[Xcb] socket handoff problems

[Clemens Eisserer]

Hi again

> There must be a way to read stuff back; I'm just not sure
> what Jamey and Josh have provided.  Handing off the socket
> would be pretty meaningless without a way to read stuff back
> on it, seems to me.
I am not sure this is really intended - wouldn't reading stuff back
myself break compatibility with existing C based code?  E.g. in my
mixed libX11+self-generated-protocol case, what would orrcur if I
would read/consume events which belong to the C part. Would I have to
manually call the event listener methods?

> On the contrary, sorry we've been so lame at getting your
> bug fixed.  I've been pretty swamped lately, I'm afraid.
Thanks for your patience :)

>> Do you know if there exists some documentation about this?
>
> Not that I know of. :-)
Hmm, that doesn't make it any easier ;-)

> It's done with some magic selects and a condition variable
> buried in the bowels of the XCB bottom half.  You might want
> to skim our old Usenix ATC paper entitled "X Meets Z" to get
> some insight.
Am I right with the assumption that never reading back is only a
problem if the x-server's output buffer is filled by e.g. events or
messages?

Do you have any idea why it hangs when calling the C interface again?
Could it be the XServer has suspended the client, because it behaved
not as expected - and thats the reason why xcb waits for the server
forever?
Do you know if Xorg provides some options to enable debug messages
about bad behaving client?

I know I asked that before in a different way, but do you think the
socket handoff code will be in a stable version at beginning 2010. I
work on OpenJDK, and my xrender-backend where I would like to use
socket handoff will be in version 7, which will be released in early
2010.

Thanks, Clemens