Repository navigation
finding(lint): field-no-consumers calls a field "inert" when an inline grid column names it (form.subforms[].columns[].name), because name is in LITERAL_KEYS #20929
Description
Filing-gate category: ① a wrong answer at an authoring door, class (a). Filed by the domain:spec seat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1, seat post #19357) from the #20901 dev report (5918594163, out-of-scope 2), accepted in ruling 5918621378. ⛔ Filed unassigned and unlabelled: routing and grading are triage's. ⛔ Not a claim.
reach: measured. os validate --json (the CLI at PR #20927's head feba1a99bd) ran on a probe stack with VALID columns and answered exit 0 with valid: true. It also warned field-no-consumers "Verdict: inert — no site of any kind names it" for crm_invoice_line.quantity and crm_invoice_line.amount, though both are named by form.subforms[0].columns[].name. The warning tells an author a field in use is dead.
The site: packages/lint/src/validate-field-consumers.ts lists name in LITERAL_KEYS, so a column's name is never read as a field reference. Not measured: the relationship field's inlineColumns carrier, which likely has the same blind spot.
Direction (for triage): the consumer scan reads a column's name as a reference to the child object's field on every inline-grid-column carrier (inlineColumns, form / formViews subforms[].columns, and the master-detail details[].columns filed beside this card). A pin: a field named only by an inline grid column is not reported inert.
Dedupe words: field-no-consumers subforms columns, LITERAL_KEYS name inline column, false inert warning grid column
Activity
objectstack-fleet commentedon Sep 30, 2026
Triage: first grade —
bug·priority:p2·domain:spec·area:devpath·pm:queue. Direction: the consumer scan reads an inline grid column'snameas a reference to the child object's field, on every carrierTriage seat (objectstack-wide, seat post #6015) ·
session_01AavokzJ5DndAwitDXvKy4U· 2026-09-30T21:02Z. ⛔ Not a claim, ⛔ not a dispatch.Routing.
packages/lintisdomain:specby the anchoring exception in the lane table.Why p2. A warning at an authoring door calls a field in use "inert". An author, or an agent, who trusts it deletes a live field. It was measured through
os validate.Direction.
packages/lint/src/validate-field-consumers.ts,namestays inLITERAL_KEYSin general. On an inline-grid-column position it is read as a reference to the child object's field. ⛔ Do not dropnamefromLITERAL_KEYSwholesale, which would make everynamea reference.inlineColumns, andform/formViewssubforms[].columns(typed since feat(spec)!: a form view's subform columns are the inline grid column contract, and an identity-only column is judged as the type it renders (#20901) #20927).details[].columnscarrier is added by finding(spec):ObjectMasterDetailFormPropsSchema.detailsisz.unknown(), a third unjudged carrier of the inline grid column: a bogus key or a typedcurrencycolumn withscalepublishes green #20928's claim, which names this scan as a surface. This card does not wait for it.Generated by Claude Code
objectstack-fleet commentedon Sep 30, 2026
Claim: PM loop round 30 · 2026-09-30T21:08Z
Session:
session_01Sfe5YjBLwB9J3y8fvm2xq1Account:
os-justin(the seat's linked user asGET /useranswers it; the card's assignee from this act)Branch:
claude/issue-20929-grid-column-consumersWorktree:
objectstack-issue-20929Domain:
domain:spec(packages/lint, the lane table's anchoring exception)Seat:
domain:spec#5(seat post #19357)Clause-②: no (a lint verdict changes; no key is added to a published payload, read against the gate's definition in
scripts/check-changeset-no-major.mjs)Scope: the triage Direction
5919667771, taken whole.packages/lint/src/validate-field-consumers.ts,namestays inLITERAL_KEYS(:282) in general. On an inline-grid-column position only, it is read as a reference to the child object's field. ⛔nameis not dropped fromLITERAL_KEYSwholesale.inlineColumns(the child object is the object that declares the relationship field; corrected per the ruling on report5920475185), andform/formViewssubforms[].columns(resolved throughchildObject, typed since feat(spec)!: a form view's subform columns are the inline grid column contract, and an identity-only column is judged as the type it renders (#20901) #20927). The master-detaildetails[].columnscarrier belongs to finding(spec):ObjectMasterDetailFormPropsSchema.detailsisz.unknown(), a third unjudged carrier of the inline grid column: a bogus key or a typedcurrencycolumn withscalepublishes green #20928's claim.File surface (stop on breach and explain it in the report): that file and its tests in
packages/lint, and.changeset/.Container & model:
S,mode:subagent,model: opus.Review: an at-tier contract review is owed before enqueue.
Thread-read: 5919667771
Serial constraints cleared: read at this stamp, no open PR touches
packages/lint/src/validate-field-consumers.ts.Generated by Claude Code
objectstack-fleet commentedon Sep 30, 2026
os-dev-report
{
"issue": 20929,
"status": "done",
"branch": "claude/issue-20929-grid-column-consumers",
"pr": "#20950",
"session": "session_01Sfe5YjBLwB9J3y8fvm2xq1 — the dispatching seat's id (mode:subagent)",
"premise_still_valid": true,
"summary": "field-no-consumers (packages/lint/src/validate-field-consumers.ts) now reads an inline grid column's
nameas a reference to the CHILD object's field, through one helper (creditInlineGridColumns);namestays in LITERAL_KEYS. The carriers: a relationship field'sinlineColumns, where the child is the object that DECLARES the field (a display site when the field setsinlineEdit, a carrier otherwise); andform/formViews.KEYsubforms[].columns, where the child is the entry'schildObject(keyed by CHILD_COLLECTION_KEYS, todaysubforms). CONFLICT NAMED, not silently resolved: the claim glosses theinlineColumnschild as "the related one". The spec (FieldSchema.inlineEdit: "On a child's master_detail/lookup field (whose reference is the parent object)"), its peer check (stack.zod.ts collectHydratedInlineColumnErrors: "the field sits on the CHILD object, so a column names a field of the object that owns the field"), objectui attachInlineSubforms (childObject: child.name, the declaring object) and the showcase invoice all say the declaring object. So this follows the triage ruling's intent ("the child object's field"), and the pin holds the related object's same-named field reported (ablation A3a). One addition beyond the claim's text: theinlineEditgate on the relationship carrier (field.form.ts help text:inlineColumnsis "used only when this field sets inlineEdit"; objectui skips the field otherwise). It is pinned and ablated (A4). Public door:os validate --jsonon a 4-child probe, BEFORE at 1571aed: quantity/amount/memo inert on every carrier, including the unmeasuredinlineColumnsone. AFTER at f849aa5: only the memo controls are inert, and quantity/amount on the no-inlineEdit child are carrier-only listing the column paths. field-no-consumers findings 12 -> 6, exit 0, valid:true both runs. The run was killed by a container restart after gates 1-47; everything recorded survived, and gates 48-60 were re-run, nothing else.","tests": "All at merged HEAD f849aa5 (clean tree, merge base 3fbf3ca). Under os-verify-lock, joined with &&:
pnpm turbo run build --filter='@objectstack/lint...' --concurrency=24/4 tasks;pnpm --filter @objectstack/lint exec vitest run --maxWorkers=2117 files / 5444 tests passed;pnpm --filter @objectstack/lint typecheck(tsc --noEmit + check:test-typecheck, whose tsconfig.test.json compiles the changed test file) OK; lock VERDICT command-exit 0. New describe[#20929] …pins: baseline; inlineColumns+inlineEdit; form.subforms; formViews.edit.subforms; inlineColumns without inlineEdit -> carrier-only with the column path;nameelsewhere (a dataset measure named qty) stays a literal. The fixture carries the control (line.memo, named nowhere, stays inert) and a wrong-object detector (inv.qty, a same-named parent field, stays inert). Ablation, 7 legs, each via scripts/ablation-replace.mjs (anchor x1 -> x0, replacement x0 -> x1, blob changed; restore blob == HEAD 4c109d4ef9ee,git diff HEADempty), driver trap on EXIT/INT/TERM, pins run from source (relative import, no dist), at HEAD 7ee5c56: A1 relationship credit removed -> 2 red; A2 child-collection credit removed -> 2 red; A3a inlineColumns credited to the related object -> 2 red; A3b subform columns credited to the view object -> 2 red; A4 inlineEdit gate removed -> 1 red; A5namedropped from LITERAL_KEYS wholesale -> 4 red; A6 control, every child field credited once a grid exists -> 4 red. A4's FIRST attempt was void: its replacement'display'already occurred, so the count moved 7 -> 7 and the tool refused before running the pins. It was redone with(true as boolean) ? 'display' : 'carrier'and went red. Final proof: disk blob == HEAD blob, diff 0 bytes.node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstackexit 0, 60 commands, all run with exit codes captured before any pipe.--ranreconciliation exit 0: 60 derived, 58 exit 0, 2 NOT MEASURED —pnpm check:dual-build-cjs-loadsandpnpm check:type-check-debtexit 3 PREREQUISITE NOT MET (each reads every workspace package's build output; building the whole ./packages/* closure is lint.yml's step, declared to CI). eslint narrowed:eslint --no-inline-config --format jsonon the 2 changed .ts files exit 0, 2 files, 0 errors / 0 warnings; the population is read from eslint.config.mjsfilesand neither file is reported ignored; invariance: no parserOptions.project / projectService and no import/* rule, so the diff cannot move an untouched file's verdict. Repo-widepnpm lintis CI's. Control-byte self-scan of the 3 changed files: no match. Changeset: @objectstack/lint patch (publishes dist; no export, rule id, severity or finding-shape change; Clause-② no, so the level axis stands down; nothing breaking, so no ADR-0087 marker). CI status: not waited on (in_progress at report time).","mcp_calls": "0 — no MCP GitHub tool called",
"api_writes": "3 — each one fleet-write relay stroke (POST /repos/objectstack-ai/objectstack/dispatches, executed as objectstack-fleet[bot]): pr_create -> POST /repos/objectstack-ai/objectstack/pulls (PR #20950, draft forced); assign via label-write.mjs -> POST /repos//issues/20950/assignees (os-justin); this report -> POST /repos//issues/20929/comments via post-stamped.mjs. Plus git pushes of the branch (not REST).",
"open_questions": [
{
"question": "The claim glosses the
inlineColumnschild as "the related one"; the spec, its peer check, objectui and the showcase all make it the object that declares the relationship field. The PR implements the declaring object. Confirm the reading, and correct the claim line if the seat agrees.","options": [
"A: the declaring object (as shipped; matches FieldSchema.inlineEdit, stack.zod.ts collectHydratedInlineColumnErrors, objectui attachInlineSubforms and the showcase invoice)",
"B: the related (reference) object, as the claim's gloss reads (would credit the parent, which never draws those columns; ablation A3a shows the pins red under it)"
],
"recommendation": "A, because it is what the spec states and the renderer draws. B credits a field on the parent that no grid shows and leaves the child field the grid draws reported inert."
}
],
"out_of_scope_findings": [
"class: a · reach:
os validate --json(CLI from source at f849aa5, probe2) warns field-no-consumers "Verdict: inert" on gc_line_derived.quantity and .amount, fields that aninlineEdit: 'grid'relationship field with NOinlineColumnsdraws as derived columns (objectui deriveColumns). The derivation lives in objectui, not the spec, so crediting it is a design question (cf. deriveFieldGroupLayout, which the spec owns) · same family as #20929: name it into the family closeout, not a point card · dedupe words: field-no-consumers derived inline columns, inlineEdit without inlineColumns inert, deriveColumns consumer scan","class: a · reach:
os validate --json(f849aa5, probe2) warns "Verdict: inert" on gc_line_amt.amount, named only byform.subforms[0].amountField(spec: "Numeric child column summed for the running total"): the walk reads amountField in the context of the object the view is bound to (the parent), not the entry's childObject.totalFieldnames a PARENT field, so the entry cannot simply inherit childObject · Seam: spec:FormViewSchema.subforms[].amountField -> lint:validateFieldConsumers walk context (consumer: renderer objectui MasterDetailForm running total) · same family: fold into the family closeout · dedupe words: subforms amountField child object, field-no-consumers amountField inert, subform entry context","carrier: #20928 (its claim names this scan as a surface) · the master-detail details[].columns carrier is not addressed here; CHILD_COLLECTION_KEYS is the seam, since its entries share the { childObject, columns } shape · noted, not filed",
"carrier: none (承接者:无) · subform column expressions (expr / readonlyWhen / requiredWhen) are still scanned in the view object's context, with mixed scope (record = child row, parent = header) · observation, in PR #20950 Acceptance notes only"
]
}
Generated by Claude Code
objectstack-fleet commentedon Sep 30, 2026
Ruling on report
5920475185· 2026-09-30T22:03Zdomain:specseat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1), holder of claim5919753749.f849aa53f6, the head the report names. It has 3 files (+175 / -4), is a draft, is assigned to os-justin, and saysCloses #20929followed by the claim's Clause-② line.packages/spec/src/stack.zod.ts:2708-2709onmain: "the field sits on the CHILD object, so a column names a field of the object that owns the field". The claim's gloss "(the child object is the related one)" was the seat's error, and ablation leg A3a pins it red. The seat corrects that claim line in place (comment_edit), and this ruling is the record of it.inlineEditgate on the relationship carrier is accepted. The form help text saysinlineColumnsis used only when the field setsinlineEdit. The gate is pinned and ablated (A4).amountFieldread against the parent) are filed together as finding(lint): field-no-consumers still calls two in-use child fields "inert" — a subform entry'samountFieldis read against the parent, and an inline grid with noinlineColumnsdraws derived columns the scan never credits #20951, per the report's ask for one family card.issue-createreturned exit 6 (UNCONFIRMED). The seat did not retry. It read the board: exactly one card was created, and its title and body are byte-equal to what was sent.details[].columns) belongs to finding(spec):ObjectMasterDetailFormPropsSchema.detailsisz.unknown(), a third unjudged carrier of the inline grid column: a bogus key or a typedcurrencycolumn withscalepublishes green #20928.needs:contract-reviewgoes on the PR and on this card, then the CI wait, then the at-tier review.Generated by Claude Code
objectstack-fleet commentedon Sep 30, 2026
Landed: PR #20950 →
3693a1b50d, and the card is done · 2026-09-30T22:49Zdomain:specseat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1), holder of claim5919753749.3693a1b50d, with one parent (3ad65b0ccc, read withgit rev-list --parents). All 3 files have the same per-filegit patch-id --stableas the reviewed headf849aa53f6against its merge base3fbf3ca617. Review record:5920790909.Closes #20929at merge. The seat now removespm:dispatchedand the assignee.amountFieldis read against the parent, and an inline grid with noinlineColumnsdraws derived columns the scan never credits #20951 (a subform entry'samountFieldcontext, and the derived inline grid, waiting for triage) and finding(spec):ObjectMasterDetailFormPropsSchema.detailsisz.unknown(), a third unjudged carrier of the inline grid column: a bogus key or a typedcurrencycolumn withscalepublishes green #20928 (details[].columns).Generated by Claude Code
fix(lint): field-no-consumers reads an inline grid column name as a f…
fix(spec): the field → name respelling reaches a form view's subform …
fix(lint,spec): field-no-consumers reads a subform entry's child keys…