Re: X11R7.0

[Zephaniah E. H. <wa...@ae...>]

On Wed, Feb 01, 2006 at 09:34:20AM +0200, Aivils Stoss wrote:
> On Otrdiena, 31. Janv=C4=81ris 2006 22:19, Zephaniah E. Hull wrote:
> > > Also one version before latest will not work proper if i create/delete
> > > devices in the runtime. It is not official but i provide this one:
> > > http://www.ltn.lv/~aivils/files/xentity-0.01.tar.bz2
> > > This package contains xmodule, which allow load/remove X server
> > > modules and xdevice, which allow create/delete X server input devices
> > > at runtime. Existing input drivers are compatible with xdevice,
> > > but Your not. Even xdevice never goes into main it make life easy
> > > during development. You don't need server restart on new driver
> > > version or on device parameter change. Fatal errors still are fatal.
> > > Syntaxis like this:
> > > # xmodule -l evdev
> > > # xdevice -c e1,evdev,Phys,isa0060/serio0/input0,AlwaysCore,1
> > > You driver changes device indentifier unwarranted. You must search new
> > > device
> > > # xsetpointer -l
> > > normal device delete uses same identifier
> > > # xdevice -d e1
> >
> > That looks handy, however due to hotplug support my driver is going to
> > cause, interesting issues.
> >
> > Mainly, we create new driver instances for additional devices that are
> > plugged in that match the specification for current devices.
> >
> > Because we need the new entries to be uniquely identifiable, we append
> > the phys to the name.
> >
> > At this time there is no way around this.
> >
> > Also at the moment I'm not using a modular tree, which makes using that
> > a bit harder even for testing purposes.
>=20
> That package didn't use compile form scratch. All what You need is X serv=
er
> source for headers, runing X binary for some variables, gcc, X server res=
tart
> is necessary. I use it under my 6.8.2.
>=20
> xdevice was developed keeping in mind buch of all input devices. It does
> "hotplug" in large-scale without touching driver source.
> Seems SIGHUP is sacred. Even aiptek tablet developers did not replay.
> From my point of view looks very stupid - plug new input device and then
> restart X for new config.

After taking a closer look at it, it's a nice idea, but I'd suggest
sitting down and considering the design and all implications for a
while.

Among other things, the kernel ioctl interface is a rather poor one to
emulate, while at first glance it looks fairly easy to deal with, it has
some noticeable scale issues as far as maintance goes.

Take a look at some of the protocol discussions in the X.org wiki on the
Hotplug subject, and keep trying.

(The evdev driver is nowhere even close to being the first, second, or
probably even eighth design for it that I actually wrote the code for,
never mind that I've had.  And it still has some design issues that I
need to fix.)

Zephaniah E. Hull.
(Spelling not guaranteed this time around, I'm tired and going to bed.)

--=20
	  1024D/E65A7801 Zephaniah E. Hull <wa...@ae...>
	   92ED 94E4 B1E6 3624 226D  5727 4453 008B E65A 7801
	    CCs of replies from mailing lists are requested.

<nonlinear> .net is microsofts perverted version of a java networked
             environment uglified for windows-specific crap
	-- nonlinear in #opengl