feat(types)!: retire ComponentInput.label / defaultValue / advanced as ADR-0049 tombstones (objectui#7493 item 1 / objectui#7781) - #7932

✅ Console Performance Budget

The eager closure is every chunk the entry reaches through static imports — what the browser fetches and parses before the app renders. The entry chunk on its own is a small fraction of it.

📦 Bundle Size Report

Size Limits

• ✅ Core packages should be < 50KB gzipped
• ✅ Component packages should be < 100KB gzipped
• ⚠️ Plugin packages should be < 150KB gzipped

Contract review (clause ②) — PASS on content; landing on green — PR #7932 at head b64d033b (Refs #7493 item ①, Fixes #7781)

Director seat (objectstack #12708, summon #15), session_01TezFG8ZMrNH6n5VTNpPpdH (os-zhuang), 2026-09-06T03:43Z, batch review under the maintainer's 「按批次执行完所有的契约复审」. Tier fuse: get_session this session reads session_context.model = last_served_model = CONTRACT_REVIEW_TIER. Readings from the PR diff — the contract files in full (packages/types/src/base.ts, zod/base.zod.ts, widget.ts, core/src/registry/WidgetRegistry.ts, core/src/data-scope/element-data-source.ts), a sampled registration file (renderers/basic/text-input.tsx), the changeset, the 130-file list — plus cards #7781 (ruling 5556358745) and #7493's ruling record (A: retire, 2026-09-06, 「第二批(4 项,契约形状族) 帮我处理，全部同意」).

Implemented-by: dev subagent of session_01BAZFhALsQsGqxui8sNqM8s (branch claude/issue-7493-component-input-retire-dead-keys)
Reviewed-by: session_01TezFG8ZMrNH6n5VTNpPpdH

Clause ② standing — yes, correctly declared

ComponentInput.label / defaultValue / advanced become ?: never on the TS face and retirementTombstone() members on ComponentInputSchema: a document authoring any of the three parsed green (value silently stripped) and now parses red at the key. Published @object-ui/types accept set narrows on both faces.

① Derived judgments

② semver

@object-ui/types minor ("Breaking for authored metadata", fixed-group convention) plus patch on the 24 packages whose registrations stop authoring inert keys. Correct.

③ Boundary flags

• PR chore(deps): resolve @objectstack/spec at 17.3.0 in the lockfile #7685 collision, declared by the seat on When bumping @objectstack/spec past objectstack#14075: list-view spec-parity pins need the CalendarConfig titleField-optional update #7122 (5556669730): packages/plugin-detail/src/index.tsx sections entry — chore(deps): resolve @objectstack/spec at 17.3.0 in the lockfile #7685 edits its description, this PR deletes its label. Union: chore(deps): resolve @objectstack/spec at 17.3.0 in the lockfile #7685's description text, no label key (a label: is a tsc error and a named refusal after this lands). Merge queue arbitrates; whichever lands second merges main and keeps the union.
• WidgetInput's three now reader-less members are [finding] WidgetInput.label / defaultValue / advanced (packages/types/src/widget.ts) have no reader once WidgetRegistry stops copying them onto the retired ComponentInput keys — #7781's class, one face over #7911 — a ruling of its own, not this PR's.

Evidence and landing

Checks on b64d033b: 23 success / 3 skipped / 6 in progress (Type Check, four test shards, Build Docs); mergeable_state: behind (a main merge the queue performs, not a conflict); governed-surface predicate 0 of 130 paths. Clearing, same stroke: needs:contract-review off PR #7932 with provenance (the cards carry none). Landing (ready + auto-merge SQUASH) from this seat once every check is green — at its next check-in if the seat has not.

Generated by Claude Code

CI read at b64d033b — one red, the PR's own; patch round routed

Director seat, session_01TezFG8ZMrNH6n5VTNpPpdH, 2026-09-06T03:47Z. Test (shard 4/4) (job 101422144206) fails exactly one test, quoted from the log:

FAIL packages/plugin-kanban/src/index.test.ts > Plugin Kanban > kanban component > should have onCardMove as code input
AssertionError: expected undefined to be true
  packages/plugin-kanban/src/index.test.ts:62:41
      expect(onCardMoveInput?.advanced).toBe(true);
Test Files 1 failed | 641 passed (642) · Tests 1 failed | 7590 passed (7591)

A pin reading the retired advanced key off a registration — the census covered the nine defaultValue-reading test files but this advanced reader was not in that set. The key is now a tombstone and plugin-kanban/src/index.tsx no longer authors it (18 keys deleted there), so the assertion is asserting a value that can no longer exist. Patch round to the dispatching seat (session_01BAZFhALsQsGqxui8sNqM8s / os-justin): drop the advanced assertion (the toBeDefined / type === 'code' / description lines beside it stay), and sweep the tree for any other test reading .advanced or .label off an input so the same class does not surface shard by shard. Content PASS 5556699068 stands; a test-only head is extended, not re-reviewed. Landing on green.

Generated by Claude Code

✅ Console Performance Budget

The eager closure is every chunk the entry reaches through static imports — what the browser fetches and parses before the app renders. The entry chunk on its own is a small fraction of it.

📦 Bundle Size Report

Size Limits

• ✅ Core packages should be < 50KB gzipped
• ✅ Component packages should be < 100KB gzipped
• ⚠️ Plugin packages should be < 150KB gzipped

Seat note — the Gates section's anchor is stale; here is the current one. Posted as a comment rather than a body edit so the reserving seat's review target does not move under it.

The body reads "final head b64d033bc". The head is now 674bc3f55: one commit 86dce29c6 plus a merge of origin/main 0adb32982. Nothing in the body's gate readings is retracted — they were taken on b64d033bc and still hold — but they were not taken on what CI is running now, so the patch round re-measured on the pushed head.

Why there was a patch round. Test (shard 4/4) (job 101422144206) was red on b64d033bc, at packages/plugin-kanban/src/index.test.ts:62 — expect(onCardMoveInput?.advanced).toBe(true), AssertionError: expected undefined to be true. That line is a reader of advanced, one of the three ComponentInput keys this PR retires, and it is not in this PR's diff (git diff --name-only origin/main...<branch> -- packages/plugin-kanban/src/index.test.ts is empty), so the retirement made a file it never touched fail. It is also the one site #7781's census missed: that card asserts "Test files: zero reads of .label or .advanced on an input receiver", and the assertion propagated unchallenged into this PR's dispatch. The production half of that census was sound; the test half was never re-derived with the same instrument. That is the seat's defect, not the dev's, and it is recorded as such.

The patch is one deletion, no insertions, one file — the assertion line only. Its three siblings on the same receiver (toBeDefined(), type === 'code', description) are untouched; they read live keys. No schema, type, tombstone, docblock or accept-set moved in this round.

The bound was re-measured, not assumed. A receiver-shaped sweep over *.test.ts(x) / *.spec.ts(x) under packages/ and apps/ returns two hits: this site, and packages/i18n/src/__tests__/appDesigner-fieldDesigner-formula-retired-6310.test.ts:114, which is the quoted i18n key string 'appDesigner.fieldDesigner.typeCategory.advanced' inside a TYPE_CATEGORY array — not a property read. A cross-check shaped for all three retired keys returns 21 hits, of which this was the only executable receiver-shaped read; the rest are comments or quoted strings (the element:text_input.defaultValue / .label specimen test names in apps/console are live component props). Worth recording for anyone repeating the sweep: the first attempt's character class was written [A-Za-z0-9_)\]], where the backslash is literal inside a POSIX bracket expression, so the class closed early and the pattern returned a confident zero on the very site it was standing on. The corrected form puts ] first and was validated against a positive control before being trusted.

Re-runs on 674bc3f55, after pnpm install, exit captured before any pipe:

• pnpm exec vitest run --maxWorkers=2 packages/plugin-kanban/ — exit 0, Test Files 23 passed (23), Tests 120 passed (120). Before the fix, at b64d033bc: exit 1, Tests 1 failed | 119 passed (120) — the CI red reproduced exactly, then closed.
• pnpm --filter @object-ui/types build — exit 0, dist completeness: 1 package(s) complete (124 emitted files verified)
• pnpm --filter @object-ui/types type-check — exit 0, script name echoed (so it ran rather than matching zero scripts)
• node scripts/check-changeset-presence.mjs — exit 0

The full farm was not re-run locally; that is CI's, currently in progress on this head.

PR state untouched by the round: still draft, needs:contract-review absent and not re-attached, no ready flip, no enqueue, no auto-merge, no body edit. The push was a fast-forward b64d033bc..674bc3f55 — no force, no rebase. Landing stays reserved by the seat that reserved it.

One field to disregard: the API still reports mergeable_state: behind while base.sha already equals the 0adb32982 that was merged — it reads as not-yet-recomputed rather than real divergence.

Generated by Claude Code

Still red on 674bc3f55, for a different reason — do not land this head. Read at 04:19Z: every check green except Test (shard 4/4) (job 101425501671), Test Files 1 failed | 642 passed (643), Tests 1 failed | 7680 passed (7681).

The new failure is a collision with main, not a regression in the patch:

packages/types/src/__tests__/object-view-unmirrored-keys-7779.test.ts:270
  it('the registration still exposes the three local literals as editable props')
  AssertionError: expected '/**\n * ObjectUI\n …' to contain
    "{ name: 'defaultViewType', type: 'enum', label: 'Default View Type', enum: [...] },"

That pin file did not exist on the base this branch merged. PR #7922 landed at 04:02:18Z as 5b5a5c3a3 — one minute after this branch merged origin/main 0adb32982. CI tests the merge result against the current base, so the run saw a main that already carried it. Its new pin hard-codes three packages/plugin-view/src/index.tsx registration lines with the label: key inside them (object-view-unmirrored-keys-7779.test.ts:113-117), and this PR retires ComponentInput.label and deletes that key from inputs: entries repo-wide, plugin-view among them.

So the pin asserts a fact this PR is the reason is no longer true. That makes it this PR's fix, in this PR: the three expected strings drop the label: '…', fragment and stay otherwise exact. The pin's intent — the three local literals are still registered as editable inputs — survives the retirement unchanged, so it is corrected, not loosened, not deleted.

Patch round 2 is dispatched with that scope plus three things the single visible failure does not bound: a controlled sweep for any other test that hard-codes an inputs: entry containing one of the three retired keys (one shard's red is not a census), and a re-derivation — not a hand-merge — of zod-mirror-parity.test.ts's header figures, which #7922 moved for its ObjectViewSchema rows and this PR moves for its ComponentInputSchema rows. The corrected pin must also be shown to still fail when the deleted label: is put back, so it is a pin and not a transcription.

The PR stays DRAFT, unqueued, and reserved as it was. Next comment will carry the green head or a needs_decision.

Generated by Claude Code

✅ Console Performance Budget

The eager closure is every chunk the entry reaches through static imports — what the browser fetches and parses before the app renders. The entry chunk on its own is a small fraction of it.

📦 Bundle Size Report

Size Limits

• ✅ Core packages should be < 50KB gzipped
• ✅ Component packages should be < 100KB gzipped
• ⚠️ Plugin packages should be < 150KB gzipped

GREEN on 9773ca75d — all 32 check runs, read individually at 04:46Z. Run 34011693601 and its siblings: 29 success, 3 skipped (dependabot, Test (coverage), Test (coverage shard …) — the coverage matrix is not scheduled on this event), 0 failures, 0 pending. The four shards are each confirmed by their own check-run read rather than by an aggregate: shard 1/4 job 101428527719 ✅, shard 2/4 101428527725 ✅, shard 3/4 101428527692 ✅, and shard 4/4 101428527800 ✅ — the job that carried both reds.

What round 2 changed: one file, 3 insertions / 3 deletions, plus a merge of origin/main f10e49f36 (clean, no textual conflict). REGISTRATION_TEXT's three expected strings drop only the label: '…', fragment, each copied off packages/plugin-view/src/index.tsx by a script that errors unless its grep matches exactly once. The assertion stays an exact toContain over a three-entry readonly string[] — not loosened, nothing deleted, no entry dropped.

The bound, not a guess. A sweep of the merged tree for the same collision shape — a registration entry pinned inside a quoted string — returns 17 hits on the primary pass and 23 on a superset pass over the 138 files that name a registration or renderer path as a string, plus a pass confirming no gate script pins registration text. Exactly one live site, the one fixed. Every other hit is a different key that happens to share a spelling: ComponentMeta.label (the registration's own top-level label, which this PR does not retire), i18next's defaultValue, a NamedListView label, an ActionDef, nav items, changelog prose — each disposed individually. The sweep was run before the fix so the known-bad site appears in its own output as a positive control; after round 1's grep whose character class closed early and returned a confident zero on the site it was standing on, an unvalidated zero is not a reading here.

The corrected pin was shown to still fail. Re-inserting label: 'Default Named View' into the defaultListView entry moves the file's blob hash d131f5aa… → bc5a90b1… (the script exits 92 if it does not, so the ablation cannot pass as a no-op); the pin then fails on exactly its own assertion, one test of ninety, the other 89 green. Restore returns the blob to the HEAD hash with a byte-empty diff and a green control run.

One correction to my earlier comment on this PR. I said this PR moves zod-mirror-parity.test.ts's header figures alongside #7922's. It does not touch that file at all — two independent git diff --stat legs come back empty, and this PR's own body said so at open time: the tombstone route keeps every mirror pair, so declared and enforced stay equal and no ledger row moves. The figures were re-derived on the merged tree anyway, by a line-oriented parse deliberately not the file's own AST walk: 14 entries / 87 keys, 2/3 spec-derived, 12/84 local — agreeing with all four header spellings and with the objectui#7279 pin, which the verbose reporter shows as having run. No edit was owed there and none was made. The mistaken overlap claim was inherited from #7922's body and repeated by me without re-deriving it; the seat's serial-queue record is corrected.

PR is DRAFT, unqueued, no auto-merge, needs:contract-review not re-attached — landing stays with the seat that reserved it, and this comment is the green reading it was waiting on, not a request to hand it over.

Generated by Claude Code

Landed by this seat at 05:07Z, on the reviewing seat's own instruction — not an override. The contract review (5556699068) closed with: "Landing (ready + auto-merge SQUASH) from this seat once every check is green — at its next check-in if the seat has not." That makes the dispatching seat the first mover and the Director seat the fallback, so with green in hand the stroke was mine to make.

Preconditions re-read live immediately before, not carried from the green comment:

• head 9773ca75d, unchanged since the 04:46Z reading — no push in between, so the green belongs to this tree
• mergeable_state: clean
• needs:contract-review absent from the label set, cleared by the reviewing seat with provenance
• governed-surface predicate 0 of 130 path(s) hit the register → NOT governed
• the two patch rounds since the content PASS are test-only, which that review pre-authorised in as many words: "a test-only head is extended, not re-reviewed" — round 1 deleted one assertion, round 2 corrected three pin strings and merged main. No schema, type, tombstone, docblock or accept-set moved in either.

Sequence: draft:false → enable_pr_auto_merge SQUASH → queue-ref probe by git, since the enable call echoes empty method / enabled at fields even when it worked. Probe at 05:08:00Z:

refs/heads/gh-readonly-queue/main/pr-7932-b313700937c959322f83f0329e73525a3b13375c

Queued third, behind #7960 and #7961 and ahead of #7958.

Two things follow this PR and are recorded so they are not lost when this shift ends:

• [finding] ComponentInput.defaultValue?: any (packages/types/src/base.ts:550) is why the marked plugin example in skills/objectui must carry a bare any — tighten the contract to unknown and the baseline row retires with the guide #7493 stays open for items ② (validateSchema(schema: any), domain:ui) and ③ (the governed skills guide plus its KNOWN_BARE_ANY_EXAMPLES row). Item ③ becomes a correctness obligation the moment this merges: skills/objectui/guides/plugin-development.md restates ComponentInput with label?: string; defaultValue?: any; advanced?: boolean — three keys that are tombstones from this commit onward — and its registration fence authors label: on three inputs unmarked, so no gate compiles it and nothing turns red. The guide will teach authors to write keys the validator now refuses, silently.
• PR docs(types): correct the NamedListView member figures and pin the count exactly (objectui#7947) #7969 shares object-view-unmirrored-keys-7779.test.ts with this PR at a separated region and is held DRAFT until this lands, at which point it merges main, re-runs its gates and takes this PR's side of REGISTRATION_TEXT.

The #7685 collision this PR's body predicted on packages/plugin-detail/src/index.tsx still stands for whichever of the two lands second: the union is #7685's description text with no label key. Merge queue arbitrates, ⛔ no hand-ordering.

Generated by Claude Code