workbench/devs/diskimage/expat_library/xmltok.c

   1 /* Copyright (c) 1998, 1999 Thai Open Source Software Center Ltd
   2    See the file COPYING for copying permission.
   3 */
   4
   5 #include "amigaconfig.h"
   6
   7 #include <stddef.h>
   8
   9 #ifdef COMPILED_FROM_DSP
  10 #include "winconfig.h"
  11 #elif defined(MACOS_CLASSIC)
  12 #include "macconfig.h"
  13 #elif defined(__amigaos4__)
  14 #include "amigaconfig.h"
  15 #elif defined(__WATCOMC__)
  16 #include "watcomconfig.h"
  17 #else
  18 #ifdef HAVE_EXPAT_CONFIG_H
  19 #include <expat_config.h>
  20 #endif
  21 #endif /* ndef COMPILED_FROM_DSP */
  22
  23 #include "expat_external.h"
  24 #include "internal.h"
  25 #include "xmltok.h"
  26 #include "nametab.h"
  27
  28 #ifdef XML_DTD
  29 #define IGNORE_SECTION_TOK_VTABLE , PREFIX(ignoreSectionTok)
  30 #else
  31 #define IGNORE_SECTION_TOK_VTABLE /* as nothing */
  32 #endif
  33
  34 #define VTABLE1 \
  35   { PREFIX(prologTok), PREFIX(contentTok), \
  36     PREFIX(cdataSectionTok) IGNORE_SECTION_TOK_VTABLE }, \
  37   { PREFIX(attributeValueTok), PREFIX(entityValueTok) }, \
  38   PREFIX(sameName), \
  39   PREFIX(nameMatchesAscii), \
  40   PREFIX(nameLength), \
  41   PREFIX(skipS), \
  42   PREFIX(getAtts), \
  43   PREFIX(charRefNumber), \
  44   PREFIX(predefinedEntityName), \
  45   PREFIX(updatePosition), \
  46   PREFIX(isPublicId)
  47
  48 #define VTABLE VTABLE1, PREFIX(toUtf8), PREFIX(toUtf16)
  49
  50 #define UCS2_GET_NAMING(pages, hi, lo) \
  51    (namingBitmap[(pages[hi] << 3) + ((lo) >> 5)] & (1 << ((lo) & 0x1F)))
  52
  53 /* A 2 byte UTF-8 representation splits the characters 11 bits between
  54    the bottom 5 and 6 bits of the bytes.  We need 8 bits to index into
  55    pages, 3 bits to add to that index and 5 bits to generate the mask.
  56 */
  57 #define UTF8_GET_NAMING2(pages, byte) \
  58     (namingBitmap[((pages)[(((byte)[0]) >> 2) & 7] << 3) \
  59                       + ((((byte)[0]) & 3) << 1) \
  60                       + ((((byte)[1]) >> 5) & 1)] \
  61          & (1 << (((byte)[1]) & 0x1F)))
  62
  63 /* A 3 byte UTF-8 representation splits the characters 16 bits between
  64    the bottom 4, 6 and 6 bits of the bytes.  We need 8 bits to index
  65    into pages, 3 bits to add to that index and 5 bits to generate the
  66    mask.
  67 */
  68 #define UTF8_GET_NAMING3(pages, byte) \
  69   (namingBitmap[((pages)[((((byte)[0]) & 0xF) << 4) \
  70                              + ((((byte)[1]) >> 2) & 0xF)] \
  71                        << 3) \
  72                       + ((((byte)[1]) & 3) << 1) \
  73                       + ((((byte)[2]) >> 5) & 1)] \
  74          & (1 << (((byte)[2]) & 0x1F)))
  75
  76 #define UTF8_GET_NAMING(pages, p, n) \
  77   ((n) == 2 \
  78   ? UTF8_GET_NAMING2(pages, (const unsigned char *)(p)) \
  79   : ((n) == 3 \
  80      ? UTF8_GET_NAMING3(pages, (const unsigned char *)(p)) \
  81      : 0))
  82
  83 /* Detection of invalid UTF-8 sequences is based on Table 3.1B
  84    of Unicode 3.2: http://www.unicode.org/unicode/reports/tr28/
  85    with the additional restriction of not allowing the Unicode
  86    code points 0xFFFF and 0xFFFE (sequences EF,BF,BF and EF,BF,BE).
  87    Implementation details:
  88      (A & 0x80) == 0     means A < 0x80
  89    and
  90      (A & 0xC0) == 0xC0  means A > 0xBF
  91 */
  92
  93 #define UTF8_INVALID2(p) \
  94   ((*p) < 0xC2 || ((p)[1] & 0x80) == 0 || ((p)[1] & 0xC0) == 0xC0)
  95
  96 #define UTF8_INVALID3(p) \
  97   (((p)[2] & 0x80) == 0 \
  98   || \
  99   ((*p) == 0xEF && (p)[1] == 0xBF \
 100     ? \
 101     (p)[2] > 0xBD \
 102     : \
 103     ((p)[2] & 0xC0) == 0xC0) \
 104   || \
 105   ((*p) == 0xE0 \
 106     ? \
 107     (p)[1] < 0xA0 || ((p)[1] & 0xC0) == 0xC0 \
 108     : \
 109     ((p)[1] & 0x80) == 0 \
 110     || \
 111     ((*p) == 0xED ? (p)[1] > 0x9F : ((p)[1] & 0xC0) == 0xC0)))
 112
 113 #define UTF8_INVALID4(p) \
 114   (((p)[3] & 0x80) == 0 || ((p)[3] & 0xC0) == 0xC0 \
 115   || \
 116   ((p)[2] & 0x80) == 0 || ((p)[2] & 0xC0) == 0xC0 \
 117   || \
 118   ((*p) == 0xF0 \
 119     ? \
 120     (p)[1] < 0x90 || ((p)[1] & 0xC0) == 0xC0 \
 121     : \
 122     ((p)[1] & 0x80) == 0 \
 123     || \
 124     ((*p) == 0xF4 ? (p)[1] > 0x8F : ((p)[1] & 0xC0) == 0xC0)))
 125
 126 static int PTRFASTCALL
 127 isNever(const ENCODING *enc, const char *p)
 128 {
 129   return 0;
 130 }
 131
 132 static int PTRFASTCALL
 133 utf8_isName2(const ENCODING *enc, const char *p)
 134 {
 135   return UTF8_GET_NAMING2(namePages, (const unsigned char *)p);
 136 }
 137
 138 static int PTRFASTCALL
 139 utf8_isName3(const ENCODING *enc, const char *p)
 140 {
 141   return UTF8_GET_NAMING3(namePages, (const unsigned char *)p);
 142 }
 143
 144 #define utf8_isName4 isNever
 145
 146 static int PTRFASTCALL
 147 utf8_isNmstrt2(const ENCODING *enc, const char *p)
 148 {
 149   return UTF8_GET_NAMING2(nmstrtPages, (const unsigned char *)p);
 150 }
 151
 152 static int PTRFASTCALL
 153 utf8_isNmstrt3(const ENCODING *enc, const char *p)
 154 {
 155   return UTF8_GET_NAMING3(nmstrtPages, (const unsigned char *)p);
 156 }
 157
 158 #define utf8_isNmstrt4 isNever
 159
 160 static int PTRFASTCALL
 161 utf8_isInvalid2(const ENCODING *enc, const char *p)
 162 {
 163   return UTF8_INVALID2((const unsigned char *)p);
 164 }
 165
 166 static int PTRFASTCALL
 167 utf8_isInvalid3(const ENCODING *enc, const char *p)
 168 {
 169   return UTF8_INVALID3((const unsigned char *)p);
 170 }
 171
 172 static int PTRFASTCALL
 173 utf8_isInvalid4(const ENCODING *enc, const char *p)
 174 {
 175   return UTF8_INVALID4((const unsigned char *)p);
 176 }
 177
 178 struct normal_encoding {
 179   ENCODING enc;
 180   unsigned char type[256];
 181 #ifdef XML_MIN_SIZE
 182   int (PTRFASTCALL *byteType)(const ENCODING *, const char *);
 183   int (PTRFASTCALL *isNameMin)(const ENCODING *, const char *);
 184   int (PTRFASTCALL *isNmstrtMin)(const ENCODING *, const char *);
 185   int (PTRFASTCALL *byteToAscii)(const ENCODING *, const char *);
 186   int (PTRCALL *charMatches)(const ENCODING *, const char *, int);
 187 #endif /* XML_MIN_SIZE */
 188   int (PTRFASTCALL *isName2)(const ENCODING *, const char *);
 189   int (PTRFASTCALL *isName3)(const ENCODING *, const char *);
 190   int (PTRFASTCALL *isName4)(const ENCODING *, const char *);
 191   int (PTRFASTCALL *isNmstrt2)(const ENCODING *, const char *);
 192   int (PTRFASTCALL *isNmstrt3)(const ENCODING *, const char *);
 193   int (PTRFASTCALL *isNmstrt4)(const ENCODING *, const char *);
 194   int (PTRFASTCALL *isInvalid2)(const ENCODING *, const char *);
 195   int (PTRFASTCALL *isInvalid3)(const ENCODING *, const char *);
 196   int (PTRFASTCALL *isInvalid4)(const ENCODING *, const char *);
 197 };
 198
 199 #define AS_NORMAL_ENCODING(enc)   ((const struct normal_encoding *) (enc))
 200
 201 #ifdef XML_MIN_SIZE
 202
 203 #define STANDARD_VTABLE(E) \
 204  E ## byteType, \
 205  E ## isNameMin, \
 206  E ## isNmstrtMin, \
 207  E ## byteToAscii, \
 208  E ## charMatches,
 209
 210 #else
 211
 212 #define STANDARD_VTABLE(E) /* as nothing */
 213
 214 #endif
 215
 216 #define NORMAL_VTABLE(E) \
 217  E ## isName2, \
 218  E ## isName3, \
 219  E ## isName4, \
 220  E ## isNmstrt2, \
 221  E ## isNmstrt3, \
 222  E ## isNmstrt4, \
 223  E ## isInvalid2, \
 224  E ## isInvalid3, \
 225  E ## isInvalid4
 226
 227 static int FASTCALL checkCharRefNumber(int);
 228
 229 #include "xmltok_impl.h"
 230 #include "ascii.h"
 231
 232 #ifdef XML_MIN_SIZE
 233 #define sb_isNameMin isNever
 234 #define sb_isNmstrtMin isNever
 235 #endif
 236
 237 #ifdef XML_MIN_SIZE
 238 #define MINBPC(enc) ((enc)->minBytesPerChar)
 239 #else
 240 /* minimum bytes per character */
 241 #define MINBPC(enc) 1
 242 #endif
 243
 244 #define SB_BYTE_TYPE(enc, p) \
 245   (((struct normal_encoding *)(enc))->type[(unsigned char)*(p)])
 246
 247 #ifdef XML_MIN_SIZE
 248 static int PTRFASTCALL
 249 sb_byteType(const ENCODING *enc, const char *p)
 250 {
 251   return SB_BYTE_TYPE(enc, p);
 252 }
 253 #define BYTE_TYPE(enc, p) \
 254  (AS_NORMAL_ENCODING(enc)->byteType(enc, p))
 255 #else
 256 #define BYTE_TYPE(enc, p) SB_BYTE_TYPE(enc, p)
 257 #endif
 258
 259 #ifdef XML_MIN_SIZE
 260 #define BYTE_TO_ASCII(enc, p) \
 261  (AS_NORMAL_ENCODING(enc)->byteToAscii(enc, p))
 262 static int PTRFASTCALL
 263 sb_byteToAscii(const ENCODING *enc, const char *p)
 264 {
 265   return *p;
 266 }
 267 #else
 268 #define BYTE_TO_ASCII(enc, p) (*(p))
 269 #endif
 270
 271 #define IS_NAME_CHAR(enc, p, n) \
 272  (AS_NORMAL_ENCODING(enc)->isName ## n(enc, p))
 273 #define IS_NMSTRT_CHAR(enc, p, n) \
 274  (AS_NORMAL_ENCODING(enc)->isNmstrt ## n(enc, p))
 275 #define IS_INVALID_CHAR(enc, p, n) \
 276  (AS_NORMAL_ENCODING(enc)->isInvalid ## n(enc, p))
 277
 278 #ifdef XML_MIN_SIZE
 279 #define IS_NAME_CHAR_MINBPC(enc, p) \
 280  (AS_NORMAL_ENCODING(enc)->isNameMin(enc, p))
 281 #define IS_NMSTRT_CHAR_MINBPC(enc, p) \
 282  (AS_NORMAL_ENCODING(enc)->isNmstrtMin(enc, p))
 283 #else
 284 #define IS_NAME_CHAR_MINBPC(enc, p) (0)
 285 #define IS_NMSTRT_CHAR_MINBPC(enc, p) (0)
 286 #endif
 287
 288 #ifdef XML_MIN_SIZE
 289 #define CHAR_MATCHES(enc, p, c) \
 290  (AS_NORMAL_ENCODING(enc)->charMatches(enc, p, c))
 291 static int PTRCALL
 292 sb_charMatches(const ENCODING *enc, const char *p, int c)
 293 {
 294   return *p == c;
 295 }
 296 #else
 297 /* c is an ASCII character */
 298 #define CHAR_MATCHES(enc, p, c) (*(p) == c)
 299 #endif
 300
 301 #define PREFIX(ident) normal_ ## ident
 302 #define XML_TOK_IMPL_C
 303 #include "xmltok_impl.c"
 304 #undef XML_TOK_IMPL_C
 305
 306 #undef MINBPC
 307 #undef BYTE_TYPE
 308 #undef BYTE_TO_ASCII
 309 #undef CHAR_MATCHES
 310 #undef IS_NAME_CHAR
 311 #undef IS_NAME_CHAR_MINBPC
 312 #undef IS_NMSTRT_CHAR
 313 #undef IS_NMSTRT_CHAR_MINBPC
 314 #undef IS_INVALID_CHAR
 315
 316 enum {  /* UTF8_cvalN is value of masked first byte of N byte sequence */
 317   UTF8_cval1 = 0x00,
 318   UTF8_cval2 = 0xc0,
 319   UTF8_cval3 = 0xe0,
 320   UTF8_cval4 = 0xf0
 321 };
 322
 323 static void PTRCALL
 324 utf8_toUtf8(const ENCODING *enc,
 325             const char **fromP, const char *fromLim,
 326             char **toP, const char *toLim)
 327 {
 328   char *to;
 329   const char *from;
 330   if (fromLim - *fromP > toLim - *toP) {
 331     /* Avoid copying partial characters. */
 332     for (fromLim = *fromP + (toLim - *toP); fromLim > *fromP; fromLim--)
 333       if (((unsigned char)fromLim[-1] & 0xc0) != 0x80)
 334         break;
 335   }
 336   for (to = *toP, from = *fromP; from != fromLim; from++, to++)
 337     *to = *from;
 338   *fromP = from;
 339   *toP = to;
 340 }
 341
 342 static void PTRCALL
 343 utf8_toUtf16(const ENCODING *enc,
 344              const char **fromP, const char *fromLim,
 345              unsigned short **toP, const unsigned short *toLim)
 346 {
 347   unsigned short *to = *toP;
 348   const char *from = *fromP;
 349   while (from != fromLim && to != toLim) {
 350     switch (((struct normal_encoding *)enc)->type[(unsigned char)*from]) {
 351     case BT_LEAD2:
 352       *to++ = (unsigned short)(((from[0] & 0x1f) << 6) | (from[1] & 0x3f));
 353       from += 2;
 354       break;
 355     case BT_LEAD3:
 356       *to++ = (unsigned short)(((from[0] & 0xf) << 12)
 357                                | ((from[1] & 0x3f) << 6) | (from[2] & 0x3f));
 358       from += 3;
 359       break;
 360     case BT_LEAD4:
 361       {
 362         unsigned long n;
 363         if (to + 1 == toLim)
 364           goto after;
 365         n = ((from[0] & 0x7) << 18) | ((from[1] & 0x3f) << 12)
 366             | ((from[2] & 0x3f) << 6) | (from[3] & 0x3f);
 367         n -= 0x10000;
 368         to[0] = (unsigned short)((n >> 10) | 0xD800);
 369         to[1] = (unsigned short)((n & 0x3FF) | 0xDC00);
 370         to += 2;
 371         from += 4;
 372       }
 373       break;
 374     default:
 375       *to++ = *from++;
 376       break;
 377     }
 378   }
 379 after:
 380   *fromP = from;
 381   *toP = to;
 382 }
 383
 384 #ifdef XML_NS
 385 static const struct normal_encoding utf8_encoding_ns = {
 386   { VTABLE1, utf8_toUtf8, utf8_toUtf16, 1, 1, 0 },
 387   {
 388 #include "asciitab.h"
 389 #include "utf8tab.h"
 390   },
 391   STANDARD_VTABLE(sb_) NORMAL_VTABLE(utf8_)
 392 };
 393 #endif
 394
 395 static const struct normal_encoding utf8_encoding = {
 396   { VTABLE1, utf8_toUtf8, utf8_toUtf16, 1, 1, 0 },
 397   {
 398 #define BT_COLON BT_NMSTRT
 399 #include "asciitab.h"
 400 #undef BT_COLON
 401 #include "utf8tab.h"
 402   },
 403   STANDARD_VTABLE(sb_) NORMAL_VTABLE(utf8_)
 404 };
 405
 406 #ifdef XML_NS
 407
 408 static const struct normal_encoding internal_utf8_encoding_ns = {
 409   { VTABLE1, utf8_toUtf8, utf8_toUtf16, 1, 1, 0 },
 410   {
 411 #include "iasciitab.h"
 412 #include "utf8tab.h"
 413   },
 414   STANDARD_VTABLE(sb_) NORMAL_VTABLE(utf8_)
 415 };
 416
 417 #endif
 418
 419 static const struct normal_encoding internal_utf8_encoding = {
 420   { VTABLE1, utf8_toUtf8, utf8_toUtf16, 1, 1, 0 },
 421   {
 422 #define BT_COLON BT_NMSTRT
 423 #include "iasciitab.h"
 424 #undef BT_COLON
 425 #include "utf8tab.h"
 426   },
 427   STANDARD_VTABLE(sb_) NORMAL_VTABLE(utf8_)
 428 };
 429
 430 static void PTRCALL
 431 latin1_toUtf8(const ENCODING *enc,
 432               const char **fromP, const char *fromLim,
 433               char **toP, const char *toLim)
 434 {
 435   for (;;) {
 436     unsigned char c;
 437     if (*fromP == fromLim)
 438       break;
 439     c = (unsigned char)**fromP;
 440     if (c & 0x80) {
 441       if (toLim - *toP < 2)
 442         break;
 443       *(*toP)++ = (char)((c >> 6) | UTF8_cval2);
 444       *(*toP)++ = (char)((c & 0x3f) | 0x80);
 445       (*fromP)++;
 446     }
 447     else {
 448       if (*toP == toLim)
 449         break;
 450       *(*toP)++ = *(*fromP)++;
 451     }
 452   }
 453 }
 454
 455 static void PTRCALL
 456 latin1_toUtf16(const ENCODING *enc,
 457                const char **fromP, const char *fromLim,
 458                unsigned short **toP, const unsigned short *toLim)
 459 {
 460   while (*fromP != fromLim && *toP != toLim)
 461     *(*toP)++ = (unsigned char)*(*fromP)++;
 462 }
 463
 464 #ifdef XML_NS
 465
 466 static const struct normal_encoding latin1_encoding_ns = {
 467   { VTABLE1, latin1_toUtf8, latin1_toUtf16, 1, 0, 0 },
 468   {
 469 #include "asciitab.h"
 470 #include "latin1tab.h"
 471   },
 472   STANDARD_VTABLE(sb_)
 473 };
 474
 475 #endif
 476
 477 static const struct normal_encoding latin1_encoding = {
 478   { VTABLE1, latin1_toUtf8, latin1_toUtf16, 1, 0, 0 },
 479   {
 480 #define BT_COLON BT_NMSTRT
 481 #include "asciitab.h"
 482 #undef BT_COLON
 483 #include "latin1tab.h"
 484   },
 485   STANDARD_VTABLE(sb_)
 486 };
 487
 488 static void PTRCALL
 489 ascii_toUtf8(const ENCODING *enc,
 490              const char **fromP, const char *fromLim,
 491              char **toP, const char *toLim)
 492 {
 493   while (*fromP != fromLim && *toP != toLim)
 494     *(*toP)++ = *(*fromP)++;
 495 }
 496
 497 #ifdef XML_NS
 498
 499 static const struct normal_encoding ascii_encoding_ns = {
 500   { VTABLE1, ascii_toUtf8, latin1_toUtf16, 1, 1, 0 },
 501   {
 502 #include "asciitab.h"
 503 /* BT_NONXML == 0 */
 504   },
 505   STANDARD_VTABLE(sb_)
 506 };
 507
 508 #endif
 509
 510 static const struct normal_encoding ascii_encoding = {
 511   { VTABLE1, ascii_toUtf8, latin1_toUtf16, 1, 1, 0 },
 512   {
 513 #define BT_COLON BT_NMSTRT
 514 #include "asciitab.h"
 515 #undef BT_COLON
 516 /* BT_NONXML == 0 */
 517   },
 518   STANDARD_VTABLE(sb_)
 519 };
 520
 521 static int PTRFASTCALL
 522 unicode_byte_type(char hi, char lo)
 523 {
 524   switch ((unsigned char)hi) {
 525   case 0xD8: case 0xD9: case 0xDA: case 0xDB:
 526     return BT_LEAD4;
 527   case 0xDC: case 0xDD: case 0xDE: case 0xDF:
 528     return BT_TRAIL;
 529   case 0xFF:
 530     switch ((unsigned char)lo) {
 531     case 0xFF:
 532     case 0xFE:
 533       return BT_NONXML;
 534     }
 535     break;
 536   }
 537   return BT_NONASCII;
 538 }
 539
 540 #define DEFINE_UTF16_TO_UTF8(E) \
 541 static void  PTRCALL \
 542 E ## toUtf8(const ENCODING *enc, \
 543             const char **fromP, const char *fromLim, \
 544             char **toP, const char *toLim) \
 545 { \
 546   const char *from; \
 547   for (from = *fromP; from != fromLim; from += 2) { \
 548     int plane; \
 549     unsigned char lo2; \
 550     unsigned char lo = GET_LO(from); \
 551     unsigned char hi = GET_HI(from); \
 552     switch (hi) { \
 553     case 0: \
 554       if (lo < 0x80) { \
 555         if (*toP == toLim) { \
 556           *fromP = from; \
 557           return; \
 558         } \
 559         *(*toP)++ = lo; \
 560         break; \
 561       } \
 562       /* fall through */ \
 563     case 0x1: case 0x2: case 0x3: \
 564     case 0x4: case 0x5: case 0x6: case 0x7: \
 565       if (toLim -  *toP < 2) { \
 566         *fromP = from; \
 567         return; \
 568       } \
 569       *(*toP)++ = ((lo >> 6) | (hi << 2) |  UTF8_cval2); \
 570       *(*toP)++ = ((lo & 0x3f) | 0x80); \
 571       break; \
 572     default: \
 573       if (toLim -  *toP < 3)  { \
 574         *fromP = from; \
 575         return; \
 576       } \
 577       /* 16 bits divided 4, 6, 6 amongst 3 bytes */ \
 578       *(*toP)++ = ((hi >> 4) | UTF8_cval3); \
 579       *(*toP)++ = (((hi & 0xf) << 2) | (lo >> 6) | 0x80); \
 580       *(*toP)++ = ((lo & 0x3f) | 0x80); \
 581       break; \
 582     case 0xD8: case 0xD9: case 0xDA: case 0xDB: \
 583       if (toLim -  *toP < 4) { \
 584         *fromP = from; \
 585         return; \
 586       } \
 587       plane = (((hi & 0x3) << 2) | ((lo >> 6) & 0x3)) + 1; \
 588       *(*toP)++ = ((plane >> 2) | UTF8_cval4); \
 589       *(*toP)++ = (((lo >> 2) & 0xF) | ((plane & 0x3) << 4) | 0x80); \
 590       from += 2; \
 591       lo2 = GET_LO(from); \
 592       *(*toP)++ = (((lo & 0x3) << 4) \
 593                    | ((GET_HI(from) & 0x3) << 2) \
 594                    | (lo2 >> 6) \
 595                    | 0x80); \
 596       *(*toP)++ = ((lo2 & 0x3f) | 0x80); \
 597       break; \
 598     } \
 599   } \
 600   *fromP = from; \
 601 }
 602
 603 #define DEFINE_UTF16_TO_UTF16(E) \
 604 static void  PTRCALL \
 605 E ## toUtf16(const ENCODING *enc, \
 606              const char **fromP, const char *fromLim, \
 607              unsigned short **toP, const unsigned short *toLim) \
 608 { \
 609   /* Avoid copying first half only of surrogate */ \
 610   if (fromLim - *fromP > ((toLim - *toP) << 1) \
 611       && (GET_HI(fromLim - 2) & 0xF8) == 0xD8) \
 612     fromLim -= 2; \
 613   for (; *fromP != fromLim && *toP != toLim; *fromP += 2) \
 614     *(*toP)++ = (GET_HI(*fromP) << 8) | GET_LO(*fromP); \
 615 }
 616
 617 #define SET2(ptr, ch) \
 618   (((ptr)[0] = ((ch) & 0xff)), ((ptr)[1] = ((ch) >> 8)))
 619 #define GET_LO(ptr) ((unsigned char)(ptr)[0])
 620 #define GET_HI(ptr) ((unsigned char)(ptr)[1])
 621
 622 DEFINE_UTF16_TO_UTF8(little2_)
 623 DEFINE_UTF16_TO_UTF16(little2_)
 624
 625 #undef SET2
 626 #undef GET_LO
 627 #undef GET_HI
 628
 629 #define SET2(ptr, ch) \
 630   (((ptr)[0] = ((ch) >> 8)), ((ptr)[1] = ((ch) & 0xFF)))
 631 #define GET_LO(ptr) ((unsigned char)(ptr)[1])
 632 #define GET_HI(ptr) ((unsigned char)(ptr)[0])
 633
 634 DEFINE_UTF16_TO_UTF8(big2_)
 635 DEFINE_UTF16_TO_UTF16(big2_)
 636
 637 #undef SET2
 638 #undef GET_LO
 639 #undef GET_HI
 640
 641 #define LITTLE2_BYTE_TYPE(enc, p) \
 642  ((p)[1] == 0 \
 643   ? ((struct normal_encoding *)(enc))->type[(unsigned char)*(p)] \
 644   : unicode_byte_type((p)[1], (p)[0]))
 645 #define LITTLE2_BYTE_TO_ASCII(enc, p) ((p)[1] == 0 ? (p)[0] : -1)
 646 #define LITTLE2_CHAR_MATCHES(enc, p, c) ((p)[1] == 0 && (p)[0] == c)
 647 #define LITTLE2_IS_NAME_CHAR_MINBPC(enc, p) \
 648   UCS2_GET_NAMING(namePages, (unsigned char)p[1], (unsigned char)p[0])
 649 #define LITTLE2_IS_NMSTRT_CHAR_MINBPC(enc, p) \
 650   UCS2_GET_NAMING(nmstrtPages, (unsigned char)p[1], (unsigned char)p[0])
 651
 652 #ifdef XML_MIN_SIZE
 653
 654 static int PTRFASTCALL
 655 little2_byteType(const ENCODING *enc, const char *p)
 656 {
 657   return LITTLE2_BYTE_TYPE(enc, p);
 658 }
 659
 660 static int PTRFASTCALL
 661 little2_byteToAscii(const ENCODING *enc, const char *p)
 662 {
 663   return LITTLE2_BYTE_TO_ASCII(enc, p);
 664 }
 665
 666 static int PTRCALL
 667 little2_charMatches(const ENCODING *enc, const char *p, int c)
 668 {
 669   return LITTLE2_CHAR_MATCHES(enc, p, c);
 670 }
 671
 672 static int PTRFASTCALL
 673 little2_isNameMin(const ENCODING *enc, const char *p)
 674 {
 675   return LITTLE2_IS_NAME_CHAR_MINBPC(enc, p);
 676 }
 677
 678 static int PTRFASTCALL
 679 little2_isNmstrtMin(const ENCODING *enc, const char *p)
 680 {
 681   return LITTLE2_IS_NMSTRT_CHAR_MINBPC(enc, p);
 682 }
 683
 684 #undef VTABLE
 685 #define VTABLE VTABLE1, little2_toUtf8, little2_toUtf16
 686
 687 #else /* not XML_MIN_SIZE */
 688
 689 #undef PREFIX
 690 #define PREFIX(ident) little2_ ## ident
 691 #define MINBPC(enc) 2
 692 /* CHAR_MATCHES is guaranteed to have MINBPC bytes available. */
 693 #define BYTE_TYPE(enc, p) LITTLE2_BYTE_TYPE(enc, p)
 694 #define BYTE_TO_ASCII(enc, p) LITTLE2_BYTE_TO_ASCII(enc, p)
 695 #define CHAR_MATCHES(enc, p, c) LITTLE2_CHAR_MATCHES(enc, p, c)
 696 #define IS_NAME_CHAR(enc, p, n) 0
 697 #define IS_NAME_CHAR_MINBPC(enc, p) LITTLE2_IS_NAME_CHAR_MINBPC(enc, p)
 698 #define IS_NMSTRT_CHAR(enc, p, n) (0)
 699 #define IS_NMSTRT_CHAR_MINBPC(enc, p) LITTLE2_IS_NMSTRT_CHAR_MINBPC(enc, p)
 700
 701 #define XML_TOK_IMPL_C
 702 #include "xmltok_impl.c"
 703 #undef XML_TOK_IMPL_C
 704
 705 #undef MINBPC
 706 #undef BYTE_TYPE
 707 #undef BYTE_TO_ASCII
 708 #undef CHAR_MATCHES
 709 #undef IS_NAME_CHAR
 710 #undef IS_NAME_CHAR_MINBPC
 711 #undef IS_NMSTRT_CHAR
 712 #undef IS_NMSTRT_CHAR_MINBPC
 713 #undef IS_INVALID_CHAR
 714
 715 #endif /* not XML_MIN_SIZE */
 716
 717 #ifdef XML_NS
 718
 719 static const struct normal_encoding little2_encoding_ns = {
 720   { VTABLE, 2, 0,
 721 #if BYTEORDER == 1234
 722     1
 723 #else
 724     0
 725 #endif
 726   },
 727   {
 728 #include "asciitab.h"
 729 #include "latin1tab.h"
 730   },
 731   STANDARD_VTABLE(little2_)
 732 };
 733
 734 #endif
 735
 736 static const struct normal_encoding little2_encoding = {
 737   { VTABLE, 2, 0,
 738 #if BYTEORDER == 1234
 739     1
 740 #else
 741     0
 742 #endif
 743   },
 744   {
 745 #define BT_COLON BT_NMSTRT
 746 #include "asciitab.h"
 747 #undef BT_COLON
 748 #include "latin1tab.h"
 749   },
 750   STANDARD_VTABLE(little2_)
 751 };
 752
 753 #if BYTEORDER != 4321
 754
 755 #ifdef XML_NS
 756
 757 static const struct normal_encoding internal_little2_encoding_ns = {
 758   { VTABLE, 2, 0, 1 },
 759   {
 760 #include "iasciitab.h"
 761 #include "latin1tab.h"
 762   },
 763   STANDARD_VTABLE(little2_)
 764 };
 765
 766 #endif
 767
 768 static const struct normal_encoding internal_little2_encoding = {
 769   { VTABLE, 2, 0, 1 },
 770   {
 771 #define BT_COLON BT_NMSTRT
 772 #include "iasciitab.h"
 773 #undef BT_COLON
 774 #include "latin1tab.h"
 775   },
 776   STANDARD_VTABLE(little2_)
 777 };
 778
 779 #endif
 780
 781
 782 #define BIG2_BYTE_TYPE(enc, p) \
 783  ((p)[0] == 0 \
 784   ? ((struct normal_encoding *)(enc))->type[(unsigned char)(p)[1]] \
 785   : unicode_byte_type((p)[0], (p)[1]))
 786 #define BIG2_BYTE_TO_ASCII(enc, p) ((p)[0] == 0 ? (p)[1] : -1)
 787 #define BIG2_CHAR_MATCHES(enc, p, c) ((p)[0] == 0 && (p)[1] == c)
 788 #define BIG2_IS_NAME_CHAR_MINBPC(enc, p) \
 789   UCS2_GET_NAMING(namePages, (unsigned char)p[0], (unsigned char)p[1])
 790 #define BIG2_IS_NMSTRT_CHAR_MINBPC(enc, p) \
 791   UCS2_GET_NAMING(nmstrtPages, (unsigned char)p[0], (unsigned char)p[1])
 792
 793 #ifdef XML_MIN_SIZE
 794
 795 static int PTRFASTCALL
 796 big2_byteType(const ENCODING *enc, const char *p)
 797 {
 798   return BIG2_BYTE_TYPE(enc, p);
 799 }
 800
 801 static int PTRFASTCALL
 802 big2_byteToAscii(const ENCODING *enc, const char *p)
 803 {
 804   return BIG2_BYTE_TO_ASCII(enc, p);
 805 }
 806
 807 static int PTRCALL
 808 big2_charMatches(const ENCODING *enc, const char *p, int c)
 809 {
 810   return BIG2_CHAR_MATCHES(enc, p, c);
 811 }
 812
 813 static int PTRFASTCALL
 814 big2_isNameMin(const ENCODING *enc, const char *p)
 815 {
 816   return BIG2_IS_NAME_CHAR_MINBPC(enc, p);
 817 }
 818
 819 static int PTRFASTCALL
 820 big2_isNmstrtMin(const ENCODING *enc, const char *p)
 821 {
 822   return BIG2_IS_NMSTRT_CHAR_MINBPC(enc, p);
 823 }
 824
 825 #undef VTABLE
 826 #define VTABLE VTABLE1, big2_toUtf8, big2_toUtf16
 827
 828 #else /* not XML_MIN_SIZE */
 829
 830 #undef PREFIX
 831 #define PREFIX(ident) big2_ ## ident
 832 #define MINBPC(enc) 2
 833 /* CHAR_MATCHES is guaranteed to have MINBPC bytes available. */
 834 #define BYTE_TYPE(enc, p) BIG2_BYTE_TYPE(enc, p)
 835 #define BYTE_TO_ASCII(enc, p) BIG2_BYTE_TO_ASCII(enc, p)
 836 #define CHAR_MATCHES(enc, p, c) BIG2_CHAR_MATCHES(enc, p, c)
 837 #define IS_NAME_CHAR(enc, p, n) 0
 838 #define IS_NAME_CHAR_MINBPC(enc, p) BIG2_IS_NAME_CHAR_MINBPC(enc, p)
 839 #define IS_NMSTRT_CHAR(enc, p, n) (0)
 840 #define IS_NMSTRT_CHAR_MINBPC(enc, p) BIG2_IS_NMSTRT_CHAR_MINBPC(enc, p)
 841
 842 #define XML_TOK_IMPL_C
 843 #include "xmltok_impl.c"
 844 #undef XML_TOK_IMPL_C
 845
 846 #undef MINBPC
 847 #undef BYTE_TYPE
 848 #undef BYTE_TO_ASCII
 849 #undef CHAR_MATCHES
 850 #undef IS_NAME_CHAR
 851 #undef IS_NAME_CHAR_MINBPC
 852 #undef IS_NMSTRT_CHAR
 853 #undef IS_NMSTRT_CHAR_MINBPC
 854 #undef IS_INVALID_CHAR
 855
 856 #endif /* not XML_MIN_SIZE */
 857
 858 #ifdef XML_NS
 859
 860 static const struct normal_encoding big2_encoding_ns = {
 861   { VTABLE, 2, 0,
 862 #if BYTEORDER == 4321
 863   1
 864 #else
 865   0
 866 #endif
 867   },
 868   {
 869 #include "asciitab.h"
 870 #include "latin1tab.h"
 871   },
 872   STANDARD_VTABLE(big2_)
 873 };
 874
 875 #endif
 876
 877 static const struct normal_encoding big2_encoding = {
 878   { VTABLE, 2, 0,
 879 #if BYTEORDER == 4321
 880   1
 881 #else
 882   0
 883 #endif
 884   },
 885   {
 886 #define BT_COLON BT_NMSTRT
 887 #include "asciitab.h"
 888 #undef BT_COLON
 889 #include "latin1tab.h"
 890   },
 891   STANDARD_VTABLE(big2_)
 892 };
 893
 894 #if BYTEORDER != 1234
 895
 896 #ifdef XML_NS
 897
 898 static const struct normal_encoding internal_big2_encoding_ns = {
 899   { VTABLE, 2, 0, 1 },
 900   {
 901 #include "iasciitab.h"
 902 #include "latin1tab.h"
 903   },
 904   STANDARD_VTABLE(big2_)
 905 };
 906
 907 #endif
 908
 909 static const struct normal_encoding internal_big2_encoding = {
 910   { VTABLE, 2, 0, 1 },
 911   {
 912 #define BT_COLON BT_NMSTRT
 913 #include "iasciitab.h"
 914 #undef BT_COLON
 915 #include "latin1tab.h"
 916   },
 917   STANDARD_VTABLE(big2_)
 918 };
 919
 920 #endif
 921
 922 #undef PREFIX
 923
 924 static int FASTCALL
 925 streqci(const char *s1, const char *s2)
 926 {
 927   for (;;) {
 928     char c1 = *s1++;
 929     char c2 = *s2++;
 930     if (ASCII_a <= c1 && c1 <= ASCII_z)
 931       c1 += ASCII_A - ASCII_a;
 932     if (ASCII_a <= c2 && c2 <= ASCII_z)
 933       c2 += ASCII_A - ASCII_a;
 934     if (c1 != c2)
 935       return 0;
 936     if (!c1)
 937       break;
 938   }
 939   return 1;
 940 }
 941
 942 static void PTRCALL
 943 initUpdatePosition(const ENCODING *enc, const char *ptr,
 944                    const char *end, POSITION *pos)
 945 {
 946   normal_updatePosition(&utf8_encoding.enc, ptr, end, pos);
 947 }
 948
 949 static int
 950 toAscii(const ENCODING *enc, const char *ptr, const char *end)
 951 {
 952   char buf[1];
 953   char *p = buf;
 954   XmlUtf8Convert(enc, &ptr, end, &p, p + 1);
 955   if (p == buf)
 956     return -1;
 957   else
 958     return buf[0];
 959 }
 960
 961 static int FASTCALL
 962 isSpace(int c)
 963 {
 964   switch (c) {
 965   case 0x20:
 966   case 0xD:
 967   case 0xA:
 968   case 0x9:
 969     return 1;
 970   }
 971   return 0;
 972 }
 973
 974 /* Return 1 if there's just optional white space or there's an S
 975    followed by name=val.
 976 */
 977 static int
 978 parsePseudoAttribute(const ENCODING *enc,
 979                      const char *ptr,
 980                      const char *end,
 981                      const char **namePtr,
 982                      const char **nameEndPtr,
 983                      const char **valPtr,
 984                      const char **nextTokPtr)
 985 {
 986   int c;
 987   char open;
 988   if (ptr == end) {
 989     *namePtr = NULL;
 990     return 1;
 991   }
 992   if (!isSpace(toAscii(enc, ptr, end))) {
 993     *nextTokPtr = ptr;
 994     return 0;
 995   }
 996   do {
 997     ptr += enc->minBytesPerChar;
 998   } while (isSpace(toAscii(enc, ptr, end)));
 999   if (ptr == end) {
1000     *namePtr = NULL;
1001     return 1;
1002   }
1003   *namePtr = ptr;
1004   for (;;) {
1005     c = toAscii(enc, ptr, end);
1006     if (c == -1) {
1007       *nextTokPtr = ptr;
1008       return 0;
1009     }
1010     if (c == ASCII_EQUALS) {
1011       *nameEndPtr = ptr;
1012       break;
1013     }
1014     if (isSpace(c)) {
1015       *nameEndPtr = ptr;
1016       do {
1017         ptr += enc->minBytesPerChar;
1018       } while (isSpace(c = toAscii(enc, ptr, end)));
1019       if (c != ASCII_EQUALS) {
1020         *nextTokPtr = ptr;
1021         return 0;
1022       }
1023       break;
1024     }
1025     ptr += enc->minBytesPerChar;
1026   }
1027   if (ptr == *namePtr) {
1028     *nextTokPtr = ptr;
1029     return 0;
1030   }
1031   ptr += enc->minBytesPerChar;
1032   c = toAscii(enc, ptr, end);
1033   while (isSpace(c)) {
1034     ptr += enc->minBytesPerChar;
1035     c = toAscii(enc, ptr, end);
1036   }
1037   if (c != ASCII_QUOT && c != ASCII_APOS) {
1038     *nextTokPtr = ptr;
1039     return 0;
1040   }
1041   open = (char)c;
1042   ptr += enc->minBytesPerChar;
1043   *valPtr = ptr;
1044   for (;; ptr += enc->minBytesPerChar) {
1045     c = toAscii(enc, ptr, end);
1046     if (c == open)
1047       break;
1048     if (!(ASCII_a <= c && c <= ASCII_z)
1049         && !(ASCII_A <= c && c <= ASCII_Z)
1050         && !(ASCII_0 <= c && c <= ASCII_9)
1051         && c != ASCII_PERIOD
1052         && c != ASCII_MINUS
1053         && c != ASCII_UNDERSCORE) {
1054       *nextTokPtr = ptr;
1055       return 0;
1056     }
1057   }
1058   *nextTokPtr = ptr + enc->minBytesPerChar;
1059   return 1;
1060 }
1061
1062 static const char KW_version[] = {
1063   ASCII_v, ASCII_e, ASCII_r, ASCII_s, ASCII_i, ASCII_o, ASCII_n, '\0'
1064 };
1065
1066 static const char KW_encoding[] = {
1067   ASCII_e, ASCII_n, ASCII_c, ASCII_o, ASCII_d, ASCII_i, ASCII_n, ASCII_g, '\0'
1068 };
1069
1070 static const char KW_standalone[] = {
1071   ASCII_s, ASCII_t, ASCII_a, ASCII_n, ASCII_d, ASCII_a, ASCII_l, ASCII_o,
1072   ASCII_n, ASCII_e, '\0'
1073 };
1074
1075 static const char KW_yes[] = {
1076   ASCII_y, ASCII_e, ASCII_s,  '\0'
1077 };
1078
1079 static const char KW_no[] = {
1080   ASCII_n, ASCII_o,  '\0'
1081 };
1082
1083 static int
1084 doParseXmlDecl(const ENCODING *(*encodingFinder)(const ENCODING *,
1085                                                  const char *,
1086                                                  const char *),
1087                int isGeneralTextEntity,
1088                const ENCODING *enc,
1089                const char *ptr,
1090                const char *end,
1091                const char **badPtr,
1092                const char **versionPtr,
1093                const char **versionEndPtr,
1094                const char **encodingName,
1095                const ENCODING **encoding,
1096                int *standalone)
1097 {
1098   const char *val = NULL;
1099   const char *name = NULL;
1100   const char *nameEnd = NULL;
1101   ptr += 5 * enc->minBytesPerChar;
1102   end -= 2 * enc->minBytesPerChar;
1103   if (!parsePseudoAttribute(enc, ptr, end, &name, &nameEnd, &val, &ptr)
1104       || !name) {
1105     *badPtr = ptr;
1106     return 0;
1107   }
1108   if (!XmlNameMatchesAscii(enc, name, nameEnd, KW_version)) {
1109     if (!isGeneralTextEntity) {
1110       *badPtr = name;
1111       return 0;
1112     }
1113   }
1114   else {
1115     if (versionPtr)
1116       *versionPtr = val;
1117     if (versionEndPtr)
1118       *versionEndPtr = ptr;
1119     if (!parsePseudoAttribute(enc, ptr, end, &name, &nameEnd, &val, &ptr)) {
1120       *badPtr = ptr;
1121       return 0;
1122     }
1123     if (!name) {
1124       if (isGeneralTextEntity) {
1125         /* a TextDecl must have an EncodingDecl */
1126         *badPtr = ptr;
1127         return 0;
1128       }
1129       return 1;
1130     }
1131   }
1132   if (XmlNameMatchesAscii(enc, name, nameEnd, KW_encoding)) {
1133     int c = toAscii(enc, val, end);
1134     if (!(ASCII_a <= c && c <= ASCII_z) && !(ASCII_A <= c && c <= ASCII_Z)) {
1135       *badPtr = val;
1136       return 0;
1137     }
1138     if (encodingName)
1139       *encodingName = val;
1140     if (encoding)
1141       *encoding = encodingFinder(enc, val, ptr - enc->minBytesPerChar);
1142     if (!parsePseudoAttribute(enc, ptr, end, &name, &nameEnd, &val, &ptr)) {
1143       *badPtr = ptr;
1144       return 0;
1145     }
1146     if (!name)
1147       return 1;
1148   }
1149   if (!XmlNameMatchesAscii(enc, name, nameEnd, KW_standalone)
1150       || isGeneralTextEntity) {
1151     *badPtr = name;
1152     return 0;
1153   }
1154   if (XmlNameMatchesAscii(enc, val, ptr - enc->minBytesPerChar, KW_yes)) {
1155     if (standalone)
1156       *standalone = 1;
1157   }
1158   else if (XmlNameMatchesAscii(enc, val, ptr - enc->minBytesPerChar, KW_no)) {
1159     if (standalone)
1160       *standalone = 0;
1161   }
1162   else {
1163     *badPtr = val;
1164     return 0;
1165   }
1166   while (isSpace(toAscii(enc, ptr, end)))
1167     ptr += enc->minBytesPerChar;
1168   if (ptr != end) {
1169     *badPtr = ptr;
1170     return 0;
1171   }
1172   return 1;
1173 }
1174
1175 static int FASTCALL
1176 checkCharRefNumber(int result)
1177 {
1178   switch (result >> 8) {
1179   case 0xD8: case 0xD9: case 0xDA: case 0xDB:
1180   case 0xDC: case 0xDD: case 0xDE: case 0xDF:
1181     return -1;
1182   case 0:
1183     if (latin1_encoding.type[result] == BT_NONXML)
1184       return -1;
1185     break;
1186   case 0xFF:
1187     if (result == 0xFFFE || result == 0xFFFF)
1188       return -1;
1189     break;
1190   }
1191   return result;
1192 }
1193
1194 int FASTCALL
1195 XmlUtf8Encode(int c, char *buf)
1196 {
1197   enum {
1198     /* minN is minimum legal resulting value for N byte sequence */
1199     min2 = 0x80,
1200     min3 = 0x800,
1201     min4 = 0x10000
1202   };
1203
1204   if (c < 0)
1205     return 0;
1206   if (c < min2) {
1207     buf[0] = (char)(c | UTF8_cval1);
1208     return 1;
1209   }
1210   if (c < min3) {
1211     buf[0] = (char)((c >> 6) | UTF8_cval2);
1212     buf[1] = (char)((c & 0x3f) | 0x80);
1213     return 2;
1214   }
1215   if (c < min4) {
1216     buf[0] = (char)((c >> 12) | UTF8_cval3);
1217     buf[1] = (char)(((c >> 6) & 0x3f) | 0x80);
1218     buf[2] = (char)((c & 0x3f) | 0x80);
1219     return 3;
1220   }
1221   if (c < 0x110000) {
1222     buf[0] = (char)((c >> 18) | UTF8_cval4);
1223     buf[1] = (char)(((c >> 12) & 0x3f) | 0x80);
1224     buf[2] = (char)(((c >> 6) & 0x3f) | 0x80);
1225     buf[3] = (char)((c & 0x3f) | 0x80);
1226     return 4;
1227   }
1228   return 0;
1229 }
1230
1231 int FASTCALL
1232 XmlUtf16Encode(int charNum, unsigned short *buf)
1233 {
1234   if (charNum < 0)
1235     return 0;
1236   if (charNum < 0x10000) {
1237     buf[0] = (unsigned short)charNum;
1238     return 1;
1239   }
1240   if (charNum < 0x110000) {
1241     charNum -= 0x10000;
1242     buf[0] = (unsigned short)((charNum >> 10) + 0xD800);
1243     buf[1] = (unsigned short)((charNum & 0x3FF) + 0xDC00);
1244     return 2;
1245   }
1246   return 0;
1247 }
1248
1249 struct unknown_encoding {
1250   struct normal_encoding normal;
1251   CONVERTER convert;
1252   void *userData;
1253   unsigned short utf16[256];
1254   char utf8[256][4];
1255 };
1256
1257 #define AS_UNKNOWN_ENCODING(enc)  ((const struct unknown_encoding *) (enc))
1258
1259 int
1260 XmlSizeOfUnknownEncoding(void)
1261 {
1262   return sizeof(struct unknown_encoding);
1263 }
1264
1265 static int PTRFASTCALL
1266 unknown_isName(const ENCODING *enc, const char *p)
1267 {
1268   const struct unknown_encoding *uenc = AS_UNKNOWN_ENCODING(enc);
1269   int c = uenc->convert(uenc->userData, p);
1270   if (c & ~0xFFFF)
1271     return 0;
1272   return UCS2_GET_NAMING(namePages, c >> 8, c & 0xFF);
1273 }
1274
1275 static int PTRFASTCALL
1276 unknown_isNmstrt(const ENCODING *enc, const char *p)
1277 {
1278   const struct unknown_encoding *uenc = AS_UNKNOWN_ENCODING(enc);
1279   int c = uenc->convert(uenc->userData, p);
1280   if (c & ~0xFFFF)
1281     return 0;
1282   return UCS2_GET_NAMING(nmstrtPages, c >> 8, c & 0xFF);
1283 }
1284
1285 static int PTRFASTCALL
1286 unknown_isInvalid(const ENCODING *enc, const char *p)
1287 {
1288   const struct unknown_encoding *uenc = AS_UNKNOWN_ENCODING(enc);
1289   int c = uenc->convert(uenc->userData, p);
1290   return (c & ~0xFFFF) || checkCharRefNumber(c) < 0;
1291 }
1292
1293 static void PTRCALL
1294 unknown_toUtf8(const ENCODING *enc,
1295                const char **fromP, const char *fromLim,
1296                char **toP, const char *toLim)
1297 {
1298   const struct unknown_encoding *uenc = AS_UNKNOWN_ENCODING(enc);
1299   char buf[XML_UTF8_ENCODE_MAX];
1300   for (;;) {
1301     const char *utf8;
1302     int n;
1303     if (*fromP == fromLim)
1304       break;
1305     utf8 = uenc->utf8[(unsigned char)**fromP];
1306     n = *utf8++;
1307     if (n == 0) {
1308       int c = uenc->convert(uenc->userData, *fromP);
1309       n = XmlUtf8Encode(c, buf);
1310       if (n > toLim - *toP)
1311         break;
1312       utf8 = buf;
1313       *fromP += (AS_NORMAL_ENCODING(enc)->type[(unsigned char)**fromP]
1314                  - (BT_LEAD2 - 2));
1315     }
1316     else {
1317       if (n > toLim - *toP)
1318         break;
1319       (*fromP)++;
1320     }
1321     do {
1322       *(*toP)++ = *utf8++;
1323     } while (--n != 0);
1324   }
1325 }
1326
1327 static void PTRCALL
1328 unknown_toUtf16(const ENCODING *enc,
1329                 const char **fromP, const char *fromLim,
1330                 unsigned short **toP, const unsigned short *toLim)
1331 {
1332   const struct unknown_encoding *uenc = AS_UNKNOWN_ENCODING(enc);
1333   while (*fromP != fromLim && *toP != toLim) {
1334     unsigned short c = uenc->utf16[(unsigned char)**fromP];
1335     if (c == 0) {
1336       c = (unsigned short)
1337           uenc->convert(uenc->userData, *fromP);
1338       *fromP += (AS_NORMAL_ENCODING(enc)->type[(unsigned char)**fromP]
1339                  - (BT_LEAD2 - 2));
1340     }
1341     else
1342       (*fromP)++;
1343     *(*toP)++ = c;
1344   }
1345 }
1346
1347 ENCODING *
1348 XmlInitUnknownEncoding(void *mem,
1349                        int *table,
1350                        CONVERTER convert,
1351                        void *userData)
1352 {
1353   int i;
1354   struct unknown_encoding *e = (struct unknown_encoding *)mem;
1355   for (i = 0; i < (int)sizeof(struct normal_encoding); i++)
1356     ((char *)mem)[i] = ((char *)&latin1_encoding)[i];
1357   for (i = 0; i < 128; i++)
1358     if (latin1_encoding.type[i] != BT_OTHER
1359         && latin1_encoding.type[i] != BT_NONXML
1360         && table[i] != i)
1361       return 0;
1362   for (i = 0; i < 256; i++) {
1363     int c = table[i];
1364     if (c == -1) {
1365       e->normal.type[i] = BT_MALFORM;
1366       /* This shouldn't really get used. */
1367       e->utf16[i] = 0xFFFF;
1368       e->utf8[i][0] = 1;
1369       e->utf8[i][1] = 0;
1370     }
1371     else if (c < 0) {
1372       if (c < -4)
1373         return 0;
1374       e->normal.type[i] = (unsigned char)(BT_LEAD2 - (c + 2));
1375       e->utf8[i][0] = 0;
1376       e->utf16[i] = 0;
1377     }
1378     else if (c < 0x80) {
1379       if (latin1_encoding.type[c] != BT_OTHER
1380           && latin1_encoding.type[c] != BT_NONXML
1381           && c != i)
1382         return 0;
1383       e->normal.type[i] = latin1_encoding.type[c];
1384       e->utf8[i][0] = 1;
1385       e->utf8[i][1] = (char)c;
1386       e->utf16[i] = (unsigned short)(c == 0 ? 0xFFFF : c);
1387     }
1388     else if (checkCharRefNumber(c) < 0) {
1389       e->normal.type[i] = BT_NONXML;
1390       /* This shouldn't really get used. */
1391       e->utf16[i] = 0xFFFF;
1392       e->utf8[i][0] = 1;
1393       e->utf8[i][1] = 0;
1394     }
1395     else {
1396       if (c > 0xFFFF)
1397         return 0;
1398       if (UCS2_GET_NAMING(nmstrtPages, c >> 8, c & 0xff))
1399         e->normal.type[i] = BT_NMSTRT;
1400       else if (UCS2_GET_NAMING(namePages, c >> 8, c & 0xff))
1401         e->normal.type[i] = BT_NAME;
1402       else
1403         e->normal.type[i] = BT_OTHER;
1404       e->utf8[i][0] = (char)XmlUtf8Encode(c, e->utf8[i] + 1);
1405       e->utf16[i] = (unsigned short)c;
1406     }
1407   }
1408   e->userData = userData;
1409   e->convert = convert;
1410   if (convert) {
1411     e->normal.isName2 = unknown_isName;
1412     e->normal.isName3 = unknown_isName;
1413     e->normal.isName4 = unknown_isName;
1414     e->normal.isNmstrt2 = unknown_isNmstrt;
1415     e->normal.isNmstrt3 = unknown_isNmstrt;
1416     e->normal.isNmstrt4 = unknown_isNmstrt;
1417     e->normal.isInvalid2 = unknown_isInvalid;
1418     e->normal.isInvalid3 = unknown_isInvalid;
1419     e->normal.isInvalid4 = unknown_isInvalid;
1420   }
1421   e->normal.enc.utf8Convert = unknown_toUtf8;
1422   e->normal.enc.utf16Convert = unknown_toUtf16;
1423   return &(e->normal.enc);
1424 }
1425
1426 /* If this enumeration is changed, getEncodingIndex and encodings
1427 must also be changed. */
1428 enum {
1429   UNKNOWN_ENC = -1,
1430   ISO_8859_1_ENC = 0,
1431   US_ASCII_ENC,
1432   UTF_8_ENC,
1433   UTF_16_ENC,
1434   UTF_16BE_ENC,
1435   UTF_16LE_ENC,
1436   /* must match encodingNames up to here */
1437   NO_ENC
1438 };
1439
1440 static const char KW_ISO_8859_1[] = {
1441   ASCII_I, ASCII_S, ASCII_O, ASCII_MINUS, ASCII_8, ASCII_8, ASCII_5, ASCII_9,
1442   ASCII_MINUS, ASCII_1, '\0'
1443 };
1444 static const char KW_US_ASCII[] = {
1445   ASCII_U, ASCII_S, ASCII_MINUS, ASCII_A, ASCII_S, ASCII_C, ASCII_I, ASCII_I,
1446   '\0'
1447 };
1448 static const char KW_UTF_8[] =  {
1449   ASCII_U, ASCII_T, ASCII_F, ASCII_MINUS, ASCII_8, '\0'
1450 };
1451 static const char KW_UTF_16[] = {
1452   ASCII_U, ASCII_T, ASCII_F, ASCII_MINUS, ASCII_1, ASCII_6, '\0'
1453 };
1454 static const char KW_UTF_16BE[] = {
1455   ASCII_U, ASCII_T, ASCII_F, ASCII_MINUS, ASCII_1, ASCII_6, ASCII_B, ASCII_E,
1456   '\0'
1457 };
1458 static const char KW_UTF_16LE[] = {
1459   ASCII_U, ASCII_T, ASCII_F, ASCII_MINUS, ASCII_1, ASCII_6, ASCII_L, ASCII_E,
1460   '\0'
1461 };
1462
1463 static int FASTCALL
1464 getEncodingIndex(const char *name)
1465 {
1466   static const char * const encodingNames[] = {
1467     KW_ISO_8859_1,
1468     KW_US_ASCII,
1469     KW_UTF_8,
1470     KW_UTF_16,
1471     KW_UTF_16BE,
1472     KW_UTF_16LE,
1473   };
1474   int i;
1475   if (name == NULL)
1476     return NO_ENC;
1477   for (i = 0; i < (int)(sizeof(encodingNames)/sizeof(encodingNames[0])); i++)
1478     if (streqci(name, encodingNames[i]))
1479       return i;
1480   return UNKNOWN_ENC;
1481 }
1482
1483 /* For binary compatibility, we store the index of the encoding
1484    specified at initialization in the isUtf16 member.
1485 */
1486
1487 #define INIT_ENC_INDEX(enc) ((int)(enc)->initEnc.isUtf16)
1488 #define SET_INIT_ENC_INDEX(enc, i) ((enc)->initEnc.isUtf16 = (char)i)
1489
1490 /* This is what detects the encoding.  encodingTable maps from
1491    encoding indices to encodings; INIT_ENC_INDEX(enc) is the index of
1492    the external (protocol) specified encoding; state is
1493    XML_CONTENT_STATE if we're parsing an external text entity, and
1494    XML_PROLOG_STATE otherwise.
1495 */
1496
1497
1498 static int
1499 initScan(const ENCODING * const *encodingTable,
1500          const INIT_ENCODING *enc,
1501          int state,
1502          const char *ptr,
1503          const char *end,
1504          const char **nextTokPtr)
1505 {
1506   const ENCODING **encPtr;
1507
1508   if (ptr == end)
1509     return XML_TOK_NONE;
1510   encPtr = enc->encPtr;
1511   if (ptr + 1 == end) {
1512     /* only a single byte available for auto-detection */
1513 #ifndef XML_DTD /* FIXME */
1514     /* a well-formed document entity must have more than one byte */
1515     if (state != XML_CONTENT_STATE)
1516       return XML_TOK_PARTIAL;
1517 #endif
1518     /* so we're parsing an external text entity... */
1519     /* if UTF-16 was externally specified, then we need at least 2 bytes */
1520     switch (INIT_ENC_INDEX(enc)) {
1521     case UTF_16_ENC:
1522     case UTF_16LE_ENC:
1523     case UTF_16BE_ENC:
1524       return XML_TOK_PARTIAL;
1525     }
1526     switch ((unsigned char)*ptr) {
1527     case 0xFE:
1528     case 0xFF:
1529     case 0xEF: /* possibly first byte of UTF-8 BOM */
1530       if (INIT_ENC_INDEX(enc) == ISO_8859_1_ENC
1531           && state == XML_CONTENT_STATE)
1532         break;
1533       /* fall through */
1534     case 0x00:
1535     case 0x3C:
1536       return XML_TOK_PARTIAL;
1537     }
1538   }
1539   else {
1540     switch (((unsigned char)ptr[0] << 8) | (unsigned char)ptr[1]) {
1541     case 0xFEFF:
1542       if (INIT_ENC_INDEX(enc) == ISO_8859_1_ENC
1543           && state == XML_CONTENT_STATE)
1544         break;
1545       *nextTokPtr = ptr + 2;
1546       *encPtr = encodingTable[UTF_16BE_ENC];
1547       return XML_TOK_BOM;
1548     /* 00 3C is handled in the default case */
1549     case 0x3C00:
1550       if ((INIT_ENC_INDEX(enc) == UTF_16BE_ENC
1551            || INIT_ENC_INDEX(enc) == UTF_16_ENC)
1552           && state == XML_CONTENT_STATE)
1553         break;
1554       *encPtr = encodingTable[UTF_16LE_ENC];
1555       return XmlTok(*encPtr, state, ptr, end, nextTokPtr);
1556     case 0xFFFE:
1557       if (INIT_ENC_INDEX(enc) == ISO_8859_1_ENC
1558           && state == XML_CONTENT_STATE)
1559         break;
1560       *nextTokPtr = ptr + 2;
1561       *encPtr = encodingTable[UTF_16LE_ENC];
1562       return XML_TOK_BOM;
1563     case 0xEFBB:
1564       /* Maybe a UTF-8 BOM (EF BB BF) */
1565       /* If there's an explicitly specified (external) encoding
1566          of ISO-8859-1 or some flavour of UTF-16
1567          and this is an external text entity,
1568          don't look for the BOM,
1569          because it might be a legal data.
1570       */
1571       if (state == XML_CONTENT_STATE) {
1572         int e = INIT_ENC_INDEX(enc);
1573         if (e == ISO_8859_1_ENC || e == UTF_16BE_ENC
1574             || e == UTF_16LE_ENC || e == UTF_16_ENC)
1575           break;
1576       }
1577       if (ptr + 2 == end)
1578         return XML_TOK_PARTIAL;
1579       if ((unsigned char)ptr[2] == 0xBF) {
1580         *nextTokPtr = ptr + 3;
1581         *encPtr = encodingTable[UTF_8_ENC];
1582         return XML_TOK_BOM;
1583       }
1584       break;
1585     default:
1586       if (ptr[0] == '\0') {
1587         /* 0 isn't a legal data character. Furthermore a document
1588            entity can only start with ASCII characters.  So the only
1589            way this can fail to be big-endian UTF-16 if it it's an
1590            external parsed general entity that's labelled as
1591            UTF-16LE.
1592         */
1593         if (state == XML_CONTENT_STATE && INIT_ENC_INDEX(enc) == UTF_16LE_ENC)
1594           break;
1595         *encPtr = encodingTable[UTF_16BE_ENC];
1596         return XmlTok(*encPtr, state, ptr, end, nextTokPtr);
1597       }
1598       else if (ptr[1] == '\0') {
1599         /* We could recover here in the case:
1600             - parsing an external entity
1601             - second byte is 0
1602             - no externally specified encoding
1603             - no encoding declaration
1604            by assuming UTF-16LE.  But we don't, because this would mean when
1605            presented just with a single byte, we couldn't reliably determine
1606            whether we needed further bytes.
1607         */
1608         if (state == XML_CONTENT_STATE)
1609           break;
1610         *encPtr = encodingTable[UTF_16LE_ENC];
1611         return XmlTok(*encPtr, state, ptr, end, nextTokPtr);
1612       }
1613       break;
1614     }
1615   }
1616   *encPtr = encodingTable[INIT_ENC_INDEX(enc)];
1617   return XmlTok(*encPtr, state, ptr, end, nextTokPtr);
1618 }
1619
1620
1621 #define NS(x) x
1622 #define ns(x) x
1623 #define XML_TOK_NS_C
1624 #include "xmltok_ns.c"
1625 #undef XML_TOK_NS_C
1626 #undef NS
1627 #undef ns
1628
1629 #ifdef XML_NS
1630
1631 #define NS(x) x ## NS
1632 #define ns(x) x ## _ns
1633
1634 #define XML_TOK_NS_C
1635 #include "xmltok_ns.c"
1636 #undef XML_TOK_NS_C
1637
1638 #undef NS
1639 #undef ns
1640
1641 ENCODING *
1642 XmlInitUnknownEncodingNS(void *mem,
1643                          int *table,
1644                          CONVERTER convert,
1645                          void *userData)
1646 {
1647   ENCODING *enc = XmlInitUnknownEncoding(mem, table, convert, userData);
1648   if (enc)
1649     ((struct normal_encoding *)enc)->type[ASCII_COLON] = BT_COLON;
1650   return enc;
1651 }
1652
1653 #endif /* XML_NS */