[PATCHv6] staging/iio/adc: change the MXS touchscreen driver implementation

[Jürgen Beisert]

Hi Alexandre,

On Thursday 09 January 2014 14:31:22 Alexandre Belloni wrote:
> Sorry to chime in only now but it seems that this series is breaking the
> touchscreen calibration on 3.13 (and -rc7 is out so it might be too
> late).
>
> At first, I though I became a terrible clicker ;) but I found some
> evidences:
>
> xinput_calibrator is complaining about misclicks, debug output:
> DEBUG: Adding click 0 (X=105, Y=59)
> DEBUG: Not adding click 1 (X=100, Y=59): within 7 pixels of previous click
> DEBUG: Adding click 1 (X=696, Y=58)
> DEBUG: Not adding click 2 (X=700, Y=55): within 7 pixels of previous click
> DEBUG: Adding click 2 (X=103, Y=422)
> DEBUG: Mis-click detected, click 3 (X=438, Y=415) not aligned with click 1 (X=696, Y=58) or click 2 (X=103, Y=422) (threshold=15)
> DEBUG: Adding click 0 (X=424, Y=411)
>
> Do you see the confusion ? At some point one of the coordinates is not
> updated. Successful calibration with 3.12.6 gives:
> DEBUG: Adding click 0 (X=126, Y=405)
> DEBUG: Adding click 1 (X=684, Y=399)
> DEBUG: Adding click 2 (X=128, Y=77)
> DEBUG: Adding click 3 (X=683, Y=77)
>
>
> With ts_calibrate:
> xres = 800, yres = 480
> Took 1 samples...
> Top left : X =  435 Y = 3572
> Took 2 samples...
> Top right : X = 2094 Y = 3553
> Took 2 samples...
> Bot right : X = 2939 Y = 2077
> Took 2 samples...
> Bot left : X = 2060 Y =  644
> Took 2 samples...
> Center : X = 1279 Y = 1346
>
> We experience the same thing, when changing position on an axis, we get
> an average between the previous and the new position (probably because
> we got 2 samples):
>  - Top Left is ok
>  - Top right is almost ok: X should be around 3500/3700
>  - Bot right is starting to get really wrong: X is getting better (it
>    should still be around 3500/3700) but Y should be quite lower
>  - Bot left: Y is ok but X should be lower
>  - Center is completely wrong, both values should be around 2000
>
> Last test with evtest which is always difficult because it outputs a lot
> of debug. Two separate touchs, bottom left and then top right (note that
> this is with fsl,ave-ctrl = <8>;):
>
> Event: time 1389210862.451000, type 3 (Absolute), code 0 (X), value 2183
> Event: time 1389210862.451000, type 3 (Absolute), code 1 (Y), value 720
> Event: time 1389210862.451000, type 3 (Absolute), code 24 (Pressure), value 6
> Event: time 1389210862.451000, type 1 (Key), code 330 (Touch), value 1 
> Event: time 1389210862.451000, -------------- Report Sync ------------
> Event: time 1389210862.477721, type 3 (Absolute), code 0 (X), value 448
> Event: time 1389210862.477721, type 3 (Absolute), code 1 (Y), value 667
> Event: time 1389210862.477721, type 3 (Absolute), code 24 (Pressure), value 595
> Event: time 1389210862.477721, -------------- Report Sync ------------ 
> Event: time 1389210862.504718, type 3 (Absolute), code 0 (X), value 434
> Event: time 1389210862.504718, type 3 (Absolute), code 1 (Y), value 704
> Event: time 1389210862.504718, type 3 (Absolute), code 24 (Pressure), value 580
> Event: time 1389210862.504718, -------------- Report Sync ------------ 
> Event: time 1389210862.536688, type 1 (Key), code 330 (Touch), value 0
> Event: time 1389210862.536688, -------------- Report Sync ------------
> Event: time 1389210863.359697, type 3 (Absolute), code 0 (X), value 432
> Event: time 1389210863.359697, type 3 (Absolute), code 1 (Y), value 726
> Event: time 1389210863.359697, type 3 (Absolute), code 24 (Pressure), value 210
> Event: time 1389210863.359697, type 1 (Key), code 330 (Touch), value 1 
> Event: time 1389210863.359697, -------------- Report Sync ------------
> Event: time 1389210863.391687, type 1 (Key), code 330 (Touch), value 0
> Event: time 1389210863.391687, -------------- Report Sync ------------
>
> We get 3 reports for bottom left, then pen up, then one report for top
> right that is almost exactly the same a bottom left !
>
> Output from 3.12:
> Event: time 1389210022.146809, type 3 (Absolute), code 0 (X), value 286
> Event: time 1389210022.146809, type 3 (Absolute), code 1 (Y), value 504
> Event: time 1389210022.146809, type 3 (Absolute), code 24 (Pressure), value 3045
> Event: time 1389210022.146809, type 1 (Key), code 330 (Touch), value 1 
> Event: time 1389210022.146809, -------------- Report Sync ------------
> Event: time 1389210022.166830, type 3 (Absolute), code 0 (X), value 256
> Event: time 1389210022.166830, type 3 (Absolute), code 1 (Y), value 489
> Event: time 1389210022.166830, type 3 (Absolute), code 24 (Pressure), value 3033
> Event: time 1389210022.166830, -------------- Report Sync ------------ 
> Event: time 1389210022.186812, type 3 (Absolute), code 24 (Pressure), value 0
> Event: time 1389210022.186812, type 1 (Key), code 330 (Touch), value 0 
> Event: time 1389210022.186812, -------------- Report Sync ------------
> Event: time 1389210025.196808, type 3 (Absolute), code 0 (X), value 3812
> Event: time 1389210025.196808, type 3 (Absolute), code 1 (Y), value 3637
> Event: time 1389210025.196808, type 3 (Absolute), code 24 (Pressure), value 4032
> Event: time 1389210025.196808, type 1 (Key), code 330 (Touch), value 1 
> Event: time 1389210025.196808, -------------- Report Sync ------------
> Event: time 1389210025.216819, type 3 (Absolute), code 24 (Pressure), value 0
> Event: time 1389210025.216819, type 1 (Key), code 330 (Touch), value 0 
> Event: time 1389210025.216819, -------------- Report Sync ------------
>
>
> While this is not necessarily an issue with ts_calibrate (it is possible
> to click longer so it collects more samples), I don't see that working
> with xinput_calibrator.
>
> While I don't have much experience with the TS part of the code but I
> can investigate if you don't have any idea.

You are right. I have seen the same behaviour here a few times. Currently I'm
short in time to dig deeper into this issue, but I will try to do so.

Juergen

-- 
Pengutronix e.K.                              | Juergen Beisert             |
Linux Solutions for Science and Industry      | http://www.pengutronix.de/  |