Bug#437273: xserver-xorg-input-synaptics: Synaptics touchpad doesn't work on 'switch user'

[Peter Osterlund <petero2@telia.com>]

On Sat, 11 Aug 2007, Mattia Dongili wrote:

> On Sat, Aug 11, 2007 at 12:40:55PM +0100, Neil Wilson wrote:
> > Package: xserver-xorg-input-synaptics
> > Version: 0.14.6-0ubuntu7
> > Severity: normal
> > Tags: patch
> ...
> > Running Ubuntu - fault in versions for Feisty and Gutsy. Fault also
> > exists in current Debian release.
> >
> > After a 'switch user' on the console, the mouse locks for the second
> > user. This is because the software does not release the console mouse on
> > switch.
> >
> > patch attached.
> >
> > Ubuntu bug is https://bugs.launchpad.net/bugs/68370
>
> > diff -ur xserver-xorg-input-synaptics-0.14.6/eventcomm.c xserver-xorg-input-synaptics-0.14.6-ds/eventcomm.c
> > --- xserver-xorg-input-synaptics-0.14.6/eventcomm.c	2007-07-09 19:42:55.000000000 -0700
> > +++ xserver-xorg-input-synaptics-0.14.6-ds/eventcomm.c	2007-07-09 19:11:03.000000000 -0700
> > @@ -56,6 +56,12 @@
> >  static void
> >  EventDeviceOffHook(LocalDevicePtr local)
> >  {
> > +    int ret;
>
> in the debian version (that'a git snapshot) you'll need to look at
> grab_event_parameter (it doesn't hurt)
>
> > +    SYSCALL(ret = ioctl(local->fd, EVIOCGRAB, (pointer)1));
>
> and maybe you want to pass 0 as parameter.
>
> > +    if (ret < 0) {
> > +	xf86Msg(X_WARNING, "%s can't ungrab event device, errno=%d\n",
> > +		local->name, errno);
> > +    }
> >  }
>
> Anyway, Peter does this make sense to you?

Unfortunately it doesn't, because the synaptics driver closes the event 
device anyway in the DeviceOff function. Closing the device will cause an 
automatic ungrab, so the patch at best only changes the timings. Maybe a 
wait-and-retry loop in EventDeviceOnHook would help.

I think what happens is that the driver running in the second X server 
tries to grab the device before the driver in the first X server has 
released it. I don't know if the X server is supposed to synchronize this 
somehow, or if it is up to the synaptics driver to deal with the 
situation.

The latest git version has an option GrabEventDevice that can be set to 
false to bypass the grab/ungrab. This would probably help, except that it 
doesn't work well if the X config file also uses /dev/input/mice. (Which 
most configs do.)

--
Peter Osterlund - petero2@telia.com
http://web.telia.com/~u89404340