Message-Id: <200703110020.l2B0Khlh005758@tigris.pounder.sol.net>
From: cygzw AT trodman DOT com (Tom Rodman)
Reply-to: cygwin AT cygwin DOT com
To: cygwin AT cygwin DOT com
Subject: Re: bash process substitution "<(list)"
In-reply-to: <45F3376B.2080507@byu.net>
References: <200703101514 DOT l2AFEkOW002199 AT tigris DOT pounder DOT sol DOT net> <45F3376B DOT 2080507 AT byu DOT net>
Date: Sat, 10 Mar 2007 18:20:43 -0600
Mailing-List: contact cygwin-help AT cygwin DOT com; run by ezmlm
Precedence: bulk
Sender: cygwin-owner AT cygwin DOT com
Mail-Followup-To: cygwin AT cygwin DOT com

On Sat 3/10/07 15:55 MST Eric Blake wrote:
> According to Tom Rodman on 3/10/2007 8:14 AM:
> >   ~ $ /tmp/foo <(echo ABC)
> >   + test -s /proc/self/fd/63
> >   + echo 1
> > 
> > The 'test -s "$file"' in the test run above should return 0.
> 
> How do you figure?  pipes are special file types, and st_size is
> unspecified in a stat() call on a pipe, so the -s test is unreliable in
> the first place.  Not to mention there is an inherent race in your script
> (did the 'echo ABC' process run before or after the 'test -s
> /proc/self/fd/63' builtin of the script?).  I maintain that there is no
> bug except in your script.

Coming at this from a strictly end user view point, not having
read the "process substitution" bash man page section carefully I
wrongly assumed the <(list) mechanism in bash would support
the test -s.  Thanks for straightening me out.  So the key issue
is that we're dealing w/named pipes, not files.

--
*thanks*,
Tom

--
PS

It's interesting that the "test -s" works in the linux I have for
the "full pipe" case, and for this "empty pipe" case:

  ~ $ cat /tmp/foo
  #!/bin/bash
  file=$1
  TMPF=$(mktemp /tmp/XXXXXX)

  set -x
  test -s "$file" || echo $?

  cp $file $TMPF
  cat $TMPF

  ~ $  /tmp/foo <(true)
  + test -s /dev/fd/63
  + echo 1
  1
  + cp /dev/fd/63 /tmp/WJvtLk
  + cat /tmp/WJvtLk
  ~ $ date;uname -a
  Sat Mar 10 17:53:22 CST 2007
  Linux myhost.blah.nil 2.4.18-14 #1 Wed Sep 4 13:35:50 EDT 2002 i686 i686 i386 GNU/Linux

I did not run any of these test cases more than two or three
times, maybe after repeated tries, that race condition would
come into play.

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/