Re: [Linuxwacom-discuss] Xorg 7.3 crashes when clicking on button after xsetwacom

[Magnus V. <Mag...@ip...>]

On torsdag 21 februari 2008, Frederik wrote:
> Onsdag 20. februar 2008, skrev  Magnus Vigerlöf <Magnus.Vigerlof AT 
ipbo.se>:
> > > Fatal server error:
> > > Impossible keyboard event
> >
> > Looks like an Xorg bug.. Maybe it was introduced the same time as the
> > pointer reporting regression was introduced. I'll dig some more.
> >
> > I get the impression that is should crash no matter what you set. Does
> > any combination work at all?
>
> thanks a lot.
>
> I haven't got any combination working.
>
> Another thing i noticed is that the listing in xsetwacom don't work if you
> don't comment out the (2) lines that is about IsXExtensionDevice in
> WacomConfigListDevices in wacomcfg.c . I don't know if that is releated
>
> An old version of xsetwacom sets numbers, (at least get ... returns
> numbers). Then i have noticed that sometimes the numbers are negative, an
> that's when xorg crashed. If i set the same key multiple times sometimes it
> turns positive (after a seemingly random number of times). (I don't
> remember quite if it crashes if the numbers are positive, but I don't think
> so.)

In the git-repo [1] there are some commits that adresses the problem xsetwacom 
has with seeing the devices in Xorg 7.3. I don't think it's related, no.

Regarding the crash when sending keys as a result of a button press...

 It's caused by a change in X.org 7.3, but I wouldn't say it's a regression at 
the moment... More the result of an improvement... So the question would be 
how we should handle it in the driver..

With 7.3 X.org introduced a 'virtual' keyboard and pointer that is always 
present but not connected to any hardware. Previously it was required to have 
a defined 'core' keyboard and 'core' pointer (the latter could be 
against /dev/input/mice and that way hotplugging of mice was handled in an 
ugly way :)
So the inputInfo.keyboard is more or less a dummy device (not a real device as 
it was before) and it looks like this should not be used as we do it. I get 
the impression that each device who wish to send keyboard events should 
define its own keymap that the xserver should be able to use for translation, 
and use it self as sending device and not the virtual core keyboard.

I'm not yet sure how everything is connected when it comes to sending 
key-events but troubleshooting things like this is a great motivation for 
learning :)

Currently I'm thinking that it might be best to properly define the keymap and 
start sending key events pretty much like the evdev driver does, and not go 
through the virtual core keyboard that may or may not have a keymap that 
works for us.

Unfortunately I don't see any workaround until we fix this in a good way in 
the driver.

Cheers
  Magnus

[1] git://git.debian.org/git/collab-maint/linux-wacom.git