/[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 218 by ph10, Thu Aug 16 10:13:23 2007 UTC revision 426 by ph10, Wed Aug 26 15:38:32 2009 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) 1997-2009 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 43  supporting internal functions that are n Line 43  supporting internal functions that are n
43    
44    
45  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
46  #include <config.h>  #include "config.h"
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
# Line 97  are simple data values; negative values Line 97  are simple data values; negative values
97  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
98  is invalid. */  is invalid. */
99    
100  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
101    
102    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
103    in UTF-8 mode. */
104    
105  static const short int escapes[] = {  static const short int escapes[] = {
106       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
107       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
108     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
109  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
110  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
111  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
112     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
113  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
114  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
115       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
116         -ESC_D,                  -ESC_E,
117         0,                       -ESC_G,
118         -ESC_H,                  0,
119         0,                       -ESC_K,
120         0,                       0,
121         0,                       0,
122         -ESC_P,                  -ESC_Q,
123         -ESC_R,                  -ESC_S,
124         0,                       0,
125         -ESC_V,                  -ESC_W,
126         -ESC_X,                  0,
127         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
128         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
129         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
130         CHAR_GRAVE_ACCENT,       7,
131         -ESC_b,                  0,
132         -ESC_d,                  ESC_e,
133         ESC_f,                   0,
134         -ESC_h,                  0,
135         0,                       -ESC_k,
136         0,                       0,
137         ESC_n,                   0,
138         -ESC_p,                  0,
139         ESC_r,                   -ESC_s,
140         ESC_tee,                 0,
141         -ESC_v,                  -ESC_w,
142         0,                       0,
143         -ESC_z
144  };  };
145    
146  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
147    
148    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
149    
150  static const short int escapes[] = {  static const short int escapes[] = {
151  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
152  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 140  static const short int escapes[] = { Line 175  static const short int escapes[] = {
175  #endif  #endif
176    
177    
178  /* Table of special "verbs" like (*PRUNE) */  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
179    searched linearly. Put all the names into a single string, in order to reduce
180    the number of relocations when a shared library is dynamically linked. The
181    string is built from string macros so that it works in UTF-8 mode on EBCDIC
182    platforms. */
183    
184  typedef struct verbitem {  typedef struct verbitem {
   const char *name;  
185    int   len;    int   len;
186    int   op;    int   op;
187  } verbitem;  } verbitem;
188    
189  static verbitem verbs[] = {  static const char verbnames[] =
190    { "ACCEPT", 6, OP_ACCEPT },    STRING_ACCEPT0
191    { "COMMIT", 6, OP_COMMIT },    STRING_COMMIT0
192    { "F",      1, OP_FAIL },    STRING_F0
193    { "FAIL",   4, OP_FAIL },    STRING_FAIL0
194    { "PRUNE",  5, OP_PRUNE },    STRING_PRUNE0
195    { "SKIP",   4, OP_SKIP  },    STRING_SKIP0
196    { "THEN",   4, OP_THEN  }    STRING_THEN;
197    
198    static const verbitem verbs[] = {
199      { 6, OP_ACCEPT },
200      { 6, OP_COMMIT },
201      { 1, OP_FAIL },
202      { 4, OP_FAIL },
203      { 5, OP_PRUNE },
204      { 4, OP_SKIP  },
205      { 4, OP_THEN  }
206  };  };
207    
208  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
   
209    
 /* Tables of names of POSIX character classes and their lengths. The list is  
 terminated by a zero length entry. The first three must be alpha, lower, upper,  
 as this is assumed for handling case independence. */  
210    
211  static const char *const posix_names[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
212    "alpha", "lower", "upper",  now all in a single string, to reduce the number of relocations when a shared
213    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  library is dynamically loaded. The list of lengths is terminated by a zero
214    "print", "punct", "space", "word",  "xdigit" };  length entry. The first three must be alpha, lower, upper, as this is assumed
215    for handling case independence. */
216    
217    static const char posix_names[] =
218      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
219      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
220      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
221      STRING_word0  STRING_xdigit;
222    
223  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
224    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 207  static const int posix_class_maps[] = { Line 257  static const int posix_class_maps[] = {
257  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
258  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
259  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
260  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
261    the number of relocations needed when a shared library is loaded dynamically,
262  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
263    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
264    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
265    "\\c at end of pattern",  because these strings are used only when there is a compilation error. */
266    "unrecognized character follows \\",  
267    "numbers out of order in {} quantifier",  static const char error_texts[] =
268      "no error\0"
269      "\\ at end of pattern\0"
270      "\\c at end of pattern\0"
271      "unrecognized character follows \\\0"
272      "numbers out of order in {} quantifier\0"
273    /* 5 */    /* 5 */
274    "number too big in {} quantifier",    "number too big in {} quantifier\0"
275    "missing terminating ] for character class",    "missing terminating ] for character class\0"
276    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
277    "range out of order in character class",    "range out of order in character class\0"
278    "nothing to repeat",    "nothing to repeat\0"
279    /* 10 */    /* 10 */
280    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
281    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
282    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
283    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
284    "missing )",    "missing )\0"
285    /* 15 */    /* 15 */
286    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
287    "erroffset passed as NULL",    "erroffset passed as NULL\0"
288    "unknown option bit(s) set",    "unknown option bit(s) set\0"
289    "missing ) after comment",    "missing ) after comment\0"
290    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
291    /* 20 */    /* 20 */
292    "regular expression is too large",    "regular expression is too large\0"
293    "failed to get memory",    "failed to get memory\0"
294    "unmatched parentheses",    "unmatched parentheses\0"
295    "internal error: code overflow",    "internal error: code overflow\0"
296    "unrecognized character after (?<",    "unrecognized character after (?<\0"
297    /* 25 */    /* 25 */
298    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
299    "malformed number or name after (?(",    "malformed number or name after (?(\0"
300    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
301    "assertion expected after (?(",    "assertion expected after (?(\0"
302    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
303    /* 30 */    /* 30 */
304    "unknown POSIX class name",    "unknown POSIX class name\0"
305    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
306    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
307    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
308    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
309    /* 35 */    /* 35 */
310    "invalid condition (?(0)",    "invalid condition (?(0)\0"
311    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
312    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
313    "number after (?C is > 255",    "number after (?C is > 255\0"
314    "closing ) for (?C expected",    "closing ) for (?C expected\0"
315    /* 40 */    /* 40 */
316    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
317    "unrecognized character after (?P",    "unrecognized character after (?P\0"
318    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
319    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
320    "invalid UTF-8 string",    "invalid UTF-8 string\0"
321    /* 45 */    /* 45 */
322    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
323    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
324    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
325    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
326    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
327    /* 50 */    /* 50 */
328    "repeated subpattern is too long",    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
329    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
330    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
331    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
332    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
333    /* 55 */    /* 55 */
334    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
335    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
336    "\\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"
337    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number",    "a numbered reference must not be zero\0"
338    "(*VERB) with an argument is not supported",    "(*VERB) with an argument is not supported\0"
339    /* 60 */    /* 60 */
340    "(*VERB) not recognized",    "(*VERB) not recognized\0"
341    "number is too big"    "number is too big\0"
342  };    "subpattern name expected\0"
343      "digit expected after (?+\0"
344      "] is an invalid data character in JavaScript compatibility mode";
345    
346    
347  /* 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 303  For convenience, we use the same bit def Line 360  For convenience, we use the same bit def
360    
361  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
362    
363  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
364    
365    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
366    UTF-8 mode. */
367    
368  static const unsigned char digitab[] =  static const unsigned char digitab[] =
369    {    {
370    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 339  static const unsigned char digitab[] = Line 400  static const unsigned char digitab[] =
400    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
401    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
402    
403  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
404    
405    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
406    
407  static const unsigned char digitab[] =  static const unsigned char digitab[] =
408    {    {
409    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 420  static BOOL Line 484  static BOOL
484    
485    
486  /*************************************************  /*************************************************
487    *            Find an error text                  *
488    *************************************************/
489    
490    /* The error texts are now all in one long string, to save on relocations. As
491    some of the text is of unknown length, we can't use a table of offsets.
492    Instead, just count through the strings. This is not a performance issue
493    because it happens only when there has been a compilation error.
494    
495    Argument:   the error number
496    Returns:    pointer to the error string
497    */
498    
499    static const char *
500    find_error_text(int n)
501    {
502    const char *s = error_texts;
503    for (; n > 0; n--) while (*s++ != 0) {};
504    return s;
505    }
506    
507    
508    /*************************************************
509  *            Handle escapes                      *  *            Handle escapes                      *
510  *************************************************/  *************************************************/
511    
# Line 458  ptr--;                            /* Set Line 544  ptr--;                            /* Set
544    
545  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
546    
547  /* 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
548  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.
549  Otherwise further processing may be required. */  Otherwise further processing may be required. */
550    
551  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
552  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
553  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
554    
555  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
556  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
557  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
558  #endif  #endif
559    
# Line 483  else Line 569  else
569      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
570      error. */      error. */
571    
572      case 'l':      case CHAR_l:
573      case 'L':      case CHAR_L:
574      case 'N':      case CHAR_N:
575      case 'u':      case CHAR_u:
576      case 'U':      case CHAR_U:
577      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
578      break;      break;
579    
580      /* \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:
581      is an absolute backreference. If negative, it is a relative backreference.  
582      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
583      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
584      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
585      fudge it up by pretending it really was \k. */  
586        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
587        is part of Perl's movement towards a unified syntax for back references. As
588        this is synonymous with \k{name}, we fudge it up by pretending it really
589        was \k.
590    
591        (3) For Oniguruma compatibility we also support \g followed by a name or a
592        number either in angle brackets or in single quotes. However, these are
593        (possibly recursive) subroutine calls, _not_ backreferences. Just return
594        the -ESC_g code (cf \k). */
595    
596      case 'g':      case CHAR_g:
597      if (ptr[1] == '{')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
598          {
599          c = -ESC_g;
600          break;
601          }
602    
603        /* Handle the Perl-compatible cases */
604    
605        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
606        {        {
607        const uschar *p;        const uschar *p;
608        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
609          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
610        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
611          {          {
612          c = -ESC_k;          c = -ESC_k;
613          break;          break;
# Line 514  else Line 617  else
617        }        }
618      else braced = FALSE;      else braced = FALSE;
619    
620      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
621        {        {
622        negated = TRUE;        negated = TRUE;
623        ptr++;        ptr++;
# Line 523  else Line 626  else
626    
627      c = 0;      c = 0;
628      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
629        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
630    
631      if (c < 0)      if (c < 0)   /* Integer overflow */
632        {        {
633        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
634        break;        break;
635        }        }
636    
637      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
638        {        {
639        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
640        break;        break;
641        }        }
642    
643        if (c == 0)
644          {
645          *errorcodeptr = ERR58;
646          break;
647          }
648    
649      if (negated)      if (negated)
650        {        {
651        if (c > bracount)        if (c > bracount)
# Line 562  else Line 671  else
671      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
672      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
673    
674      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
675      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
676    
677      if (!isclass)      if (!isclass)
678        {        {
679        oldptr = ptr;        oldptr = ptr;
680        c -= '0';        c -= CHAR_0;
681        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
682          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
683        if (c < 0)        if (c < 0)    /* Integer overflow */
684          {          {
685          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
686          break;          break;
687          }          }
688        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
689          {          {
690          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 588  else Line 697  else
697      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
698      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
699    
700      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
701        {        {
702        ptr--;        ptr--;
703        c = 0;        c = 0;
# Line 601  else Line 710  else
710      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
711      than 3 octal digits. */      than 3 octal digits. */
712    
713      case '0':      case CHAR_0:
714      c -= '0';      c -= CHAR_0;
715      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
716          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
717      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
718      break;      break;
719    
# Line 612  else Line 721  else
721      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
722      treated as a data character. */      treated as a data character. */
723    
724      case 'x':      case CHAR_x:
725      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
726        {        {
727        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
728        int count = 0;        int count = 0;
# Line 622  else Line 731  else
731        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
732          {          {
733          register int cc = *pt++;          register int cc = *pt++;
734          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
735          count++;          count++;
736    
737  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
738          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
739          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
740  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
741          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
742          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
743  #endif  #endif
744          }          }
745    
746        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
747          {          {
748          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
749          ptr = pt;          ptr = pt;
# Line 650  else Line 759  else
759      c = 0;      c = 0;
760      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
761        {        {
762        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
763        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
764  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
765        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
766        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
767  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
768        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
769        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
770  #endif  #endif
771        }        }
772      break;      break;
# Line 666  else Line 775  else
775      This coding is ASCII-specific, but then the whole concept of \cx is      This coding is ASCII-specific, but then the whole concept of \cx is
776      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
777    
778      case 'c':      case CHAR_c:
779      c = *(++ptr);      c = *(++ptr);
780      if (c == 0)      if (c == 0)
781        {        {
# Line 674  else Line 783  else
783        break;        break;
784        }        }
785    
786  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
787      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
788      c ^= 0x40;      c ^= 0x40;
789  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
790      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
791      c ^= 0xC0;      c ^= 0xC0;
792  #endif  #endif
793      break;      break;
794    
795      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
796      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
797      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
798      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
799      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
800    
801      default:      default:
802      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 740  if (c == 0) goto ERROR_RETURN; Line 849  if (c == 0) goto ERROR_RETURN;
849  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
850  negation. */  negation. */
851    
852  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
853    {    {
854    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
855      {      {
856      *negptr = TRUE;      *negptr = TRUE;
857      ptr++;      ptr++;
# Line 751  if (c == '{') Line 860  if (c == '{')
860      {      {
861      c = *(++ptr);      c = *(++ptr);
862      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
863      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
864      name[i] = c;      name[i] = c;
865      }      }
866    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
867    name[i] = 0;    name[i] = 0;
868    }    }
869    
# Line 776  top = _pcre_utt_size; Line 885  top = _pcre_utt_size;
885  while (bot < top)  while (bot < top)
886    {    {
887    i = (bot + top) >> 1;    i = (bot + top) >> 1;
888    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
889    if (c == 0)    if (c == 0)
890      {      {
891      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 819  is_counted_repeat(const uschar *p) Line 928  is_counted_repeat(const uschar *p)
928  {  {
929  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
930  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
931  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
932    
933  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
934  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
935    
936  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
937  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
938    
939  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
940  }  }
941    
942    
# Line 860  int max = -1; Line 969  int max = -1;
969  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
970  an integer overflow. */  an integer overflow. */
971    
972  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
973  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
974    {    {
975    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 870  if (min < 0 || min > 65535) Line 979  if (min < 0 || min > 65535)
979  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
980  Also, max must not be less than min. */  Also, max must not be less than min. */
981    
982  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
983    {    {
984    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
985      {      {
986      max = 0;      max = 0;
987      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
988      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
989        {        {
990        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 900  return p; Line 1009  return p;
1009    
1010    
1011  /*************************************************  /*************************************************
1012  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1013  *************************************************/  *************************************************/
1014    
1015  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1016    top-level call starts at the beginning of the pattern. All other calls must
1017    start at a parenthesis. It scans along a pattern's text looking for capturing
1018  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1019  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1020  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. We know that if (?P< is
1021  references to subpatterns. We know that if (?P< is encountered, the name will  encountered, the name will be terminated by '>' because that is checked in the
1022  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1023    capturing group numbers - the (?| feature.
1024    
1025  Arguments:  Arguments:
1026    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1027    count        current count of capturing parens so far encountered    cd           compile background data
1028    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1029    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1030    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1031      count        pointer to the current capturing subpattern number (updated)
1032    
1033  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1034  */  */
1035    
1036  static int  static int
1037  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1038    BOOL xmode)    BOOL xmode, int *count)
1039  {  {
1040  const uschar *thisname;  uschar *ptr = *ptrptr;
1041    int start_count = *count;
1042    int hwm_count = start_count;
1043    BOOL dup_parens = FALSE;
1044    
1045  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1046    dealing with. The very first call may not start with a parenthesis. */
1047    
1048    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1049    {    {
1050    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1051          ptr[2] == CHAR_VERTICAL_LINE)
1052        {
1053        ptr += 3;
1054        dup_parens = TRUE;
1055        }
1056    
1057      /* Handle a normal, unnamed capturing parenthesis */
1058    
1059      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1060        {
1061        *count += 1;
1062        if (name == NULL && *count == lorn) return *count;
1063        ptr++;
1064        }
1065    
1066      /* Handle a condition. If it is an assertion, just carry on so that it
1067      is processed as normal. If not, skip to the closing parenthesis of the
1068      condition (there can't be any nested parens. */
1069    
1070      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1071        {
1072        ptr += 2;
1073        if (ptr[1] != CHAR_QUESTION_MARK)
1074          {
1075          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1076          if (*ptr != 0) ptr++;
1077          }
1078        }
1079    
1080      /* We have either (? or (* and not a condition */
1081    
1082      else
1083        {
1084        ptr += 2;
1085        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1086    
1087        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1088    
1089        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1090            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1091          {
1092          int term;
1093          const uschar *thisname;
1094          *count += 1;
1095          if (name == NULL && *count == lorn) return *count;
1096          term = *ptr++;
1097          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1098          thisname = ptr;
1099          while (*ptr != term) ptr++;
1100          if (name != NULL && lorn == ptr - thisname &&
1101              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1102            return *count;
1103          }
1104        }
1105      }
1106    
1107    /* Past any initial parenthesis handling, scan for parentheses or vertical
1108    bars. */
1109    
1110    for (; *ptr != 0; ptr++)
1111      {
1112    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1113    
1114    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1115      {      {
1116      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1117      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1118        {        {
1119        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1120        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1121        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1122        }        }
1123      continue;      continue;
1124      }      }
1125    
1126    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1127      are handled for real. If the first character is '^', skip it. Also, if the
1128      first few characters (either before or after ^) are \Q\E or \E we skip them
1129      too. This makes for compatibility with Perl. Note the use of STR macros to
1130      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1131    
1132    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1133      {      {
1134      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1135        for (;;)
1136          {
1137          int c = *(++ptr);
1138          if (c == CHAR_BACKSLASH)
1139            {
1140            if (ptr[1] == CHAR_E)
1141              ptr++;
1142            else if (strncmp((const char *)ptr+1,
1143                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1144              ptr += 3;
1145            else
1146              break;
1147            }
1148          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
1149            negate_class = TRUE;
1150          else break;
1151          }
1152    
1153        /* If the next character is ']', it is a data character that must be
1154        skipped, except in JavaScript compatibility mode. */
1155    
1156        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1157            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1158          ptr++;
1159    
1160        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1161        {        {
1162        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1163        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1164          {          {
1165          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1166          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1167            {            {
1168            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1169            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1170            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1171            }            }
1172          continue;          continue;
1173          }          }
# Line 968  for (; *ptr != 0; ptr++) Line 1177  for (; *ptr != 0; ptr++)
1177    
1178    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1179    
1180    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1181      {      {
1182      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1183      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1184      continue;      continue;
1185      }      }
1186    
1187    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1188    
1189    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1190      {      {
1191      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1192      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1193      continue;      if (*ptr == 0) goto FAIL_EXIT;
1194        }
1195    
1196      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1197        {
1198        if (dup_parens && *count < hwm_count) *count = hwm_count;
1199        *ptrptr = ptr;
1200        return -1;
1201      }      }
1202    
1203    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1204    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1205        if (*count > hwm_count) hwm_count = *count;
1206        *count = start_count;
1207        }
1208      }
1209    
1210    /* We have to disambiguate (?<! and (?<= from (?<name> */  FAIL_EXIT:
1211    *ptrptr = ptr;
1212    return -1;
1213    }
1214    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1215    
   count++;  
1216    
1217    if (name == NULL && count == lorn) return count;  
1218    term = *ptr++;  /*************************************************
1219    if (term == '<') term = '>';  *       Find forward referenced subpattern       *
1220    thisname = ptr;  *************************************************/
1221    while (*ptr != term) ptr++;  
1222    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1223        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1224      return count;  name it is given, it returns its number. Alternatively, if the name is NULL, it
1225    returns when it reaches a given numbered subpattern. This is used for forward
1226    references to subpatterns. We used to be able to start this scan from the
1227    current compiling point, using the current count value from cd->bracount, and
1228    do it all in a single loop, but the addition of the possibility of duplicate
1229    subpattern numbers means that we have to scan from the very start, in order to
1230    take account of such duplicates, and to use a recursive function to keep track
1231    of the different types of group.
1232    
1233    Arguments:
1234      cd           compile background data
1235      name         name to seek, or NULL if seeking a numbered subpattern
1236      lorn         name length, or subpattern number if name is NULL
1237      xmode        TRUE if we are in /x mode
1238    
1239    Returns:       the number of the found subpattern, or -1 if not found
1240    */
1241    
1242    static int
1243    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1244    {
1245    uschar *ptr = (uschar *)cd->start_pattern;
1246    int count = 0;
1247    int rc;
1248    
1249    /* If the pattern does not start with an opening parenthesis, the first call
1250    to find_parens_sub() will scan right to the end (if necessary). However, if it
1251    does start with a parenthesis, find_parens_sub() will return when it hits the
1252    matching closing parens. That is why we have to have a loop. */
1253    
1254    for (;;)
1255      {
1256      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1257      if (rc > 0 || *ptr++ == 0) break;
1258    }    }
1259    
1260  return -1;  return rc;
1261  }  }
1262    
1263    
1264    
1265    
1266  /*************************************************  /*************************************************
1267  *      Find first significant op code            *  *      Find first significant op code            *
1268  *************************************************/  *************************************************/
# Line 1171  for (;;) Line 1423  for (;;)
1423      branchlength++;      branchlength++;
1424      cc += 2;      cc += 2;
1425  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1426      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1427        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1428  #endif  #endif
1429      break;      break;
1430    
# Line 1185  for (;;) Line 1435  for (;;)
1435      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1436      cc += 4;      cc += 4;
1437  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1438      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1439        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1440  #endif  #endif
1441      break;      break;
1442    
1443      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1444      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1445      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1446      cc += 4;      cc += 4;
1447      break;      break;
1448    
# Line 1212  for (;;) Line 1460  for (;;)
1460      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1461      case OP_WORDCHAR:      case OP_WORDCHAR:
1462      case OP_ANY:      case OP_ANY:
1463        case OP_ALLANY:
1464      branchlength++;      branchlength++;
1465      cc++;      cc++;
1466      break;      break;
# Line 1303  for (;;) Line 1552  for (;;)
1552      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1553      }      }
1554    
1555    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1556    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1557    two bytes of parameters. */    two bytes of parameters. */
1558    
1559    else    else
# Line 1317  for (;;) Line 1566  for (;;)
1566        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
1567        case OP_TYPEQUERY:        case OP_TYPEQUERY:
1568        case OP_TYPEMINQUERY:        case OP_TYPEMINQUERY:
       case OP_TYPEUPTO:  
       case OP_TYPEMINUPTO:  
       case OP_TYPEEXACT:  
1569        case OP_TYPEPOSSTAR:        case OP_TYPEPOSSTAR:
1570        case OP_TYPEPOSPLUS:        case OP_TYPEPOSPLUS:
1571        case OP_TYPEPOSQUERY:        case OP_TYPEPOSQUERY:
       case OP_TYPEPOSUPTO:  
1572        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1573        break;        break;
1574        }  
1575          case OP_TYPEUPTO:
1576          case OP_TYPEMINUPTO:
1577          case OP_TYPEEXACT:
1578          case OP_TYPEPOSUPTO:
1579          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1580          break;
1581          }
1582    
1583      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
1584    
1585      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1586    
1587    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1588    a multi-byte character. The length in the table is a minimum, so we have to    a multi-byte character. The length in the table is a minimum, so we have to
1589    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
1590    
1591  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1592      if (utf8) switch(c)      if (utf8) switch(c)
1593        {        {
# Line 1357  for (;;) Line 1609  for (;;)
1609        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1610        break;        break;
1611        }        }
1612    #else
1613        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1614  #endif  #endif
1615      }      }
1616    }    }
# Line 1386  for (;;) Line 1640  for (;;)
1640    register int c = *code;    register int c = *code;
1641    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1642    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1643    
1644    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
1645    map. This includes negated single high-valued characters. The length in    map. This includes negated single high-valued characters. The length in
1646    the table is zero; the actual length is stored in the compiled code. */    the table is zero; the actual length is stored in the compiled code. */
1647    
1648    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1649    
1650    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1651    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1652    two bytes of parameters. */    two bytes of parameters. */
1653    
1654    else    else
# Line 1407  for (;;) Line 1661  for (;;)
1661        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
1662        case OP_TYPEQUERY:        case OP_TYPEQUERY:
1663        case OP_TYPEMINQUERY:        case OP_TYPEMINQUERY:
       case OP_TYPEUPTO:  
       case OP_TYPEMINUPTO:  
       case OP_TYPEEXACT:  
1664        case OP_TYPEPOSSTAR:        case OP_TYPEPOSSTAR:
1665        case OP_TYPEPOSPLUS:        case OP_TYPEPOSPLUS:
1666        case OP_TYPEPOSQUERY:        case OP_TYPEPOSQUERY:
       case OP_TYPEPOSUPTO:  
1667        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1668        break;        break;
1669        }  
1670          case OP_TYPEPOSUPTO:
1671          case OP_TYPEUPTO:
1672          case OP_TYPEMINUPTO:
1673          case OP_TYPEEXACT:
1674          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1675          break;
1676          }
1677    
1678      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
1679    
1680      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1681    
1682      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
1683      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
1684      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
1685    
1686  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1687      if (utf8) switch(c)      if (utf8) switch(c)
1688        {        {
# Line 1447  for (;;) Line 1704  for (;;)
1704        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1705        break;        break;
1706        }        }
1707    #else
1708        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1709  #endif  #endif
1710      }      }
1711    }    }
# Line 1462  for (;;) Line 1721  for (;;)
1721  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()
1722  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
1723  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1724  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
1725  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
1726    bracket whose current branch will already have been scanned.
1727    
1728  Arguments:  Arguments:
1729    code        points to start of search    code        points to start of search
# Line 1485  for (code = first_significant_code(code Line 1745  for (code = first_significant_code(code
1745    
1746    c = *code;    c = *code;
1747    
1748      /* Skip over forward assertions; the other assertions are skipped by
1749      first_significant_code() with a TRUE final argument. */
1750    
1751      if (c == OP_ASSERT)
1752        {
1753        do code += GET(code, 1); while (*code == OP_ALT);
1754        c = *code;
1755        continue;
1756        }
1757    
1758    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1759    
1760    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1761      {      {
1762      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1763      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1502  for (code = first_significant_code(code Line 1772  for (code = first_significant_code(code
1772      BOOL empty_branch;      BOOL empty_branch;
1773      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1774    
1775      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1776        empty branch, so just skip over the conditional, because it could be empty.
1777        Otherwise, scan the individual branches of the group. */
1778    
1779      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
1780        code += GET(code, 1);        code += GET(code, 1);
1781        else
1782          {
1783          empty_branch = FALSE;
1784          do
1785            {
1786            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1787              empty_branch = TRUE;
1788            code += GET(code, 1);
1789            }
1790          while (*code == OP_ALT);
1791          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1792        }        }
1793      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1794      c = *code;      c = *code;
1795      continue;      continue;
1796      }      }
# Line 1524  for (code = first_significant_code(code Line 1802  for (code = first_significant_code(code
1802      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
1803      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
1804      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1805      actual length is stored in the compiled code, so we must update "code"      actual length is stored in the compiled code, so we must update "code"
1806      here. */      here. */
1807    
1808  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1573  for (code = first_significant_code(code Line 1851  for (code = first_significant_code(code
1851      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1852      case OP_WORDCHAR:      case OP_WORDCHAR:
1853      case OP_ANY:      case OP_ANY:
1854        case OP_ALLANY:
1855      case OP_ANYBYTE:      case OP_ANYBYTE:
1856      case OP_CHAR:      case OP_CHAR:
1857      case OP_CHARNC:      case OP_CHARNC:
# Line 1591  for (code = first_significant_code(code Line 1870  for (code = first_significant_code(code
1870      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1871      return FALSE;      return FALSE;
1872    
1873        /* These are going to continue, as they may be empty, but we have to
1874        fudge the length for the \p and \P cases. */
1875    
1876        case OP_TYPESTAR:
1877        case OP_TYPEMINSTAR:
1878        case OP_TYPEPOSSTAR:
1879        case OP_TYPEQUERY:
1880        case OP_TYPEMINQUERY:
1881        case OP_TYPEPOSQUERY:
1882        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1883        break;
1884    
1885        /* Same for these */
1886    
1887        case OP_TYPEUPTO:
1888        case OP_TYPEMINUPTO:
1889        case OP_TYPEPOSUPTO:
1890        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1891        break;
1892    
1893      /* End of branch */      /* End of branch */
1894    
1895      case OP_KET:      case OP_KET:
# Line 1609  for (code = first_significant_code(code Line 1908  for (code = first_significant_code(code
1908      case OP_QUERY:      case OP_QUERY:
1909      case OP_MINQUERY:      case OP_MINQUERY:
1910      case OP_POSQUERY:      case OP_POSQUERY:
1911        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1912        break;
1913    
1914      case OP_UPTO:      case OP_UPTO:
1915      case OP_MINUPTO:      case OP_MINUPTO:
1916      case OP_POSUPTO:      case OP_POSUPTO:
1917      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1918      break;      break;
1919  #endif  #endif
1920      }      }
# Line 1660  return TRUE; Line 1962  return TRUE;
1962  *************************************************/  *************************************************/
1963    
1964  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1965  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
1966  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1967  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1968    
1969    Originally, this function only recognized a sequence of letters between the
1970    terminators, but it seems that Perl recognizes any sequence of characters,
1971    though of course unknown POSIX names are subsequently rejected. Perl gives an
1972    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1973    didn't consider this to be a POSIX class. Likewise for [:1234:].
1974    
1975    The problem in trying to be exactly like Perl is in the handling of escapes. We
1976    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1977    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1978    below handles the special case of \], but does not try to do any other escape
1979    processing. This makes it different from Perl for cases such as [:l\ower:]
1980    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1981    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1982    I think.
1983    
1984  Argument:  Arguments:
1985    ptr      pointer to the initial [    ptr      pointer to the initial [
1986    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1987    
1988  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1989  */  */
1990    
1991  static BOOL  static BOOL
1992  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1993  {  {
1994  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1995  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1996  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1997    {    {
1998    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
1999    return TRUE;      {
2000        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2001        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2002          {
2003          *endptr = ptr;
2004          return TRUE;
2005          }
2006        }
2007    }    }
2008  return FALSE;  return FALSE;
2009  }  }
# Line 1707  Returns:     a value representing the na Line 2028  Returns:     a value representing the na
2028  static int  static int
2029  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2030  {  {
2031    const char *pn = posix_names;
2032  register int yield = 0;  register int yield = 0;
2033  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2034    {    {
2035    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2036      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2037      pn += posix_name_lengths[yield] + 1;
2038    yield++;    yield++;
2039    }    }
2040  return -1;  return -1;
# Line 1726  return -1; Line 2049  return -1;
2049  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2050  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2051  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
2052  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
2053  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
2054  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
2055  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
2056  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2057    OP_END.
2058    
2059  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2060  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 1753  adjust_recurse(uschar *group, int adjust Line 2077  adjust_recurse(uschar *group, int adjust
2077    uschar *save_hwm)    uschar *save_hwm)
2078  {  {
2079  uschar *ptr = group;  uschar *ptr = group;
2080    
2081  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2082    {    {
2083    int offset;    int offset;
# Line 1864  get_othercase_range(unsigned int *cptr, Line 2189  get_othercase_range(unsigned int *cptr,
2189  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2190    
2191  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2192    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2193    
2194  if (c > d) return FALSE;  if (c > d) return FALSE;
2195    
# Line 1873  next = othercase + 1; Line 2198  next = othercase + 1;
2198    
2199  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2200    {    {
2201    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2202    next++;    next++;
2203    }    }
2204    
# Line 1919  if ((options & PCRE_EXTENDED) != 0) Line 2244  if ((options & PCRE_EXTENDED) != 0)
2244    for (;;)    for (;;)
2245      {      {
2246      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2247      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2248        {        {
2249        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2250          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1931  if ((options & PCRE_EXTENDED) != 0) Line 2256  if ((options & PCRE_EXTENDED) != 0)
2256  /* If the next item is one that we can handle, get its value. A non-negative  /* If the next item is one that we can handle, get its value. A non-negative
2257  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2258    
2259  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2260    {    {
2261    int temperrorcode = 0;    int temperrorcode = 0;
2262    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1956  if ((options & PCRE_EXTENDED) != 0) Line 2281  if ((options & PCRE_EXTENDED) != 0)
2281    for (;;)    for (;;)
2282      {      {
2283      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2284      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2285        {        {
2286        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2287          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1967  if ((options & PCRE_EXTENDED) != 0) Line 2292  if ((options & PCRE_EXTENDED) != 0)
2292    
2293  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2294    
2295  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2296    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2297        return FALSE;
2298    
2299  /* Now compare the next item with the previous opcode. If the previous is a  /* Now compare the next item with the previous opcode. If the previous is a
2300  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 1983  if (next >= 0) switch(op_code) Line 2309  if (next >= 0) switch(op_code)
2309    case OP_CHAR:    case OP_CHAR:
2310  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2311    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2312    #else
2313      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2314  #endif  #endif
2315    return item != next;    return item != next;
2316    
# Line 2001  if (next >= 0) switch(op_code) Line 2329  if (next >= 0) switch(op_code)
2329      unsigned int othercase;      unsigned int othercase;
2330      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2331  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2332      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2333  #else  #else
2334      othercase = NOTACHAR;      othercase = NOTACHAR;
2335  #endif  #endif
# Line 2014  if (next >= 0) switch(op_code) Line 2342  if (next >= 0) switch(op_code)
2342    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2343    
2344    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2345    if (item == next) return TRUE;    if (item == next) return TRUE;
2346    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2347  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2023  if (next >= 0) switch(op_code) Line 2350  if (next >= 0) switch(op_code)
2350      unsigned int othercase;      unsigned int othercase;
2351      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2352  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2353      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2354  #else  #else
2355      othercase = NOTACHAR;      othercase = NOTACHAR;
2356  #endif  #endif
# Line 2277  uschar classbits[32]; Line 2604  uschar classbits[32];
2604  BOOL class_utf8;  BOOL class_utf8;
2605  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2606  uschar *class_utf8data;  uschar *class_utf8data;
2607    uschar *class_utf8data_base;
2608  uschar utf8_char[6];  uschar utf8_char[6];
2609  #else  #else
2610  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2316  req_caseopt = ((options & PCRE_CASELESS) Line 2644  req_caseopt = ((options & PCRE_CASELESS)
2644  for (;; ptr++)  for (;; ptr++)
2645    {    {
2646    BOOL negate_class;    BOOL negate_class;
2647      BOOL should_flip_negation;
2648    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2649    BOOL is_quantifier;    BOOL is_quantifier;
2650    BOOL is_recurse;    BOOL is_recurse;
# Line 2403  for (;; ptr++) Line 2732  for (;; ptr++)
2732    
2733    if (inescq && c != 0)    if (inescq && c != 0)
2734      {      {
2735      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2736        {        {
2737        inescq = FALSE;        inescq = FALSE;
2738        ptr++;        ptr++;
# Line 2429  for (;; ptr++) Line 2758  for (;; ptr++)
2758    /* Fill in length of a previous callout, except when the next thing is    /* Fill in length of a previous callout, except when the next thing is
2759    a quantifier. */    a quantifier. */
2760    
2761    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2762      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2763        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2764    
2765    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2766         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2445  for (;; ptr++) Line 2775  for (;; ptr++)
2775    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2776      {      {
2777      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2778      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2779        {        {
2780        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2781          {          {
# Line 2470  for (;; ptr++) Line 2800  for (;; ptr++)
2800      {      {
2801      /* ===================================================================*/      /* ===================================================================*/
2802      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2803      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2804      case ')':      case CHAR_RIGHT_PARENTHESIS:
2805      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2806      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2807      *codeptr = code;      *codeptr = code;
# Line 2493  for (;; ptr++) Line 2823  for (;; ptr++)
2823      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2824      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2825    
2826      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2827      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2828        {        {
2829        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2502  for (;; ptr++) Line 2832  for (;; ptr++)
2832      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2833      break;      break;
2834    
2835      case '$':      case CHAR_DOLLAR_SIGN:
2836      previous = NULL;      previous = NULL;
2837      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2838      break;      break;
# Line 2510  for (;; ptr++) Line 2840  for (;; ptr++)
2840      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
2841      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2842    
2843      case '.':      case CHAR_DOT:
2844      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2845      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2846      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2847      previous = code;      previous = code;
2848      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2849      break;      break;
2850    
2851    
# Line 2530  for (;; ptr++) Line 2860  for (;; ptr++)
2860      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
2861      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2862      whether the bitmap is present, and whether this is a negated class or not.      whether the bitmap is present, and whether this is a negated class or not.
     */  
2863    
2864      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2865        default (Perl) mode, it is treated as a data character. */
2866    
2867        case CHAR_RIGHT_SQUARE_BRACKET:
2868        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2869          {
2870          *errorcodeptr = ERR64;
2871          goto FAILED;
2872          }
2873        goto NORMAL_CHAR;
2874    
2875        case CHAR_LEFT_SQUARE_BRACKET:
2876      previous = code;      previous = code;
2877    
2878      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
2879      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. */
2880    
2881      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2882          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2883            check_posix_syntax(ptr, &tempptr))
2884        {        {
2885        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2886        goto FAILED;        goto FAILED;
2887        }        }
2888    
# Line 2553  for (;; ptr++) Line 2894  for (;; ptr++)
2894      for (;;)      for (;;)
2895        {        {
2896        c = *(++ptr);        c = *(++ptr);
2897        if (c == '\\')        if (c == CHAR_BACKSLASH)
2898          {          {
2899          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
2900            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
2901              else break;          else if (strncmp((const char *)ptr+1,
2902                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2903              ptr += 3;
2904            else
2905              break;
2906          }          }
2907        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2908          negate_class = TRUE;          negate_class = TRUE;
2909        else break;        else break;
2910        }        }
2911    
2912        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2913        an initial ']' is taken as a data character -- the code below handles
2914        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2915        [^] must match any character, so generate OP_ALLANY. */
2916    
2917        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2918            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2919          {
2920          *code++ = negate_class? OP_ALLANY : OP_FAIL;
2921          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2922          zerofirstbyte = firstbyte;
2923          break;
2924          }
2925    
2926        /* If a class contains a negative special such as \S, we need to flip the
2927        negation flag at the end, so that support for characters > 255 works
2928        correctly (they are all included in the class). */
2929    
2930        should_flip_negation = FALSE;
2931    
2932      /* 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
2933      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
2934      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2581  for (;; ptr++) Line 2946  for (;; ptr++)
2946  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2947      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2948      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2949        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2950  #endif  #endif
2951    
2952      /* 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 2596  for (;; ptr++) Line 2962  for (;; ptr++)
2962          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2963          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2964          }          }
2965    
2966          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2967          data and reset the pointer. This is so that very large classes that
2968          contain a zillion UTF-8 characters no longer overwrite the work space
2969          (which is on the stack). */
2970    
2971          if (lengthptr != NULL)
2972            {
2973            *lengthptr += class_utf8data - class_utf8data_base;
2974            class_utf8data = class_utf8data_base;
2975            }
2976    
2977  #endif  #endif
2978    
2979        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
2980    
2981        if (inescq)        if (inescq)
2982          {          {
2983          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
2984            {            {
2985            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
2986            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2617  for (;; ptr++) Line 2995  for (;; ptr++)
2995        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
2996        5.6 and 5.8 do. */        5.6 and 5.8 do. */
2997    
2998        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
2999            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3000            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3001          {          {
3002          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3003          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3004          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3005          uschar pbits[32];          uschar pbits[32];
3006    
3007          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3008            {            {
3009            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3010            goto FAILED;            goto FAILED;
3011            }            }
3012    
3013          ptr += 2;          ptr += 2;
3014          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3015            {            {
3016            local_negate = TRUE;            local_negate = TRUE;
3017              should_flip_negation = TRUE;  /* Note negative special */
3018            ptr++;            ptr++;
3019            }            }
3020    
# Line 2705  for (;; ptr++) Line 3084  for (;; ptr++)
3084        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
3085        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3086    
3087        if (c == '\\')        if (c == CHAR_BACKSLASH)
3088          {          {
3089          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3090          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3091    
3092          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = CHAR_BS;       /* \b is backspace in a class */
3093          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */
3094          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */          else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3095          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3096            {            {
3097            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3098              {              {
3099              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3100              }              }
3101            else inescq = TRUE;            else inescq = TRUE;
3102            continue;            continue;
3103            }            }
3104          else if (-c == ESC_E) continue;  /* Ignore orphan \E */          else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3105    
3106          if (c < 0)          if (c < 0)
3107            {            {
# Line 2738  for (;; ptr++) Line 3117  for (;; ptr++)
3117              continue;              continue;
3118    
3119              case ESC_D:              case ESC_D:
3120                should_flip_negation = TRUE;
3121              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3122              continue;              continue;
3123    
# Line 2746  for (;; ptr++) Line 3126  for (;; ptr++)
3126              continue;              continue;
3127    
3128              case ESC_W:              case ESC_W:
3129                should_flip_negation = TRUE;
3130              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3131              continue;              continue;
3132    
# Line 2755  for (;; ptr++) Line 3136  for (;; ptr++)
3136              continue;              continue;
3137    
3138              case ESC_S:              case ESC_S:
3139                should_flip_negation = TRUE;
3140              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3141              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3142              continue;              continue;
3143    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
3144              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
3145              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
3146              }              }
# Line 2941  for (;; ptr++) Line 3320  for (;; ptr++)
3320        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3321    
3322        CHECK_RANGE:        CHECK_RANGE:
3323        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3324          {          {
3325          inescq = FALSE;          inescq = FALSE;
3326          ptr += 2;          ptr += 2;
# Line 2949  for (;; ptr++) Line 3328  for (;; ptr++)
3328    
3329        oldptr = ptr;        oldptr = ptr;
3330    
3331        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3332    
3333          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3334    
3335          /* Check for range */
3336    
3337          if (!inescq && ptr[1] == CHAR_MINUS)
3338          {          {
3339          int d;          int d;
3340          ptr += 2;          ptr += 2;
3341          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3342    
3343          /* If we hit \Q (not followed by \E) at this point, go into escaped          /* If we hit \Q (not followed by \E) at this point, go into escaped
3344          mode. */          mode. */
3345    
3346          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3347            {            {
3348            ptr += 2;            ptr += 2;
3349            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3350                { ptr += 2; continue; }
3351            inescq = TRUE;            inescq = TRUE;
3352            break;            break;
3353            }            }
3354    
3355          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3356            {            {
3357            ptr = oldptr;            ptr = oldptr;
3358            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2985  for (;; ptr++) Line 3371  for (;; ptr++)
3371          not any of the other escapes. Perl 5.6 treats a hyphen as a literal          not any of the other escapes. Perl 5.6 treats a hyphen as a literal
3372          in such circumstances. */          in such circumstances. */
3373    
3374          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3375            {            {
3376            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3377            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3378    
3379            /* \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
3380            special means the '-' was literal */            special means the '-' was literal */
3381    
3382            if (d < 0)            if (d < 0)
3383              {              {
3384              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3385              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3386              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3387                {                {
3388                ptr = oldptr;                ptr = oldptr;
3389                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3016  for (;; ptr++) Line 3402  for (;; ptr++)
3402    
3403          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3404    
3405            /* Remember \r or \n */
3406    
3407            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3408    
3409          /* 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
3410          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3411          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3134  for (;; ptr++) Line 3524  for (;; ptr++)
3524          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3525            {            {
3526            unsigned int othercase;            unsigned int othercase;
3527            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3528              {              {
3529              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3530              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3161  for (;; ptr++) Line 3551  for (;; ptr++)
3551    
3552      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3553    
3554      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3555    
3556      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3557        {        {
# Line 3169  for (;; ptr++) Line 3559  for (;; ptr++)
3559        goto FAILED;        goto FAILED;
3560        }        }
3561    
3562    
3563    /* This code has been disabled because it would mean that \s counts as
3564    an explicit \r or \n reference, and that's not really what is wanted. Now
3565    we set the flag only if there is a literal "\r" or "\n" in the class. */
3566    
3567    #if 0
3568        /* Remember whether \r or \n are in this class */
3569    
3570        if (negate_class)
3571          {
3572          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3573          }
3574        else
3575          {
3576          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3577          }
3578    #endif
3579    
3580    
3581      /* 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
3582      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
3583      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
3584      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3585      single-bytes only. This is an historical hangover. Maybe one day we can  
3586      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3587        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3588        operate on single-bytes only. This is an historical hangover. Maybe one day
3589        we can tidy these opcodes to handle multi-byte characters.
3590    
3591      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
3592      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 3184  for (;; ptr++) Line 3596  for (;; ptr++)
3596      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3597    
3598  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3599      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3600            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3601  #else  #else
3602      if (class_charcount == 1)      if (class_charcount == 1)
3603  #endif  #endif
# Line 3230  for (;; ptr++) Line 3640  for (;; ptr++)
3640      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3641    
3642      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3643      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3644      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
3645        the class, so any that were explicitly given as well can be ignored. If
3646        (when there are explicit characters > 255 that must be listed) there are no
3647        characters < 256, we can omit the bitmap in the actual compiled code. */
3648    
3649  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3650      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3651        {        {
3652        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3653        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3260  for (;; ptr++) Line 3673  for (;; ptr++)
3673        }        }
3674  #endif  #endif
3675    
3676      /* 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
3677      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
3678      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
3679      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3680    
3681        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3682      if (negate_class)      if (negate_class)
3683        {        {
       *code++ = OP_NCLASS;  
3684        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3685          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3686        }        }
3687      else      else
3688        {        {
       *code++ = OP_CLASS;  
3689        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3690        }        }
3691      code += 32;      code += 32;
# Line 3284  for (;; ptr++) Line 3696  for (;; ptr++)
3696      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3697      has been tested above. */      has been tested above. */
3698    
3699      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3700      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3701      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3702      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3703      goto REPEAT;      goto REPEAT;
3704    
3705      case '*':      case CHAR_ASTERISK:
3706      repeat_min = 0;      repeat_min = 0;
3707      repeat_max = -1;      repeat_max = -1;
3708      goto REPEAT;      goto REPEAT;
3709    
3710      case '+':      case CHAR_PLUS:
3711      repeat_min = 1;      repeat_min = 1;
3712      repeat_max = -1;      repeat_max = -1;
3713      goto REPEAT;      goto REPEAT;
3714    
3715      case '?':      case CHAR_QUESTION_MARK:
3716      repeat_min = 0;      repeat_min = 0;
3717      repeat_max = 1;      repeat_max = 1;
3718    
# Line 3335  for (;; ptr++) Line 3747  for (;; ptr++)
3747      but if PCRE_UNGREEDY is set, it works the other way round. We change the      but if PCRE_UNGREEDY is set, it works the other way round. We change the
3748      repeat type to the non-default. */      repeat type to the non-default. */
3749    
3750      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3751        {        {
3752        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3753        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3754        ptr++;        ptr++;
3755        }        }
3756      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3757        {        {
3758        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3759        ptr++;        ptr++;
# Line 3456  for (;; ptr++) Line 3868  for (;; ptr++)
3868    
3869        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3870    
3871          /*--------------------------------------------------------------------*/
3872          /* This code is obsolete from release 8.00; the restriction was finally
3873          removed: */
3874    
3875        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3876        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3877    
3878        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3879          /*--------------------------------------------------------------------*/
3880    
3881        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3882    
# Line 3606  for (;; ptr++) Line 4023  for (;; ptr++)
4023          goto END_REPEAT;          goto END_REPEAT;
4024          }          }
4025    
4026          /*--------------------------------------------------------------------*/
4027          /* This code is obsolete from release 8.00; the restriction was finally
4028          removed: */
4029    
4030        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4031        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4032    
4033        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4034          /*--------------------------------------------------------------------*/
4035    
4036        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4037          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3667  for (;; ptr++) Line 4089  for (;; ptr++)
4089    
4090        if (repeat_min == 0)        if (repeat_min == 0)
4091          {          {
4092          /* 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
4093          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4094    
4095          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4096          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4097          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4098          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4099          the start of the whole regex. Temporarily terminate the pattern while          **   }
4100          doing this. */  
4101            However, that fails when a group is referenced as a subroutine from
4102            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4103            so that it is skipped on execution. As we don't have a list of which
4104            groups are referenced, we cannot do this selectively.
4105    
4106            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4107            and do no more at this point. However, we do need to adjust any
4108            OP_RECURSE calls inside the group that refer to the group itself or any
4109            internal or forward referenced group, because the offset is from the
4110            start of the whole regex. Temporarily terminate the pattern while doing
4111            this. */
4112    
4113          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4114            {            {
4115            *code = OP_END;            *code = OP_END;
4116            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4117            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4118            code++;            code++;
4119              if (repeat_max == 0)
4120                {
4121                *previous++ = OP_SKIPZERO;
4122                goto END_REPEAT;
4123                }
4124            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4125            }            }
4126    
# Line 3883  for (;; ptr++) Line 4315  for (;; ptr++)
4315          }          }
4316        }        }
4317    
4318        /* If previous is OP_FAIL, it was generated by an empty class [] in
4319        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4320        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4321        error above. We can just ignore the repeat in JS case. */
4322    
4323        else if (*previous == OP_FAIL) goto END_REPEAT;
4324    
4325      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4326    
4327      else      else
# Line 3907  for (;; ptr++) Line 4346  for (;; ptr++)
4346      if (possessive_quantifier)      if (possessive_quantifier)
4347        {        {
4348        int len;        int len;
4349        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4350            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4351            tempcode += _pcre_OP_lengths[*tempcode] +
4352              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4353    
4354          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4355            {
4356          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4357    #ifdef SUPPORT_UTF8
4358            if (utf8 && tempcode[-1] >= 0xc0)
4359              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4360    #endif
4361            }
4362    
4363        len = code - tempcode;        len = code - tempcode;
4364        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4365          {          {
# Line 3955  for (;; ptr++) Line 4405  for (;; ptr++)
4405      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4406      parenthesis forms.  */      parenthesis forms.  */
4407    
4408      case '(':      case CHAR_LEFT_PARENTHESIS:
4409      newoptions = options;      newoptions = options;
4410      skipbytes = 0;      skipbytes = 0;
4411      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 3964  for (;; ptr++) Line 4414  for (;; ptr++)
4414    
4415      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4416    
4417      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4418        {        {
4419        int i, namelen;        int i, namelen;
4420          const char *vn = verbnames;
4421        const uschar *name = ++ptr;        const uschar *name = ++ptr;
4422        previous = NULL;        previous = NULL;
4423        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4424        if (*ptr == ':')        if (*ptr == CHAR_COLON)
4425          {          {
4426          *errorcodeptr = ERR59;   /* Not supported */          *errorcodeptr = ERR59;   /* Not supported */
4427          goto FAILED;          goto FAILED;
4428          }          }
4429        if (*ptr != ')')        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4430          {          {
4431          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4432          goto FAILED;          goto FAILED;
# Line 3984  for (;; ptr++) Line 4435  for (;; ptr++)
4435        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4436          {          {
4437          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4438              strncmp((char *)name, verbs[i].name, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4439            {            {
4440            *code = verbs[i].op;            *code = verbs[i].op;
4441            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4442            break;            break;
4443            }            }
4444            vn += verbs[i].len + 1;
4445          }          }
4446        if (i < verbcount) continue;        if (i < verbcount) continue;
4447        *errorcodeptr = ERR60;        *errorcodeptr = ERR60;
# Line 3999  for (;; ptr++) Line 4451  for (;; ptr++)
4451      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4452      appearance of any of them means that this is not a capturing group. */      appearance of any of them means that this is not a capturing group. */
4453    
4454      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4455        {        {
4456        int i, set, unset, namelen;        int i, set, unset, namelen;
4457        int *optset;        int *optset;
# Line 4008  for (;; ptr++) Line 4460  for (;; ptr++)
4460    
4461        switch (*(++ptr))        switch (*(++ptr))
4462          {          {
4463          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4464          ptr++;          ptr++;
4465          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4466          if (*ptr == 0)          if (*ptr == 0)
4467            {            {
4468            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4020  for (;; ptr++) Line 4472  for (;; ptr++)
4472    
4473    
4474          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4475          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4476          reset_bracount = TRUE;          reset_bracount = TRUE;
4477          /* Fall through */          /* Fall through */
4478    
4479          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4480          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4481          bravalue = OP_BRA;          bravalue = OP_BRA;
4482          ptr++;          ptr++;
4483          break;          break;
4484    
4485    
4486          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4487          case '(':          case CHAR_LEFT_PARENTHESIS:
4488          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4489    
4490          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
# Line 4052  for (;; ptr++) Line 4504  for (;; ptr++)
4504          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4505          including assertions, are processed. */          including assertions, are processed. */
4506    
4507          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4508                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4509            break;            break;
4510    
4511          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Most other conditions use OP_CREF (a couple change to OP_RREF
# Line 4064  for (;; ptr++) Line 4517  for (;; ptr++)
4517    
4518          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4519    
4520          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4521            {            {
4522            terminator = -1;            terminator = -1;
4523            ptr += 2;            ptr += 2;
# Line 4074  for (;; ptr++) Line 4527  for (;; ptr++)
4527          /* Check for a test for a named group's having been set, using the Perl          /* Check for a test for a named group's having been set, using the Perl
4528          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4529    
4530          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4531            {            {
4532            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4533            ptr++;            ptr++;
4534            }            }
4535          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4536            {            {
4537            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4538            ptr++;            ptr++;
4539            }            }
4540          else          else
4541            {            {
4542            terminator = 0;            terminator = 0;
4543            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4544            }            }
4545    
4546          /* We now expect to read a name; any thing else is an error */          /* We now expect to read a name; any thing else is an error */
# Line 4107  for (;; ptr++) Line 4560  for (;; ptr++)
4560            {            {
4561            if (recno >= 0)            if (recno >= 0)
4562              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4563                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4564            ptr++;            ptr++;
4565            }            }
4566          namelen = ptr - name;          namelen = ptr - name;
4567    
4568          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4569                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4570            {            {
4571            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4572            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4134  for (;; ptr++) Line 4588  for (;; ptr++)
4588              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4589              goto FAILED;              goto FAILED;
4590              }              }
4591            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
4592                cd->bracount - recno + 1 : recno +cd->bracount;
4593              if (recno <= 0 || recno > cd->final_bracount)
4594              {              {
4595              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4596              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4597              }              }
           else recno += cd->bracount;  
4598            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4599            break;            break;
4600            }            }
# Line 4168  for (;; ptr++) Line 4619  for (;; ptr++)
4619    
4620          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4621    
4622          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4623                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4624            {            {
4625            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
# Line 4189  for (;; ptr++) Line 4640  for (;; ptr++)
4640          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4641          specific group number. */          specific group number. */
4642    
4643          else if (*name == 'R')          else if (*name == CHAR_R)
4644            {            {
4645            recno = 0;            recno = 0;
4646            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4199  for (;; ptr++) Line 4650  for (;; ptr++)
4650                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4651                goto FAILED;                goto FAILED;
4652                }                }
4653              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4654              }              }
4655            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4656            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4209  for (;; ptr++) Line 4660  for (;; ptr++)
4660          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4661          false. */          false. */
4662    
4663          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4664            {            {
4665            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4666            skipbytes = 1;            skipbytes = 1;
4667            }            }
4668    
4669          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4670            in the second pass here, so final_bracount is set. */
4671    
4672          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4673            {            {
4674            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4675            }            }
# Line 4233  for (;; ptr++) Line 4685  for (;; ptr++)
4685    
4686    
4687          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4688          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4689          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4690          ptr++;          ptr++;
4691          break;          break;
4692    
4693    
4694          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4695          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
4696          ptr++;          ptr++;
4697          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4698            {            {
4699            *code++ = OP_FAIL;            *code++ = OP_FAIL;
4700            previous = NULL;            previous = NULL;
# Line 4253  for (;; ptr++) Line 4705  for (;; ptr++)
4705    
4706    
4707          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4708          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4709          switch (ptr[1])          switch (ptr[1])
4710            {            {
4711            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4712            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4713            ptr += 2;            ptr += 2;
4714            break;            break;
4715    
4716            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4717            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4718            ptr += 2;            ptr += 2;
4719            break;            break;
# Line 4276  for (;; ptr++) Line 4728  for (;; ptr++)
4728    
4729    
4730          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4731          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4732          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4733          ptr++;          ptr++;
4734          break;          break;
4735    
4736    
4737          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4738          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4739          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4740          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4741          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4742            {            {
4743            int n = 0;            int n = 0;
4744            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4745              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4746            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4747              {              {
4748              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4749              goto FAILED;              goto FAILED;
# Line 4311  for (;; ptr++) Line 4763  for (;; ptr++)
4763    
4764    
4765          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4766          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4767          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4768                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4769            {            {
4770            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4771            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4772            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4773            }            }
4774          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4775            {            {
4776            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4777            goto FAILED;            goto FAILED;
# Line 4328  for (;; ptr++) Line 4781  for (;; ptr++)
4781    
4782          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4783          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4784          case '\'':          case CHAR_APOSTROPHE:
4785            {            {
4786            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4787                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4788            name = ++ptr;            name = ++ptr;
4789    
4790            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4404  for (;; ptr++) Line 4858  for (;; ptr++)
4858    
4859    
4860          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4861          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4862          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4863          is_recurse = TRUE;          is_recurse = TRUE;
4864          /* Fall through */          /* Fall through */
4865    
4866          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4867          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4868          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4869            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4870            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4871    
4872          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4873          name = ++ptr;          name = ++ptr;
# Line 4423  for (;; ptr++) Line 4879  for (;; ptr++)
4879    
4880          if (lengthptr != NULL)          if (lengthptr != NULL)
4881            {            {
4882              if (namelen == 0)
4883                {
4884                *errorcodeptr = ERR62;
4885                goto FAILED;
4886                }
4887            if (*ptr != terminator)            if (*ptr != terminator)
4888              {              {
4889              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4436  for (;; ptr++) Line 4897  for (;; ptr++)
4897            recno = 0;            recno = 0;
4898            }            }
4899    
4900          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4901            first, and then check that we have reached the end of the name in the
4902            table. That way, if the name that is longer than any in the table,
4903            the comparison will fail without reading beyond the table entry. */
4904    
4905          else          else
4906            {            {
4907            slot = cd->name_table;            slot = cd->name_table;
4908            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4909              {              {
4910              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4911                    slot[2+namelen] == 0)
4912                  break;
4913              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4914              }              }
4915    
# Line 4452  for (;; ptr++) Line 4918  for (;; ptr++)
4918              recno = GET2(slot, 0);              recno = GET2(slot, 0);
4919              }              }
4920            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
4921                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
4922                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
4923              {              {
4924              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4468  for (;; ptr++) Line 4934  for (;; ptr++)
4934    
4935    
4936          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4937          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
4938          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
4939          /* Fall through */          /* Fall through */
4940    
4941    
4942          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4943          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
4944          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
4945          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
4946            {            {
4947            const uschar *called;            const uschar *called;
4948              terminator = CHAR_RIGHT_PARENTHESIS;
4949    
4950            if ((refsign = *ptr) == '+') ptr++;            /* Come here from the \g<...> and \g'...' code (Oniguruma
4951            else if (refsign == '-')            compatibility). However, the syntax has been checked to ensure that
4952              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
4953              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
4954              ever be taken. */
4955    
4956              HANDLE_NUMERICAL_RECURSION:
4957    
4958              if ((refsign = *ptr) == CHAR_PLUS)
4959                {
4960                ptr++;
4961                if ((digitab[*ptr] & ctype_digit) == 0)
4962                  {
4963                  *errorcodeptr = ERR63;
4964                  goto FAILED;
4965                  }
4966                }
4967              else if (refsign == CHAR_MINUS)
4968              {              {
4969              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
4970                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4490  for (;; ptr++) Line 4973  for (;; ptr++)
4973    
4974            recno = 0;            recno = 0;
4975            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4976              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
4977    
4978            if (*ptr != ')')            if (*ptr != terminator)
4979              {              {
4980              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
4981              goto FAILED;              goto FAILED;
4982              }              }
4983    
4984            if (refsign == '-')            if (refsign == CHAR_MINUS)
4985              {              {
4986              if (recno == 0)              if (recno == 0)
4987                {                {
# Line 4512  for (;; ptr++) Line 4995  for (;; ptr++)
4995                goto FAILED;                goto FAILED;
4996                }                }
4997              }              }
4998            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
4999              {              {
5000              if (recno == 0)              if (recno == 0)
5001                {                {
# Line 4545  for (;; ptr++) Line 5028  for (;; ptr++)
5028    
5029              if (called == NULL)              if (called == NULL)
5030                {                {
5031                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5032                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5033                  {                  {
5034                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5035                  goto FAILED;                  goto FAILED;
# Line 4598  for (;; ptr++) Line 5081  for (;; ptr++)
5081          set = unset = 0;          set = unset = 0;
5082          optset = &set;          optset = &set;
5083    
5084          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5085            {            {
5086            switch (*ptr++)            switch (*ptr++)
5087              {              {
5088              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5089    
5090              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5091              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5092              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5093              break;              break;
5094    
5095              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5096              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5097              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5098              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5099              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5100              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5101    
5102              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5103                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4645  for (;; ptr++) Line 5128  for (;; ptr++)
5128          both phases.          both phases.
5129    
5130          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, compile code to change the ims
5131          options if this setting actually changes any of them. We also pass the          options if this setting actually changes any of them, and reset the
5132          new setting back so that it can be put at the start of any following          greedy defaults and the case value for firstbyte and reqbyte. */
         branches, and when this group ends (if we are in a group), a resetting  
         item can be compiled. */  
5133    
5134          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5135            {            {
5136            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5137                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5138              {              {
5139              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5140              }              }
5141           else           else
5142              {              {
# Line 4665  for (;; ptr++) Line 5145  for (;; ptr++)
5145                *code++ = OP_OPT;                *code++ = OP_OPT;
5146                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5147                }                }
   
             /* Change options at this level, and pass them back for use  
             in subsequent branches. Reset the greedy defaults and the case  
             value for firstbyte and reqbyte. */  
   
             *optionsptr = options = newoptions;  
5148              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5149              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5150              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5151              }              }
5152    
5153              /* Change options at this level, and pass them back for use
5154              in subsequent branches. When not at the start of the pattern, this
5155              information is also necessary so that a resetting item can be
5156              compiled at the end of a group (if we are in a group). */
5157    
5158              *optionsptr = options = newoptions;
5159            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5160            continue;              /* It is complete */            continue;              /* It is complete */
5161            }            }
# Line 4791  for (;; ptr++) Line 5271  for (;; ptr++)
5271    
5272      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5273    
5274      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5275        {        {
5276        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5277        goto FAILED;        goto FAILED;
5278        }        }
5279    
5280      /* In the pre-compile phase, update the length by the length of the group,      /* In the pre-compile phase, update the length by the length of the group,
5281      less the brackets at either end. Then reduce the compiled code to just a      less the brackets at either end. Then reduce the compiled code to just a
5282      set of non-capturing brackets so that it doesn't use much memory if it is      set of non-capturing brackets so that it doesn't use much memory if it is
5283      duplicated by a quantifier.*/      duplicated by a quantifier.*/
5284    
# Line 4810  for (;; ptr++) Line 5290  for (;; ptr++)
5290          goto FAILED;          goto FAILED;
5291          }          }
5292        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5293        *code++ = OP_BRA;        *code++ = OP_BRA;
5294        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5295        *code++ = OP_KET;        *code++ = OP_KET;
5296        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 4889  for (;; ptr++) Line 5369  for (;; ptr++)
5369      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
5370      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5371    
5372      case '\\':      case CHAR_BACKSLASH:
5373      tempptr = ptr;      tempptr = ptr;
5374      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5375      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 4898  for (;; ptr++) Line 5378  for (;; ptr++)
5378        {        {
5379        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5380          {          {
5381          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5382            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5383                else inescq = TRUE;
5384          continue;          continue;
5385          }          }
5386    
# Line 4916  for (;; ptr++) Line 5397  for (;; ptr++)
5397        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5398        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5399    
5400          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5401          is a subroutine call by number (Oniguruma syntax). In fact, the value
5402          -ESC_g is returned only for these cases. So we don't need to check for <
5403          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5404          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5405          that is a synonym for a named back reference). */
5406    
5407          if (-c == ESC_g)
5408            {
5409            const uschar *p;
5410            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5411            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5412              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5413    
5414            /* These two statements stop the compiler for warning about possibly
5415            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5416            fact, because we actually check for a number below, the paths that
5417            would actually be in error are never taken. */
5418    
5419            skipbytes = 0;
5420            reset_bracount = FALSE;
5421    
5422            /* Test for a name */
5423    
5424            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5425              {
5426              BOOL isnumber = TRUE;
5427              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5428                {
5429                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5430                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5431                }
5432              if (*p != terminator)
5433                {
5434                *errorcodeptr = ERR57;
5435                break;
5436                }
5437              if (isnumber)
5438                {
5439                ptr++;
5440                goto HANDLE_NUMERICAL_RECURSION;
5441                }
5442              is_recurse = TRUE;
5443              goto NAMED_REF_OR_RECURSE;
5444              }
5445    
5446            /* Test a signed number in angle brackets or quotes. */
5447    
5448            p = ptr + 2;
5449            while ((digitab[*p] & ctype_digit) != 0) p++;
5450            if (*p != terminator)
5451              {
5452              *errorcodeptr = ERR57;
5453              break;
5454              }
5455            ptr++;
5456            goto HANDLE_NUMERICAL_RECURSION;
5457            }
5458    
5459        /* \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).
5460        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5461    
5462        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5463              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5464          {          {
5465          is_recurse = FALSE;          is_recurse = FALSE;
5466          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5467              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5468              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5469          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5470          }          }
5471    
# Line 5023  for (;; ptr++) Line 5566  for (;; ptr++)
5566      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5567      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5568    
5569        /* Remember if \r or \n were seen */
5570    
5571        if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5572          cd->external_flags |= PCRE_HASCRORLF;
5573    
5574      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5575      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.
5576      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 5265  for (;;) Line 5813  for (;;)
5813    compile a resetting op-code following, except at the very end of the pattern.    compile a resetting op-code following, except at the very end of the pattern.
5814    Return leaving the pointer at the terminating char. */    Return leaving the pointer at the terminating char. */
5815    
5816    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
5817      {      {
5818      if (lengthptr == NULL)      if (lengthptr == NULL)
5819        {        {
# Line 5288  for (;;) Line 5836  for (;;)
5836    
5837      /* Resetting option if needed */      /* Resetting option if needed */
5838    
5839      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
5840        {        {
5841        *code++ = OP_OPT;        *code++ = OP_OPT;
5842        *code++ = oldims;        *code++ = oldims;
# Line 5417  do { Line 5965  do {
5965       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
5966       }       }
5967    
5968     /* .* is not anchored unless DOTALL is set and it isn't in brackets that     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
5969     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
5970    
5971     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
5972               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
5973       {       {
5974       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
5975           return FALSE;
5976       }       }
5977    
5978     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5470  do { Line 6018  do {
6018       NULL, 0, FALSE);       NULL, 0, FALSE);
6019     register int op = *scode;     register int op = *scode;
6020    
6021       /* If we are at the start of a conditional assertion group, *both* the
6022       conditional assertion *and* what follows the condition must satisfy the test
6023       for start of line. Other kinds of condition fail. Note that there may be an
6024       auto-callout at the start of a condition. */
6025    
6026       if (op == OP_COND)
6027         {
6028         scode += 1 + LINK_SIZE;
6029         if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6030         switch (*scode)
6031           {
6032           case OP_CREF:
6033           case OP_RREF:
6034           case OP_DEF:
6035           return FALSE;
6036    
6037           default:     /* Assertion */
6038           if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6039           do scode += GET(scode, 1); while (*scode == OP_ALT);
6040           scode += 1 + LINK_SIZE;
6041           break;
6042           }
6043         scode = first_significant_code(scode, NULL, 0, FALSE);
6044         op = *scode;
6045         }
6046    
6047     /* Non-capturing brackets */     /* Non-capturing brackets */
6048    
6049     if (op == OP_BRA)     if (op == OP_BRA)
# Line 5488  do { Line 6062  do {
6062    
6063     /* Other brackets */     /* Other brackets */
6064    
6065     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
6066       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
6067         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6068         }
6069    
6070     /* .* means "start at start or after \n" if it isn't in brackets that     /* .* means "start at start or after \n" if it isn't in brackets that
6071     may be referenced. */     may be referenced. */
# Line 5606  Returns:        pointer to compiled data Line 6182  Returns:        pointer to compiled data
6182                  with errorptr and erroroffset set                  with errorptr and erroroffset set
6183  */  */
6184    
6185  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6186  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
6187    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
6188  {  {
# Line 5614  return pcre_compile2(pattern, options, N Line 6190  return pcre_compile2(pattern, options, N
6190  }  }
6191    
6192    
6193  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6194  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
6195    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
6196  {  {
# Line 5622  real_pcre *re; Line 6198  real_pcre *re;
6198  int length = 1;  /* For final END opcode */  int length = 1;  /* For final END opcode */
6199  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6200  int errorcode = 0;  int errorcode = 0;
6201    int skipatstart = 0;
6202  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
6203  BOOL utf8;  BOOL utf8;
6204  #endif  #endif
# Line 5640  to fill in forward references to subpatt Line 6217  to fill in forward references to subpatt
6217    
6218  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
6219    
   
6220  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
6221    
6222  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;
# Line 5668  if (erroroffset == NULL) Line 6244  if (erroroffset == NULL)
6244    
6245  *erroroffset = 0;  *erroroffset = 0;
6246    
6247    /* Set up pointers to the individual character tables */
6248    
6249    if (tables == NULL) tables = _pcre_default_tables;
6250    cd->lcc = tables + lcc_offset;
6251    cd->fcc = tables + fcc_offset;
6252    cd->cbits = tables + cbits_offset;
6253    cd->ctypes = tables + ctypes_offset;
6254    
6255    /* Check that all undefined public option bits are zero */
6256    
6257    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6258      {
6259      errorcode = ERR17;
6260      goto PCRE_EARLY_ERROR_RETURN;
6261      }
6262    
6263    /* Check for global one-time settings at the start of the pattern, and remember
6264    the offset for later. */
6265    
6266    while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6267           ptr[skipatstart+1] == CHAR_ASTERISK)
6268      {
6269      int newnl = 0;
6270      int newbsr = 0;
6271    
6272      if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6273        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6274    
6275      if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6276        { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6277      else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
6278        { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6279      else if (strncmp((char *)(ptr+skipatstart+2), STRING_CRLF_RIGHTPAR, 5)  == 0)
6280        { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6281      else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANY_RIGHTPAR, 4) == 0)
6282        { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6283      else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANYCRLF_RIGHTPAR, 8) == 0)
6284        { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6285    
6286      else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
6287        { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6288      else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
6289        { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6290    
6291      if (newnl != 0)
6292        options = (options & ~PCRE_NEWLINE_BITS) | newnl;
6293      else if (newbsr != 0)
6294        options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
6295      else break;
6296      }
6297    
6298  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6299    
6300  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 5686  if ((options & PCRE_UTF8) != 0) Line 6313  if ((options & PCRE_UTF8) != 0)
6313    }    }
6314  #endif  #endif
6315    
6316  if ((options & ~PUBLIC_OPTIONS) != 0)  /* Check validity of \R options. */
6317    
6318    switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6319    {    {
6320    errorcode = ERR17;    case 0:
6321    goto PCRE_EARLY_ERROR_RETURN;    case PCRE_BSR_ANYCRLF:
6322      case PCRE_BSR_UNICODE:
6323      break;
6324      default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
6325    }    }
6326    
 /* Set up pointers to the individual character tables */  
   
 if (tables == NULL) tables = _pcre_default_tables;  
 cd->lcc = tables + lcc_offset;  
 cd->fcc = tables + fcc_offset;  
 cd->cbits = tables + cbits_offset;  
 cd->ctypes = tables + ctypes_offset;  
   
6327  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
6328  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
6329  "anycrlf". */  "anycrlf". */
6330    
6331  switch (options & (PCRE_NEWLINE_CRLF | PCRE_NEWLINE_ANY))  switch (options & PCRE_NEWLINE_BITS)
6332    {    {
6333    case 0: newline = NEWLINE; break;   /* Compile-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6334    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6335    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6336    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6337         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6338    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6339    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6340    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
# Line 5759  to compile parts of the pattern into; th Line 6383  to compile parts of the pattern into; th
6383  no longer needed, so hopefully this workspace will never overflow, though there  no longer needed, so hopefully this workspace will never overflow, though there
6384  is a test for its doing so. */  is a test for its doing so. */
6385    
6386  cd->bracount = 0;  cd->bracount = cd->final_bracount = 0;
6387  cd->names_found = 0;  cd->names_found = 0;
6388  cd->name_entry_size = 0;  cd->name_entry_size = 0;
6389  cd->name_table = NULL;  cd->name_table = NULL;
# Line 5769  cd->hwm = cworkspace; Line 6393  cd->hwm = cworkspace;
6393  cd->start_pattern = (const uschar *)pattern;  cd->start_pattern = (const uschar *)pattern;
6394  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));
6395  cd->req_varyopt = 0;  cd->req_varyopt = 0;
 cd->nopartial = FALSE;  
6396  cd->external_options = options;  cd->external_options = options;
6397    cd->external_flags = 0;
6398    
6399  /* 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
6400  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 5778  been put into the cd block so that they Line 6402  been put into the cd block so that they
6402  found within the regex right at the beginning. Bringing initial option settings  found within the regex right at the beginning. Bringing initial option settings
6403  outside can help speed up starting point checks. */  outside can help speed up starting point checks. */
6404    
6405    ptr += skipatstart;
6406  code = cworkspace;  code = cworkspace;
6407  *code = OP_BRA;  *code = OP_BRA;
6408  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,
# Line 5808  if (re == NULL) Line 6433  if (re == NULL)
6433    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
6434    }    }
6435    
6436  /* 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
6437  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
6438  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
6439  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
6440    pointers. */
6441    
6442  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
6443  re->size = size;  re->size = size;
6444  re->options = cd->external_options;  re->options = cd->external_options;
6445    re->flags = cd->external_flags;
6446  re->dummy1 = 0;  re->dummy1 = 0;
6447  re->first_byte = 0;  re->first_byte = 0;
6448  re->req_byte = 0;  re->req_byte = 0;
# Line 5833  field. Reset the bracket count and the n Line 6460  field. Reset the bracket count and the n
6460  field; this time it's used for remembering forward references to subpatterns.  field; this time it's used for remembering forward references to subpatterns.
6461  */  */
6462    
6463    cd->final_bracount = cd->bracount;  /* Save for checking forward references */
6464  cd->bracount = 0;  cd->bracount = 0;
6465  cd->names_found = 0;  cd->names_found = 0;
6466  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 5840  codestart = cd->name_table + re->name_en Line 6468  codestart = cd->name_table + re->name_en
6468  cd->start_code = codestart;  cd->start_code = codestart;
6469  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6470  cd->req_varyopt = 0;  cd->req_varyopt = 0;
 cd->nopartial = FALSE;  
6471  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6472    
6473  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6474  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
6475  of the function here. */  of the function here. */
6476    
6477  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern + skipatstart;
6478  code = (uschar *)codestart;  code = (uschar *)codestart;
6479  *code = OP_BRA;  *code = OP_BRA;
6480  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,
6481    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);
6482  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
6483  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
6484    re->flags = cd->external_flags;
6485    
 if (cd->nopartial) re->options |= PCRE_NOPARTIAL;  
6486  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */
6487    
6488  /* 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. */
# Line 5898  if (errorcode != 0) Line 6525  if (errorcode != 0)
6525    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
6526    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = ptr - (const uschar *)pattern;
6527    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
6528    *errorptr = error_texts[errorcode];    *errorptr = find_error_text(errorcode);
6529    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
6530    return NULL;    return NULL;
6531    }    }
# Line 5927  if ((re->options & PCRE_ANCHORED) == 0) Line 6554  if ((re->options & PCRE_ANCHORED) == 0)
6554        int ch = firstbyte & 255;        int ch = firstbyte & 255;
6555        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&
6556           cd->fcc[ch] == ch)? ch : firstbyte;           cd->fcc[ch] == ch)? ch : firstbyte;
6557        re->options |= PCRE_FIRSTSET;        re->flags |= PCRE_FIRSTSET;
6558        }        }
6559      else if (is_startline(codestart, 0, cd->backref_map))      else if (is_startline(codestart, 0, cd->backref_map))
6560        re->options |= PCRE_STARTLINE;        re->flags |= PCRE_STARTLINE;
6561      }      }
6562    }    }
6563    
# Line 5944  if (reqbyte >= 0 && Line 6571  if (reqbyte >= 0 &&
6571    int ch = reqbyte & 255;    int ch = reqbyte & 255;
6572    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&
6573      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;
6574    re->options |= PCRE_REQCHSET;    re->flags |= PCRE_REQCHSET;
6575    }    }
6576    
6577  /* 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 5955  case when building a production library. Line 6582  case when building a production library.
6582  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6583    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6584    
6585  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 " : "");  
   }  
6586    
6587  if ((re->options & PCRE_FIRSTSET) != 0)  if ((re->flags & PCRE_FIRSTSET) != 0)
6588    {    {
6589    int ch = re->first_byte & 255;    int ch = re->first_byte & 255;
6590    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?
# Line 5978  if ((re->options & PCRE_FIRSTSET) != 0) Line 6593  if ((re->options & PCRE_FIRSTSET) != 0)
6593      else printf("First char = \\x%02x%s\n", ch, caseless);      else printf("First char = \\x%02x%s\n", ch, caseless);
6594    }    }
6595    
6596  if ((re->options & PCRE_REQCHSET) != 0)  if ((re->flags & PCRE_REQCHSET) != 0)
6597    {    {
6598    int ch = re->req_byte & 255;    int ch = re->req_byte & 255;
6599    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?
# Line 5995  was compiled can be seen. */ Line 6610  was compiled can be seen. */
6610  if (code - codestart > length)  if (code - codestart > length)
6611    {    {
6612    (pcre_free)(re);    (pcre_free)(re);
6613    *errorptr = error_texts[ERR23];    *errorptr = find_error_text(ERR23);
6614    *erroroffset = ptr - (uschar *)pattern;    *erroroffset = ptr - (uschar *)pattern;
6615    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6616    return NULL;    return NULL;

Legend:
Removed from v.218  
changed lines
  Added in v.426

  ViewVC Help
Powered by ViewVC 1.1.5