Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!sdd.hp.com!caen!uwm.edu!bionet!agate!ucbvax!CADFX.CCAD.UIOWA.EDU!koppes
From: koppes@CADFX.CCAD.UIOWA.EDU (Beth Koppes)
Newsgroups: comp.windows.x.motif
Subject: RSH Problems
Message-ID: <9104161928.AA03622@cadfx.ccad.uiowa.edu>
Date: 16 Apr 91 19:28:58 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: inet
Organization: CAD-Research, University of Iowa, Iowa City, Iowa
Lines: 27


	Platform:  Vax Station 3100
	Motif Version:	1.03a

I have discovered that a call to MrmOpenHierarchy() is interfering with
a call to system(). The Mrm...() call fetches 18 uid files. The
system() call executes rsh.

If I put the system() call before the Mrm...() call, the remote command
executed by rsh completes, the rsh process completes, and system()
returns to my program.

If I put the system() call after the Mrm...() call, the remote command
completes, but the rsh process does not. The result is that system()
does not return. A netstat on the local machine shows two network
connections still present with statuss of CLOSE_WAIT. A netstat on the
remote machine shows statuses of FIN_WAIT_2. It appears that the rsh
process is holding the connection open.

If I press [Return] (thus sending a newline to standard input of my
program) the rsh process completes and system() returns. I don't know
why the [Return] satisfies the rsh process. It does not matter what
remote command is executed -- ls will do it. I eliminated the local
.cshrc and remote .cshrc to ensure that no unexpected reads were coming
from them.

Can anyone suggest a cure for this problem?