Input Devices (was: Re: [Xgl/Xegl] Input Devices )

[Jim Gettys]

On Fri, 2005-07-29 at 14:10 +0200, Waldo Bastian wrote:

> > It probably is better to use DBUS, for much of it, though we'll have to
> > figure out some way to send the file descriptor.  The issue here is that
> > the current internals of the X server do no have the ability for the X
> > server to become a client of another server (in this case, DBUS); it
> > might deadlock.
> 
> Even though the X server would communicate with the dbus server, wouldn't, 
> conceptually speaking, the X server still be the server in this case? The 
> dbus connection would just be another connection for clients to connect to 
> the xserver and send commands to the X server.
> 
> I can see deadlock potential if the X server was to make calls across DBUS and 
> blocked waiting for a response, but I don't think there is a need for that. 
> First of all, with a configuration client, the X server itself would not need 
> to be making any calls. Second, if it would need to make a call it could make 
> the call asynchronously and the response would come back later, pretty much 
> like other incoming dbus calls.
> 
> All that is needed is a way to register a fd and have the X server activate a 
> callback when something happens on that fd. Is that possible at the moment, 
> or does that require reorganizing the X server internals first?
> 

Not really; the server implementation has all combinations of facilities
needed *except* for the one we need, and the code is pretty gross.
Kristian took a close look at this last fall before he was diverted to
helping complete Cairo (a correct call, btw; the pdf and postscript
backends are vital for it).

Rather than just hacking something in, I'd prefer to build a proper
general facility, use it, and then we can gradually migrate all the
existing uglyness over to it, resulting in a much cleaner server
implementation.  I suspect it is a few weeks to a month's work to build
the new select/poll facility.
				Regards,
					- Jim