libs-base/Source/Additions/Unicode.m

/** Support functions for Unicode implementation
   Function to determine default c string encoding for
   GNUstep based on GNUSTEP_STRING_ENCODING environment variable.

   Copyright (C) 1997 Free Software Foundation, Inc.

   Written by: Stevo Crvenkovski < stevo@btinternet.com >
   Date: March 1997
   Merged with GetDefEncoding.m and iconv by: Fred Kiefer <fredkiefer@gmx.de>
   Date: September 2000
   Rewrite by: Richard Frith-Macdonald <rfm@gnu.org>

   This file is part of the GNUstep Base Library.

   This 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 3 of the License, or (at your option) any later version.

   This 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
   Library General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with this library; if not, write to the Free
   Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
   Boston, MA 02111 USA.
*/

#include "config.h"
#ifndef NeXT_Foundation_LIBRARY
#include <Foundation/NSArray.h>
#include <Foundation/NSBundle.h>
#include <Foundation/NSDictionary.h>
#include <Foundation/NSError.h>
#include <Foundation/NSException.h>
#include <Foundation/NSString.h>
#include <Foundation/NSLock.h>
#include <Foundation/NSPathUtilities.h>
#else
#include <Foundation/Foundation.h>
#endif

#include "GNUstepBase/GSLock.h"
#include "GNUstepBase/GSMime.h"
#include "GNUstepBase/GSCategories.h"
#include "GNUstepBase/Unicode.h"

#include "../GSPrivate.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#if HAVE_LANGINFO_CODESET
#include <langinfo.h>
#endif

typedef struct {unichar from; unsigned char to;} _ucc_;

#include "unicode/cyrillic.h"
#include "unicode/latin2.h"
#include "unicode/latin9.h"
#include "unicode/nextstep.h"
#include "unicode/caseconv.h"
#include "unicode/cop.h"
#include "unicode/decomp.h"
#include "unicode/gsm0338.h"
#include "unicode/thai.h"

#ifdef HAVE_ICONV
#ifdef HAVE_GICONV_H
#include <giconv.h>
#else
#include <iconv.h>
#endif
#include <errno.h>

/*
 * The whole of the GNUstep code stores UNICODE in internal byte order,
 * so we do the same. We have switched to using UTF16 so the defines here
 * recognise this. We use the endian specific versions of UTF16 so that
 * iconv does not introduce a BOM where we do not want it.
 * If UTF16 does not work, we revert to UCS-2-INTERNAL.
 */
#ifdef WORDS_BIGENDIAN
#define UNICODE_UTF16 "UTF-16BE"
#define UNICODE_INT "UNICODEBIG"
#else
#define UNICODE_UTF16 "UTF-16LE"
#define UNICODE_INT "UNICODELITTLE"
#endif

#define UNICODE_ENC ((unicode_enc) ? unicode_enc : internal_unicode_enc())

static const char *unicode_enc = NULL;

/* Check to see what type of internal unicode format the library supports */
static const char *
internal_unicode_enc(void)
{
  iconv_t conv;

  unicode_enc = UNICODE_UTF16;
  conv = iconv_open(unicode_enc, "ASCII");
  if (conv != (iconv_t)-1)
    {
      iconv_close(conv);
      return unicode_enc;
    }
  fprintf(stderr, "Could not initialise iconv() for UTF16, using UCS-2\n");
  fprintf(stderr, "Using characters outside 16 bits may give bad results.\n");

  unicode_enc = UNICODE_INT;
  conv = iconv_open(unicode_enc, "ASCII");
  if (conv != (iconv_t)-1)
    {
      iconv_close(conv);
      return unicode_enc;
    }
  unicode_enc = "UCS-2-INTERNAL";
  conv = iconv_open(unicode_enc, "ASCII");
  if (conv != (iconv_t)-1)
    {
      iconv_close(conv);
      return unicode_enc;
    }
  unicode_enc = "UCS-2";
  /* This had better work */
  return unicode_enc;
}

#endif

static GSLazyLock *local_lock = nil;

typedef	unsigned char	unc;
static NSStringEncoding	defEnc = GSUndefinedEncoding;
static NSStringEncoding	natEnc = GSUndefinedEncoding;
static NSStringEncoding *_availableEncodings = 0;

struct _strenc_ {
  NSStringEncoding	enc;		// Constant representing the encoding.
  const char		*ename;		// ASCII string representation of name.
  const char		*iconv;		/* Iconv name of encoding.  If this
					 * is the empty string, we cannot use
					 * iconv perform conversions to/from
					 * this encoding.
					 * NB. do not put a null pointer in this
					 * field in the table, use "" instread.
					 */
  BOOL			eightBit;	/* Flag to say whether this encoding
					 * can be stored in a byte array ...
					 * ie whether the encoding consists
					 * entirely of single byte characters
					 * and the first 128 are identical to
					 * the ASCII character set.
					 */
  char			supported;	/* Is this supported?  Some encodings
					 * have builtin conversion to/from
					 * unicode, but for others we must
					 * check with iconv to see if it
					 * supports them on this platform.
					 * A one means supported.
					 * A negative means unsupported.
					 * A zero means not yet checked.
					 */
  const char		*lossy;		/* Iconv name for lossy encoding */
};

/*
 * The str_encoding_table is a compact representation of all the string
 * encoding information we might need.  It gets modified at runtime.
 */
static struct _strenc_ str_encoding_table[] = {
  {NSASCIIStringEncoding,
    "NSASCIIStringEncoding","ASCII",1,1,0},
  {NSNEXTSTEPStringEncoding,
    "NSNEXTSTEPStringEncoding","NEXTSTEP",1,1,0},
  {NSJapaneseEUCStringEncoding,
    "NSJapaneseEUCStringEncoding","EUC-JP",0,0,0},
  {NSUTF8StringEncoding,
    "NSUTF8StringEncoding","UTF-8",0,1,0},
  {NSISOLatin1StringEncoding,
    "NSISOLatin1StringEncoding","ISO-8859-1",1,1,0},
  {NSSymbolStringEncoding,
    "NSSymbolStringEncoding","",0,0,0},
  {NSNonLossyASCIIStringEncoding,
    "NSNonLossyASCIIStringEncoding","",1,1,0},
  {NSShiftJISStringEncoding,
    "NSShiftJISStringEncoding","SHIFT-JIS",0,0,0},
  {NSISOLatin2StringEncoding,
    "NSISOLatin2StringEncoding","ISO-8859-2",1,1,0},
  {NSUnicodeStringEncoding,
    "NSUnicodeStringEncoding","",0,1,0},
  {NSWindowsCP1251StringEncoding,
    "NSWindowsCP1251StringEncoding","CP1251",0,0,0},
  {NSWindowsCP1252StringEncoding,
    "NSWindowsCP1252StringEncoding","CP1252",0,0,0},
  {NSWindowsCP1253StringEncoding,
    "NSWindowsCP1253StringEncoding","CP1253",0,0,0},
  {NSWindowsCP1254StringEncoding,
    "NSWindowsCP1254StringEncoding","CP1254",0,0,0},
  {NSWindowsCP1250StringEncoding,
    "NSWindowsCP1250StringEncoding","CP1250",0,0,0},
  {NSISO2022JPStringEncoding,
    "NSISO2022JPStringEncoding","ISO-2022-JP",0,0,0},
  {NSMacOSRomanStringEncoding,
    "NSMacOSRomanStringEncoding","MACINTOSH",0,0,0},
  {NSProprietaryStringEncoding,
    "NSProprietaryStringEncoding","",0,0,0},

// GNUstep additions
  {NSISOCyrillicStringEncoding,
    "NSISOCyrillicStringEncoding","ISO-8859-5",0,1,0},
  {NSKOI8RStringEncoding,
    "NSKOI8RStringEncoding","KOI8-R",0,0,0},
  {NSISOLatin3StringEncoding,
    "NSISOLatin3StringEncoding","ISO-8859-3",0,0,0},
  {NSISOLatin4StringEncoding,
    "NSISOLatin4StringEncoding","ISO-8859-4",0,0,0},
  {NSISOArabicStringEncoding,
    "NSISOArabicStringEncoding","ISO-8859-6",0,0,0},
  {NSISOGreekStringEncoding,
    "NSISOGreekStringEncoding","ISO-8859-7",0,0,0},
  {NSISOHebrewStringEncoding,
    "NSISOHebrewStringEncoding","ISO-8859-8",0,0,0},
  {NSISOLatin5StringEncoding,
    "NSISOLatin5StringEncoding","ISO-8859-9",0,0,0},
  {NSISOLatin6StringEncoding,
    "NSISOLatin6StringEncoding","ISO-8859-10",0,0,0},
  {NSISOThaiStringEncoding,
    "NSISOThaiStringEncoding","ISO-8859-11",1,1,0},
  {NSISOLatin7StringEncoding,
    "NSISOLatin7StringEncoding","ISO-8859-13",0,0,0},
  {NSISOLatin8StringEncoding,
    "NSISOLatin8StringEncoding","ISO-8859-14",0,0,0},
  {NSISOLatin9StringEncoding,
    "NSISOLatin9StringEncoding","ISO-8859-15",1,1,0},
  {NSUTF7StringEncoding,
    "NSUTF7StringEncoding","UTF-7",0,0,0},
  {NSGB2312StringEncoding,
    "NSGB2312StringEncoding","EUC-CN",0,0,0},
  {NSGSM0338StringEncoding,
    "NSGSM0338StringEncoding","",0,1,0},
  {NSBIG5StringEncoding,
    "NSBIG5StringEncoding","BIG5",0,0,0},
  {NSKoreanEUCStringEncoding,
    "NSKoreanEUCStringEncoding","EUC-KR",0,0,0},

  {0,"Unknown encoding","",0,0,0}
};

static struct _strenc_	**encodingTable = 0;
static unsigned		encTableSize = 0;

static void GSSetupEncodingTable(void)
{
  if (encodingTable == 0)
    {
      [GS_INITIALIZED_LOCK(local_lock, GSLazyLock) lock];
      if (encodingTable == 0)
	{
	  static struct _strenc_	**encTable = 0;
	  unsigned		count;
	  unsigned		i;

	  /*
	   * We want to store pointers to our string encoding info in a
	   * large table so we can do efficient lookup by encoding value.
	   */
#define	MAX_ENCODING	128
	  count = sizeof(str_encoding_table) / sizeof(struct _strenc_);

	  /*
	   * First determine the largest encoding value and create a
	   * large enough table of pointers.
	   */
	  encTableSize = 0;
	  for (i = 0; i < count; i++)
	    {
	      unsigned	tmp = str_encoding_table[i].enc;

	      if (tmp >= MAX_ENCODING)
		{
		  fprintf(stderr, "ERROR ... illegal NSStringEncoding "
		    "value in str_encoding_table. Ignored\n");
		}
	      else if (tmp > encTableSize)
		{
		  encTableSize = tmp;
		}
	    }
	  encTable = objc_malloc((encTableSize+1)*sizeof(struct _strenc_ *));
	  memset(encTable, 0, (encTableSize+1)*sizeof(struct _strenc_ *));

	  /*
	   * Now set up the pointers at the correct location in the table.
	   */
	  for (i = 0; i < count; i++)
	    {
	      unsigned	tmp = str_encoding_table[i].enc;

	      if (tmp < MAX_ENCODING)
		{
		  encTable[tmp] = &str_encoding_table[i];
#ifdef HAVE_ICONV
		  if (encTable[tmp]->iconv != 0 && *(encTable[tmp]->iconv) != 0)
		    {
		      iconv_t	c;
		      char	*lossy;

		      /*
		       * See if we can do a lossy conversion.
		       */
		      lossy = objc_malloc(strlen(encTable[tmp]->iconv) + 12);
		      strcpy(lossy, encTable[tmp]->iconv);
		      strcat(lossy, "//TRANSLIT");
		      c = iconv_open(UNICODE_ENC, encTable[tmp]->iconv);
		      if (c == (iconv_t)-1)
			{
			  objc_free(lossy);
			}
		      else
			{
			  encTable[tmp]->lossy = lossy;
			  iconv_close(c);
			}
		    }
#endif
		}
	    }
	  encodingTable = encTable;
	}
      [local_lock unlock];
    }
}

BOOL
GSPrivateIsEncodingSupported(NSStringEncoding enc)
{
  GSSetupEncodingTable();

  if (enc == 0 || enc > encTableSize || encodingTable[enc] == 0)
    {
      return NO;
    }
#ifdef HAVE_ICONV
  if (encodingTable[enc]->iconv != 0 && encodingTable[enc]->supported == 0)
    {
      if (enc == NSUnicodeStringEncoding)
	{
	  encodingTable[enc]->iconv = UNICODE_ENC;
	  encodingTable[enc]->supported = 1;
	}
      else if (encodingTable[enc]->iconv[0] == 0)
        {
	  /* explicitly check for empty encoding name since some systems
	   * have buggy iconv_open() code which succeeds on an empty name.
	   */
	  encodingTable[enc]->supported = -1;
	}
      else
	{
	  iconv_t	c;

	  c = iconv_open(UNICODE_ENC, encodingTable[enc]->iconv);
	  if (c == (iconv_t)-1)
	    {
	      encodingTable[enc]->supported = -1;
	    }
	  else
	    {
	      iconv_close(c);
	      c = iconv_open(encodingTable[enc]->iconv, UNICODE_ENC);
	      if (c == (iconv_t)-1)
		{
		  encodingTable[enc]->supported = -1;
		}
	      else
		{
		  iconv_close(c);
		  encodingTable[enc]->supported = 1;
		}
	    }
	}
    }
#endif
  if (encodingTable[enc]->supported == 1)
    {
      return YES;
    }
  return NO;
}

/** Returns the NSStringEncoding that matches the specified
 *  character set registry and encoding information. For instance,
 *  for the iso8859-5 character set, the registry is iso8859 and
 *  the encoding is 5, and the returned NSStringEncoding is
 *  NSISOCyrillicStringEncoding. If there is no specific encoding,
 *  use @"0". Returns GSUndefinedEncoding if there is no match.
 */
NSStringEncoding
GSEncodingForRegistry (NSString *registry, NSString *encoding)
{
  NSString	*charset = registry;

  if ([encoding length] > 0 && [encoding isEqualToString: @"0"] == NO)
    {
      charset = [NSString stringWithFormat: @"%@-%@", registry, encoding];
    }
  return [GSMimeDocument encodingFromCharset: charset];
}

/** Try to deduce the string encoding from the locale string
 *  clocale. This function looks in the Locale.encodings file
 *  installed as part of GNUstep Base if the encoding cannot be
 *  deduced from the clocale string itself. If  clocale isn't set or
 *  no match can be found, returns GSUndefinedEncoding.
 */
/* It would be really nice if this could be used in +defaultCStringEncoding, but
 * there are too many dependancies on other parts of the library to
 * make this practical (even if everything possible was written in C,
 * we'd still need some way to find the Locale.encodings file).
 */
NSStringEncoding
GSEncodingFromLocale(const char *clocale)
{
  NSStringEncoding	encoding = GSUndefinedEncoding;
  NSString		*encodstr;

  if (clocale == NULL || strcmp(clocale, "C") == 0
    || strcmp(clocale, "POSIX") == 0)
    {
      /* Don't make any assumptions. Let caller handle that */
      return encoding;
    }

  if (strchr (clocale, '.') != NULL)
    {
      /* Locale contains the 'codeset' section. Parse it and see
	 if we know what encoding this cooresponds to */
      NSString	*registry;
      NSString	*charset;
      NSArray	*array;
      char	*s;

      s = strchr (clocale, '.');
      registry = [[NSString stringWithUTF8String: s+1] lowercaseString];
      array = [registry componentsSeparatedByString: @"-"];
      registry = [array objectAtIndex: 0];
      if ([array count] > 1)
	{
	  charset = [NSString stringWithFormat: @"%@-%@",
	    registry, [array lastObject]];
	}
      else
	{
	  charset = registry;
	}

      encoding = [GSMimeDocument encodingFromCharset: charset];
    }
  else
    {
      /* Look up the locale in our table of encodings */
      NSBundle *gbundle;
      NSString *table;

#ifdef GNUSTEP
      gbundle = [NSBundle bundleForLibrary: @"gnustep-base"];
#else
      gbundle = [NSBundle bundleForClass: NSClassFromString(@"GSXMLNode")];
#endif
      table = [gbundle pathForResource: @"Locale"
		                ofType: @"encodings"
		           inDirectory: @"Languages"];
      if (table != nil)
	{
	  unsigned	count;
	  NSDictionary	*dict;
	
	  dict = [NSDictionary dictionaryWithContentsOfFile: table];
	  encodstr = [dict objectForKey:
			     [NSString stringWithUTF8String: clocale]];
	  if (encodstr == nil)
	    return GSUndefinedEncoding;

	  /* Find the matching encoding */
	  count = 0;
	  while (str_encoding_table[count].enc
	    && strcmp(str_encoding_table[count].ename, [encodstr lossyCString]))
	    {
	      count++;
	    }
	  if (str_encoding_table[count].enc)
	    {
	      encoding = str_encoding_table[count].enc;
	    }
	  if (encoding == GSUndefinedEncoding)
	    {
	      NSLog(@"No known GNUstep encoding for %s = %@",
		    clocale, encodstr);
	    }
	}
    }

  return encoding;
}

/**
 * Uses direct access into a two-level table to map cases.<br />
 * The two-level table method is less space efficient (but still not bad) than
 * a single table and a linear search, but it reduces the number of
 * conditional statements to just one.
 */
unichar
uni_tolower(unichar ch)
{
  unichar result = gs_tolower_map[ch / 256][ch % 256];

  return result ? result : ch;
}

/**
 * Uses direct access into a two-level table to map cases.<br />
 * The two-level table method is less space efficient (but still not bad) than
 * a single table and a linear search, but it reduces the number of
 * conditional statements to just one.
 */
unichar
uni_toupper(unichar ch)
{
  unichar result = gs_toupper_map[ch / 256][ch % 256];

  return result ? result : ch;
}

unsigned char
GSPrivateUniCop(unichar u)
{
  if (u < uni_cop_table[0].code)
    {
      return 0;	// Special case for latin1
    }
  else
    {
      unichar	code;
      unichar	count = 0;
      unichar	first = 0;
      unichar	last = uni_cop_table_size;

      while (first <= last)
	{
	  if (first != last)
	    {
	      count = (first + last) / 2;
	      code = uni_cop_table[count].code;
	      if (code < u)
		{
		  first = count+1;
		}
	      else if (code > u)
		{
		  last = count-1;
		}
	      else
		{
		  return uni_cop_table[count].cop;
		}
	    }
	  else  /* first == last */
	    {
	      if (u == uni_cop_table[first].code)
		{
		  return uni_cop_table[first].cop;
		}
	      return 0;
	    }
	}
      return 0;
    }
}

unsigned char
uni_cop(unichar u)
{
  return GSPrivateUniCop(u);
}

BOOL
uni_isnonsp(unichar u)
{
  /*
   * Treating upper surrogates as non-spacing is a convenient solution
   * to a number of issues with UTF-16
   */
  if ((u >= 0xdc00) && (u <= 0xdfff))
    return YES;

// FIXME check is uni_cop good for this
  if (GSPrivateUniCop(u))
    return YES;
  else
    return NO;
}

unichar*
uni_is_decomp(unichar u)
{
  if (u < uni_dec_table[0].code)
    {
      return 0;		// Special case for latin1
    }
  else
    {
      unichar	code;
      unichar	count = 0;
      unichar	first = 0;
      unichar	last = uni_dec_table_size;

      while (first <= last)
	{
	  if (first != last)
	    {
	      count = (first + last) / 2;
	      code = uni_dec_table[count].code;
	      if (code < u)
		{
		  first = count+1;
		}
	      else if (code > u)
		{
		  last = count-1;
		}
	      else
		{
		  return uni_dec_table[count].decomp;
		}
	    }
	  else  /* first == last */
	    {
	      if (u == uni_dec_table[first].code)
		{
		  return uni_dec_table[first].decomp;
		}
	      return 0;
	    }
	}
      return 0;
    }
}

/**
 * Function to check a block of data for validity as a unicode string and
 * say whether it contains solely ASCII or solely Latin1 data.<br />
 * Any leading BOM must already have been removed and the data must already
 * be in native byte order.<br />
 * Returns the number of characters which were found valid.
 */
unsigned
GSUnicode(const unichar *chars, unsigned length,
  BOOL *isASCII, BOOL *isLatin1)
{
  unsigned	i = 0;
  unichar	c;

  if (isASCII) *isASCII = YES;
  if (isLatin1) *isLatin1 = YES;
  while (i < length)
    {
      if ((c = chars[i++]) > 127)
        {
	  if (isASCII) *isASCII = NO;
	  i--;
	  while (i < length)
	    {
	      if ((c = chars[i++]) > 255)
		{
		  if (isLatin1) *isLatin1 = NO;
		  i--;
		  while (i < length)
		    {
		      c = chars[i++];
		      if (c == 0xfffe || c == 0xffff
			|| (c >= 0xfdd0 && c <= 0xfdef))
			{
			  return i - 1;	// Non-characters.
			}
		      if (c >= 0xdc00 && c <= 0xdfff)
		        {
			  return i - 1;	// Second half of a surrogate pair.
		        }
		      if (c >= 0xd800 && c <= 0xdbff)
		        {
			  // First half of a surrogate pair.
			  if (i >= length)
			    {
			      return i - 1;	// Second half missing
			    }
			  c = chars[i];
			  if (c < 0xdc00 || c > 0xdfff)
			    {
			      return i - 1;	// Second half missing
			    }
			  i++;		// Step past second half
		        }
		    }
		}
	    }
        }
    }
  return i;
}

#define	GROW() \
if (dst == 0) \
  { \
    /* \
     * Data is just being discarded anyway, so we can \
     * adjust the offset into the local buffer on the \
     * stack and pretend the buffer has grown. \
     */ \
    if (extra == 0) \
      { \
	ptr -= BUFSIZ; \
	bsize += BUFSIZ; \
      } \
    else \
      { \
	ptr -= BUFSIZ-1; \
	bsize += BUFSIZ-1; \
      } \
  } \
else if (zone == 0) \
  { \
    result = NO; /* No buffer growth possible ... fail. */ \
    goto done; \
  } \
else \
  { \
    unsigned	grow = slen; \
\
    if (grow < bsize + BUFSIZ) \
      { \
	grow = bsize + BUFSIZ; \
      } \
    grow *= sizeof(unichar); \
\
    if (ptr == buf || ptr == *dst) \
      { \
	unichar	*tmp; \
\
	tmp = NSZoneMalloc(zone, grow + extra); \
	if (tmp != 0) \
	  { \
	    memcpy(tmp, ptr, bsize * sizeof(unichar)); \
	  } \
	ptr = tmp; \
      } \
    else \
      { \
	ptr = NSZoneRealloc(zone, ptr, grow + extra); \
      } \
    if (ptr == 0) \
      { \
	result = NO;	/* Not enough memory */ \
	break; \
      } \
    bsize = grow / sizeof(unichar); \
  }

/**
 * Function to convert from 8-bit data to 16-bit unicode characters.
 * <p>The dst argument is a pointer to a pointer to a buffer in which the
 * converted string is to be stored.  If it is a null pointer, this function
 * discards converted data, and is used only to determine the length of the
 * converted string.  If the zone argument is non-nul, the function is free
 * to allocate a larger buffer if necessary, and store this new buffer in
 * the dst argument.  It will *NOT* deallocate the original buffer!
 * </p>
 * <p>The size argument is a pointer to the initial size of the destination
 * buffer.  If the function changes the buffer size, this value will be
 * altered to the new size.  This is measured in 16-bit unicode characters,
 * not bytes.
 * </p>
 * <p>The src argument is a pointer to the byte sequence which is
 * to be converted to 16-bit unicode.
 * </p>
 * <p>The slen argument is the length of the byte sequence
 * which is to be converted to 16-bit unicode.
 * This is measured in bytes.
 * </p>
 * <p>The enc argument specifies the encoding type of the 8-bit byte sequence
 * which is to be converted to 16-bit unicode.
 * </p>
 * <p>The zone argument specifies a memory zone in which the function may
 * allocate a buffer to return data in.
 * If this is nul, the function will fail if the originally supplied buffer
 * is not big enough (unless dst is a null pointer ... indicating that
 * converted data is to be discarded).
 * </p>
 * The options argument controls some special behavior.
 * <list>
 * <item>If GSUniTerminate is set, the function is expected to null terminate
 * the output string, and will assume that it is safe to place the nul
 * just beyond the end of the stated buffer size.
 * Also, if the function grows the buffer, it will allow for an extra
 * termination character.</item>
 * <item>If GSUniTemporary is set, the function will return the results in
 * an autoreleased buffer rather than in a buffer that the caller must
 * release.</item>
 * <item>If GSUniBOM is set, the function will write the first unicode
 * character as a byte order marker.</item>
 * <item>If GSUniShortOk is set, the function will return a buffer containing
 * any decoded characters even if the whole conversion fails.</item>
 * </list>
 * <p>On return, the function result is a flag indicating success (YES)
 * or failure (NO), and on success, the value stored in size is the number
 * of characters in the converted string.  The converted string itself is
 * stored in the location given by dst.<br />
 * NB. If the value stored in dst has been changed, it is a pointer to
 * allocated memory which the caller is responsible for freeing, and the
 * caller is <em>still</em> responsible for freeing the original buffer.
 * </p>
 */
BOOL
GSToUnicode(unichar **dst, unsigned int *size, const unsigned char *src,
  unsigned int slen, NSStringEncoding enc, NSZone *zone,
  unsigned int options)
{
  unichar	buf[BUFSIZ];
  unichar	*ptr;
  unsigned	bsize;
  unsigned	dpos = 0;	// Offset into destination buffer.
  unsigned	spos = 0;	// Offset into source buffer.
  unsigned	extra = (options & GSUniTerminate) ? sizeof(unichar) : 0;
  unichar	base = 0;
  unichar	*table = 0;
  BOOL		result = YES;

  /*
   * Ensure we have an initial buffer set up to decode data into.
   */
  if (dst == 0 || *size == 0)
    {
      ptr = buf;
      bsize = (extra != 0) ? BUFSIZ - 1 : BUFSIZ;
    }
  else
    {
      ptr = *dst;
      bsize = *size;
    }

  if (options & GSUniBOM)
    {
      while (dpos >= bsize)
	{
	  GROW();
	}
      ptr[dpos++] = (unichar)0xFEFF;	// Insert byte order marker.
    }

  switch (enc)
    {
      case NSUTF8StringEncoding:
	{
	  while (spos < slen)
	    {
	      unsigned char	c = src[spos];
	      unsigned long	u = c;

	      if (c > 0x7f)
                {
                  int i, sle = 0;

		  /* calculated the expected sequence length */
                  while (c & 0x80)
                    {
                      c = c << 1;
                      sle++;
                    }

		  /* legal ? */
		  if ((sle < 2) || (sle > 6))
                    {
	               result = NO;
		       goto done;
	            }

		  /* do we have enough bytes ? */
		  if ((spos + sle) > slen)
                    {
	               result = NO;
		       goto done;
	            }

		  /* get the codepoint */
		  for (i = 1; i < sle; i++)
		    {
		      if (src[spos + i] < 0x80 || src[spos + i] >= 0xc0)
			break;
		      u = (u << 6) | (src[spos + i] & 0x3f);
		    }
		  if (i < sle)
		    {
		      result = NO;
		      goto done;
		    }
	          u = u & ~(0xffffffff << ((5 * sle) + 1));
		  spos += sle;

		  /*
		   * We discard invalid codepoints here.
		   */
		  if (u > 0x10ffff || u == 0xfffe || u == 0xffff
		    || (u >= 0xfdd0 && u <= 0xfdef))
		    {
		      result = NO;	// Invalid character.
		      goto done;
		    }

		  if ((u >= 0xd800) && (u <= 0xdfff))
		    {
		      result = NO;	// Unmatched half of surrogate pair.
		      goto done;
		    }
                }
              else
		{
		  spos++;
		}

	      /*
	       * Add codepoint as either a single unichar for BMP
	       * or as a pair of surrogates for codepoints over 16 bits.
	       */

	      if (dpos >= bsize)
		{
		  GROW();
		}
	      if (u < 0x10000)
	        {
	          ptr[dpos++] = u;
	        }
	      else
	        {
                  unichar ul, uh;

                  u -= 0x10000;
                  ul = u & 0x3ff;
                  uh = (u >> 10) & 0x3ff;

	          ptr[dpos++] = uh + 0xd800;
	          if (dpos >= bsize)
		    {
		      GROW();
		    }
	          ptr[dpos++] = ul + 0xdc00;
	        }
	    }
	}
	break;

      case NSNonLossyASCIIStringEncoding:
      case NSASCIIStringEncoding:
	while (spos < slen)
	  {
	    unichar	c = (unichar)((unc)src[spos++]);

	    if (c > 127)
	      {
		result = NO;	// Non-ascii data found in input.
		goto done;
	      }
	    if (dpos >= bsize)
	      {
		GROW();
	      }
	    ptr[dpos++] = c;
	  }
	break;

      case NSISOLatin1StringEncoding:
      case NSUnicodeStringEncoding:
	while (spos < slen)
	  {
	    if (dpos >= bsize)
	      {
		GROW();
	      }
	    ptr[dpos++] = (unichar)((unc)src[spos++]);
	  }
	break;

      case NSNEXTSTEPStringEncoding:
	base = Next_conv_base;
	table = Next_char_to_uni_table;
	goto tables;

      case NSISOCyrillicStringEncoding:
	base = Cyrillic_conv_base;
	table = Cyrillic_char_to_uni_table;
	goto tables;

      case NSISOLatin2StringEncoding:
	base = Latin2_conv_base;
	table = Latin2_char_to_uni_table;
	goto tables;

      case NSISOLatin9StringEncoding:
	base = Latin9_conv_base;
	table = Latin9_char_to_uni_table;
	goto tables;

      case NSISOThaiStringEncoding:
        base = Thai_conv_base;
	table = Thai_char_to_uni_table;
	goto tables;
	
#if 0
      case NSSymbolStringEncoding:
	base = Symbol_conv_base;
	table = Symbol_char_to_uni_table;
	goto tables;
#endif

tables:
	while (spos < slen)
	  {
	    unc c = (unc)src[spos];

	    if (dpos >= bsize)
	      {
		GROW();
	      }
	    if (c < base)
	      {
		ptr[dpos++] = c;
	      }
	    else
	      {
		ptr[dpos++] = table[c - base];
	      }
	    spos++;
	  }
	break;

      case NSGSM0338StringEncoding:
	while (spos < slen)
	  {
	    unc c = (unc)src[spos];

	    if (dpos >= bsize)
	      {
		GROW();
	      }

	    ptr[dpos] = GSM0338_char_to_uni_table[c];
	    if (c == 0x1b && spos < slen)
	      {
		unsigned	i = 0;

		c = (unc)src[spos+1];
		while (i < sizeof(GSM0338_escapes)/sizeof(GSM0338_escapes[0]))
		  {
		    if (GSM0338_escapes[i].to == c)
		      {
			ptr[dpos] = GSM0338_escapes[i].from;
			spos++;
			break;
		      }
		    i++;
		  }
	      }
	    dpos++;
	    spos++;
	  }
	break;

      default:
#ifdef HAVE_ICONV
	{
	  unsigned char	*inbuf;
	  unsigned char	*outbuf;
	  size_t	inbytesleft;
	  size_t	outbytesleft;
	  size_t	rval;
	  iconv_t	cd;
	  const char	*estr = 0;
	  BOOL		done = NO;

	  if (GSPrivateIsEncodingSupported(enc) == YES)
	    {
	      estr = encodingTable[enc]->iconv;
	    }
	  /* explicitly check for empty encoding name since some systems
	   * have buggy iconv_open() code which succeeds on an empty name.
	   */
	  if (estr == 0)
	    {
	      NSLog(@"No iconv for encoding x%02x", enc);
	      result = NO;
	      goto done;
	    }
	  if (slen == 0)
	    {
	      break;	// Nothing to do
	    }
	  cd = iconv_open(UNICODE_ENC, estr);
	  if (cd == (iconv_t)-1)
	    {
	      NSLog(@"No iconv for encoding %@ tried to use %s",
		GSPrivateEncodingName(enc), estr);
	      result = NO;
	      goto done;
	    }

	  inbuf = (unsigned char*)src;
	  inbytesleft = slen;
	  outbuf = (unsigned char*)ptr;
	  outbytesleft = bsize * sizeof(unichar);
	  do
	    {
	      if (inbytesleft == 0)
		{
		  done = YES;	// Flush iconv
		  rval = iconv(cd, 0, 0, (void*)&outbuf, &outbytesleft);
		}
	      else
		{
		  rval = iconv(cd,
		    (void*)&inbuf, &inbytesleft, (void*)&outbuf, &outbytesleft);
		}
	      dpos = (bsize * sizeof(unichar) - outbytesleft) / sizeof(unichar);
	      if (rval == (size_t)-1)
		{
		  if (errno == E2BIG)
		    {
		      unsigned	old = bsize;

		      GROW();
		      outbuf = (unsigned char*)&ptr[dpos];
		      outbytesleft += (bsize - old) * sizeof(unichar);
		    }
		  else
		    {
		      result = NO;
		      goto done;
		    }
		}
	    } while (!done || rval != 0);
	  // close the converter
	  iconv_close(cd);
	}
#else
	result = NO;
#endif
    }

done:

  /*
   * Post conversion ... set output values.
   */
  if (extra != 0)
    {
      ptr[dpos] = (unichar)0;
    }
  *size = dpos;
  if (dst != 0 && (result == YES || (options & GSUniShortOk)))
    {
      if (options & GSUniTemporary)
	{
	  unsigned	bytes = dpos * sizeof(unichar) + extra;
	  void		*r;

	  /*
	   * Temporary string was requested ... make one.
	   */
	  r = GSAutoreleasedBuffer(bytes);
	  memcpy(r, ptr, bytes);
	  if (ptr != buf && (dst == 0 || ptr != *dst))
	    {
	      NSZoneFree(zone, ptr);
	    }
	  ptr = r;
	  *dst = ptr;
	}
      else if (zone != 0 && (ptr == buf || bsize > dpos))
	{
	  unsigned	bytes = dpos * sizeof(unichar) + extra;

	  /*
	   * Resizing is permitted, try ensure we return a buffer which
	   * is just big enough to hold the converted string.
	   */
	  if (ptr == buf || ptr == *dst)
	    {
	      unichar	*tmp;

	      tmp = NSZoneMalloc(zone, bytes);
	      if (tmp != 0)
		{
		  memcpy(tmp, ptr, bytes);
		}
	      ptr = tmp;
	    }
	  else
	    {
	      ptr = NSZoneRealloc(zone, ptr, bytes);
	    }
	  *dst = ptr;
	}
      else if (ptr == buf)
	{
	  ptr = NULL;
	  result = NO;
	}
      else
	{
	  *dst = ptr;
	}
    }
  else if (ptr != buf && dst != 0 && ptr != *dst)
    {
      NSZoneFree(zone, ptr);
    }

  if (dst)
    NSCAssert(*dst != buf, @"attempted to pass out pointer to internal buffer");

  return result;
}

#undef	GROW


#define	GROW() \
if (dst == 0) \
  { \
    /* \
     * Data is just being discarded anyway, so we can \
     * adjust the offset into the local buffer on the \
     * stack and pretend the buffer has grown. \
     */ \
    if (extra == 0) \
      { \
	ptr -= BUFSIZ; \
	bsize += BUFSIZ; \
      } \
    else \
      { \
	ptr -= BUFSIZ-1; \
	bsize += BUFSIZ-1; \
      } \
  } \
else if (zone == 0) \
  { \
    result = NO; /* No buffer growth possible ... fail. */ \
    goto done; \
  } \
else \
  { \
    unsigned	grow = slen; \
\
    if (grow < bsize + BUFSIZ) \
      { \
	grow = bsize + BUFSIZ; \
      } \
\
    if (ptr == buf || ptr == *dst) \
      { \
	unsigned char	*tmp; \
\
	tmp = NSZoneMalloc(zone, grow + extra); \
	if (tmp != 0) \
	  { \
	    memcpy(tmp, ptr, bsize); \
	  } \
	ptr = tmp; \
      } \
    else \
      { \
	ptr = NSZoneRealloc(zone, ptr, grow + extra); \
      } \
    if (ptr == 0) \
      { \
	result = NO;	/* Not enough memory */ \
	break; \
      } \
    bsize = grow; \
  }


static inline int chop(unichar c, _ucc_ *table, int hi)
{
  int	lo = 0;

  while (hi > lo)
    {
      int	i = (hi + lo) / 2;
      unichar	from = table[i].from;

      if (from < c)
        {
	  lo = i + 1;
	}
      else if (from > c)
        {
	  hi = i;
	}
      else
        {
	  return i;	// Found
	}
    }
  return -1;		// Not found
}

/**
 * Function to convert from 16-bit unicode to 8-bit data.
 * <p>The dst argument is a pointer to a pointer to a buffer in which the
 * converted data is to be stored.  If it is a null pointer, this function
 * discards converted data, and is used only to determine the length of the
 * converted data.  If the zone argument is non-nul, the function is free
 * to allocate a larger buffer if necessary, and store this new buffer in
 * the dst argument.  It will *NOT* deallocate the original buffer!
 * </p>
 * <p>The size argument is a pointer to the initial size of the destination
 * buffer.  If the function changes the buffer size, this value will be
 * altered to the new size.  This is measured in bytes.
 * </p>
 * <p>The src argument is a pointer to the 16-bit unicode string which is
 * to be converted to 8-bit data.
 * </p>
 * <p>The slen argument is the length of the 16-bit unicode string
 * which is to be converted to 8-bit data.
 * This is measured in 16-bit characters, not bytes.
 * </p>
 * <p>The enc argument specifies the encoding type of the 8-bit byte sequence
 * which is to be produced from the 16-bit unicode.
 * </p>
 * <p>The zone argument specifies a memory zone in which the function may
 * allocate a buffer to return data in.
 * If this is nul, the function will fail if the originally supplied buffer
 * is not big enough (unless dst is a null pointer ... indicating that
 * converted data is to be discarded).
 * </p>
 * The options argument controls some special behavior.
 * <list>
 * <item>If GSUniStrict is set, the function will fail if a character is
 * encountered in the source which can't be converted.  Otherwise, some
 * approximation or marker will be placed in the destination.</item>
 * <item>If GSUniTerminate is set, the function is expected to nul terminate
 * the output data, and will assume that it is safe to place the nul
 * just beyond the end of the stated buffer size.
 * Also, if the function grows the buffer, it will allow for an extra
 * termination byte.</item>
 * <item>If GSUniTemporary is set, the function will return the results in
 * an autoreleased buffer rather than in a buffer that the caller must
 * release.</item>
 * <item>If GSUniBOM is set, the function will read the first unicode
 * character as a byte order marker.</item>
 * <item>If GSUniShortOk is set, the function will return a buffer containing
 * any decoded characters even if the whole conversion fails.</item>
 * </list>
 * <p>On return, the function result is a flag indicating success (YES)
 * or failure (NO), and on success, the value stored in size is the number
 * of bytes in the converted data.  The converted data itself is
 * stored in the location given by dst.<br />
 * NB. If the value stored in dst has been changed, it is a pointer to
 * allocated memory which the caller is responsible for freeing, and the
 * caller is <em>still</em> responsible for freeing the original buffer.
 * </p>
 */
BOOL
GSFromUnicode(unsigned char **dst, unsigned int *size, const unichar *src,
  unsigned int slen, NSStringEncoding enc, NSZone *zone,
  unsigned int options)
{
  unsigned char	buf[BUFSIZ];
  unsigned char	*ptr;
  unsigned	bsize;
  unsigned	dpos = 0;	// Offset into destination buffer.
  unsigned	spos = 0;	// Offset into source buffer.
  unsigned	extra = (options & GSUniTerminate) ? 1 : 0;
  BOOL		strict = (options & GSUniStrict) ? YES : NO;
  unichar	base = 0;
  _ucc_		*table = 0;
  unsigned	tsize = 0;
  unsigned char	escape = 0;
  _ucc_		*etable = 0;
  unsigned	etsize = 0;
  _ucc_		*ltable = 0;
  unsigned	ltsize = 0;
  BOOL		swapped = NO;
  BOOL		result = YES;

  if (options & GSUniBOM)
    {
      if (slen == 0)
	{
	  *size = 0;
	  result = NO;	// Missing byte order marker.
	}
      else
	{
	  unichar	c;

	  c = *src++;
	  slen--;
	  if (c != 0xFEFF)
	    {
	      if (c == 0xFFFE)
		{
		  swapped = YES;
		}
	      else
		{
		  *size = 0;
		  result = NO;	// Illegal byte order marker.
		}
	    }
	}
    }

  /*
   * Ensure we have an initial buffer set up to decode data into.
   */
  if (dst == 0 || *size == 0)
    {
      ptr = buf;
      bsize = (extra != 0) ? BUFSIZ - 1 : BUFSIZ;
    }
  else
    {
      ptr = *dst;
      bsize = *size;
    }

  if (result == NO)
    {
      goto done;
    }

#ifdef HAVE_ICONV
  if (strict == NO
    && enc != NSUTF8StringEncoding
    && enc != NSGSM0338StringEncoding)
    {
      goto iconv_start;	// For lossy conversion
    }
#endif

  switch (enc)
    {
      case NSUTF8StringEncoding:
	{
	  while (spos < slen)
	    {
	      unichar 		u1, u2;
	      unsigned long	u;
	      int		sl = 0;

	      /* get first unichar */
	      u1 = src[spos++];
	      if (swapped == YES)
		{
		  u1 = (((u1 & 0xff00) >> 8) + ((u1 & 0x00ff) << 8));
		}
	      // 0xfeff is a zero-width-no-break-space inside text (not a BOM).
	      if (u1 == 0xfffe				// unexpected BOM
	        || u1 == 0xffff				// not a character
		|| (u1 >= 0xfdd0 && u1 <= 0xfdef)	// invalid character
		|| (u1 >= 0xdc00 && u1 <= 0xdfff))	// bad pairing
	        {
		  if (strict)
		    {
		      result = NO;
		      goto done;
                    }
		  continue;	// Skip invalid character.
	        }

	      /* possibly get second character and calculate 'u' */
	      if ((u1 >= 0xd800) && (u1 < 0xdc00))
                {
	  	  if (spos >= slen)
                    {
		      if (strict)
			{
			  result = NO;
			  goto done;
			}
		      continue;	// At end.
                    }

	          /* get second unichar */
	          u2 = src[spos++];
	          if (swapped == YES)
		    {
		      u2 = (((u2 & 0xff00) >> 8) + ((u2 & 0x00ff) << 8));
		    }

	          if ((u2 < 0xdc00) && (u2 > 0xdfff))
                    {
		      spos--;
		      if (strict)
			{
			  result = NO;
			  goto done;
			}
		      continue;		// Skip bad half of surrogate pair.
                    }

                  /* make the full value */
		  u = ((unsigned long)(u1 - 0xd800) * 0x400)
		    + (u2 - 0xdc00) + 0x10000;
                }
              else
		{
		  u = u1;
		}

              /* calculate the sequence length */
              if (u <= 0x7f)
		{
		  sl = 1;
		}
              else if (u <= 0x7ff)
		{
		  sl = 2;
		}
              else if (u <= 0xffff)
		{
		  sl = 3;
		}
              else if (u <= 0x1fffff)
		{
		  sl = 4;
		}
              else if (u <= 0x3ffffff)
		{
		  sl = 5;
		}
              else
		{
		  sl = 6;
		}

              /* make sure we have enough space for it */
	      while (dpos + sl >= bsize)
		{
		  GROW();
		}

	      if (sl == 1)
                {
	          ptr[dpos++] = u & 0x7f;
                }
              else
                {
                  int		i;
                  unsigned char	reversed[8];

                  /* split value into reversed array */
                  for (i = 0; i < sl; i++)
                    {
                      reversed[i] = (u & 0x3f);
                      u = u >> 6;
                    }

	          ptr[dpos++] = reversed[sl-1] | ((0xff << (8-sl)) & 0xff);
                  /* add bytes into the output sequence */
                  for (i = sl - 2; i >= 0; i--)
		    {
		      ptr[dpos++] = reversed[i] | 0x80;
		    }
                }
	    }
        }
        break;

      case NSNonLossyASCIIStringEncoding:
      case NSASCIIStringEncoding:
	base = 128;
	goto bases;

      case NSISOLatin1StringEncoding:
      case NSUnicodeStringEncoding:
	base = 256;
	goto bases;

bases:
	if (strict == NO)
	  {
	    while (spos < slen)
	      {
		unichar	u = src[spos++];

		if (swapped == YES)
		  {
		    u = (((u & 0xff00) >> 8) + ((u & 0x00ff) << 8));
		  }
		
		if (dpos >= bsize)
		  {
		    GROW();
		  }
		if (u < base)
		  {
		    ptr[dpos++] = (unsigned char)u;
		  }
		else
		  {
		    ptr[dpos++] =  '?';
		  }
	      }
	  }
	else
	  {
	    while (spos < slen)
	      {
		unichar	u = src[spos++];

		if (swapped == YES)
		  {
		    u = (((u & 0xff00) >> 8) + ((u & 0x00ff) << 8));
		  }
		if (dpos >= bsize)
		  {
		    GROW();
		  }
		if (u < base)
		  {
		    ptr[dpos++] = (unsigned char)u;
		  }
		else
		  {
		    result = NO;
		    goto done;
		  }
	      }
	  }
	break;

      case NSNEXTSTEPStringEncoding:
	base = Next_conv_base;
	table = Next_uni_to_char_table;
	tsize = Next_uni_to_char_table_size;
	goto tables;

      case NSISOCyrillicStringEncoding:
	base = Cyrillic_conv_base;
	table = Cyrillic_uni_to_char_table;
	tsize = Cyrillic_uni_to_char_table_size;
	goto tables;

      case NSISOLatin2StringEncoding:
	base = Latin2_conv_base;
	table = Latin2_uni_to_char_table;
	tsize = Latin2_uni_to_char_table_size;
	goto tables;

      case NSISOLatin9StringEncoding:
	base = Latin9_conv_base;
	table = Latin9_uni_to_char_table;
	tsize = Latin9_uni_to_char_table_size;
	goto tables;

      case NSISOThaiStringEncoding:
        base = Thai_conv_base;
	table = Thai_uni_to_char_table;
	tsize = Thai_uni_to_char_table_size;
	goto tables;

#if 0
      case NSSymbolStringEncoding:
	base = Symbol_conv_base;
	table = Symbol_uni_to_char_table;
	tsize = Symbol_uni_to_char_table_size;
	goto tables;
#endif

      case NSGSM0338StringEncoding:
	base = 0;
	table = GSM0338_uni_to_char_table;
	tsize = GSM0338_tsize;
        escape = 0x1b;
	etable = GSM0338_escapes;
	etsize = GSM0338_esize;
	if (strict == NO)
	  {
	    ltable = GSM0338_lossy;
	    ltsize = GSM0338_lsize;
	  }
	goto tables;

tables:
	while (spos < slen)
	  {
	    unichar	u = src[spos++];
	    int	i;

	    /* Swap byte order if necessary */
	    if (swapped == YES)
	      {
		u = (((u & 0xff00) >> 8) + ((u & 0x00ff) << 8));
	      }

	    /* Grow output buffer to make room if necessary */
	    if (dpos >= bsize)
	      {
		GROW();
	      }

	    if (u < base)
	      {
		/*
		 * The character set has a lower section whose contents
		 * are identical to unicode, so no mapping is needed.
		 */
		ptr[dpos++] = (unsigned char)u;
	      }
	    else if (table != 0 && (i = chop(u, table, tsize)) >= 0)
	      {
		/*
		 * The character mapping is found in a basic table.
		 */
		ptr[dpos++] = table[i].to;
	      }
	    else if (etable != 0 && (i = chop(u, etable, etsize)) >= 0)
	      {
		/*
		 * The character mapping is found in a table of simple
		 * escape sequences consisting of an escape byte followed
		 * by another single byte.
		 */
		ptr[dpos++] = escape;
		if (dpos >= bsize)
		  {
		    GROW();
		  }
		ptr[dpos++] = etable[i].to;
	      }
	    else if (ltable != 0 && (i = chop(u, ltable, ltsize)) >= 0)
	      {
		/*
		 * The character is found in a lossy mapping table.
		 */
		ptr[dpos++] = ltable[i].to;
	      }
	    else if (strict == NO)
	      {
		/*
		 * The default lossy mapping generates a question mark.
		 */
		ptr[dpos++] = '?';
	      }
	    else
	      {
		/*
		 * No mapping has been found.
		 */
		result = NO;
		spos = slen;
		goto done;
	      }
	  }
	break;

      default:
#ifdef HAVE_ICONV
iconv_start:
	{
	  iconv_t	cd;
	  unsigned char	*inbuf;
	  unsigned char	*outbuf;
	  size_t	inbytesleft;
	  size_t	outbytesleft;
	  size_t	rval;
	  const char	*estr = 0;
	  BOOL		done = NO;

	  if (GSPrivateIsEncodingSupported(enc) == YES)
	    {
	      if (strict == NO)
		{
		  /*
		   * Try to transliterate where no direct conversion
		   * is available.
		   */
		  estr = encodingTable[enc]->lossy;
		}
	      if (estr == 0)
		{
		  estr = encodingTable[enc]->iconv;
		}
	    }
	  
	  /* explicitly check for empty encoding name since some systems
	   * have buggy iconv_open() code which succeeds on an empty name.
	   */
	  if (estr == 0)
	    {
	      NSLog(@"No iconv for encoding x%02x", enc);
	      result = NO;
	      goto done;
	    }
	  if (slen == 0)
	    {
	      break;	// Nothing to convert.
	    }
	  cd = iconv_open(estr, UNICODE_ENC);
	  if (cd == (iconv_t)-1)
	    {
	      NSLog(@"No iconv for encoding %@ tried to use %s",
		GSPrivateEncodingName(enc), estr);
	      result = NO;
	      goto done;
	    }

	  inbuf = (unsigned char*)src;
	  inbytesleft = slen * sizeof(unichar);
	  outbuf = (unsigned char*)ptr;
	  outbytesleft = bsize;
	  do
	    {
	      if (inbytesleft == 0)
		{
		  done = YES; // Flush buffer
		  rval = iconv(cd, 0, 0, (void*)&outbuf, &outbytesleft);
		}
	      else
		{
		  rval = iconv(cd,
		    (void*)&inbuf, &inbytesleft, (void*)&outbuf, &outbytesleft);
		}
	      dpos = bsize - outbytesleft;
	      if (rval != 0)
		{
		  if (rval == (size_t)-1)
		    {
		      if (errno == E2BIG)
			{
			  unsigned	old = bsize;

			  GROW();
			  outbuf = (unsigned char*)&ptr[dpos];
			  outbytesleft += (bsize - old);
			}
		      else if (errno == EILSEQ)
			{
			  if (strict == YES)
			    {
			      result = NO;
			      goto done;
			    }
			  /*
			   * If we are allowing lossy conversion, we replace any
			   * unconvertable character with a question mark.
			   */
			  if (outbytesleft > 0)
			    {
			      *outbuf++ = '?';
			      outbytesleft--;
			      inbuf += sizeof(unichar);
			      inbytesleft -= sizeof(unichar);
			    }
			}
		      else
			{
			  result = NO;
			  goto done;
			}
		    }
		  else if (strict == YES)
		    {
		      /*
		       * A positive return from iconv indicates some
		       * irreversible (ie lossy) conversions took place,
		       * so if we are doing strict conversions we must fail.
		       */
		      result = NO;
		      goto done;
		    }
		}
	    } while (!done || rval != 0);
	  // close the converter
	  iconv_close(cd);
	}
#else
	result = NO;
	goto done;
#endif
    }

  done:

  /*
   * Post conversion ... set output values.
   */
  if (extra != 0)
    {
      ptr[dpos] = (unsigned char)0;
    }
  *size = dpos;
  if (dst != 0 && (result == YES || (options & GSUniShortOk)))
    {
      if (options & GSUniTemporary)
	{
	  unsigned	bytes = dpos + extra;
	  void		*r;

	  /*
	   * Temporary string was requested ... make one.
	   */
	  r = GSAutoreleasedBuffer(bytes);
	  memcpy(r, ptr, bytes);
	  if (ptr != buf && (dst == 0 || ptr != *dst))
	    {
	      NSZoneFree(zone, ptr);
	    }
	  ptr = r;
	  *dst = ptr;
	}
      else if (zone != 0 && (ptr == buf || bsize > dpos))
	{
	  unsigned	bytes = dpos + extra;

	  /*
	   * Resizing is permitted - try ensure we return a buffer
	   * which is just big enough to hold the converted string.
	   */
	  if (ptr == buf || ptr == *dst)
	    {
	      unsigned char	*tmp;

	      tmp = NSZoneMalloc(zone, bytes);
	      if (tmp != 0)
		{
		  memcpy(tmp, ptr, bytes);
		}
	      ptr = tmp;
	    }
	  else
	    {
	      ptr = NSZoneRealloc(zone, ptr, bytes);
	    }
	  *dst = ptr;
	}
      else if (ptr == buf)
	{
	  ptr = NULL;
	  result = NO;
	}
      else
	{
	  *dst = ptr;
	}
    }
  else if (ptr != buf && dst != 0 && ptr != *dst)
    {
      NSZoneFree(zone, ptr);
    }

  if (dst)
    NSCAssert(*dst != buf, @"attempted to pass out pointer to internal buffer");

  return result;
}

#undef	GROW



NSStringEncoding*
GSPrivateAvailableEncodings()
{
  if (_availableEncodings == 0)
    {
      GSSetupEncodingTable();
      [GS_INITIALIZED_LOCK(local_lock, GSLazyLock) lock];
      if (_availableEncodings == 0)
	{
	  NSStringEncoding	*encodings;
	  unsigned		pos;
	  unsigned		i;

	  /*
	   * Now build up a list of supported encodings ... in the
	   * format needed to support [NSString+availableStringEncodings]
	   * Check to see what iconv support we have as we go along.
	   * This is also the place where we determine the name we use
	   * for iconv to support unicode.
	   */
	  encodings = objc_malloc(sizeof(NSStringEncoding) * (encTableSize+1));
	  pos = 0;
	  for (i = 0; i < encTableSize+1; i++)
	    {
	      if (GSPrivateIsEncodingSupported(i) == YES)
		{
		  encodings[pos++] = i;
		}
	    }
	  encodings[pos] = 0;
	  _availableEncodings = encodings;
	}
      [local_lock unlock];
    }
  return _availableEncodings;
}

NSStringEncoding
GSPrivateDefaultCStringEncoding()
{
  if (defEnc == GSUndefinedEncoding)
    {
      char		*encoding;
      unsigned int	count;

      GSSetupEncodingTable();

      [GS_INITIALIZED_LOCK(local_lock, GSLazyLock) lock];
      if (defEnc != GSUndefinedEncoding)
	{
	  [local_lock unlock];
	  return defEnc;
	}

      if (natEnc == GSUndefinedEncoding)
	{
	  /* Encoding not set */
#if HAVE_LANGINFO_CODESET
	  /* Take it from the system locale information.  */
	  encoding = nl_langinfo(CODESET);
	  /*
	   * First handle the fallback response from nl_langinfo() ...
	   * if we are getting the default value we can't assume that
	   * the user has set anything up at all, so we must use the
	   * OpenStep/GNUstep default encopding ... latin1, even though
	   * the nl_langinfo() stuff would say default is ascii.
	   */
	  if (strcmp(encoding, "ANSI_X3.4-1968") == 0 /* glibc */
	    || strcmp(encoding, "ISO_646.IRV:1983") == 0 /* glibc */
	    || strcmp(encoding, "646") == 0 /* Solaris NetBSD */)
	    natEnc = NSISOLatin1StringEncoding;
	  else if (strcmp(encoding, "EUC-JP") == 0 /* glibc */
	    /* HP-UX IRIX OSF/1 Solaris NetBSD */
	    || strcmp(encoding, "eucJP") == 0
	    || strcmp(encoding, "IBM-eucJP") == 0 /* AIX */)
	    natEnc = NSJapaneseEUCStringEncoding;
	  else if (strcmp(encoding, "UTF-8") == 0 /* glibc AIX OSF/1 Solaris */
	    || strcmp(encoding, "utf8") == 0 /* HP-UX */)
	    natEnc = NSUTF8StringEncoding;
	  else if (strcmp(encoding, "ISO-8859-1") == 0 /* glibc */
	    /* AIX IRIX OSF/1 Solaris NetBSD */
	    || strcmp(encoding, "ISO8859-1") == 0
	    || strcmp(encoding, "iso88591") == 0 /* HP-UX */)
	    natEnc = NSISOLatin1StringEncoding;
	  else if (strcmp(encoding, "IBM-932") == 0 /* AIX */
	    || strcmp(encoding, "SJIS") == 0 /* HP-UX OSF/1 NetBSD */
	    || strcmp(encoding, "PCK") == 0 /* Solaris */)
	    natEnc = NSShiftJISStringEncoding;
	  else if (strcmp(encoding, "ISO-8859-2") == 0 /* glibc */
	    /* AIX IRIX OSF/1 Solaris NetBSD */
	    || strcmp(encoding, "ISO8859-2") == 0
	    || strcmp(encoding, "iso88592") == 0 /* HP-UX */)
	    natEnc = NSISOLatin2StringEncoding;
	  else if (strcmp(encoding, "CP1251") == 0 /* glibc */
	    || strcmp(encoding, "ansi-1251") == 0 /* Solaris */)
	    natEnc = NSWindowsCP1251StringEncoding;
	  else if (strcmp(encoding, "CP1252") == 0 /*  */
	    || strcmp(encoding, "IBM-1252") == 0 /* AIX */)
	    natEnc = NSWindowsCP1252StringEncoding;
	  else if (strcmp(encoding, "ISO-8859-5") == 0 /* glibc */
	    /* AIX IRIX OSF/1 Solaris NetBSD */
	    || strcmp(encoding, "ISO8859-5") == 0
	    || strcmp(encoding, "iso88595") == 0 /* HP-UX */)
	    natEnc = NSISOCyrillicStringEncoding;
	  else if (strcmp(encoding, "KOI8-R") == 0 /* glibc */
	    || strcmp(encoding, "koi8-r") == 0 /* Solaris */)
	    natEnc = NSKOI8RStringEncoding;
	  else if (strcmp(encoding, "ISO-8859-3") == 0 /* glibc */
	    || strcmp(encoding, "ISO8859-3") == 0 /* Solaris */)
	    natEnc = NSISOLatin3StringEncoding;
	  else if (strcmp(encoding, "ISO-8859-4") == 0 /*  */
	    || strcmp(encoding, "ISO8859-4") == 0 /* OSF/1 Solaris NetBSD */)
	    natEnc = NSISOLatin4StringEncoding;
	  else if (strcmp(encoding, "ISO-8859-6") == 0 /* glibc */
	    || strcmp(encoding, "ISO8859-6") == 0 /* AIX Solaris */
	    || strcmp(encoding, "iso88596") == 0 /* HP-UX */)
	    natEnc = NSISOArabicStringEncoding;
	  else if (strcmp(encoding, "ISO-8859-7") == 0 /* glibc */
	    || strcmp(encoding, "ISO8859-7") == 0 /* AIX IRIX OSF/1 Solaris */
	    || strcmp(encoding, "iso88597") == 0 /* HP-UX */)
	    natEnc = NSISOGreekStringEncoding;
	  else if (strcmp(encoding, "ISO-8859-8") == 0 /* glibc */
	    || strcmp(encoding, "ISO8859-8") == 0 /* AIX OSF/1 Solaris */
	    || strcmp(encoding, "iso88598") == 0 /* HP-UX */)
	    natEnc = NSISOHebrewStringEncoding;
	  else if (strcmp(encoding, "ISO-8859-9") == 0 /* glibc */
	    || strcmp(encoding, "ISO8859-9") == 0 /* AIX IRIX OSF/1 Solaris */
	    || strcmp(encoding, "iso88599") == 0 /* HP-UX */)
	    natEnc = NSISOLatin5StringEncoding;
	  else if (strcmp(encoding, "ISO-8859-10") == 0 /*  */
	    || strcmp(encoding, "ISO8859-10") == 0 /*  */)
	    natEnc = NSISOLatin6StringEncoding;
	  else if (strcmp(encoding, "TIS-620") == 0 /* glibc AIX */
	    || strcmp(encoding, "tis620") == 0 /* HP-UX */
	    || strcmp(encoding, "TIS620.2533") == 0 /* Solaris */
	    || strcmp(encoding, "TACTIS") == 0 /* OSF/1 */)
	    natEnc = NSISOThaiStringEncoding;
	  else if (strcmp(encoding, "ISO-8859-13") == 0 /* glibc */
	    || strcmp(encoding, "ISO8859-13") == 0 /*  */
	    || strcmp(encoding, "IBM-921") == 0 /* AIX */)
	    natEnc = NSISOLatin7StringEncoding;
	  else if (strcmp(encoding, "ISO-8859-14") == 0 /* glibc */
	    || strcmp(encoding, "ISO8859-14") == 0 /*  */)
	    natEnc = NSISOLatin8StringEncoding;
	  else if (strcmp(encoding, "ISO-8859-15") == 0 /* glibc */
	    /* AIX OSF/1 Solaris NetBSD */
	    || strcmp(encoding, "ISO8859-15") == 0
	    || strcmp(encoding, "iso885915") == 0 /* HP-UX */)
	    natEnc = NSISOLatin9StringEncoding;
	  else if (strcmp(encoding, "GB2312") == 0 /* glibc */
	    || strcmp(encoding, "gb2312") == 0 /* Solaris */
	    || strcmp(encoding, "eucCN") == 0 /* IRIX NetBSD */
	    || strcmp(encoding, "IBM-eucCN") == 0 /* AIX */
	    || strcmp(encoding, "hp15CN") == 0 /* HP-UX */)
	    natEnc = NSGB2312StringEncoding;
	  else if (strcmp(encoding, "BIG5") == 0 /* glibc Solaris NetBSD */
	    || strcmp(encoding, "big5") == 0 /* AIX HP-UX OSF/1 */)
	    natEnc = NSBIG5StringEncoding;
	  else if (strcmp(encoding, "EUC-KR") == 0 /* glibc */
	    || strcmp(encoding, "eucKR") == 0 /* HP-UX IRIX OSF/1 NetBSD */
	    || strcmp(encoding, "IBM-eucKR") == 0 /* AIX */
	    || strcmp(encoding, "5601") == 0 /* Solaris */)
	    natEnc = NSKoreanEUCStringEncoding;
#endif
	}

      encoding = getenv("GNUSTEP_STRING_ENCODING");
      if (encoding != 0)
	{
	  count = 0;
	  while (str_encoding_table[count].enc
	    && strcasecmp(str_encoding_table[count].ename, encoding)
	    && strcasecmp(str_encoding_table[count].iconv, encoding))
	    {
	      count++;
	    }
	  if (str_encoding_table[count].enc)
	    {
	      defEnc = str_encoding_table[count].enc;
	    }
	  else
	    {
	      fprintf(stderr,
		      "WARNING: %s - encoding not supported.\n", encoding);
	      fprintf(stderr,
		      "  NSISOLatin1StringEncoding set as default.\n");
	      defEnc = NSISOLatin1StringEncoding;
	    }
	}
      if (defEnc == GSUndefinedEncoding)
	{
	  defEnc = natEnc;
	}
      if (defEnc == GSUndefinedEncoding)
	{
	  defEnc = NSISOLatin1StringEncoding;
	}
      else if (GSPrivateIsEncodingSupported(defEnc) == NO)
	{
	  fprintf(stderr, "WARNING: %s - encoding not implemented as "
		  "default c string encoding.\n", encoding);
	  fprintf(stderr,
		  "  NSISOLatin1StringEncoding set as default.\n");
	  defEnc = NSISOLatin1StringEncoding;
	}

      if (natEnc == GSUndefinedEncoding)
        {
	  natEnc = defEnc;
	}

      [local_lock unlock];
    }
  return defEnc;
}

NSString*
GSPrivateEncodingName(NSStringEncoding encoding)
{
  if (GSPrivateIsEncodingSupported(encoding) == NO)
    {
      return @"Unknown encoding";
    }
  return [NSString stringWithUTF8String: encodingTable[encoding]->ename];
}

BOOL
GSPrivateIsByteEncoding(NSStringEncoding encoding)
{
  if (GSPrivateIsEncodingSupported(encoding) == NO)
    {
      return NO;
    }
  return encodingTable[encoding]->eightBit;
}

NSStringEncoding
GSPrivateNativeCStringEncoding()
{
  if (natEnc == GSUndefinedEncoding)
    {
      /* GSPrivateDefaultCStringEncoding() will actually set the encoding.
       */
      GSPrivateDefaultCStringEncoding();
    }
  return natEnc;
}