According to Hans de Hartog:
>
>xdpyinfo says:
>

You have MIT-SHM.... good.


>ipcs tells me:
>

that you have ipcs in the kernel.  I assume the shared memory segments
here are xmms dying and not cleaning up properly.

>
>My kernel config contains:
>

Looks good too.

>
>Would it help to increase the default 1024 for SHMMAXPGS?
>

I believe that 1024 pages translates into about 4Meg of shared
memory.  Unless you are using more I doubt if it would help.

>
>I'm not sure about this. Changed my DISPLAY to :0.0 and every-
>thing worked (except for xmms, same message and symptoms).

Well, if you are sitting at the console of the machine your are
running xmms on then it is a pretty safe bet that shared memory should
work.  Believe it or not I have seen someone complain about not being
able to use the shared memory extension when they were remote
displaying a x session to another xserver.

>I guess (based on ipfstat-output) that all my X-traffic is
>via lo0 and not using shmem:
>

That is normal.  Shared memory is used as a speed hack which works by
the x server sharing a bit of memory with the x application, the app
updates the memory with what it wants on the screen and the x server
puts it there. This bypasses the normal x protocol traffic path which
is network based.


You may have to start digging into the xmms source and see what is
going on.  Try looking for the error message in the code and see what
could be causing it to produce that message.

-- 
===============================================================================
Brett Lymn, Computer Systems Administrator, BAE SYSTEMS
===============================================================================