src/expat/lib/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 <stddef.h>
   6
   7 #ifdef COMPILED_FROM_DSP
   8 #include "winconfig.h"
   9 #elif defined(OS2_32)
  10 #include "os2config.h"
  11 #elif defined(__MSDOS__)
  12 #include "dosconfig.h"
  13 #elif defined(MACOS_CLASSIC)
  14 #include "macconfig.h"
  15 #elif defined(__amigaos__)
  16 #include "amigaconfig.h"
  17 #elif defined(__WATCOMC__)
  18 #include "watcomconfig.h"
  19 #else
  20 #ifdef HAVE_EXPAT_CONFIG_H
  21 #include <expat_config.h>
  22 #endif
  23 #endif /* ndef COMPILED_FROM_DSP */
  24
  25 #include "expat_external.h"
  26 #include "internal.h"
  27 #include "xmltok.h"
  28 #include "nametab.h"
  29
  30 #ifdef XML_DTD
  31 #define IGNORE_SECTION_TOK_VTABLE , PREFIX(ignoreSectionTok)
  32 #else
  33 #define IGNORE_SECTION_TOK_VTABLE /* as nothing */
  34 #endif
  35
  36 #define VTABLE1 \
  37   { PREFIX(prologTok), PREFIX(contentTok), \
  38     PREFIX(cdataSectionTok) IGNORE_SECTION_TOK_VTABLE }, \
  39   { PREFIX(attributeValueTok), PREFIX(entityValueTok) }, \
  40   PREFIX(sameName), \
  41   PREFIX(nameMatchesAscii), \
  42   PREFIX(nameLength), \
  43   PREFIX(skipS), \
  44   PREFIX(getAtts), \
  45   PREFIX(charRefNumber), \
  46   PREFIX(predefinedEntityName), \
  47   PREFIX(updatePosition), \
  48   PREFIX(isPublicId)
  49
  50 #define VTABLE VTABLE1, PREFIX(toUtf8), PREFIX(toUtf16)
  51
  52 #define UCS2_GET_NAMING(pages, hi, lo) \
  53    (namingBitmap[(pages[hi] << 3) + ((lo) >> 5)] & (1 << ((lo) & 0x1F)))
  54
  55 /* A 2 byte UTF-8 representation splits the characters 11 bits between
  56    the bottom 5 and 6 bits of the bytes.  We need 8 bits to index into
  57    pages, 3 bits to add to that index and 5 bits to generate the mask.
  58 */
  59 #define UTF8_GET_NAMING2(pages, byte) \
  60     (namingBitmap[((pages)[(((byte)[0]) >> 2) & 7] << 3) \
  61                       + ((((byte)[0]) & 3) << 1) \
  62                       + ((((byte)[1]) >> 5) & 1)] \
  63          & (1 << (((byte)[1]) & 0x1F)))
  64
  65 /* A 3 byte UTF-8 representation splits the characters 16 bits between
  66    the bottom 4, 6 and 6 bits of the bytes.  We need 8 bits to index
  67    into pages, 3 bits to add to that index and 5 bits to generate the
  68    mask.
  69 */
  70 #define UTF8_GET_NAMING3(pages, byte) \
  71   (namingBitmap[((pages)[((((byte)[0]) & 0xF) << 4) \
  72                              + ((((byte)[1]) >> 2) & 0xF)] \
  73                        << 3) \
  74                       + ((((byte)[1]) & 3) << 1) \
  75                       + ((((byte)[2]) >> 5) & 1)] \
  76          & (1 << (((byte)[2]) & 0x1F)))
  77
  78 #define UTF8_GET_NAMING(pages, p, n) \
  79   ((n) == 2 \
  80   ? UTF8_GET_NAMING2(pages, (const unsigned char *)(p)) \
  81   : ((n) == 3 \
  82      ? UTF8_GET_NAMING3(pages, (const unsigned char *)(p)) \
  83      : 0))
  84
  85 /* Detection of invalid UTF-8 sequences is based on Table 3.1B
  86    of Unicode 3.2: http://www.unicode.org/unicode/reports/tr28/
  87    with the additional restriction of not allowing the Unicode
  88    code points 0xFFFF and 0xFFFE (sequences EF,BF,BF and EF,BF,BE).
  89    Implementation details:
  90      (A & 0x80) == 0     means A < 0x80
  91    and
  92      (A & 0xC0) == 0xC0  means A > 0xBF
  93 */
  94
  95 #define UTF8_INVALID2(p) \
  96   ((*p) < 0xC2 || ((p)[1] & 0x80) == 0 || ((p)[1] & 0xC0) == 0xC0)
  97
  98 #define UTF8_INVALID3(p) \
  99   (((p)[2] & 0x80) == 0 \
 100   || \
 101   ((*p) == 0xEF && (p)[1] == 0xBF \
 102     ? \
 103     (p)[2] > 0xBD \
 104     : \
 105     ((p)[2] & 0xC0) == 0xC0) \
 106   || \
 107   ((*p) == 0xE0 \
 108     ? \
 109     (p)[1] < 0xA0 || ((p)[1] & 0xC0) == 0xC0 \
 110     : \
 111     ((p)[1] & 0x80) == 0 \
 112     || \
 113     ((*p) == 0xED ? (p)[1] > 0x9F : ((p)[1] & 0xC0) == 0xC0)))
 114
 115 #define UTF8_INVALID4(p) \
 116   (((p)[3] & 0x80) == 0 || ((p)[3] & 0xC0) == 0xC0 \
 117   || \
 118   ((p)[2] & 0x80) == 0 || ((p)[2] & 0xC0) == 0xC0 \
 119   || \
 120   ((*p) == 0xF0 \
 121     ? \
 122     (p)[1] < 0x90 || ((p)[1] & 0xC0) == 0xC0 \
 123     : \
 124     ((p)[1] & 0x80) == 0 \
 125     || \
 126     ((*p) == 0xF4 ? (p)[1] > 0x8F : ((p)[1] & 0xC0) == 0xC0)))
 127
 128 static int PTRFASTCALL
 129 isNever(const ENCODING *enc, const char *p)
 130 {
 131   return 0;
 132 }
 133
 134 static int PTRFASTCALL
 135 utf8_isName2(const ENCODING *enc, const char *p)
 136 {
 137   return UTF8_GET_NAMING2(namePages, (const unsigned char *)p);
 138 }
 139
 140 static int PTRFASTCALL
 141 utf8_isName3(const ENCODING *enc, const char *p)
 142 {
 143   return UTF8_GET_NAMING3(namePages, (const unsigned char *)p);
 144 }
 145
 146 #define utf8_isName4 isNever
 147
 148 static int PTRFASTCALL
 149 utf8_isNmstrt2(const ENCODING *enc, const char *p)
 150 {
 151   return UTF8_GET_NAMING2(nmstrtPages, (const unsigned char *)p);
 152 }
 153
 154 static int PTRFASTCALL
 155 utf8_isNmstrt3(const ENCODING *enc, const char *p)
 156 {
 157   return UTF8_GET_NAMING3(nmstrtPages, (const unsigned char *)p);
 158 }
 159
 160 #define utf8_isNmstrt4 isNever
 161
 162 static int PTRFASTCALL
 163 utf8_isInvalid2(const ENCODING *enc, const char *p)
 164 {
 165   return UTF8_INVALID2((const unsigned char *)p);
 166 }
 167
 168 static int PTRFASTCALL
 169 utf8_isInvalid3(const ENCODING *enc, const char *p)
 170 {
 171   return UTF8_INVALID3((const unsigned char *)p);
 172 }
 173
 174 static int PTRFASTCALL
 175 utf8_isInvalid4(const ENCODING *enc, const char *p)
 176 {
 177   return UTF8_INVALID4((const unsigned char *)p);
 178 }
 179
 180 struct normal_encoding {
 181   ENCODING enc;
 182   unsigned char type[256];
 183 #ifdef XML_MIN_SIZE
 184   int (PTRFASTCALL *byteType)(const ENCODING *, const char *);
 185   int (PTRFASTCALL *isNameMin)(const ENCODING *, const char *);
 186   int (PTRFASTCALL *isNmstrtMin)(const ENCODING *, const char *);
 187   int (PTRFASTCALL *byteToAscii)(const ENCODING *, const char *);
 188   int (PTRCALL *charMatches)(const ENCODING *, const char *, int);
 189 #endif /* XML_MIN_SIZE */
 190   int (PTRFASTCALL *isName2)(const ENCODING *, const char *);
 191   int (PTRFASTCALL *isName3)(const ENCODING *, const char *);
 192   int (PTRFASTCALL *isName4)(const ENCODING *, const char *);
 193   int (PTRFASTCALL *isNmstrt2)(const ENCODING *, const char *);
 194   int (PTRFASTCALL *isNmstrt3)(const ENCODING *, const char *);
 195   int (PTRFASTCALL *isNmstrt4)(const ENCODING *, const char *);
 196   int (PTRFASTCALL *isInvalid2)(const ENCODING *, const char *);
 197   int (PTRFASTCALL *isInvalid3)(const ENCODING *, const char *);
 198   int (PTRFASTCALL *isInvalid4)(const ENCODING *, const char *);
 199 };
 200
 201 #define AS_NORMAL_ENCODING(enc)   ((const struct normal_encoding *) (enc))
 202
 203 #ifdef XML_MIN_SIZE
 204
 205 #define STANDARD_VTABLE(E) \
 206  E ## byteType, \
 207  E ## isNameMin, \
 208  E ## isNmstrtMin, \
 209  E ## byteToAscii, \
 210  E ## charMatches,
 211
 212 #else
 213
 214 #define STANDARD_VTABLE(E) /* as nothing */
 215
 216 #endif
 217
 218 #define NORMAL_VTABLE(E) \
 219  E ## isName2, \
 220  E ## isName3, \
 221  E ## isName4, \
 222  E ## isNmstrt2, \
 223  E ## isNmstrt3, \
 224  E ## isNmstrt4, \
 225  E ## isInvalid2, \
 226  E ## isInvalid3, \
 227  E ## isInvalid4
 228
 229 static int FASTCALL checkCharRefNumber(int);
 230
 231 #include "xmltok_impl.h"
 232 #include "ascii.h"
 233
 234 #ifdef XML_MIN_SIZE
 235 #define sb_isNameMin isNever
 236 #define sb_isNmstrtMin isNever
 237 #endif
 238
 239 #ifdef XML_MIN_SIZE
 240 #define MINBPC(enc) ((enc)->minBytesPerChar)
 241 #else
 242 /* minimum bytes per character */
 243 #define MINBPC(enc) 1
 244 #endif
 245
 246 #define SB_BYTE_TYPE(enc, p) \
 247   (((struct normal_encoding *)(enc))->type[(unsigned char)*(p)])
 248
 249 #ifdef XML_MIN_SIZE
 250 static int PTRFASTCALL
 251 sb_byteType(const ENCODING *enc, const char *p)
 252 {
 253   return SB_BYTE_TYPE(enc, p);
 254 }
 255 #define BYTE_TYPE(enc, p) \
 256  (AS_NORMAL_ENCODING(enc)->byteType(enc, p))
 257 #else
 258 #define BYTE_TYPE(enc, p) SB_BYTE_TYPE(enc, p)
 259 #endif
 260
 261 #ifdef XML_MIN_SIZE
 262 #define BYTE_TO_ASCII(enc, p) \
 263  (AS_NORMAL_ENCODING(enc)->byteToAscii(enc, p))
 264 static int PTRFASTCALL
 265 sb_byteToAscii(const ENCODING *enc, const char *p)
 266 {
 267   return *p;
 268 }
 269 #else
 270 #define BYTE_TO_ASCII(enc, p) (*(p))
 271 #endif
 272
 273 #define IS_NAME_CHAR(enc, p, n) \
 274  (AS_NORMAL_ENCODING(enc)->isName ## n(enc, p))
 275 #define IS_NMSTRT_CHAR(enc, p, n) \
 276  (AS_NORMAL_ENCODING(enc)->isNmstrt ## n(enc, p))
 277 #define IS_INVALID_CHAR(enc, p, n) \
 278  (AS_NORMAL_ENCODING(enc)->isInvalid ## n(enc, p))
 279
 280 #ifdef XML_MIN_SIZE
 281 #define IS_NAME_CHAR_MINBPC(enc, p) \
 282  (AS_NORMAL_ENCODING(enc)->isNameMin(enc, p))
 283 #define IS_NMSTRT_CHAR_MINBPC(enc, p) \
 284  (AS_NORMAL_ENCODING(enc)->isNmstrtMin(enc, p))
 285 #else
 286 #define IS_NAME_CHAR_MINBPC(enc, p) (0)
 287 #define IS_NMSTRT_CHAR_MINBPC(enc, p) (0)
 288 #endif
 289
 290 #ifdef XML_MIN_SIZE
 291 #define CHAR_MATCHES(enc, p, c) \
 292  (AS_NORMAL_ENCODING(enc)->charMatches(enc, p, c))
 293 static int PTRCALL
 294 sb_charMatches(const ENCODING *enc, const char *p, int c)
 295 {
 296   return *p == c;
 297 }
 298 #else
 299 /* c is an ASCII character */
 300 #define CHAR_MATCHES(enc, p, c) (*(p) == c)
 301 #endif
 302
 303 #define PREFIX(ident) normal_ ## ident
 304 #define XML_TOK_IMPL_C
 305 #include "xmltok_impl.c"
 306 #undef XML_TOK_IMPL_C
 307
 308 #undef MINBPC
 309 #undef BYTE_TYPE
 310 #undef BYTE_TO_ASCII
 311 #undef CHAR_MATCHES
 312 #undef IS_NAME_CHAR
 313 #undef IS_NAME_CHAR_MINBPC
 314 #undef IS_NMSTRT_CHAR
 315 #undef IS_NMSTRT_CHAR_MINBPC
 316 #undef IS_INVALID_CHAR
 317
 318 enum {  /* UTF8_cvalN is value of masked first byte of N byte sequence */
 319   UTF8_cval1 = 0x00,
 320   UTF8_cval2 = 0xc0,
 321   UTF8_cval3 = 0xe0,
 322   UTF8_cval4 = 0xf0
 323 };
 324
 325 static void PTRCALL
 326 utf8_toUtf8(const ENCODING *enc,
 327             const char **fromP, const char *fromLim,
 328             char **toP, const char *toLim)
 329 {
 330   char *to;
 331   const char *from;
 332   if (fromLim - *fromP > toLim - *toP) {
 333     /* Avoid copying partial characters. */
 334     for (fromLim = *fromP + (toLim - *toP); fromLim > *fromP; fromLim--)
 335       if (((unsigned char)fromLim[-1] & 0xc0) != 0x80)
 336         break;
 337   }
 338   for (to = *toP, from = *fromP; from != fromLim; from++, to++)
 339     *to = *from;
 340   *fromP = from;
 341   *toP = to;
 342 }
 343
 344 static void PTRCALL
 345 utf8_toUtf16(const ENCODING *enc,
 346              const char **fromP, const char *fromLim,
 347              unsigned short **toP, const unsigned short *toLim)
 348 {
 349   unsigned short *to = *toP;
 350   const char *from = *fromP;
 351   while (from != fromLim && to != toLim) {
 352     switch (((struct normal_encoding *)enc)->type[(unsigned char)*from]) {
 353     case BT_LEAD2:
 354       *to++ = (unsigned short)(((from[0] & 0x1f) << 6) | (from[1] & 0x3f));
 355       from += 2;
 356       break;
 357     case BT_LEAD3:
 358       *to++ = (unsigned short)(((from[0] & 0xf) << 12)
 359                                | ((from[1] & 0x3f) << 6) | (from[2] & 0x3f));
 360       from += 3;
 361       break;
 362     case BT_LEAD4:
 363       {
 364         unsigned long n;
 365         if (to + 1 == toLim)
 366           goto after;
 367         n = ((from[0] & 0x7) << 18) | ((from[1] & 0x3f) << 12)
 368             | ((from[2] & 0x3f) << 6) | (from[3] & 0x3f);
 369         n -= 0x10000;
 370         to[0] = (unsigned short)((n >> 10) | 0xD800);
 371         to[1] = (unsigned short)((n & 0x3FF) | 0xDC00);
 372         to += 2;
 373         from += 4;
 374       }
 375       break;
 376     default:
 377       *to++ = *from++;
 378       break;
 379     }
 380   }
 381 after:
 382   *fromP = from;
 383   *toP = to;
 384 }
 385
 386 #ifdef XML_NS
 387 static const struct normal_encoding utf8_encoding_ns = {
 388   { VTABLE1, utf8_toUtf8, utf8_toUtf16, 1, 1, 0 },
 389   {
 390 #include "asciitab.h"
 391 #include "utf8tab.h"
 392   },
 393   STANDARD_VTABLE(sb_) NORMAL_VTABLE(utf8_)
 394 };
 395 #endif
 396
 397 static const struct normal_encoding utf8_encoding = {
 398   { VTABLE1, utf8_toUtf8, utf8_toUtf16, 1, 1, 0 },
 399   {
 400 #define BT_COLON BT_NMSTRT
 401 #include "asciitab.h"
 402 #undef BT_COLON
 403 #include "utf8tab.h"
 404   },
 405   STANDARD_VTABLE(sb_) NORMAL_VTABLE(utf8_)
 406 };
 407
 408 #ifdef XML_NS
 409
 410 static const struct normal_encoding internal_utf8_encoding_ns = {
 411   { VTABLE1, utf8_toUtf8, utf8_toUtf16, 1, 1, 0 },
 412   {
 413 #include "iasciitab.h"
 414 #include "utf8tab.h"
 415   },
 416   STANDARD_VTABLE(sb_) NORMAL_VTABLE(utf8_)
 417 };
 418
 419 #endif
 420
 421 static const struct normal_encoding internal_utf8_encoding = {
 422   { VTABLE1, utf8_toUtf8, utf8_toUtf16, 1, 1, 0 },
 423   {
 424 #define BT_COLON BT_NMSTRT
 425 #include "iasciitab.h"
 426 #undef BT_COLON
 427 #include "utf8tab.h"
 428   },
 429   STANDARD_VTABLE(sb_) NORMAL_VTABLE(utf8_)
 430 };
 431
 432 static void PTRCALL
 433 latin1_toUtf8(const ENCODING *enc,
 434               const char **fromP, const char *fromLim,
 435               char **toP, const char *toLim)
 436 {
 437   for (;;) {
 438     unsigned char c;
 439     if (*fromP == fromLim)
 440       break;
 441     c = (unsigned char)**fromP;
 442     if (c & 0x80) {
 443       if (toLim - *toP < 2)
 444         break;
 445       *(*toP)++ = (char)((c >> 6) | UTF8_cval2);
 446       *(*toP)++ = (char)((c & 0x3f) | 0x80);
 447       (*fromP)++;
 448     }
 449     else {
 450       if (*toP == toLim)
 451         break;
 452       *(*toP)++ = *(*fromP)++;
 453     }
 454   }
 455 }
 456
 457 static void PTRCALL
 458 latin1_toUtf16(const ENCODING *enc,
 459                const char **fromP, const char *fromLim,
 460                unsigned short **toP, const unsigned short *toLim)
 461 {
 462   while (*fromP != fromLim && *toP != toLim)
 463     *(*toP)++ = (unsigned char)*(*fromP)++;
 464 }
 465
 466 #ifdef XML_NS
 467
 468 static const struct normal_encoding latin1_encoding_ns = {
 469   { VTABLE1, latin1_toUtf8, latin1_toUtf16, 1, 0, 0 },
 470   {
 471 #include "asciitab.h"
 472 #include "latin1tab.h"
 473   },
 474   STANDARD_VTABLE(sb_)
 475 };
 476
 477 #endif
 478
 479 static const struct normal_encoding latin1_encoding = {
 480   { VTABLE1, latin1_toUtf8, latin1_toUtf16, 1, 0, 0 },
 481   {
 482 #define BT_COLON BT_NMSTRT
 483 #include "asciitab.h"
 484 #undef BT_COLON
 485 #include "latin1tab.h"
 486   },
 487   STANDARD_VTABLE(sb_)
 488 };
 489
 490 static void PTRCALL
 491 ascii_toUtf8(const ENCODING *enc,
 492              const char **fromP, const char *fromLim,
 493              char **toP, const char *toLim)
 494 {
 495   while (*fromP != fromLim && *toP != toLim)
 496     *(*toP)++ = *(*fromP)++;
 497 }
 498
 499 #ifdef XML_NS
 500
 501 static const struct normal_encoding ascii_encoding_ns = {
 502   { VTABLE1, ascii_toUtf8, latin1_toUtf16, 1, 1, 0 },
 503   {
 504 #include "asciitab.h"
 505 /* BT_NONXML == 0 */
 506   },
 507   STANDARD_VTABLE(sb_)
 508 };
 509
 510 #endif
 511
 512 static const struct normal_encoding ascii_encoding = {
 513   { VTABLE1, ascii_toUtf8, latin1_toUtf16, 1, 1, 0 },
 514   {
 515 #define BT_COLON BT_NMSTRT
 516 #include "asciitab.h"
 517 #undef BT_COLON
 518 /* BT_NONXML == 0 */
 519   },
 520   STANDARD_VTABLE(sb_)
 521 };
 522
 523 static int PTRFASTCALL
 524 unicode_byte_type(char hi, char lo)
 525 {
 526   switch ((unsigned char)hi) {
 527   case 0xD8: case 0xD9: case 0xDA: case 0xDB:
 528     return BT_LEAD4;
 529   case 0xDC: case 0xDD: case 0xDE: case 0xDF:
 530     return BT_TRAIL;
 531   case 0xFF:
 532     switch ((unsigned char)lo) {
 533     case 0xFF:
 534     case 0xFE:
 535       return BT_NONXML;
 536     }
 537     break;
 538   }
 539   return BT_NONASCII;
 540 }
 541
 542 #define DEFINE_UTF16_TO_UTF8(E) \
 543 static void  PTRCALL \
 544 E ## toUtf8(const ENCODING *enc, \
 545             const char **fromP, const char *fromLim, \
 546             char **toP, const char *toLim) \
 547 { \
 548   const char *from; \
 549   for (from = *fromP; from != fromLim; from += 2) { \
 550     int plane; \
 551     unsigned char lo2; \
 552     unsigned char lo = GET_LO(from); \
 553     unsigned char hi = GET_HI(from); \
 554     switch (hi) { \
 555     case 0: \
 556       if (lo < 0x80) { \
 557         if (*toP == toLim) { \
 558           *fromP = from; \
 559           return; \
 560         } \
 561         *(*toP)++ = lo; \
 562         break; \
 563       } \
 564       /* fall through */ \
 565     case 0x1: case 0x2: case 0x3: \
 566     case 0x4: case 0x5: case 0x6: case 0x7: \
 567       if (toLim -  *toP < 2) { \
 568         *fromP = from; \
 569         return; \
 570       } \
 571       *(*toP)++ = ((lo >> 6) | (hi << 2) |  UTF8_cval2); \
 572       *(*toP)++ = ((lo & 0x3f) | 0x80); \
 573       break; \
 574     default: \
 575       if (toLim -  *toP < 3)  { \
 576         *fromP = from; \
 577         return; \
 578       } \
 579       /* 16 bits divided 4, 6, 6 amongst 3 bytes */ \
 580       *(*toP)++ = ((hi >> 4) | UTF8_cval3); \
 581       *(*toP)++ = (((hi & 0xf) << 2) | (lo >> 6) | 0x80); \
 582       *(*toP)++ = ((lo & 0x3f) | 0x80); \
 583       break; \
 584     case 0xD8: case 0xD9: case 0xDA: case 0xDB: \
 585       if (toLim -  *toP < 4) { \
 586         *fromP = from; \
 587         return; \
 588       } \
 589       plane = (((hi & 0x3) << 2) | ((lo >> 6) & 0x3)) + 1; \
 590       *(*toP)++ = ((plane >> 2) | UTF8_cval4); \
 591       *(*toP)++ = (((lo >> 2) & 0xF) | ((plane & 0x3) << 4) | 0x80); \
 592       from += 2; \
 593       lo2 = GET_LO(from); \
 594       *(*toP)++ = (((lo & 0x3) << 4) \
 595                    | ((GET_HI(from) & 0x3) << 2) \
 596                    | (lo2 >> 6) \
 597                    | 0x80); \
 598       *(*toP)++ = ((lo2 & 0x3f) | 0x80); \
 599       break; \
 600     } \
 601   } \
 602   *fromP = from; \
 603 }
 604
 605 #define DEFINE_UTF16_TO_UTF16(E) \
 606 static void  PTRCALL \
 607 E ## toUtf16(const ENCODING *enc, \
 608              const char **fromP, const char *fromLim, \
 609              unsigned short **toP, const unsigned short *toLim) \
 610 { \
 611   /* Avoid copying first half only of surrogate */ \
 612   if (fromLim - *fromP > ((toLim - *toP) << 1) \
 613       && (GET_HI(fromLim - 2) & 0xF8) == 0xD8) \
 614     fromLim -= 2; \
 615   for (; *fromP != fromLim && *toP != toLim; *fromP += 2) \
 616     *(*toP)++ = (GET_HI(*fromP) << 8) | GET_LO(*fromP); \
 617 }
 618
 619 #define SET2(ptr, ch) \
 620   (((ptr)[0] = ((ch) & 0xff)), ((ptr)[1] = ((ch) >> 8)))
 621 #define GET_LO(ptr) ((unsigned char)(ptr)[0])
 622 #define GET_HI(ptr) ((unsigned char)(ptr)[1])
 623
 624 DEFINE_UTF16_TO_UTF8(little2_)
 625 DEFINE_UTF16_TO_UTF16(little2_)
 626
 627 #undef SET2
 628 #undef GET_LO
 629 #undef GET_HI
 630
 631 #define SET2(ptr, ch) \
 632   (((ptr)[0] = ((ch) >> 8)), ((ptr)[1] = ((ch) & 0xFF)))
 633 #define GET_LO(ptr) ((unsigned char)(ptr)[1])
 634 #define GET_HI(ptr) ((unsigned char)(ptr)[0])
 635
 636 DEFINE_UTF16_TO_UTF8(big2_)
 637 DEFINE_UTF16_TO_UTF16(big2_)
 638
 639 #undef SET2
 640 #undef GET_LO
 641 #undef GET_HI
 642
 643 #define LITTLE2_BYTE_TYPE(enc, p) \
 644  ((p)[1] == 0 \
 645   ? ((struct normal_encoding *)(enc))->type[(unsigned char)*(p)] \
 646   : unicode_byte_type((p)[1], (p)[0]))
 647 #define LITTLE2_BYTE_TO_ASCII(enc, p) ((p)[1] == 0 ? (p)[0] : -1)
 648 #define LITTLE2_CHAR_MATCHES(enc, p, c) ((p)[1] == 0 && (p)[0] == c)
 649 #define LITTLE2_IS_NAME_CHAR_MINBPC(enc, p) \
 650   UCS2_GET_NAMING(namePages, (unsigned char)p[1], (unsigned char)p[0])
 651 #define LITTLE2_IS_NMSTRT_CHAR_MINBPC(enc, p) \
 652   UCS2_GET_NAMING(nmstrtPages, (unsigned char)p[1], (unsigned char)p[0])
 653
 654 #ifdef XML_MIN_SIZE
 655
 656 static int PTRFASTCALL
 657 little2_byteType(const ENCODING *enc, const char *p)
 658 {
 659   return LITTLE2_BYTE_TYPE(enc, p);
 660 }
 661
 662 static int PTRFASTCALL
 663 little2_byteToAscii(const ENCODING *enc, const char *p)
 664 {
 665   return LITTLE2_BYTE_TO_ASCII(enc, p);
 666 }
 667
 668 static int PTRCALL
 669 little2_charMatches(const ENCODING *enc, const char *p, int c)
 670 {
 671   return LITTLE2_CHAR_MATCHES(enc, p, c);
 672 }
 673
 674 static int PTRFASTCALL
 675 little2_isNameMin(const ENCODING *enc, const char *p)
 676 {
 677   return LITTLE2_IS_NAME_CHAR_MINBPC(enc, p);
 678 }
 679
 680 static int PTRFASTCALL
 681 little2_isNmstrtMin(const ENCODING *enc, const char *p)
 682 {
 683   return LITTLE2_IS_NMSTRT_CHAR_MINBPC(enc, p);
 684 }
 685
 686 #undef VTABLE
 687 #define VTABLE VTABLE1, little2_toUtf8, little2_toUtf16
 688
 689 #else /* not XML_MIN_SIZE */
 690
 691 #undef PREFIX
 692 #define PREFIX(ident) little2_ ## ident
 693 #define MINBPC(enc) 2
 694 /* CHAR_MATCHES is guaranteed to have MINBPC bytes available. */
 695 #define BYTE_TYPE(enc, p) LITTLE2_BYTE_TYPE(enc, p)
 696 #define BYTE_TO_ASCII(enc, p) LITTLE2_BYTE_TO_ASCII(enc, p)
 697 #define CHAR_MATCHES(enc, p, c) LITTLE2_CHAR_MATCHES(enc, p, c)
 698 #define IS_NAME_CHAR(enc, p, n) 0
 699 #define IS_NAME_CHAR_MINBPC(enc, p) LITTLE2_IS_NAME_CHAR_MINBPC(enc, p)
 700 #define IS_NMSTRT_CHAR(enc, p, n) (0)
 701 #define IS_NMSTRT_CHAR_MINBPC(enc, p) LITTLE2_IS_NMSTRT_CHAR_MINBPC(enc, p)
 702
 703 #define XML_TOK_IMPL_C
 704 #include "xmltok_impl.c"
 705 #undef XML_TOK_IMPL_C
 706
 707 #undef MINBPC
 708 #undef BYTE_TYPE
 709 #undef BYTE_TO_ASCII
 710 #undef CHAR_MATCHES
 711 #undef IS_NAME_CHAR
 712 #undef IS_NAME_CHAR_MINBPC
 713 #undef IS_NMSTRT_CHAR
 714 #undef IS_NMSTRT_CHAR_MINBPC
 715 #undef IS_INVALID_CHAR
 716
 717 #endif /* not XML_MIN_SIZE */
 718
 719 #ifdef XML_NS
 720
 721 static const struct normal_encoding little2_encoding_ns = {
 722   { VTABLE, 2, 0,
 723 #if BYTEORDER == 1234
 724     1
 725 #else
 726     0
 727 #endif
 728   },
 729   {
 730 #include "asciitab.h"
 731 #include "latin1tab.h"
 732   },
 733   STANDARD_VTABLE(little2_)
 734 };
 735
 736 #endif
 737
 738 static const struct normal_encoding little2_encoding = {
 739   { VTABLE, 2, 0,
 740 #if BYTEORDER == 1234
 741     1
 742 #else
 743     0
 744 #endif
 745   },
 746   {
 747 #define BT_COLON BT_NMSTRT
 748 #include "asciitab.h"
 749 #undef BT_COLON
 750 #include "latin1tab.h"
 751   },
 752   STANDARD_VTABLE(little2_)
 753 };
 754
 755 #if BYTEORDER != 4321
 756
 757 #ifdef XML_NS
 758
 759 static const struct normal_encoding internal_little2_encoding_ns = {
 760   { VTABLE, 2, 0, 1 },
 761   {
 762 #include "iasciitab.h"
 763 #include "latin1tab.h"
 764   },
 765   STANDARD_VTABLE(little2_)
 766 };
 767
 768 #endif
 769
 770 static const struct normal_encoding internal_little2_encoding = {
 771   { VTABLE, 2, 0, 1 },
 772   {
 773 #define BT_COLON BT_NMSTRT
 774 #include "iasciitab.h"
 775 #undef BT_COLON
 776 #include "latin1tab.h"
 777   },
 778   STANDARD_VTABLE(little2_)
 779 };
 780
 781 #endif
 782
 783
 784 #define BIG2_BYTE_TYPE(enc, p) \
 785  ((p)[0] == 0 \
 786   ? ((struct normal_encoding *)(enc))->type[(unsigned char)(p)[1]] \
 787   : unicode_byte_type((p)[0], (p)[1]))
 788 #define BIG2_BYTE_TO_ASCII(enc, p) ((p)[0] == 0 ? (p)[1] : -1)
 789 #define BIG2_CHAR_MATCHES(enc, p, c) ((p)[0] == 0 && (p)[1] == c)
 790 #define BIG2_IS_NAME_CHAR_MINBPC(enc, p) \
 791   UCS2_GET_NAMING(namePages, (unsigned char)p[0], (unsigned char)p[1])
 792 #define BIG2_IS_NMSTRT_CHAR_MINBPC(enc, p) \
 793   UCS2_GET_NAMING(nmstrtPages, (unsigned char)p[0], (unsigned char)p[1])
 794
 795 #ifdef XML_MIN_SIZE
 796
 797 static int PTRFASTCALL
 798 big2_byteType(const ENCODING *enc, const char *p)
 799 {
 800   return BIG2_BYTE_TYPE(enc, p);
 801 }
 802
 803 static int PTRFASTCALL
 804 big2_byteToAscii(const ENCODING *enc, const char *p)
 805 {
 806   return BIG2_BYTE_TO_ASCII(enc, p);
 807 }
 808
 809 static int PTRCALL
 810 big2_charMatches(const ENCODING *enc, const char *p, int c)
 811 {
 812   return BIG2_CHAR_MATCHES(enc, p, c);
 813 }
 814
 815 static int PTRFASTCALL
 816 big2_isNameMin(const ENCODING *enc, const char *p)
 817 {
 818   return BIG2_IS_NAME_CHAR_MINBPC(enc, p);
 819 }
 820
 821 static int PTRFASTCALL
 822 big2_isNmstrtMin(const ENCODING *enc, const char *p)
 823 {
 824   return BIG2_IS_NMSTRT_CHAR_MINBPC(enc, p);
 825 }
 826
 827 #undef VTABLE
 828 #define VTABLE VTABLE1, big2_toUtf8, big2_toUtf16
 829
 830 #else /* not XML_MIN_SIZE */
 831
 832 #undef PREFIX
 833 #define PREFIX(ident) big2_ ## ident
 834 #define MINBPC(enc) 2
 835 /* CHAR_MATCHES is guaranteed to have MINBPC bytes available. */
 836 #define BYTE_TYPE(enc, p) BIG2_BYTE_TYPE(enc, p)
 837 #define BYTE_TO_ASCII(enc, p) BIG2_BYTE_TO_ASCII(enc, p)
 838 #define CHAR_MATCHES(enc, p, c) BIG2_CHAR_MATCHES(enc, p, c)
 839 #define IS_NAME_CHAR(enc, p, n) 0
 840 #define IS_NAME_CHAR_MINBPC(enc, p) BIG2_IS_NAME_CHAR_MINBPC(enc, p)
 841 #define IS_NMSTRT_CHAR(enc, p, n) (0)
 842 #define IS_NMSTRT_CHAR_MINBPC(enc, p) BIG2_IS_NMSTRT_CHAR_MINBPC(enc, p)
 843
 844 #define XML_TOK_IMPL_C
 845 #include "xmltok_impl.c"
 846 #undef XML_TOK_IMPL_C
 847
 848 #undef MINBPC
 849 #undef BYTE_TYPE
 850 #undef BYTE_TO_ASCII
 851 #undef CHAR_MATCHES
 852 #undef IS_NAME_CHAR
 853 #undef IS_NAME_CHAR_MINBPC
 854 #undef IS_NMSTRT_CHAR
 855 #undef IS_NMSTRT_CHAR_MINBPC
 856 #undef IS_INVALID_CHAR
 857
 858 #endif /* not XML_MIN_SIZE */
 859
 860 #ifdef XML_NS
 861
 862 static const struct normal_encoding big2_encoding_ns = {
 863   { VTABLE, 2, 0,
 864 #if BYTEORDER == 4321
 865   1
 866 #else
 867   0
 868 #endif
 869   },
 870   {
 871 #include "asciitab.h"
 872 #include "latin1tab.h"
 873   },
 874   STANDARD_VTABLE(big2_)
 875 };
 876
 877 #endif
 878
 879 static const struct normal_encoding big2_encoding = {
 880   { VTABLE, 2, 0,
 881 #if BYTEORDER == 4321
 882   1
 883 #else
 884   0
 885 #endif
 886   },
 887   {
 888 #define BT_COLON BT_NMSTRT
 889 #include "asciitab.h"
 890 #undef BT_COLON
 891 #include "latin1tab.h"
 892   },
 893   STANDARD_VTABLE(big2_)
 894 };
 895
 896 #if BYTEORDER != 1234
 897
 898 #ifdef XML_NS
 899
 900 static const struct normal_encoding internal_big2_encoding_ns = {
 901   { VTABLE, 2, 0, 1 },
 902   {
 903 #include "iasciitab.h"
 904 #include "latin1tab.h"
 905   },
 906   STANDARD_VTABLE(big2_)
 907 };
 908
 909 #endif
 910
 911 static const struct normal_encoding internal_big2_encoding = {
 912   { VTABLE, 2, 0, 1 },
 913   {
 914 #define BT_COLON BT_NMSTRT
 915 #include "iasciitab.h"
 916 #undef BT_COLON
 917 #include "latin1tab.h"
 918   },
 919   STANDARD_VTABLE(big2_)
 920 };
 921
 922 #endif
 923
 924 #undef PREFIX
 925
 926 static int FASTCALL
 927 streqci(const char *s1, const char *s2)
 928 {
 929   for (;;) {
 930     char c1 = *s1++;
 931     char c2 = *s2++;
 932     if (ASCII_a <= c1 && c1 <= ASCII_z)
 933       c1 += ASCII_A - ASCII_a;
 934     if (ASCII_a <= c2 && c2 <= ASCII_z)
 935       c2 += ASCII_A - ASCII_a;
 936     if (c1 != c2)
 937       return 0;
 938     if (!c1)
 939       break;
 940   }
 941   return 1;
 942 }
 943
 944 static void PTRCALL
 945 initUpdatePosition(const ENCODING *enc, const char *ptr,
 946                    const char *end, POSITION *pos)
 947 {
 948   normal_updatePosition(&utf8_encoding.enc, ptr, end, pos);
 949 }
 950
 951 static int
 952 toAscii(const ENCODING *enc, const char *ptr, const char *end)
 953 {
 954   char buf[1];
 955   char *p = buf;
 956   XmlUtf8Convert(enc, &ptr, end, &p, p + 1);
 957   if (p == buf)
 958     return -1;
 959   else
 960     return buf[0];
 961 }
 962
 963 static int FASTCALL
 964 isSpace(int c)
 965 {
 966   switch (c) {
 967   case 0x20:
 968   case 0xD:
 969   case 0xA:
 970   case 0x9:
 971     return 1;
 972   }
 973   return 0;
 974 }
 975
 976 /* Return 1 if there's just optional white space or there's an S
 977    followed by name=val.
 978 */
 979 static int
 980 parsePseudoAttribute(const ENCODING *enc,
 981                      const char *ptr,
 982                      const char *end,
 983                      const char **namePtr,
 984                      const char **nameEndPtr,
 985                      const char **valPtr,
 986                      const char **nextTokPtr)
 987 {
 988   int c;
 989   char open;
 990   if (ptr == end) {
 991     *namePtr = NULL;
 992     return 1;
 993   }
 994   if (!isSpace(toAscii(enc, ptr, end))) {
 995     *nextTokPtr = ptr;
 996     return 0;
 997   }
 998   do {
 999     ptr += enc->minBytesPerChar;
1000   } while (isSpace(toAscii(enc, ptr, end)));
1001   if (ptr == end) {
1002     *namePtr = NULL;
1003     return 1;
1004   }
1005   *namePtr = ptr;
1006   for (;;) {
1007     c = toAscii(enc, ptr, end);
1008     if (c == -1) {
1009       *nextTokPtr = ptr;
1010       return 0;
1011     }
1012     if (c == ASCII_EQUALS) {
1013       *nameEndPtr = ptr;
1014       break;
1015     }
1016     if (isSpace(c)) {
1017       *nameEndPtr = ptr;
1018       do {
1019         ptr += enc->minBytesPerChar;
1020       } while (isSpace(c = toAscii(enc, ptr, end)));
1021       if (c != ASCII_EQUALS) {
1022         *nextTokPtr = ptr;
1023         return 0;
1024       }
1025       break;
1026     }
1027     ptr += enc->minBytesPerChar;
1028   }
1029   if (ptr == *namePtr) {
1030     *nextTokPtr = ptr;
1031     return 0;
1032   }
1033   ptr += enc->minBytesPerChar;
1034   c = toAscii(enc, ptr, end);
1035   while (isSpace(c)) {
1036     ptr += enc->minBytesPerChar;
1037     c = toAscii(enc, ptr, end);
1038   }
1039   if (c != ASCII_QUOT && c != ASCII_APOS) {
1040     *nextTokPtr = ptr;
1041     return 0;
1042   }
1043   open = (char)c;
1044   ptr += enc->minBytesPerChar;
1045   *valPtr = ptr;
1046   for (;; ptr += enc->minBytesPerChar) {
1047     c = toAscii(enc, ptr, end);
1048     if (c == open)
1049       break;
1050     if (!(ASCII_a <= c && c <= ASCII_z)
1051         && !(ASCII_A <= c && c <= ASCII_Z)
1052         && !(ASCII_0 <= c && c <= ASCII_9)
1053         && c != ASCII_PERIOD
1054         && c != ASCII_MINUS
1055         && c != ASCII_UNDERSCORE) {
1056       *nextTokPtr = ptr;
1057       return 0;
1058     }
1059   }
1060   *nextTokPtr = ptr + enc->minBytesPerChar;
1061   return 1;
1062 }
1063
1064 static const char KW_version[] = {
1065   ASCII_v, ASCII_e, ASCII_r, ASCII_s, ASCII_i, ASCII_o, ASCII_n, '\0'
1066 };
1067
1068 static const char KW_encoding[] = {
1069   ASCII_e, ASCII_n, ASCII_c, ASCII_o, ASCII_d, ASCII_i, ASCII_n, ASCII_g, '\0'
1070 };
1071
1072 static const char KW_standalone[] = {
1073   ASCII_s, ASCII_t, ASCII_a, ASCII_n, ASCII_d, ASCII_a, ASCII_l, ASCII_o,
1074   ASCII_n, ASCII_e, '\0'
1075 };
1076
1077 static const char KW_yes[] = {
1078   ASCII_y, ASCII_e, ASCII_s,  '\0'
1079 };
1080
1081 static const char KW_no[] = {
1082   ASCII_n, ASCII_o,  '\0'
1083 };
1084
1085 static int
1086 doParseXmlDecl(const ENCODING *(*encodingFinder)(const ENCODING *,
1087                                                  const char *,
1088                                                  const char *),
1089                int isGeneralTextEntity,
1090                const ENCODING *enc,
1091                const char *ptr,
1092                const char *end,
1093                const char **badPtr,
1094                const char **versionPtr,
1095                const char **versionEndPtr,
1096                const char **encodingName,
1097                const ENCODING **encoding,
1098                int *standalone)
1099 {
1100   const char *val = NULL;
1101   const char *name = NULL;
1102   const char *nameEnd = NULL;
1103   ptr += 5 * enc->minBytesPerChar;
1104   end -= 2 * enc->minBytesPerChar;
1105   if (!parsePseudoAttribute(enc, ptr, end, &name, &nameEnd, &val, &ptr)
1106       || !name) {
1107     *badPtr = ptr;
1108     return 0;
1109   }
1110   if (!XmlNameMatchesAscii(enc, name, nameEnd, KW_version)) {
1111     if (!isGeneralTextEntity) {
1112       *badPtr = name;
1113       return 0;
1114     }
1115   }
1116   else {
1117     if (versionPtr)
1118       *versionPtr = val;
1119     if (versionEndPtr)
1120       *versionEndPtr = ptr;
1121     if (!parsePseudoAttribute(enc, ptr, end, &name, &nameEnd, &val, &ptr)) {
1122       *badPtr = ptr;
1123       return 0;
1124     }
1125     if (!name) {
1126       if (isGeneralTextEntity) {
1127         /* a TextDecl must have an EncodingDecl */
1128         *badPtr = ptr;
1129         return 0;
1130       }
1131       return 1;
1132     }
1133   }
1134   if (XmlNameMatchesAscii(enc, name, nameEnd, KW_encoding)) {
1135     int c = toAscii(enc, val, end);
1136     if (!(ASCII_a <= c && c <= ASCII_z) && !(ASCII_A <= c && c <= ASCII_Z)) {
1137       *badPtr = val;
1138       return 0;
1139     }
1140     if (encodingName)
1141       *encodingName = val;
1142     if (encoding)
1143       *encoding = encodingFinder(enc, val, ptr - enc->minBytesPerChar);
1144     if (!parsePseudoAttribute(enc, ptr, end, &name, &nameEnd, &val, &ptr)) {
1145       *badPtr = ptr;
1146       return 0;
1147     }
1148     if (!name)
1149       return 1;
1150   }
1151   if (!XmlNameMatchesAscii(enc, name, nameEnd, KW_standalone)
1152       || isGeneralTextEntity) {
1153     *badPtr = name;
1154     return 0;
1155   }
1156   if (XmlNameMatchesAscii(enc, val, ptr - enc->minBytesPerChar, KW_yes)) {
1157     if (standalone)
1158       *standalone = 1;
1159   }
1160   else if (XmlNameMatchesAscii(enc, val, ptr - enc->minBytesPerChar, KW_no)) {
1161     if (standalone)
1162       *standalone = 0;
1163   }
1164   else {
1165     *badPtr = val;
1166     return 0;
1167   }
1168   while (isSpace(toAscii(enc, ptr, end)))
1169     ptr += enc->minBytesPerChar;
1170   if (ptr != end) {
1171     *badPtr = ptr;
1172     return 0;
1173   }
1174   return 1;
1175 }
1176
1177 static int FASTCALL
1178 checkCharRefNumber(int result)
1179 {
1180   switch (result >> 8) {
1181   case 0xD8: case 0xD9: case 0xDA: case 0xDB:
1182   case 0xDC: case 0xDD: case 0xDE: case 0xDF:
1183     return -1;
1184   case 0:
1185     if (latin1_encoding.type[result] == BT_NONXML)
1186       return -1;
1187     break;
1188   case 0xFF:
1189     if (result == 0xFFFE || result == 0xFFFF)
1190       return -1;
1191     break;
1192   }
1193   return result;
1194 }
1195
1196 int FASTCALL
1197 XmlUtf8Encode(int c, char *buf)
1198 {
1199   enum {
1200     /* minN is minimum legal resulting value for N byte sequence */
1201     min2 = 0x80,
1202     min3 = 0x800,
1203     min4 = 0x10000
1204   };
1205
1206   if (c < 0)
1207     return 0;
1208   if (c < min2) {
1209     buf[0] = (char)(c | UTF8_cval1);
1210     return 1;
1211   }
1212   if (c < min3) {
1213     buf[0] = (char)((c >> 6) | UTF8_cval2);
1214     buf[1] = (char)((c & 0x3f) | 0x80);
1215     return 2;
1216   }
1217   if (c < min4) {
1218     buf[0] = (char)((c >> 12) | UTF8_cval3);
1219     buf[1] = (char)(((c >> 6) & 0x3f) | 0x80);
1220     buf[2] = (char)((c & 0x3f) | 0x80);
1221     return 3;
1222   }
1223   if (c < 0x110000) {
1224     buf[0] = (char)((c >> 18) | UTF8_cval4);
1225     buf[1] = (char)(((c >> 12) & 0x3f) | 0x80);
1226     buf[2] = (char)(((c >> 6) & 0x3f) | 0x80);
1227     buf[3] = (char)((c & 0x3f) | 0x80);
1228     return 4;
1229   }
1230   return 0;
1231 }
1232
1233 int FASTCALL
1234 XmlUtf16Encode(int charNum, unsigned short *buf)
1235 {
1236   if (charNum < 0)
1237     return 0;
1238   if (charNum < 0x10000) {
1239     buf[0] = (unsigned short)charNum;
1240     return 1;
1241   }
1242   if (charNum < 0x110000) {
1243     charNum -= 0x10000;
1244     buf[0] = (unsigned short)((charNum >> 10) + 0xD800);
1245     buf[1] = (unsigned short)((charNum & 0x3FF) + 0xDC00);
1246     return 2;
1247   }
1248   return 0;
1249 }
1250
1251 struct unknown_encoding {
1252   struct normal_encoding normal;
1253   CONVERTER convert;
1254   void *userData;
1255   unsigned short utf16[256];
1256   char utf8[256][4];
1257 };
1258
1259 #define AS_UNKNOWN_ENCODING(enc)  ((const struct unknown_encoding *) (enc))
1260
1261 int
1262 XmlSizeOfUnknownEncoding(void)
1263 {
1264   return sizeof(struct unknown_encoding);
1265 }
1266
1267 static int PTRFASTCALL
1268 unknown_isName(const ENCODING *enc, const char *p)
1269 {
1270   const struct unknown_encoding *uenc = AS_UNKNOWN_ENCODING(enc);
1271   int c = uenc->convert(uenc->userData, p);
1272   if (c & ~0xFFFF)
1273     return 0;
1274   return UCS2_GET_NAMING(namePages, c >> 8, c & 0xFF);
1275 }
1276
1277 static int PTRFASTCALL
1278 unknown_isNmstrt(const ENCODING *enc, const char *p)
1279 {
1280   const struct unknown_encoding *uenc = AS_UNKNOWN_ENCODING(enc);
1281   int c = uenc->convert(uenc->userData, p);
1282   if (c & ~0xFFFF)
1283     return 0;
1284   return UCS2_GET_NAMING(nmstrtPages, c >> 8, c & 0xFF);
1285 }
1286
1287 static int PTRFASTCALL
1288 unknown_isInvalid(const ENCODING *enc, const char *p)
1289 {
1290   const struct unknown_encoding *uenc = AS_UNKNOWN_ENCODING(enc);
1291   int c = uenc->convert(uenc->userData, p);
1292   return (c & ~0xFFFF) || checkCharRefNumber(c) < 0;
1293 }
1294
1295 static void PTRCALL
1296 unknown_toUtf8(const ENCODING *enc,
1297                const char **fromP, const char *fromLim,
1298                char **toP, const char *toLim)
1299 {
1300   const struct unknown_encoding *uenc = AS_UNKNOWN_ENCODING(enc);
1301   char buf[XML_UTF8_ENCODE_MAX];
1302   for (;;) {
1303     const char *utf8;
1304     int n;
1305     if (*fromP == fromLim)
1306       break;
1307     utf8 = uenc->utf8[(unsigned char)**fromP];
1308     n = *utf8++;
1309     if (n == 0) {
1310       int c = uenc->convert(uenc->userData, *fromP);
1311       n = XmlUtf8Encode(c, buf);
1312       if (n > toLim - *toP)
1313         break;
1314       utf8 = buf;
1315       *fromP += (AS_NORMAL_ENCODING(enc)->type[(unsigned char)**fromP]
1316                  - (BT_LEAD2 - 2));
1317     }
1318     else {
1319       if (n > toLim - *toP)
1320         break;
1321       (*fromP)++;
1322     }
1323     do {
1324       *(*toP)++ = *utf8++;
1325     } while (--n != 0);
1326   }
1327 }
1328
1329 static void PTRCALL
1330 unknown_toUtf16(const ENCODING *enc,
1331                 const char **fromP, const char *fromLim,
1332                 unsigned short **toP, const unsigned short *toLim)
1333 {
1334   const struct unknown_encoding *uenc = AS_UNKNOWN_ENCODING(enc);
1335   while (*fromP != fromLim && *toP != toLim) {
1336     unsigned short c = uenc->utf16[(unsigned char)**fromP];
1337     if (c == 0) {
1338       c = (unsigned short)
1339           uenc->convert(uenc->userData, *fromP);
1340       *fromP += (AS_NORMAL_ENCODING(enc)->type[(unsigned char)**fromP]
1341                  - (BT_LEAD2 - 2));
1342     }
1343     else
1344       (*fromP)++;
1345     *(*toP)++ = c;
1346   }
1347 }
1348
1349 ENCODING *
1350 XmlInitUnknownEncoding(void *mem,
1351                        int *table,
1352                        CONVERTER convert,
1353                        void *userData)
1354 {
1355   int i;
1356   struct unknown_encoding *e = (struct unknown_encoding *)mem;
1357   for (i = 0; i < (int)sizeof(struct normal_encoding); i++)
1358     ((char *)mem)[i] = ((char *)&latin1_encoding)[i];
1359   for (i = 0; i < 128; i++)
1360     if (latin1_encoding.type[i] != BT_OTHER
1361         && latin1_encoding.type[i] != BT_NONXML
1362         && table[i] != i)
1363       return 0;
1364   for (i = 0; i < 256; i++) {
1365     int c = table[i];
1366     if (c == -1) {
1367       e->normal.type[i] = BT_MALFORM;
1368       /* This shouldn't really get used. */
1369       e->utf16[i] = 0xFFFF;
1370       e->utf8[i][0] = 1;
1371       e->utf8[i][1] = 0;
1372     }
1373     else if (c < 0) {
1374       if (c < -4)
1375         return 0;
1376       e->normal.type[i] = (unsigned char)(BT_LEAD2 - (c + 2));
1377       e->utf8[i][0] = 0;
1378       e->utf16[i] = 0;
1379     }
1380     else if (c < 0x80) {
1381       if (latin1_encoding.type[c] != BT_OTHER
1382           && latin1_encoding.type[c] != BT_NONXML
1383           && c != i)
1384         return 0;
1385       e->normal.type[i] = latin1_encoding.type[c];
1386       e->utf8[i][0] = 1;
1387       e->utf8[i][1] = (char)c;
1388       e->utf16[i] = (unsigned short)(c == 0 ? 0xFFFF : c);
1389     }
1390     else if (checkCharRefNumber(c) < 0) {
1391       e->normal.type[i] = BT_NONXML;
1392       /* This shouldn't really get used. */
1393       e->utf16[i] = 0xFFFF;
1394       e->utf8[i][0] = 1;
1395       e->utf8[i][1] = 0;
1396     }
1397     else {
1398       if (c > 0xFFFF)
1399         return 0;
1400       if (UCS2_GET_NAMING(nmstrtPages, c >> 8, c & 0xff))
1401         e->normal.type[i] = BT_NMSTRT;
1402       else if (UCS2_GET_NAMING(namePages, c >> 8, c & 0xff))
1403         e->normal.type[i] = BT_NAME;
1404       else
1405         e->normal.type[i] = BT_OTHER;
1406       e->utf8[i][0] = (char)XmlUtf8Encode(c, e->utf8[i] + 1);
1407       e->utf16[i] = (unsigned short)c;
1408     }
1409   }
1410   e->userData = userData;
1411   e->convert = convert;
1412   if (convert) {
1413     e->normal.isName2 = unknown_isName;
1414     e->normal.isName3 = unknown_isName;
1415     e->normal.isName4 = unknown_isName;
1416     e->normal.isNmstrt2 = unknown_isNmstrt;
1417     e->normal.isNmstrt3 = unknown_isNmstrt;
1418     e->normal.isNmstrt4 = unknown_isNmstrt;
1419     e->normal.isInvalid2 = unknown_isInvalid;
1420     e->normal.isInvalid3 = unknown_isInvalid;
1421     e->normal.isInvalid4 = unknown_isInvalid;
1422   }
1423   e->normal.enc.utf8Convert = unknown_toUtf8;
1424   e->normal.enc.utf16Convert = unknown_toUtf16;
1425   return &(e->normal.enc);
1426 }
1427
1428 /* If this enumeration is changed, getEncodingIndex and encodings
1429 must also be changed. */
1430 enum {
1431   UNKNOWN_ENC = -1,
1432   ISO_8859_1_ENC = 0,
1433   US_ASCII_ENC,
1434   UTF_8_ENC,
1435   UTF_16_ENC,
1436   UTF_16BE_ENC,
1437   UTF_16LE_ENC,
1438   /* must match encodingNames up to here */
1439   NO_ENC
1440 };
1441
1442 static const char KW_ISO_8859_1[] = {
1443   ASCII_I, ASCII_S, ASCII_O, ASCII_MINUS, ASCII_8, ASCII_8, ASCII_5, ASCII_9,
1444   ASCII_MINUS, ASCII_1, '\0'
1445 };
1446 static const char KW_US_ASCII[] = {
1447   ASCII_U, ASCII_S, ASCII_MINUS, ASCII_A, ASCII_S, ASCII_C, ASCII_I, ASCII_I,
1448   '\0'
1449 };
1450 static const char KW_UTF_8[] =  {
1451   ASCII_U, ASCII_T, ASCII_F, ASCII_MINUS, ASCII_8, '\0'
1452 };
1453 static const char KW_UTF_16[] = {
1454   ASCII_U, ASCII_T, ASCII_F, ASCII_MINUS, ASCII_1, ASCII_6, '\0'
1455 };
1456 static const char KW_UTF_16BE[] = {
1457   ASCII_U, ASCII_T, ASCII_F, ASCII_MINUS, ASCII_1, ASCII_6, ASCII_B, ASCII_E,
1458   '\0'
1459 };
1460 static const char KW_UTF_16LE[] = {
1461   ASCII_U, ASCII_T, ASCII_F, ASCII_MINUS, ASCII_1, ASCII_6, ASCII_L, ASCII_E,
1462   '\0'
1463 };
1464
1465 static int FASTCALL
1466 getEncodingIndex(const char *name)
1467 {
1468   static const char * const encodingNames[] = {
1469     KW_ISO_8859_1,
1470     KW_US_ASCII,
1471     KW_UTF_8,
1472     KW_UTF_16,
1473     KW_UTF_16BE,
1474     KW_UTF_16LE,
1475   };
1476   int i;
1477   if (name == NULL)
1478     return NO_ENC;
1479   for (i = 0; i < (int)(sizeof(encodingNames)/sizeof(encodingNames[0])); i++)
1480     if (streqci(name, encodingNames[i]))
1481       return i;
1482   return UNKNOWN_ENC;
1483 }
1484
1485 /* For binary compatibility, we store the index of the encoding
1486    specified at initialization in the isUtf16 member.
1487 */
1488
1489 #define INIT_ENC_INDEX(enc) ((int)(enc)->initEnc.isUtf16)
1490 #define SET_INIT_ENC_INDEX(enc, i) ((enc)->initEnc.isUtf16 = (char)i)
1491
1492 /* This is what detects the encoding.  encodingTable maps from
1493    encoding indices to encodings; INIT_ENC_INDEX(enc) is the index of
1494    the external (protocol) specified encoding; state is
1495    XML_CONTENT_STATE if we're parsing an external text entity, and
1496    XML_PROLOG_STATE otherwise.
1497 */
1498
1499
1500 static int
1501 initScan(const ENCODING * const *encodingTable,
1502          const INIT_ENCODING *enc,
1503          int state,
1504          const char *ptr,
1505          const char *end,
1506          const char **nextTokPtr)
1507 {
1508   const ENCODING **encPtr;
1509
1510   if (ptr == end)
1511     return XML_TOK_NONE;
1512   encPtr = enc->encPtr;
1513   if (ptr + 1 == end) {
1514     /* only a single byte available for auto-detection */
1515 #ifndef XML_DTD /* FIXME */
1516     /* a well-formed document entity must have more than one byte */
1517     if (state != XML_CONTENT_STATE)
1518       return XML_TOK_PARTIAL;
1519 #endif
1520     /* so we're parsing an external text entity... */
1521     /* if UTF-16 was externally specified, then we need at least 2 bytes */
1522     switch (INIT_ENC_INDEX(enc)) {
1523     case UTF_16_ENC:
1524     case UTF_16LE_ENC:
1525     case UTF_16BE_ENC:
1526       return XML_TOK_PARTIAL;
1527     }
1528     switch ((unsigned char)*ptr) {
1529     case 0xFE:
1530     case 0xFF:
1531     case 0xEF: /* possibly first byte of UTF-8 BOM */
1532       if (INIT_ENC_INDEX(enc) == ISO_8859_1_ENC
1533           && state == XML_CONTENT_STATE)
1534         break;
1535       /* fall through */
1536     case 0x00:
1537     case 0x3C:
1538       return XML_TOK_PARTIAL;
1539     }
1540   }
1541   else {
1542     switch (((unsigned char)ptr[0] << 8) | (unsigned char)ptr[1]) {
1543     case 0xFEFF:
1544       if (INIT_ENC_INDEX(enc) == ISO_8859_1_ENC
1545           && state == XML_CONTENT_STATE)
1546         break;
1547       *nextTokPtr = ptr + 2;
1548       *encPtr = encodingTable[UTF_16BE_ENC];
1549       return XML_TOK_BOM;
1550     /* 00 3C is handled in the default case */
1551     case 0x3C00:
1552       if ((INIT_ENC_INDEX(enc) == UTF_16BE_ENC
1553            || INIT_ENC_INDEX(enc) == UTF_16_ENC)
1554           && state == XML_CONTENT_STATE)
1555         break;
1556       *encPtr = encodingTable[UTF_16LE_ENC];
1557       return XmlTok(*encPtr, state, ptr, end, nextTokPtr);
1558     case 0xFFFE:
1559       if (INIT_ENC_INDEX(enc) == ISO_8859_1_ENC
1560           && state == XML_CONTENT_STATE)
1561         break;
1562       *nextTokPtr = ptr + 2;
1563       *encPtr = encodingTable[UTF_16LE_ENC];
1564       return XML_TOK_BOM;
1565     case 0xEFBB:
1566       /* Maybe a UTF-8 BOM (EF BB BF) */
1567       /* If there's an explicitly specified (external) encoding
1568          of ISO-8859-1 or some flavour of UTF-16
1569          and this is an external text entity,
1570          don't look for the BOM,
1571          because it might be a legal data.
1572       */
1573       if (state == XML_CONTENT_STATE) {
1574         int e = INIT_ENC_INDEX(enc);
1575         if (e == ISO_8859_1_ENC || e == UTF_16BE_ENC
1576             || e == UTF_16LE_ENC || e == UTF_16_ENC)
1577           break;
1578       }
1579       if (ptr + 2 == end)
1580         return XML_TOK_PARTIAL;
1581       if ((unsigned char)ptr[2] == 0xBF) {
1582         *nextTokPtr = ptr + 3;
1583         *encPtr = encodingTable[UTF_8_ENC];
1584         return XML_TOK_BOM;
1585       }
1586       break;
1587     default:
1588       if (ptr[0] == '\0') {
1589         /* 0 isn't a legal data character. Furthermore a document
1590            entity can only start with ASCII characters.  So the only
1591            way this can fail to be big-endian UTF-16 if it it's an
1592            external parsed general entity that's labelled as
1593            UTF-16LE.
1594         */
1595         if (state == XML_CONTENT_STATE && INIT_ENC_INDEX(enc) == UTF_16LE_ENC)
1596           break;
1597         *encPtr = encodingTable[UTF_16BE_ENC];
1598         return XmlTok(*encPtr, state, ptr, end, nextTokPtr);
1599       }
1600       else if (ptr[1] == '\0') {
1601         /* We could recover here in the case:
1602             - parsing an external entity
1603             - second byte is 0
1604             - no externally specified encoding
1605             - no encoding declaration
1606            by assuming UTF-16LE.  But we don't, because this would mean when
1607            presented just with a single byte, we couldn't reliably determine
1608            whether we needed further bytes.
1609         */
1610         if (state == XML_CONTENT_STATE)
1611           break;
1612         *encPtr = encodingTable[UTF_16LE_ENC];
1613         return XmlTok(*encPtr, state, ptr, end, nextTokPtr);
1614       }
1615       break;
1616     }
1617   }
1618   *encPtr = encodingTable[INIT_ENC_INDEX(enc)];
1619   return XmlTok(*encPtr, state, ptr, end, nextTokPtr);
1620 }
1621
1622
1623 #define NS(x) x
1624 #define ns(x) x
1625 #define XML_TOK_NS_C
1626 #include "xmltok_ns.c"
1627 #undef XML_TOK_NS_C
1628 #undef NS
1629 #undef ns
1630
1631 #ifdef XML_NS
1632
1633 #define NS(x) x ## NS
1634 #define ns(x) x ## _ns
1635
1636 #define XML_TOK_NS_C
1637 #include "xmltok_ns.c"
1638 #undef XML_TOK_NS_C
1639
1640 #undef NS
1641 #undef ns
1642
1643 ENCODING *
1644 XmlInitUnknownEncodingNS(void *mem,
1645                          int *table,
1646                          CONVERTER convert,
1647                          void *userData)
1648 {
1649   ENCODING *enc = XmlInitUnknownEncoding(mem, table, convert, userData);
1650   if (enc)
1651     ((struct normal_encoding *)enc)->type[ASCII_COLON] = BT_COLON;
1652   return enc;
1653 }
1654
1655 #endif /* XML_NS */