Add support for SDL-compatible joypad GUID and Input.get_joy_info() on Android

[Nintorch]

Related to #56181 (if this PR doesn't fix this issue, then we need to create a mapping for Xbox controllers ourselves)

This PR adds support for SDL-compatible joypad GUIDs on Android, so now we should be able to use the mappings from SDL controller mapping database for Android controllers that start with 0500, while still keeping the compatibility with mappings that use the older GUID format (if a mapping with the new format can't be found, the code tries to find one with the old format).
This PR adds support for Input.get_joy_info() for Android as well.

See also SDL's Android_AddJoystick, SDL's SDLControllerManager, and the way GUID was constructed in Godot's Linux joypad driver prior to the SDL3 joystick driver PR here.

DualShock 4 GUID before this PR

(Ignore that there's no fingerprint sensor here, it's a screenshot with a different PR applied)

DualShock 4 GUID after this PR + Input.get_joy_info()

TODO:

• Use SDL's internal mapping database
• Add a note saying that changing the GUID-generating code to be different from SDL will make some GUIDs incompatible
• Fix swapped triggers issue by looking at SDL's Android code (for separate PR)
• Fix triggers being mapped to [-1;1] on my Xbox One controller without breaking my Dualshock 4 triggers
• If a mapping for an old GUID exists while a mapping for a new one doesn't, fallback to old mapping format
• Don't introduce regressions for Nvidia Shield axis mapping (for separate PR)
• Make the CI pass

[m4gr3d]

Alongside manual testing, can you add unit tests to validate the added logic.

[Nintorch]

Thank you for your review, m4gr3d! I will be able to fix the code later today!

[Alex2782]

This branch has conflicts that must be resolved

I'll probably have time again on Monday when an Xbox controller needs to be tested on Android.

https://developer.android.com/reference/android/view/KeyEvent#KEYCODE_MEDIA_RECORD

I don't see KEYCODE_MEDIA_RECORD in the JOYPAD_BUTTON_MASK_KEYS array yet. (for misc1:b15)
I think the share button isn't being recognized yet, or I'm missing something.

[Nintorch]

I think adding a new key in JOYPAD_BUTTON_MASK_KEYS would make the GUIDs sometimes incompatible (since we're adding something new that wasn't in the original GUID generation code) :(

[Alex2782]

All keys are recognized, but LT and RT are swapped.

Godot Engine v4.7.dev.gh-114338.b25248d22 (2026-02-06 11:08:17 UTC) - https://godotengine.org
OpenGL API OpenGL ES 3.2 V@0502.0 (GIT@193b2ee, I593c16c433, 1633593732) (Date:10/07/21) - Compatibility - Using Device: Qualcomm - Adreno (TM) 650

+ Found newly connected joypad #0: Xbox Series X Controller - 050000005e040000130b0000ffff3f00
{ "raw_name": "Xbox Wireless Controller 4416", "vendor_id": "1118", "product_id": "2835" }

I think this is coming from the header file, but when I tested your SDL-PR a few days ago using the Android SDL test app, everything was OK, so it seems it's not compatible with Godot.

"050000005e040000130b0000ffff3f00,Xbox Series X Controller,a:b0,b:b1,back:b4,misc1:b15,dpdown:b12,dpleft:b13,dpright:b14,dpup:b11,guide:b5,leftshoulder:b9,leftstick:b7,lefttrigger:a4,leftx:a0,lefty:a1,rightshoulder:b10,rightstick:b8,righttrigger:a5,rightx:a2,righty:a3,start:b6,x:b2,y:b3,",

---

And I think we need to change the order; godotcontrollerdb.txt needs to be at the very bottom, right? So that we can override it properly.

controller_databases = [
    "gamecontrollerdb.txt",
    "godotcontrollerdb.txt",
    "../../thirdparty/sdl/joystick/SDL_gamepad_db.h",
]

[Nintorch]

Thank you for your review! I will see what I can do to fix this issue later when I can!

[Alex2782]

lefttrigger:a4,righttrigger:a5

I also checked the official demo app, where they are simply mapped with Axes, a4 is on the left and a5 is on the right.
One possible solution would be to override (swap) the settings in the file godotcontrollerdb.txt.

[Nintorch]

That does sound like a good solution that doesn't introduce any more regressions (besides, well, what this PR does for controller mappings that don't start with 0500...)
Thanks!

[Nintorch]

I just noticed that the triggers for the older GUID for your controller are mapped with +s in the godotcontrollerdb.txt mapping, while in SDL's internal mapping they are not. Maybe that's causing the problem? Maybe our Godot code can't map the triggers properly unless there are +s in the mapping?

I will see what I can do.

[Alex2782]

I just noticed that the triggers for the older GUID for your controller are mapped with +s in the godotcontrollerdb.txt mapping, while in SDL's internal mapping they are not. Maybe that's causing the problem? Maybe our Godot code can't map the triggers properly unless there are +s in the mapping?

I will see what I can do.

In my PR I used this to solve a different problem; that shouldn't be necessary anymore.

Issue #99191 (lefttrigger:+a4,righttrigger:+a5)

On Android, Controller Triggers once pressed, stay pressed until the game is restarted

[Nintorch]

But that's still strange that the trigger mapping hasn't changed between the previous used mapping and the current one aside of the +, but now they're swapped 😅

[Alex2782]

I found an old custom template on my hard drive and tried it with Godot 4.4.1, but even then the buttons/triggers were swapped. And when released, they remained pressed at 0.5. That's what the + configuration was for.

I think back then I always pressed both triggers simultaneously during testing and didn't notice that they were swapped 😃

Godot Engine v4.3.beta.custom_build.7fea02856 (2024-06-06 15:14:24 UTC) - https://godotengine.org
OpenGL API OpenGL ES 3.2 V@0502.0 (GIT@193b2ee, I593c16c433, 1633593732) (Date:10/07/21) - Compatibility - Using Device: Qualcomm - Adreno (TM) 650

+ Found newly connected joypad #0: Xbox One Controller - 58626f7820576972656c65737320436f

[Nintorch]

Ohh I see, thank you for testing! :D
I will see if I'm able to reproduce it with my Xbox controller too and I will see what I can do, maybe I will have to port more code from SDL.

EDIT: The triggers are not swapped on my controller, but they are mapped to [-1;1], this will also need to be fixed without breaking my Dualshock 4 triggers

[Nintorch]

So I found some differences between how Godot handles Android joypad axes and how SDL handles them.
Godot checks for duplicate axes:

godot/platform/android/java/lib/src/main/java/org/godotengine/godot/input/GodotInputHandler.java

Lines 389 to 407 in 4044808

	Set<Integer> already = new HashSet<>();
	for (InputDevice.MotionRange range : device.getMotionRanges()) {
	boolean isJoystick = range.isFromSource(InputDevice.SOURCE_JOYSTICK);
	boolean isGamepad = range.isFromSource(InputDevice.SOURCE_GAMEPAD);
	if (!isJoystick && !isGamepad) {
	continue;
	}
	final int axis = range.getAxis();
	if (axis == MotionEvent.AXIS_HAT_X || axis == MotionEvent.AXIS_HAT_Y) {
	joystick.hasAxisHat = true;
	} else {
	if (!already.contains(axis)) {
	already.add(axis);
	joystick.axes.add(axis);
	} else {
	Log.w(TAG, " - DUPLICATE AXIS VALUE IN LIST: " + axis);
	}
	}
	}

SDL does not:
https://github.com/libsdl-org/SDL/blob/4e2fd57e77fb4a28c0eeef0670fc4121cc2cf1f9/android-project/app/src/main/java/org/libsdl/app/SDLControllerManager.java#L230-L240

                    List<InputDevice.MotionRange> ranges = joystickDevice.getMotionRanges();
                    Collections.sort(ranges, new RangeComparator());
                    for (InputDevice.MotionRange range : ranges) {
                        if ((range.getSource() & InputDevice.SOURCE_CLASS_JOYSTICK) != 0) {
                            if (range.getAxis() == MotionEvent.AXIS_HAT_X || range.getAxis() == MotionEvent.AXIS_HAT_Y) {
                                joystick.hats.add(range);
                            } else {
                                joystick.axes.add(range);
                            }
                        }
                    }

Currently in master my Xbox triggers are mapped to [-1;1]. If I remove the duplicate checking functionality, they work just fine (Dualshock 4 triggers too).
Maybe that's also the reason the triggers are swapped in Godot for you?

This functionality was added in #45771 to fix #45712 , but to me the original issue looks more like a mapping issue that needs to be checked in SDL testcontroller too, just in case. It also provides a way to setup a mapping for the device, maybe that would be helpful too! Or if it doesn't help, we can leave the duplicate checking functionality only for Nvidia Shield controller.

I'm gonna add a commit to remove the duplicate checking functionality here. May I ask if you can test your controller with that commit too, please?

And I also just checked and #56181 was reported only after #45771 was merged. It may be a coincidence, or may be not :D

[Nintorch]

I came up with a way to keep the joypad mapping system backwards-compatible with the old GUID format, see the changes in core/input/input.cpp. Basically, if a mapping with the new GUID format wasn't found, the system tries to find one with the old GUID format.
(If both fail, well, the joypad doesn't have a mapping :D)
I will need to test it later though.

[Nintorch]

I feel like the triggers issues should be fixed in a separate PR since this PR is about the mappings and GUID, but I feel like it should still be easy enough to fix.
For now I will need to make the CI pass.