Gamepad input example doesn't detect buttons, only some axes

[Blechlawine]

Bevy version

0.16.0

What you did

I copied the code from https://github.com/bevyengine/bevy/blob/main/examples/input/gamepad_input_events.rs and ran it locally.

What went wrong

My joystick (Virpil WarBRD-D (base) and Virpil Alpha Prime(stick)) get detected correctly and 5 out of 6 axes work and print their events. But none of the buttons show anything and the last (6th axis also doesn't do anything)

I also tried the example from gilrs (https://gitlab.com/gilrs-project/gilrs/-/blob/master/gilrs/examples/ev.rs?ref_type=heads) with gilrs version 0.10.10, which works perfectly, all axes and buttons print events there.

I also tested the gamepad viewer in the browser (https://bevyengine.org/examples/tools/gamepad-viewer/), there the axes are mapped differently from when I run that example locally, and in the browser some buttons are detected, but locally they aren't again.

https://hardwaretester.com/gamepad also detects all buttons and axes correctly.

Additional information

I'm on Linux, Zorin OS 17 (based on Ubuntu).
I have calibrated and configured the joystick under windows, so there shouldn't be an issue with the hardware (also since the gilrs example works)

[Blechlawine]

I think i found the cause of this issue, the events that I'm seeing with the gilrs example are for Unknown buttons and axies and in

bevy/crates/bevy_gilrs/src/converter.rs

Line 24 in 7f04906

gilrs::Button::Unknown => None,

and

bevy/crates/bevy_gilrs/src/converter.rs

Line 39 in 7f04906

gilrs::Axis::Unknown | gilrs::Axis::DPadX | gilrs::Axis::DPadY => None,

events for unknown buttons and axies are discarded, due to the continue here:

bevy/crates/bevy_gilrs/src/gilrs_system.rs

Line 84 in 7f04906

continue;

and here:

bevy/crates/bevy_gilrs/src/gilrs_system.rs

Line 100 in 7f04906

continue;

I don't know what the solution would look like. I'd guess the unknown events could map to GamepadButton::Other(u8) (https://docs.rs/bevy/latest/bevy/input/gamepad/enum.GamepadButton.html#variant.Other) and GamepadAxis::Other(u8) (https://docs.rs/bevy/latest/bevy/input/gamepad/enum.GamepadAxis.html#variant.Other)

This is a slice of the output I see when running the gilrs example

Event { id: GamepadId(0), event: ButtonRepeated(Unknown, Code(EvCode(EvCode { kind: 1, code: 288 }))), time: SystemTime { tv_sec: 1745856256, tv_nsec: 396988000 } }
Event { id: GamepadId(0), event: AxisChanged(Unknown, -0.952, Code(EvCode(EvCode { kind: 3, code: 6 }))), time: SystemTime { tv_sec: 1745856256, tv_nsec: 405011000 } }
Power info of gamepad 0(VIRPIL Controls 20241226 L-VPC Stick WarBRD-D): Wired
Event { id: GamepadId(0), event: AxisChanged(Unknown, -0.9569143, Code(EvCode(EvCode { kind: 3, code: 6 }))), time: SystemTime { tv_sec: 1745856256, tv_nsec: 425011000 } }
Event { id: GamepadId(0), event: ButtonRepeated(Unknown, Code(EvCode(EvCode { kind: 1, code: 288 }))), time: SystemTime { tv_sec: 1745856256, tv_nsec: 426988000 } }
Event { id: GamepadId(0), event: ButtonRepeated(Unknown, Code(EvCode(EvCode { kind: 1, code: 288 }))), time: SystemTime { tv_sec: 1745856256, tv_nsec: 456988000 } }
Event { id: GamepadId(0), event: AxisChanged(Unknown, -0.9629714, Code(EvCode(EvCode { kind: 3, code: 6 }))), time: SystemTime { tv_sec: 1745856256, tv_nsec: 461011000 } }
Event { id: GamepadId(0), event: ButtonRepeated(Unknown, Code(EvCode(EvCode { kind: 1, code: 288 }))), time: SystemTime { tv_sec: 1745856256, tv_nsec: 486988000 } }
Event { id: GamepadId(0), event: ButtonRepeated(Unknown, Code(EvCode(EvCode { kind: 1, code: 288 }))), time: SystemTime { tv_sec: 1745856256, tv_nsec: 516988000 } }
Power info of gamepad 0(VIRPIL Controls 20241226 L-VPC Stick WarBRD-D): Wired
Event { id: GamepadId(0), event: ButtonRepeated(Unknown, Code(EvCode(EvCode { kind: 1, code: 288 }))), time: SystemTime { tv_sec: 1745856256, tv_nsec: 546988000 } }
Event { id: GamepadId(0), event: ButtonRepeated(Unknown, Code(EvCode(EvCode { kind: 1, code: 288 }))), time: SystemTime { tv_sec: 1745856256, tv_nsec: 576988000 } }

I noticed that every button and axis has a different value in EvCode::code, but that value doesn't fit into a u8 (as can be seen here due to the 288)

---

I found a similar issue from 2021 and a pull request from 2023:
#1916
#8560

and i did some testing myself and these changes seem to fix it:
main...Blechlawine:bevy:fix/gamepad-buttons-issue-18958

checking over the differences between platforms in implementation in gilrs, it doesn't seem like there should be any compatibility issues across platforms, other than the code for each button/axis being different between platforms.

[clangdo]

I am using a steam controller on Linux, and I lack analog trigger input events. After looking into the code, and testing with bare gilrs, I can be pretty sure that it's the same issue. Gilrs works fine, but Bevy's gilrs plugin just drops everything that's not for one of the standard buttons/axes, as you pointed out.

What's potentially worse, the steam controller is firing the appropriate event codes! These are standard 0x12, 0x13, 0x14, 0x15, for controllers on Linux ABS_HAT1X, ABS_HAT1Y, ABS_HAT2X, ABS_HAT2Y, according to https://docs.kernel.org/6.0/input/gamepad.html. But these do not yield axes recognized by gilrs at all, which only recognizes the few you mentioned. Because the raw events never reach the bus, there's no way to capture these without getting into the weeds and rewriting the gilrs plugin. There's an argument that this issue could also be created in the gilrs repo, and I honestly think the analog triggers on modern gamepads are common enough that an issue should be raised there as well. Still, I think some improvements should be made on Bevy's side as well. People like to support lots of strange (even custom?) controllers for that game they tinker on, and bevy is a great fit for building those projects!

I have some, potentially overwrought, concerns about registering all unknown axis events by default, because with HID devices that have a large number of axes, there could be knock-ons when bevy goes to cache everything, particularly when a new game developer runs their game for the first time with a full sim-pit connected and wonders why their input system is so slow. Still, putting this behind a flag, or perhaps allowing a whitelist for specific event codes could be a good step?

This would be a relatively simple change that would involve (the OP already has a PR for a bunch of this)

1. Storing a set of codes the developer wants to listen to, potentially per-controller
2. Binding the code (third field of the AxisChanged variant) in

   bevy/crates/bevy_gilrs/src/gilrs_system.rs

   Line 98 in 7f04906

   EventType::AxisChanged(gilrs_axis, raw_value, _) => {

3. Passing the whitelist and code down to the filter, which could then make a better determination about whether to drop it when the event is Unknown.

Finally, doing the same thing for convert_button, because lots of controllers (like the OP's virpil, very nice by the way), will trigger button presses that will just be dropped too. In fact, back "paddles", as well as the right trackpad click are unrecognized on my steam controller.

I think it's worth noting that this should probably be documented somewhere in the higher level input subsystem.

Something along the lines of "If you are using DefaultPlugins, the Gilrs plugin is responsible for dispatching RawGamepadAxisChangedEvents, and non-standard event codes will only write when whitelisted…",

TLDR; thanks for posting and investigating this issue, I suffer from it too, albeit with an entirely different device, and I appreciate your pull request, which I skimmed! I hope you get it merged, with or without the whitelist.

Also I don't think the PR caches any of the custom inputs on bevy's side. They would need to be cached manually by the developer, I think, but maybe this is desirable?