Skip to content

docs(scripts): 按真实机制改写两处 Lychee 门禁描述,并删掉 judgeHref 重复注释 (#3587) (#3648) - #3656

Merged
yinlianghui merged 1 commit into
mainfrom
claude/issue-3587-lychee-comment-truth
Aug 7, 2026
Merged

docs(scripts): 按真实机制改写两处 Lychee 门禁描述,并删掉 judgeHref 重复注释 (#3587) (#3648)#3656
yinlianghui merged 1 commit into
mainfrom
claude/issue-3587-lychee-comment-truth

Conversation

@yinlianghui

@yinlianghuiyinlianghui commented Aug 7, 2026

Copy link
Copy Markdown
Collaborator

Fixes#3587
Fixes#3648

纯注释改动,零行为变化,两文件三处。

前提复核(先按内容锚定,再改)

正文引的行号确已大漂,但三处失真都还在origin/main(93c2619,含 #3649):

处issue 说的行今天实际结论
#3587 spot 1check-doc-links.mjs:157仍是 :157存在,未被改写
#3587 spot 2check-doc-links.test.ts:853漂到 :886存在,未被改写
#3648 重复注释check-doc-links.mjs:689-696漂到 :764-771存在,git blame 指向 0e4ea07(#3629)

.github/workflows/check-links.yml 现状我重测过,不是照抄正文:

所以「gates nothing」这个结论对,机制反了。

改动一:机制正确的表述(#3587)

scripts/check-doc-links.mjs(§2 自仓 blob/tree URL 那段):

改前

 * thing that would resolve it — is a weekly cron with `continue-on-error`, so
* it gates nothing.

改后

 * thing that would resolve it — is a `schedule` + `workflow_dispatch` workflow
* with no PR trigger, so it blocks nobody. (It is `fail: true`, not a
* soft-failing job: what keeps it off pull requests is the absent trigger, not
* a tolerated failure. #3213 ruling B says keep it that way — uncommenting
* `pull_request:` would turn it into a hard gate over the network.)

措辞对齐同一文件 :194 起已有的正确版本(#3589 写的头注释:"that workflow is schedule + workflow_dispatch only … so it blocks nobody")——同一文件里两种说法并存,本来就是下一个读者踩坑的地方。括号那句是 issue 的核心指控落地:按旧注释行事的人会去摘一行不存在的 continue-on-error,而真实风险方向相反。

scripts/__tests__/check-doc-links.test.ts:886,某断言上方的说明文字,只改文字,断言一行未动:

改前

 // scheme, and lychee (weekly cron, continue-on-error) gates nothing.

改后

 // scheme, and lychee (`schedule` + `workflow_dispatch`, no PR trigger) blocks
// nobody.

改动二:删掉重复粘贴的 4 行(#3648)

judgeHref() 里 "A URL on this site is an internal route wearing an origin…" 那 4 行出现两遍、一字不差,删掉后一份。

先预测后验证

预测改前实测改后实测结果
grep -rn 'continue-on-error' scripts/ 归零20✅
grep -rn 'weekly cron' scripts/ 归零20✅
重复注释特征行 2 → 121✅
node scripts/check-doc-links.mjs 前后同输出Links are valid across 7 scan roots. exit 0同上,exit 0✅ 行为零变化
vitest 前后同绿同数82 passed82 passed(连 check-links-workflow.test.ts 共 89 passed / 2 files)✅
pnpm type-check:scripts 绿exit 0exit 0✅
pnpm check:control-bytes 绿—OK (scanned 3661 tracked text file(s)),另加 grep -naP 自扫两文件干净✅

一处没按预测走、如实记下:草稿里我曾在新注释中写 "carries no continue-on-error" 来正面否认旧说法,那会让 grep 'continue-on-error' 停在 1 而不是 0。改用 "not a soft-failing job" 表达同一个意思,既保留了防误读的价值,又让 grep 真正归零。

围栏

  • content/docs/guide/ci-cd-pipeline.md未动——它对该工作流的描述本身正确(issue 已确认)。
  • .github/workflows/check-links.yml未动——本单只修注释,不改门禁行为。
  • 无 changeset:scripts/ 不是发布包,且改动纯注释、零用户可见变化。
  • 与在途 i18n 门禁:en 文案变更时其余九包必须同批跟改(或显式挂账)——#3582/#3625 族缺陷缺的那道不变量 #3650(在 scripts/ 新增 en-drift 门禁文件)实测零相交。
  • 仓内 prettier 说明:pnpm exec prettier --check 对这两文件报 warn,但改前的 origin/main 版本同样报(默认配置下 389 / 1646 行差异),仓内既无 prettier 配置也无任何工作流跑 prettier——非本 PR 引入,不处理。

🤖 Generated with Claude Code

https://claude.ai/code/session_01GTRjn8xBqp75dk7kFupVRt

Generated by Claude Code

@claude
docs(scripts): 按真实机制改写两处 Lychee 门禁描述,并删掉 judgeHref 重复注释 (#3587) (#3648)
9ffae06
两处注释称 lychee 是「weekly cron with continue-on-error」,结论「gates
nothing」对,机制反了:check-links.yml 里没有任何 continue-on-error,第 77 行
是 fail: true,`on:` 只有 workflow_dispatch + schedule('17 4 * * 0'),push 与
pull_request 被注释掉并附 ⛔ Do NOT enable(#3213 ruling B)。按错误理由行事的
人会去摘一行不存在的 continue-on-error,而真实风险相反 —— 取消注释
pull_request: 会让它立刻变成走网络的硬门禁。
改成同一文件 :194 起已有的正确表述(#3589 头注释措辞):schedule +
workflow_dispatch、无 PR 触发器,所以它谁也拦不住。
同时删掉 judgeHref() 里被逐字复制两遍的那 4 行注释(#3629 合并时的重复粘贴)。
纯注释改动,零行为变化;test 那处只改说明文字,断言未动。scripts/ 非发布包且
无用户可见变化,故无 changeset。
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GTRjn8xBqp75dk7kFupVRt
@vercel

vercelBot commented Aug 7, 2026

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment
ProjectDeploymentActionsUpdated (UTC)
objectuiIgnoredIgnoredAug 7, 2026 5:33pm

Request Review

This was referenced Aug 7, 2026
Closed
Closed
@yinlianghui
yinlianghui marked this pull request as ready for review August 7, 2026 17:42
@yinlianghuiyinlianghui mentioned this pull request Aug 7, 2026
Closed
yinlianghui pushed a commit that referenced this pull request Aug 7, 2026
@claude
docs(app-shell): rewrite three docblocks this branch falsified (#3666)
bd04651
The two sidebars' `sys-datasources` entry now names the canonical
`…/metadata/datasource` route, so three comments claiming the setup left-nav
points at the legacy alias became false as of the previous commit.
Rewritten as positive statements of what is true now. Per #3656, none of them
re-plants the alias URL in a denial sentence — a comment that says "no longer
points at X" keeps a grep for X returning the very hit the rewrite was meant
to clear.
`console/AppContent.tsx` is a comment-only change: the two alias route
declarations, and every other line of code, are untouched. Its #3610 history
is preserved and extended rather than replaced — #3610 declined to re-point
the navigation because a zero-app-only spelling would have given the alias a
second canonical destination, and #3660 re-pointed it at the shared route, so
that reasoning still holds.
Fixes#3666
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GTRjn8xBqp75dk7kFupVRt
This was referenced Aug 7, 2026
Merged
Closed
@yinlianghui
yinlianghui added this pull request to the merge queueAug 7, 2026
Merged via the queue into main with commit f4b8288Aug 7, 2026
17 checks passed
@yinlianghui
yinlianghui deleted the claude/issue-3587-lychee-comment-truth branch August 7, 2026 18:54
This was referenced Aug 7, 2026
Closed
Merged
Closed
Merged
Closed
Merged
Closed
Closed
Merged
Closed
Merged
Closed
Merged
Merged
Merged
github-merge-queueBot pushed a commit that referenced this pull request Aug 8, 2026
@yinlianghui@claude
docs(app-shell): 两个 AppContent 路由测试的生产端叙述改写为「历史 + 现状」两段 (#3749) (#3860)
7fbef94
* docs(app-shell): 两个 AppContent 路由测试的生产端叙述改写为「历史 + 现状」两段 (#3749)
三处注释以现在时枚举「侧边栏 / QuickActions 现在发哪些 URL」,而这些 URL 已被
#3660(`sys-datasources`)与 #3739(`sys-objects` + 首页 Manage Objects 卡片)改指
metadata-admin 引擎的规范路由。断言全部正确且全绿 —— 陈旧的只有叙述,以及两个以
生产端命名的 `it` 标题:它们实际量的是别名路由仍能解析且只跳一次。
手法照 #3666:每处拆成「#3610 当时如此」(过去时)与「#3660/#3739 之后如此」
(现在时)两段,现状段正面陈述两条别名今天的身份 —— 不再是任何导航的目标,而是
书签与外部链接的到达路径,这正是重定向必须继续工作的理由。按 #3656,现状段不靠
「不再指向 X」的否认句复述别名 URL。
- `AppContent.noAppComponentRoutes.test.tsx`:头注拆段;两个 `it` 标题从
`sys-datasources:` / `sys-objects:` 改为 `shell alias:` / `host alias:`,即两条
别名各自的改写者(shell 自己的 `LegacyMetadataRedirect` vs 宿主的
`MetadataRedirect`)—— 文件正文本来就画了这条区分。用例体内两处重申同一陈旧
断言的行内注释同步(`sys-datasources` item still points straight at this
spelling / this case now measures what production actually does),否则改完标题
的文件会自相矛盾。顺带把同一段落里 `isMetadataRoute` 的现在时「substring test」
收敛为过去时 —— #3638 起它是 `pathSegments.includes('metadata')` 段测试。
- `AppContent.pseudoRouteSegments.test.tsx`:生产端表格按实读重写(逐条读自
`AppSidebar` / `UnifiedSidebar` / `QuickActions` / `HomePage` / `InboxPopover`),
并补一段说明两条旧拼写并未消失,只是从 `navigation` 行移到了上面两行的到达面。
`metadata` 两种拼写下都是完整路径段,所以本文件要证伪的论断不受影响。
- `AppContent.tsx`(#3610 那段):收尾句补上 #3739 也已重指,并注明 `sys-objects`
的链路早在 #3658 就离开了这两条路由。该段前半的过去时叙述保留。
零行为改动:两个测试文件跑前跑后同为 43 passed,断言与用例数未动。
无 changeset —— 注释-only,照 #3666(bd04651)与 #3785(f9d70a7)先例。
Fixes#3749
* chore(changeset): 为 #3749 注释改写声明「不发布任何东西」(空 frontmatter)
CI 的 Changeset Declaration 门(scripts/check-changeset-presence.mjs)守的是
「fixed 组内任一包的 `<pkg>/src/**` 被改动 -> 必须有 `.changeset/*.md` 声明」,
本次三个改动文件全部落在 packages/app-shell/src 下,所以门要求声明。
先前照 #3666(bd04651)与 #3785(f9d70a7)判断「注释-only 不带 changeset」是
读错了先例:那两个 PR 早于这道门(#3387 引入),且 #3785 只动 content/docs/**,
根本不在守护面内。门自己写明了正确出口 —— 「If this change really should release
nothing, say so — that is a pass, not a workaround」,即空 frontmatter,照
.changeset/registry-inputs-spec-parity-gate-3797.md 的先例写法。
空 frontmatter 而非 patch:AppContent.tsx 只改 JSX 注释,两条
LegacyMetadataRedirect 路由声明与其余每一行代码未动;两个测试文件断言与用例数
逐字节未变(前后同为 43 passed)。确无可发布的行为改动。
本地验证:
node scripts/check-changeset-presence.mjs -> exit 0(识别为空 frontmatter 豁免)
node scripts/check-changeset-no-major.mjs -> exit 0
---------
Co-authored-by: Claude <noreply@anthropic.com>
akarma-synetal pushed a commit to akarma-synetal/objectui that referenced this pull request Aug 10, 2026
@yinlianghui@claude
fix(nav): 导航侧最后 4 处生产端直发规范 metadata 路由,不再绕已废弃的 component/metadata 别名 (o…
d2fd044
…bjectstack-ai#3660) (objectstack-ai#3668)
* fix(nav): the last four producers emit canonical metadata routes, not the deprecated alias (objectstack-ai#3660)
The System hub's "Metadata" and "Datasources" cards, and the `sys-datasources`
entry in both AppSidebar and UnifiedSidebar, aimed at
`component/metadata/{directory,resource?type=datasource}`. app-shell declares
those spellings as legacy aliases whose route element is
`LegacyMetadataRedirect` — a bare Navigate onto `metadata` and
`metadata/datasource`. Each click paid a redundant hop plus a re-render.
All four now name the destination directly. Endpoints are byte-identical to
what the alias hop computed; only the intermediate hop is gone. The alias
routes are untouched and stay reachable for bookmarks and external links.
Completes objectstack-ai#3639, which fixed the console host's two redirects and enumerated
these four as the remainder.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GTRjn8xBqp75dk7kFupVRt
* docs(app-shell): rewrite three docblocks this branch falsified (objectstack-ai#3666)
The two sidebars' `sys-datasources` entry now names the canonical
`…/metadata/datasource` route, so three comments claiming the setup left-nav
points at the legacy alias became false as of the previous commit.
Rewritten as positive statements of what is true now. Per objectstack-ai#3656, none of them
re-plants the alias URL in a denial sentence — a comment that says "no longer
points at X" keeps a grep for X returning the very hit the rewrite was meant
to clear.
`console/AppContent.tsx` is a comment-only change: the two alias route
declarations, and every other line of code, are untouched. Its objectstack-ai#3610 history
is preserved and extended rather than replaced — objectstack-ai#3610 declined to re-point
the navigation because a zero-app-only spelling would have given the alias a
second canonical destination, and objectstack-ai#3660 re-pointed it at the shared route, so
that reasoning still holds.
Fixesobjectstack-ai#3666
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GTRjn8xBqp75dk7kFupVRt
---------
Co-authored-by: Claude <noreply@anthropic.com>
Sign up for freeto join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

tests

Projects

None yet

Milestone

No milestone

2 participants

@yinlianghui@claude