Replacement for tslib?

Replacement for tslib?

[Enrico Scholz]

Hello,

tslib is more or less mandatory for resistive touchscreens but it was
removed some days ago.  This breaks e.g. meta-qt5 layer which depends on
it.

Is this removal really final (which would be really bad, because resistive
touchscreens are used e.g. in industrial environments)?


Enrico

Re: Replacement for tslib?

[Richard Purdie]

On Wed, 2016-08-31 at 17:42 +0200, Enrico Scholz wrote:
> Hello,
> 
> tslib is more or less mandatory for resistive touchscreens but it was
> removed some days ago.  This breaks e.g. meta-qt5 layer which depends 
> on it.
> 
> Is this removal really final (which would be really bad, because
> resistive
> touchscreens are used e.g. in industrial environments)?

The underlying pressure was from xcalibrate never having been merged
into x11 and hence needing to replace xtscal by xinput-calibrate. With
that change, the need for tslib wasn't clear.

Its was also frustrating having two sets of pointercal files, one for
xinput-calibrate and one for tslib.

That said, I can see the need for tslib and if that is the only way to
support these touchscreens in qt5/qt4 we might have to reconsider that.

Are you sure the kernel interfaces that xinput is using don't work for
qt4/qt5? If not, we may need to reconsider tslib...

Cheers,

Richard

Re: Replacement for tslib?

[Enrico Scholz]

Richard Purdie <richard.purdie@linuxfoundation.org> writes:

>> tslib is more or less mandatory for resistive touchscreens but it was
>> removed some days ago.  This breaks e.g. meta-qt5 layer which depends 
>> on it.
>> 
>> Is this removal really final (which would be really bad, because
>> resistive touchscreens are used e.g. in industrial environments)?
> [...]
> Are you sure the kernel interfaces that xinput is using don't work for
> qt4/qt5?

I am not aware of any way how /dev/input/eventX values can be calibrated
within the kernel.

Most capacitive touchscreens are doing this calibration in hardware
so that values correspond 1:1 to screen coordinates.  But resistive
touchscreens return usually only the raw 10-12 bit ADC values in
ABS_X/Y.



Enrico

Re: Replacement for tslib?

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 500 bytes --]

On 1 September 2016 at 00:53, Enrico Scholz <enrico.scholz@sigma-chemnitz.de
> wrote:

> I am not aware of any way how /dev/input/eventX values can be calibrated
> within the kernel.
>
> Most capacitive touchscreens are doing this calibration in hardware
> so that values correspond 1:1 to screen coordinates.  But resistive
> touchscreens return usually only the raw 10-12 bit ADC values in
> ABS_X/Y.
>

tslib is still available in meta-oe, or I'd suggest looking at libinput.

Ross

[-- Attachment #2: Type: text/html, Size: 929 bytes --]

Re: Replacement for tslib?

[Otavio Salvador]

On Fri, Sep 2, 2016 at 8:56 AM, Burton, Ross <ross.burton@intel.com> wrote:
>
> On 1 September 2016 at 00:53, Enrico Scholz
> <enrico.scholz@sigma-chemnitz.de> wrote:
>>
>> I am not aware of any way how /dev/input/eventX values can be calibrated
>> within the kernel.
>>
>> Most capacitive touchscreens are doing this calibration in hardware
>> so that values correspond 1:1 to screen coordinates.  But resistive
>> touchscreens return usually only the raw 10-12 bit ADC values in
>> ABS_X/Y.
>
>
> tslib is still available in meta-oe, or I'd suggest looking at libinput.

The biggest problem with libinput at this moment is the lack of an
utility to calibrate it.

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

Re: Replacement for tslib?

[Enrico Scholz]

"Burton, Ross" <ross.burton@intel.com> writes:

> tslib is still available in meta-oe,

really?  In fef7b1feb79c7b43623760e2cdf4e68bd0dbd526, a

| find -name '*tslib*'

returns only xf86-input-tslib, but not plain tslib.


Enrico

Re: Replacement for tslib?

[Maxin B. John]

Hi Enrico,

On Fri, Sep 02, 2016 at 02:43:03PM +0200, Enrico Scholz wrote:
> "Burton, Ross" <ross.burton@intel.com> writes:
> 
> > tslib is still available in meta-oe,
> 
> really?  In fef7b1feb79c7b43623760e2cdf4e68bd0dbd526, a
> 
> | find -name '*tslib*'
> 
> returns only xf86-input-tslib, but not plain tslib.

Patches are on it's way:
http://lists.openembedded.org/pipermail/openembedded-devel/2016-September/108945.html

Initial discussions here:
http://lists.openembedded.org/pipermail/openembedded-devel/2016-August/108428.html

Sorry for the trouble. 

> 
> Enrico

Best Regards,
Maxin

Re: Replacement for tslib?

[Mike Looijmans]

On 31-08-16 18:57, Richard Purdie wrote:
> On Wed, 2016-08-31 at 17:42 +0200, Enrico Scholz wrote:
>> Hello,
>>
>> tslib is more or less mandatory for resistive touchscreens but it was
>> removed some days ago.  This breaks e.g. meta-qt5 layer which depends
>> on it.
>>
>> Is this removal really final (which would be really bad, because
>> resistive
>> touchscreens are used e.g. in industrial environments)?
>
> The underlying pressure was from xcalibrate never having been merged
> into x11 and hence needing to replace xtscal by xinput-calibrate. With
> that change, the need for tslib wasn't clear.
>
> Its was also frustrating having two sets of pointercal files, one for
> xinput-calibrate and one for tslib.
>
> That said, I can see the need for tslib and if that is the only way to
> support these touchscreens in qt5/qt4 we might have to reconsider that.
>
> Are you sure the kernel interfaces that xinput is using don't work for
> qt4/qt5? If not, we may need to reconsider tslib...

Once reason to use Qt is the fact that you don't need X11 for it. And tslib is 
the only calibration tool that seems to work properly with resistive touchscreens.