New input hotness not happening for 1.5

Peter Hutterer mailinglists at who-t.net
Thu Dec 6 19:02:18 PST 2007


Daniel Stone wrote:
> Despite putting in the XDS2007 notes that input transformation, XKB 2
> and Xi 2 would happen for 1.5, they just aren't going to, to be honest.
> Doing these sensibly requires some fairly sweeping changes to the way we
> _process_ events (input-hotplug was all about how we generate them) that
> also requires sorting out the interactions with MPX, particularly the
> grab bits, which are going to be tough.  That and I badly need a break.

FWIW, here's a quick update on the state of the "rest of XI2" (i.e. MPX).
First, MPX will not happen for 1.5, but 1.6 may be realistic.

The most recent change added the device hierarchy as put forward by 
Keith Packard at the XDS07.

This allows you to create and remove new cursors and keyboard foci 
(master devices) at runtime and attach the physical devices (slave 
devices) to any of these. And change them around as you need.

By default, the devices are all assigned to a single cursor/focus which 
resembles exactly the current behaviour of the server. this is a good 
thing, unless you know mpx is there and start using it, it's invisible.

Aside from that, it's seems to be working well so far but as we have 
seen from 1.4 if it is working on one box doesn't mean it caters for all 
configurations.

I _think_ most of the grab-related issues are sorted out, although more 
testing is required. One missing thing is that I need to go through all 
extensions and replicate their APIs with device-specific ones if 
possible. XTest comes to my mind here.

The blob stuff (i.e. touch screen support) is in a decent state now too, 
allowing clients to query devices for the support etc. I'm quite keen to 
merge it into mpx. can I please get a heads-up whether we want that in 
the first version of mpx?

What I would like to see is people actually trying to write some apps 
with multiple devices. The number of these apps is still so limited that 
considering the API stable is a bit of a speculation.

Cheers,
   Peter



More information about the xorg mailing list