iconvdata/utf-16.c

   1 /* Conversion module for UTF-16.
   2    Copyright (C) 1999, 2000 Free Software Foundation, Inc.
   3    This file is part of the GNU C Library.
   4    Contributed by Ulrich Drepper <drepper@cygnus.com>, 1999.
   5
   6    The GNU C Library is free software; you can redistribute it and/or
   7    modify it under the terms of the GNU Library General Public License as
   8    published by the Free Software Foundation; either version 2 of the
   9    License, or (at your option) any later version.
  10
  11    The GNU C Library is distributed in the hope that it will be useful,
  12    but WITHOUT ANY WARRANTY; without even the implied warranty of
  13    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14    Library General Public License for more details.
  15
  16    You should have received a copy of the GNU Library General Public
  17    License along with the GNU C Library; see the file COPYING.LIB.  If not,
  18    write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  19    Boston, MA 02111-1307, USA.  */
  20
  21 #include <byteswap.h>
  22 #include <dlfcn.h>
  23 #include <gconv.h>
  24 #include <stddef.h>
  25 #include <stdint.h>
  26 #include <stdlib.h>
  27 #include <string.h>
  28
  29 /* This is the Byte Order Mark character (BOM).  */
  30 #define BOM     0xfeff
  31 /* And in the other byte order.  */
  32 #define BOM_OE  0xfffe
  33
  34
  35 /* Definitions used in the body of the `gconv' function.  */
  36 #define FROM_LOOP               from_utf16_loop
  37 #define TO_LOOP                 to_utf16_loop
  38 #define DEFINE_INIT             0
  39 #define DEFINE_FINI             0
  40 #define MIN_NEEDED_FROM         2
  41 #define MAX_NEEDED_FROM         4
  42 #define MIN_NEEDED_TO           4
  43 #define FROM_DIRECTION          (dir == from_utf16)
  44 #define PREPARE_LOOP \
  45   enum direction dir = ((struct utf16_data *) step->__data)->dir;             \
  46   enum variant var = ((struct utf16_data *) step->__data)->var;               \
  47   int swap = ((struct utf16_data *) step->__data)->swap;                      \
  48   if (FROM_DIRECTION && var == UTF_16)                                        \
  49     {                                                                         \
  50       if (data->__invocation_counter == 0)                                    \
  51         {                                                                     \
  52           /* We have to find out which byte order the file is encoded in.  */ \
  53           if (inptr + 2 > inend)                                              \
  54             return __GCONV_EMPTY_INPUT;                                       \
  55                                                                               \
  56           if (get16u (inptr) == BOM)                                          \
  57             /* Simply ignore the BOM character.  */                           \
  58             inptr += 2;                                                       \
  59           else if (get16u (inptr) == BOM_OE)                                  \
  60             {                                                                 \
  61               ((struct utf16_data *) step->__data)->swap = 1;                 \
  62               inptr += 2;                                                     \
  63             }                                                                 \
  64         }                                                                     \
  65     }                                                                         \
  66   else if (!FROM_DIRECTION && var == UTF_16 && !data->__internal_use          \
  67            && data->__invocation_counter == 0)                                \
  68     {                                                                         \
  69       /* Emit the Byte Order Mark.  */                                        \
  70       if (__builtin_expect (outbuf + 2 > outend, 0))                          \
  71         return __GCONV_FULL_OUTPUT;                                           \
  72                                                                               \
  73       put16u (outbuf, BOM);                                                   \
  74       outbuf += 2;                                                            \
  75     }
  76 #define EXTRA_LOOP_ARGS         , var, swap
  77
  78
  79 /* Direction of the transformation.  */
  80 enum direction
  81 {
  82   illegal_dir,
  83   to_utf16,
  84   from_utf16
  85 };
  86
  87 enum variant
  88 {
  89   illegal_var,
  90   UTF_16,
  91   UTF_16LE,
  92   UTF_16BE
  93 };
  94
  95 struct utf16_data
  96 {
  97   enum direction dir;
  98   enum variant var;
  99   int swap;
 100 };
 101
 102
 103 int
 104 gconv_init (struct __gconv_step *step)
 105 {
 106   /* Determine which direction.  */
 107   struct utf16_data *new_data;
 108   enum direction dir = illegal_dir;
 109   enum variant var = illegal_var;
 110   int result;
 111
 112   if (__strcasecmp (step->__from_name, "UTF-16//") == 0)
 113     {
 114       dir = from_utf16;
 115       var = UTF_16;
 116     }
 117   else if (__strcasecmp (step->__to_name, "UTF-16//") == 0)
 118     {
 119       dir = to_utf16;
 120       var = UTF_16;
 121     }
 122   else if (__strcasecmp (step->__from_name, "UTF-16BE//") == 0)
 123     {
 124       dir = from_utf16;
 125       var = UTF_16BE;
 126     }
 127   else if (__strcasecmp (step->__to_name, "UTF-16BE//") == 0)
 128     {
 129       dir = to_utf16;
 130       var = UTF_16BE;
 131     }
 132   else if (__strcasecmp (step->__from_name, "UTF-16LE//") == 0)
 133     {
 134       dir = from_utf16;
 135       var = UTF_16LE;
 136     }
 137   else if (__strcasecmp (step->__to_name, "UTF-16LE//") == 0)
 138     {
 139       dir = to_utf16;
 140       var = UTF_16LE;
 141     }
 142
 143   result = __GCONV_NOCONV;
 144   if (__builtin_expect (dir, to_utf16) != illegal_dir)
 145     {
 146       new_data = (struct utf16_data *) malloc (sizeof (struct utf16_data));
 147
 148       result = __GCONV_NOMEM;
 149       if (new_data != NULL)
 150         {
 151           new_data->dir = dir;
 152           new_data->var = var;
 153           new_data->swap = ((var == UTF_16LE && BYTE_ORDER == BIG_ENDIAN)
 154                             || (var == UTF_16BE
 155                                 && BYTE_ORDER == LITTLE_ENDIAN));
 156           step->__data = new_data;
 157
 158           if (dir == from_utf16)
 159             {
 160               step->__min_needed_from = MIN_NEEDED_FROM;
 161               step->__max_needed_from = MIN_NEEDED_FROM;
 162               step->__min_needed_to = MIN_NEEDED_TO;
 163               step->__max_needed_to = MIN_NEEDED_TO;
 164             }
 165           else
 166             {
 167               step->__min_needed_from = MIN_NEEDED_TO;
 168               step->__max_needed_from = MIN_NEEDED_TO;
 169               step->__min_needed_to = MIN_NEEDED_FROM;
 170               step->__max_needed_to = MIN_NEEDED_FROM;
 171             }
 172
 173           step->__stateful = 0;
 174
 175           result = __GCONV_OK;
 176         }
 177     }
 178
 179   return result;
 180 }
 181
 182
 183 void
 184 gconv_end (struct __gconv_step *data)
 185 {
 186   free (data->__data);
 187 }
 188
 189
 190 /* Convert from the internal (UCS4-like) format to UTF-16.  */
 191 #define MIN_NEEDED_INPUT        MIN_NEEDED_TO
 192 #define MIN_NEEDED_OUTPUT       MIN_NEEDED_FROM
 193 #define MAX_NEEDED_OUTPUT       MAX_NEEDED_FROM
 194 #define LOOPFCT                 TO_LOOP
 195 #define BODY \
 196   {                                                                           \
 197     uint32_t c = get32 (inptr);                                               \
 198                                                                               \
 199     if (__builtin_expect (c >= 0xd800 && c < 0xe000, 0))                      \
 200       {                                                                       \
 201         /* Surrogate characters in UCS-4 input are not valid.                 \
 202            We must catch this.  If we let surrogates pass through,            \
 203            attackers could make a security hole exploit by                    \
 204            synthesizing any desired plane 1-16 character.  */                 \
 205         if (! ignore_errors_p ())                                             \
 206           {                                                                   \
 207             result = __GCONV_ILLEGAL_INPUT;                                   \
 208             break;                                                            \
 209           }                                                                   \
 210         inptr += 4;                                                           \
 211         ++*irreversible;                                                      \
 212         continue;                                                             \
 213       }                                                                       \
 214                                                                               \
 215     if (swap)                                                                 \
 216       {                                                                       \
 217         if (__builtin_expect (c, 0) >= 0x10000)                               \
 218           {                                                                   \
 219             if (__builtin_expect (c, 0) >= 0x110000)                          \
 220               {                                                               \
 221                 STANDARD_ERR_HANDLER (4);                                     \
 222               }                                                               \
 223                                                                               \
 224             /* Generate a surrogate character.  */                            \
 225             if (__builtin_expect (outptr + 4 > outend, 0))                    \
 226               {                                                               \
 227                 /* Overflow in the output buffer.  */                         \
 228                 result = __GCONV_FULL_OUTPUT;                                 \
 229                 break;                                                        \
 230               }                                                               \
 231                                                                               \
 232             put16 (outptr, bswap_16 (0xd7c0 + (c >> 10)));                    \
 233             outptr += 2;                                                      \
 234             put16 (outptr, bswap_16 (0xdc00 + (c & 0x3ff)));                  \
 235           }                                                                   \
 236         else                                                                  \
 237           put16 (outptr, bswap_16 (c));                                       \
 238       }                                                                       \
 239     else                                                                      \
 240       {                                                                       \
 241         if (__builtin_expect (c, 0) >= 0x10000)                               \
 242           {                                                                   \
 243             if (__builtin_expect (c, 0) >= 0x110000)                          \
 244               {                                                               \
 245                 STANDARD_ERR_HANDLER (4);                                     \
 246               }                                                               \
 247                                                                               \
 248             /* Generate a surrogate character.  */                            \
 249             if (__builtin_expect (outptr + 4 > outend, 0))                    \
 250               {                                                               \
 251                 /* Overflow in the output buffer.  */                         \
 252                 result = __GCONV_FULL_OUTPUT;                                 \
 253                 break;                                                        \
 254               }                                                               \
 255                                                                               \
 256             put16 (outptr, 0xd7c0 + (c >> 10));                               \
 257             outptr += 2;                                                      \
 258             put16 (outptr, 0xdc00 + (c & 0x3ff));                             \
 259           }                                                                   \
 260         else                                                                  \
 261           put16 (outptr, c);                                                  \
 262       }                                                                       \
 263     outptr += 2;                                                              \
 264     inptr += 4;                                                               \
 265   }
 266 #define LOOP_NEED_FLAGS
 267 #define EXTRA_LOOP_DECLS \
 268         , enum variant var, int swap
 269 #include <iconv/loop.c>
 270
 271
 272 /* Convert from UTF-16 to the internal (UCS4-like) format.  */
 273 #define MIN_NEEDED_INPUT        MIN_NEEDED_FROM
 274 #define MAX_NEEDED_INPUT        MAX_NEEDED_FROM
 275 #define MIN_NEEDED_OUTPUT       MIN_NEEDED_TO
 276 #define LOOPFCT                 FROM_LOOP
 277 #define BODY \
 278   {                                                                           \
 279     uint16_t u1 = get16 (inptr);                                              \
 280                                                                               \
 281     if (swap)                                                                 \
 282       {                                                                       \
 283         u1 = bswap_16 (u1);                                                   \
 284                                                                               \
 285         if (__builtin_expect (u1, 0) < 0xd800 || u1 > 0xdfff)                 \
 286           {                                                                   \
 287             /* No surrogate.  */                                              \
 288             put32 (outptr, u1);                                               \
 289             inptr += 2;                                                       \
 290           }                                                                   \
 291         else                                                                  \
 292           {                                                                   \
 293             uint16_t u2;                                                      \
 294                                                                               \
 295             /* It's a surrogate character.  At least the first word says      \
 296                it is.  */                                                     \
 297             if (__builtin_expect (inptr + 4 > inend, 0))                      \
 298               {                                                               \
 299                 /* We don't have enough input for another complete input      \
 300                    character.  */                                             \
 301                 result = __GCONV_INCOMPLETE_INPUT;                            \
 302                 break;                                                        \
 303               }                                                               \
 304                                                                               \
 305             inptr += 2;                                                       \
 306             u2 = bswap_16 (get16 (inptr));                                    \
 307             if (__builtin_expect (u2, 0xdc00) < 0xdc00                        \
 308                 || __builtin_expect (u2, 0xdc00) >= 0xdfff)                   \
 309               {                                                               \
 310                 /* This is no valid second word for a surrogate.  */          \
 311                 if (! ignore_errors_p ())                                     \
 312                   {                                                           \
 313                     inptr -= 2;                                               \
 314                     result = __GCONV_ILLEGAL_INPUT;                           \
 315                     break;                                                    \
 316                   }                                                           \
 317                                                                               \
 318                 ++*irreversible;                                              \
 319                 continue;                                                     \
 320               }                                                               \
 321                                                                               \
 322             put32 (outptr, ((u1 - 0xd7c0) << 10) + (u2 - 0xdc00));            \
 323             inptr += 2;                                                       \
 324           }                                                                   \
 325       }                                                                       \
 326     else                                                                      \
 327       {                                                                       \
 328         if (__builtin_expect (u1, 0) < 0xd800 || u1 > 0xdfff)                 \
 329           {                                                                   \
 330             /* No surrogate.  */                                              \
 331             put32 (outptr, u1);                                               \
 332             inptr += 2;                                                       \
 333           }                                                                   \
 334         else                                                                  \
 335           {                                                                   \
 336             uint16_t u2;                                                      \
 337                                                                               \
 338             /* It's a surrogate character.  At least the first word says      \
 339                it is.  */                                                     \
 340             if (__builtin_expect (inptr + 4 > inend, 0))                      \
 341               {                                                               \
 342                 /* We don't have enough input for another complete input      \
 343                    character.  */                                             \
 344                 result = __GCONV_INCOMPLETE_INPUT;                            \
 345                 break;                                                        \
 346               }                                                               \
 347                                                                               \
 348             inptr += 2;                                                       \
 349             u2 = get16 (inptr);                                               \
 350             if (__builtin_expect (u2, 0xdc00) < 0xdc00                        \
 351                 || __builtin_expect (u2, 0xdc00) >= 0xdfff)                   \
 352               {                                                               \
 353                 /* This is no valid second word for a surrogate.  */          \
 354                 if (! ignore_errors_p ())                                     \
 355                   {                                                           \
 356                     inptr -= 2;                                               \
 357                     result = __GCONV_ILLEGAL_INPUT;                           \
 358                     break;                                                    \
 359                   }                                                           \
 360                                                                               \
 361                 ++*irreversible;                                              \
 362                 continue;                                                     \
 363               }                                                               \
 364                                                                               \
 365             put32 (outptr, ((u1 - 0xd7c0) << 10) + (u2 - 0xdc00));            \
 366             inptr += 2;                                                       \
 367           }                                                                   \
 368       }                                                                       \
 369     outptr += 4;                                                              \
 370   }
 371 #define LOOP_NEED_FLAGS
 372 #define EXTRA_LOOP_DECLS \
 373         , enum variant var, int swap
 374 #include <iconv/loop.c>
 375
 376
 377 /* Now define the toplevel functions.  */
 378 #include <iconv/skeleton.c>