Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!columbia!rutgers!clyde!cuae2!ihnp4!inuxc!pur-ee!uiucdcs!uiucuxc!roma
From: roma@uiucuxc.CSO.UIUC.EDU
Newsgroups: comp.unix.questions
Subject: Re: Query on TCP/IP, SIGPIPE, ECONNRESE
Message-ID: <176200002@uiucuxc>
Date: Mon, 17-Nov-86 23:48:00 EST
Article-I.D.: uiucuxc.176200002
Posted: Mon Nov 17 23:48:00 1986
Date-Received: Wed, 19-Nov-86 06:20:57 EST
References: <231@sbcs.UUCP>
Lines: 26
Nf-ID: #R:sbcs.UUCP:231:uiucuxc:176200002:000:1136
Nf-From: uiucuxc.CSO.UIUC.EDU!roma    Nov 17 22:48:00 1986


> /* Written  7:04 pm  Nov 14, 1986 by sen@sbcs.UUCP in */
> /* ---------- "Query on TCP/IP, SIGPIPE, ECONNRESE" ---------- */
> 
> Instead, if I kill P2, P1 remains in loop, recv succeeds everytime i.e.
> number of bytes received is not less than 0. In fact, it *is* zero. P1
> stays alive. Qn: Why?
> 
> I know that the SIGPIPE is supposed to be raised at the
> sender process if it tries to send msg on a broken connection. How will
> the reciver ever know about the sender's death? By checking number of
> bytes received = 0?

I've had similar questions while working on a TCP/IP project.  From what I can
tell, it's rather easy to check for termination by looping on a read() or recv()
as long as bytes > 0.  However, I *WANT* the receiver to be informed of a
sender's untimely death, as opposed to when he's simply finished sending data.
Do I have to do something with out-of-band data to signal the receiver that
the sender died?

Jon Roma
Computing Services Office, University of Illinois at Urbana-Champaign

ARPANET: roma%uiucuxc@a.cs.uiuc.edu
CSNET:	 roma%uiucuxc@uiuc.csnet
UUCP:	 {ihnp4,pur-ee,convex}!uiucdcs!uiucuxc!roma