Honour USBMUXD_SOCKET_ADDRESS in the device listener

[Folkman]

The problem

libusbmuxd reads USBMUXD_SOCKET_ADDRESS so the idevice_* tools can drive a device attached to a different machine. dl_connect() doesn't go through libusbmuxd — the comment above it explains that usbmuxd_subscribe() is threaded and does blocking reads while this listener wants a select()-able fd — so it opens the usbmuxd socket itself, and always used the local Unix socket.

On a host configured for a remote usbmuxd, that produces a confusing split:

$ export USBMUXD_SOCKET_ADDRESS=10.0.0.2:27015
$ idevice_id -l
<udid>
$ ios_webkit_debug_proxy
No device found, is it plugged in?

Same machine, same library, same phone. This comes up whenever the device is attached to a Windows or macOS box and the proxy runs on a Linux host (or in a VM/container) reaching usbmuxd over TCP.

The change

dl_connect() now reads the variable, mirroring libusbmuxd's connect_usbmuxd_socket() so the two agree on what an address means:

USBMUXD_SOCKET_ADDRESS	Behaviour
unset / empty	local Unix socket (unchanged)
UNIX:/path	that Unix socket — exact case, as upstream
host:port	TCP; port must be an entire, valid number in 1..65535
[::1]:27015	brackets stripped before getaddrinfo()
no port, host:, host:abc, host:99999	falls back to the local socket, as upstream

Deliberately matching libusbmuxd rather than being more permissive: a bare host with no port does not default to 27015, and a malformed address falls back rather than failing, because that is what every other libimobiledevice tool does with the same string. UNIX: is matched case-sensitively for the same reason.

On provenance: the behaviour is matched to libusbmuxd's connect_usbmuxd_socket() (LGPL-2.1) so the two agree on what an address means, but the implementation here was written independently against this BSD-licensed file — no code was copied from it. The differences are visible in the approach: this version parses in place without allocating, and calls getaddrinfo() directly rather than going through a socket_connect() helper.

Addresses resolve through getaddrinfo() with AF_UNSPEC, so IPv6 remotes work. A resolver failure now reports gai_strerror() instead of a bare message.

Testing

Built with the project's own -Wall -Werror, no new warnings.

Verified against a real device (iPhone SE, iOS 15.8.8) attached to a Windows host running usbmuxd, with the proxy on Linux:

• <host>:27015 — device found, proxy serves tabs on 9222.
• [::1]:27015 — before this change, cannot resolve "[::1]:27015"; after, it reaches connect() (Connection refused, nothing listening there), confirming the brackets are stripped.
• <host> / <host>: / <host>:abc / <host>:99999 / UNIX:/var/run/usbmuxd / lowercase unix:... / unset — all fall back to the local socket.
• With the variable unset the Unix-socket path is byte-for-byte the original code.

---

Written with Claude Code.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.