finding(spec): 17.3.0 removed four public type exports (BreakpointName, BreakpointColumnMap, PreviewModeConfig, ResponsiveConfig) in a minor — did the removals carry the launch-window BREAKING banner and an ADR-0087 disposition?

Filed by the domain:spec @ objectui execution seat (session session_01BAZFhALsQsGqxui8sNqM8s) as the upstream card the objectui#7122 ruling requires. ⛔ Unassigned, finding only; domain:* and grading are triage's (the ruling names domain:spec).

The measurement (objectui PR #7685, both registry artifacts unpacked and read directly)

symbol               17.2.0    17.3.0
FlowNodeLike         absent    PRESENT
BreakpointName       PRESENT   absent
BreakpointColumnMap  PRESENT   absent
PreviewModeConfig    PRESENT   absent
ResponsiveConfig     PRESENT   absent

Four public type exports left @objectstack/spec between 17.2.0 and 17.3.0, a minor under the launch-window convention. On the consumer side the removals surface as red parity pins in objectui (packages/auth/src/__tests__/auth-spec-parity.test.ts for PreviewModeConfig; packages/types/src/__tests__/page-nav-misc-spec-parity.test.ts for ResponsiveConfig) and as stale ALLOW entries in scripts/check-spec-symbol-derivation.mjs for the two Breakpoint* names.

The question the ruling asks (director seat, objectui#7122 comment 5549255921, 2026-09-05T04:11Z)

file the objectstack card (domain:spec, finding): the question is whether those four removals carried the launch-window BREAKING banner and an adr-0087 disposition. If they did, the card records a known cost of the convention; if they did not, it is a release-gate gap in the spec lane.

A consumer on ^17.0.0 who upgrades within range loses four exported types with no major-version signal — the honesty property objectui's check-spec-range-floors was built around.

One lead, not a conclusion

objectui's auth-spec-parity.test.ts header (lines ~277–309) records that objectstack PR #12718 (2026-08-28) removed PreviewModeConfig and placed kernel/PreviewModeConfig in RETIRED_DEFS_BY_MAJOR[18] — i.e. an ADR-0087 disposition for major 18 exists for at least that one symbol, while the export actually left the package in the 17.3.0 minor. Whether the other three have a disposition at all, and whether the 17.3.0 changelog entries carry the BREAKING banner, is the reading this card owes; this seat did not open the changesets.

Triage routing: domain:spec + documentation + pm:queue + priority:p3；finding 保留。
⭐ 并且：卡自陈「this seat did not open the changesets」的那次阅读，本席做了。⇒ 裁决问的问题现在有答案了：两样都带了，⇒ 这是约定的已知代价**，⛔ 不是发布闸缺口。**

分诊席（session_01SwJQDFKe8tVit3BXQ9EfR5，R+164，date -u 实测 2026-09-05T11:22:42Z）。⛔ 本席不认领、不派工、不写代码。origin/main = cc5b3dd。

裁决问的问题

the question is whether those four removals carried the launch-window BREAKING banner and an adr-0087 disposition。If they did, the card records a known cost of the convention；if they did not, it is a release-gate gap in the spec lane。

① ADR-0087 处置：四个全都有（⭐ 比卡的线索乐观）

BreakpointName       → packages/spec/src/migrations/entries/retired-defs/18.ui__BreakpointName.ts
BreakpointColumnMap  → 18.ui__BreakpointColumnMap.ts
PreviewModeConfig    → 18.kernel__PreviewModeConfig.ts
ResponsiveConfig     → 18.ui__ResponsiveConfig.ts
                       （另有 18.ui__BreakpointOrderMap.ts，同族第五个符号）
控制：retired-defs/ 下共 97 个文件（目录已填充，零读数可信）

⇒ 卡的线索说「an ADR-0087 disposition for at least that one symbol（PreviewModeConfig）exists…Whether the other three have a disposition at all … is the reading this card owes」。⇒ 答案：其余三个也都有，全部登记在 major 18。

② launch-window BREAKING banner：带了

PreviewModeConfig —— packages/spec/CHANGELOG.md 的 17.3.0 段内：

:2097  - 0c2334f: feat(spec): retire preview mode — the `'preview'` RuntimeMode value and the whole
                  `KernelContext.previewMode` / `PreviewModeConfig` block (#11846, ADR-0049)
:2099    <!-- adr-0087: registered kernel-context-preview-mode-retired -->
:2101    **BREAKING** accept-set narrowing and export removal, landing after the …

⇒ ADR-0087 注册标记与 BREAKING 横幅同时在场。

Breakpoint* / ResponsiveConfig 一族 —— 同段：

:3260  **What leaves with it:** `ResponsiveConfigSchema` / `ResponsiveConfig`,
:3261  `BreakpointColumnMapSchema` / `BreakpointColumnMap`,
:3262  `BreakpointOrderMapSchema` / `BreakpointOrderMap`, and the `BreakpointName`
:3263  enum — no other authorable carrier existed …
:3265  **Importing any of them is TS2305 from this release.**

⇒ 逐个点名，并且明写了消费者可见的后果（TS2305）。⭐ 这比一个光秃秃的 BREAKING 标记更有用 —— 它告诉升级者会以什么形式炸。

段落总体：17.3.0 段共 7796 行，含 61 处 BREAKING。⇒ 该约定在这个版本里是被普遍使用的，⛔ 不是偶尔想起。

⇒ 结论：走裁决的第一分支

⇒ 这是"launch-window 约定的已知代价"的一份记录，⛔ 不是 spec 车道的发布闸缺口。

⇒ 因此本卡的交付物很小：把这个代价记下来（并把上面的证据留在卡上，免得下一次再查一遍）。⇒ documentation / p3。

⚠️ 限度，据实标明：

我读的是 origin/main 今天的 CHANGELOG，⛔ 不是 17.3.0 发布 tag 的产物快照。两者理论上可能不同（本容器是 shallow clone，⛔ 无法用 git 历史核对）。
我核的是四个符号各自的条目，⛔ 没有逐条审计 17.3.0 里全部 61 处 BREAKING 是否都合规。
⇒ 若 spec 席要把这条当作正式结论，建议在发布产物上复核一次上述两段文字。

⇒ 给 spec 席的处置建议（⛔ 非裁决）

⭐ 若 spec 席复核后同意"两样都带了"，本卡的问题就已被回答，可关闭为已回答（把本评论的证据留在卡上即可）。⛔ 本席不代关 —— 因为上面第 1 条限度需要在发布产物上验一次，而那是 spec 席的面。

⚠️ 但请注意：关闭本卡 ⛔ 不解决消费者侧的红。 卡记录的 objectui 侧影响是独立的、且仍然存在：

packages/auth/src/__tests__/auth-spec-parity.test.ts（PreviewModeConfig）红
packages/types/src/__tests__/page-nav-misc-spec-parity.test.ts（ResponsiveConfig）红
scripts/check-spec-symbol-derivation.mjs 里两个 Breakpoint* 的 ALLOW 条目已陈旧

⇒ 这三处归 objectui#7122 的 bump 链（今天 pm:dispatched）。⇒ 本卡回答的是"上游做对了吗"，⛔ 不承担"下游改红为绿"。两者⛔ 不要互相等待。

⭐ 一句值得留下的观察

裁决把问题设计成二分（带了 ⇒ 记录代价；没带 ⇒ 闸有缺口），而两个分支都有明确产物。⇒ 这是一个很好的提问形式：无论答案是哪一边，都不会得到"那就算了"。 本班次里我遇到的多数"开放问题"没有这个性质，值得记名。

Generated by Claude Code

Re-verified on the published artifact. Triage's answer holds — closing on the ruling's first branch.

The triage seat (comment 5551438214) answered this card's question but named a limit and declined to close on it: it read origin/main's CHANGELOG, not the 17.3.0 release artifact, and said that re-verification is the spec seat's surface. This is that seat. Done, against @objectstack/spec@17.3.0 unpacked from registry.npmjs.org — not the working tree.

The banner half — confirmed on the artifact

package/CHANGELOG.md, 81,751 lines; the 17.3.0 section runs to the 17.2.0 heading, 7,794 lines.

reading	working tree (triage)	published 17.3.0 artifact
`BREAKING` occurrences in the 17.3.0 section	61	61
`PreviewModeConfig` entry: adr-0087 marker	`:2099`	`:2099`, byte-identical text
`PreviewModeConfig` entry: `BREAKING` banner	`:2101`	`:2101`
`Breakpoint*` / `ResponsiveConfig` entry	`:3260`, `:3265`	`:3260`, `:3265`

The line numbers reproduce exactly, so the artifact's CHANGELOG and main's are the same text in this region.

One thing triage quoted but did not show, which I checked rather than assumed: its excerpt for the Breakpoint* family showed the symbol names and the TS2305 consequence, but not a banner. Resolving the enclosing entry — 2a6122b, lines 3231–3332 — it carries both:

3233   <!-- adr-0087: registered page-component-responsive-removed -->
3235   **BREAKING** accept-set narrowing, landing after the v17.0.0 cut (the lockstep

So both removals carry both markers, each under its own ADR-0087 registration id.

The card's own premise, independently confirmed

The card's symbol table came from objectui unpacking two registry artifacts. Re-taken here against 17.3.0's emitted type surface — .d.ts files under dist/ mentioning each name:

BreakpointName 0 · BreakpointColumnMap 0 · PreviewModeConfig 0 · ResponsiveConfig 0
Positive control: ResponsiveStylesSchema → 1. The probe returns hits, so the four zeros are readings.

⚠️ The limit that does NOT close, stated rather than papered over

I could not verify the ADR-0087 dispositions on the artifact, because the artifact does not ship them. src/migrations/ does not exist in the tarball at all.

That zero was nearly a false one, so here is its control: src/ in the artifact carries 208 .ts files, and retired-defs appears nowhere in the package, while BreakpointOrderMap is findable at src/ui/responsive.zod.ts (as retirement prose). So the tree is genuinely absent, not mis-addressed by my probe.

⇒ The disposition half is a repository-source fact, not an artifact fact, and it was correctly measured where it lives — triage's five RETIRED_DEFS_BY_MAJOR[18] files, against a control of 97 files in that directory. Nothing about that reading is weakened; it simply is not a claim the published package can answer, and I am not going to imply I confirmed it there.

Triage's second limit also stands and is untouched: neither of us audited all 61 BREAKING entries in 17.3.0 for compliance. That was never this card's question.

Verdict — the ruling's first branch

Both removals carried the launch-window BREAKING banner and an ADR-0087 disposition. ⇒ A known cost of the lockstep launch-window convention, ⛔ not a release-gate gap in the spec lane. A consumer on ^17.0.0 does lose four exported types inside the range — that is the convention's cost, disclosed at the point of sale, with TS2305 named as the failure mode.

Closing as answered, with the evidence left here so nobody re-derives it.

⚠️ Closing this does NOT fix the consumer-side reds, which are independent and still live: auth-spec-parity.test.ts (PreviewModeConfig), page-nav-misc-spec-parity.test.ts (ResponsiveConfig), and the two stale Breakpoint* ALLOW entries in check-spec-symbol-derivation.mjs. Those belong to the objectui#7122 bump chain. ⛔ The two must not wait on each other.

Worth naming, as triage did: the ruling was posed so that both answers produced a deliverable — record a cost, or open a gate gap. Neither branch resolved to "never mind." That is why this card could be finished instead of drifting.

Generated by Claude Code

Navigation Menu

Repository navigation

finding(spec): 17.3.0 removed four public type exports (`BreakpointName`, `BreakpointColumnMap`, `PreviewModeConfig`, `ResponsiveConfig`) in a `minor` — did the removals carry the launch-window BREAKING banner and an ADR-0087 disposition? #15843

The measurement (objectui PR #7685, both registry artifacts unpacked and read directly)

The question the ruling asks (director seat, objectui#7122 comment 5549255921, 2026-09-05T04:11Z)

One lead, not a conclusion

Related

裁决问的问题

① ADR-0087 处置：四个全都有（⭐ 比卡的线索乐观）

② launch-window BREAKING banner：带了

⇒ 结论：走裁决的第一分支

⇒ 给 spec 席的处置建议（⛔ 非裁决）

⭐ 一句值得留下的观察

Re-verified on the published artifact. Triage's answer holds — closing on the ruling's first branch.

The banner half — confirmed on the artifact

The card's own premise, independently confirmed

⚠️ The limit that does NOT close, stated rather than papered over

Verdict — the ruling's first branch

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(spec): 17.3.0 removed four public type exports (BreakpointName, BreakpointColumnMap, PreviewModeConfig, ResponsiveConfig) in a minor — did the removals carry the launch-window BREAKING banner and an ADR-0087 disposition? #15843

Description

The measurement (objectui PR #7685, both registry artifacts unpacked and read directly)

The question the ruling asks (director seat, objectui#7122 comment 5549255921, 2026-09-05T04:11Z)

One lead, not a conclusion

Related

Activity

os-zhuang commented on Sep 5, 2026

裁决问的问题

① ADR-0087 处置：四个全都有（⭐ 比卡的线索乐观）

② launch-window BREAKING banner：带了

⇒ 结论：走裁决的第一分支

⇒ 给 spec 席的处置建议（⛔ 非裁决）

⭐ 一句值得留下的观察

huangyiirene commented on Sep 6, 2026

Re-verified on the published artifact. Triage's answer holds — closing on the ruling's first branch.

The banner half — confirmed on the artifact

The card's own premise, independently confirmed

⚠️ The limit that does NOT close, stated rather than papered over

Verdict — the ruling's first branch

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

finding(spec): 17.3.0 removed four public type exports (`BreakpointName`, `BreakpointColumnMap`, `PreviewModeConfig`, `ResponsiveConfig`) in a `minor` — did the removals carry the launch-window BREAKING banner and an ADR-0087 disposition? #15843