From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <zsh-users-return-18061-mason-zsh=primenet.com.au@zsh.org>
Received: (qmail 8217 invoked by alias); 27 Oct 2013 17:01:40 -0000
Mailing-List: contact zsh-users-help@zsh.org; run by ezmlm
Precedence: bulk
X-No-Archive: yes
List-Id: Zsh Users List <zsh-users.zsh.org>
List-Post: <mailto:zsh-users@zsh.org>
List-Help: <mailto:zsh-users-help@zsh.org>
X-Seq: 18061
Received: (qmail 8589 invoked from network); 27 Oct 2013 17:01:35 -0000
X-Spam-Checker-Version: SpamAssassin 3.3.2 (2011-06-06) on f.primenet.com.au
X-Spam-Level: 
X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00,RCVD_IN_DNSWL_NONE
	autolearn=ham version=3.3.2
From: Bart Schaefer <schaefer@brasslantern.com>
Message-id: <131027100137.ZM4100@torch.brasslantern.com>
Date: Sun, 27 Oct 2013 10:01:37 -0700
In-reply-to: <20131027145917.GA5509@localhost.localdomain>
Comments: In reply to Han Pingtian <hanpt@linux.vnet.ibm.com>
 "multios doesn't work with 2>&1" (Oct 27, 10:59pm)
References: <20131027145917.GA5509@localhost.localdomain>
X-Mailer: OpenZMail Classic (0.9.2 24April2005)
To: zsh-users@zsh.org
Subject: Re: multios doesn't work with 2>&1
MIME-version: 1.0
Content-type: text/plain; charset=us-ascii

On Oct 27, 10:59pm, Han Pingtian wrote:
} 
} According to the zsh guide's "2.5.8 `Go faster' options for power
} users", page 44:
} 
}     % echo foo 2>&1 >/dev/null | sed 's/foo/bar/'
} 
} should output "bar" if multios has been set. But it outputs nothing

Hmm, looks like this has been broken for a LONG time.

If I back out workers/20666 (Jan 2005), then this example works again.

Note that

    % print -u2 foo 2>&1 >/dev/null | sed 's/foo/bar/'

does print "bar", so the problem is that the original fd 1 is directed
to /dev/null while the multio fd created by 2>&1 is is directed to the
pipe.  You can see the reasoning for this in the "references" thread
going back from http://www.zsh.org/mla/workers/2005/msg00026.html

The workaround is to either explicitly add the original stdout to the
multio: 

    % echo foo 2>&1 >&1 >/dev/null | sed 's/foo/bar/'

Or to make sure the multio isn't created until the pipe is seen:

    % { echo foo 2>&1 } >/dev/null | sed 's/foo/bar/'

If there's a way to fix the original complaint (that >&- doesn't really
close the descriptor when a multio is involved, so unnecessary multios
result) without the above example remaining broken, I may need some help
to find it.

Otherwise we have to decide whether to revert to the pre-2005 behavior,
or document around it.