## Summary:
The C++ spec header generated for a C++ TurboModule names each method's static invoker `__<method>` and registers it under that name in `methodMap_`. Names starting with a double underscore are reserved in C++, so a library that builds with `-Wreserved-identifier` under `-Werror` can't include its own generated `<Module>JSI.h`. The invokers are now named `invoke_<method>`. They are private to the spec class, and the `methodMap_` keys, arity and calls into the module stay the same.
Fixes#59843
## Changelog:
[GENERAL] [FIXED] - C++ TurboModule codegen no longer names method invokers with reserved identifiers
## Test Plan:
Updated the `GenerateModuleH` snapshots; `jest packages/react-native-codegen` passes. clang++ with `-Werror -Wreserved-identifier` rejects `__add` and accepts `invoke_add`.
Differential Revision: D132179193
Pull Request resolved: https://github.com/react/react-native/pull/59844
Summary:
X-link: https://github.com/react/metro/pull/2224
@public
Upgrade `flow-parser`, `flow-estree`, `flow-eslint`, `flow-transform` and `flow-api-translator` to 0.340.0, and adapt the code generators, codemods, lint rules and parser adapters that read the ESTree output those releases produce.
Four shapes moved since the versions in use, and each one silently changed behavior rather than failing loudly:
- Generic type annotations carry their arguments in `typeArguments` instead of `typeParameters`, so consumers that still read the old field dropped every type argument. Readers now take the canonical field whenever the node carries it, including when it is `null`, and fall back to the previous field otherwise, so the same source keeps working with the older releases that remain installed elsewhere.
- Enum declarations expose a single `EnumBody` carrying an explicit type instead of one node type per value kind. Enum readers accept both and resolve the value kind the same way, so the supported enums and the existing rejection of boolean and symbol enums are unchanged.
- `JSXAttribute.name` widened to also allow a namespaced name, so transforms that compare an attribute name against a list now match only plain identifiers and leave namespaced attributes alone, which is what they already did at runtime.
- Object type properties, array elements, import phases and several other fields are now nullable or newly required. Readers guard them and report the existing unsupported-input errors instead of failing on a missing property, and import declarations are built from constant parser-owned templates so one source satisfies both the old and the new builder.
The lint rules, the declaration generators and the `jscodeshift`/recast boundary are adapted the same way, and the published type mirrors for the lint and transform packages are refreshed.
Changelog: [Internal]
Reviewed By: SamChou19815, panagosg7
Differential Revision: D131489493
Pull Request resolved: https://github.com/react/react-native/pull/59983
Summary:
Labels unlabeled dictionary keys: `{[string]: T}` becomes `{[key: string]: T}`. In TypeScript, `{[string]: T}` is a computed property, not an index signature, so this prepares the code for `experimental.require_dictionary_key_label`, which requires the label. The label is documentation only and does not change any type.
Generated by the `flow/nameIndexerKeys` codemod, run only on the keys that the flag reports, so computed keys like `{['a']: T}` are left alone. Keys inside comment types, and files whose formatting the codemod's printer would change, were labeled with a plain text insertion instead.
Changelog: [Internal]
Differential Revision: D130071787
Pull Request resolved: https://github.com/react/react-native/pull/59818
Summary:
TypeScript declares `parseInt(string: string, radix?: number)` and `isNaN(number: number)`, and Flow is adopting the same signatures. Calls that pass another type rely on the function's implicit conversion and start to error.
Wrap each such argument in `String(...)` (`parseInt`) or `Number(...)` (`isNaN`), the conversion the function already applies, so behavior doesn't change. The calls were found from Flow's errors and rewritten by a codemod. 8 arguments in 6 files in React Native and Metro.
Changelog: [Internal]
Differential Revision: D129644365
Pull Request resolved: https://github.com/react/react-native/pull/59799
Summary:
Gives every unnamed parameter in a Flow function type annotation a name: `v` for a lone parameter, `v<index>` otherwise, and `rest` for a rest parameter. Type-only change with no runtime behavior change.
Changelog: [Internal]
Reviewed By: panagosg7
Differential Revision: D129441819
Pull Request resolved: https://github.com/react/react-native/pull/59790
Summary:
Invalid Android pointerEvents values fall back to `AUTO` and log a warning instead of throwing. Null still resets the default silently, and valid case/hyphen variants retain their existing behaviour.
Changelog:
[Android][Fixed] - Default invalid pointerEvents values to auto with a warning
Differential Revision: D129603915
Pull Request resolved: https://github.com/react/react-native/pull/59794
## Summary:
This is a follow-up to #59414, suggested by @javache in [its review](https://github.com/facebook/react-native/pull/59414#discussion_r4399176566): keep a method's `jmethodID` in the same per-module cache as its parsed JNI signature, and index that cache by method name, instead of having two caches.
Before this change, a Java TurboModule method's ID was cached by its caller: every generated host function kept a `static jmethodID cachedMethodId` and passed it to `invokeJavaMethod` by reference, and `JavaInteropTurboModule` kept a `methodIDs_` vector for the same purpose. `JavaTurboModule` then used that ID as the key of its `MethodInfo` cache (#59414).
With this change:
- `JavaTurboModule::MethodInfo` also holds the method's `jmethodID`, and `methodInfos_` is keyed by the method name (a `std::string`, looked up through a `std::string_view` without allocating). On a method's first call, `getMethodInfo` looks the method up with `GetMethodID` and parses its signature. A signature that doesn't match still turns into a JS error (redbox) before anything is called, and nothing is cached for it.
- `invokeJavaMethod` gets an overload that takes the method name and JNI signature as C strings and no `cachedMethodID`. Generated code passes string literals, so a call no longer builds two `std::string`s; the JNI signature is usually longer than the small-string buffer, so that was a heap allocation on every call.
- The previous overload (`const std::string&` name and signature, `jmethodID&`) is kept, for code generated by earlier codegen versions and for modules that call it directly, and marked `[[deprecated]]`. It forwards to the new one; `cachedMethodID` is no longer read or written. Libraries that ship pre-generated JNI code build it with `-Wall -Werror` (the codegen CMake template applies `target_compile_reactnative_options`), so their builds will fail on the deprecation warning until they regenerate that code.
- The codegen template for Java module JNI files (`GenerateModuleJniCpp`) no longer emits `static jmethodID cachedMethodId`, and `JavaInteropTurboModule` drops `methodIDs_`.
Besides one cache instead of two, every generated host function gets smaller: RNTester's `libreactnative.so` (which includes `FBReactNativeSpec`) is 61 KB smaller and its `libappmodules.so` 9.4 KB smaller.
## Changelog:
[ANDROID] [CHANGED] - Java TurboModules cache each method's ID with its parsed signature, by name, instead of in every generated host function
[ANDROID] [DEPRECATED] - `JavaTurboModule::invokeJavaMethod` with a `cachedMethodID` argument; use the overload without it
## Test Plan:
- `yarn jest packages/react-native-codegen` passes (64 suites, 3,130 tests); the `GenerateModuleJniCpp` snapshots are updated for the template change.
- `node scripts/clang-format.js --check` reports no changes for the touched C++ files.
- C++ API snapshots: the new `invokeJavaMethod` overload is added to the four `ReactAndroid*Cxx.api` snapshots by hand, where the generator sorts it; `scripts/cxx-api` can't run on current `main` locally (its codegen step fails with `SyntaxError: Unexpected token '{'` while processing `FBReactNativeSpec`, without this change too).
- `./gradlew :packages:rn-tester:android:app:assembleRelease -PreactNativeArchitectures=arm64-v8a` succeeds with this change; RNTester's generated `FBReactNativeSpec` and app module code use the new overload.
- **On a device.** Two RNTester release APKs built from the same local checkout (`main` at f7de20301c with local benchmark screens), without and with this change, installed alternately on a moto g57 power (Android 16), three installs each with two launches. A Playground screen times synchronous calls to `NativeSampleTurboModule` (20,000 calls each, 5,000 for void, callback and promise calls), median of the six launches:
| Call (JNI signature) | before | this PR | change |
|---|--:|--:|--:|
| `getBool` (`(Z)Z`) | 1.10 µs | 1.15 µs | +4% |
| `getNumber` (`(D)D`) | 1.46 µs | 1.53 µs | +5%, ranges overlap |
| `getString` (`(Ljava/lang/String;)Ljava/lang/String;`) | 2.05 µs | 1.92 µs | −6% |
| `getArray`, 3 items | 10.34 µs | 10.06 µs | −3% |
| `getObject`, 3 keys | 9.90 µs | 9.58 µs | −3% |
| `getObject`, 20 keys | 35.67 µs | 35.28 µs | −1% |
| `getValue(number, string, object)` | 13.94 µs | 14.21 µs | +2%, within noise |
| `voidFunc` | 4.55 µs | 4.59 µs | within noise |
| `getValueWithCallback` | 8.54 µs | 7.99 µs | −6.5% |
| `getValueWithPromise` | 18.26 µs | 17.38 µs | −5% |
Calls whose JNI signature is longer than the small-string buffer no longer allocate it, and are 5–6% cheaper. The cheapest calls, whose short signatures never allocated, now hash the method name instead of a pointer, about 50 ns more. Native library sizes are from the same two APKs.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Differential Revision: D128349302
Pull Request resolved: https://github.com/react/react-native/pull/59619
Summary:
Test only change. RN's own Jest transform (`jest/preprocessor.js`, for this repo's own tests, not public) doesn't include `babel-plugin-jest-hoist`, so `jest.mock` calls run where they're written. Several tests only saw their mocks because inline requires deferred the `require` of the module under test (or of its dependencies) until first use.
That only works for destructured requires like `const {foo} = require('./foo')` while the Babel preset lowers destructuring, which in this config it currently does only as a side effect of lowering generators with regenerator. https://github.com/react/react-native/pull/59303 removes that, and 114 tests across these 16 suites fail.
This makes the tests install their mocks before the modules under test load, so they pass either way, without diverging Jest's transform from the app's:
- Require modules under test after the `jest.mock` calls that affect them.
- Use mock factories instead of spies or automocks where the code under test binds a function when it loads, or loads an ESM-only package (`octokit/rest`).
- Re-require modules after `jest.resetModules()` where tests reset the registry (`publish-npm`, `Animated`).
The release script tests in particular are dangerous when their mocks silently stop applying - locally, they ran the real version bump against the checkout.
## Changelog:
[INTERNAL] [CHANGED] - Don't rely on inline requires to apply mocks in JS tests
Pull Request resolved: https://github.com/react/react-native/pull/59307
Test Plan: The full Jest suite passes on this PR (6,284 tests), and with https://github.com/react/react-native/pull/59303 applied on top (6,285), both under Node 24 with `--no-cache`. Without this PR, https://github.com/react/react-native/issues/59303 fails 114 tests in 16 suites, matching its CI.
Reviewed By: christophpurrer
Differential Revision: D126757286
Pulled By: javache
fbshipit-source-id: ce39217579aa9805f74786ee9b669bf411e97191
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/59164
X-link: https://github.com/react/metro/pull/2075
Upgrade `flow-parser`, `flow-estree`, `flow-eslint`, `flow-transform`, and `flow-api-translator` across the xplat and AR/VR JavaScript tooling consumers. Refresh dependency locks, installed runtime payloads, and shared offline archives while preserving the independently pinned Fox toolchains and locally maintained type shims.
Build on D125751591's `flow-remove-types` upgrade. Keep Gauntlet in its own child because it has an independent release boundary.
Support reviewed, exact-root Yarn installation options in the upgrade policy. Recharts' forced macOS `fsevents` resolution needs platform checking disabled and optional packages omitted on Linux; no other installation roots or dependency-verification checks are exempted.
The installation-policy correction is an intentional prerequisite for executing the supplied playbook on Linux, not a separate feature rollout. The unchanged 0.335 Recharts dependency installation reproduced the forced macOS fsevents failure, including with --ignore-optional alone; the exact-root --ignore-platform/--ignore-optional combination passes without changing the manifest or lock. Keeping this tested, allowlisted exception in source makes the upgrade reproducible while preserving dependency/version checks and the other installation roots.
Changelog: [internal] Upgrade xplat and AR/VR Flow JavaScript tooling to 0.336.0.
Reviewed By: SamChou19815
Differential Revision: D125765307
fbshipit-source-id: 39a67c3f295bce809af89d1ee53bc51876214870
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/59094
With `experimental.function_and_object_type_is_no_longer_any`, Flow resolves `Object` and `Function` to the builtin class types instead of `any`.
Changelog: [Internal]
Reviewed By: panagosg7
Differential Revision: D125312021
fbshipit-source-id: 1646441aa3c2728313cf56b347d04e62d69712e4
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/59086
Flow will soon give `Object` the same meaning as TypeScript's `Object`, rather than treating it as `any`. React Native sources that use `Object` as an untyped escape hatch need to move off it. But the generated TypeScript API (`ReactNativeApi.d.ts`) relies on Flow `Object` mapping to TS `Object`, and it should keep emitting `Object`.
This adds a `LegacyObject` global type to bridge the two:
- Flow: `flow/LegacyObject.js` declares `LegacyObject` as `any`, next to `Stringish`.
- TypeScript: build-types replaces `LegacyObject` with `Object` before translation, via a new `replaceLegacyObjectWithObject` pre-transform that mirrors `replaceStringishWithString`.
- Codegen: native module specs accept `LegacyObject` as a generic object, like `Object` and `UnsafeObject`.
`CodegenTypes.UnsafeObject`, the event object types in `EventEmitter` and `NativeEventEmitter`, `PushNotificationIOS`, and the native module specs that used bare `Object` now use `LegacyObject`. Neither the generated TypeScript API nor the codegen schema changes. The codegen parser's internal AST aliases (`PropAST`, `ASTNode`) switch to plain `any`.
Changelog: [Internal]
Reviewed By: panagosg7
Differential Revision: D125245815
fbshipit-source-id: 98d9ded6523186a7efbcb8e6d1b25b0c6369307b
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/59051
X-link: https://github.com/react/metro/pull/2063
Upgrade `flow-parser`, `flow-estree`, `flow-eslint`, `flow-transform`, and `flow-api-translator` to 0.335.0 across `xplat/js` and `arvr/js`.
(Excluding the Fox toolchains in `tools/fox`, `tools/fox-haas-bundler`, and `tools/react-fox`).
Refresh 39 dependency manifests, 18 Yarn lockfiles, the recharts npm lockfile, checked-in runtime packages, Flow type mirrors, and the five shared offline archives.
Gauntlet's upgrade is kept in the stacked diff D125076658 because its framework has an independent release system.
Use canonical Yarn's hoisted root parser instead of redundant nested rpc/jscodeshift copies. Repair the legacy codegen lock's missing `nullthrows@^1.1.1` entry for its existing dependency declaration, without changing that declaration. Bootstrap GazeHands' declared build dependencies before CI validates its generated outputs, matching the existing build script.
Teach license lint to recognize the existing MIT headers in imported Flow type mirrors rather than rewriting them as internal headers; add narrow path assertions.
Changelog: [Internal]
Reviewed By: SamChou19815, panagosg7
Differential Revision: D125071452
fbshipit-source-id: 85cccf5242913fc559030d7e43f9a7266a54425a
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/58730
Codegen output is compiled in consumer targets, so it now includes React Native through module umbrella headers (`<React/RendererCore.h>`, `<React/Graphics.h>`, `<React/NativeModuleCore.h>`, etc.) instead of fine-grained paths, which hard-error under `RN_STRICT_API`.
The one exception is `<react/renderer/debug/debugStringConvertibleUtils.h>`, which `Props.cpp` still includes directly alongside `<React/RendererDebug.h>`. It is a private header not re-exported by any umbrella, but it is only emitted together with the `getDebugProps` implementation, which is generated solely for React Native's own core components library (`FBReactNativeSpec`). That library is built as part of React Native (`RN_BUILDING`), so the private-header guard does not apply.
Changelog:
[General][Changed] - Codegen-generated C++ includes React Native through module umbrella headers
Reviewed By: cortinico
Differential Revision: D122317911
fbshipit-source-id: 53b3275acf148c665c27470a929cd0397bdce6d9
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/58762
This bumps the Flow tooling packages from 0.331.0 to 0.333.0, matching `flow-bin`. From 0.332.0, `flow-api-translator` translates `declare module.exports` to TypeScript's `export =` (https://github.com/facebook/flow/pull/9491), which https://github.com/react/metro/pull/1996 uses to generate definitions for CommonJS modules.
It also declares `babel/parser` as a root dev dependency. The API snapshot script has imported it since https://github.com/react/metro/pull/1988, but it only resolved because other packages hoist it. The range matches theirs, so it doesn't change the lockfile.
Changelog: [Internal]
X-link: https://github.com/react/metro/pull/1998
Test Plan: Generated `.d.ts` (gitignored) and all 14 API snapshots are unchanged on the new versions.
Reviewed By: javache
Differential Revision: D122383746
Pulled By: vzaidman
fbshipit-source-id: f9d47c8e8fb60b10302b7ce5f8378db6ba59b786
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/58651
Generated view configs currently emit repeated property accesses from React Native entry-point requires and route commands through a renderer namespace. Consolidate imports by source and expose `dispatchCommand` directly from the unstable internals entry point, reducing generated Hermes bytecode while preserving existing stable root exports. Metro's inline-require transform would otherwise expand the consolidated unstable-internals binding back into repeated `require(...)` calls, so keep that dependency non-inlined while leaving other generated imports eligible for lazy loading. Make the unstable `HMRClient` export follow React Native's normal development/production selection so production consumers receive `HMRClientProdShim` instead of including HMR tooling.
Changelog:
[General][Fixed] - Reduce generated native component bundle overhead
Reviewed By: javache
Differential Revision: D121383217
fbshipit-source-id: 3b81db620bcff172d43d0493dd4ca22b30f305d6
Summary:
X-link: https://github.com/react/metro/pull/1962
Pull Request resolved: https://github.com/react/react-native/pull/58641
This picks up https://github.com/babel/babel/pull/18241, fixing parsing of arrow function params inside ternaries with Flow (eg `cond ? async (...args) => {} : null`).
(I remember hitting this one internally and not realising where it came from - IIRC it also applies to arrows within conditional types)
The update brings `babel/types` 7.29.7 -> 7.29.8, requiring regeneration of generated flow types. ~~**The `xplat/js/yarn.lock` version will likely need updating to match, or the Babel types sync test will fail**~~ - I've updated specifiers to bring everyone onto that version.
Changelog: [Internal]
X-link: https://github.com/react/metro/pull/1958
Test Plan: CI
Reviewed By: javache
Differential Revision: D121188416
Pulled By: vzaidman
fbshipit-source-id: 60f9c3514b5d6018825423644c47e66b3610b652
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/58603
Generate view configs and Codegen fixtures using supported React Native entry points or values derived from their exports. Generated view config modules can be emitted into arbitrary consumer package locations, so they cannot use a package-relative path back into React Native. Keep `ConditionallyIgnoredEventHandlers` centralized and expose it from the explicitly unstable compatibility entry point instead of duplicating its platform-specific semantics in generated output.
Changelog: [Internal]
Reviewed By: cipolleschi
Differential Revision: D120717571
fbshipit-source-id: 9f3ea06d393eb8702bbb3ef2767c2d7c5fde1315
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/58558
Generate TurboModule interfaces using the public Bridging C++ entry point so generated consumers remain compatible with the strict API boundary.
Export the bridging module include root from its CMake target so source-built consumers can resolve <React/Bridging.h>.
Changelog:
[Internal]
Reviewed By: cortinico
Differential Revision: D120357994
fbshipit-source-id: 072d67197daf19acffbb66c110833a10a7094cfb
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/58572
Migrate Codegen and compatibility fixtures from React Native deep imports to the public `react-native` entry point. Reference codegen-only types through the exported `CodegenTypes` namespace.
Changelog: [Internal]
Reviewed By: christophpurrer
Differential Revision: D120537584
fbshipit-source-id: f97ee85fb5f16d23a998d25c1544134ee08bc3cc
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/58539
Update Fabric component codegen to consume the public View module through <React/View.h> for generated props, event emitters, shadow nodes, and dimension conversions. Refresh generated snapshots and checked-in popup-menu bindings, and add strict consumer compile coverage that does not use RN_BUILDING.
Make the View umbrella compatible with force-CXX platform configurations discovered by the broader generated-header coverage. On Android, gate the ViewPropsInterpolation rawProps workaround on both ANDROID and RN_SERIALIZABLE_STATE, since rawProps only exists when serializable state is enabled; this preserves existing Android behavior without enabling the runtime/ABI feature globally. For Android, macOS, and Windows platform types, rely on the Buck-selected HostPlatformViewProps and HostPlatformViewEventEmitter headers instead of compiler platform macros: the native implementations already expose NativeDrawable, HostPlatformViewEvents, KeyEvent, MouseEvent, and WindowsViewEvents, while force-CXX builds intentionally do not export those platform-only headers.
Changelog:
[Internal].
Reviewed By: cipolleschi
Differential Revision: D119490247
fbshipit-source-id: 7266baa57f5629e56134b65f2a6b591cb1e0b155
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/58554
Generate component descriptor headers through the public <React/ComponentRegistry.h> umbrella instead of directly including ComponentDescriptorProviderRegistry.h. Refresh codegen snapshots and the checked-in popup-menu binding, and export the ComponentRegistry dependency needed by that public generated header.
Changelog:
[Internal].
Reviewed By: cipolleschi
Differential Revision: D120317127
fbshipit-source-id: fedea53bbe65e22ae53ed00a7fde787b2581bfab
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/58486
Generate TurboModule interfaces using the public Bridging C++ entry point so generated consumers remain compatible with the strict API boundary.
Changelog:
[Internal]
Reviewed By: cipolleschi
Differential Revision: D119658518
fbshipit-source-id: 42b2c23c8a1c4264704a916b0c4e19660110a86d
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/58391
Regenerate API snapshots after the bump. `flow-api-translator` picks up the fix in https://github.com/facebook/flow/pull/9486, which stops treating `async` functions with no explicit return annotation as `void` — updating `DependencyGraph#end` and `MetroServer#end` to their correct `Promise<void>` return type in `packages/metro/API.md`.
Changelog: [Internal]
Changelog: **[Types]**: `DependencyGraph#end` and `MetroServer#end` are now correctly declared as returning `Promise<void>` rather than `void`
X-link: https://github.com/react/metro/pull/1909
Test Plan:
```
yarn typecheck
yarn lint
yarn build
yarn run verify-api-snapshots
```
All pass.
Reviewed By: mweststrate
Differential Revision: D119090467
Pulled By: javache
fbshipit-source-id: 21b65f90ed511e0cb035917eaf1b0ca399ece855
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/58063
Codegen's `EventEmitter<T>` support for TurboModules diverged between platforms
in two ways:
- The explicit number types `Double`, `Float` and `Int32` worked as emitter
payloads on Android and C++, but iOS rejected them at codegen time (plain
`number` worked). iOS now maps all of them to `NSNumber *_Nonnull`, matching
how plain `number` is already emitted.
- `ArrayBuffer` was rejected on Android and iOS, but the C++ generator silently
accepted it and produced a `jsi::ArrayBuffer` emitter. `ArrayBuffer` is not
emittable on any platform — Android emitters always carry a `folly::dynamic`
payload, which cannot hold raw bytes — and the schema type
`NativeModuleEventEmitterBaseTypeAnnotation` already excluded it. It is now
rejected everywhere.
`ArrayBuffer` payloads are rejected in the shared parser, so Flow and TypeScript
specs produce the same error, and the three generators keep an equivalent guard
so schemas that are constructed without going through the parser fail the same
way. `ArrayBuffer` remains supported as a method argument and as a synchronous
return value.
Changelog:
[iOS][Added] - Support `Double`, `Float` and `Int32` payloads for TurboModule `EventEmitter`s
[General][Breaking] - Reject `ArrayBuffer` as a TurboModule `EventEmitter` payload on all platforms
Reviewed By: GijsWeterings
Differential Revision: D116952150
fbshipit-source-id: e148a394c4916b3ebfbd3f559b190a5910b29de1
Summary:
`GeneratePropsJavaDelegate` emits the default value for an optional `DoubleTypeAnnotation` prop with an `f` (float) suffix, but the setter it calls takes a `double`:
```js
// GeneratePropsJavaDelegate.js
case 'DoubleTypeAnnotation':
if (prop.optional) {
return `value == null ? ${typeAnnotation.default}f : ((Double) value).doubleValue()`;
}
```
```js
// GeneratePropsJavaInterface.js — the setter this is passed to
case 'DoubleTypeAnnotation':
return 'double value';
```
The `f` literal is parsed as a `float`, then widened back to `double` at the call site, so the default silently arrives rounded to float precision. It compiles without a warning, which is why it has gone unnoticed.
This is visible in the repo's own snapshots today. Same fixture, same prop, two platforms:
| Generator | Output for `blurRadius3?: WithDefault<Double, 2.1>` |
| --- | --- |
| `GeneratePropsH` (C++) | `double blurRadius3{2.1};` |
| `GeneratePropsJavaDelegate` | `value == null ? 2.1f : ((Double) value).doubleValue()` |
So a component using the C++ renderer gets `2.1` and the same component on the Android delegate path gets `2.0999999046325684`.
Compiling the generated shape confirms it:
```java
static void setBlurRadius3(double value) { System.out.println(value); }
setBlurRadius3(value == null ? 2.1f : ...); // 2.0999999046325684
setBlurRadius3(value == null ? 123456789f : ...); // 1.23456792E8
```
The integer case is the clearest damage: any `Double` default above 2^24 is not representable as a `float`, so `123456789` arrives as `123456792`. Fractional defaults are wrong from the first value that is not a dyadic rational — `0.1`, `2.1`, and the fixture's own `0.001` all shift.
`FloatTypeAnnotation` on the line below is correct — it targets a `float` setter, so `f` is right there. Only the `Double` branch has the wrong suffix, which is consistent with it having been copied from the `Float` branch.
Changed to a `d` suffix, mirroring the existing `f` for float rather than dropping the suffix entirely, so the literal's type is stated rather than left to numeric promotion.
`DoubleTypeAnnotation` is never boxed by codegen (it is always a primitive `double`, defaulting to `Double.NaN` when non-optional), so there is no nullable-double path to consider.
## Changelog:
[ANDROID] [FIXED] - Fix `Double` prop defaults being rounded to float precision in generated `ViewManager` delegates
Pull Request resolved: https://github.com/react/react-native/pull/58070
Test Plan:
The generator is snapshot-tested, so the snapshot is the test. Updating it changes only the `Double` component and leaves the `Float` component untouched:
```diff
public class DoublePropNativeComponentManagerDelegate<T extends View, U extends ...
case "blurRadius2":
- mViewManager.setBlurRadius2(view, value == null ? 0.001f : ((Double) value).doubleValue());
+ mViewManager.setBlurRadius2(view, value == null ? 0.001d : ((Double) value).doubleValue());
case "blurRadius3":
- mViewManager.setBlurRadius3(view, value == null ? 2.1f : ((Double) value).doubleValue());
+ mViewManager.setBlurRadius3(view, value == null ? 2.1d : ((Double) value).doubleValue());
case "blurRadius4":
- mViewManager.setBlurRadius4(view, value == null ? 0f : ((Double) value).doubleValue());
+ mViewManager.setBlurRadius4(view, value == null ? 0d : ((Double) value).doubleValue());
case "blurRadius5":
- mViewManager.setBlurRadius5(view, value == null ? 1f : ((Double) value).doubleValue());
+ mViewManager.setBlurRadius5(view, value == null ? 1d : ((Double) value).doubleValue());
case "blurRadius6":
- mViewManager.setBlurRadius6(view, value == null ? 0f : ((Double) value).doubleValue());
+ mViewManager.setBlurRadius6(view, value == null ? 0d : ((Double) value).doubleValue());
```
```
$ yarn jest packages/react-native-codegen
Test Suites: 64 passed, 64 total
Tests: 3122 passed, 3122 total
Snapshots: 1277 passed, 1277 total
$ yarn flow-check
Found 0 errors
$ npx eslint packages/react-native-codegen/src/generators/components/GeneratePropsJavaDelegate.js
(no output)
```
`0d`, `1d` and `2.1d` are valid Java double literals; the proof snippet above compiles and runs under `java Proof.java`.
No committed generated sources carry the old output — `grep -rn "f : ((Double) value).doubleValue()" --include='*.java' --include='*.kt'` returns nothing outside the snapshot.
Reviewed By: christophpurrer
Differential Revision: D117082006
Pulled By: fabriziocucci
fbshipit-source-id: 7b61b16d36eea0dcedb9c2b047bf57d23c0132ca
Summary:
Fixes https://github.com/react/react-native/issues/57956.
A spec containing a type alias whose name matches the aliased `CodegenTypes` member hangs codegen forever:
```js
import type {CodegenTypes} from 'react-native';
type Double = CodegenTypes.Double;
```
`getResolvedTypeAnnotation` resolves `Double` to the alias's RHS (`CodegenTypes.Double`), but `getTypeAnnotationName` strips the namespace qualifier before the local-alias lookup — so the RHS resolves straight back to the same alias, forever. This hangs the Node process of anything driving the parser, most visibly ESLint via `react-native/eslint-plugin-specs`. Contrary to the issue's initial isolation, **both** the Flow and the TypeScript parsers hang (verified with a subprocess-timeout harness against both).
Two guards in each parser's resolution loop:
1. A qualified name (`CodegenTypes.Double` / `TSQualifiedName`) names a namespace member, never a local alias — stop local resolution there. Downstream translation already handles qualified names, so the alias now resolves to the correct primitive (`DoubleTypeAnnotation`).
2. Track resolved alias names, so genuinely cyclic aliases (`type A = B; type B = A;`) stop resolving and fail through the existing `UnsupportedGenericParserError` (`Unrecognized generic type 'A'`) instead of hanging.
## Changelog:
[GENERAL] [FIXED] - Codegen no longer hangs on type aliases that shadow CodegenTypes member names
Pull Request resolved: https://github.com/react/react-native/pull/57986
Test Plan:
- New `NAMESPACED_NATIVE_MODULE_WITH_LOCAL_TYPE_ALIASES` fixture in both the Flow and TypeScript module snapshot suites — before the fix these hang the test runner; after it they produce the correct schema (aliased `Double`/`Float` resolve to `DoubleTypeAnnotation`/`FloatTypeAnnotation`, snapshots included, cross-parser consistency suite passes).
- `jest packages/react-native-codegen/src/parsers`: 13 suites, 1965 tests, 186 snapshots — all passing.
- `jest packages/eslint-plugin-specs`: passing.
- Verified `type A = B; type B = A;` now throws `UnsupportedGenericParserError` instead of hanging.
Reviewed By: javache
Differential Revision: D116386584
Pulled By: christophpurrer
fbshipit-source-id: edbf04c8531d4dc35655318d6245918080ae55b7
Summary:
iOS TurboModules mapped a JS `ArrayBuffer` to `NSData` on arguments and `NSMutableData` on returns, so every crossing copied — and `NSMutableData` cannot alias foreign memory, so there was no way to express "these bytes live somewhere else".
This adds `RCTArrayBuffer` (`packages/react-native/React/Base/`) as the ObjC representation of an `ArrayBuffer`. It carries an `isOwningBytes` flag: an owning buffer can be stored and read from any thread, a non-owning one aliases bytes valid only for the synchronous call that produced it.
Codegen now emits `RCTArrayBuffer *` for `ArrayBufferTypeAnnotation` params (was `NSData *`) and returns (was `NSMutableData *`).
## Changelog:
[IOS] [BREAKING] - Add `RCTArrayBuffer`, the ObjC representation of a JS `ArrayBuffer` for TurboModules, with an explicit byte-ownership contract
Pull Request resolved: https://github.com/react/react-native/pull/57879
Test Plan:
- `RCTTurboModuleArrayBufferTests` — 9 tests over the sync in-place path, `isOwningBytes` on a sync argument, returning one's own argument, the void/Promise copy paths, nesting, and a zero-length round trip.
- `RCTTurboModuleTests.mm` adds `testNativeBackedArrayBufferIsAliasedAndKeepsBackingStoreAlive`.
- `RCTSampleTurboModule` doubles its argument in place and returns the same buffer, covering the path end to end.
- Codegen and C++ API snapshots regenerated.
Reviewed By: cipolleschi
Differential Revision: D115629409
Pulled By: christophpurrer
fbshipit-source-id: a7ee6eadf99fd5f9f82fd2a95b0d3ff1369f4ddf
Summary:
Android TurboModules mapped a JS `ArrayBuffer` to `java.nio.ByteBuffer`, copying every argument into a direct buffer — and `ByteBuffer` carries no ownership contract, so there was no way to express aliased or borrowed bytes for synchronous in-place access.
This adds `ArrayBuffer` (`packages/react-native/ReactAndroid/src/main/java/com/facebook/react/bridge/ArrayBuffer.kt`) as the Java representation of an `ArrayBuffer`. It carries an `isOwningBytes` flag: an owning buffer can be stored and returned to JS, a non-owning one aliases bytes valid only for the synchronous call that produced it.
Codegen now emits `ArrayBuffer` for `ArrayBufferTypeAnnotation` params (was `ByteBuffer`) and returns (was `ByteBuffer`).
## Changelog:
[ANDROID] [ADDED] - Add `ArrayBuffer`, the Java representation of a JS `ArrayBuffer` for TurboModules, with an explicit byte-ownership contract
Pull Request resolved: https://github.com/react/react-native/pull/57897
Test Plan:
- Codegen Java spec and JNI C++ snapshot tests updated for `ArrayBuffer` param/return signatures.
- `SampleTurboModule` doubles its sync argument in place and returns the same buffer, covering the zero-copy path end to end; `createNativeBuffer` allocates via `ArrayBuffer`.
- C++ API snapshots regenerated.
Reviewed By: javache
Differential Revision: D115755247
Pulled By: christophpurrer
fbshipit-source-id: de067789ad145b7202da721a02f838358c82f4d8
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/57893
The codegen'd TurboModule `emitOn<Event>` methods invoked their `EventEmitterCallback` without checking it was set. That callback is installed by the generated `*SpecJSI` constructor, which runs when JS first looks the module up — so a native module that emits before that point invoked an empty `std::function` on iOS (`std::bad_function_call`) or a null field on Android (NPE). Native code commonly holds the module instance and pushes events well before any JS surface mounts, so call sites had to wrap every emit in a try/catch to stay crash-free.
Both generators now read the callback into a local and no-op when it is absent:
- ObjC++ (`serializeEventEmitter.js`): copies `_eventEmitterCallback`, calls it only if non-empty.
- Java (`GenerateModuleJavaSpec.js`): copies the `Nullable CxxCallbackImpl` field and null-checks it.
The local is for readability, not synchronization: the callback is installed once and never cleared, and Java reference reads are already atomic.
The `setEventEmitterCallback` lambdas had a separate lifetime bug: they captured `eventEmitterMap_` by reference and looked events up with `operator[]`. The Java/ObjC module owns the callback and can outlive the C++ `*SpecJSI` that installed it, so a stale callback dereferenced a dangling map; and `operator[]` silently default-inserted a null `shared_ptr` for an unknown event name, which the next line dereferenced. They now capture a copy of the map and use a checked `find` (`serializeModule.js`, `JavaTurboModule.cpp`). Every emitter is registered before the callback is installed, so the copy is complete.
Note the scope of the guarantee: it covers the ObjC++ and Java generators, which route through an `EventEmitterCallback`. A C++-only TurboModule (`<Module>CxxSpec`, from `GenerateModuleH.js`) has no callback to check — it emits through `eventEmitterMap_` entries its own constructor registers — so the "emit unconditionally" guidance is about the callback, not about emitting before construction finishes.
Changelog:
[General][Fixed] - TurboModule event emitters no longer throw when an event is emitted before the emitter callback is installed
Reviewed By: javache
Differential Revision: D115130767
fbshipit-source-id: 7d7b9a706b8983968e3e1207b22c66322f04a223
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/57851
The flow-* packages have support for all latest syntax and is better maintained.
Changelog: [Internal]
Reviewed By: huntie
Differential Revision: D115064040
fbshipit-source-id: 7265eb722910a460c76d0dbde0603cf962fe4e4e
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/57752
## Rationale
Flow lib defs that declare global types (available anywhere, without an `import`) must be referenced in `.flowconfig` and can't be maintained incrementally. That's not too bad for very stable APIs and it's necessary for environment/runtime globals, but for 3P libraries it makes the lib defs much more difficult to maintain for little benefit (we have to import the runtime APIs anyway). Secondarily, it's a problem for generating TypeScript types, as TS doesn't declare any 3P library globally.
Babel is one of few cases where a library declares Flow globals - every one has an importable equivalent.
## This diff
Replaces usages of Babel global types across xplat/js with their `babel/types` equivalents
Changelog: [Internal]
Reviewed By: javache
Differential Revision: D113574665
fbshipit-source-id: be668d968345a60e76a3515d19d8eaa002d53274
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/57769
Eventually, we will replace the hermes ones with the `flow-{estree,eslint,parser,transform}` packages. This diff first installs the flow ones.
Changelog: [Internal]
Reviewed By: javache
Differential Revision: D114226573
fbshipit-source-id: a693cdac57d2ce77c002a8ac455aceabaadbcf20
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/57596
The original implementation used `NSMutableData` for both directions of the
JSI<->ObjC ArrayBuffer conversion. That is the wrong contract on the argument
(JS -> ObjC/Native) side: an inbound buffer is owned by the caller, not the
native module, and `NSMutableData` must own a resizable backing store it can
`realloc`/`free`, so it cannot durably alias a foreign buffer.
This diff keeps `NSMutableData` for the ObjC/Native -> JS (return) path, but
switches the JS -> ObjC/Native (argument) path to the immutable `NSData`:
- Codegen: `getParamObjCType` now maps `ArrayBufferTypeAnnotation` params to
`NSData *` (return type stays `NSMutableData *`).
- Runtime: `convertJSIArrayBufferToNSMutableData` is renamed to
`convertJSIArrayBufferToNSData` and returns an immutable `NSData`. The bytes
are still eagerly copied, which keeps the result safe to retain, store, or
dispatch to another thread regardless of whether the source bytes were owned
by JS or by a native `MutableBuffer`.
- Updated the ObjC unit tests and the codegen `GenerateModuleHObjCpp` snapshot.
Changelog: [INTERNAL]
Reviewed By: javache
Differential Revision: D112582384
fbshipit-source-id: 8ed2dfb3f23e7f39677dfb1824c5530f378705bc
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/57611
Directly follows D112733139 (Metro).
**Motivation**
- New Node builtins are *only* available under the prefix (e.g. `node:sqlite`), as this allows Node to introduce them without ecosystem-breaking changes, so this is the direction of travel and the only choice that'll allow consistency.
- Encouraging them and grouping them separately makes it easier to reason about a module's 3rd party dependencies.
Changelog: [Internal]
Reviewed By: robhogan
Differential Revision: D112803684
fbshipit-source-id: 40668d746a7151b3aa4800ff8af997902a18d198
Summary:
I'm doing CI/CD that includes React Native inside of https://tart.run/. Tart uses Apple Virtualization Framework, which in turn mounts VirtioFS volumes.
This feels like a common enough use case to contribute upstream, since I can envision many more people using Apple's virtualization features.
## Changelog:
[GENERAL] [FIXED] - react-native-codegen build.sh on AppleVirtIOFS volumes
Pull Request resolved: https://github.com/react/react-native/pull/57555
Test Plan: Run `packages/react-native-codegen/scripts/oss/build.sh` inside a virtual machine that uses Apple Virtualization Framework such as Tart
Reviewed By: cipolleschi
Differential Revision: D112091671
Pulled By: cortinico
fbshipit-source-id: 800e6ac54175aaf7a9dfb4b99950cbc6da86e085
Summary:
Adds ArrayBuffer support to ObjC TurboModules, following the C++ ArrayBuffer PR ([`226ef2e`](https://github.com/facebook/react-native/commit/226ef2e7c5d1928d5696dc23efc1b8950ba00e37)).
- Codegen support for `ArrayBufferTypeAnnotation` in ObjC module specs (`NSMutableData *` params/returns, new `ArrayBufferKind`)
- JSI↔ObjC conversion wraps native-backed buffers zero-copy via `-[NSMutableData initWithBytesNoCopy:length:deallocator:]`; the deallocator retains the backing store so the bytes stay valid even if the `NSMutableData` escapes the call or the source ArrayBuffer is garbage-collected
- JS-backed buffers are copied, which is safe on both the synchronous and asynchronous paths
This PR is iOS-only; Android support follows in a separate PR.
## Changelog:
[IOS] [ADDED] - Add ArrayBuffer support to ObjC TurboModules
X-link: https://github.com/facebook/react-native/pull/56986
Reviewed By: javache
Differential Revision: D106846249
Pulled By: christophpurrer
fbshipit-source-id: 3393d5d6f31a1412f5d52328c90e51205aa6b153
Summary:
Pull Request resolved: https://github.com/react/react-native/pull/57273
**Motivation**
A README is the first thing an npm visitor sees on each package; internal-repo boilerplate like the "Testing" sections and the "We're using yarn…" note doesn't belong on a published package page.
**Changes**
- Standardize every README: a `# react-native/<name>` heading, a user-facing description, and a blue version badge plus a green monthly-downloads badge.
- Keep the "internal dependency" prelude only on `community-cli-plugin`, `virtualized-lists`, and `js-polyfills`.
- Drop the monorepo-template "Testing" sections and yarn note.
- Rework `metro-config` around the Configuring Metro guide; add missing READMEs for `metro-babel-transformer` and `popup-menu-android`.
Changelog: [Internal]
Reviewed By: emily8rown
Differential Revision: D109017270
fbshipit-source-id: 6208294c4ad2e6235a6605ae54f22d730f0476e7