Consume the spec section group-reference form (objectstack#13855, ruled B) — page `record:details` sections and view `form.sections` resolve `group:` keys via `deriveFieldGroupLayout`

[huangyiirene]

Renderer half of objectstack#13855 (maintainer ruling 2026-08-31, verbatim 「直接处理b」): the spec gains a delta form for layout sections — a section may reference an object field group by key (shape like group: contact_info) and inherit membership + presentation via deriveFieldGroupLayout (ADR-0085 §5), instead of re-enumerating fields by hand.

Deliverable (this repo)

The two section consumers resolve the new form:

1. the record page record:details sections renderer;
2. the view-level form.sections renderer (if the spec half lands both surfaces — follow the spec design, do not invent a divergent reading).

Resolution goes through the platform's deriveFieldGroupLayout single source — ⛔ no re-implementation of assembly rules (declared order, empty-group drop, ungrouped trailing bucket, collapse + visibleWhen passthrough) on the objectui side. Mixing semantics (group-reference sections beside field-enumerating sections) follow the spec half's ruled design verbatim.

⛔ Sequencing gate — do not start before

objectstack#13855's spec PR is merged AND objectui has adopted the @objectstack/* version carrying it. The spec half is being dispatched directly by the director seat (maintainer: 「直接处理」).

Why this exists

hotcrm#806 measured the three-way field-grouping divergence this form retires: both escape hatches were wholesale + field-enumerating + derivation-free, so every hand-curated page re-copies membership and silently drifts as the object grows. The maintainer ruled the divergence a platform problem (「元数据应用应该是简化的业务实现…要处理也是平台统一处理」) and chose the delta form (B) over a lint diagnostic (A, not taken).

Refs: objectstack#13855 (ruling + design constraints) · hotcrm#806 (measured divergence, app half already queued) · objectstack#13848 (application-repo principles).

[huangyiirene]

Sequencing-gate update: first half SATISFIED — spec half is merged

项目总监席 · session session_01KGtaLpkW1mycWgkbSb3H6t

The gate in this card's body has two halves. As of 2026-08-31:

1. objectstack#13855's spec PR merged ✅ — objectstack#13897 landed on objectstack main as squash commit 39404f3d9 (contract review PASS recorded on objectstack#13855; the card is closed completed). Note for verifiers: squash merging makes an ancestor check on the branch head return false — confirm landing by grepping main's log for the PR number, not by ancestry.
2. objectui has adopted the @objectstack/* version carrying it ❌ — still outstanding. The spec change shipped with a @objectstack/spec minor changeset, so it reaches this repo only via a spec release plus a version bump here.

What the adopted version must carry (adoption check for the dispatching seat): the shared module packages/spec/src/shared/section-group-reference.ts (exports FIELD_GROUP_KEY_PATTERN), the group key accepted on both section surfaces (page record:details sections and view form.sections, aliases fieldGroup/groupKey folded to group), and the two reference diagnostics page-section-group-unknown / form-section-group-unknown in @objectstack/lint.

Design facts the renderer half must follow verbatim (from the merged spec, not to be re-derived): group ⊕ fields exactly-one per section; group-owned presentation keys are REFUSED beside group (no override semantics); mixed section kinds coexist in declared array order; group is banned on wizard steps; resolution goes through deriveFieldGroupLayout (ADR-0085 §5) as the single assembler.

Card stays pm:queue; do not dispatch until half 2 is verified against the checks above.

---

Generated by Claude Code

[os-warren]

Gate re-measured by the domain:ui seat (session session_012wwHa4aaFybxXrfmfHioDM): half cleared, still blocked → pm:queue → pm:blocked

This card was selected for dispatch this round. Its ⛔ sequencing gate has two conditions, and they have now diverged — so recording the measurement rather than leaving the next round to redo it.

⛔ objectstack#13855's spec PR is merged AND objectui has adopted the @objectstack/* version carrying it.

Condition 1 — spec PR merged: ✅ YES

objectstack#13855 is closed completed (2026-08-31T19:16:03Z), and the code is on objectstack main (2bff79ae). Measured, not inferred from the closure:

landed	path
the shared mixing rule	packages/spec/src/shared/section-group-reference.ts
view-level section	FormSectionSchema.group — packages/spec/src/ui/view.zod.ts
record page section	RecordDetailsProps.sections[].group — packages/spec/src/ui/component.zod.ts
key-grammar single source	FIELD_GROUP_KEY_PATTERN, packages/spec/src/data/field-group-layout.ts
dangling-reference lint	validate-form-layout.ts + validate-page-field-bindings.ts (+ both test files)
spec-side pins	packages/spec/src/ui/section-group-reference.test.ts

⭐ This also answers the open question in this card's own Deliverable §2. The card said "if the spec half lands both surfaces" — it did. field-group-layout.ts says so explicitly: the key "is now written on TWO surfaces: the group that DECLARES it (ObjectFieldGroupSchema.key) and the layout sections that REFERENCE it (FormSectionSchema.group, RecordDetailsProps.sections[].group — #13855)." So both consumers are in scope, and the mixing rule is single-sourced in section-group-reference.ts — which is also the answer to this card's ⛔ "no re-implementation of assembly rules on the objectui side." There is one module to resolve against.

Condition 2 — objectui adopted a version carrying it: ⛔ NO

• @objectstack/spec on objectstack main is version: 17.2.0 — read from packages/spec/package.json at 2bff79ae.
• @objectstack/spec resolved by objectui is 17.2.0 — pnpm-lock.yaml ('@objectstack/spec@17.2.0'), consistent with PR test(plugin-grid): render what the inline lookup picker actually receives #7167's independent reading today ("@objectstack/spec 17.2.0 as installed") on a fresh head.
• The feature is sitting in an unconsumed changeset: .changeset/section-field-group-reference.md, "feat(spec,lint): a layout section can reference a declared field group instead of copying its members (#13855)", declared as "Additive accept widening."

⇒ The same version number on both sides is not adoption — it is the feature being merged but not yet released. Dispatching now would send a developer to consume a form that the installed package does not declare; the honest outcome would be a premise-false report, and the dishonest one would be a hand-rolled reading of the spec source that violates this card's own ⛔.

Unblock condition — recorded machine-findably, per #7041 / #6653

Unblock-when: @objectstack/spec resolves to ≥ 17.3.0 in objectui's pnpm-lock.yaml and section-group-reference is present in the installed package.

Checkable in two commands, no judgement required:

grep -E "'@objectstack/spec@[0-9]" pnpm-lock.yaml
ls node_modules/@objectstack/spec/dist/shared/ | grep section-group-reference

The bump is a minor (additive accept widening), so ^17.0.0 / ^17.2.0 ranges will take it on the next install — no manifest edit is needed, only the release plus a lockfile refresh.

Note for whoever takes this when it unblocks

The objectstack half was "offered for contract review" (Clause ②) because it widens what the contract accepts. ⭐ The objectui half is not that. Consuming a form the spec already declares is renderer work — it does not widen any objectui surface and does not change what objectui's own contracts accept or reject. So this card is dispatchable at normal tier when it unblocks; it is not waiting on the CONTRACT_REVIEW_TIER quota that is currently holding the retirement pile.

Refs: objectstack#13855 (ruled B, 「直接处理b」) · hotcrm#806 · ADR-0085 §5.

---

Generated by Claude Code