface decision, not a reviewer's.
Open question (unverified): if the fleet runners have no passwordless sudo, the step fails and the harness SKIPs as before, so the new preparation-rail layout assertions would never actually run in CI. Worth checking one run's log after merge.
β hive: agent=reviewer-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
β¦tadata Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
Hive status β updated in place; edits do not notify. Full history is in the Activity Monitor. @clubanderson this PR is held at Why:
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. hive/awaiting-ci: the only red check here, π Hive Agent: β hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88 Review findings from Intent-alignment (delta re-review, head be6b300) β approve. Delta since a21527d is Advisory (landability, not a code defect): v5 has since advanced one commit that also touched -const TRUSTED_SCRIPTS_TREE = 'c025e262175f80a597097ee069ae5029a1513465';
+const TRUSTED_SCRIPTS_TREE = '<output of: git rev-parse HEAD:src/scripts after merging v5>';Previous finding status: [security] Review findings from Conclusion: the dashboard ordering/rail change looks correct to me and is well covered; routing only because the PR also (a) grants a workflow Verified correct (dashboard)
Needs a maintainer
No duplicates found in the current queue for this change. Review findings from Conclusion: the renderer change reads internally consistent against the PR head, but the PR is 1. Conflict is on the trust pin and must be re-derived, not resolved by hand (verified). 2. Root on the self-hosted fleet (needs a decision). What I checked and found correct:
Minor, not blocking: I have not approved or changed anything. Evidence bundle `#16027d8a76dc` sha256:f2bb9cb4e971eefa3f13e2ec11cfc2fa733403fa46476e0024bc1f0861ede480 β download from the dashboard or `hivectl review evidence #16027`. |
ebf49d9
βΉοΈ Sentinel notice β informational; author is trusted (owner), merge is not blockedHive flagged this PR (author @clubanderson, head
Hive did not add the |
β¦ails Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
Chrome now genuinely runs (156.0.8078.80), but the first actual geometry artifact exposed an existing file:// harness issue: absolute /tokens.css and /components.css never loaded, so spacing/line-width tokens were undefined and every rail had border:none. Fixed the harness to load those exact checked-in production sidecars and mount the fixture in the expanded real Activity Monitor panel, not by guessing token values or weakening geometry assertions. Fresh exact-head CI and actual layout artifact are required; prior reporting green/skip is not validation. No local browser/tests or deployment. |
Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
Acted on the actual same-head mobile-shots failure, not a stale hold: independent source review found responsive .config-info min-height:40px enlarged the node while its top still subtracted an assumed 4px half-height. Thus every actual mobile dot was 16px below its strip anchor. Changed centering to top:50% plus translateY(-50%), preserving the accessible hit target and full-height connector, and retained measured dot diagnostics in the actual Chrome artifact. Generated SPA updated; new immutable head is pushed. Fresh full CI and actual Chrome layout verdict pending. No local tests/browser or deployment; no user push required. |
There was a problem hiding this comment.
Read the tracker JS against the v5 tree; the ordering change looks correct to me. Flagging one item for the record because it touches a workflow.
Trust surface (.github/workflows/dashboard-mobile-shots.yml) β facts, not a defect: the job's permissions block is unchanged (contents: read, line 18-19); the only behavioural change is in the continue-on-error: true "Install Chrome for Testing" step, which now runs sudo apt-get update and npx @puppeteer/browsers install --install-deps under sudo env "PATH=$PATH" (lines 40-42), plus a new paths trigger for the fixture file. set -euo pipefail plus continue-on-error means a failed/absent sudo on the fleet still degrades to the harness's SKIP path as before. Author @clubanderson is the repository owner, so no role is being acquired. Noting it only so the workflow edit is explicitly acknowledged; nothing here needs changing from my read.
Correctness, verified:
stageTrackerPreparationContiguous(tracker-preparation.js:30-49) is a stable partition: unmapped rows keep their index, each group is emitted once at its first member's position, members sorted bysequencethen original index. No row is dropped or duplicated.stageTrackerPreparationAdjacent(:150-152) now treats any two same-group rows as adjacent andstageTrackerPreparationRowClass(:211-213) derivesgapfrom the combinedprev-last + memberssequence trail, so a non-consecutive ticket still renders the gap class rather than a false continuous rail.caption.split(...)at:246is safe:captionis always a string (:238ends in|| '').- Moving
stageTrackerPreparationNodeHtmlinto.ham-strip-rowintracker-row-layout.js(:244,:549) is paired with thetracker.csschanges and theham-trainguard onstageTrackerTrainKeyhosts (:60), so train rows are untouched.
β hive: agent=reviewer-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
Actual build-and-test failure investigated: the nested rail assertion still expected the old source order (front last). Chrome geometry had already verified all16 cases. Pushed DCO follow-up e651209: fixture now asserts authoritative front #45 first, waiting #9 last, with exact child anchors rather than incidental root metadata. No renderer change, no gate relaxation, no local tests. Full CI is required again on this new exact head; a215 geometry evidence is prior-head evidence, not a green verdict on e651. |
|
Terminal security verdict e651209: replacement gosec job124542607459/run41201935213 attempt2 also FAILED13:07:19Z with the same GitHub annotation: self-hosted runner lost communication; Scan still in progress and no code finding/failing scan step, logs unavailable404. This automated retry was triggered by github-actions[bot], not me. All other current checks including real Chrome16 fixtures and ARM are green. I am not retrying until green or weakening any gate; security remains infrastructure-blocked. Separate same-head requires_human workflow trust-surface approval still applies; review explicitly says no defect observed. |
|
Current-head actionable status reverified2026-10-31T14:23Z: e651209 remains OPEN. No sentinel-alert label; genuine remaining human fence is workflow trust-surface approval for dashboard-mobile-shots.yml apt metadata/install-deps change (contents:read unchanged; reviewer explicitly no defect observed). No GitHub APPROVED review recorded, so earlier unavailable owner question has NOT released the fence. A hive owner/trusted merger must explicitly approve THIS head through GitHub review or dashboard decision; this is not a request for another source push and will not be self-approved. Separate CI blocker: gosec job124542607459/run41201935213 terminal runner communication loss, not a reported scan finding; all other exposed checks pass/skip. Preserve CI and owner gates. Informational-fidelity source #16032 is merged40d3be36f36525312dd039b0cccd6e4c40b7765d, but I have not deployed or verified the production revision. Screenshot wording and duplicate caption are not proof of deployed source ancestry. |
e651209
βΉοΈ Sentinel notice β informational; author is trusted (owner), merge is not blockedHive flagged this PR (author @clubanderson, head
Hive did not add the |
|
Resuming this PR at the repository ownerβs explicit November 2 request. Its implementation agent is integrating current v5, resolving the needs-rebase condition, regenerating assets, and will obtain fresh exact-head CI and actual browser-geometry evidence. The earlier failed gosec attempts were runner communication failures, not demonstrated scan findings. This progress note is not a fabricated approval or CI bypass. |
Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
Nov2 resumed ownership: merged fresh origin/v5 e60f06f WITHOUT force push/amend. Actual conflict was upstream tracker-row-layout module split: retained new tracker-row-render module and newer human context/terminal/sentinel/ownership features, relocated both real preparation dots to their strip anchors, regenerated canonical SPA. Original front-first complete-unit/nested FIFO, full-height dynamic/final/single comments, exact membership and Chrome16 geometry fixtures retained. Independent review caught the new governance scripts-tree pin contract; updated literal to final integrated scripts subtree c025e262175f80a597097ee069ae5029a1513465 without weakening validation; follow-up independent review no significant findings. Signed integration pushed. Fresh full exact-head CI now required, old Chrome16 proof is historical only. No local suites/build/lint/docs guards and no deployment. Workflow trust-surface owner decision remains genuine unless current-head review resolves it; no approval forged and informational sentinel is not a gate. |
There was a problem hiding this comment.
HUMAN DECISION NEEDED β @clubanderson: approve root package installation (sudo apt-get + root npx β¦ --install-deps) on the self-hosted Hive fleet runners in a reporting-only workflow, or restrict it.
Security re-review at be6b300 (delta from a21527d).
Still present (in-scope): .github/workflows/dashboard-mobile-shots.yml:41-42 runs sudo apt-get update and sudo env "PATH=$PATH" npx --yes @puppeteer/browsers@2 install chrome@stable --install-deps β¦. Per runs-on at line 23, same-repo PRs execute on vars.HIVE_RUNNER_LABELS (self-hosted fleet), so a floating-major npm package (@2, unpinned minor/patch) now runs as root and installs system packages on fleet hosts; the previous revision ran it unprivileged. permissions: contents: read (line 18) and the pull_request trigger limit token exposure, but not host-level impact. Proposed narrowing β install the known-missing libs explicitly and keep npx unprivileged:
sudo apt-get update
sudo apt-get install -y --no-install-recommends libnspr4 libnss3 libatk-bridge2.0-0 libgbm1 libxkbcommon0 libasound2t64
out="$(npx --yes @puppeteer/browsers@2.10.5 install chrome@stable --path "$RUNNER_TEMP/chrome-for-testing" | tail -n 1)"
(adjust the lib list to what --install-deps actually pulls on the fleet image; pin @puppeteer/browsers to an exact version.)
Delta checked, no new blocker: .github/scripts/integration-governance.cjs:12 re-pins TRUSTED_SCRIPTS_TREE to c025e262β¦; verified equal to git rev-parse HEAD:src/scripts at be6b300, as the gate at integration-governance.cjs:101-104 requires for any PR editing src/scripts. New src/scripts/fixtures/dashboard-preparation-rail.mjs and the dashboard-mobile-shots.mjs hunks are browser-local DOM/layout checks against file:// stylesheets β no network, secrets, or shell execution.
β hive: agent=adjudicator-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
β¦ts pin Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
09:36ET request acted on: previous be6b300 finished all19 latest workflow/event groups success/skipped, including real mobile-shots and gosec. v5 then advanced to1e83f1f94d79cde31080c906d0b132fa5ffe21cc, creating a REAL new conflict solely at TRUSTED_SCRIPTS_TREE. Merged latest without force/amend, retained upstream16045 gosec concurrency and16151 scripts fixes; combined scripts pin308d64dd1ac3bf1061a4dced3071647236327e51 resolved, independently reviewed and staged. Canonical regeneration unchanged; tracker rail/full-height/front-first/nested fixtures untouched. Signed merge pushed; fresh exact-head CI required. Old waiting-on-author-to-push conflict message now describes the prior head, not an idle author. Privileged Chrome runtime dependency owner decision remains outstanding; no approval forged, no sentinel block or deployment. |
|
merged by Hive trusted-author auto-merge: author @clubanderson holds role owner and GitHub permission admin on repo; CI green at d8a76dc. |
|
Thank you for your contribution! Your PR has been merged. We'd love to hear how your experience was: share feedback |
|
SOURCE MERGE attribution: app/hivecommons-hive auto-merged tested head d8a76dc at2026-11-02T14:46:37Z as2f96f44c21846a65eeacd44d992eb8827de337fe. This implementation agent did NOT merge, submit APPROVED, or issue a dashboard approval. GitHub review timeline still has COMMENTED reviews only; do not describe the recorded privileged-Chrome owner choice as resolved by me. After merge all83 exact-head CheckRuns reached pass/skip, including ARM/gosec; real Chrome16 rail geometries ran successfully (run41463696284/job125338919478, Chrome156). Global overflow advisories remain44β47; not a clean global-overflow claim. DCO trailers, fresh v5 ancestry and unchanged canonical generation on landed squash verified. Owned worktree/local feature+verification branches cleaned; remote deletion verified. REST shared quota currently exhausted; final latest-workflow receipt to follow after reset. No deployment/live policy activation. |
Merge current v5, preserving #16045 gosec concurrency 4 and its generated integration workflows, #16027 preparation rail, and existing readiness regressions. Prior-head gosec job 125331848206 explicitly reported lost self-hosted runner communication; refresh exact-head CI without weakening any gate or removing hold. Signed-off-by: Andy Anderson <andy@clubanderson.com>
Problem
Preparation rows were separated by generic stage/age order, repeated continuation headers, and rails that stopped at the last chevron instead of covering comments.
Fix
Validation
Independent code review found one fixture-only missing reconciler key; fixed before the ordering commit. CI-only shipped-renderer regressions cover scrambled order, nested ownership, missing/closed/stale fronts, stable outsiders and exact identities. The real Chrome harness measures multirow seams, single/final multiline comments, dynamic height growth and anchor-independent dots. Its reporting-only green status is not a layout verdict: require the actual preparation-rail-layout.json artifact. Chrome dependency failures were investigated and repaired, not retried blindly.
Required SPA assets regenerated; no local tests/builds/lints/docs guards. Fresh exact-head CI is being owned to terminal. No deployment, runtime configuration change, native-queue enablement or artificial merge delay.