G27 Shifter Adapter gets mapped as gamepad

[AlexanderWKoenig]

Hello there,

I've bought an aftermarket USB-Adapter for the Logitech G27 Shifter.
lsusb shows this:
Bus 001 Device 008: ID 1209:f00d Generic RomanStepanov Shifter/Pedals Adapter

This device gets mapped in Proton/wine or native linux games like BeamNG.drive via SDL as a xinput gamepad.
My current workaround is to use a custom mapping to map the essential buttons to useable gamepad buttons:

SDL_GAMECONTROLLERCONFIG="03008f6a091200000df0000011010000,InterBiometrics Logitech@ Shifter V2,a:b7,b:b8,x:b9,y:b10,back:b11,guide:b0,start:b13,leftstick:b14,rightstick:b12,leftshoulder:b1,rightshoulder:b2,dpup:b5,dpdown:b7,dpleft:b4,dpright:b6,platform:Linux,crc:6a8f,"

But this is not really pratical and not all 15 buttons can be used.
jstest --normal /dev/input/js3 outputs this:

Driver version is 2.1.0.
Joystick (InterBiometrics Logitech@ Shifter V2) has 1 axes ((null))
and 15 buttons (BtnA, BtnB, BtnC, BtnX, BtnY, BtnZ, BtnTL, BtnTR, BtnTL2, BtnTR2, BtnSelect, BtnStart, BtnMode, BtnThumbL, BtnThumbR).
Testing ... (interrupt to exit)
Axes:  0: -1365 Buttons:  0:off  1:off  2:off  3:off  4:off  5:off  6:off  7:off  8:off  9:off 10:off 11:off 12:off 13:off 14:off

Is there a way to tell SDL to stop promoting this device as an Gamepad and use it as a Joystick with all available buttons?
Thanks for looking into this.

[slouken]

Does this patch work?

diff --git a/src/joystick/linux/SDL_sysjoystick.c b/src/joystick/linux/SDL_sysjoystick.c
index ea73821c0..3fd601709 100644
--- a/src/joystick/linux/SDL_sysjoystick.c
+++ b/src/joystick/linux/SDL_sysjoystick.c
@@ -2287,7 +2287,7 @@ static bool LINUX_JoystickGetGamepadMapping(int device_index, SDL_GamepadMapping

     // it is now safe to call LINUX_JoystickClose on this fake joystick.

-    if (!joystick->hwdata->has_key[BTN_GAMEPAD]) {
+    if (!joystick->hwdata->has_key[BTN_GAMEPAD] || joystick->naxes < 2) {
         // Not a gamepad according to the specs.
         goto done;
     }

[AlexanderWKoenig]

Thanks for the reply. But this patch does not work for me, atleast not with BeamNG.drive
I've used the AUR sdl3-git pkgbuild and added the patch in the prepare funktion.

patching file src/joystick/linux/SDL_sysjoystick.c
patch unexpectedly ends in middle of line
Hunk #1 succeeded at 2287 with fuzz 1.

I confirmed that in the file src/joystick/linux/SDL_sysjoystick.c the line 2290 looks like this

    if (!joystick->hwdata->has_key[BTN_GAMEPAD] || joystick->naxes < 2) {
        // Not a gamepad according to the specs.
        goto done;
    }

Builded and installed sdl3-git succesful.

But native BeamNG and Assetto Corsa via Proton still seem to detect the device as gamepad.

Is there a better way to test how SDL detects an device on linux? Could the sdl2-compat package been an issue?

[slouken]

Can you debug testcontroller in the SDL3 package and see why it's being detected as a gamepad? Is LINUX_JoystickGetGamepadMapping() getting called?

[AlexanderWKoenig]

Guess I could debug this if you can give me a hint how. Im just a PHP developer and linux nerd. So I dont know much about C and how to debug SDL. Do I have to use gdb for this?

[slouken]

Yep, gdb is your friend. Go ahead and look up commands for stepping and setting break points. The function I mentioned above is a good place to start, to see if it's getting called.

[AlexanderWKoenig]

Thanks, tried to debug with gdb, but without luck.
Think the issue is that libSDL3.so is build without debugging info

file /usr/lib/libSDL3.so.0.3.0
/usr/lib/libSDL3.so.0.3.0: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, BuildID[sha1]=271a7e2f4bbf26a93c64e6c272b30120d8e5973d, stripped

I've added the flags -D CMAKE_C_FLAGS="-g2" and -D CMAKE_BUILD_TYPE=Debug so the build function looks like this:

build() {
	CFLAGS+=" -ffat-lto-objects"
	cmake -S SDL -B build -G Ninja \
	-D CMAKE_C_FLAGS="-g2" \
	-D CMAKE_BUILD_TYPE=Debug \
	-D SDL_HIDAPI_LIBUSB=ON \
	-D CMAKE_INSTALL_PREFIX=/usr \
	-D SDL_STATIC=OFF \
	-D SDL_RPATH=OFF
	cmake --build build
}

But this does not seem to help gdb finding the breakpoint LINUX_JoystickGetGamepadMapping. The file command still outputs that the library is stripped.
I've used rpcs3 to debug this, since im sure this emulator has SDL3 as dependency. gdb info sharedLibrary also confirms that /usr/lib/libSDL3.so.0 gets loaded.

But setting the breakpoint with break LINUX_JoystickGetGamepadMapping and then run in gdb does only start rpcs3 but does not break at anypoint.

Sorry for theese beginner questions, but what did I miss?

[slouken]

Presumably the libSDL3.so.0 in /usr is not the one that you built... Try using git to checkout SDL from the main repo and build and run testcontroller locally?

[AlexanderWKoenig]

Thanks for the feedback. My issue was makepkg from arch has some custom options to enable debugging info:
https://wiki.archlinux.org/title/Debugging/Getting_traces#Compilation_options

Building with options=(debug !strip) in PKGBUILD and installing it, now allows gdb to find the breakpoint:

gdb /usr/bin/rpcs3
.....
run
....
(gdb) break LINUX_JoystickGetGamepadMapping
Breakpoint 1 at 0x7fffeb930e23: file /usr/src/debug/sdl3-git/SDL/src/joystick/linux/SDL_sysjoystick.c, line 2236.

But on the next run gdb wont break at anypoint..

I've also used a clean git clone from the SDL main repo, build without the patch and run testcontroller locally. This is the output:

Should I build this with your patch again and try testcontroller to see if anything changes?

[slouken]

Should I build this with your patch again and try testcontroller to see if anything changes?

Yes please

[AlexanderWKoenig]

Alright. I've applied the patch. Double checked line 2290 in SDL_sysjoystick.c and builded again with:

cmake -S . -B build -DSDL_TESTS=ON
cmake --build build

And run testcontroller again. But I dont think that something has changed:

[slouken]

Can you use gdb and step into LINUX_JoystickGetGamepadMapping(), and see what joystick->naxes is at the line that was changed by the patch?

[AlexanderWKoenig]

Got gdb working with testcontroller.
But LINUX_JoystickGetGamepadMapping() gets only called once and this is not cause of the shifter adapter.
I also have a pedal set adapter connected with 3 axes, this is what triggers LINUX_JoystickGetGamepadMapping and joystick->naxes ist 3 on line 2290. But my pedals work fine and dont get registered as a gamepad.

If I disconnect this pedal set, then gdb does not break anymore. Guess we are missing something before the function LINUX_JoystickGetGamepadMapping

[slouken]

What is the text output of testcontroller when just the shifter is connected?