Error: 'gh pr view' is disabled in your shell. Instead, use your builtin git_view_pr command.

[jkelleyrtp]

Summary

Every one-shot ADB invocation in AdbClient used blocking std::process::Command::output() with no timeout, so each ADB-backed discovery/orientation/button/accessibility call parked a thread until adb exited (and hung forever if it never did). This replaces that with a single async runner on tokio::process::Command and propagates async through the Android callers.

The core change is one runner that owns stdio, child cleanup, and the timeout:

pub const DEFAULT_ADB_TIMEOUT: Duration = Duration::from_secs(30);

async fn run(&self, kind: &str, mut cmd: Command) -> Result<Output> {
    cmd.stdin(null()).stdout(piped()).stderr(piped()).kill_on_drop(true);
    let child = cmd.spawn().with_context(|| /* "ADB binary not found at '{path}'..." */)?;
    timeout(self.timeout, child.wait_with_output()).await
        .map_err(|_| anyhow!("adb {kind} command timed out after {}s (adb path: '{}')", ..))?
        ...
}

kill_on_drop(true) means a timed-out adb is terminated rather than leaked. The timeout is configurable per client via AdbClient::with_timeout(Duration); connected_devices still runs adb devices without -s <serial>.

All one-shot methods (shell, shell_raw, exec_out, command, check_connection, connected_devices, resolved_serial, get_screen_size, screenshot, tap, swipe, key_event, input_text, dump_ui*, launch_app, stop_app, get_current_activity) are now async, and the UI-dump retry backoff uses tokio::time::sleep.

Propagation, driven by the above:

• AccessibilityReader::capture_screen becomes future-returning (Android capture calls ADB); macOS/Windows/X11/iOS-Simulator impls just wrap their existing synchronous native capture, and TargetedAccessibility::capture_screen gets an async dispatch macro.
• AndroidAccessibility::new / with_adb_path / refresh_screen_size, set_device_orientation (orientation polling now uses tokio::time::sleep), spawn_input_worker, spawn_ax_worker, and EmulatorSession::{start, set_orientation, send_input} are async; the workers keep their dedicated-thread + current-thread-runtime architecture.
• Emulator discovery file/dir reads use tokio::fs; CLI screenshot writes use tokio::fs::write.

Deliberately unchanged: screenrecord/video workers (long-lived children whose stdout is pumped by dedicated threads), macOS/iOS xcrun calls behind synchronous native APIs, and ListenerHandle::stop_blocking.

New tests cover the timeout path (fake adb = /bin/sleep, asserting the call returns fast with a "timed out" error) and the missing-binary context.

Link to Devin session: https://dioxus.staging.devinenterprise.com/sessions/d88fc07ea8064b38a4b4e5a7481ffebc
Open in Devin Desktop: https://dioxus.staging.devinenterprise.com/desktop/session/d88fc07ea8064b38a4b4e5a7481ffebc?variant=devin-insiders
Requested by: @jkelleyrtp

[staging-devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR that start with 'DevinAI' or '@devin'.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[staging-devin-ai-integration]

Folded into #8, which now contains these commits and targets main.