Skip to content
This repository was archived by the owner on Nov 4, 2026. It is now read-only.

Repository navigation
This repository was archived by the owner on Nov 4, 2026. It is now read-only.

AppContent.pseudoRouteSegments.test.tsx 里还有第二份 MetadataRedirectStub,同样在 #3658 后失真(自称逐字转录,仍走废弃别名) #3669

Closed
#3691
Closed
AppContent.pseudoRouteSegments.test.tsx 里还有第二份 MetadataRedirectStub,同样在 #3658 后失真(自称逐字转录,仍走废弃别名)#3669
#3691
Assignees
yinlianghui
Labels
pm:dispatchedtests

Description

@yinlianghui
yinlianghui
opened on Aug 7, 2026

越界发现,记录于 #3661 / PR 修 stub 失真期间。#3661 的文件面被明确限定为 AppContent.noAppComponentRoutes.test.tsx 单文件,本文件在围栏外,按纪律只报不改、未认领。

观察类:测试是绿的,CI 不会报,没有任何用户会撞上。故打 finding、不打 pm:queue。

事实

packages/app-shell/src/console/__tests__/AppContent.pseudoRouteSegments.test.tsx:218-233 有第二份 MetadataRedirectStub,docblock 与 #3661 修掉的那份措辞几乎一致:

> VERBATIM transcription of apps/console/src/AppContent.tsx's MetadataRedirect — the system/metadata/* legs of systemRoutes. Keep the regex and the target construction identical to the original.

而它的 target 构造仍是 #3658 之前的别名拼法:

  const base = `${prefix}/component/metadata/resource`;
  const target = !metadataType
    ? `${prefix}/component/metadata/directory`
    : itemName
      ? `${base}/${itemName}?type=${metadataType}`
      : `${base}?type=${metadataType}`;

宿主自 7883c0250(PR #3658)起已改为直接构造 …/metadata/:type[/:name],所以这份转录同样不再逐字,而 docblock 仍在指示后来者「keep identical to the original」。

受影响的用例

这三行是被这份 stub 喂的,现在走的是生产中已不存在的两跳链:

  • :300 ['/apps/ghost/system/metadata/object', 'metadata-resource-list-page', '/apps/ghost/metadata/object'] —— 注释还明写 "Two-hop: host rewrite (system/metadata/:type) -> shell alias -> canonical",该注释现已是事实错误;
  • :321 ['/apps/setup/system/metadata/object', 'metadata-resource-list-page'](setup family 块)。

它们仍绿,理由与 #3661 完全相同:断言只钉最终落点(或只钉 testid),而落点在一跳与两跳下逐字节相同。绿的理由已与被测对象脱钩。

与 #3661 的关系

同一机制、同一来源改动(#3658)、同一处置形态,但不同文件,且 #3661 的派发单把文件面收得很紧,所以没有一并修。#3661 的 PR 里已经把可复用的手法落地了:LocationProbe 收集整条重定向链,断言 toEqual 整个数组而非只断终点 —— 逆向验证实测为「回退 stub → 恰好两条用例转红,diff 各多出一行别名,首尾条目不变」。这份 stub 照抄该手法即可,不需要重新设计。

⚠️ 一处需要接手者注意的连带项:本文件 :299 的 ['/apps/ghost/component/metadata/directory', …] 是直接进入别名,不经 stub,所以同步 stub 不会让它失去覆盖(与 #3661 那边不同 —— 那边零应用分支的 component/metadata/directory 只有 stub 的 typeless 臂在跑,PR 已补了一条直接进入的用例把覆盖接住)。同步前请确认本文件对应分支的别名覆盖同样有着落。

关联

已就关键词(MetadataRedirectStub / pseudoRouteSegments / verbatim transcription / component/metadata)搜过本仓开放 issue 与 PR,无同源单(#3660 / #3666 是生产端导航与 docblock,不含本文件)。

Generated by Claude Code

Activity

  1. yinlianghui
    added
    tests
    finding
    pm:dispatched
    on Aug 7, 2026
  2. yinlianghui
    self-assigned this
    on Aug 7, 2026

  3. yinlianghui commented on Aug 7, 2026

    @yinlianghui
    yinlianghui
    on Aug 7, 2026 – with Claude
    ContributorAuthor

    认领(PM 派发)。

    Generated by Claude Code

  4. yinlianghui commented on Aug 7, 2026

    @yinlianghui
    yinlianghui
    on Aug 7, 2026 – with Claude
    ContributorAuthor

    验收 ACCEPT → PR #3691(session_01GTRjn8xBqp75dk7kFupVRt)。

    已核:CI 17 项 15 成功 2 跳过零失败;逐字转录用剥注释机器 diff 验证(IDENTICAL);逆向验证恰 2 红且 diff 各只多一行别名中间站 — 「旧断言结构上不可能红」的直接证据再次成立。

    裁决记录:

    1. 覆盖审计的对比论证记入方法库:四条别名声明逐条落实覆盖来源,「直接进入 vs stub 途经」的区分决定要不要抢救 — 与 test(app-shell): 把 MetadataRedirectStub 同步回宿主实现,并用整链断言钉住转录真实性 (#3661) #3671 的差异不是省略而是论证出的不需要。
    2. 探针取数新技法收录(reporter 吞 console.log → 必失败断言让实际值进 diff)。
    3. stub 两条从未执行的臂如实记录不冒领。

    #3661/#3669 双 stub 同步闭环。流程注:本单完成通知曾在 worker 重启窗口丢失,探活找回 — 交付本身零瑕疵。

    Generated by Claude Code

  5. os-zhuang
    removed
    finding
    on Aug 8, 2026

  6. os-zhuang commented on Aug 8, 2026

    @os-zhuang
    os-zhuang
    on Aug 8, 2026 – with Claude
    Contributor

    Label hygiene only — finding removed. ⛔ Not a re-grade, not a claim, and nothing here touches the lane's work or its PR.

    This card carried finding + pm:dispatched simultaneously, which is the half-state label discipline forbids: finding means "held, not dispatchable until graded", and pm:dispatched means the opposite. The two readers disagree — the findings-triage round sees a card awaiting a verdict, while the queue view sees one in flight. Removing finding aligns the labels with the lane's own decision, already recorded on this card (claim at 18:33Z, session session_01GTRjn8xBqp75dk7kFupVRt, branch claude/issue-3669-second-stub-sync; ACCEPT at 21:46Z on PR #3691). It is not this seat overriding a lane call — it is this seat writing the lane's call into the label the machine reads. tests and pm:dispatched are untouched.

    State as of this round, for the record: PR #3691 is accepted but not yet on main — the newest commit touching packages/app-shell/src/console/__tests__/AppContent.pseudoRouteSegments.test.tsx on origin/main (0cf8f0f) is still 5f752a0 (#3638 via #3654), which is the change the second stub drifted from. The card closes when that PR merges; the owning lane keeps it, as the claim protocol requires.

    Side effect worth stating plainly, since it moves a number this seat reports: objectui's open finding count goes 20 → 19, and none of that is a grading decision — it is one card that was never actually held.

    本评论来自分诊座位 Routine（#5474 试点），不构成认领。

    Generated by Claude Code

  7. os-project-manager
    closed this as completedin #3691on Aug 8, 2026
  8. akarma-synetal
    added a commit that references this issue on Aug 10, 2026

    test(app-shell): 同步 pseudoRouteSegments 里的第二份 MetadataRedirectStub,并用…

    ae3bd96
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

pm:dispatchedtests

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions