coreutils: Add various bug fixes

Add a number of bug fixes, mostly imported from Fedora and Wind River
Linux.

cp-i-u: fix unnecessary prompting
fix-install: Fix installing to a dangling symlink
i18n: li18nux/lsb compliance
ls-x: Fix incorrect output
overflow: Fix potential overflow in who command

Signed-off-by: Mark Hatle <mark.hatle@windriver.com>

6 files changed, 4403 insertions, 0 deletions

diff --git a/meta/recipes-core/coreutils/coreutils-6.9/coreutils-6.9-cp-i-u.patch b/meta/recipes-core/coreutils/coreutils-6.9/coreutils-6.9-cp-i-u.patch
new file mode 100644
index 0000000000..6fec683bc3
--- /dev/null
+++ b/meta/recipes-core/coreutils/coreutils-6.9/coreutils-6.9-cp-i-u.patch
@@ -0,0 +1,118 @@
+This patch was imported from the Fedora Core 8 coreutils-6.9-9 package.
+
+The package is stated as being Licensed as GPLv2+.
+
+Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
+
+----
+
+When "cp -i --update old new" would do nothing because "new" is
+newer than "old", cp would nonetheless prompt for whether it is
+ok to overwrite "new".  Then, regardless of the response (because
+of the --update option), cp would do nothing.
+
+The following patch eliminates the unnecessary prompt in that case.
+
+diff --git a/src/copy.c b/src/copy.c
+index b7bf73b..0e549d2 100644
+--- a/src/copy.c
++++ b/src/copy.c
+@@ -1210,6 +1210,30 @@ copy_internal (char const *src_name, char const *dst_name,
+ 	      return false;
+ 	    }
+
++	  if (!S_ISDIR (src_mode) && x->update)
++	    {
++	      /* When preserving time stamps (but not moving within a file
++		 system), don't worry if the destination time stamp is
++		 less than the source merely because of time stamp
++		 truncation.  */
++	      int options = ((x->preserve_timestamps
++			      && ! (x->move_mode
++				    && dst_sb.st_dev == src_sb.st_dev))
++			     ? UTIMECMP_TRUNCATE_SOURCE
++			     : 0);
++
++	      if (0 <= utimecmp (dst_name, &dst_sb, &src_sb, options))
++		{
++		  /* We're using --update and the destination is not older
++		     than the source, so do not copy or move.  Pretend the
++		     rename succeeded, so the caller (if it's mv) doesn't
++		     end up removing the source file.  */
++		  if (rename_succeeded)
++		    *rename_succeeded = true;
++		  return true;
++		}
++	    }
++
+ 	  /* When there is an existing destination file, we may end up
+ 	     returning early, and hence not copying/moving the file.
+ 	     This may be due to an interactive `negative' reply to the
+@@ -1302,30 +1326,6 @@ copy_internal (char const *src_name, char const *dst_name,
+ 		      return false;
+ 		    }
+ 		}
+-
+-	      if (x->update)
+-		{
+-		  /* When preserving time stamps (but not moving within a file
+-		     system), don't worry if the destination time stamp is
+-		     less than the source merely because of time stamp
+-		     truncation.  */
+-		  int options = ((x->preserve_timestamps
+-				  && ! (x->move_mode
+-					&& dst_sb.st_dev == src_sb.st_dev))
+-				 ? UTIMECMP_TRUNCATE_SOURCE
+-				 : 0);
+-
+-		  if (0 <= utimecmp (dst_name, &dst_sb, &src_sb, options))
+-		    {
+-		      /* We're using --update and the destination is not older
+-			 than the source, so do not copy or move.  Pretend the
+-			 rename succeeded, so the caller (if it's mv) doesn't
+-			 end up removing the source file.  */
+-		      if (rename_succeeded)
+-			*rename_succeeded = true;
+-		      return true;
+-		    }
+-		}
+ 	    }
+
+ 	  if (x->move_mode)
+diff --git a/tests/mv/update b/tests/mv/update
+index 0c06024..6c3d149 100755
+--- a/tests/mv/update
++++ b/tests/mv/update
+@@ -1,7 +1,7 @@
+ #!/bin/sh
+ # make sure --update works as advertised
+
+-# Copyright (C) 2001, 2004, 2006 Free Software Foundation, Inc.
++# Copyright (C) 2001, 2004, 2006-2007 Free Software Foundation, Inc.
+
+ # This program is free software; you can redistribute it and/or modify
+ # it under the terms of the GNU General Public License as published by
+@@ -46,11 +46,16 @@ fi
+
+ fail=0
+
+-for cp_or_mv in cp mv; do
+-  # This is a no-op.
+-  $cp_or_mv --update old new || fail=1
+-  case "`cat new`" in new) ;; *) fail=1 ;; esac
+-  case "`cat old`" in old) ;; *) fail=1 ;; esac
++for interactive in '' -i; do
++  for cp_or_mv in cp mv; do
++    # This is a no-op, with no prompt.
++    # With coreutils-6.9 and earlier, using --update with -i would
++    # mistakenly elicit a prompt.
++    $cp_or_mv $interactive --update old new < /dev/null > out 2>&1 || fail=1
++    test -s out && fail=1
++    case "`cat new`" in new) ;; *) fail=1 ;; esac
++    case "`cat old`" in old) ;; *) fail=1 ;; esac
++  done
+ done
+
+ # This will actually perform the rename.
+--
+1.5.3.rc1.16.g9d6f
diff --git a/meta/recipes-core/coreutils/coreutils-6.9/coreutils-fix-install.patch b/meta/recipes-core/coreutils/coreutils-6.9/coreutils-fix-install.patch
new file mode 100644
index 0000000000..1b989f5311
--- /dev/null
+++ b/meta/recipes-core/coreutils/coreutils-6.9/coreutils-fix-install.patch
@@ -0,0 +1,99 @@
+The install command doesn't over write the dangling symlink, for
+example:
+
+$ install fileA /tmp/fileA
+
+If /tmp/fileA is a dangling symlink, there would be an error:
+
+install: cannot create regular file '/tmp/fileA': File exists
+
+This is because of the following code in copy.c:
+
+  if (!new_dst)
+    {
+      if (XSTAT (x, dst_name, &dst_sb) != 0)
+        {
+          if (errno != ENOENT)
+            {
+              error (0, errno, _("cannot stat %s"), quote (dst_name));
+              return false;
+            }
+          else
+            {
+              new_dst = true;
+            }
+        }
+
+XSTAT() use stat() for dst_name(the dangling symlink /tmp/fileA) when
+install.c invokes it, and stat will set errno to ENOENT, and then
+new_dst will be set to true which means that /tmp/fileA doesn't exist,
+then we will create /tmp/fileA without remove it first, so the error
+comes.
+
+This is fixed in a way which adds the member cmd_install in
+struct cp_options to make sure my change only affected to the install
+command and use lstat to fix the problem.
+    
+Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
+Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
+
+---
+ src/copy.c    |   10 +++++++++-
+ src/copy.h    |    3 +++
+ src/install.c |    1 +
+ 3 files changed, 13 insertions(+), 1 deletions(-)
+
+diff --git a/src/copy.c b/src/copy.c
+--- a/src/copy.c
++++ b/src/copy.c
+@@ -1029,6 +1029,7 @@ copy_internal (char const *src_name, char const *dst_name,
+   bool delayed_ok;
+   bool copied_as_regular = false;
+   bool preserve_metadata;
++  int dst_stat_result;
+ 
+   if (x->move_mode && rename_succeeded)
+     *rename_succeeded = false;
+@@ -1069,7 +1070,14 @@ copy_internal (char const *src_name, char const *dst_name,
+ 
+   if (!new_dst)
+     {
+-      if (XSTAT (x, dst_name, &dst_sb) != 0)
++      if ( x->cmd_install && ( x->backup_type == no_backups))
++        dst_stat_result = lstat (dst_name, &dst_sb);
++      else
++        {
++          dst_stat_result = XSTAT (x, dst_name, &dst_sb);
++        }
++
++      if (dst_stat_result != 0)
+ 	{
+ 	  if (errno != ENOENT)
+ 	    {
+diff --git a/src/copy.h b/src/copy.h
+--- a/src/copy.h
++++ b/src/copy.h
+@@ -114,6 +114,9 @@ struct cp_options
+      If that fails, then resort to copying.  */
+   bool move_mode;
+ 
++  /* For the install command */
++  bool cmd_install;
++
+   /* Whether this process has appropriate privileges to chown a file
+      whose owner is not the effective user ID.  */
+   bool chown_privileges;
+diff --git a/src/install.c b/src/install.c
+--- a/src/install.c
++++ b/src/install.c
+@@ -149,6 +149,7 @@ cp_option_init (struct cp_options *x)
+   x->hard_link = false;
+   x->interactive = I_UNSPECIFIED;
+   x->move_mode = false;
++  x->cmd_install = true;
+   x->chown_privileges = chown_privileges ();
+   x->one_file_system = false;
+   x->preserve_ownership = false;
+-- 
+1.7.0.1
+
diff --git a/meta/recipes-core/coreutils/coreutils-6.9/coreutils-i18n.patch b/meta/recipes-core/coreutils/coreutils-6.9/coreutils-i18n.patch
new file mode 100644
index 0000000000..3587186af1
--- /dev/null
+++ b/meta/recipes-core/coreutils/coreutils-6.9/coreutils-i18n.patch
@@ -0,0 +1,4049 @@
+This patch was imported from the Fedora Core 8 coreutils-6.9-9 package.
+
+The package is stated as being Licensed as GPLv2+.
+
+The comment indicates that the purpose is lin18nux/lsb compliance.
+
+Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
+
+--- /dev/null	2007-03-01 09:16:39.219409909 +0000
++++ coreutils-6.8+/tests/sort/sort-mb-tests	2007-03-01 15:08:24.000000000 +0000
+@@ -0,0 +1,58 @@
++#! /bin/sh
++case $# in
++  0) xx='../../src/sort';;
++  *) xx="$1";;
++esac
++test "$VERBOSE" && echo=echo || echo=:
++$echo testing program: $xx
++errors=0
++test "$srcdir" || srcdir=.
++test "$VERBOSE" && $xx --version 2> /dev/null
++
++export LC_ALL=en_US.UTF-8
++locale -k LC_CTYPE 2>&1 | grep -q charmap.*UTF-8 || exit 77
++errors=0
++
++$xx -t ＠ -k2 -n mb1.I > mb1.O
++code=$?
++if test $code != 0; then
++  $echo "Test mb1 failed: $xx return code $code differs from expected value 0" 1>&2
++  errors=`expr $errors + 1`
++else
++  cmp mb1.O $srcdir/mb1.X > /dev/null 2>&1
++  case $? in
++    0) if test "$VERBOSE"; then $echo "passed mb1"; fi;;
++    1) $echo "Test mb1 failed: files mb1.O and $srcdir/mb1.X differ" 1>&2
++       (diff -c mb1.O $srcdir/mb1.X) 2> /dev/null
++       errors=`expr $errors + 1`;;
++    2) $echo "Test mb1 may have failed." 1>&2
++       $echo The command "cmp mb1.O $srcdir/mb1.X" failed. 1>&2
++       errors=`expr $errors + 1`;;
++  esac
++fi
++
++$xx -t ＠ -k4 -n mb2.I > mb2.O
++code=$?
++if test $code != 0; then
++  $echo "Test mb2 failed: $xx return code $code differs from expected value 0" 1>&2
++  errors=`expr $errors + 1`
++else
++  cmp mb2.O $srcdir/mb2.X > /dev/null 2>&1
++  case $? in
++    0) if test "$VERBOSE"; then $echo "passed mb2"; fi;;
++    1) $echo "Test mb2 failed: files mb2.O and $srcdir/mb2.X differ" 1>&2
++       (diff -c mb2.O $srcdir/mb2.X) 2> /dev/null
++       errors=`expr $errors + 1`;;
++    2) $echo "Test mb2 may have failed." 1>&2
++       $echo The command "cmp mb2.O $srcdir/mb2.X" failed. 1>&2
++       errors=`expr $errors + 1`;;
++  esac
++fi
++
++if test $errors = 0; then
++  $echo Passed all 113 tests. 1>&2
++else
++  $echo Failed $errors tests. 1>&2
++fi
++test $errors = 0 || errors=1
++exit $errors
+--- /dev/null	2007-03-01 09:16:39.219409909 +0000
++++ coreutils-6.8+/tests/sort/mb2.I	2007-03-01 15:08:24.000000000 +0000
+@@ -0,0 +1,4 @@
++Apple＠ＡＡ10＠＠20
++Banana＠ＡＡ5＠＠30
++Citrus＠ＡＡ20＠＠5
++Cherry＠ＡＡ30＠＠10
+--- /dev/null	2007-03-01 09:16:39.219409909 +0000
++++ coreutils-6.8+/tests/sort/mb2.X	2007-03-01 15:08:24.000000000 +0000
+@@ -0,0 +1,4 @@
++Citrus＠ＡＡ20＠＠5
++Cherry＠ＡＡ30＠＠10
++Apple＠ＡＡ10＠＠20
++Banana＠ＡＡ5＠＠30
+--- /dev/null	2007-03-01 09:16:39.219409909 +0000
++++ coreutils-6.8+/tests/sort/mb1.I	2007-03-01 15:08:24.000000000 +0000
+@@ -0,0 +1,4 @@
++Apple＠10
++Banana＠5
++Citrus＠20
++Cherry＠30
+--- /dev/null	2007-03-01 09:16:39.219409909 +0000
++++ coreutils-6.8+/tests/sort/mb1.X	2007-03-01 15:08:24.000000000 +0000
+@@ -0,0 +1,4 @@
++Banana＠5
++Apple＠10
++Citrus＠20
++Cherry＠30
+--- coreutils-6.8+/tests/sort/Makefile.am.i18n	2007-01-24 07:47:37.000000000 +0000
++++ coreutils-6.8+/tests/sort/Makefile.am	2007-03-01 15:09:59.000000000 +0000
+@@ -66,15 +66,17 @@
+ bigfield.O bigfield.E
+ ##test-files-end
+ 
+-EXTRA_DIST = Test.pm $x-tests $(explicit) $(maint_gen)
+-noinst_SCRIPTS = $x-tests
++run_gen += mb1.0 mb2.0
++
++EXTRA_DIST = Test.pm $x-tests $(explicit) $(maint_gen) mb1.I mb1.X mb2.I mb2.X
++noinst_SCRIPTS = $x-tests # $x-mb-tests
+ TESTS_ENVIRONMENT = \
+   CU_TEST_NAME=`basename $(abs_srcdir)`,$$tst \
+   PATH="$(VG_PATH_PREFIX)`pwd`/../../src$(PATH_SEPARATOR)$$PATH"
+ 
+ editpl = sed -e 's,@''PERL''@,$(PERL),g' -e 's,@''srcdir''@,$(srcdir),g'
+ 
+-TESTS = $x-tests
++TESTS = $x-tests $x-mb-tests
+ 
+ mk_script = $(srcdir)/../mk-script
+ $(srcdir)/$x-tests: $(mk_script) Test.pm Makefile.am
+--- coreutils-6.8+/lib/linebuffer.h.i18n	2005-05-14 07:44:24.000000000 +0100
++++ coreutils-6.8+/lib/linebuffer.h	2007-03-01 15:08:24.000000000 +0000
+@@ -22,6 +22,11 @@
+ 
+ # include <stdio.h>
+ 
++/* Get mbstate_t.  */
++# if HAVE_WCHAR_H
++#  include <wchar.h>
++# endif
++
+ /* A `struct linebuffer' holds a line of text. */
+ 
+ struct linebuffer
+@@ -29,6 +34,9 @@
+   size_t size;			/* Allocated. */
+   size_t length;		/* Used. */
+   char *buffer;
++# if HAVE_WCHAR_H
++  mbstate_t state;
++# endif
+ };
+ 
+ /* Initialize linebuffer LINEBUFFER for use. */
+--- coreutils-6.8+/src/expand.c.i18n	2007-01-14 15:41:28.000000000 +0000
++++ coreutils-6.8+/src/expand.c	2007-03-01 15:08:24.000000000 +0000
+@@ -38,11 +38,28 @@
+ #include <stdio.h>
+ #include <getopt.h>
+ #include <sys/types.h>
++
++/* Get mbstate_t, mbrtowc(), wcwidth(). */
++#if HAVE_WCHAR_H
++# include <wchar.h>
++#endif
++
+ #include "system.h"
+ #include "error.h"
+ #include "quote.h"
+ #include "xstrndup.h"
+ 
++/* MB_LEN_MAX is incorrectly defined to be 1 in at least one GCC
++   installation; work around this configuration error.  */
++#if !defined MB_LEN_MAX || MB_LEN_MAX < 2
++# define MB_LEN_MAX 16
++#endif
++
++/* Some systems, like BeOS, have multibyte encodings but lack mbstate_t.  */
++#if HAVE_MBRTOWC && defined mbstate_t
++# define mbrtowc(pwc, s, n, ps) (mbrtowc) (pwc, s, n, 0)
++#endif
++
+ /* The official name of this program (e.g., no `g' prefix).  */
+ #define PROGRAM_NAME "expand"
+ 
+@@ -183,6 +200,7 @@
+ 	      stops = num_start + len - 1;
+ 	    }
+ 	}
++
+       else
+ 	{
+ 	  error (0, 0, _("tab size contains invalid character(s): %s"),
+@@ -365,6 +383,142 @@
+     }
+ }
+ 
++#if HAVE_MBRTOWC
++static void
++expand_multibyte (void)
++{
++  FILE *fp;			/* Input strem. */
++  mbstate_t i_state;		/* Current shift state of the input stream. */
++  mbstate_t i_state_bak;	/* Back up the I_STATE. */
++  mbstate_t o_state;		/* Current shift state of the output stream. */
++  char buf[MB_LEN_MAX + BUFSIZ];  /* For spooling a read byte sequence. */
++  char *bufpos;			/* Next read position of BUF. */
++  size_t buflen = 0;		/* The length of the byte sequence in buf. */
++  wchar_t wc;			/* A gotten wide character. */
++  size_t mblength;		/* The byte size of a multibyte character
++				   which shows as same character as WC. */
++  int tab_index = 0;		/* Index in `tab_list' of next tabstop. */
++  int column = 0;		/* Column on screen of the next char. */
++  int next_tab_column;		/* Column the next tab stop is on. */
++  int convert = 1;		/* If nonzero, perform translations. */
++
++  fp = next_file ((FILE *) NULL);
++  if (fp == NULL)
++    return;
++
++  memset (&o_state, '\0', sizeof(mbstate_t));
++  memset (&i_state, '\0', sizeof(mbstate_t));
++
++  for (;;)
++    {
++      /* Refill the buffer BUF. */
++      if (buflen < MB_LEN_MAX && !feof(fp) && !ferror(fp))
++	{
++	  memmove (buf, bufpos, buflen);
++	  buflen += fread (buf + buflen, sizeof(char), BUFSIZ, fp);
++	  bufpos = buf;
++	}
++
++      /* No character is left in BUF. */
++      if (buflen < 1)
++	{
++	  fp = next_file (fp);
++
++	  if (fp == NULL)
++	    break;		/* No more files. */
++	  else
++	    {
++	      memset (&i_state, '\0', sizeof(mbstate_t));
++	      continue;
++	    }
++	}
++
++      /* Get a wide character. */
++      i_state_bak = i_state;
++      mblength = mbrtowc (&wc, bufpos, buflen, &i_state);
++
++      switch (mblength)
++	{
++	case (size_t)-1:	/* illegal byte sequence. */
++	case (size_t)-2:
++	  mblength = 1;
++	  i_state = i_state_bak;
++	  if (convert)
++	    {
++	      ++column;
++	      if (convert_entire_line == 0)
++		convert = 0;
++	    }
++	  putchar (*bufpos);
++	  break;
++
++	case 0:		/* null. */
++	  mblength = 1;
++	  if (convert && convert_entire_line == 0)
++	    convert = 0;
++	  putchar ('\0');
++	  break;
++
++	default:
++	  if (wc == L'\n')   /* LF. */
++	    {
++	      tab_index = 0;
++	      column = 0;
++	      convert = 1;
++	      putchar ('\n');
++	    }
++	  else if (wc == L'\t' && convert)	/* Tab. */
++	    {
++	      if (tab_size == 0)
++		{
++		  /* Do not let tab_index == first_free_tab;
++		     stop when it is 1 less. */
++		  while (tab_index < first_free_tab - 1
++		      && column >= tab_list[tab_index])
++		    tab_index++;
++		  next_tab_column = tab_list[tab_index];
++		  if (tab_index < first_free_tab - 1)
++		    tab_index++;
++		  if (column >= next_tab_column)
++		    next_tab_column = column + 1;
++		}
++	      else
++		next_tab_column = column + tab_size - column % tab_size;
++
++	      while (column < next_tab_column)
++		{
++		  putchar (' ');
++		  ++column;
++		}
++	    }
++	  else  /* Others. */
++	    {
++	      if (convert)
++		{
++		  if (wc == L'\b')
++		    {
++		      if (column > 0)
++			--column;
++		    }
++		  else
++		    {
++		      int width;		/* The width of WC. */
++
++		      width = wcwidth (wc);
++		      column += (width > 0) ? width : 0;
++		      if (convert_entire_line == 0)
++			convert = 0;
++		    }
++		}
++	      fwrite (bufpos, sizeof(char), mblength, stdout);
++	    }
++	}
++      buflen -= mblength;
++      bufpos += mblength;
++    }
++}
++#endif
++
+ int
+ main (int argc, char **argv)
+ {
+@@ -429,7 +583,12 @@
+ 
+   file_list = (optind < argc ? &argv[optind] : stdin_argv);
+ 
+-  expand ();
++#if HAVE_MBRTOWC
++  if (MB_CUR_MAX > 1)
++    expand_multibyte ();
++  else
++#endif
++    expand ();
+ 
+   if (have_read_stdin && fclose (stdin) != 0)
+     error (EXIT_FAILURE, errno, "-");
+--- coreutils-6.8+/src/join.c.i18n	2007-01-14 15:41:28.000000000 +0000
++++ coreutils-6.8+/src/join.c	2007-03-01 15:08:24.000000000 +0000
+@@ -23,16 +23,30 @@
+ #include <sys/types.h>
+ #include <getopt.h>
+ 
++/* Get mbstate_t, mbrtowc(), mbrtowc(), wcwidth().  */
++#if HAVE_WCHAR_H
++# include <wchar.h>
++#endif
++
++/* Get iswblank(), towupper.  */
++#if HAVE_WCTYPE_H
++# include <wctype.h>
++#endif
++
+ #include "system.h"
+ #include "error.h"
+ #include "hard-locale.h"
+ #include "linebuffer.h"
+-#include "memcasecmp.h"
+ #include "quote.h"
+ #include "stdio--.h"
+ #include "xmemcoll.h"
+ #include "xstrtol.h"
+ 
++/* Some systems, like BeOS, have multibyte encodings but lack mbstate_t.  */
++#if HAVE_MBRTOWC && defined mbstate_t
++# define mbrtowc(pwc, s, n, ps) (mbrtowc) (pwc, s, n, 0)
++#endif
++
+ /* The official name of this program (e.g., no `g' prefix).  */
+ #define PROGRAM_NAME "join"
+ 
+@@ -104,10 +118,12 @@
+ /* Last element in `outlist', where a new element can be added.  */
+ static struct outlist *outlist_end = &outlist_head;
+ 
+-/* Tab character separating fields.  If negative, fields are separated
+-   by any nonempty string of blanks, otherwise by exactly one
+-   tab character whose value (when cast to unsigned char) equals TAB.  */
+-static int tab = -1;
++/* Tab character separating fields.  If NULL, fields are separated
++   by any nonempty string of blanks.  */
++static char *tab = NULL;
++
++/* The number of bytes used for tab. */
++static size_t tablen = 0;
+ 
+ static struct option const longopts[] =
+ {
+@@ -190,6 +206,8 @@
+ 
+ /* Fill in the `fields' structure in LINE.  */
+ 
++/* Fill in the `fields' structure in LINE.  */
++
+ static void
+ xfields (struct line *line)
+ {
+@@ -199,10 +217,11 @@
+   if (ptr == lim)
+     return;
+ 
+-  if (0 <= tab)
++  if (tab != NULL)
+     {
++      unsigned char t = tab[0];
+       char *sep;
+-      for (; (sep = memchr (ptr, tab, lim - ptr)) != NULL; ptr = sep + 1)
++      for (; (sep = memchr (ptr, t, lim - ptr)) != NULL; ptr = sep + 1)
+ 	extract_field (line, ptr, sep - ptr);
+     }
+   else
+@@ -229,6 +248,148 @@
+   extract_field (line, ptr, lim - ptr);
+ }
+ 
++#if HAVE_MBRTOWC
++static void
++xfields_multibyte (struct line *line)
++{
++  char *ptr = line->buf.buffer;
++  char const *lim = ptr + line->buf.length - 1;
++  wchar_t wc = 0;
++  size_t mblength = 1;
++  mbstate_t state, state_bak;
++
++  memset (&state, 0, sizeof (mbstate_t));
++
++  if (ptr == lim)
++    return;
++
++  if (tab != NULL)
++    {
++      unsigned char t = tab[0];
++      char *sep = ptr;
++      for (; ptr < lim; ptr = sep + mblength)
++	{
++	  sep = ptr;
++	  while (sep < lim)
++	    {
++	      state_bak = state;
++	      mblength = mbrtowc (&wc, sep, lim - sep + 1, &state);
++
++	      if (mblength == (size_t)-1 || mblength == (size_t)-2)
++		{
++		  mblength = 1;
++		  state = state_bak;
++		}
++	      mblength = (mblength < 1) ? 1 : mblength;
++
++	      if (mblength == tablen && !memcmp (sep, tab, mblength))
++		break;
++	      else
++		{
++		  sep += mblength;
++		  continue;
++		}
++	    }
++
++	  if (sep == lim)
++	    break;
++
++	  extract_field (line, ptr, sep - ptr);
++	}
++    }
++  else
++    {
++      /* Skip leading blanks before the first field.  */
++      while(ptr < lim)
++      {
++        state_bak = state;
++        mblength = mbrtowc (&wc, ptr, lim - ptr + 1, &state);
++
++        if (mblength == (size_t)-1 || mblength == (size_t)-2)
++          {
++            mblength = 1;
++            state = state_bak;
++            break;
++          }
++        mblength = (mblength < 1) ? 1 : mblength;
++
++        if (!iswblank(wc))
++          break;
++        ptr += mblength;
++      }
++
++      do
++	{
++	  char *sep;
++	  state_bak = state;
++	  mblength = mbrtowc (&wc, ptr, lim - ptr + 1, &state);
++	  if (mblength == (size_t)-1 || mblength == (size_t)-2)
++	    {
++	      mblength = 1;
++	      state = state_bak;
++	      break;
++	    }
++	  mblength = (mblength < 1) ? 1 : mblength;
++
++	  sep = ptr + mblength;
++	  while (sep != lim)
++	    {
++	      state_bak = state;
++	      mblength = mbrtowc (&wc, sep, lim - sep + 1, &state);
++	      if (mblength == (size_t)-1 || mblength == (size_t)-2)
++		{
++		  mblength = 1;
++		  state = state_bak;
++		  break;
++		}
++	      mblength = (mblength < 1) ? 1 : mblength;
++
++	      if (iswblank (wc))
++		break;
++
++	      sep += mblength;
++	    }
++
++	  extract_field (line, ptr, sep - ptr);
++	  if (sep == lim)
++	    return;
++
++	  state_bak = state;
++	  mblength = mbrtowc (&wc, sep, lim - sep + 1, &state);
++	  if (mblength == (size_t)-1 || mblength == (size_t)-2)
++	    {
++	      mblength = 1;
++	      state = state_bak;
++	      break;
++	    }
++	  mblength = (mblength < 1) ? 1 : mblength;
++
++	  ptr = sep + mblength;
++	  while (ptr != lim)
++	    {
++	      state_bak = state;
++	      mblength = mbrtowc (&wc, ptr, lim - ptr + 1, &state);
++	      if (mblength == (size_t)-1 || mblength == (size_t)-2)
++		{
++		  mblength = 1;
++		  state = state_bak;
++		  break;
++		}
++	      mblength = (mblength < 1) ? 1 : mblength;
++
++	      if (!iswblank (wc))
++		break;
++
++	      ptr += mblength;
++	    }
++	}
++      while (ptr != lim);
++    }
++
++  extract_field (line, ptr, lim - ptr);
++}
++#endif
++
+ /* Read a line from FP into LINE and split it into fields.
+    Return true if successful.  */
+ 
+@@ -249,6 +410,11 @@
+   line->nfields_allocated = 0;
+   line->nfields = 0;
+   line->fields = NULL;
++#if HAVE_MBRTOWC
++  if (MB_CUR_MAX > 1)
++    xfields_multibyte (line);
++  else
++#endif
+   xfields (line);
+   return true;
+ }
+@@ -303,56 +469,114 @@
+ keycmp (struct line const *line1, struct line const *line2)
+ {
+   /* Start of field to compare in each file.  */
+-  char *beg1;
+-  char *beg2;
+-
+-  size_t len1;
+-  size_t len2;		/* Length of fields to compare.  */
++  char *beg[2];
++  char *copy[2];
++  size_t len[2];	/* Length of fields to compare.  */
+   int diff;
++  int i, j;
+ 
+   if (join_field_1 < line1->nfields)
+     {
+-      beg1 = line1->fields[join_field_1].beg;
+-      len1 = line1->fields[join_field_1].len;
++      beg[0] = line1->fields[join_field_1].beg;
++      len[0] = line1->fields[join_field_1].len;
+     }
+   else
+     {
+-      beg1 = NULL;
+-      len1 = 0;
++      beg[0] = NULL;
++      len[0] = 0;
+     }
+ 
+   if (join_field_2 < line2->nfields)
+     {
+-      beg2 = line2->fields[join_field_2].beg;
+-      len2 = line2->fields[join_field_2].len;
++      beg[1] = line2->fields[join_field_2].beg;
++      len[1] = line2->fields[join_field_2].len;
+     }
+   else
+     {
+-      beg2 = NULL;
+-      len2 = 0;
++      beg[1] = NULL;
++      len[1] = 0;
+     }
+ 
+-  if (len1 == 0)
+-    return len2 == 0 ? 0 : -1;
+-  if (len2 == 0)
++  if (len[0] == 0)
++    return len[1] == 0 ? 0 : -1;
++  if (len[1] == 0)
+     return 1;
+ 
+   if (ignore_case)
+     {
+-      /* FIXME: ignore_case does not work with NLS (in particular,
+-         with multibyte chars).  */
+-      diff = memcasecmp (beg1, beg2, MIN (len1, len2));
++#ifdef HAVE_MBRTOWC
++      if (MB_CUR_MAX > 1)
++      {
++        size_t mblength;
++        wchar_t wc, uwc;
++        mbstate_t state, state_bak;
++
++        memset (&state, '\0', sizeof (mbstate_t));
++
++        for (i = 0; i < 2; i++)
++          {
++            copy[i] = alloca (len[i] + 1);
++
++            for (j = 0; j < MIN (len[0], len[1]);)
++              {
++                state_bak = state;
++                mblength = mbrtowc (&wc, beg[i] + j, len[i] - j, &state);
++
++                switch (mblength)
++                  {
++                  case (size_t) -1:
++                  case (size_t) -2:
++                    state = state_bak;
++                    /* Fall through */
++                  case 0:
++                    mblength = 1;
++                    break;
++
++                  default:
++                    uwc = towupper (wc);
++
++                    if (uwc != wc)
++                      {
++                        mbstate_t state_wc;
++
++                        memset (&state_wc, '\0', sizeof (mbstate_t));
++                        wcrtomb (copy[i] + j, uwc, &state_wc);
++                      }
++                    else
++                      memcpy (copy[i] + j, beg[i] + j, mblength);
++                  }
++                j += mblength;
++              }
++            copy[i][j] = '\0';
++          }
++      }
++      else
++#endif
++      {
++        for (i = 0; i < 2; i++)
++          {
++            copy[i] = alloca (len[i] + 1);
++
++            for (j = 0; j < MIN (len[0], len[1]); j++)
++              copy[i][j] = toupper (beg[i][j]);
++
++            copy[i][j] = '\0';
++          }
++      }
+     }
+   else
+     {
+-      if (hard_LC_COLLATE)
+-	return xmemcoll (beg1, len1, beg2, len2);
+-      diff = memcmp (beg1, beg2, MIN (len1, len2));
++      copy[0] = (unsigned char *) beg[0];
++      copy[1] = (unsigned char *) beg[1];
+     }
+ 
++  if (hard_LC_COLLATE)
++    return xmemcoll ((char *) copy[0], len[0], (char *) copy[1], len[1]);
++  diff = memcmp (copy[0], copy[1], MIN (len[0], len[1]));
++
+   if (diff)
+     return diff;
+-  return len1 < len2 ? -1 : len1 != len2;
++  return len[0] - len[1];
+ }
+ 
+ /* Print field N of LINE if it exists and is nonempty, otherwise
+@@ -377,11 +601,18 @@
+ 
+ /* Print the join of LINE1 and LINE2.  */
+ 
++#define PUT_TAB_CHAR							\
++  do									\
++    {									\
++      (tab != NULL) ?							\
++	fwrite(tab, sizeof(char), tablen, stdout) : putchar (' ');	\
++    }									\
++  while (0)								
++
+ static void
+ prjoin (struct line const *line1, struct line const *line2)
+ {
+   const struct outlist *outlist;
+-  char output_separator = tab < 0 ? ' ' : tab;
+ 
+   outlist = outlist_head.next;
+   if (outlist)
+@@ -397,12 +628,12 @@
+ 	  if (o->file == 0)
+ 	    {
+ 	      if (line1 == &uni_blank)
+-	        {
++		{
+ 		  line = line2;
+ 		  field = join_field_2;
+ 		}
+ 	      else
+-	        {
++		{
+ 		  line = line1;
+ 		  field = join_field_1;
+ 		}
+@@ -416,7 +647,7 @@
+ 	  o = o->next;
+ 	  if (o == NULL)
+ 	    break;
+-	  putchar (output_separator);
++	  PUT_TAB_CHAR;
+ 	}
+       putchar ('\n');
+     }
+@@ -434,23 +665,23 @@
+       prfield (join_field_1, line1);
+       for (i = 0; i < join_field_1 && i < line1->nfields; ++i)
+ 	{
+-	  putchar (output_separator);
++	  PUT_TAB_CHAR;
+ 	  prfield (i, line1);
+ 	}
+       for (i = join_field_1 + 1; i < line1->nfields; ++i)
+ 	{
+-	  putchar (output_separator);
++	  PUT_TAB_CHAR;
+ 	  prfield (i, line1);
+ 	}
+ 
+       for (i = 0; i < join_field_2 && i < line2->nfields; ++i)
+ 	{
+-	  putchar (output_separator);
++	  PUT_TAB_CHAR;
+ 	  prfield (i, line2);
+ 	}
+       for (i = join_field_2 + 1; i < line2->nfields; ++i)
+ 	{
+-	  putchar (output_separator);
++	  PUT_TAB_CHAR;
+ 	  prfield (i, line2);
+ 	}
+       putchar ('\n');
+@@ -859,20 +1090,41 @@
+ 
+ 	case 't':
+ 	  {
+-	    unsigned char newtab = optarg[0];
+-	    if (! newtab)
++	    char *newtab;
++	    size_t newtablen;
++	    if (! optarg[0])
+ 	      error (EXIT_FAILURE, 0, _("empty tab"));
+-	    if (optarg[1])
++	    newtab = xstrdup (optarg);
++#if HAVE_MBRTOWC
++	    if (MB_CUR_MAX > 1)
++	      {
++		mbstate_t state;
++
++		memset (&state, 0, sizeof (mbstate_t));
++		newtablen = mbrtowc (NULL, newtab,
++				     strnlen (newtab, MB_LEN_MAX),
++				     &state);
++		if (newtablen == (size_t) 0
++		    || newtablen == (size_t) -1
++		    || newtablen == (size_t) -2)
++		  newtablen = 1;
++	      }
++	    else
++#endif
++	      newtablen = 1;
++		
++	    if (newtablen == 1 && newtab[1])
++	      {
++		if (STREQ (newtab, "\\0"))
++		  newtab[0] = '\0';
++	      }
++	    if (tab != NULL && strcmp (tab, newtab))
+ 	      {
+-		if (STREQ (optarg, "\\0"))
+-		  newtab = '\0';
+-		else
+-		  error (EXIT_FAILURE, 0, _("multi-character tab %s"),
+-			 quote (optarg));
++		free (newtab);
++		error (EXIT_FAILURE, 0, _("incompatible tabs"));
+ 	      }
+-	    if (0 <= tab && tab != newtab)
+-	      error (EXIT_FAILURE, 0, _("incompatible tabs"));
+ 	    tab = newtab;
++	    tablen = newtablen;
+ 	  }
+ 	  break;
+ 
+--- coreutils-6.8+/src/uniq.c.i18n	2007-01-14 15:41:28.000000000 +0000
++++ coreutils-6.8+/src/uniq.c	2007-03-01 15:08:24.000000000 +0000
+@@ -23,6 +23,16 @@
+ #include <getopt.h>
+ #include <sys/types.h>
+ 
++/* Get mbstate_t, mbrtowc(). */