[all components] Fix synthetic event target regressions

[atomiks]

React's synthetic hover, pointer, and focus events can intentionally report a different event.target than the native event's composed-path leaf. These handlers need to keep using the synthetic target when they are making component-level active/inside checks.

Changes

• Restore synthetic event.target checks for ScrollArea hover and scrollbar track interactions.
• Restore synthetic event.target checks in hover and composite handlers that were switched to getTarget(event.nativeEvent).
• Add regression coverage for cases where the synthetic target and native composed path intentionally differ.

[pkg-pr-new]

• base-ui-tanstack-start

• vite-css-base-ui-example

  pnpm add https://pkg.pr.new/mui/base-ui/@base-ui/react@4516
  

  pnpm add https://pkg.pr.new/mui/base-ui/@base-ui/utils@4516
  

commit: 9f602bb

[mui-bot]

Bundle size report

Bundle	Parsed size	Gzip size
@base-ui/react	🔺+41B(+0.01%)	🔺+17B(+0.01%)

Details of bundle changes

---

Check out the code infra dashboard for more information about this PR.

[netlify]

✅ Deploy Preview for base-ui ready!

Name	Link
🔨 Latest commit	9f602bb
🔍 Latest deploy log	https://app.netlify.com/projects/base-ui/deploys/69d37c5c5e293800089707f2
😎 Deploy Preview	https://deploy-preview-4516--base-ui.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[flaviendelangle]

PR Review: [all components] Fix synthetic event target regressions

PR: #4516
Author: atomiks
Labels: type: bug, type: regression, component: scroll area, internal

Context

This PR reverts getTarget(event.nativeEvent) back to event.target in 4 source files. React's synthetic events intentionally normalize event.target to the React component boundary, which is correct for containment checks like contains(ref, target). Using getTarget(nativeEvent) bypasses this and can return a shadow DOM leaf or wrong element, breaking component-level hover/focus/click checks.

The fix is clean and consistent across all 4 locations. Import cleanup is done correctly in both ScrollArea files.

---

Critical Issues (1 found)

• [test-analyzer] Missing test for ScrollAreaScrollbar.tsx thumb click change. The PR modifies the event.currentTarget !== event.target check (line 129) but no test covers this path. A test should render a ScrollArea with a visible thumb, dispatch a pointerdown on the thumb with a divergent composedPath, and verify the component still correctly distinguishes thumb clicks from track clicks.

Important Issues (1 found)

• [code-reviewer] Possible unused getTarget import / nativeEvent destructuring in useHoverReferenceInteraction.ts and useHover.ts. The diff removes the only visible usage of getTarget(nativeEvent) in these handlers but doesn't show whether the import or destructuring is cleaned up. If getTarget/nativeEvent aren't used elsewhere in those files, they should be removed to avoid lint warnings.

Suggestions (2 found)

• [test-analyzer] The useHover.test.tsx composedPath returns [document.body, child] which is an unrealistic ordering (body before child). While it doesn't affect correctness, a more realistic path like [child, button, document.body, ...] would avoid confusing future readers.
• [test-analyzer] The CompositeRoot test asserts select-all behavior (selectionStart=0, selectionEnd=4) which is an implementation detail. A comment explaining why select-all is expected would help, since the primary assertion is that ArrowRight doesn't move focus.

Strengths

• Correct and consistent fix across all 4 locations
• Import cleanup done properly in ScrollArea files
• 3 well-written regression tests with clear naming and consistent patterns
• Tests verify user-visible behavior (focus stability, tooltip persistence, data-hovering), not internals
• Proper async handling with flushMicrotasks()
• The CompositeRoot test covers both focusin and keydown paths in one scenario

Recommended Action

1. Add a regression test for the ScrollAreaScrollbar.tsx thumb click path
2. Verify getTarget/nativeEvent aren't left as dead code in useHover.ts and useHoverReferenceInteraction.ts
3. Consider the cosmetic suggestions for test clarity

[atomiks]

Codex Review (GPT-5.4)

This is a bug-fix PR with secondary tests and react-ui changes. Reviewing the full branch against a freshly fetched master, I do not see any remaining branch-relevant issues: the final patch now makes the right distinction between React-synthetic target checks for hover logic and native composed-path target checks for true thumb-hit detection.

1. Bugs / Issues (None)

I did not find any concrete correctness, regression, or validation gaps that still need action before merge.

Root Cause & Patch Assessment

The final code path split looks correct now. ScrollAreaRoot, useHover, and useHoverReferenceInteraction all switch back to event.target specifically in React synthetic handlers where React’s retargeting is the semantic source of truth for “inside / active trigger” checks, while ScrollAreaScrollbar intentionally keeps a native target lookup for thumb-hit detection, which is the one place where the actual composed-path leaf matters more than the synthetic host target.

That same distinction is reflected in the tests now. The hover regressions cover the synthetic/native target mismatch directly, and the scrollbar regression covers the previously missing thumb-click path rather than only a happy-path hover case.

Test Coverage Assessment

I ran the targeted JSDOM suites for the touched areas:

• pnpm test:jsdom ScrollAreaScrollbar --no-watch
• pnpm test:jsdom useHover --no-watch
• pnpm test:jsdom CompositeRoot --no-watch

Those checks line up well with the actual risk in this PR: hover-state retargeting, wrapper-trigger hover logic, scrollbar thumb-vs-track detection, and the shadow-hosted native-input behavior guarded by the CompositeRoot regression test.

Recommendation

Approve ✅

The bug-fix path now looks internally consistent, the earlier scrollbar test gap has been closed, and the final regression coverage is strong for the exact failure modes this PR is addressing.