libintl.cpp

PIXIL is a small footprint operating environment, complete with PDA PIM applications, a browser and

#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

/* XPG3 defines the result of `setlocale (category, NULL)' as:
   ``Directs `setlocale()' to query `category' and return the current
     setting of `local'.''
   However it does not specify the exact format.  And even worse: POSIX
   defines this not at all.  So we can use this feature only on selected
   system (e.g. those using GNU C Library).  */

/* 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[] = "messages";

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

/* Contains the default location of the message catalogs.  */
const char k_nl_default_dirname[] = GNULOCALEDIR;

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

/* Prototypes for local functions.  */
static char *find_msg  (struct loaded_l10nfile *domain_file,
			const char *msgid);

/* 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 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);							      \
    }									      \
  } while (0)
# undef alloca
# define alloca(size) (malloc (size))
#endif	/* have alloca */

/* Look up MSGID in the DOMAINNAME message catalog for the current CATEGORY
   locale.  */
char *
k_dcgettext (const char *domainname, const char *msgid, const char *categoryvalue)
{
#ifndef HAVE_ALLOCA
  struct block_list *block_list = NULL;
#endif
  struct loaded_l10nfile *domain;
  struct binding *binding;
  char *dirname, *xdomainname;
  const char *categoryname;  
  char *single_locale;
  char *retval;
  int saved_errno = errno;
  const char *_domainname = (domainname == 0L) ? k_nl_current_default_domain : domainname;     

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

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

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

      __set_errno (0);
      while ((ret = getcwd (dirname, path_max)) == NULL && errno == ERANGE)
	{
	  path_max += PATH_INCR;
	  dirname = (char *) alloca (path_max + dirname_len);
	  ADD_BLOCK (block_list, dirname);
	  __set_errno (0);
	}

      if (ret == NULL)
	{
	  /* We cannot get the current working directory.  Don't signal an
	     error but simply return the default string.  */
	  FREE_BLOCKS (block_list);
	  __set_errno (saved_errno);
	  return (char *) msgid;
	}

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

  /* Now determine the symbolic name of CATEGORY and its value.  */
  /* Stephan: we don't use this
     categoryname = category_to_name (category);
     categoryvalue = guess_category_value (category, categoryname);
  */
  categoryname = "LC_MESSAGES";

  xdomainname = (char *) alloca (strlen (categoryname)
				 + strlen (_domainname) + 5);
  ADD_BLOCK (block_list, xdomainname);
  /* 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.  */
  stpcpy (stpcpy (stpcpy (stpcpy (xdomainname, categoryname), "/"),
		  _domainname),
	  ".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 th 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';
	}

      /* 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)
	{
	  FREE_BLOCKS (block_list);
	  __set_errno (saved_errno);
	  return (char *) msgid;
	}


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

      if (domain != NULL)
	{
	  retval = find_msg (domain, msgid);

	  if (retval == NULL)
	    {
	      int cnt;

	      for (cnt = 0; domain->successor[cnt] != NULL; ++cnt)
		{
		  retval = find_msg (domain->successor[cnt], msgid);

		  if (retval != NULL)
		    break;
		}
	    }

	  if (retval != NULL)
	    {
	      FREE_BLOCKS (block_list);
	      __set_errno (saved_errno);
	      return retval;
	    }
	}
    }
  /* NOTREACHED */
}

static char *
find_msg (struct loaded_l10nfile *domain_file, const char *msgid)
{
  size_t top, act, bottom;
  struct loaded_domain *domain;

  if (domain_file->decided == 0)
    k_nl_load_domain (domain_file);

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

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

  /* Locate the MSGID and its translation.  */
  if (domain->hash_size > 2 && 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));
      nls_uint32 nstr = W (domain->must_swap, domain->hash_tab[idx]);

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

      if (W (domain->must_swap, domain->orig_tab[nstr - 1].length) == len
	  && strcmp (msgid,
		     domain->data + W (domain->must_swap,
				       domain->orig_tab[nstr - 1].offset)) == 0)
	return (char *) domain->data + W (domain->must_swap,
					  domain->trans_tab[nstr - 1].offset);

      while (1)
	{
	  if (idx >= domain->hash_size - incr)
	    idx -= domain->hash_size - incr;
	  else
	    idx += incr;

	  nstr = W (domain->must_swap, domain->hash_tab[idx]);
	  if (nstr == 0)
	    /* Hash table entry is empty.  */
	    return NULL;

	  if (W (domain->must_swap, domain->orig_tab[nstr - 1].length) == len
	      && strcmp (msgid,
			 domain->data + W (domain->must_swap,
					   domain->orig_tab[nstr - 1].offset))
	         == 0)
	    return (char *) domain->data
	      + W (domain->must_swap, domain->trans_tab[nstr - 1].offset);
	}
      /* NOTREACHED */
    }

  /* Now we try the default method:  binary search in the sorted
     array of messages.  */
  bottom = 0;
  top = domain->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
	break;
    }

  /* If an translation is found return this.  */
  return bottom >= top ? NULL : (char *) domain->data
                                + W (domain->must_swap,
				     domain->trans_tab[act].offset);
}

/* @@ 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 !HAVE_STPCPY
static char *
stpcpy (char *dest, const char *src)
{
  while ((*dest++ = *src++) != '\0')
    /* Do nothing. */ ;
  return dest - 1;
}
#endif


int
_nl_explode_name (char * name, const char **language, 
		  const char **modifier, const char **territory, 
		  const char **codeset, 
		  const char **normalized_codeset, 
		  const char **special, const char **sponsor, 
		  const char **revision)
     
{
  enum { undecided, xpg, cen } syntax;
  char *cp;
  int mask;

  *modifier = NULL;
  *territory = NULL;
  *codeset = NULL;
  *normalized_codeset = NULL;
  *special = NULL;
  *sponsor = NULL;
  *revision = NULL;

  /* Now we determine the single parts of the locale name.  First
     look for the language.  Termination symbols are `_' and `@' if
     we use XPG4 style, and `_', `+', and `,' if we use CEN syntax.  */
  mask = 0;
  syntax = undecided;
  *language = cp = name;
  while (cp[0] != '\0' && cp[0] != '_' && cp[0] != '@'
	 && cp[0] != '+' && cp[0] != ',')
    ++cp;

  if (*language == cp)
    /* This does not make sense: language has to be specified.  Use
       this entry as it is without exploding.  Perhaps it is an alias.  */
    cp = strchr (*language, '\0');
  else if (cp[0] == '_')
    {
      /* Next is the territory.  */
      cp[0] = '\0';
      *territory = ++cp;

      while (cp[0] != '\0' && cp[0] != '.' && cp[0] != '@'
	     && cp[0] != '+' && cp[0] != ',' && cp[0] != '_')
	++cp;

      mask |= TERRITORY;

      if (cp[0] == '.')
	{
	  /* Next is the codeset.  */
	  syntax = xpg;
	  cp[0] = '\0';
	  *codeset = ++cp;

	  while (cp[0] != '\0' && cp[0] != '@')
	    ++cp;

	  mask |= XPG_CODESET;

	  if (*codeset != cp && (*codeset)[0] != '\0')
	    {
	      *normalized_codeset = _nl_normalize_codeset (*codeset,
							   cp - *codeset);
	      if (strcmp (*codeset, *normalized_codeset) == 0)
		free ((char *) *normalized_codeset);
	      else
		mask |= XPG_NORM_CODESET;
	    }
	}
    }

  if (cp[0] == '@' || (syntax != xpg && cp[0] == '+'))
    {
      /* Next is the modifier.  */
      syntax = cp[0] == '@' ? xpg : cen;
      cp[0] = '\0';
      *modifier = ++cp;

      while (syntax == cen && cp[0] != '\0' && cp[0] != '+'
	     && cp[0] != ',' && cp[0] != '_')
	++cp;

      mask |= XPG_MODIFIER | CEN_AUDIENCE;
    }

  if (syntax != xpg && (cp[0] == '+' || cp[0] == ',' || cp[0] == '_'))
    {
      syntax = cen;

      if (cp[0] == '+')
	{
 	  /* Next is special application (CEN syntax).  */
	  cp[0] = '\0';
	  *special = ++cp;

	  while (cp[0] != '\0' && cp[0] != ',' && cp[0] != '_')
	    ++cp;

	  mask |= CEN_SPECIAL;
	}

      if (cp[0] == ',')
	{
 	  /* Next is sponsor (CEN syntax).  */
	  cp[0] = '\0';
	  *sponsor = ++cp;

	  while (cp[0] != '\0' && cp[0] != '_')
	    ++cp;

	  mask |= CEN_SPONSOR;
	}