Remove absolute device class

Simon Thum simon.thum at gmx.de
Sun Mar 20 08:28:21 PDT 2011


I made an attempt to remove the absolute device class from the server. It was, to the best of my knowledge,
unused and quite misleading since all is there except for real useage. The actual removal (patch 5+6)
is split up so the API is kept (more or less) for the benefit of out-of-tree drivers.
"More or less" refers to the AbsoluteClassRec/Ptr pair which is being removed, but
no-one out of the server seems to have used it. The protocol and libXi parts remain
untouched so GIMP won't break.

Also, the server is more honest to clients asking for AbsoluteClass support, i.e. now you
should get protocol errors instead of silent ignorance.

The real removal merits an ABI bump, but TBH I was unsure how to do it properly and
whether or not it's overkill. No driver I'm aware of used it.

A possible alternative would be actually using this stuff, but there's lots of overlap
with existing input properties and the class does x/y only. The only senario I could imagine
was making better use of raw absolute events, but since we have sub-pixel events
that's a questionable use case.

[PATCH 1/6] xquartz: simplify ChangeDeviceControl
[PATCH 2/6] kdrive: don't pretent to support DEVICE_ABS_* in ChangeDeviceControl
[PATCH 3/6] xf86: don't pretend to support DEVICE_ABS_* in ChangeDeviceControl
[PATCH 4/6] simplify ChangeDeviceControl in stubs
[PATCH 5/6] xserver: remove AbsoluteClassRec keeping the ABI
[PATCH 6/6] xserver: remove AbsoluteClass, breaking the A(P|B)I


More information about the xorg-devel mailing list