lib/libc/citrus/modules/citrus_utf1632.c

   1 /* $NetBSD: citrus_utf1632.c,v 1.8 2008/03/20 11:47:45 tnozaki Exp $ */
   2 /* $DragonFly: src/lib/libc/citrus/modules/citrus_utf1632.c,v 1.2 2008/04/10 10:21:02 hasso Exp $ */
   3
   4 /*-
   5  * Copyright (c)2003 Citrus Project,
   6  * All rights reserved.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  */
  29
  30 #include <sys/types.h>
  31 #include <sys/endian.h>
  32 #include <assert.h>
  33 #include <errno.h>
  34 #include <limits.h>
  35 #include <locale.h>
  36 #include <stddef.h>
  37 #include <stdio.h>
  38 #include <stdlib.h>
  39 #include <string.h>
  40 #include <wchar.h>
  41
  42 #include "citrus_namespace.h"
  43 #include "citrus_types.h"
  44 #include "citrus_module.h"
  45 #include "citrus_stdenc.h"
  46 #include "citrus_bcs.h"
  47
  48 #include "citrus_utf1632.h"
  49
  50
  51 /* ----------------------------------------------------------------------
  52  * private stuffs used by templates
  53  */
  54
  55 typedef struct {
  56         u_int8_t                ch[4];
  57         int                     chlen;
  58         int                     current_endian;
  59 } _UTF1632State;
  60
  61 typedef struct {
  62         int             preffered_endian;
  63         unsigned int    cur_max;
  64 #define _ENDIAN_UNKNOWN 0
  65 #define _ENDIAN_BIG     1
  66 #define _ENDIAN_LITTLE  2
  67         u_int32_t       mode;
  68 #define _MODE_UTF32             0x00000001U
  69 #define _MODE_FORCE_ENDIAN      0x00000002U
  70 } _UTF1632EncodingInfo;
  71
  72 #define _FUNCNAME(m)                    _citrus_UTF1632_##m
  73 #define _ENCODING_INFO                  _UTF1632EncodingInfo
  74 #define _ENCODING_STATE                 _UTF1632State
  75 #define _ENCODING_MB_CUR_MAX(_ei_)      ((_ei_)->cur_max)
  76 #define _ENCODING_IS_STATE_DEPENDENT    0
  77 #define _STATE_NEEDS_EXPLICIT_INIT(_ps_)        0
  78
  79
  80 static __inline void
  81 /*ARGSUSED*/
  82 _citrus_UTF1632_init_state(_UTF1632EncodingInfo *ei, _UTF1632State *s)
  83 {
  84         memset(s, 0, sizeof(*s));
  85 }
  86
  87 static int
  88 _citrus_UTF1632_mbrtowc_priv(_UTF1632EncodingInfo *ei, wchar_t *pwc,
  89                              const char **s, size_t n, _UTF1632State *psenc,
  90                              size_t *nresult)
  91 {
  92         int chlenbak, endian, needlen;
  93         wchar_t wc;
  94         size_t result;
  95         const char *s0;
  96
  97         _DIAGASSERT(nresult != 0);
  98         _DIAGASSERT(ei != NULL);
  99         _DIAGASSERT(s != NULL);
 100         _DIAGASSERT(psenc != NULL);
 101
 102         s0 = *s;
 103
 104         if (s0 == NULL) {
 105                 _citrus_UTF1632_init_state(ei, psenc);
 106                 *nresult = 0; /* state independent */
 107                 return (0);
 108         }
 109
 110         result = 0;
 111         chlenbak = psenc->chlen;
 112
 113 refetch:
 114         if ((ei->mode & _MODE_UTF32) != 0 || chlenbak>=2)
 115                 needlen = 4;
 116         else
 117                 needlen = 2;
 118
 119         while (chlenbak < needlen) {
 120                 if (n==0)
 121                         goto restart;
 122                 psenc->ch[chlenbak++] = *s0++;
 123                 n--;
 124                 result++;
 125         }
 126
 127         /* judge endian marker */
 128         if ((ei->mode & _MODE_UTF32) == 0) {
 129                 /* UTF16 */
 130                 if (psenc->ch[0]==0xFE && psenc->ch[1]==0xFF) {
 131                         psenc->current_endian = _ENDIAN_BIG;
 132                         chlenbak = 0;
 133                         goto refetch;
 134                 } else if (psenc->ch[0]==0xFF && psenc->ch[1]==0xFE) {
 135                         psenc->current_endian = _ENDIAN_LITTLE;
 136                         chlenbak = 0;
 137                         goto refetch;
 138                 }
 139         } else {
 140                 /* UTF32 */
 141                 if (psenc->ch[0]==0x00 && psenc->ch[1]==0x00 &&
 142                     psenc->ch[2]==0xFE && psenc->ch[3]==0xFF) {
 143                         psenc->current_endian = _ENDIAN_BIG;
 144                         chlenbak = 0;
 145                         goto refetch;
 146                 } else if (psenc->ch[0]==0xFF && psenc->ch[1]==0xFE &&
 147                            psenc->ch[2]==0x00 && psenc->ch[3]==0x00) {
 148                         psenc->current_endian = _ENDIAN_LITTLE;
 149                         chlenbak = 0;
 150                         goto refetch;
 151                 }
 152         }
 153         if ((ei->mode & _MODE_FORCE_ENDIAN) != 0 ||
 154             psenc->current_endian == _ENDIAN_UNKNOWN)
 155                 endian = ei->preffered_endian;
 156         else
 157                 endian = psenc->current_endian;
 158
 159         /* get wc */
 160         if ((ei->mode & _MODE_UTF32) == 0) {
 161                 /* UTF16 */
 162                 if (needlen==2) {
 163                         switch (endian) {
 164                         case _ENDIAN_LITTLE:
 165                                 wc = (psenc->ch[0] |
 166                                       ((wchar_t)psenc->ch[1] << 8));
 167                                 break;
 168                         case _ENDIAN_BIG:
 169                                 wc = (psenc->ch[1] |
 170                                       ((wchar_t)psenc->ch[0] << 8));
 171                                 break;
 172                         default:
 173                                 goto ilseq;
 174                         }
 175                         if (wc >= 0xD800 && wc <= 0xDBFF) {
 176                                 /* surrogate high */
 177                                 needlen=4;
 178                                 goto refetch;
 179                         }
 180                 } else {
 181                         /* surrogate low */
 182                         wc -= 0xD800; /* wc : surrogate high (see above) */
 183                         wc <<= 10;
 184                         switch (endian) {
 185                         case _ENDIAN_LITTLE:
 186                                 if (psenc->ch[2]<0xDC || psenc->ch[2]>0xDF)
 187                                         goto ilseq;
 188                                 wc |= psenc->ch[2];
 189                                 wc |= (wchar_t)(psenc->ch[3] & 3) << 8;
 190                                 break;
 191                         case _ENDIAN_BIG:
 192                                 if (psenc->ch[3]<0xDC || psenc->ch[3]>0xDF)
 193                                         goto ilseq;
 194                                 wc |= psenc->ch[3];
 195                                 wc |= (wchar_t)(psenc->ch[2] & 3) << 8;
 196                                 break;
 197                         default:
 198                                 goto ilseq;
 199                         }
 200                         wc += 0x10000;
 201                 }
 202         } else {
 203                 /* UTF32 */
 204                 switch (endian) {
 205                 case _ENDIAN_LITTLE:
 206                         wc = (psenc->ch[0] |
 207                               ((wchar_t)psenc->ch[1] << 8) |
 208                               ((wchar_t)psenc->ch[2] << 16) |
 209                               ((wchar_t)psenc->ch[3] << 24));
 210                         break;
 211                 case _ENDIAN_BIG:
 212                         wc = (psenc->ch[3] |
 213                               ((wchar_t)psenc->ch[2] << 8) |
 214                               ((wchar_t)psenc->ch[1] << 16) |
 215                               ((wchar_t)psenc->ch[0] << 24));
 216                         break;
 217                 default:
 218                         goto ilseq;
 219                 }
 220                 if (wc >= 0xD800 && wc <= 0xDFFF)
 221                         goto ilseq;
 222         }
 223
 224
 225         *pwc = wc;
 226         psenc->chlen = 0;
 227         *nresult = result;
 228         *s = s0;
 229
 230         return (0);
 231
 232 ilseq:
 233         *nresult = (size_t)-1;
 234         psenc->chlen = 0;
 235         return (EILSEQ);
 236
 237 restart:
 238         *nresult = (size_t)-2;
 239         psenc->chlen = chlenbak;
 240         *s = s0;
 241         return (0);
 242 }
 243
 244 static int
 245 _citrus_UTF1632_wcrtomb_priv(_UTF1632EncodingInfo *ei, char *s, size_t n,
 246                              wchar_t wc, _UTF1632State *psenc,
 247                              size_t *nresult)
 248 {
 249         wchar_t wc2;
 250         static const char _bom[4] = {
 251 #if BYTE_ORDER == BIG_ENDIAN
 252             0x00, 0x00, 0xFE, 0xFF,
 253 #else
 254             0xFF, 0xFE, 0x00, 0x00,
 255 #endif
 256         };
 257         const char *bom = &_bom[0];
 258         size_t cnt;
 259
 260         _DIAGASSERT(ei != NULL);
 261         _DIAGASSERT(nresult != 0);
 262         _DIAGASSERT(s != NULL);
 263
 264         cnt = (size_t)0;
 265         if (psenc->current_endian == _ENDIAN_UNKNOWN) {
 266                 if ((ei->mode & _MODE_FORCE_ENDIAN) == 0) {
 267                         if (ei->mode & _MODE_UTF32) {
 268                                 cnt = 4;
 269                         } else {
 270                                 cnt = 2;
 271 #if BYTE_ORDER == BIG_ENDIAN
 272                                 bom += 2;
 273 #endif
 274                         }
 275                         if (n < cnt)
 276                                 goto e2big;
 277                         memcpy(s, bom, cnt);
 278                         s += cnt, n -= cnt;
 279                 }
 280                 psenc->current_endian = ei->preffered_endian;
 281         }
 282
 283         wc2 = 0;
 284         if ((ei->mode & _MODE_UTF32)==0) {
 285                 /* UTF16 */
 286                 if (wc>0xFFFF) {
 287                         /* surrogate */
 288                         if (wc>0x10FFFF)
 289                                 goto ilseq;
 290                         if (n < 4)
 291                                 goto e2big;
 292                         cnt += 4;
 293                         wc -= 0x10000;
 294                         wc2 = (wc & 0x3FF) | 0xDC00;
 295                         wc = (wc>>10) | 0xD800;
 296                 } else {
 297                         if (n < 2)
 298                                 goto e2big;
 299                         cnt += 2;
 300                 }
 301
 302 surrogate:
 303                 switch (psenc->current_endian) {
 304                 case _ENDIAN_BIG:
 305                         s[1] = wc;
 306                         s[0] = (wc >>= 8);
 307                         break;
 308                 case _ENDIAN_LITTLE:
 309                         s[0] = wc;
 310                         s[1] = (wc >>= 8);
 311                         break;
 312                 }
 313                 if (wc2!=0) {
 314                         wc = wc2;
 315                         wc2 = 0;
 316                         s += 2;
 317                         goto surrogate;
 318                 }
 319         } else {
 320                 /* UTF32 */
 321                 if (wc >= 0xD800 && wc <= 0xDFFF)
 322                         goto ilseq;
 323                 if (n < 4)
 324                         goto e2big;
 325                 cnt += 4;
 326                 switch (psenc->current_endian) {
 327                 case _ENDIAN_BIG:
 328                         s[3] = wc;
 329                         s[2] = (wc >>= 8);
 330                         s[1] = (wc >>= 8);
 331                         s[0] = (wc >>= 8);
 332                         break;
 333                 case _ENDIAN_LITTLE:
 334                         s[0] = wc;
 335                         s[1] = (wc >>= 8);
 336                         s[2] = (wc >>= 8);
 337                         s[3] = (wc >>= 8);
 338                         break;
 339                 }
 340         }
 341         *nresult = cnt;
 342
 343         return 0;
 344
 345 ilseq:
 346         *nresult = (size_t)-1;
 347         return EILSEQ;
 348 e2big:
 349         *nresult = (size_t)-1;
 350         return E2BIG;
 351 }
 352
 353 static void
 354 parse_variable(_UTF1632EncodingInfo * __restrict ei,
 355                const void * __restrict var, size_t lenvar)
 356 {
 357 #define MATCH(x, act)                                           \
 358 do {                                                            \
 359         if (lenvar >= (sizeof(#x)-1) &&                         \
 360             _bcs_strncasecmp(p, #x, sizeof(#x)-1) == 0) {       \
 361                 act;                                            \
 362                 lenvar -= sizeof(#x)-1;                         \
 363                 p += sizeof(#x)-1;                              \
 364         }                                                       \
 365 } while (/*CONSTCOND*/0)
 366         const char *p;
 367         p = var;
 368         while (lenvar>0) {
 369                 switch (*p) {
 370                 case 'B':
 371                 case 'b':
 372                         MATCH(big, ei->preffered_endian = _ENDIAN_BIG);
 373                         break;
 374                 case 'L':
 375                 case 'l':
 376                         MATCH(little, ei->preffered_endian = _ENDIAN_LITTLE);
 377                         break;
 378                 case 'F':
 379                 case 'f':
 380                         MATCH(force, ei->mode |= _MODE_FORCE_ENDIAN);
 381                         break;
 382                 case 'U':
 383                 case 'u':
 384                         MATCH(utf32, ei->mode |= _MODE_UTF32);
 385                         break;
 386                 }
 387                 p++;
 388                 lenvar--;
 389         }
 390 }
 391
 392 static int
 393 /*ARGSUSED*/
 394 _citrus_UTF1632_encoding_module_init(_UTF1632EncodingInfo * __restrict ei,
 395                                      const void * __restrict var,
 396                                      size_t lenvar)
 397 {
 398         _DIAGASSERT(ei != NULL);
 399
 400         memset((void *)ei, 0, sizeof(*ei));
 401
 402         parse_variable(ei, var, lenvar);
 403
 404         if ((ei->mode&_MODE_UTF32)==0)
 405                 ei->cur_max = 6; /* endian + surrogate */
 406         else
 407                 ei->cur_max = 8; /* endian + normal */
 408
 409         if (ei->preffered_endian == _ENDIAN_UNKNOWN) {
 410 #if BYTE_ORDER == BIG_ENDIAN
 411                 ei->preffered_endian = _ENDIAN_BIG;
 412 #else
 413                 ei->preffered_endian = _ENDIAN_LITTLE;
 414 #endif
 415         }
 416
 417         return (0);
 418 }
 419
 420 static void
 421 /*ARGSUSED*/
 422 _citrus_UTF1632_encoding_module_uninit(_UTF1632EncodingInfo *ei)
 423 {
 424 }
 425
 426 static __inline int
 427 /*ARGSUSED*/
 428 _citrus_UTF1632_stdenc_wctocs(_UTF1632EncodingInfo * __restrict ei,
 429                               _csid_t * __restrict csid,
 430                               _index_t * __restrict idx,
 431                               _wc_t wc)
 432 {
 433
 434         _DIAGASSERT(csid != NULL && idx != NULL);
 435
 436         *csid = 0;
 437         *idx = (_index_t)wc;
 438
 439         return (0);
 440 }
 441
 442 static __inline int
 443 /*ARGSUSED*/
 444 _citrus_UTF1632_stdenc_cstowc(_UTF1632EncodingInfo * __restrict ei,
 445                               _wc_t * __restrict wc,
 446                               _csid_t csid, _index_t idx)
 447 {
 448
 449         _DIAGASSERT(wc != NULL);
 450
 451         if (csid != 0)
 452                 return (EILSEQ);
 453
 454         *wc = (_wc_t)idx;
 455
 456         return (0);
 457 }
 458
 459 static __inline int
 460 /*ARGSUSED*/
 461 _citrus_UTF1632_stdenc_get_state_desc_generic(_UTF1632EncodingInfo * __restrict ei,
 462                                               _UTF1632State * __restrict psenc,
 463                                               int * __restrict rstate)
 464 {
 465
 466         if (psenc->chlen == 0)
 467                 *rstate = _STDENC_SDGEN_INITIAL;
 468         else
 469                 *rstate = _STDENC_SDGEN_INCOMPLETE_CHAR;
 470
 471         return 0;
 472 }
 473
 474 /* ----------------------------------------------------------------------
 475  * public interface for stdenc
 476  */
 477
 478 _CITRUS_STDENC_DECLS(UTF1632);
 479 _CITRUS_STDENC_DEF_OPS(UTF1632);
 480
 481 #include "citrus_stdenc_template.h"