retire(studio): drop the canonical page:header icon field, restate the alias's on its renderer read (#3829) - #4794
Merged
Conversation
…e alias's on its renderer read (#3829) objectui#3829 was filed as a three-way fork over `page:header.icon` and `page:card.actions` (wire them / declare them with a KNOWN GAP / retire them upstream). The maintainer ruled route (c) on 2026-08-09 and objectstack#6946 / PR objectstack#7115 executed it: both keys are ADR-0087 D2 tombstones in @objectstack/spec 17.0.0, live on the rc.6 this repo pins. This is the objectui follow-through, and the measurement behind it found the retirement's "zero producers" premise held for one key and not the other. - previews/block-config.ts + both locale tables: the canonical `page:header` icon box and its two now-orphaned i18n keys. This was the one live producer. Before the retirement an authored value was silently dropped; after it the platform rejects the node BY NAME, so a designer field left standing is strictly worse than the gap it papered over. Negative pin (plus the positive non-vacuity half and the spec-side tombstone assertion) in previews/__tests__/block-config.test.ts. Precedent followed verbatim: 38ab505 / PR #4169's global_nav cleanup, "the palette option and its now-orphaned i18n key". - page:card.actions had no producer at all — it is a clean retirement, and only the exemption prose changes for it. - registry-inputs-spec-parity.test.ts: both exemptions rewritten from the "wire it or declare it" MENU to the retired-upstream class, templated on the `record:details.layout` entry. They are NOT deleted: D2 retirement replaces the member with `z.never()` rather than removing it, so `Object.keys(shape)` still reports both keys and the reverse direction still demands cover. They resolve via objectui#3809's tombstone recognition, which the entries now say. Stale line numbers repaired (822-1570 -> 973-1677, 666-695 -> 703-745) and the block header re-based on rc.6. - layout's `page-header` / `layout:page-header` alias KEEPS its `icon` input (route A). That is a different renderer which really reads and draws it (PageHeader.tsx:123, :231-233), the docs page publishes it and the repo's only live demo writes it — withdrawing it would delete a working capability. What changes is the LICENCE: the comment now rests on the renderer read instead of spec parity, and page-header-authorable-keys.test.tsx skips tombstones when it derives the spec key set, so `icon` stops passing the dialect guard for the wrong reason. The carve-out is a named, issue-backed, self-clearing list. The renderer read points themselves are untouched. - content/docs/layout/page-header.mdx: line 40 still taught `icon` as a member of `PageHeaderProps`. Rewritten to rc.6 reality, and the component prop is now distinguished from the retired spec key rather than conflated with it. - PageBlockInspector.i18n.test.tsx: the #3979 placeholder pair re-pointed from `page:header` to `record:alert`, which reproduces the symptom (prose hint under a translated label) on a key the renderer really reads. Same move and same reason as objectui#4649's re-point a few tests above. Part of #3829 Co-authored-by: Claude <noreply@anthropic.com>
Contributor
✅ Console Performance Budget
📦 Bundle Size Report
Size Limits
|
…n literal (#3829) CI red on `scripts/__tests__/doc-version-claims.test.ts`, self-inflicted by the previous commit: the new PageHeader paragraph wrote "retired in `@objectstack/spec` 17.0.0", and a first-party package name followed by a dotted version IS a version claim on a scanned surface (`content/docs` is a SCAN_ROOT). Nothing in this repo can check that literal, so the ratchet demands it be deleted or registered. Deleted, which is the ratchet's own stated preference and the right one here: the sentence already cites objectstack#6946 and ADR-0087 D2, so the version number carried no fact those two do not — while being exactly the kind of literal objectui#3645 froze at `^3.3.0` for thirteen majors. The actionable form of "which release" is already in the same paragraph as `os migrate meta --from 16`, which is a migration command rather than a claim about a dependency's current version. The same literal appears in this PR's test comments, block-config.ts and the changeset and does NOT trip the gate — those are not scanned surfaces. Only the mdx is, so only the mdx changes here. Verified against the gate's own matcher rather than by eye: FIRST_PARTY + SEP + VERSION, where SEP cannot cross an opening parenthesis, so "`@objectstack/spec` (objectstack#6946" is not a claim. Re-ran the file through a probe built from the test's exact regex fragments (1 hit before, 0 after) and then the test itself. Part of #3829 Co-authored-by: Claude <noreply@anthropic.com>
Contributor
✅ Console Performance Budget
📦 Bundle Size Report
Size Limits
|
Collaborator
Author
|
【PM 验收 · ACCEPT】objectui 分片 PM(session_01GTRjn8xBqp75dk7kFupVRt),#3829 route (c) follow-through 四件套(exemption 重写 / 设计器 icon 字段退役 / docs / alias A 案)。 实物核验(首轮已过,本轮增量):返工 commit 50f690e 仅动 CI 终态(PM 亲读):head 50f690e 全部 19 项 completed、零 failure(17 success + dependabot/coverage path-filter skipped;原红的 Test shard 4/4 已 success @ 08:51:24)。 返工记录(公开记账):首轮唯一红是 doc-version-claims 棘轮 —— 本 PR 在 mdx 新增段落写入 回传处置:四条 stale-pin exemption 文案的 rc.5→rc.6 余数归 #3809/pin bump 线,不另立卡;棘轮扫描面边界(测试注释/changeset 不扫)记为已知合理行为。 处置:undraft + auto-merge(SQUASH)。 Generated by Claude Code |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #3829
落地本卡 2026-08-16 的「PM 裁定 · 测量回贴」:route (c) 的 objectui follow-through,
alias 半边按代裁的 A 案。维护者 2026-08-09 已裁 route (c),上游 objectstack#6946 /
PR objectstack#7115 已执行,并随本仓现锁的
@objectstack/spec@17.0.0-rc.6生效 ——PageHeaderProps.icon与PageCardProps.actions均已是 ADR-0087 D2 墓碑(成员被替换为z.never(),带具名迁移文案)。为什么 objectui 这半边比原裁定预想的大
退役 prescription 写的是「零生产者」。这对
page:card.actions成立 —— 它是干净完整的退役;对
page:header.icon不成立:metadata-admin 设计器的 BLOCK_CONFIG 仍在为canonical
page:header发布icon编辑框。退役前作者填进去的值被静默丢弃,退役后平台按名拒绝整个节点 —— 留着生产面比退役前更糟,因为它在教作者(尤其是 AI 作者)写出
解析失败的元数据。
原裁定描述的完成机制也需要修正:D2 退役是把成员替换成
z.never(),不是删除,两键因此仍留在
Object.keys(shape)里 —— pin bump 并不会让那两条 exemption 变 stale(gate 源码自己写明了这一点)。所以正确的 sequencing 是重写而不是删除,删除权归
objectui#3809。
四件事
1. 两条 exemption 重写(
apps/console/src/__tests__/registry-inputs-spec-parity.test.ts)从「wire it or declare it」菜单改写为 retired-upstream 类,模板逐字取自同文件的
record:details.layout条目:引用 objectstack#6946 / PR objectstack#7115、写明墓碑留在Object.keys(shape)、删除权交 objectui#3809。顺带修陈旧行号(822-1570 改为 973-1677、666-695 改为 703-745,均按当前
containers.tsx自核)与块头的 rc.5 注。两条 exemption 不删 —— 删了 gate 会反向变红。
2. 移除设计器 canonical
page:header的 icon 字段previews/block-config.ts的字段,加上此时已成孤儿的两条 i18n 键(en / zh 同一次改动,两张表键集保持一致)。先例逐字:commit 38ab505(PR #4169)对 global_nav 的同类清理
(「the palette option and its now-orphaned i18n key」)。防回潮钉在
previews/__tests__/block-config.test.ts:正面非空半 + 「不提供 icon」负钉 + spec 侧墓碑断言(含探针自身的非空性)+ 两条 i18n 键在两个 locale 都已无译文。
3. docs 修正(
content/docs/layout/page-header.mdx)第 40 行仍把
icon教成PageHeaderProps的成员。改按 rc.6 现实重写,并把组件自有prop 与已退役的 spec 键如实区分 —— 这页文档写的是
@object-ui/layout的PageHeader组件,它的icon是 renderer 自有能力;canonical 节点上同名键已被按名拒绝。未碰
content/docs/releases/。4. alias 半边(A 案)
packages/layout的page-header/layout:page-header的iconinput 保留:那是另一个渲染器,它真读真画(
PageHeader.tsx:123读、:231-233画),文档发布它,本仓唯一的活 demo(
layout-page-header/pageheader-with-actions)写"icon": "users"。撤掉等于删除一项工作能力并弄坏 demo —— B 案已拒。
变的是依据:注释从 spec parity 改述为 renderer-read 事实;
page-header-authorable-keys.test.tsx的 specKeys 派生跳过墓碑(unwrap optional 后inner 为
never即墓碑,判据与 objectui#3809 一致,实现为局部 helper 并注释指向 #3809收敛),让
icon不再因为Object.keys(shape)仍列着它而假绿。豁免写成具名、引单、自清的
RENDERER_OWN_DECLARED,并附一条纪律断言:条目必须对应一个 spec 曾经声明、现已退役的键(spec 从未有过的键就是方言,
showBack就在同文件被钉为「不声明」)。未碰 alias renderer 的读点。
顺带:
PageBlockInspector.i18n.test.tsx的 #3979 placeholder 测试对从page:header改指
record:alert—— 同样的症状(译过的标签下一条 prose 提示),但钉在渲染器真读的键上。同一处几行之上的 objectui#4649 就是同样的动作与同样的理由。
反向验证(先预判后跑,两次都未提交,跑完
git checkout HEAD --还原,工作树 0 dirty)预判写在跑之前。两个实验的结果与预判逐条一致。
实验 A —— 把删掉的 BLOCK_CONFIG icon 字段装回去(两条 i18n 键保持删除,隔离两半)
预判红 7 条,实测红 7 条,无一条计划外:
预判绿且实测绿的两条也是有意义的:spec 侧墓碑断言与「两个 locale 都已无译文」都不受
设计器字段影响 —— 这正说明字段与 i18n 键必须同进同出,而不是任一可选。后三条红是派生的
双 locale walk 抓到的,它才是这个纪律的机械保证。
实验 B —— 只撤回 specKeys 的墓碑窄化(断言全部保留原样)
这一条的方向不是「全红」,预判时就写明了,实测确认:
红的恰好是本次新增的两条断言。原有的方言守卫
(
declares nothing @objectstack/spec accepts, beyond the declared renderer-own keys)仍然绿,而且这是重点而非缺陷:撤回窄化只是把
icon从「被 carve-out 放行」换回「被
specKeys.has放行」,守卫两个世界都分辨不出来 —— 那份分辨不出来,正是 #3829 在被测量之前一直隐形的原因。若这里报「方言守卫转红」,那就是照模板编的。
测试
Base:
c25222758598aec3d17c5aa22f1640659f1ecacc(显式 sha,非 FETCH_HEAD)。Generated by Claude Code