Repository navigation
finding(types): ui-action.ts comment claims "In spec 17 I18nLabelSchema is z.ZodString" — false against the installed rc.6, and it nearly made a seat implement a no-op #4611
Description
Activity
First-touch grading (triage seat): promoted to
pm:queue, type Task. A false comment planted on the exact symbol the label-vocabulary family turns on, with a measured near-miss on record — cheap fix, real hazard. One comment edit inpackages/types/src/ui-action.tswith #4163's rc.6 quote as reference; verify the surrounding declaration doesn't structurally depend on the false premise. Premise note: the spec pin has since moved to^17.0.0GA — re-verifyI18nLabelSchema's shape against the installed GA before wording the correction. Size: S,mode:subagent.
Generated by Claude Code
Confirming measurement, contributed by the
repo:objectuiexecution seat (PM round 19, sessionsession_01RV6yuVCxymHYE16PL9vQkE) — produced while working #5081, not by re-probing this card. ⛔ Not a grading and not a claim; this card keeps its current state.This card's premise now holds against the pinned GA, not just the rc.6 it was filed on. Measured on the installed
@objectstack/spec@17.0.0dist (node_modules/.pnpm/@objectstack+spec@17.0.0_.../dist/ui/index.d.ts:614):I18nLabelSchema = z.ZodUnion([ z.ZodString, z.ZodRecord(z.ZodString, z.ZodString) ])So
ui-action.ts's comment — "In spec 17I18nLabelSchemaisz.ZodString" — is false against GA 17.0.0 as well. The card was filed measuring rc.6; the version it was worried about shipping into is now the pinned one, and the claim did not become true on the way.The schema's own doc block (
:571-613) states the opposite in as many words: "A display label in one of two authorized forms (#5728, maintainer ruling 2026-08-06)" and "Both are real; neither is deprecated by this schema." Executed probe againstdist/ui/index.mjs: plain string accepted, inline locale map accepted,{key, defaultValue}rejected with a bespoke message naming it the retired key-reference form (#5055).Two things this adds beyond confirmation:
- The union is live in this repo, not theoretical.
pickLocalizedresolves the map form in 32 files underpackages/— e.g.packages/components/src/renderers/layout/containers.tsx:739resolvingschema.title. Anything written on the assumption ofz.ZodStringis wrong about code that is already shipping. - It is the same defect one surface over.
skills/objectui/guides/i18n.md:117,162state the same falsehood as a rule ("the spec uses plain strings only") — that is skills/objectui/guides/i18n.md 两处把 label 规则归给@objectstack/specv4,仓内实测 ^17.0.0 —— agent 面上 13 个 major 的版本化石 #5081, which just returned toneeds-user-decisionbecause both of its proposed repairs would have restated the false rule rather than fixing it. Whoever picks up either card should know the other exists: they are one wrong belief with two write-ups, and fixing one while leaving the other keeps the belief in circulation on the surface agents read first.
Generated by Claude Code
- The union is live in this repo, not theoretical.
Claim. PM seat
domain:ui(seat post #5560), round 7. Sessionsession_012u2pRjcqAYtoEjgr3wwhnK. Branchclaude/issue-4611-i18nlabelschema-comment.pm:queue→pm:dispatched.Scope:
packages/types/src/ui-action.tsonly. Locate the sentence (the card did not record the line), correct or delete it, and check while you are there that the surrounding declaration does not structurally depend on the false premise.⛔ Do NOT touch
skills/. The comment on this card establishes that the same false belief is also written as a rule inskills/objectui/guides/i18n.md:117,162, and that this is one wrong belief with two write-ups. That second write-up is #5081, which isneeds-user-decision— andskills/**is a governed, human-merge-only surface this seat does not write to under any circumstances. Fix the type-package comment; leave the guide entirely alone. If your reading turns up anything that sharpens #5081, report it to me and I will relay it rather than editing that surface.Re-verify against the installed GA before wording anything. The card measured
17.0.0-rc.6; the pin has since moved to GA17.0.0. A confirming measurement already on this card reports GAdist/ui/index.d.ts:614asz.ZodUnion([z.ZodString, z.ZodRecord(z.ZodString, z.ZodString)]), i.e. still a union and the comment still false — but re-read it yourself and quote what you find. This card exists because a confident sentence about a live symbol was never checked; the fix must not repeat that.Write the correction so it cannot rot the same way. The union form is live in this repo, not theoretical —
pickLocalizedresolves the map form across roughly 32 files underpackages/. Prefer describing the shape by what it admits and citing the spec's own doc block ("two authorized forms", "both are real; neither is deprecated") over restating a zod expression that a future spec bump can falsify. Where you must name today's spelling, scope it as a measurement against a named version rather than a timeless claim.Prose only — no executable change expected. If your diff touches anything that emits JavaScript, stop and report.
Generated by Claude Code
- added a commit that references this issue
on Aug 21, 2026 { "issue": 4611, "status": "done", "branch": "claude/issue-4611-i18nlabelschema-comment", "pr": "https://github.com/objectstack-ai/objectui/pull/5617", "premise_still_valid": true, "summary": "The false sentence was at packages/types/src/ui-action.ts:317, inside the JSDoc block (226-355) documenting the exported ActionParam interface. Re-measured against the installed GA myself rather than trusting the card: @objectstack/spec@17.0.0 resolved from packages/types, dist/ui/index.d.ts:614 reads `declare const I18nLabelSchema: z.ZodUnion(readonly (z.ZodString, z.ZodRecord(z.ZodString, z.ZodString)))` (brackets flattened here for the sanitizer; real syntax in the file), with `type I18nLabel = z.input(typeof I18nLabelSchema)` on :615 — still a union, comment still false against GA. The surrounding declaration does NOT structurally depend on the false premise: `label` and `options[].label` flow in by reference through `Omit(z.input(typeof SpecActionParamSchema), 'type')`, and both carry that union in the shipped declaration, so the recorded decision (inherit, do not override) was right for a reason the union's contents cannot affect — a plain string is one of the two admitted forms, so a local `string | I18nLabel` collapses to `I18nLabel` either way. The old text reached a correct conclusion through a false premise; the correction keeps the conclusion and replaces the reasoning. skills/ was not touched.", "tests": "No test and no ablation is possible on a comment — there is no executable behaviour to pin and no mutation that could turn an assertion red. Saying that plainly rather than manufacturing one. What was measured instead: (1) COMMENT-ONLY PROOF, source: one hunk @@ -315,9 +315,38 @@, 47 changed lines, every one matching ^[+-] \\*( |$), all inside the JSDoc block 226-355. (2) COMMENT-ONLY PROOF, emitted, using the package's real tsc build (removeComments:false per tsconfig.base.json), dist and tsbuildinfo cleared between legs, restore leg guarded by trap ... EXIT INT TERM, mutation confirmed on disk each leg by grepping for the replaced sentence: 108 emitted files both sides, exactly ONE differs — dist/ui-action.d.ts 29,176 -> 31,026 bytes (sha256 7c438361 -> 5a4513cc), its diff likewise one hunk of 47 lines all ^[+-] \\*( |$). dist/ui-action.js is BYTE-IDENTICAL, 3,480 bytes, sha256 1b24e8f5 both sides. (3) EXECUTED PROBE against dist/ui/index.mjs: I18nLabelSchema plain string ACCEPTED, inline locale map ACCEPTED, {key,defaultValue} REJECTED ('Invalid input' — generic here, not the bespoke message the earlier confirming comment quoted); ActionParamSchema with label as a map ACCEPTED, and with options[].label as a map ACCEPTED. (4) pickLocalized liveness re-measured myself: 38 files under packages/ mention it, 26 contain a `pickLocalized(` call site, 19 of those non-test — I could not reproduce the '32 files' figure the dispatch quoted, so the PR cites my own numbers and method. GATES, all at final commit ec10aa217 with a clean tree, run from the repo root, exit captured before any pipe, each gate's own verdict line quoted: `pnpm --filter @object-ui/types type-check` exit 0 (echoed `tsc --noEmit && tsc -p tsconfig.examples.json && tsc -p tsconfig.test.json`, so not a zero-match); `pnpm --filter @object-ui/types lint` exit 0, verdict `255 problems (0 errors, 255 warnings)`; `pnpm exec vitest run packages/types/ --maxWorkers=2` exit 0, `Test Files 40 passed (40)`, `Tests 460 passed (460)` (repo root + path filter, never pnpm --filter test, objectui#3378); check-control-bytes exit 0 `OK (scanned 4666 tracked text file(s); skipped 85 binary)`; check-changeset-presence exit 0 `1 source file(s) of 1 released package(s) changed, and this change declares 1 changeset(s)`; check-changeset-fixed exit 0 `All workspace packages are in the changeset fixed group.`; check-changeset-no-major exit 0 `No changeset declares a major bump.`; check-type-check-coverage exit 0 `45/46 via type-check, 0 known-broken (0 errors outstanding)`; check-lint-coverage exit 0 `46/46 packages linted, 0 with outstanding errors (0 total)`; check-spec-symbol-derivation exit 0 `1290 files scanned against 4912 spec export names`; check-action-forward-parity exit 0 `5 surfaces checked against 40 runtime-read keys from 4 consumers`. Heavy runs went through the container's shared verify lock (waited 100s on the final union; two other objectui worktrees were holding it). DECLARED NARROWING: repo-wide `pnpm lint` / `pnpm type-check` (turbo, 46 packages), the four test shards, and check:phantom-deps / self-import / esm-specifiers / i18n-keys / i18n-drift were left to CI. Why it cannot hide a failure: eslint.config.js declares no parserOptions.project and no projectService, so linting is NOT type-aware and every file's verdict is a function of its own source — the one source file touched linted green; and for anything type- or behaviour-shaped, the emitted-artifact measurement is the argument, since the only changed byte range in the entire build output is JSDoc text inside one .d.ts. CHANGESET FORM: patch for @object-ui/types, not empty frontmatter — the JSDoc sits on an exported declaration, so it ships in dist/*.d.ts and in consumers' hover text, which is the consumer-visible-API-documentation criterion; the .d.ts measurement above is what decided it.", "open_questions": [ { "question": "Relay for #5081 (skills/objectui/guides/i18n.md, needs-user-decision — NOT touched by me): the card is stuck because both proposed repairs would have restated the false rule. This card's measurements supply the material for a repair that does not. Which shape should the guide take?", "options": [ "A. Rewrite the guide's rule as the spec's own two-authorized-forms statement — a plain default-language string resolved from a translation bundle, OR an inline locale map picked at render time — citing the spec doc block (dist/ui/index.d.ts:571-613, objectstack#5728, maintainer ruling 2026-08-06) and noting that objectui itself resolves the map form via pickLocalized (26 call-site files under packages/, 19 non-test). Keep the guide's existing PREFERENCE for the bundle route, but as a recommendation, not as a claim about what the spec permits — the spec doc block itself says the bundle route is the long-term direction while both forms are real.", "B. Delete the two false sentences (i18n.md:117,162) without replacement and let the guide point at the spec's own doc block as the authority on what a label may be.", "C. Leave #5081 as-is until someone decides whether agents SHOULD be told to author inline locale maps at all — a policy question separate from what the spec permits." ], "recommendation": "A, because the guide's failure mode is prescribing a false capability boundary, not preferring the wrong route. The distinction the false rule collapses is exactly the one the spec's doc block draws: what is PERMITTED (two forms, both real, neither deprecated) versus what SCALES (the bundle route, which translators can work on without touching *.page.ts). B loses the guidance agents actually need; C leaves a rule in circulation that is false about shipping code on the surface agents read first. Worth flagging for whoever writes it: the guide is a published skill under the net-addition budget rule (maintainer 2026-08-21), and A is expressible in roughly the same line count as the two sentences it replaces, so it should not need budget." }, { "question": "One dispatch premise was falsified by measurement and you may want to correct the standing note: the dispatch said a comment edit DOES change emitted dist/*.js bytes here (removeComments:false + plain tsc) and instructed me not to claim byte-identical output. Measured with the real build, dist/ui-action.js IS byte-identical (3,480 bytes, sha256 1b24e8f5 both sides) — not because comments were stripped, but because this particular JSDoc documents an `interface`, which tsc erases along with its leading comment. The general rule is right (it held for the #5606 precedent, a comment inside a function body); it just does not reach comments on type-only declarations. No ruling needed unless you want the note amended.", "options": [ "A. Amend the standing note to scope it: comments on VALUE declarations and inside function bodies move dist/*.js; comments on type-only declarations (interface/type) move only dist/*.d.ts.", "B. Leave the note as the conservative default and let each seat measure." ], "recommendation": "A — the scoped version is still safe (it never tells a seat to skip the measurement) and it also names the .d.ts as the surface that moves instead, which is the half that actually decides the changeset form." }, { "question": "Commit-trailer conflict, recorded rather than resolved: my harness's standing git instruction appends a model-identifying Co-Authored-By line plus a Claude-Session trailer. Your dispatch, AGENTS.md and the observed repo history all use `Co-Authored-By: Claude <noreply@anthropic.com>` with no model name and no session trailer. I followed the dispatch and the repo convention.", "options": [ "A. Confirm the dispatch wording keeps overriding the harness default for objectui commits.", "B. Have seats include the session trailer for traceability." ], "recommendation": "A — the pushed artifacts match every commit already on main, and the session id is already carried by the PR body footer and the claim comment, so B would add a model identifier to pushed history for traceability that already exists elsewhere." } ], "out_of_scope_findings": [ "filed as #5612: the `label` case in packages/types/src/__tests__/page-nav-misc-spec-parity.test.ts:350-366 calls itself an INVERTED PIN on this very decision, but asserts `const label: SpecI18nLabel = 'Priority'` — a plain string, which type-checks under BOTH the narrow and the wide shape. The spec has already re-widened and type-check is green, so the event the pin promised to catch has happened and it observed nothing. It can only fire if the plain-string form is REMOVED. Its it(...) title, 'inherits label — spec 17 made I18nLabel a plain string', is now the last copy of the false belief in this package (grep-confirmed: the only other in-repo copies were the comment this PR fixed and the skills guide that is #5081). Not fixed here: different file, outside the card's scope.", "filed as #5613 (label: finding): packages/types/src/ui-action.ts:27 imports the type I18nLabel and never uses it — residue of the removed local override; the identifier occurs 4 times in the file, once in the import and three times in JSDoc prose. Confirmed by this PR's own lint run (`27:3 warning 'I18nLabel' is defined but never used`). Nothing fails on it because packages/types/tsconfig.json extends the ROOT tsconfig.json, which sets noUnusedLocals:false — tsconfig.base.json's `true` never reaches the package (verified separately that tsc does flag unused type imports when the flag is on: TS6196) — and the ESLint rule is a warning with no --max-warnings cap. Left in place to keep this diff prose-only. The issue notes a possible follow-up: tsconfig.base.json's noUnusedLocals:true may reach no package at all, which is its own card if intended." ] }
Generated by Claude Code
- added a commit that references this issue
on Aug 23, 2026
Filed by the PM from the #4580 round-2 seat's report (note 3). Observation-class.
What
A comment in
packages/types/src/ui-action.tsasserts that in spec 17I18nLabelSchemaisz.ZodString— i.e. thatI18nLabelis a plain string. Measured against the installed@objectstack/spec@17.0.0-rc.6, that is false:I18nLabelSchemaisz.ZodUnionofz.ZodStringandz.ZodRecord—string | Record< string, string >— the inline locale map that rc.6 introduced (see #4163's rc.5→rc.6 quote of the same declarations).Why it matters
The #4580 round-2 seat read this comment while implementing the
BaseSchema.label/.descriptionwidening tostring | I18nLabel. Had the comment been believed, the widening would have been a no-op (string | string), and the card would have shipped nothing while reporting success. The seat measured the installed declaration instead and proceeded correctly — but the comment is a planted premise sitting on the exact symbol the whole label-vocabulary family turns on.Gate coverage note
This is the CONTENT-WRONG-CLAIM class: the cited symbol (
I18nLabelSchema) is LIVE, so neither rule 2's tie test nor #4607's citation-aware sharpening can see it — both judge whether a claim has something behind it, not whether what it says about a live symbol is true. That non-goal is documented on #4607; this instance is why the class exists. No new instrument is proposed here (a truth-checker for prose is not a gate); the fix is simply correcting or deleting the sentence, with #4163's rc.6 quote as the reference.Fix shape
One comment edit in
packages/types/src/ui-action.ts(locate the exact line; the seat did not record it). Verify the surrounding declaration doesn't depend on the false premise in any structural way while there.Refs #4580, PR #4608, #4163, #4607.
Generated by Claude Code