On Tue, 4 Mar 2008, Ville Syrjälä wrote:

quoted

Actually, this is the very same issue we discussed with respect to 
ati_remote2 changes ... this device has Keyboard usages as well as 
Mouse/Pointer. As we set the EV_REP (because of the Keyboard usage), 
this inputdevice-wide flag applies also to the mouse button events and 
they get repeated.

Except this device does actually register two input devices since the
endpoints for the keyboard and mouse are separate, and here it even
makes sense since they are physically separate devices.

Yes, but unfortunately both interfaces contain Keyboard usage in their 
report descriptor -- please see the debug output you sent me. The first 
interface's descriptor states to send Keyboard usages in Field(1) of INPUT 
report, the second one presents it in Field(1) of INPUT[1] report.

I guess that one of the report descriptors contains redundant information, 
not both interfaces are actually sending Keyboard/Consumer/etc events, 
right?

The possibility for this particular device might be fixing the report 
descriptor on the fly, before it enters the parser. We do it for certain 
broken HID devices already.

-- 
Jiri Kosina
SUSE Labs