Automatically kill running ADB server before opening USB devices

[Averroeskw]

Fixes #1

Problem

A running stock ADB server captures connected USB devices, so adb_remote_auth usb … fails to claim the interface until the user manually runs adb kill-server.

Change

• adb_client: adds ADBServer::kill_if_running(). It probes the server address (default 127.0.0.1:5037) with a plain TCP connect and, only if something is listening, issues host:kill over the ADB smart protocol. Unlike the existing ADBServer::kill(), it never goes through connect(), which would spawn adb start-server first — so no adb binary is required and nothing is started just to be killed. If no server is listening it logs at debug level and returns Ok(()).
• adb_cli: the usb subcommand now calls ADBServer::default().kill_if_running() before opening the USB device. Failures are logged as a warning and do not abort the command (the device open path will surface any real problem).

The local subcommand is untouched — it needs the server and still starts it.

Verification

Tested on macOS (no Ai Pin needed for this one):

• With a running server (adb start-server, confirmed listening on :5037): adb_cli usb shell ls logs killing currently running ADB server..., the server process is gone afterwards, and the command proceeds to USB autodetection.
• With no server running: logs no ADB server listening on 127.0.0.1:5037 at debug level and proceeds directly, no error and no server spawned.
• cargo build --release and cargo clippy --release are clean (no new warnings vs main).

🤖 Generated with Claude Code

[agg23]

I'm going back and forth on whether or not this is a good idea. I guess this would need to be a thing exclusively because this is a special version of the ADB server that can do our auth, but at the same time it seems weird to patch the client to shut down other servers.

[Averroeskw]

Fair point — killing another server unprompted is surprising behavior. Reworked: by default a running server now just produces an actionable error (a running ADB server is holding USB devices. Stop it with adb kill-server or re-run with --kill-server.), and the kill only happens with an explicit --kill-server flag. The probe (ADBServer::is_running()) is side-effect-free and nothing is ever started or stopped without asking.

One request: please hold off merging either PR for now — my Pin's interposer is still inbound, and I'd like to verify everything end-to-end on the actual device before this lands.

[Averroeskw]

Closing for now — pulling this back until I've verified the full stack end-to-end on the physical device (interposer inbound). Will reopen once it's hardware-verified. The branch lives on in my fork history if anyone needs the diff in the meantime.

[agg23]

Neither of these changes require an Ai Pin or an interposer; you just need an Android device (and you could even use the emulator if you so desire).