[comp.mail.uucp] Summary: Terrible uucp problem

rcw@upas.UUCP (Robert White) (05/17/87)

This line == can of raid

In article <94@upas.UUCP> I write:
>
> I'll be happy if one or the other of these problems can be resolved!
>
>1> We are experiencing a rather frustrating problem with uucico on
>a Wicat 160 running Uniplus System V.2. I am trying to set up news
>on this system, but when other systems call up, uucico occasionally
>hangs the port so that the machine must be rebooted. Even a "kill -9
><pid>" has no effect. (Incidentally, user terminals hang periodically
>as well, especially when the line is accidentally disconnected then
>reconnected to the terminal).
>
>2> OK, so if they can't call me, I'll call out. Wrong! Because we must
>use our two phone lines at night, we needed a way to drop getty on the
>phone line ports some time late at night. Customer support at Wicat
>suggest writing a cron task that would edit /etc/inittab then do a 
>telinit q. For some strange reason unknown to me, I get (NO DEVICE)
>listed in the uucp accounting files when the port is supposed to be .
>free. Can anyone help me? Send to the path below, and I'll summarize.
> Robert White

Thank you for your responses. I received many insightful comments.
These are summaries, not reprints.

*Preferred solution* 

John Cornelius writes:
> If Bell didn't supply Unisoft->Wicat->you with uugetty, then I
> suggest you shove the problem back through the pipe. It's part
> of System V.2 Basic Networking Package.
>     Solution 2 should be avoided (editing inittab). Results were
> not satisfactory since not everyone used locks.
> John Cornelius
> (...!sdcvax!piaget!jc)

Good point about the locks. On our small system, we might get 
away with it, but I'd rather not. I've heard that under V.3, 
uucp setup is radically different.

*Bits and pieces, important and not to be forgotten*

Robert (Usenet News at Boulder) writes:
> You should write a script that will disable the port, call
> out, then re-enable the tty port. This script should monitor
> the presence of LCK..sysname so that when it disappears, you
> can execute 'enable tty??' This should solve your problems.
> News Administrator, {whatever}!boulder!cu-den!netnews

This is part of it, but read on:

Many people wrote and mentioned that the permissions on the
modem port should be at least 0666. However, Amos Shapir
writes to give me the solution to my problem:

> It's not enough to disable the line - login chown's it to
> the last user id, and chmod's it to 622; this setup remains
> until the next user logs in. To enable uucp to open it for
> both read/write you should also *** chown it to uucp ***.
> Amos Shapier amos%nsta@nsc.com @{hplabs,pyramid,sun,decwrl}
> 38.48'E 32.10N

Kudos to Jack Jansen, <isis!seismo!mcvax!cwi.nl!jack> for also
providing the correct answer and insight into my first problem
with a hung port. Jack writes:

> One of the things you can do to track the problem down is see
> what happens when you do an stty to that terminal. If
> stty hangs and is also unkillable, there's probably a 
> kernel bug.
> Jack Jansen, seismo!mcvax!cwi.nl!jack

Stty hangs on our system big time, making that terminal
unusable.  Apparently, the problem is even worse than we 
first thought. We use a Calcomp 1075 plotter on one of the 
ports. It uses xon/xoff protocol for data, drops DTR when off. 
If it is turned off while plotting (Say to clear a paper jam), the
port will hang. Haven't yet been able to decode this problem
with the box, but suspect it's the same problem.  Wicat has
finally admitted to the kernel bug, and it'll be fixed "real
soon now." 8-).

One last thing: I received a couple of annoying flames that
didn't help me in the least. If I wanted flames, I'd subscribe
to flamenet. Consider this: if we knew all the answers, and were
all gurus, there wouldn't be the need for a costly network.
If I am wrong, tell me, just don't start out a letter with:
That's a stupid question but here's the answer for you 
anyway, you nonentity.
'Nuff said

Robert C. White, Jr. Graphics Information, Inc.
UUCP: seismo!{nbires!isis|hao}!scicom!qetzal!rcw
USPS: 3067 Robin Way, Denver, CO 80222
ATT : +1 303 759-3666
Disclaimer: I take the 5th on the grounds that
	    what I say may incriminate me.
		-Oliver North, and many others.