Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!rutgers!cs.utexas.edu!uunet!mcvax!ukc!cam-cl!tmal
From: tmal@cl.cam.ac.uk (Mark Lomas)
Newsgroups: comp.sources.bugs
Subject: Re: inconsistent behavior of sed
Summary: This behaviour is deliberate
Keywords: substitute delete botched confusion
Message-ID: <839@scaup.cl.cam.ac.uk>
Date: 12 Jul 89 12:39:25 GMT
References: <377@proexam.UUCP>
Sender: news@cl.cam.ac.uk
Reply-To: tmal@cl.cam.ac.uk (Mark Lomas)
Organization: U of Cambridge Comp Lab, UK
Lines: 21

In article <377@proexam.UUCP> glen@proexam.UUCP (Glen Brydon) writes:
>	sed 's/.*/&/p;d'  (this is a minimal condensation)
>
>It seems obvious to me that this should simply pass all information,
>however, some versions of sed actually delete. This seems to be always
>true of sysV sed and recently true of bsd sed (post 4.0 sunOS).

The manual entry for sed which is supplied with HP-UX, a System V
derivative, indicates that this change in behaviour is deliberate.

In previous versions of sed the `p' flag at the end of a substitute
command caused the result of the substitution to be printed.  In the
current version the state of the `-n' switch is also checked.  Only
if output is disabled by default will the `p' flag have any effect.

This means that the example which you gave may be corrected by adding
the `-n' switch to the command line:

	sed -n 's/.*/&/p;d'

		Mark Lomas (tmal@cl.cam.ac.uk)