glibc/intl/dcigettext.c

/* Implementation of the internal dcigettext function.
   Copyright (C) 1995-2014 Free Software Foundation, Inc.
   This file is part of the GNU C Library.

   The GNU C Library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2.1 of the License, or (at your option) any later version.

   The GNU C Library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with the GNU C Library; if not, see
   <http://www.gnu.org/licenses/>.  */

/* Tell glibc's <string.h> to provide a prototype for mempcpy().
   This must come before <config.h> because <config.h> may include
   <features.h>, and once <features.h> has been included, it's too late.  */
#ifndef _GNU_SOURCE
# define _GNU_SOURCE	1
#endif

#ifdef HAVE_CONFIG_H
# include <config.h>
#endif

#include <sys/types.h>

#ifdef __GNUC__
# define alloca __builtin_alloca
# define HAVE_ALLOCA 1
#else
# if defined HAVE_ALLOCA_H || defined _LIBC
#  include <alloca.h>
# else
#  ifdef _AIX
 #pragma alloca
#  else
#   ifndef alloca
char *alloca ();
#   endif
#  endif
# endif
#endif

#include <errno.h>
#ifndef errno
extern int errno;
#endif
#ifndef __set_errno
# define __set_errno(val) errno = (val)
#endif

#include <stddef.h>
#include <stdlib.h>
#include <string.h>

#if defined HAVE_UNISTD_H || defined _LIBC
# include <unistd.h>
#endif

#include <locale.h>

#if defined HAVE_SYS_PARAM_H || defined _LIBC
# include <sys/param.h>
#endif

#include "gettextP.h"
#include "plural-exp.h"
#ifdef _LIBC
# include <libintl.h>
#else
# include "libgnuintl.h"
#endif
#include "hash-string.h"

/* Thread safetyness.  */
#ifdef _LIBC
# include <bits/libc-lock.h>
#else
/* Provide dummy implementation if this is outside glibc.  */
# define __libc_lock_define_initialized(CLASS, NAME)
# define __libc_lock_lock(NAME)
# define __libc_lock_unlock(NAME)
# define __libc_rwlock_define_initialized(CLASS, NAME)
# define __libc_rwlock_rdlock(NAME)
# define __libc_rwlock_unlock(NAME)
#endif

/* Alignment of types.  */
#if defined __GNUC__ && __GNUC__ >= 2
# define alignof(TYPE) __alignof__ (TYPE)
#else
# define alignof(TYPE) \
    ((int) &((struct { char dummy1; TYPE dummy2; } *) 0)->dummy2)
#endif

/* The internal variables in the standalone libintl.a must have different
   names than the internal variables in GNU libc, otherwise programs
   using libintl.a cannot be linked statically.  */
#if !defined _LIBC
# define _nl_default_default_domain libintl_nl_default_default_domain
# define _nl_current_default_domain libintl_nl_current_default_domain
# define _nl_default_dirname libintl_nl_default_dirname
# define _nl_domain_bindings libintl_nl_domain_bindings
#endif

/* Some compilers, like SunOS4 cc, don't have offsetof in <stddef.h>.  */
#ifndef offsetof
# define offsetof(type,ident) ((size_t)&(((type*)0)->ident))
#endif

/* @@ end of prolog @@ */

#ifdef _LIBC
/* Rename the non ANSI C functions.  This is required by the standard
   because some ANSI C functions will require linking with this object
   file and the name space must not be polluted.  */
# define getcwd __getcwd
# ifndef stpcpy
#  define stpcpy __stpcpy
# endif
# define tfind __tfind
#else
# if !defined HAVE_GETCWD
char *getwd ();
#  define getcwd(buf, max) getwd (buf)
# else
char *getcwd ();
# endif
# ifndef HAVE_STPCPY
static char *stpcpy PARAMS ((char *dest, const char *src));
# endif
# ifndef HAVE_MEMPCPY
static void *mempcpy PARAMS ((void *dest, const void *src, size_t n));
# endif
#endif

/* Amount to increase buffer size by in each try.  */
#define PATH_INCR 32

/* The following is from pathmax.h.  */
/* Non-POSIX BSD systems might have gcc's limits.h, which doesn't define
   PATH_MAX but might cause redefinition warnings when sys/param.h is
   later included (as on MORE/BSD 4.3).  */
#if defined _POSIX_VERSION || (defined HAVE_LIMITS_H && !defined __GNUC__)
# include <limits.h>
#endif

#ifndef _POSIX_PATH_MAX
# define _POSIX_PATH_MAX 255
#endif

#if !defined PATH_MAX && defined _PC_PATH_MAX
# define PATH_MAX (pathconf ("/", _PC_PATH_MAX) < 1 ? 1024 : pathconf ("/", _PC_PATH_MAX))
#endif

/* Don't include sys/param.h if it already has been.  */
#if defined HAVE_SYS_PARAM_H && !defined PATH_MAX && !defined MAXPATHLEN
# include <sys/param.h>
#endif

#if !defined PATH_MAX && defined MAXPATHLEN
# define PATH_MAX MAXPATHLEN
#endif

#ifndef PATH_MAX
# define PATH_MAX _POSIX_PATH_MAX
#endif

/* Whether to support different locales in different threads.  */
#if defined _LIBC || HAVE_NL_LOCALE_NAME
# define HAVE_PER_THREAD_LOCALE
#endif

/* This is the type used for the search tree where known translations
   are stored.  */
struct known_translation_t
{
  /* Domain in which to search.  */
  const char *domainname;

  /* The category.  */
  int category;

#ifdef HAVE_PER_THREAD_LOCALE
  /* Name of the relevant locale category, or "" for the global locale.  */
  const char *localename;
#endif

  /* State of the catalog counter at the point the string was found.  */
  int counter;

  /* Catalog where the string was found.  */
  struct loaded_l10nfile *domain;

  /* And finally the translation.  */
  const char *translation;
  size_t translation_length;

  /* Pointer to the string in question.  */
  union
    {
      char appended[ZERO];  /* used if domain != NULL */
      const char *ptr;      /* used if domain == NULL */
    }
  msgid;
};

/* Root of the search tree with known translations.  We can use this
   only if the system provides the `tsearch' function family.  */
#if defined HAVE_TSEARCH || defined _LIBC
# include <search.h>

static void *root;

# ifdef _LIBC
#  define tsearch __tsearch
# endif

/* Function to compare two entries in the table of known translations.  */
static int transcmp PARAMS ((const void *p1, const void *p2));
static int
transcmp (p1, p2)
     const void *p1;
     const void *p2;
{
  const struct known_translation_t *s1;
  const struct known_translation_t *s2;
  int result;

  s1 = (const struct known_translation_t *) p1;
  s2 = (const struct known_translation_t *) p2;

  result = strcmp (s1->domain != NULL ? s1->msgid.appended : s1->msgid.ptr,
		   s2->domain != NULL ? s2->msgid.appended : s2->msgid.ptr);
  if (result == 0)
    {
      result = strcmp (s1->domainname, s2->domainname);
      if (result == 0)
	{
#ifdef HAVE_PER_THREAD_LOCALE
	  result = strcmp (s1->localename, s2->localename);
	  if (result == 0)
#endif
	    /* We compare the category last (though this is the cheapest
	       operation) since it is hopefully always the same (namely
	       LC_MESSAGES).  */
	    result = s1->category - s2->category;
	}
    }

  return result;
}
#endif

/* Name of the default domain used for gettext(3) prior any call to
   textdomain(3).  The default value for this is "messages".  */
const char _nl_default_default_domain[] attribute_hidden = "messages";

/* Value used as the default domain for gettext(3).  */
const char *_nl_current_default_domain attribute_hidden
     = _nl_default_default_domain;

/* Contains the default location of the message catalogs.  */

#ifdef _LIBC
extern const char _nl_default_dirname[];
libc_hidden_proto (_nl_default_dirname)
#endif
const char _nl_default_dirname[] = LOCALEDIR;
#ifdef _LIBC
libc_hidden_data_def (_nl_default_dirname)
#endif

/* List with bindings of specific domains created by bindtextdomain()
   calls.  */
struct binding *_nl_domain_bindings;

/* Prototypes for local functions.  */
static char *plural_lookup PARAMS ((struct loaded_l10nfile *domain,
				    unsigned long int n,
				    const char *translation,
				    size_t translation_len))
     internal_function;
static const char *guess_category_value PARAMS ((int category,
						 const char *categoryname))
     internal_function;
#ifdef _LIBC
# include "../locale/localeinfo.h"
# define category_to_name(category) \
  _nl_category_names.str + _nl_category_name_idxs[category]
#else
static const char *category_to_name PARAMS ((int category)) internal_function;
#endif


/* For those loosing systems which don't have `alloca' we have to add
   some additional code emulating it.  */
#ifdef HAVE_ALLOCA
/* Nothing has to be done.  */
# define freea(p) /* nothing */
# define ADD_BLOCK(list, address) /* nothing */
# define FREE_BLOCKS(list) /* nothing */
#else
struct block_list
{
  void *address;
  struct block_list *next;
};
# define ADD_BLOCK(list, addr)						      \
  do {									      \
    struct block_list *newp = (struct block_list *) malloc (sizeof (*newp));  \
    /* If we cannot get a free block we cannot add the new element to	      \
       the list.  */							      \
    if (newp != NULL) {							      \
      newp->address = (addr);						      \
      newp->next = (list);						      \
      (list) = newp;							      \
    }									      \
  } while (0)
# define FREE_BLOCKS(list)						      \
  do {									      \
    while (list != NULL) {						      \
      struct block_list *old = list;					      \
      list = list->next;						      \
      free (old->address);						      \
      free (old);							      \
    }									      \
  } while (0)
# undef alloca
# define alloca(size) (malloc (size))
# define freea(p) free (p)
#endif	/* have alloca */


#ifdef _LIBC
/* List of blocks allocated for translations.  */
typedef struct transmem_list
{
  struct transmem_list *next;
  char data[ZERO];
} transmem_block_t;
static struct transmem_list *transmem_list;
#else
typedef unsigned char transmem_block_t;
#endif
#if defined _LIBC || HAVE_ICONV
static const char *get_output_charset PARAMS ((struct binding *domainbinding))
     internal_function;
#endif


/* Names for the libintl functions are a problem.  They must not clash
   with existing names and they should follow ANSI C.  But this source
   code is also used in GNU C Library where the names have a __
   prefix.  So we have to make a difference here.  */
#ifdef _LIBC
# define DCIGETTEXT __dcigettext
#else
# define DCIGETTEXT libintl_dcigettext
#endif

/* Lock variable to protect the global data in the gettext implementation.  */
#ifdef _LIBC
__libc_rwlock_define_initialized (, _nl_state_lock attribute_hidden)
#endif

/* Checking whether the binaries runs SUID must be done and glibc provides
   easier methods therefore we make a difference here.  */
#ifdef _LIBC
# define ENABLE_SECURE __libc_enable_secure
# define DETERMINE_SECURE
#else
# ifndef HAVE_GETUID
#  define getuid() 0
# endif
# ifndef HAVE_GETGID
#  define getgid() 0
# endif
# ifndef HAVE_GETEUID
#  define geteuid() getuid()
# endif
# ifndef HAVE_GETEGID
#  define getegid() getgid()
# endif
static int enable_secure;
# define ENABLE_SECURE (enable_secure == 1)
# define DETERMINE_SECURE \
  if (enable_secure == 0)						      \
    {									      \
      if (getuid () != geteuid () || getgid () != getegid ())		      \
	enable_secure = 1;						      \
      else								      \
	enable_secure = -1;						      \
    }
#endif

/* Get the function to evaluate the plural expression.  */
#include "plural-eval.c"

/* Look up MSGID in the DOMAINNAME message catalog for the current
   CATEGORY locale and, if PLURAL is nonzero, search over string
   depending on the plural form determined by N.  */
char *
DCIGETTEXT (domainname, msgid1, msgid2, plural, n, category)
     const char *domainname;
     const char *msgid1;
     const char *msgid2;
     int plural;
     unsigned long int n;
     int category;
{
#ifndef HAVE_ALLOCA
  struct block_list *block_list = NULL;
#endif
  struct loaded_l10nfile *domain;
  struct binding *binding;
  const char *categoryname;
  const char *categoryvalue;
  char *dirname, *xdomainname;
  char *single_locale;
  char *retval;
  size_t retlen;
  int saved_errno;
#if defined HAVE_TSEARCH || defined _LIBC
  struct known_translation_t search;
  struct known_translation_t **foundp = NULL;
# ifdef HAVE_PER_THREAD_LOCALE
  const char *localename;
# endif
#endif
  size_t domainname_len;

  /* If no real MSGID is given return NULL.  */
  if (msgid1 == NULL)
    return NULL;

#ifdef _LIBC
  if (category < 0 || category >= __LC_LAST || category == LC_ALL)
    /* Bogus.  */
    return (plural == 0
	    ? (char *) msgid1
	    /* Use the Germanic plural rule.  */
	    : n == 1 ? (char *) msgid1 : (char *) msgid2);
#endif

#ifdef _LIBC
  __libc_rwlock_define (extern, __libc_setlocale_lock attribute_hidden)
  __libc_rwlock_rdlock (__libc_setlocale_lock);
#endif

  __libc_rwlock_rdlock (_nl_state_lock);

  /* If DOMAINNAME is NULL, we are interested in the default domain.  If
     CATEGORY is not LC_MESSAGES this might not make much sense but the
     definition left this undefined.  */
  if (domainname == NULL)
    domainname = _nl_current_default_domain;

#if defined HAVE_TSEARCH || defined _LIBC
  /* Try to find the translation among those which we found at
     some time.  */
  search.domain = NULL;
  search.msgid.ptr = msgid1;
  search.domainname = domainname;
  search.category = category;
# ifdef HAVE_PER_THREAD_LOCALE
#  ifdef _LIBC
  localename = strdupa (__current_locale_name (category));
#  endif
  search.localename = localename;
# endif

  /* Since tfind/tsearch manage a balanced tree, concurrent tfind and
     tsearch calls can be fatal.  */
  __libc_rwlock_define_initialized (static, tree_lock);
  __libc_rwlock_rdlock (tree_lock);

  foundp = (struct known_translation_t **) tfind (&search, &root, transcmp);

  __libc_rwlock_unlock (tree_lock);

  if (foundp != NULL && (*foundp)->counter == _nl_msg_cat_cntr)
    {
      /* Now deal with plural.  */
      if (plural)
	retval = plural_lookup ((*foundp)->domain, n, (*foundp)->translation,
				(*foundp)->translation_length);
      else
	retval = (char *) (*foundp)->translation;

# ifdef _LIBC
      __libc_rwlock_unlock (__libc_setlocale_lock);
# endif
      __libc_rwlock_unlock (_nl_state_lock);
      return retval;
    }
#endif

  /* Preserve the `errno' value.  */
  saved_errno = errno;

  /* See whether this is a SUID binary or not.  */
  DETERMINE_SECURE;

  /* First find matching binding.  */
  for (binding = _nl_domain_bindings; binding != NULL; binding = binding->next)
    {
      int compare = strcmp (domainname, binding->domainname);
      if (compare == 0)
	/* We found it!  */
	break;
      if (compare < 0)
	{
	  /* It is not in the list.  */
	  binding = NULL;
	  break;
	}
    }

  if (binding == NULL)
    dirname = (char *) _nl_default_dirname;
  else if (binding->dirname[0] == '/')
    dirname = binding->dirname;
  else
    {
      /* We have a relative path.  Make it absolute now.  */
      size_t dirname_len = strlen (binding->dirname) + 1;
      size_t path_max;
      char *ret;

      path_max = (unsigned int) PATH_MAX;
      path_max += 2;		/* The getcwd docs say to do this.  */

      for (;;)
	{
	  dirname = (char *) alloca (path_max + dirname_len);
	  ADD_BLOCK (block_list, dirname);

	  __set_errno (0);
	  ret = getcwd (dirname, path_max);
	  if (ret != NULL || errno != ERANGE)
	    break;

	  path_max += path_max / 2;
	  path_max += PATH_INCR;
	}

      if (ret == NULL)
	goto no_translation;

      stpcpy (stpcpy (strchr (dirname, '\0'), "/"), binding->dirname);
    }

  /* Now determine the symbolic name of CATEGORY and its value.  */
  categoryname = category_to_name (category);
  categoryvalue = guess_category_value (category, categoryname);

  domainname_len = strlen (domainname);
  xdomainname = (char *) alloca (strlen (categoryname)
				 + domainname_len + 5);
  ADD_BLOCK (block_list, xdomainname);

  stpcpy (mempcpy (stpcpy (stpcpy (xdomainname, categoryname), "/"),
		  domainname, domainname_len),
	  ".mo");

  /* Creating working area.  */
  single_locale = (char *) alloca (strlen (categoryvalue) + 1);
  ADD_BLOCK (block_list, single_locale);


  /* Search for the given string.  This is a loop because we perhaps
     got an ordered list of languages to consider for the translation.  */
  while (1)
    {
      /* Make CATEGORYVALUE point to the next element of the list.  */
      while (categoryvalue[0] != '\0' && categoryvalue[0] == ':')
	++categoryvalue;
      if (categoryvalue[0] == '\0')
	{
	  /* The whole contents of CATEGORYVALUE has been searched but
	     no valid entry has been found.  We solve this situation
	     by implicitly appending a "C" entry, i.e. no translation
	     will take place.  */
	  single_locale[0] = 'C';
	  single_locale[1] = '\0';
	}
      else
	{
	  char *cp = single_locale;
	  while (categoryvalue[0] != '\0' && categoryvalue[0] != ':')
	    *cp++ = *categoryvalue++;
	  *cp = '\0';

	  /* When this is a SUID binary we must not allow accessing files
	     outside the dedicated directories.  */
	  if (ENABLE_SECURE && strchr (single_locale, '/') != NULL)
	    /* Ingore this entry.  */
	    continue;
	}

      /* If the current locale value is C (or POSIX) we don't load a
	 domain.  Return the MSGID.  */
      if (strcmp (single_locale, "C") == 0
	  || strcmp (single_locale, "POSIX") == 0)
	{
	no_translation:
	  FREE_BLOCKS (block_list);
	  __libc_rwlock_unlock (__libc_setlocale_lock);
	  __libc_rwlock_unlock (_nl_state_lock);
	  __set_errno (saved_errno);
	  return (plural == 0
		  ? (char *) msgid1
		  /* Use the Germanic plural rule.  */
		  : n == 1 ? (char *) msgid1 : (char *) msgid2);
	}


      /* Find structure describing the message catalog matching the
	 DOMAINNAME and CATEGORY.  */
      domain = _nl_find_domain (dirname, single_locale, xdomainname, binding);

      if (domain != NULL)
	{
	  retval = _nl_find_msg (domain, binding, msgid1, 1, &retlen);

	  if (retval == NULL)
	    {
	      int cnt;

	      for (cnt = 0; domain->successor[cnt] != NULL; ++cnt)
		{
		  retval = _nl_find_msg (domain->successor[cnt], binding,
					 msgid1, 1, &retlen);

		  /* Resource problems are not fatal, instead we return no
		     translation.  */
		  if (__glibc_unlikely (retval == (char *) -1))
		    goto no_translation;

		  if (retval != NULL)
		    {
		      domain = domain->successor[cnt];
		      break;
		    }
		}
	    }

	  /* Returning -1 means that some resource problem exists
	     (likely memory) and that the strings could not be
	     converted.  Return the original strings.  */
	  if (__glibc_unlikely (retval == (char *) -1))
	    goto no_translation;

	  if (retval != NULL)
	    {
	      /* Found the translation of MSGID1 in domain DOMAIN:
		 starting at RETVAL, RETLEN bytes.  */
	      FREE_BLOCKS (block_list);
#if defined HAVE_TSEARCH || defined _LIBC
	      if (foundp == NULL)
		{
		  /* Create a new entry and add it to the search tree.  */
		  size_t msgid_len;
		  size_t size;
		  struct known_translation_t *newp;

		  msgid_len = strlen (msgid1) + 1;
		  size = offsetof (struct known_translation_t, msgid)
			 + msgid_len + domainname_len + 1;
# ifdef HAVE_PER_THREAD_LOCALE
		  size += strlen (localename) + 1;
# endif
		  newp = (struct known_translation_t *) malloc (size);
		  if (newp != NULL)
		    {
		      char *new_domainname;
# ifdef HAVE_PER_THREAD_LOCALE
		      char *new_localename;
# endif

		      new_domainname =
			mempcpy (newp->msgid.appended, msgid1, msgid_len);
		      memcpy (new_domainname, domainname, domainname_len + 1);
# ifdef HAVE_PER_THREAD_LOCALE
		      new_localename = new_domainname + domainname_len + 1;
		      strcpy (new_localename, localename);
# endif
		      newp->domainname = new_domainname;
		      newp->category = category;
# ifdef HAVE_PER_THREAD_LOCALE
		      newp->localename = new_localename;
# endif
		      newp->counter = _nl_msg_cat_cntr;
		      newp->domain = domain;
		      newp->translation = retval;
		      newp->translation_length = retlen;

		      __libc_rwlock_wrlock (tree_lock);

		      /* Insert the entry in the search tree.  */
		      foundp = (struct known_translation_t **)
			tsearch (newp, &root, transcmp);

		      __libc_rwlock_unlock (tree_lock);

		      if (foundp == NULL
			  || __builtin_expect (*foundp != newp, 0))
			/* The insert failed.  */
			free (newp);
		    }
		}
	      else
		{
		  /* We can update the existing entry.  */
		  (*foundp)->counter = _nl_msg_cat_cntr;
		  (*foundp)->domain = domain;
		  (*foundp)->translation = retval;
		  (*foundp)->translation_length = retlen;
		}
#endif
	      __set_errno (saved_errno);

	      /* Now deal with plural.  */
	      if (plural)
		retval = plural_lookup (domain, n, retval, retlen);

	      __libc_rwlock_unlock (__libc_setlocale_lock);
	      __libc_rwlock_unlock (_nl_state_lock);
	      return retval;
	    }
	}
    }
  /* NOTREACHED */
}


char *
internal_function
_nl_find_msg (domain_file, domainbinding, msgid, convert, lengthp)
     struct loaded_l10nfile *domain_file;
     struct binding *domainbinding;
     const char *msgid;
     int convert;
     size_t *lengthp;
{
  struct loaded_domain *domain;
  nls_uint32 nstrings;
  size_t act;
  char *result;
  size_t resultlen;

  if (domain_file->decided <= 0)
    _nl_load_domain (domain_file, domainbinding);

  if (domain_file->data == NULL)
    return NULL;

  domain = (struct loaded_domain *) domain_file->data;

  nstrings = domain->nstrings;

  /* Locate the MSGID and its translation.  */
  if (domain->hash_tab != NULL)
    {
      /* Use the hashing table.  */
      nls_uint32 len = strlen (msgid);
      nls_uint32 hash_val = __hash_string (msgid);
      nls_uint32 idx = hash_val % domain->hash_size;
      nls_uint32 incr = 1 + (hash_val % (domain->hash_size - 2));

      while (1)
	{
	  nls_uint32 nstr =
	    W (domain->must_swap_hash_tab, domain->hash_tab[idx]);

	  if (nstr == 0)
	    /* Hash table entry is empty.  */
	    return NULL;

	  nstr--;

	  /* Compare msgid with the original string at index nstr.
	     We compare the lengths with >=, not ==, because plural entries
	     are represented by strings with an embedded NUL.  */
	  if (nstr < nstrings
	      ? W (domain->must_swap, domain->orig_tab[nstr].length) >= len
		&& (strcmp (msgid,
			    domain->data + W (domain->must_swap,
					      domain->orig_tab[nstr].offset))
		    == 0)
	      : domain->orig_sysdep_tab[nstr - nstrings].length > len
		&& (strcmp (msgid,
			    domain->orig_sysdep_tab[nstr - nstrings].pointer)
		    == 0))
	    {
	      act = nstr;
	      goto found;
	    }

	  if (idx >= domain->hash_size - incr)
	    idx -= domain->hash_size - incr;
	  else
	    idx += incr;
	}
      /* NOTREACHED */
    }
  else
    {
      /* Try the default method:  binary search in the sorted array of
	 messages.  */
      size_t top, bottom;

      bottom = 0;
      top = nstrings;
      while (bottom < top)
	{
	  int cmp_val;

	  act = (bottom + top) / 2;
	  cmp_val = strcmp (msgid, (domain->data
				    + W (domain->must_swap,
					 domain->orig_tab[act].offset)));
	  if (cmp_val < 0)
	    top = act;
	  else if (cmp_val > 0)
	    bottom = act + 1;
	  else
	    goto found;
	}
      /* No translation was found.  */
      return NULL;
    }

 found:
  /* The translation was found at index ACT.  If we have to convert the
     string to use a different character set, this is the time.  */
  if (act < nstrings)
    {
      result = (char *)
	(domain->data + W (domain->must_swap, domain->trans_tab[act].offset));
      resultlen = W (domain->must_swap, domain->trans_tab[act].length) + 1;
    }
  else
    {
      result = (char *) domain->trans_sysdep_tab[act - nstrings].pointer;
      resultlen = domain->trans_sysdep_tab[act - nstrings].length;
    }

#if defined _LIBC || HAVE_ICONV
  if (convert)
    {
      /* We are supposed to do a conversion.  */
      const char *encoding = get_output_charset (domainbinding);

      /* Protect against reallocation of the table.  */
      __libc_rwlock_rdlock (domain->conversions_lock);

      /* Search whether a table with converted translations for this
	 encoding has already been allocated.  */
      size_t nconversions = domain->nconversions;
      struct converted_domain *convd = NULL;
      size_t i;

      for (i = nconversions; i > 0; )
	{
	  i--;
	  if (strcmp (domain->conversions[i].encoding, encoding) == 0)
	    {
	      convd = &domain->conversions[i];
	      break;
	    }
	}

      __libc_rwlock_unlock (domain->conversions_lock);

      if (convd == NULL)
	{
	  /* We have to allocate a new conversions table.  */
	  __libc_rwlock_wrlock (domain->conversions_lock);
	  nconversions = domain->nconversions;

	  /* Maybe in the meantime somebody added the translation.
	     Recheck.  */
	  for (i = nconversions; i > 0; )
	    {
	      i--;
	      if (strcmp (domain->conversions[i].encoding, encoding) == 0)
		{
		  convd = &domain->conversions[i];
		  goto found_convd;
		}
	    }

	  /* Allocate a table for the converted translations for this
	     encoding.  */
	  struct converted_domain *new_conversions =
	    (struct converted_domain *)
	    realloc (domain->conversions,
		     (nconversions + 1) * sizeof (struct converted_domain));

	  if (__glibc_unlikely (new_conversions == NULL))
	    {
	      /* Nothing we can do, no more memory.  We cannot use the
		 translation because it might be encoded incorrectly.  */
	    unlock_fail:
	      __libc_rwlock_unlock (domain->conversions_lock);
	      return (char *) -1;
	    }

	  domain->conversions = new_conversions;

	  /* Copy the 'encoding' string to permanent storage.  */
	  encoding = strdup (encoding);
	  if (__glibc_unlikely (encoding == NULL))
	    /* Nothing we can do, no more memory.  We cannot use the
	       translation because it might be encoded incorrectly.  */
	    goto unlock_fail;

	  convd = &new_conversions[nconversions];
	  convd->encoding = encoding;

	  /* Find out about the character set the file is encoded with.
	     This can be found (in textual form) in the entry "".  If this
	     entry does not exist or if this does not contain the 'charset='
	     information, we will assume the charset matches the one the
	     current locale and we don't have to perform any conversion.  */
# ifdef _LIBC
	  convd->conv = (__gconv_t) -1;
# else
#  if HAVE_ICONV
	  convd->conv = (iconv_t) -1;
#  endif
# endif
	  {
	    char *nullentry;
	    size_t nullentrylen;

	    /* Get the header entry.  This is a recursion, but it doesn't
	       reallocate domain->conversions because we pass convert = 0.  */
	    nullentry =
	      _nl_find_msg (domain_file, domainbinding, "", 0, &nullentrylen);

	    /* Resource problems are fatal.  If we continue onwards we will
	       only attempt to calloc a new conv_tab and fail later.  */
	    if (__glibc_unlikely (nullentry == (char *) -1))
	      return (char *) -1;

	    if (nullentry != NULL)
	      {
		const char *charsetstr;

		charsetstr = strstr (nullentry, "charset=");
		if (charsetstr != NULL)
		  {
		    size_t len;
		    char *charset;
		    const char *outcharset;

		    charsetstr += strlen ("charset=");
		    len = strcspn (charsetstr, " \t\n");

		    charset = (char *) alloca (len + 1);
# if defined _LIBC || HAVE_MEMPCPY
		    *((char *) mempcpy (charset, charsetstr, len)) = '\0';
# else
		    memcpy (charset, charsetstr, len);
		    charset[len] = '\0';
# endif

		    outcharset = encoding;

# ifdef _LIBC
		    /* We always want to use transliteration.  */
		    outcharset = norm_add_slashes (outcharset, "TRANSLIT");
		    charset = norm_add_slashes (charset, "");
		    int r = __gconv_open (outcharset, charset, &convd->conv,
					  GCONV_AVOID_NOCONV);
		    if (__glibc_unlikely (r != __GCONV_OK))
		      {
			/* If the output encoding is the same there is
			   nothing to do.  Otherwise do not use the
			   translation at all.  */
			if (__glibc_likely (r != __GCONV_NULCONV))
			  {
			    __libc_rwlock_unlock (domain->conversions_lock);
			    free ((char *) encoding);
			    return NULL;
			  }

			convd->conv = (__gconv_t) -1;
		      }
# else
#  if HAVE_ICONV
		    /* When using GNU libc >= 2.2 or GNU libiconv >= 1.5,
		       we want to use transliteration.  */
#   if (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 2) || __GLIBC__ > 2 \
       || _LIBICONV_VERSION >= 0x0105
		    if (strchr (outcharset, '/') == NULL)
		      {
			char *tmp;

			len = strlen (outcharset);
			tmp = (char *) alloca (len + 10 + 1);
			memcpy (tmp, outcharset, len);
			memcpy (tmp + len, "//TRANSLIT", 10 + 1);
			outcharset = tmp;

			convd->conv = iconv_open (outcharset, charset);

			freea (outcharset);
		      }
		    else
#   endif
		      convd->conv = iconv_open (outcharset, charset);
#  endif
# endif

		    freea (charset);
		  }
	      }
	  }
	  convd->conv_tab = NULL;
	  /* Here domain->conversions is still == new_conversions.  */
	  domain->nconversions++;

	found_convd:
	  __libc_rwlock_unlock (domain->conversions_lock);
	}

      if (
# ifdef _LIBC
	  convd->conv != (__gconv_t) -1
# else
#  if HAVE_ICONV
	  convd->conv != (iconv_t) -1
#  endif
# endif
	  )
	{
	  __libc_lock_define_initialized (static, lock)
	  /* We are supposed to do a conversion.  First allocate an
	     appropriate table with the same structure as the table
	     of translations in the file, where we can put the pointers
	     to the converted strings in.
	     There is a slight complication with plural entries.  They
	     are represented by consecutive NUL terminated strings.  We
	     handle this case by converting RESULTLEN bytes, including
	     NULs.  */

	  if (__glibc_unlikely (convd->conv_tab == NULL))
	    {
	      __libc_lock_lock (lock);
	      if (convd->conv_tab == NULL)
		{
		  convd->conv_tab
		    = calloc (nstrings + domain->n_sysdep_strings,
			      sizeof (char *));
		  if (convd->conv_tab != NULL)
		    goto not_translated_yet;
		  /* Mark that we didn't succeed allocating a table.  */
		  convd->conv_tab = (char **) -1;
		}
	      __libc_lock_unlock (lock);
	    }

	  if (__glibc_unlikely (convd->conv_tab == (char **) -1))
	    /* Nothing we can do, no more memory.  We cannot use the
	       translation because it might be encoded incorrectly.  */
	    return (char *) -1;

	  if (convd->conv_tab[act] == NULL)
	    {
	      __libc_lock_lock (lock);
	    not_translated_yet:;

	      /* We haven't used this string so far, so it is not
		 translated yet.  Do this now.  */
	      /* We use a bit more efficient memory handling.
		 We allocate always larger blocks which get used over
		 time.  This is faster than many small allocations.   */
# define INITIAL_BLOCK_SIZE	4080
	      static unsigned char *freemem;
	      static size_t freemem_size;

	      const unsigned char *inbuf;
	      unsigned char *outbuf;
	      int malloc_count;
# ifndef _LIBC
	      transmem_block_t *transmem_list = NULL;
# endif

	      inbuf = (const unsigned char *) result;
	      outbuf = freemem + sizeof (size_t);

	      malloc_count = 0;
	      while (1)
		{
		  transmem_block_t *newmem;
# ifdef _LIBC
		  size_t non_reversible;
		  int res;

		  if (freemem_size < sizeof (size_t))
		    goto resize_freemem;

		  res = __gconv (convd->conv,
				 &inbuf, inbuf + resultlen,
				 &outbuf,
				 outbuf + freemem_size - sizeof (size_t),
				 &non_reversible);

		  if (res == __GCONV_OK || res == __GCONV_EMPTY_INPUT)
		    break;

		  if (res != __GCONV_FULL_OUTPUT)
		    {
		      /* We should not use the translation at all, it
			 is incorrectly encoded.  */
		      __libc_lock_unlock (lock);
		      return NULL;
		    }

		  inbuf = (const unsigned char *) result;
# else
#  if HAVE_ICONV
		  const char *inptr = (const char *) inbuf;
		  size_t inleft = resultlen;
		  char *outptr = (char *) outbuf;
		  size_t outleft;

		  if (freemem_size < sizeof (size_t))
		    goto resize_freemem;

		  outleft = freemem_size - sizeof (size_t);
		  if (iconv (convd->conv,
			     (ICONV_CONST char **) &inptr, &inleft,
			     &outptr, &outleft)
		      != (size_t) (-1))
		    {
		      outbuf = (unsigned char *) outptr;
		      break;
		    }
		  if (errno != E2BIG)
		    {
		      __libc_lock_unlock (lock);
		      return NULL;
		    }
#  endif
# endif

		resize_freemem:
		  /* We must allocate a new buffer or resize the old one.  */
		  if (malloc_count > 0)
		    {
		      ++malloc_count;
		      freemem_size = malloc_count * INITIAL_BLOCK_SIZE;
		      newmem = (transmem_block_t *) realloc (transmem_list,
							     freemem_size);
# ifdef _LIBC
		      if (newmem != NULL)
			transmem_list = newmem;
		      else
			{
			  struct transmem_list *old = transmem_list;

			  transmem_list = transmem_list->next;
			  free (old);
			}
# endif
		    }
		  else
		    {
		      malloc_count = 1;
		      freemem_size = INITIAL_BLOCK_SIZE;
		      newmem = (transmem_block_t *) malloc (freemem_size);
# ifdef _LIBC
		      if (newmem != NULL)
			{
			  /* Add the block to the list of blocks we have to free
			     at some point.  */
			  newmem->next = transmem_list;
			  transmem_list = newmem;
			}
		      /* Fall through and return -1.  */
# endif
		    }
		  if (__glibc_unlikely (newmem == NULL))
		    {
		      freemem = NULL;
		      freemem_size = 0;
		      __libc_lock_unlock (lock);
		      return (char *) -1;
		    }

# ifdef _LIBC
		  freemem = (unsigned char *) newmem->data;
		  freemem_size -= offsetof (struct transmem_list, data);
# else
		  transmem_list = newmem;
		  freemem = newmem;
# endif

		  outbuf = freemem + sizeof (size_t);
		}

	      /* We have now in our buffer a converted string.  Put this
		 into the table of conversions.  */
	      *(size_t *) freemem = outbuf - freemem - sizeof (size_t);
	      convd->conv_tab[act] = (char *) freemem;
	      /* Shrink freemem, but keep it aligned.  */
	      freemem_size -= outbuf - freemem;
	      freemem = outbuf;
	      freemem += freemem_size & (alignof (size_t) - 1);
	      freemem_size = freemem_size & ~ (alignof (size_t) - 1);

	      __libc_lock_unlock (lock);
	    }

	  /* Now convd->conv_tab[act] contains the translation of all
	     the plural variants.  */
	  result = convd->conv_tab[act] + sizeof (size_t);
	  resultlen = *(size_t *) convd->conv_tab[act];
	}
    }

  /* The result string is converted.  */

#endif /* _LIBC || HAVE_ICONV */

  *lengthp = resultlen;
  return result;
}


/* Look up a plural variant.  */
static char *
internal_function
plural_lookup (domain, n, translation, translation_len)
     struct loaded_l10nfile *domain;
     unsigned long int n;
     const char *translation;
     size_t translation_len;
{
  struct loaded_domain *domaindata = (struct loaded_domain *) domain->data;
  unsigned long int index;
  const char *p;

  index = plural_eval (domaindata->plural, n);
  if (index >= domaindata->nplurals)
    /* This should never happen.  It means the plural expression and the
       given maximum value do not match.  */
    index = 0;

  /* Skip INDEX strings at TRANSLATION.  */
  p = translation;
  while (index-- > 0)
    {
#ifdef _LIBC
      p = __rawmemchr (p, '\0');
#else
      p = strchr (p, '\0');
#endif
      /* And skip over the NUL byte.  */
      p++;

      if (p >= translation + translation_len)
	/* This should never happen.  It means the plural expression
	   evaluated to a value larger than the number of variants
	   available for MSGID1.  */
	return (char *) translation;
    }
  return (char *) p;
}

#ifndef _LIBC
/* Return string representation of locale CATEGORY.  */
static const char *
internal_function
category_to_name (category)
     int category;
{
  const char *retval;

  switch (category)
  {
#ifdef LC_COLLATE
  case LC_COLLATE:
    retval = "LC_COLLATE";
    break;
#endif
#ifdef LC_CTYPE
  case LC_CTYPE:
    retval = "LC_CTYPE";
    break;
#endif
#ifdef LC_MONETARY
  case LC_MONETARY:
    retval = "LC_MONETARY";
    break;
#endif
#ifdef LC_NUMERIC
  case LC_NUMERIC:
    retval = "LC_NUMERIC";
    break;
#endif
#ifdef LC_TIME
  case LC_TIME:
    retval = "LC_TIME";
    break;
#endif
#ifdef LC_MESSAGES
  case LC_MESSAGES:
    retval = "LC_MESSAGES";
    break;
#endif
#ifdef LC_RESPONSE
  case LC_RESPONSE:
    retval = "LC_RESPONSE";
    break;
#endif
#ifdef LC_ALL
  case LC_ALL:
    /* This might not make sense but is perhaps better than any other
       value.  */
    retval = "LC_ALL";
    break;
#endif
  default:
    /* If you have a better idea for a default value let me know.  */
    retval = "LC_XXX";
  }

  return retval;
}
#endif

/* Guess value of current locale from value of the environment variables.  */
static const char *
internal_function
guess_category_value (category, categoryname)
     int category;
     const char *categoryname;
{
  const char *language;
  const char *retval;

  /* The highest priority value is the `LANGUAGE' environment
     variable.  But we don't use the value if the currently selected
     locale is the C locale.  This is a GNU extension.  */
  language = getenv ("LANGUAGE");
  if (language != NULL && language[0] == '\0')
    language = NULL;

  /* We have to proceed with the POSIX methods of looking to `LC_ALL',
     `LC_xxx', and `LANG'.  On some systems this can be done by the
     `setlocale' function itself.  */
#ifdef _LIBC
  retval = __current_locale_name (category);
#else
  retval = _nl_locale_name (category, categoryname);
#endif

  return language != NULL && strcmp (retval, "C") != 0 ? language : retval;
}

#if defined _LIBC || HAVE_ICONV
/* Returns the output charset.  */
static const char *
internal_function
get_output_charset (domainbinding)
     struct binding *domainbinding;
{
  /* The output charset should normally be determined by the locale.  But
     sometimes the locale is not used or not correctly set up, so we provide
     a possibility for the user to override this: the OUTPUT_CHARSET
     environment variable.  Moreover, the value specified through
     bind_textdomain_codeset overrides both.  */
  if (domainbinding != NULL && domainbinding->codeset != NULL)
    return domainbinding->codeset;
  else
    {
      /* For speed reasons, we look at the value of OUTPUT_CHARSET only
	 once.  This is a user variable that is not supposed to change
	 during a program run.  */
      static char *output_charset_cache;
      static int output_charset_cached;

      if (!output_charset_cached)
	{
	  const char *value = getenv ("OUTPUT_CHARSET");

	  if (value != NULL && value[0] != '\0')
	    {
	      size_t len = strlen (value) + 1;
	      char *value_copy = (char *) malloc (len);

	      if (value_copy != NULL)
		memcpy (value_copy, value, len);
	      output_charset_cache = value_copy;
	    }
	  output_charset_cached = 1;
	}

      if (output_charset_cache != NULL)
	return output_charset_cache;
      else
	{
# ifdef _LIBC
	  return _NL_CURRENT (LC_CTYPE, CODESET);
# else
#  if HAVE_ICONV
	  extern const char *locale_charset PARAMS ((void));
	  return locale_charset ();
#  endif
# endif
	}
    }
}
#endif

/* @@ begin of epilog @@ */

/* We don't want libintl.a to depend on any other library.  So we
   avoid the non-standard function stpcpy.  In GNU C Library this
   function is available, though.  Also allow the symbol HAVE_STPCPY
   to be defined.  */
#if !_LIBC && !HAVE_STPCPY
static char *
stpcpy (dest, src)
     char *dest;
     const char *src;
{
  while ((*dest++ = *src++) != '\0')
    /* Do nothing. */ ;
  return dest - 1;
}
#endif

#if !_LIBC && !HAVE_MEMPCPY
static void *
mempcpy (dest, src, n)
     void *dest;
     const void *src;
     size_t n;
{
  return (void *) ((char *) memcpy (dest, src, n) + n);
}
#endif


#ifdef _LIBC
/* If we want to free all resources we have to do some work at
   program's end.  */
libc_freeres_fn (free_mem)
{
  void *old;

  while (_nl_domain_bindings != NULL)
    {
      struct binding *oldp = _nl_domain_bindings;
      _nl_domain_bindings = _nl_domain_bindings->next;
      if (oldp->dirname != _nl_default_dirname)
	/* Yes, this is a pointer comparison.  */
	free (oldp->dirname);
      free (oldp->codeset);
      free (oldp);
    }

  if (_nl_current_default_domain != _nl_default_default_domain)
    /* Yes, again a pointer comparison.  */
    free ((char *) _nl_current_default_domain);

  /* Remove the search tree with the known translations.  */
  __tdestroy (root, free);
  root = NULL;

  while (transmem_list != NULL)
    {
      old = transmem_list;
      transmem_list = transmem_list->next;
      free (old);
    }
}
#endif