search

univrsal / input-overlay

Show keyboard, gamepad and mouse input on stream
GNU General Public License v2.0
3.01k stars 248 forks source link

Gamepad issues #297

Open univrsal opened 1 year ago

univrsal commented 1 year ago

I've just finished some tests with SDL2, and it seems like I could replace the gamepad library with it. Before I do that, it would be good to have some people test SDL2 with their game pads to see if the mappings are working. SDL2 is used in a lot of games so I assume there shouldn't be any issues. I've attached the SDL2 gamepad test.

sdl2.gamepadtest.zip image Holding shift will show the back of the controller with additional buttons.

I'll also close all gamepad related issues and we'll just use this issue to collect information regarding gamepads.

SDL2 tests so far

Previous gamepad issues:

294

286, #253

Should be fixed when using SDL2

260

Should work with SDL2, if not I won't spend any additional time on this for now

208

Apparently a crash with the gamepad library, should be fixed when using SDL2.

195, #191

Mapping/ D-Pad issue should be fixed with SDL2

183

SDL2 should support xbox one gamepads just fine

175

More mapping issues once again SDL2 should handle them correctly

241

Another dpad issue

116

General issue with gamepad being detected

213

Razer Rajiu not being detected, apparently that is a controller, so sdl will most likely handle it just fine.

196 DirectInput implementation seemed to have caused stalls (probably waiting for a mutex)

Semi related issues:

285 The incorrect position of the analog sticks is an issue with the library using different value ranges for linux and windows

199 The binding tab will be removed as sdl2 should take care of correct bindings

Arbel-arad commented 1 year ago

seems to work fine for me, even with everything else connected

univrsal commented 1 year ago

seems to work fine for me, even with everything else connected

Thanks for testing, since your issue was a crash we'll have to see if the problem was really fixed once I switch the plugin to sdl

clarfonthey commented 1 year ago

I don't think that testing via WINE would be a fair test of how the native Linux SDL performs, but SDL2 stuff works perfectly on my system and I think porting to SDL2 would probably fix all my issues.

Would be happy to help test anything or provide some code changes as well depending on what's needed.

FWIW, since Nintendo Switch controller support is relatively new on Linux, a lot of things don't properly support it (mapping the buttons wrong under the assumption it's just an Xbox controller), although SDL2 actually does support it correctly.

univrsal commented 1 year ago

I don't think that testing via WINE would be a fair test of how the native Linux SDL performs, but SDL2 stuff works perfectly on my system and I think porting to SDL2 would probably fix all my issues.

Would be happy to help test anything or provide some code changes as well depending on what's needed.

FWIW, since Nintendo Switch controller support is relatively new on Linux, a lot of things don't properly support it (mapping the buttons wrong under the assumption it's just an Xbox controller), although SDL2 actually does support it correctly.

You can compile the program yourself if you want, it's just https://github.com/univrsal/SDL/blob/main/test/testgamecontroller.c which is part of the sdl2 tests

HereInPlainSight commented 1 year ago

You can compile the program yourself if you want, it's just https://github.com/univrsal/SDL/blob/main/test/testgamecontroller.c which is part of the sdl2 tests

Tested in Linux with a PS3, PS4, and PS5 controller, seems to work fine on all.

sanyer commented 1 year ago

I can test it for Xbox and Xbox Elite controllers later today or tomorrow. Additionally, I will also submit new preset/design for the Xbox controller.

univrsal commented 1 year ago

You can compile the program yourself if you want, it's just https://github.com/univrsal/SDL/blob/main/test/testgamecontroller.c which is part of the sdl2 tests

Tested in Linux with a PS3, PS4, and PS5 controller, seems to work fine on all.

The ps4 and ps4 controller have a touch pad, right? Did sdl2 detect any input when you used them? (There probably won't be any visualization since the display uses an xbox gamepad, but maybe the console window shows something). Also did the button with the playstation logo work?

I can test it for Xbox and Xbox Elite controllers later today or tomorrow. Additionally, I will also submit new preset/design for the Xbox controller.

Can you also test if the guide button and the additional buttons work?

sanyer commented 1 year ago

Can you also test if the guide button and the additional buttons work?

Sure, will do.

sanyer commented 1 year ago

All buttons (including Guide) work. And I don't have controller with Share button, so couldn't test that. Sticks register direction correctly, but stick coordinates are all over the place, or at least it is hard to get consistent numbers, but perhaps it is just the internal thing as overlay always shows correct direction.

univrsal commented 1 year ago

All buttons (including Guide) work. And I don't have controller with Share button, so couldn't test that. Sticks register direction correctly, but stick coordinates are all over the place, or at least it is hard to get consistent numbers, but perhaps it is just the internal thing as overlay always shows correct direction.

Thanks for testing. I assume there's a way to convert the axis values if the interface shows the movements correctly.

HereInPlainSight commented 1 year ago

The ps4 and ps4 controller have a touch pad, right? Did sdl2 detect any input when you used them? (There probably won't be any visualization since the display uses an xbox gamepad, but maybe the console window shows something).

Yes, as you said, no visual indicator. (I had to filter out the gyro \ accelerometer lines via ./testgamecontroller 2>&1 | grep -v "gyro\|accelerometer", for any other Linux user looking to pull out useful output.) This was as quick a straight press on the touchpad on a PS5 controller as possible. Note that only the first line and the 'button touchpad release' had to do with me pressing down, the rest is all 'We see a finger!' output. It's very spammy.

INFO: Controller 5 button touchpad pressed
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.48, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.48, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 button touchpad released
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.48, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.48, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.48, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.48, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.49, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.49, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.49, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.50, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.50, 1.00
INFO: Controller 5 touchpad 0 finger 0 released at 0.85, 0.50, 0.00

Also did the button with the playstation logo work?

Yes. It showed as the giant center button in the picture (output indicated the 'guide' button), and the microphone mute button was reflected in the button just under that (the 'misc1' button).

Unrelated to the rest, since people were talking about switch controllers, I found an off-brand pro version I had in the house, and it works as expected. Y and X are swapped visually, but report correctly, the same as B and A, since Nintendo and Microsoft seem to have reversed positioning on those compared to each other.

Edit: I forgot the switch controller had buttons on the back / bottom. These do not seem to report anything at all when pressed.

univrsal commented 1 year ago

The ps4 and ps4 controller have a touch pad, right? Did sdl2 detect any input when you used them? (There probably won't be any visualization since the display uses an xbox gamepad, but maybe the console window shows something).

Yes, as you said, no visual indicator. (I had to filter out the gyro \ accelerometer lines via ./testgamecontroller 2>&1 | grep -v "gyro\|accelerometer", for any other Linux user looking to pull out useful output.) This was as quick a straight press on the touchpad on a PS5 controller as possible. Note that only the first line and the 'button touchpad release' had to do with me pressing down, the rest is all 'We see a finger!' output. It's very spammy.

INFO: Controller 5 button touchpad pressed
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.48, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.48, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 button touchpad released
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.47, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.48, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.48, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.48, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.48, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.49, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.49, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.49, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.50, 1.00
INFO: Controller 5 touchpad 0 finger 0 moved to 0.85, 0.50, 1.00
INFO: Controller 5 touchpad 0 finger 0 released at 0.85, 0.50, 0.00

Also did the button with the playstation logo work?

Yes. It showed as the giant center button in the picture (output indicated the 'guide' button), and the microphone mute button was reflected in the button just under that (the 'misc1' button).

Thanks for the feedback. I assume that the movement is limited to values between 0 and 1. So I guess input overlay could visualize and forward that via the websocket server. The swapped X/Y are a bit annoying, but the preset for the switch can just switch (hah) them, alternatively it seems like there is an option to change this in sdl

Unrelated to the rest, since people were talking about switch controllers, I found an off-brand pro version I had in the house, and it works as expected. Y and X are swapped visually, but report correctly, the same as B and A, since Nintendo and Microsoft seem to have reversed positioning on those compared to each other.

Edit: I forgot the switch controller had buttons on the back / bottom. These do not seem to report anything at all when pressed.

Are those extra buttons also on a normal switch pro controller? If not it's probably an issue with SDL assuming that it is a switch pro controller

HereInPlainSight commented 1 year ago

I assume that the movement is limited to values between 0 and 1.

Correct. Touchpad values on the PS5 touchpad seem to be 0.00 through 1.00 on the X and Y axis, with a Z value that seems to always be '1.00' when the touchpad's being touched, and '0.00' as the very last value as the touch pad is released. (You can see that in my output above.) No indication that I can see that the Z value is ever anything other than a binary value besides it being shown as having a decimal point. 0.00 0.00 is the top-left coordinates, 1.00 1.00 is the bottom-right. I can test the PS4 controller if you want, but I suspect it's the exact same given everything else is.

Also, while it only lists 'finger 0' above, it looks like the touchpad can register two fingers (0 and 1) at the same time.

Are those extra buttons also on a normal switch pro controller?

I don't see anything that looks similar on the pictures of a real switch pro controller, so I guess not. I have no idea what they're for, but since it's an off-brand and doesn't exist on the official pro controller and they're not registering in SDL, I doubt it's worth caring about.

univrsal commented 1 year ago

Correct. Touchpad values on the PS5 touchpad seem to be 0.00 through 1.00 on the X and Y axis, with a Z value that seems to always be '1.00' when the touchpad's being touched, and '0.00' as the very last value as the touch pad is released. (You can see that in my output above.) No indication that I can see that the Z value is ever anything other than a binary value besides it being shown as having a decimal point. 0.00 0.00 is the top-left coordinates, 1.00 1.00 is the bottom-right. I can test the PS4 controller if you want, but I suspect it's the exact same given everything else is.

The code references the third value as pressure so maybe it could technically be in between 0 and 1 just not with playstation controllers.

ngeTaco commented 1 year ago

I assume that the movement is limited to values between 0 and 1.

Correct. Touchpad values on the PS5 touchpad seem to be 0.00 through 1.00 on the X and Y axis, with a Z value that seems to always be '1.00' when the touchpad's being touched, and '0.00' as the very last value as the touch pad is released. (You can see that in my output above.) No indication that I can see that the Z value is ever anything other than a binary value besides it being shown as having a decimal point. 0.00 0.00 is the top-left coordinates, 1.00 1.00 is the bottom-right. I can test the PS4 controller if you want, but I suspect it's the exact same given everything else is.

Also, while it only lists 'finger 0' above, it looks like the touchpad can register two fingers (0 and 1) at the same time.

Are those extra buttons also on a normal switch pro controller?

I don't see anything that looks similar on the pictures of a real switch pro controller, so I guess not. I have no idea what they're for, but since it's an off-brand and doesn't exist on the official pro controller and they're not registering in SDL, I doubt it's worth caring about.

Those buttons probably have to be assigned to mimick another face or shoulder button, which is non-standard for the base model of controlers for microsoft, sony, and nintendo, so no need to worry. I would assume that if theyre assigned, SDL2 would show the mimicked button being activated and not those back buttons, but not something I think would be a deal breaker for people.

ItsRogueRen commented 1 year ago

I don't think that testing via WINE would be a fair test of how the native Linux SDL performs, but SDL2 stuff works perfectly on my system and I think porting to SDL2 would probably fix all my issues. Would be happy to help test anything or provide some code changes as well depending on what's needed. FWIW, since Nintendo Switch controller support is relatively new on Linux, a lot of things don't properly support it (mapping the buttons wrong under the assumption it's just an Xbox controller), although SDL2 actually does support it correctly.

You can compile the program yourself if you want, it's just https://github.com/univrsal/SDL/blob/main/test/testgamecontroller.c which is part of the sdl2 tests

Are there build instructions for this?

HereInPlainSight commented 1 year ago

Are there build instructions for this?

Clone the repo and change directory to it.

mkdir build
cd build
cmake -DSDL_TESTS=ON ../
cmake --build .

testgamecontroller should now be inside the test subdirectory. If you've got a gyroscope in your controller and you want to verify output on the command line, you might want to filter the gyroscope info out with something like ./testgamecontroller 2>&1 | grep -v "gyro\|accelerometer", it's super spammy.

GameKyuubi commented 1 year ago

https://github.com/univrsal/input-overlay/issues/241 Another dpad issue

Seems like the main dpad issue imo. Dpad being POV-hat seems to be the standard going forward. Previous version worked with this paradigm, if I recall correctly something about that library used an "is_hat" boolean to distinguish pov hat situations.

univrsal commented 1 year ago

Have you tried the gamepad test program?

GameKyuubi commented 1 year ago

I'm on linux so I can't use the attachment, but it's the sdl2-gamepad branch?

univrsal commented 1 year ago

the test program is part of the sdl2 tests, but the branch should also work if you want to test the plugin with sdl2. It doesn't build on ubuntu for some reason though

univrsal commented 1 year ago

https://github.com/univrsal/input-overlay/actions/runs/4166499103 Contains a windows and linux build using sdl2 for gamepad input. The presets in the master branch should be updated to use the new keycodes. If you have presets that you've made yourself you can just load them in the config tool and then export it again to convert it to the new keycodes.

GameKyuubi commented 1 year ago

@univrsal I'm on Arch, when I try to build that commit I get pages of errors. Same with latest master. Pacman reports installed version is 5.0.0.r9.g05b175f-1.

univrsal commented 1 year ago

yeah It's something with simd headers clashing with sdl2, debug builds don't have that issue for some reason

andreottigiulio commented 1 year ago

XBOX 360 Controller Not Working properly UNDER OBS: leftshoulder,rightshoulder,x,y,a,b,start,back NOT WORKING in input-overlay object dpad,left and right stick,button rightstick,button leftstick, axis righttrigger,axis lefttrigger WORKING NORMALLY

univrsal commented 1 year ago

Did you test it with the test program?

andreottigiulio commented 1 year ago

i wrote "sdl2.gamepadtest.zip Working PERFECTLY" trying to be as complete as possible. sorry for bothering but i really have no more ideas

univrsal commented 1 year ago

no worries, I missed that. If the test program works the next release should fix the issues

univrsal commented 1 year ago

Anybody who wants can try the new gamepad and network backend with this. If there's no more bugs I'll close this issue.

patlefort commented 1 year ago

Xbox elite controller 2 here: Y and X buttons are swapped using the gamepad preset. Swapping them in the profile was simple enough.

univrsal commented 1 year ago

Strange that the buttons are swapped, I assume they're fine in the gamepad test program (linked in the issue description)?

patlefort commented 1 year ago

Strange that the buttons are swapped, I assume they're fine in the gamepad test program (linked in the issue description)?

Indeed it is working, Y is Y and X is X.

clarfonthey commented 1 year ago

yeah It's something with simd headers clashing with sdl2, debug builds don't have that issue for some reason

I'm glad that you know the reason for this, because I honestly thought it was just some weird quirk of my system and didn't want to mention it. When you say "debug builds," how are you building it differently? (as in, is there a way for me to modify it so I can build a debug build, and maybe I can help figure out the issue this way)

univrsal commented 1 year ago

yeah It's something with simd headers clashing with sdl2, debug builds don't have that issue for some reason

I'm glad that you know the reason for this, because I honestly thought it was just some weird quirk of my system and didn't want to mention it. When you say "debug builds," how are you building it differently? (as in, is there a way for me to modify it so I can build a debug build, and maybe I can help figure out the issue this way)

Just the normal cmake debug target, i.e. cmake -S <path to git repo> -DCMAKE_BUILD_TYPE=Debug

You can also do what I did for the CI builds and tamper with the headers that cause the issue, but I wouldn't recommend that: https://github.com/univrsal/input-overlay/blob/master/.github/scripts/.build.zsh#L182-L183

clarfonthey commented 1 year ago

Yeah, I wouldn't want to actually mess with library headers; that sounds like a lost cause.

I will try and see what's different in the debug build, though; my guess is that there might be some extra include flags that conflict, but we'll see. Not sure if I'll get to this over the weekend or not, but I'll share what I find if I find anything.

patlefort commented 1 year ago

On Arch, makepkg will default to x86-64 arch which result in the wall of errors with intrinsics. Adding -msse4.1 work around the problem.

univrsal commented 1 year ago

Where did you add that?

patlefort commented 1 year ago

It's a compile flag for GCC. For makepkg, you'd have to add it into /etc/makepkg.conf in the CFLAGS and CXXFLAGS. Alternatively I think it's better to change -march=x86-64 to -march=native so it can take full advantage of your hardware.

univrsal commented 1 year ago

So I guess set(CMAKE_CXX_FLAGS "-march=native") should do the trick

patlefort commented 1 year ago

I see what's going on: OBS is defining things that are defined in standard intrinsincs headers because they aren't included yet and compiler flags indicate that they aren't available. And then SDL include the file <immintrin.h> which include all intrinsics headers regardless of flags. I think what OBS is doing is wrong but to fix it, including SDL before obs headers should work.

univrsal commented 1 year ago

Strange that the buttons are swapped, I assume they're fine in the gamepad test program (linked in the issue description)?

Indeed it is working, Y is Y and X is X.

I figured it out. Since the scroll wheel and right mouse button are switched on linux I switched keycodes 2 and 3. That should only happen for mouse button elements, but I did it for all elements so the gamepad codes 2 and 3 where also switched. Should be fixed with 4cb687f16b99067aaefbf016796aefb35c58ced2

patlefort commented 1 year ago

So I guess set(CMAKE_CXX_FLAGS "-march=native") should do the trick

You shouldn't force it in your source. For a packager, it would cause issues as binaries packages are expected to target an architecture which isn't always to one of the machine compiling. Ideally it should have the widest compatibility by default. Also it won't work for msvc.

univrsal commented 1 year ago

Well windows never had the issue to begin with. I'll just leave it the way it is for now.

GameKyuubi commented 1 year ago

Went from dpad not working to everything not working now. New issue?

univrsal commented 1 year ago

nah just post the full log here, ideally while running obs with --verbose

GameKyuubi commented 1 year ago

obs.log

univrsal commented 1 year ago

Try the a6d212fb1306835dd9743e12cee908147a63df17 commit. Ideally with only one gamepad connected first. I already heard that multiple seem to cause issues.

GameKyuubi commented 1 year ago

I only have one connected, not sure why it's detecting two. jstest only shows one.

univrsal commented 1 year ago

Might be that it first thinks it's an xbox 360 gamepad and then identifies it as an xbox one gamepad. I'll try and see if I can find another gamepad so I can reproduce the bug

AlexEndris commented 1 year ago

I don't know if my issue is in anyway related to this. I'm using a GuliKit gamepad over Bluetooth. It's recognized as an XBox One controller but the only inputs that are shown are the left trigger and right trigger. None of the others light up or move. Is there anything I can provide you in terms of logs? Should I create another issue for this?

Edit: Maybe as a side-note: With your SDL2 implementation, everything is recognised as it should be.