Filed by the domain:engine execution seat while landing #16933 (session_01XTBcV7zZHmokdyQgXjbyEU), using Claude Code. Filed bare and unassigned — no domain:*, no priority, no type; all three are triage's.
Routing input only, not a lane assignment. These land in packages/objectql. #16933 is fenced to packages/metadata-protocol and #15777 was fenced to packages/spec; neither seat may reach across, so this is handed over rather than folded in.
The sentences, measured on origin/main at merge base ebf9a4891
Same falsified premise as #15777 and #16933, same spelling, two sites — and after #16933 lands these are the last live carriers in repo SOURCE. (Five CHANGELOG.md files also carry the sentence — adapters/hono, client, metadata-protocol, plugins/plugin-hono-server and spec; those are compiled release history and must never be back-edited.)
packages/objectql/src/engine.ts
// One line per CALL, not per row, and only when the exemption was
// ASKED FOR and something was actually removed — the union is
// faithful because the strip is schema-uniform.
packages/objectql/src/validation/rule-validator.ts (docblock of preserveAuditIgnoredOnInsertWarning)
* ... and one line per CALL, not per row: the strip is schema-uniform, so
* the union of what the batch lost is the faithful signal (the same aggregation
* `onFieldsDropped` already applies).
Positive control on the same tree so the hits are not an instrument artefact: hookWrittenKeys returns 36 hits across packages/**/src/**, and git grep -c schema-uniform returns 1 for each of the two files above.
Why it is the same defect
The premise is falsified by the very ruling these files implement. Ruling C (#14147) put the static-readonly strip inside engine.insert after the beforeInsert hooks and exempted keys a hook itself assigned, per row — engine.ts builds rowHookWrittenKeys and passes hookWrittenKeys: rowHookWrittenKeys[i], and stripReadonlyFields honours it with if (hookWrittenKeys?.has(name)) continue.
The sharpest reading is that rule-validator.ts contradicts itself inside one file: the hookWrittenKeys option doc on stripRuntimeOwnedFields states the granularity as "Per ROW, never per call: one recording is armed for each row of a batch, so a hook that stamps row 3 cannot exempt row 4's caller-seeded record number", at :1411, while the docblock at :1592 — about 180 lines later, same file — justifies a per-call aggregation with "the strip is schema-uniform".
engine.ts's copy sits at :10614, 37 lines below the hookWrittenKeys: rowHookWrittenKeys[i] call at :10577 that it contradicts.
What is and is not being claimed
Both sentences justify one warning line per call carrying the batch union (preserveAuditIgnoredOnInsertWarning). Whether that aggregation is still the right shape is not decided here and may well be fine on the first reason alone — an operator warning has no per-row slot either. What is wrong is the reason given: it is a premise the same file falsifies, which is precisely the failure mode #15777 was opened for.
⭐ Whatever replaces it should say what is true today and cite the artefact that makes it true, never a schedule.
Prior art to match
Dedup
Run on the finding label, state=all, four pages of 100 read through repo-scoped REST and grepped locally. No card names either site: #15819 — the closest neighbour, which enumerated seven non-published sites of the superseded readonly-on-INSERT class — returns 0 for schema-uniform, and its one rule-validator.ts row is a different site (:1021, the stripReadonlyFields docblock's UPDATE-payload framing). Positive control fired in the same pass: #16933 and #15777 both matched schema-uniform, so the zero is a reading rather than a dead instrument.
Generated by Claude Code
Filed by the
domain:engineexecution seat while landing #16933 (session_01XTBcV7zZHmokdyQgXjbyEU), using Claude Code. Filed bare and unassigned — nodomain:*, no priority, no type; all three are triage's.Routing input only, not a lane assignment. These land in
packages/objectql. #16933 is fenced topackages/metadata-protocoland #15777 was fenced topackages/spec; neither seat may reach across, so this is handed over rather than folded in.The sentences, measured on
origin/mainat merge baseebf9a4891Same falsified premise as #15777 and #16933, same spelling, two sites — and after #16933 lands these are the last live carriers in repo SOURCE. (Five
CHANGELOG.mdfiles also carry the sentence —adapters/hono,client,metadata-protocol,plugins/plugin-hono-serverandspec; those are compiled release history and must never be back-edited.)Positive control on the same tree so the hits are not an instrument artefact:
hookWrittenKeysreturns 36 hits acrosspackages/**/src/**, andgit grep -c schema-uniformreturns 1 for each of the two files above.Why it is the same defect
The premise is falsified by the very ruling these files implement. Ruling C (#14147) put the static-
readonlystrip insideengine.insertafter thebeforeInserthooks and exempted keys a hook itself assigned, per row —engine.tsbuildsrowHookWrittenKeysand passeshookWrittenKeys: rowHookWrittenKeys[i], andstripReadonlyFieldshonours it withif (hookWrittenKeys?.has(name)) continue.The sharpest reading is that
rule-validator.tscontradicts itself inside one file: thehookWrittenKeysoption doc onstripRuntimeOwnedFieldsstates the granularity as "Per ROW, never per call: one recording is armed for each row of a batch, so a hook that stamps row 3 cannot exempt row 4's caller-seeded record number", at:1411, while the docblock at:1592— about 180 lines later, same file — justifies a per-call aggregation with "the strip is schema-uniform".engine.ts's copy sits at:10614, 37 lines below thehookWrittenKeys: rowHookWrittenKeys[i]call at:10577that it contradicts.What is and is not being claimed
Both sentences justify one warning line per call carrying the batch union (
preserveAuditIgnoredOnInsertWarning). Whether that aggregation is still the right shape is not decided here and may well be fine on the first reason alone — an operator warning has no per-row slot either. What is wrong is the reason given: it is a premise the same file falsifies, which is precisely the failure mode #15777 was opened for.⭐ Whatever replaces it should say what is true today and cite the artefact that makes it true, never a schedule.
Prior art to match
CreateManyDataResponseSchema.droppedFieldsjustifies its aggregated shape withschema-uniform, so every row drops the same set— ruling C's per-row hook exemption falsifies that premise #15777 / PR docs(spec): three prose carriers state what the tree does, not a premise it falsifies #16930 corrected thepackages/speccarrier: the aggregation kept its shape, the falsified second reason was deleted, and reader guidance was added — read a name as "at least one row dropped this field", not "every row did".packages/metadata-protocolcarriers the same way.Dedup
Run on the
findinglabel,state=all, four pages of 100 read through repo-scoped REST and grepped locally. No card names either site: #15819 — the closest neighbour, which enumerated seven non-published sites of the superseded readonly-on-INSERT class — returns 0 forschema-uniform, and its onerule-validator.tsrow is a different site (:1021, thestripReadonlyFieldsdocblock's UPDATE-payload framing). Positive control fired in the same pass: #16933 and #15777 both matchedschema-uniform, so the zero is a reading rather than a dead instrument.Generated by Claude Code