test(react): Lock concurrent rendering GOALS on public APIs

[ntucker]

Fixes # .

Motivation

Lock the GOALS bullet Fully concurrent rendering on the latest React with tests against public @data-client/react (DataProvider, useSuspense, useCache, useLive, useController) plus React itself. No packages/* source changes and no changeset.

This is a new PR off master. It does not push to #4090, does not copy #4092 waiter / RSC-first contracts, and does not assert a universal “all live Context publishes preserve pending siblings.”

Existing probes establish two conditional cases, not one universal rule:

• Master’s stable, gate-before-consumer shape (publish through memoized DataStore while a sibling whose data consumer has not rendered past an outer gate is <!--$?-->) remains dehydrated on 19.2.3 / 19.3.0 and later hydrates in place.
• PR 4090’s two-pending live-publish shape client-renders the sibling on 19.2.3 and 19.3.0; startTransition did not rescue that composition.

Shape 1 is a green regression guard for the narrower gate-before-consumer statement on 19.x. Shapes 2 and 3 are causal controls whose outcomes are recorded per React version.

Solution

• __tests__/reactCommitProbe.ts + __tests__/streamingHarness.tsx: DevTools commit-priority probe (self-calibrated) and two-phase renderToPipeableStream / hydrateRoot replay. The Node server build is required so jsdom’s browser export on React 18 still exposes renderToPipeableStream; executeScripts hoists 18’s $RC function declarations onto the global.
• concurrent-probe-controls.web.tsx: uSES is Immediate even inside startTransition; plain setState is Normal; discrete setResponse is Immediate; recreating Suspense client-renders a pending boundary. On React 18 that control logs “This Suspense boundary received an update before it finished hydrating… wrap the original update in startTransition.” The recreate test expects that warning and drains it so the suite afterEach empty-console assert stays strict. Hydration and transition suites do not filter it. recordConsoleErrors is unchanged.
• concurrent-transition.web.tsx: startTransition over useSuspense keeps the committed UI, accepts urgent input, then commits B; isolated store publishes are non-Immediate; zero flushSync / useSyncExternalStore.
• concurrent-hydration.web.tsx: shape 1 green lock on 19.x; shapes 2–3 recorded per version.
• useCacheState.web.tsx: hoisted throwing useSyncExternalStore; React 19 use(StateContext); React 18 useContext labeled fallback.
• CircleCI unit_tests-19.3: delete root React resolutions, yarn up react@19.3.0, run the focused ReactDOM suites. The latest row stays on the 19.2.3 pin.

Simplify (same PR, no assertion or CI-row changes): shared replayRest / getHydrateRoot / probe calibration helpers; hydration streamPage + publishAOnTransition; collapsed identical shape 2/3 table rows.

Per-version results

Focused suites: packages/react/src/__tests__/concurrent-.*\\.web\\.tsx + useCacheState.web.tsx. 24 passed, 1 skipped on every version below. No it.failing.

React	Where	Result
16 / 17	CI unit_tests-^17	Suites describe.skip (no react-dom/client / useTransition). useCacheState 16/17 isolate still runs.
18.3.1	CI unit_tests-^18 (local yarn-up)	Recreate-Suspense client-renders pending B and logs the hydration-update warning (expected on 18 only). After expecting/draining that warning (2c592e7d30): probe-controls 4 passed (3.621 s); all concurrent suites 20 passed; useCacheState 4 passed, 1 skipped. Shape 1 transition publish: Normal commit, B stays pending, host title stays todo A. Shape 1 default-lane: title becomes v2, B is client-rendered.
19.2.3	repo pin; CI latest	4 suites, 24 passed, 1 skipped. Shape 1 green (identity, interactivity, in-place reveal). Probe-controls re-ran after the 18 warning allowance: 4 passed, 2.307 s.
19.3.0	new CI unit_tests-19.3 (local yarn-up)	4 suites, 24 passed, 1 skipped, 3.705 s. Same shape 1 green lock as 19.2.3.

The adapter-level flushSync assertion needs #4090 code and is out of scope.

Shape 2 / shape 3 recorded table (public DataProvider)

Version	Shape	pending after publish	B rendered after publish	streamed node adopted	fetchIds
18.3.1	2 read-then-suspend	B stays pending	no	yes	[]
18.3.1	3 two-pending	A and B stay pending	no	yes	[]
19.2.3	2 read-then-suspend	B stays pending	no	yes	[]
19.2.3	3 two-pending	A and B stay pending	no	yes	[]
19.3.0	2 read-then-suspend	B stays pending	no	yes	[]
19.3.0	3 two-pending	A and B stay pending	no	yes	[]

Composition difference vs #4090: on public master, the two-pending live-publish topology (shape 3) does not client-render B on 18.3.1, 19.2.3, or 19.3.0 (fetchIds: [], streamed node adopted). PR 4090’s adapter-shaped live HYDRATE did client-render the sibling (clientFetches: ["B"]). That difference is reported, not rewritten.

Open questions

None for this PR. Doc patch intents for layering/integrity/19.3 spike docs belong to other subagents.

  

[changeset-bot]

⚠️ No Changeset found

Latest commit: 2c592e7

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
docs-site	Ignored	Preview	Sep 18, 2026 4:17am UTC

[github-actions]

Size Change: 0 B

Total Size: 81 kB

ℹ️ View Unchanged

Filename	Size
examples/test-bundlesize/dist/App.js	1.46 kB
examples/test-bundlesize/dist/polyfill.js	307 B
examples/test-bundlesize/dist/rdcClient.js	10.9 kB
examples/test-bundlesize/dist/rdcEndpoint.js	8.07 kB
examples/test-bundlesize/dist/react.js	59.6 kB
examples/test-bundlesize/dist/webpack-runtime.js	784 B

_{compressed-size-action}

[github-actions]

Benchmark React

Details

Benchmark suite	Current: 2c592e7	Previous: 3fa47c8	Ratio
data-client: getlist-100	126.58 ops/s (± 4.5%)	137.94 ops/s (± 4.5%)	1.09
data-client: getlist-500	41.07 ops/s (± 6.0%)	44.05 ops/s (± 3.8%)	1.07
data-client: update-entity	322.58 ops/s (± 7.6%)	384.62 ops/s (± 9.3%)	1.19
data-client: update-user	312.5 ops/s (± 6.9%)	333.33 ops/s (± 8.6%)	1.07
data-client: getlist-500-sorted	43.01 ops/s (± 10.3%)	43.11 ops/s (± 9.4%)	1.00
data-client: update-entity-sorted	285.71 ops/s (± 8.8%)	312.5 ops/s (± 7.9%)	1.09
data-client: update-entity-multi-view	285.71 ops/s (± 6.7%)	327.96 ops/s (± 7.5%)	1.15
data-client: list-detail-switch-10	7.16 ops/s (± 6.4%)	9.9 ops/s (± 9.4%)	1.38
data-client: update-user-10000	72.46 ops/s (± 5.6%)	72.2 ops/s (± 14.8%)	1.00
data-client: invalidate-and-resolve	34.01 ops/s (± 5.4%)	37.67 ops/s (± 5.6%)	1.11
data-client: unshift-item	204.08 ops/s (± 5.5%)	219.81 ops/s (± 6.0%)	1.08
data-client: delete-item	270.27 ops/s (± 4.7%)	285.71 ops/s (± 3.8%)	1.06
data-client: move-item	161.29 ops/s (± 8.4%)	177.01 ops/s (± 7.9%)	1.10

This comment was automatically generated by workflow using github-action-benchmark.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.84%. Comparing base (44cb7e6) to head (2c592e7).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #4094   +/-   ##
=======================================
  Coverage   97.84%   97.84%           
=======================================
  Files         156      156           
  Lines        3057     3057           
  Branches      612      612           
=======================================
  Hits         2991     2991           
  Misses         18       18           
  Partials       48       48           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ntucker]

FOLLOW_UP (Staff Architect) — does not block merge after Ready / /simplify.

Solid GOALS lock: public @data-client/react only, no packages/* source, honest conditional cases (shape 1 green vs shapes 2–3 recorded). Harnesses + self-calibrated DevTools probe are the right architecture; 19.3 CI row is correctly focused.

After land, please consider:

1. Drop duplicate calibration — concurrent-hydration.web.tsx and concurrent-transition.web.tsx still copy the uSES-Immediate / plain-setState-Normal blocks that concurrent-probe-controls.web.tsx already owns post-0acb4b2. Keep one suite as the probe control; let the others rely on it (or a shared beforeAll helper).

2. Soften B:0 / B:1 coupling — Shape 1’s green path pins React streaming Suspense template ids. If a future React minor renumbers boundaries, prefer asserting dehydrated pending via <!--$?--> + “B not in renders / no fetch” where that still proves the contract.

3. Retire unit_tests-19.3 when latest pins ≥19.3 — Avoid a permanent extra matrix job once the repo pin catches up.

4. DevTools priority maintenance — Treat Immediate/Normal remaps as probe recalibration, not a data-client product break (already self-calibrated; call that out in the spike/docs follow-ups).

No CHANGE_THIS_PR. Draft is fine to review; merge gate remains Ready after /simplify + green checks.

[ntucker]

FOLLOW_UP (no change required on this PR)

Direction looks right for a public-API concurrent-rendering goal lock: shape 1 as the green regression guard, shapes 2/3 as per-minor recorded controls (not #4090 adapter contracts), shared probe + streaming harness after simplify, and a focused unit_tests-19.3 row that deletes resolutions rather than fighting the 19.2.3 pin.

When you next touch this area (not blocking Ready/merge):

1. Soften B:0 / B:1 — prefer discovering pending template[id] from the installed shell (or fallback-text + <!--$?-->) so a React HTML-protocol id tweak doesn’t brick the goal suite.
2. DevTools hook — consider installing __REACT_DEVTOOLS_GLOBAL_HOOK__ from ReactDOM scripts/testSetup.js so suites can drop the import/order ritual and stay safe if a jest.mock('react-dom') ever requireActuals client early. Keep per-suite calibration fail-loud either way.
3. Fixture drift — shared Todo/Endpoint/mockInitialState helper; keep Immediate/Normal calibration primarily in concurrent-probe-controls unless a file’s mocks need a local smoke.
4. 19.3 → pin — when 19.3 becomes latest, widen the CI row past the concurrent path glob and add shape 2/3 table rows for each new React minor.