[menu] Prevent pointerleave from stealing dialog focus

[atomiks]

• I have followed (at least) the PR section of the contributing guide.

When clicking a menu item that's a dialog trigger with a pointer, focus would be taken from the dialog by onPointerLeave.

[pkg-pr-new]

• base-ui-tanstack-start

• vite-css-base-ui-example

  pnpm add https://pkg.pr.new/@base-ui/react@4125
  

  pnpm add https://pkg.pr.new/@base-ui/utils@4125
  

commit: d3d7c49

[mui-bot]

Bundle size report

Bundle	Parsed size	Gzip size
@base-ui/react	🔺+48B(+0.01%)	🔺+21B(+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	19756af
🔍 Latest deploy log	https://app.netlify.com/projects/base-ui/deploys/69a7e475d1b5df00075ec7da
😎 Deploy Preview	https://deploy-preview-4125--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.

[atomiks]

Codex Review

Overview

This PR updates pointer-leave handling in list navigation so menu focus is only restored when focus is still inside the menu floating element, which prevents nested alert dialog focus from being stolen. It also adds a Chromium-only regression test covering a nested AlertDialog trigger inside Menu.

Findings (None)

No blocking issues found in this patch.

Confidence: 4/5

High confidence from a full branch-diff review (master...c750efaa), targeted unit coverage checks, and a focused hot-path perf spot-check.

Notes

• Full branch scope reviewed (single commit branch).
• Test evidence: pnpm test:jsdom MenuRoot --no-watch and pnpm test:chromium MenuRoot --no-watch both passed.
• Perf spot-check method: Chromium micro-benchmark via @playwright/test comparing unconditional focus vs activeElement + contains + conditional focus over 5 runs × 100,000 iterations.
• Perf spot-check result: mean overhead was 6.26ms per 100,000 iterations (0.063µs per event), which is negligible in practice.