Repository navigation
tooling(objectui-readiness): make the objectui tree merge-ready in its own repo before cutover (objectstack#25685) - #13472
Merged
Merged
Conversation
…to 19.3.0 (objectstack#25685 item 1) Every manifest that pins `react` / `react-dom` moves 19.2.8 -> 19.3.0 and every `@types/react` / `@types/react-dom` pin moves to 19.3.0, matching the root `pnpm.overrides`. objectstack's `@objectstack/client-react` requires react `^19.3.0`, and a second `@types/react` copy broke `@object-ui/plugin-view`'s declaration build in the merged `ui/` tree. The CLI scaffold (`REACT_RANGE` and the shared dev-dependency map) quotes the same pins, as its anchor test requires. Lockfile regenerated with `pnpm install --no-frozen-lockfile`; it also carries the single-zod-range resolution that item 2 declares in the manifests. Clause-②: no Metadata-protocol: none Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RSDL3SFtJaKzcymvCD8RsU
…ge that lists it (objectstack#25685 item 2) `@object-ui/types`, `components`, `app-shell`, `test-support` and `plugin-timeline` declared three different zod ranges (`^4.4.3`, `^4.6.1`); in the merged `ui/` tree that resolved to two zod copies and the `z.ZodType` identity split between them. All five now declare `^4.6.5`, the range objectstack's own packages use, so the lockfile holds one zod 4.x for the workspace (commit 1's regenerated lockfile already carries that resolution). Clause-②: no Metadata-protocol: none Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RSDL3SFtJaKzcymvCD8RsU
… `.git` or `pnpm-workspace.yaml` (objectstack#25685 item 3) In the merged tree `ui/` has no `.git` and no `pnpm-workspace.yaml` of its own, so every test and script that located the repo root by walking up to one of those markers (or by `git rev-parse --show-toplevel`) would resolve the OBJECTSTACK root instead. `scripts/ui-root.mjs` exports `UI_ROOT` / `uiRoot()`, derived from the helper's own location and nothing else. Migrated: fifteen `repoRoot()` walkers, the sdui-parser census IIFE, four `pnpm-workspace.yaml` existence assertions, and four scripts that shelled out to `git rev-parse --show-toplevel`. `check-new-cross-file-line-citations.mjs` now diffs `--relative` and reads blobs as `rev:./path`, so it keeps working when the git toplevel is a parent of the ui root. Two instruments keep it that way: `scripts/check-ui-root-lookups.mjs` (`pnpm check:ui-root-lookups`, with `--self-test`) scans `scripts/`, `packages/` and `apps/` with the TypeScript AST for marker probes, `--show-toplevel` / `--git-common-dir` and `find-up`, against a ledger allowlist of the four CLI product predicates that legitimately probe the USER's cwd; and `check-test-path-roots.mjs` now classifies `uiRoot()` / `UI_ROOT` as the self root. Clause-②: no Metadata-protocol: none Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RSDL3SFtJaKzcymvCD8RsU
…h-heap.mjs` (objectstack#25685 item 4) `@object-ui/app-shell` and `@object-ui/console` type-check against the whole objectstack tree once merged, and both OOM at Node's default old-space limit there. `scripts/with-heap.mjs CMD...` re-runs its arguments with `--max-old-space-size=4096` appended to NODE_OPTIONS (an old-space size the caller already set wins), propagates the child's exit code, and is what the two packages' `build` and `type-check` scripts now run `tsc` through — a cross-platform spelling instead of an inline `NODE_OPTIONS=...` prefix. Measured in this tree with `tsc --noEmit --extendedDiagnostics`: app-shell 2 234 110K, console 2 355 117K of heap. The turbo `build` inputs guard derives each package's build program from its script; it now understands the wrapper (recorded as part of the program AND looked through, so the wrapped `tsc` still contributes its tsconfig chain), and `turbo.json` lists `$TURBO_ROOT$/scripts/with-heap.mjs` as a `build` input. Clause-②: no Metadata-protocol: none Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RSDL3SFtJaKzcymvCD8RsU
…workspace layout (objectstack#25685 item 5) `apps/console/vite.config.ts` recognised `@objectstack/*` modules for its vendor / spec / lazy-linter chunk rules only by their `node_modules` and pnpm-store path spellings. In the merged tree those packages are workspace siblings resolved from `<root>/packages/<name>/dist/...`, so every rule would have silently missed them. `scripts/vite-objectstack-workspace.ts` exports `frameworkPackagesDir(consoleDir)` (`../../../packages` from `ui/apps/console`) and `workspaceChunkTests(dir)`, which widens each of the three tests with an anchored, escaped `<dir>/` alternative (vendor excludes `lint/`, which stays lazy). The baseline spellings are kept, so the standalone layout is unchanged. Ported from spike 3efb112efb with the unit test it lacked: a standalone `node_modules/@objectstack/spec/...` id stays matched, `<root>/packages/spec/dist/x.mjs` is matched by spec and vendor, `<root>/packages/lint/...` is excluded from vendor and matched by lint, and a sibling `packages-other/` is not matched. Clause-②: no Metadata-protocol: none Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RSDL3SFtJaKzcymvCD8RsU
…floor readers `workspace:` ranges (objectstack#25685 item 6) Two small pre-conditions for the merged tree. The root `typecheck` script is an alias of `type-check` (`turbo run type-check`), the spelling objectstack's fleet tooling invokes. And the three readers of `@objectstack/spec` RANGES — `scripts/check-spec-range-floors.mjs`, `scripts/objectstack-floor.mjs` and the `declaredSpecFloor-9012` test — understand the `workspace:` protocol: `workspace:*` / `^` / `~` / bare mean "the workspace package's own version", read from the installed `@objectstack/spec/package.json` (`workspaceSpecVersion`, `installedVersion`) and passed into `minVersionOf` / `effectiveFloor` / `planRanges` as `workspaceVersion`; `workspace:^18.1.0` reads its range tail. A `workspace:` range met without that version throws rather than silently passing, and the gate's self-tests cover both shapes. Clause-②: no Metadata-protocol: none Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RSDL3SFtJaKzcymvCD8RsU
github-actions
Bot
added
configuration
data-adapter
dependencies
package: core
package: types
package: react
package: components
plugin
plugin: charts
plugin: editor
plugin: kanban
plugin: markdown
examples
apps
tests
package: app-shell
package: fields
package: cli
package: auth
package: layout
package: runner
package: i18n
package: mobile
labels
Contributor
✅ 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
|
…rivate regex (objectstack#25685 item 3) The `uiRoot()` test strips comments from `scripts/ui-root.mjs` before asserting what the helper spells and does not spell. It did so with its own `/\/\*[\s\S]*?\*\//` + `//`-to-end-of-line pair, which the `check-hand-rolled-comment-mask` gate rightly refuses: that is the naive-regex family the tree retired, and a `//` inside a string would open a phantom comment. The test reports neither a line nor an offset, so it now calls `stripComments` from `scripts/js-comment-mask.mjs`, the one answer to "is this span a comment". Verified locally: the gate test and `ui-root.test.ts` (26 tests) pass. Clause-②: no Metadata-protocol: none Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RSDL3SFtJaKzcymvCD8RsU
Contributor
✅ 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
|
… not a root wrapper script (objectstack#25685 item 4) `scripts/with-heap.mjs` was a file OUTSIDE the two packages whose build runs it, and this tree's gates chain from that fact to a governed surface: `turbo-build-inputs.test.ts` requires every out-of-package file a build program reads to be a `$TURBO_ROOT$` input of `build`, and `docs-build-trigger.test.ts` requires every such input to be in the docs job's pathspec in `ci.yml` -- a `.github/workflows/**` edit the maintainer must approve. Shard 15 went red on the second gate. A heap ceiling is not a build input: the spelling `packages/spec` in objectstack uses, `NODE_OPTIONS=--max-old-space-size=4096 tsc` inside the package script, needs no file, no turbo input and no workflow line, so app-shell and the console carry it that way. `build-program.ts` skips the leading `NAME=value` words of a command before classifying the tool, so the two wrapped builds are still tsc runs in the inputs sweep; the wrapper, its test and its turbo.json line are gone. Clause-②: no Metadata-protocol: none Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RSDL3SFtJaKzcymvCD8RsU
…e monorepo packages dir exists (objectstack#25685 item 5) `vite-objectstack-spec-dist.test.ts` pins the `vendor-objectstack` test's literal as one of objectui#4854's four flagged surfaces, and shard 7 went red because the workspace alternative was prepended unconditionally: standalone it matched nothing, but the literal had moved. `frameworkPackagesDir()` now returns `null` when `<console>/../../../packages` is not a directory, and `workspaceChunkTests(null)` returns `BASELINE_CHUNK_TESTS`, the baseline literals byte for byte -- so standalone the config is unchanged and the pin holds, and under `ui/` the directory exists and the arm is added. The helper's own test pins the literal equality, a real monorepo-shaped temp tree, and that a FILE named `packages` is not the directory. Clause-②: no Metadata-protocol: none Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RSDL3SFtJaKzcymvCD8RsU
…e cwd, narrowed to the objectui tree (objectstack#25685 item 3) The gate's self-test drives it against a scratch repository under `node_modules/`, so its subject is the repository at the cwd; rooting it on `uiRoot()` alone made `git merge-base` run in the objectui checkout against a scratch commit (shard 12: "no merge base"). `gateRoot()` reads the toplevel of the cwd's repository and narrows to `uiRoot()` only when the objectui tree lies inside it: standalone the two coincide, under the objectstack monorepo the narrowing is what `--relative` and `rev:./path` rest on, and in the scratch repository the toplevel is read. The one `--show-toplevel` this adds is on the `check-ui-root-lookups` allowlist with that reason. Clause-②: no Metadata-protocol: none Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RSDL3SFtJaKzcymvCD8RsU
…s the module (objectstack#25685 item 3) `check-lint-rule-coverage` reported `scripts/ui-root.d.mts` as a file no ESLint config reaches and no ledger row declares (shard 13). The twin was never needed: `tsconfig.scripts.json` reads `.mjs` with `allowJs`, so the JSDoc on `ui-root.mjs` is its typing, exactly as for every other `.mjs` helper the script tests import. The test now pins the export surface and the twin's absence. Clause-②: no Metadata-protocol: none Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RSDL3SFtJaKzcymvCD8RsU
Contributor
✅ 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
|
…ave no allowJs — and ledger it for the lint census (objectstack#25685 item 3) 2195392 deleted the declaration twin because `tsconfig.scripts.json` reads `.mjs` with `allowJs`. The PACKAGE tsconfigs do not: `@object-ui/data-objectstack`'s `type-check` failed with TS7016 on `import { uiRoot } from '../../../scripts/ui-root.mjs'` in `readme-filter-operator-table.test.ts`, and every migrated package test imports the helper the same way. So the twin is back, with its pin test, and `check-lint-rule-coverage`'s predicate-2 ledger (`UNREACHED_GROUPS`, exact paths only) carries a row for it: two `export declare` lines no ESLint config reaches, declared rather than repaired because reaching a `.d.mts` means widening a rule-bearing `files` glob to `.mts`, the rule-strength decision objectui#8337 kept out of this gate. The gate's own test pins the unreached population as these two files. Clause-②: no Metadata-protocol: none Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RSDL3SFtJaKzcymvCD8RsU
Contributor
✅ 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
|
Contributor
Author
维护者速读改了什么 —— 这是 objectui 并入 objectstack 为什么改 —— 导入 风险与代价(含回滚) —— 超过 3000 行人工审核线的主因是 席位意见 —— 建议批准。无法拆到 3000 行以下:锁文件是整体文件,拆开 react/zod 两项只会产生两个各自超线的锁文件 diff。路径面未命中任何受管面( 你要做的(一个动作) —— 在本 PR 上留一条 APPROVED review(os-zhuang / hotlong 任一),之后由本席翻 ready、挂 auto-merge 走合并队列;或直接合并也可。 席位核对
ACCEPT for objectstack#25685 points here. |
os-zhuang
approved these changes
Oct 24, 2026
os-zhuang
marked this pull request as ready for review
October 24, 2026 22:47
os-zhuang
enabled auto-merge
October 24, 2026 22:47
This was referenced Oct 25, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes objectstack-ai/objectstack#25685
Part of objectstack-ai/objectstack#25677 (sub-issue 2/6).
Clause-②: no
Metadata-protocol: none
One draft PR for the whole card, one commit per scope item in card order. Branch
claude/project-thread-dwhq6r, basemainat a3c0da9 (chore: release packages, #13289). Nothing under.github/,AGENTS.md,.changeset/config.jsonor any other governed surface is touched; the objectui tree is prepared in its own repo so theui/import (ruling D1 / A+) merges without a follow-up fix-up commit.What changed, per item
1. One react line (commit 8474f4d)
package.jsondevDependencies andpnpm.overrides, everypackages/*,apps/console,apps/site, the three examples):react/react-dom19.2.8 to 19.3.0,@types/react19.2.18 to 19.3.0,@types/react-dom19.2.4 to 19.3.0.packages/cli/src/utils/scaffold-dependencies.ts:REACT_RANGE = '19.3.0'and the@types/react*entries of the shared dev-dependency map, which the CLI anchor test compares to the root manifest.pnpm-lock.yamlregenerated withpnpm install --no-frozen-lockfile(1337 lines added / 1337 removed). It also carries item 2's single-zod resolution, because one install produced both..changeset/25685-one-react-line.md: patch for the 30 published packages whose manifests moved.Before:
grep -c '"react": "19.2.8"'over the workspace manifests = 33; after = 0, andpnpm ls react --depth 0across the workspace resolves onereact@19.3.0.2. One zod range (commit 1d6a038)
packages/types,components,app-shell,test-support,plugin-timeline:"zod": "^4.6.5"(was^4.4.3x3 and^4.6.1x2)..changeset/25685-one-zod-range.mdfor the four published packages.Before: three distinct declared ranges; after: one, and the lockfile holds a single
zod@4.6.7for the workspace (the unrelatedzod@3.25.76is a transitive dependency of a third-party tool and is unchanged).3.
uiRoot()instead of.git/pnpm-workspace.yamllookups (commit 5cc5354)scripts/ui-root.mjs(UI_ROOT,uiRoot(), derived from the helper's own location; node builtins only) and its declaration twinscripts/ui-root.d.mts: package tsconfigs have noallowJs, so without the twin every migrated package test's import is TS7016 (2195392 removed it, the Type Check job proved it needed, ae7fcc5 restored it).check-lint-rule-coveragereports a.d.mtsas a file no ESLint config reaches, so its predicate-2 ledger (UNREACHED_GROUPS, exact paths) carries a row for it, owned by this PR; the gate's own test pins the unreached population asscripts/ui-root.d.mts+vitest.config.mts.repoRoot()walkers in package tests (types x7, components, plugin-calendar x2, plugin-gantt, app-shell x2, data-objectstack, core), the sdui-parser census IIFE, fourexistsSync(join(root, 'pnpm-workspace.yaml'))sanity assertions (nowexpect(root).toBe(uiRoot())), and four scripts that shelled out togit rev-parse --show-toplevel(cross-file-line-citation-census.mjs,dollar-dialect-alias-census.mjs,check-new-cross-file-line-citations.mjs,check-cross-repo-closer-outcome.mjs).check-new-cross-file-line-citations.mjsnow diffs with--relativeand reads blobs asrev:./path, so it keeps working when the git toplevel is a parent of the ui root; its root isgateRoot()(7537206): the repository at the cwd via--show-toplevel— its self-test drives the gate against a scratch repository — narrowed touiRoot()when the objectui tree lies inside it.scripts/check-ui-root-lookups.mjs(pnpm check:ui-root-lookups;--list,--json,--self-testwith 13 cases): a TypeScript-AST scan ofscripts/,packages/,apps/for marker probes (existsSync/statSync/lstatSync/accessSyncon.gitorpnpm-workspace.yaml, directly, via a const binding or a template literal),--show-toplevel/--git-common-dirliterals, andfind-up. Allowlist is a ledger (a stale entry is red) with five entries:scripts/objectstack-pin.mjs(deleted at cutover),scripts/check-new-cross-file-line-citations.mjs(its subject is the repository at the cwd, above) and three CLI product predicates that probe the USER's cwd (app-generator.ts,workspace-vite.ts,app-generator.test.ts).scripts/check-test-path-roots.mjsclassifiesuiRoot()/UI_ROOTas the self root (two new self-test cases).Before: on
mainthe fifteen walkers and four shell-outs resolved the root through.git/pnpm-workspace.yaml/git rev-parse, andcheck-test-path-rootsreported 1264 self / 590 unclassified roots. After:check-ui-root-lookups: OK (7796 files scanned, 55 parsed, 6 probes all on the allowlist, 5 allowlist entries live);check-test-path-roots: OK, 1264 self / 579 unclassified. The card's crude grep (A4) re-derived to 55 unique files onmain(the card quoted 71); 20 package test files and 4 scripts were migrated, and the residue is comments,readFileSyncDATA reads of the workspace globs, fixture writes, two'.git'directory skip-list entries and the allowlisted CLI predicates — the gate, not the grep, is the instrument that fails when a new lookup lands.4. tsc heap for app-shell and console (commits 02eaff5, c464a4a)
packages/app-shell/package.jsonandapps/console/package.json: thetscofbuildandtype-checkruns asNODE_OPTIONS=--max-old-space-size=4096 tsc ...— the spelling objectstack'spackages/specuses for the same purpose.scripts/__tests__/helpers/build-program.ts:stripEnvAssignments()skips the leadingNAME=valuewords of a command before classifying the tool, so the two builds are still tsc runs in theturbo-build-inputssweep and contribute their tsconfig chain.scripts/with-heap.mjs(02eaff5) is a file outside the two packages, soturbo-build-inputs.test.tsrequires it as a$TURBO_ROOT$input ofbuild, anddocs-build-trigger.test.tsthen requires every such input in the docs job's pathspec inci.yml— a.github/workflows/**edit the maintainer must approve. Shard 15 went red on the second gate; c464a4a removes the wrapper, its test and its turbo.json line. A heap ceiling is not a build input.Measured (A3) with
tsc --noEmit --extendedDiagnosticsin this tree: app-shellMemory used: 2234110K(3915 files, 50.7 s), consoleMemory used: 2355117K(4122 files, 25.4 s). Both exceed what the default old-space limit left on the merged-tree runner in the spike measurement; 4096 MB has headroom.5. Vite chunk tests derived from the workspace layout (commit 5c76fe3)
scripts/vite-objectstack-workspace.ts:candidateFrameworkPackagesDir(consoleDir)(../../../packagesfromui/apps/console),frameworkPackagesDir(consoleDir)(that directory when it exists, elsenull) andworkspaceChunkTests(dir | null)returning{ vendor, spec, lint }regexes that keep the baselinenode_modules/ pnpm-store spellings and add an anchored, escapeddir/alternative (vendor excludeslint/, which must stay lazy). Withnull— standalone — it returnsBASELINE_CHUNK_TESTS, the baseline literals byte for byte (dde5e6d):vite-objectstack-spec-dist.test.tspins the vendor literal as one of objectui#4854's flagged surfaces, and an unconditional arm had moved it (shard 7).apps/console/vite.config.tsuses them forVENDOR_OBJECTSTACK_TEST,SPEC_MODULE_TESTand the lazy-linter assertion.scripts/__tests__/vite-objectstack-workspace.test.ts: standalone the three tests are the baseline literals; anode_modules/@objectstack/spec/...id is matched in both layouts;<monorepo>/packages/spec/dist/x.mjsis matched by spec and vendor;<monorepo>/packages/lint/...is excluded from vendor and matched by lint; a sibling directory, a FILE namedpackagesand objectui's ownui/packages/*are not matched.scripts/vite-objectstack-client-dist.tsjudged and left alone: its chunk test has nonode_modulesassumption, andpackageResolvesFromwalksnode_modules/name, which workspace links satisfy.6.
pnpm typecheckalias andworkspace:spec ranges (commit 11bc209)package.json:"typecheck": "turbo run type-check".scripts/check-spec-range-floors.mjs:WORKSPACE_PROTOCOL;minVersionOf(range, { workspaceVersion })readsworkspace:*/^/~/ bare as the workspace package's own version andworkspace:^18.1.0as its range tail;workspaceSpecVersion(root)reads the installed@objectstack/spec/package.json; aworkspace:range met without that version throws instead of passing.scripts/objectstack-floor.mjs:installedVersion,floorOfRange,planRanges(..., { workspaceVersionOf }).packages/core/src/utils/__tests__/normalize-list-view.declaredSpecFloor-9012.test.ts:floorOfreads aworkspace:range as the installed spec's own version (the oldworkspace:* is not a floorassertion replaced by that test).scripts/__tests__/check-spec-range-floors.test.tsandscripts/__tests__/objectstack-floor.test.ts.Other readers of
@objectstack/specranges that were NOT changed (they do not fail onworkspace:today):scripts/check-changeset-no-major.mjs(majorOfRange) andscripts/sync-quick-reference-release.mjs.Verification (local, this worktree)
Install:
pnpm install --no-frozen-lockfileagainst a stand-inOBJECTSTACK_ROOT(sibling objectstack checkout at 3efb112efb; pin manifests and sibling manifests have identical dependency names and exports, all 18.1.0, so.objectstack-pinand thefile:lockfile snapshots are unchanged). react / react-dom / @types 19.3.0 resolved.Under the shared verify lock (
--concurrency=2, 4096 MB heap):turbo run buildfor the dependency closure of console, cli and app-shell: 34 tasks, 34 successful.pnpm --filter @object-ui/app-shell type-check,pnpm --filter @object-ui/console build(vite build + sdui manifest + plugin),pnpm --filter @object-ui/cli build,type-check,test: all exit 0.declaredSpecFloor-9012, cliapp-generator,console-vite-alias-closure-4925,doc-version-claims): 24 passed, 325 tests.Unlocked (the dev's run, before the five red shards below):
pnpm type-check:scriptsexit 0;scripts/__tests__suites naming an edited script (ui-root,with-heap,check-ui-root-lookups,check-test-path-roots,vite-objectstack-workspace,objectstack-floor,check-spec-range-floors,react-peer-range-norm-3741,quick-reference-commands-4149,turbo-type-check-inputs,turbo-build-inputs,turbo-test-inputs,turbo-lint-inputs,scripts-type-check): all green. Gates exit 0:check:lockfile-integrity,check:lockfile-dedupe,check:phantom-deps,check:unused-deps,check:installed-pin-claims,check:test-path-roots,check:ui-root-lookups,check:entry-guard,check:pre-install-import-graph,check:control-bytes,type-check:coverage,check:new-line-citations,check-changeset-presence,check-changeset-no-major,check-changeset-fixed.eslintover all 41 changed source files: 0 errors (4 pre-existingno-explicit-anywarnings on lines this PR did not touch).pnpm check:spec-floorsafter the build: exit 1 with twofloor-too-lowfindings (@object-ui/authMEMBERSHIP_ROLE_EXTERNAL,@object-ui/typesnavItemVisibleDataNeedsDataSourceagainst the FETCHED@objectstack/spec@18.0.0-next.x). Identical verdict with the unmodified gate script frommainon the same dists, and both symbols are referenced bymain's source, so this is pre-existing and not introduced here; CI'sspec-range-floors.ymlruns the gate with--cross-check, which is the measurement that counts.NOT MEASURED locally (CI's): the full
turbo run test,lintandtype-checkacross the workspace,check:published-dist, performance budgets, the react 19.3.0 runtime behaviour beyond building and the touched test files.Epic PM follow-up (commits 6f3d9d5, c464a4a, dde5e6d, 7537206, 2195392, ae7fcc5)
The first CI run on 11bc209 had five red shards, all in
scripts/__tests__:check-hand-rolled-comment-mask(theuiRoot()test stripped comments with its own regex),vite-objectstack-spec-dist(vendor literal moved),check-new-cross-file-line-citations(gate rooted off its scratch repository),check-lint-rule-coverage(ui-root.d.mtsunreached),docs-build-trigger(new turbo input not in the governed docs pathspec). The second run's Type Check job then failed on@object-ui/data-objectstack(TS7016, the twin deleted by 2195392 was needed); ae7fcc5 restores it with a ledger row. Fixed as described in items 3–5; verified locally afterpnpm run setup(the pin built):vitest run scripts/__tests__194 files passed / 2 skipped,vite-objectstack-spec-dist.test.ts84 passed,check-lint-rule-coverage+ui-roottests green with the twin,pnpm type-check:scriptsexit 0,turbo run type-check --filter=@object-ui/data-objectstack --filter=@object-ui/types --filter=@object-ui/core6/6 tasks,check:ui-root-lookupsOK.Acceptance notes
pnpm installregenerates the whole file); commit 2 holds the manifest hunks and the changeset.NODE_OPTIONS=prefix rather than the card's cross-platform wrapper: a root wrapper is a build input, and a new build input must be in the governed docs pathspec (item 4). The flag changes the heap, not the program, so turbo's cache verdicts are unaffected.pnpm-workspace.yamlglobs inscripts/objectstack-pin.mjs,release-github-releases.mjsand the workspace-reading tests;'.git'in two directory skip lists) is deliberate: those are not root lookups. Thepnpm-workspace.yamlDATA readers will need a path onceui/has no workspace file of its own; that belongs to the carrier sub-issue, and is listed in the report.--write-manifestswas not passed.Generated by Claude Code