Edward O'Callaghan wrote:
> Right..
> Not sure what you mean about xcb being large and complex ?

Fully integrating xcb includes replacing the transport &
multi-threaded locking mechanisms of libX11.   Those
interactions are complex and need to be well understood and
tested.

Questions off the top of my head that we need answers to:

 - Solaris libX11 includes supports for transport types not
   used on Linux & BSD, including Named Pipes (also available
   in upstream X.Org, but only used on SysV systems) and
   the SUN_SME shared memory transport (a Solaris-ism we've
   never gotten around to pushing upstream) - does xcb support
   pipes?  What do we need to do to add SME support, or does this
   force us to drop it?

 - Solaris libX11 connection code also has special support for
   the IA process control extension that I've not gotten upstream
   yet - what needs to happen to xcb to make this work?

 - Linux distros that adopted libxcb discovered the threading
   changes caused Java to deadlock - how do we make sure we
   avoid this on Solaris?   What Solaris Java versions may be
   affected?   (Just saying "Use latest Java" doesn't help
   people who want to install Oracle which forces its install
   gui to use their bundled JRE.)

   http://bugs.sun.com/view_bug.do?bug_id=6532373

   Are any other programs going to be similarly affected?
   For instance, I doubt anyone on the planet has ever run CDE
   with an xcb-ified libX11 before.

 - Is there an increase or decrease in application performance
   when using the xcb-ified libX11?

 - Our libX11 has a bunch of thread/locking changes that aren't
   upstream - are these still correct with libxcb?  Do we need
   to change them?

 - We have several extensions that haven't been upstreamed,
   do we need to create libxcb xml protocol descriptions of them?

> So what am I to do to accelerate this process ?

Either help us do all this, or tell me why this is important to
you and give me a reason to prioritize this work over the other
stuff we're working on.  Is this more important than finishing
the migration of our old code base to current open source
releases that we can publish the sources for on opensolaris.org?
More important than upgrading Xorg server to 1.5.x?   I can't
make more people appear to do more work, and no one is sitting
around here with nothing to do, so would have to ask someone
to stop working on other things to work on this.

-- 
        -Alan Coopersmith-           alan.coopersmith at sun.com
         Sun Microsystems, Inc. - X Window System Engineering


Reply via email to