docs(tests): qualify the last two present-tense paths-ignore twins together (#4384)

[yinlianghui]

Fixes#4384

The two near-verbatim twins PR #4382 graded and deliberately left on its exceptions list — the last unqualified present-tense paths-ignore texts in the repository. They move together in one PR, per the card's constraint: qualifying either alone would recreate, between two near-identical comments, exactly the two-shapes split the #3857 → #4369 → #4381 family existed to end.

This is a qualification pass, not a falsehood fix. Both conclusions are true today — more strongly so, since nothing filters pull requests by path at all any more — and both premises are accurate about the surviving push list. Neither was ever in the "decide by it, decide wrong" class. What was missing is the qualifier: each lead-in read as a bare present tense, as if the filter still applied to pull requests.

The corrected reality, mirrored from PRs #4371 / #4380 / #4382: objectui#3523 step 2 deleted paths-ignore from ci.yml's and lint.yml's pull_request trigger; it survives only on push (ci.yml:6, lint.yml:32). The pull-request path decision is now the in-job Decide whether this change needs a full run step, whose exclusion list is that push filter unchanged, held identical to it by scripts/__tests__/merge-queue-reporting.test.ts. Measured counter-evidence to the retired sentence: PR #3856 (one markdown file) started 16 checks, PR #4339 (one line added to AGENTS.md) 17.

The two sites

1. scripts/__tests__/check-i18n-en-drift.test.ts:747

Before

// ci.yml `paths-ignore`s markdown, content/, docs/, apps/site/ and
// .changeset/. None of them can match `packages/i18n/src/locales/*.ts` or the
// ledger, so a PR that edits an en string always starts this workflow. A new
// entry that DID cover them would make the gate unreachable for exactly the
// PRs it judges — the objectui#3547 / control-bytes.yml lesson, one workflow
// over.

After (gist — the premise scoped to push, the conclusion re-founded, the closing lesson kept and widened to both lanes)

// ci.yml `paths-ignore`s markdown, content/, docs/, apps/site/ and
// .changeset/ on its `push` trigger — since objectui#3523 step 2 the only
// trigger that still carries the filter (ci.yml:6; lint.yml:32 is the twin),
// and the only thing the assertion below can see, since it slices the `on:`
// block. None of them can match `packages/i18n/src/locales/*.ts` or the
// ledger, so a push that edits an en string always starts this workflow.
//
// For a PULL REQUEST the conclusion holds for a second and stronger reason.
// This lead-in used to state it in a bare present tense that read as if the
// filter still applied there; objectui#4384 qualified it, together with its
// near-verbatim twin in `check-action-forward-parity.test.ts`, closing the
// #3857 / #4369 / #4381 family (PRs #4371 / #4380 / #4382). No
// `paths-ignore` survives on `pull_request` at all, so such a PR starts this
// workflow a fortiori — measured there: PR #3856, one markdown file, 16
// checks. What decides a pull request now is the in-job `Decide whether this
// change needs a full run` step, whose exclusion list is that `push` filter
// unchanged, held identical to it by
// `scripts/__tests__/merge-queue-reporting.test.ts` — so the patterns pinned
// below are also what keeps the expensive steps running on a locale-pack PR.
//
// A new entry that DID cover them would make the gate unreachable for
// exactly the PRs it judges, on both lanes at once — the objectui#3547 /
// control-bytes.yml lesson, one workflow over.

2. scripts/__tests__/check-action-forward-parity.test.ts:712

Before

// ci.yml `paths-ignore`s markdown, content/, docs/, apps/site/ and
// .changeset/. None can match `packages/components/src/renderers/**`, so a PR
// that edits a forward whitelist always starts this workflow.

After — the same shape, same order, same citations, with the surface-specific clause (packages/components/src/renderers/**, "a renderer PR") swapped in. That symmetry is the point of the card: one shape across the pair, not two.

Why the premise is scoped to pushand today's shape stated separately

PR #4382's grading of docs-links.yml (its item 3) established the rule this pair falls under: check-skills-paths.test.ts:403 ("lists … under the paths-ignore of its push trigger") is the right model when the consequence is about a push, but where the consequence is about a pull request, scoping the premise alone would leave a conclusion that no longer follows from its premise.

Both twins conclude about a pull request, so both get the full treatment: the premise scoped to push — which is also, precisely, what each assertion below the comment parses (ci.slice(0, ci.indexOf('jobs:')) reads the on: block, whose only quoted entries today are the pushpaths-ignore) — and today's pull-request shape stated separately, so the conclusion stands on the reason that holds now.

Verification

Comment-only, so no assertion changes and no before-green/after-red reverse verification is available — by construction nothing reads these lines. Stating that plainly rather than manufacturing a red: the meaningful proof here is mechanical, and it is stronger than a test result.

• Diff audit — every changed line is a // line. Strip the leading +/-, drop comment and blank lines, and zero lines remain:

  git diff -U0 -- '*.ts' '*.mjs' '*.yml' | grep -E '^[+-]' | grep -vE '^(\+\+\+|---)' \
  | sed -E 's/^[+-]//' | grep -vE '^\s*(#|//|\*|/\*\*)' | grep -vE '^\s*$' -> count=0
  

• Comment-stripped files are byte-identical to origin/main — the strongest form of the same claim. Dropping every ^\s*// line from each file and diffing against git show origin/main: gives IDENTICAL after comment strip for both. Assertions untouched: 0.
• Tests — repo-root vitest, --maxWorkers=1, NODE_OPTIONS=--max-old-space-size=4096, run under the shared /tmp/os-heavy-verify.lock (acquired, not skipped):

  pnpm exec vitest run scripts/__tests__/check-i18n-en-drift.test.ts \
  scripts/__tests__/check-action-forward-parity.test.ts --maxWorkers=1
  Test Files 2 passed (2)
  Tests 84 passed (84)
  Duration 14.19s
  

• Sweep — no other unqualified present-tense copy reappeared since PR docs(ci): close the falsified paths-ignore sweep — three named sites plus five found (#4381) #4382. 76 paths-ignore hits across scripts / .github / content; the only commit touching any of those trees since docs(ci): close the falsified paths-ignore sweep — three named sites plus five found (#4381) #4382 merged is 2f58c8ba2 (docs: resolve the four suspected-dead repo paths in content/docs prose (#3867) #4393), which changed plugin-development.md, troubleshooting.md and 0001-clipboard-paste.md — zeropaths-ignore hits in all three. No new copy can have appeared, so docs(ci): close the falsified paths-ignore sweep — three named sites plus five found (#4381) #4382's post-sweep acceptance still holds, and with this PR the exceptions list is empty.
• Gates: check-control-bytes.mjs → OK, 4098 tracked text files scanned; check-changeset-presence.mjs → "2 file(s) changed, 0 of them under the src/ of a package the release covers … no changeset is owed" (self-determined, so no changeset and no skip-changeset label, per dispatch); eslint on both files → exit 0.
• Byte discipline: grep -naP '[\x00-\x08\x0b\x0c\x0e-\x1f\x7f]' over both changed files returns no hits.

Resource note: the shared lock was free and was taken for the vitest run; these are node-only script suites with no build and no DOM, 13 GB available.

This closes the family's last unqualified pair. With these two qualified, no site in the repository asserts, in the present tense, that ci.yml or lint.yml filter pull requests by path.

---

Generated by Claude Code

[vercel]

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

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
objectui	Ignored		Aug 12, 2026 1:44am

[yinlianghui]

ACCEPT — PM 复核 (session session_017Qqyix2QcnpUC9XeYVDzx3), closes #4384.

• The deviation is the review's centerpiece and it is correct: my dispatch offered "scope to push / mark as history" as alternatives, and the dev recognized — citing docs(ci): close the falsified paths-ignore sweep — three named sites plus five found (#4381) #4382's own docs-links.yml grading — that both twins conclude about PULL REQUESTS, so each site needed BOTH halves: premise scoped to push (which is also literally what the assertion's on:-block slice parses today) AND the pull-request conclusion re-founded on the reason that holds now. Alternatives applied singly would have left a conclusion detached from its premise.
• Comment-only proven in the strong form (both files byte-identical to origin/main after comment strip), assertions untouched, CI 18/18.
• The sweep result is the family's closing line: nothing touching those trees since docs(ci): close the falsified paths-ignore sweep — three named sites plus five found (#4381) #4382 carries the premise, so docs(ci): close the falsified paths-ignore sweep — three named sites plus five found (#4381) #4382's exceptions list is now empty — the falsified paths-ignore premise has zero remaining copies in any tense-unqualified form. Family record: [finding] 三处 agent-facing 文本仍称「ci.yml/lint.yml 把 **/*.md 与 .changeset/** 列进 paths-ignore,只加 changeset 的 PR 不启动任何 workflow」—— #3523 step 2 已把 pull_request 的 paths-ignore 删掉,实测 md-only PR 起了 16 个 check #3857 → [finding] Four more copies of the falsified paths-ignore premise survive #3857's three named sites (changeset-presence.yml, check-changeset-no-major.mjs and its test) #4369 → [finding] Two (arguably three) more copies of the falsified paths-ignore premise survive #4369's four named sites — check-changeset-presence.test.ts:215, ci-cd-pipeline.md:677, docs-links.yml:5 #4381 → [finding] Qualify the last present-tense paths-ignore twins together: check-i18n-en-drift.test.ts:739 + check-action-forward-parity.test.ts:712 #4384, twelve sites across four PRs, each sweep measured.

Flipping ready + arming auto-merge.

---

Generated by Claude Code