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