>Number:         539
>Category:       kern
>Synopsis:       UUCP reports "line disconnected" without the line being disconnected
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    gnats-admin (Kernel Bug People)
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sun Oct 30 00:05:01 1994
>Originator:     Martin Husemann
>Organization:
private
>Release:        1.0
>Environment:
	
System: NetBSD euterpe.owl.de 1.0 NetBSD 1.0 (EUTERPE) #1: Wed Oct 26 21:53:49 MET 1994 root@euterpe.owl.de:/usr/src/sys/arch/i386/compile/EUTERPE i386
	Sources sup'ed Oct 26, world and kernel completely rebuild after
	the sup. A 486/66 notebook, no fifo chips, 19200 baud connected
	US Robotics modem, V32bis/V42/V42bis connection to another NetBSD
	system (some older current with sources prior to the first UUCP
	Problems).

>Description:
	When calling a system via UUCP uucico hangs up a few seconds after
	it is connected. It logs 

uucico bi-link - (1994-10-29 12:50:22.80 2888) Calling system bi-link (port tty01)
uucico bi-link - (1994-10-29 12:50:42.24 2888) ERROR: Line disconnected

	which is simply not true. It then disconnects the line itself.

	When trying to debug it further the problem vanishes as soon
	as "-x chat" is activated (that's how I got this mail out).
	
	These are exactly the same symptoms as with the previous UUCP
	bug (read returning 0), but I thought this had been fixed?

>How-To-Repeat:
	Place a uucp call without debug output.

>Fix:
	Two workarounds: a patch to libunix.c had been posted to current-users
	a while ago to treat 0 returned by read() as no error. Or: use
	uucico's "-x chat,handshake" option, which magicaly lets the problem
	disapear.
	
	The real bug is in the kernel, I suppose.
>Audit-Trail:
>Unformatted: