Aron Griffis <[EMAIL PROTECTED]> wrote:
> Additionally, I noted that the Single UNIX Specification does not make
> application recommendations for tail, although it refers to the older
> syntax as obsolescent.

SUSv2 is pretty old.
POSIX 1003.1-2001 is more relevant these days.
Usage like `tail -1', `head +2', `sort +1' has been classified as
obsolescent for many years. Recently, it has finally become
officially `obsolete'.

A little background:
A long time ago, POSIX decided that certain types of
options (like `-1' +6, etc.) should eventually be considered invalid.
That's why people have been moving away from uses of e.g., `sort +N' to
`sort -kM,M'.  Less well known is the fact that commands like these
  head -1
  tail +2
  tail -4
  uniq -3
(and more)
are all technically obsolete, now.
And if you build the GNU coreutils on a system that conforms to
POSIX 1003.1-2001, then you'll find that the above all fail, e.g. like this:

  $ head -1
  head: `-1' option is obsolete; use `-n 1'
  Try `head --help' for more information.
  [Exit 1]

>From the coreutils documentation:

  For example, if you are running older software that
  assumes an older version of POSIX and uses `sort +1', `head -1', or
  `tail +1', you can work around the compatibility problems by setting
  `_POSIX2_VERSION=199209' in your environment.

> The pain I'm seeing in the Gentoo Linux distribution, along with the
> above research, leads me to question the rationale behind removing the
> -number syntax in the GNU coreutils.  Personally, I don't see a
> motivation for their removal, since the older syntax is unambiguous and
> in heavy use by thousands of UNIX scripts already in existence.  It

Short answer: it's required for conformance to POSIX 1003.1-2001.
See the `Standards Conformance' section in the coreutils documentation.
E.g., run `info coreutils standards'.

Since you're encountering this problem, it's almost certainly the result
of your building coreutils with a C library and headers that indicate the
system conforms to POSIX 1003.1-2001 by virtue of the definition of
_POSIX2_VERSION in <unistd.h>.

> doesn't seem fruitful for the Gentoo developers (along with developers
> of other distributions) to spend our time making this trivial change in
> hundreds of packages, and trying to push the changes upstream.

You may find solace in the fact that you're not alone :-)
All distributions that care about standards conformance are (or will be)
facing the same battle.  Some began making such changes a few months ago,
so you may have less work than you think.

> Would the coreutils maintainers consider reinstating the older syntax to
> save us the unnecessary toil?  If that is not considered an option,
> would you mind explaining the rationale behind making this change to the
> head/tail programs?  I checked the coreutils FAQ but didn't find
> coverage of this topic.

I suggest you do something like what Debian has done:

Give yourselves a little breathing room by making it so coreutils'
lib/posixver.c is compiled with _POSIX2_VERSION #defined to a value
so that it doesn't cause all that trouble just yet.
Then, try to convince gentoo/upstream developers to set
_POSIX2_VERSION=200112 in their environment and to fix any
problems that are exposed.

Bug-coreutils mailing list

Reply via email to