[JakePre]

What and why

With Press a button to join on (DM_PAD_PRESS_TO_JOIN=1), a Switch 2 controller couldn't be used. Holding Sync paired it and it connected, then it was dropped at once. A button press only woke it to be dropped again. On an iPhone, the log showed:

controller 'Joy-Con 2' (Bluetooth) connected: press a button to play (DM_PAD_PRESS_TO_JOIN)
[sw2] Joy-Con 2 (L): no free player, not used
[sw2] Joy-Con 2 (L) disconnected

The front end deliberately gives a new pad no player until its first button (s2pad_attach, then s2pad_read through pad_join). But the transport took "no player after attach" to mean "no free player" and let the controller go. The Mac has the same path. It wasn't seen because every run I made had the setting off. That's on me: it came in with #1346.

The change:

• The front end's attach now returns whether it keeps the pad (pad_switch2_apple.h): 1 with a player or waiting for its first press, 0 with no free player. Only 0 lets the controller go.
• A pad that gets its player later, at that press, gets its Joy-Con styles (dm_joy_attach through s2_sync_sides) and its player lights at the next report. Before, both came only at connection.
• A Joy-Con split off a pair (SL + SR held) waits for a press the same way when the setting is on. With no free player, it goes back into the pair as before.
• The log says a player once a button is pressed for a pad that waits.

front_mac.m returns the new value. #1900's iOS wiring does the same.

Checked

• Tests: make -C dm/host/tests GAME=0100000000010000 run passes. lint_c.py tidy: 0 findings above the baseline.
• Mac (M1 Pro, macOS 27.2): Nintendo Switch 2 Welcome Tour (040057C02159A000, with game: Nintendo Switch 2 Welcome Tour plays on the Mac #1425), a left and a right Joy-Con 2, DM_PAD_PRESS_TO_JOIN=1.
  • Sync on each: both connected and stayed (a player once a button is pressed, the right one joining the left as a pair).
  • The first press: a button pressed: player 1 (npad 0).
  • Rumble played, and the mouse moved the cursor. No disconnects until quitting with Home.
• iPhone 17 (with ios: Switch 2 controllers on iPhone and iPad #1900): Super Mario Galaxy (010099C022B96000, v327680), the same two Joy-Cons.
  • Sync on each, then a press: player 1, as a pair.
  • Splitting one off (SL + SR) made it player 2 at once (its held buttons count as the press). L + R held made them one pair again.
• Known detail: while a pad waits, its lights show player 1, since the light code's fallback for "no player" is the first pattern. They change at the join if it becomes another player. I didn't pick an "off" or flashing pattern, since I don't know what the Joy-Con 2 takes for that.
• Not checked: the Pro Controller 2; more than two controllers waiting at once.

Docs

• Nothing for the player docs: they already say a controller joins at its first press
• No chain tool changed
• No game files, keys or extracted game data

🤖 Generated with Claude Code

[seedds]

Landed on main as 7ef8339, with you as co-author. Thanks for the diagnosis and the fix.

Your three files are in as written, with one change in pad_switch2_apple.m: the player check you added in report: is now a helper, playerSync, which leavePair calls too. With press to join on, a Joy-Con split off a pair (SL + SR) becomes a player inside read(), since the held buttons are the press. In the PR, s2_sync_sides ran before that and _player was set right after, so the check in report: never fired and the split-off Joy-Con never got dm_joy_attach for its new player.

I have no Joy-Con 2 here, so this is checked by the host build and the host tests only. Could you retest on main with press to join on: Sync, first press, then a split, and whether the game sees the split-off one as a single Joy-Con?

#1900 needs a rebase onto main now; its host part is this commit.

[JakePre]

Retested on main as asked (7ef8339, built at dfc5a15), on the Mac (M1 Pro) with a left and a right Joy-Con 2, DM_SW2=1 and DM_PAD_PRESS_TO_JOIN=1. I added two log lines locally, not in any PR: one in dm_joy_attach and one in styles_sync where it tells an npad a new style. The split-off Joy-Con gets its single style now:

Super Smash Bros. Ultimate (supported styles 0x3f):

• Both Joy-Cons connected and waited ("a player once a button is pressed"). At the first press they became player 1: dm_joy_attach player 1 sides 3, and npad 0 went FullKey → JoyDual.
• SL + SR on the right one: dm_joy_attach player 1 sides 1, and npad 0 went JoyDual → JoyLeft. Then "a button pressed: player 2", dm_joy_attach player 2 sides 2, and npad 1 went 0 → FullKey → JoyRight. The log then said "Joy-Con 2 (R) plays alone now: player 2 (SL + SR held); the other Joy-Con stays player 1".
• One small thing: npad 1 is FullKey for the moment between read() giving it the player and playerSync's dm_joy_attach, so the game gets two style signals for it in a row. Smash took it without trouble.

Nintendo Switch 2 Welcome Tour (supported styles 0x7, no single Joy-Con styles): the same joins and split, with the same dm_joy_attach calls. Both npads went to FullKey after the split, as npad_resolve maps an unsupported JoyLeft / JoyRight. Its controller applet then asked for 1 player and folded player 2 into player 1.

Not run: the Pro Controller 2, and more than two controllers waiting.