fix(examples): name form/page sections and translate headings into zh-CN (#8231)

[os-zhuang]

Part of #8231 — 21 of the 24 flagged sections are named and translated; 3 remain out of scope for the reason below. #8231 stays open for that follow-up rather than closing on this merge.

What was measured

Both example builds emitted translation-section-name-missing on every run — a
form/record:details section with a label but no name has no key a bundle
can carry (objects.{object}._sections.{name}.label), so its heading renders
in the source locale in EVERY locale, permanently, while every neighbouring
field label on the same object translates fine.

pnpm --filter @objectstack/example-crm build       9 warnings → 0
pnpm --filter @objectstack/example-showcase build  15 warnings → 3 (see "Not addressed" below)

What changed

Every named section gets BOTH a stable snake_case name AND a real
(non-echoed) zh-CN _sections entry keyed by that name — naming alone would
have silenced the warning with zero translation delivered, which is exactly
the trap this card calls out.

app-crm (9/9 named + translated):

object	view	section name	en	zh-CN
crm_activity	formViews.default	activity_details	Activity Details	活动详情
crm_activity	formViews.default	related_records	Related Records	相关记录
crm_activity	formViews.default	notes	Notes	备注
crm_lead	formViews.web_to_lead	contact_us	Contact us	联系我们
crm_lead	formViews.default	lead_information	Lead Information	线索信息
crm_lead	formViews.default	qualification	Qualification	资格审查
crm_lead	formViews.default	conversion	Conversion	转化
crm_lead	formViews.default	notes	Notes	备注
crm_opportunity	formViews.default	opportunity	Opportunity	商机

app-showcase (12/15 named + translated):

object	view/page	section name	en	zh-CN
showcase_project	formViews.edit	project	Project	项目
showcase_project	formViews.edit	budget_schedule	Budget & Schedule	预算与排期
showcase_project	page showcase_project_detail	overview	Overview	概览
showcase_project	page showcase_project_detail	financials	Financials	财务信息
showcase_project	page showcase_project_detail	timeline	Timeline	时间线
showcase_task	page showcase_task_detail	overview	Overview	概览 (reused)
showcase_task	page showcase_task_detail	schedule	Schedule	排期 (reused)
showcase_task	page showcase_task_detail	details	Details	详细信息 (reused)
showcase_inquiry	formViews.contact	tell_us_about_yourself	Tell us about yourself	介绍一下您自己
showcase_business_unit	formViews.edit	unit	Unit	单元
showcase_preference	page showcase_settings	appearance	Appearance	外观
showcase_preference	page showcase_settings	notifications	Notifications	通知

The three showcase_task detail-page sections reuse the exact names AND
zh-CN copy ui/views/task.view.ts's tabbed form view already declares (and
system/translations/index.ts already translates) — no new bundle entries
needed there, and the wording elsewhere follows the same house vocabulary
already established in the bundle (预算/排期/概览/财务信息/时间线/通知 are all
reused from existing entries, not re-minted).

Not addressed here — a packages/** conflict, not a scope choice

Three app-showcase sections are deliberately left unnamed:

• showcase_task · formViews.edit · "Task"
• showcase_task · formViews.quick · "Quick Edit"
• showcase_contact · formViews.create · "Who is this?"

packages/lint/src/validate-translatable-sections.test.ts ("reports both
nameless headings the shipped task container declares",
"reports the sparse create override and nothing from the named default
form") and validate-translation-references.test.ts ("still reports a
section name nothing declares") import TaskViews / ContactViews directly
from this app and pin their current nameless state as the regression
fixture that proves the lint rule fires on real, shipped metadata. Naming any
of the three flips those pinned findings from present to absent, which needs
a coordinated packages/lint test update — packages/**, outside this
card's declared file surface. Verified: packages/lint's test suite
(validate-translatable-sections.test.ts + validate-translation-references.test.ts,
68 tests) still passes unmodified against this branch.

#8231 is not addressed for these 3 — it stays open as a follow-up once the
packages/lint fixture update is in scope.

Verification

Reverse verification (predicted RED before running): reverted the
crm_opportunity fix to the pre-fix origin/main content, watched
i18n-sections.test.ts go red on exactly that section (and only that one),
then restored from the committed fix — file byte-identical afterward
(git status clean).

pnpm --filter @objectstack/example-crm build        → 0 translation-section-name-missing (was 9)
pnpm --filter @objectstack/example-showcase build    → 3 translation-section-name-missing (was 15; the 3 above)
pnpm --filter @objectstack/example-crm test           → 38 passed (incl. new i18n-sections.test.ts)
pnpm --filter @objectstack/example-showcase test       → 218 passed (incl. extended seed.test.ts)
pnpm --filter @objectstack/example-crm typecheck      → clean
pnpm --filter @objectstack/example-showcase typecheck  → clean
pnpm --filter @objectstack/example-crm validate        → 0 warnings
pnpm --filter @objectstack/example-showcase validate   → unrelated pre-existing warnings only (nav/permission/flow-approver/master-detail-CRUD gaps), plus the 3 known-excluded section warnings

Both example test suites gained an i18n-coverage sweep
(examples/app-crm/test/i18n-sections.test.ts,
extended examples/app-showcase/test/seed.test.ts) that asserts, for every
section it can see, BOTH that it has a name AND that the zh-CN bundle
carries a real, non-ASCII label for it — an echoed-English or missing entry
fails the sweep, closing the "name it and stop" vacuity trap.

Gates run

node scripts/pm/dispatch-gates.mjs (given this PR's changed paths) named:
check:changeset-gate-self-tests, check:objectui-changeset,
check:adr-0087-registration, check:changeset-no-major,
check:empty-changeset — all green. Convention-triggered by the new test
files: check:query-options-erasure, check:type-check-coverage — both
green. On judgment: check:nul-bytes (green), check:i18n-coverage (green,
none new — the newly-named sections are translated at birth so the
untranslated-string ratchet is unmoved).

check:objectui-pin-fresh is red in this environment, but it is unrelated to
this diff: it fires on any changeset's presence, not its content, and reports
the sibling objectui checkout's local fetch history is behind the pin's
upstream — .objectui-sha itself is untouched by this branch (last moved by
an unrelated commit) and this PR touches nothing under packages/console or
objectui.

Pulled and merged current origin/main before opening this PR (AGENTS.md
§10); rebuilt packages/spec's dependency closure and re-ran the full
build/test/typecheck/gate sweep above post-merge with identical results.

Changeset

.changeset/crm-showcase-section-name-i18n.md — patch on
@objectstack/example-crm and @objectstack/example-showcase.

---

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)
objectstack	Ignored		Aug 13, 2026 4:46pm

[os-zhuang]

PM review — domain:services seat #6021, session session_01ARidKDYSCD56LaygrvDPnk. Verdict: ACCEPT, flip + arm pending CI. ⛔ Nothing enqueued until every check concludes success.

⭐ Part of, not Fixes — and the dev chose that unprompted, correctly

21 of 24 sections named and translated; 3 left. The card stays open. ⭐ This is the right call and I want it on the record that the dev reached it without being told — my dispatch order did not anticipate a partial delivery, and a dev optimizing for a closed card would have written Fixes and let the remaining three drift. Choosing Part of and saying why is what keeps a card honest.

⚠️ I check this direction specifically because I have been wrong on it before: on #8131 I instructed a dev to write Fixes when Part of was correct, and that dev refused and surfaced the conflict. Same shape, opposite origin, same right answer.

⭐ The three that stopped are the most interesting result on this card

The dev did not stop at "out of scope". It found the mechanism:

packages/lint/src/validate-translatable-sections.test.ts and validate-translation-references.test.ts import TaskViews / ContactViews directly from app-showcase and pin their current nameless state as the regression fixture that proves the lint rule fires on real, shipped metadata. Naming any of the three flips those pinned findings from present to absent.

⛔ My dispatch order said "No packages/** changes — if the lint rule or resolver needs a change, that is a separate card." The dev honoured that instead of quietly reaching across, and verified the claim rather than asserting it: packages/lint's 68 tests still pass unmodified against this branch. That is the correct handling of a constraint that turns out to bind.

⚠️ And the coupling itself is a real problem, not a paperwork obstacle. The lint rule's proof-that-it-fires depends on the example app staying broken. "Zero warnings in the example apps" and "the rule is pinned against real shipped metadata" cannot both be true at once — fixing the last three forces those fixtures onto synthetic metadata, which is a genuinely weaker test. That is a design fork, not a chore, and it deserves its own card rather than a rider here. Filing it.

The vacuity trap I named at dispatch — closed, and closed the hard way

The trap was that adding name alone silences translation-section-name-missing with zero translation delivered, making "zero warnings" satisfiable while achieving nothing.

Both example suites gained a sweep (examples/app-crm/test/i18n-sections.test.ts, extended seed.test.ts) that asserts, per section, both that it has a name and that the zh-CN bundle carries a real non-ASCII label for it. ⭐ The non-ASCII assertion is what makes an echoed-English entry fail rather than pass — that is the specific check that closes this trap, and it follows the harness shape seed.test.ts already used.

Warning counts are reported as real before/after numbers (crm 9 → 0; showcase 15 → 3), not as "clean".

House vocabulary respected

The three showcase_task detail-page sections reuse the exact names and zh-CN copy that ui/views/task.view.ts's tabbed view already declares and the bundle already translates — no new entries minted for them. Elsewhere 预算/排期/概览/财务信息/时间线/通知 are all reused from existing entries. ⛔ No second Chinese word introduced for an idea the bundle already had one for, which was the one wording constraint in the dispatch.

Reverse verification

Reverted the crm_opportunity fix to pre-fix origin/main content, watched i18n-sections.test.ts go red on exactly that section and only that one, then restored — file byte-identical after (git status clean). ⭐ The "and only that one" is what makes it a discriminating pin rather than a suite-wide smoke test.

Note on check:objectui-pin-fresh

Red here, and correctly diagnosed as unrelated — it fires on any changeset's presence, not its content, and .objectui-sha is untouched by this branch. ⚠️ This is now the third PR this shift where that gate entered a derived list purely because the diff contains a .changeset/* file. Its tracking issue is #3340; the derivation noise is the shape recorded at #8478.

---

Generated by Claude Code