/[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 178 by ph10, Wed Jun 13 08:44:34 2007 UTC revision 513 by ph10, Mon May 3 11:13:37 2010 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) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 62  used by pcretest. DEBUG is not defined w Line 66  used by pcretest. DEBUG is not defined w
66    
67  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 81  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
103  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
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         0,                       0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 120  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
186  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
192      int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194    } verbitem;
195    
196    static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199      STRING_ACCEPT0
200      STRING_COMMIT0
201      STRING_F0
202      STRING_FAIL0
203      STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217    };
218    
219    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222    /* Tables of names of POSIX character classes and their lengths. The names are
223    now all in a single string, to reduce the number of relocations when a shared
224    library is dynamically loaded. The list of lengths is terminated by a zero
225    length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    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 176  static const int posix_class_maps[] = { Line 268  static const int posix_class_maps[] = {
268  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
269  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
270  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
271  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
272    the number of relocations needed when a shared library is loaded dynamically,
273  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
274    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
275    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
276    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
277    "unrecognized character follows \\",  
278    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
279    substring, so that the whole string ends with \0\0, which can be detected when
280    counting through. */
281    
282    static const char error_texts[] =
283      "no error\0"
284      "\\ at end of pattern\0"
285      "\\c at end of pattern\0"
286      "unrecognized character follows \\\0"
287      "numbers out of order in {} quantifier\0"
288    /* 5 */    /* 5 */
289    "number too big in {} quantifier",    "number too big in {} quantifier\0"
290    "missing terminating ] for character class",    "missing terminating ] for character class\0"
291    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
292    "range out of order in character class",    "range out of order in character class\0"
293    "nothing to repeat",    "nothing to repeat\0"
294    /* 10 */    /* 10 */
295    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
296    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
297    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
298    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
299    "missing )",    "missing )\0"
300    /* 15 */    /* 15 */
301    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
302    "erroffset passed as NULL",    "erroffset passed as NULL\0"
303    "unknown option bit(s) set",    "unknown option bit(s) set\0"
304    "missing ) after comment",    "missing ) after comment\0"
305    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
306    /* 20 */    /* 20 */
307    "regular expression too large",    "regular expression is too large\0"
308    "failed to get memory",    "failed to get memory\0"
309    "unmatched parentheses",    "unmatched parentheses\0"
310    "internal error: code overflow",    "internal error: code overflow\0"
311    "unrecognized character after (?<",    "unrecognized character after (?<\0"
312    /* 25 */    /* 25 */
313    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
314    "malformed number or name after (?(",    "malformed number or name after (?(\0"
315    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
316    "assertion expected after (?(",    "assertion expected after (?(\0"
317    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
318    /* 30 */    /* 30 */
319    "unknown POSIX class name",    "unknown POSIX class name\0"
320    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
321    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
322    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
323    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
324    /* 35 */    /* 35 */
325    "invalid condition (?(0)",    "invalid condition (?(0)\0"
326    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
327    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
328    "number after (?C is > 255",    "number after (?C is > 255\0"
329    "closing ) for (?C expected",    "closing ) for (?C expected\0"
330    /* 40 */    /* 40 */
331    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
332    "unrecognized character after (?P",    "unrecognized character after (?P\0"
333    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
334    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
335    "invalid UTF-8 string",    "invalid UTF-8 string\0"
336    /* 45 */    /* 45 */
337    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
338    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
339    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
340    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
341    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
342    /* 50 */    /* 50 */
343    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
344    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
345    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
346    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
347    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
348    /* 55 */    /* 55 */
349    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
350    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
351    "\\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"
352    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"    "a numbered reference must not be zero\0"
353  };    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
354      /* 60 */
355      "(*VERB) not recognized\0"
356      "number is too big\0"
357      "subpattern name expected\0"
358      "digit expected after (?+\0"
359      "] is an invalid data character in JavaScript compatibility mode\0"
360      /* 65 */
361      "different names for subpatterns of the same number are not allowed\0"
362      "(*MARK) must have an argument\0"
363      ;
364    
365  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
366  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 268  For convenience, we use the same bit def Line 378  For convenience, we use the same bit def
378    
379  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
380    
381  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
382    
383    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
384    UTF-8 mode. */
385    
386  static const unsigned char digitab[] =  static const unsigned char digitab[] =
387    {    {
388    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 304  static const unsigned char digitab[] = Line 418  static const unsigned char digitab[] =
418    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
419    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
420    
421  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
422    
423    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
424    
425  static const unsigned char digitab[] =  static const unsigned char digitab[] =
426    {    {
427    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 379  static const unsigned char ebcdic_charta Line 496  static const unsigned char ebcdic_charta
496  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
497    
498  static BOOL  static BOOL
499    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
500      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
501    
502    
503    
504  /*************************************************  /*************************************************
505    *            Find an error text                  *
506    *************************************************/
507    
508    /* The error texts are now all in one long string, to save on relocations. As
509    some of the text is of unknown length, we can't use a table of offsets.
510    Instead, just count through the strings. This is not a performance issue
511    because it happens only when there has been a compilation error.
512    
513    Argument:   the error number
514    Returns:    pointer to the error string
515    */
516    
517    static const char *
518    find_error_text(int n)
519    {
520    const char *s = error_texts;
521    for (; n > 0; n--)
522      {
523      while (*s++ != 0) {};
524      if (*s == 0) return "Error text not found (please report)";
525      }
526    return s;
527    }
528    
529    
530    /*************************************************
531  *            Handle escapes                      *  *            Handle escapes                      *
532  *************************************************/  *************************************************/
533    
# Line 405  Arguments: Line 548  Arguments:
548    
549  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
550                   negative => a special escape sequence                   negative => a special escape sequence
551                   on error, errorptr is set                   on error, errorcodeptr is set
552  */  */
553    
554  static int  static int
# Line 423  ptr--; /* Set Line 566  ptr--; /* Set
566    
567  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
568    
569  /* 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
570  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.
571  Otherwise further processing may be required. */  Otherwise further processing may be required. */
572    
573  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
574  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
575  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
576    
577  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
578  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
579  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
580  #endif  #endif
581    
# Line 448  else Line 591  else
591      /* 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
592      error. */      error. */
593    
594      case 'l':      case CHAR_l:
595      case 'L':      case CHAR_L:
596      case 'N':      case CHAR_N:
597      case 'u':      case CHAR_u:
598      case 'U':      case CHAR_U:
599      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
600      break;      break;
601    
602      /* \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:
603      is an absolute backreference. If negative, it is a relative backreference.  
604      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
605      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
606      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
607      fudge it up by pretending it really was \k. */  
608        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
609        is part of Perl's movement towards a unified syntax for back references. As
610        this is synonymous with \k{name}, we fudge it up by pretending it really
611        was \k.
612    
613        (3) For Oniguruma compatibility we also support \g followed by a name or a
614        number either in angle brackets or in single quotes. However, these are
615        (possibly recursive) subroutine calls, _not_ backreferences. Just return
616        the -ESC_g code (cf \k). */
617    
618        case CHAR_g:
619        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
620          {
621          c = -ESC_g;
622          break;
623          }
624    
625        /* Handle the Perl-compatible cases */
626    
627      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
628        {        {
629        const uschar *p;        const uschar *p;
630        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
631          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
632        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
633          {          {
634          c = -ESC_k;          c = -ESC_k;
635          break;          break;
# Line 479  else Line 639  else
639        }        }
640      else braced = FALSE;      else braced = FALSE;
641    
642      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
643        {        {
644        negated = TRUE;        negated = TRUE;
645        ptr++;        ptr++;
# Line 488  else Line 648  else
648    
649      c = 0;      c = 0;
650      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
651        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
652    
653      if (c == 0 || (braced && *(++ptr) != '}'))      if (c < 0)   /* Integer overflow */
654          {
655          *errorcodeptr = ERR61;
656          break;
657          }
658    
659        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
660        {        {
661        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
662        return 0;        break;
663          }
664    
665        if (c == 0)
666          {
667          *errorcodeptr = ERR58;
668          break;
669        }        }
670    
671      if (negated)      if (negated)
# Line 501  else Line 673  else
673        if (c > bracount)        if (c > bracount)
674          {          {
675          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
676          return 0;          break;
677          }          }
678        c = bracount - (c - 1);        c = bracount - (c - 1);
679        }        }
# Line 521  else Line 693  else
693      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
694      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
695    
696      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:
697      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
698    
699      if (!isclass)      if (!isclass)
700        {        {
701        oldptr = ptr;        oldptr = ptr;
702        c -= '0';        c -= CHAR_0;
703        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
704          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
705          if (c < 0)    /* Integer overflow */
706            {
707            *errorcodeptr = ERR61;
708            break;
709            }
710        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
711          {          {
712          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 542  else Line 719  else
719      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.
720      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
721    
722      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
723        {        {
724        ptr--;        ptr--;
725        c = 0;        c = 0;
# Line 555  else Line 732  else
732      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
733      than 3 octal digits. */      than 3 octal digits. */
734    
735      case '0':      case CHAR_0:
736      c -= '0';      c -= CHAR_0;
737      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
738          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
739      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
740      break;      break;
741    
# Line 566  else Line 743  else
743      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
744      treated as a data character. */      treated as a data character. */
745    
746      case 'x':      case CHAR_x:
747      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
748        {        {
749        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
750        int count = 0;        int count = 0;
# Line 576  else Line 753  else
753        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
754          {          {
755          register int cc = *pt++;          register int cc = *pt++;
756          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
757          count++;          count++;
758    
759  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
760          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
761          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
762  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
763          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
764          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
765  #endif  #endif
766          }          }
767    
768        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
769          {          {
770          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
771          ptr = pt;          ptr = pt;
# Line 604  else Line 781  else
781      c = 0;      c = 0;
782      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
783        {        {
784        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
785        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
786  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
787        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
788        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
789  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
790        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
791        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
792  #endif  #endif
793        }        }
794      break;      break;
# Line 620  else Line 797  else
797      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
798      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
799    
800      case 'c':      case CHAR_c:
801      c = *(++ptr);      c = *(++ptr);
802      if (c == 0)      if (c == 0)
803        {        {
804        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
805        return 0;        break;
806        }        }
807    
808  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
809      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
810      c ^= 0x40;      c ^= 0x40;
811  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
812      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
813      c ^= 0xC0;      c ^= 0xC0;
814  #endif  #endif
815      break;      break;
816    
817      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
818      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
819      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
820      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
821      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
822    
823      default:      default:
824      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 694  if (c == 0) goto ERROR_RETURN; Line 871  if (c == 0) goto ERROR_RETURN;
871  /* \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
872  negation. */  negation. */
873    
874  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
875    {    {
876    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
877      {      {
878      *negptr = TRUE;      *negptr = TRUE;
879      ptr++;      ptr++;
880      }      }
881    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
882      {      {
883      c = *(++ptr);      c = *(++ptr);
884      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
885      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
886      name[i] = c;      name[i] = c;
887      }      }
888    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
889    name[i] = 0;    name[i] = 0;
890    }    }
891    
# Line 730  top = _pcre_utt_size; Line 907  top = _pcre_utt_size;
907  while (bot < top)  while (bot < top)
908    {    {
909    i = (bot + top) >> 1;    i = (bot + top) >> 1;
910    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
911    if (c == 0)    if (c == 0)
912      {      {
913      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 773  is_counted_repeat(const uschar *p) Line 950  is_counted_repeat(const uschar *p)
950  {  {
951  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
952  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
953  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
954    
955  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
956  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
957    
958  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
959  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
960    
961  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
962  }  }
963    
964    
# Line 814  int max = -1; Line 991  int max = -1;
991  /* 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
992  an integer overflow. */  an integer overflow. */
993    
994  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
995  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
996    {    {
997    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 824  if (min < 0 || min > 65535) Line 1001  if (min < 0 || min > 65535)
1001  /* 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.
1002  Also, max must not be less than min. */  Also, max must not be less than min. */
1003    
1004  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1005    {    {
1006    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1007      {      {
1008      max = 0;      max = 0;
1009      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1010      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1011        {        {
1012        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 854  return p; Line 1031  return p;
1031    
1032    
1033  /*************************************************  /*************************************************
1034  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1035  *************************************************/  *************************************************/
1036    
1037  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1038    top-level call starts at the beginning of the pattern. All other calls must
1039    start at a parenthesis. It scans along a pattern's text looking for capturing
1040  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1041  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1042  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. We know that if (?P< is
1043  references to subpatterns. We know that if (?P< is encountered, the name will  encountered, the name will be terminated by '>' because that is checked in the
1044  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1045    capturing group numbers - the (?| feature.
1046    
1047  Arguments:  Arguments:
1048    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1049    count        current count of capturing parens so far encountered    cd           compile background data
1050    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1051    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1052    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1053      count        pointer to the current capturing subpattern number (updated)
1054    
1055  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1056  */  */
1057    
1058  static int  static int
1059  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1060    BOOL xmode)    BOOL xmode, int *count)
1061  {  {
1062  const uschar *thisname;  uschar *ptr = *ptrptr;
1063    int start_count = *count;
1064    int hwm_count = start_count;
1065    BOOL dup_parens = FALSE;
1066    
1067  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1068    dealing with. The very first call may not start with a parenthesis. */
1069    
1070    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1071    {    {
1072    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1073          ptr[2] == CHAR_VERTICAL_LINE)
1074        {
1075        ptr += 3;
1076        dup_parens = TRUE;
1077        }
1078    
1079      /* Handle a normal, unnamed capturing parenthesis */
1080    
1081      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1082        {
1083        *count += 1;
1084        if (name == NULL && *count == lorn) return *count;
1085        ptr++;
1086        }
1087    
1088      /* Handle a condition. If it is an assertion, just carry on so that it
1089      is processed as normal. If not, skip to the closing parenthesis of the
1090      condition (there can't be any nested parens. */
1091    
1092      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1093        {
1094        ptr += 2;
1095        if (ptr[1] != CHAR_QUESTION_MARK)
1096          {
1097          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1098          if (*ptr != 0) ptr++;
1099          }
1100        }
1101    
1102      /* We have either (? or (* and not a condition */
1103    
1104      else
1105        {
1106        ptr += 2;
1107        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1108    
1109        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1110    
1111        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1112            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1113          {
1114          int term;
1115          const uschar *thisname;
1116          *count += 1;
1117          if (name == NULL && *count == lorn) return *count;
1118          term = *ptr++;
1119          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1120          thisname = ptr;
1121          while (*ptr != term) ptr++;
1122          if (name != NULL && lorn == ptr - thisname &&
1123              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1124            return *count;
1125          term++;
1126          }
1127        }
1128      }
1129    
1130    /* Past any initial parenthesis handling, scan for parentheses or vertical
1131    bars. */
1132    
1133    for (; *ptr != 0; ptr++)
1134      {
1135    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1136    
1137    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1138      {      {
1139      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1140      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1141        {        {
1142        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1143        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1144        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1145        }        }
1146      continue;      continue;
1147      }      }
1148    
1149    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1150      are handled for real. If the first character is '^', skip it. Also, if the
1151      first few characters (either before or after ^) are \Q\E or \E we skip them
1152      too. This makes for compatibility with Perl. Note the use of STR macros to
1153      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1154    
1155      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1156        {
1157        BOOL negate_class = FALSE;
1158        for (;;)
1159          {
1160          if (ptr[1] == CHAR_BACKSLASH)
1161            {
1162            if (ptr[2] == CHAR_E)
1163              ptr+= 2;
1164            else if (strncmp((const char *)ptr+2,
1165                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1166              ptr += 4;
1167            else
1168              break;
1169            }
1170          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1171            {
1172            negate_class = TRUE;
1173            ptr++;
1174            }
1175          else break;
1176          }
1177    
1178    if (*ptr == '[')      /* If the next character is ']', it is a data character that must be
1179      {      skipped, except in JavaScript compatibility mode. */
1180      while (*(++ptr) != ']')  
1181        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1182            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1183          ptr++;
1184    
1185        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1186        {        {
1187        if (*ptr == '\\')        if (*ptr == 0) return -1;
1188          if (*ptr == CHAR_BACKSLASH)
1189          {          {
1190          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1191          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1192            {            {
1193            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1194            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1195            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1196            }            }
1197          continue;          continue;
1198          }          }
# Line 921  for (; *ptr != 0; ptr++) Line 1202  for (; *ptr != 0; ptr++)
1202    
1203    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1204    
1205    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1206      {      {
1207      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1208      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1209      continue;      continue;
1210      }      }
1211    
1212    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1213    
1214    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1215      {      {
1216      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1217      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1218      continue;      if (*ptr == 0) goto FAIL_EXIT;
1219        }
1220    
1221      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1222        {
1223        if (dup_parens && *count < hwm_count) *count = hwm_count;
1224        *ptrptr = ptr;
1225        return -1;
1226      }      }
1227    
1228    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1229    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1230        if (*count > hwm_count) hwm_count = *count;
1231        *count = start_count;
1232        }
1233      }
1234    
1235    FAIL_EXIT:
1236    *ptrptr = ptr;
1237    return -1;
1238    }
1239    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1240    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1241    
   count++;  
1242    
1243    if (name == NULL && count == lorn) return count;  /*************************************************
1244    term = *ptr++;  *       Find forward referenced subpattern       *
1245    if (term == '<') term = '>';  *************************************************/
1246    thisname = ptr;  
1247    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1248    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1249        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1250      return count;  returns when it reaches a given numbered subpattern. This is used for forward
1251    references to subpatterns. We used to be able to start this scan from the
1252    current compiling point, using the current count value from cd->bracount, and
1253    do it all in a single loop, but the addition of the possibility of duplicate
1254    subpattern numbers means that we have to scan from the very start, in order to
1255    take account of such duplicates, and to use a recursive function to keep track
1256    of the different types of group.
1257    
1258    Arguments:
1259      cd           compile background data
1260      name         name to seek, or NULL if seeking a numbered subpattern
1261      lorn         name length, or subpattern number if name is NULL
1262      xmode        TRUE if we are in /x mode
1263    
1264    Returns:       the number of the found subpattern, or -1 if not found
1265    */
1266    
1267    static int
1268    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1269    {
1270    uschar *ptr = (uschar *)cd->start_pattern;
1271    int count = 0;
1272    int rc;
1273    
1274    /* If the pattern does not start with an opening parenthesis, the first call
1275    to find_parens_sub() will scan right to the end (if necessary). However, if it
1276    does start with a parenthesis, find_parens_sub() will return when it hits the
1277    matching closing parens. That is why we have to have a loop. */
1278    
1279    for (;;)
1280      {
1281      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1282      if (rc > 0 || *ptr++ == 0) break;
1283    }    }
1284    
1285  return -1;  return rc;
1286  }  }
1287    
1288    
1289    
1290    
1291  /*************************************************  /*************************************************
1292  *      Find first significant op code            *  *      Find first significant op code            *
1293  *************************************************/  *************************************************/
# Line 1013  for (;;) Line 1337  for (;;)
1337    
1338      case OP_CALLOUT:      case OP_CALLOUT:
1339      case OP_CREF:      case OP_CREF:
1340        case OP_NCREF:
1341      case OP_RREF:      case OP_RREF:
1342        case OP_NRREF:
1343      case OP_DEF:      case OP_DEF:
1344      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1345      break;      break;
# Line 1029  for (;;) Line 1355  for (;;)
1355    
1356    
1357  /*************************************************  /*************************************************
1358  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1359  *************************************************/  *************************************************/
1360    
1361  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1362  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1363  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1364    temporarily terminated with OP_END when this function is called.
1365    
1366    This function is called when a backward assertion is encountered, so that if it
1367    fails, the error message can point to the correct place in the pattern.
1368    However, we cannot do this when the assertion contains subroutine calls,
1369    because they can be forward references. We solve this by remembering this case
1370    and doing the check at the end; a flag specifies which mode we are running in.
1371    
1372  Arguments:  Arguments:
1373    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1374    options  the compiling options    options  the compiling options
1375      atend    TRUE if called when the pattern is complete
1376      cd       the "compile data" structure
1377    
1378  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1379                 or -1 if there is no fixed length,
1380               or -2 if \C was encountered               or -2 if \C was encountered
1381                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1382  */  */
1383    
1384  static int  static int
1385  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1386  {  {
1387  int length = -1;  int length = -1;
1388    
# Line 1058  branch, check the length against that of Line 1395  branch, check the length against that of
1395  for (;;)  for (;;)
1396    {    {
1397    int d;    int d;
1398      uschar *ce, *cs;
1399    register int op = *cc;    register int op = *cc;
   
1400    switch (op)    switch (op)
1401      {      {
1402      case OP_CBRA:      case OP_CBRA:
1403      case OP_BRA:      case OP_BRA:
1404      case OP_ONCE:      case OP_ONCE:
1405      case OP_COND:      case OP_COND:
1406      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1407      if (d < 0) return d;      if (d < 0) return d;
1408      branchlength += d;      branchlength += d;
1409      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1089  for (;;) Line 1426  for (;;)
1426      branchlength = 0;      branchlength = 0;
1427      break;      break;
1428    
1429        /* A true recursion implies not fixed length, but a subroutine call may
1430        be OK. If the subroutine is a forward reference, we can't deal with
1431        it until the end of the pattern, so return -3. */
1432    
1433        case OP_RECURSE:
1434        if (!atend) return -3;
1435        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1436        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1437        if (cc > cs && cc < ce) return -1;                /* Recursion */
1438        d = find_fixedlength(cs + 2, options, atend, cd);
1439        if (d < 0) return d;
1440        branchlength += d;
1441        cc += 1 + LINK_SIZE;
1442        break;
1443    
1444      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1445    
1446      case OP_ASSERT:      case OP_ASSERT:
# Line 1102  for (;;) Line 1454  for (;;)
1454    
1455      case OP_REVERSE:      case OP_REVERSE:
1456      case OP_CREF:      case OP_CREF:
1457        case OP_NCREF:
1458      case OP_RREF:      case OP_RREF:
1459        case OP_NRREF:
1460      case OP_DEF:      case OP_DEF:
1461      case OP_OPT:      case OP_OPT:
1462      case OP_CALLOUT:      case OP_CALLOUT:
1463      case OP_SOD:      case OP_SOD:
1464      case OP_SOM:      case OP_SOM:
1465        case OP_SET_SOM:
1466      case OP_EOD:      case OP_EOD:
1467      case OP_EODN:      case OP_EODN:
1468      case OP_CIRC:      case OP_CIRC:
# Line 1125  for (;;) Line 1480  for (;;)
1480      branchlength++;      branchlength++;
1481      cc += 2;      cc += 2;
1482  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1483      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1484        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1485  #endif  #endif
1486      break;      break;
1487    
# Line 1139  for (;;) Line 1492  for (;;)
1492      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1493      cc += 4;      cc += 4;
1494  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1495      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1496        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1497  #endif  #endif
1498      break;      break;
1499    
1500      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1501      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1502        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1503      cc += 4;      cc += 4;
1504      break;      break;
1505    
# Line 1165  for (;;) Line 1517  for (;;)
1517      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1518      case OP_WORDCHAR:      case OP_WORDCHAR:
1519      case OP_ANY:      case OP_ANY:
1520        case OP_ALLANY:
1521      branchlength++;      branchlength++;
1522      cc++;      cc++;
1523      break;      break;
# Line 1219  for (;;) Line 1572  for (;;)
1572    
1573    
1574  /*************************************************  /*************************************************
1575  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1576  *************************************************/  *************************************************/
1577    
1578  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1579  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1580    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1581    so that it can be called from pcre_study() when finding the minimum matching
1582    length.
1583    
1584  Arguments:  Arguments:
1585    code        points to start of expression    code        points to start of expression
1586    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1587    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1588    
1589  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
1590  */  */
1591    
1592  static const uschar *  const uschar *
1593  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1594  {  {
1595  for (;;)  for (;;)
1596    {    {
# Line 1247  for (;;) Line 1603  for (;;)
1603    
1604    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1605    
1606      /* Handle recursion */
1607    
1608      else if (c == OP_REVERSE)
1609        {
1610        if (number < 0) return (uschar *)code;
1611        code += _pcre_OP_lengths[c];
1612        }
1613    
1614    /* Handle capturing bracket */    /* Handle capturing bracket */
1615    
1616    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1256  for (;;) Line 1620  for (;;)
1620      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1621      }      }
1622    
1623    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* Otherwise, we can get the item's length from the table, except that for
1624    a multi-byte character. The length in the table is a minimum, so we have to    repeated character types, we have to test for \p and \P, which have an extra
1625    arrange to skip the extra bytes. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1626      must add in its length. */
1627    
1628    else    else
1629      {      {
1630        switch(c)
1631          {
1632          case OP_TYPESTAR:
1633          case OP_TYPEMINSTAR:
1634          case OP_TYPEPLUS:
1635          case OP_TYPEMINPLUS:
1636          case OP_TYPEQUERY:
1637          case OP_TYPEMINQUERY:
1638          case OP_TYPEPOSSTAR:
1639          case OP_TYPEPOSPLUS:
1640          case OP_TYPEPOSQUERY:
1641          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1642          break;
1643    
1644          case OP_TYPEUPTO:
1645          case OP_TYPEMINUPTO:
1646          case OP_TYPEEXACT:
1647          case OP_TYPEPOSUPTO:
1648          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1649          break;
1650    
1651          case OP_MARK:
1652          case OP_PRUNE_ARG:
1653          case OP_SKIP_ARG:
1654          case OP_THEN_ARG:
1655          code += code[1];
1656          break;
1657          }
1658    
1659        /* Add in the fixed length from the table */
1660    
1661      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1662    
1663      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1664      a multi-byte character. The length in the table is a minimum, so we have to
1665      arrange to skip the extra bytes. */
1666    
1667  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1668      if (utf8) switch(c)      if (utf8) switch(c)
1669        {        {
# Line 1284  for (;;) Line 1685  for (;;)
1685        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1686        break;        break;
1687        }        }
1688    #else
1689        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1690  #endif  #endif
1691      }      }
1692    }    }
# Line 1320  for (;;) Line 1723  for (;;)
1723    
1724    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1725    
1726    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
1727    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
1728    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1729    bytes. */    must add in its length. */
1730    
1731    else    else
1732      {      {
1733        switch(c)
1734          {
1735          case OP_TYPESTAR:
1736          case OP_TYPEMINSTAR:
1737          case OP_TYPEPLUS:
1738          case OP_TYPEMINPLUS:
1739          case OP_TYPEQUERY:
1740          case OP_TYPEMINQUERY:
1741          case OP_TYPEPOSSTAR:
1742          case OP_TYPEPOSPLUS:
1743          case OP_TYPEPOSQUERY:
1744          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1745          break;
1746    
1747          case OP_TYPEPOSUPTO:
1748          case OP_TYPEUPTO:
1749          case OP_TYPEMINUPTO:
1750          case OP_TYPEEXACT:
1751          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1752          break;
1753    
1754          case OP_MARK:
1755          case OP_PRUNE_ARG:
1756          case OP_SKIP_ARG:
1757          case OP_THEN_ARG:
1758          code += code[1];
1759          break;
1760          }
1761    
1762        /* Add in the fixed length from the table */
1763    
1764      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1765    
1766        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1767        by a multi-byte character. The length in the table is a minimum, so we have
1768        to arrange to skip the extra bytes. */
1769    
1770  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1771      if (utf8) switch(c)      if (utf8) switch(c)
1772        {        {
# Line 1349  for (;;) Line 1788  for (;;)
1788        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1789        break;        break;
1790        }        }
1791    #else
1792        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1793  #endif  #endif
1794      }      }
1795    }    }
# Line 1364  for (;;) Line 1805  for (;;)
1805  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()
1806  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
1807  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1808  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
1809  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
1810    bracket whose current branch will already have been scanned.
1811    
1812  Arguments:  Arguments:
1813    code        points to start of search    code        points to start of search
1814    endcode     points to where to stop    endcode     points to where to stop
1815    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1816      cd          contains pointers to tables etc.
1817    
1818  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1819  */  */
1820    
1821  static BOOL  static BOOL
1822  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1823      compile_data *cd)
1824  {  {
1825  register int c;  register int c;
1826  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
# Line 1387  for (code = first_significant_code(code Line 1831  for (code = first_significant_code(code
1831    
1832    c = *code;    c = *code;
1833    
1834      /* Skip over forward assertions; the other assertions are skipped by
1835      first_significant_code() with a TRUE final argument. */
1836    
1837      if (c == OP_ASSERT)
1838        {
1839        do code += GET(code, 1); while (*code == OP_ALT);
1840        c = *code;
1841        continue;
1842        }
1843    
1844    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1845    
1846    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1847      {      {
1848      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1849      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1397  for (code = first_significant_code(code Line 1851  for (code = first_significant_code(code
1851      continue;      continue;
1852      }      }
1853    
1854      /* For a recursion/subroutine call, if its end has been reached, which
1855      implies a subroutine call, we can scan it. */
1856    
1857      if (c == OP_RECURSE)
1858        {
1859        BOOL empty_branch = FALSE;
1860        const uschar *scode = cd->start_code + GET(code, 1);
1861        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1862        do
1863          {
1864          if (could_be_empty_branch(scode, endcode, utf8, cd))
1865            {
1866            empty_branch = TRUE;
1867            break;
1868            }
1869          scode += GET(scode, 1);
1870          }
1871        while (*scode == OP_ALT);
1872        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1873        continue;
1874        }
1875    
1876    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1877    
1878    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1879      {      {
1880      BOOL empty_branch;      BOOL empty_branch;
1881      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1882    
1883      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1884        empty branch, so just skip over the conditional, because it could be empty.
1885        Otherwise, scan the individual branches of the group. */
1886    
1887      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;  
1888        code += GET(code, 1);        code += GET(code, 1);
1889        else
1890          {
1891          empty_branch = FALSE;
1892          do
1893            {
1894            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1895              empty_branch = TRUE;
1896            code += GET(code, 1);
1897            }
1898          while (*code == OP_ALT);
1899          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1900        }        }
1901      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1902      c = *code;      c = *code;
1903      continue;      continue;
1904      }      }
# Line 1423  for (code = first_significant_code(code Line 1907  for (code = first_significant_code(code
1907    
1908    switch (c)    switch (c)
1909      {      {
1910      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1911        cannot be represented just by a bit map. This includes negated single
1912        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1913        actual length is stored in the compiled code, so we must update "code"
1914        here. */
1915    
1916  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1917      case OP_XCLASS:      case OP_XCLASS:
1918      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1919      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1920  #endif  #endif
1921    
# Line 1471  for (code = first_significant_code(code Line 1959  for (code = first_significant_code(code
1959      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1960      case OP_WORDCHAR:      case OP_WORDCHAR:
1961      case OP_ANY:      case OP_ANY:
1962        case OP_ALLANY:
1963      case OP_ANYBYTE:      case OP_ANYBYTE:
1964      case OP_CHAR:      case OP_CHAR:
1965      case OP_CHARNC:      case OP_CHARNC:
# Line 1489  for (code = first_significant_code(code Line 1978  for (code = first_significant_code(code
1978      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1979      return FALSE;      return FALSE;
1980    
1981        /* These are going to continue, as they may be empty, but we have to
1982        fudge the length for the \p and \P cases. */
1983    
1984        case OP_TYPESTAR:
1985        case OP_TYPEMINSTAR:
1986        case OP_TYPEPOSSTAR:
1987        case OP_TYPEQUERY:
1988        case OP_TYPEMINQUERY:
1989        case OP_TYPEPOSQUERY:
1990        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1991        break;
1992    
1993        /* Same for these */
1994    
1995        case OP_TYPEUPTO:
1996        case OP_TYPEMINUPTO:
1997        case OP_TYPEPOSUPTO:
1998        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1999        break;
2000    
2001      /* End of branch */      /* End of branch */
2002    
2003      case OP_KET:      case OP_KET:
# Line 1507  for (code = first_significant_code(code Line 2016  for (code = first_significant_code(code
2016      case OP_QUERY:      case OP_QUERY:
2017      case OP_MINQUERY:      case OP_MINQUERY:
2018      case OP_POSQUERY:      case OP_POSQUERY:
2019        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2020        break;
2021    
2022      case OP_UPTO:      case OP_UPTO:
2023      case OP_MINUPTO:      case OP_MINUPTO:
2024      case OP_POSUPTO:      case OP_POSUPTO:
2025      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2026      break;      break;
2027  #endif  #endif
2028    
2029        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2030        string. */
2031    
2032        case OP_MARK:
2033        case OP_PRUNE_ARG:
2034        case OP_SKIP_ARG:
2035        case OP_THEN_ARG:
2036        code += code[1];
2037        break;
2038    
2039        /* None of the remaining opcodes are required to match a character. */
2040    
2041        default:
2042        break;
2043      }      }
2044    }    }
2045    
# Line 1535  Arguments: Line 2062  Arguments:
2062    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2063    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2064    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2065      cd          pointers to tables etc
2066    
2067  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2068  */  */
2069    
2070  static BOOL  static BOOL
2071  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2072    BOOL utf8)    BOOL utf8, compile_data *cd)
2073  {  {
2074  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2075    {    {
2076    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2077        return FALSE;
2078    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2079    }    }
2080  return TRUE;  return TRUE;
# Line 1558  return TRUE; Line 2087  return TRUE;
2087  *************************************************/  *************************************************/
2088    
2089  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2090  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
2091  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2092  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2093    
2094    Originally, this function only recognized a sequence of letters between the
2095    terminators, but it seems that Perl recognizes any sequence of characters,
2096    though of course unknown POSIX names are subsequently rejected. Perl gives an
2097    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2098    didn't consider this to be a POSIX class. Likewise for [:1234:].
2099    
2100    The problem in trying to be exactly like Perl is in the handling of escapes. We
2101    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2102    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2103    below handles the special case of \], but does not try to do any other escape
2104    processing. This makes it different from Perl for cases such as [:l\ower:]
2105    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2106    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2107    I think.
2108    
2109  Argument:  Arguments:
2110    ptr      pointer to the initial [    ptr      pointer to the initial [
2111    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2112    
2113  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2114  */  */
2115    
2116  static BOOL  static BOOL
2117  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2118  {  {
2119  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2120  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2121  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2122    {    {
2123    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2124    return TRUE;      {
2125        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2126        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2127          {
2128          *endptr = ptr;
2129          return TRUE;
2130          }
2131        }
2132    }    }
2133  return FALSE;  return FALSE;
2134  }  }
# Line 1605  Returns: a value representing the na Line 2153  Returns: a value representing the na
2153  static int  static int
2154  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2155  {  {
2156    const char *pn = posix_names;
2157  register int yield = 0;  register int yield = 0;
2158  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2159    {    {
2160    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2161      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2162      pn += posix_name_lengths[yield] + 1;
2163    yield++;    yield++;
2164    }    }
2165  return -1;  return -1;
# Line 1624  return -1; Line 2174  return -1;
2174  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2175  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2176  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
2177  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
2178  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
2179  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
2180  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
2181  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2182    OP_END.
2183    
2184  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2185  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 1651  adjust_recurse(uschar *group, int adjust Line 2202  adjust_recurse(uschar *group, int adjust
2202    uschar *save_hwm)    uschar *save_hwm)
2203  {  {
2204  uschar *ptr = group;  uschar *ptr = group;
2205    
2206  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2207    {    {
2208    int offset;    int offset;
# Line 1762  get_othercase_range(unsigned int *cptr, Line 2314  get_othercase_range(unsigned int *cptr,
2314  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2315    
2316  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2317    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2318    
2319  if (c > d) return FALSE;  if (c > d) return FALSE;
2320    
# Line 1771  next = othercase + 1; Line 2323  next = othercase + 1;
2323    
2324  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2325    {    {
2326    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2327    next++;    next++;
2328    }    }
2329    
# Line 1817  if ((options & PCRE_EXTENDED) != 0) Line 2369  if ((options & PCRE_EXTENDED) != 0)
2369    for (;;)    for (;;)
2370      {      {
2371      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2372      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2373        {        {
2374        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2375          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1829  if ((options & PCRE_EXTENDED) != 0) Line 2381  if ((options & PCRE_EXTENDED) != 0)
2381  /* 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
2382  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2383    
2384  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2385    {    {
2386    int temperrorcode = 0;    int temperrorcode = 0;
2387    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1854  if ((options & PCRE_EXTENDED) != 0) Line 2406  if ((options & PCRE_EXTENDED) != 0)
2406    for (;;)    for (;;)
2407      {      {
2408      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2409      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2410        {        {
2411        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2412          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1865  if ((options & PCRE_EXTENDED) != 0) Line 2417  if ((options & PCRE_EXTENDED) != 0)
2417    
2418  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2419    
2420  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2421    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2422        return FALSE;
2423    
2424  /* 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
2425  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 1881  if (next >= 0) switch(op_code) Line 2434  if (next >= 0) switch(op_code)
2434    case OP_CHAR:    case OP_CHAR:
2435  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2436    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2437    #else
2438      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2439  #endif  #endif
2440    return item != next;    return item != next;
2441    
# Line 1899  if (next >= 0) switch(op_code) Line 2454  if (next >= 0) switch(op_code)
2454      unsigned int othercase;      unsigned int othercase;
2455      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2456  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2457      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2458  #else  #else
2459      othercase = NOTACHAR;      othercase = NOTACHAR;
2460  #endif  #endif
# Line 1912  if (next >= 0) switch(op_code) Line 2467  if (next >= 0) switch(op_code)
2467    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2468    
2469    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2470    if (item == next) return TRUE;    if (item == next) return TRUE;
2471    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2472  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1921  if (next >= 0) switch(op_code) Line 2475  if (next >= 0) switch(op_code)
2475      unsigned int othercase;      unsigned int othercase;
2476      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2477  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2478      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2479  #else  #else
2480      othercase = NOTACHAR;      othercase = NOTACHAR;
2481  #endif  #endif
# Line 1949  if (next >= 0) switch(op_code) Line 2503  if (next >= 0) switch(op_code)
2503    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2504    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2505    
2506      case OP_HSPACE:
2507      case OP_NOT_HSPACE:
2508      switch(next)
2509        {
2510        case 0x09:
2511        case 0x20:
2512        case 0xa0:
2513        case 0x1680:
2514        case 0x180e:
2515        case 0x2000:
2516        case 0x2001:
2517        case 0x2002:
2518        case 0x2003:
2519        case 0x2004:
2520        case 0x2005:
2521        case 0x2006:
2522        case 0x2007:
2523        case 0x2008:
2524        case 0x2009:
2525        case 0x200A:
2526        case 0x202f:
2527        case 0x205f:
2528        case 0x3000:
2529        return op_code != OP_HSPACE;
2530        default:
2531        return op_code == OP_HSPACE;
2532        }
2533    
2534      case OP_VSPACE:
2535      case OP_NOT_VSPACE:
2536      switch(next)
2537        {
2538        case 0x0a:
2539        case 0x0b:
2540        case 0x0c:
2541        case 0x0d:
2542        case 0x85:
2543        case 0x2028:
2544        case 0x2029:
2545        return op_code != OP_VSPACE;
2546        default:
2547        return op_code == OP_VSPACE;
2548        }
2549    
2550    default:    default:
2551    return FALSE;    return FALSE;
2552    }    }
# Line 1983  switch(op_code) Line 2581  switch(op_code)
2581      case ESC_W:      case ESC_W:
2582      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2583    
2584        case ESC_h:
2585        case ESC_H:
2586        switch(item)
2587          {
2588          case 0x09:
2589          case 0x20:
2590          case 0xa0:
2591          case 0x1680:
2592          case 0x180e:
2593          case 0x2000:
2594          case 0x2001:
2595          case 0x2002:
2596          case 0x2003:
2597          case 0x2004:
2598          case 0x2005:
2599          case 0x2006:
2600          case 0x2007:
2601          case 0x2008:
2602          case 0x2009:
2603          case 0x200A:
2604          case 0x202f:
2605          case 0x205f:
2606          case 0x3000:
2607          return -next != ESC_h;
2608          default:
2609          return -next == ESC_h;
2610          }
2611    
2612        case ESC_v:
2613        case ESC_V:
2614        switch(item)
2615          {
2616          case 0x0a:
2617          case 0x0b:
2618          case 0x0c:
2619          case 0x0d:
2620          case 0x85:
2621          case 0x2028:
2622          case 0x2029:
2623          return -next != ESC_v;
2624          default:
2625          return -next == ESC_v;
2626          }
2627    
2628      default:      default:
2629      return FALSE;      return FALSE;
2630      }      }
2631    
2632    case OP_DIGIT:    case OP_DIGIT:
2633    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2634             next == -ESC_h || next == -ESC_v;
2635    
2636    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2637    return next == -ESC_d;    return next == -ESC_d;
# Line 1997  switch(op_code) Line 2640  switch(op_code)
2640    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2641    
2642    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2643    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2644    
2645      case OP_HSPACE:
2646      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2647    
2648      case OP_NOT_HSPACE:
2649      return next == -ESC_h;
2650    
2651      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2652      case OP_VSPACE:
2653      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2654    
2655      case OP_NOT_VSPACE:
2656      return next == -ESC_v;
2657    
2658    case OP_WORDCHAR:    case OP_WORDCHAR:
2659    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2660    
2661    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2662    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2073  uschar classbits[32]; Line 2729  uschar classbits[32];
2729  BOOL class_utf8;  BOOL class_utf8;
2730  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2731  uschar *class_utf8data;  uschar *class_utf8data;
2732    uschar *class_utf8data_base;
2733  uschar utf8_char[6];  uschar utf8_char[6];
2734  #else  #else
2735  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2736  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2737  #endif  #endif
2738    
2739  #ifdef DEBUG  #ifdef PCRE_DEBUG
2740  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2741  #endif  #endif
2742    
# Line 2112  req_caseopt = ((options & PCRE_CASELESS) Line 2769  req_caseopt = ((options & PCRE_CASELESS)
2769  for (;; ptr++)  for (;; ptr++)
2770    {    {
2771    BOOL negate_class;    BOOL negate_class;
2772      BOOL should_flip_negation;
2773    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2774    BOOL is_quantifier;    BOOL is_quantifier;
2775    BOOL is_recurse;    BOOL is_recurse;
2776    BOOL reset_bracount;    BOOL reset_bracount;
2777    int class_charcount;    int class_charcount;
2778    int class_lastchar;    int class_lastchar;
2779    int newoptions;    int newoptions;
# Line 2137  for (;; ptr++) Line 2795  for (;; ptr++)
2795    
2796    if (lengthptr != NULL)    if (lengthptr != NULL)
2797      {      {
2798  #ifdef DEBUG  #ifdef PCRE_DEBUG
2799      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2800  #endif  #endif
2801      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
2802        {        {
2803        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
2804        goto FAILED;        goto FAILED;
# Line 2153  for (;; ptr++) Line 2811  for (;; ptr++)
2811      */      */
2812    
2813      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2814    
2815        /* Paranoid check for integer overflow */
2816    
2817        if (OFLOW_MAX - *lengthptr < code - last_code)
2818          {
2819          *errorcodeptr = ERR20;
2820          goto FAILED;
2821          }
2822    
2823      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2824      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2825    
# Line 2180  for (;; ptr++) Line 2847  for (;; ptr++)
2847    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
2848    reference list. */    reference list. */
2849    
2850    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
2851      {      {
2852      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
2853      goto FAILED;      goto FAILED;
# Line 2190  for (;; ptr++) Line 2857  for (;; ptr++)
2857    
2858    if (inescq && c != 0)    if (inescq && c != 0)
2859      {      {
2860      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2861        {        {
2862        inescq = FALSE;        inescq = FALSE;
2863        ptr++;        ptr++;
# Line 2216  for (;; ptr++) Line 2883  for (;; ptr++)
2883    /* 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
2884    a quantifier. */    a quantifier. */
2885    
2886    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2887      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2888        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2889    
2890    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2891         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2232  for (;; ptr++) Line 2900  for (;; ptr++)
2900    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2901      {      {
2902      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2903      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2904        {        {
2905        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2906          {          {
# Line 2257  for (;; ptr++) Line 2925  for (;; ptr++)
2925      {      {
2926      /* ===================================================================*/      /* ===================================================================*/
2927      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2928      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2929      case ')':      case CHAR_RIGHT_PARENTHESIS:
2930      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2931      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2932      *codeptr = code;      *codeptr = code;
2933      *ptrptr = ptr;      *ptrptr = ptr;
2934      if (lengthptr != NULL)      if (lengthptr != NULL)
2935        {        {
2936          if (OFLOW_MAX - *lengthptr < code - last_code)
2937            {
2938            *errorcodeptr = ERR20;
2939            goto FAILED;
2940            }
2941        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2942        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2943        }        }
# Line 2275  for (;; ptr++) Line 2948  for (;; ptr++)
2948      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2949      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2950    
2951      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2952      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2953        {        {
2954        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2284  for (;; ptr++) Line 2957  for (;; ptr++)
2957      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2958      break;      break;
2959    
2960      case '$':      case CHAR_DOLLAR_SIGN:
2961      previous = NULL;      previous = NULL;
2962      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2963      break;      break;
# Line 2292  for (;; ptr++) Line 2965  for (;; ptr++)
2965      /* 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
2966      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2967    
2968      case '.':      case CHAR_DOT:
2969      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2970      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2971      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2972      previous = code;      previous = code;
2973      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2974      break;      break;
2975    
2976    
# Line 2312  for (;; ptr++) Line 2985  for (;; ptr++)
2985      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,
2986      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2987      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.
     */  
2988    
2989      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2990        default (Perl) mode, it is treated as a data character. */
2991    
2992        case CHAR_RIGHT_SQUARE_BRACKET:
2993        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2994          {
2995          *errorcodeptr = ERR64;
2996          goto FAILED;
2997          }
2998        goto NORMAL_CHAR;
2999    
3000        case CHAR_LEFT_SQUARE_BRACKET:
3001      previous = code;      previous = code;
3002    
3003      /* 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
3004      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. */
3005    
3006      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3007          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3008            check_posix_syntax(ptr, &tempptr))
3009        {        {
3010        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3011        goto FAILED;        goto FAILED;
3012        }        }
3013    
3014      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
3015        if the first few characters (either before or after ^) are \Q\E or \E we
3016        skip them too. This makes for compatibility with Perl. */
3017    
3018      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3019        for (;;)
3020        {        {
       negate_class = TRUE;  
3021        c = *(++ptr);        c = *(++ptr);
3022          if (c == CHAR_BACKSLASH)
3023            {
3024            if (ptr[1] == CHAR_E)
3025              ptr++;
3026            else if (strncmp((const char *)ptr+1,
3027                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3028              ptr += 3;
3029            else
3030              break;
3031            }
3032          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3033            negate_class = TRUE;
3034          else break;
3035        }        }
3036      else  
3037        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3038        an initial ']' is taken as a data character -- the code below handles
3039        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3040        [^] must match any character, so generate OP_ALLANY. */
3041    
3042        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3043            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3044        {        {
3045        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3046          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3047          zerofirstbyte = firstbyte;
3048          break;
3049        }        }
3050    
3051        /* If a class contains a negative special such as \S, we need to flip the
3052        negation flag at the end, so that support for characters > 255 works
3053        correctly (they are all included in the class). */
3054    
3055        should_flip_negation = FALSE;
3056    
3057      /* 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
3058      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
3059      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2356  for (;; ptr++) Line 3071  for (;; ptr++)
3071  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3072      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3073      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3074        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3075  #endif  #endif
3076    
3077      /* 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 2371  for (;; ptr++) Line 3087  for (;; ptr++)
3087          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3088          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3089          }          }
3090    
3091          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3092          data and reset the pointer. This is so that very large classes that
3093          contain a zillion UTF-8 characters no longer overwrite the work space
3094          (which is on the stack). */
3095    
3096          if (lengthptr != NULL)
3097            {
3098            *lengthptr += class_utf8data - class_utf8data_base;
3099            class_utf8data = class_utf8data_base;
3100            }
3101    
3102  #endif  #endif
3103    
3104        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3105    
3106        if (inescq)        if (inescq)
3107          {          {
3108          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3109            {            {
3110            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3111            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2392  for (;; ptr++) Line 3120  for (;; ptr++)
3120        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3121        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3122    
3123        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3124            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3125            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3126          {          {
3127          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3128          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3129          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3130          uschar pbits[32];          uschar pbits[32];
3131    
3132          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3133            {            {
3134            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3135            goto FAILED;            goto FAILED;
3136            }            }
3137    
3138          ptr += 2;          ptr += 2;
3139          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3140            {            {
3141            local_negate = TRUE;            local_negate = TRUE;
3142              should_flip_negation = TRUE;  /* Note negative special */
3143            ptr++;            ptr++;
3144            }            }
3145    
# Line 2475  for (;; ptr++) Line 3204  for (;; ptr++)
3204    
3205        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3206        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3207        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3208        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so set
3209        to or into the one we are building. We assume they have more than one        class_charcount bigger than one. Unrecognized escapes fall through and
3210        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3211          PCRE_EXTRA is set. */
3212    
3213        if (c == '\\')        if (c == CHAR_BACKSLASH)
3214          {          {
3215          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3216          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3217    
3218          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 */
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */  
3219          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3220            {            {
3221            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3222              {              {
3223              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3224              }              }
3225            else inescq = TRUE;            else inescq = TRUE;
3226            continue;            continue;
3227            }            }
3228            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3229    
3230          if (c < 0)          if (c < 0)
3231            {            {
# Line 2512  for (;; ptr++) Line 3241  for (;; ptr++)
3241              continue;              continue;
3242    
3243              case ESC_D:              case ESC_D:
3244                should_flip_negation = TRUE;
3245              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3246              continue;              continue;
3247    
# Line 2520  for (;; ptr++) Line 3250  for (;; ptr++)
3250              continue;              continue;
3251    
3252              case ESC_W:              case ESC_W:
3253                should_flip_negation = TRUE;
3254              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3255              continue;              continue;
3256    
# Line 2529  for (;; ptr++) Line 3260  for (;; ptr++)
3260              continue;              continue;
3261    
3262              case ESC_S:              case ESC_S:
3263                should_flip_negation = TRUE;
3264              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3265              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3266              continue;              continue;
3267    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
3268              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
3269              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
3270              }              }
# Line 2544  for (;; ptr++) Line 3273  for (;; ptr++)
3273    
3274            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3275                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3276    
3277            /* We need to deal with \H, \h, \V, and \v in both phases because            /* We need to deal with \H, \h, \V, and \v in both phases because
3278            they use extra memory. */            they use extra memory. */
3279    
3280            if (-c == ESC_h)            if (-c == ESC_h)
3281              {              {
3282              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3283              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3284              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
3285  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3286              if (utf8)              if (utf8)
3287                {                {
3288                class_utf8 = TRUE;                class_utf8 = TRUE;
3289                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3290                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3291                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3292                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3293                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3294                class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3295                class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3296                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3297                class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3298                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3299                class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3300                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3301                class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3302                }                }
3303  #endif  #endif
3304              continue;              continue;
3305              }              }
3306    
3307            if (-c == ESC_H)            if (-c == ESC_H)
3308              {              {
# Line 2581  for (;; ptr++) Line 3310  for (;; ptr++)
3310                {                {
3311                int x = 0xff;                int x = 0xff;
3312                switch (c)                switch (c)
3313                  {                  {
3314                  case 0x09/8: x ^= 1 << (0x09%8); break;                  case 0x09/8: x ^= 1 << (0x09%8); break;
3315                  case 0x20/8: x ^= 1 << (0x20%8); break;                  case 0x20/8: x ^= 1 << (0x20%8); break;
3316                  case 0xa0/8: x ^= 1 << (0xa0%8); break;                  case 0xa0/8: x ^= 1 << (0xa0%8); break;
3317                  default: break;                  default: break;
3318                  }                  }
3319                classbits[c] |= x;                classbits[c] |= x;
3320                }                }
3321    
3322  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3323              if (utf8)              if (utf8)
3324                {                {
3325                class_utf8 = TRUE;                class_utf8 = TRUE;
3326                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3327                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3328                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3329                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3330                class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3331                class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3332                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3333                class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3334                class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3335                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3336                class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3337                class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3338                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3339                class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3340                class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3341                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3342                class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3343                class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3344                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3345                class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3346                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3347                }                }
3348  #endif  #endif
3349              continue;              continue;
3350              }              }
3351    
3352            if (-c == ESC_v)            if (-c == ESC_v)
3353              {              {
3354              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3355              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3356              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
3357              SETBIT(classbits, 0x0d); /* CR */              SETBIT(classbits, 0x0d); /* CR */
3358              SETBIT(classbits, 0x85); /* NEL */              SETBIT(classbits, 0x85); /* NEL */
3359  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3360              if (utf8)              if (utf8)
3361                {                {
3362                class_utf8 = TRUE;                class_utf8 = TRUE;
3363                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3364                class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3365                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3366                }                }
3367  #endif  #endif
3368              continue;              continue;
3369              }              }
3370    
3371            if (-c == ESC_V)            if (-c == ESC_V)
3372              {              {
# Line 2645  for (;; ptr++) Line 3374  for (;; ptr++)
3374                {                {
3375                int x = 0xff;                int x = 0xff;
3376                switch (c)                switch (c)
3377                  {                  {
3378                  case 0x0a/8: x ^= 1 << (0x0a%8);                  case 0x0a/8: x ^= 1 << (0x0a%8);
3379                               x ^= 1 << (0x0b%8);                               x ^= 1 << (0x0b%8);
3380                               x ^= 1 << (0x0c%8);                               x ^= 1 << (0x0c%8);
3381                               x ^= 1 << (0x0d%8);                               x ^= 1 << (0x0d%8);
3382                               break;                               break;
3383                  case 0x85/8: x ^= 1 << (0x85%8); break;                  case 0x85/8: x ^= 1 << (0x85%8); break;
3384                  default: break;                  default: break;
3385                  }                  }
3386                classbits[c] |= x;                classbits[c] |= x;
3387                }                }
3388    
3389  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3390              if (utf8)              if (utf8)
3391                {                {
3392                class_utf8 = TRUE;                class_utf8 = TRUE;
3393                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3394                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3395                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3396                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3397                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3398                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3399                }                }
3400  #endif  #endif
3401              continue;              continue;
3402              }              }
3403    
3404            /* We need to deal with \P and \p in both phases. */            /* We need to deal with \P and \p in both phases. */
3405    
# Line 2715  for (;; ptr++) Line 3444  for (;; ptr++)
3444        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3445    
3446        CHECK_RANGE:        CHECK_RANGE:
3447        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3448          {          {
3449          inescq = FALSE;          inescq = FALSE;
3450          ptr += 2;          ptr += 2;
# Line 2723  for (;; ptr++) Line 3452  for (;; ptr++)
3452    
3453        oldptr = ptr;        oldptr = ptr;
3454    
3455        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3456    
3457          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3458    
3459          /* Check for range */
3460    
3461          if (!inescq && ptr[1] == CHAR_MINUS)
3462          {          {
3463          int d;          int d;
3464          ptr += 2;          ptr += 2;
3465          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3466    
3467          /* 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
3468          mode. */          mode. */
3469    
3470          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3471            {            {
3472            ptr += 2;            ptr += 2;
3473            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3474                { ptr += 2; continue; }
3475            inescq = TRUE;            inescq = TRUE;
3476            break;            break;
3477            }            }
3478    
3479          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3480            {            {
3481            ptr = oldptr;            ptr = oldptr;
3482            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2759  for (;; ptr++) Line 3495  for (;; ptr++)
3495          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
3496          in such circumstances. */          in such circumstances. */
3497    
3498          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3499            {            {
3500            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3501            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3502    
3503            /* \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
3504            special means the '-' was literal */            special means the '-' was literal */
3505    
3506            if (d < 0)            if (d < 0)
3507              {              {
3508              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3509              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3510              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3511                {                {
3512                ptr = oldptr;                ptr = oldptr;
3513                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2790  for (;; ptr++) Line 3526  for (;; ptr++)
3526    
3527          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3528    
3529            /* Remember \r or \n */
3530    
3531            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3532    
3533          /* 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
3534          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3535          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2812  for (;; ptr++) Line 3552  for (;; ptr++)
3552              unsigned int origd = d;              unsigned int origd = d;
3553              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3554                {                {
3555                if (occ >= (unsigned int)c &&                if (occ >= (unsigned int)c &&
3556                    ocd <= (unsigned int)d)                    ocd <= (unsigned int)d)
3557                  continue;                          /* Skip embedded ranges */                  continue;                          /* Skip embedded ranges */
3558    
3559                if (occ < (unsigned int)c  &&                if (occ < (unsigned int)c  &&
3560                    ocd >= (unsigned int)c - 1)      /* Extend the basic range */                    ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3561                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3562                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3563                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3564                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3565                if (ocd > (unsigned int)d &&                if (ocd > (unsigned int)d &&
3566                    occ <= (unsigned int)d + 1)      /* always shorter than    */                    occ <= (unsigned int)d + 1)      /* always shorter than    */
3567                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3568                  d = ocd;                  d = ocd;
# Line 2908  for (;; ptr++) Line 3648  for (;; ptr++)
3648          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3649            {            {
3650            unsigned int othercase;            unsigned int othercase;
3651            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3652              {              {
3653              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3654              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2935  for (;; ptr++) Line 3675  for (;; ptr++)
3675    
3676      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3677    
3678      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3679    
3680      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3681        {        {
# Line 2943  for (;; ptr++) Line 3683  for (;; ptr++)
3683        goto FAILED;        goto FAILED;
3684        }        }
3685    
3686    
3687    /* This code has been disabled because it would mean that \s counts as
3688    an explicit \r or \n reference, and that's not really what is wanted. Now
3689    we set the flag only if there is a literal "\r" or "\n" in the class. */
3690    
3691    #if 0
3692        /* Remember whether \r or \n are in this class */
3693    
3694        if (negate_class)
3695          {
3696          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3697          }
3698        else
3699          {
3700          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3701          }
3702    #endif
3703    
3704    
3705      /* 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
3706      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
3707      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
3708      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3709      single-bytes only. This is an historical hangover. Maybe one day we can  
3710      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3711        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3712        operate on single-bytes only. This is an historical hangover. Maybe one day
3713        we can tidy these opcodes to handle multi-byte characters.
3714    
3715      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
3716      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2958  for (;; ptr++) Line 3720  for (;; ptr++)
3720      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3721    
3722  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3723      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3724            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3725  #else  #else
3726      if (class_charcount == 1)      if (class_charcount == 1)
3727  #endif  #endif
# Line 3004  for (;; ptr++) Line 3764  for (;; ptr++)
3764      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3765    
3766      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3767      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3768      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
3769        the class, so any that were explicitly given as well can be ignored. If
3770        (when there are explicit characters > 255 that must be listed) there are no
3771        characters < 256, we can omit the bitmap in the actual compiled code. */
3772    
3773  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3774      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3775        {        {
3776        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3777        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3034  for (;; ptr++) Line 3797  for (;; ptr++)
3797        }        }
3798  #endif  #endif
3799    
3800      /* 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
3801      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
3802      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
3803      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3804    
3805        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3806      if (negate_class)      if (negate_class)
3807        {        {
       *code++ = OP_NCLASS;  
3808        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3809          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3810        }        }
3811      else      else
3812        {        {
       *code++ = OP_CLASS;  
3813        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3814        }        }
3815      code += 32;      code += 32;
# Line 3058  for (;; ptr++) Line 3820  for (;; ptr++)
3820      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3821      has been tested above. */      has been tested above. */
3822    
3823      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3824      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3825      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3826      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3827      goto REPEAT;      goto REPEAT;
3828    
3829      case '*':      case CHAR_ASTERISK:
3830      repeat_min = 0;      repeat_min = 0;
3831      repeat_max = -1;      repeat_max = -1;
3832      goto REPEAT;      goto REPEAT;
3833    
3834      case '+':      case CHAR_PLUS:
3835      repeat_min = 1;      repeat_min = 1;
3836      repeat_max = -1;      repeat_max = -1;
3837      goto REPEAT;      goto REPEAT;
3838    
3839      case '?':      case CHAR_QUESTION_MARK:
3840      repeat_min = 0;      repeat_min = 0;
3841      repeat_max = 1;      repeat_max = 1;
3842    
# Line 3109  for (;; ptr++) Line 3871  for (;; ptr++)
3871      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
3872      repeat type to the non-default. */      repeat type to the non-default. */
3873    
3874      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3875        {        {
3876        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3877        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3878        ptr++;        ptr++;
3879        }        }
3880      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3881        {        {
3882        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3883        ptr++;        ptr++;
# Line 3230  for (;; ptr++) Line 3992  for (;; ptr++)
3992    
3993        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3994    
3995          /*--------------------------------------------------------------------*/
3996          /* This code is obsolete from release 8.00; the restriction was finally
3997          removed: */
3998    
3999        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4000        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4001    
4002        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4003          /*--------------------------------------------------------------------*/
4004    
4005        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4006    
# Line 3380  for (;; ptr++) Line 4147  for (;; ptr++)
4147          goto END_REPEAT;          goto END_REPEAT;
4148          }          }
4149    
4150          /*--------------------------------------------------------------------*/
4151          /* This code is obsolete from release 8.00; the restriction was finally
4152          removed: */
4153    
4154        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4155        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4156    
4157        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4158          /*--------------------------------------------------------------------*/
4159    
4160        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4161          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3419  for (;; ptr++) Line 4191  for (;; ptr++)
4191          goto FAILED;          goto FAILED;
4192          }          }
4193    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
4194        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
4195        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
4196        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 3449  for (;; ptr++) Line 4213  for (;; ptr++)
4213    
4214        if (repeat_min == 0)        if (repeat_min == 0)
4215          {          {
4216          /* 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
4217          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4218    
4219          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4220          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4221          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4222          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4223          the start of the whole regex. Temporarily terminate the pattern while          **   }
4224          doing this. */  
4225            However, that fails when a group is referenced as a subroutine from
4226            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4227            so that it is skipped on execution. As we don't have a list of which
4228            groups are referenced, we cannot do this selectively.
4229    
4230            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4231            and do no more at this point. However, we do need to adjust any
4232            OP_RECURSE calls inside the group that refer to the group itself or any
4233            internal or forward referenced group, because the offset is from the
4234            start of the whole regex. Temporarily terminate the pattern while doing
4235            this. */
4236    
4237          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4238            {            {
4239            *code = OP_END;            *code = OP_END;
4240            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4241            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4242            code++;            code++;
4243              if (repeat_max == 0)
4244                {
4245                *previous++ = OP_SKIPZERO;
4246                goto END_REPEAT;
4247                }
4248            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4249            }            }
4250    
# Line 3515  for (;; ptr++) Line 4289  for (;; ptr++)
4289          if (repeat_min > 1)          if (repeat_min > 1)
4290            {            {
4291            /* In the pre-compile phase, we don't actually do the replication. We            /* In the pre-compile phase, we don't actually do the replication. We
4292            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4293              potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4294              integer type when available, otherwise double. */
4295    
4296            if (lengthptr != NULL)            if (lengthptr != NULL)
4297              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4298                int delta = (repeat_min - 1)*length_prevgroup;
4299                if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4300                      (INT64_OR_DOUBLE)length_prevgroup >
4301                        (INT64_OR_DOUBLE)INT_MAX ||
4302                    OFLOW_MAX - *lengthptr < delta)
4303                  {
4304                  *errorcodeptr = ERR20;
4305                  goto FAILED;
4306                  }
4307                *lengthptr += delta;
4308                }
4309    
4310            /* This is compiling for real */            /* This is compiling for real */
4311    
# Line 3556  for (;; ptr++) Line 4343  for (;; ptr++)
4343          /* In the pre-compile phase, we don't actually do the replication. We          /* In the pre-compile phase, we don't actually do the replication. We
4344          just adjust the length as if we had. For each repetition we must add 1          just adjust the length as if we had. For each repetition we must add 1
4345          to the length for BRAZERO and for all but the last repetition we must          to the length for BRAZERO and for all but the last repetition we must
4346          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4347            paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4348            a 64-bit integer type when available, otherwise double. */
4349    
4350          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4351            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4352              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4353                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4354              if ((INT64_OR_DOUBLE)repeat_max *
4355                    (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4356                      > (INT64_OR_DOUBLE)INT_MAX ||
4357                  OFLOW_MAX - *lengthptr < delta)
4358                {
4359                *errorcodeptr = ERR20;
4360                goto FAILED;
4361                }
4362              *lengthptr += delta;
4363              }
4364    
4365          /* This is compiling for real */          /* This is compiling for real */
4366    
# Line 3630  for (;; ptr++) Line 4430  for (;; ptr++)
4430            uschar *scode = bracode;            uschar *scode = bracode;
4431            do            do
4432              {              {
4433              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4434                {                {
4435                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4436                break;                break;
# Line 3642  for (;; ptr++) Line 4442  for (;; ptr++)
4442          }          }
4443        }        }
4444    
4445        /* If previous is OP_FAIL, it was generated by an empty class [] in
4446        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4447        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4448        error above. We can just ignore the repeat in JS case. */
4449    
4450        else if (*previous == OP_FAIL) goto END_REPEAT;
4451    
4452      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4453    
4454      else      else
# Line 3666  for (;; ptr++) Line 4473  for (;; ptr++)
4473      if (possessive_quantifier)      if (possessive_quantifier)
4474        {        {
4475        int len;        int len;
4476        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4477            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4478            tempcode += _pcre_OP_lengths[*tempcode] +
4479              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4480    
4481          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4482            {
4483          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4484    #ifdef SUPPORT_UTF8
4485            if (utf8 && tempcode[-1] >= 0xc0)
4486              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4487    #endif
4488            }
4489    
4490        len = code - tempcode;        len = code - tempcode;
4491        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4492          {          {
# Line 3687  for (;; ptr++) Line 4505  for (;; ptr++)
4505          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4506          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4507    
4508            /* Because we are moving code along, we must ensure that any
4509            pending recursive references are updated. */
4510    
4511          default:          default:
4512            *code = OP_END;
4513            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4514          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4515          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4516          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 3712  for (;; ptr++) Line 4535  for (;; ptr++)
4535      /* ===================================================================*/      /* ===================================================================*/
4536      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4537      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4538      parenthesis forms. First deal with the specials; all are introduced by ?,      parenthesis forms.  */
     and the appearance of any of them means that this is not a capturing  
     group. */  
4539    
4540      case '(':      case CHAR_LEFT_PARENTHESIS:
4541      newoptions = options;      newoptions = options;
4542      skipbytes = 0;      skipbytes = 0;
4543      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4544      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4545      reset_bracount = FALSE;      reset_bracount = FALSE;
4546    
4547        /* First deal with various "verbs" that can be introduced by '*'. */
4548    
4549        if (*(++ptr) == CHAR_ASTERISK &&
4550             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4551          {
4552          int i, namelen;
4553          int arglen = 0;
4554          const char *vn = verbnames;
4555          const uschar *name = ptr + 1;
4556          const uschar *arg = NULL;
4557          previous = NULL;
4558          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4559          namelen = ptr - name;
4560    
4561          if (*ptr == CHAR_COLON)
4562            {
4563            arg = ++ptr;
4564            while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4565              || *ptr == '_') ptr++;
4566            arglen = ptr - arg;
4567            }
4568    
4569          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4570            {
4571            *errorcodeptr = ERR60;
4572            goto FAILED;
4573            }
4574    
4575          /* Scan the table of verb names */
4576    
4577          for (i = 0; i < verbcount; i++)
4578            {
4579            if (namelen == verbs[i].len &&
4580                strncmp((char *)name, vn, namelen) == 0)
4581              {
4582              /* Check for open captures before ACCEPT */
4583    
4584              if (verbs[i].op == OP_ACCEPT)
4585                {
4586                open_capitem *oc;
4587                cd->had_accept = TRUE;
4588                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4589                  {
4590                  *code++ = OP_CLOSE;
4591                  PUT2INC(code, 0, oc->number);
4592                  }
4593                }
4594    
4595              /* Handle the cases with/without an argument */
4596    
4597              if (arglen == 0)
4598                {
4599                if (verbs[i].op < 0)   /* Argument is mandatory */
4600                  {
4601                  *errorcodeptr = ERR66;
4602                  goto FAILED;
4603                  }
4604                *code++ = verbs[i].op;
4605                }
4606    
4607              else
4608                {
4609                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4610                  {
4611                  *errorcodeptr = ERR59;
4612                  goto FAILED;
4613                  }
4614                *code++ = verbs[i].op_arg;
4615                *code++ = arglen;
4616                memcpy(code, arg, arglen);
4617                code += arglen;
4618                *code++ = 0;
4619                }
4620    
4621              break;  /* Found verb, exit loop */
4622              }
4623    
4624            vn += verbs[i].len + 1;
4625            }
4626    
4627      if (*(++ptr) == '?')        if (i < verbcount) continue;    /* Successfully handled a verb */
4628          *errorcodeptr = ERR60;          /* Verb not recognized */
4629          goto FAILED;
4630          }
4631    
4632        /* Deal with the extended parentheses; all are introduced by '?', and the
4633        appearance of any of them means that this is not a capturing group. */
4634    
4635        else if (*ptr == CHAR_QUESTION_MARK)
4636        {        {
4637        int i, set, unset, namelen;        int i, set, unset, namelen;
4638        int *optset;        int *optset;
# Line 3732  for (;; ptr++) Line 4641  for (;; ptr++)
4641    
4642        switch (*(++ptr))        switch (*(++ptr))
4643          {          {
4644          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4645          ptr++;          ptr++;
4646          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4647          if (*ptr == 0)          if (*ptr == 0)
4648            {            {
4649            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3744  for (;; ptr++) Line 4653  for (;; ptr++)
4653    
4654    
4655          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4656          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4657          reset_bracount = TRUE;          reset_bracount = TRUE;
4658          /* Fall through */          /* Fall through */
4659    
4660          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4661          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4662          bravalue = OP_BRA;          bravalue = OP_BRA;
4663          ptr++;          ptr++;
4664          break;          break;
4665    
4666    
4667          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4668          case '(':          case CHAR_LEFT_PARENTHESIS:
4669          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4670    
4671          /* 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 3776  for (;; ptr++) Line 4685  for (;; ptr++)
4685          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4686          including assertions, are processed. */          including assertions, are processed. */
4687    
4688          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4689                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4690            break;            break;
4691    
4692          /* 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 3788  for (;; ptr++) Line 4698  for (;; ptr++)
4698    
4699          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4700    
4701          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4702            {            {
4703            terminator = -1;            terminator = -1;
4704            ptr += 2;            ptr += 2;
# Line 3798  for (;; ptr++) Line 4708  for (;; ptr++)
4708          /* 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
4709          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4710    
4711          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4712            {            {
4713            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4714            ptr++;            ptr++;
4715            }            }
4716          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4717            {            {
4718            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4719            ptr++;            ptr++;
4720            }            }
4721          else          else
4722            {            {
4723            terminator = 0;            terminator = 0;
4724            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4725            }            }
4726    
4727          /* 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 3831  for (;; ptr++) Line 4741  for (;; ptr++)
4741            {            {
4742            if (recno >= 0)            if (recno >= 0)
4743              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4744                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4745            ptr++;            ptr++;
4746            }            }
4747          namelen = ptr - name;          namelen = ptr - name;
4748    
4749          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4750                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4751            {            {
4752            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4753            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 3858  for (;; ptr++) Line 4769  for (;; ptr++)
4769              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4770              goto FAILED;              goto FAILED;
4771              }              }
4772            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
4773                cd->bracount - recno + 1 : recno +cd->bracount;
4774              if (recno <= 0 || recno > cd->final_bracount)
4775              {              {
4776              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4777              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4778              }              }
           else recno += cd->bracount;  
4779            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4780            break;            break;
4781            }            }
4782    
4783          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4784          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4785            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4786            except they record that the reference was originally to a name. The
4787            information is used to check duplicate names. */
4788    
4789          slot = cd->name_table;          slot = cd->name_table;
4790          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 3888  for (;; ptr++) Line 4799  for (;; ptr++)
4799            {            {
4800            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4801            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4802              code[1+LINK_SIZE]++;
4803            }            }
4804    
4805          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4806    
4807          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4808                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4809            {            {
4810            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4811              code[1+LINK_SIZE]++;
4812            }            }
4813    
4814          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 3913  for (;; ptr++) Line 4826  for (;; ptr++)
4826          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4827          specific group number. */          specific group number. */
4828    
4829          else if (*name == 'R')          else if (*name == CHAR_R)
4830            {            {
4831            recno = 0;            recno = 0;
4832            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 3923  for (;; ptr++) Line 4836  for (;; ptr++)
4836                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4837                goto FAILED;                goto FAILED;
4838                }                }
4839              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4840              }              }
4841            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4842            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 3933  for (;; ptr++) Line 4846  for (;; ptr++)
4846          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4847          false. */          false. */
4848    
4849          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4850            {            {
4851            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4852            skipbytes = 1;            skipbytes = 1;
4853            }            }
4854    
4855          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4856            in the second pass here, so final_bracount is set. */
4857    
4858          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4859            {            {
4860            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4861            }            }
# Line 3957  for (;; ptr++) Line 4871  for (;; ptr++)
4871    
4872    
4873          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4874          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4875          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4876          ptr++;          ptr++;
4877          break;          break;
4878    
4879    
4880          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4881          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4882          ptr++;          ptr++;
4883            if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4884              {
4885              *code++ = OP_FAIL;
4886              previous = NULL;
4887              continue;
4888              }
4889            bravalue = OP_ASSERT_NOT;
4890          break;          break;
4891    
4892    
4893          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4894          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4895          switch (ptr[1])          switch (ptr[1])
4896            {            {
4897            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4898            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4899            ptr += 2;            ptr += 2;
4900            break;            break;
4901    
4902            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4903            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4904            ptr += 2;            ptr += 2;
4905            break;            break;
# Line 3994  for (;; ptr++) Line 4914  for (;; ptr++)
4914    
4915    
4916          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4917          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4918          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4919          ptr++;          ptr++;
4920          break;          break;
4921    
4922    
4923          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4924          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4925          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4926          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4927          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4928            {            {
4929            int n = 0;            int n = 0;
4930            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4931              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4932            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4933              {              {
4934              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4935              goto FAILED;              goto FAILED;
# Line 4029  for (;; ptr++) Line 4949  for (;; ptr++)
4949    
4950    
4951          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4952          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4953          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4954                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4955            {            {
4956            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4957            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4958            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4959            }            }
4960          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4961            {            {
4962            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4963            goto FAILED;            goto FAILED;
# Line 4046  for (;; ptr++) Line 4967  for (;; ptr++)
4967    
4968          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4969          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4970          case '\'':          case CHAR_APOSTROPHE:
4971            {            {
4972            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4973                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4974            name = ++ptr;            name = ++ptr;
4975    
4976            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4079  for (;; ptr++) Line 5001  for (;; ptr++)
5001                }                }
5002              }              }
5003    
5004            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5005              alphabetical order. Duplicate names for different numbers are
5006              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5007              number are always OK. (An existing number can be re-used if (?|
5008              appears in the pattern.) In either event, a duplicate name results in
5009              a duplicate entry in the table, even if the number is the same. This
5010              is because the number of names, and hence the table size, is computed
5011              in the pre-compile, and it affects various numbers and pointers which
5012              would all have to be modified, and the compiled code moved down, if
5013              duplicates with the same number were omitted from the table. This
5014              doesn't seem worth the hassle. However, *different* names for the
5015              same number are not permitted. */
5016   &nbs