/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 202 by ph10, Fri Aug 3 09:44:26 2007 UTC revision 359 by ph10, Wed Jul 9 16:20:19 2008 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2008 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 43  supporting internal functions that are n Line 43  supporting internal functions that are n
43    
44    
45  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
46  #include <config.h>  #include "config.h"
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
# Line 140  static const short int escapes[] = { Line 140  static const short int escapes[] = {
140  #endif  #endif
141    
142    
143  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
144  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
145  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. */
146    
147  static const char *const posix_names[] = {  typedef struct verbitem {
148    "alpha", "lower", "upper",    int   len;
149    "alnum", "ascii", "blank", "cntrl", "digit", "graph",    int   op;
150    "print", "punct", "space", "word",  "xdigit" };  } verbitem;
151    
152    static const char verbnames[] =
153      "ACCEPT\0"
154      "COMMIT\0"
155      "F\0"
156      "FAIL\0"
157      "PRUNE\0"
158      "SKIP\0"
159      "THEN";
160    
161    static const verbitem verbs[] = {
162      { 6, OP_ACCEPT },
163      { 6, OP_COMMIT },
164      { 1, OP_FAIL },
165      { 4, OP_FAIL },
166      { 5, OP_PRUNE },
167      { 4, OP_SKIP  },
168      { 4, OP_THEN  }
169    };
170    
171    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
172    
173    
174    /* Tables of names of POSIX character classes and their lengths. The names are
175    now all in a single string, to reduce the number of relocations when a shared
176    library is dynamically loaded. The list of lengths is terminated by a zero
177    length entry. The first three must be alpha, lower, upper, as this is assumed
178    for handling case independence. */
179    
180    static const char posix_names[] =
181      "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"
182      "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"
183      "word\0"   "xdigit";
184    
185  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
186    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 186  static const int posix_class_maps[] = { Line 219  static const int posix_class_maps[] = {
219  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
220  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
221  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
222  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
223    the number of relocations needed when a shared library is loaded dynamically,
224  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
225    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
226    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
227    "\\c at end of pattern",  because these strings are used only when there is a compilation error. */
228    "unrecognized character follows \\",  
229    "numbers out of order in {} quantifier",  static const char error_texts[] =
230      "no error\0"
231      "\\ at end of pattern\0"
232      "\\c at end of pattern\0"
233      "unrecognized character follows \\\0"
234      "numbers out of order in {} quantifier\0"
235    /* 5 */    /* 5 */
236    "number too big in {} quantifier",    "number too big in {} quantifier\0"
237    "missing terminating ] for character class",    "missing terminating ] for character class\0"
238    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
239    "range out of order in character class",    "range out of order in character class\0"
240    "nothing to repeat",    "nothing to repeat\0"
241    /* 10 */    /* 10 */
242    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
243    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
244    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
245    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
246    "missing )",    "missing )\0"
247    /* 15 */    /* 15 */
248    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
249    "erroffset passed as NULL",    "erroffset passed as NULL\0"
250    "unknown option bit(s) set",    "unknown option bit(s) set\0"
251    "missing ) after comment",    "missing ) after comment\0"
252    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
253    /* 20 */    /* 20 */
254    "regular expression is too large",    "regular expression is too large\0"
255    "failed to get memory",    "failed to get memory\0"
256    "unmatched parentheses",    "unmatched parentheses\0"
257    "internal error: code overflow",    "internal error: code overflow\0"
258    "unrecognized character after (?<",    "unrecognized character after (?<\0"
259    /* 25 */    /* 25 */
260    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
261    "malformed number or name after (?(",    "malformed number or name after (?(\0"
262    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
263    "assertion expected after (?(",    "assertion expected after (?(\0"
264    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
265    /* 30 */    /* 30 */
266    "unknown POSIX class name",    "unknown POSIX class name\0"
267    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
268    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
269    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
270    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
271    /* 35 */    /* 35 */
272    "invalid condition (?(0)",    "invalid condition (?(0)\0"
273    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
274    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
275    "number after (?C is > 255",    "number after (?C is > 255\0"
276    "closing ) for (?C expected",    "closing ) for (?C expected\0"
277    /* 40 */    /* 40 */
278    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
279    "unrecognized character after (?P",    "unrecognized character after (?P\0"
280    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
281    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
282    "invalid UTF-8 string",    "invalid UTF-8 string\0"
283    /* 45 */    /* 45 */
284    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
285    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
286    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
287    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
288    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
289    /* 50 */    /* 50 */
290    "repeated subpattern is too long",    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
291    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
292    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
293    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
294    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
295    /* 55 */    /* 55 */
296    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
297    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
298    "\\g is not followed by a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
299    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"    "a numbered reference must not be zero\0"
300  };    "(*VERB) with an argument is not supported\0"
301      /* 60 */
302      "(*VERB) not recognized\0"
303      "number is too big\0"
304      "subpattern name expected\0"
305      "digit expected after (?+\0"
306      "] is an invalid data character in JavaScript compatibility mode";
307    
308    
309  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 395  static BOOL Line 439  static BOOL
439    
440    
441  /*************************************************  /*************************************************
442    *            Find an error text                  *
443    *************************************************/
444    
445    /* The error texts are now all in one long string, to save on relocations. As
446    some of the text is of unknown length, we can't use a table of offsets.
447    Instead, just count through the strings. This is not a performance issue
448    because it happens only when there has been a compilation error.
449    
450    Argument:   the error number
451    Returns:    pointer to the error string
452    */
453    
454    static const char *
455    find_error_text(int n)
456    {
457    const char *s = error_texts;
458    for (; n > 0; n--) while (*s++ != 0);
459    return s;
460    }
461    
462    
463    /*************************************************
464  *            Handle escapes                      *  *            Handle escapes                      *
465  *************************************************/  *************************************************/
466    
# Line 415  Arguments: Line 481  Arguments:
481    
482  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
483                   negative => a special escape sequence                   negative => a special escape sequence
484                   on error, errorptr is set                   on error, errorcodeptr is set
485  */  */
486    
487  static int  static int
# Line 433  ptr--; /* Set Line 499  ptr--; /* Set
499    
500  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
501    
502  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
503  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
504  Otherwise further processing may be required. */  Otherwise further processing may be required. */
505    
506  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
507  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */
508  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
509    
510  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
511  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
512  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
513  #endif  #endif
514    
# Line 466  else Line 532  else
532      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
533      break;      break;
534    
535      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
536      is an absolute backreference. If negative, it is a relative backreference.  
537      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
538      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
539      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
540      fudge it up by pretending it really was \k. */  
541        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
542        is part of Perl's movement towards a unified syntax for back references. As
543        this is synonymous with \k{name}, we fudge it up by pretending it really
544        was \k.
545    
546        (3) For Oniguruma compatibility we also support \g followed by a name or a
547        number either in angle brackets or in single quotes. However, these are
548        (possibly recursive) subroutine calls, _not_ backreferences. Just return
549        the -ESC_g code (cf \k). */
550    
551      case 'g':      case 'g':
552        if (ptr[1] == '<' || ptr[1] == '\'')
553          {
554          c = -ESC_g;
555          break;
556          }
557    
558        /* Handle the Perl-compatible cases */
559    
560      if (ptr[1] == '{')      if (ptr[1] == '{')
561        {        {
562        const uschar *p;        const uschar *p;
# Line 500  else Line 583  else
583      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
584        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - '0';
585    
586      if (c == 0 || (braced && *(++ptr) != '}'))      if (c < 0)   /* Integer overflow */
587          {
588          *errorcodeptr = ERR61;
589          break;
590          }
591    
592        if (braced && *(++ptr) != '}')
593        {        {
594        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
595        return 0;        break;
596          }
597    
598        if (c == 0)
599          {
600          *errorcodeptr = ERR58;
601          break;
602        }        }
603    
604      if (negated)      if (negated)
# Line 511  else Line 606  else
606        if (c > bracount)        if (c > bracount)
607          {          {
608          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
609          return 0;          break;
610          }          }
611        c = bracount - (c - 1);        c = bracount - (c - 1);
612        }        }
# Line 540  else Line 635  else
635        c -= '0';        c -= '0';
636        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
637          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
638          if (c < 0)    /* Integer overflow */
639            {
640            *errorcodeptr = ERR61;
641            break;
642            }
643        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
644          {          {
645          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 635  else Line 735  else
735      if (c == 0)      if (c == 0)
736        {        {
737        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
738        return 0;        break;
739        }        }
740    
741  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
# Line 648  else Line 748  else
748      break;      break;
749    
750      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
751      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
752      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
753      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
754      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
755    
756      default:      default:
757      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 740  top = _pcre_utt_size; Line 840  top = _pcre_utt_size;
840  while (bot < top)  while (bot < top)
841    {    {
842    i = (bot + top) >> 1;    i = (bot + top) >> 1;
843    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
844    if (c == 0)    if (c == 0)
845      {      {
846      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 876  be terminated by '>' because that is che Line 976  be terminated by '>' because that is che
976    
977  Arguments:  Arguments:
978    ptr          current position in the pattern    ptr          current position in the pattern
979    count        current count of capturing parens so far encountered    cd           compile background data
980    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
981    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
982    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
# Line 885  Returns: the number of the named s Line 985  Returns: the number of the named s
985  */  */
986    
987  static int  static int
988  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens(const uschar *ptr, compile_data *cd, const uschar *name, int lorn,
989    BOOL xmode)    BOOL xmode)
990  {  {
991  const uschar *thisname;  const uschar *thisname;
992    int count = cd->bracount;
993    
994  for (; *ptr != 0; ptr++)  for (; *ptr != 0; ptr++)
995    {    {
# Line 908  for (; *ptr != 0; ptr++) Line 1009  for (; *ptr != 0; ptr++)
1009      continue;      continue;
1010      }      }
1011    
1012    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1013      are handled for real. If the first character is '^', skip it. Also, if the
1014      first few characters (either before or after ^) are \Q\E or \E we skip them
1015      too. This makes for compatibility with Perl. */
1016    
1017    if (*ptr == '[')    if (*ptr == '[')
1018      {      {
1019        BOOL negate_class = FALSE;
1020        for (;;)
1021          {
1022          int c = *(++ptr);
1023          if (c == '\\')
1024            {
1025            if (ptr[1] == 'E') ptr++;
1026              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
1027                else break;
1028            }
1029          else if (!negate_class && c == '^')
1030            negate_class = TRUE;
1031          else break;
1032          }
1033    
1034        /* If the next character is ']', it is a data character that must be
1035        skipped, except in JavaScript compatibility mode. */
1036    
1037        if (ptr[1] == ']' && (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1038          ptr++;
1039    
1040      while (*(++ptr) != ']')      while (*(++ptr) != ']')
1041        {        {
1042          if (*ptr == 0) return -1;
1043        if (*ptr == '\\')        if (*ptr == '\\')
1044          {          {
1045          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) return -1;
# Line 941  for (; *ptr != 0; ptr++) Line 1067  for (; *ptr != 0; ptr++)
1067    /* An opening parens must now be a real metacharacter */    /* An opening parens must now be a real metacharacter */
1068    
1069    if (*ptr != '(') continue;    if (*ptr != '(') continue;
1070    if (ptr[1] != '?')    if (ptr[1] != '?' && ptr[1] != '*')
1071      {      {
1072      count++;      count++;
1073      if (name == NULL && count == lorn) return count;      if (name == NULL && count == lorn) return count;
# Line 1069  for (;;) Line 1195  for (;;)
1195    {    {
1196    int d;    int d;
1197    register int op = *cc;    register int op = *cc;
   
1198    switch (op)    switch (op)
1199      {      {
1200      case OP_CBRA:      case OP_CBRA:
# Line 1158  for (;;) Line 1283  for (;;)
1283    
1284      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1285      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1286        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1287      cc += 4;      cc += 4;
1288      break;      break;
1289    
# Line 1175  for (;;) Line 1301  for (;;)
1301      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1302      case OP_WORDCHAR:      case OP_WORDCHAR:
1303      case OP_ANY:      case OP_ANY:
1304        case OP_ALLANY:
1305      branchlength++;      branchlength++;
1306      cc++;      cc++;
1307      break;      break;
# Line 1266  for (;;) Line 1393  for (;;)
1393      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1394      }      }
1395    
1396    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* Otherwise, we can get the item's length from the table, except that for
1397    a multi-byte character. The length in the table is a minimum, so we have to    repeated character types, we have to test for \p and \P, which have an extra
1398    arrange to skip the extra bytes. */    two bytes of parameters. */
1399    
1400    else    else
1401      {      {
1402        switch(c)
1403          {
1404          case OP_TYPESTAR:
1405          case OP_TYPEMINSTAR:
1406          case OP_TYPEPLUS:
1407          case OP_TYPEMINPLUS:
1408          case OP_TYPEQUERY:
1409          case OP_TYPEMINQUERY:
1410          case OP_TYPEPOSSTAR:
1411          case OP_TYPEPOSPLUS:
1412          case OP_TYPEPOSQUERY:
1413          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1414          break;
1415    
1416          case OP_TYPEUPTO:
1417          case OP_TYPEMINUPTO:
1418          case OP_TYPEEXACT:
1419          case OP_TYPEPOSUPTO:
1420          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1421          break;
1422          }
1423    
1424        /* Add in the fixed length from the table */
1425    
1426      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1427    
1428      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1429      a multi-byte character. The length in the table is a minimum, so we have to
1430      arrange to skip the extra bytes. */
1431    
1432  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1433      if (utf8) switch(c)      if (utf8) switch(c)
1434        {        {
# Line 1330  for (;;) Line 1486  for (;;)
1486    
1487    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1488    
1489    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
1490    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
1491    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1492    
1493    else    else
1494      {      {
1495        switch(c)
1496          {
1497          case OP_TYPESTAR:
1498          case OP_TYPEMINSTAR:
1499          case OP_TYPEPLUS:
1500          case OP_TYPEMINPLUS:
1501          case OP_TYPEQUERY:
1502          case OP_TYPEMINQUERY:
1503          case OP_TYPEPOSSTAR:
1504          case OP_TYPEPOSPLUS:
1505          case OP_TYPEPOSQUERY:
1506          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1507          break;
1508    
1509          case OP_TYPEPOSUPTO:
1510          case OP_TYPEUPTO:
1511          case OP_TYPEMINUPTO:
1512          case OP_TYPEEXACT:
1513          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1514          break;
1515          }
1516    
1517        /* Add in the fixed length from the table */
1518    
1519      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1520    
1521        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1522        by a multi-byte character. The length in the table is a minimum, so we have
1523        to arrange to skip the extra bytes. */
1524    
1525  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1526      if (utf8) switch(c)      if (utf8) switch(c)
1527        {        {
# Line 1374  for (;;) Line 1558  for (;;)
1558  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1559  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
1560  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1561  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
1562  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
1563    bracket whose current branch will already have been scanned.
1564    
1565  Arguments:  Arguments:
1566    code        points to start of search    code        points to start of search
# Line 1397  for (code = first_significant_code(code Line 1582  for (code = first_significant_code(code
1582    
1583    c = *code;    c = *code;
1584    
1585      /* Skip over forward assertions; the other assertions are skipped by
1586      first_significant_code() with a TRUE final argument. */
1587    
1588      if (c == OP_ASSERT)
1589        {
1590        do code += GET(code, 1); while (*code == OP_ALT);
1591        c = *code;
1592        continue;
1593        }
1594    
1595    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1596    
1597    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1598      {      {
1599      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1600      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1409  for (code = first_significant_code(code Line 1604  for (code = first_significant_code(code
1604    
1605    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1606    
1607    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1608      {      {
1609      BOOL empty_branch;      BOOL empty_branch;
1610      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1433  for (code = first_significant_code(code Line 1628  for (code = first_significant_code(code
1628    
1629    switch (c)    switch (c)
1630      {      {
1631      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1632        cannot be represented just by a bit map. This includes negated single
1633        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1634        actual length is stored in the compiled code, so we must update "code"
1635        here. */
1636    
1637  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1638      case OP_XCLASS:      case OP_XCLASS:
1639      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1640      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1641  #endif  #endif
1642    
# Line 1481  for (code = first_significant_code(code Line 1680  for (code = first_significant_code(code
1680      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1681      case OP_WORDCHAR:      case OP_WORDCHAR:
1682      case OP_ANY:      case OP_ANY:
1683        case OP_ALLANY:
1684      case OP_ANYBYTE:      case OP_ANYBYTE:
1685      case OP_CHAR:      case OP_CHAR:
1686      case OP_CHARNC:      case OP_CHARNC:
# Line 1499  for (code = first_significant_code(code Line 1699  for (code = first_significant_code(code
1699      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1700      return FALSE;      return FALSE;
1701    
1702        /* These are going to continue, as they may be empty, but we have to
1703        fudge the length for the \p and \P cases. */
1704    
1705        case OP_TYPESTAR:
1706        case OP_TYPEMINSTAR:
1707        case OP_TYPEPOSSTAR:
1708        case OP_TYPEQUERY:
1709        case OP_TYPEMINQUERY:
1710        case OP_TYPEPOSQUERY:
1711        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1712        break;
1713    
1714        /* Same for these */
1715    
1716        case OP_TYPEUPTO:
1717        case OP_TYPEMINUPTO:
1718        case OP_TYPEPOSUPTO:
1719        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1720        break;
1721    
1722      /* End of branch */      /* End of branch */
1723    
1724      case OP_KET:      case OP_KET:
# Line 1568  return TRUE; Line 1788  return TRUE;
1788  *************************************************/  *************************************************/
1789    
1790  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1791  encountered in a character class. It checks whether this is followed by an  encountered in a character class. It checks whether this is followed by a
1792  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1793  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1794    
1795    Originally, this function only recognized a sequence of letters between the
1796    terminators, but it seems that Perl recognizes any sequence of characters,
1797    though of course unknown POSIX names are subsequently rejected. Perl gives an
1798    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1799    didn't consider this to be a POSIX class. Likewise for [:1234:].
1800    
1801    The problem in trying to be exactly like Perl is in the handling of escapes. We
1802    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1803    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1804    below handles the special case of \], but does not try to do any other escape
1805    processing. This makes it different from Perl for cases such as [:l\ower:]
1806    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1807    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1808    I think.
1809    
1810  Argument:  Arguments:
1811    ptr      pointer to the initial [    ptr      pointer to the initial [
1812    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1813    
1814  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1815  */  */
1816    
1817  static BOOL  static BOOL
1818  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1819  {  {
1820  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1821  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1822  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1823    {    {
1824    *endptr = ptr;    if (*ptr == '\\' && ptr[1] == ']') ptr++; else
1825    return TRUE;      {
1826        if (*ptr == ']') return FALSE;
1827        if (*ptr == terminator && ptr[1] == ']')
1828          {
1829          *endptr = ptr;
1830          return TRUE;
1831          }
1832        }
1833    }    }
1834  return FALSE;  return FALSE;
1835  }  }
# Line 1615  Returns: a value representing the na Line 1854  Returns: a value representing the na
1854  static int  static int
1855  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1856  {  {
1857    const char *pn = posix_names;
1858  register int yield = 0;  register int yield = 0;
1859  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1860    {    {
1861    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1862      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1863      pn += posix_name_lengths[yield] + 1;
1864    yield++;    yield++;
1865    }    }
1866  return -1;  return -1;
# Line 1634  return -1; Line 1875  return -1;
1875  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
1876  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
1877  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
1878  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
1879  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
1880  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
1881  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
1882  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
1883    OP_END.
1884    
1885  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
1886  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1661  adjust_recurse(uschar *group, int adjust Line 1903  adjust_recurse(uschar *group, int adjust
1903    uschar *save_hwm)    uschar *save_hwm)
1904  {  {
1905  uschar *ptr = group;  uschar *ptr = group;
1906    
1907  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1908    {    {
1909    int offset;    int offset;
# Line 1772  get_othercase_range(unsigned int *cptr, Line 2015  get_othercase_range(unsigned int *cptr,
2015  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2016    
2017  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2018    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2019    
2020  if (c > d) return FALSE;  if (c > d) return FALSE;
2021    
# Line 1781  next = othercase + 1; Line 2024  next = othercase + 1;
2024    
2025  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2026    {    {
2027    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2028    next++;    next++;
2029    }    }
2030    
# Line 1909  if (next >= 0) switch(op_code) Line 2152  if (next >= 0) switch(op_code)
2152      unsigned int othercase;      unsigned int othercase;
2153      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2154  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2155      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2156  #else  #else
2157      othercase = NOTACHAR;      othercase = NOTACHAR;
2158  #endif  #endif
# Line 1922  if (next >= 0) switch(op_code) Line 2165  if (next >= 0) switch(op_code)
2165    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2166    
2167    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2168    if (item == next) return TRUE;    if (item == next) return TRUE;
2169    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2170  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1931  if (next >= 0) switch(op_code) Line 2173  if (next >= 0) switch(op_code)
2173      unsigned int othercase;      unsigned int othercase;
2174      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2175  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2176      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2177  #else  #else
2178      othercase = NOTACHAR;      othercase = NOTACHAR;
2179  #endif  #endif
# Line 2185  uschar classbits[32]; Line 2427  uschar classbits[32];
2427  BOOL class_utf8;  BOOL class_utf8;
2428  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2429  uschar *class_utf8data;  uschar *class_utf8data;
2430    uschar *class_utf8data_base;
2431  uschar utf8_char[6];  uschar utf8_char[6];
2432  #else  #else
2433  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2224  req_caseopt = ((options & PCRE_CASELESS) Line 2467  req_caseopt = ((options & PCRE_CASELESS)
2467  for (;; ptr++)  for (;; ptr++)
2468    {    {
2469    BOOL negate_class;    BOOL negate_class;
2470      BOOL should_flip_negation;
2471    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2472    BOOL is_quantifier;    BOOL is_quantifier;
2473    BOOL is_recurse;    BOOL is_recurse;
# Line 2423  for (;; ptr++) Line 2667  for (;; ptr++)
2667      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2668      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2669      previous = code;      previous = code;
2670      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2671      break;      break;
2672    
2673    
# Line 2438  for (;; ptr++) Line 2682  for (;; ptr++)
2682      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
2683      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2684      whether the bitmap is present, and whether this is a negated class or not.      whether the bitmap is present, and whether this is a negated class or not.
2685      */  
2686        In JavaScript compatibility mode, an isolated ']' causes an error. In
2687        default (Perl) mode, it is treated as a data character. */
2688    
2689        case ']':
2690        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2691          {
2692          *errorcodeptr = ERR64;
2693          goto FAILED;
2694          }
2695        goto NORMAL_CHAR;
2696    
2697      case '[':      case '[':
2698      previous = code;      previous = code;
# Line 2447  for (;; ptr++) Line 2701  for (;; ptr++)
2701      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
2702    
2703      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2704          check_posix_syntax(ptr, &tempptr, cd))          check_posix_syntax(ptr, &tempptr))
2705        {        {
2706        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;
2707        goto FAILED;        goto FAILED;
2708        }        }
2709    
2710      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
2711        if the first few characters (either before or after ^) are \Q\E or \E we
2712        skip them too. This makes for compatibility with Perl. */
2713    
2714      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2715        for (;;)
2716        {        {
       negate_class = TRUE;  
2717        c = *(++ptr);        c = *(++ptr);
2718          if (c == '\\')
2719            {
2720            if (ptr[1] == 'E') ptr++;
2721              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2722                else break;
2723            }
2724          else if (!negate_class && c == '^')
2725            negate_class = TRUE;
2726          else break;
2727        }        }
2728      else  
2729        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2730        an initial ']' is taken as a data character -- the code below handles
2731        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2732        [^] must match any character, so generate OP_ALLANY. */
2733    
2734        if (c ==']' && (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2735        {        {
2736        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2737          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2738          zerofirstbyte = firstbyte;
2739          break;
2740        }        }
2741    
2742        /* If a class contains a negative special such as \S, we need to flip the
2743        negation flag at the end, so that support for characters > 255 works
2744        correctly (they are all included in the class). */
2745    
2746        should_flip_negation = FALSE;
2747    
2748      /* Keep a count of chars with values < 256 so that we can optimize the case      /* Keep a count of chars with values < 256 so that we can optimize the case
2749      of just a single character (as long as it's < 256). However, For higher      of just a single character (as long as it's < 256). However, For higher
2750      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2482  for (;; ptr++) Line 2762  for (;; ptr++)
2762  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2763      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2764      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2765        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2766  #endif  #endif
2767    
2768      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2497  for (;; ptr++) Line 2778  for (;; ptr++)
2778          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2779          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2780          }          }
2781    
2782          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2783          data and reset the pointer. This is so that very large classes that
2784          contain a zillion UTF-8 characters no longer overwrite the work space
2785          (which is on the stack). */
2786    
2787          if (lengthptr != NULL)
2788            {
2789            *lengthptr += class_utf8data - class_utf8data_base;
2790            class_utf8data = class_utf8data_base;
2791            }
2792    
2793  #endif  #endif
2794    
2795        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 2520  for (;; ptr++) Line 2813  for (;; ptr++)
2813    
2814        if (c == '[' &&        if (c == '[' &&
2815            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2816            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr))
2817          {          {
2818          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2819          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
# Line 2537  for (;; ptr++) Line 2830  for (;; ptr++)
2830          if (*ptr == '^')          if (*ptr == '^')
2831            {            {
2832            local_negate = TRUE;            local_negate = TRUE;
2833              should_flip_negation = TRUE;  /* Note negative special */
2834            ptr++;            ptr++;
2835            }            }
2836    
# Line 2603  for (;; ptr++) Line 2897  for (;; ptr++)
2897        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
2898        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
2899        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2900        to or into the one we are building. We assume they have more than one        to 'or' into the one we are building. We assume they have more than one
2901        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2902    
2903        if (c == '\\')        if (c == '\\')
# Line 2611  for (;; ptr++) Line 2905  for (;; ptr++)
2905          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2906          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
2907    
2908          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
2909          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */
2910          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2911          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
# Line 2623  for (;; ptr++) Line 2917  for (;; ptr++)
2917            else inescq = TRUE;            else inescq = TRUE;
2918            continue;            continue;
2919            }            }
2920            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2921    
2922          if (c < 0)          if (c < 0)
2923            {            {
# Line 2638  for (;; ptr++) Line 2933  for (;; ptr++)
2933              continue;              continue;
2934    
2935              case ESC_D:              case ESC_D:
2936                should_flip_negation = TRUE;
2937              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2938              continue;              continue;
2939    
# Line 2646  for (;; ptr++) Line 2942  for (;; ptr++)
2942              continue;              continue;
2943    
2944              case ESC_W:              case ESC_W:
2945                should_flip_negation = TRUE;
2946              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2947              continue;              continue;
2948    
# Line 2655  for (;; ptr++) Line 2952  for (;; ptr++)
2952              continue;              continue;
2953    
2954              case ESC_S:              case ESC_S:
2955                should_flip_negation = TRUE;
2956              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2957              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2958              continue;              continue;
2959    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
2960              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
2961              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
2962              }              }
# Line 2849  for (;; ptr++) Line 3144  for (;; ptr++)
3144    
3145        oldptr = ptr;        oldptr = ptr;
3146    
3147          /* Remember \r or \n */
3148    
3149          if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3150    
3151          /* Check for range */
3152    
3153        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == '-')
3154          {          {
3155          int d;          int d;
# Line 2890  for (;; ptr++) Line 3191  for (;; ptr++)
3191            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3192            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3193    
3194            /* \b is backslash; \X is literal X; \R is literal R; any other            /* \b is backspace; \X is literal X; \R is literal R; any other
3195            special means the '-' was literal */            special means the '-' was literal */
3196    
3197            if (d < 0)            if (d < 0)
# Line 2916  for (;; ptr++) Line 3217  for (;; ptr++)
3217    
3218          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3219    
3220            /* Remember \r or \n */
3221    
3222            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3223    
3224          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
3225          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3226          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3034  for (;; ptr++) Line 3339  for (;; ptr++)
3339          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3340            {            {
3341            unsigned int othercase;            unsigned int othercase;
3342            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3343              {              {
3344              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3345              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3069  for (;; ptr++) Line 3374  for (;; ptr++)
3374        goto FAILED;        goto FAILED;
3375        }        }
3376    
3377    
3378    /* This code has been disabled because it would mean that \s counts as
3379    an explicit \r or \n reference, and that's not really what is wanted. Now
3380    we set the flag only if there is a literal "\r" or "\n" in the class. */
3381    
3382    #if 0
3383        /* Remember whether \r or \n are in this class */
3384    
3385        if (negate_class)
3386          {
3387          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3388          }
3389        else
3390          {
3391          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3392          }
3393    #endif
3394    
3395    
3396      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
3397      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
3398      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
3399      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3400      single-bytes only. This is an historical hangover. Maybe one day we can  
3401      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3402        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3403        operate on single-bytes only. This is an historical hangover. Maybe one day
3404        we can tidy these opcodes to handle multi-byte characters.
3405    
3406      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
3407      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 3084  for (;; ptr++) Line 3411  for (;; ptr++)
3411      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3412    
3413  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3414      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3415            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3416  #else  #else
3417      if (class_charcount == 1)      if (class_charcount == 1)
3418  #endif  #endif
# Line 3130  for (;; ptr++) Line 3455  for (;; ptr++)
3455      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3456    
3457      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3458      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3459      we can omit the bitmap in the actual compiled code. */      such as \S in the class, because in that case all characters > 255 are in
3460        the class, so any that were explicitly given as well can be ignored. If
3461        (when there are explicit characters > 255 that must be listed) there are no
3462        characters < 256, we can omit the bitmap in the actual compiled code. */
3463    
3464  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3465      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3466        {        {
3467        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3468        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3160  for (;; ptr++) Line 3488  for (;; ptr++)
3488        }        }
3489  #endif  #endif
3490    
3491      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, set the opcode to OP_CLASS or
3492      and copy it into the code vector. If this is the first thing in the branch,      OP_NCLASS, depending on whether the whole class was negated and whether
3493      there can be no first char setting, whatever the repeat count. Any reqbyte      there were negative specials such as \S in the class. Then copy the 32-byte
3494      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3495    
3496        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3497      if (negate_class)      if (negate_class)
3498        {        {
       *code++ = OP_NCLASS;  
3499        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3500          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3501        }        }
3502      else      else
3503        {        {
       *code++ = OP_CLASS;  
3504        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3505        }        }
3506      code += 32;      code += 32;
# Line 3359  for (;; ptr++) Line 3686  for (;; ptr++)
3686        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3687        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3688    
3689        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3690    
3691        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3692    
# Line 3509  for (;; ptr++) Line 3836  for (;; ptr++)
3836        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3837        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3838    
3839        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3840    
3841        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3842          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3567  for (;; ptr++) Line 3894  for (;; ptr++)
3894    
3895        if (repeat_min == 0)        if (repeat_min == 0)
3896          {          {
3897          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
3898          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
3899    
3900          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
3901          BRAZERO and do no more at this point. However, we do need to adjust          **   {
3902          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
3903          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
3904          the start of the whole regex. Temporarily terminate the pattern while          **   }
3905          doing this. */  
3906            However, that fails when a group is referenced as a subroutine from
3907            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
3908            so that it is skipped on execution. As we don't have a list of which
3909            groups are referenced, we cannot do this selectively.
3910    
3911            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
3912            and do no more at this point. However, we do need to adjust any
3913            OP_RECURSE calls inside the group that refer to the group itself or any
3914            internal or forward referenced group, because the offset is from the
3915            start of the whole regex. Temporarily terminate the pattern while doing
3916            this. */
3917    
3918          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
3919            {            {
3920            *code = OP_END;            *code = OP_END;
3921            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3922            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3923            code++;            code++;
3924              if (repeat_max == 0)
3925                {
3926                *previous++ = OP_SKIPZERO;
3927                goto END_REPEAT;
3928                }
3929            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
3930            }            }
3931    
# Line 3783  for (;; ptr++) Line 4120  for (;; ptr++)
4120          }          }
4121        }        }
4122    
4123        /* If previous is OP_FAIL, it was generated by an empty class [] in
4124        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4125        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4126        error above. We can just ignore the repeat in JS case. */
4127    
4128        else if (*previous == OP_FAIL) goto END_REPEAT;
4129    
4130      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4131    
4132      else      else
# Line 3809  for (;; ptr++) Line 4153  for (;; ptr++)
4153        int len;        int len;
4154        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4155            *tempcode == OP_NOTEXACT)            *tempcode == OP_NOTEXACT)
4156          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode] +
4157              ((*tempcode == OP_TYPEEXACT &&
4158                 (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4159        len = code - tempcode;        len = code - tempcode;
4160        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4161          {          {
# Line 3853  for (;; ptr++) Line 4199  for (;; ptr++)
4199      /* ===================================================================*/      /* ===================================================================*/
4200      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4201      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4202      parenthesis forms. First deal with the specials; all are introduced by ?,      parenthesis forms.  */
     and the appearance of any of them means that this is not a capturing  
     group. */  
4203    
4204      case '(':      case '(':
4205      newoptions = options;      newoptions = options;
# Line 3864  for (;; ptr++) Line 4208  for (;; ptr++)
4208      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4209      reset_bracount = FALSE;      reset_bracount = FALSE;
4210    
4211      if (*(++ptr) == '?')      /* First deal with various "verbs" that can be introduced by '*'. */
4212    
4213        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4214          {
4215          int i, namelen;
4216          const char *vn = verbnames;
4217          const uschar *name = ++ptr;
4218          previous = NULL;
4219          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4220          if (*ptr == ':')
4221            {
4222            *errorcodeptr = ERR59;   /* Not supported */
4223            goto FAILED;
4224            }
4225          if (*ptr != ')')
4226            {
4227            *errorcodeptr = ERR60;
4228            goto FAILED;
4229            }
4230          namelen = ptr - name;
4231          for (i = 0; i < verbcount; i++)
4232            {
4233            if (namelen == verbs[i].len &&
4234                strncmp((char *)name, vn, namelen) == 0)
4235              {
4236              *code = verbs[i].op;
4237              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4238              break;
4239              }
4240            vn += verbs[i].len + 1;
4241            }
4242          if (i < verbcount) continue;
4243          *errorcodeptr = ERR60;
4244          goto FAILED;
4245          }
4246    
4247        /* Deal with the extended parentheses; all are introduced by '?', and the
4248        appearance of any of them means that this is not a capturing group. */
4249    
4250        else if (*ptr == '?')
4251        {        {
4252        int i, set, unset, namelen;        int i, set, unset, namelen;
4253        int *optset;        int *optset;
# Line 3999  for (;; ptr++) Line 4382  for (;; ptr++)
4382              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4383              goto FAILED;              goto FAILED;
4384              }              }
4385            if (refsign == '-')            recno = (refsign == '-')?
4386                cd->bracount - recno + 1 : recno +cd->bracount;
4387              if (recno <= 0 || recno > cd->final_bracount)
4388              {              {
4389              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4390              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4391              }              }
           else recno += cd->bracount;  
4392            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4393            break;            break;
4394            }            }
# Line 4033  for (;; ptr++) Line 4413  for (;; ptr++)
4413    
4414          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4415    
4416          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(ptr, cd, name, namelen,
4417                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4418            {            {
4419            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
# Line 4080  for (;; ptr++) Line 4460  for (;; ptr++)
4460            skipbytes = 1;            skipbytes = 1;
4461            }            }
4462    
4463          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4464            in the second pass here, so final_bracount is set. */
4465    
4466          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4467            {            {
4468            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4469            }            }
# Line 4106  for (;; ptr++) Line 4487  for (;; ptr++)
4487    
4488          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4489          case '!':                 /* Negative lookahead */          case '!':                 /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4490          ptr++;          ptr++;
4491            if (*ptr == ')')          /* Optimize (?!) */
4492              {
4493              *code++ = OP_FAIL;
4494              previous = NULL;
4495              continue;
4496              }
4497            bravalue = OP_ASSERT_NOT;
4498          break;          break;
4499    
4500    
# Line 4270  for (;; ptr++) Line 4657  for (;; ptr++)
4657    
4658          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4659          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4660          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4661            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4662            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4663    
4664          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4665          name = ++ptr;          name = ++ptr;
# Line 4282  for (;; ptr++) Line 4671  for (;; ptr++)
4671    
4672          if (lengthptr != NULL)          if (lengthptr != NULL)
4673            {            {
4674              if (namelen == 0)
4675                {
4676                *errorcodeptr = ERR62;
4677                goto FAILED;
4678                }
4679            if (*ptr != terminator)            if (*ptr != terminator)
4680              {              {
4681              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4295  for (;; ptr++) Line 4689  for (;; ptr++)
4689            recno = 0;            recno = 0;
4690            }            }
4691    
4692          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4693            first, and then check that we have reached the end of the name in the
4694            table. That way, if the name that is longer than any in the table,
4695            the comparison will fail without reading beyond the table entry. */
4696    
4697          else          else
4698            {            {
4699            slot = cd->name_table;            slot = cd->name_table;
4700            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4701              {              {
4702              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4703                    slot[2+namelen] == 0)
4704                  break;
4705              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4706              }              }
4707    
# Line 4311  for (;; ptr++) Line 4710  for (;; ptr++)
4710              recno = GET2(slot, 0);              recno = GET2(slot, 0);
4711              }              }
4712            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
4713                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(ptr, cd, name, namelen,
4714                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
4715              {              {
4716              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4338  for (;; ptr++) Line 4737  for (;; ptr++)
4737          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case '5': case '6': case '7': case '8': case '9':   /* subroutine */
4738            {            {
4739            const uschar *called;            const uschar *called;
4740              terminator = ')';
4741    
4742            if ((refsign = *ptr) == '+') ptr++;            /* Come here from the \g<...> and \g'...' code (Oniguruma
4743              compatibility). However, the syntax has been checked to ensure that
4744              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
4745              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
4746              ever be taken. */
4747    
4748              HANDLE_NUMERICAL_RECURSION:
4749    
4750              if ((refsign = *ptr) == '+')
4751                {
4752                ptr++;
4753                if ((digitab[*ptr] & ctype_digit) == 0)
4754                  {
4755                  *errorcodeptr = ERR63;
4756                  goto FAILED;
4757                  }
4758                }
4759            else if (refsign == '-')            else if (refsign == '-')
4760              {              {
4761              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
# Line 4351  for (;; ptr++) Line 4767  for (;; ptr++)
4767            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4768              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - '0';
4769    
4770            if (*ptr != ')')            if (*ptr != terminator)
4771              {              {
4772              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
4773              goto FAILED;              goto FAILED;
# Line 4404  for (;; ptr++) Line 4820  for (;; ptr++)
4820    
4821              if (called == NULL)              if (called == NULL)
4822                {                {
4823                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(ptr, cd, NULL, recno,
4824                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
4825                  {                  {
4826                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
4827                  goto FAILED;                  goto FAILED;
# Line 4465  for (;; ptr++) Line 4881  for (;; ptr++)
4881    
4882              case 'J':    /* Record that it changed in the external options */              case 'J':    /* Record that it changed in the external options */
4883              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
4884              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
4885              break;              break;
4886    
4887              case 'i': *optset |= PCRE_CASELESS; break;              case 'i': *optset |= PCRE_CASELESS; break;
# Line 4656  for (;; ptr++) Line 5072  for (;; ptr++)
5072        goto FAILED;        goto FAILED;
5073        }        }
5074    
5075      /* In the pre-compile phase, update the length by the length of the nested      /* In the pre-compile phase, update the length by the length of the group,
5076      group, less the brackets at either end. Then reduce the compiled code to      less the brackets at either end. Then reduce the compiled code to just a
5077      just the brackets so that it doesn't use much memory if it is duplicated by      set of non-capturing brackets so that it doesn't use much memory if it is
5078      a quantifier. */      duplicated by a quantifier.*/
5079    
5080      if (lengthptr != NULL)      if (lengthptr != NULL)
5081        {        {
# Line 4669  for (;; ptr++) Line 5085  for (;; ptr++)
5085          goto FAILED;          goto FAILED;
5086          }          }
5087        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5088        code++;        *code++ = OP_BRA;
5089        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5090        *code++ = OP_KET;        *code++ = OP_KET;
5091        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5092          break;    /* No need to waste time with special character handling */
5093        }        }
5094    
5095      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
5096    
5097      else code = tempcode;      code = tempcode;
5098    
5099      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
5100      relevant. */      relevant. */
# Line 4774  for (;; ptr++) Line 5191  for (;; ptr++)
5191        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5192        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5193    
5194          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5195          is a subroutine call by number (Oniguruma syntax). In fact, the value
5196          -ESC_g is returned only for these cases. So we don't need to check for <
5197          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5198          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5199          that is a synonym for a named back reference). */
5200    
5201          if (-c == ESC_g)
5202            {
5203            const uschar *p;
5204            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5205            terminator = (*(++ptr) == '<')? '>' : '\'';
5206    
5207            /* These two statements stop the compiler for warning about possibly
5208            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5209            fact, because we actually check for a number below, the paths that
5210            would actually be in error are never taken. */
5211    
5212            skipbytes = 0;
5213            reset_bracount = FALSE;
5214    
5215            /* Test for a name */
5216    
5217            if (ptr[1] != '+' && ptr[1] != '-')
5218              {
5219              BOOL isnumber = TRUE;
5220              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5221                {
5222                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5223                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5224                }
5225              if (*p != terminator)
5226                {
5227                *errorcodeptr = ERR57;
5228                break;
5229                }
5230              if (isnumber)
5231                {
5232                ptr++;
5233                goto HANDLE_NUMERICAL_RECURSION;
5234                }
5235              is_recurse = TRUE;
5236              goto NAMED_REF_OR_RECURSE;
5237              }
5238    
5239            /* Test a signed number in angle brackets or quotes. */
5240    
5241            p = ptr + 2;
5242            while ((digitab[*p] & ctype_digit) != 0) p++;
5243            if (*p != terminator)
5244              {
5245              *errorcodeptr = ERR57;
5246              break;
5247              }
5248            ptr++;
5249            goto HANDLE_NUMERICAL_RECURSION;
5250            }
5251    
5252        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
5253        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5254    
# Line 4881  for (;; ptr++) Line 5356  for (;; ptr++)
5356      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5357      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5358    
5359        /* Remember if \r or \n were seen */
5360    
5361        if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')
5362          cd->external_flags |= PCRE_HASCRORLF;
5363    
5364      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5365      byte, set it from this character, but revert to none on a zero repeat.      byte, set it from this character, but revert to none on a zero repeat.
5366      Otherwise, leave the firstbyte value alone, and don't change it on a zero      Otherwise, leave the firstbyte value alone, and don't change it on a zero
# Line 5275  do { Line 5755  do {
5755       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
5756       }       }
5757    
5758     /* .* is not anchored unless DOTALL is set and it isn't in brackets that     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
5759     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
5760    
5761     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
5762               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
5763       {       {
5764       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
5765           return FALSE;
5766       }       }
5767    
5768     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5464  Returns: pointer to compiled data Line 5944  Returns: pointer to compiled data
5944                  with errorptr and erroroffset set                  with errorptr and erroroffset set
5945  */  */
5946    
5947  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
5948  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
5949    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
5950  {  {
# Line 5472  return pcre_compile2(pattern, options, N Line 5952  return pcre_compile2(pattern, options, N
5952  }  }
5953    
5954    
5955  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
5956  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
5957    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
5958  {  {
# Line 5480  real_pcre *re; Line 5960  real_pcre *re;
5960  int length = 1;  /* For final END opcode */  int length = 1;  /* For final END opcode */
5961  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
5962  int errorcode = 0;  int errorcode = 0;
5963    int skipatstart = 0;
5964  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5965  BOOL utf8;  BOOL utf8;
5966  #endif  #endif
# Line 5498  to fill in forward references to subpatt Line 5979  to fill in forward references to subpatt
5979    
5980  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
5981    
   
5982  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
5983    
5984  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;
# Line 5558  cd->fcc = tables + fcc_offset; Line 6038  cd->fcc = tables + fcc_offset;
6038  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6039  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6040    
6041    /* Check for global one-time settings at the start of the pattern, and remember
6042    the offset for later. */
6043    
6044    while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')
6045      {
6046      int newnl = 0;
6047      int newbsr = 0;
6048    
6049      if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)
6050        { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6051      else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)
6052        { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6053      else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)
6054        { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6055      else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)
6056        { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6057      else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)
6058        { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6059    
6060      else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)
6061        { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6062      else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)
6063        { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6064    
6065      if (newnl != 0)
6066        options = (options & ~PCRE_NEWLINE_BITS) | newnl;
6067      else if (newbsr != 0)
6068        options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
6069      else break;
6070      }
6071    
6072    /* Check validity of \R options. */
6073    
6074    switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6075      {
6076      case 0:
6077      case PCRE_BSR_ANYCRLF:
6078      case PCRE_BSR_UNICODE:
6079      break;
6080      default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
6081      }
6082    
6083  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
6084  current code allows for fixed one- or two-byte sequences, plus "any" and  current code allows for fixed one- or two-byte sequences, plus "any" and
6085  "anycrlf". */  "anycrlf". */
6086    
6087  switch (options & (PCRE_NEWLINE_CRLF | PCRE_NEWLINE_ANY))  switch (options & PCRE_NEWLINE_BITS)
6088    {    {
6089    case 0: newline = NEWLINE; break;   /* Compile-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6090    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = '\r'; break;
6091    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = '\n'; break;
6092    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
# Line 5617  to compile parts of the pattern into; th Line 6139  to compile parts of the pattern into; th
6139  no longer needed, so hopefully this workspace will never overflow, though there  no longer needed, so hopefully this workspace will never overflow, though there
6140  is a test for its doing so. */  is a test for its doing so. */
6141    
6142  cd->bracount = 0;  cd->bracount = cd->final_bracount = 0;
6143  cd->names_found = 0;  cd->names_found = 0;
6144  cd->name_entry_size = 0;  cd->name_entry_size = 0;
6145  cd->name_table = NULL;  cd->name_table = NULL;
# Line 5627  cd->hwm = cworkspace; Line 6149  cd->hwm = cworkspace;
6149  cd->start_pattern = (const uschar *)pattern;  cd->start_pattern = (const uschar *)pattern;
6150  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));
6151  cd->req_varyopt = 0;  cd->req_varyopt = 0;
 cd->nopartial = FALSE;  
6152  cd->external_options = options;  cd->external_options = options;
6153    cd->external_flags = 0;
6154    
6155  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we
6156  don't need to look at the result of the function here. The initial options have  don't need to look at the result of the function here. The initial options have
# Line 5636  been put into the cd block so that they Line 6158  been put into the cd block so that they
6158  found within the regex right at the beginning. Bringing initial option settings  found within the regex right at the beginning. Bringing initial option settings
6159  outside can help speed up starting point checks. */  outside can help speed up starting point checks. */
6160    
6161    ptr += skipatstart;
6162  code = cworkspace;  code = cworkspace;
6163  *code = OP_BRA;  *code = OP_BRA;
6164  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,
# Line 5666  if (re == NULL) Line 6189  if (re == NULL)
6189    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
6190    }    }
6191    
6192  /* Put in the magic number, and save the sizes, initial options, and character  /* Put in the magic number, and save the sizes, initial options, internal
6193  table pointer. NULL is used for the default character tables. The nullpad field  flags, and character table pointer. NULL is used for the default character
6194  is at the end; it's there to help in the case when a regex compiled on a system  tables. The nullpad field is at the end; it's there to help in the case when a
6195  with 4-byte pointers is run on another with 8-byte pointers. */  regex compiled on a system with 4-byte pointers is run on another with 8-byte
6196    pointers. */
6197    
6198  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
6199  re->size = size;  re->size = size;
6200  re->options = cd->external_options;  re->options = cd->external_options;
6201    re->flags = cd->external_flags;
6202  re->dummy1 = 0;  re->dummy1 = 0;
6203  re->first_byte = 0;  re->first_byte = 0;
6204  re->req_byte = 0;  re->req_byte = 0;
# Line 5691  field. Reset the bracket count and the n Line 6216  field. Reset the bracket count and the n
6216  field; this time it's used for remembering forward references to subpatterns.  field; this time it's used for remembering forward references to subpatterns.
6217  */  */
6218    
6219    cd->final_bracount = cd->bracount;  /* Save for checking forward references */
6220  cd->bracount = 0;  cd->bracount = 0;
6221  cd->names_found = 0;  cd->names_found = 0;
6222  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 5698  codestart = cd->name_table + re->name_en Line 6224  codestart = cd->name_table + re->name_en
6224  cd->start_code = codestart;  cd->start_code = codestart;
6225  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6226  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6227  cd->nopartial = FALSE;  cd->had_accept = FALSE;
6228    
6229  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6230  error, errorcode will be set non-zero, so we don't need to look at the result  error, errorcode will be set non-zero, so we don't need to look at the result
6231  of the function here. */  of the function here. */
6232    
6233  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern + skipatstart;
6234  code = (uschar *)codestart;  code = (uschar *)codestart;
6235  *code = OP_BRA;  *code = OP_BRA;
6236  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,
6237    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);
6238  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
6239  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
6240    re->flags = cd->external_flags;
6241    
6242  if (cd->nopartial) re->options |= PCRE_NOPARTIAL;  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */
6243    
6244  /* If not reached end of pattern on success, there's an excess bracket. */  /* If not reached end of pattern on success, there's an excess bracket. */
6245    
# Line 5754  if (errorcode != 0) Line 6281  if (errorcode != 0)
6281    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
6282    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = ptr - (const uschar *)pattern;
6283    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
6284    *errorptr = error_texts[errorcode];    *errorptr = find_error_text(errorcode);
6285    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
6286    return NULL;    return NULL;
6287    }    }
# Line 5783  if ((re->options & PCRE_ANCHORED) == 0) Line 6310  if ((re->options & PCRE_ANCHORED) == 0)
6310        int ch = firstbyte & 255;        int ch = firstbyte & 255;
6311        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&
6312           cd->fcc[ch] == ch)? ch : firstbyte;           cd->fcc[ch] == ch)? ch : firstbyte;
6313        re->options |= PCRE_FIRSTSET;        re->flags |= PCRE_FIRSTSET;
6314        }        }
6315      else if (is_startline(codestart, 0, cd->backref_map))      else if (is_startline(codestart, 0, cd->backref_map))
6316        re->options |= PCRE_STARTLINE;        re->flags |= PCRE_STARTLINE;
6317      }      }
6318    }    }
6319    
# Line 5800  if (reqbyte >= 0 && Line 6327  if (reqbyte >= 0 &&
6327    int ch = reqbyte & 255;    int ch = reqbyte & 255;
6328    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&
6329      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;
6330    re->options |= PCRE_REQCHSET;    re->flags |= PCRE_REQCHSET;
6331    }    }
6332    
6333  /* Print out the compiled data if debugging is enabled. This is never the  /* Print out the compiled data if debugging is enabled. This is never the
# Line 5811  case when building a production library. Line 6338  case when building a production library.
6338  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6339    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6340    
6341  if (re->options != 0)  printf("Options=%08x\n", re->options);
   {  
   printf("%s%s%s%s%s%s%s%s%s\n",  
     ((re->options & PCRE_NOPARTIAL) != 0)? "nopartial " : "",  
     ((re->options & PCRE_ANCHORED) != 0)? "anchored " : "",  
     ((re->options & PCRE_CASELESS) != 0)? "caseless " : "",  
     ((re->options & PCRE_EXTENDED) != 0)? "extended " : "",  
     ((re->options & PCRE_MULTILINE) != 0)? "multiline " : "",  
     ((re->options & PCRE_DOTALL) != 0)? "dotall " : "",  
     ((re->options & PCRE_DOLLAR_ENDONLY) != 0)? "endonly " : "",  
     ((re->options & PCRE_EXTRA) != 0)? "extra " : "",  
     ((re->options & PCRE_UNGREEDY) != 0)? "ungreedy " : "");  
   }  
6342    
6343  if ((re->options & PCRE_FIRSTSET) != 0)  if ((re->flags & PCRE_FIRSTSET) != 0)
6344    {    {
6345    int ch = re->first_byte & 255;    int ch = re->first_byte & 255;
6346    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?
# Line 5834  if ((re->options & PCRE_FIRSTSET) != 0) Line 6349  if ((re->options & PCRE_FIRSTSET) != 0)
6349      else printf("First char = \\x%02x%s\n", ch, caseless);      else printf("First char = \\x%02x%s\n", ch, caseless);
6350    }    }
6351    
6352  if ((re->options & PCRE_REQCHSET) != 0)  if ((re->flags & PCRE_REQCHSET) != 0)
6353    {    {
6354    int ch = re->req_byte & 255;    int ch = re->req_byte & 255;
6355    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?
# Line 5851  was compiled can be seen. */ Line 6366  was compiled can be seen. */
6366  if (code - codestart > length)  if (code - codestart > length)
6367    {    {
6368    (pcre_free)(re);    (pcre_free)(re);
6369    *errorptr = error_texts[ERR23];    *errorptr = find_error_text(ERR23);
6370    *erroroffset = ptr - (uschar *)pattern;    *erroroffset = ptr - (uschar *)pattern;
6371    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6372    return NULL;    return NULL;

Legend:
Removed from v.202  
changed lines
  Added in v.359

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12