Re: Question regarding multitouch input on Linux kernel

[Nuno Santos <nsantos@displax.com>]

Hi Benjamin,

 From my understanding, with upstreamed you mean, putting the source on 
the kernel tree.

In fact, I would love to have the driver on the kernel source but right 
now we still rely on a internal lib for data processing and touch 
tracking which we wont make it public. If it is ok to make an the 
upstream of this driver accompanied with a static lib, we might consider 
that case.

We are working on complete standalone device that will make touch data 
processing in place communicating via HID to the host. By that time that 
driver will be on kernel source.

Regarding the questions arised, i'm happy and sad at the same time. 
Happy to know that this is not a bug from the driver. Sad to know that 
this is an Ubuntu problem due to their decisions.

Thanks for the quick reply.

Regards,

Nuno Santos

On 02/25/2013 03:31 PM, Benjamin Tissoires wrote:
> Hi Nuno,
>
> On Mon, Feb 25, 2013 at 4:00 PM, Nuno Santos <nsantos@displax.com> wrote:
>> Hi,
>>
>> I have been experiencing an issue with a Linux driver for multitouch input
>> that i'm responsible for maintaining.
> Side note. Your driver does not seems to be upstreamed (or I missed
> it). You should really consider put it upstream. If we make changes in
> the multitouch API, we can change your driver too, whereas here, you
> will have to maintain several releases of your driver, one per kernel
> version.
>
>> The issue is basically the following:
>>
>> - I load the driver and the mouse works just fine
>> - I touch the screen and the first touch input is delivered to the system
>> - On that very same moment I can't use mouse left button down to click on
>> folders on nautilus. I can only selected them using drag select. I also
>> can't get a folder to get selected with a single touch input.
>> - The user experience with the mouse gets inconsistent.
>> - Unloading the module doesn't return the good experience
>> - Restarting X fixes the problem until I report a touch input again with
>> this driver
>> - If I only use common pointer input, the issue doesn't occur.
>>
>> My questions resides in if the problem is due to bad touch reporting, or due
>> to a bug in X/nautilus.
> Definitively X and nautilus problems. The very same kind of problems
> were observed in Fedora 17 and fixed in the X.org shipped in Fedora
> 18.
>
> Ubuntu is relying on an older X.org release, which contains several
> bugs related to multitouch.
>
>> I have been analyzing a lot of examples under kernel tree for multitouch
>> input under Linux an it seems I'm doing what is necessary.
>>
>> I'm using Ubuntu 12.04 but this happened with Ubuntu 11.XX already
> And in 12.10 also IIRC. I really hope that they will rebase X.org in 13.04.
>
>> This what I do in order to declare device capabilities:
>>
>> input_set_abs_params(input_dev, ABS_X, 0, 6300, 0, 0);
>> input_set_abs_params(input_dev, ABS_Y, 0, 6300, 0, 0);
>>
>> input_mt_init_slots(input_dev, DPX_TOUCH_MAX_COUNT);
> input_mt_init_slots has been changed recently, it takes an extra arg: 'flags'.
>
>> input_set_abs_params(input_dev, ABS_MT_POSITION_X, 0, 6300, 0, 0);
>> input_set_abs_params(input_dev, ABS_MT_POSITION_Y, 0, 6300, 0, 0);
> Not required IIRC. The params are copied from their single-touch equivalent.
>
>> This is what I do in order to report touches:
>>
>> for (currentTouch=0;currentTouch<DPX_TOUCH_MAX_COUNT &&
>> touchCount<priv->context->State.Acquisition.TouchPoints;++currentTouch)
>>      {
>>          Touch = priv->context->State.Touches + currentTouch;
>>
>>          x = Touch->CalibratedPoint.Position.X;
>>          y = Touch->CalibratedPoint.Position.Y;
>>
>>          input_mt_slot(usbtouch->input, currentTouch);
>>
>>          // touch down
>>          if(Touch->CurrentState==DPX_TOUCH_STATE_ACTIVE &&
>> Touch->ReportedState==DPX_TOUCH_STATE_INACTIVE)
>>          {
>>              Touch->ReportedState = DPX_TOUCH_STATE_ACTIVE;
>>
>>              input_mt_report_slot_state(usbtouch->input, MT_TOOL_FINGER, 1);
>>
>>              input_report_abs(usbtouch->input, ABS_MT_POSITION_X, x);
>>              input_report_abs(usbtouch->input, ABS_MT_POSITION_Y, y);
>>
>>              touchCount++;
>>          }
>>          // touch move
>>          else if (Touch->CurrentState==DPX_TOUCH_STATE_ACTIVE &&
>> Touch->ReportedState==DPX_TOUCH_STATE_ACTIVE)
>>          {
>>              Touch->ReportedState = DPX_TOUCH_STATE_ACTIVE;
>>
>>              input_mt_report_slot_state(usbtouch->input, MT_TOOL_FINGER, 1);
>>
>>              input_report_abs(usbtouch->input, ABS_MT_POSITION_X, x);
>>              input_report_abs(usbtouch->input, ABS_MT_POSITION_Y, y);
>>
>>              touchCount++;
>>          }
>>          // touch up
>>          else
>>          {
>>              Touch->ReportedState = DPX_TOUCH_STATE_INACTIVE;
>>
>>              input_mt_report_slot_state(usbtouch->input, MT_TOOL_FINGER, 0);
>>
>>              input_report_abs(usbtouch->input, ABS_MT_POSITION_X, x);
>>              input_report_abs(usbtouch->input, ABS_MT_POSITION_Y, y);
> No need to update ABS_MT_POSITION_X/Y in this case: they should not be
> sent to the user space according to the multitouch protocol.
>
>>              touchCount++;
>>          }
>>      }
>>
>>      if (touchCount>0)
> Looks like this test is always true.
>
>>      {
>>          input_mt_report_pointer_emulation(usbtouch->input, true);
> In the latest version of the kernel tree, you should rely on the
> input_mt_sync_frame() now. It will call
> input_mt_report_pointer_emulation() plus other things depending of the
> flags you passed to input_mt_init_slots().
>
>>          input_sync(usbtouch->input);
>>      }
>>
>>
> Cheers,
> Benjamin
> --
> To unsubscribe from this list: send the line "unsubscribe linux-input" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html