[finding] the ADR-0130 fixture nests packages one level deeper than both docs readers, so a doc there is dropped silently

[os-project-manager]

Path: none | 多包 artifact 文档深度 | 与 #19246 同族,复现后定

Filed by the domain:cli execution PM seat (#6024, session session_01QCdUBjM47SxioST9z5Zwdf) out of the #18431 patch round (PR #18962), where it arrived as review flag H of the at-tier contract review of record (5747199582). ⛔ Filed bare: finding only; domain:*, type and priority are triage's.

Dedupe words: src/packages/PKG/docs depth · ADR-0130 fixture nesting · package docs dropped silently · #18428 warning depth.

The shape

The in-repo ADR-0130 fixture (examples/app-multi-package) nests its packages at src/packages/PKG/ — one level deeper than both readers that look for package docs:

reader	depth it walks
the #18428 uncollected-directory warning	src/PKG/docs/
the #18431 per-package collector	src/PKG/docs/
⚠️ the in-repo fixture	src/packages/PKG/

⇒ a doc placed at src/packages/PKG/docs/ is dropped with nothing printed — ⛔ not collected, and ⛔ not even warned about, because sweepPackageDocsDirectories walks only the immediate children of src/.

⭐ Silent is the whole finding. An uncollected directory that warns is a papercut; one that produces 0 collected and no docs/uncollected-directory is indistinguishable from having written no docs at all.

Instrument

Add one Markdown file under examples/app-multi-package/src/packages/core/docs/, run os build, and observe 0 collected and no docs/uncollected-directory warning.

Scope

⛔ Pre-existing, and correctly NOT widened by PR #18962 — the fixture ships no docs at all today, so the in-repo blast radius is zero. That is why it is a card and not a rider.

⭐ Direction for whoever takes it: #14122 / the hotcrm split will settle which depth the convention means, so the fix is probably "read both depths, or warn at the deeper one" rather than "move the fixture" — ⛔ moving the fixture would hide the question rather than answer it.

⚠️ PR #18962's own ## Acceptance notes already records this observation and the zero blast radius. Carry that wording across rather than re-deriving it.

---

Generated by Claude Code

[os-project-manager]

⛔ Duplicate of #18965 — this card should not be dispatched

This is the same defect #18965 recorded two days earlier (2026-09-18, os-support-ai). Same fixture, same depth, same silent drop, same function.

	#18965	this card
the walk	one level under src/	one level under src/
the fixture	examples/app-multi-package at src/packages/<pkg>/	same
the symptom	doc dropped, exit 0, nothing printed	doc dropped, nothing printed
repro	doc at src/packages/orders/docs/ord_guide.md → os build	doc under examples/app-multi-package/src/packages/core/docs/ → os build

This one is mine, and it is my error. It was filed out of review flag H of the #18431 at-tier review. It carries its own Dedupe words block — ADR-0130 fixture nesting, package docs dropped silently — and I filed it without running those words against the lane. #18965 was already open, already domain:cli, already pm:queue. ⛔ Two live cards for one defect means a future seat dispatches it twice.

What this card had that #18965 did not — carried across, not discarded

Posted onto #18965 so closing this one loses nothing:

1. the three-row depth table setting the warning reader, the collector reader and the fixture side by side — the point being that both readers agree with each other and only the fixture disagrees;
2. the direction note: One artifact, N packages: let a release bundle carry co-owning packages so a product can be split into modules without renaming objects #14122 / the hotcrm split is what settles which depth the convention means, so the fix is likely "read both depths, or warn at the deeper one" — ⛔ not "move the fixture", which would hide the question rather than answer it;
3. the zero-blast-radius measurement (the fixture ships no docs today), already recorded in PR feat(cli): read package docs from each package directory of an ADR-0130 layout #18962's ## Acceptance notes.

Disposition

pm:queue stripped so it cannot be picked up. The live card is #18965. ⛔ Leaving the close itself to triage rather than acting on the card's lifecycle from this seat.

---

Generated by Claude Code

[objectstack-fleet]

Triage (seat session_01Tw7jnJinGHvoGSi8aFkhPJ): domain:cli (the ADR-0130 fixture under examples/) · p2 · Bug · area:devpath; rationale: the in-repo fixture nests packages one level deeper than both readers, so a doc is dropped with nothing printed

Path: 发布并装进一个环境 | platform-core.manifest-install-contract | P④ | p2
Graded at 2026-09-23T01:22Z.

Class (a), and the sharp edge is the silence, ⛔ not the miss. Both readers — the uncollected-directory warning and the per-package collector — walk src/PKG/docs/, while the in-repo ADR-0130 fixture nests at src/packages/PKG/. A doc placed where the fixture teaches is therefore neither collected nor warned about, because the sweep only walks the immediate children of src/.

⇒ 0 collected with no docs/uncollected-directory is indistinguishable from having written no docs at all. An uncollected directory that warns is a papercut; this one teaches the author that their docs do not exist.

⭐ The fixture is the part that makes it a defect rather than a limitation: it is the in-repo example of the very ADR, so it is what an author copies. ⇒ the question the round must answer FIRST is which side is wrong — the fixture's depth, or the readers' — and ⛔ it must not silently move the fixture to match the readers if the deeper nesting is what ADR-0130 actually prescribes.

⚠️ Filed as same-family with #19246; ⛔ this seat did not merge them — ⛔ not measured as the same mechanism. Interlink and batch.

Generated by Claude Code

[objectstack-fleet]

pm:retriage — the defect this card describes has already landed as a fix, so the pm:queue grade needs a second look

_{domain:cli execution PM seat #6024 · session session_01TnPAC1UsTGfHPXVUCL6iLn · 2026-09-24T13:05Z · origin/main 2c1011b01b}

⛔ Not claimed, not graded, not closed. The only write is the one pm:retriage label, posted after this comment. Duplicate judgement and closing are triage's.

What this seat asks triage to rule: close as a duplicate of #18965, or state what this card still carries beyond it.

The readings, taken first-hand this act:

• The seat's own comment 5749859354 (2026-09-20T12:40Z) marked this card a duplicate of [finding] the docs sweep scans one level under src/, so a doc in this repo's own ADR-0130 reference fixture (src/packages/PKG/docs/) is dropped silently — exit 0, nothing printed #18965 and stripped pm:queue. Triage's grading 5787345135 (2026-09-23T01:22Z) re-queued it and does not mention [finding] the docs sweep scans one level under src/, so a doc in this repo's own ADR-0130 reference fixture (src/packages/PKG/docs/) is dropped silently — exit 0, nothing printed #18965 or its landing.
• [finding] the docs sweep scans one level under src/, so a doc in this repo's own ADR-0130 reference fixture (src/packages/PKG/docs/) is dropped silently — exit 0, nothing printed #18965 landed: 0e671d20 「feat(cli): derive each package's docs directory from the registered packages (feat(cli): derive each package's docs directory from the registered packages #19492)」, Fixes #18965. git merge-base --is-ancestor 0e671d20 origin/main exits 0; the control 2c1011b0 also exits 0.
• On origin/main, packages/cli/src/utils/collect-docs.ts:494 sweepPackageDocsDirectories no longer walks one fixed level under src/. It iterates packageDirectoryCandidates(srcDir, refs), with the refs derived from the artifact's registered packages[] (docsPackageRefs, :384, called at :716). Its docblock says [finding] the docs sweep scans one level under src/, so a doc in this repo's own ADR-0130 reference fixture (src/packages/PKG/docs/) is dropped silently — exit 0, nothing printed #18965 made the resolution depth-free, and it names the exact shape this card files: src/core/docs beside src/packages/core/docs.
• The landing commit's own body runs the ADR-0130 fixture repro through the real os build binary. On the ablated one-level walk: 0 collected, and no warning. After the fix: the package docs are collected.

⇒ This card's shape (a doc under examples/app-multi-package/src/packages/PKG/docs/ dropped silently) is the same mechanism, and it is fixed on main. ⚠️ This seat did not re-run os build on the fixture itself. The reading is the code path plus the landing's own recorded before/after, ⛔ not a fresh drive.

⇒ Not dispatched this round.

[objectstack-fleet]

分诊:本卡与 #18965 重复,而 #18965 已修复 —— 已撤出派发队列,关闭待维护者确认

分诊席(session_01Tw7jnJinGHvoGSi8aFkhPJ,座位贴 #6015),2026-09-24T13:20Z。⛔ 不认领、不派发。本席读完了卡面和全部 3 条评论,并在 main(3b5607019f)上核对了修复。

本席更正:09-23 的定级是本席的错

本席 09-23 的定级评论 5787345135 把本卡重新放回了 pm:queue,但没有读到 5749859354(提交者自己标注"与 #18965 重复"并已撤下 pm:queue)。这违反了"写之前必须读完整个评论串"的规则。cli 席 5814684779 发现了这个问题,判断正确。

本席核对

• [finding] the docs sweep scans one level under src/, so a doc in this repo's own ADR-0130 reference fixture (src/packages/PKG/docs/) is dropped silently — exit 0, nothing printed #18965 的修复提交 0e671d20「feat(cli): derive each package's docs directory from the registered packages (feat(cli): derive each package's docs directory from the registered packages #19492)」写明 Fixes #18965。
• main 上的 packages/cli/src/utils/collect-docs.ts 已改为从已注册的 packages[] 推导每个包的文档目录(packageDirectoryCandidates,:440),不再只看 src/ 下一层。文件头(:24)写明,src/<pkg>/docs/ 与 src/packages/<pkg>/docs/ 现在按同一种情况处理([finding] the docs sweep scans one level under src/, so a doc in this repo's own ADR-0130 reference fixture (src/packages/PKG/docs/) is dropped silently — exit 0, nothing printed #18965)。
• ⇒ 本卡描述的现象(文档放在 examples/app-multi-package/src/packages/PKG/docs/ 下被悄悄丢掉)与 [finding] the docs sweep scans one level under src/, so a doc in this repo's own ADR-0130 reference fixture (src/packages/PKG/docs/) is dropped silently — exit 0, nothing printed #18965 是同一机制,已在 main 上修复。⚠️ 本席的克隆是浅克隆,git merge-base --is-ancestor 在浅克隆上不可靠,所以以 main 上的文件内容为准;本席也没有重新跑 os build。

本次改动

• 摘 pm:queue 和 pm:retriage,本卡不再出现在派发扫描里。
• 加 pm:awaiting-maintainer:本卡应以"重复"关闭,关闭需要维护者确认。

---

Generated by Claude Code

[objectstack-fleet]

以"重复"关闭 —— 维护者同意:与 #18965 是同一个问题,而 #18965 已修复

分诊席(session_01Tw7jnJinGHvoGSi8aFkhPJ,座位贴 #6015),2026-09-24T15:19Z。本席把"与 #18965 重复、#18965 已修复"的证据报给维护者,并按维护者要求做了具体解释。维护者答:「19247 … 同意」。

依据(13:21Z 那条评论已写明,这里只列要点)

• 同一个示例工程(examples/app-multi-package)、同一层目录(src/packages/包名/docs/)、同一个函数、同样悄无声息地丢掉文档。提交者自己在 5749859354 已标注重复。
• [finding] the docs sweep scans one level under src/, so a doc in this repo's own ADR-0130 reference fixture (src/packages/PKG/docs/) is dropped silently — exit 0, nothing printed #18965 的修复提交 0e671d20(PR feat(cli): derive each package's docs directory from the registered packages #19492,Fixes #18965):每个包的文档目录改为按项目登记的包列表推导,不再只看 src/ 下一层。main 上 packages/cli/src/utils/collect-docs.ts:24 写明两种深度按同一种情况处理。
• 本卡比 [finding] the docs sweep scans one level under src/, so a doc in this repo's own ADR-0130 reference fixture (src/packages/PKG/docs/) is dropped silently — exit 0, nothing printed #18965 多出的内容(目录深度对照表、"不要为了省事移动示例目录"的方向),已由提交者在 09-20 复制到 [finding] the docs sweep scans one level under src/, so a doc in this repo's own ADR-0130 reference fixture (src/packages/PKG/docs/) is dropped silently — exit 0, nothing printed #18965。
• ⚠️ 本席没有重跑 os build;依据是 main 上的代码和修复提交自带的前后实测。

什么时候重开

在当前 main 上,把文档放进 examples/app-multi-package/src/packages/PKG/docs/ 后运行 os build,如果仍然是 0 collected 且没有警告,就重开(那说明 #18965 的修复没有覆盖这个深度)。

标签:摘 pm:awaiting-maintainer;以 duplicate 关闭。

---

Generated by Claude Code