[Xcb] patches from Arnaud Fontaine

[Arnaud Fontaine]

>>>>> "Ian" == Ian Osgood <iano at quirkster.com> writes:

Hi,

    Ian> I think  it is time to split some of  the more useful libraries
    Ian> out of util into their own versioned repositories.

    Ian>   1.  aux:   most  programs   use   xcb_aux_get_screen().  Most
    Ian>   simple-minded  Xlib  ports   use  xcb_aux_sync()   and  other
    Ian> convenience functions.

    Ian> 2. renderutil:  cairo's XCB backend depends on  this now, so it
    Ian> would be nice to have it independently versioned.

    Ian> 3.  image: Bart's implementation  is now mature  and featureful
    Ian> enough  for general  use. (One question:  do we need  a wrapper
    Ian>  that hides  the  socket/shm transport  mechanism? I'm  unclear
    Ian> whether Xlib does this.)

I agree with this, I think  it should be quite convenient to split these
libraries.

    Bart> Probably  what needs to  happen is that  we need to  split off
    Bart>  icccm and  wm  from the  rest  of util  and  make Arnaud  the
    Bart>  maintainer of  it.  I  think he's  the person  currently most
    Bart>  interested  in it,  and  seems  to  be making  progress  with
    Bart> it---I'd be up for this.

    Bart> In  the meantime, Arnaud  should start the process  by cloning
    Bart> util, deleting the stuff  other than icccm and wm, and pushing
    Bart> the result to  a private repo on people.freedesktop.org.  When
    Bart> we're  happy with  what's going on  there, we'll move  it back
    Bart> into  the canonical tree, and  in the meantime  he can proceed
    Bart> with his work.

    Ian> 4.  atom and icccm:  most non-trivial X applications  will want
    Ian> these.  Maybe get a wider API review first.

I am  alright to take care  of icccm as I  use it for Awesome  even if I
will probably  ask questions sometimes in  case of problems.  There is a
lot of things to work on in icccm though.

Concerning xcb-util/atom, I would like  to add an atom cache (e.g. maybe
by extending xcb_intern_atom_fast in order to allow atoms for a specific
application to  be cached),  but I would  like some feedbacks  about the
implementation. Does  someone has some ideas about  implementating it in
an efficient way?

Regards,
Arnaud Fontaine