Severe X breakage heads up

Bernardo Innocenti bernie at codewiz.org
Tue Nov 6 15:09:58 PST 2007


(cc davidz, hal ml)

On 11/06/07 10:00, Nicolas Mailhot wrote:

> 95% of the user-reported problems with evdev are due to the user
> switching himself and failing to do it properly (because switching to
> evdev requires setting an awful lot of xorg parameters to "evdev", and
> if you forget to change one xorg won't fail cleanly but work in a
> strange hybrid buggy PS/2+evdev input mode, and users will assume this
> buggy mode is evdev itself).

With input hotplug, we have exactly *zero* instances of
"evdev" in xorg.conf :-)

And the new evdev 1.2 also deduces the fancy bits stuff
directly from /dev/input/event* devices, as set by the
drivers.

This is where an OLPC-specific bug creeped: the combined
driver for the touchpad and glide sensor was improperly
setting the bits that describe the axes supported by
the input device.

HAL also is too strict: to detect a device as a touchpad,
it requires abs X/Y axes and PRESSURE.  We lack pressure
in the glide-sensor, so it was not being recognized
properly.

Now we're faking the PRESSURE axis just to satisfy hal.
If we just remove the check for PRESSURE, hal would
confuse analog joysticks with touchpads.

In order to detect a joysticks, we could look for the
presence of buttons.  I don't have one to test with.
Can anyone do a quick test and report what bits are
set in evdev?


> Just switching to evdev can fix an awful lot of multimedia key problem
> reports.

Moreover, it would let applications to discern multiple
input devices from one another.  Very useful for music and
videogames.

-- 
 \___/
 |___|   Bernardo Innocenti - http://www.codewiz.org/
  \___\  One Laptop Per Child - http://www.laptop.org/


More information about the hal mailing list