[patch] XInput support for GTK

[Owen Taylor <owt1 cornell edu>]

I've finished (at least roughly) migrating XInput support into
the GTK core. The monster patch is available from my gsumi web page
at:

   http://www.msc.cornell.edu/~otaylor/gsumi/gsumi.html

Because of the number of patches floating around, and to save
people some difficulty, I've also put a copy of my gtk+ source
tree at the same location. It includes all bug-fix patches I know
about, plus Li-Da Lho's I18N patch, and my paned window widgets,
but not the DND or modal dialog stuff, and not Molaro Paolo's
multi-line patch, and not gtkgamma ... other than that it's
complete ;-) This is purely meant as a convenience thing, and
should not be construed as any sort of release.

[ In this context, I should confess that the paned window patch
  I posted a pointer to yesterday does not work cleanly. But
  if you take out the patch to gtk.h and do that manually, it
  should work. ]

As a companion to this I've also released gsumi-0.5 which
is gsumi-0.4 minus everything that's been moved into gtk/gdk.

-----

The interface to the patch is pretty similar to what Raph, then
I suggested a couple of weeks ago. The type of XInput support 
is controlled by a configure option

   --with-xinput=[no/gxi/xfree]


Where 'gxi' indicates support for device switching via the
included gxid daemon, and 'xfree' support via the mechanisms
introduced in XFree86 3.3. (Because of bugs, the later is not
useful with the stock release of 3.3)

[ The design decision I have the most doubts about was to
  call everything gdk_input_ ... based on the general trend
  of turning XBlah into gtk_blah. Hopefully, people won't
  get too confused by gtk_input_add now ... ]


The patch also includes an Input Device control dialog widget for
gtk, and a new example program 'testinput.c' in the gtk/
directory which demonstrates some of the basics of using the
extension. (It's not quite as simple as I would like, because of
the necessity of drawing the cursor manually with the gxid style
switching.)

I haven't implemented motion history yet, and the details of
tilt support are waiting for someone with a device that supports
it, but otherwise (I hope), it's in a pretty finished form -
it should be good enough to start work the GIMP.

As always, complaints, bug reports, patches, (and a complete
implementation of the rest of XInput support for the GIMP) would
be appreciated.

Regards,
                                        Owen