Problems with panel startup

[fraioli dg-rtp dg com (Marc J. Fraioli)]

I'm having trouble getting panel to start after compiling a bunch of
the required Gnome stuff on DG/UX, a SYSV-ish UNIX.  I'm trying to
start it by just invoking "panel" from the shell as a test, which has
brought it up for me under Linux in the past (with X already running,
obviously.  DG/UX has X11R5, and I'm using an old fvwm as the WM
currently.  I think it's version 1.24 or something).  In DG/UX,
however, doing that does nothing-- the program doesn't core dump, but
neither does the panel itself ever appear on the screen.  Running under
the debugger, I get this stack trace:

(debug) walk
frame 0, pc _select+12
frame 1, line 1016, scope \connection\giop_main_next_message_2, pc 0x8011d2b9
frame 2, line 863, scope \giop-msg-buffer\giop_recv_reply_buffer_use_multiple_2, pc 0x8011e62b
frame 3, line 913, scope \giop-msg-buffer\giop_recv_reply_buffer_use_2, pc 0x8011e75a
frame 4, line 696, scope \CosNaming-stubs\CosNaming_NamingContext_resolve, pc 0x8009ebd3
frame 5, line 1006, scope \goad\goad_server_register, pc 0x80183c02
frame 6, line 762, scope \extern\panel_corba_gtk_init, pc 0x80616a7
frame 7, line 114, scope \main\main, pc 0x805145f
frame 8, pc _start+105
(debug) 

This is actually missing a frame (probably because of optimization):
there is also a frame "0.5" which is in giop_check_connections(), in
ORBit/src/IIOP/connection.c.  This is where the select() call actually
comes from.  Looking at the select() call, it appears that it was set
up intentionally to block indefinitely (passes NULL for the timeval
pointer).

So I have a couple questions.  Is there some other component that's
failing that should be writing to one of the descriptors being selected
on, and panel is doing what it should?  Is this code supposed to be
executed from a separate thread, so the main window won't block, but
this isn't happening (I noticed that configure didn't find my thread
library-- you get it with -lthread on DG/UX, but configure only checked
for -lpthread.  I was intending to fix that later, but do I have to
have working threads to get this to work at all?)?  Or, as a slight
variation on my first question, perhaps some important component is
simply missing, rather than failing?  I didn't build all of Gnome.  I
followed the guide at:

http://www.linuxpower.org/display_item.phtml?id=83

And only built the components it described as 'necessary' (plus all the
things it didn't mention but which you need in order to compile those
necessary things, like the Berkeley db, xpm, etc.).  So looking at my
build area, I have the following Gnome components built and installed:
ORBit, glib, gnome-core, gnome-libs, gtk+, gtk-engines, imlib, and mc.
I built this out of CVS over the last week or so.  Did I miss something
critical?

Thanks for any insight,

	Marc