Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!bu.edu!dartvax!Michael.J.Glenn
From: Michael.J.Glenn@dartmouth.edu (Michael J. Glenn)
Newsgroups: comp.sys.next
Subject: write() problem with Mach 2.0
Summary: A network game of mine exits occasionally on a call to 
 write(), and I can't figure it out.  Help!
Keywords: Mach 2.0, LPMud, write()
Message-ID: <1991Apr5.162455.24305@dartvax.dartmouth.edu>
Date: 5 Apr 91 16:24:55 GMT
Sender: news@dartvax.dartmouth.edu (The News Manager)
Reply-To: Fermat@Dartmouth.edu
Organization: Dartmouth College, Hanover, NH
Lines: 25


I run an LPMud (a network game) on a NeXT running Mach 2.0, and I seem
to have a few problems that I didn't have with Mach 1.0.  The main
problem is, there is a write() statement in the C source code that
occasionally fails (randomly) and causes the program to exit.  What
could cause this?  It seems to happen most often when there is a large
network load (30 to 40 users playing at a time), but I have trapped it
even with a smaller network load.  The line reads

      if ((n = write (ip->socket, buff2 + offset, chunk)) == -1) { ...
}

where the ... checks errno.  This problem never seemed to happen with
Mach 1.0, or on other machines. (I have run on Vax, IBM)  One thing
only I changed when compiling in 2.0, and that is I had to add the
"-bsd" option, otherwise it seemed to have some problem with
inet_ntoa().  The -bsd made it compile (and seem to run) fine.

Does anyone know what could cause this write() to force the program to
exit, and does this -bsd option have anything to do with it?  I'd be
very appreciative of any help/advice.

Fermat@Dartmouth.edu

(admin. Dartmouth LP, telnet wings.dartmouth.edu 2000, 129.170,16.22)