2004-06-06 17:48:44

by Vojtech Pavlik

[permalink] [raw]
Subject: Re: [RFC/RFT] Metter mixing of relative & absolute devices by mousedev

On Thu, May 06, 2004 at 01:44:52AM -0500, Dmitry Torokhov wrote:
> Hi,
>
> Currently mousedev re-interprets every event for every client (reader) that
> has either /dev/input/mouseX or /dev/input/mice opened. For absolute events
> mousedev converts them into relative motion data using device's size, screen
> size and previous absolute position. For /dev/input/mice that gets events
> all devices this poses a small problem as true relative events do not adjust
> previous position - so if you were using touchscreen, then a mouse and were
> to touch the thouchscreen again the cursor would jump before getting to your
> finger.
>
> Please consider the following patch - it processes every event only once and
> converts it into relative motion. The mousedev clients are fed with relative
> motion data only this eliminating "jumping cursor" problem.

I see I didn't comment on this one yet: Yes, it makes a lot of sense.
Can you prepare a pull for me, or do I have it already?

--
Vojtech Pavlik
SuSE Labs, SuSE CR


2004-06-06 18:49:02

by Dmitry Torokhov

[permalink] [raw]
Subject: Re: [RFC/RFT] Metter mixing of relative & absolute devices by mousedev

On Sunday 06 June 2004 12:48 pm, Vojtech Pavlik wrote:
> On Thu, May 06, 2004 at 01:44:52AM -0500, Dmitry Torokhov wrote:
> > Hi,
> >
> > Currently mousedev re-interprets every event for every client (reader) that
> > has either /dev/input/mouseX or /dev/input/mice opened. For absolute events
> > mousedev converts them into relative motion data using device's size, screen
> > size and previous absolute position. For /dev/input/mice that gets events
> > all devices this poses a small problem as true relative events do not adjust
> > previous position - so if you were using touchscreen, then a mouse and were
> > to touch the thouchscreen again the cursor would jump before getting to your
> > finger.
> >
> > Please consider the following patch - it processes every event only once and
> > converts it into relative motion. The mousedev clients are fed with relative
> > motion data only this eliminating "jumping cursor" problem.
>
> I see I didn't comment on this one yet: Yes, it makes a lot of sense.
> Can you prepare a pull for me, or do I have it already?
>

You should have pulled it last time (before today).
--
Dmitry