bret@codonics.COM (Bret Orsburn) (06/30/90)
In article <632@aio.jesnet.jsc.nasa.gov> poirot@aio.jesnet.jsc.nasa.gov (Daniel Poirot) writes: >Howdy folks, > >I have been trying to get Thomas Roell's recently published X11R4 >patches to work under Bell Technologies SysVR3.2u. I have the Lachman >TCP/IP package that I would like to use to get network clients >running. I'm running ISC 386/ix 2.0.2 with Lachman TCP. As far as I can tell, Roell's code does not support TCP transport. He has built a nice STREAMS pipes mechanism for local connections, but his changes seem to make it more difficult than before to bring up TLI STREAMS TCP. (This used to be possible with a few header changes, -DSTREAMSCONN, and by running the att-nameserver code). (Anybody have any better luck with this?) I finally bit the bullet and hacked the osx layer to use the 386/ix socket emulation library. (Sorry, I backed up to pre-Roell osx source, so no local streams connections.) Now I can run remote clients, but not local. :-( The connection is very bursty: large cat's to xterm and ico's run a couple seconds, then pause a couple seconds, then run a couple seconds. Anybody know offhand where the bottleneck is? I found that shared libs were a bad idea *in the server*. They introduce a lot of problems (gdb single-stepping breaks) and save almost nothing (Xlib is linked to get about four functions). BTW: If you are failing in InitOutput, make sure you have read permissions on /dev/console. Be aware that a lot of this code continues after (fatal) errors. Watch the console messages. Some of them are fatal error messages. Also -- get gdb. I got version 3.5 and it came up very easily. Max out your MAXUMEM kernel parameter or you will get (v)fork errors when debugging the server. -- ------------------- bret@codonics.com uunet!codonics!bret Bret Orsburn