[PATCH libinput] RFC on quirk handling in libinput

[Peter Hutterer]

On 22/08/2015 06:33 , Andreas Pokorny wrote:
> Hi,
> The patch in the follow up mail describes a quirk needed to handle the
> touchscreen for a range of mediatek based devices. Depending on how the touch
> screen is wired to the soc, it always would report maximum == minimum == 0 for
> ABS_MT_TRACKING_ID.
>
> Whats the best way to handle this sort of broken drivers (next to fixing the
> kernel vendor)?
>
> If adding an entry to the hwdb would help - how do I get to the strings that
> identify the device?

if it's a unique ID, the lookup is something like: 
evdev:input:b0003v05ACp0259* which is just bustype, vendor and product 
ID as uppercase 4-digit hex.
otherwise, you can use the dmi info together with the device name,
see
https://github.com/systemd/systemd/blob/master/hwdb/60-evdev.hwdb
for a couple of examples.

that aside, this patch seems like something that's better suited for 
libevdev. the tracking ID is specific to the protocol B and the drivers 
are clearly buggy but it won't have any impact on users of the evdev 
protocol otherwise. so we should fix this transparently at the libevdev 
level, print a bug, correct the range and proceed as normal.

Cheers,
   Peter