feat(macos): merge RN through RedBox 2.0 cutoff and port RedBox 2.0 to AppKit - #3054
Draft
Saadnajmi wants to merge 566 commits into
Draft
feat(macos): merge RN through RedBox 2.0 cutoff and port RedBox 2.0 to AppKit#3054Saadnajmi wants to merge 566 commits into
Saadnajmi wants to merge 566 commits into
Conversation
Summary: Pull Request resolved: react#56042 Adds CI workflow for validating whether the current C++ API snapshot is equivalent with the generated one. Changelog: [Internal] Reviewed By: cortinico Differential Revision: D95963515 fbshipit-source-id: 4629999e2d09dbdcfb9fd3e6308a9254dd4f0237
Summary: Pull Request resolved: react#56173 Removes leaked test code from the snapshot. Changelog: [Internal] Reviewed By: cortinico Differential Revision: D97457090 fbshipit-source-id: ac6aedfa9f4fda8c7f7371fc29a91ce813916777
Summary: Pull Request resolved: react#56248 Refactor and add global exclude patterns to the C++ stable API snapshot config. This enables better modularization of shared exclude patterns coming from the assumption that ReactAndroid and ReactApple consists of ReactCommon in their inputs and some paths should not be processed on each platform. Changelog: [Internal] Reviewed By: cortinico Differential Revision: D97465431 fbshipit-source-id: e71daef90853dad8003a0f41b1ea78b8c56bee02
…56247) Summary: Pull Request resolved: react#56247 Removes fantom test specific methods, namely: - NativeFantomTestSpecificMethods, - FantomForcedCloneCommitHook from the C++ API snapshot. Changelog: [Internal] Reviewed By: cortinico Differential Revision: D97466201 fbshipit-source-id: b8464ddd242056a1e21db409b738414814550b30
Summary: Pull Request resolved: react#56252 Removes stubs from the snapshot which should be irrelevant in context of the public surface. Changelog: [Internal] Reviewed By: cortinico Differential Revision: D97478132 fbshipit-source-id: da95c86a02b056e9ece0fbdcbbbced2300c9fbae
…#56238) Summary: Pull Request resolved: react#56238 Currently, all xml artifacts generated by doxygen are stored in the tmp dir and deleted at the end of the snapshot generation. For debugging reasons, this diff adds `--xml` flag to persist generated artifacts, so that they can be analyzed. Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98289956 fbshipit-source-id: 73eb0185f7112389e589d3e8f8a2004efe325e8d
…zations (react#56239) Summary: Pull Request resolved: react#56239 Doxygen incorrectly merges base classes from primary templates into their partial specializations. In C++, a specialization's inheritance list completely replaces the primary template's, but Doxygen combines both into a single basecompoundref list. - **Contradictory type traits** — is_optional<std::optional<T>> and is_variant_of_data_types<std::variant<Ts...>> showed inheritance from both std::false_type (primary) and std::true_type (specialization) - **Duplicate base classes** — Converter<jsi::Object> listed ConverterBase<jsi::Object> twice (once from the primary template after substitution, once from the specialization) This diff fixes both issues with two complementary mechanisms: - **Dedup-by-name (Extendable._deduplicate_base_classes)**: removes exact duplicate base classes, keeping the last occurrence. Handles cases where Doxygen's template argument substitution produces identical names. - **Primary template base subtraction (StructLikeScopeKind._remove_merged_primary_bases)**: for partial specializations, looks up the primary template among sibling scopes and performs count-based subtraction of its bases. Count-based (rather than set-based) subtraction correctly preserves bases that a specialization explicitly re-inherits from the same class as the primary. Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98291360 fbshipit-source-id: 82b34969b13c7df8babf2c5efc222b29979d6361
Summary: Pull Request resolved: react#56246 Fixes return type to receive correct generic types. Before: ``` static std::pair<T1, T1> fromJs(facebook::jsi::Runtime &rt, const jsi::Array &array, const std::shared_ptr<CallInvoker> &jsInvoker) ``` After: ``` static std::pair<T1, T2> fromJs(facebook::jsi::Runtime &rt, const jsi::Array &array, const std::shared_ptr<CallInvoker> &jsInvoker) ``` Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98292820 fbshipit-source-id: bb7791ace2391240dd50ece55b2a2ce905a51115
Summary: Pull Request resolved: react#56240 There are cases in which we want to strip particular symbols from the snapshot. This diff adds `exclude_symbols` to the parser config to enable definitions of symbol patterns that should be excluded from the snapshot. Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98295409 fbshipit-source-id: 47414c1e5060a01dce76d1fc4fdd8ea355bb8d37
Summary: Pull Request resolved: react#56241 Strips remaining Fantom symbols from the public API snapshots generated by codegen. Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98295521 fbshipit-source-id: 24568b825b2aaa062e2b11dc77a018e403c68ddd
…56243) Summary: Pull Request resolved: react#56243 Modifies C++ API parser config to accept `exclude_symbols` at the platform level and sets `Android` symbols to be excluded from the ReactApple snapshot. Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98298766 fbshipit-source-id: 70b39fe4935e6791701c173e7c9e12532c2817a5
Summary: Pull Request resolved: react#56245 Removes feature flags from the C++ public API snapshots as they are considered private. This greatly reduces snapshots size. Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98312529 fbshipit-source-id: e6aad9d930c7c77300329c25668dc2fc3a4686f2
Summary: Pull Request resolved: react#56242 Excluded symbols that contain "UnstableLegacy" from public stable C++ API snapshot, that is `UnstableLegacyViewManagerAutomaticComponentDescriptor` and `UnstableLegacyViewManagerInteropComponentDescriptor`. Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98317003 fbshipit-source-id: deb4631a26ecb138007cf4958dbabb57ef3739cc
…56244) Summary: Pull Request resolved: react#56244 Excludes symbols marked as Experimental from the C++ public API snapshots. Types with the "Experimental" prefix are explicitly not stable API — they represent features still under active development and are subject to change or removal without notice. Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98454721 fbshipit-source-id: 7f7488fab6bc94c52b94a3ae2d9a7a7914815d63
…react#56249) Summary: Pull Request resolved: react#56249 Prevents iOS Switch component leaking into the ReactAndroid snapshot by adding `*/components/switch/iosswitch/*` path to the `exclude_patterns` in the parser config. Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98473064 fbshipit-source-id: de25c4cbe839f8a065b8465e48a2b8c2a5ba81d7
…ot (react#56251) Summary: Pull Request resolved: react#56251 Excludes InputAccessory which is an iOS-only component with no Android support at any layer. In the `InputAccessoryView.js` there is explicit iOS guard which logs warning on Android. Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98478125 fbshipit-source-id: 871923e001024971167a7bb2eb900ca3867983a4
…eact#56250) Summary: Pull Request resolved: react#56250 Excludes iOS-specific symbols from the ReactAndroid C++ API snapshot. The snapshot contains iOS native modules CxxSpec templates, their data structs, and bridging structs that have no relevance on Android. These symbols leaked because the Android codegen generates CxxSpec templates for ALL JavaScript module specs regardless of platform. Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98480080 fbshipit-source-id: a5990c4769456ab756095129851eb4e8dbeda912
…#56253) Summary: Pull Request resolved: react#56253 Renames flag for verifying that the C++ public API snapshot is with sync with the actual public API surface to match the JS API command. Changelog: [Internal] Reviewed By: huntie Differential Revision: D98484129 fbshipit-source-id: ec806353f263b00e5060d9e7f0333bcf4cced148
Summary: Pull Request resolved: react#56286 [changelog](https://github.com/facebook/flow/blob/main/Changelog.md) Changelog: [Internal] Reviewed By: gkz Differential Revision: D98860515 fbshipit-source-id: 196ffafcee417ac2949822848fe7752070a2c30a
Summary: Pull Request resolved: react#56207 Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D96101927 fbshipit-source-id: 06c6dd11c145be44bcdd97f14683a222c66c96c6
Summary: Add Changelog for 0.85.0-rc.7 ## Changelog: [Internal] - Add Changelog for 0.85.0-rc.7 Pull Request resolved: react#56276 Test Plan: N/A Reviewed By: cipolleschi Differential Revision: D98792883 Pulled By: alanleedev fbshipit-source-id: c21c189e35d3eca3a013e2e924c87725cec86153
Summary: Adds missing `mixBlendMode: 'plus-lighter'` which has full support on CSS across all browsers (as opposed to, for example, `plus-darker`) iOS: ended up using linear dodge blend mode which should be identical, I tried using `@"addition"` which should've used `CIAdditionCompositing` but that didn't work, although I did confirm `maximum` does work and apply `CIMaximumCompositing`, funny stuff, can't wait for SwiftUI layer to take this over Android works as expected with its `BlendMode.PLUS` ## Changelog: [GENERAL] [ADDED] - Add mixBlendMode plus-lighter Pull Request resolved: react#56278 Test Plan: <img width="726" height="822" alt="image" src="https://github.com/user-attachments/assets/6f8a235c-f79b-4739-9d6c-90c989bad5f5" /> Reviewed By: javache, jorge-cab Differential Revision: D98842036 Pulled By: NickGerleman fbshipit-source-id: 3f4bf83918276579c48c784479d2a20bf2a2ba4c
Summary: Pull Request resolved: react#56282 Fixes react#56185 Fixes react#56186 Fixes react#56187 Delete the `fixTextClippingAndroid15useBoundsForWidth()` feature flag and all code gated behind it. This flag was never shipped, and the new code path has been implicated in several bugs. The change removes: - The feature flag definition and all generated plumbing (C++, JNI, Kotlin, JS) - The visual-bounds-based width measurement path in `TextLayoutManager.kt` - The `setUseBoundsForWidth(true)` call in `buildLayout()` - The MobileConfig parameter (`react_fabric.fix_text_clipping_android_15_use_bounds_for_width`) - The Facebook Android and OSS Experimental overrides The old advance-based width logic is retained as the sole code path. Changelog: [Internal] Reviewed By: cortinico Differential Revision: D98427218 fbshipit-source-id: 7c322353d9c0a0548386a486ecd4280b73a66872
…fined with props 2.0 (react#56288) Summary: When Android props 2.0 diffs a `View` where `zIndex` changes from a value to `undefined`, `HostPlatformViewProps::getDiffProps` calls `zIndex.value()` and throws `bad_optional_access`: https://github.com/user-attachments/assets/89752243-ff48-42c7-a584-b47a091b4a4d This change emits `folly::dynamic(nullptr)` when `zIndex` is cleared instead. ## Changelog: [ANDROID] [FIXED] - Fixed a crash when clearing `zIndex` with props 2.0 enabled Pull Request resolved: react#56288 Test Plan: Use the repro in `packages/rn-tester/js/examples/Playground/RNTesterPlayground.js`: ```js function Playground() { const [toggle, setToggle] = React.useState(false); return ( <View style={styles.container}> <Button title="Toggle" onPress={() => setToggle(t => !t)} /> <View style={{ backgroundColor: toggle ? 'blue' : 'red', height: 100, width: 100, position: 'absolute', zIndex: toggle ? undefined : 1, }} /> </View> ); } ``` 1. Enable Android props 2.0: ``` override fun enableAccumulatedUpdatesInRawPropsAndroid(): Boolean = true override fun enableExclusivePropsUpdateAndroid(): Boolean = true override fun enablePropsUpdateReconciliationAndroid(): Boolean = true ``` 2. run RNTester Playground on Android, and press `Toggle` to switch `zIndex` between `1` and `undefined`. This is with the issue fixed: https://github.com/user-attachments/assets/11253c23-b235-4e70-89a3-0d7bd07937f5 Reviewed By: javache Differential Revision: D98960669 Pulled By: NickGerleman fbshipit-source-id: d4a98434c599d2298c5cfc8c6895988ed0743af6
Summary: Pull Request resolved: react#56293 Changelog: [Internal] Reviewed By: mdvacca Differential Revision: D98942266 fbshipit-source-id: b6df078e4e10191b464b3aebc70d55cca5f12fa2
Summary: Pull Request resolved: react#56300 Changelog: [Internal] Reviewed By: marcoww6 Differential Revision: D99004126 fbshipit-source-id: dfedc13c867cb768777e2ba65256a2c40ffb1d08
Summary: Pull Request resolved: react#56289 Changelog: [GENERAL][FIXED] Link VirtualView component I saw D98039999 and I wanted to check how the new components are working in OSS, and currently they don't. VirtualView component isn't correctly linked in OSS. This diff addresses that. Note that VirtualView in OSS requires `babel-plugin-transform-flow-enums` which is not part of the default template. Reviewed By: CalixTang Differential Revision: D98887952 fbshipit-source-id: f06ec9d4581c6fa3f2185641cb4b9f7acb8f557e
…56292) Summary: Pull Request resolved: react#56292 The TinyMap is a helper class under internal directory, not referenced anywhere in the snapshot, and considered unsafe (from the comment). It is not part of any module, so it is better to remove that from the snapshot. The `packAnimatedProps` method is used in `AnimationBackend` for translation from `AnimatedProps` to `folly::dynamic` and should not be used outside of React Native. Changelog: [Internal] Reviewed By: cipolleschi Differential Revision: D98907855 fbshipit-source-id: e3a371cc088000a23f1ce1121fda808494979140
…EY (react#56304) Summary: Pull Request resolved: react#56304 Add support for specifying a custom `timeStamp` in the `Event` constructor options via a private symbol key (`EVENT_INIT_TIMESTAMP_KEY`) defined in `EventInternals`. This is for internal construction of events using a custom timestamp (instead of event object creation), for use cases like dispatching events from the host platform using the original timestamps. - Added `EVENT_INIT_TIMESTAMP_KEY` symbol to `EventInternals` - Added `setEventInitTimeStamp(eventInit, timeStamp)` helper function in `EventInternals` that validates the type and sets the symbol-keyed property on the event init object - Updated the `Event` constructor to read the custom timestamp from options if present, falling back to `performance.now()` - Added Fantom tests for custom timestamp and zero as a valid timestamp Changelog: [Internal] Reviewed By: huntie Differential Revision: D99076591 fbshipit-source-id: 2ea00394de885a5f338c1b3fe4792d5b4768cd7b
Summary: Pull Request resolved: react#56295 Changelog: [Android][Added] - Pass event object to `BackHandler` `hardwareBackPress` events to access timeStamp from native event. BackHandler on Android emits the `hardwareBackPress` event with a `null` payload — no timestamp or event data. This means there is no way for handlers to know when the user actually pressed the back button, which prevents accurate performance tracing for back navigations. This diff: - Sends `SystemClock.uptimeMillis()` as `timestamp` in the native `DeviceEventManagerModule.emitHardwareBackPressed()` event payload - Creates a new `HardwareBackPressEvent` class extending the DOM `Event` class, which overrides the `timeStamp` getter to return the native timestamp (falling back to `performance.now()` if unavailable) - Makes the public constructor throw `TypeError("Illegal constructor")` following the `PerformanceEntry` pattern - Updates all BackHandler platform files (Android, iOS, macOS, Windows) and type definitions (Flow, TypeScript) to create and pass the event to handlers - Updates the BackHandler mock to pass events in `mockPressBack` - Adds Fantom integration tests for both `HardwareBackPressEvent` and `BackHandler` Reviewed By: huntie Differential Revision: D98941079 fbshipit-source-id: baca35947cd7e8f598e9e040059579fb6e19553e
Summary: Pull Request resolved: react#56486 Follow-up to react#55855 (D100437445), which replaced the legacy height-based IME detector in `ReactRootView.CustomGlobalLayoutListener` with a visibility-only detector (`WindowInsetsCompat.Type.ime()`). The legacy code also re-emitted `keyboardDidShow` whenever the IME *height* changed while still visible (e.g., user toggles the emoji panel or predictive bar). Removing that left these JS consumers reading stale `endCoordinates`: - `KeyboardAvoidingView` — caches `_keyboardEvent`, no longer resizes on IME height changes. - `ScrollView` — caches `_keyboardMetrics`; focused `TextInput` auto-scroll uses stale `screenY`/`height`. - `Keyboard.metrics()` — public API returning the cached payload from the last `keyboardDidShow`. Fix: track the last reported IME height and re-emit `keyboardDidShow` when the keyboard is visible AND (it just appeared OR the height changed), restoring pre-D100437445 semantics. The hide branch is gated on `mKeyboardIsVisible` so it still fires exactly once per cycle. Changelog: [Android][Fixed] KeyboardAvoidingView and TextInput auto-scroll not responding to IME height changes (e.g., when toggling emoji panel or predictive bar) Reviewed By: mdvacca, javache Differential Revision: D101385688 fbshipit-source-id: 071707bc6fb9e7173e491135b27a66dc177c14c5
…eact#56604) Summary: Pull Request resolved: react#56604 ## Changelog: [Internal] [Fixed] - keep old snapshot node position updated at applyViewTransitionName `createViewTransitionInstance` is only called once in a shadowNode's lifecycle but `applyViewTransitionName` will be called multiple times at each transition for the same shadowNode. Reviewed By: christophpurrer Differential Revision: D102226352 fbshipit-source-id: ac1ba1fe53dce199d6a2ff7e83a22e5249336a01
…orking) (react#56602) Summary: Pull Request resolved: react#56602 Changelog: [iOS][Fixed] - Removing unused ivars Reviewed By: Murf-o Differential Revision: D102315121 fbshipit-source-id: 51075b1460fda1d7d044f6788fc0b2d390bed778
Summary: Pull Request resolved: react#56603 Upgrade transitive dependency lodash from 4.17.21 to 4.18.1 to remediate CVE-2026-4800 (Improper Control of Generation of Code / Code Injection). - Added `"lodash": "4.18.1"` to `resolutions` in package.json to force all lodash ranges (including `~4.17.15`) to resolve to 4.18.1 - Updated yarn.lock entry to resolve all lodash ranges to 4.18.1 Without the resolution override, the `~4.17.15` range would stay at 4.17.21 (vulnerable) since `~` only allows patch-level updates and 4.18.1 is a minor bump. Changelog: [Internal] Reviewed By: christophpurrer Differential Revision: D102273666 fbshipit-source-id: d3a17ba480b5d44c048b2b1d814bc3989abcd06a
… dispatch + fix unsigned retainCount_ underflow (react#56537) Summary: **1) `EventQueueProcessor::flushEvents` — release on every exit path.** `flushEvents` retains every `EventTarget` in the batch, then for each event calls `eventPipe_` (which dispatches into JS via `UIManagerBinding::dispatchEventToJS` and can throw `jsi::JSError`), then `eventPipeConclusion_`, and finally runs a trailing loop that calls `event.eventTarget->release(runtime)`. If either pipe call throws, stack unwinding skips the release loop, leaving `strongInstanceHandle_` (a `jsi::Value` strong reference to the JS event-target object) pinned and `retainCount_` desynchronised. This change replaces the trailing release loop with a folly `SCOPE_EXIT` block immediately after the retain pass, so every retained target is released on every exit path — normal return, `eventPipe_` throw, or `eventPipeConclusion_` throw. (Earlier revisions used a hand-rolled RAII class; switched to `SCOPE_EXIT` per reviewer feedback. `folly/ScopeGuard.h` is already in the OSS RN folly subset — `RCT-Folly.podspec`, `ReactAndroid/.../folly/CMakeLists.txt`, `react-native-fantom/.../folly/CMakeLists.txt` — and is a transitive dep of `//xplat/folly:dynamic`, so no new build edges.) Behaviour-preserving on the happy path: same release ordering (after `eventPipeConclusion_`), same no-`DispatchMutex` policy for release (the `events` vector still holds the strong `shared_ptr`s), zero added allocation. **2) `EventTarget::release` — guard the unsigned decrement.** `retainCount_` is `size_t`. `retain()` early-returns when the target is disabled, but the matching `release()` always runs, so `release()` can be entered with `retainCount_ == 0`; the unconditional `--retainCount_` then wraps to `SIZE_MAX`, the `== 0` check never fires again, and `strongInstanceHandle_` is never cleared. The trailing `react_native_assert(retainCount_ >= 0)` is a tautology on an unsigned type and is removed. This change guards the decrement (`if (retainCount_ > 0) --retainCount_;`) and drops the dead assert and the now-unused `react_native_assert.h` include. Adds `EventQueueProcessorTest.releasesEventTargetsWhenDispatchThrows` (proves the strong handle is released during unwind when `eventPipe_` throws; fails on the pre-fix trailing-loop code) and `EventTargetTests.releaseWhileDisabledDoesNotCorruptRetainCount` (retain/release while disabled, then enable+retain must still acquire the handle; fails on the pre-fix unconditional decrement). Changelog: [Internal] Pull Request resolved: react#56537 Reviewed By: javache Differential Revision: D100280132 fbshipit-source-id: 8c7128082d8c967a150952434a7791b62586659f
Summary: Add Changelog for 0.83.8 ## Changelog: [Internal] - Add Changelog for 0.83.8 Pull Request resolved: react#56596 Test Plan: N/A Reviewed By: christophpurrer Differential Revision: D102339274 Pulled By: motiz88 fbshipit-source-id: c83710b74ad615cd389d33b5ee6c519de72cb32d
Summary: - Adds retry jobs (up to 2 retries) for all 4 E2E test jobs on PRs to handle flakiness - Original E2E jobs use `continue-on-error` on PRs so failures don't block the workflow; on `main`, behavior is unchanged and the existing `rerun-failed-jobs` mechanism continues to work - Each retry runs on a fresh runner (addressing environment-level flakes) and is triggered via step-level outcome captured as a job output - Added `overwrite: true` to artifact uploads in maestro composite actions so retry jobs don't conflict on artifact names ### How it works - **On PRs:** E2E job fails → `retry_1` triggers → if that fails → `retry_2` triggers. All have `continue-on-error` so the workflow stays green. - **On `main`:** `continue-on-error` is `false`, retry jobs are skipped (PR-only), and `rerun-failed-jobs` handles retries as before. ### Known limitation Since these are matrix jobs (Debug/Release), the job output uses the last-to-complete matrix combination's value. If only one flavor fails and the passing one finishes last, the retry may not trigger. In the common flakiness pattern (environment-level issues), both flavors tend to be affected, so this works well in practice. ## Changelog: [Internal] - Add E2E test retry jobs for PRs Pull Request resolved: react#56581 Test Plan: - CI will validate the workflow syntax - On a PR where E2E tests pass: retry jobs should be skipped - On a PR where E2E tests fail due to flakiness: retry jobs should trigger and (ideally) pass on a fresh runner - On pushes to `main`: existing `rerun-failed-jobs` behavior is preserved (retry jobs are skipped) Reviewed By: cortinico Differential Revision: D102327900 Pulled By: cipolleschi fbshipit-source-id: e257e012357fdba442c302085ceed0667355272e
…gn null checks and error message code style (react#56150) Summary: This PR continue the work I'm doing to improve the `AccessibilityInfo` module (react#56066 and connected PRs). In this new PR I added the missing tests for all the Promise based methods in the `AccessibilityInfo`, to prevent regressions similar to those fixed in the previous PRs. This PR also includes a couple of small code style consistency improvement: - aligned the null check in `getRecommendedTimeoutMillis` to the other ones already present in the file - aligned the error messages returned from `isGrayscaleEnabled`, `isInvertColorsEnabled` and `isReduceMotionEnabled` to the other ones - added the docs links where was missing (related to the PR I opened on the react-native doc repo) ## Changelog: [GENERAL] [ADDED] - AccessibilityInfo: added missing tests for Promise based methods, align null checks and error message code style Pull Request resolved: react#56150 Test Plan: I run the tests that I added to verify their correctness. Some of them were failing because of the change of the error messages, and I made them green by changing the message in the production code. For this others, I verified their correctness by breaking the production code to see if they were failing. Reviewed By: cipolleschi Differential Revision: D102190403 Pulled By: cortinico fbshipit-source-id: f5fe5cacfe0136b25a93d05f3b136c2bc49bd0e2
Summary: Pull Request resolved: react#56623 Fix the return types of toJSON() methods in PerformanceEntry, PerformanceServerTiming, and Performance to return an object representation ({[string]: unknown}) instead of a string, per the WebIDL spec ([Default] object toJSON()). Changelog: [internal] Reviewed By: fabriziocucci Differential Revision: D102597516 fbshipit-source-id: 9f063d16693e1b44f89a298c08f25548f8a05c76
…eact#56624) Summary: Pull Request resolved: react#56624 Adds Flow type definitions for `MutationObserver` (and its related types `MutationRecord` and `MutationObserverInit`) and `IntersectionObserver` (and its related types `IntersectionObserverEntry`, `IntersectionObserverCallback` and `IntersectionObserverOptions`) to React Native's library definitions, reflecting the features supported by the React Native implementations. Changelog: [internal] Reviewed By: fabriziocucci Differential Revision: D102597515 fbshipit-source-id: 86ff8d90aeb26f7f950a968b93596808da834951
…flag) (react#56622) Summary: Pull Request resolved: react#56622 Changelog: [internal] Adds examples for MutationObserver if the API is available (which isn't normally as it's gated). Reviewed By: fabriziocucci Differential Revision: D102593725 fbshipit-source-id: cdcfe613b485bf9b015a342cb51003ce43455096
Summary: Add a new feature flag to gate overriding getClipBounds on Android views to return the padding box when overflow is hidden. This communicates the view's clipping region to the Android framework for correct getGlobalVisibleRect calculations (T253147322, T263753590). Changelog: [Internal] Reviewed By: twasilczyk Differential Revision: D102353841 fbshipit-source-id: d5f9c5428c342061c0d45bfcc000db356048f930
…eact#56606) Summary: Pull Request resolved: react#56606 Override `getClipBounds()` on ReactViewGroup to return the padding box rect when `overflow` is hidden or scroll, gated behind the `syncAndroidClipBoundsWithOverflow` feature flag. This exposes the view's clipping behavior to the Android framework via the standard `View.getClipBounds()` API: - `getClipBounds(Rect): Boolean` returns `true` and populates the rect with the padding box when overflow is hidden/scroll, or `false` when overflow is visible - `getClipBounds(): Rect?` returns the padding box rect or `null` respectively This allows framework-level systems to inspect a view and determine whether it clips its children by checking `getClipBounds() != null`, and to obtain the actual clipping region for `getGlobalVisibleRect()` calculations. ## Why not `setClipBounds()`? An alternative approach would be to call `View.setClipBounds(Rect)` which sets the `mClipBounds` field directly. This has the advantage of being picked up by AOSP's `ViewGroup.getChildVisibleRect()` (which reads `mClipBounds` via direct field access rather than calling `getClipBounds()`). However, `setClipBounds()` also propagates the clip rect to the RenderNode via `mRenderNode.setClipRect()`, which clips the view's own rendering — including its borders. It also requires maintaining a Rect that must be kept in sync across size changes, border width changes, and overflow changes. For now, the `getClipBounds()` override is sufficient because harvesting systems query the method rather than the field. If we find that AOSP's field-based path is also needed, we can switch to `setClipBounds()` in the future. Adds `BackgroundStyleApplicator.getPaddingBoxRect()` (internal via buck friend_paths) which computes the padding box (view bounds minus border insets) reusing the same border inset resolution logic as `clipToPaddingBox()`. Changelog: [Internal] Reviewed By: twasilczyk, javache Differential Revision: D102353840 fbshipit-source-id: fe1d52e3c000ba1b5e2d21eac75928be24a86645
…s out of `react-native`" (react#56621) Summary: Pull Request resolved: react#56621 Changelog: [General][Deprecated] Deprecated second argument for hook function passed to AppRegistry.setComponentProviderInstrumentationHook as it's not usable by apps. Passed stub that warns when used. Re-applying react#56580 that was reverted. Reviewed By: javache Differential Revision: D102593753 fbshipit-source-id: 7714260a494a62d55880f10887cd440d217a4672
Summary: Pull Request resolved: react#56608 ## Changelog: [Internal] [Fixed] - Fix PixelCopy snapshot for partially off-screen PixelCopy captures only the visible portion of the window surface. When a view is partially off-screen, the capture rect extends beyond the window bounds, resulting in a bitmap where only the visible region has content. This partial bitmap then gets stretched to fill the full-size pseudo-element, causing visual distortion. Fix by clamping the PixelCopy rect to the window bounds and compositing the clamped capture into a full-size bitmap at the correct offset. The off-screen portions remain transparent instead of being stretched. If the view is entirely off-screen, skip capture entirely — the pseudo-element will have no snapshot applied (didMountItems skips tags without a captured bitmap). Reviewed By: javache Differential Revision: D102360642 fbshipit-source-id: 6da9ddce0eef4bac011f25494a6d916c8311058e
Summary: Pull Request resolved: react#56601 `ACTION_SCROLL` events (mouse wheel, trackpad, joystick) were not participating in FPS performance logging. Touch-driven scrolls already called `enableFpsListener`/`disableFpsListener` through `handlePostTouchScrolling`, which uses a stable-frames heuristic to detect when scrolling has settled (3 consecutive frames with no position change). ACTION_SCROLL events bypassed this entirely. This adds FPS listener support to the `ACTION_SCROLL` path by calling `enableFpsListener` when a scroll event arrives, then delegating to `handlePostTouchScrolling` for lifecycle management: - **Snapping ScrollViews (`ReactScrollView`, `ReactHorizontalScrollView`):** The existing debounce runnable calls `flingAndSnap` to animate to a snap point. After the snap, `handlePostTouchScrolling` monitors the animation via the stable-frames mechanism and disables the FPS listener once settled. - **Non-snapping ScrollViews (`ReactScrollView`):** `handlePostTouchScrolling` is called directly. Its re-entry guard (`mPostTouchRunnable != null`) naturally deduplicates across rapid scroll events, while `onScrollChanged` keeps the monitor alive by setting `mActivelyScrolling`. Also nulls `mPostTouchRunnable` before reassignment in the debounce cancel path to avoid stale references. Changelog: [Android][Fixed] - Dispatch FPS performance events for mouse wheel and trackpad scroll interactions in ScrollView Reviewed By: javache, jehartzog Differential Revision: D102347043 fbshipit-source-id: aab91858404db54b1c44146d4d071193baaae7a1
Summary: This PR (initially created for edge-to-edge opt-in support, rebased multiple times) fixes the `Dimensions` API `window` values on Android < 15, when edge-to-edge is enabled. Currently the window height doesn't include the status and navigation bar heights (but it does on Android >= 15): <img width="300" alt="Screenshot 2025-06-27 at 16 23 02" src="https://github.com/user-attachments/assets/c7d11334-9298-4f7f-a75c-590df8cc2d8a" /> Using `WindowMetricsCalculator` from AndroidX: <img width="300" alt="Screenshot 2025-06-27 at 16 34 01" src="https://github.com/user-attachments/assets/7a4e3dc7-a83b-421b-8f6d-fd1344f5fe81" /> Fixes react#47080 ## Changelog: [Android] [Fixed] Fix `Dimensions` `window` values on Android < 15 when edge-to-edge is enabled Pull Request resolved: react#53254 Test Plan: Run the example app on an Android < 15 device. Reviewed By: javache Differential Revision: D80237818 Pulled By: alanleedev fbshipit-source-id: f05e546cd38c7bd539b86d30fe7b663715a34a99
…() (react#56627) Summary: Pull Request resolved: react#56627 ## Changelog: [Android] [Fixed] - make sure view width and height are non 0 before calling createBitmap() Reviewed By: Abbondanzo Differential Revision: D102657602 fbshipit-source-id: 5e93e110af0e5a795e23c0006a9833934c5e5985
…t#56630) Summary: Pull Request resolved: react#56630 Catch `OutOfMemoryError` in the `onResponse` callback's response processing try-catch block in `NetworkingModule.kt`. Previously, only `IOException` was caught — `OutOfMemoryError` (which extends `Error`, not `Exception`) would propagate uncaught and crash the app. When the Java heap is near its limit (e.g., 256MB with <1% free), any allocation during response processing (buffer creation in `readWithProgress`, `responseBody.bytes()`, `responseBody.string()`) can trigger OOM. The fix converts this to a network request error event sent to JS, allowing the app to handle the failure gracefully (retry, show error message) instead of crashing. **Crash:** `android_crash:java.lang.OutOfMemoryError:com.oculus.igvr` **MID:** `024a57cfb70eef9a61af42d845c981a1` — 246 crashes in 7 days **Task:** T212267841 Logview link: [024a57cfb70eef9a61af42d845c981a1](https://www.internalfb.com/logview/system_vros_crashes/024a57cfb70eef9a61af42d845c981a1) Changelog: [Android][Fixed] - Catch OutOfMemoryError in NetworkingModule response processing to prevent app crashes Reviewed By: cortinico Differential Revision: D102236869 fbshipit-source-id: de6624640b3f2343ca8e516fe7677eecae23a932
…0) (react#56586) Summary: Fixes react#50797. A view with a non-invertible transform (e.g. `transform: [{scaleY: 0}]` or `{scaleX: 0}`) still received touches on both Android and iOS. On Android the view appeared to "inherit" a hit region from another view in the hierarchy; on iOS the view registered taps along the collapsed axis. Both symptoms collapse to the same native behaviour: when a transform can't be inverted, the platform APIs silently fall back to stale or degenerate data during hit testing. ### Root cause **Android.** `TouchTargetHelper.getChildPoint` calls `Matrix.invert(inverseMatrix)` without checking its return value. `Matrix.invert` returns `false` for a non-invertible matrix and **leaves its destination unchanged**. `inverseMatrix` is a class-level field reused across every child-point conversion, so on failure we apply *the previous view's* inverse to the current view's touch point: the exact "inherits from another view" behavior. It is also why shrinking a view from `scaleY: 0.9` to `0.0` leaves a 90 % hit region: the 0.9 frame populated the cache. **iOS.** `CGAffineTransformInvert` is documented to return the original matrix when it can't invert. `-[UIView convertPoint:fromView:]` uses that "inverse" during hit testing, so the touch point gets collapsed onto the degenerate axis (e.g. y = 0 for `scaleY: 0`) and the inherited `pointInside:` still reports a hit against a visually invisible view. ### Fix - **Android** (`TouchTargetHelper.kt`): `getChildPoint` now returns `Boolean`. It returns `false` when `Matrix.invert` fails; the child iteration loop skips such children. One behaviour change, no epsilon tuning, no shared state leaks. Resolves both symptom variants via the same code path. - **iOS** (`RCTViewComponentView.mm`, Fabric): add a small `RCTLayerTransformCollapsesAxis` helper that tests the determinant of the 2 × 2 XY projection of `layer.transform` and return `NO` from the existing `pointInside:` override when it collapses. - **Tests.** `TouchTargetHelperTest` (new, Robolectric): initial `scaleX`/`scaleY: 0`, zero-scale parent, the "inherits from sibling" regression, the 0.9 to 0.0 transition, and the touch-path accumulator. `RCTViewComponentViewTests` gets four parallel iOS cases. All pass locally. - **RNTester.** New Transforms entry "Zero-scale hit test (regression for react#50797)". Reproduces both variants and exposes a scale input so you can step through the transition interactively. ### Why not react#53769? PR react#53769 solved the Android side with a `hasZeroScale(view)` early-return plus an epsilon comparison on individual matrix entries, and handled the `invert` failure in `getChildPoint` by *falling back to untransformed coordinates*. That still lets the degenerate view be hit with its pre-transform bounds; the early-return then has to catch the miss. This branch fixes the root cause once: if the matrix can't be inverted we don't descend into the child. The iOS fix is the matching change on the Fabric side. ## Changelog: [GENERAL] [FIXED] - Views with a non-invertible transform (e.g. `scaleX: 0` or `scaleY: 0`) no longer receive touches on Android or iOS. Pull Request resolved: react#56586 Test Plan: ### Unit tests - `./gradlew :packages:react-native:ReactAndroid:testDebugUnitTest --tests 'com.facebook.react.uimanager.TouchTargetHelperTest*'`: 7/7 pass. - iOS build-for-testing of `RNTesterUnitTests` scheme succeeds; the four new `RCTViewComponentViewTests` cases live next to the existing cases in `React/Tests/Mounting/`. - `./gradlew ktfmtCheck`, `clang-format --dry-run -Werror`, `prettier --check`, `flow`, and `eslint` all pass. ### Manual verification (Android) RNTester → Components → Transforms → "Zero-scale hit test (regression for react#50797)". **Variant 1: tap the `scaleY: 0` row directly.** "Last tapped:" must stay `(none)`. | Before | After | | --- | --- | | <img src="https://raw.githubusercontent.com/qflen/react-native/6f3799dbb597630a91f43e460437d636796dbbe7/assets/50797/before-tap-zero-scale.png" width="260"> | <img src="https://raw.githubusercontent.com/qflen/react-native/6f3799dbb597630a91f43e460437d636796dbbe7/assets/50797/after-tap-zero-scale.png" width="260"> | | `Last tapped: zero-scale` (bug) | `Last tapped: (none)` | **Variant 2: tap the `scaleY: 0.5` row (①), then tap the `scaleY: 0` row (②).** Without the fix, the shared `inverseMatrix` cache populated by tap ① leaks into the zero-scale hit test on tap ②. After the fix, tap ② is correctly ignored and "Last tapped" still reflects tap ①. | Before | After | | --- | --- | | <img src="https://raw.githubusercontent.com/qflen/react-native/6f3799dbb597630a91f43e460437d636796dbbe7/assets/50797/before-tap-zero-after-05.png" width="260"> | <img src="https://raw.githubusercontent.com/qflen/react-native/6f3799dbb597630a91f43e460437d636796dbbe7/assets/50797/after-tap-zero-after-05.png" width="260"> | | `Last tapped: zero-scale` (stale cache) | `Last tapped: variable (scaleY=0.5)` | iOS behaves similarly. I tried the example on iPhone 17 Pro (Simulator, iOS 26.1) and the zero-scale pressable is correctly untouchable; driving synthetic taps on the simulator wasn't scripted here, so the iOS signal is the four XCTest cases plus manual verification. Reviewed By: fabriziocucci Differential Revision: D102590386 Pulled By: javache fbshipit-source-id: 863a653f17db72ff85135fd8eb344ccd1528eda7
…ent (react#56633) Summary: Pull Request resolved: react#56633 Changelog: [internal] TSIA Reviewed By: javache Differential Revision: D102604056 fbshipit-source-id: 4d965a04b7f3d99a15bf11c8fa9f448425f63e6d
Summary: Pull Request resolved: react#56636 General typing improvements to the `renderApplication` module: - Refactors `renderApplication` to take a single options object instead of positional arguments, and removes the unused `fabric` and `_unused` parameters. The `fabric` parameter was already always treated as `true` internally, and the `_unused` slot was a reserved positional placeholder for a removed `scopedPerformanceLogger` argument. - Migrates the file to `flow strict-local`. - Types `RootComponent` and `WrapperComponent` using component syntax instead of `React.ComponentType`. - Replaces `rootTag: any` with `number | RootTag`, normalizing via the existing `createRootTag` helper for `<AppContainer>` and `Number()` for `Renderer.renderElement`. - Simplifies `ActivityType` (drops the unnecessary spread) and tightens null/boolean checks (`debugName != null`, `useOffscreen === true`). Exports a new `RenderApplicationOptions<Props>` type and updates all internal callers accordingly. Changelog: [Internal] Reviewed By: javache Differential Revision: D102325691 fbshipit-source-id: 9b4f0b16d8508f6eee33c2e6cb8d5b7157c0742a
Summary: Pull Request resolved: react#56628 This test file contains a single `it.skip` test guarded by the comment "Enable once synchronous C++ state updates are re-introduced." That feature was deleted entirely (not just disabled) since 2025-07-15, with the reason: *"The current design of enableSynchronousStateUpdates is not correct and breaks `<Modal />` on Android. let's delete it. Findings: A scan of `packages/react-native/src/private/featureflags`, `scripts/featureflags`, and `ReactCommon/react/featureflags` returns no replacement flag for synchronous C++ state updates. The `enableSynchronousStateUpdates` flag was removed from every layer (JS config, C++, Android Kotlin/JNI, iOS overrides, MC schemas) and the guarded code path in `EventQueue.cpp` was deleted as well. Empirical confirmation: Removed the `it.skip` and ran the test 10x via stress runs. All 10 runs failed deterministically. The test is asserting the *presence* of the bug — i.e., that an intermediate state mutation leaks through when a UI-thread commit pulls transactions from an in-flight JS commit. Without the synchronous-state-updates codepath, that leak no longer occurs, so the test's expected log never matches the actual output: Expected: [Update ScrollView, Update intermediate-state-should-not-be-visible, Update view] Actual: [Update view, Update ScrollView] The test cannot be un-skipped (the bug it documents is no longer reproducible), and rewriting the assertion would defeat the purpose of the test. Deleting it. Changelog: [Internal] Reviewed By: pixelfreak Differential Revision: D102661539 fbshipit-source-id: fca5d552a80cad572ac43047ff6709403de294f1
…eact#56609) Summary: Pull Request resolved: react#56609 `mFpsListener` is injected via `ReactHorizontalScrollView`'s constructor and should live for the lifetime of the view. However, it was declared as a mutable field and nulled in `initView()`, which is called both during construction and when the view is recycled in a virtualized list. After recycling, `isScrollPerfLoggingEnabled()` returned false because `mFpsListener` was null, silently disabling FPS performance tracking for the rest of that view instance's lifecycle. The fix moves `mFpsListener` to a `final` field initialized in the constructor, and removes the `mFpsListener = null` assignment from `initView()`. Applied to `ReactScrollView`, `ReactHorizontalScrollView`, and `ReactNestedScrollView`. This is a safe change because both view managers are instantiated with a single instance of FPS listener, so recycled view instances will always reference the same FPS listener instance. Changelog: [Android][Fixed] - Fix FPS performance listener being cleared on ScrollView recycle Reviewed By: jehartzog Differential Revision: D102359747 fbshipit-source-id: 2707e2d853620ff6980c68edaf9d11354b0825be
Summary: Same fix as react#56584 (targeting 0.83-stable), applied to main. Add a `redbox` subspec to `React-debug.podspec` with its own `header_dir`, following the same pattern `React-Fabric` uses for its subdirectories. Without this, static library builds (without `USE_FRAMEWORKS`) flatten all headers into the top-level `header_dir` (`react/debug`), so imports like `<react/debug/redbox/RedBoxErrorParser.h>` fail because the `redbox/` subdirectory is lost. On main this is currently masked because `test_e2e_ios_templateapp` uses prebuilt `React.framework`, but it would break any from-source static library build that includes `React-CoreModules`. Changelog: [Internal] Pull Request resolved: react#56585 Test Plan: CI Reviewed By: cortinico Differential Revision: D102809818 Pulled By: motiz88 fbshipit-source-id: 9871706a8e028c27f64f4d0dd9942d7c408c6b3b
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Set RCT_HERMES_V1_ENABLED=1 on the resolve-hermes job so the resolver reads HERMES_V1_VERSION_NAME (250829098.0.12) instead of HERMES_VERSION_NAME (1000.0.0), which 404s and falls back to a stale hermes-v0.14.0 engine lacking the facebook::jsi IRuntime ABI. Align the build-from-source fallback tag in .hermesv1version to the same version. Fixes the missing IRuntime symbol link failures in Prebuild macOS Core. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
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
20 participants
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.
Summary
This draft genuinely merges React Native through
78a740d40aa2f9284b4568cb14307628e0f1d3f1(react#56585, the complete RedBox 2.0 introduction), preserves React Native macOS adaptations, regenerates broad-merge artifacts, and ports RedBox 2.0 to AppKit.This replacement supersedes #3047 without modifying or closing it: #3047 uses cherry-pick topology, while this draft preserves the real upstream two-parent merge and its provenance. #3047 remains open and unmodified.
Topology
0.85-mergebasedcd3d60324c220393aae7729bb538235f3c7041678a740d40aa2f9284b4568cb14307628e0f1d3f1c1798b51c75f09c2b8585bd6ceb8ccc2dee31dc756c1f7c25cb4e86c322dc55d48102816508b1315dcd3d603…then78a740d4…e3351d0afc5432314e586f4fd6fd8d516e51355f127f67ee477700dabf536454e9a1f389fe344848c1798b51…c616cd7a66125ca03c04d30c8dd44c2738ef178050b97dd5df97a82b5605e2cdfaab6667dccb6ccfe3351d0a…c499d672e6a43bcb5ba7c00b5ecde6112905f9f9778136a143540a71a21ed42406ccd8590458cbe6c616cd7a…The first-parent count from Microsoft
0.85-mergeto the PR head is exactly four. No commit was rebased, squashed, reworded, or regenerated.Merge scope
git rev-list --count dcd3d60324c220393aae7729bb538235f3c70416..78a740d40aa2f9284b4568cb14307628e0f1d3f1returns 561.The deterministic path-scoped query
git log --oneline \ dcd3d60324c220393aae7729bb538235f3c70416..78a740d40aa2f9284b4568cb14307628e0f1d3f1 \ -- 'packages/react-native/React/CoreModules/RCTRedBox*'returns 10 of 561 commits that touch
RCTRedBox*; 551 do not. The cutoff commit78a740d4(react#56585,Add redbox subspec to React-debug pod) is within those 551 because it does not touch that pathspec.The fetched React Native 0.86 fork point is
d1cdce24cca2224f8d37873d24a49523a8271a0a, derived bygit merge-base origin/0.86-merge facebook/0.86-stable. The cutoff is its ancestor, and:Conflict and correction decisions
A fresh-clone-only
git merge-tree --write-tree dcd3d603… 78a740d4…re-derived 49 conflict paths: 46 content and 3 modify/delete, with 144 stage records. The generated merge tree contains 100 conflict blocks / 300 exact marker terms.The 50-path correction layer is exact-set checked against the finished merge diff:
yarn featureflags --updateoutputyarn build-typessnapshot22 RNM native/SwiftPM adaptations
Exact path list
Full correction inventory
All 50 paths, categories, and provenance
Hermes, Fantom, and generated-artifact rationale
hermes-estreeis realigned to the cutoff's Hermes 0.35 toolchain. CocoaPods'hermes-enginechecksum embeds an absoluteHERMES_CLI_PATH, so the two checksum lines are checkout-path-dependent—not a dependency change. The corrected Pod gate also hides Git from podspec evaluation, preventing 74 podspecs from injecting the ephemeral PR-head SHA into portable checksums.@react-native/fantomdeclares its directistanbul-lib-coverageimport, required by Yarn's strict pnpm linker.The tracked JavaScript API snapshots at
e3351d0aandc499d672are byte-identical (SHA-256a95b55e7c1ef4185d8bf82edf8e93d6a16ad9a4621ae101bba6d7db1c2691127), so the five-file AppKit delta has no JS API change. The repository's diff tool still emits its known historical nine-itemBREAKINGclassification when given those identical inputs; that exact output is retained in gate evidence rather than misrepresented as an AppKit delta.Five-file AppKit delta
The post-correction AppKit delta is exactly five modified files under
packages/react-native/React/CoreModules/: 579 insertions / 15 deletions.[macOS]tagsRCTRedBox.mmRCTRedBox2AnsiParser+Internal.hRCTRedBox2AnsiParser.mmRCTRedBox2Controller+Internal.hRCTRedBox2Controller.mmc616cd7ais the four-file AppKit implementation.c499d672is the separately reviewable reachability fix: it removes the transplanted 0.86 macOS exclusion so flag ON instantiatesRCTRedBox2Controller, while flag OFF retainsRCTRedBoxController.Commands and gates
All dependency and JavaScript commands used a tracked-only archive of tree
778136a1…, Node22.6.0, checked-in Yarn4.12.0, an isolated HOME/XDG/config/cache environment, no auth variables, and effective npm/Yarn registryhttps://registry.npmjs.org. The public-registry host wassaads-macbook-pro; no Office/private registry was loaded.yarn install --immutablejest packages/react-native/Libraries/LogBox --runInBandyarn test-ciGIT_DIR=<nonexistent> RCT_SKIP_CODEGEN=0 bundle _2.5.16_ exec pod installxcodebuild … -sdk macosx -scheme RNTester-macOS -destination generic/platform=macOS … buildxcodebuild … -sdk iphonesimulator -scheme RNTester -destination id=0A9B… … buildxcodebuild … -sdk xrsimulator -scheme RNTester-visionOS -destination id=320E… … build1.16.1;python -m scripts.cxx-api.parser --validateswift package dump-packageinpackages/react-nativeReact, 71 targets, zero dependenciesgit diff --check, exact conflict-marker scan, tag counts, iOS projectionThe correction-only
git diff --check c1798b51… e3351d0a…exits 2 on one four-space line inRCTBackedTextInputDelegateAdapter.mm. Byte evidence shows the identical four spaces already exist in first parentdcd3d603, so the first-parent PR diff and five-file AppKit diff both pass.GUI evidence — PARTIAL, not a full release smoke
The disposable harness used tree
778136a143540a71a21ed42406ccd8590458cbe6, exactly the PR-head tree.RCTRedBox2ControllerRCTRedBoxController5c3f2d0baf1afb011febbfd1e9da07f113bb6ad2209955bc2b9894bfd693202ad9c455accc54cb0a287601fd7130fb33294121606ae70806be68b17a32ada74fThis is partial evidence, not a full release smoke. The residual GUI checklist remains a declared blocker, which is why this PR must remain a draft.
Evidence availability
The correction disclosure is fully inline above. No gist, upload, attachment comment, or follow-up publication action is required. Durable local evidence is at:
That local artifact tree is not available on GitHub. Its exact SHA-256 manifest is supplied to the independent judge before publication.
Supersession
#3047 remains open and unmodified. Close it manually only after this replacement draft is reviewed.
Review checklist
dcd3d603then78a740d4)c499d672reachability fix reviewed (ON=RedBox2 AppKit, OFF=legacy)