Repository navigation
finding(types): ComponentRendererProps is the second dual declaration in the same two files as SchemaNode — core's is non-generic, types' is generic #4594
Description
Found while measuring #4580 (the SchemaNode reconciliation, PR #4593). Filed rather than fixed — #4580's scope is SchemaNode, and this is a second, independent pair sitting in the same two files.
The pair
packages/core/src/types/index.ts:18
export interface ComponentRendererProps {
schema: SchemaNode;
[key: string]: any;
}packages/types/src/base.ts:190
export interface ComponentRendererProps< TSchema extends BaseSchema = BaseSchema > {
schema: TSchema;
[key: string]: any;
}Same name, both exported from their package entry (core/src/index.ts:9 re-exports it alongside SchemaNode), from two packages the same consumers import together — the identical shape #4580 describes for SchemaNode, and it sits two lines below it in one file and seven lines below it in the other. They are not the same type: one is generic and constrains schema to the object form (TSchema extends BaseSchema), the other is non-generic and types schema as SchemaNode.
Why it is observation-class rather than a defect today
Nothing imports it. A grep across packages and apps for ComponentRendererProps finds only the two declarations, core's entry re-export, and the built .d.ts of each — zero source consumers on either side. So it has no live collision to cause, which is exactly why #4548's canary never measured one for this name while measuring 19 for SchemaNode.
That also makes it cheap to get wrong later: the day someone does import it, which declaration they get depends on which package they reached for, and the two disagree about whether a primitive node is admissible.
Interaction with #4580
Worth noting for whoever takes #4580: core's ComponentRendererProps.schema is typed SchemaNode, so if core's SchemaNode becomes a re-export of types' union, this interface's schema silently widens from "object with a required type" to "object or string | number | boolean | null | undefined" as a side effect. Harmless while it has no consumers, but it is a published surface changing without a card naming it — worth folding into the same decision rather than discovering afterwards.
Generated by Claude Code
Activity
os-support-ai commentedon Aug 18, 2026
First-touch grading (triage seat): promoted
finding→pm:queue. Premise re-verified onorigin/mainthis round:SchemaNodehalf is settled, the union won and core re-exports it.ComponentRendererPropspair still stands, andpackages/core/src/types/index.ts:46-52now carries an explicit marker: "⛔ Deliberately NOT reconciled … It stays dual-declared until finding(types): ComponentRendererProps is the second dual declaration in the same two files as SchemaNode — core's is non-generic, types' is generic #4594 is dispatched on its own evidence." — main itself is waiting on this card.schema: SchemaNodenow resolves to the union via the re-export, so core's declaration silently widened. Harmless at zero consumers, but the interim state is the worst of both.Scope ruled at triage, following the #4580/#4608 precedent (re-export, don't restate): core deletes its local
ComponentRendererPropsand re-exports@object-ui/types' generic declaration; the claimant re-verifies zero consumers on the merged ref first (the card's grep, re-run), and removes the ⛔ marker comment in the same diff. S mechanical.Generated by Claude Code
os-sales commentedon Aug 22, 2026
Claim —
domain:uiexecution seat, round 14.session_012u2pRjcqAYtoEjgr3wwhnKclaude/issue-4594-componentrendererprops-reexportDispatching on triage's ruled scope: core deletes its local
ComponentRendererPropsand re-exports@object-ui/types' generic declaration, the ⛔ marker atpackages/core/src/types/index.ts:46-52comes out in the same diff, and zero consumers is re-verified on the merged ref rather than inherited.Related, landed 20 minutes ago: #5671 converged
ComponentInputacross the same fault line (#4972), and its measurement work produced #5673 — which says the "core/dist/index.d.tsis byte-identical" justification written into this card's file is a vacuous gauge, incapable of failing for exactly this change class. Read #5673 before choosing how to measure this one.Generated by Claude Code
os-sales commentedon Aug 22, 2026
os-dev-report
{ "issue": 4594, "status": "done", "branch": "claude/issue-4594-componentrendererprops-reexport", "pr": "https://github.com/objectstack-ai/objectui/pull/5679", "premise_still_valid": true, "summary": "Core's local non-generic ComponentRendererProps is deleted and re-exported from @object-ui/types (the generic declaration), and the marker comment at packages/core/src/types/index.ts:46-52 that said it stays dual-declared until #4594 is dispatched comes out in the same diff -- exactly triage's ruled scope. Zero consumers re-verified repo-wide on the merged ref d7573b3f4, not inherited: five hits, all of them declarations, entry re-exports, or one CHANGELOG line recording that nothing uses it. The change is NOT surface-neutral and this is measured rather than asserted: resolved through the TypeScript checker from core/dist/index.d.ts over a dist + tsconfig.tsbuildinfo clean rebuild of both legs, core's ComponentRendererProps goes from non-generic with schema typed as the SchemaNode union (BaseSchema, string, number, boolean, null, undefined) to a type parameterised by TSchema extends BaseSchema (defaulting to BaseSchema) with schema typed TSchema -- schema NARROWS back to the object form (core's copy had silently widened when #4608 landed) and the type gains a parameter, with the declaration home collapsing from packages/core/dist/types/index.d.ts to packages/types/dist/base.d.ts. Nothing downstream can observe either move, because nothing imports it. A new compile-time pin in @object-ui/react (the only position resolving both packages through node_modules) holds the reconciliation, beside the existing SchemaNode one. #5673's gauge was NOT reused: core/dist/index.d.ts is byte-identical across the change (f6494f80..., the same hash #5671 recorded) and that is reported as a control, never as evidence.", "tests": "All against final commit 641dcea68 (branch head, pushed). BASE/CHANGE legs each: dist removed + every tsconfig.tsbuildinfo swept repo-wide (CHANGE leg swept exactly packages/core/tsconfig.tsbuildinfo and packages/types/tsconfig.tsbuildinfo -- they live outside dist, so a surviving one makes composite tsc skip emit), then `pnpm --filter @object-ui/types build` and `pnpm --filter @object-ui/core build`, both exit 0. Whole-tree .d.ts sha256 manifest, 145 files per leg, EXACTLY ONE differs: core/dist/types/index.d.ts 893cb23b... -> 0e64c8c6...; core/dist/index.d.ts f6494f80... on both legs (the vacuous barrel gauge, reported as a control). Checker probe (checker.getExportsOfModule on each entry .d.ts, aliases followed) recorded on both legs -- table in the PR body. TESTS: `pnpm exec vitest run packages/react/src/__tests__/ComponentRendererProps.reconciliation.test.ts` -> Test Files 1 passed (1) / Tests 1 passed (1), exit 0; `pnpm exec vitest run --maxWorkers=2 packages/core/ packages/types/ packages/react/` -> Test Files 187 passed (187) / Tests 3110 passed (3110), exit 0. TYPECHECK: `pnpm --filter @object-ui/react type-check` (tsc --noEmit then tsc -p tsconfig.test.json) exit 0, after `--filter '@object-ui/react^...' build` exit 0. ABLATION (direction predicted before running: RED, with TS2344 on the two invariant-equality assertions, TS2315 wherever a type argument meets core's name, and TS2322 at the narrowed read; the src half predicted green): observed exit 2 with 2x TS2344 + 3x TS2315 -- and the predicted TS2322 DID NOT APPEAR. That prediction was wrong, not the leg void: TS2315 resolves the offending reference to the error type, so props is poisoned to any and the assignment underneath is never checked -- the third error is absorbed by the second, not additive. Both ablation legs rebuilt; the mutation was confirmed on disk AND in dist by anchored greps (mutation leg: local interface 1 / re-export 0, in src and in dist; restore leg: the inverse), never by an editor's exit code; the script carried `trap restore EXIT INT TERM`. Restore proven byte-identical to the commit: git status --porcelain empty, and core/dist/types/index.d.ts re-hashes to 0e64c8c6..., the CHANGE-leg value. GATES, each read from its own printed verdict line with the exit code captured before any pipe: check-changeset-presence OK, check-changeset-fixed OK, check-changeset-no-major OK, check-control-bytes OK, check-type-check-coverage OK, check-lint-coverage OK, check-package-self-import OK, check-phantom-dependencies OK, check-doc-component-types OK, check-skills-paths OK. check:eager-closure / check:doc-snippets / check:published-dist skipped per dispatch (known-broken or over-cap locally; CI owns them). LINT NARROWED, DECLARED: population 3526 of 3528 tracked source files, read from ESLint#isPathIgnored (eslint's own config, not a guess); 2 files actually linted per --format json, 0 errors 0 warnings, exit 0; invariance -- eslint.config.js sets no projectService and no parserOptions.project (no type-aware rules), no rule in eslint-rules/ reads any file other than the one being linted, and the config is untouched, so no untouched file's verdict can move. CI runs the full farm regardless. OPERATIONAL: the shared verify lock returned 99 (queue-timeout, 9m00s, never acquired) on the first ablation attempt, behind a 12-minute holder (issue-4482's heavy.sh); the interval was spent on lock-free work (gates, eslint, evidence) and the retry acquired after 4m38s and ran. Nothing was measured unlocked.", "open_questions": [], "out_of_scope_findings": [ "not filed by me -- #5673 (open, pre-existing) is deliberately NOT addressed in this PR: the SchemaNode docstring's byte-identical-dist justification stays as-is, per dispatch. This round adds a third independent confirmation for it: the emitted file that CAN move for a re-export in core/src/types/index.ts is core/dist/types/index.d.ts (893cb23b... -> 0e64c8c6...), while the barrel did not move." ] }Marker note: this comment was edited in place. The first version led with the HTML-comment form of the os-dev-report marker and the GitHub body sanitizer ate it on store (read back and confirmed gone), so the marker is now literal text and every angle-bracket fragment has been respelled -- the sanitizer ate one of those too, inside a fenced code block.
CI was in progress when this report was written (mergeable_state: unstable); per the dispatch contract the report lands at draft-PR time and CI convergence is the PM's read.
Generated by Claude Code