FTC #14273 Updates

[tkroon]

Sense & Sound 14273, 27050, 32859

[tkroon]

It took some time to get the development environment setup but once done adding the systemcore to an existing mecanum chassis based on actobotics parts went very well.

[tkroon]

We tried for the smallest robot with the Gobuilda channel and parts we had spare. We used the development environment originally setup for Hybrid to work with the new hardware. It is all running on a macbook including the driver station.

The most dificult part was getting the motors to accept firmware updates. A full cycle of power on every thing allowed us to get the motors to turn. The other tricky part was using our logitech controllers. The driver that we were able to use does not consider the triggers as analog but only digital. Also the names of the controlles as part of the driver class did not properly read the correct controls. The ultimate solution was to use the elastic dashboard to send all the controll signals by their HID number then use numbered controlls to make sure joysticks and triggers were not getting mixed up. The 500RPM gear boxes make a really fast robot.

[tkroon]

I will give that another try. Every time I switched to X it would not recognize the joysticks at all.

[ThadHouse]

What OS and what controllers?

[ThadHouse]

Ah. I realized you said macbook. The F310 requires the switch to be in D mode on macOS to show up at all. And then yes as you've seen the triggers are digital. There is nothing we can do to solve that, macOS gamepad support is very troublesome.

The buttons being different is not something I'm seeing however. All the buttons have correct mappings. Which ones are you seeing wrong? And you were using the "Gamepad" class, right? The "NiDs" prefixed classes will not work, do not use them.

[qwertychouskie]

It seems https://github.com/360Controller/360Controller was the most common solution in the past, but doesn't work on modern macOS (360Controller/360Controller#1139).

A comment near the end of that thread pointed to https://github.com/meghrathod/xinput-controller, and while I couldn't run the pre-compiled GUI app (required macOS 26), it includes a CLI tool, xinput_usb_tool.

Running ./xinput-usb-tool listen 100 1133 49693, my F310 turns on in Xinput mode, and controller events are printed, though sadly it still doesn't seem to be detected as a gamepad by macOS itself, at least on my system. That said, since the events are received, perhaps the data could just be passed directly to the DS, ignoring macOS's gamepad support?

I then found and tried https://github.com/xsyetopz/OpenJoystickDriver/. It actually detects and activates the F310 in Xinput mode, and the built-in tester showed the controller buttons/sticks/etc working, though some buttons are mis-mapped, and for some reason, the part that actually exposes the virtual gamepad to applications does not seem to be working on my system (SDL3's testcontroller saw the virtual gamepad, but inputs weren't being passed to it).

I just opened libsdl-org/SDL#15790 with my findings in hope that SDL3 might eventually integrate a userspace driver directly, but until then, OpenJoystickDriver seems the closest to getting the F310 (along with other not-Apple-supported controllers) properly working on macOS.

[qwertychouskie]

👀

Turns out SDL3 has the code necessary for it to work just fine, but it's deferring to macOS's gamepad APIs despite macOS not supporting this gamepad. Running the DS with SDL_JOYSTICK_MFI=0 SDL_HIDAPI_LIBUSB_WHITELIST=0 would allow it to work, except the logic for the xbox360-class gamepads only checks whether MFI is compiled into SDL itself, not whether the user/app has requested to disable it.

I just opened libsdl-org/SDL#15791 to fix SDL_HINT_JOYSTICK_MFI not being respected. @ThadHouse is it relatively easy to patch the copy of SDL3 built into the DS app? Or is it better to just wait until all the needed fixes are upstream?

[tkroon]

One of the team members and I figured a way to use the parts that came with the A301's to run this test. However, the mecanum wheels aren't a solid fit onto the new axels. We are looking forward to clamping hubs that screw onto the wheels and clamp onto the axels.