/[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 199 by ph10, Tue Jul 31 14:39:09 2007 UTC revision 335 by ph10, Sat Apr 12 14:36:14 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 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
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 */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
# Line 65  used by pcretest. DEBUG is not defined w Line 65  used by pcretest. DEBUG is not defined w
65    
66  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 133  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 179  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 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",    "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 (?+";
306    
307    
308  /* 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 388  static BOOL Line 438  static BOOL
438    
439    
440  /*************************************************  /*************************************************
441    *            Find an error text                  *
442    *************************************************/
443    
444    /* The error texts are now all in one long string, to save on relocations. As
445    some of the text is of unknown length, we can't use a table of offsets.
446    Instead, just count through the strings. This is not a performance issue
447    because it happens only when there has been a compilation error.
448    
449    Argument:   the error number
450    Returns:    pointer to the error string
451    */
452    
453    static const char *
454    find_error_text(int n)
455    {
456    const char *s = error_texts;
457    for (; n > 0; n--) while (*s++ != 0);
458    return s;
459    }
460    
461    
462    /*************************************************
463  *            Handle escapes                      *  *            Handle escapes                      *
464  *************************************************/  *************************************************/
465    
# Line 408  Arguments: Line 480  Arguments:
480    
481  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
482                   negative => a special escape sequence                   negative => a special escape sequence
483                   on error, errorptr is set                   on error, errorcodeptr is set
484  */  */
485    
486  static int  static int
# Line 426  ptr--;                            /* Set Line 498  ptr--;                            /* Set
498    
499  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
500    
501  /* 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
502  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.
503  Otherwise further processing may be required. */  Otherwise further processing may be required. */
504    
505  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
506  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */
507  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
508    
509  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
510  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
511  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
512  #endif  #endif
513    
# Line 459  else Line 531  else
531      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
532      break;      break;
533    
534      /* \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:
535      is an absolute backreference. If negative, it is a relative backreference.  
536      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
537      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
538      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
539      fudge it up by pretending it really was \k. */  
540        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
541        is part of Perl's movement towards a unified syntax for back references. As
542        this is synonymous with \k{name}, we fudge it up by pretending it really
543        was \k.
544    
545        (3) For Oniguruma compatibility we also support \g followed by a name or a
546        number either in angle brackets or in single quotes. However, these are
547        (possibly recursive) subroutine calls, _not_ backreferences. Just return
548        the -ESC_g code (cf \k). */
549    
550      case 'g':      case 'g':
551        if (ptr[1] == '<' || ptr[1] == '\'')
552          {
553          c = -ESC_g;
554          break;
555          }
556    
557        /* Handle the Perl-compatible cases */
558    
559      if (ptr[1] == '{')      if (ptr[1] == '{')
560        {        {
561        const uschar *p;        const uschar *p;
# Line 493  else Line 582  else
582      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
583        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - '0';
584    
585      if (c == 0 || (braced && *(++ptr) != '}'))      if (c < 0)   /* Integer overflow */
586          {
587          *errorcodeptr = ERR61;
588          break;
589          }
590    
591        if (braced && *(++ptr) != '}')
592        {        {
593        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
594        return 0;        break;
595        }        }
596    
597        if (c == 0)
598          {
599          *errorcodeptr = ERR58;
600          break;
601          }
602    
603      if (negated)      if (negated)
604        {        {
605        if (c > bracount)        if (c > bracount)
606          {          {
607          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
608          return 0;          break;
609          }          }
610        c = bracount - (c - 1);        c = bracount - (c - 1);
611        }        }
# Line 533  else Line 634  else
634        c -= '0';        c -= '0';
635        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
636          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
637          if (c < 0)    /* Integer overflow */
638            {
639            *errorcodeptr = ERR61;
640            break;
641            }
642        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
643          {          {
644          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 628  else Line 734  else
734      if (c == 0)      if (c == 0)
735        {        {
736        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
737        return 0;        break;
738        }        }
739    
740  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
# Line 641  else Line 747  else
747      break;      break;
748    
749      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
750      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
751      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
752      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
753      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
754    
755      default:      default:
756      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 733  top = _pcre_utt_size; Line 839  top = _pcre_utt_size;
839  while (bot < top)  while (bot < top)
840    {    {
841    i = (bot + top) >> 1;    i = (bot + top) >> 1;
842    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
843    if (c == 0)    if (c == 0)
844      {      {
845      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 907  for (; *ptr != 0; ptr++) Line 1013  for (; *ptr != 0; ptr++)
1013      {      {
1014      while (*(++ptr) != ']')      while (*(++ptr) != ']')
1015        {        {
1016          if (*ptr == 0) return -1;
1017        if (*ptr == '\\')        if (*ptr == '\\')
1018          {          {
1019          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) return -1;
# Line 934  for (; *ptr != 0; ptr++) Line 1041  for (; *ptr != 0; ptr++)
1041    /* An opening parens must now be a real metacharacter */    /* An opening parens must now be a real metacharacter */
1042    
1043    if (*ptr != '(') continue;    if (*ptr != '(') continue;
1044    if (ptr[1] != '?')    if (ptr[1] != '?' && ptr[1] != '*')
1045      {      {
1046      count++;      count++;
1047      if (name == NULL && count == lorn) return count;      if (name == NULL && count == lorn) return count;
# Line 1062  for (;;) Line 1169  for (;;)
1169    {    {
1170    int d;    int d;
1171    register int op = *cc;    register int op = *cc;
   
1172    switch (op)    switch (op)
1173      {      {
1174      case OP_CBRA:      case OP_CBRA:
# Line 1151  for (;;) Line 1257  for (;;)
1257    
1258      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1259      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1260        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1261      cc += 4;      cc += 4;
1262      break;      break;
1263    
# Line 1259  for (;;) Line 1366  for (;;)
1366      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1367      }      }
1368    
1369    /* 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
1370    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
1371    arrange to skip the extra bytes. */    two bytes of parameters. */
1372    
1373    else    else
1374      {      {
1375        switch(c)
1376          {
1377          case OP_TYPESTAR:
1378          case OP_TYPEMINSTAR:
1379          case OP_TYPEPLUS:
1380          case OP_TYPEMINPLUS:
1381          case OP_TYPEQUERY:
1382          case OP_TYPEMINQUERY:
1383          case OP_TYPEPOSSTAR:
1384          case OP_TYPEPOSPLUS:
1385          case OP_TYPEPOSQUERY:
1386          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1387          break;
1388    
1389          case OP_TYPEUPTO:
1390          case OP_TYPEMINUPTO:
1391          case OP_TYPEEXACT:
1392          case OP_TYPEPOSUPTO:
1393          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1394          break;
1395          }
1396    
1397        /* Add in the fixed length from the table */
1398    
1399      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1400    
1401      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1402      a multi-byte character. The length in the table is a minimum, so we have to
1403      arrange to skip the extra bytes. */
1404    
1405  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1406      if (utf8) switch(c)      if (utf8) switch(c)
1407        {        {
# Line 1323  for (;;) Line 1459  for (;;)
1459    
1460    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1461    
1462    /* 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
1463    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
1464    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1465    
1466    else    else
1467      {      {
1468        switch(c)
1469          {
1470          case OP_TYPESTAR:
1471          case OP_TYPEMINSTAR:
1472          case OP_TYPEPLUS:
1473          case OP_TYPEMINPLUS:
1474          case OP_TYPEQUERY:
1475          case OP_TYPEMINQUERY:
1476          case OP_TYPEPOSSTAR:
1477          case OP_TYPEPOSPLUS:
1478          case OP_TYPEPOSQUERY:
1479          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1480          break;
1481    
1482          case OP_TYPEPOSUPTO:
1483          case OP_TYPEUPTO:
1484          case OP_TYPEMINUPTO:
1485          case OP_TYPEEXACT:
1486          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1487          break;
1488          }
1489    
1490        /* Add in the fixed length from the table */
1491    
1492      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1493    
1494        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1495        by a multi-byte character. The length in the table is a minimum, so we have
1496        to arrange to skip the extra bytes. */
1497    
1498  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1499      if (utf8) switch(c)      if (utf8) switch(c)
1500        {        {
# Line 1367  for (;;) Line 1531  for (;;)
1531  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()
1532  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
1533  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1534  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
1535  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
1536    bracket whose current branch will already have been scanned.
1537    
1538  Arguments:  Arguments:
1539    code        points to start of search    code        points to start of search
# Line 1390  for (code = first_significant_code(code Line 1555  for (code = first_significant_code(code
1555    
1556    c = *code;    c = *code;
1557    
1558      /* Skip over forward assertions; the other assertions are skipped by
1559      first_significant_code() with a TRUE final argument. */
1560    
1561      if (c == OP_ASSERT)
1562        {
1563        do code += GET(code, 1); while (*code == OP_ALT);
1564        c = *code;
1565        continue;
1566        }
1567    
1568    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1569    
1570    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1571      {      {
1572      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1573      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1402  for (code = first_significant_code(code Line 1577  for (code = first_significant_code(code
1577    
1578    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1579    
1580    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1581      {      {
1582      BOOL empty_branch;      BOOL empty_branch;
1583      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1426  for (code = first_significant_code(code Line 1601  for (code = first_significant_code(code
1601    
1602    switch (c)    switch (c)
1603      {      {
1604      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1605        cannot be represented just by a bit map. This includes negated single
1606        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1607        actual length is stored in the compiled code, so we must update "code"
1608        here. */
1609    
1610  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1611      case OP_XCLASS:      case OP_XCLASS:
1612      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1613      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1614  #endif  #endif
1615    
# Line 1492  for (code = first_significant_code(code Line 1671  for (code = first_significant_code(code
1671      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1672      return FALSE;      return FALSE;
1673    
1674        /* These are going to continue, as they may be empty, but we have to
1675        fudge the length for the \p and \P cases. */
1676    
1677        case OP_TYPESTAR:
1678        case OP_TYPEMINSTAR:
1679        case OP_TYPEPOSSTAR:
1680        case OP_TYPEQUERY:
1681        case OP_TYPEMINQUERY:
1682        case OP_TYPEPOSQUERY:
1683        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1684        break;
1685    
1686        /* Same for these */
1687    
1688        case OP_TYPEUPTO:
1689        case OP_TYPEMINUPTO:
1690        case OP_TYPEPOSUPTO:
1691        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1692        break;
1693    
1694      /* End of branch */      /* End of branch */
1695    
1696      case OP_KET:      case OP_KET:
# Line 1561  return TRUE; Line 1760  return TRUE;
1760  *************************************************/  *************************************************/
1761    
1762  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1763  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
1764  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1765  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1766    
1767    Originally, this function only recognized a sequence of letters between the
1768    terminators, but it seems that Perl recognizes any sequence of characters,
1769    though of course unknown POSIX names are subsequently rejected. Perl gives an
1770    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1771    didn't consider this to be a POSIX class. Likewise for [:1234:].
1772    
1773    The problem in trying to be exactly like Perl is in the handling of escapes. We
1774    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1775    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1776    below handles the special case of \], but does not try to do any other escape
1777    processing. This makes it different from Perl for cases such as [:l\ower:]
1778    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1779    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1780    I think.
1781    
1782  Argument:  Arguments:
1783    ptr      pointer to the initial [    ptr      pointer to the initial [
1784    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1785    
1786  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1787  */  */
1788    
1789  static BOOL  static BOOL
1790  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1791  {  {
1792  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1793  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1794  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1795    {    {
1796    *endptr = ptr;    if (*ptr == '\\' && ptr[1] == ']') ptr++; else
1797    return TRUE;      {
1798        if (*ptr == ']') return FALSE;
1799        if (*ptr == terminator && ptr[1] == ']')
1800          {
1801          *endptr = ptr;
1802          return TRUE;
1803          }
1804        }
1805    }    }
1806  return FALSE;  return FALSE;
1807  }  }
# Line 1608  Returns:     a value representing the na Line 1826  Returns:     a value representing the na
1826  static int  static int
1827  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1828  {  {
1829    const char *pn = posix_names;
1830  register int yield = 0;  register int yield = 0;
1831  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1832    {    {
1833    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1834      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1835      pn += posix_name_lengths[yield] + 1;
1836    yield++;    yield++;
1837    }    }
1838  return -1;  return -1;
# Line 1627  return -1; Line 1847  return -1;
1847  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
1848  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
1849  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
1850  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
1851  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
1852  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
1853  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
1854  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
1855    OP_END.
1856    
1857  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
1858  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 1654  adjust_recurse(uschar *group, int adjust Line 1875  adjust_recurse(uschar *group, int adjust
1875    uschar *save_hwm)    uschar *save_hwm)
1876  {  {
1877  uschar *ptr = group;  uschar *ptr = group;
1878    
1879  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1880    {    {
1881    int offset;    int offset;
# Line 1661  while ((ptr = (uschar *)find_recurse(ptr Line 1883  while ((ptr = (uschar *)find_recurse(ptr
1883    
1884    /* See if this recursion is on the forward reference list. If so, adjust the    /* See if this recursion is on the forward reference list. If so, adjust the
1885    reference. */    reference. */
1886    
1887    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1888      {      {
1889      offset = GET(hc, 0);      offset = GET(hc, 0);
# Line 1915  if (next >= 0) switch(op_code) Line 2137  if (next >= 0) switch(op_code)
2137    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2138    
2139    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2140    if (item == next) return TRUE;    if (item == next) return TRUE;
2141    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2142  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2178  uschar classbits[32]; Line 2399  uschar classbits[32];
2399  BOOL class_utf8;  BOOL class_utf8;
2400  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2401  uschar *class_utf8data;  uschar *class_utf8data;
2402    uschar *class_utf8data_base;
2403  uschar utf8_char[6];  uschar utf8_char[6];
2404  #else  #else
2405  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2217  req_caseopt = ((options & PCRE_CASELESS) Line 2439  req_caseopt = ((options & PCRE_CASELESS)
2439  for (;; ptr++)  for (;; ptr++)
2440    {    {
2441    BOOL negate_class;    BOOL negate_class;
2442      BOOL should_flip_negation;
2443    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2444    BOOL is_quantifier;    BOOL is_quantifier;
2445    BOOL is_recurse;    BOOL is_recurse;
# Line 2236  for (;; ptr++) Line 2459  for (;; ptr++)
2459    /* Get next byte in the pattern */    /* Get next byte in the pattern */
2460    
2461    c = *ptr;    c = *ptr;
2462    
2463    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
2464    previous cycle of this loop. */    previous cycle of this loop. */
2465    
# Line 2258  for (;; ptr++) Line 2481  for (;; ptr++)
2481      */      */
2482    
2483      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2484    
2485        /* Paranoid check for integer overflow */
2486    
2487        if (OFLOW_MAX - *lengthptr < code - last_code)
2488          {
2489          *errorcodeptr = ERR20;
2490          goto FAILED;
2491          }
2492    
2493      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2494      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2495    
# Line 2370  for (;; ptr++) Line 2602  for (;; ptr++)
2602      *ptrptr = ptr;      *ptrptr = ptr;
2603      if (lengthptr != NULL)      if (lengthptr != NULL)
2604        {        {
2605          if (OFLOW_MAX - *lengthptr < code - last_code)
2606            {
2607            *errorcodeptr = ERR20;
2608            goto FAILED;
2609            }
2610        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2611        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2612        }        }
# Line 2426  for (;; ptr++) Line 2663  for (;; ptr++)
2663      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. */
2664    
2665      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2666          check_posix_syntax(ptr, &tempptr, cd))          check_posix_syntax(ptr, &tempptr))
2667        {        {
2668        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;
2669        goto FAILED;        goto FAILED;
2670        }        }
2671    
2672      /* 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,
2673        if the first few characters (either before or after ^) are \Q\E or \E we
2674        skip them too. This makes for compatibility with Perl. */
2675    
2676      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2677        for (;;)
2678        {        {
       negate_class = TRUE;  
2679        c = *(++ptr);        c = *(++ptr);
2680          if (c == '\\')
2681            {
2682            if (ptr[1] == 'E') ptr++;
2683              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2684                else break;
2685            }
2686          else if (!negate_class && c == '^')
2687            negate_class = TRUE;
2688          else break;
2689        }        }
2690      else  
2691        {      /* If a class contains a negative special such as \S, we need to flip the
2692        negate_class = FALSE;      negation flag at the end, so that support for characters > 255 works
2693        }      correctly (they are all included in the class). */
2694    
2695        should_flip_negation = FALSE;
2696    
2697      /* 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
2698      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
# Line 2461  for (;; ptr++) Line 2711  for (;; ptr++)
2711  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2712      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2713      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2714        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2715  #endif  #endif
2716    
2717      /* 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 2476  for (;; ptr++) Line 2727  for (;; ptr++)
2727          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2728          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2729          }          }
2730    
2731          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2732          data and reset the pointer. This is so that very large classes that
2733          contain a zillion UTF-8 characters no longer overwrite the work space
2734          (which is on the stack). */
2735    
2736          if (lengthptr != NULL)
2737            {
2738            *lengthptr += class_utf8data - class_utf8data_base;
2739            class_utf8data = class_utf8data_base;
2740            }
2741    
2742  #endif  #endif
2743    
2744        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 2499  for (;; ptr++) Line 2762  for (;; ptr++)
2762    
2763        if (c == '[' &&        if (c == '[' &&
2764            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2765            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr))
2766          {          {
2767          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2768          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
# Line 2516  for (;; ptr++) Line 2779  for (;; ptr++)
2779          if (*ptr == '^')          if (*ptr == '^')
2780            {            {
2781            local_negate = TRUE;            local_negate = TRUE;
2782              should_flip_negation = TRUE;  /* Note negative special */
2783            ptr++;            ptr++;
2784            }            }
2785    
# Line 2582  for (;; ptr++) Line 2846  for (;; ptr++)
2846        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
2847        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
2848        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2849        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
2850        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2851    
2852        if (c == '\\')        if (c == '\\')
# Line 2590  for (;; ptr++) Line 2854  for (;; ptr++)
2854          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2855          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
2856    
2857          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
2858          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 */
2859          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 */
2860          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
# Line 2602  for (;; ptr++) Line 2866  for (;; ptr++)
2866            else inescq = TRUE;            else inescq = TRUE;
2867            continue;            continue;
2868            }            }
2869            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2870    
2871          if (c < 0)          if (c < 0)
2872            {            {
# Line 2617  for (;; ptr++) Line 2882  for (;; ptr++)
2882              continue;              continue;
2883    
2884              case ESC_D:              case ESC_D:
2885                should_flip_negation = TRUE;
2886              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2887              continue;              continue;
2888    
# Line 2625  for (;; ptr++) Line 2891  for (;; ptr++)
2891              continue;              continue;
2892    
2893              case ESC_W:              case ESC_W:
2894                should_flip_negation = TRUE;
2895              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2896              continue;              continue;
2897    
# Line 2634  for (;; ptr++) Line 2901  for (;; ptr++)
2901              continue;              continue;
2902    
2903              case ESC_S:              case ESC_S:
2904                should_flip_negation = TRUE;
2905              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2906              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2907              continue;              continue;
2908    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
2909              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
2910              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
2911              }              }
# Line 2828  for (;; ptr++) Line 3093  for (;; ptr++)
3093    
3094        oldptr = ptr;        oldptr = ptr;
3095    
3096          /* Remember \r or \n */
3097    
3098          if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3099    
3100          /* Check for range */
3101    
3102        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == '-')
3103          {          {
3104          int d;          int d;
# Line 2869  for (;; ptr++) Line 3140  for (;; ptr++)
3140            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3141            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3142    
3143            /* \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
3144            special means the '-' was literal */            special means the '-' was literal */
3145    
3146            if (d < 0)            if (d < 0)
# Line 2895  for (;; ptr++) Line 3166  for (;; ptr++)
3166    
3167          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3168    
3169            /* Remember \r or \n */
3170    
3171            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3172    
3173          /* 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
3174          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3175          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3048  for (;; ptr++) Line 3323  for (;; ptr++)
3323        goto FAILED;        goto FAILED;
3324        }        }
3325    
3326    
3327    /* This code has been disabled because it would mean that \s counts as
3328    an explicit \r or \n reference, and that's not really what is wanted. Now
3329    we set the flag only if there is a literal "\r" or "\n" in the class. */
3330    
3331    #if 0
3332        /* Remember whether \r or \n are in this class */
3333    
3334        if (negate_class)
3335          {
3336          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3337          }
3338        else
3339          {
3340          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3341          }
3342    #endif
3343    
3344    
3345      /* 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
3346      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
3347      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
3348      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3349      single-bytes only. This is an historical hangover. Maybe one day we can  
3350      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3351        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3352        operate on single-bytes only. This is an historical hangover. Maybe one day
3353        we can tidy these opcodes to handle multi-byte characters.
3354    
3355      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
3356      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 3063  for (;; ptr++) Line 3360  for (;; ptr++)
3360      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3361    
3362  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3363      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3364            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3365  #else  #else
3366      if (class_charcount == 1)      if (class_charcount == 1)
3367  #endif  #endif
# Line 3109  for (;; ptr++) Line 3404  for (;; ptr++)
3404      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3405    
3406      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3407      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3408      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
3409        the class, so any that were explicitly given as well can be ignored. If
3410        (when there are explicit characters > 255 that must be listed) there are no
3411        characters < 256, we can omit the bitmap in the actual compiled code. */
3412    
3413  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3414      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3415        {        {
3416        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3417        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3139  for (;; ptr++) Line 3437  for (;; ptr++)
3437        }        }
3438  #endif  #endif
3439    
3440      /* 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
3441      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
3442      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
3443      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3444    
3445        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3446      if (negate_class)      if (negate_class)
3447        {        {
       *code++ = OP_NCLASS;  
3448        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3449          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3450        }        }
3451      else      else
3452        {        {
       *code++ = OP_CLASS;  
3453        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3454        }        }
3455      code += 32;      code += 32;
# Line 3338  for (;; ptr++) Line 3635  for (;; ptr++)
3635        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3636        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3637    
3638        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3639    
3640        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3641    
# Line 3488  for (;; ptr++) Line 3785  for (;; ptr++)
3785        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3786        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3787    
3788        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3789    
3790        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3791          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3524  for (;; ptr++) Line 3821  for (;; ptr++)
3821          goto FAILED;          goto FAILED;
3822          }          }
3823    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
3824        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
3825        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
3826        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 3554  for (;; ptr++) Line 3843  for (;; ptr++)
3843    
3844        if (repeat_min == 0)        if (repeat_min == 0)
3845          {          {
3846          /* 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
3847          altogether. */          output altogether, like this:
3848    
3849          if (repeat_max == 0)          ** if (repeat_max == 0)
3850            {          **   {
3851            code = previous;          **   code = previous;
3852            goto END_REPEAT;          **   goto END_REPEAT;
3853            }          **   }
3854    
3855          /* If the maximum is 1 or unlimited, we just have to stick in the          However, that fails when a group is referenced as a subroutine from
3856          BRAZERO and do no more at this point. However, we do need to adjust          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
3857          any OP_RECURSE calls inside the group that refer to the group itself or          so that it is skipped on execution. As we don't have a list of which
3858          any internal or forward referenced group, because the offset is from          groups are referenced, we cannot do this selectively.
3859          the start of the whole regex. Temporarily terminate the pattern while  
3860          doing this. */          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
3861            and do no more at this point. However, we do need to adjust any
3862            OP_RECURSE calls inside the group that refer to the group itself or any
3863            internal or forward referenced group, because the offset is from the
3864            start of the whole regex. Temporarily terminate the pattern while doing
3865            this. */
3866    
3867          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
3868            {            {
3869            *code = OP_END;            *code = OP_END;
3870            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3871            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3872            code++;            code++;
3873              if (repeat_max == 0)
3874                {
3875                *previous++ = OP_SKIPZERO;
3876                goto END_REPEAT;
3877                }
3878            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
3879            }            }
3880    
# Line 3620  for (;; ptr++) Line 3919  for (;; ptr++)
3919          if (repeat_min > 1)          if (repeat_min > 1)
3920            {            {
3921            /* In the pre-compile phase, we don't actually do the replication. We            /* In the pre-compile phase, we don't actually do the replication. We
3922            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
3923              potential integer overflow. */
3924    
3925            if (lengthptr != NULL)            if (lengthptr != NULL)
3926              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
3927                int delta = (repeat_min - 1)*length_prevgroup;
3928                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3929                                                                (double)INT_MAX ||
3930                    OFLOW_MAX - *lengthptr < delta)
3931                  {
3932                  *errorcodeptr = ERR20;
3933                  goto FAILED;
3934                  }
3935                *lengthptr += delta;
3936                }
3937    
3938            /* This is compiling for real */            /* This is compiling for real */
3939    
# Line 3661  for (;; ptr++) Line 3971  for (;; ptr++)
3971          /* In the pre-compile phase, we don't actually do the replication. We          /* In the pre-compile phase, we don't actually do the replication. We
3972          just adjust the length as if we had. For each repetition we must add 1          just adjust the length as if we had. For each repetition we must add 1
3973          to the length for BRAZERO and for all but the last repetition we must          to the length for BRAZERO and for all but the last repetition we must
3974          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3975            paranoid checks to avoid integer overflow. */
3976    
3977          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
3978            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
3979              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3980                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3981              if ((double)repeat_max *
3982                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3983                      > (double)INT_MAX ||
3984                  OFLOW_MAX - *lengthptr < delta)
3985                {
3986                *errorcodeptr = ERR20;
3987                goto FAILED;
3988                }
3989              *lengthptr += delta;
3990              }
3991    
3992          /* This is compiling for real */          /* This is compiling for real */
3993    
# Line 3773  for (;; ptr++) Line 4095  for (;; ptr++)
4095        int len;        int len;
4096        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4097            *tempcode == OP_NOTEXACT)            *tempcode == OP_NOTEXACT)
4098          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode] +
4099              ((*tempcode == OP_TYPEEXACT &&
4100                 (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4101        len = code - tempcode;        len = code - tempcode;
4102        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4103          {          {
# Line 3817  for (;; ptr++) Line 4141  for (;; ptr++)
4141      /* ===================================================================*/      /* ===================================================================*/
4142      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4143      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4144      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. */  
4145    
4146      case '(':      case '(':
4147      newoptions = options;      newoptions = options;
# Line 3827  for (;; ptr++) Line 4149  for (;; ptr++)
4149      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4150      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4151      reset_bracount = FALSE;      reset_bracount = FALSE;
4152    
4153        /* First deal with various "verbs" that can be introduced by '*'. */
4154    
4155      if (*(++ptr) == '?')      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4156          {
4157          int i, namelen;
4158          const char *vn = verbnames;
4159          const uschar *name = ++ptr;
4160          previous = NULL;
4161          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4162          if (*ptr == ':')
4163            {
4164            *errorcodeptr = ERR59;   /* Not supported */
4165            goto FAILED;
4166            }
4167          if (*ptr != ')')
4168            {
4169            *errorcodeptr = ERR60;
4170            goto FAILED;
4171            }
4172          namelen = ptr - name;
4173          for (i = 0; i < verbcount; i++)
4174            {
4175            if (namelen == verbs[i].len &&
4176                strncmp((char *)name, vn, namelen) == 0)
4177              {
4178              *code = verbs[i].op;
4179              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4180              break;
4181              }
4182            vn += verbs[i].len + 1;
4183            }
4184          if (i < verbcount) continue;
4185          *errorcodeptr = ERR60;
4186          goto FAILED;
4187          }
4188    
4189        /* Deal with the extended parentheses; all are introduced by '?', and the
4190        appearance of any of them means that this is not a capturing group. */
4191    
4192        else if (*ptr == '?')
4193        {        {
4194        int i, set, unset, namelen;        int i, set, unset, namelen;
4195        int *optset;        int *optset;
# Line 3963  for (;; ptr++) Line 4324  for (;; ptr++)
4324              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4325              goto FAILED;              goto FAILED;
4326              }              }
4327            if (refsign == '-')            recno = (refsign == '-')?
4328                cd->bracount - recno + 1 : recno +cd->bracount;
4329              if (recno <= 0 || recno > cd->final_bracount)
4330              {              {
4331              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4332              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4333              }              }
           else recno += cd->bracount;  
4334            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4335            break;            break;
4336            }            }
# Line 4044  for (;; ptr++) Line 4402  for (;; ptr++)
4402            skipbytes = 1;            skipbytes = 1;
4403            }            }
4404    
4405          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4406            in the second pass here, so final_bracount is set. */
4407    
4408          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4409            {            {
4410            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4411            }            }
# Line 4070  for (;; ptr++) Line 4429  for (;; ptr++)
4429    
4430          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4431          case '!':                 /* Negative lookahead */          case '!':                 /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4432          ptr++;          ptr++;
4433            if (*ptr == ')')          /* Optimize (?!) */
4434              {
4435              *code++ = OP_FAIL;
4436              previous = NULL;
4437              continue;
4438              }
4439            bravalue = OP_ASSERT_NOT;
4440          break;          break;
4441    
4442    
# Line 4234  for (;; ptr++) Line 4599  for (;; ptr++)
4599    
4600          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4601          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4602          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4603            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4604            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4605    
4606          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4607          name = ++ptr;          name = ++ptr;
# Line 4246  for (;; ptr++) Line 4613  for (;; ptr++)
4613    
4614          if (lengthptr != NULL)          if (lengthptr != NULL)
4615            {            {
4616              if (namelen == 0)
4617                {
4618                *errorcodeptr = ERR62;
4619                goto FAILED;
4620                }
4621            if (*ptr != terminator)            if (*ptr != terminator)
4622              {              {
4623              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4259  for (;; ptr++) Line 4631  for (;; ptr++)
4631            recno = 0;            recno = 0;
4632            }            }
4633    
4634          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4635            first, and then check that we have reached the end of the name in the
4636            table. That way, if the name that is longer than any in the table,
4637            the comparison will fail without reading beyond the table entry. */
4638    
4639          else          else
4640            {            {
4641            slot = cd->name_table;            slot = cd->name_table;
4642            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4643              {              {
4644              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4645                    slot[2+namelen] == 0)
4646                  break;
4647              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4648              }              }
4649    
# Line 4302  for (;; ptr++) Line 4679  for (;; ptr++)
4679          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case '5': case '6': case '7': case '8': case '9':   /* subroutine */
4680            {            {
4681            const uschar *called;            const uschar *called;
4682              terminator = ')';
4683    
4684              /* Come here from the \g<...> and \g'...' code (Oniguruma
4685              compatibility). However, the syntax has been checked to ensure that
4686              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
4687              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
4688              ever be taken. */
4689    
4690              HANDLE_NUMERICAL_RECURSION:
4691    
4692            if ((refsign = *ptr) == '+') ptr++;            if ((refsign = *ptr) == '+')
4693                {
4694                ptr++;
4695                if ((digitab[*ptr] & ctype_digit) == 0)
4696                  {
4697                  *errorcodeptr = ERR63;
4698                  goto FAILED;
4699                  }
4700                }
4701            else if (refsign == '-')            else if (refsign == '-')
4702              {              {
4703              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
# Line 4315  for (;; ptr++) Line 4709  for (;; ptr++)
4709            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4710              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - '0';
4711    
4712            if (*ptr != ')')            if (*ptr != terminator)
4713              {              {
4714              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
4715              goto FAILED;              goto FAILED;
# Line 4429  for (;; ptr++) Line 4823  for (;; ptr++)
4823    
4824              case 'J':    /* Record that it changed in the external options */              case 'J':    /* Record that it changed in the external options */
4825              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
4826              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
4827              break;              break;
4828    
4829              case 'i': *optset |= PCRE_CASELESS; break;              case 'i': *optset |= PCRE_CASELESS; break;
# Line 4620  for (;; ptr++) Line 5014  for (;; ptr++)
5014        goto FAILED;        goto FAILED;
5015        }        }
5016    
5017      /* 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,
5018      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
5019      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
5020      a quantifier. */      duplicated by a quantifier.*/
5021    
5022      if (lengthptr != NULL)      if (lengthptr != NULL)
5023        {        {
5024          if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
5025            {
5026            *errorcodeptr = ERR20;
5027            goto FAILED;
5028            }
5029        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5030        code++;        *code++ = OP_BRA;
5031        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5032        *code++ = OP_KET;        *code++ = OP_KET;
5033        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5034          break;    /* No need to waste time with special character handling */
5035        }        }
5036    
5037      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
5038    
5039      else code = tempcode;      code = tempcode;
5040    
5041      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
5042      relevant. */      relevant. */
# Line 4705  for (;; ptr++) Line 5105  for (;; ptr++)
5105      back references and those types that consume a character may be repeated.      back references and those types that consume a character may be repeated.
5106      We can test for values between ESC_b and ESC_Z for the latter; this may      We can test for values between ESC_b and ESC_Z for the latter; this may
5107      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5108    
5109      case '\\':      case '\\':
5110      tempptr = ptr;      tempptr = ptr;
5111      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
# Line 4732  for (;; ptr++) Line 5132  for (;; ptr++)
5132    
5133        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5134        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5135    
5136          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5137          is a subroutine call by number (Oniguruma syntax). In fact, the value
5138          -ESC_g is returned only for these cases. So we don't need to check for <
5139          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5140          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5141          that is a synonym for a named back reference). */
5142    
5143          if (-c == ESC_g)
5144            {
5145            const uschar *p;
5146            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5147            terminator = (*(++ptr) == '<')? '>' : '\'';
5148    
5149            /* These two statements stop the compiler for warning about possibly
5150            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5151            fact, because we actually check for a number below, the paths that
5152            would actually be in error are never taken. */
5153    
5154            skipbytes = 0;
5155            reset_bracount = FALSE;
5156    
5157            /* Test for a name */
5158    
5159            if (ptr[1] != '+' && ptr[1] != '-')
5160              {
5161              BOOL isnumber = TRUE;
5162              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5163                {
5164                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5165                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5166                }
5167              if (*p != terminator)
5168                {
5169                *errorcodeptr = ERR57;
5170                break;
5171                }
5172              if (isnumber)
5173                {
5174                ptr++;
5175                goto HANDLE_NUMERICAL_RECURSION;
5176                }
5177              is_recurse = TRUE;
5178              goto NAMED_REF_OR_RECURSE;
5179              }
5180    
5181            /* Test a signed number in angle brackets or quotes. */
5182    
5183            p = ptr + 2;
5184            while ((digitab[*p] & ctype_digit) != 0) p++;
5185            if (*p != terminator)
5186              {
5187              *errorcodeptr = ERR57;
5188              break;
5189              }
5190            ptr++;
5191            goto HANDLE_NUMERICAL_RECURSION;
5192            }
5193    
5194        /* \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).
5195        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
# Line 4840  for (;; ptr++) Line 5298  for (;; ptr++)
5298      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5299      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5300    
5301        /* Remember if \r or \n were seen */
5302    
5303        if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')
5304          cd->external_flags |= PCRE_HASCRORLF;
5305    
5306      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5307      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.
5308      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 5122  for (;;) Line 5585  for (;;)
5585      *ptrptr = ptr;      *ptrptr = ptr;
5586      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
5587      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
5588      if (lengthptr != NULL) *lengthptr += length;      if (lengthptr != NULL)
5589          {
5590          if (OFLOW_MAX - *lengthptr < length)
5591            {
5592            *errorcodeptr = ERR20;
5593            return FALSE;
5594            }
5595          *lengthptr += length;
5596          }
5597      return TRUE;      return TRUE;
5598      }      }
5599    
# Line 5431  real_pcre *re; Line 5902  real_pcre *re;
5902  int length = 1;  /* For final END opcode */  int length = 1;  /* For final END opcode */
5903  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
5904  int errorcode = 0;  int errorcode = 0;
5905    int skipatstart = 0;
5906  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5907  BOOL utf8;  BOOL utf8;
5908  #endif  #endif
# Line 5449  to fill in forward references to subpatt Line 5921  to fill in forward references to subpatt
5921    
5922  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
5923    
   
5924  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
5925    
5926  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;
# Line 5509  cd->fcc = tables + fcc_offset; Line 5980  cd->fcc = tables + fcc_offset;
5980  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
5981  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
5982    
5983    /* Check for global one-time settings at the start of the pattern, and remember
5984    the offset for later. */
5985    
5986    while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')
5987      {
5988      int newnl = 0;
5989      int newbsr = 0;
5990    
5991      if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)
5992        { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
5993      else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)
5994        { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
5995      else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)
5996        { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
5997      else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)
5998        { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
5999      else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)
6000        { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6001    
6002      else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)
6003        { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6004      else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)
6005        { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6006    
6007      if (newnl != 0)
6008        options = (options & ~PCRE_NEWLINE_BITS) | newnl;
6009      else if (newbsr != 0)
6010        options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
6011      else break;
6012      }
6013    
6014    /* Check validity of \R options. */
6015    
6016    switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6017      {
6018      case 0:
6019      case PCRE_BSR_ANYCRLF:
6020      case PCRE_BSR_UNICODE:
6021      break;
6022      default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
6023      }
6024    
6025  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
6026  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
6027  "anycrlf". */  "anycrlf". */
6028    
6029  switch (options & (PCRE_NEWLINE_CRLF | PCRE_NEWLINE_ANY))  switch (options & PCRE_NEWLINE_BITS)
6030    {    {
6031    case 0: newline = NEWLINE; break;   /* Compile-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6032    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = '\r'; break;
6033    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = '\n'; break;
6034    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
# Line 5568  to compile parts of the pattern into; th Line 6081  to compile parts of the pattern into; th
6081  no longer needed, so hopefully this workspace will never overflow, though there  no longer needed, so hopefully this workspace will never overflow, though there
6082  is a test for its doing so. */  is a test for its doing so. */
6083    
6084  cd->bracount = 0;  cd->bracount = cd->final_bracount = 0;
6085  cd->names_found = 0;  cd->names_found = 0;
6086  cd->name_entry_size = 0;  cd->name_entry_size = 0;
6087  cd->name_table = NULL;  cd->name_table = NULL;
# Line 5578  cd->hwm = cworkspace; Line 6091  cd->hwm = cworkspace;
6091  cd->start_pattern = (const uschar *)pattern;  cd->start_pattern = (const uschar *)pattern;
6092  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));
6093  cd->req_varyopt = 0;  cd->req_varyopt = 0;
 cd->nopartial = FALSE;  
6094  cd->external_options = options;  cd->external_options = options;
6095    cd->external_flags = 0;
6096    
6097  /* 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
6098  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 5587  been put into the cd block so that they Line 6100  been put into the cd block so that they
6100  found within the regex right at the beginning. Bringing initial option settings  found within the regex right at the beginning. Bringing initial option settings
6101  outside can help speed up starting point checks. */  outside can help speed up starting point checks. */
6102    
6103    ptr += skipatstart;
6104  code = cworkspace;  code = cworkspace;
6105  *code = OP_BRA;  *code = OP_BRA;
6106  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,
# Line 5617  if (re == NULL) Line 6131  if (re == NULL)
6131    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
6132    }    }
6133    
6134  /* 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
6135  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
6136  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
6137  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
6138    pointers. */
6139    
6140  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
6141  re->size = size;  re->size = size;
6142  re->options = cd->external_options;  re->options = cd->external_options;
6143    re->flags = cd->external_flags;
6144  re->dummy1 = 0;  re->dummy1 = 0;
6145  re->first_byte = 0;  re->first_byte = 0;
6146  re->req_byte = 0;  re->req_byte = 0;
# Line 5642  field. Reset the bracket count and the n Line 6158  field. Reset the bracket count and the n
6158  field; this time it's used for remembering forward references to subpatterns.  field; this time it's used for remembering forward references to subpatterns.
6159  */  */
6160    
6161    cd->final_bracount = cd->bracount;  /* Save for checking forward references */
6162  cd->bracount = 0;  cd->bracount = 0;
6163  cd->names_found = 0;  cd->names_found = 0;
6164  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 5649  codestart = cd->name_table + re->name_en Line 6166  codestart = cd->name_table + re->name_en
6166  cd->start_code = codestart;  cd->start_code = codestart;
6167  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6168  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6169  cd->nopartial = FALSE;  cd->had_accept = FALSE;
6170    
6171  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6172  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
6173  of the function here. */  of the function here. */
6174    
6175  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern + skipatstart;
6176  code = (uschar *)codestart;  code = (uschar *)codestart;
6177  *code = OP_BRA;  *code = OP_BRA;
6178  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,
6179    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);
6180  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
6181  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
6182    re->flags = cd->external_flags;
6183    
6184  if (cd->nopartial) re->options |= PCRE_NOPARTIAL;  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */
6185    
6186  /* 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. */
6187    
# Line 5691  while (errorcode == 0 && cd->hwm > cwork Line 6209  while (errorcode == 0 && cd->hwm > cwork
6209    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6210      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6211    }    }
6212    
6213  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
6214  subpattern. */  subpattern. */
6215    
# Line 5705  if (errorcode != 0) Line 6223  if (errorcode != 0)
6223    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
6224    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = ptr - (const uschar *)pattern;
6225    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
6226    *errorptr = error_texts[errorcode];    *errorptr = find_error_text(errorcode);
6227    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
6228    return NULL;    return NULL;
6229    }    }
# Line 5734  if ((re->options & PCRE_ANCHORED) == 0) Line 6252  if ((re->options & PCRE_ANCHORED) == 0)
6252        int ch = firstbyte & 255;        int ch = firstbyte & 255;
6253        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&
6254           cd->fcc[ch] == ch)? ch : firstbyte;           cd->fcc[ch] == ch)? ch : firstbyte;
6255        re->options |= PCRE_FIRSTSET;        re->flags |= PCRE_FIRSTSET;
6256        }        }
6257      else if (is_startline(codestart, 0, cd->backref_map))      else if (is_startline(codestart, 0, cd->backref_map))
6258        re->options |= PCRE_STARTLINE;        re->flags |= PCRE_STARTLINE;
6259      }      }
6260    }    }
6261    
# Line 5751  if (reqbyte >= 0 && Line 6269  if (reqbyte >= 0 &&
6269    int ch = reqbyte & 255;    int ch = reqbyte & 255;
6270    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&
6271      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;
6272    re->options |= PCRE_REQCHSET;    re->flags |= PCRE_REQCHSET;
6273    }    }
6274    
6275  /* 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 5762  case when building a production library. Line 6280  case when building a production library.
6280  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6281    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6282    
6283  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 " : "");  
   }  
6284    
6285  if ((re->options & PCRE_FIRSTSET) != 0)  if ((re->flags & PCRE_FIRSTSET) != 0)
6286    {    {
6287    int ch = re->first_byte & 255;    int ch = re->first_byte & 255;
6288    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?
# Line 5785  if ((re->options & PCRE_FIRSTSET) != 0) Line 6291  if ((re->options & PCRE_FIRSTSET) != 0)
6291      else printf("First char = \\x%02x%s\n", ch, caseless);      else printf("First char = \\x%02x%s\n", ch, caseless);
6292    }    }
6293    
6294  if ((re->options & PCRE_REQCHSET) != 0)  if ((re->flags & PCRE_REQCHSET) != 0)
6295    {    {
6296    int ch = re->req_byte & 255;    int ch = re->req_byte & 255;
6297    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?
# Line 5802  was compiled can be seen. */ Line 6308  was compiled can be seen. */
6308  if (code - codestart > length)  if (code - codestart > length)
6309    {    {
6310    (pcre_free)(re);    (pcre_free)(re);
6311    *errorptr = error_texts[ERR23];    *errorptr = find_error_text(ERR23);
6312    *erroroffset = ptr - (uschar *)pattern;    *erroroffset = ptr - (uschar *)pattern;
6313    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6314    return NULL;    return NULL;

Legend:
Removed from v.199  
changed lines
  Added in v.335

  ViewVC Help
Powered by ViewVC 1.1.5