Improve modern controller support

[Zizin13]

DOS game only supports two controllers with one stick and two buttons each. We should be able to configure any number of axes and buttons.

[schlangz]

This would be wonderful. In the meantime, is there any tool that can map a wheel axis to a button?

[Zizin13]

Wheel and pedal support very soon. Idk about force feedback but basic control is doable.

[Zizin13]

Wheels are now supported with PR #207. There is a new option PLAYER # WHEEL CONTROLS that has an additional sub-menu after applying an axis to allow greater configuration so you don't have to do anything weird like awkwardly hold your pedals in the middle until they are assigned like with DOSBox. Gamepad controller support is also improved with all axes and buttons available for assignment

[schlangz]

roller-crash-20260530-031810.zip

Thanks for the update but my game is crashing when I connect my simrig controls

PROCESS_NAME:  roller.exe

ERROR_CODE: (NTSTATUS) 0xc000001d - {EXCEPTION}  Illegal Instruction  An attempt was made to execute an illegal instruction.

EXCEPTION_CODE_STR:  c000001d

STACK_TEXT:  
000000a4 ac3fea20 00007ff6 cadc3fad     : 000001c0 e2a3a070 000001c0 e2a3a050 00000000 00000000 000000a4 ac3fee7f : roller!UpdateDINPUTJoystickState_Polled+0x408
000000a4 ac3febb0 00007ff6 cabb384e     : 00000000 00000007 00000000 000001e0 00007ff6 caef0b80 000001c0 e2a40cc0 : roller!SDL_DINPUT_JoystickUpdate+0x1c5
000000a4 ac3ff860 00007ff6 cab03315     : 000000a4 ac3ff980 00007ff6 cab7dab9 00007ff6 caef0b80 00000000 00000001 : roller!SDL_OpenJoystick_REAL+0xb06
(Inline Function) -------- --------     : -------- -------- -------- -------- -------- -------- -------- -------- : roller!InputOpenDevice+0xc5
000000a4 ac3ff950 00007ff6 cab00cf2     : 00000000 00000001 00007ff6 caef0b80 00000000 00000048 00007ffb 7f35b44d : roller!InputRefreshDevices+0x1c5
000000a4 ac3ff9e0 00007ff6 caa7bd05     : 000001c0 d5420120 006e0069 00630061 006f0072 005c0067 00720065 006c006c : roller!InitSDL+0x242
000000a4 ac3ffb90 00007ff6 cab1b89d     : 00007ffb 7cbf1270 00007ff6 cab58e20 00007ff6 00000002 00000000 00000002 : roller!main+0x475
000000a4 ac3ffd80 00007ff6 cab1b6bc     : 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 : roller!__tmainCRTStartup+0x1cd
000000a4 ac3ffde0 00007ffb 7e367374     : 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 : roller!WinMainCRTStartup+0x1c
000000a4 ac3ffe10 00007ffb 7f37cc91     : 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 : kernel32!BaseThreadInitThunk+0x14
000000a4 ac3ffe40 00000000 00000000     : 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 : ntdll!RtlUserThreadStart+0x21

[Zizin13]

Will take a look, glad my crash dump feature worked at least 😎

[Zizin13]

This may actually be a bug in SDL3. I ran into another dinput related bug in SDL3 that I made a workaround for when I was making this feature.

[schlangz]

I tried the latest build and it's still happening. I unplugged all my controls one by one, and the culprit is my Fanatec CSL DD - it doesn't even matter if there is a wheel attached to it, as soon as the DD wheel base is connected via USB I get the crash on start.

[Zizin13]

Nothing related to this was fixed last night, I just investigated the matter for awhile and it looked like the SDL3 team may have recently fixed it. Unfortunately I will have to update a bunch of other stuff as well to get the latest SDL3 including some repos owned by others so it may take a bit. Or maybe I'll just fork SDL3. Might have to do that anyways to fix the problem myself if they have not actually fixed it lol.

[schlangz]

I'm trying to see what the issue could be. E.g. when I set these environment parameters before starting roller.exe the game doesn't crash:

SET SDL_JOYSTICK_WGI=1
SET SDL_JOYSTICK_RAWINPUT=0
SET SDL_JOYSTICK_GAMEINPUT=0
SET SDL_JOYSTICK_DIRECTINPUT=0

[schlangz]

Tests performed with Fanatec CSL DD and latest Fanatec App/Driver (1.4.2.3).
Tested with all other joystick devices disconnected (VKB and Virpil devices unplugged).

• Wheel is detected correctly in Windows (joy.cpl) and all axes/buttons work normally.
• ROLLER starts normally when the wheel is powered off.
• As soon as the CSL DD is powered on, ROLLER immediately crashes.
• The crash occurs with the CSL DD in both native PC mode and in Compatibility Mode (CSW V2.5 mode).
• Tested the following SDL environment variables without success, the crash still occurs immediately when the wheel is detected:

SDL_JOYSTICK_HIDAPI=0
SDL_JOYSTICK_RAWINPUT=1
SDL_JOYSTICK_WGI=0
SDL_JOYSTICK_GAMEINPUT=0

• Disabling DirectInput (SDL_JOYSTICK_DIRECTINPUT=0) prevents any of my joysticks to be recognized, including the wheel, and ROLLER doesn't crash.

[Zizin13]

@schlangz There does seem to be a bug in SDL3, and it is not yet fixed in the latest version. However I would like some more information to make sure ROLLER does not also have a bug like a heap corruption or buffer overrun (common mistake in the decomp, but I thought I got most of them). Can you run the attached programs in a console and upload what they output? I want to know if your device is actually enumerating more than 256 inputs and if so, what they are. You can either run the exes built for windows or build them yourself, build instructions included in the comments at the top.

di_probe.zip

[Zizin13]

Also, if there is a problem within ROLLER and if you can build a debug version and run with a debugger attached (see video in readme for build instructions) with Application Verifier running we may be able to determine exactly where the problem is.

[schlangz]

Thanks for the tools, attached the output from both files even though I don't see a device with more than 256 inputs.
di_probe_verbose.txt
di_probe.txt

Based on the application verifier log generated by the debug version the crash is happening in the Fanatec force feedback driver initialization.
roller.exe.0.dat.xml

VScode debugging points me to the Fanatec driver logs:

vscode.txt
FW-Error.log
FW-FFB.log

Edit: according to Fanatec website the DIERR_DEVICENOTREG trace log entry is normal, see https://forum.fanatec.com/topic/13039-fw-ffb-error-tracing-for-cprogram-files-x86iracingiracingsim64dx11exe/