Re: [PATCH] add new sort option --xargs (-x)

[Jim Meyering]

"Bo Borgerson" <[EMAIL PROTECTED]> wrote:
> The number of inputs that can be handled by the sort utility is
> currently limited by what may be passed in argv.
>
> Due to the nature of sort, this limit can't be stepped around with
> `xargs' as it could be with some other utilities.
>
> My solution to this locally has been to add an option to the sort
> utility, --xargs, which causes sort to treat STDIN as a source of
> newline-separated arguments that supplement those on the command-line
> (please see attached patch).

I suppose you have a real application where this is useful?
If so, please describe it -- motivation/justification helps ;-)

> Is this an option that might be worth including in a future release?

I think so.  du and wc each have the --files0-from=F option, added for
the same reason.  Any such option in sort should have the same name and
be implemented in the same way.

[haven't forgotten about --nmerge.  will get to it eventually ]


___
Bug-coreutils mailing list
Bug-coreutils@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-coreutils

Re: [PATCH] add new sort option --xargs (-x)

[Bo Borgerson]

On Thu, Apr 3, 2008 at 12:18 PM, Jim Meyering <[EMAIL PROTECTED]> wrote:
>  I suppose you have a real application where this is useful?
>  If so, please describe it -- motivation/justification helps ;-)
>

Just a merge with a lot of source files.  It's the same motivation as
the nmerge patch.  I've actually got another patch as well that I'll
clean up and offer soon that allows a merge of > nmerge files to be
divided among sub-processes whose output is then merged by the parent,
which provides a performance benefit (if you've got the resources for
it).  I've got yet another patch that adds an option to open
compressed files through a decompression program, so I don't have to
set up fifos for a merge of gzipped files.

I've been maintaining these patches against sort for a while,
re-patching whenever a new release is published.  I figured it would
be worth a shot seeing if I could get any them incorporated into the
package upstream. :)

I'm also just generally interested in helping out with maintenance.  I
think I'm probably less experienced than most of the regular
contributors but I can help with simple stuff and when it comes to
coding I think doing is the best way of learning.

>  I think so.  du and wc each have the --files0-from=F option, added for
>  the same reason.  Any such option in sort should have the same name and
>  be implemented in the same way.

That seems reasonable enough.  Looks like readtokens0 does most of the
work for me. :)

How would you feel about also including a --filesn-from=F option to
support pipelines like the one in my example where the input is
newline separated?

>  [haven't forgotten about --nmerge.  will get to it eventually ]

Thanks. :)

Bo


___
Bug-coreutils mailing list
Bug-coreutils@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-coreutils

Re: [PATCH] add new sort option --xargs (-x)

[Jim Meyering]

"Bo Borgerson" <[EMAIL PROTECTED]> wrote:
>>  I think so.  du and wc each have the --files0-from=F option, added for
>>  the same reason.  Any such option in sort should have the same name and
>>  be implemented in the same way.
>
> That seems reasonable enough.  Looks like readtokens0 does most of the
> work for me. :)
>
> How would you feel about also including a --filesn-from=F option to
> support pipelines like the one in my example where the input is
> newline separated?

I'd rather not.  Instead, just pipe your list through tr '\n' '\0' first.
We had the same discussion back when I added --files0-from=F to du and wc.


___
Bug-coreutils mailing list
Bug-coreutils@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-coreutils

Re: [PATCH] add new sort option --xargs (-x)

[Jim Meyering]

"Bo Borgerson" <[EMAIL PROTECTED]> wrote:
> On Thu, Apr 3, 2008 at 12:18 PM, Jim Meyering <[EMAIL PROTECTED]> wrote:
>>  I suppose you have a real application where this is useful?
>>  If so, please describe it -- motivation/justification helps ;-)
>
> Just a merge with a lot of source files.  It's the same motivation as
> the nmerge patch.  I've actually got another patch as well that I'll
> clean up and offer soon that allows a merge of > nmerge files to be
> divided among sub-processes whose output is then merged by the parent,
> which provides a performance benefit (if you've got the resources for
> it).  I've got yet another patch that adds an option to open
> compressed files through a decompression program, so I don't have to
> set up fifos for a merge of gzipped files.

Sounds interesting.
I suppose it can work with an arbitrary decompressor?

Note this relatively new option:

  --compress-program=PROG  compress temporaries with PROG;
  decompress them with PROG -d


___
Bug-coreutils mailing list
Bug-coreutils@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-coreutils

Re: [PATCH] add new sort option --xargs (-x)

[Bo Borgerson]

On Thu, Apr 3, 2008 at 2:05 PM, Jim Meyering <[EMAIL PROTECTED]> wrote:
 >  Sounds interesting.
 >  I suppose it can work with an arbitrary decompressor?
 >
 >  Note this relatively new option:
 >
 >   --compress-program=PROG  compress temporaries with PROG;
 >   decompress them with PROG -d
 >

 Yep.

 My current convention is:

  --magic-open=PROG[,PROG]...

 So if you want to merge a gzip'd file with a bzip2'd file you can use
 --magic-open=gzip,bzip2 (or just --magic-open, which enables all).

 For each regular file it checks magic and if it looks like a type that
 can be handled by one of PROG it opens a PROG -d -c -f (the -f is just
 in case the magic was a false-positive).

 Of course this re-introduces findprog into sort (for find_in_path),
 which may not be desirable.  A convention more similar to that used
 for --compress-program would eliminate this (and the magic-checking),
 but limit a given merge to files compressed with a single program.

 Bo


___
Bug-coreutils mailing list
Bug-coreutils@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-coreutils

Re: [PATCH] add new sort option --xargs (-x)

[Bo Borgerson]

Okay, here's a version that supports argument input in --files0-from=F style.

Bo
From 3108b79cbbb5d6c2fe3c2f8d5037f166cb0f1ca6 Mon Sep 17 00:00:00 2001
From: Bo Borgerson <[EMAIL PROTECTED]>
Date: Thu, 3 Apr 2008 18:42:57 -0400
Subject: [PATCH] Add new sort option --files0-from=F

src/sort.c: support new option
tests/misc/sort-files0-from: test new option
tests/misc/Makefile.am: indicate new test
docs/coreutils.texti: explain new option
NEWS: advertise new option

Signed-off-by: Bo Borgerson <[EMAIL PROTECTED]>
---
 NEWS|5 ++
 doc/coreutils.texi  |   16 +++
 src/sort.c  |   58 +++-
 tests/misc/Makefile.am  |1 +
 tests/misc/sort-files0-from |  105 +++
 5 files changed, 183 insertions(+), 2 deletions(-)
 create mode 100755 tests/misc/sort-files0-from

diff --git a/NEWS b/NEWS
index e208b30..492c4e9 100644
--- a/NEWS
+++ b/NEWS
@@ -55,6 +55,11 @@ GNU coreutils NEWS-*- outline -*-
   options --general-numeric-sort/-g, --month-sort/-M, --numeric-sort/-n
   and --random-sort/-R, resp.
 
+  sort accepts a new option, --files0-from=F, that specifies a file
+  containing a null-separated list of files to sort.  This list is used
+  instead of filenames passed on the command-line to avoid problems with
+  maximum command-line (argv) length.
+
 ** Improvements
 
   id and groups work around an AFS-related bug whereby those programs
diff --git a/doc/coreutils.texi b/doc/coreutils.texi
index ee7dbb2..5415394 100644
--- a/doc/coreutils.texi
+++ b/doc/coreutils.texi
@@ -3667,6 +3667,22 @@ Terminate with an error if @var{prog} exits with nonzero status.
 Whitespace and the backslash character should not appear in
 @var{prog}; they are reserved for future use.
 
[EMAIL PROTECTED] [EMAIL PROTECTED]
[EMAIL PROTECTED] [EMAIL PROTECTED]
[EMAIL PROTECTED] including files from @command{du}
+Rather than processing files named on the command line, process those
+named in file @var{FILE}; each name is terminated by a null byte.
+This is useful when the list of file names is so long that it may exceed
+a command line length limitation.
+In such cases, running @command{sort} via @command{xargs} is undesirable
+because it splits the list into pieces and gives each piece to a different
+instance of @command{sort}, with the resulting output being multiple sets
+of sorted data concatenated together.
+One way to produce a list of null-byte-terminated file names is with @sc{gnu}
[EMAIL PROTECTED], using its @option{-print0} predicate.
+
+Do not specify any @var{FILE} on the command line when using this option.
+
 @item -k @var{pos1}[,@var{pos2}]
 @itemx [EMAIL PROTECTED],@var{pos2}]
 @opindex -k
diff --git a/src/sort.c b/src/sort.c
index 8b2eec5..8342399 100644
--- a/src/sort.c
+++ b/src/sort.c
@@ -37,6 +37,7 @@
 #include "posixver.h"
 #include "quote.h"
 #include "randread.h"
+#include "readtokens0.h"
 #include "stdio--.h"
 #include "stdlib--.h"
 #include "strnumcmp.h"
@@ -304,8 +305,9 @@ usage (int status)
 {
   printf (_("\
 Usage: %s [OPTION]... [FILE]...\n\
+  or:  %s [OPTION]... --files0-from=F\n\
 "),
-	  program_name);
+	  program_name, program_name);
   fputs (_("\
 Write sorted concatenation of all FILE(s) to standard output.\n\
 \n\
@@ -342,6 +344,9 @@ Other options:\n\
   -C, --check=quiet, --check=silent  like -c, but do not report first bad line\n\
   --compress-program=PROG  compress temporaries with PROG;\n\
   decompress them with PROG -d\n\
+  --files0-from=Fread input from the files specified by\n\
+   NUL-terminated names in file F\n\
+  -L, --max-line-length  print the length of the longest line\n\
   -k, --key=POS1[,POS2] start a key at POS1, end it at POS2 (origin 1)\n\
   -m, --merge   merge already sorted files; do not sort\n\
 "), stdout);
@@ -395,7 +400,8 @@ enum
   CHECK_OPTION = CHAR_MAX + 1,
   COMPRESS_PROGRAM_OPTION,
   RANDOM_SOURCE_OPTION,
-  SORT_OPTION
+  SORT_OPTION,
+  FILES0_FROM_OPTION
 };
 
 static char const short_options[] = "-bcCdfgik:mMno:rRsS:t:T:uy:z";
@@ -407,6 +413,7 @@ static struct option const long_options[] =
   {"compress-program", required_argument, NULL, COMPRESS_PROGRAM_OPTION},
   {"dictionary-order", no_argument, NULL, 'd'},
   {"ignore-case", no_argument, NULL, 'f'},
+  {"files0-from", required_argument, NULL, FILES0_FROM_OPTION},
   {"general-numeric-sort", no_argument, NULL, 'g'},
   {"ignore-nonprinting", no_argument, NULL, 'i'},
   {"key", required_argument, NULL, 'k'},
@@ -2752,6 +2759,8 @@ main (int argc, char **argv)
   bool posixly_correct = (getenv ("POSIXLY_CORRECT") != NULL);
   bool obsolete_usage = (posix2_version () < 200112);
   char **files;
+  char *files_from = NULL;
+  struct Tokens tok;
   char const *outfile = NULL;
 
   initialize_main (&argc, &argv);
@@ -2955,6 +2964,10 @@ main (int argc, char **argv)

Re: [PATCH] add new sort option --xargs (-x)

[Bo Borgerson]

I had a capitalized error message in this patch.
I also didn't use a correct commit message format.

Thanks

Bo
From 9a37b547bcc892d1d5e2542c43d77b13497318db Mon Sep 17 00:00:00 2001
From: Bo Borgerson <[EMAIL PROTECTED]>
Date: Thu, 3 Apr 2008 18:42:57 -0400
Subject: [PATCH] Add new sort option --files0-from=F

* src/sort.c: support new option
* tests/misc/sort-files0-from: test new option
* tests/misc/Makefile.am: indicate new test
* docs/coreutils.texti: explain new option
* NEWS: advertise new option

Signed-off-by: Bo Borgerson <[EMAIL PROTECTED]>
---
 NEWS|5 ++
 doc/coreutils.texi  |   16 +++
 src/sort.c  |   57 ++-
 tests/misc/Makefile.am  |1 +
 tests/misc/sort-files0-from |  105 +++
 5 files changed, 182 insertions(+), 2 deletions(-)
 create mode 100755 tests/misc/sort-files0-from

diff --git a/NEWS b/NEWS
index e208b30..492c4e9 100644
--- a/NEWS
+++ b/NEWS
@@ -55,6 +55,11 @@ GNU coreutils NEWS-*- outline -*-
   options --general-numeric-sort/-g, --month-sort/-M, --numeric-sort/-n
   and --random-sort/-R, resp.
 
+  sort accepts a new option, --files0-from=F, that specifies a file
+  containing a null-separated list of files to sort.  This list is used
+  instead of filenames passed on the command-line to avoid problems with
+  maximum command-line (argv) length.
+
 ** Improvements
 
   id and groups work around an AFS-related bug whereby those programs
diff --git a/doc/coreutils.texi b/doc/coreutils.texi
index ee7dbb2..5415394 100644
--- a/doc/coreutils.texi
+++ b/doc/coreutils.texi
@@ -3667,6 +3667,22 @@ Terminate with an error if @var{prog} exits with nonzero status.
 Whitespace and the backslash character should not appear in
 @var{prog}; they are reserved for future use.
 
[EMAIL PROTECTED] [EMAIL PROTECTED]
[EMAIL PROTECTED] [EMAIL PROTECTED]
[EMAIL PROTECTED] including files from @command{du}
+Rather than processing files named on the command line, process those
+named in file @var{FILE}; each name is terminated by a null byte.
+This is useful when the list of file names is so long that it may exceed
+a command line length limitation.
+In such cases, running @command{sort} via @command{xargs} is undesirable
+because it splits the list into pieces and gives each piece to a different
+instance of @command{sort}, with the resulting output being multiple sets
+of sorted data concatenated together.
+One way to produce a list of null-byte-terminated file names is with @sc{gnu}
[EMAIL PROTECTED], using its @option{-print0} predicate.
+
+Do not specify any @var{FILE} on the command line when using this option.
+
 @item -k @var{pos1}[,@var{pos2}]
 @itemx [EMAIL PROTECTED],@var{pos2}]
 @opindex -k
diff --git a/src/sort.c b/src/sort.c
index 8b2eec5..c14a8d3 100644
--- a/src/sort.c
+++ b/src/sort.c
@@ -37,6 +37,7 @@
 #include "posixver.h"
 #include "quote.h"
 #include "randread.h"
+#include "readtokens0.h"
 #include "stdio--.h"
 #include "stdlib--.h"
 #include "strnumcmp.h"
@@ -304,8 +305,9 @@ usage (int status)
 {
   printf (_("\
 Usage: %s [OPTION]... [FILE]...\n\
+  or:  %s [OPTION]... --files0-from=F\n\
 "),
-	  program_name);
+	  program_name, program_name);
   fputs (_("\
 Write sorted concatenation of all FILE(s) to standard output.\n\
 \n\
@@ -342,6 +344,8 @@ Other options:\n\
   -C, --check=quiet, --check=silent  like -c, but do not report first bad line\n\
   --compress-program=PROG  compress temporaries with PROG;\n\
   decompress them with PROG -d\n\
+  --files0-from=F   read input from the files specified by\n\
+NUL-terminated names in file F\n\
   -k, --key=POS1[,POS2] start a key at POS1, end it at POS2 (origin 1)\n\
   -m, --merge   merge already sorted files; do not sort\n\
 "), stdout);
@@ -395,7 +399,8 @@ enum
   CHECK_OPTION = CHAR_MAX + 1,
   COMPRESS_PROGRAM_OPTION,
   RANDOM_SOURCE_OPTION,
-  SORT_OPTION
+  SORT_OPTION,
+  FILES0_FROM_OPTION
 };
 
 static char const short_options[] = "-bcCdfgik:mMno:rRsS:t:T:uy:z";
@@ -407,6 +412,7 @@ static struct option const long_options[] =
   {"compress-program", required_argument, NULL, COMPRESS_PROGRAM_OPTION},
   {"dictionary-order", no_argument, NULL, 'd'},
   {"ignore-case", no_argument, NULL, 'f'},
+  {"files0-from", required_argument, NULL, FILES0_FROM_OPTION},
   {"general-numeric-sort", no_argument, NULL, 'g'},
   {"ignore-nonprinting", no_argument, NULL, 'i'},
   {"key", required_argument, NULL, 'k'},
@@ -2752,6 +2758,8 @@ main (int argc, char **argv)
   bool posixly_correct = (getenv ("POSIXLY_CORRECT") != NULL);
   bool obsolete_usage = (posix2_version () < 200112);
   char **files;
+  char *files_from = NULL;
+  struct Tokens tok;
   char const *outfile = NULL;
 
   initialize_main (&argc, &argv);
@@ -2955,6 +2963,10 @@ main (int argc, char **argv)
 	  compress_program = 

Re: [PATCH] add new sort option --xargs (-x)

[Jim Meyering]

"Bo Borgerson" <[EMAIL PROTECTED]> wrote:
> I had a capitalized error message in this patch.
> I also didn't use a correct commit message format.

Thanks for noticing and correcting.

> Subject: [PATCH] Add new sort option --files0-from=F
>
> * src/sort.c: support new option
> * tests/misc/sort-files0-from: test new option
> * tests/misc/Makefile.am: indicate new test
> * docs/coreutils.texti: explain new option

s/texti/texi/

> * NEWS: advertise new option

Please use capitals and periods in ChangeLogs. ;-)
Follow existing style -- there are plenty of examples.

> diff --git a/NEWS b/NEWS
> index e208b30..492c4e9 100644
> --- a/NEWS
> +++ b/NEWS
> @@ -55,6 +55,11 @@ GNU coreutils NEWS-*- 
> outline -*-
>options --general-numeric-sort/-g, --month-sort/-M, --numeric-sort/-n
>and --random-sort/-R, resp.
>
> +  sort accepts a new option, --files0-from=F, that specifies a file
> +  containing a null-separated list of files to sort.  This list is used

s/null/NUL/

> +  instead of filenames passed on the command-line to avoid problems with
> +  maximum command-line (argv) length.
> +
>  ** Improvements
>
>id and groups work around an AFS-related bug whereby those programs
> diff --git a/doc/coreutils.texi b/doc/coreutils.texi
> index ee7dbb2..5415394 100644
> --- a/doc/coreutils.texi
> +++ b/doc/coreutils.texi
> @@ -3667,6 +3667,22 @@ Terminate with an error if @var{prog} exits with 
> nonzero status.
>  Whitespace and the backslash character should not appear in
>  @var{prog}; they are reserved for future use.
>
> [EMAIL PROTECTED] [EMAIL PROTECTED]
> [EMAIL PROTECTED] [EMAIL PROTECTED]
> [EMAIL PROTECTED] including files from @command{du}

s/du/sort/

If this text is verbatim or nearly identical to that for
wc and/or du, please see if you can use a macro to factor it out.
Hmm... I went to check and spotted the same error (use of `du' in
wc's section) so went ahead and factored out the duplication.

Now, your doc change will be to add this line:

@files0fromOption{sort,}

...
>--compress-program=PROG  compress temporaries with PROG;\n\
>decompress them with PROG -d\n\
> +  --files0-from=F   read input from the files specified by\n\
> +NUL-terminated names in file F\n\

Split the string.  Otherwise, your addition pushes its length beyond
a portability limit whose exact number I forget but it's around 500.

>-k, --key=POS1[,POS2] start a key at POS1, end it at POS2 (origin 1)\n\
>-m, --merge   merge already sorted files; do not sort\n\
>  "), stdout);
> @@ -395,7 +399,8 @@ enum
>CHECK_OPTION = CHAR_MAX + 1,
>COMPRESS_PROGRAM_OPTION,
>RANDOM_SOURCE_OPTION,
> -  SORT_OPTION
> +  SORT_OPTION,
> +  FILES0_FROM_OPTION

No big deal, but it's good practice to alphabetize.

...
> diff --git a/tests/misc/sort-files0-from b/tests/misc/sort-files0-from
> new file mode 100755
> index 000..a96ab1a
> --- /dev/null
> +++ b/tests/misc/sort-files0-from
> @@ -0,0 +1,105 @@
> +#!/bin/sh
> +# Test "sort --files0-from=F".
> +
> +# Copyright (C) 2002, 2003, 2005-2008 Free Software Foundation, Inc.

This should have only 1 year number: 2008.

If the file is based on some other, please indicate that.
That will help me as reviewer, and future maintainers.

E.g., I put this comment in the wc test of --files0-from:

  # This file bears a striking resemblance to tests/du/files0-from.


___
Bug-coreutils mailing list
Bug-coreutils@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-coreutils

Re: [PATCH] add new sort option --xargs (-x)

[Bo Borgerson]

On Sun, Apr 6, 2008 at 4:30 PM, Jim Meyering <[EMAIL PROTECTED]> wrote:
>  s/texti/texi/
>  Please use capitals and periods in ChangeLogs. ;-)
>  s/null/NUL/
>  Split the string.  Otherwise, your addition pushes its length beyond
>  a portability limit whose exact number I forget but it's around 500.
>  No big deal, but it's good practice to alphabetize.
>  This should have only 1 year number: 2008.

Thanks.  I'll try to catch this sort of thing myself in the future.


>  Now, your doc change will be to add this line:
>
>  @files0fromOption{sort,}

I added an argument to the macro that specifies output for sub-lists,
since it's 'a total' for wc and du, but 'sorted output' for sort.


>  If the file is based on some other, please indicate that.
>  That will help me as reviewer, and future maintainers.
>
>  E.g., I put this comment in the wc test of --files0-from:
>
>   # This file bears a striking resemblance to tests/du/files0-from.

Unfortunately I didn't just copy one of the relevant test files.  If
it would be easier for maintenance to have a more direct copy I can
redo it.  I added a line at the top indicating that this test script
covers a lot of the same ground as the wc-files0-from tests.

Thanks,

Bo
From 404e23daf6874e4d36e2048de569bcac057b7400 Mon Sep 17 00:00:00 2001
From: Bo Borgerson <[EMAIL PROTECTED]>
Date: Thu, 3 Apr 2008 18:42:57 -0400
Subject: [PATCH] Add new sort option --files0-from=F

* src/sort.c: Support new option.
* tests/misc/sort-files0-from: Test new option.
* tests/misc/Makefile.am: Indicate new test.
* docs/coreutils.texi: Explain new option.
* NEWS: Advertise new option.

Signed-off-by: Bo Borgerson <[EMAIL PROTECTED]>
---
 NEWS|5 ++
 doc/coreutils.texi  |   12 +++--
 src/sort.c  |   65 ---
 tests/misc/Makefile.am  |1 +
 tests/misc/sort-files0-from |  106 +++
 5 files changed, 178 insertions(+), 11 deletions(-)
 create mode 100755 tests/misc/sort-files0-from

diff --git a/NEWS b/NEWS
index e208b30..492c4e9 100644
--- a/NEWS
+++ b/NEWS
@@ -55,6 +55,11 @@ GNU coreutils NEWS-*- outline -*-
   options --general-numeric-sort/-g, --month-sort/-M, --numeric-sort/-n
   and --random-sort/-R, resp.
 
+  sort accepts a new option, --files0-from=F, that specifies a file
+  containing a null-separated list of files to sort.  This list is used
+  instead of filenames passed on the command-line to avoid problems with
+  maximum command-line (argv) length.
+
 ** Improvements
 
   id and groups work around an AFS-related bug whereby those programs
diff --git a/doc/coreutils.texi b/doc/coreutils.texi
index 5a6f2c3..9ac7bbf 100644
--- a/doc/coreutils.texi
+++ b/doc/coreutils.texi
@@ -3074,7 +3074,7 @@ Print only the newline counts.
 @opindex --max-line-length
 Print only the maximum line lengths.
 
[EMAIL PROTECTED] files0fromOption{cmd,withTotalOption}
[EMAIL PROTECTED] files0fromOption{cmd,withTotalOption,subListOutput}
 @itemx [EMAIL PROTECTED]
 @opindex [EMAIL PROTECTED]
 @cindex including files from @command{\cmd\}
@@ -3084,13 +3084,13 @@ This is useful \withTotalOption\
 when the list of file names is so long that it may exceed a command line
 length limitation.
 In such cases, running @command{\cmd\} via @command{xargs} is undesirable
-because it splits the list into pieces and makes @command{\cmd\} print a
-total for each sublist rather than for the entire list.
+because it splits the list into pieces and makes @command{\cmd\} print
+\subListOutput\ for each sublist rather than for the entire list.
 One way to produce a list of null-byte-terminated file names is with @sc{gnu}
 @command{find}, using its @option{-print0} predicate.
 Do not specify any @var{FILE} on the command line when using this option.
 @end macro
[EMAIL PROTECTED],}
[EMAIL PROTECTED],,a total}
 
 For example, to find the length of the longest line in any @file{.c} or
 @file{.h} file in the current hierarchy, do this:
@@ -3670,6 +3670,8 @@ Terminate with an error if @var{prog} exits with nonzero status.
 Whitespace and the backslash character should not appear in
 @var{prog}; they are reserved for future use.
 
[EMAIL PROTECTED],,sorted output}
+
 @item -k @var{pos1}[,@var{pos2}]
 @itemx [EMAIL PROTECTED],@var{pos2}]
 @opindex -k
@@ -9757,7 +9759,7 @@ Does not affect other symbolic links.  This is helpful for finding
 out the disk usage of directories, such as @file{/usr/tmp}, which
 are often symbolic links.
 
[EMAIL PROTECTED], with the @option{--total} (@option{-c}) option}
[EMAIL PROTECTED], with the @option{--total} (@option{-c}) option,a total}
 
 @optHumanReadable
 
diff --git a/src/sort.c b/src/sort.c
index 8b2eec5..e67ce80 100644
--- a/src/sort.c
+++ b/src/sort.c
@@ -37,6 +37,7 @@
 #include "posixver.h"
 #include "quote.h"
 #include "randread.h"
+#include "readtokens0.h"
 #include "stdio--.h"
 #include "stdlib--.h"
 #include "strnumcmp.h"
@@ -304,8 +305