Process responder events for touches whose target unmounted by lazerg · Pull Request #59522 · react/react-native

lazerg · 2026-10-27T01:05:25Z

Summary:

With enableNativeEventTargetEventDispatching on, a touch that ends after its view unmounted never reaches the responder, so the old responder sticks around and trackedTouchCount stays one too high, which drops the next tap. processResponderEvent already handles a null target, but dispatchNativeEvent only accepts a non-null one and would crash in dispatchTrustedEvent if it got null. It now takes EventTarget | null and returns right after the responder step when there is no target. The renderer's dispatchEvent (ReactFabricEventEmitter.js in react/react) still skips the call for a null target, so that guard has to go too before the bug is fully fixed, and this change needs to land first.

Fixes #59519

Changelog:

[GENERAL] [FIXED] - Fix the responder not being released when a touch ends after its view unmounted

Test Plan:

Added a Fantom test to ResponderEventTarget-itest.js that grants the responder, sends topTouchEnd with a null target through dispatchNativeEvent, and checks that the release fires and the next touch is granted again. I couldn't build the Fantom runner locally, so I also ran the null-target case as a throwaway Jest test: on main it throws TypeError: Cannot read properties of null from dispatchTrustedEvent, and it passes with this change. flow check, eslint and prettier are clean.

github-actions · 2026-10-27T09:05:23Z

Visual regression

1 visual difference(s) at 2d382197e434. Visual differences need maintainer approval. Review the report, then apply visual-change-approved.

ios `scroll-view/horizontal-scroll-to/end`: Changed, 69504 px (18.123%)

Expected and Captured

Expected	Captured

meta-codesync · 2026-10-27T11:39:12Z

@rubennorte has imported this pull request. If you are a Meta employee, you can view this in D127796330.

rubennorte · 2026-10-27T11:40:44Z

    expect(onResponderRelease).toHaveBeenCalledTimes(1);
  });

+  (ReactNativeFeatureFlags.enableNativeEventTargetEventDispatching()


If the behavior should be the same as before, this test should run for both versions of the feature flag and behave the same way

I left this as is for now because the flag-on run can't pass in this PR alone. With the flag off, a touch end with a null target already goes through the legacy responder plugin and releases the responder. With the flag on, ReactFabric's dispatchEvent drops events with a null target, so they never reach dispatchNativeEvent (react/react#38006). After that guard is removed, the test can run for both flag values with the same expectations. Should I keep it here so it lands with the renderer sync, or add it in a follow-up?

I flipped the gate in 6208d24, so the test now runs with the flag off and is skipped with the flag on. With the flag on, the synced renderer still drops touch events whose target is null before they reach dispatchNativeEvent, so that run can't pass until the renderer sync lands (react/react#38006). I left a TODO on the skip to turn it on after the sync.

rubennorte · 2026-10-27T11:43:39Z

+      expect(onResponderGrant).toHaveBeenCalledTimes(1);
+
+      // The target is null when the touched view unmounted during the touch.
+      dispatchNativeEvent(null, 'topTouchEnd', {


This is an internal API and we should avoid using them in Fantom tests that should be more e2e. We should find a way to test and verify this using externally visible behaviors (e.g.: dispatching the event while unmounting the component it in the test).

I'll remove the dispatchNativeEvent import. This case doesn't need a Fantom API change: I can queue the touch end with Fantom.enqueueNativeEvent, unmount the view, and then flush the queue. The native side then sends the event with a null target, the same as on a device. The flag-on run has the blocker from the other thread, so I'll push this test after we agree on how to handle it.

Done in 6208d24. The test now uses only Fantom APIs. It queues the touch end, unmounts the view before the event is flushed, and then checks that a touch on a new view gets the responder.

rubennorte · 2026-10-27T11:45:56Z

 */
 export default function dispatchNativeEvent(
-  target: EventTarget,
+  target: EventTarget | null,


I'd use target: ?EventTarget in case we slip an undefined here from React instead of null.

Changed to ?EventTarget in 6208d24, also in processResponderEvent. The existing == null check already treats undefined like null.

rubennorte · 2026-10-27T11:46:39Z

Good catch! Thanks for finding this and submitting a potential fix. Could you please address the comments so we can merge it?

…d target

meta-codesync · 2026-10-27T15:01:19Z

@rubennorte merged this pull request in f754dbd.

react-native-bot · 2026-10-27T15:01:21Z

This pull request was successfully merged by @lazerg in f754dbd

^{When will my fix make it into a release? | How to file a pick request?}

## Summary With `enableNativeEventTargetEventDispatching` on, the Fabric renderer skipped `dispatchNativeEvent` when the touched view had already unmounted, so the touch end never reached the JS responder and the next tap got lost. This passes the null target through instead, so React Native can still release the responder. It needs react/react-native#59522 to land first, since `dispatchNativeEvent` there throws on a null target today. Fixes #38006 ## How did you test this change? Added a test in `ReactFabric-test.internal.js` that unmounts a view, dispatches a touch end to its old instance handle, and expects `dispatchNativeEvent` to get a null target. It fails on main and passes with this change. Also ran `yarn test packages/react-native-renderer` (default, `--prod`, www-modern), `yarn flow fabric`, `yarn linc` and `yarn prettier-check`.

## Summary With `enableNativeEventTargetEventDispatching` on, the Fabric renderer skipped `dispatchNativeEvent` when the touched view had already unmounted, so the touch end never reached the JS responder and the next tap got lost. This passes the null target through instead, so React Native can still release the responder. It needs react/react-native#59522 to land first, since `dispatchNativeEvent` there throws on a null target today. Fixes #38006 ## How did you test this change? Added a test in `ReactFabric-test.internal.js` that unmounts a view, dispatches a touch end to its old instance handle, and expects `dispatchNativeEvent` to get a null target. It fails on main and passes with this change. Also ran `yarn test packages/react-native-renderer` (default, `--prod`, www-modern), `yarn flow fabric`, `yarn linc` and `yarn prettier-check`. DiffTrain build for [dc9aed3](dc9aed3)

Process responder events for touches whose target unmounted

2d38219

meta-cla Bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Oct 27, 2026

facebook-github-tools Bot added the Shared with Meta Applied via automation to indicate that an Issue or Pull Request has been shared with the team. label Oct 27, 2026

lazerg mentioned this pull request Oct 27, 2026

Bug: Fabric renderer should call dispatchNativeEvent when the event target is null react/react#38006

Closed

cursor Bot mentioned this pull request Oct 27, 2026

Bug: Fabric renderer should call dispatchNativeEvent when the event target is null ghemlick/issue-triage#139

Open

lazerg mentioned this pull request Oct 27, 2026

[RN] Call dispatchNativeEvent when the event target is null react/react#38007

Merged

rubennorte reviewed Oct 27, 2026

View reviewed changes

Test the unmounted touch target through Fantom and accept an undefine…

6208d24

…d target

meta-codesync Bot closed this in f754dbd Oct 27, 2026

meta-codesync Bot added the Merged This PR has been merged. label Oct 27, 2026

Conversation

lazerg commented Oct 27, 2026

Summary:

Changelog:

Test Plan:

Uh oh!

github-actions Bot commented Oct 27, 2026

Visual regression

ios scroll-view/horizontal-scroll-to/end: Changed, 69504 px (18.123%)

Uh oh!

meta-codesync Bot commented Oct 27, 2026

Uh oh!

rubennorte Oct 27, 2026

Choose a reason for hiding this comment

Uh oh!

lazerg Oct 27, 2026

Choose a reason for hiding this comment

Uh oh!

lazerg Oct 27, 2026

Choose a reason for hiding this comment

Uh oh!

rubennorte Oct 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lazerg Oct 27, 2026

Choose a reason for hiding this comment

Uh oh!

lazerg Oct 27, 2026

Choose a reason for hiding this comment

Uh oh!

rubennorte Oct 27, 2026

Choose a reason for hiding this comment

Uh oh!

lazerg Oct 27, 2026

Choose a reason for hiding this comment

Uh oh!

rubennorte commented Oct 27, 2026

Uh oh!

meta-codesync Bot commented Oct 27, 2026

Uh oh!

react-native-bot commented Oct 27, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

ios `scroll-view/horizontal-scroll-to/end`: Changed, 69504 px (18.123%)

rubennorte Oct 27, 2026 •

edited

Loading