feat(cua-driver-rs)(macos): enable Chromium/Electron AX trees (AXManualAccessibility) for get_window_state

[f-trycua]

Problem

Chromium-based apps (Arc, VS Code, Electron shells) ship their web-content accessibility tree off and only build it once an assistive client requests it. Without enablement, the first get_window_state AX walk of such an app returns an empty / title-bar-only tree — the macOS/AX analog of the empty-tree symptom tracked in #1616 (VS Code exposing only its title-bar elements). With enablement the full tree materializes.

Change

In walk_tree's app-setup path (right after AXUIElementCreateApplication(pid)):

1. Enable — set AXManualAccessibility = true on the app root (the modern, side-effect-free opt-in). If the app reports the attribute unsupported (kAXErrorAttributeUnsupported / -25205), fall back to AXEnhancedUserInterface = true (the legacy flag some Electron builds expose instead). Native Cocoa apps (Finder, Calculator, TextEdit) reject both — they take no further action.
2. Settle — the tree is built asynchronously over IPC once the app detects the client, so a walk that starts immediately still sees only the chrome. When the flip actually took, pump the run loop ~500ms (reusing the existing pump_run_loop_briefly helper) so the tree has time to materialize before the walk reads it.
3. Per-pid cache — a process-global HashSet<pid> (guarded by a Mutex in a OnceLock) records pids we've already enabled + settled. Repeat snapshots of the same app skip the settle; the tree is already built and stays built for the life of the process.

The settle delay fires only when enablement actually flips an app from unsupported→supported, so apps that never needed it (every native Cocoa app) pay zero added latency.

Hard dependency

This is intentionally higher-risk and must land after the Phase 1 messaging-timeout PR. Enablement makes the tree bigger (fully materialized instead of empty/partial) — which is exactly why it depends on the per-element AX messaging timeout (so a wedged element fails fast rather than stalling the now-larger walk) plus the node cap to keep the materialized tree bounded. Shipping this without those bounds would trade "empty/partial" for "complete but potentially huge."

Test

• cargo build --release -p cua-driver passes.
• cargo test --no-run -p platform-macos compiles cleanly (the crate's test binary has a pre-existing, unrelated libswift_Concurrency.dylib rpath issue at the run step only).
• macOS-only change; platform-windows / platform-linux are untouched.

Refs #1616

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs	Ignored	Preview	May 29, 2026 4:38pm

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: abe4a1c9-6d29-45e2-bc43-d4e15ec1e093

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/ax-chromium-enablement

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}