x, y in struct _SynapticsSHM is ABS_X, ABS_Y, why not ABS_MT_POSITION_X, ABS_MT_POSITION_Y?

[Peter Hutterer]

On Wed, Jan 15, 2014 at 05:43:31PM +0800, Leslie Zhai wrote:
> Hi Peter,
> 
> I am debugging geis-2.2.16 these days, and also read about synaptics
> 1.7.x source code, in 1.7.x, it uses xf86PostTouchEvent, post
> multitouch event to X, then the client application, such as
> touchegg, is able to get XEvent in a while loop, is that right?

as said in the other email, I don't know much about geis and synaptics only
posts touch events in certain configurations.
 
> But in 1.7.x src/eventcomm.c, about the 576 line, it still treats
> ABS_MT_POSITION_X as cumulative_dx, and I read about your
> libtouchpad source code hosting on github src/touchpad-events.c,
> about the 42 line, it correctly treats ABS_MT_POSITION_X as x.
> 
> Take Linux multitouch protocol type B for example,
> 
> ABS_MT_SLOT 0
> ABS_MT_TRACKING_ID 45
> ABS_MT_POSITION_X x[0]
> ABS_MT_POSITION_Y y[0]
> ABS_MT_SLOT 1
> ABS_MT_TRACKING_ID 46
> ABS_MT_POSITION_X x[1]
> ABS_MT_POSITION_Y y[1]
> SYN_REPORT
> 
> During a SYN_REPORT period, there are might several
> (ABS_MT_POSITION_X, ABS_MT_POSITION_Y) 2D points to indicate the
> positions where multi-fingers touched, but synaptics 1.7.x only
> maintain 1 (ABS_MT_POSITION_X, ABS_MT_POSITION_Y) point, it might
> lost other multitouch points.

yes, I'm aware of all this - synaptics just doesn't have the code, it's too
old and never got updated properly for a multitude of reasons.

> It is able to switch to tty2, because synaptics process used device
> file descriptor, run evtest /dev/input/eventX to print_absdata, and
> I have no idea how to directly read the device fd under tty7, such
> as gnome shell, kde, sort of DE... thanks a lot :)

I don't know what the question is here, sorry. the device node doesn't
change if you change vts, it's always /dev/input/eventX.

you may be running into the issue that synaptics calls an EVIOCGRAB on the
device, hence you won't see events on the device while the synaptics driver
is active. there's a GrabEventDevice "off" option that helps with that.

Cheers,
   Peter

> 
> >On Wed, Jan 08, 2014 at 09:28:29PM +0800, Zhai Xiang wrote:
> >>Hi Peter,
> >>
> >>Sorry for my late reply because I wen to Beijing for a short business
> >>travel.
> >>
> >>Yes, I read about the source code about 1.7.x by apt-get source
> >>xserver-xorg-input-synaptics under Linux Deepin 2014, the SHM has been
> >>completely removed.
> >generally I recommend looking at the git sources before you hack on any
> >software project, much more likely to be up-to-date than the package a
> >distribution ships.
> >
> >>And I do not want to hack the 1.6.x source code under Linux Deepin 2013 to
> >>support ABS_MT_SLOT, ABS_MT_POSITION_X && ABS_MT_POSITION_Y sort of
> >>multitouch support, even it supports multitouch and gesture recognition,
> >>such as 2 fingers to zoom out/in, 4 fingers swipe up to show all open
> >>windows, 4 fingers swipe right to switch to 2nd work space, 4 fingers swipe
> >>left to switch to 1st work space, 3 fingers to move active window, the demo
> >>video shown as http://v.youku.com/v_show/id_XNjU5MTk4MTk2.html
> >>
> >>You said you are writing the driver with multitouch support using a
> >>different base design, can I ask whether or not the driver`s source code
> >>will be accepted by xorg official maintainer?
> >well, given that I'm the input maintainer I'm gonna say "yes" :)
> >
> >>Would it be OK for you if I also contribute to your driver? If it is OK,
> >>may I have your driver source code version control link such as github or
> >>something else?
> >it's called libinput but currently spread across some personal github
> >repositories. it'll be on freedesktop soon, but it's still very early
> >stages with a lot of the top-level work to be sorted out first so it's a bit
> >too early to go full volume on hacking on it. I've got an experimental
> >driver in https://github.com/whot/libtouchpad which is a test-bed for a
> >couple of things that I'll move over to libinput eventually.
> >
> >Cheers,
> >    Peter
> >
> >>2014/1/6 Peter Hutterer <peter.hutterer at who-t.net>
> >>
> >>>On Sat, Jan 04, 2014 at 04:25:51PM +0800, Leslie Zhai wrote:
> >>>>Hi xorg developers,
> >>>>
> >>>>in xserver-xorg-input-synaptics-1.6.2/include/synaptics.h
> >>>>the struct _SynapticsSHM is shown as below:
> >>>SHM has been completely removed in synaptics 1.7, it's largely a leftover
> >>>from before device properties where it was needed for run-time
> >>>configuration. it had't been updated in a while, so any code that deals
> >>>with
> >>>it is likely outdated.
> >>>
> >>>updating synaptics to support multitouch properly is a bit of a larger
> >>>issue, it's likely easier to write the driver from scratch using a
> >>>different
> >>>base design (which I'm currently in the process of doing).
> >>>
> >>>Cheers,
> >>>    Peter
> >>>
> >>>>```
> >>>>typedef struct _SynapticsSHM {
> >>>>int version; /* Driver version */
> >>>>
> >>>>/* Current device state */
> >>>>int x, y; /* actual x, y coordinates */
> >>>>int z; /* pressure value */
> >>>>int numFingers; /* number of fingers */
> >>>>int fingerWidth; /* finger width value */
> >>>>int left, right, up, down; /* left/right/up/down buttons */
> >>>>Bool multi[8];
> >>>>Bool middle;
> >>>>} SynapticsSHM;
> >>>>```
> >>>>
> >>>>there are x and y, because I want to know what they are, then read about
> >>>>src/eventcomm.c
> >>>>```
> >>>>if (ev.code < ABS_MT_SLOT) {
> >>>>switch (ev.code) {
> >>>>case ABS_X:
> >>>>hw->x = apply_st_scaling(proto_data, ev.value, 0);
> >>>>break;
> >>>>case ABS_Y:
> >>>>hw->y = apply_st_scaling(proto_data, ev.value, 1);
> >>>>break;
> >>>>case ABS_PRESSURE:
> >>>>hw->z = ev.value;
> >>>>break;
> >>>>case ABS_TOOL_WIDTH:
> >>>>hw->fingerWidth = ev.value;
> >>>>break;
> >>>>}
> >>>>}
> >>>>```
> >>>>
> >>>>Why use ABS_X for hw->x?
> >>>>I want to recognize multi-touch gesture, such as zoom in/out, with
> >>>>ABS_MT_SLOT, ABS_MT_POSITION_X and ABS_MT_POSITION_Y based on Linux
> >>>>Multi-touch (MT) Protocol.
> >>>>But struct _SynapticsSHM only provides ABS_X and ABS_Y, even there is
> >>>>numFingers, it is still unable to distinguish which finger touching the
> >>>>point1, point2 or pointN...
> >>>>
> >>>>So I hacked the struct _SynapticsSHM, added int slot, mt_x, mt_y, shown
> >>>>as below:
> >>>>```
> >>>>typedef struct _SynapticsSHM {
> >>>>int version; /* Driver version */
> >>>>
> >>>>/* Current device state */
> >>>>int x, y; /* actual x, y coordinates */
> >>>>int z; /* pressure value */
> >>>>int numFingers; /* number of fingers */
> >>>>int fingerWidth; /* finger width value */
> >>>>int left, right, up, down; /* left/right/up/down buttons */
> >>>>Bool multi[8];
> >>>>Bool middle;
> >>>>
> >>>>int slot;
> >>>>int mt_x;
> >>>>int mt_y;
> >>>>} SynapticsSHM;
> >>>>```
> >>>>
> >>>>and also hacked src/eventcomm.c, using ABS_MT_POSITION_X for mt_x, shown
> >>>>as below:
> >>>>```
> >>>>case EV_ABS:
> >>>>if (ev.code == ABS_MT_SLOT)
> >>>>hw->slot = ev.value;
> >>>>if (ev.code == ABS_MT_POSITION_X)
> >>>>hw->mt_x = ev.value;
> >>>>if (ev.code == ABS_MT_POSITION_Y)
> >>>>hw->mt_y = ev.value;
> >>>>```
> >>>>
> >>>>Then I can use slot, mt_x, mt_y to recognize multi-touch gesture such as
> >>>>zoom in, rotate ... but that is hacking way, perhaps there are some
> >>>>synaptics developers already considered about multi-touch gesture
> >>>>recognization requirement, it might be in TODO list.
> >>>>
> >>>>Please someone give me some advice, thanks a lot!
> >>>>
> >>>>Leslie Zhai
> >>>>_______________________________________________
> >>>>xorg-devel at lists.x.org: X.Org development
> >>>>Archives: http://lists.x.org/archives/xorg-devel
> >>>>Info: http://lists.x.org/mailman/listinfo/xorg-devel
> >>>>
>