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

Diff of /code/trunk/pcre_compile.c

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

revision 172 by ph10, Tue Jun 5 10:40:13 2007 UTC revision 454 by ph10, Tue Sep 22 09:42:11 2009 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2009 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 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    
# Line 58  used by pcretest. DEBUG is not defined w Line 61  used by pcretest. DEBUG is not defined w
61  #endif  #endif
62    
63    
64    /* Macro for setting individual bits in class bitmaps. */
65    
66    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
78  *************************************************/  *************************************************/
# Line 82  are simple data values; negative values Line 97  are simple data values; negative values
97  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
98  is invalid. */  is invalid. */
99    
100  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
101    
102    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
103    in UTF-8 mode. */
104    
105  static const short int escapes[] = {  static const short int escapes[] = {
106       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
107       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
108     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
109       0,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
110  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
111  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
112     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
113       0,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
114  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
115       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
116         -ESC_D,                  -ESC_E,
117         0,                       -ESC_G,
118         -ESC_H,                  0,
119         0,                       -ESC_K,
120         0,                       0,
121         0,                       0,
122         -ESC_P,                  -ESC_Q,
123         -ESC_R,                  -ESC_S,
124         0,                       0,
125         -ESC_V,                  -ESC_W,
126         -ESC_X,                  0,
127         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
128         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
129         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
130         CHAR_GRAVE_ACCENT,       7,
131         -ESC_b,                  0,
132         -ESC_d,                  ESC_e,
133         ESC_f,                   0,
134         -ESC_h,                  0,
135         0,                       -ESC_k,
136         0,                       0,
137         ESC_n,                   0,
138         -ESC_p,                  0,
139         ESC_r,                   -ESC_s,
140         ESC_tee,                 0,
141         -ESC_v,                  -ESC_w,
142         0,                       0,
143         -ESC_z
144  };  };
145    
146  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
147    
148    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
149    
150  static const short int escapes[] = {  static const short int escapes[] = {
151  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
152  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 106  static const short int escapes[] = { Line 156  static const short int escapes[] = {
156  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
157  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
158  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
159  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
160  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
161  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
162  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
163  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
164  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
165  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
166  /*  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,
167  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
168  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
169  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
170  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
171  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
172  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
173  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 125  static const short int escapes[] = { Line 175  static const short int escapes[] = {
175  #endif  #endif
176    
177    
178  /* 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
179  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
180  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
181    string is built from string macros so that it works in UTF-8 mode on EBCDIC
182  static const char *const posix_names[] = {  platforms. */
183    "alpha", "lower", "upper",  
184    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
185    "print", "punct", "space", "word",  "xdigit" };    int   len;
186      int   op;
187    } verbitem;
188    
189    static const char verbnames[] =
190      STRING_ACCEPT0
191      STRING_COMMIT0
192      STRING_F0
193      STRING_FAIL0
194      STRING_PRUNE0
195      STRING_SKIP0
196      STRING_THEN;
197    
198    static const verbitem verbs[] = {
199      { 6, OP_ACCEPT },
200      { 6, OP_COMMIT },
201      { 1, OP_FAIL },
202      { 4, OP_FAIL },
203      { 5, OP_PRUNE },
204      { 4, OP_SKIP  },
205      { 4, OP_THEN  }
206    };
207    
208    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
209    
210    
211    /* Tables of names of POSIX character classes and their lengths. The names are
212    now all in a single string, to reduce the number of relocations when a shared
213    library is dynamically loaded. The list of lengths is terminated by a zero
214    length entry. The first three must be alpha, lower, upper, as this is assumed
215    for handling case independence. */
216    
217    static const char posix_names[] =
218      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
219      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
220      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
221      STRING_word0  STRING_xdigit;
222    
223  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
224    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 171  static const int posix_class_maps[] = { Line 257  static const int posix_class_maps[] = {
257  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
258  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
259  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
260  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
261    the number of relocations needed when a shared library is loaded dynamically,
262  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
263    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
264    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
265    "\\c at end of pattern",  because these strings are used only when there is a compilation error. */
266    "unrecognized character follows \\",  
267    "numbers out of order in {} quantifier",  static const char error_texts[] =
268      "no error\0"
269      "\\ at end of pattern\0"
270      "\\c at end of pattern\0"
271      "unrecognized character follows \\\0"
272      "numbers out of order in {} quantifier\0"
273    /* 5 */    /* 5 */
274    "number too big in {} quantifier",    "number too big in {} quantifier\0"
275    "missing terminating ] for character class",    "missing terminating ] for character class\0"
276    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
277    "range out of order in character class",    "range out of order in character class\0"
278    "nothing to repeat",    "nothing to repeat\0"
279    /* 10 */    /* 10 */
280    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
281    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
282    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
283    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
284    "missing )",    "missing )\0"
285    /* 15 */    /* 15 */
286    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
287    "erroffset passed as NULL",    "erroffset passed as NULL\0"
288    "unknown option bit(s) set",    "unknown option bit(s) set\0"
289    "missing ) after comment",    "missing ) after comment\0"
290    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
291    /* 20 */    /* 20 */
292    "regular expression too large",    "regular expression is too large\0"
293    "failed to get memory",    "failed to get memory\0"
294    "unmatched parentheses",    "unmatched parentheses\0"
295    "internal error: code overflow",    "internal error: code overflow\0"
296    "unrecognized character after (?<",    "unrecognized character after (?<\0"
297    /* 25 */    /* 25 */
298    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
299    "malformed number or name after (?(",    "malformed number or name after (?(\0"
300    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
301    "assertion expected after (?(",    "assertion expected after (?(\0"
302    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
303    /* 30 */    /* 30 */
304    "unknown POSIX class name",    "unknown POSIX class name\0"
305    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
306    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
307    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
308    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
309    /* 35 */    /* 35 */
310    "invalid condition (?(0)",    "invalid condition (?(0)\0"
311    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
312    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
313    "number after (?C is > 255",    "number after (?C is > 255\0"
314    "closing ) for (?C expected",    "closing ) for (?C expected\0"
315    /* 40 */    /* 40 */
316    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
317    "unrecognized character after (?P",    "unrecognized character after (?P\0"
318    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
319    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
320    "invalid UTF-8 string",    "invalid UTF-8 string\0"
321    /* 45 */    /* 45 */
322    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
323    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
324    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
325    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
326    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
327    /* 50 */    /* 50 */
328    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
329    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
330    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
331    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
332    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
333    /* 55 */    /* 55 */
334    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
335    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
336    "\\g is not followed by a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
337    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"    "a numbered reference must not be zero\0"
338  };    "(*VERB) with an argument is not supported\0"
339      /* 60 */
340      "(*VERB) not recognized\0"
341      "number is too big\0"
342      "subpattern name expected\0"
343      "digit expected after (?+\0"
344      "] is an invalid data character in JavaScript compatibility mode";
345    
346    
347  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 263  For convenience, we use the same bit def Line 360  For convenience, we use the same bit def
360    
361  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
362    
363  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
364    
365    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
366    UTF-8 mode. */
367    
368  static const unsigned char digitab[] =  static const unsigned char digitab[] =
369    {    {
370    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 299  static const unsigned char digitab[] = Line 400  static const unsigned char digitab[] =
400    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
401    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
402    
403  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
404    
405    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
406    
407  static const unsigned char digitab[] =  static const unsigned char digitab[] =
408    {    {
409    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 374  static const unsigned char ebcdic_charta Line 478  static const unsigned char ebcdic_charta
478  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
479    
480  static BOOL  static BOOL
481    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
482      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
483    
484    
485    
486    /*************************************************
487    *            Find an error text                  *
488    *************************************************/
489    
490    /* The error texts are now all in one long string, to save on relocations. As
491    some of the text is of unknown length, we can't use a table of offsets.
492    Instead, just count through the strings. This is not a performance issue
493    because it happens only when there has been a compilation error.
494    
495    Argument:   the error number
496    Returns:    pointer to the error string
497    */
498    
499    static const char *
500    find_error_text(int n)
501    {
502    const char *s = error_texts;
503    for (; n > 0; n--) while (*s++ != 0) {};
504    return s;
505    }
506    
507    
508  /*************************************************  /*************************************************
# Line 400  Arguments: Line 526  Arguments:
526    
527  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
528                   negative => a special escape sequence                   negative => a special escape sequence
529                   on error, errorptr is set                   on error, errorcodeptr is set
530  */  */
531    
532  static int  static int
# Line 418  ptr--;                            /* Set Line 544  ptr--;                            /* Set
544    
545  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
546    
547  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
548  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
549  Otherwise further processing may be required. */  Otherwise further processing may be required. */
550    
551  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
552  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
553  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
554    
555  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
556  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
557  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
558  #endif  #endif
559    
# Line 443  else Line 569  else
569      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
570      error. */      error. */
571    
572      case 'l':      case CHAR_l:
573      case 'L':      case CHAR_L:
574      case 'N':      case CHAR_N:
575      case 'u':      case CHAR_u:
576      case 'U':      case CHAR_U:
577      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
578      break;      break;
579    
580      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
581      is an absolute backreference. If negative, it is a relative backreference.  
582      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
583      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
584      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
585      fudge it up by pretending it really was \k. */  
586        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
587        is part of Perl's movement towards a unified syntax for back references. As
588        this is synonymous with \k{name}, we fudge it up by pretending it really
589        was \k.
590    
591        (3) For Oniguruma compatibility we also support \g followed by a name or a
592        number either in angle brackets or in single quotes. However, these are
593        (possibly recursive) subroutine calls, _not_ backreferences. Just return
594        the -ESC_g code (cf \k). */
595    
596        case CHAR_g:
597        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
598          {
599          c = -ESC_g;
600          break;
601          }
602    
603        /* Handle the Perl-compatible cases */
604    
605      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
606        {        {
607        const uschar *p;        const uschar *p;
608        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
609          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
610        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
611          {          {
612          c = -ESC_k;          c = -ESC_k;
613          break;          break;
# Line 474  else Line 617  else
617        }        }
618      else braced = FALSE;      else braced = FALSE;
619    
620      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
621        {        {
622        negated = TRUE;        negated = TRUE;
623        ptr++;        ptr++;
# Line 483  else Line 626  else
626    
627      c = 0;      c = 0;
628      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
629        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
630    
631        if (c < 0)   /* Integer overflow */
632          {
633          *errorcodeptr = ERR61;
634          break;
635          }
636    
637      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
638        {        {
639        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
640        return 0;        break;
641          }
642    
643        if (c == 0)
644          {
645          *errorcodeptr = ERR58;
646          break;
647        }        }
648    
649      if (negated)      if (negated)
# Line 496  else Line 651  else
651        if (c > bracount)        if (c > bracount)
652          {          {
653          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
654          return 0;          break;
655          }          }
656        c = bracount - (c - 1);        c = bracount - (c - 1);
657        }        }
# Line 516  else Line 671  else
671      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
672      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
673    
674      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
675      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
676    
677      if (!isclass)      if (!isclass)
678        {        {
679        oldptr = ptr;        oldptr = ptr;
680        c -= '0';        c -= CHAR_0;
681        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
682          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
683          if (c < 0)    /* Integer overflow */
684            {
685            *errorcodeptr = ERR61;
686            break;
687            }
688        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
689          {          {
690          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 537  else Line 697  else
697      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
698      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
699    
700      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
701        {        {
702        ptr--;        ptr--;
703        c = 0;        c = 0;
# Line 550  else Line 710  else
710      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
711      than 3 octal digits. */      than 3 octal digits. */
712    
713      case '0':      case CHAR_0:
714      c -= '0';      c -= CHAR_0;
715      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
716          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
717      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
718      break;      break;
719    
# Line 561  else Line 721  else
721      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
722      treated as a data character. */      treated as a data character. */
723    
724      case 'x':      case CHAR_x:
725      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
726        {        {
727        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
728        int count = 0;        int count = 0;
# Line 571  else Line 731  else
731        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
732          {          {
733          register int cc = *pt++;          register int cc = *pt++;
734          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
735          count++;          count++;
736    
737  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
738          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
739          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
740  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
741          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
742          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
743  #endif  #endif
744          }          }
745    
746        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
747          {          {
748          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
749          ptr = pt;          ptr = pt;
# Line 599  else Line 759  else
759      c = 0;      c = 0;
760      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
761        {        {
762        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
763        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
764  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
765        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
766        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
767  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
768        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
769        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
770  #endif  #endif
771        }        }
772      break;      break;
# Line 615  else Line 775  else
775      This coding is ASCII-specific, but then the whole concept of \cx is      This coding is ASCII-specific, but then the whole concept of \cx is
776      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
777    
778      case 'c':      case CHAR_c:
779      c = *(++ptr);      c = *(++ptr);
780      if (c == 0)      if (c == 0)
781        {        {
782        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
783        return 0;        break;
784        }        }
785    
786  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
787      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
788      c ^= 0x40;      c ^= 0x40;
789  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
790      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
791      c ^= 0xC0;      c ^= 0xC0;
792  #endif  #endif
793      break;      break;
794    
795      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
796      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
797      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
798      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
799      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
800    
801      default:      default:
802      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 689  if (c == 0) goto ERROR_RETURN; Line 849  if (c == 0) goto ERROR_RETURN;
849  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
850  negation. */  negation. */
851    
852  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
853    {    {
854    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
855      {      {
856      *negptr = TRUE;      *negptr = TRUE;
857      ptr++;      ptr++;
858      }      }
859    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
860      {      {
861      c = *(++ptr);      c = *(++ptr);
862      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
863      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
864      name[i] = c;      name[i] = c;
865      }      }
866    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
867    name[i] = 0;    name[i] = 0;
868    }    }
869    
# Line 725  top = _pcre_utt_size; Line 885  top = _pcre_utt_size;
885  while (bot < top)  while (bot < top)
886    {    {
887    i = (bot + top) >> 1;    i = (bot + top) >> 1;
888    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
889    if (c == 0)    if (c == 0)
890      {      {
891      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 768  is_counted_repeat(const uschar *p) Line 928  is_counted_repeat(const uschar *p)
928  {  {
929  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
930  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
931  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
932    
933  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
934  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
935    
936  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
937  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
938    
939  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
940  }  }
941    
942    
# Line 809  int max = -1; Line 969  int max = -1;
969  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
970  an integer overflow. */  an integer overflow. */
971    
972  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
973  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
974    {    {
975    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 819  if (min < 0 || min > 65535) Line 979  if (min < 0 || min > 65535)
979  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
980  Also, max must not be less than min. */  Also, max must not be less than min. */
981    
982  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
983    {    {
984    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
985      {      {
986      max = 0;      max = 0;
987      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
988      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
989        {        {
990        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 849  return p; Line 1009  return p;
1009    
1010    
1011  /*************************************************  /*************************************************
1012  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1013  *************************************************/  *************************************************/
1014    
1015  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1016    top-level call starts at the beginning of the pattern. All other calls must
1017    start at a parenthesis. It scans along a pattern's text looking for capturing
1018  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1019  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1020  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. We know that if (?P< is
1021  references to subpatterns. We know that if (?P< is encountered, the name will  encountered, the name will be terminated by '>' because that is checked in the
1022  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1023    capturing group numbers - the (?| feature.
1024    
1025  Arguments:  Arguments:
1026    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1027    count        current count of capturing parens so far encountered    cd           compile background data
1028    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1029    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1030    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1031      count        pointer to the current capturing subpattern number (updated)
1032    
1033  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1034  */  */
1035    
1036  static int  static int
1037  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1038    BOOL xmode)    BOOL xmode, int *count)
1039  {  {
1040  const uschar *thisname;  uschar *ptr = *ptrptr;
1041    int start_count = *count;
1042    int hwm_count = start_count;
1043    BOOL dup_parens = FALSE;
1044    
1045  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1046    dealing with. The very first call may not start with a parenthesis. */
1047    
1048    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1049    {    {
1050    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1051          ptr[2] == CHAR_VERTICAL_LINE)
1052        {
1053        ptr += 3;
1054        dup_parens = TRUE;
1055        }
1056    
1057      /* Handle a normal, unnamed capturing parenthesis */
1058    
1059      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1060        {
1061        *count += 1;
1062        if (name == NULL && *count == lorn) return *count;
1063        ptr++;
1064        }
1065    
1066      /* Handle a condition. If it is an assertion, just carry on so that it
1067      is processed as normal. If not, skip to the closing parenthesis of the
1068      condition (there can't be any nested parens. */
1069    
1070      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1071        {
1072        ptr += 2;
1073        if (ptr[1] != CHAR_QUESTION_MARK)
1074          {
1075          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1076          if (*ptr != 0) ptr++;
1077          }
1078        }
1079    
1080      /* We have either (? or (* and not a condition */
1081    
1082      else
1083        {
1084        ptr += 2;
1085        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1086    
1087        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1088    
1089        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1090            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1091          {
1092          int term;
1093          const uschar *thisname;
1094          *count += 1;
1095          if (name == NULL && *count == lorn) return *count;
1096          term = *ptr++;
1097          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1098          thisname = ptr;
1099          while (*ptr != term) ptr++;
1100          if (name != NULL && lorn == ptr - thisname &&
1101              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1102            return *count;
1103          term++;
1104          }
1105        }
1106      }
1107    
1108    /* Past any initial parenthesis handling, scan for parentheses or vertical
1109    bars. */
1110    
1111    for (; *ptr != 0; ptr++)
1112      {
1113    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1114    
1115    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1116      {      {
1117      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1118      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1119        {        {
1120        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1121        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1122        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1123        }        }
1124      continue;      continue;
1125      }      }
1126    
1127    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1128      are handled for real. If the first character is '^', skip it. Also, if the
1129      first few characters (either before or after ^) are \Q\E or \E we skip them
1130      too. This makes for compatibility with Perl. Note the use of STR macros to
1131      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1132    
1133      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1134        {
1135        BOOL negate_class = FALSE;
1136        for (;;)
1137          {
1138          if (ptr[1] == CHAR_BACKSLASH)
1139            {
1140            if (ptr[2] == CHAR_E)
1141              ptr+= 2;
1142            else if (strncmp((const char *)ptr+2,
1143                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1144              ptr += 4;
1145            else
1146              break;
1147            }
1148          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1149            {
1150            negate_class = TRUE;
1151            ptr++;
1152            }
1153          else break;
1154          }
1155    
1156    if (*ptr == '[')      /* If the next character is ']', it is a data character that must be
1157      {      skipped, except in JavaScript compatibility mode. */
1158      while (*(++ptr) != ']')  
1159        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1160            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1161          ptr++;
1162    
1163        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1164        {        {
1165        if (*ptr == '\\')        if (*ptr == 0) return -1;
1166          if (*ptr == CHAR_BACKSLASH)
1167          {          {
1168          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1169          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1170            {            {
1171            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1172            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1173            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1174            }            }
1175          continue;          continue;
1176          }          }
# Line 916  for (; *ptr != 0; ptr++) Line 1180  for (; *ptr != 0; ptr++)
1180    
1181    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1182    
1183    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1184      {      {
1185      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1186      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1187      continue;      continue;
1188      }      }
1189    
1190    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1191    
1192    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1193      {      {
1194      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1195      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1196      continue;      if (*ptr == 0) goto FAIL_EXIT;
1197      }      }
1198    
1199    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1200    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1201        if (dup_parens && *count < hwm_count) *count = hwm_count;
1202        *ptrptr = ptr;
1203        return -1;
1204        }
1205    
1206    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1207        {
1208        if (*count > hwm_count) hwm_count = *count;
1209        *count = start_count;
1210        }
1211      }
1212    
1213    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1214         *ptr != '\'')  *ptrptr = ptr;
1215      continue;  return -1;
1216    }
1217    
   count++;  
1218    
1219    if (name == NULL && count == lorn) return count;  
1220    term = *ptr++;  
1221    if (term == '<') term = '>';  /*************************************************
1222    thisname = ptr;  *       Find forward referenced subpattern       *
1223    while (*ptr != term) ptr++;  *************************************************/
1224    if (name != NULL && lorn == ptr - thisname &&  
1225        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  /* This function scans along a pattern's text looking for capturing
1226      return count;  subpatterns, and counting them. If it finds a named pattern that matches the
1227    name it is given, it returns its number. Alternatively, if the name is NULL, it
1228    returns when it reaches a given numbered subpattern. This is used for forward
1229    references to subpatterns. We used to be able to start this scan from the
1230    current compiling point, using the current count value from cd->bracount, and
1231    do it all in a single loop, but the addition of the possibility of duplicate
1232    subpattern numbers means that we have to scan from the very start, in order to
1233    take account of such duplicates, and to use a recursive function to keep track
1234    of the different types of group.
1235    
1236    Arguments:
1237      cd           compile background data
1238      name         name to seek, or NULL if seeking a numbered subpattern
1239      lorn         name length, or subpattern number if name is NULL
1240      xmode        TRUE if we are in /x mode
1241    
1242    Returns:       the number of the found subpattern, or -1 if not found
1243    */
1244    
1245    static int
1246    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1247    {
1248    uschar *ptr = (uschar *)cd->start_pattern;
1249    int count = 0;
1250    int rc;
1251    
1252    /* If the pattern does not start with an opening parenthesis, the first call
1253    to find_parens_sub() will scan right to the end (if necessary). However, if it
1254    does start with a parenthesis, find_parens_sub() will return when it hits the
1255    matching closing parens. That is why we have to have a loop. */
1256    
1257    for (;;)
1258      {
1259      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1260      if (rc > 0 || *ptr++ == 0) break;
1261    }    }
1262    
1263  return -1;  return rc;
1264  }  }
1265    
1266    
1267    
1268    
1269  /*************************************************  /*************************************************
1270  *      Find first significant op code            *  *      Find first significant op code            *
1271  *************************************************/  *************************************************/
# Line 1024  for (;;) Line 1331  for (;;)
1331    
1332    
1333  /*************************************************  /*************************************************
1334  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1335  *************************************************/  *************************************************/
1336    
1337  /* 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,
1338  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.
1339  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
1340    temporarily terminated with OP_END when this function is called.
1341    
1342    This function is called when a backward assertion is encountered, so that if it
1343    fails, the error message can point to the correct place in the pattern.
1344    However, we cannot do this when the assertion contains subroutine calls,
1345    because they can be forward references. We solve this by remembering this case
1346    and doing the check at the end; a flag specifies which mode we are running in.
1347    
1348  Arguments:  Arguments:
1349    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1350    options  the compiling options    options  the compiling options
1351      atend    TRUE if called when the pattern is complete
1352      cd       the "compile data" structure
1353    
1354  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1355                 or -1 if there is no fixed length,
1356               or -2 if \C was encountered               or -2 if \C was encountered
1357                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1358  */  */
1359    
1360  static int  static int
1361  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1362  {  {
1363  int length = -1;  int length = -1;
1364    
# Line 1053  branch, check the length against that of Line 1371  branch, check the length against that of
1371  for (;;)  for (;;)
1372    {    {
1373    int d;    int d;
1374      uschar *ce, *cs;
1375    register int op = *cc;    register int op = *cc;
   
1376    switch (op)    switch (op)
1377      {      {
1378      case OP_CBRA:      case OP_CBRA:
1379      case OP_BRA:      case OP_BRA:
1380      case OP_ONCE:      case OP_ONCE:
1381      case OP_COND:      case OP_COND:
1382      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1383      if (d < 0) return d;      if (d < 0) return d;
1384      branchlength += d;      branchlength += d;
1385      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1083  for (;;) Line 1401  for (;;)
1401      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1402      branchlength = 0;      branchlength = 0;
1403      break;      break;
1404    
1405        /* A true recursion implies not fixed length, but a subroutine call may
1406        be OK. If the subroutine is a forward reference, we can't deal with
1407        it until the end of the pattern, so return -3. */
1408    
1409        case OP_RECURSE:
1410        if (!atend) return -3;
1411        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1412        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1413        if (cc > cs && cc < ce) return -1;                /* Recursion */
1414        d = find_fixedlength(cs + 2, options, atend, cd);
1415        if (d < 0) return d;
1416        branchlength += d;
1417        cc += 1 + LINK_SIZE;
1418        break;
1419    
1420      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1421    
# Line 1120  for (;;) Line 1453  for (;;)
1453      branchlength++;      branchlength++;
1454      cc += 2;      cc += 2;
1455  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1456      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1457        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1458  #endif  #endif
1459      break;      break;
1460    
# Line 1134  for (;;) Line 1465  for (;;)
1465      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1466      cc += 4;      cc += 4;
1467  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1468      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1469        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1470  #endif  #endif
1471      break;      break;
1472    
1473      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1474      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1475        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1476      cc += 4;      cc += 4;
1477      break;      break;
1478    
# Line 1160  for (;;) Line 1490  for (;;)
1490      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1491      case OP_WORDCHAR:      case OP_WORDCHAR:
1492      case OP_ANY:      case OP_ANY:
1493        case OP_ALLANY:
1494      branchlength++;      branchlength++;
1495      cc++;      cc++;
1496      break;      break;
# Line 1214  for (;;) Line 1545  for (;;)
1545    
1546    
1547  /*************************************************  /*************************************************
1548  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1549  *************************************************/  *************************************************/
1550    
1551  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1552  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1553    instance of OP_REVERSE for a lookbehind.
1554    
1555  Arguments:  Arguments:
1556    code        points to start of expression    code        points to start of expression
1557    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1558    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1559    
1560  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
1561  */  */
# Line 1241  for (;;) Line 1573  for (;;)
1573    the table is zero; the actual length is stored in the compiled code. */    the table is zero; the actual length is stored in the compiled code. */
1574    
1575    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1576    
1577      /* Handle recursion */
1578    
1579      else if (c == OP_REVERSE)
1580        {
1581        if (number < 0) return (uschar *)code;
1582        code += _pcre_OP_lengths[c];
1583        }
1584    
1585    /* Handle capturing bracket */    /* Handle capturing bracket */
1586    
# Line 1251  for (;;) Line 1591  for (;;)
1591      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1592      }      }
1593    
1594    /* 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
1595    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
1596    arrange to skip the extra bytes. */    two bytes of parameters. */
1597    
1598    else    else
1599      {      {
1600        switch(c)
1601          {
1602          case OP_TYPESTAR:
1603          case OP_TYPEMINSTAR:
1604          case OP_TYPEPLUS:
1605          case OP_TYPEMINPLUS:
1606          case OP_TYPEQUERY:
1607          case OP_TYPEMINQUERY:
1608          case OP_TYPEPOSSTAR:
1609          case OP_TYPEPOSPLUS:
1610          case OP_TYPEPOSQUERY:
1611          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1612          break;
1613    
1614          case OP_TYPEUPTO:
1615          case OP_TYPEMINUPTO:
1616          case OP_TYPEEXACT:
1617          case OP_TYPEPOSUPTO:
1618          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1619          break;
1620          }
1621    
1622        /* Add in the fixed length from the table */
1623    
1624      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1625    
1626      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1627      a multi-byte character. The length in the table is a minimum, so we have to
1628      arrange to skip the extra bytes. */
1629    
1630  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1631      if (utf8) switch(c)      if (utf8) switch(c)
1632        {        {
# Line 1279  for (;;) Line 1648  for (;;)
1648        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1649        break;        break;
1650        }        }
1651    #else
1652        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1653  #endif  #endif
1654      }      }
1655    }    }
# Line 1315  for (;;) Line 1686  for (;;)
1686    
1687    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1688    
1689    /* 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
1690    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
1691    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1692    
1693    else    else
1694      {      {
1695        switch(c)
1696          {
1697          case OP_TYPESTAR:
1698          case OP_TYPEMINSTAR:
1699          case OP_TYPEPLUS:
1700          case OP_TYPEMINPLUS:
1701          case OP_TYPEQUERY:
1702          case OP_TYPEMINQUERY:
1703          case OP_TYPEPOSSTAR:
1704          case OP_TYPEPOSPLUS:
1705          case OP_TYPEPOSQUERY:
1706          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1707          break;
1708    
1709          case OP_TYPEPOSUPTO:
1710          case OP_TYPEUPTO:
1711          case OP_TYPEMINUPTO:
1712          case OP_TYPEEXACT:
1713          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1714          break;
1715          }
1716    
1717        /* Add in the fixed length from the table */
1718    
1719      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1720    
1721        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1722        by a multi-byte character. The length in the table is a minimum, so we have
1723        to arrange to skip the extra bytes. */
1724    
1725  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1726      if (utf8) switch(c)      if (utf8) switch(c)
1727        {        {
# Line 1344  for (;;) Line 1743  for (;;)
1743        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1744        break;        break;
1745        }        }
1746    #else
1747        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1748  #endif  #endif
1749      }      }
1750    }    }
# Line 1359  for (;;) Line 1760  for (;;)
1760  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()
1761  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
1762  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1763  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
1764  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
1765    bracket whose current branch will already have been scanned.
1766    
1767  Arguments:  Arguments:
1768    code        points to start of search    code        points to start of search
# Line 1382  for (code = first_significant_code(code Line 1784  for (code = first_significant_code(code
1784    
1785    c = *code;    c = *code;
1786    
1787      /* Skip over forward assertions; the other assertions are skipped by
1788      first_significant_code() with a TRUE final argument. */
1789    
1790      if (c == OP_ASSERT)
1791        {
1792        do code += GET(code, 1); while (*code == OP_ALT);
1793        c = *code;
1794        continue;
1795        }
1796    
1797    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1798    
1799    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1800      {      {
1801      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1802      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1394  for (code = first_significant_code(code Line 1806  for (code = first_significant_code(code
1806    
1807    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1808    
1809    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1810      {      {
1811      BOOL empty_branch;      BOOL empty_branch;
1812      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1813    
1814      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1815        empty branch, so just skip over the conditional, because it could be empty.
1816        Otherwise, scan the individual branches of the group. */
1817    
1818      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;  
1819        code += GET(code, 1);        code += GET(code, 1);
1820        else
1821          {
1822          empty_branch = FALSE;
1823          do
1824            {
1825            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1826              empty_branch = TRUE;
1827            code += GET(code, 1);
1828            }
1829          while (*code == OP_ALT);
1830          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1831        }        }
1832      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1833      c = *code;      c = *code;
1834      continue;      continue;
1835      }      }
# Line 1418  for (code = first_significant_code(code Line 1838  for (code = first_significant_code(code
1838    
1839    switch (c)    switch (c)
1840      {      {
1841      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1842        cannot be represented just by a bit map. This includes negated single
1843        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1844        actual length is stored in the compiled code, so we must update "code"
1845        here. */
1846    
1847  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1848      case OP_XCLASS:      case OP_XCLASS:
1849      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1850      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1851  #endif  #endif
1852    
# Line 1466  for (code = first_significant_code(code Line 1890  for (code = first_significant_code(code
1890      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1891      case OP_WORDCHAR:      case OP_WORDCHAR:
1892      case OP_ANY:      case OP_ANY:
1893        case OP_ALLANY:
1894      case OP_ANYBYTE:      case OP_ANYBYTE:
1895      case OP_CHAR:      case OP_CHAR:
1896      case OP_CHARNC:      case OP_CHARNC:
# Line 1484  for (code = first_significant_code(code Line 1909  for (code = first_significant_code(code
1909      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1910      return FALSE;      return FALSE;
1911    
1912        /* These are going to continue, as they may be empty, but we have to
1913        fudge the length for the \p and \P cases. */
1914    
1915        case OP_TYPESTAR:
1916        case OP_TYPEMINSTAR:
1917        case OP_TYPEPOSSTAR:
1918        case OP_TYPEQUERY:
1919        case OP_TYPEMINQUERY:
1920        case OP_TYPEPOSQUERY:
1921        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1922        break;
1923    
1924        /* Same for these */
1925    
1926        case OP_TYPEUPTO:
1927        case OP_TYPEMINUPTO:
1928        case OP_TYPEPOSUPTO:
1929        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1930        break;
1931    
1932      /* End of branch */      /* End of branch */
1933    
1934      case OP_KET:      case OP_KET:
# Line 1502  for (code = first_significant_code(code Line 1947  for (code = first_significant_code(code
1947      case OP_QUERY:      case OP_QUERY:
1948      case OP_MINQUERY:      case OP_MINQUERY:
1949      case OP_POSQUERY:      case OP_POSQUERY:
1950        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1951        break;
1952    
1953      case OP_UPTO:      case OP_UPTO:
1954      case OP_MINUPTO:      case OP_MINUPTO:
1955      case OP_POSUPTO:      case OP_POSUPTO:
1956      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1957      break;      break;
1958  #endif  #endif
1959      }      }
# Line 1553  return TRUE; Line 2001  return TRUE;
2001  *************************************************/  *************************************************/
2002    
2003  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2004  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
2005  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2006  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2007    
2008    Originally, this function only recognized a sequence of letters between the
2009    terminators, but it seems that Perl recognizes any sequence of characters,
2010    though of course unknown POSIX names are subsequently rejected. Perl gives an
2011    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2012    didn't consider this to be a POSIX class. Likewise for [:1234:].
2013    
2014    The problem in trying to be exactly like Perl is in the handling of escapes. We
2015    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2016    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2017    below handles the special case of \], but does not try to do any other escape
2018    processing. This makes it different from Perl for cases such as [:l\ower:]
2019    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2020    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2021    I think.
2022    
2023  Argument:  Arguments:
2024    ptr      pointer to the initial [    ptr      pointer to the initial [
2025    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2026    
2027  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2028  */  */
2029    
2030  static BOOL  static BOOL
2031  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2032  {  {
2033  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2034  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2035  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2036    {    {
2037    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2038    return TRUE;      {
2039        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2040        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2041          {
2042          *endptr = ptr;
2043          return TRUE;
2044          }
2045        }
2046    }    }
2047  return FALSE;  return FALSE;
2048  }  }
# Line 1600  Returns:     a value representing the na Line 2067  Returns:     a value representing the na
2067  static int  static int
2068  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2069  {  {
2070    const char *pn = posix_names;
2071  register int yield = 0;  register int yield = 0;
2072  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2073    {    {
2074    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2075      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2076      pn += posix_name_lengths[yield] + 1;
2077    yield++;    yield++;
2078    }    }
2079  return -1;  return -1;
# Line 1619  return -1; Line 2088  return -1;
2088  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2089  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2090  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
2091  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
2092  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
2093  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
2094  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
2095  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2096    OP_END.
2097    
2098  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2099  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 1646  adjust_recurse(uschar *group, int adjust Line 2116  adjust_recurse(uschar *group, int adjust
2116    uschar *save_hwm)    uschar *save_hwm)
2117  {  {
2118  uschar *ptr = group;  uschar *ptr = group;
2119    
2120  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2121    {    {
2122    int offset;    int offset;
# Line 1757  get_othercase_range(unsigned int *cptr, Line 2228  get_othercase_range(unsigned int *cptr,
2228  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2229    
2230  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2231    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2232    
2233  if (c > d) return FALSE;  if (c > d) return FALSE;
2234    
# Line 1766  next = othercase + 1; Line 2237  next = othercase + 1;
2237    
2238  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2239    {    {
2240    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2241    next++;    next++;
2242    }    }
2243    
# Line 1812  if ((options & PCRE_EXTENDED) != 0) Line 2283  if ((options & PCRE_EXTENDED) != 0)
2283    for (;;)    for (;;)
2284      {      {
2285      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2286      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2287        {        {
2288        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2289          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1824  if ((options & PCRE_EXTENDED) != 0) Line 2295  if ((options & PCRE_EXTENDED) != 0)
2295  /* 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
2296  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2297    
2298  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2299    {    {
2300    int temperrorcode = 0;    int temperrorcode = 0;
2301    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1849  if ((options & PCRE_EXTENDED) != 0) Line 2320  if ((options & PCRE_EXTENDED) != 0)
2320    for (;;)    for (;;)
2321      {      {
2322      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2323      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2324        {        {
2325        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2326          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1860  if ((options & PCRE_EXTENDED) != 0) Line 2331  if ((options & PCRE_EXTENDED) != 0)
2331    
2332  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2333    
2334  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2335    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2336        return FALSE;
2337    
2338  /* 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
2339  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 1876  if (next >= 0) switch(op_code) Line 2348  if (next >= 0) switch(op_code)
2348    case OP_CHAR:    case OP_CHAR:
2349  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2350    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2351    #else
2352      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2353  #endif  #endif
2354    return item != next;    return item != next;
2355    
# Line 1894  if (next >= 0) switch(op_code) Line 2368  if (next >= 0) switch(op_code)
2368      unsigned int othercase;      unsigned int othercase;
2369      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2370  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2371      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2372  #else  #else
2373      othercase = NOTACHAR;      othercase = NOTACHAR;
2374  #endif  #endif
# Line 1907  if (next >= 0) switch(op_code) Line 2381  if (next >= 0) switch(op_code)
2381    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2382    
2383    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2384    if (item == next) return TRUE;    if (item == next) return TRUE;
2385    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2386  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1916  if (next >= 0) switch(op_code) Line 2389  if (next >= 0) switch(op_code)
2389      unsigned int othercase;      unsigned int othercase;
2390      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2391  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2392      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2393  #else  #else
2394      othercase = NOTACHAR;      othercase = NOTACHAR;
2395  #endif  #endif
# Line 1944  if (next >= 0) switch(op_code) Line 2417  if (next >= 0) switch(op_code)
2417    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2418    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2419    
2420      case OP_HSPACE:
2421      case OP_NOT_HSPACE:
2422      switch(next)
2423        {
2424        case 0x09:
2425        case 0x20:
2426        case 0xa0:
2427        case 0x1680:
2428        case 0x180e:
2429        case 0x2000:
2430        case 0x2001:
2431        case 0x2002:
2432        case 0x2003:
2433        case 0x2004:
2434        case 0x2005:
2435        case 0x2006:
2436        case 0x2007:
2437        case 0x2008:
2438        case 0x2009:
2439        case 0x200A:
2440        case 0x202f:
2441        case 0x205f:
2442        case 0x3000:
2443        return op_code != OP_HSPACE;
2444        default:
2445        return op_code == OP_HSPACE;
2446        }
2447    
2448      case OP_VSPACE:
2449      case OP_NOT_VSPACE:
2450      switch(next)
2451        {
2452        case 0x0a:
2453        case 0x0b:
2454        case 0x0c:
2455        case 0x0d:
2456        case 0x85:
2457        case 0x2028:
2458        case 0x2029:
2459        return op_code != OP_VSPACE;
2460        default:
2461        return op_code == OP_VSPACE;
2462        }
2463    
2464    default:    default:
2465    return FALSE;    return FALSE;
2466    }    }
# Line 1978  switch(op_code) Line 2495  switch(op_code)
2495      case ESC_W:      case ESC_W:
2496      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2497    
2498        case ESC_h:
2499        case ESC_H:
2500        switch(item)
2501          {
2502          case 0x09:
2503          case 0x20:
2504          case 0xa0:
2505          case 0x1680:
2506          case 0x180e:
2507          case 0x2000:
2508          case 0x2001:
2509          case 0x2002:
2510          case 0x2003:
2511          case 0x2004:
2512          case 0x2005:
2513          case 0x2006:
2514          case 0x2007:
2515          case 0x2008:
2516          case 0x2009:
2517          case 0x200A:
2518          case 0x202f:
2519          case 0x205f:
2520          case 0x3000:
2521          return -next != ESC_h;
2522          default:
2523          return -next == ESC_h;
2524          }
2525    
2526        case ESC_v:
2527        case ESC_V:
2528        switch(item)
2529          {
2530          case 0x0a:
2531          case 0x0b:
2532          case 0x0c:
2533          case 0x0d:
2534          case 0x85:
2535          case 0x2028:
2536          case 0x2029:
2537          return -next != ESC_v;
2538          default:
2539          return -next == ESC_v;
2540          }
2541    
2542      default:      default:
2543      return FALSE;      return FALSE;
2544      }      }
2545    
2546    case OP_DIGIT:    case OP_DIGIT:
2547    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2548             next == -ESC_h || next == -ESC_v;
2549    
2550    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2551    return next == -ESC_d;    return next == -ESC_d;
# Line 1992  switch(op_code) Line 2554  switch(op_code)
2554    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2555    
2556    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2557    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2558    
2559      case OP_HSPACE:
2560      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2561    
2562      case OP_NOT_HSPACE:
2563      return next == -ESC_h;
2564    
2565      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2566      case OP_VSPACE:
2567      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2568    
2569      case OP_NOT_VSPACE:
2570      return next == -ESC_v;
2571    
2572    case OP_WORDCHAR:    case OP_WORDCHAR:
2573    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2574    
2575    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2576    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2068  uschar classbits[32]; Line 2643  uschar classbits[32];
2643  BOOL class_utf8;  BOOL class_utf8;
2644  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2645  uschar *class_utf8data;  uschar *class_utf8data;
2646    uschar *class_utf8data_base;
2647  uschar utf8_char[6];  uschar utf8_char[6];
2648  #else  #else
2649  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2107  req_caseopt = ((options & PCRE_CASELESS) Line 2683  req_caseopt = ((options & PCRE_CASELESS)
2683  for (;; ptr++)  for (;; ptr++)
2684    {    {
2685    BOOL negate_class;    BOOL negate_class;
2686      BOOL should_flip_negation;
2687    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2688    BOOL is_quantifier;    BOOL is_quantifier;
2689    BOOL is_recurse;    BOOL is_recurse;
2690      BOOL reset_bracount;
2691    int class_charcount;    int class_charcount;
2692    int class_lastchar;    int class_lastchar;
2693    int newoptions;    int newoptions;
# Line 2147  for (;; ptr++) Line 2725  for (;; ptr++)
2725      */      */
2726    
2727      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2728    
2729        /* Paranoid check for integer overflow */
2730    
2731        if (OFLOW_MAX - *lengthptr < code - last_code)
2732          {
2733          *errorcodeptr = ERR20;
2734          goto FAILED;
2735          }
2736    
2737      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2738      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));
2739    
# Line 2184  for (;; ptr++) Line 2771  for (;; ptr++)
2771    
2772    if (inescq && c != 0)    if (inescq && c != 0)
2773      {      {
2774      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2775        {        {
2776        inescq = FALSE;        inescq = FALSE;
2777        ptr++;        ptr++;
# Line 2210  for (;; ptr++) Line 2797  for (;; ptr++)
2797    /* 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
2798    a quantifier. */    a quantifier. */
2799    
2800    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2801      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2802        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2803    
2804    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2805         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2226  for (;; ptr++) Line 2814  for (;; ptr++)
2814    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2815      {      {
2816      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2817      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2818        {        {
2819        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2820          {          {
# Line 2251  for (;; ptr++) Line 2839  for (;; ptr++)
2839      {      {
2840      /* ===================================================================*/      /* ===================================================================*/
2841      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2842      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2843      case ')':      case CHAR_RIGHT_PARENTHESIS:
2844      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2845      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2846      *codeptr = code;      *codeptr = code;
2847      *ptrptr = ptr;      *ptrptr = ptr;
2848      if (lengthptr != NULL)      if (lengthptr != NULL)
2849        {        {
2850          if (OFLOW_MAX - *lengthptr < code - last_code)
2851            {
2852            *errorcodeptr = ERR20;
2853            goto FAILED;
2854            }
2855        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2856        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2857        }        }
# Line 2269  for (;; ptr++) Line 2862  for (;; ptr++)
2862      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2863      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2864    
2865      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2866      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2867        {        {
2868        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2278  for (;; ptr++) Line 2871  for (;; ptr++)
2871      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2872      break;      break;
2873    
2874      case '$':      case CHAR_DOLLAR_SIGN:
2875      previous = NULL;      previous = NULL;
2876      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2877      break;      break;
# Line 2286  for (;; ptr++) Line 2879  for (;; ptr++)
2879      /* 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
2880      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2881    
2882      case '.':      case CHAR_DOT:
2883      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2884      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2885      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2886      previous = code;      previous = code;
2887      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2888      break;      break;
2889    
2890    
# Line 2306  for (;; ptr++) Line 2899  for (;; ptr++)
2899      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,
2900      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2901      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.
     */  
2902    
2903      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2904        default (Perl) mode, it is treated as a data character. */
2905    
2906        case CHAR_RIGHT_SQUARE_BRACKET:
2907        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2908          {
2909          *errorcodeptr = ERR64;
2910          goto FAILED;
2911          }
2912        goto NORMAL_CHAR;
2913    
2914        case CHAR_LEFT_SQUARE_BRACKET:
2915      previous = code;      previous = code;
2916    
2917      /* 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
2918      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. */
2919    
2920      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2921          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2922            check_posix_syntax(ptr, &tempptr))
2923        {        {
2924        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2925        goto FAILED;        goto FAILED;
2926        }        }
2927    
2928      /* 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,
2929        if the first few characters (either before or after ^) are \Q\E or \E we
2930        skip them too. This makes for compatibility with Perl. */
2931    
2932      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2933        for (;;)
2934        {        {
       negate_class = TRUE;  
2935        c = *(++ptr);        c = *(++ptr);
2936          if (c == CHAR_BACKSLASH)
2937            {
2938            if (ptr[1] == CHAR_E)
2939              ptr++;
2940            else if (strncmp((const char *)ptr+1,
2941                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2942              ptr += 3;
2943            else
2944              break;
2945            }
2946          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2947            negate_class = TRUE;
2948          else break;
2949        }        }
2950      else  
2951        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2952        an initial ']' is taken as a data character -- the code below handles
2953        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2954        [^] must match any character, so generate OP_ALLANY. */
2955    
2956        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2957            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2958        {        {
2959        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2960          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2961          zerofirstbyte = firstbyte;
2962          break;
2963        }        }
2964    
2965        /* If a class contains a negative special such as \S, we need to flip the
2966        negation flag at the end, so that support for characters > 255 works
2967        correctly (they are all included in the class). */
2968    
2969        should_flip_negation = FALSE;
2970    
2971      /* 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
2972      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
2973      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2350  for (;; ptr++) Line 2985  for (;; ptr++)
2985  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2986      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2987      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2988        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2989  #endif  #endif
2990    
2991      /* 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 2365  for (;; ptr++) Line 3001  for (;; ptr++)
3001          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3002          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3003          }          }
3004    
3005          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3006          data and reset the pointer. This is so that very large classes that
3007          contain a zillion UTF-8 characters no longer overwrite the work space
3008          (which is on the stack). */
3009    
3010          if (lengthptr != NULL)
3011            {
3012            *lengthptr += class_utf8data - class_utf8data_base;
3013            class_utf8data = class_utf8data_base;
3014            }
3015    
3016  #endif  #endif
3017    
3018        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3019    
3020        if (inescq)        if (inescq)
3021          {          {
3022          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3023            {            {
3024            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3025            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2386  for (;; ptr++) Line 3034  for (;; ptr++)
3034        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3035        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3036    
3037        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3038            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3039            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3040          {          {
3041          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3042          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3043          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3044          uschar pbits[32];          uschar pbits[32];
3045    
3046          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3047            {            {
3048            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3049            goto FAILED;            goto FAILED;
3050            }            }
3051    
3052          ptr += 2;          ptr += 2;
3053          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3054            {            {
3055            local_negate = TRUE;            local_negate = TRUE;
3056              should_flip_negation = TRUE;  /* Note negative special */
3057            ptr++;            ptr++;
3058            }            }
3059    
# Line 2471  for (;; ptr++) Line 3120  for (;; ptr++)
3120        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
3121        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
3122        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3123        to or into the one we are building. We assume they have more than one        to 'or' into the one we are building. We assume they have more than one
3124        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3125    
3126        if (c == '\\')        if (c == CHAR_BACKSLASH)
3127          {          {
3128          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3129          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3130    
3131          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 */
3132          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */
3133          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */          else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3134          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3135            {            {
3136            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3137              {              {
3138              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3139              }              }
3140            else inescq = TRUE;            else inescq = TRUE;
3141            continue;            continue;
3142            }            }
3143            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3144    
3145          if (c < 0)          if (c < 0)
3146            {            {
# Line 2506  for (;; ptr++) Line 3156  for (;; ptr++)
3156              continue;              continue;
3157    
3158              case ESC_D:              case ESC_D:
3159                should_flip_negation = TRUE;
3160              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3161              continue;              continue;
3162    
# Line 2514  for (;; ptr++) Line 3165  for (;; ptr++)
3165              continue;              continue;
3166    
3167              case ESC_W:              case ESC_W:
3168                should_flip_negation = TRUE;
3169              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3170              continue;              continue;
3171    
# Line 2523  for (;; ptr++) Line 3175  for (;; ptr++)
3175              continue;              continue;
3176    
3177              case ESC_S:              case ESC_S:
3178                should_flip_negation = TRUE;
3179              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3180              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3181              continue;              continue;
3182    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
3183              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
3184              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
3185              }              }
# Line 2539  for (;; ptr++) Line 3189  for (;; ptr++)
3189            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3190                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3191    
3192              /* We need to deal with \H, \h, \V, and \v in both phases because
3193              they use extra memory. */
3194    
3195              if (-c == ESC_h)
3196                {
3197                SETBIT(classbits, 0x09); /* VT */
3198                SETBIT(classbits, 0x20); /* SPACE */
3199                SETBIT(classbits, 0xa0); /* NSBP */
3200    #ifdef SUPPORT_UTF8
3201                if (utf8)
3202                  {
3203                  class_utf8 = TRUE;
3204                  *class_utf8data++ = XCL_SINGLE;
3205                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3206                  *class_utf8data++ = XCL_SINGLE;
3207                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3208                  *class_utf8data++ = XCL_RANGE;
3209                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3210                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3211                  *class_utf8data++ = XCL_SINGLE;
3212                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3213                  *class_utf8data++ = XCL_SINGLE;
3214                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3215                  *class_utf8data++ = XCL_SINGLE;
3216                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3217                  }
3218    #endif
3219                continue;
3220                }
3221    
3222              if (-c == ESC_H)
3223                {
3224                for (c = 0; c < 32; c++)
3225                  {
3226                  int x = 0xff;
3227                  switch (c)
3228                    {
3229                    case 0x09/8: x ^= 1 << (0x09%8); break;
3230                    case 0x20/8: x ^= 1 << (0x20%8); break;
3231                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3232                    default: break;
3233                    }
3234                  classbits[c] |= x;
3235                  }
3236    
3237    #ifdef SUPPORT_UTF8
3238                if (utf8)
3239                  {
3240                  class_utf8 = TRUE;
3241                  *class_utf8data++ = XCL_RANGE;
3242                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3243                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3244                  *class_utf8data++ = XCL_RANGE;
3245                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3246                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3247                  *class_utf8data++ = XCL_RANGE;
3248                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3249                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3250                  *class_utf8data++ = XCL_RANGE;
3251                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3252                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3253                  *class_utf8data++ = XCL_RANGE;
3254                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3255                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3256                  *class_utf8data++ = XCL_RANGE;
3257                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3258                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3259                  *class_utf8data++ = XCL_RANGE;
3260                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3261                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3262                  }
3263    #endif
3264                continue;
3265                }
3266    
3267              if (-c == ESC_v)
3268                {
3269                SETBIT(classbits, 0x0a); /* LF */
3270                SETBIT(classbits, 0x0b); /* VT */
3271                SETBIT(classbits, 0x0c); /* FF */
3272                SETBIT(classbits, 0x0d); /* CR */
3273                SETBIT(classbits, 0x85); /* NEL */
3274    #ifdef SUPPORT_UTF8
3275                if (utf8)
3276                  {
3277                  class_utf8 = TRUE;
3278                  *class_utf8data++ = XCL_RANGE;
3279                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3280                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3281                  }
3282    #endif
3283                continue;
3284                }
3285    
3286              if (-c == ESC_V)
3287                {
3288                for (c = 0; c < 32; c++)
3289                  {
3290                  int x = 0xff;
3291                  switch (c)
3292                    {
3293                    case 0x0a/8: x ^= 1 << (0x0a%8);
3294                                 x ^= 1 << (0x0b%8);
3295                                 x ^= 1 << (0x0c%8);
3296                                 x ^= 1 << (0x0d%8);
3297                                 break;
3298                    case 0x85/8: x ^= 1 << (0x85%8); break;
3299                    default: break;
3300                    }
3301                  classbits[c] |= x;
3302                  }
3303    
3304    #ifdef SUPPORT_UTF8
3305                if (utf8)
3306                  {
3307                  class_utf8 = TRUE;
3308                  *class_utf8data++ = XCL_RANGE;
3309                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3310                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3311                  *class_utf8data++ = XCL_RANGE;
3312                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3313                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3314                  }
3315    #endif
3316                continue;
3317                }
3318    
3319            /* We need to deal with \P and \p in both phases. */            /* We need to deal with \P and \p in both phases. */
3320    
3321  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 2582  for (;; ptr++) Line 3359  for (;; ptr++)
3359        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3360    
3361        CHECK_RANGE:        CHECK_RANGE:
3362        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3363          {          {
3364          inescq = FALSE;          inescq = FALSE;
3365          ptr += 2;          ptr += 2;
# Line 2590  for (;; ptr++) Line 3367  for (;; ptr++)
3367    
3368        oldptr = ptr;        oldptr = ptr;
3369    
3370        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3371    
3372          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3373    
3374          /* Check for range */
3375    
3376          if (!inescq && ptr[1] == CHAR_MINUS)
3377          {          {
3378          int d;          int d;
3379          ptr += 2;          ptr += 2;
3380          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3381    
3382          /* 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
3383          mode. */          mode. */
3384    
3385          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3386            {            {
3387            ptr += 2;            ptr += 2;
3388            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3389                { ptr += 2; continue; }
3390            inescq = TRUE;            inescq = TRUE;
3391            break;            break;
3392            }            }
3393    
3394          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3395            {            {
3396            ptr = oldptr;            ptr = oldptr;
3397            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2626  for (;; ptr++) Line 3410  for (;; ptr++)
3410          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
3411          in such circumstances. */          in such circumstances. */
3412    
3413          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3414            {            {
3415            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3416            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3417    
3418            /* \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
3419            special means the '-' was literal */            special means the '-' was literal */
3420    
3421            if (d < 0)            if (d < 0)
3422              {              {
3423              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3424              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3425              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3426                {                {
3427                ptr = oldptr;                ptr = oldptr;
3428                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2657  for (;; ptr++) Line 3441  for (;; ptr++)
3441    
3442          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3443    
3444            /* Remember \r or \n */
3445    
3446            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3447    
3448          /* 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
3449          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3450          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2679  for (;; ptr++) Line 3467  for (;; ptr++)
3467              unsigned int origd = d;              unsigned int origd = d;
3468              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3469                {                {
3470                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3471                      ocd <= (unsigned int)d)
3472                    continue;                          /* Skip embedded ranges */
3473    
3474                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3475                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3476                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3477                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3478                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3479                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3480                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3481                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3482                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3483                  d = ocd;                  d = ocd;
3484                  continue;                  continue;
# Line 2771  for (;; ptr++) Line 3563  for (;; ptr++)
3563          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3564            {            {
3565            unsigned int othercase;            unsigned int othercase;
3566            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3567              {              {
3568              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3569              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2798  for (;; ptr++) Line 3590  for (;; ptr++)
3590    
3591      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3592    
3593      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3594    
3595      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3596        {        {
# Line 2806  for (;; ptr++) Line 3598  for (;; ptr++)
3598        goto FAILED;        goto FAILED;
3599        }        }
3600    
3601    
3602    /* This code has been disabled because it would mean that \s counts as
3603    an explicit \r or \n reference, and that's not really what is wanted. Now
3604    we set the flag only if there is a literal "\r" or "\n" in the class. */
3605    
3606    #if 0
3607        /* Remember whether \r or \n are in this class */
3608    
3609        if (negate_class)
3610          {
3611          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3612          }
3613        else
3614          {
3615          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3616          }
3617    #endif
3618    
3619    
3620      /* 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
3621      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
3622      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
3623      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3624      single-bytes only. This is an historical hangover. Maybe one day we can  
3625      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3626        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3627        operate on single-bytes only. This is an historical hangover. Maybe one day
3628        we can tidy these opcodes to handle multi-byte characters.
3629    
3630      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
3631      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 2821  for (;; ptr++) Line 3635  for (;; ptr++)
3635      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3636    
3637  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3638      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3639            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3640  #else  #else
3641      if (class_charcount == 1)      if (class_charcount == 1)
3642  #endif  #endif
# Line 2867  for (;; ptr++) Line 3679  for (;; ptr++)
3679      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3680    
3681      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3682      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3683      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
3684        the class, so any that were explicitly given as well can be ignored. If
3685        (when there are explicit characters > 255 that must be listed) there are no
3686        characters < 256, we can omit the bitmap in the actual compiled code. */
3687    
3688  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3689      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3690        {        {
3691        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3692        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2897  for (;; ptr++) Line 3712  for (;; ptr++)
3712        }        }
3713  #endif  #endif
3714    
3715      /* 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
3716      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
3717      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
3718      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3719    
3720        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3721      if (negate_class)      if (negate_class)
3722        {        {
       *code++ = OP_NCLASS;  
3723        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3724          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3725        }        }
3726      else      else
3727        {        {
       *code++ = OP_CLASS;  
3728        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3729        }        }
3730      code += 32;      code += 32;
# Line 2921  for (;; ptr++) Line 3735  for (;; ptr++)
3735      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3736      has been tested above. */      has been tested above. */
3737    
3738      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3739      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3740      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3741      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3742      goto REPEAT;      goto REPEAT;
3743    
3744      case '*':      case CHAR_ASTERISK:
3745      repeat_min = 0;      repeat_min = 0;
3746      repeat_max = -1;      repeat_max = -1;
3747      goto REPEAT;      goto REPEAT;
3748    
3749      case '+':      case CHAR_PLUS:
3750      repeat_min = 1;      repeat_min = 1;
3751      repeat_max = -1;      repeat_max = -1;
3752      goto REPEAT;      goto REPEAT;
3753    
3754      case '?':      case CHAR_QUESTION_MARK:
3755      repeat_min = 0;      repeat_min = 0;
3756      repeat_max = 1;      repeat_max = 1;
3757    
# Line 2972  for (;; ptr++) Line 3786  for (;; ptr++)
3786      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
3787      repeat type to the non-default. */      repeat type to the non-default. */
3788    
3789      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3790        {        {
3791        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3792        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3793        ptr++;        ptr++;
3794        }        }
3795      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3796        {        {
3797        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3798        ptr++;        ptr++;
# Line 3093  for (;; ptr++) Line 3907  for (;; ptr++)
3907    
3908        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3909    
3910          /*--------------------------------------------------------------------*/
3911          /* This code is obsolete from release 8.00; the restriction was finally
3912          removed: */
3913    
3914        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3915        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3916    
3917        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3918          /*--------------------------------------------------------------------*/
3919    
3920        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3921    
# Line 3243  for (;; ptr++) Line 4062  for (;; ptr++)
4062          goto END_REPEAT;          goto END_REPEAT;
4063          }          }
4064    
4065          /*--------------------------------------------------------------------*/
4066          /* This code is obsolete from release 8.00; the restriction was finally
4067          removed: */
4068    
4069        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4070        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4071    
4072        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4073          /*--------------------------------------------------------------------*/
4074    
4075        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4076          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3282  for (;; ptr++) Line 4106  for (;; ptr++)
4106          goto FAILED;          goto FAILED;
4107          }          }
4108    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
4109        /* 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
4110        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
4111        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 3312  for (;; ptr++) Line 4128  for (;; ptr++)
4128    
4129        if (repeat_min == 0)        if (repeat_min == 0)
4130          {          {
4131          /* 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
4132          altogether. */          output altogether, like this:
4133    
4134          if (repeat_max == 0)          ** if (repeat_max == 0)
4135            {          **   {
4136            code = previous;          **   code = previous;
4137            goto END_REPEAT;          **   goto END_REPEAT;
4138            }          **   }
4139    
4140            However, that fails when a group is referenced as a subroutine from
4141            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4142            so that it is skipped on execution. As we don't have a list of which
4143            groups are referenced, we cannot do this selectively.
4144    
4145            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4146            and do no more at this point. However, we do need to adjust any
4147            OP_RECURSE calls inside the group that refer to the group itself or any
4148            internal or forward referenced group, because the offset is from the
4149            start of the whole regex. Temporarily terminate the pattern while doing
4150            this. */
4151    
4152          /* If the maximum is 1 or unlimited, we just have to stick in the          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
         BRAZERO and do no more at this point. However, we do need to adjust  
         any OP_RECURSE calls inside the group that refer to the group itself or  
         any internal or forward referenced group, because the offset is from  
         the start of the whole regex. Temporarily terminate the pattern while  
         doing this. */  
   
         if (repeat_max <= 1)  
4153            {            {
4154            *code = OP_END;            *code = OP_END;
4155            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4156            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4157            code++;            code++;
4158              if (repeat_max == 0)
4159                {
4160                *previous++ = OP_SKIPZERO;
4161                goto END_REPEAT;
4162                }
4163            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4164            }            }
4165    
# Line 3378  for (;; ptr++) Line 4204  for (;; ptr++)
4204          if (repeat_min > 1)          if (repeat_min > 1)
4205            {            {
4206            /* 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
4207            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4208              potential integer overflow. */
4209    
4210            if (lengthptr != NULL)            if (lengthptr != NULL)
4211              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4212                int delta = (repeat_min - 1)*length_prevgroup;
4213                if ((double)(repeat_min - 1)*(double)length_prevgroup >
4214                                                                (double)INT_MAX ||
4215                    OFLOW_MAX - *lengthptr < delta)
4216                  {
4217                  *errorcodeptr = ERR20;
4218                  goto FAILED;
4219                  }
4220                *lengthptr += delta;
4221                }
4222    
4223            /* This is compiling for real */            /* This is compiling for real */
4224    
# Line 3419  for (;; ptr++) Line 4256  for (;; ptr++)
4256          /* 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
4257          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
4258          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
4259          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4260            paranoid checks to avoid integer overflow. */
4261    
4262          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4263            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4264              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4265                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4266              if ((double)repeat_max *
4267                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4268                      > (double)INT_MAX ||
4269                  OFLOW_MAX - *lengthptr < delta)
4270                {
4271                *errorcodeptr = ERR20;
4272                goto FAILED;
4273                }
4274              *lengthptr += delta;
4275              }
4276    
4277          /* This is compiling for real */          /* This is compiling for real */
4278    
# Line 3505  for (;; ptr++) Line 4354  for (;; ptr++)
4354          }          }
4355        }        }
4356    
4357        /* If previous is OP_FAIL, it was generated by an empty class [] in
4358        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4359        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4360        error above. We can just ignore the repeat in JS case. */
4361    
4362        else if (*previous == OP_FAIL) goto END_REPEAT;
4363    
4364      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4365    
4366      else      else
# Line 3529  for (;; ptr++) Line 4385  for (;; ptr++)
4385      if (possessive_quantifier)      if (possessive_quantifier)
4386        {        {
4387        int len;        int len;
4388        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4389            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4390            tempcode += _pcre_OP_lengths[*tempcode] +
4391              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4392    
4393          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4394            {
4395          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4396    #ifdef SUPPORT_UTF8
4397            if (utf8 && tempcode[-1] >= 0xc0)
4398              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4399    #endif
4400            }
4401    
4402        len = code - tempcode;        len = code - tempcode;
4403        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4404          {          {
# Line 3575  for (;; ptr++) Line 4442  for (;; ptr++)
4442      /* ===================================================================*/      /* ===================================================================*/
4443      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4444      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4445      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. */  
4446    
4447      case '(':      case CHAR_LEFT_PARENTHESIS:
4448      newoptions = options;      newoptions = options;
4449      skipbytes = 0;      skipbytes = 0;
4450      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4451      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4452        reset_bracount = FALSE;
4453    
4454        /* First deal with various "verbs" that can be introduced by '*'. */
4455    
4456        if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4457          {
4458          int i, namelen;
4459          const char *vn = verbnames;
4460          const uschar *name = ++ptr;
4461          previous = NULL;
4462          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4463          if (*ptr == CHAR_COLON)
4464            {
4465            *errorcodeptr = ERR59;   /* Not supported */
4466            goto FAILED;
4467            }
4468          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4469            {
4470            *errorcodeptr = ERR60;
4471            goto FAILED;
4472            }
4473          namelen = ptr - name;
4474          for (i = 0; i < verbcount; i++)
4475            {
4476            if (namelen == verbs[i].len &&
4477                strncmp((char *)name, vn, namelen) == 0)
4478              {
4479              /* Check for open captures before ACCEPT */
4480    
4481              if (verbs[i].op == OP_ACCEPT)
4482                {
4483                open_capitem *oc;
4484                cd->had_accept = TRUE;
4485                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4486                  {
4487                  *code++ = OP_CLOSE;
4488                  PUT2INC(code, 0, oc->number);
4489                  }
4490                }
4491              *code++ = verbs[i].op;
4492              break;
4493              }
4494            vn += verbs[i].len + 1;
4495            }
4496          if (i < verbcount) continue;
4497          *errorcodeptr = ERR60;
4498          goto FAILED;
4499          }
4500    
4501      if (*(++ptr) == '?')      /* Deal with the extended parentheses; all are introduced by '?', and the
4502        appearance of any of them means that this is not a capturing group. */
4503    
4504        else if (*ptr == CHAR_QUESTION_MARK)
4505        {        {
4506        int i, set, unset, namelen;        int i, set, unset, namelen;
4507        int *optset;        int *optset;
# Line 3594  for (;; ptr++) Line 4510  for (;; ptr++)
4510    
4511        switch (*(++ptr))        switch (*(++ptr))
4512          {          {
4513          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4514          ptr++;          ptr++;
4515          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4516          if (*ptr == 0)          if (*ptr == 0)
4517            {            {
4518            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3606  for (;; ptr++) Line 4522  for (;; ptr++)
4522    
4523    
4524          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4525          case ':':                 /* Non-capturing bracket */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4526            reset_bracount = TRUE;
4527            /* Fall through */
4528    
4529            /* ------------------------------------------------------------ */
4530            case CHAR_COLON:          /* Non-capturing bracket */
4531          bravalue = OP_BRA;          bravalue = OP_BRA;
4532          ptr++;          ptr++;
4533          break;          break;
4534    
4535    
4536          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4537          case '(':          case CHAR_LEFT_PARENTHESIS:
4538          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4539    
4540          /* 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 3633  for (;; ptr++) Line 4554  for (;; ptr++)
4554          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4555          including assertions, are processed. */          including assertions, are processed. */
4556    
4557          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4558                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4559            break;            break;
4560    
4561          /* 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 3645  for (;; ptr++) Line 4567  for (;; ptr++)
4567    
4568          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4569    
4570          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4571            {            {
4572            terminator = -1;            terminator = -1;
4573            ptr += 2;            ptr += 2;
# Line 3655  for (;; ptr++) Line 4577  for (;; ptr++)
4577          /* 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
4578          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4579    
4580          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4581            {            {
4582            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4583            ptr++;            ptr++;
4584            }            }
4585          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4586            {            {
4587            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4588            ptr++;            ptr++;
4589            }            }
4590          else          else
4591            {            {
4592            terminator = 0;            terminator = 0;
4593            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4594            }            }
4595    
4596          /* 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 3688  for (;; ptr++) Line 4610  for (;; ptr++)
4610            {            {
4611            if (recno >= 0)            if (recno >= 0)
4612              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4613                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4614            ptr++;            ptr++;
4615            }            }
4616          namelen = ptr - name;          namelen = ptr - name;
4617    
4618          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4619                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4620            {            {
4621            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4622            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 3715  for (;; ptr++) Line 4638  for (;; ptr++)
4638              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4639              goto FAILED;              goto FAILED;
4640              }              }
4641            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
4642                cd->bracount - recno + 1 : recno +cd->bracount;
4643              if (recno <= 0 || recno > cd->final_bracount)
4644              {              {
4645              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4646              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4647              }              }
           else recno += cd->bracount;  
4648            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4649            break;            break;
4650            }            }
# Line 3749  for (;; ptr++) Line 4669  for (;; ptr++)
4669    
4670          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4671    
4672          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4673                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4674            {            {
4675            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
# Line 3770  for (;; ptr++) Line 4690  for (;; ptr++)
4690          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4691          specific group number. */          specific group number. */
4692    
4693          else if (*name == 'R')          else if (*name == CHAR_R)
4694            {            {
4695            recno = 0;            recno = 0;
4696            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 3780  for (;; ptr++) Line 4700  for (;; ptr++)
4700                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4701                goto FAILED;                goto FAILED;
4702                }                }
4703              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4704              }              }
4705            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4706            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 3790  for (;; ptr++) Line 4710  for (;; ptr++)
4710          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4711          false. */          false. */
4712    
4713          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4714            {            {
4715            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4716            skipbytes = 1;            skipbytes = 1;
4717            }            }
4718    
4719          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4720            in the second pass here, so final_bracount is set. */
4721    
4722          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4723            {            {
4724            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4725            }            }
# Line 3814  for (;; ptr++) Line 4735  for (;; ptr++)
4735    
4736    
4737          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4738          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4739          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4740          ptr++;          ptr++;
4741          break;          break;
4742    
4743    
4744          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4745          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4746          ptr++;          ptr++;
4747            if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4748              {
4749              *code++ = OP_FAIL;
4750              previous = NULL;
4751              continue;
4752              }
4753            bravalue = OP_ASSERT_NOT;
4754          break;          break;
4755    
4756    
4757          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4758          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4759          switch (ptr[1])          switch (ptr[1])
4760            {            {
4761            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4762            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4763            ptr += 2;            ptr += 2;
4764            break;            break;
4765    
4766            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4767            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4768            ptr += 2;            ptr += 2;
4769            break;            break;
# Line 3851  for (;; ptr++) Line 4778  for (;; ptr++)
4778    
4779    
4780          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4781          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4782          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4783          ptr++;          ptr++;
4784          break;          break;
4785    
4786    
4787          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4788          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4789          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4790          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4791          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4792            {            {
4793            int n = 0;            int n = 0;
4794            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4795              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4796            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4797              {              {
4798              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4799              goto FAILED;              goto FAILED;
# Line 3886  for (;; ptr++) Line 4813  for (;; ptr++)
4813    
4814    
4815          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4816          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4817          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4818                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4819            {            {
4820            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4821            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4822            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4823            }            }
4824          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4825            {            {
4826            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4827            goto FAILED;            goto FAILED;
# Line 3903  for (;; ptr++) Line 4831  for (;; ptr++)
4831    
4832          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4833          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4834          case '\'':          case CHAR_APOSTROPHE:
4835            {            {
4836            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4837                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4838            name = ++ptr;            name = ++ptr;
4839    
4840            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 3979  for (;; ptr++) Line 4908  for (;; ptr++)
4908    
4909    
4910          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4911          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4912          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4913          is_recurse = TRUE;          is_recurse = TRUE;
4914          /* Fall through */          /* Fall through */
4915    
4916          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4917          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4918          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4919            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4920            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4921    
4922          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4923          name = ++ptr;          name = ++ptr;
# Line 3998  for (;; ptr++) Line 4929  for (;; ptr++)
4929    
4930          if (lengthptr != NULL)          if (lengthptr != NULL)
4931            {            {
4932              if (namelen == 0)
4933                {
4934                *errorcodeptr = ERR62;
4935                goto FAILED;
4936                }
4937            if (*ptr != terminator)            if (*ptr != terminator)
4938              {              {
4939              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4011  for (;; ptr++) Line 4947  for (;; ptr++)
4947            recno = 0;            recno = 0;
4948            }            }
4949    
4950          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4951            first, and then check that we have reached the end of the name in the
4952            table. That way, if the name that is longer than any in the table,
4953            the comparison will fail without reading beyond the table entry. */
4954    
4955          else          else
4956            {            {
4957            slot = cd->name_table;            slot = cd->name_table;
4958            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4959              {              {
4960              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4961                    slot[2+namelen] == 0)
4962                  break;
4963              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4964              }              }
4965    
# Line 4027  for (;; ptr++) Line 4968  for (;; ptr++)
4968              recno = GET2(slot, 0);              recno = GET2(slot, 0);
4969              }              }
4970            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
4971                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
4972                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
4973              {              {
4974              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4043  for (;; ptr++) Line 4984  for (;; ptr++)
4984    
4985    
4986          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4987          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
4988          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
4989          /* Fall through */          /* Fall through */
4990    
4991    
4992          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4993          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
4994          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
4995          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
4996            {            {
4997            const uschar *called;            const uschar *called;
4998              terminator = CHAR_RIGHT_PARENTHESIS;
4999    
5000              /* Come here from the \g<...> and \g'...' code (Oniguruma
5001              compatibility). However, the syntax has been checked to ensure that
5002              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5003              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5004              ever be taken. */
5005    
5006              HANDLE_NUMERICAL_RECURSION:
5007    
5008            if ((refsign = *ptr) == '+') ptr++;            if ((refsign = *ptr) == CHAR_PLUS)
5009            else if (refsign == '-')              {
5010                ptr++;
5011                if ((digitab[*ptr] & ctype_digit) == 0)
5012                  {
5013                  *errorcodeptr = ERR63;
5014                  goto FAILED;
5015                  }
5016                }
5017              else if (refsign == CHAR_MINUS)
5018              {              {
5019              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5020                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4065  for (;; ptr++) Line 5023  for (;; ptr++)
5023    
5024            recno = 0;            recno = 0;
5025            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5026              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5027    
5028            if (*ptr != ')')            if (*ptr != terminator)
5029              {              {
5030              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5031              goto FAILED;              goto FAILED;
5032              }              }
5033    
5034            if (refsign == '-')            if (refsign == CHAR_MINUS)
5035              {              {
5036              if (recno == 0)              if (recno == 0)
5037                {                {
# Line 4087  for (;; ptr++) Line 5045  for (;; ptr++)
5045                goto FAILED;                goto FAILED;
5046                }                }
5047              }              }
5048            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5049              {              {
5050              if (recno == 0)              if (recno == 0)
5051                {                {
# Line 4120  for (;; ptr++) Line 5078  for (;; ptr++)
5078    
5079              if (called == NULL)              if (called == NULL)
5080                {                {
5081                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5082                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5083                  {                  {
5084                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5085                  goto FAILED;                  goto FAILED;
# Line 4173  for (;; ptr++) Line 5131  for (;; ptr++)
5131          set = unset = 0;          set = unset = 0;
5132          optset = &set;          optset = &set;
5133    
5134          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5135            {            {
5136            switch (*ptr++)            switch (*ptr++)
5137              {              {
5138              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5139    
5140              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5141              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5142              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5143              break;              break;
5144    
5145              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5146              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5147              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5148              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5149              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5150              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5151    
5152              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5153                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4220  for (;; ptr++) Line 5178  for (;; ptr++)
5178          both phases.          both phases.
5179    
5180          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, compile code to change the ims
5181          options if this setting actually changes any of them. We also pass the          options if this setting actually changes any of them, and reset the
5182          new setting back so that it can be put at the start of any following          greedy defaults and the case value for firstbyte and reqbyte. */
         branches, and when this group ends (if we are in a group), a resetting  
         item can be compiled. */  
5183    
5184          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5185            {            {
5186            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5187                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5188              {              {
5189              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5190              }              }
5191           else           else
5192              {              {
# Line 4240  for (;; ptr++) Line 5195  for (;; ptr++)
5195                *code++ = OP_OPT;                *code++ = OP_OPT;
5196                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5197                }                }
   
             /* Change options at this level, and pass them back for use  
             in subsequent branches. Reset the greedy defaults and the case  
             value for firstbyte and reqbyte. */  
   
             *optionsptr = options = newoptions;  
5198              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5199              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5200              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5201              }              }
5202    
5203              /* Change options at this level, and pass them back for use
5204              in subsequent branches. When not at the start of the pattern, this
5205              information is also necessary so that a resetting item can be
5206              compiled at the end of a group (if we are in a group). */
5207    
5208              *optionsptr = options = newoptions;
5209            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5210            continue;              /* It is complete */            continue;              /* It is complete */
5211            }            }
# Line 4304  for (;; ptr++) Line 5259  for (;; ptr++)
5259           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
5260           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5261            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5262             reset_bracount,               /* True if (?| group */
5263           skipbytes,                    /* Skip over bracket number */           skipbytes,                    /* Skip over bracket number */
5264           &subfirstbyte,                /* For possible first char */           &subfirstbyte,                /* For possible first char */
5265           &subreqbyte,                  /* For possible last char */           &subreqbyte,                  /* For possible last char */
# Line 4365  for (;; ptr++) Line 5321  for (;; ptr++)
5321    
5322      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5323    
5324      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5325        {        {
5326        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5327        goto FAILED;        goto FAILED;
5328        }        }
5329    
5330      /* In the pre-compile phase, update the length by the length of the nested      /* In the pre-compile phase, update the length by the length of the group,
5331      group, less the brackets at either end. Then reduce the compiled code to      less the brackets at either end. Then reduce the compiled code to just a
5332      just the brackets so that it doesn't use much memory if it is duplicated by      set of non-capturing brackets so that it doesn't use much memory if it is
5333      a quantifier. */      duplicated by a quantifier.*/
5334    
5335      if (lengthptr != NULL)      if (lengthptr != NULL)
5336        {        {
5337          if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
5338            {
5339            *errorcodeptr = ERR20;
5340            goto FAILED;
5341            }
5342        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5343        code++;        *code++ = OP_BRA;
5344        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5345        *code++ = OP_KET;        *code++ = OP_KET;
5346        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5347          break;    /* No need to waste time with special character handling */
5348        }        }
5349    
5350      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
5351    
5352      else code = tempcode;      code = tempcode;
5353    
5354      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
5355      relevant. */      relevant. */
# Line 4457  for (;; ptr++) Line 5419  for (;; ptr++)
5419      We can test for values between ESC_b and ESC_Z for the latter; this may      We can test for values between ESC_b and ESC_Z for the latter; this may
5420      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5421    
5422      case '\\':      case CHAR_BACKSLASH:
5423      tempptr = ptr;      tempptr = ptr;
5424      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5425      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 4466  for (;; ptr++) Line 5428  for (;; ptr++)
5428        {        {
5429        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5430          {          {
5431          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5432            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5433                else inescq = TRUE;
5434          continue;          continue;
5435          }          }
5436    
# Line 4484  for (;; ptr++) Line 5447  for (;; ptr++)
5447        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5448        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5449    
5450          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5451          is a subroutine call by number (Oniguruma syntax). In fact, the value
5452          -ESC_g is returned only for these cases. So we don't need to check for <
5453          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5454          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5455          that is a synonym for a named back reference). */
5456    
5457          if (-c == ESC_g)
5458            {
5459            const uschar *p;
5460            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5461            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5462              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5463    
5464            /* These two statements stop the compiler for warning about possibly
5465            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5466            fact, because we actually check for a number below, the paths that
5467            would actually be in error are never taken. */
5468    
5469            skipbytes = 0;
5470            reset_bracount = FALSE;
5471    
5472            /* Test for a name */
5473    
5474            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5475              {
5476              BOOL isnumber = TRUE;
5477              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5478                {
5479                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5480                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5481                }
5482              if (*p != terminator)
5483                {
5484                *errorcodeptr = ERR57;
5485                break;
5486                }
5487              if (isnumber)
5488                {
5489                ptr++;
5490                goto HANDLE_NUMERICAL_RECURSION;
5491                }
5492              is_recurse = TRUE;
5493              goto NAMED_REF_OR_RECURSE;
5494              }
5495    
5496            /* Test a signed number in angle brackets or quotes. */
5497    
5498            p = ptr + 2;
5499            while ((digitab[*p] & ctype_digit) != 0) p++;
5500            if (*p != terminator)
5501              {
5502              *errorcodeptr = ERR57;
5503              break;
5504              }
5505            ptr++;
5506            goto HANDLE_NUMERICAL_RECURSION;
5507            }
5508    
5509        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
5510        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5511    
5512        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5513              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5514          {          {
5515          is_recurse = FALSE;          is_recurse = FALSE;
5516          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5517              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5518              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5519          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5520          }          }
5521    
# Line 4591  for (;; ptr++) Line 5616  for (;; ptr++)
5616      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5617      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5618    
5619        /* Remember if \r or \n were seen */
5620    
5621        if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5622          cd->external_flags |= PCRE_HASCRORLF;
5623    
5624      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5625      byte, set it from this character, but revert to none on a zero repeat.      byte, set it from this character, but revert to none on a zero repeat.
5626      Otherwise, leave the firstbyte value alone, and don't change it on a zero      Otherwise, leave the firstbyte value alone, and don't change it on a zero
# Line 4663  Arguments: Line 5693  Arguments:
5693    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
5694    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
5695    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
5696      reset_bracount TRUE to reset the count for each branch
5697    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
5698    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
5699    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
# Line 4676  Returns:         TRUE on success Line 5707  Returns:         TRUE on success
5707    
5708  static BOOL  static BOOL
5709  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,
5710    int *errorcodeptr, BOOL lookbehind, int skipbytes, int *firstbyteptr,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
5711    int *reqbyteptr, branch_chain *bcptr, compile_data *cd, int *lengthptr)    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
5712      int *lengthptr)
5713  {  {
5714  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
5715  uschar *code = *codeptr;  uschar *code = *codeptr;
5716  uschar *last_branch = code;  uschar *last_branch = code;
5717  uschar *start_bracket = code;  uschar *start_bracket = code;
5718  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5719    open_capitem capitem;
5720    int capnumber = 0;
5721  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5722  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5723  int length;  int length;
5724    int orig_bracount;
5725    int max_bracount;
5726  branch_chain bc;  branch_chain bc;
5727    
5728  bc.outer = bcptr;  bc.outer = bcptr;
# Line 4708  the code that abstracts option settings Line 5744  the code that abstracts option settings
5744  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the
5745  pre-compile phase to find out whether anything has yet been compiled or not. */  pre-compile phase to find out whether anything has yet been compiled or not. */
5746    
5747    /* If this is a capturing subpattern, add to the chain of open capturing items
5748    so that we can detect them if (*ACCEPT) is encountered. */
5749    
5750    if (*code == OP_CBRA)
5751      {
5752      capnumber = GET2(code, 1 + LINK_SIZE);
5753      capitem.number = capnumber;
5754      capitem.next = cd->open_caps;
5755      cd->open_caps = &capitem;
5756      }
5757    
5758  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5759    
5760  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 4715  code += 1 + LINK_SIZE + skipbytes; Line 5762  code += 1 + LINK_SIZE + skipbytes;
5762    
5763  /* Loop for each alternative branch */  /* Loop for each alternative branch */
5764    
5765    orig_bracount = max_bracount = cd->bracount;
5766  for (;;)  for (;;)
5767    {    {
5768      /* For a (?| group, reset the capturing bracket count so that each branch
5769      uses the same numbers. */
5770    
5771      if (reset_bracount) cd->bracount = orig_bracount;
5772    
5773    /* Handle a change of ims options at the start of the branch */    /* Handle a change of ims options at the start of the branch */
5774    
5775    if ((options & PCRE_IMS) != oldims)    if ((options & PCRE_IMS) != oldims)
# Line 4746  for (;;) Line 5799  for (;;)
5799      return FALSE;      return FALSE;
5800      }      }
5801    
5802      /* Keep the highest bracket count in case (?| was used and some branch
5803      has fewer than the rest. */
5804    
5805      if (cd->bracount > max_bracount) max_bracount = cd->bracount;
5806    
5807    /* In the real compile phase, there is some post-processing to be done. */    /* In the real compile phase, there is some post-processing to be done. */
5808    
5809    if (lengthptr == NULL)    if (lengthptr == NULL)
# Line 4791  for (;;) Line 5849  for (;;)
5849    
5850      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
5851      put the length into the OP_REVERSE item. Temporarily mark the end of the      put the length into the OP_REVERSE item. Temporarily mark the end of the
5852      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
5853        because there may be forward references that we can't check here. Set a
5854        flag to cause another lookbehind check at the end. Why not do it all at the
5855        end? Because common, erroneous checks are picked up here and the offset of
5856        the problem can be shown. */
5857    
5858      if (lookbehind)      if (lookbehind)
5859        {        {
5860        int fixed_length;        int fixed_length;
5861        *code = OP_END;        *code = OP_END;
5862        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
5863        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
5864        if (fixed_length < 0)        if (fixed_length == -3)
5865            {
5866            cd->check_lookbehind = TRUE;
5867            }
5868          else if (fixed_length < 0)
5869          {          {
5870          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
5871          *ptrptr = ptr;          *ptrptr = ptr;
5872          return FALSE;          return FALSE;
5873          }          }
5874        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
5875        }        }
5876      }      }
5877    
# Line 4818  for (;;) Line 5884  for (;;)
5884    compile a resetting op-code following, except at the very end of the pattern.    compile a resetting op-code following, except at the very end of the pattern.
5885    Return leaving the pointer at the terminating char. */    Return leaving the pointer at the terminating char. */
5886    
5887    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
5888      {      {
5889      if (lengthptr == NULL)      if (lengthptr == NULL)
5890        {        {
# Line 4832  for (;;) Line 5898  for (;;)
5898          }          }
5899        while (branch_length > 0);        while (branch_length > 0);
5900        }        }
5901    
5902        /* If it was a capturing subpattern, remove it from the chain. */
5903    
5904        if (capnumber > 0) cd->open_caps = cd->open_caps->next;
5905    
5906      /* Fill in the ket */      /* Fill in the ket */
5907    
# Line 4841  for (;;) Line 5911  for (;;)
5911    
5912      /* Resetting option if needed */      /* Resetting option if needed */
5913    
5914      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
5915        {        {
5916        *code++ = OP_OPT;        *code++ = OP_OPT;
5917        *code++ = oldims;        *code++ = oldims;
5918        length += 2;        length += 2;
5919        }        }
5920    
5921        /* Retain the highest bracket number, in case resetting was used. */
5922    
5923        cd->bracount = max_bracount;
5924    
5925      /* Set values to pass back */      /* Set values to pass back */
5926    
5927      *codeptr = code;      *codeptr = code;
5928      *ptrptr = ptr;      *ptrptr = ptr;
5929      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
5930      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
5931      if (lengthptr != NULL) *lengthptr += length;      if (lengthptr != NULL)
5932          {
5933          if (OFLOW_MAX - *lengthptr < length)
5934            {
5935            *errorcodeptr = ERR20;
5936            return FALSE;
5937            }
5938          *lengthptr += length;
5939          }
5940      return TRUE;      return TRUE;
5941      }      }
5942    
# Line 4958  do { Line 6040  do {
6040       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;