/[pcre]/code/branches/pcre16/pcre_compile.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_compile.c

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

revision 230 by ph10, Mon Sep 10 13:23:56 2007 UTC revision 406 by ph10, Mon Mar 23 12:05:43 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;
# 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 912  be terminated by '>' because that is che Line 1021  be terminated by '>' because that is che
1021    
1022  Arguments:  Arguments:
1023    ptr          current position in the pattern    ptr          current position in the pattern
1024    count        current count of capturing parens so far encountered    cd           compile background data
1025    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1026    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1027    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
# Line 921  Returns:       the number of the named s Line 1030  Returns:       the number of the named s
1030  */  */
1031    
1032  static int  static int
1033  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens(const uschar *ptr, compile_data *cd, const uschar *name, int lorn,
1034    BOOL xmode)    BOOL xmode)
1035  {  {
1036  const uschar *thisname;  const uschar *thisname;
1037    int count = cd->bracount;
1038    
1039  for (; *ptr != 0; ptr++)  for (; *ptr != 0; ptr++)
1040    {    {
# Line 932  for (; *ptr != 0; ptr++) Line 1042  for (; *ptr != 0; ptr++)
1042    
1043    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1044    
1045    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1046      {      {
1047      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) return -1;
1048      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1049        {        {
1050        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1051        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1052        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1053        }        }
1054      continue;      continue;
1055      }      }
1056    
1057    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1058      are handled for real. If the first character is '^', skip it. Also, if the
1059      first few characters (either before or after ^) are \Q\E or \E we skip them
1060      too. This makes for compatibility with Perl. Note the use of STR macros to
1061      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1062    
1063    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1064      {      {
1065      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1066        for (;;)
1067          {
1068          int c = *(++ptr);
1069          if (c == CHAR_BACKSLASH)
1070            {
1071            if (ptr[1] == CHAR_E)
1072              ptr++;
1073            else if (strncmp((const char *)ptr+1,
1074                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1075              ptr += 3;
1076            else
1077              break;
1078            }
1079          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
1080            negate_class = TRUE;
1081          else break;
1082          }
1083    
1084        /* If the next character is ']', it is a data character that must be
1085        skipped, except in JavaScript compatibility mode. */
1086    
1087        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1088            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1089          ptr++;
1090    
1091        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1092        {        {
1093        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1094        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1095          {          {
1096          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) return -1;
1097          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1098            {            {
1099            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1100            if (*ptr == 0) return -1;            if (*ptr == 0) return -1;
1101            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1102            }            }
1103          continue;          continue;
1104          }          }
# Line 968  for (; *ptr != 0; ptr++) Line 1108  for (; *ptr != 0; ptr++)
1108    
1109    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1110    
1111    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1112      {      {
1113      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1114      if (*ptr == 0) return -1;      if (*ptr == 0) return -1;
1115      continue;      continue;
1116      }      }
1117    
1118    /* An opening parens must now be a real metacharacter */    /* An opening parens must now be a real metacharacter */
1119    
1120    if (*ptr != '(') continue;    if (*ptr != CHAR_LEFT_PARENTHESIS) continue;
1121    if (ptr[1] != '?' && ptr[1] != '*')    if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1122      {      {
1123      count++;      count++;
1124      if (name == NULL && count == lorn) return count;      if (name == NULL && count == lorn) return count;
# Line 986  for (; *ptr != 0; ptr++) Line 1126  for (; *ptr != 0; ptr++)
1126      }      }
1127    
1128    ptr += 2;    ptr += 2;
1129    if (*ptr == 'P') ptr++;                      /* Allow optional P */    if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1130    
1131    /* We have to disambiguate (?<! and (?<= from (?<name> */    /* We have to disambiguate (?<! and (?<= from (?<name> */
1132    
1133    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&    if ((*ptr != CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_EXCLAMATION_MARK ||
1134         *ptr != '\'')        ptr[1] == CHAR_EQUALS_SIGN) && *ptr != CHAR_APOSTROPHE)
1135      continue;      continue;
1136    
1137    count++;    count++;
1138    
1139    if (name == NULL && count == lorn) return count;    if (name == NULL && count == lorn) return count;
1140    term = *ptr++;    term = *ptr++;
1141    if (term == '<') term = '>';    if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1142    thisname = ptr;    thisname = ptr;
1143    while (*ptr != term) ptr++;    while (*ptr != term) ptr++;
1144    if (name != NULL && lorn == ptr - thisname &&    if (name != NULL && lorn == ptr - thisname &&
# Line 1212  for (;;) Line 1352  for (;;)
1352      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1353      case OP_WORDCHAR:      case OP_WORDCHAR:
1354      case OP_ANY:      case OP_ANY:
1355        case OP_ALLANY:
1356      branchlength++;      branchlength++;
1357      cc++;      cc++;
1358      break;      break;
# Line 1360  for (;;) Line 1501  for (;;)
1501        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1502        break;        break;
1503        }        }
1504    #else
1505        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1506  #endif  #endif
1507      }      }
1508    }    }
# Line 1453  for (;;) Line 1596  for (;;)
1596        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1597        break;        break;
1598        }        }
1599    #else
1600        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1601  #endif  #endif
1602      }      }
1603    }    }
# Line 1468  for (;;) Line 1613  for (;;)
1613  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()
1614  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
1615  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1616  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
1617  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
1618    bracket whose current branch will already have been scanned.
1619    
1620  Arguments:  Arguments:
1621    code        points to start of search    code        points to start of search
# Line 1491  for (code = first_significant_code(code Line 1637  for (code = first_significant_code(code
1637    
1638    c = *code;    c = *code;
1639    
1640      /* Skip over forward assertions; the other assertions are skipped by
1641      first_significant_code() with a TRUE final argument. */
1642    
1643      if (c == OP_ASSERT)
1644        {
1645        do code += GET(code, 1); while (*code == OP_ALT);
1646        c = *code;
1647        continue;
1648        }
1649    
1650    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1651    
1652    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1653      {      {
1654      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1655      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1508  for (code = first_significant_code(code Line 1664  for (code = first_significant_code(code
1664      BOOL empty_branch;      BOOL empty_branch;
1665      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1666    
1667      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1668        empty branch, so just skip over the conditional, because it could be empty.
1669        Otherwise, scan the individual branches of the group. */
1670    
1671      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;  
1672        code += GET(code, 1);        code += GET(code, 1);
1673        else
1674          {
1675          empty_branch = FALSE;
1676          do
1677            {
1678            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1679              empty_branch = TRUE;
1680            code += GET(code, 1);
1681            }
1682          while (*code == OP_ALT);
1683          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1684        }        }
1685      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1686      c = *code;      c = *code;
1687      continue;      continue;
1688      }      }
# Line 1579  for (code = first_significant_code(code Line 1743  for (code = first_significant_code(code
1743      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1744      case OP_WORDCHAR:      case OP_WORDCHAR:
1745      case OP_ANY:      case OP_ANY:
1746        case OP_ALLANY:
1747      case OP_ANYBYTE:      case OP_ANYBYTE:
1748      case OP_CHAR:      case OP_CHAR:
1749      case OP_CHARNC:      case OP_CHARNC:
# Line 1686  return TRUE; Line 1851  return TRUE;
1851  *************************************************/  *************************************************/
1852    
1853  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1854  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
1855  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1856  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1857    
1858    Originally, this function only recognized a sequence of letters between the
1859    terminators, but it seems that Perl recognizes any sequence of characters,
1860    though of course unknown POSIX names are subsequently rejected. Perl gives an
1861    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1862    didn't consider this to be a POSIX class. Likewise for [:1234:].
1863    
1864    The problem in trying to be exactly like Perl is in the handling of escapes. We
1865    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1866    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1867    below handles the special case of \], but does not try to do any other escape
1868    processing. This makes it different from Perl for cases such as [:l\ower:]
1869    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1870    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1871    I think.
1872    
1873  Argument:  Arguments:
1874    ptr      pointer to the initial [    ptr      pointer to the initial [
1875    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1876    
1877  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1878  */  */
1879    
1880  static BOOL  static BOOL
1881  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1882  {  {
1883  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1884  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1885  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1886    {    {
1887    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
1888    return TRUE;      {
1889        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
1890        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
1891          {
1892          *endptr = ptr;
1893          return TRUE;
1894          }
1895        }
1896    }    }
1897  return FALSE;  return FALSE;
1898  }  }
# Line 1733  Returns:     a value representing the na Line 1917  Returns:     a value representing the na
1917  static int  static int
1918  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1919  {  {
1920    const char *pn = posix_names;
1921  register int yield = 0;  register int yield = 0;
1922  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1923    {    {
1924    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1925      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1926      pn += posix_name_lengths[yield] + 1;
1927    yield++;    yield++;
1928    }    }
1929  return -1;  return -1;
# Line 1752  return -1; Line 1938  return -1;
1938  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
1939  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
1940  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
1941  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
1942  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
1943  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
1944  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
1945  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
1946    OP_END.
1947    
1948  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
1949  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 1891  get_othercase_range(unsigned int *cptr, Line 2078  get_othercase_range(unsigned int *cptr,
2078  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2079    
2080  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2081    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2082    
2083  if (c > d) return FALSE;  if (c > d) return FALSE;
2084    
# Line 1900  next = othercase + 1; Line 2087  next = othercase + 1;
2087    
2088  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2089    {    {
2090    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2091    next++;    next++;
2092    }    }
2093    
# Line 1946  if ((options & PCRE_EXTENDED) != 0) Line 2133  if ((options & PCRE_EXTENDED) != 0)
2133    for (;;)    for (;;)
2134      {      {
2135      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2136      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2137        {        {
2138        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2139          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1958  if ((options & PCRE_EXTENDED) != 0) Line 2145  if ((options & PCRE_EXTENDED) != 0)
2145  /* 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
2146  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2147    
2148  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2149    {    {
2150    int temperrorcode = 0;    int temperrorcode = 0;
2151    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1983  if ((options & PCRE_EXTENDED) != 0) Line 2170  if ((options & PCRE_EXTENDED) != 0)
2170    for (;;)    for (;;)
2171      {      {
2172      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2173      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2174        {        {
2175        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2176          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1994  if ((options & PCRE_EXTENDED) != 0) Line 2181  if ((options & PCRE_EXTENDED) != 0)
2181    
2182  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2183    
2184  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2185    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2186        return FALSE;
2187    
2188  /* 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
2189  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 2010  if (next >= 0) switch(op_code) Line 2198  if (next >= 0) switch(op_code)
2198    case OP_CHAR:    case OP_CHAR:
2199  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2200    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2201    #else
2202      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2203  #endif  #endif
2204    return item != next;    return item != next;
2205    
# Line 2028  if (next >= 0) switch(op_code) Line 2218  if (next >= 0) switch(op_code)
2218      unsigned int othercase;      unsigned int othercase;
2219      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2220  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2221      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2222  #else  #else
2223      othercase = NOTACHAR;      othercase = NOTACHAR;
2224  #endif  #endif
# Line 2041  if (next >= 0) switch(op_code) Line 2231  if (next >= 0) switch(op_code)
2231    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2232    
2233    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2234    if (item == next) return TRUE;    if (item == next) return TRUE;
2235    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2236  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2050  if (next >= 0) switch(op_code) Line 2239  if (next >= 0) switch(op_code)
2239      unsigned int othercase;      unsigned int othercase;
2240      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2241  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2242      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2243  #else  #else
2244      othercase = NOTACHAR;      othercase = NOTACHAR;
2245  #endif  #endif
# Line 2304  uschar classbits[32]; Line 2493  uschar classbits[32];
2493  BOOL class_utf8;  BOOL class_utf8;
2494  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2495  uschar *class_utf8data;  uschar *class_utf8data;
2496    uschar *class_utf8data_base;
2497  uschar utf8_char[6];  uschar utf8_char[6];
2498  #else  #else
2499  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2343  req_caseopt = ((options & PCRE_CASELESS) Line 2533  req_caseopt = ((options & PCRE_CASELESS)
2533  for (;; ptr++)  for (;; ptr++)
2534    {    {
2535    BOOL negate_class;    BOOL negate_class;
2536      BOOL should_flip_negation;
2537    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2538    BOOL is_quantifier;    BOOL is_quantifier;
2539    BOOL is_recurse;    BOOL is_recurse;
# Line 2430  for (;; ptr++) Line 2621  for (;; ptr++)
2621    
2622    if (inescq && c != 0)    if (inescq && c != 0)
2623      {      {
2624      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2625        {        {
2626        inescq = FALSE;        inescq = FALSE;
2627        ptr++;        ptr++;
# Line 2456  for (;; ptr++) Line 2647  for (;; ptr++)
2647    /* 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
2648    a quantifier. */    a quantifier. */
2649    
2650    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2651      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2652        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2653    
2654    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2655         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2472  for (;; ptr++) Line 2664  for (;; ptr++)
2664    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2665      {      {
2666      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2667      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2668        {        {
2669        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2670          {          {
# Line 2497  for (;; ptr++) Line 2689  for (;; ptr++)
2689      {      {
2690      /* ===================================================================*/      /* ===================================================================*/
2691      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2692      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2693      case ')':      case CHAR_RIGHT_PARENTHESIS:
2694      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2695      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2696      *codeptr = code;      *codeptr = code;
# Line 2520  for (;; ptr++) Line 2712  for (;; ptr++)
2712      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2713      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2714    
2715      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2716      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2717        {        {
2718        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2529  for (;; ptr++) Line 2721  for (;; ptr++)
2721      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2722      break;      break;
2723    
2724      case '$':      case CHAR_DOLLAR_SIGN:
2725      previous = NULL;      previous = NULL;
2726      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2727      break;      break;
# Line 2537  for (;; ptr++) Line 2729  for (;; ptr++)
2729      /* 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
2730      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2731    
2732      case '.':      case CHAR_DOT:
2733      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2734      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2735      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2736      previous = code;      previous = code;
2737      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2738      break;      break;
2739    
2740    
# Line 2557  for (;; ptr++) Line 2749  for (;; ptr++)
2749      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,
2750      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2751      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.
     */  
2752    
2753      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2754        default (Perl) mode, it is treated as a data character. */
2755    
2756        case CHAR_RIGHT_SQUARE_BRACKET:
2757        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2758          {
2759          *errorcodeptr = ERR64;
2760          goto FAILED;
2761          }
2762        goto NORMAL_CHAR;
2763    
2764        case CHAR_LEFT_SQUARE_BRACKET:
2765      previous = code;      previous = code;
2766    
2767      /* 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
2768      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. */
2769    
2770      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2771          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2772            check_posix_syntax(ptr, &tempptr))
2773        {        {
2774        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2775        goto FAILED;        goto FAILED;
2776        }        }
2777    
# Line 2580  for (;; ptr++) Line 2783  for (;; ptr++)
2783      for (;;)      for (;;)
2784        {        {
2785        c = *(++ptr);        c = *(++ptr);
2786        if (c == '\\')        if (c == CHAR_BACKSLASH)
2787          {          {
2788          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
2789            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
2790              else break;          else if (strncmp((const char *)ptr+1,
2791                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2792              ptr += 3;
2793            else
2794              break;
2795          }          }
2796        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2797          negate_class = TRUE;          negate_class = TRUE;
2798        else break;        else break;
2799        }        }
2800    
2801        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2802        an initial ']' is taken as a data character -- the code below handles
2803        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2804        [^] must match any character, so generate OP_ALLANY. */
2805    
2806        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2807            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2808          {
2809          *code++ = negate_class? OP_ALLANY : OP_FAIL;
2810          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2811          zerofirstbyte = firstbyte;
2812          break;
2813          }
2814    
2815        /* If a class contains a negative special such as \S, we need to flip the
2816        negation flag at the end, so that support for characters > 255 works
2817        correctly (they are all included in the class). */
2818    
2819        should_flip_negation = FALSE;
2820    
2821      /* 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
2822      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
2823      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2608  for (;; ptr++) Line 2835  for (;; ptr++)
2835  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2836      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2837      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2838        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2839  #endif  #endif
2840    
2841      /* 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 2623  for (;; ptr++) Line 2851  for (;; ptr++)
2851          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2852          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2853          }          }
2854    
2855          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2856          data and reset the pointer. This is so that very large classes that
2857          contain a zillion UTF-8 characters no longer overwrite the work space
2858          (which is on the stack). */
2859    
2860          if (lengthptr != NULL)
2861            {
2862            *lengthptr += class_utf8data - class_utf8data_base;
2863            class_utf8data = class_utf8data_base;
2864            }
2865    
2866  #endif  #endif
2867    
2868        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
2869    
2870        if (inescq)        if (inescq)
2871          {          {
2872          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
2873            {            {
2874            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
2875            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2644  for (;; ptr++) Line 2884  for (;; ptr++)
2884        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
2885        5.6 and 5.8 do. */        5.6 and 5.8 do. */
2886    
2887        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
2888            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2889            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
2890          {          {
2891          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2892          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
2893          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
2894          uschar pbits[32];          uschar pbits[32];
2895    
2896          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
2897            {            {
2898            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
2899            goto FAILED;            goto FAILED;
2900            }            }
2901    
2902          ptr += 2;          ptr += 2;
2903          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
2904            {            {
2905            local_negate = TRUE;            local_negate = TRUE;
2906              should_flip_negation = TRUE;  /* Note negative special */
2907            ptr++;            ptr++;
2908            }            }
2909    
# Line 2732  for (;; ptr++) Line 2973  for (;; ptr++)
2973        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
2974        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2975    
2976        if (c == '\\')        if (c == CHAR_BACKSLASH)
2977          {          {
2978          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2979          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
2980    
2981          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 */
2982          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 */
2983          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 */
2984          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
2985            {            {
2986            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
2987              {              {
2988              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
2989              }              }
# Line 2765  for (;; ptr++) Line 3006  for (;; ptr++)
3006              continue;              continue;
3007    
3008              case ESC_D:              case ESC_D:
3009                should_flip_negation = TRUE;
3010              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3011              continue;              continue;
3012    
# Line 2773  for (;; ptr++) Line 3015  for (;; ptr++)
3015              continue;              continue;
3016    
3017              case ESC_W:              case ESC_W:
3018                should_flip_negation = TRUE;
3019              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3020              continue;              continue;
3021    
# Line 2782  for (;; ptr++) Line 3025  for (;; ptr++)
3025              continue;              continue;
3026    
3027              case ESC_S:              case ESC_S:
3028                should_flip_negation = TRUE;
3029              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3030              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3031              continue;              continue;
3032    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
3033              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
3034              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
3035              }              }
# Line 2968  for (;; ptr++) Line 3209  for (;; ptr++)
3209        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3210    
3211        CHECK_RANGE:        CHECK_RANGE:
3212        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3213          {          {
3214          inescq = FALSE;          inescq = FALSE;
3215          ptr += 2;          ptr += 2;
3216          }          }
3217    
3218        oldptr = ptr;        oldptr = ptr;
3219    
3220        /* Remember \r or \n */        /* Remember \r or \n */
3221    
3222        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3223    
3224        /* Check for range */        /* Check for range */
3225    
3226        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3227          {          {
3228          int d;          int d;
3229          ptr += 2;          ptr += 2;
3230          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3231    
3232          /* 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
3233          mode. */          mode. */
3234    
3235          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3236            {            {
3237            ptr += 2;            ptr += 2;
3238            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3239                { ptr += 2; continue; }
3240            inescq = TRUE;            inescq = TRUE;
3241            break;            break;
3242            }            }
3243    
3244          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3245            {            {
3246            ptr = oldptr;            ptr = oldptr;
3247            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3018  for (;; ptr++) Line 3260  for (;; ptr++)
3260          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
3261          in such circumstances. */          in such circumstances. */
3262    
3263          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3264            {            {
3265            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3266            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3267    
3268            /* \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
3269            special means the '-' was literal */            special means the '-' was literal */
3270    
3271            if (d < 0)            if (d < 0)
3272              {              {
3273              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3274              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3275              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3276                {                {
3277                ptr = oldptr;                ptr = oldptr;
3278                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3050  for (;; ptr++) Line 3292  for (;; ptr++)
3292          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3293    
3294          /* Remember \r or \n */          /* Remember \r or \n */
3295    
3296          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3297    
3298          /* 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
3299          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3300          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3157  for (;; ptr++) Line 3399  for (;; ptr++)
3399        apparent range that isn't. */        apparent range that isn't. */
3400    
3401        LONE_SINGLE_CHARACTER:        LONE_SINGLE_CHARACTER:
3402    
3403        /* Handle a character that cannot go in the bit map */        /* Handle a character that cannot go in the bit map */
3404    
3405  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 3171  for (;; ptr++) Line 3413  for (;; ptr++)
3413          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3414            {            {
3415            unsigned int othercase;            unsigned int othercase;
3416            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3417              {              {
3418              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3419              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3198  for (;; ptr++) Line 3440  for (;; ptr++)
3440    
3441      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3442    
3443      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3444    
3445      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3446        {        {
3447        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3448        goto FAILED;        goto FAILED;
3449        }        }
3450    
3451    
3452  /* This code has been disabled because it would mean that \s counts as  /* This code has been disabled because it would mean that \s counts as
3453  an explicit \r or \n reference, and that's not really what is wanted. Now  an explicit \r or \n reference, and that's not really what is wanted. Now
3454  we set the flag only if there is a literal "\r" or "\n" in the class. */  we set the flag only if there is a literal "\r" or "\n" in the class. */
# Line 3223  we set the flag only if there is a liter Line 3465  we set the flag only if there is a liter
3465        if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;        if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3466        }        }
3467  #endif  #endif
3468    
3469    
3470      /* 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
3471      less than 256. As long as there were no characters >= 128 and there was no      less than 256. As long as there were no characters >= 128 and there was no
# Line 3287  we set the flag only if there is a liter Line 3529  we set the flag only if there is a liter
3529      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3530    
3531      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3532      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3533      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
3534        the class, so any that were explicitly given as well can be ignored. If
3535        (when there are explicit characters > 255 that must be listed) there are no
3536        characters < 256, we can omit the bitmap in the actual compiled code. */
3537    
3538  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3539      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3540        {        {
3541        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3542        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3317  we set the flag only if there is a liter Line 3562  we set the flag only if there is a liter
3562        }        }
3563  #endif  #endif
3564    
3565      /* 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
3566      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
3567      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
3568      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3569    
3570        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3571      if (negate_class)      if (negate_class)
3572        {        {
       *code++ = OP_NCLASS;  
3573        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3574          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3575        }        }
3576      else      else
3577        {        {
       *code++ = OP_CLASS;  
3578        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3579        }        }
3580      code += 32;      code += 32;
# Line 3341  we set the flag only if there is a liter Line 3585  we set the flag only if there is a liter
3585      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3586      has been tested above. */      has been tested above. */
3587    
3588      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3589      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3590      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3591      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3592      goto REPEAT;      goto REPEAT;
3593    
3594      case '*':      case CHAR_ASTERISK:
3595      repeat_min = 0;      repeat_min = 0;
3596      repeat_max = -1;      repeat_max = -1;
3597      goto REPEAT;      goto REPEAT;
3598    
3599      case '+':      case CHAR_PLUS:
3600      repeat_min = 1;      repeat_min = 1;
3601      repeat_max = -1;      repeat_max = -1;
3602      goto REPEAT;      goto REPEAT;
3603    
3604      case '?':      case CHAR_QUESTION_MARK:
3605      repeat_min = 0;      repeat_min = 0;
3606      repeat_max = 1;      repeat_max = 1;
3607    
# Line 3392  we set the flag only if there is a liter Line 3636  we set the flag only if there is a liter
3636      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
3637      repeat type to the non-default. */      repeat type to the non-default. */
3638    
3639      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3640        {        {
3641        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3642        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3643        ptr++;        ptr++;
3644        }        }
3645      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3646        {        {
3647        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3648        ptr++;        ptr++;
# Line 3724  we set the flag only if there is a liter Line 3968  we set the flag only if there is a liter
3968    
3969        if (repeat_min == 0)        if (repeat_min == 0)
3970          {          {
3971          /* 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
3972          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
3973    
3974          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
3975          BRAZERO and do no more at this point. However, we do need to adjust          **   {
3976          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
3977          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
3978          the start of the whole regex. Temporarily terminate the pattern while          **   }
3979          doing this. */  
3980            However, that fails when a group is referenced as a subroutine from
3981            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
3982            so that it is skipped on execution. As we don't have a list of which
3983            groups are referenced, we cannot do this selectively.
3984    
3985            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
3986            and do no more at this point. However, we do need to adjust any
3987            OP_RECURSE calls inside the group that refer to the group itself or any
3988            internal or forward referenced group, because the offset is from the
3989            start of the whole regex. Temporarily terminate the pattern while doing
3990            this. */
3991    
3992          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
3993            {            {
3994            *code = OP_END;            *code = OP_END;
3995            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3996            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3997            code++;            code++;
3998              if (repeat_max == 0)
3999                {
4000                *previous++ = OP_SKIPZERO;
4001                goto END_REPEAT;
4002                }
4003            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4004            }            }
4005    
# Line 3940  we set the flag only if there is a liter Line 4194  we set the flag only if there is a liter
4194          }          }
4195        }        }
4196    
4197        /* If previous is OP_FAIL, it was generated by an empty class [] in
4198        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4199        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4200        error above. We can just ignore the repeat in JS case. */
4201    
4202        else if (*previous == OP_FAIL) goto END_REPEAT;
4203    
4204      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4205    
4206      else      else
# Line 3966  we set the flag only if there is a liter Line 4227  we set the flag only if there is a liter
4227        int len;        int len;
4228        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4229            *tempcode == OP_NOTEXACT)            *tempcode == OP_NOTEXACT)
4230          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode] +
4231              ((*tempcode == OP_TYPEEXACT &&
4232                 (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4233        len = code - tempcode;        len = code - tempcode;
4234        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4235          {          {
# Line 4012  we set the flag only if there is a liter Line 4275  we set the flag only if there is a liter
4275      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4276      parenthesis forms.  */      parenthesis forms.  */
4277    
4278      case '(':      case CHAR_LEFT_PARENTHESIS:
4279      newoptions = options;      newoptions = options;
4280      skipbytes = 0;      skipbytes = 0;
4281      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4021  we set the flag only if there is a liter Line 4284  we set the flag only if there is a liter
4284    
4285      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4286    
4287      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4288        {        {
4289        int i, namelen;        int i, namelen;
4290          const char *vn = verbnames;
4291        const uschar *name = ++ptr;        const uschar *name = ++ptr;
4292        previous = NULL;        previous = NULL;
4293        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4294        if (*ptr == ':')        if (*ptr == CHAR_COLON)
4295          {          {
4296          *errorcodeptr = ERR59;   /* Not supported */          *errorcodeptr = ERR59;   /* Not supported */
4297          goto FAILED;          goto FAILED;
4298          }          }
4299        if (*ptr != ')')        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4300          {          {
4301          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4302          goto FAILED;          goto FAILED;
# Line 4041  we set the flag only if there is a liter Line 4305  we set the flag only if there is a liter
4305        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4306          {          {
4307          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4308              strncmp((char *)name, verbs[i].name, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4309            {            {
4310            *code = verbs[i].op;            *code = verbs[i].op;
4311            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4312            break;            break;
4313            }            }
4314            vn += verbs[i].len + 1;
4315          }          }
4316        if (i < verbcount) continue;        if (i < verbcount) continue;
4317        *errorcodeptr = ERR60;        *errorcodeptr = ERR60;
# Line 4056  we set the flag only if there is a liter Line 4321  we set the flag only if there is a liter
4321      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4322      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. */
4323    
4324      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4325        {        {
4326        int i, set, unset, namelen;        int i, set, unset, namelen;
4327        int *optset;        int *optset;
# Line 4065  we set the flag only if there is a liter Line 4330  we set the flag only if there is a liter
4330    
4331        switch (*(++ptr))        switch (*(++ptr))
4332          {          {
4333          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4334          ptr++;          ptr++;
4335          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4336          if (*ptr == 0)          if (*ptr == 0)
4337            {            {
4338            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4077  we set the flag only if there is a liter Line 4342  we set the flag only if there is a liter
4342    
4343    
4344          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4345          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4346          reset_bracount = TRUE;          reset_bracount = TRUE;
4347          /* Fall through */          /* Fall through */
4348    
4349          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4350          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4351          bravalue = OP_BRA;          bravalue = OP_BRA;
4352          ptr++;          ptr++;
4353          break;          break;
4354    
4355    
4356          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4357          case '(':          case CHAR_LEFT_PARENTHESIS:
4358          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4359    
4360          /* 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 4109  we set the flag only if there is a liter Line 4374  we set the flag only if there is a liter
4374          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4375          including assertions, are processed. */          including assertions, are processed. */
4376    
4377          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4378                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4379            break;            break;
4380    
4381          /* 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 4121  we set the flag only if there is a liter Line 4387  we set the flag only if there is a liter
4387    
4388          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4389    
4390          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4391            {            {
4392            terminator = -1;            terminator = -1;
4393            ptr += 2;            ptr += 2;
# Line 4131  we set the flag only if there is a liter Line 4397  we set the flag only if there is a liter
4397          /* 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
4398          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4399    
4400          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4401            {            {
4402            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4403            ptr++;            ptr++;
4404            }            }
4405          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4406            {            {
4407            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4408            ptr++;            ptr++;
4409            }            }
4410          else          else
4411            {            {
4412            terminator = 0;            terminator = 0;
4413            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4414            }            }
4415    
4416          /* 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 4164  we set the flag only if there is a liter Line 4430  we set the flag only if there is a liter
4430            {            {
4431            if (recno >= 0)            if (recno >= 0)
4432              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4433                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4434            ptr++;            ptr++;
4435            }            }
4436          namelen = ptr - name;          namelen = ptr - name;
4437    
4438          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4439                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4440            {            {
4441            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4442            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4191  we set the flag only if there is a liter Line 4458  we set the flag only if there is a liter
4458              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4459              goto FAILED;              goto FAILED;
4460              }              }
4461            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
4462                cd->bracount - recno + 1 : recno +cd->bracount;
4463              if (recno <= 0 || recno > cd->final_bracount)
4464              {              {
4465              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4466              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4467              }              }
           else recno += cd->bracount;  
4468            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4469            break;            break;
4470            }            }
# Line 4225  we set the flag only if there is a liter Line 4489  we set the flag only if there is a liter
4489    
4490          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4491    
4492          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(ptr, cd, name, namelen,
4493                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4494            {            {
4495            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
# Line 4246  we set the flag only if there is a liter Line 4510  we set the flag only if there is a liter
4510          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4511          specific group number. */          specific group number. */
4512    
4513          else if (*name == 'R')          else if (*name == CHAR_R)
4514            {            {
4515            recno = 0;            recno = 0;
4516            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4256  we set the flag only if there is a liter Line 4520  we set the flag only if there is a liter
4520                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4521                goto FAILED;                goto FAILED;
4522                }                }
4523              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4524              }              }
4525            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4526            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4266  we set the flag only if there is a liter Line 4530  we set the flag only if there is a liter
4530          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4531          false. */          false. */
4532    
4533          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4534            {            {
4535            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4536            skipbytes = 1;            skipbytes = 1;
4537            }            }
4538    
4539          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4540            in the second pass here, so final_bracount is set. */
4541    
4542          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4543            {            {
4544            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4545            }            }
# Line 4290  we set the flag only if there is a liter Line 4555  we set the flag only if there is a liter
4555    
4556    
4557          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4558          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4559          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4560          ptr++;          ptr++;
4561          break;          break;
4562    
4563    
4564          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4565          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
4566          ptr++;          ptr++;
4567          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4568            {            {
4569            *code++ = OP_FAIL;            *code++ = OP_FAIL;
4570            previous = NULL;            previous = NULL;
# Line 4310  we set the flag only if there is a liter Line 4575  we set the flag only if there is a liter
4575    
4576    
4577          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4578          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4579          switch (ptr[1])          switch (ptr[1])
4580            {            {
4581            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4582            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4583            ptr += 2;            ptr += 2;
4584            break;            break;
4585    
4586            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4587            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4588            ptr += 2;            ptr += 2;
4589            break;            break;
# Line 4333  we set the flag only if there is a liter Line 4598  we set the flag only if there is a liter
4598    
4599    
4600          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4601          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4602          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4603          ptr++;          ptr++;
4604          break;          break;
4605    
4606    
4607          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4608          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4609          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4610          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4611          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4612            {            {
4613            int n = 0;            int n = 0;
4614            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4615              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4616            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4617              {              {
4618              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4619              goto FAILED;              goto FAILED;
# Line 4368  we set the flag only if there is a liter Line 4633  we set the flag only if there is a liter
4633    
4634    
4635          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4636          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4637          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4638                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4639            {            {
4640            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4641            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4642            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4643            }            }
4644          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4645            {            {
4646            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4647            goto FAILED;            goto FAILED;
# Line 4385  we set the flag only if there is a liter Line 4651  we set the flag only if there is a liter
4651    
4652          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4653          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4654          case '\'':          case CHAR_APOSTROPHE:
4655            {            {
4656            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4657                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4658            name = ++ptr;            name = ++ptr;
4659    
4660            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4461  we set the flag only if there is a liter Line 4728  we set the flag only if there is a liter
4728    
4729    
4730          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4731          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4732          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4733          is_recurse = TRUE;          is_recurse = TRUE;
4734          /* Fall through */          /* Fall through */
4735    
4736          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4737          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4738          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4739            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4740            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4741    
4742          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4743          name = ++ptr;          name = ++ptr;
# Line 4480  we set the flag only if there is a liter Line 4749  we set the flag only if there is a liter
4749    
4750          if (lengthptr != NULL)          if (lengthptr != NULL)
4751            {            {
4752              if (namelen == 0)
4753                {
4754                *errorcodeptr = ERR62;
4755                goto FAILED;
4756                }
4757            if (*ptr != terminator)            if (*ptr != terminator)
4758              {              {
4759              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4493  we set the flag only if there is a liter Line 4767  we set the flag only if there is a liter
4767            recno = 0;            recno = 0;
4768            }            }
4769    
4770          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4771            first, and then check that we have reached the end of the name in the
4772            table. That way, if the name that is longer than any in the table,
4773            the comparison will fail without reading beyond the table entry. */
4774    
4775          else          else
4776            {            {
4777            slot = cd->name_table;            slot = cd->name_table;
4778            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4779              {              {
4780              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4781                    slot[2+namelen] == 0)
4782                  break;
4783              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4784              }              }
4785    
# Line 4509  we set the flag only if there is a liter Line 4788  we set the flag only if there is a liter
4788              recno = GET2(slot, 0);              recno = GET2(slot, 0);
4789              }              }
4790            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
4791                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(ptr, cd, name, namelen,
4792                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
4793              {              {
4794              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4525  we set the flag only if there is a liter Line 4804  we set the flag only if there is a liter
4804    
4805    
4806          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4807          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
4808          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
4809          /* Fall through */          /* Fall through */
4810    
4811    
4812          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4813          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
4814          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:
4815          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:
4816            {            {
4817            const uschar *called;            const uschar *called;
4818              terminator = CHAR_RIGHT_PARENTHESIS;
4819    
4820              /* Come here from the \g<...> and \g'...' code (Oniguruma
4821              compatibility). However, the syntax has been checked to ensure that
4822              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
4823              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
4824              ever be taken. */
4825    
4826            if ((refsign = *ptr) == '+') ptr++;            HANDLE_NUMERICAL_RECURSION:
4827            else if (refsign == '-')  
4828              if ((refsign = *ptr) == CHAR_PLUS)
4829                {
4830                ptr++;
4831                if ((digitab[*ptr] & ctype_digit) == 0)
4832                  {
4833                  *errorcodeptr = ERR63;
4834                  goto FAILED;
4835                  }
4836                }
4837              else if (refsign == CHAR_MINUS)
4838              {              {
4839              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
4840                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4547  we set the flag only if there is a liter Line 4843  we set the flag only if there is a liter
4843    
4844            recno = 0;            recno = 0;
4845            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4846              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
4847    
4848            if (*ptr != ')')            if (*ptr != terminator)
4849              {              {
4850              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
4851              goto FAILED;              goto FAILED;
4852              }              }
4853    
4854            if (refsign == '-')            if (refsign == CHAR_MINUS)
4855              {              {
4856              if (recno == 0)              if (recno == 0)
4857                {                {
# Line 4569  we set the flag only if there is a liter Line 4865  we set the flag only if there is a liter
4865                goto FAILED;                goto FAILED;
4866                }                }
4867              }              }
4868            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
4869              {              {
4870              if (recno == 0)              if (recno == 0)
4871                {                {
# Line 4602  we set the flag only if there is a liter Line 4898  we set the flag only if there is a liter
4898    
4899              if (called == NULL)              if (called == NULL)
4900                {                {
4901                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(ptr, cd, NULL, recno,
4902                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
4903                  {                  {
4904                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
4905                  goto FAILED;                  goto FAILED;
# Line 4655  we set the flag only if there is a liter Line 4951  we set the flag only if there is a liter
4951          set = unset = 0;          set = unset = 0;
4952          optset = &set;          optset = &set;
4953    
4954          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
4955            {            {
4956            switch (*ptr++)            switch (*ptr++)
4957              {              {
4958              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
4959    
4960              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
4961              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
4962              cd->external_flags |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
4963              break;              break;
4964    
4965              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
4966              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
4967              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
4968              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
4969              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
4970              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
4971    
4972              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
4973                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4702  we set the flag only if there is a liter Line 4998  we set the flag only if there is a liter
4998          both phases.          both phases.
4999    
5000          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
5001          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
5002          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. */  
5003    
5004          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5005            {            {
5006            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5007                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5008              {              {
5009              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5010              }              }
5011           else           else
5012              {              {
# Line 4722  we set the flag only if there is a liter Line 5015  we set the flag only if there is a liter
5015                *code++ = OP_OPT;                *code++ = OP_OPT;
5016                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5017                }                }
   
             /* 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;  
5018              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5019              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5020              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5021              }              }
5022    
5023              /* Change options at this level, and pass them back for use
5024              in subsequent branches. When not at the start of the pattern, this
5025              information is also necessary so that a resetting item can be
5026              compiled at the end of a group (if we are in a group). */
5027    
5028              *optionsptr = options = newoptions;
5029            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5030            continue;              /* It is complete */            continue;              /* It is complete */
5031            }            }
# Line 4848  we set the flag only if there is a liter Line 5141  we set the flag only if there is a liter
5141    
5142      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5143    
5144      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5145        {        {
5146        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5147        goto FAILED;        goto FAILED;
# Line 4946  we set the flag only if there is a liter Line 5239  we set the flag only if there is a liter
5239      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
5240      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5241    
5242      case '\\':      case CHAR_BACKSLASH:
5243      tempptr = ptr;      tempptr = ptr;
5244      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5245      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 4955  we set the flag only if there is a liter Line 5248  we set the flag only if there is a liter
5248        {        {
5249        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5250          {          {
5251          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5252            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5253                else inescq = TRUE;
5254          continue;          continue;
5255          }          }
5256    
# Line 4973  we set the flag only if there is a liter Line 5267  we set the flag only if there is a liter
5267        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5268        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5269    
5270          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5271          is a subroutine call by number (Oniguruma syntax). In fact, the value
5272          -ESC_g is returned only for these cases. So we don't need to check for <
5273          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5274          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5275          that is a synonym for a named back reference). */
5276    
5277          if (-c == ESC_g)
5278            {
5279            const uschar *p;
5280            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5281            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5282              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5283    
5284            /* These two statements stop the compiler for warning about possibly
5285            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5286            fact, because we actually check for a number below, the paths that
5287            would actually be in error are never taken. */
5288    
5289            skipbytes = 0;
5290            reset_bracount = FALSE;
5291    
5292            /* Test for a name */
5293    
5294            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5295              {
5296              BOOL isnumber = TRUE;
5297              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5298                {
5299                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5300                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5301                }
5302              if (*p != terminator)
5303                {
5304                *errorcodeptr = ERR57;
5305                break;
5306                }
5307              if (isnumber)
5308                {
5309                ptr++;
5310                goto HANDLE_NUMERICAL_RECURSION;
5311                }
5312              is_recurse = TRUE;
5313              goto NAMED_REF_OR_RECURSE;
5314              }
5315    
5316            /* Test a signed number in angle brackets or quotes. */
5317    
5318            p = ptr + 2;
5319            while ((digitab[*p] & ctype_digit) != 0) p++;
5320            if (*p != terminator)
5321              {
5322              *errorcodeptr = ERR57;
5323              break;
5324              }
5325            ptr++;
5326            goto HANDLE_NUMERICAL_RECURSION;
5327            }
5328    
5329        /* \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).
5330        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5331    
5332        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5333              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5334          {          {
5335          is_recurse = FALSE;          is_recurse = FALSE;
5336          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5337              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5338              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5339          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5340          }          }
5341    
# Line 5082  we set the flag only if there is a liter Line 5438  we set the flag only if there is a liter
5438    
5439      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
5440    
5441      if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')      if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5442        cd->external_flags |= PCRE_HASCRORLF;        cd->external_flags |= PCRE_HASCRORLF;
5443    
5444      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
# Line 5327  for (;;) Line 5683  for (;;)
5683    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.
5684    Return leaving the pointer at the terminating char. */    Return leaving the pointer at the terminating char. */
5685    
5686    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
5687      {      {
5688      if (lengthptr == NULL)      if (lengthptr == NULL)
5689        {        {
# Line 5350  for (;;) Line 5706  for (;;)
5706    
5707      /* Resetting option if needed */      /* Resetting option if needed */
5708    
5709      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
5710        {        {
5711        *code++ = OP_OPT;        *code++ = OP_OPT;
5712        *code++ = oldims;        *code++ = oldims;
# Line 5479  do { Line 5835  do {
5835       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
5836       }       }
5837    
5838     /* .* 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
5839     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
5840    
5841     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
5842               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
5843       {       {
5844       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
5845           return FALSE;
5846       }       }
5847    
5848     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5532  do { Line 5888  do {
5888       NULL, 0, FALSE);       NULL, 0, FALSE);
5889     register int op = *scode;     register int op = *scode;
5890    
5891       /* If we are at the start of a conditional assertion group, *both* the
5892       conditional assertion *and* what follows the condition must satisfy the test
5893       for start of line. Other kinds of condition fail. Note that there may be an
5894       auto-callout at the start of a condition. */
5895    
5896       if (op == OP_COND)
5897         {
5898         scode += 1 + LINK_SIZE;
5899         if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
5900         switch (*scode)
5901           {
5902           case OP_CREF:
5903           case OP_RREF:
5904           case OP_DEF:
5905           return FALSE;
5906    
5907           default:     /* Assertion */
5908           if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
5909           do scode += GET(scode, 1); while (*scode == OP_ALT);
5910           scode += 1 + LINK_SIZE;
5911           break;
5912           }
5913         scode = first_significant_code(scode, NULL, 0, FALSE);
5914         op = *scode;
5915         }
5916    
5917     /* Non-capturing brackets */     /* Non-capturing brackets */
5918    
5919     if (op == OP_BRA)     if (op == OP_BRA)
# Line 5550  do { Line 5932  do {
5932    
5933     /* Other brackets */     /* Other brackets */
5934    
5935     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
5936       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
5937         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
5938         }
5939    
5940     /* .* 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
5941     may be referenced. */     may be referenced. */
# Line 5668  Returns:        pointer to compiled data Line 6052  Returns:        pointer to compiled data
6052                  with errorptr and erroroffset set                  with errorptr and erroroffset set
6053  */  */
6054    
6055  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6056  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
6057    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
6058  {  {
# Line 5676  return pcre_compile2(pattern, options, N Line 6060  return pcre_compile2(pattern, options, N
6060  }  }
6061    
6062    
6063  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6064  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
6065    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
6066  {  {
# Line 5703  to fill in forward references to subpatt Line 6087  to fill in forward references to subpatt
6087    
6088  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
6089    
   
6090  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
6091    
6092  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;
# Line 5749  if ((options & PCRE_UTF8) != 0) Line 6132  if ((options & PCRE_UTF8) != 0)
6132    }    }
6133  #endif  #endif
6134    
6135  if ((options & ~PUBLIC_OPTIONS) != 0)  if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6136    {    {
6137    errorcode = ERR17;    errorcode = ERR17;
6138    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
# Line 5763  cd->fcc = tables + fcc_offset; Line 6146  cd->fcc = tables + fcc_offset;
6146  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6147  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6148    
6149  /* Check for newline settings at the start of the pattern, and remember the  /* Check for global one-time settings at the start of the pattern, and remember
6150  offset for later. */  the offset for later. */
6151    
6152  if (ptr[0] == '(' && ptr[1] == '*')  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6153           ptr[skipatstart+1] == CHAR_ASTERISK)
6154    {    {
6155    int newnl = 0;    int newnl = 0;
6156    if (strncmp((char *)(ptr+2), "CR)", 3) == 0)    int newbsr = 0;
6157      { skipatstart = 5; newnl = PCRE_NEWLINE_CR; }  
6158    else if (strncmp((char *)(ptr+2), "LF)", 3)  == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6159      { skipatstart = 5; newnl = PCRE_NEWLINE_LF; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6160    else if (strncmp((char *)(ptr+2), "CRLF)", 5)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
6161      { skipatstart = 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6162    else if (strncmp((char *)(ptr+2), "ANY)", 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_CRLF_RIGHTPAR, 5)  == 0)
6163      { skipatstart = 6; newnl = PCRE_NEWLINE_ANY; }      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6164    else if (strncmp((char *)(ptr+2), "ANYCRLF)", 8)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANY_RIGHTPAR, 4) == 0)
6165      { skipatstart = 10; newnl = PCRE_NEWLINE_ANYCRLF; }      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6166    if (skipatstart > 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANYCRLF_RIGHTPAR, 8) == 0)
6167        { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6168    
6169      else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
6170        { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6171      else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
6172        { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6173    
6174      if (newnl != 0)
6175      options = (options & ~PCRE_NEWLINE_BITS) | newnl;      options = (options & ~PCRE_NEWLINE_BITS) | newnl;
6176      else if (newbsr != 0)
6177        options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
6178      else break;
6179      }
6180    
6181    /* Check validity of \R options. */
6182    
6183    switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6184      {
6185      case 0:
6186      case PCRE_BSR_ANYCRLF:
6187      case PCRE_BSR_UNICODE:
6188      break;
6189      default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
6190    }    }
6191    
6192  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
# Line 5790  current code allows for fixed one- or tw Line 6196  current code allows for fixed one- or tw
6196  switch (options & PCRE_NEWLINE_BITS)  switch (options & PCRE_NEWLINE_BITS)
6197    {    {
6198    case 0: newline = NEWLINE; break;   /* Build-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6199    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6200    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6201    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6202         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6203    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6204    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6205    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
# Line 5842  to compile parts of the pattern into; th Line 6248  to compile parts of the pattern into; th
6248  no longer needed, so hopefully this workspace will never overflow, though there  no longer needed, so hopefully this workspace will never overflow, though there
6249  is a test for its doing so. */  is a test for its doing so. */
6250    
6251  cd->bracount = 0;  cd->bracount = cd->final_bracount = 0;
6252  cd->names_found = 0;  cd->names_found = 0;
6253  cd->name_entry_size = 0;  cd->name_entry_size = 0;
6254  cd->name_table = NULL;  cd->name_table = NULL;
# Line 5892  if (re == NULL) Line 6298  if (re == NULL)
6298    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
6299    }    }
6300    
6301  /* Put in the magic number, and save the sizes, initial options, internal  /* Put in the magic number, and save the sizes, initial options, internal
6302  flags, and character table pointer. NULL is used for the default character  flags, and character table pointer. NULL is used for the default character
6303  tables. The nullpad field is at the end; it's there to help in the case when a  tables. The nullpad field is at the end; it's there to help in the case when a
6304  regex compiled on a system with 4-byte pointers is run on another with 8-byte  regex compiled on a system with 4-byte pointers is run on another with 8-byte
# Line 5919  field. Reset the bracket count and the n Line 6325  field. Reset the bracket count and the n
6325  field; this time it's used for remembering forward references to subpatterns.  field; this time it's used for remembering forward references to subpatterns.
6326  */  */
6327    
6328    cd->final_bracount = cd->bracount;  /* Save for checking forward references */
6329  cd->bracount = 0;  cd->bracount = 0;
6330  cd->names_found = 0;  cd->names_found = 0;
6331  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 5983  if (errorcode != 0) Line 6390  if (errorcode != 0)
6390    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
6391    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = ptr - (const uschar *)pattern;
6392    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
6393    *errorptr = error_texts[errorcode];    *errorptr = find_error_text(errorcode);
6394    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
6395    return NULL;    return NULL;
6396    }    }
# Line 6068  was compiled can be seen. */ Line 6475  was compiled can be seen. */
6475  if (code - codestart > length)  if (code - codestart > length)
6476    {    {
6477    (pcre_free)(re);    (pcre_free)(re);
6478    *errorptr = error_texts[ERR23];    *errorptr = find_error_text(ERR23);
6479    *erroroffset = ptr - (uschar *)pattern;    *erroroffset = ptr - (uschar *)pattern;
6480    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6481    return NULL;    return NULL;

Legend:
Removed from v.230  
changed lines
  Added in v.406

  ViewVC Help
Powered by ViewVC 1.1.5