GNU bug report logs - #77462
"/s" instability? I think this is a bug.

Previous Next

Full log

View this message in rfc822 format

From: Jim Meyering <jim <at> meyering.net>
To: "gnudborgonly <at> s-epost.no" <gnudborgonly <at> s-epost.no>
Cc: 77462 <at> debbugs.gnu.org
Subject: bug#77462: "/s" instability? I think this is a bug.
Date: Wed, 2 Apr 2025 13:07:14 -0700

On Wed, Apr 2, 2025 at 8:41 AM gnudborgonly <at> s-epost.no via
<bug-sed <at> gnu.org> wrote:
> This seems to qualify as a bug:

Thanks for the report.
You can fix your usage by not putting "[...]" around those uses of "\s" or "\S".

> The sed version included in my Linux seems to be unstable when using
> the '\s' and/or '\S' regex extensions:
>
> Example:
>
>    id <at> pc:~$ echo '[ subCA2 ]' | sed -n 's|^\[\s\+\([^\s]\+[0-9]\+\)\s\+\]\s*$|\1|p'
>    id <at> pc:~$ echo '[ subCA2 ]' | sed -n 's:^\[\s\+\([\S]\+[0-9]\+\)\s\+\]\s*$:\1:p'

Drop the square brackets and it works. I.e., change the latter to this:

  $ echo '[ subCA2 ]' | sed -n 's:^\[\s\+\(\S\+[0-9]\+\)\s\+\]\s*$:\1:p'
  subCA2

Or better still, use sed's -E option to make the regular expression
more readable, eliding **six** backslashes:

  echo '[ subCA2 ]' | sed -nE 's:^\[\s+(\S+[0-9]+)\s+\]\s*$:\1:p'

I admit this is an unpleasant irregularity about GNU sed's "\S" extension,
since it's different from how things work in PCRE.
This is one of the reasons I urge people use Perl instead of sed
(another is because PCRE lets you use "\d" and non-greedy modifiers
like "\S+?" below):

  $ echo '[ subCA2 ]' | perl -nle 'm{^\[\s+(\S+?\d+)\s+\]\s*$} and print $1'
  subCA2

 Searching Sed's sources/docs for references to \S and \s vs ranges, I
found no trace, but did see this 4.1 NEWS entry:

  * removed documentation for \s and \S which worked incorrectly

I'll leave this bug report open, because this is a wart that needs to
be documented.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.