Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!rpi!crdgw1!uunet!mcsun!hp4nl!star.cs.vu.nl!maart
From: maart@cs.vu.nl (Maarten Litmaath)
Newsgroups: comp.unix.questions
Subject: Re: Return values in pipelines
Message-ID: <7378@star.cs.vu.nl>
Date: 24 Aug 90 15:30:48 GMT
References: <1990Aug22.211057.19850@agate.berkeley.edu> <7372@star.cs.vu.nl> <1990Aug23.215612.12422@iwarp.intel.com>
Sender: news@cs.vu.nl
Reply-To: maart@cs.vu.nl (Maarten Litmaath)
Organization: VU Dept. of Computer Science, Amsterdam, The Netherlands
Lines: 36

In article <1990Aug23.215612.12422@iwarp.intel.com>,
	merlyn@iwarp.intel.com (Randal Schwartz) writes:
)In article <7372@star.cs.vu.nl>, maart@cs (Maarten Litmaath) writes:
					^^
					Still broken software, Randal?

)| In article <1990Aug22.211057.19850@agate.berkeley.edu>,
)| )	      How can I detect if something went wrong on the previous
)| )commands in the pipeline (prog2, prog3)?
)| 
)| You can do something like this:
)[program using fd 3 and /bin/sh fd duping deleted]
)
)Hmm.  Similar to what I was going to do, except that I would have used
)temp files (remember... a temp file is just a pipe with an attitude
)and a strong will to live).

Yes, and that's why I _hate_ 'em!  You can always get screwed if someone
else decides to ``rm /tmp/*''...
Furthermore pipes have nice blocking & signaling semantics.

)...
)With temp files, you won't get a race condition if both processes
)decide to exit at *exactly* the same time (rare, but it'll happen!).

There is _no_ race condition with my solution, as each of the write()s
to the pipe is atomic, because the number of bytes is smaller than
4096 (<= PIPESIZE, or whatever it's called).

Only the _order_ in which the various messages arrive is undetermined.

BTW, I expected a Perl solution from you...  :-)
Larry?
--
        "[Your C code] seems about as portable as the Rock of Gibraltar."
                                 (Wayne Throop)