[FreeNX-kNX] Call to bind failed for font TCP port 12000

[Alastair Johnson]

On Wednesday 01 July 2009, Neal Becker wrote:
> Alastair Johnson wrote:
> > On Tuesday 30 June 2009, Neal Becker wrote:
> >> Alastair Johnson wrote:
> >> > On Tuesday 30 June 2009, Neal Becker wrote:
> >> >> server:
> >> >> freenx-server-0.7.3-12.fc11.x86_64
> >> >> nx-3.2.0-32.fc11.x86_64
> >> >>
> >> >> client:
> >> >> nxclient-3.3.0-6.x86_64
> >> >>
> >> >> I can connect to 1 server, but simultaneous connect to 2 servers
> >> >> fails:
> >> >>
> >> >> NXPROXY - Version 3.3.0
> >> >>
> >> >> Copyright (C) 2001, 2007 NoMachine.
> >> >> See http://www.nomachine.com/ for more information.
> >> >>
> >> >> Info: Proxy running in client mode with pid '20092'.
> >> >> Session: Starting session at 'Tue Jun 30 10:26:17 2009'.
> >> >> Info: Connection with remote proxy completed.
> >> >> Warning: Unrecognized session type 'unix-gnome'. Assuming agent
> >> >> session. Error: Call to bind failed for font TCP port 12000. Error is
> >> >> 98 'Address already in use'.
> >> >> Session: Session terminated at 'Tue Jun 30 10:26:17 2009'.
> >> >
> >> > IIRC you need to change the DISPLAY_BASE for one of the servers in
> >> > node.conf. You should increase it by roughly the number of concurrent
> >> > users, say from 1000 to 1030.
> >>
> >> non-free nxserver doesn't have this problem.  Must be a better way.
> >
> > Are you sure the non-free one doesn't have this problem? It used to, and
> > the bug is still marked open.
> >
> > http://www.nomachine.com/tr/view.php?id=TR05E01682
>
> I am pretty sure I've had more than one server simultaneously connected to
> my one client using non-free.  I used all default settings on both client
> and servers.

That's perfectly possible as there is only a problem when you have the same 
session number on both machines. If you have the 5th session on server A and 
the 7th on server B then there's no problem. 

Section 13 of the nomachine admin guide talks about adjustment of a similarly 
named variable for exactly this reason, so it looks like they still have the 
same issue. There may be differences in how session numbers are recycled that 
make a success on the next attempt more likely in nomachine's nxserver than 
freenx I suppose.

The 'better way' would be to extend the protocol to let the client tell the 
server which session numbers were unavailable. It might be possible to include 
this in one of the open clients like qtnx, but the problem would remain with 
the nomachine client.