On Thu, Nov 08, 2001 at 12:04:00PM +0100, Volker Borchert wrote:
> 
> Hi,
> 
> <INTRO type="optional">
> I am currently consolidating two Sparc-2 file/boot ervers running
> SunOS-4.1.4 and OpenBSD-2.3 onto one Sparc-10 fileserver running
> NetBSD-1.5.1. So far I have migrated rarp/bootparam and about half
> the file systems without problems. The last action was to move the
> file system that contains all clients' /var/tmp, and this was when
> the following problem came up:
> </INTRO>
> 
> A diskless SunOS-4.1.4 client mounts serveral file systems, including
> /var/tmp, from a NetBSD-1.5.1 server. /tmp is a symlink to /var/tmp.
> The X server starts and succeeds to establish the unix domain socket
> for display :0, /tmp/.X11-unix/X0. But any clients that try to
> connect(<fd>, "\0\1/tmp/.X11-unix/X0.", 19) get an ENOTSOCK error.
> 
> This used to work while /var/tmp was mounted from the SunOS-4.1.4
> server.
> 
> With most X apps, I could work around this by using hostname:0.0 as
> $DISPLAY, thus using the X server's inet socket. But xinit insists
> on using the unix socket. This makes it nearly impossible to get an
> X session up.
> 
> What am I missing?

All SunOS boxes share the same /var/tmp, also with the NetBSD server ?
If so, all of them want to create a /tmp/.X11-unix/X0 socket, but
NetBSD and SunOS don't do it the same way I suspect.
So you need a separate /var/tmp for the clients.

--
Manuel Bouyer <bouyer@antioche.eu.org>
--