Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!rutgers!seismo!mcvax!diku!thorinn
From: thorinn@diku.UUCP
Newsgroups: comp.bugs.4bsd,comp.unix.wizards
Subject: Re: ex doesn't wait for all its children
Message-ID: <2931@diku.UUCP>
Date: Mon, 19-Jan-87 10:53:53 EST
Article-I.D.: diku.2931
Posted: Mon Jan 19 10:53:53 1987
Date-Received: Sun, 1-Feb-87 16:04:28 EST
References: <10069@tektronix.TEK.COM>
Organization: DIKU, U of Copenhagen, DK
Lines: 17
Keywords: wait zombie
Xref: watmath comp.bugs.4bsd:164 comp.unix.wizards:778

In article <10069@tektronix.TEK.COM> tonyb@tektronix.TEK.COM (Tony Birnseth) writes:
>	The only change that was really necessary was to add a call 
>	to waitfor() after the call to setrupt() if forking to write the 'io' 
>	stream.

>	Of course it's entirely possible that I'm missing something here.  I 
>	think the fix should be done correctly by waiting for each child after
>	it is created.
>
  I think that you missed one important point: If the region written to the
pipe contains more than 4K bytes, the first child won't exit if no one reads
the pipe. So the correct thing would be to remember the pid of the writer and
wait for it *after* the reader exits -- even if the writer didn't get every-
thing into the pipe, it will get a SIGPIPE and die when the pipe is closed.
--
Lars Mathiesen, DIKU, U of Copenhagen, Denmark		..mcvax!diku!thorinn
Institute of Datalogy -- we're scientists, not engineers.