[finding] Two inline copies of the expected-read-refusal capture still wrap only the engine `error` channel — one of them can no longer fire since #13273 moved the frame to `debug`

[os-trump]

Found while working #13273; out of that card's scope, filed unassigned. Nothing is red — this is dead suppression plus stale prose, recorded so the next reader is not misled by it.

What changed underneath them

#13273 makes ObjectQL.find pick its failure log level from the cause: a read that positively identifies as "relation does not exist" (isMissingTableError) is logged at debug with a reason: 'table-not-provisioned' meta and no stack; every other read failure keeps error with the stack.

The shared helper packages/runtime/src/expected-read-refusal-noise.ts was updated in that PR: its captureEngine now proxies both error and debug, so the ~20 fixtures that assert silentChannels() keep counting the frame on whichever channel the engine chose. Those are fine.

The two inline copies that were not

captureExpectedReadRefusals was factored out (#10629) precisely to end the per-fixture copies, but two files still carry their own, and both wrap error only:

• packages/runtime/src/notification-schema-conformance.integration.test.ts — captureExpectedAbsentTableNoise().wrapEngine, gated on a pending counter fed by no such table: <t> driver refusals. Since os migrate plan prints 4 ERROR records with stack traces on a not-yet-migrated database — every one of them is a read whose caller treats absence as a normal answer #13273 those engine frames arrive on debug, so the recognition arm can no longer match: the frames fall through to target.debug and are dropped by the fixture's own kernel level. The file stays green because everything it asserts (withheld) is fed by the DRIVER channel, which is untouched. Its header's measurement — "63 [sql-driver] DATABASE_ERROR … lines and 63 matching ERROR Find operation failed frames" — is now a historical statement about origin/main at [finding] Two PASSING @objectstack/runtime tests print Acquire connection error and ~150 ERROR-level sql-driver DATABASE_ERROR lines into the shared Test Core log — they were adopted verbatim as a p1 flake signature #10380 time, not about this tree.
• packages/runtime/src/metadata-list-ambient-vs-bare-transaction.integration.test.ts — engineError. ⚠️ This one is NOT affected: its declared fault is Timeout acquiring a connection, which is not a missing table, so its frame legitimately stays on error and its withheld.engineFind assertion still fires. Named here only so a reader does not "fix" it by symmetry.

Suggested disposition

Migrate the first file onto the shared captureExpectedReadRefusals helper — the reason the helper exists, and it carries the debug arm already — and re-measure the header's counts against the current tree while doing it. Leave the second alone.

Low priority: no assertion is weakened, and no noise leaks (the frames are quiet by construction now rather than by suppression).

[os-trump]

Claim: domain:cli PM seat · session session_01TvqBFLRzXdSPcbusDoED9k · branch claude/issue-13325-migrate-inline-refusal-capture

Prerequisite verified, not assumed: this card's premise is that #13273's frame moved to debug and that the shared helper gained its debug arm. That shipped in PR #13327, merged 2026-08-30T04:42:09Z. Before that merge this card described a future; now it describes main.

Serial constraints cleared: ✅ yes. packages/runtime/src/expected-read-refusal-noise.ts was under a same-file hold by #13327 — released by that same merge — and this card consumes the helper rather than editing it. notification-schema-conformance.integration.test.ts is touched by no open PR from this lane.

⛔ Fences

• ⛔ Leave metadata-list-ambient-vs-bare-transaction.integration.test.ts alone. The card names it only so nobody repairs it by symmetry: its declared fault is Timeout acquiring a connection, which is not a missing table, so its frame legitimately stays on error and its withheld.engineFind assertion still fires. Touching it would break a working pin in the name of consistency.
• ⛔ Nothing here is red, and nothing may become red. This is dead suppression plus stale prose. No assertion may be weakened, dropped, or relaxed in the migration — the file's withheld assertions are fed by the driver channel and must keep passing exactly as they do now.
• ⚠️ Re-measure, do not transcribe. The header's "63 [sql-driver] DATABASE_ERROR … lines and 63 matching ERROR Find operation failed frames" is now a historical statement about origin/main at [finding] Two PASSING @objectstack/runtime tests print Acquire connection error and ~150 ERROR-level sql-driver DATABASE_ERROR lines into the shared Test Core log — they were adopted verbatim as a p1 flake signature #10380 time, not about this tree. Whatever number replaces it must come from a run on the current tree, and the prose must not keep asserting a frame arrives on error when it now arrives on debug.

Why it is worth doing at p2 even though nothing fails: the recognition arm can no longer match, so the fixture's engine-side suppression is dead code that reads as live protection. That is the same shape as the defect #13271 and #13318 just fixed one package over — a mechanism whose prose says it is doing something it cannot do — and the reason captureExpectedReadRefusals was factored out in #10629 was precisely to end per-fixture copies like this one.

---

Generated by Claude Code

[os-trump]

os-dev-report

{
  "issue": 13325,
  "status": "done",
  "branch": "claude/issue-13325-migrate-inline-refusal-capture",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/13352",
  "premise_still_valid": true,
  "summary": "Verified the premise before writing code: origin/main carries PR #13327 (3a86a65e7), the built @objectstack/objectql dist carries the 'table-not-provisioned' marker, and the shared helper's debug arm is present. Migrated packages/runtime/src/notification-schema-conformance.integration.test.ts off its 70-line inline captureExpectedAbsentTableNoise copy and onto the shared captureExpectedReadRefusals: captureDriver/captureEngine replace install/wrapEngine, and the afterAll assertion moves from the driver-only withheld.has(table) loop to silentChannels(ALWAYS_READ_AUTHZ_TABLES), which is strictly stronger because it also requires the engine channel to have fired per table. The header prose is rewritten: it no longer claims the engine frame arrives on error, and its counts are re-measured on this tree rather than transcribed. FENCE HELD: metadata-list-ambient-vs-bare-transaction.integration.test.ts is not in the diff (git diff --name-only against merge base 0783d7b80 returns exactly one path; the remote PR reports changed_files 1). expected-read-refusal-noise.ts is not edited either -- it was ablated and restored byte-identically as a measurement only. No changeset: test-fixture-only diff that publishes nothing; skip-changeset applied. PROCESS INCIDENT, self-caught and repaired: my first PR-page label extractor was written for the ISSUE page link form and had no positive control on a PR page, so it read the PR's labels as empty when they were size/m and tests; the whole-set MCP label write then stripped both. The comparative read-back caught it, and a union write re-hung all three (read-back: size/m, skip-changeset, tests -- none missing, none extra). Nothing else on the PR was touched.",
  "tests": "All measurements on the tree at 5293ddc76 (final commit), clean working tree. DEAD-ARM PROOF (before, at merge base 0783d7b80): instrumented the inline capture with per-channel counters; mutation proven on disk (marker present, removed anchor at 0 occurrences, blob ad54f793a vs HEAD blob 2c4e4aecd) and restore proven the same way (blob back to 2c4e4aecd, 0 residue, git diff HEAD empty, git status --porcelain empty). Result: driverRefusals 63, errorCalls 0, errorFindFrames 0, errorMatches 0, debugCalls 417, debugFindFrames 63, debugWouldMatch 63. POSITIVE CONTROL for the zeros: the arm's own predicate matched 63 of 63 frames on debug, so the zero is the wrong channel, not a broken predicate. ALIVE-AFTER PROOF: post-migration tally read off the capture itself -- totalRefusals 63, totalEngineFrames 63, per table 12/12/12/12/12/3 on both channels, silentChannels() = []. REVERSE VERIFICATION (direction predicted first: RED): ablated the shared helper's debug arm, restoring its pre-#13327 shape. Observed exactly the predicted red -- 'expected [ 5 items ] to deeply equal []' naming the engine's Find operation failed frame for sys_user, sys_member, sys_user_position, sys_user_permission_set, sys_position; Test Files 1 failed, Tests 8 passed (only the newly live afterAll assertion reds; the driver-fed cases stay green). Ablation rebuild condition: NOT required for this leg -- the helper is a relative same-package import vitest resolves to source, not an exports entry into dist; the ENGINE under test IS resolved through dist, so @objectstack/objectql was built first (pnpm --filter '@objectstack/runtime^...' build, exit 0) and its dist was confirmed to carry 'table-not-provisioned' before any measurement. Mutation proven on disk (marker injected once, debug-arm anchor at 0 occurrences, blob 982555d52 vs HEAD blob 064a19632); restore proven (blob back to 064a19632, 0 residue, git diff HEAD empty, clean status), and the migrated file re-run green afterwards. GREEN RUNS: migrated file 'Test Files 1 passed (1) / Tests 8 passed (8)'; migrated file plus notifications.hono.integration.test.ts plus the helper's two pin suites 'Test Files 4 passed (4) / Tests 28 passed (28)'; whole affected package 'Test Files 201 passed (201) / Tests 2968 passed (2968)'; repo-wide pnpm lint (eslint . --no-inline-config) exit 0, no narrowing needed. GATES derived with 'node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack' (it reported the change set as 1 path vs merge base 0783d7b80, three-dot, at commit 5293ddc76) and then run. Green, quoting each gate's own verdict line: check:nul-bytes 'OK (scanned 7390 text file(s) ... no raw ASCII control bytes)'; check:cross-package-test-inputs 'OK: 24 package(s) read outside themselves, all declared'; check:engine-double-contract 'OK - 709 pinned, 134 in the DEBT ledger, 3 exempt'; check:where-matcher '316 matcher(s) discovered ... 0 silently-wrong and 0 unjudged ... none new'; check:query-options-erasure 'ratchet holds: 67 unswept non-test site(s) ... none new'; check:test-source-alias 'OK - 72 packages with tests scanned'; check:type-check-coverage 'OK - 66/78 workspace packages type-checked'; check:type-check-debt '--re-measure: OK - 30 ledger entr(ies) re-measured in 234.3s, 1558 raw tsc error(s) total, none above its recorded number' with 'surplus: none'; check:dual-build-cjs-loads '102 published require entry point(s) across 66 package(s) load; 610 emitted CommonJS file(s) parse' (after a full turbo build, 71 successful / 71 total); plus check:logger-receiver-detach, check:objectql-double-limit, check:page-declaration-shape, check:published-files, check:slot-lookup, check:type-source-resolution and the nine node scripts/ gates (ci-filter-parity, comment-mask-adoption, cross-package-test-inputs, keyed-text-bounds, plugin-teardown-shape, shard-attestation, undeclared-dep-imports, docs-audit/check-affected-docs, docs-audit/check-drift-comment) all exit 0. Exit codes captured before any pipe. NOT MEASURED, recorded as such and never folded into the green list: (1) node scripts/check-test-completeness.mjs exit 3, its own text 'PREREQUISITE NOT MET - this gate grades a saved turbo run test log, and no log was named' plus 'running the family locally, record this gate as NOT MEASURED'; (2) pnpm --filter @objectstack/runtime typecheck exits 0 but says nothing about the edited file -- packages/runtime/tsconfig.json excludes **/*.test.ts and tsc --noEmit --listFiles returns 0 occurrences of the edited path. That hidden test layer IS measured, by check:type-check-debt --re-measure, which re-ran the @objectstack/runtime TEST_DEBT entry (217) and reported it not above its recorded number.",
  "mcp_calls": "5 - create_pull_request 1, pull_request_read 1, issue_write 2 (label write plus the union repair), add_issue_comment 1 for this report. Issue body, comments and the PR read-backs went through the zero-quota public-repo page payload; the repo-scoped REST probe returned 403 for this seat and gh is absent, so MCP was used for writes only.",
  "open_questions": [],
  "out_of_scope_findings": []
}

---

Generated by Claude Code