Input queue lock broken ==> Fatal Error?

[Loren Buck Buchanan]

Hi,

We have been having a problem with NeWS getting confused to the point
where we have to kill the news_server process (thus killing all proceeses
associated with windows).  We had assumed it was something in the code
we are currently developing that caused this.  Last week, we moved our
machine for a demo, and while at the demo site we disabled all of the
networking code (inetd, named, nfs, etc.) and in 30 runs of our program
we never had to kill the news_server once.

The initial indication that the news_server is confused is when grcond
logs the following (somewhat edited for brevity) to /usr/adm/SYSLOG:

	grcond[980]: CIO: ***ERROR*** 
	grcond[980]: CIO: Process: 0x100223A4   Error: undefined
	grcond[980]: CIO: Stack:
	grcond[980]: CIO: Executing: input
	grcond[980]: CIO: At: Reading file('%stdio',W,R)
	grcond[980]: CIO: *****
	grcond[980]: CIO: ***ERROR*** 
	grcond[980]: CIO: Process: 0x100223A4   Error: undefined
	grcond[980]: CIO: Stack:
	grcond[980]: CIO: Executing: queue
	grcond[980]: CIO: At: Reading file('%stdio',W,R)
	grcond[980]: CIO: *****
	grcond[980]: CIO: ***ERROR*** 
	grcond[980]: CIO: Process: 0x100223A4   Error: undefined
	grcond[980]: CIO: Stack:
	grcond[980]: CIO: Executing: lock
	grcond[980]: CIO: At: Reading file('%stdio',W,R)
	grcond[980]: CIO: *****
	grcond[980]: CIO: ***ERROR*** 
	grcond[980]: CIO: Process: 0x100223A4   Error: undefined
	grcond[980]: CIO: Stack:
	grcond[980]: CIO: Executing: broken!
	grcond[980]: CIO: At: Reading file('%stdio',W,R)
	grcond[980]: CIO: *****

The thing that strikes me odd about this, is that the error

	input queue lock broken

is supposed to be harmless, but somehow it is getting fed to the
news_server as a PostScript program to be run.  It really gets bad
when the error messages grcond is logging are also sent to news_server
to be run.  This sounds like the start of an infinite loop, but 
fortunately it does stop being logged to SYSLOG after a while.  At
that point the only recovery is to kill the news_server.  Killing
grcond at this point is useless, because doing so will require a
press of the "reset" button to recover.

We are using a 4D/20 running IRIX 3.2

Thankx for any help in solving this problem.

B Cing U

Buck

Loren Buchanan     | buck at drax.gsfc.nasa.gov   | #include <std_disclaimer.h> 
CSC, 1100 West St. | ...!ames!dftsrv!drax!buck | typedef int by
Laurel, MD 20707   | (301) 497-2531            | void where_prohibited(by law){}
CD International lists over 40,000 pop music CDs, collect the whole set.