SUMMARY: <defunct> processes

From: Richard F. Sidley, GCRC Sys Mgr (sidley@med.unc.edu)
Date: Tue Aug 06 1991 - 10:20:44 CDT


Received only one response with a solution -- but you only need one if it
works! Thanks to dlg@skydesign.arc.nasa.gov.

*********************
Here is the response:

We have noticed the same sort of zombie behavior. The answer is the way the
rsh's to start clients are handled. The local side hangs around, in the
zombie state until (either or both) stdin and stdout are closed, normally
when the client quits. We have found two answers.

1. append the string " >/dev/null </dev/null " to the end of the client
invocations. I even have the following in my .cshrc:

        set C='</dev/null >&/dev/null '

and then my client invocation lines look like:

        rsh somehost -n xclient ... $C &

2. The solution I NOW use is the utility xrsh, which affirmatively closes
stdin and stdout causing the local side of the rsh to exit, hence no
zombies. I have forgotten where the sources came from, but uunet might be a
good start as a place to look.

--> Found the Shell Script in the anonymous ftp directory 'contrib' at
    export.lcs.mit.edu

********************
My original message:

After installing the MIT X11R4 system and the 18 fixes which came with it, I
have started experiencing <defunct> processes. This hasn't occurred be
fore. I can't pin it down to any abnormal abort either. Any ideas?

        Config: Sun SPARCstation 1
                SunOS 4.1 (no patches)
                MIT X11R4 (18 patches)
                    running xdm which is started in rc.local



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:20 CDT