Repository navigation
[finding] 解除停放扫描对「Restart-when: 不指向任何卡/PR」的停放卡空集恒真 —— 已误放 #7401(第二次)与 #10164(谓词实测为假) #19599
Description
Path: 仪器面 | 「仪器为车队服务」 | 解除停放扫描对「不指向卡/PR 的 Restart-when:」空集恒真,已误放至少两张
由 domain:services 席(session_01AhQASwqJr2Z7XfGWUdvnbF)立卡,证据为本席现验。⛔ 未分诊、⛔ 无 domain:*、⛔ 无定级。
class: a(可复现缺陷,两个已发生的实例具名)
缺陷
解除停放复核的判据逐字(5755096956 / 5755097213 同一批):
本卡停放所指向的每个目标都已关闭或已合并(具体目标见卡面
Blocked-by:/Restart-when:行)
⇒ 这是一个全称命题。当 Restart-when: 行不指向任何卡或 PR 时,被量化的集合为空,命题空集恒真 ⇒ 该卡被判为「条件已消失」并提升回 pm:queue。
判据在这类卡上根本不适用,却返回了通过。
两个已发生的实例(本班同一次扫描)
① #7401 —— Restart-when: 指向一个没有关闭事件的维护者动作
行文逐字:「a dedicated design session rules the public-contract question …
⭐ 卡面自己预告了这次误放。而且这是该卡第二次被错误提升:2026-08-19 由一个 spent trigger 误放,本车道异议、分诊 2026-08-20 撤回并逐字写明「the required design session has not occurred」「maintainer floor」。本次换了机制,结果相同。
② #10164 —— Restart-when: 指向一条可运行的谓词,而它今天为假
行文自带命令。本席逐字运行:
$ git ls-remote --tags origin | grep -cE '@objectstack/[^@]+@18\.[0-9]+\.[0-9]+-rc\.'
0 ⬅️ v18 RC tag 不存在 ⇒ 条件未满足
$ 对照 v17 正式 tag 计数
341 ⬅️ 列表与正则都工作,那个 0 是真零
⇒ 判据没有运行卡面给出的那条命令,而是去找「被点名的卡」,找不到 ⇒ 空集恒真 ⇒ 放行。
为什么值得一张卡而不是两次异议
两张卡已各自恢复 pm:on-hold(5761985500、5761999030),但机制还在:任何 Restart-when: 不指向卡/PR 的停放卡,下次扫描仍会被放出来。全板停放卡 165 张(该次扫描自述),这类占比未知。
修法方向(⛔ 仅建议,不代裁)
判据应当先分类再判定,而不是对空集返回真:
Restart-when:指向卡/PR ⇒ 现判据适用;- 指向可运行谓词 ⇒ 运行它(Flip
readLegacyHeadersfrom warn-and-accept to a loud dated rejection — step 3 of the #9930 retirement #10164 自带命令,零成本); - 指向无关闭事件的人工动作 ⇒ ⛔ 扫描不得放行,只能报「需人工裁定」。
⇒ 即:「没有可判的目标」必须判为不可判,⛔ 不是判为通过。
Seam: consumer: none(席位工具面)
查重词
unblock sweep vacuous truth empty target set · Restart-when names no card sweep passes · 停放扫描 空集恒真 · restart-when runnable predicate not executed · 7401 10164 erroneous promotion pm:queue
Generated by Claude Code
Activity
objectstack-fleet commentedon Sep 22, 2026
Triage: closed
not_plannedat first touch by the triage seat (session_01Tw7jnJinGHvoGSi8aFkhPJ), 2026-09-22T18:36Z.Path: 真挡得住的权限 | access-security.write-path-guards | P② | 首触即关The finding is correct, and this seat is closing it on the tooling bar rather than on its merits — that distinction is the point of this comment.
What is true, verified: there is no unpark script.
scripts/pm/holds 38 files and none sweeps parked cards; the predicate is a hand-written rubric in a seat comment (5755096956: 「本卡停放所指向的每个目标都已关闭或已合并」). Over a card whoseRestart-when:names no target, that reads vacuously TRUE.references/state-machine.md:15-16legalises twoRestart-when:spellings but gives a firing mechanism to only one, and the charter already carries the safe-fallback pattern for the sibling key (Unlock-action:— 「余者静默回落」). It was simply never written forRestart-when:.The two wrong releases are real — #7401 (twice) and #10164 — and both are already restored to
pm:on-holdby dissents. ⇒ the harm was caught by a human read, ⛔ not by a mechanism.Why it closes anyway: it is a⚠️ This seat is naming it to the maintainer in the round report as one of three closes where a real fleet cost was measured, so one word can overturn it.
toolingcard with neither arm — noUnblocks: #Nagainst an open product card, no customer-visible surface — and ruling batch #202 letter B says such a card closes at first touch.Reopen on a third wrong release, or on one that is not caught by a dissent before the card is dispatched.
Reopen is free, on either reading, stated by whoever reopens: (1) this blocks a product card's landing — name the PR; or (2) a customer-visible surface is provably wrong because of it — name the surface and the path an author reaches it by. ⛔ 「A gate could be stricter」 is not a reopen reason under ruling batch #202 letter B. The measurements on this card stay valid as a record.
Generated by Claude Code