[YevheniiKotyrlo]

Problem

The example keeps its UI frame loop running after the app goes to the background: worklets 0.8.3's WorkletsModule never registers for host lifecycle events (software-mansion/react-native-reanimated#10196, fixed in worklets 0.12.0). On a Mi MIX 3 (Android 10), the Reanimated screen keeps 20–22% of a core busy in the background. Every worklets release with the fix needs React Native 0.83 or later, and the example is on 0.81.4.

Solution

• The example moves to React Native 0.83.10 with React 19.2.3, the @react-native/* 0.83.10 tooling, react-native-test-app 5.4.13 and Gradle 9.0.0, and to Reanimated 4.6.0 and worklets 0.12.2; the docs and packages/webgpu's dev dependencies follow. The library's peer ranges are unchanged.
• apps/expo-webgpu keeps its dependencies in its own node_modules (installConfig.hoistingLimits: "workspaces"). @react-native/metro-config resolves react-native/Libraries/Core/InitializeCore from its own location, so the root has to hold the example's React Native; with the Expo template's 0.86.3 hoisted there instead, the example's bundle failed ("Unable to determine event arguments for onModeChange"). React Native, React, Reanimated and worklets now have one copy each at the root.
• The docs' twoslash samples resolve their imports from the docs app (vfsRoot), which declares the packages they import, as it already does typegpu; from the repo root, the vision-camera sample failed with TS2307.
• TransparencyDemo uses StyleSheet.absoluteFill: React Native 0.83's types no longer declare absoluteFillObject.
• The Podfiles stop copying their fmt and RCT-Folly podspecs over React Native's. React Native 0.83 ships fmt 12.1.0 itself (Bump fmt to 12.1.0 to fix xcode 26.4 react/react-native#56099) and the same Folly podspec, and pod install without the copy resolves both with the checksums main's lock records.

react-native-skia's example gets the same upgrade in wcandillon/react-native-skia#4254.

Verification

On a Mi MIX 3 (Android 10), release builds, the Reanimated screen keeps 20.4–21.9% of a core busy after HOME on main and 0.0–0.1% with this branch (measurement).

yarn lint 0 (rewrites nothing) · yarn tsc 0 · package and web-template builds 0 · yarn test:plugin 0 · yarn check-dawn 0 · yarn build:docs 0

CI's Android e2e step on an API 35 emulator, this branch and main one after the other: 219 of 221 passed on both, the YUV cases of ExternalTextureColorSpace and NativeBufferWrapper failing on both. CI's iOS xcodebuild (Debug, iOS Simulator SDK) after pod install: BUILD SUCCEEDED, built without debug symbols to fit the disk of the Mac I built on.

Known limits

The macOS Podfile.lock is left as on main, which still records React Native 0.81.4, Reanimated 4.2.1 and worklets 0.7.1; CI builds no macOS app.

[YevheniiKotyrlo]

Device measurement — before / after

main — after HOME, the Reanimated screen keeps 20.4–21.9% of a core busy: worklets 0.8.3 never pauses its UI frame loop.

This PR — 0.0–0.1% after HOME; in the foreground the two builds are within each other's spread.

% of a core, Mi MIX 3 (Android 10, API 29)	main 46f42a6b	this PR 210c4b35
background, six 10 s windows	21.4 (20.4–21.9)	0.0 (0.0–0.1)
foreground, one 10 s window per install	25.9 / 24.8	22.0 / 24.5

Release builds of the example opening Reanimated's UI-thread screen on an opaque canvas, installed in the order main, this PR, this PR, main. Each install reads the app process's user and system time from /proc/<pid>/stat over one 10 s window in the foreground, then three 10 s windows after HOME, as a share of one core; the background column is the median and range of each build's six windows.

[github-actions]

🎉 This PR is included in version 0.16.1 🎉

The release is available on:

• v0.16.1
• GitHub release

Your semantic-release bot 📦🚀