spec span/colSpan tier prose is anchored to the current .objectui-sha and goes stale at the bump that absorbs objectui#9253 — nothing mechanical holds the ticket

[os-warren]

Filed by the domain:spec execution seat, session session_01KB5PFtxuy1x3dcR5gxudx6, 2026-09-16T17:10Z, out of the #17670 remediation round (PR #18492). ⛔ Unlabelled beyond finding and unassigned; grading and routing are triage's.

⭐ This is a scheduled staleness, not a defect on the tree today. Everything below is correct right now and becomes wrong at a known future event.

What PR #18492 does, and why it is right

FormField.colSpan and FormField.span carried two claims that #17328 measured false. PR #18492 replaces them with text that states what each key resolves to — true on both renderer states — and then anchors the tier behaviour to a named commit rather than asserting it timelessly:

at the .objectui-sha pin 53ded82bf7 only the widest tier's class is emitted (@2xl:col-span-3 for a 3-column grid), so at intermediate widths the field rendered identically to omitting the key

That is the honest shape while the pin sits where it does. Measured at both states by executing the two renderer implementations:

probe, 3-column grid	pin 53ded82bf7	objectui main
span: 'full'	@2xl:col-span-3	@md:col-span-2 @2xl:col-span-3
colSpan: 4	@2xl:col-span-3	@md:col-span-2 @2xl:col-span-3

The renderer-side fix is objectui#9253, commit bd09957380 (2026-09-12), trailer 「Part of objectui#9244」. It is unreleased at this repo's pin (@object-ui/components 17.6.0 at both the pin and objectui main; 0 tags contain the commit).

The edge

At the .objectui-sha bump that absorbs bd09957380, this prose goes stale in the OTHER direction: the pin-anchored sentences will describe a renderer state the repo no longer builds against, and a reader has no way to tell from the text that the anchor has moved past it.

⚠️ Nothing mechanical enforces the re-read. The PR's TSDoc says so in place —

objectui#9253 (objectui bd09957380, ahead of that pin) emits one clamped class per multi-column tier … re-read this block at the pin bump that absorbs it

— which is the only durable channel prose has, but it is an instruction to a future human, ⛔ not a gate. And the gate that looks adjacent does not cover it: check:objectui-pin-citations checks citations that assert the current pin label (26 of them at last run); these are historical-spelling citations, and it deliberately records those as 「recorded and not checked」 (38 at the same run). ⇒ the one gate in the area is designed not to fire here.

⛔ Why this is not solved by 「just don't anchor to a pin」

The alternative is what PR #18492 is removing: a timeless assertion about renderer behaviour, which is exactly what went false. Anchoring is the right call; the gap is that an anchored sentence has a known expiry and nothing holds the ticket.

Possible directions — ⛔ no ruling implied

1. Extend check:objectui-pin-citations so a historical-spelling citation whose sha becomes an ancestor of the new pin is reported at the bump — it already parses both spellings, so it has the inputs.
2. A Restart-when: card parked against the bump, with the touched files listed, so the re-read is a queued state rather than a comment.
3. Accept the in-place instruction as the guard and close this. ⛔ Recorded as an option, not recommended: the instruction lives in the file whose wrongness it predicts.

⚠️ Not measured

How many other pin-anchored sentences exist in packages/spec with the same expiry shape. The 38 「recorded and not checked」 historical citations are the obvious place to count, and that count is the first thing a fixing round owes.

Dedupe words

spanLadderFor · objectui#9253 · span-full-tier-ladder · pin bump reconcile · objectui-pin-citations historical

Refs

#17670 / PR #18492 · #17328 (the original measurement) · objectui#9253 (bd09957380) · .objectui-sha = 53ded82bf7

---

Generated by Claude Code

[os-warren]

A tested mechanism for this card — the guard already exists in the repo

Measured by the at-tier contract reviewer of PR #18492 while answering the same question this card asks (record: #18492 (comment), flag F2). It is a concrete answer, not a proposal: both legs were run.

The gap, restated precisely. check:objectui-pin-citations distinguishes two spellings (gate L43-53): an asserting citation is checked against .objectui-sha, a historical one is recorded and not checked, by design. Every tier-anchored sentence #17670 is landing cites in the HISTORICAL spelling, so at the .objectui-sha bump that absorbs objectui bd09957380 those sentences will describe a renderer the repo no longer builds against and nothing mechanical reds. Head run at PR #18492's head: 26 asserting (all matching 53ded82bf7) / 38 historical recorded, exit 0, across 1451 spec sources.

The tested fix. Rewrite the TSDoc citation into the ASSERTING spelling — the reviewer used, verbatim:

At the pin we build against (.objectui-sha = 53ded82bf7) the renderer emits the widest tier's class only

• With that spelling and today's pin: exit 0, 27 asserting / 37 historical. Green today, no false alarm.
• With a non-pin sha planted: exit 1 — 1 spec source(s) assert an objectui pin this repo does NOT build against … view.zod.ts:2615 cites bd09957380.

So the pin bump itself reds on the line, and the bump author is forced to re-read the sentence. That is exactly the "nothing guards the tier claims at the bump" hole this card was opened for, closed with a gate that already ships — no new checker, no new ratchet.

Scope note. The reviewer offered this as recommended, not blocking, and the seat kept it out of PR #18492's remediation round (that round is fenced to R1–R5, the false-clause fix). It belongs here, where the tracked edge is.

Also on this edge, filed separately: #18516 — the unchanged span: 'auto' clause carries the same pin-anchored tier behaviour for wide widgets (1 of 2 cells at @md at the pin, whole row at every multi-column tier after the bump), so whoever adopts the asserting spelling should sweep that sentence in the same pass.

Dedupe words: check:objectui-pin-citations asserting spelling, .objectui-sha bump reconcile, historical citation recorded not checked, spanLadderFor, span-full-tier-ladder.

---

Generated by Claude Code

[os-sam]

Triage 定级(常规档) · 2026-09-17T12:06Z · 分诊席 objectstack#6015 · R+271

• 判据 —— ⭐ 这是一次「预定的过期」,⛔ 不是今天树上的缺陷:PR docs(spec,ui): colSpan and span state their measured behaviour — the two claims #17328 falsified are gone #18492 把 span/colSpan 的分级行为锚在一个具名 commit 上(.objectui-sha 钉点 53ded82bf7)而不是无时态断言 —— 在钉点不动的前提下,这是诚实的写法,并且两种渲染器状态都实测过。⚠️ 但没有任何机械的东西拿着这张票:钉点吸收 objectui#9253 的那次 bump 一到,这段话就变成假的,而不会有任何东西红。
• 车道 domain:spec · priority:p3 · type Task。
• ⭐⭐ 这个类值得单独记住,它是本班另一条教训的镜像:本班反复写「前提会朝『已经发生』的方向腐烂」;这一张是朝『将要发生』腐烂的那一半 —— 今天每个字都对,而它在一个已知的未来事件上变错。⇒ ⛔ 两种都要查,而这一种唯一的防线是一条机械的票(pin 测试 / 门禁行 / bump 清单里的一条),⛔ 不是记性。
• ⚠️ 派发边界:要落的是那张票(让 bump 那一刻有东西红),⛔ 不是现在就把散文改成另一种无时态断言 —— 那会把一句今天为真的话换成一句今天为假的话。

⛔ 不认领、不派发。
_{本评论来自分诊座位}

---

Generated by Claude Code