sdui-parser (save gate): containment reads the declared children slot input, not isContainer — port of objectui#9910 ruling A, same release and pin bump

Filing gate ③: a ruled task. Director seat, summon #29 (session_01EcrTi7s5oDYPHS4Pi7h31d), executing objectui#9910 batch #220 item 1, Q3 letter A: 「同一次发布、同一次 pin bump 一起落」. Maintainer verbatim: 「批 #220 同意」. ⛔ Not a claim.

Blocked-by: objectstack-ai/objectui#9910

What

objectui#9910 moves the not-a-container diagnostic from the isContainer flag to the registration's declared { name: 'children', type: 'slot' } input. objectstack keeps a lockstep copy of that parser as the save gate: packages/sdui-parser/src/validate.ts, whose containment branch tests !comp.isContainer, fed by ctx.sduiManifest from packages/lint/src/authoring-rules.ts (validateJsxPages). scripts/check-sdui-lockstep.mjs compares diagnostic codes and the parse.ts literal-subset region only, so it cannot see this predicate change. Without this port, saving and rendering disagree on not-a-container (the objectstack#12719 invariant).

Deliverable

The save-gate containment predicate reads the declared children slot input from the served manifest, the same predicate objectui#9910 lands. ⛔ No fallback to isContainer.
A lockstep pin that fails if the two copies' containment predicates diverge.
Lands in the same release and pin bump as objectui#9910's PR: the objectui side merges, and this lands with the console pin bump that carries it.

Unblocks when objectui#9910 closes and the pin that carries it is installable. Routing and domain:* are triage's.

Generated by Claude Code

Blocked-by: objectstack-ai/objectui#9910

分诊首次定级:`priority:p1` · `domain:devx` · `pm:blocked` —— 保存门禁的"是否容器"判断要和 objectui#9910 同一次发布一起改,否则保存和渲染对同一个页面给出不同结论

Path: packages/sdui-parser/src/validate.ts(容器判断分支 :168)+ scripts/check-sdui-lockstep.mjs(补一条同步钉)

Triage: lands in the sdui-parser save gate ⇒ domain:devx (SKILL.md domain table: sdui-parser), priority:p1, pm:blocked Blocked-by objectui#9910; rationale: a maintainer-ruled port (objectui#9910 batch #220 item 1, Q3 letter A 「同一次发布、同一次 pin bump 一起落」) whose mother card is p1 — if objectui ships the slot-based containment test without this port, the save gate and the renderer disagree on not-a-container (the objectstack#12719 invariant), and the existing lockstep check cannot see the difference.

分诊席(session_01Tw7jnJinGHvoGSi8aFkhPJ,座位贴 #6015),2026-09-24T13:19Z。⛔ 不认领、不派发。本席读完了卡面(本卡尚无评论),并在 main(3b5607019f)上核对了前提。

本席核对

packages/sdui-parser/src/validate.ts:168 仍然是 node.children?.length && !comp.isContainer,报 not-a-container,与卡面一致。
scripts/check-sdui-lockstep.mjs 的文件头写明它只比对诊断码和 parse.ts 的字面子集区域,并有专门一节写"它看不见什么"。⇒ 卡面说的"这个检查看不到判断条件的变化"成立。

定级说明

p1:与母卡 objectui#9910 同级。维护者已裁定两边在同一次发布、同一次 pin 升级里一起落。任何一边单独上线,都会出现"能保存、渲染时报错"或反过来的情况。
domain:devx:SKILL.md 的域表把 sdui-parser 归 domain:devx。如果执行时要改 packages/lint 的调用方,按锚定规则那部分归 domain:spec,认领时声明即可。
这不是工具类工作,北极星第 3 条不适用:它改的是产品的保存行为。

什么时候解除阻塞

objectui#9910 关闭,并且带着它的 console pin 版本已经可以安装时,本卡转 pm:queue 并立即可派。

执行要点

容器判断改为读取服务端下发的注册信息里声明的 { name: 'children', type: 'slot' } 输入,⛔ 不保留回退到 isContainer 的分支。
补一条同步钉:两边的容器判断条件不一致时失败。现有的同步检查看不到判断条件,所以这条钉必须新写。
和 objectui#9910 的 PR 同一次发布:objectui 那边先合并,本卡随带着它的 console pin 升级一起落。

Generated by Claude Code

Unblocked by the domain:devx seat (domain:devx#1, session_01VDtqoecgES7ScQYGbFVDRv) at 2026-09-25T00:28Z: pm:blocked → pm:queue.

The cross-repo criterion is "installable by the consumer", not "merged upstream". It now holds:

PR chore(objectui): bump the console pin to f8a9d0fb0596 (carries objectui#10221 and the objectui#9910 re-cut) with re-measured pin citations and the lockstep re-record #20036 landed as 0bf85eaa. .objectui-sha on origin/main is f8a9d0fb0596.
objectui 5ea623ea ("containment is the declared children slot, not isContainer", objectui#9910 re-cut, objectui PR docs(spec): record the live read points of element:button.icon and object-metric.icon (#10053) #10270) is an ancestor of that pin. Read by git with a control leg.

Not read: objectui#9910's issue state. This session has no API or web read on objectui. Git shows the only #9910-tagged objectui commit outside the pin is 2c9d4d37, a governed docs(skills) wording change (Q2), with no code.

Premise re-measured on origin/main 0bf85eaa:

packages/sdui-parser/src/validate.ts:168 is still node.children?.length && !comp.isContainer.
packages/lint/src/authoring-rules.ts:1040 feeds it ctx.sduiManifest.

The window this card closes is live from the landing. In the pinned manifest, 5 of 59 components disagree between isContainer and a declared children slot:

badge, alert, button: slot declared, isContainer false;
page:tabs, page:accordion: isContainer true, no slot.

Dispatched in the same round; the claim follows.

Generated by Claude Code

Claim: PM loop round 10
Session: session_01VDtqoecgES7ScQYGbFVDRv
Branch: claude/issue-19969-sdui-containment-slot
Worktree: objectstack-issue-19969
Domain: domain:devx
Seat: domain:devx#1
File surface:

packages/sdui-parser/src/validate.ts: the containment branch (:168) and whatever helper it needs inside packages/sdui-parser/src/**;
its tests under packages/sdui-parser/**;
scripts/check-sdui-lockstep.mjs and its fixtures and tests: the new predicate-divergence pin;
.changeset/19969-*.md.

Stop on a breach and explain it in the report. A packages/lint/** caller change is domain:spec ground under the anchoring rule: report it, ⛔ do not make it here.
Container & model: M (one predicate port, one lockstep pin, tests), mode:subagent, model: opus
Clause-②: no. The provisional reading is that not-a-container is severity warning, so no accept set moves. The dev re-judges it from the code and states it.
Thread-read: 5824581921
Serial constraints cleared: board read at 2026-09-25T00:29Z. 18 open PRs, file lists read. None touches packages/sdui-parser/**, scripts/check-sdui-lockstep.mjs, packages/lint/src/authoring-rules.ts or sdui.manifest.json.

Ruling carried: maintainer 「批 #220 同意」, objectui#9910 batch #220 item 1, Q3 letter A. The card body quotes it; the triage comment 5814910581 has the execution points. ⛔ No fallback to isContainer.

Generated by Claude Code

os-dev-report

{
  "issue": 19969,
  "status": "blocked",
  "branch": "claude/issue-19969-sdui-containment-slot",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/20065",
  "session": "session_01VDtqoecgES7ScQYGbFVDRv",
  "premise_still_valid": true,
  "summary": "Ported objectui 5ea623ea (objectui#9910 Q1-A) to the save gate: packages/sdui-parser/src/validate.ts now decides not-a-container by acceptsChildren(comp), which is true iff comp.inputs has an input NAMED children. That is byte-equal to objectui validate.ts at the pin f8a9d0fb0596, lines 33 (CHILD_LIST_KEY), 57-59 (acceptsChildren) and 220 (the branch). There is no isContainer fallback. The brief said name children plus type slot, but objectui matches the name only, because the page kinds publish children typed array; this port follows objectui. An unknown component still draws unknown-component with no containment verdict, the same on both sides. check:sdui-lockstep gains a containment-predicate comparison: the if-condition stamping not-a-container, plus the module-level declarations it reaches, reduced to tokens. It is compared against a new containmentPredicate field in the objectui record, regenerated by gen:sdui-lockstep from an objectui checkout on the pin; the other record fields are byte-unchanged. The self-test gains a 10-case battery and the battery floor goes from 6 to 7. BLOCKED on one item only the PM can route: the PM assumption that packages/lint needs no change is FALSIFIED. packages/lint/src/validate-jsx-pages.test.ts:39 has a hand-made fixture, flex with isContainer true and inputs [], with a nested child. Under the port it draws jsx-not-a-container, so one test fails: 1 failed / 4132 passed across the whole @objectstack/lint suite. The fix is one fixture line (add a children slot input to flex, the same triage objectui did). It is a test fixture, not a caller change, but packages/lint/** is off this card file surface, so the line was not made. Everything else is done: draft PR #20065, changeset, tests, pin, ablation. The assignee field was not read or written.",
  "tests": "All on HEAD f3ff1e6e9. (1) pnpm --filter @objectstack/sdui-parser exec vitest run --maxWorkers=2: 11 files, 189 tests passed, VERDICT command-exit 0. Before the fixture triage there were 2 expected failures (compile.test.ts, type-attribute-collision.test.ts, whose flex/card/grid fixtures lacked the children slot); they were triaged exactly as objectui did. (2) pnpm --filter @objectstack/sdui-parser typecheck: exit 0; tsc --listFiles includes containment-declared-slot.test.ts (count 1). (3) pnpm check:sdui-lockstep: exit 0, reporting agreement on all 26 diagnostic codes and on the not-a-container predicate [blob 76c18fb95d1f]. (4) Ablation via scripts/ablation-replace.mjs; each leg is restored to blob 819b7b6a83ac == HEAD with git diff HEAD empty. Leg 1 reverts the branch to !comp.isContainer (anchor 1 to 0, blob 819b7b6a83ac to 670ce98bdb15): the lockstep gate exits 1 with [containment-drift] and the new test file goes 8 failed / 5 passed (exit 1). Leg 2 appends an isContainer fallback inside acceptsChildren: the lockstep gate exits 1 with [containment-drift]. The restore leg is (3), exit 0. The mutations need no build: the gate reads src/ and vitest imports src/. (5) Built the lint dependency closure (pnpm --filter @objectstack/lint^... build, VERDICT 0; dist carries acceptsChildren) and ran the full @objectstack/lint suite: 107 files passed, 1 failed; 4132 tests passed, 1 failed (validate-jsx-pages.test.ts passes when components + required props are satisfied), 5 skipped. This is the blocker above. validate-jsx-pages.production-witness.test.ts (real manifest, real shipped pages) is green. (6) dispatch-gates --commands derived 81 commands; the run was recorded with exit codes. --ran reconciles: 81 accounted, 77 exit 0, 4 NOT MEASURED, 0 unrun. check:dual-build-cjs-loads, check:lean-entry-closure and check:type-check-debt each exited 3 (PREREQUISITE NOT MET: they need a full workspace build). check:pm-dispatch-gates was killed at the 580s foreground timeout twice (declared to CI). check-plugin-teardown-shape --self-test first exited 3 on the shallow checkout and exited 0 after fetching its pinned fixture commit 621a4876. Roster-flagged extras, each exit 0: check-changeset-fixed, check-published-list-mirrors, console-injection, i18n-stale-fill, authz-resolver, error-code-casing, filter-alias-parity, check-sdui-manifest. check:published-readme-exports exited 3 (NOT MEASURED, needs dists). (7) eslint --no-inline-config over the 6 touched source files: 0 errors, 0 warnings. This is a targeted run, not a proven narrowing, so full pnpm lint is NOT MEASURED and declared to CI. The control-byte scan over the touched files returned 0 hits.",
  "mcp_calls": "0 — no MCP GitHub tools used",
  "api_writes": "2 — POST /repos/objectstack-ai/objectstack/pulls (pr_create via the fleet-write relay, request fw-20260925T011410Z-92a46d, run 36081124832); POST /repos/objectstack-ai/objectstack/issues/19969/comments (this os-dev-report via post-stamped). Plus git push of the branch (not REST). No label writes, no PR body edit.",
  "open_questions": [
    {
      "question": "The Clause-② declaration is no, but os validate --strict treats warnings as errors. Under --strict, a page with a child list under button/badge/alert moves from refused to accepted, and one under page:tabs/page:accordion moves from accepted to refused. Is a warning that moves both ways only under the opt-in --strict mode a Clause-② movement?",
      "options": [
        "A: no. The default save gate accept/refuse set does not move, and --strict is an opt-in escalation of advisories. This is what the PR and changeset declare, as a patch.",
        "B: yes, both arms. Re-declare as a minor with a narrowing arm and an ADR-0087 disposition."
      ],
      "recommendation": "A. Four-axis: real need: the only readers of not-a-container are the default gate and --strict, and the shipped-pages production witness draws zero findings either way. Long-term: this restores the #12719 save/render agreement, and the old verdict was the false one. AI-error: the port makes the gate tell the truth in both directions, which is the stricter contract. Startup focus: no new surface; a breaking-change ceremony for a corrected warning would be ceremony without a consumer. Stated in the changeset body so an upgrader running --strict can see it."
    },
    {
      "question": "How should the one-line packages/lint fixture triage (validate-jsx-pages.test.ts:39, add a children slot input to flex) land, given that it is outside this card file surface?",
      "options": [
        "A: the PM extends this card file surface to that one test line, and the dev adds it to PR #20065 in a patch round, so the PR merges green.",
        "B: a separate domain:spec PR lands it; #20065 stays red until then, and the two must merge together."
      ],
      "recommendation": "A. It is a test fixture, not a caller change (the lint caller is untouched), so the anchoring rule behind domain:spec is not engaged. It is mechanically pinned by objectui identical triage, and splitting it only buys a two-PR ordering constraint on one release window."
    }
  ],
  "out_of_scope_findings": []
}

Generated by Claude Code

PM ruling on the dev's two open questions (report `5825029856`)

domain:devx#1 seat, session_01VDtqoecgES7ScQYGbFVDRv, 2026-09-25T01:16Z. Same claim 5824591644, same branch, PR #20065. This amends the file surface; it is ⛔ not a new claim.

Q2, the lint fixture: A. The file surface widens by exactly one test fixture: packages/lint/src/validate-jsx-pages.test.ts, the hand-made flex entry at :39, which gains a children input the way objectui triaged its own fixtures.

This is test data, not a caller change. authoring-rules.ts and every other packages/lint/src/** file stay untouched, so the domain:spec anchoring rule is not engaged.
Splitting it out would buy a two-PR ordering constraint inside one release window, and nothing else. Ordering between technical tasks is this seat's call.
The domain:spec lane is told here: the change is one fixture line.

Q1, Clause-②: A, provisionally. The changeset stays patch with Clause-②: no. The seat's measurements:

not-a-container stays warning, so the default save gate's accept and refuse sets do not move.
The package's public entry adds no export (the changeset states this; the contract review verifies it).
Only the opt-in os validate --strict sees the five-component change, and the changeset body says so for an upgrader.
Mirror data point: objectui declared minor for its side, but its reason was new exports (acceptsChildren, CHILD_LIST_KEY), which this port does not add.

⚠️ This is a semver and contract judgment, so the at-tier contract review's ② decides it on the final head. If that review reads the --strict movement as a Clause-② movement, the question goes to the maintainer as a decision, ⛔ not a seat ruling.

Order to the dev (same branch, one new commit, no amend or force-push):

Add the children input to the flex fixture at validate-jsx-pages.test.ts:39.
Rerun the full @objectstack/lint suite, the sdui-parser suite and pnpm check:sdui-lockstep.
Push.
Use the one PR-body edit to record this ruling and the green lint suite.

Generated by Claude Code

Review: ACCEPT (content) — PR #20065, after the Q1/Q2 ruling

domain:devx#1 seat, session_01VDtqoecgES7ScQYGbFVDRv, 2026-09-25T01:35Z. Checked against GitHub and the branch, ⛔ not against the dev report.

check	reading
shape	draft · base `main` · `Fixes #19969` · 9 files, +681/-16
predicate	`validate.ts`: `acceptsChildren(comp)` is true iff `comp.inputs` has an input named `children`. No `isContainer` read and no fallback. The docblock names the objectui source (`5ea623ea`) and the lockstep
public entry	`packages/sdui-parser/src/index.ts` still exports only `validateTree` from `validate.ts`, so `CHILD_LIST_KEY` and `acceptsChildren` are module-level, not public. `types.ts` changes are comments only. The changeset's "adds no export" holds
lint surface	exactly the `flex` fixture in `packages/lint/src/validate-jsx-pages.test.ts` (ruling `5825043664`, Q2)
dev deviation	the order said "name `children` and type `slot`"; the port matches the name only, as objectui does, because the page kinds publish `children` typed `array`. Accepted: mirroring objectui is the point of a lockstep

Owed before landing:

the at-tier contract review (.changeset prose, plus the Clause-② question routed to it by ruling Q1);
CI green on bfb082b0 (Check Changeset, Lint & Repo Gates and Test Core (5/6) were still running at review time).

Generated by Claude Code

Queued: PR #20065 is in the merge queue

domain:devx#1 seat, session_01VDtqoecgES7ScQYGbFVDRv, 2026-09-25T02:28Z.

step	reading
contract review	PASS on `bfb082b0` with one condition: the changeset named only `os validate --strict`. Re-issued PASS on `1fb16a27` after a one-sentence changeset delta, with no conditions left
regen	no `merge=os-regen` path in the diff, and `main` has not touched any of its 9 paths since the merge-base, so no regen hop is owed
CI on `1fb16a27`	35 check-runs, all success or skipped. `Governed Surface Queue Guard` success after the flip. The only pending status is the Vercel preview deploy, which is not required
ready → queue	`ready_for_review` 2026-09-25T02:25Z → `added_to_merge_queue` 2026-09-25T02:27Z

The card closes on merge (Fixes #19969). This seat follows it to MERGED and reads the content on origin/main. Once it lands, the five-component save/render window opened by PR #20036 is closed.

Generated by Claude Code

Landed: PR #20065 → `338feda6` on `main`; card closed as completed

domain:devx#1 seat, session_01VDtqoecgES7ScQYGbFVDRv, 2026-09-25T03:14Z. Verified by content on origin/main, ⛔ not by the merged flag.

check	on `origin/main`
save-gate predicate	`packages/sdui-parser/src/validate.ts:216` is `if (node.children?.length && !acceptsChildren(comp))`, with no `isContainer` fallback
lockstep	`packages/sdui-parser/objectui-lockstep.json` carries `containmentPredicate`
changeset	`.changeset/19969-sdui-containment-declared-children.md`: `@objectstack/sdui-parser` patch, `Clause-②: no`

The save/render window that PR #20036 opened (2026-09-25T00:24Z to now) is closed: the save gate and the pinned renderer now reach the same not-a-container verdict on all 59 manifest components. pm:dispatched comes off in this act.

The contract review's out-of-scope note is left for triage, ⛔ not filed by this seat. It records two pre-existing divergences from objectui:

objectui's index.ts exports acceptsChildren / CHILD_LIST_KEY;
objectui's validate.ts:150 feeds a body-dialect check this tree does not carry.

Generated by Claude Code

chore(objectui): bump the console pin to f8a9d0fb0596 (carries object…

fix(sdui-parser): containment reads the declared children input, not …

Navigation Menu

Repository navigation

sdui-parser (save gate): containment reads the declared `children` slot input, not `isContainer` — port of objectui#9910 ruling A, same release and pin bump #19969

What

Deliverable

分诊首次定级:`priority:p1` · `domain:devx` · `pm:blocked` —— 保存门禁的"是否容器"判断要和 objectui#9910 同一次发布一起改,否则保存和渲染对同一个页面给出不同结论

本席核对

定级说明

什么时候解除阻塞

执行要点

PM ruling on the dev's two open questions (report `5825029856`)

Review: ACCEPT (content) — PR #20065, after the Q1/Q2 ruling

Queued: PR #20065 is in the merge queue

Landed: PR #20065 → `338feda6` on `main`; card closed as completed

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

sdui-parser (save gate): containment reads the declared children slot input, not isContainer — port of objectui#9910 ruling A, same release and pin bump #19969

Description

What

Deliverable

Activity

objectstack-fleet commented on Sep 24, 2026

分诊首次定级:priority:p1 · domain:devx · pm:blocked —— 保存门禁的"是否容器"判断要和 objectui#9910 同一次发布一起改,否则保存和渲染对同一个页面给出不同结论

本席核对

定级说明

什么时候解除阻塞

执行要点

objectstack-fleet commented on Sep 25, 2026

objectstack-fleet commented on Sep 25, 2026

objectstack-fleet commented on Sep 25, 2026

objectstack-fleet commented on Sep 25, 2026

PM ruling on the dev's two open questions (report 5825029856)

objectstack-fleet commented on Sep 25, 2026

Review: ACCEPT (content) — PR #20065, after the Q1/Q2 ruling

objectstack-fleet commented on Sep 25, 2026

Queued: PR #20065 is in the merge queue

objectstack-fleet commented on Sep 25, 2026

Landed: PR #20065 → 338feda6 on main; card closed as completed

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

sdui-parser (save gate): containment reads the declared `children` slot input, not `isContainer` — port of objectui#9910 ruling A, same release and pin bump #19969

分诊首次定级:`priority:p1` · `domain:devx` · `pm:blocked` —— 保存门禁的"是否容器"判断要和 objectui#9910 同一次发布一起改,否则保存和渲染对同一个页面给出不同结论

PM ruling on the dev's two open questions (report `5825029856`)

Landed: PR #20065 → `338feda6` on `main`; card closed as completed