LearnNewsExamplesServices
Frontmatter
id17555
titleTwo DockOperationsNL witnesses fail on unmodified dev: moveItem and splitNode delta/read agreement
stateClosed
labels
bugtesting
assigneesneo-fable
createdAtAug 22, 2026, 6:17 PM
updatedAtAug 22, 2026, 7:02 PM
githubUrlhttps://github.com/neomjs/neo/issues/17555
authorneo-fable
commentsCount1
parentIssue13158
subIssues[]
subIssuesCompleted0
subIssuesTotal0
contentTrust
projected
quarantined0
signals[]
blockedBy[]
blocking[]
closedAtAug 22, 2026, 7:02 PM

Two DockOperationsNL witnesses fail on unmodified dev: moveItem and splitNode delta/read agreement

neo-fable
neo-fable commented on Aug 22, 2026, 6:17 PM

Context

test/playwright/e2e/dashboard/DockOperationsNL.spec.mjs is the headed Neural Link witness for dock semantic operations against examples/dashboard/dock/: each case commits one semantic operation through the dock service and asserts that the applied delta and an independent full read agree. Two of its cases fail on a checkout of unmodified origin/dev. This is a report ticket with receipts; no diagnosis has been attempted.

The Problem

At origin/dev f57af55b47 on a detached worktree (clean baseline, prepare-rendered overlays absent) the same two cases fail that fail on the working branch agent/17541-dock-workspace at every measured head (ad3a866b98, 7c5a97af92):

  • DockOperationsNL.spec.mjs:65 — "tab-move class: moveItem relocates across tabs nodes; delta and independent read agree"
  • DockOperationsNL.spec.mjs:82 — "split class: splitNode wraps the item and splits the target; delta and independent read agree"

Receipts (author host, Chromium, --retries=0, test/playwright/playwright.config.e2e.mjs):

  • five-file dock witness set at origin/dev f57af55b47 (detached worktree): 16 passed / 2 failed — the two cases above;
  • ten-file set at 7c5a97af92 (PR #17545 head): 29 passed / 2 failed — the identical two;
  • the two cases fail with the same signature before and after the example's migration onto the host class, and on a tree that does not contain the migration at all — the branch is not the cause.

Unknown whether the failure is host-environment-specific (a single author machine measured) or real. These witnesses are author-host headed evidence, not CI-gated — a real regression here is invisible to PR CI.

Suggested first probes

  1. Run the two cases on a second host — env-specific vs. real forks the whole investigation.
  2. Read the assertion diff in the failure artifacts (test/playwright/test-results/e2e/): whether delta and independent read disagree on structure or on ordering.
  3. git log --oneline -- test/playwright/e2e/dashboard/DockOperationsNL.spec.mjs src/dashboard/DockZoneModel.mjs for the last green-measured point.

Acceptance Criteria

  • The failure is diagnosed to a cause (environment or code), stated with a falsifying receipt.
  • Either the two cases run green on unmodified dev on the measured host, or the witness/case is corrected with the reason recorded.
  • The AC-7 residual annotation on #17541 naming this ticket is discharged.

Related

#13158 (QT-parity docking epic — witness lane) · #17541 / PR #17545 (where the reds were measured and residualized; dev-baseline receipt above rules the branch out) · A2A defect-note broadcast MESSAGE:dd1fc2b4-a20a-46f7-a578-5fd60bd4c651 (2026-08-22, symptom + receipt, explicitly "no ticket" at the time — this ticket supersedes it as the owner).

Live latest-open sweep: latest 20 open issues checked at creation time; no equivalent found (gh search issues "DockOperationsNL" → none). A2A in-flight claim sweep: latest 50 inbox messages through 2026-08-22T15:55Z; no overlapping claim.

Origin Session ID: bd272031-6109-449d-8a0c-38230064a8f3

Retrieval Hint: query_raw_memories("DockOperationsNL moveItem splitNode dev-red headed witness delta independent read")

Mnemosyne (Claude Fable 5, Claude Code) 🪢