On Thu, Dec 13, 2012 at 9:37 PM, Peter Hutterer <peter.hutte...@who-t.net>wrote:
> On Wed, Dec 12, 2012 at 11:08:17AM -0800, Ping Cheng wrote:
> > On Wed, Dec 12, 2012 at 10:54 AM, Jason Gerecke <killert...@gmail.com
> >wrote:
> >
> > > It certainly doesn't hurt :) I think my biggest problem is that I never
> > > really understood exactly when this code was required.
> >
> >
> > I bet you don't want to know ;-). Chris knows this chunk of code way
> better
> > than me. It is a long story and it is history already. I'll share it
> > offline, with those who are curious...
>
> no, please share it online so it's archived for future generations.
> this code isn't that old, so if we can't even remember what it was written
> for we really need to document better.
>
> commit messages are free, we might as well use them properly.
>
> afaict, this code was there because before the MT protocol we used the
> BTN_TOOL_DOUBLETAP and friends to switch between tools. but because this
> was
> abusing the system a bit, switching to a different tool wouldn't update the
> other coordinates for that tool (the kernel would skip those events if they
> were on the last-sent coordinates for the previous tool), so we had to copy
> that around in the driver.
>
> except for the pad which is always in proximity and didn't rely on
> multiplexing, so we could skip that.
>
> does this sound correct?
>
> Cheers,
> Peter
>
>
Right, except that "tool" means fingers in this case. The "old kernel way"
is explained in some detail on this page under the "Touchpad Overview"
section.
https://sourceforge.net/apps/mediawiki/linuxwacom/index.php?title=Kernel_Input_Event_Overview
The concept of that deleted code is that any user land apps need to track
last values sent by kernel that are shared between tools and when switching
tools you need to continue using the last sent values because of event
filtering. Since we are storing X/Y/Pressure in different "channels", when
switching channels the X/Y/Pressure fields need to be copied over at switch
time.
For MT devices, there is not same concept of shared events so the code can
be deleted.
Protocol 5 devices and any Protocol 4 with mice do in fact need similar
logic but we don't have it; as the Protocol 5 deleted comment alludes to.
We are most likely losing data during some tool/finger switch overs for
those devices but since they mostly work in Absolute mode almost no one
notices it because of fast recovery.
FYI: I see now that I documented "Synaptics" wrong on wiki. The
X/Y/Pressure for syntaptics are really always bound to the FINGER tool and
DOUBLETAP is more an extra tool with no X/Y/Pressure associated directly
with it. Something closer to this:
* BTN_TOOL_FINGER
* BTN_TOUCH
* ABS_X
* ABS_Y
* ABS_PRESSURE
* BTN_TOOL_DOUBLETAP
* BTN_LEFT
* BTN_RIGHT
Chris
> >
> > Ping
> >
> > It makes sense that only MT using *TAP events would need it (or
> dual-track,
> > > but that's protocol 5), so I don't really see any problem given the
> testing
> > > you've done.
> > >
> > >
> > > Jason
> > >
> > > ---
> > > When you're rife with devastation / There's a simple explanation:
> > > You're a toymaker's creation / Trapped inside a crystal ball.
> > > And whichever way he tilts it / Know that we must be resilient
> > > We won't let them break our spirits / As we sing our silly song.
> > >
> > >
> > >
> > > On Wed, Dec 12, 2012 at 10:39 AM, Ping Cheng <pingli...@gmail.com>
> wrote:
> > >
> > >> On Wed, Dec 12, 2012 at 9:45 AM, Jason Gerecke <killert...@gmail.com
> >wrote:
> > >>
> > >>> Though I can't quite convince myself that this is safe, I don't see
> any
> > >>> problems with it.
> > >>>
> > >>> Acked-By: Jason Gerecke <killert...@gmail.com>
> > >>>
> > >>
> > >> Will I say it is tested on older and new systems make it easier to
> > >> convince you?
> > >>
> > >> Ping
> > >>
> > >>
> > >>> Jason
> > >>>
> > >>> ---
> > >>> When you're rife with devastation / There's a simple explanation:
> > >>> You're a toymaker's creation / Trapped inside a crystal ball.
> > >>> And whichever way he tilts it / Know that we must be resilient
> > >>> We won't let them break our spirits / As we sing our silly song.
> > >>>
> > >>>
> > >>>
> > >>> On Thu, Dec 6, 2012 at 4:20 PM, Ping Cheng <pingli...@gmail.com>
> wrote:
> > >>>
> > >>>> We use true MT protocol for MT devices in kernel now. This code
> > >>>> was introduced to deal with ABS_TOOL_*TAP events loss issue. It
> > >>>> is uncessary any more. And its existence makes it hard to support
> > >>>> generic PAD cleanly.
> > >>>>
> > >>>> Signed-off-by: Ping Cheng <pi...@wacom.com>
> > >>>> ---
> > >>>> src/wcmUSB.c | 34 ++--------------------------------
> > >>>> 1 file changed, 2 insertions(+), 32 deletions(-)
> > >>>>
> > >>>> diff --git a/src/wcmUSB.c b/src/wcmUSB.c
> > >>>> index e192489..4b5f53b 100644
> > >>>> --- a/src/wcmUSB.c
> > >>>> +++ b/src/wcmUSB.c
> > >>>> @@ -37,7 +37,6 @@ typedef struct {
> > >>>> Bool wcmPenTouch;
> > >>>> Bool wcmUseMT;
> > >>>> int wcmMTChannel;
> > >>>> - int wcmPrevChannel;
> > >>>> int wcmEventCnt;
> > >>>> struct input_event wcmEvents[MAX_USB_EVENTS];
> > >>>> int nbuttons; /* total number of buttons */
> > >>>> @@ -1601,11 +1600,8 @@ static void usbDispatchEvents(InputInfoPtr
> pInfo)
> > >>>> return;
> > >>>> }
> > >>>>
> > >>>> - /* Protocol 5 devices have some complications related to
> > >>>> DUALINPUT
> > >>>> - * support and can not use below logic to recover from input
> > >>>> - * event filtering. Instead, just live with occasional
> dropped
> > >>>> - * event. Since tools are dynamically assigned a channel
> #, the
> > >>>> - * structure must be initialized to known starting values
> > >>>> + /* Protocol 5 tools are dynamically assigned with channel
> > >>>> numbers.
> > >>>> + * The structure must be initialized to known starting
> values
> > >>>> * when first entering proximity to discard invalid data.
> > >>>> */
> > >>>> if (common->wcmProtocolLevel == WCM_PROTOCOL_5)
> > >>>> @@ -1614,32 +1610,6 @@ static void usbDispatchEvents(InputInfoPtr
> pInfo)
> > >>>> memset(&common->wcmChannel[channel],0,
> > >>>> sizeof(WacomChannel));
> > >>>> }
> > >>>> - else
> > >>>> - {
> > >>>> - /* Because of linux input filtering, each switch to
> a
> > >>>> new
> > >>>> - * tool is required to have its initial values match
> > >>>> values
> > >>>> - * of previous tool.
> > >>>> - *
> > >>>> - * For normal case, all tools are in channel 0 and
> so
> > >>>> - * no issue. Protocol 4 2FGT devices split between
> > >>>> - * two channels though and so need to copy data
> between
> > >>>> - * channels to prevent loss of events; which could
> > >>>> - * lead to cursor jumps.
> > >>>> - *
> > >>>> - * PAD device is special. It shares no events
> > >>>> - * with other channels and is always in proximity.
> > >>>> - * So it requires no copying of data from other
> > >>>> - * channels.
> > >>>> - */
> > >>>> - if (private->wcmPrevChannel != channel &&
> > >>>> - channel != PAD_CHANNEL &&
> > >>>> - private->wcmPrevChannel != PAD_CHANNEL)
> > >>>> - {
> > >>>> - common->wcmChannel[channel].work =
> > >>>> -
> > >>>> common->wcmChannel[private->wcmPrevChannel].work;
> > >>>> - private->wcmPrevChannel = channel;
> > >>>> - }
> > >>>> - }
> > >>>>
> > >>>> ds = &common->wcmChannel[channel].work;
> > >>>> dslast = common->wcmChannel[channel].valid.state;
> > >>>> --
> > >>>> 1.7.10.4
> > >>>>
> > >>>>
>
>
------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
Linuxwacom-devel mailing list
Linuxwacom-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linuxwacom-devel
