NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/comp/unix/internal/1952

Text File | 1992-11-12 | 1.6 KB | 37 lines

Newsgroups: comp.unix.internals
Path: sparky!uunet!ferkel.ucsb.edu!taco!rock!stanford.edu!bcm!cs.utexas.edu!asuvax!ennews!anasaz!duane
From: [e-mail] (Duane Morse)
Subject: setsid and streams/device driver output to controlling terminal
Organization: Anasazi, Inc.  Phoenix, Arizona USA
Date: Thu, 12 Nov 1992 18:15:09 GMT
Message-ID: <[e-mail]>
Sender: [e-mail] (Usenet News)
Lines: 26

On an NCR 3433, a program of mine which uses TCP/IP occasionally gets
an "error 22" from t_open.  Running a system call trace (truss) shows that
the errno error is 5 (EIO) upon return from ioctl(fd, I_FIND, "timod")
(looking for the Transport Interface streams module, I presume).

An NCR rep has said that my program needs to use "setsid" because the
streams driver is checking to see if it can put a message on the controlling
terminal and it returns this error when it finds it cannot (the
program and others in the same process group are running "nohup"
and the initiator has logged off).

Though invoking "setsid" during program initialization does seem to get
rid of the error return, I am reluctant to make it a permanent feature
of the software because (1) I don't know what sort of error the
ioctl is encountering, (2) I don't really understand how setsid()
is fixing the problem, and (3) since I don't understand what is going on,
I am concerned that there may be other situations with other pieces
of software in which I should be using setsid to avoid a problem which
might arise.

Any comments or clarifications?

--

Duane Morse    e-mail: [e-mail]
[phone]