[fredybp]

Closes #25

Adds Sound Connect's multipoint controls to the macOS app: see which devices are connected to the headset, which one has playback, switch playback to another device, and "fix" (pin) it — e.g. pin playback to the Mac before a call so the phone can't take over the headphones.

What's added

Devices card (v2 headsets that advertise it), right after Ambient Sound Control:

• Connected and paired devices, with the playback device (speaker icon), the fixed state (lock) and a "This Mac" tag
• Fix / Unfix playback
• Per-device menu: "Switch playback device and fix" (Sound Connect's action) and "Switch playback device"
• "Connect to 2 devices" on/off (read-only for now)
• Loading state until the first read lands; spinner in the header during refreshes and switches
• "Copy diagnostics": hex log of every multipoint request/reply, to help bring up other models

Menu-bar item

• Pin Playback to This Mac, Unfix Playback, Connect, Open SonyBridge, Quit
• The icon shows the playback device next to the headphones (phone, computer, or its slot number) and fills in while playback is fixed to this Mac
• The device list is re-read every 10 s so the icon follows changes made from the phone

Protocol

The multipoint commands live in command table 2 (frame type 0x0e, DATA_MDR_NO2). BluetoothWrapper can now send on it, match replies on it and ACK its frames; existing table 1 calls are unchanged (new default parameter).

	Request	Reply
Support functions	06 00	07 00 n (function, priority)… — needs 0x31 and 0x32/0x30
Device list	36 02 (36 00 fallback)	37 02 n {addr[17] slot cod[3] len name}… playbackSlot
Fixed state	36 01	37 01 v (0 = fixed, 1 = free)
Switch playback	3c 01 addr[17]	3d 01 result addr
Fix / unfix	38 01 00 / 38 01 01	39 01 value result
Multipoint setting (table 1)	d0 d1..d4 01 to find MULTIPOINT_SETTING, then d6 slot	d7 slot 00 v (0 = on)

Safety:

• Table 2 commands only go out with frame type 0x0e (in table 1, 0x30–0x3f are the firmware-update commands), only to v2 devices, and only if the init reply advertises table 2. v1 devices get nothing new.
• Discovery is read-only. Writes are refused locally unless the headset advertised function 0x31, and switching only targets devices the headset lists as connected.
• "Switch and fix" = unfix if fixed → switch → fix; if the switch is refused, it re-fixes the previous device. A plain switch unfixes afterwards, since the headset keeps "fixed" on across a switch.
• The state is re-read from the headset after every write.

Opcodes and layouts come from mos9527/SonyHeadphonesClient (ProtocolV2T2.hpp, generated from Sound Connect) and mehrshaad/nullpoint.

Tested

WH-1000XM5, firmware 2.5.1, on macOS:

• Device list, playback device and fixed state match Sound Connect
• Switch to the iPhone and back, switch and fix, plain switch ends unfixed
• Unfix during a call is refused by the headset (39 01 00 02, FAIL_CALLING), so playback stays fixed to the Mac
• Multipoint setting found in slot D2 (ON)

I also replayed the WF-1000XM5 captures from mos9527's repo through the parser.

Not covered: the Windows/Linux UI (the core changes are shared and keep the existing call sites, but I only built the macOS app), disconnect/unpair, and writing the multipoint on/off setting.

macOS 12 needed to build with Xcode 27.2

Xcode 27.2 no longer builds the project as-is: it rejects MACOSX_DEPLOYMENT_TARGET = 11.0 ("the range of supported deployment target versions is 12.0 to 27.2.x"). To build and test this I raised it to 12.0 in all four build configurations of SonyHeadphonesClient.xcodeproj. I left that change out of this PR so the app keeps targeting macOS 11 — it's a one-line-per-config change in this commit on my fork if you want to take it too.

Disclosure

This was mainly vibe coded (written with an AI assistant; I drove it and tested it on my headphones). If that's a problem for the project, feel free to reject it.

[AmitRajput-Dev]

Merged via #30. Thank you, multipoint was the most requested feature (#25).

A review before merging turned up a few issues, all fixed in the merge:

• Possible crash on disconnect: the multipoint queue captured a raw Headphones*, which could be freed while a read was still running. Everything now runs on the single serial I/O queue with a shared_ptr.
• Slower first read: the general-settings slot search ran before the first battery/EQ update, which could delay it by ~10 s. It now runs after the capability probes.
• Lost after a reconnect: after the headset dropped the link, a reconnect skipped the init handshake, so table 2 and multipoint went missing.
• Stale switch/fix decisions: they read the device list on tap. They're now made inside the queued write.
• Card shown too broadly: the Devices card only appears when the headset actually returned a device list or a pin state.

The menu-bar item now opens a quick-controls panel, and your menu is still there on right-click. Ships in v0.4.0.