undefined behaviour if multiple devices in the same group have similar capabilities (beta)

[jonasBoss]

The Sony DualSense controller has a mouse-pad which reports ABS_X and ABS_Y as well as the standard controller layout which also reports ABS_X and ABS_Y events. When a mapping uses one of those events as input both the joystick and the mouse-pad can trigger this mapping.

This leads to all sorts of issues, especially because the absinfo are not the same for those axis.

[jonasBoss]

https://github.com/sezanzeb/input-remapper/pull/375/commits/a622037834a824d534c8736203e3799538c9fd0b mitigates this partially by only grabbing one of the devices. This however does not help if both devices are grabbed by different mappings.

[jonasBoss]

@sezanzeb I think we need to include the origin of each event for a mapping. Otherwise we will not be able to fix both, this issue and #517

You know evdev better than I do. Is there any good parameter we can use to identify a input-device, which is unique to each device in a group but stays always the same?

[sezanzeb]

I wrote some relevant stuff down at some point: https://github.com/sezanzeb/input-remapper/blob/b3e1e4ca1976bb420017c737525151ac37099e6e/inputremapper/groups.py#L209

# - device.phys is empty sometimes and varies across virtual
#   subdevices
# - device.version varies across subdevices
# - device.uniq is empty most of the time, I don't know what this is
#   supposed to be

• Maybe create a hash from all three of them
• Or sort them and use an index

You know evdev better than I do

I actually barely know stuff about it. I just use the things it provides

[jonasBoss]

This needs a better solution than my attempt from above. #517 demonstrates a different failure mode rooted in the ambiguity of just using the event type, code and group name to identify a input.

A proper solution needs to identify a input event by type, code, group-name and sub-device.

[sezanzeb]

preset config

{
    "1,4,1+1,6,1+1,7,1+1,8,1": {
        "target_uinput": "keyboard"
        "output_symbol": "sfdgsdfgsdfgh",
        "mapping_type": "key_macro"
    },
}

to

[
    {
        "target_uinput": "keyboard"
        "combination": [
            {"type": 1, "code": 4, "value": 1, "device": "foo"},
            {"type": 1, "code": 6, "value": 1, "device": "foo"},
            {"type": 1, "code": 7, "value": 1, "device": "foo"},
            {"type": 1, "code": 8, "value": 1, "device": "bar"}
        ],
        "output_symbol": "sfdgsdfgsdfgh",
        "mapping_type": "key_macro"
    }
]

?

[sezanzeb]

some more examples for the device attributes:

{'phys': 'usb-0000:28:00.3-2.3/input1', 'uniq': '', 'version': 65537, 'name': 'Logitech USB Keyboard System Control'}
{'phys': 'usb-0000:28:00.3-2.3/input1', 'uniq': '', 'version': 65537, 'name': 'Logitech USB Keyboard Consumer Control'}
{'phys': 'usb-0000:28:00.3-2.3/input0', 'uniq': '', 'version': 65537, 'name': 'Logitech USB Keyboard'}

{'phys': 'usb-0000:06:00.0-3/input0', 'uniq': '', 'version': 65537, 'name': 'Wacom Intuos5 M Pad'}
{'phys': 'usb-0000:06:00.0-3/input0', 'uniq': '', 'version': 65537, 'name': 'Wacom Intuos5 M Pen'}

I don't have the razer naga trinity anymore, so I can't test that one.

In order to identify them, the name would work well for those two.

I haven't seen the attributes of the DualSense controller.

I don't know if the attributes stay constant after system upgrades or reboots.

[jonasBoss]

the name will not work for a razer naga trinity, but at least the phys is unique

phys = 'usb-0000:29:00.3-1/input0', uniq = '00000000001A', version = 65537, name = 'Razer Razer Naga Trinity'
phys = 'usb-0000:29:00.3-1/input1', uniq = '00000000001A', version = 65537, name = 'Razer Razer Naga Trinity'
phys = 'usb-0000:29:00.3-1/input2', uniq = '00000000001A', version = 65537, name = 'Razer Razer Naga Trinity'

It honestly looks like the different devices are just a hot mess. I am currently considering to use:

hash(str(InputDevice.capabilities()))

given the high amount of parameters in capabilities I would guess that this is always unique. At least for all my devices this is always unique.

---

[
   {
       "target_uinput": "keyboard"
       "combination": [
           {"type": 1, "code": 4, "value": 1, "device": "foo"},
           {"type": 1, "code": 6, "value": 1, "device": "foo"},
           {"type": 1, "code": 7, "value": 1, "device": "foo"},
           {"type": 1, "code": 8, "value": 1, "device": "bar"}
       ],
       "output_symbol": "sfdgsdfgsdfgh",
       "mapping_type": "key_macro"
   }
]

yeah, I was thinking something like this as well.

[sezanzeb]

yeah. Since this only happens once when the injection starts, the complexity of the hash generation doesn't matter.

hash(str(InputDevice.capabilities()) + InputDevice.phys + InputDevice.name)