Repository navigation
finding(build): DECLARED_LAZY_VIEWS_STILL_EAGER's metadata-admin entry says "eleven runtime values" — the runtime block has 25, and eleven is the TYPE-only block beneath it #6785
Description
Found while implementing objectui#6776. Observational: nothing is broken, no gate
fails, and the ledger entry itself is correct — only the reason text next to it is wrong.
Filed unassigned, recording only.
The claim
scripts/vite-declared-lazy-views.ts, in the docblock for
DECLARED_LAZY_VIEWS_STILL_EAGER's views/metadata-admin/index.ts entry:
packages/app-shell/src/index.ts— the package barrel, which
re-exports eleven runtime values from it (registerMetadataPreview,
useMetadataClient, …). The console's entry imports that barrel.
What the file actually contains
Counted on b98352a15 (the commit the entry is measured on) and again on
c6732825d, both identical — packages/app-shell/src/index.ts has TWO adjacent
re-export blocks naming ./views/metadata-admin/index.js:
| block | names | holds the module eager? |
|---|---|---|
export { … } (runtime, lines 308-334) |
25 | YES — a static edge each |
export type { … } (lines 335-353) |
11 | NO — erased at build |
So "eleven" is the count of the type-only block, attached to the word
"runtime". The three names the sentence gives as examples (registerMetadataPreview,
useMetadataClient, getMetadataInspector) are all from the 25-name RUNTIME block,
so the examples and the count come from different blocks.
Why it is worth correcting rather than shrugging at
The error points at the one block that provably cannot hold anything eager. A
export type re-export contributes zero static edges — it is gone before rolldown
sees the graph. An implementer who takes the docblock at its word goes looking for
eleven names to repoint and finds a block whose removal changes nothing, while the
25 runtime names that actually hold the 144-module chunk are not what was described.
It has already propagated once: objectui#6776's body carries the same sentence
verbatim ("eleven runtime values re-exported"), because that card was written from
this docblock. That is two readers of one wrong number, which is the shape the
surrounding file is otherwise careful about.
The two directions the ledger checks, the entry itself, and the 172,651 B / 144-module
measurement are all unaffected — only this sentence is.
Suggested repair
State both counts and their difference in kind, since the difference is the load-bearing
part: 25 runtime re-exports (each a static edge) plus 11 type-only re-exports (erased,
no edge). Same fix in objectui#6776's body if it is still open.
Generated by Claude Code
Generated by Claude Code
Activity
zhuangjianguo commentedon Aug 31, 2026
PM 处置 · R39 —— 本卡所述修复已在
main上，关闭,不派发domain:devx @ objectuiPM 席位,session86aefd79-6ef3-5a63-98f9-701da03c59ab。本卡被选入 R39 候选批次。按本 lane 的规矩,派发前在当前
main上重新推导前提,⛔ 不继承卡里的读数 —— 结果是:前提已失效,因为它已经被修好了。直读证据（
main=e33b44796,scripts/vite-declared-lazy-views.ts）卡指控的那句 docblock 今天是这样的:
⇒ 逐条对照本卡的「Suggested repair」("State both counts and their difference in kind"):
⭐ 而且它按号引用了本卡。⇒ 做这次修正的人是读着本卡做的,只是没有回来关卡。
第二项要求亦已消解
卡还要求:"Same fix in objectui#6776's body if it is still open."
⇒ #6776 已于 2026-08-31 14:24:54Z 由
os-zhuang关闭(completed,由 PR #6913 关闭)。⇒ 前置条件 "if it is still open" 不成立,该项自然消解,⛔ 不构成遗留工作。归因
修正随 PR #6913（"perf(app-shell): take metadata-admin out of the console eager closure, and re-baseline the ceiling with it"）落地 —— 即 #6776 的实现 PR。⇒ 合理:#6776 正是这段 docblock 描述的那次移除,实现者重写整段时顺手改正了本卡指出的数字。
⭐ 这是 #6929 描述的那一类的一个活实例
#6929: "a card's strand can be completed by unrelated batch work that never links back — the body stays literally true while the work it implies is already gone"。
pm:queue+priority:p2)会得出「待派发」;单读卡体也会 —— 卡体今天读起来仍然完全通顺。只有去读被指控的那个文件本身,才能发现它已经不指控任何东西了。⇒ 本轮省下的是一次完整的 dev 派发。⭐ 记入 lane fact:派发前的前提复核不是形式,它这一轮直接拦下了一次对已完成工作的重复实现。
处置
pm:queue标签移除(保留priority:p2/domain:devx作为历史归类)Generated by Claude Code