finding(fields): RichTextField reads a rows key off field metadata that neither MarkdownFieldMetadata nor HtmlFieldMetadata declares — undeclared-but-consumed, reached through an as any

Filed unassigned from objectui#6138 round 2, which ruled that rich-text.mdx's five editor keys get determined and filed with the answer in them rather than filed as a question. The answer is measured below. The five keys come out of that page's annotated Field Schema block either way; this card decides whether one of them comes back as a declared member.

The measurement — one of five is consumed, four are inert

Read: packages/fields/src/widgets/RichTextField.tsx (the single widget behind the markdown, html and richtext registry keys — objectui#5498).

key documented on `rich-text.mdx`	read by `RichTextField`?	evidence
`rows`	YES	`const rows = richField?.rows \|\| 8;`, passed to `RichTextEditorSurface` and onto `Textarea`'s `rows={fullHeight ? undefined : rows}`
`toolbar`	no	the only occurrence in `packages/fields/src` is a JSDoc sentence about what the editor might grow into — "whatever the editor grows into (a toolbar, a TipTap/Lexical instance, a preview toggle)"
`preview`	no	no read on this widget; the other `preview` hits in `packages/fields/src` belong to `VectorField`, `RecordPickerDialog` and `ImageLightbox`
`minHeight`	no	zero occurrences anywhere in `packages/fields/src`
`maxHeight`	no	zero occurrences anywhere in `packages/fields/src`

Command:

grep -rn "\btoolbar\b\|\bpreview\b\|\bminHeight\b\|\bmaxHeight\b\|\brows\b" \
  packages/fields/src --include=*.ts --include=*.tsx | grep -v __tests__

So four of the five are inert prose and are simply deleted by objectui#6138. rows is the finding.

Why `rows` is a defect rather than a working feature

The read is real and reaches the DOM, but it is undeclared on both metadata types the widget can be handed:

MarkdownFieldMetadata — 18 members, rows not among them
HtmlFieldMetadata — 18 members, rows not among them

(Key sets read from the built packages/types/dist/index.d.ts through the doc gate's own derivePackageTypePaths(), so they are the shipped surface, not source.)

The read only compiles because the widget launders the carrier first:

const richField = field as any;
const rows = richField?.rows || 8;

field is typed FieldMetadata, so without that as any the read would not compile. The cast is what lets an undeclared key be consumed silently — and it is the same cast that reads placeholder, label and mobile_fullscreen, all three of which are declared on BaseFieldMetadata. rows is the one member of that group with no declaration behind it.

Consequences, both live today:

An author cannot legally write it. A markdown field carrying rows is a key @objectstack/spec's strict schemas reject at publish, and a MarkdownFieldMetadata literal carrying it fails the excess-property check — while the running widget honours it. Declared and enforced disagree in the direction that is hardest to notice: the feature works in the app and fails at authoring time.
The sibling proves the shape. TextareaFieldMetadata does declare rows, and TextAreaField reads it the same way (const rows = textareaField?.rows || 4;). Two long-text widgets, the same capability, one declared and one not — so this is a gap in one type rather than a question about whether the capability is wanted.

The two candidate resolutions

A — declare it. Add rows?: number to MarkdownFieldMetadata and HtmlFieldMetadata, matching TextareaFieldMetadata. Makes the consumed key authorable, and rich-text.mdx's Field Schema block can then carry it. Needs the corresponding @objectstack/spec question answered — the renderer type is not the publish contract.
B — remove the read. Drop richField?.rows and let the inline editor keep its fixed height. Consistent with "declared = enforced" in the other direction, and cheaper, but it removes a capability the sibling widget has and an author may already be using.

⚠️ Not decided here, deliberately: the choice turns on whether @objectstack/spec should carry rows for markdown/html, which is a spec question and not this repo's to answer alone. What this card fixes is that the question was previously invisible.

Refs

objectui#6138 (which ruled this be filed with its answer) · objectui#5498 (the one widget, three registry keys) · objectui#3233 (field as the single metadata carrier) · objectui#3221 (why the closed contract matters).

Generated by Claude Code

Triage (triage seat, session session_01PCuhUdqwK7zkkxgtJHswTc): graded into the decision inbox — whether @objectstack/spec carries rows on markdown/html authorable metadata is an acceptance-set expansion (human floor), and the card itself rules it "a spec question and not this repo's to answer alone". → needs-user-decision, domain:spec (contract face per the 2026-08-21 three-way split), type Feature. This card is the mother card for the undeclared-but-consumed family this sweep measured: #6153's two instances (dependsOn on select metadata, options[].description on lookup) ride the same ruling frame and are parked pm:blocked on this one. 四维分析如下(中文,业务视角):

一句话问题:富文本字段的行高(rows)在应用里真实生效,但按规则书写的作者写了它会被发布校验响亮拒绝 —— 功能在跑,合法作者却写不进去;同门的多行文本字段(textarea)早就正大光明声明了同一个键。

选项 × 真实代价:

A 声明它:MarkdownFieldMetadata/HtmlFieldMetadata 加 rows?: number,并在 @objectstack/spec authorable 面同步声明 —— 业务上 = 把已经在跑的能力补进说明书,作者(人与 AI)从此可合法使用;代价:契约面 +1 键,spec 与 objectui 两仓各一笔,永久维护义务。
B 删掉读取:widget 不再读 rows,行内编辑器固定高度 —— 业务上 = 收回一个没人承诺过的旁门,和「声明即强制」同向;代价:已有作者若在库里写了 rows(经 as any 生效)会静默失去行高,且同门 textarea 有此能力、富文本没有,产品面不对称。
(家族卡 finding(fields): two more undeclared-but-consumed metadata keys — dependsOn on select and options[].description on lookup, both reached past the declared type #6153 的两个键各自套同一框架裁:dependsOn 有已声明的 snake_case 兄弟 depends_on,大概率是 B/迁移拼写;options[].description 是真被搜索消费的行为,更像 A。)

四棱:
① 项目长远合理性:两案都把 declared=enforced 拉回一致,方向都对;A 以最小增量对齐兄弟字段(textarea 先例),不新造形状;B 制造同门能力不对称。
② 实际业务拉动:rows 有真实消费(widget 读并落到 DOM),textarea 先例证明该能力是被要的;但本仓测不到今天有多少存量 app 实际写了这个键(见置信缺口)。
③ 防 AI 犯错:今天的状态最糟 —— AI 抄运行中的 app 学会 rows,发布时被拒,或反过来从类型学到「不能写」而实际有效;A 和 B 都消除这个双面教材,A 后作者面与运行面说同一句话。
④ 创业阶段不扩散:+1 键是最小义务,且是「把既成事实入册」而非新能力;若倾向从紧,B 也成立,但要接受删行为的回归风险。
推荐:A(rows 补声明,对齐 textarea 先例),回退 B。低摩擦裁决:回 A/B;「A,但只 objectui 类型层、spec 面缓议」也可,自动落为 objectui 先行 + spec 卡另立。
置信缺口:本分析看不见存量客户 app 里 rows/dependsOn/description 的真实书写量(需 hotcrm/示例仓与云端元数据抽样才知道删读取伤不伤人);也未读 #4631 三面失和总卡是否已有裁决方向。

裁后执行(你不用管):A ⇒ objectui types/zod + spec authorable 面各一笔,spec 半边转 domain:spec 席按 Clause-② 走契约复审档位;B ⇒ ui 席删读取 + 文档不回写;#6153 两键按同框架由分诊代批注继承裁决理由逐键落地,不再回决策箱。

Generated by Claude Code

Marker repair (triage seat, same session): the machine-findable facets marker in the previous comment was stripped by the sanitizer (raw HTML-comment form). Re-anchoring it here in the escaped spelling the extraction greps for — the four-facet block itself is the previous comment and stands unchanged.

Generated by Claude Code

Maintainer ruling (2026-08-25, in-session batch adjudication of the decision inbox; verbatim: 「就全部接受，然后继续下一批」— all five batch-1 recommendations accepted): Option A — declare rows?: number on both MarkdownFieldMetadata and HtmlFieldMetadata, aligning the TextareaFieldMetadata precedent, so validation and the running widget finally agree.

Scope riders accepted with it: same-stroke docs cleanup of the four inert editor keys rich-text.mdx teaches (toolbar/preview/minHeight/maxHeight) — remove or mark them not-yet-implemented in docs; if any of them is also declared in types, that is a separate enforce-or-remove item, do not fold it. The capability expansion stops at rows — the four inert keys are NOT being made real under this ruling.

State transition: needs-user-decision → pm:queue (label swap by triage; if the label still shows the old state, it is recorded rate-limit debt — this comment is the authoritative ruling record). Lane: domain:spec per the three-flow split — spec lane owns the contract half; Clause-②: yes (accepted-set expansion, ruled) — contract-review tier applies at dispatch. Same-family #6153 inherits this ruling framework in the same stroke (see its comment).

Generated by Claude Code

Claim: PM loop round 1 (2026-08-31 shift) — family chain head with #6153 (triage pre-cleared the fold: "five gates look satisfiable"; this seat answers FOLD for the SPEC HALF), spec-half-first decomposition
Session: session_01PBjwYLS6BciTQW3c9xQiD2
Branch: claude/issue-6140-field-metadata-declared-keys-spec (in objectstack — the spec half lands there; this card and #6153 stay OPEN through both halves and close on the objectui half)
Worktree: objectstack-issue-6140
Domain: domain:spec (cross-repo contract face)
File surface (spec half): packages/spec/src/data/field.zod.ts (markdown/html field metadata rows?: number; select-option description?: string) + authorable-surface/generated baselines + tests + .changeset/*. ⚠️ H17: field.zod.ts is #8345's declared Restart-touch trigger file — notification-only per that hold's own precedent (zero currency surface touched); notice posted on #8345 (stop on breach; explain in the report)
Container & model: M, mode:subagent, model: claude-fable-5 — Clause-② RULED yes on both cards (accepted-set expansion, maintainer 2026-08-25 「就全部接受」), review-tier dispatch.
Clause-②: yes (ruled)
Serial constraints cleared: zero devs in flight this seat; no open PR touches field.zod.ts (five parked PRs all merged 04:31–35Z; #13641 surface = i18n-resolver/cli, disjoint); H17 hit handled as notification per #8345's R5/R6 precedent. Fold exclusion list (gate ⑤): dependsOn is NOT declared anywhere — the inherited ruling resolves it as objectui-side "widget reads the canonical depends_on" (stop-and-fork if insufficient); the four inert rich-text editor keys (toolbar/preview/minHeight/maxHeight) are a docs-only rider on the objectui half — if any turns out DECLARED in types it becomes a separate enforce-or-remove card, never folded here.
Sequencing: objectui half (types/zod mirrors + RichTextField de-cast + SelectField depends_on read + lookup description consumption + rich-text.mdx docs rider) dispatches AFTER the spec half lands — linked halves are never co-batched. Fixes lines ride the objectui PR; the spec PR carries Part of objectstack-ai/objectui#6140 / #6153.

Parking commitment: spec PR opens DRAFT; needs:contract-review on carriers in the same stroke the PR exists. Review chain/maintainer owns enqueue.

Generated by Claude Code

Claim (spec-half dev seat, session session_01PBjwYLS6BciTQW3c9xQiD2): implementing the objectstack spec half of the 2026-08-25 Option A ruling on this card, plus the inherited per-key ruling on #6153 (description?: string on the select-option shape; dependsOn explicitly excluded — it resolves objectui-side).

Branch: claude/issue-6140-field-metadata-declared-keys-spec (objectstack repo, from origin/main). Scope: rows?: number on the markdown/html field-metadata authorable shapes matching the textarea precedent, description?: string on the select option type, regenerated artifacts, pins. Nothing objectui-side in this PR; the objectui half dispatches after this lands. Clause-② applies: the PR parks DRAFT + needs:contract-review.

Generated by Claude Code

os-dev-report

{
  "issue": 6140,
  "status": "done",
  "branch": "claude/issue-6140-field-metadata-declared-keys-spec",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/13669",
  "premise_still_valid": true,
  "summary": "Spec half of the 2026-08-25 Option A ruling on objectui#6140 plus objectui#6153's inherited per-key ruling, landed in objectstack. rows (positive integer) declared on FieldSchema for the multiline editor types textarea/markdown/html/richtext via a module-local MULTILINE_EDITOR_FIELD_TYPES set + the #11566 superRefine template (ruled pair markdown/html; textarea/richtext derived from the measured consumption set — flagged in the PR body for contract review, two-line trim if refused). description (string) declared on SelectOptionSchema (LookupField search consumer); flows to FormSelectOptionSchema by the #12868 Omit construction. dependsOn NOT declared per the inherited ruling. Both authoring forms show rows for exactly the set. Regenerated: docs references, authorable-surface shards, i18n bundles (merge mode), liveness state-counts; liveness ledger classifies field/rows live (placeholder-precedent cross-repo evidence); census --fix re-anchored one system-context.mdx citation (line rot from the insertion). Changeset @objectstack/spec minor, non-breaking, ADR-0087 gate green. PR parked DRAFT + needs:contract-review per Clause-2 (ruled).",
  "tests": "All local, at HEAD 83eeba1b (final union re-run after the last commit, via scripts/pm/os-verify-lock.sh). Full spec suite: 'Test Files 445 passed (445) / Tests 11888 passed (11888)'. Artifact gates: 'All 14 generated artifacts are up to date.' New pin file field-rows-option-description.test.ts: 28/28. Reverse verification from the COMMITTED state, no build legs needed (the pin file imports ./field.zod source-relative, so no dist resolution is in the loop): field.zod.ts checked out at base 7483afdd, mutation proved on disk (grep -c for the set const and the rows key both 0), pins went 22 failed / 6 passed — exactly the six control pins (four inert-key refusals, absent-stays-absent, icon+dependsOn refusals) staying green; restore via git checkout HEAD -- path, proved by empty git diff HEAD and git hash-object == HEAD blob (57ff40b5), pins back to 28/28. Spec typecheck: tsc + 'check:test-typecheck: OK — test layer compiles under packages/spec/tsconfig.test.json'. platform-objects typecheck green after bundle regen. pnpm lint (repo-wide eslint, full run, no narrowing): exit 0 at e665729f; the one later commit touches only liveness/field.json + state-counts.md.",
  "gates": "Derived via dispatch-gates.mjs (no paths passed; stderr line pinned the answer to objectstack-ai/objectstack at bdbe67b3; re-run implicitly covered by final union). All matched families run and exit 0, quoted verdicts archived: check:i18n 'OK (9 package(s) — all bundles in sync)'; check:i18n-coverage 'OK (12 config(s), 602 baselined untranslated string(s), none new)'; check:liveness 'every governed-type property … is classified' after classifying field/rows; check-system-context-census 'OK — 145 anchors resolve' after --fix; check:dual-build-cjs-loads and check-dev-prereqs green after their prescribed full build; changeset gates (empty/no-major/adr-0087/self-tests) green; docs family (anchors, authoring, frontmatter, single-h1, redirects, route-spelling, section-name, landing-index, affected-docs, drift-comment, audit-scope, quick-reference-counts) green; type-check-coverage/debt, nul-bytes, merge-driver, test-source-alias, cross-package-test-inputs, published-files, published-readme-links, role-word, keyed-text-bounds, spec-parsed-alias, where-matcher, slot-lookup, objectql-double-limit, page-declaration-shape, react-page-adapter, logger-receiver-detach, plugin-teardown, shard-attestation, ci-filter-parity, comment-mask-adoption, undeclared-dep-imports, pm-half-states, release-rehearsal self-test green. check:test-completeness locally NOT MEASURED by design (needs a CI test log; its own output prescribes recording exactly that).",
  "publish_door_pins": "Accept: rows preserved through parse on all four types (e.g. rows: 12 -> data.rows === 12) and rows: 1 at the bound; description preserved through the bare option shape AND the real select-field and lookup-field doors (data.options[0].description === 'Still being worked'); form-view option face accepts description via the Omit. Refuse (controls unchanged): toolbar/preview/minHeight/maxHeight on a markdown field each still unrecognized_keys; icon and dependsOn on the option shape still unrecognized_keys; rows on 10 wrong types answers a custom issue at path ['rows'] naming the set and the offending type; rows 0/-5/2.5 answer too_small/too_small/invalid_type with prescriptive messages.",
  "files_changed": ["packages/spec/src/data/field.zod.ts", "packages/spec/src/data/field.form.ts", "packages/spec/src/data/object.form.ts", "packages/spec/src/data/field-rows-option-description.test.ts", "packages/spec/liveness/field.json", "packages/spec/liveness/state-counts.md", "packages/spec/authorable-surface/data.json", "packages/spec/authorable-surface/ui.json", ".changeset/field-rows-and-option-description-declared.md", "content/docs/references/data/field.mdx", "content/docs/references/data/object.mdx", "content/docs/references/system/migration.mdx", "content/docs/references/ui/view.mdx", "content/docs/permissions/system-context.mdx", "packages/platform-objects/src/apps/translations/ (7 generated bundle files: en/es-ES/ja-JP/zh-CN metadata-forms + 3 source-hashes)"],
  "label_writes": "Cross-repo label writes SUCCEEDED via the MCP fallback (session REST gate closed — repo-scoped probe answered 403 'GitHub access is not enabled for this session', so the additive endpoint was unreachable; channel switch declared). Procedure per platform-readings: fresh payload-channel read of current labels, union, whole-group issue_write, compare read-back via payload channel, plus one delayed re-confirmation. Outcome: PR 13669 carries needs:contract-review (union preserved the auto-labeler's documentation/protocol:data/size/m/tests/tooling); objectui#6140 and objectui#6153 each carry needs:contract-review (union preserved domain:spec + pm:dispatched). All three confirmed on delayed re-read.",
  "deviations": ["rows applicability set widened from the ruled pair markdown/html to the measured consumption set + the precedent type (textarea, richtext added) — reasoned in the PR body's 'derived judgments' section, trimmable in two lines; parked for contract review by construction.", "rows value shape int().min(1): the objectui TextareaFieldMetadata precedent is boundless only because a TS interface cannot express the house count discipline this same schema applies to maxLength/minLength; no upper bound invented.", "check:test-completeness recorded NOT MEASURED locally (by the gate's own design).", "pnpm lint ran at e665729f, not final HEAD — the only later commit touches liveness/field.json + state-counts.md, neither in eslint's population.", "objectstack main moved during the run (7483afdd -> 967402a1, five merges); git merge-tree --write-tree against fresh origin/main is conflict-free, so no local merge was folded into the parked draft."],
  "blockers": [],
  "mcp_calls": "11 — 2 claim comments, 1 create_pull_request, 1 pull_request_read (PR body read-back), 3 issue_write label unions, 1 search_issues (dedup, non-empty result = live-channel proof), 1 issue_write create (finding), 2 report comments (this one + the 6153 pointer). All list/read traffic otherwise went through git and the zero-quota payload channel.",
  "open_questions": [
    {
      "question": "Contract review: keep textarea+richtext in the rows applicability set, or trim to the ruled literal pair markdown/html?",
      "options": ["A keep the four-type measured consumption set (textarea is the precedent type the ruling aligns to and its widget reads the key; richtext is served by the very widget the card measured)", "B trim to markdown/html (literal ruling; leaves the spec refusing rows on the precedent type while objectui declares and consumes it there)"],
      "recommendation": "A — declaring the ruled pair while refusing the precedent type's own declared key would manufacture a fresh declared-vs-enforced split; the PR is parked for exactly this review and B is a two-line trim."
    }
  ],
  "out_of_scope_findings": ["filed as objectstack-ai/objectstack#13671: the object.form options repeater still offers an icon input that SelectOptionSchema refuses at publish (offer-vs-door disagreement; the description half of the same repeater was made honest by this PR, icon needs its own measurement/ruling — #5016's option C measured only the action-param path)"]
}

Generated by Claude Code

PM verdict: ACCEPT (spec half, family #6140+#6153) — parked under needs:contract-review, one review point surfaced to the maintainer. (session_01PBjwYLS6BciTQW3c9xQiD2)

Verified on GitHub — objectstack PR #13669:

Form: draft ✓ · Part of both cards (no closing keyword — cards close on the objectui half) ✓ · needs:contract-review on PR and both cards (cross-repo MCP union writes, delayed read-back confirmed) ✓ · 21 files all in scope: field.zod.ts (+87), the two authoring-form offers, the new 28-pin test, 16 generated artifacts (authorable-surface, liveness, i18n bundles, docs references, one census re-anchor), changeset (spec minor).
Ruled members delivered: rows as positive integer with the fix(plugin-form): the top-level fields descriptions and the unresolved-member warning stop teaching { name } as tolerated (objectui#11550) #11566 refusal template (0/−5/2.5 refused with named codes; wrong field types get a custom issue naming the applicability set); description?: string on the select-option shape, flowing to the form-option schema by the established Omit construction. dependsOn NOT declared ✓ (inherited ruling honored); the four inert editor keys measured NOT declared and pinned refused ✓ (fence held).
Reverse verification: base-state mutation proven on disk, 22 pins red / exactly the 6 controls green, restore hash-proven; full spec suite 11888 green; repo-wide lint full-run exit 0; merge-tree conflict-free against fresh main.
THE review point (dev's open question, this seat recommends A): the ruling's letter names markdown/html; the PR declares the measured four-type multiline set (adding textarea + richtext). Grounds for A: spec today declares rows for NO type, so the literal pair would leave spec REFUSING rows on textarea — the very precedent type the ruling says to align with, whose objectui metadata already declares and whose widget consumes it — manufacturing a fresh declared-vs-enforced split on the precedent itself. B is a two-line trim if the maintainer wants the letter. Surfaced in the round report for a one-word ruling.
Out-of-scope finding objectstack#13671 (form repeater offers icon that the option schema refuses at publish — offer-vs-door, same seam this PR made honest for description) — verified filed, correct routing.

Sequencing: cards stay OPEN; the objectui half (types/zod mirrors + RichTextField de-cast + SelectField reads canonical depends_on + lookup description consumption + rich-text.mdx docs rider) dispatches after this PR LANDS. Enqueue awaits the review chain or the maintainer's word (with the A/B answer settling the one open point).

Generated by Claude Code

Contract review — PASS · objectstack PR #13669 (spec half) · the A/B point ruled: A

Director seat, summon #7 · session session_01KGtaLpkW1mycWgkbSb3H6t · tier fuse machine-read: last_served_model = claude-fable-5 = CONTRACT_REVIEW_TIER.

Verdict: PASS — the spec half is faithful to the 2026-08-25 Option A ruling (verbatim 「就全部接受，然后继续下一批」, read on this card this session), and the one surfaced judgment is ruled at tier:

rows applicability set — A (keep the four-type measured consumption set). Grounds: the ruling's own stated goal is "aligning the TextareaFieldMetadata precedent, so validation and the running widget finally agree." The spec today refuses rows on every type, including textarea — the precedent type whose objectui metadata already declares the key and whose widget reads it. Trimming to the literal markdown/html pair would leave the precedent itself in a declared≠enforced split, which is the exact condition the ruling exists to end; richtext is served by the same measured RichTextField read that motivated the ruled pair. A is the ruling's alignment applied coherently, not a widening past it. (Vetoable in a two-line trim if the maintainer reads the letter narrowly — recorded in the round report.)

Diff readings (branch, this session): MULTILINE_EDITOR_FIELD_TYPES module-local with the measured-consumption docblock; rows as int().min(1) per the house count discipline, refused outside the set via the #11566 superRefine with a set-enumerating message; description declared on SelectOptionSchema with the LookupField consumption evidence, flowing to the form face by the #12868 Omit construction; dependsOn correctly NOT declared and pinned refused; the four inert editor keys pinned still-refused (the ruling's "stops at rows" rider held). 28 pins with reverse verification; changeset @objectstack/spec minor.

Carrier disposition: needs:contract-review cleared on all three carriers (PR #13669, this card, #6153) in the same stroke as this PASS. The objectui half re-attaches its own carrier when its PR exists (no-pre-attach ruling, 2026-08-28).

Landing: the branch is currently in merge conflict with main on exactly one file — content/docs/permissions/system-context.mdx (census line-anchor rot, mechanically repaired by check-system-context-census.mjs --fix). A dispatched dev is performing the os-regen-merge.sh sync on the branch now; ready + enqueue follow once the sync is pushed and CI is green. The objectui half dispatches after the spec half lands, per the family sequencing.

Generated by Claude Code

os-dev-report

{
  "issue": 6140,
  "status": "done",
  "round": "patch (CI red on 83eeba1b; run 33362533873)",
  "branch": "claude/issue-6140-field-metadata-declared-keys-spec",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/13669",
  "head": "b6e9d27b",
  "premise_still_valid": true,
  "summary": "Two red signatures diagnosed and fixed; the maintainer's A ruling on the four-type multiline set is recorded (no trim — the shipped set stands). Signature 1 (deterministic, packages/lint): the #7010 shipped-form census pins moved by my two visibleWhen-gated rows entries — inverted IN PLACE per the #13216 precedent with the reason beside each: shipped-form predicates 49 to 51 (one row in field.form, one in object.form), object form's data.type-rooted predicates 17 to 18; the ==-literal census stays 42 because the new rows use 'data.type in' lists. Signature 2 (Temporal Conformance + Test Core shard 5): NOT the #11565 row-byte-budget hypothesis — the InnoDB os11565_narrow server-log line is the refusal that test deliberately triggers on every run, green ones included. The real assertion, recovered from the job log body: builtin-column-collision.test.ts, the #12015 classification pin — 'unclassified FieldSchema key(s): expected [rows] to deeply equal []'. That table demands every FieldSchema key be classified storage-vs-presentation deliberately; rows joins as 'presentation' (read by objectui's TextAreaField/RichTextField to size the inline editor, never by the DDL), one table row with the reason recorded.",
  "tests": "At HEAD b6e9d27b (final union re-run after the patch commit, via os-verify-lock; tree clean): repo-wide pnpm lint exit 0 (full run, no narrowing); packages/lint full suite 'Test Files 85 passed (85) / Tests 2355 passed | 5 skipped'; driver-sql full suite 'Test Files 147 passed | 9 skipped (156) / Tests 2252 passed | 132 skipped'. The two previously-red files directly, quoted: builtin-column-collision.test.ts 'Tests 11 passed (11)'; validate-predicate-path-refs.test.ts 'Tests 54 passed (54)'. Both packages' tsc --noEmit green (script echoes verified — no zero-match). Red leg witnessed by CI itself on 83eeba1b (both assertions quoted from the job log); green leg local at b6e9d27b — both directions covered without a local ablation.",
  "gates": "Path-derived re-run for the two patch files: check:driver-conformance (the one NEW family; exit 0), cross-package-test-inputs, test-source-alias, doc-authoring (80715 strings, baseline holds), keyed-text-bounds, undeclared-dep-imports, nul-bytes — all exit 0. No spec/docs/i18n surface moved this round (diff = one driver-sql source table row + one lint test file), so round-1 artifact gates stand at their sha.",
  "narrowing_and_prediction": "The live-MySQL cells (132 skipped locally) could not run here — no docker daemon. Declared narrowing, with the compensating measurement: the failing assertion is provably in the DIALECT-FREE half — it red identically in Test Core shard 5, a job with no service containers — and that half now passes locally 11/11. Prediction: both Temporal Conformance and Test Core shard 5 go green on b6e9d27b; the os11565_narrow InnoDB line will still appear in the MySQL container log on green runs (it is the refusal the per-page-limit test exists to trigger).",
  "deviations": ["The coordinator's #11565 row-byte-budget hypothesis was verified FALSE — no budget or fixture was touched; the fix landed in the #12015 classification table instead.", "Round-1 union miss acknowledged: packages/lint and driver-sql suites were not in my round-1 union because dispatch-gates path derivation names check families, not sibling packages' test suites pinned against spec shape — both are in this round's union and belong in any future spec-key round."],
  "blockers": [],
  "mcp_calls": "6 this round (2 actions_get job lookups, 2 get_job_logs, 1 workflow-logs-url attempt — archive host egress-blocked, so the assertion text came from the larger get_job_logs window — plus this comment); 12 in round 1.",
  "out_of_scope_findings": []
}

Generated by Claude Code

PM patch-round verdict: ACCEPT. (session_01PBjwYLS6BciTQW3c9xQiD2) Both CI signatures fixed at b6e9d27b: the lint census pins inverted in place with reasons (49→51 / 17→18, the #13216 precedent), and the shard-5/Temporal red root-caused to the #12015 classification pin — my dispatch's row-byte-budget hypothesis was VERIFIED FALSE by the dev (the InnoDB line is that test's deliberate per-run refusal; good falsification), the real fix being rows classified presentation in FIELD_KEY_STORAGE_CLASS with the reason recorded. Both previously-red files quoted green; repo-wide lint full-run green; live-MySQL narrowing compensated by the dialect-free proof. The maintainer's A ruling (four-type set stands) is recorded — no trim.

Handoff note of record: the contract-review seat has since PASSed PR #13669 at tier, cleared needs:contract-review on all carriers, and is driving the main-sync (one-file census-anchor conflict) with its own dev — landing belongs to that chain now; this seat's dev has stopped touching the branch, and this seat watches to MERGED only. The objectui half of #6140/#6153 dispatches once it lands.

Generated by Claude Code

Claim(半场更新): session_01PBjwYLS6BciTQW3c9xQiD2 · branch claude/issue-6140-field-metadata-objectui-half · seat domain:spec (os-warren)

spec 半 PR objectstack#13669 已 MERGED(14:19Z,维护者亲手合)——A 裁定(四型适用集)与契约复核均在案。本认领覆盖 objectui 半边(本卡与 #6153 同笔,Fixes 双卡收口)。

Clause-②: yes(裁定已裁双卡 yes)⇒ fable;PR 创建时 draft + needs:contract-review 双载体停靠,放行走席内 contract review(objectstack#13795)。范围:@object-ui/types 的 types/zod 镜像(rows 四型 + options[].description)+ RichTextField de-cast + SelectField 读 canonical depends_on(⛔ 无 camelCase 孪生)+ LookupField description 消费对齐 + rich-text.mdx 四惰性键 docs rider。installability 门:lockfile @objectstack/spec@17.2.0 不带 #13669——断言 INSTALLED spec 行为的测试按已装版本门控,PR 写清取舍。

Generated by Claude Code

os-dev-report

{
  "issue": 6140,
  "status": "needs_decision",
  "branch": "claude/issue-6140-field-metadata-objectui-half",
  "pr": "https://github.com/objectstack-ai/objectui/pull/7061",
  "premise_still_valid": true,
  "summary": "OBJECTUI HALF delivered as draft PR #7061 (Fixes #6140, Part of #6153), with task 3 of the dispatch deliberately STOPPED on the inherited ruling's own stop-and-fork clause. Per dispatch item: (1) types mirror DONE - rows?: number declared on MarkdownFieldMetadata and HtmlFieldMetadata (textarea already had it; no zod mirror exists for the field-metadata face in @object-ui/types, src/zod has no field-types file, so no zod half; richtext has NO union member at all - observation for PM, not folded) and description?: string declared on SelectOptionMetadata, both with docblocks citing the rulings and the spec keys. (2) De-cast DONE - RichTextField reads through MarkdownFieldMetadata | HtmlFieldMetadata (field as any retired); TextAreaField checked AND de-cast to TextareaFieldMetadata, its spec-face maxLength dual-read kept as a documented one-key structural read (framework#1878). (3) dependsOn convergence NOT IMPLEMENTED - FORKED: the ruling's premise ('the parent already declares depends_on' as THE canonical) is incomplete against origin/main - @objectstack/spec FieldSchema declares FIELD-LEVEL camelCase dependsOn (field.zod.ts:1304, since 2026-06-21 objectstack#2125, description explicitly covers select/radio gating AND lookup scoping; the strict door refuses snake depends_on), the spec liveness ledger classifies it LIVE naming objectui LookupField as consumer, and published example apps author it (showcase cascading-select province dependsOn country; invoice-contact dependent lookup). The widgets' camelCase metadata arms are the ONLY read path serving spec-published field documents, so deleting them per the dispatch letter would break shipped cascade/dependent-lookup behaviour and orphan a spec-declared live key. What the spec half pinned refused is dependsOn on the OPTION shape - a different declaration. No read arm was touched; decision needed (open_questions). (4) LookupField description alignment DONE - LookupOption.description docblock cites the now-declared upstream; behaviour-pinned (search widens label matching), zero behaviour change. (5) rich-text.mdx rider DONE - rows taught as declared key in the Field Schema block; the four inert keys named as NOT metadata keys; the page's false WYSIWYG/preview/auto-save feature claims rewritten to plain-textarea reality with 'not yet implemented'; lookup.mdx gap paragraph became the declared-member paragraph; select.mdx untouched on purpose (its #6153 paragraph is about the forked instance and stays true). (6) Installability gate HELD - zero assertions about the INSTALLED spec@17.2.0 (predates objectstack#13669); all pins are against objectui's own declarations and widget behaviour; form.zod.ts SelectOptionSchema needs no edit (spreads installed SpecSelectOptionSchema.shape by reference, gains description mechanically at the future dep bump). Deviation of record: PR carries Part of #6153 instead of the dispatched Fixes, because closing #6153 would bury its undecided dependsOn half.",
  "tests": "All at HEAD 10ff9ea via the shared verify lock, tree clean. Full suites of both packages: 'Test Files 208 passed (208) / Tests 3093 passed (3093)'; the three new pin files re-run AFTER the final commit at 10ff9ea: 3 files / 16 tests passed. Type-check incl. test layer: turbo type-check for @object-ui/types + @object-ui/fields 'Tasks: 12 successful, 12 total' (both chain tsc -p tsconfig.test.json). Reverse verification FROM THE COMMITTED STATE, tsc as instrument (vitest strips types so the membership pins are tsc-visible by design; no build in the loop - the types package's own tsc reads src): field-types.ts checked out at base e33b447, mutation proven ON DISK by grep counts (rows-member pattern 5 to 3, description pattern 2 to 1) before running; observed direction = @object-ui/types#type-check FAILED with TS2339/TS2353 naming exactly the new members inside the new test file ('Property rows does not exist on type MarkdownFieldMetadata', HtmlFieldMetadata, 'description does not exist in type SelectOptionMetadata') - which also proves the test project reads the new file; the @object-ui/fields red leg NOT OBSERVED (turbo halted after the types failure; recorded, not counted). Restore via git checkout HEAD, proven by empty git diff HEAD and git hash-object equal to the HEAD blob 29c9c8c9. Gates: check:doc-snippets 'Semantic phase: 271 of 271 block(s) judged, 0 failed' after building its own --build-filter closure (the rich-text.mdx rows:12 literal compiles against this branch's built dist); doc-links, doc-fences, doc-component-types, control-bytes (5849 files), changeset-presence (1 changeset / 2 released pkgs), changeset-no-major, designer-field-key-parity, self-import, phantom-deps, vi-mock-specifiers all green with their own verdict lines archived. Lint: targeted eslint over the 7 edited files exit 0, 0 errors; declared narrowing with invariance argument - eslint.config.js is NOT type-aware (no parserOptions.project), verdicts are per-file, so this diff cannot move untouched files' verdicts; the per-package farm is CI's run. NOT MEASURED: the repo-wide CI farm (CI runs it once on the PR) and anything about the installed spec.",
  "mcp_calls": "8 - 2 issue_read get_comments (locating the claim comment, absent from the SSR payload by seconds), 1 create_pull_request, 3 issue_write label unions (REST session gate probed 403 'GitHub access is not enabled for this session', so the additive endpoint was unreachable; channel switch declared - fresh payload-channel label reads, union, whole-group write, compare read-back via payload confirmed all three carriers: 6140/6153 keep domain:spec + pm:dispatched, PR keeps the auto-labeler's package: types + tests), 2 report comments (this one + the 6153 pointer). All other reads went through git and the zero-quota payload channel. Zero dedup searches because zero issues filed (dispatch routes findings to PM).",
  "open_questions": [
    {
      "question": "Field-level cascade key on the objectui METADATA face: which spelling is canonical, given @objectstack/spec FieldSchema declares camelCase dependsOn (live, ledger-classified, authored by published apps, refuses snake) while objectui BaseFieldMetadata declares snake depends_on (no framework producer; authored only by hand-written hosts and objectui tests)?",
      "options": [
        "A - mirror the spec: declare dependsOn (DependsOnInput shape) on the objectui metadata face, de-cast the four option widgets + LookupField onto it, and open a separate enforce-or-remove card for the legacy snake depends_on twin. Four-axis: real business pull is measured (shipped cascades in published apps ride the camelCase key); long-term it is contract-first (Commandment #0: the spec is the contract, no adapter translation workaround); anti-AI-error it ends the double-faced teaching material (today an AI copying a running app learns dependsOn works while the types teach depends_on - the exact split the #6140 ruling condemned); startup-scope it registers an existing fact, adds no capability.",
        "B - the dispatch letter: converge widgets onto snake depends_on only. Requires renaming the spec's field-level key (ADR-scale, breaks published apps) or an adapter-level camel-to-snake translation; both are unruled spec-shaping decisions, and as-is it breaks showcase cascade + dependent-lookup demos.",
        "C - status quo: keep the dual reads, document both spellings, fold the discrepancy into #4631 (declared surfaces that disagree). Cheapest today, leaves the trap armed."
      ],
      "recommendation": "A, because it is the only option where spec, types and runtime say one word without breaking shipped behaviour; B's letter contradicts the spec's standing declared key, and C preserves the AI-facing double teaching. The ruling's own fork clause anticipated exactly this ('declaring dependsOn as a second spelling needs its own decision - stop and fork')."
    }
  ],
  "out_of_scope_findings": [
    "For PM to file (dispatch forbids self-filing): the spec-half comment in objectstack packages/spec/src/data/field.zod.ts (option description docblock, 'the canonical spelling depends_on already exists at the field level') and the matching line in field-rows-option-description.test.ts:25 are factually wrong about the spec - the spec's own field-level key is camelCase dependsOn; that false sentence is what steered this dispatch's task 3. One-line doc corrections in objectstack, blocked on the open question's ruling.",
    "For PM to file: the richtext registry key has no FieldMetadata union member in @object-ui/types (markdown/html do) - the third key of the one widget is structurally served but undeclarable; whether it deserves a union member is its own small card.",
    "Observation, likely #4631 fodder: LookupField carries a whole camelCase-fallback family beyond dependsOn (displayField/descriptionField/allowCreate/lookupColumns/lookupPageSize/lookupFilters/avatarField dual-reads) - same legacy-snake vs spec-camel face split, one card could measure them together."
  ]
}

Generated by Claude Code

席内 contract review:PASS(交付范围) — PR #7061(head 10ff9ea),依 objectstack#13795 流程。

①:rows(markdown/html 镜像)+ option description 声明与 spec 对齐,de-cast 双 widget,四惰性键围栏三层 pin(类型级 Equal + 读点存在 + 无读扫描)✓;rich-text.mdx 把虚假 WYSIWYG/preview/auto-save 能力宣称改写为 not-yet-implemented 事实——诚实修正,在 rider 范围内 ✓;richtext 无 union 成员的一处 deliberate cast 有记录 ✓。
② semver:types minor + fields patch ✓。
③ fork(任务 3)判定正确:dev 以一手源证据(field.zod.ts:1304 camelCase dependsOn 声明+strict 拒 snake+ledger LIVE+已发布 app 在用)证明派发令「⛔无 camelCase 孪生」的前提只对 option 面成立、字段级为假——停手不动读臂是唯一对的处置。真设计分叉 ⇒ 决策箱(finding(fields): two more undeclared-but-consumed metadata keys — dependsOn on select and options[].description on lookup, both reached past the declared type #6153 承载),⛔ 本席不代裁。
偏差采纳:Fixes #6140 + Part of #6153(不埋未决半件)✓。

放行:剥标三载体 → ready → auto-merge;MERGED ⇒ 本卡关,#6153 转 needs-user-decision 承载 dependsOn 拼写裁定。

Generated by Claude Code

Navigation Menu

Repository navigation

finding(fields): `RichTextField` reads a `rows` key off field metadata that neither `MarkdownFieldMetadata` nor `HtmlFieldMetadata` declares — undeclared-but-consumed, reached through an `as any` #6140

The measurement — one of five is consumed, four are inert

Why `rows` is a defect rather than a working feature

The two candidate resolutions

Refs

Contract review — PASS · objectstack PR #13669 (spec half) · the A/B point ruled: A

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(fields): RichTextField reads a rows key off field metadata that neither MarkdownFieldMetadata nor HtmlFieldMetadata declares — undeclared-but-consumed, reached through an as any #6140

Description

The measurement — one of five is consumed, four are inert

Why rows is a defect rather than a working feature

The two candidate resolutions

Refs

Activity

os-zhuang commented on Aug 24, 2026

os-zhuang commented on Aug 24, 2026

os-steve commented on Aug 25, 2026

os-warren commented on Aug 31, 2026

os-warren commented on Aug 31, 2026

os-warren commented on Aug 31, 2026

os-warren commented on Aug 31, 2026

huangyiirene commented on Aug 31, 2026

Contract review — PASS · objectstack PR #13669 (spec half) · the A/B point ruled: A

os-warren commented on Aug 31, 2026

os-warren commented on Aug 31, 2026

os-warren commented on Aug 31, 2026

os-warren commented on Aug 31, 2026

os-warren commented on Aug 31, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

finding(fields): `RichTextField` reads a `rows` key off field metadata that neither `MarkdownFieldMetadata` nor `HtmlFieldMetadata` declares — undeclared-but-consumed, reached through an `as any` #6140

Why `rows` is a defect rather than a working feature