Re: [PATCH 1/4] Input: Add trackpoint doubletap and system debug info keycodes

From: Hans de Goede
Date: Tue Apr 16 2024 - 09:03:46 EST


Hi,

On 4/16/24 2:48 PM, Mark Pearson wrote:
> Hi Hans
>
> On Tue, Apr 16, 2024, at 4:33 AM, Hans de Goede wrote:
>> Hi Mark,
>>
>> On 4/16/24 1:57 AM, Mark Pearson wrote:
>>> Hi Dmitry,
>>>
>>> On Mon, Apr 15, 2024, at 6:54 PM, Dmitry Torokhov wrote:
>>>> On Mon, Apr 15, 2024 at 04:28:19PM -0400, Mark Pearson wrote:
>>>>> Hi
>>>>>
>>>>> On Mon, Apr 15, 2024, at 3:58 PM, Dmitry Torokhov wrote:
>>>>>> On Mon, Apr 15, 2024 at 09:50:37PM +0200, Hans de Goede wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 4/15/24 9:40 PM, Dmitry Torokhov wrote:
>>>>>>>> On Wed, Apr 10, 2024 at 10:48:10PM -0400, Mark Pearson wrote:
>>>>>>>>>
>>>>>>>>> I have a stronger preference to keep the KEY_DOUBLECLICK - that one seems less controversial as a genuine new input event.
>>>>>>>>
>>>>>>>> Please see my response to Peter's letter. I think it very much depends
>>>>>>>> on how it will be used (associated with the pointer or standalone as it
>>>>>>>> is now).
>>>>>>>>
>>>>>>>> For standalone application, recalling your statement that on Win you
>>>>>>>> have this gesture invoke configuration utility I would argue for
>>>>>>>> KEY_CONFIG for it.
>>>>>>>
>>>>>>> KEY_CONFIG is already generated by Fn + F# on some ThinkPads to launch
>>>>>>> the GNOME/KDE control center/panel and I believe that at least GNOME
>>>>>>> comes with a default binding to map KEY_CONFIG to the control-center.
>>>>>>
>>>>>> Not KEY_CONTROLPANEL?
>>>>>>
>>>>>> Are there devices that use both Fn+# and the doubleclick? Would it be an
>>>>>> acceptable behavior for the users to have them behave the same?
>>>>>>
>>>>> Catching up with the thread, thanks for all the comments.
>>>>>
>>>>> For FN+N (originally KEY_DEBUG_SYS_INFO) the proposal was to now use
>>>>> KEY_VENDOR there. My conclusion was that this is targeting vendor
>>>>> specific functionality, and that was the closest fit, if a new keycode
>>>>> was not preferred.
>>>>
>>>> Fn+N -> KEY_VENDOR mapping sounds good to me.
>>>>
>>>>>
>>>>> For the doubletap (which is a unique input event - not related to the
>>>>> pointer) I would like to keep it as a new unique event.
>>>>>
>>>>> I think it's most likely use would be for control panel, but I don't
>>>>> think it should be limited to that. I can see it being useful if users
>>>>> are able to reconfigure it to launch something different (browser or
>>>>> music player maybe?), hence it would be best if it did not conflict
>>>>> with an existing keycode function. I also can't confirm it doesn't
>>>>> clash on existing or future systems - it's possible.
>>>>
>>>> So here is the problem. Keycodes in linux input are not mere
>>>> placeholders for something that will be decided later how it is to be
>>>> used, they are supposed to communicate intent and userspace ideally does
>>>> not need to have any additional knowledge about where the event is
>>>> coming from. A keyboard either internal or external sends KEY_SCREENLOCK
>>>> and the system should lock the screen. It should not be aware that one
>>>> device was a generic USB external keyboard while another had an internal
>>>> sensor that recognized hovering palm making swiping motion to the right
>>>> because a vendor decided to make it. Otherwise you have millions of
>>>> input devices all generating unique codes and you need userspace to
>>>> decide how to interpret data coming from each device individually.
>>>>
>>>> If you truly do not have a defined use case for it you have a couple
>>>> options:
>>>>
>>>> - assign it KEY_RESERVED, ensure your driver allows remapping to an
>>>> arbitrary keycode, let user or distro assign desired keycode to it
>>>>
>>>> - assign KEY_PROG1 .. KEY_PROG4 - pretty much the same - leave it in the
>>>> hand of the user to define a shortcut in their DE to make it useful
>>>>
>>>>>
>>>>> FWIW - I wouldn't be surprised with some of the new gaming systems
>>>>> we're seeing (Steamdeck, Legion-Go, etc), that a doubletap event on a
>>>>> joystick might be useful to have, if the HW supports it?
>>>>
>>>> What would it do exactly? Once we have this answer we can define key or
>>>> button code (although I do agree that game controller buttons are
>>>> different from "normal" keys because they map to the geometry of the
>>>> controller which in turn defines their commonly understood function).
>>>>
>>>> But in any case you would not reuse the same keycode for something that
>>>> is supposed to invoke a configuration utility and also to let's say
>>>> drop a flash grenade in a game.
>>>>
>>>
>>> Understood.
>>>
>>> I don't see a path forward within your stated parameters. I did not realise that there were such limitations, so my apologies for wasting everybody's time, and thank you for your patience and guidance.
>>>
>>> I will drop this patch from the series and proceed using existing defined codes only.
>>>
>>> Hans, I'll need to rejig things a bits but I have some ideas and I think I can make it work and stay within the pdx86 tree, which will make it simpler.
>>
>> Ok this sounds good to me. For Fn + N using KEY_VENDOR sounds good for
>> me and for the doubletap any one of
>> KEY_CONFIG/KEY_CONTROLPANEL/KEY_INFO/KEY_PROG1
>> or some other suitable KEY_foo define works for me.
>>
> I think this should be a configurable input, by design. So my preference (if not allowed a new keycode, which I personally think is the better option) is for PROG1.
>
> I discussed with Peter last night and it looks likely OK on their side. I do plan on doing some testing first, so it might take a few days to get the next set of patches out.

Ok, PROG1 works for me.

Regards,

Hans