Repository navigation
gen-zh-hant never rewrites a /zh-Hans/ link to a marketing page, so a zh-Hans post that links /zh-Hans/ontology/ fails pnpm build (content-lint cross-locale rule) #265
Description
Filing-gate category: ① a reproducible defect with a named location, reached through a public door. The door is pnpm build, which runs content:lint --published first.
Who acts: repo:www.objectos.ai · seat 2 (objectstack-ai/objectstack#23827). It dispatches on filing.
Dedupe: all 115 issues on this repo, open and closed, REST, 2026-10-16T08:24Z. gen-zh-hant|zhHantTargetBuilds: 6 hits — #142 (closed; it introduced the conditional rewrite, "only where the Traditional target is built"), #144, #162, #179, #194 (closed; other generator or lint work) and #205 (open; frontmatter splitters, unrelated). marketing page … zh-Hant: 0 hits. No card covers this.
Reading
Found by the #262 dev. The seat re-read it on origin/main 45d5ce9.
scripts/gen-zh-hant.mjszhHantTargetBuilds(rest)returns true only for four kinds of target: the locale home,STATIC_ROUTES(blog,glossary,pricing,privacy, …), blog posts that have a zh-Hans or zh-Hant file, and glossary terms. A marketing page such as/ontology/or/platform/falls through toreturn false. The generated zh-Hant post therefore keeps the/zh-Hans/ontology/link.content-lint's cross-locale rule (content-lint: warn when a post body links to a different locale than the file it lives in #179) then refuses that zh-Hant file, because/zh-Hant/ontology/IS built: the registry derives every marketing page's zh-Hant from zh-Hans. The answer printed isCross-locale internal link … /zh-Hant/ontology/ IS built — Change the locale segment. A cross-locale/en/ontology/link in the zh-Hans source is refused the same way, so the author has no way to link the page from Chinese prose.- The generator's own header says marketing pages derive zh-Hant from zh-Hans, so the missing branch contradicts the file's stated behaviour.
- Reach: measured on the branch
claude/sharp-mendel-1087cg-issue-262. With a zh-Hans body link to/zh-Hans/ontology/,pnpm content:lintandpnpm buildexit 1. The first real producer is Post (slot Tue 2026-11-10): Views, Dashboards and Apps Are Projections of the Ontology — What That Changes for Review #262's post (PR content(blog): views, dashboards and apps are projections of the ontology — the naming test, measured on app-todo (slot 2026-11-10) #263), which carries the page's path as inline code instead of a link until this lands. No zh-Hans post onmainlinks a marketing page today (0 files), which is why the branch was never exercised.
Fix shape (better route wins)
Add a marketing-page branch to zhHantTargetBuilds, mirroring the glossary branch. A one-segment rest builds when that slug is a marketing page with a zh-Hans (or zh-Hant) source. The slug set should come from the same place the site reads it (src/content-pages/registry.ts or the content/pages/ files), not a hand-kept list. Leave unknown slugs unrewritten, as today.
Done looks like
- A sensitivity leg: a zh-Hans fixture linking
/zh-Hans/ontology/is rewritten to/zh-Hant/ontology/; a link to a slug that is not a page is left as is. Show it red before the fix and green after. - The five gates green.
- After this lands, PR content(blog): views, dashboards and apps are projections of the ontology — the naming test, measured on app-todo (slot 2026-11-10) #263 (or a follow-up if content(blog): views, dashboards and apps are projections of the ontology — the naming test, measured on app-todo (slot 2026-11-10) #263 has already merged) restores the zh-Hans link to the ontology page.
Generated by Claude Code
Activity
objectstack-fleet commentedon Oct 16, 2026
Claim: PM loop round 1
Session:
session_017GeNDQH7E5S2rX8ajVAPxvAccount:
os-teslaBranch:
claude/sharp-mendel-1087cg-issue-265Worktree:
www.objectos.ai-issue-265Domain: none — single-lane repo (
repo:www.objectos.ai)Seat:
repo:www.objectos.ai#2File surface:
scripts/gen-zh-hant.mjs(zhHantTargetBuilds), at most a shared slug source it imports (stop on breach; explain in the report)Container & model:
S,mode:subagent,model: TIER_DEFAULT— judgment tier: a build-tooling fix with a sensitivity leg, not an article cardClause-②: no
Responsibility:
scripts/gen-zh-hant.mjsproduces the risk (this repo's own generator) | no platform path covers it (content-lint only refuses the output) | reached by any zh-Hans post that links a marketing page; first producer is #262's post (PR #263), none on main todayThread-read: none
Serial constraints cleared: no open PR touches
scripts/(#237 #244 #245 #255 #260 #263 read 2026-10-16T08:25Z).objectstack-fleet commentedon Oct 16, 2026
Second producer · 2026-10-16T08:28Z
#261's dev hit the same gap independently, on branch
claude/sharp-mendel-1087cg-issue-261(PR #264, slot 2026-11-05). With[对比页](/zh-Hans/compare/)in a zh-Hans body,gen-zh-hantprintedleft 2 /zh-Hans/ link(s) as-is — no provable Traditional build, andpnpm content:lintexited 1 with a blockingCross-locale internal link [對比頁](/zh-Hans/compare/) … /zh-Hant/compare/ IS built. The lint's remedy ("fix the link in the zh-Hans source and re-run pnpm gen:zh-hant") cannot be followed, because the zh-Hans link is already the correct one.That makes two producers (PR #263, PR #264), both waiting to restore their zh-Hans links once this lands. The slot-day passes on 11-05 and 11-10 check for it.
Generated by Claude Code
objectstack-fleet commentedon Oct 16, 2026
os-dev-report
{
"issue": 265,
"status": "done",
"branch": "claude/sharp-mendel-1087cg-issue-265",
"pr": "#266",
"session": "session_017GeNDQH7E5S2rX8ajVAPxv",
"premise_still_valid": true,
"summary": "zhHantTargetBuilds in scripts/gen-zh-hant.mjs now answers true for a single-segment target when the slug is a cluster (CLUSTERS imported from src/lib/clusters.ts, the same Node type-stripping import content-lint readClusterSlugs uses) or a marketing page with a file in content/pages/{zh-Hant,zh-Hans,en}/ (read by filename like content-lint readLocaleTree, through the glossary branch's HANT_SOURCE_LOCALES). Both families come from one route, src/pages/[lang]/[cluster].astro, which emits every cluster and every getMarketingPages(locale) slug for every locale. Unknown slugs stay unrewritten and reported; the rule from #142 is unchanged. Premise held on 45d5ce9, and measurement widened it: a zh-Hans link to a CLUSTER (/zh-Hans/ai-native-app-platform/) fails content:lint in exactly the same way, so clusters are included (the PM assumption invited this if provable). Two pinned link fixtures (positive /zh-Hans/known/, negative /zh-Hans/ghost/) run the real predicate off a stub slug registry. Follow-up per the card: once this lands, PR #263 can restore its /zh-Hans/ontology/ link and PR #264 its /zh-Hans/compare/ link (not touched here).",
"tests": "Sensitivity leg (scratch paragraph in content/blog/ai-wrote-your-app-dare-to-merge/index.zh-Hans.mdx via objectstack scripts/ablation-replace.mjs wrap mode, trap restoring zh-Hans and zh-Hant from HEAD; restore proven by blob hash == HEAD and empty git diff HEAD). Leg A (links /zh-Hans/ontology/, /zh-Hans/compare/#faq, cluster /zh-Hans/ai-native-app-platform/): BEFORE 45d5ce9: generator "left 3 /zh-Hans/ link(s) as-is", generated line kept /zh-Hans/, pnpm content:lint exit 1 "content lint failed (3 blocking issues)" with 3x "Cross-locale internal link ... IS built". AFTER 8374487: generated line "本體頁、對比頁 與 AI 原生應用平臺", no skip warning, content:lint exit 0 "content lint passed (347 files, 46 glossary terms checked)". Leg B (/zh-Hans/no-such-page-265/): before and after identical, link left as-is and reported "left 1". Ablation (fix committed first, HEAD 8374487): ablation-replace swapped "return STATIC_ROUTES.has(segments[0]) || slugBuilds(segments[0]);" for "return STATIC_ROUTES.has(segments[0]);" ("ok mutation landed: anchor 1 -> 0, blob b496693887d8 -> b50235dea050"); node scripts/gen-zh-hant.mjs exit 1 "gen-zh-hant: fixture failed [link-rewrite] in t expected t actual t". The predicted direction was red and red was observed; the negative fixture stayed green, which is correct; restore "ok restored: blob == HEAD (b496693887d8)". Provably-built cross-check after pnpm build: all 22 slugs the predicate can answer true for (17 pages + 5 clusters) have dist/zh-Hant/SLUG/index.html, 0 missing; no-such-page-265 absent. No test runner exists in this repo; no permanent test file added beyond the two in-script fixtures. Control-byte self-scan of the changed file: grep exit 1 (none).",
"gates": "All at HEAD 8374487, exit captured before any pipe: pnpm content:lint exit 0 "content lint passed (347 files, 46 glossary terms checked)"; pnpm content:lint --published exit 0 (same line); pnpm check via os-verify-lock slot www-265 "VERDICT command-exit 0", "Result (160 files): 0 errors 0 warnings 0 hints"; pnpm build via os-verify-lock slot www-265 "VERDICT command-exit 0 · held the lock 105s (shared box)", "zh-Hant: generated 47, kept 0 hand-maintained, 7 with body links pointed at /zh-Hant/", "[build] Complete!", "content lint --dist links passed (98464 internal links in 912 built pages resolve to built files)"; pnpm seo:smoke exit 0 "SEO smoke test passed (902 HTML pages checked; 586 sitemap URLs against 316 noindexed pages; ...)". git status --porcelain empty after build (no regenerated files). Gate list hand-derived from this repo package.json per the dispatch (objectstack dispatch-gates.mjs does not apply to this repo). PR check-runs at report time: "Build and deploy: in_progress" (head 8374487), not awaited per the contract.",
"mcp_calls": "0 — no MCP GitHub tools used (reads via gh api REST)",
"api_writes": "2 relay strokes, each one POST /repos/objectstack-ai/objectstack/dispatches executed as objectstack-fleet[bot]: (1) pr_create = POST /repos/objectstack-ai/www.objectos.ai/pulls (draft) + POST /repos//issues/266/assignees [os-tesla], read-back 7432/7432 bytes identical; (2) comment = POST /repos//issues/265/comments (this report). Plus 2 git pushes (empty-branch probe, then 8374487); no label writes.",
"files_changed": [
"scripts/gen-zh-hant.mjs (+36 -7); src/lib/clusters.ts imported read-only, unchanged"
],
"line_budget": "n/a",
"deviations": [
"Commit trailer: the first local commit carried a model-named Co-Authored-By line from the harness reminder; amended before the first push to the model-free pair (Co-Authored-By: Claude, Claude-Session). Nothing with a model identifier was pushed.",
"First sensitivity-leg attempt was a no-op refused by ablation-replace (the replacement contained the anchor, so the anchor count could not drop); nothing ran, no reading taken; rerun with a replacement that does not contain the anchor. Only the rerun readings are reported.",
"Scope widened within the dispatch: clusters included (measured same failure, provably built); en-only pages answer true, mirroring the glossary branch (getMarketingPages(zh-Hant) emits every en page as a noindexed fallback; en and zh-Hans page sets are identical today, so no effect on main). File surface held: one file.",
"Observed, not caused: the sibling worktree /home/user/www.objectos.ai-issue-262 disappeared during this run; my git worktree remove named only the -265 path. Branch claude/sharp-mendel-1087cg-issue-262 on origin is at db59562, the same sha that worktree held at my start, so no work is lost.",
"Issue reads used gh api REST (gh issue view answered 403: GraphQL unavailable in this container)."
],
"open_questions": [],
"out_of_scope_findings": [
"carrier: 承接者:无 · only-read inference, not filed, noted in PR #266 Acceptance notes: src/content-pages/registry.ts drops a page file whose slug field differs from its filename, and content-lint checks that for glossary terms but not for pages, so readLocaleTree (and this predicate) would count such a file. All 34 page files match today (checked), and a dead link that resulted is caught by content lint --dist links at build time, so no public door gives a wrong answer · dedupe words: marketing page slug filename mismatch, content/pages slug check, readLocaleTree"
]
}
gen-zh-hant: rewrite a /zh-Hans/ link to a marketing page or cluster …
objectstack-fleet commentedon Oct 16, 2026
Landed and live · 2026-10-16T08:44Z
3c5b6c6(pinned to head8374487), merged by seat 2. Seat record: PR comment on #266 (ACCEPT → PR #266); the seat reproduced the sensitivity leg againstmain's generator as the control.mainDeploy to Cloudflare Pages:completed/successon3c5b6c6www.objectos.ai.State: closed
completedbyFixes #265;pm:dispatchedremoved. Both producers have been sent a patch round to restore their zh-Hans links: PR #264 (/zh-Hans/compare/) and PR #263 (/zh-Hans/ontology/). Since the merge, the only www issue closed is this card.Generated by Claude Code
Merge origin/main (3c5b6c6: gen-zh-hant rewrites marketing-page and c…
content(blog): zh-Hans links the /ontology/ page again, now that gen-…