The app was made for a DualSense and took everything else for a lesser one of those. Players
without one could play on the PC, with the headset's controllers, and not on the headset.
The headset's own controllers: the app reads them through the headset's runtime (OpenXR
actions, the layout the PC host has: A cross, B square, X circle, Y triangle and both
together to blow, grips L1/R1, triggers L2/R2, right stick the finger on the touchpad, left
menu button OPTIONS, both sticks in to reset the view), the right one is the controller in
the game with where it is and how it points, both shake with the game's rumble. The gamepad
the system makes of them for other apps is no longer taken for a gamepad: it says neither
where it is nor whether anybody holds it.
Which of the two: a small router (quest-host/cpp/pad_router.h) between the activity, which
hears of gamepads, and the session, which reads the controllers. The game is played with the
one used last. From a gamepad only a button under a thumb or a stick takes over, never a
trigger or a grip: while a gamepad is held it is hands the headset sees. What buttons mean
about the view besides (the first press of X takes the seat, OPTIONS held resets the view)
moved there from the activity, so that it is the same for both. The start-up menu takes
either; what confirmed it is not pressed in the game.
Gamepads that are not a PlayStation's:
- The right stick as the touchpad's finger is Input::StickFinger now, as on the PC: the
finger lifts where the stick was let go, which the catapult at the end of a level needs.
A button that presses the touchpad touches it in its middle.
- A gamepad without motion sensors followed the hands to where they were and kept pointing
straight ahead. Now the hands turn and tilt it (quest-host/cpp/held_pad.h): the line
between the palms is its sideways axis, the way the hands point is its tilt. The app says
so with a new flag of the pad pose message (HeldInHands), and the emulator lets go of a
controller that was a tracked device once it no longer is (it stayed where that device
was last seen for as long as no motion sensor of its own spoke up).
- Motion sensors are only taken from PlayStation controllers: which way is up for one is
each maker's own affair.
- A right stick on RX/RY, triggers that are only buttons, a Back button the system takes
for its own "back".
For trying a build on a headset whose owner plays with the installed one:
quest-host/build.sh --test-package builds the app under another name, which installs next to
it; the test tools take ASTRO_PACKAGE. The self-test can report a controller the way a
session does (HOST_PAD) and keep pictures of it (HOST_SHOTS).
Tried on a Quest 3 without wearing it: the runtime takes the bindings, the app starts and
plays with a made-up gamepad, and in the game's picture the controller is where and how the
self-test says a tracked controller or hands around a gamepad are, and comes back when they
are lost. tools/tests/pad_router_test.cpp and held_pad_test.cpp try the rest on the PC.
Not tried: any of it worn, and no gamepad other than a DualSense.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
