/[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 171 by ph10, Mon Jun 4 14:28:58 2007 UTC revision 455 by ph10, Sat Sep 26 19:12:32 2009 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2009 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 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 'g':      case CHAR_g:
597      if (ptr[1] == '{')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
598          {
599          c = -ESC_g;
600          break;
601          }
602    
603        /* Handle the Perl-compatible cases */
604    
605        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
606        {        {
607        const uschar *p;        const uschar *p;
608        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
609          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
610        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
611          {          {
612          c = -ESC_k;          c = -ESC_k;
613          break;          break;
614          }          }
615        braced = TRUE;        braced = TRUE;
616        ptr++;        ptr++;
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 == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1134      {      {
1135      while (*(++ptr) != ']')      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 the next character is ']', it is a data character that must be
1157        skipped, except in JavaScript compatibility mode. */
1158    
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      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1200        {
1201        if (dup_parens && *count < hwm_count) *count = hwm_count;
1202        *ptrptr = ptr;
1203        return -1;
1204        }
1205    
1206      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    ptr += 2;  FAIL_EXIT:
1214    if (*ptr == 'P') ptr++;                      /* Allow optional P */  *ptrptr = ptr;
1215    return -1;
1216    }
1217    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1218    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1219    
   count++;  
1220    
1221    if (name == NULL && count == lorn) return count;  /*************************************************
1222    term = *ptr++;  *       Find forward referenced subpattern       *
1223    if (term == '<') term = '>';  *************************************************/
1224    thisname = ptr;  
1225    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1226    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1227        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1228      return count;  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. The function is global in the C sense
1554    so that it can be called from pcre_study() when finding the minimum matching
1555    length.
1556    
1557  Arguments:  Arguments:
1558    code        points to start of expression    code        points to start of expression
1559    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1560    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1561    
1562  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
1563  */  */
1564    
1565  static const uschar *  const uschar *
1566  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1567  {  {
1568  for (;;)  for (;;)
1569    {    {
# Line 1241  for (;;) Line 1575  for (;;)
1575    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. */
1576    
1577    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1578    
1579      /* Handle recursion */
1580    
1581      else if (c == OP_REVERSE)
1582        {
1583        if (number < 0) return (uschar *)code;
1584        code += _pcre_OP_lengths[c];
1585        }
1586    
1587    /* Handle capturing bracket */    /* Handle capturing bracket */
1588    
# Line 1251  for (;;) Line 1593  for (;;)
1593      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1594      }      }
1595    
1596    /* 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
1597    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
1598    arrange to skip the extra bytes. */    two bytes of parameters. */
1599    
1600    else    else
1601      {      {
1602        switch(c)
1603          {
1604          case OP_TYPESTAR:
1605          case OP_TYPEMINSTAR:
1606          case OP_TYPEPLUS:
1607          case OP_TYPEMINPLUS:
1608          case OP_TYPEQUERY:
1609          case OP_TYPEMINQUERY:
1610          case OP_TYPEPOSSTAR:
1611          case OP_TYPEPOSPLUS:
1612          case OP_TYPEPOSQUERY:
1613          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1614          break;
1615    
1616          case OP_TYPEUPTO:
1617          case OP_TYPEMINUPTO:
1618          case OP_TYPEEXACT:
1619          case OP_TYPEPOSUPTO:
1620          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1621          break;
1622          }
1623    
1624        /* Add in the fixed length from the table */
1625    
1626      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1627    
1628      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1629      a multi-byte character. The length in the table is a minimum, so we have to
1630      arrange to skip the extra bytes. */
1631    
1632  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1633      if (utf8) switch(c)      if (utf8) switch(c)
1634        {        {
# Line 1279  for (;;) Line 1650  for (;;)
1650        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1651        break;        break;
1652        }        }
1653    #else
1654        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1655  #endif  #endif
1656      }      }
1657    }    }
# Line 1315  for (;;) Line 1688  for (;;)
1688    
1689    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1690    
1691    /* 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
1692    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
1693    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1694    
1695    else    else
1696      {      {
1697        switch(c)
1698          {
1699          case OP_TYPESTAR:
1700          case OP_TYPEMINSTAR:
1701          case OP_TYPEPLUS:
1702          case OP_TYPEMINPLUS:
1703          case OP_TYPEQUERY:
1704          case OP_TYPEMINQUERY:
1705          case OP_TYPEPOSSTAR:
1706          case OP_TYPEPOSPLUS:
1707          case OP_TYPEPOSQUERY:
1708          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1709          break;
1710    
1711          case OP_TYPEPOSUPTO:
1712          case OP_TYPEUPTO:
1713          case OP_TYPEMINUPTO:
1714          case OP_TYPEEXACT:
1715          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1716          break;
1717          }
1718    
1719        /* Add in the fixed length from the table */
1720    
1721      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1722    
1723        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1724        by a multi-byte character. The length in the table is a minimum, so we have
1725        to arrange to skip the extra bytes. */
1726    
1727  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1728      if (utf8) switch(c)      if (utf8) switch(c)
1729        {        {
# Line 1344  for (;;) Line 1745  for (;;)
1745        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1746        break;        break;
1747        }        }
1748    #else
1749        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1750  #endif  #endif
1751      }      }
1752    }    }
# Line 1359  for (;;) Line 1762  for (;;)
1762  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()
1763  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
1764  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1765  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
1766  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
1767    bracket whose current branch will already have been scanned.
1768    
1769  Arguments:  Arguments:
1770    code        points to start of search    code        points to start of search
# Line 1381  for (code = first_significant_code(code Line 1785  for (code = first_significant_code(code
1785    const uschar *ccode;    const uschar *ccode;
1786    
1787    c = *code;    c = *code;
1788    
1789      /* Skip over forward assertions; the other assertions are skipped by
1790      first_significant_code() with a TRUE final argument. */
1791    
1792      if (c == OP_ASSERT)
1793        {
1794        do code += GET(code, 1); while (*code == OP_ALT);
1795        c = *code;
1796        continue;
1797        }
1798    
1799    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1800    
1801    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1802      {      {
1803        code += _pcre_OP_lengths[c];
1804      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1805      c = *code;      c = *code;
1806      continue;      continue;
1807      }      }
1808    
1809    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1810    
1811    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1812      {      {
1813      BOOL empty_branch;      BOOL empty_branch;
1814      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1815    
1816      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1817        empty branch, so just skip over the conditional, because it could be empty.
1818        Otherwise, scan the individual branches of the group. */
1819    
1820      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;  
1821        code += GET(code, 1);        code += GET(code, 1);
1822        else
1823          {
1824          empty_branch = FALSE;
1825          do
1826            {
1827            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1828              empty_branch = TRUE;
1829            code += GET(code, 1);
1830            }
1831          while (*code == OP_ALT);
1832          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1833        }        }
1834      while (*code == OP_ALT);  
1835      if (!empty_branch) return FALSE;   /* All branches are non-empty */      c = *code;
     c = *code;  
1836      continue;      continue;
1837      }      }
1838    
# Line 1417  for (code = first_significant_code(code Line 1840  for (code = first_significant_code(code
1840    
1841    switch (c)    switch (c)
1842      {      {
1843      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1844        cannot be represented just by a bit map. This includes negated single
1845        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1846        actual length is stored in the compiled code, so we must update "code"
1847        here. */
1848    
1849  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1850      case OP_XCLASS:      case OP_XCLASS:
1851      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1852      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1853  #endif  #endif
1854    
# Line 1465  for (code = first_significant_code(code Line 1892  for (code = first_significant_code(code
1892      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1893      case OP_WORDCHAR:      case OP_WORDCHAR:
1894      case OP_ANY:      case OP_ANY:
1895        case OP_ALLANY:
1896      case OP_ANYBYTE:      case OP_ANYBYTE:
1897      case OP_CHAR:      case OP_CHAR:
1898      case OP_CHARNC:      case OP_CHARNC:
# Line 1483  for (code = first_significant_code(code Line 1911  for (code = first_significant_code(code
1911      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1912      return FALSE;      return FALSE;
1913    
1914        /* These are going to continue, as they may be empty, but we have to
1915        fudge the length for the \p and \P cases. */
1916    
1917        case OP_TYPESTAR:
1918        case OP_TYPEMINSTAR:
1919        case OP_TYPEPOSSTAR:
1920        case OP_TYPEQUERY:
1921        case OP_TYPEMINQUERY:
1922        case OP_TYPEPOSQUERY:
1923        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1924        break;
1925    
1926        /* Same for these */
1927    
1928        case OP_TYPEUPTO:
1929        case OP_TYPEMINUPTO:
1930        case OP_TYPEPOSUPTO:
1931        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1932        break;
1933    
1934      /* End of branch */      /* End of branch */
1935    
1936      case OP_KET:      case OP_KET:
# Line 1501  for (code = first_significant_code(code Line 1949  for (code = first_significant_code(code
1949      case OP_QUERY:      case OP_QUERY:
1950      case OP_MINQUERY:      case OP_MINQUERY:
1951      case OP_POSQUERY:      case OP_POSQUERY:
1952        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1953        break;
1954    
1955      case OP_UPTO:      case OP_UPTO:
1956      case OP_MINUPTO:      case OP_MINUPTO:
1957      case OP_POSUPTO:      case OP_POSUPTO:
1958      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1959      break;      break;
1960  #endif  #endif
1961      }      }
# Line 1552  return TRUE; Line 2003  return TRUE;
2003  *************************************************/  *************************************************/
2004    
2005  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2006  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
2007  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2008  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2009    
2010    Originally, this function only recognized a sequence of letters between the
2011    terminators, but it seems that Perl recognizes any sequence of characters,
2012    though of course unknown POSIX names are subsequently rejected. Perl gives an
2013    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2014    didn't consider this to be a POSIX class. Likewise for [:1234:].
2015    
2016    The problem in trying to be exactly like Perl is in the handling of escapes. We
2017    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2018    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2019    below handles the special case of \], but does not try to do any other escape
2020    processing. This makes it different from Perl for cases such as [:l\ower:]
2021    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2022    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2023    I think.
2024    
2025  Argument:  Arguments:
2026    ptr      pointer to the initial [    ptr      pointer to the initial [
2027    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2028    
2029  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2030  */  */
2031    
2032  static BOOL  static BOOL
2033  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2034  {  {
2035  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2036  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2037  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2038    {    {
2039    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2040    return TRUE;      {
2041        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2042        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2043          {
2044          *endptr = ptr;
2045          return TRUE;
2046          }
2047        }
2048    }    }
2049  return FALSE;  return FALSE;
2050  }  }
# Line 1599  Returns:     a value representing the na Line 2069  Returns:     a value representing the na
2069  static int  static int
2070  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2071  {  {
2072    const char *pn = posix_names;
2073  register int yield = 0;  register int yield = 0;
2074  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2075    {    {
2076    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2077      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2078      pn += posix_name_lengths[yield] + 1;
2079    yield++;    yield++;
2080    }    }
2081  return -1;  return -1;
# Line 1618  return -1; Line 2090  return -1;
2090  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2091  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2092  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
2093  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
2094  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
2095  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
2096  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
2097  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2098    OP_END.
2099    
2100  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2101  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 1645  adjust_recurse(uschar *group, int adjust Line 2118  adjust_recurse(uschar *group, int adjust
2118    uschar *save_hwm)    uschar *save_hwm)
2119  {  {
2120  uschar *ptr = group;  uschar *ptr = group;
2121    
2122  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2123    {    {
2124    int offset;    int offset;
# Line 1756  get_othercase_range(unsigned int *cptr, Line 2230  get_othercase_range(unsigned int *cptr,
2230  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2231    
2232  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2233    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2234    
2235  if (c > d) return FALSE;  if (c > d) return FALSE;
2236    
# Line 1765  next = othercase + 1; Line 2239  next = othercase + 1;
2239    
2240  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2241    {    {
2242    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2243    next++;    next++;
2244    }    }
2245    
# Line 1811  if ((options & PCRE_EXTENDED) != 0) Line 2285  if ((options & PCRE_EXTENDED) != 0)
2285    for (;;)    for (;;)
2286      {      {
2287      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2288      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2289        {        {
2290        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2291          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1823  if ((options & PCRE_EXTENDED) != 0) Line 2297  if ((options & PCRE_EXTENDED) != 0)
2297  /* 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
2298  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2299    
2300  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2301    {    {
2302    int temperrorcode = 0;    int temperrorcode = 0;
2303    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1848  if ((options & PCRE_EXTENDED) != 0) Line 2322  if ((options & PCRE_EXTENDED) != 0)
2322    for (;;)    for (;;)
2323      {      {
2324      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2325      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2326        {        {
2327        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2328          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1859  if ((options & PCRE_EXTENDED) != 0) Line 2333  if ((options & PCRE_EXTENDED) != 0)
2333    
2334  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2335    
2336  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2337    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2338        return FALSE;
2339    
2340  /* 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
2341  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 1875  if (next >= 0) switch(op_code) Line 2350  if (next >= 0) switch(op_code)
2350    case OP_CHAR:    case OP_CHAR:
2351  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2352    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2353    #else
2354      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2355  #endif  #endif
2356    return item != next;    return item != next;
2357    
# Line 1893  if (next >= 0) switch(op_code) Line 2370  if (next >= 0) switch(op_code)
2370      unsigned int othercase;      unsigned int othercase;
2371      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2372  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2373      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2374  #else  #else
2375      othercase = NOTACHAR;      othercase = NOTACHAR;
2376  #endif  #endif
# Line 1906  if (next >= 0) switch(op_code) Line 2383  if (next >= 0) switch(op_code)
2383    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2384    
2385    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2386    if (item == next) return TRUE;    if (item == next) return TRUE;
2387    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2388  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1915  if (next >= 0) switch(op_code) Line 2391  if (next >= 0) switch(op_code)
2391      unsigned int othercase;      unsigned int othercase;
2392      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2393  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2394      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2395  #else  #else
2396      othercase = NOTACHAR;      othercase = NOTACHAR;
2397  #endif  #endif
# Line 1943  if (next >= 0) switch(op_code) Line 2419  if (next >= 0) switch(op_code)
2419    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2420    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2421    
2422      case OP_HSPACE:
2423      case OP_NOT_HSPACE:
2424      switch(next)
2425        {
2426        case 0x09:
2427        case 0x20:
2428        case 0xa0:
2429        case 0x1680:
2430        case 0x180e:
2431        case 0x2000:
2432        case 0x2001:
2433        case 0x2002:
2434        case 0x2003:
2435        case 0x2004:
2436        case 0x2005:
2437        case 0x2006:
2438        case 0x2007:
2439        case 0x2008:
2440        case 0x2009:
2441        case 0x200A:
2442        case 0x202f:
2443        case 0x205f:
2444        case 0x3000:
2445        return op_code != OP_HSPACE;
2446        default:
2447        return op_code == OP_HSPACE;
2448        }
2449    
2450      case OP_VSPACE:
2451      case OP_NOT_VSPACE:
2452      switch(next)
2453        {
2454        case 0x0a:
2455        case 0x0b:
2456        case 0x0c:
2457        case 0x0d:
2458        case 0x85:
2459        case 0x2028:
2460        case 0x2029:
2461        return op_code != OP_VSPACE;
2462        default:
2463        return op_code == OP_VSPACE;
2464        }
2465    
2466    default:    default:
2467    return FALSE;    return FALSE;
2468    }    }
# Line 1977  switch(op_code) Line 2497  switch(op_code)
2497      case ESC_W:      case ESC_W:
2498      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2499    
2500        case ESC_h:
2501        case ESC_H:
2502        switch(item)
2503          {
2504          case 0x09:
2505          case 0x20:
2506          case 0xa0:
2507          case 0x1680:
2508          case 0x180e:
2509          case 0x2000:
2510          case 0x2001:
2511          case 0x2002:
2512          case 0x2003:
2513          case 0x2004:
2514          case 0x2005:
2515          case 0x2006:
2516          case 0x2007:
2517          case 0x2008:
2518          case 0x2009:
2519          case 0x200A:
2520          case 0x202f:
2521          case 0x205f:
2522          case 0x3000:
2523          return -next != ESC_h;
2524          default:
2525          return -next == ESC_h;
2526          }
2527    
2528        case ESC_v:
2529        case ESC_V:
2530        switch(item)
2531          {
2532          case 0x0a:
2533          case 0x0b:
2534          case 0x0c:
2535          case 0x0d:
2536          case 0x85:
2537          case 0x2028:
2538          case 0x2029:
2539          return -next != ESC_v;
2540          default:
2541          return -next == ESC_v;
2542          }
2543    
2544      default:      default:
2545      return FALSE;      return FALSE;
2546      }      }
2547    
2548    case OP_DIGIT:    case OP_DIGIT:
2549    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2550             next == -ESC_h || next == -ESC_v;
2551    
2552    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2553    return next == -ESC_d;    return next == -ESC_d;
# Line 1991  switch(op_code) Line 2556  switch(op_code)
2556    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2557    
2558    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2559    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2560    
2561      case OP_HSPACE:
2562      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2563    
2564      case OP_NOT_HSPACE:
2565      return next == -ESC_h;
2566    
2567      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2568      case OP_VSPACE:
2569      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2570    
2571      case OP_NOT_VSPACE:
2572      return next == -ESC_v;
2573    
2574    case OP_WORDCHAR:    case OP_WORDCHAR:
2575    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2576    
2577    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2578    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2067  uschar classbits[32]; Line 2645  uschar classbits[32];
2645  BOOL class_utf8;  BOOL class_utf8;
2646  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2647  uschar *class_utf8data;  uschar *class_utf8data;
2648    uschar *class_utf8data_base;
2649  uschar utf8_char[6];  uschar utf8_char[6];
2650  #else  #else
2651  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2106  req_caseopt = ((options & PCRE_CASELESS) Line 2685  req_caseopt = ((options & PCRE_CASELESS)
2685  for (;; ptr++)  for (;; ptr++)
2686    {    {
2687    BOOL negate_class;    BOOL negate_class;
2688      BOOL should_flip_negation;
2689    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2690    BOOL is_quantifier;    BOOL is_quantifier;
2691    BOOL is_recurse;    BOOL is_recurse;
2692      BOOL reset_bracount;
2693    int class_charcount;    int class_charcount;
2694    int class_lastchar;    int class_lastchar;
2695    int newoptions;    int newoptions;
2696    int recno;    int recno;
2697    int refsign;    int refsign;
2698    int skipbytes;    int skipbytes;
2699    int subreqbyte;    int subreqbyte;
2700    int subfirstbyte;    int subfirstbyte;
# Line 2146  for (;; ptr++) Line 2727  for (;; ptr++)
2727      */      */
2728    
2729      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2730    
2731        /* Paranoid check for integer overflow */
2732    
2733        if (OFLOW_MAX - *lengthptr < code - last_code)
2734          {
2735          *errorcodeptr = ERR20;
2736          goto FAILED;
2737          }
2738    
2739      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2740      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));
2741    
# Line 2183  for (;; ptr++) Line 2773  for (;; ptr++)
2773    
2774    if (inescq && c != 0)    if (inescq && c != 0)
2775      {      {
2776      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2777        {        {
2778        inescq = FALSE;        inescq = FALSE;
2779        ptr++;        ptr++;
# Line 2209  for (;; ptr++) Line 2799  for (;; ptr++)
2799    /* 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
2800    a quantifier. */    a quantifier. */
2801    
2802    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2803      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2804        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2805    
2806    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2807         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2225  for (;; ptr++) Line 2816  for (;; ptr++)
2816    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2817      {      {
2818      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2819      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2820        {        {
2821        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2822          {          {
# Line 2250  for (;; ptr++) Line 2841  for (;; ptr++)
2841      {      {
2842      /* ===================================================================*/      /* ===================================================================*/
2843      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2844      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2845      case ')':      case CHAR_RIGHT_PARENTHESIS:
2846      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2847      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2848      *codeptr = code;      *codeptr = code;
2849      *ptrptr = ptr;      *ptrptr = ptr;
2850      if (lengthptr != NULL)      if (lengthptr != NULL)
2851        {        {
2852          if (OFLOW_MAX - *lengthptr < code - last_code)
2853            {
2854            *errorcodeptr = ERR20;
2855            goto FAILED;
2856            }
2857        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2858        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2859        }        }
# Line 2268  for (;; ptr++) Line 2864  for (;; ptr++)
2864      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2865      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2866    
2867      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2868      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2869        {        {
2870        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2277  for (;; ptr++) Line 2873  for (;; ptr++)
2873      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2874      break;      break;
2875    
2876      case '$':      case CHAR_DOLLAR_SIGN:
2877      previous = NULL;      previous = NULL;
2878      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2879      break;      break;
# Line 2285  for (;; ptr++) Line 2881  for (;; ptr++)
2881      /* 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
2882      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2883    
2884      case '.':      case CHAR_DOT:
2885      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2886      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2887      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2888      previous = code;      previous = code;
2889      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2890      break;      break;
2891    
2892    
# Line 2305  for (;; ptr++) Line 2901  for (;; ptr++)
2901      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,
2902      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2903      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.
     */  
2904    
2905      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2906        default (Perl) mode, it is treated as a data character. */
2907    
2908        case CHAR_RIGHT_SQUARE_BRACKET:
2909        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2910          {
2911          *errorcodeptr = ERR64;
2912          goto FAILED;
2913          }
2914        goto NORMAL_CHAR;
2915    
2916        case CHAR_LEFT_SQUARE_BRACKET:
2917      previous = code;      previous = code;
2918    
2919      /* 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
2920      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. */
2921    
2922      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2923          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2924            check_posix_syntax(ptr, &tempptr))
2925        {        {
2926        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2927        goto FAILED;        goto FAILED;
2928        }        }
2929    
2930      /* 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,
2931        if the first few characters (either before or after ^) are \Q\E or \E we
2932        skip them too. This makes for compatibility with Perl. */
2933    
2934      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2935        for (;;)
2936        {        {
       negate_class = TRUE;  
2937        c = *(++ptr);        c = *(++ptr);
2938          if (c == CHAR_BACKSLASH)
2939            {
2940            if (ptr[1] == CHAR_E)
2941              ptr++;
2942            else if (strncmp((const char *)ptr+1,
2943                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2944              ptr += 3;
2945            else
2946              break;
2947            }
2948          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2949            negate_class = TRUE;
2950          else break;
2951        }        }
2952      else  
2953        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2954        an initial ']' is taken as a data character -- the code below handles
2955        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2956        [^] must match any character, so generate OP_ALLANY. */
2957    
2958        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2959            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2960        {        {
2961        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2962          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2963          zerofirstbyte = firstbyte;
2964          break;
2965        }        }
2966    
2967        /* If a class contains a negative special such as \S, we need to flip the
2968        negation flag at the end, so that support for characters > 255 works
2969        correctly (they are all included in the class). */
2970    
2971        should_flip_negation = FALSE;
2972    
2973      /* 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
2974      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
2975      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2349  for (;; ptr++) Line 2987  for (;; ptr++)
2987  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2988      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2989      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2990        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2991  #endif  #endif
2992    
2993      /* 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 2364  for (;; ptr++) Line 3003  for (;; ptr++)
3003          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3004          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3005          }          }
3006    
3007          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3008          data and reset the pointer. This is so that very large classes that
3009          contain a zillion UTF-8 characters no longer overwrite the work space
3010          (which is on the stack). */
3011    
3012          if (lengthptr != NULL)
3013            {
3014            *lengthptr += class_utf8data - class_utf8data_base;
3015            class_utf8data = class_utf8data_base;
3016            }
3017    
3018  #endif  #endif
3019    
3020        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3021    
3022        if (inescq)        if (inescq)
3023          {          {
3024          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3025            {            {
3026            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3027            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2385  for (;; ptr++) Line 3036  for (;; ptr++)
3036        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3037        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3038    
3039        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3040            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3041            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3042          {          {
3043          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3044          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3045          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3046          uschar pbits[32];          uschar pbits[32];
3047    
3048          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3049            {            {
3050            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3051            goto FAILED;            goto FAILED;
3052            }            }
3053    
3054          ptr += 2;          ptr += 2;
3055          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3056            {            {
3057            local_negate = TRUE;            local_negate = TRUE;
3058              should_flip_negation = TRUE;  /* Note negative special */
3059            ptr++;            ptr++;
3060            }            }
3061    
# Line 2470  for (;; ptr++) Line 3122  for (;; ptr++)
3122        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
3123        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
3124        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3125        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
3126        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3127    
3128        if (c == '\\')        if (c == CHAR_BACKSLASH)
3129          {          {
3130          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3131          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3132    
3133          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 */
3134          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 */
3135          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 */
3136          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3137            {            {
3138            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3139              {              {
3140              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3141              }              }
3142            else inescq = TRUE;            else inescq = TRUE;
3143            continue;            continue;
3144            }            }
3145            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3146    
3147          if (c < 0)          if (c < 0)
3148            {            {
# Line 2505  for (;; ptr++) Line 3158  for (;; ptr++)
3158              continue;              continue;
3159    
3160              case ESC_D:              case ESC_D:
3161                should_flip_negation = TRUE;
3162              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3163              continue;              continue;
3164    
# Line 2513  for (;; ptr++) Line 3167  for (;; ptr++)
3167              continue;              continue;
3168    
3169              case ESC_W:              case ESC_W:
3170                should_flip_negation = TRUE;
3171              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3172              continue;              continue;
3173    
# Line 2522  for (;; ptr++) Line 3177  for (;; ptr++)
3177              continue;              continue;
3178    
3179              case ESC_S:              case ESC_S:
3180                should_flip_negation = TRUE;
3181              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3182              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3183              continue;              continue;
3184    
3185              case ESC_E: /* Perl ignores an orphan \E */              default:    /* Not recognized; fall through */
3186                break;      /* Need "default" setting to stop compiler warning. */
3187                }
3188    
3189              /* In the pre-compile phase, just do the recognition. */
3190    
3191              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3192                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3193    
3194              /* We need to deal with \H, \h, \V, and \v in both phases because
3195              they use extra memory. */
3196    
3197              if (-c == ESC_h)
3198                {
3199                SETBIT(classbits, 0x09); /* VT */
3200                SETBIT(classbits, 0x20); /* SPACE */
3201                SETBIT(classbits, 0xa0); /* NSBP */
3202    #ifdef SUPPORT_UTF8
3203                if (utf8)
3204                  {
3205                  class_utf8 = TRUE;
3206                  *class_utf8data++ = XCL_SINGLE;
3207                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3208                  *class_utf8data++ = XCL_SINGLE;
3209                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3210                  *class_utf8data++ = XCL_RANGE;
3211                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3212                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3213                  *class_utf8data++ = XCL_SINGLE;
3214                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3215                  *class_utf8data++ = XCL_SINGLE;
3216                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3217                  *class_utf8data++ = XCL_SINGLE;
3218                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3219                  }
3220    #endif
3221                continue;
3222                }
3223    
3224              if (-c == ESC_H)
3225                {
3226                for (c = 0; c < 32; c++)
3227                  {
3228                  int x = 0xff;
3229                  switch (c)
3230                    {
3231                    case 0x09/8: x ^= 1 << (0x09%8); break;
3232                    case 0x20/8: x ^= 1 << (0x20%8); break;
3233                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3234                    default: break;
3235                    }
3236                  classbits[c] |= x;
3237                  }
3238    
3239    #ifdef SUPPORT_UTF8
3240                if (utf8)
3241                  {
3242                  class_utf8 = TRUE;
3243                  *class_utf8data++ = XCL_RANGE;
3244                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3245                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3246                  *class_utf8data++ = XCL_RANGE;
3247                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3248                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3249                  *class_utf8data++ = XCL_RANGE;
3250                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3251                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3252                  *class_utf8data++ = XCL_RANGE;
3253                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3254                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3255                  *class_utf8data++ = XCL_RANGE;
3256                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3257                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3258                  *class_utf8data++ = XCL_RANGE;
3259                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3260                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3261                  *class_utf8data++ = XCL_RANGE;
3262                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3263                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3264                  }
3265    #endif
3266              continue;              continue;
3267                }
3268    
3269              default:    /* Not recognized; fall through */            if (-c == ESC_v)
3270              break;      /* Need "default" setting to stop compiler warning. */              {
3271                SETBIT(classbits, 0x0a); /* LF */
3272                SETBIT(classbits, 0x0b); /* VT */
3273                SETBIT(classbits, 0x0c); /* FF */
3274                SETBIT(classbits, 0x0d); /* CR */
3275                SETBIT(classbits, 0x85); /* NEL */
3276    #ifdef SUPPORT_UTF8
3277                if (utf8)
3278                  {
3279                  class_utf8 = TRUE;
3280                  *class_utf8data++ = XCL_RANGE;
3281                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3282                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3283                  }
3284    #endif
3285                continue;
3286              }              }
3287    
3288            /* In the pre-compile phase, just do the recognition. */            if (-c == ESC_V)
3289                {
3290                for (c = 0; c < 32; c++)
3291                  {
3292                  int x = 0xff;
3293                  switch (c)
3294                    {
3295                    case 0x0a/8: x ^= 1 << (0x0a%8);
3296                                 x ^= 1 << (0x0b%8);
3297                                 x ^= 1 << (0x0c%8);
3298                                 x ^= 1 << (0x0d%8);
3299                                 break;
3300                    case 0x85/8: x ^= 1 << (0x85%8); break;
3301                    default: break;
3302                    }
3303                  classbits[c] |= x;
3304                  }
3305    
3306            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  #ifdef SUPPORT_UTF8
3307                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;              if (utf8)
3308                  {
3309                  class_utf8 = TRUE;
3310                  *class_utf8data++ = XCL_RANGE;
3311                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3312                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3313                  *class_utf8data++ = XCL_RANGE;
3314                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3315                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3316                  }
3317    #endif
3318                continue;
3319                }
3320    
3321            /* We need to deal with \P and \p in both phases. */            /* We need to deal with \P and \p in both phases. */
3322    
# Line 2581  for (;; ptr++) Line 3361  for (;; ptr++)
3361        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3362    
3363        CHECK_RANGE:        CHECK_RANGE:
3364        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3365          {          {
3366          inescq = FALSE;          inescq = FALSE;
3367          ptr += 2;          ptr += 2;
# Line 2589  for (;; ptr++) Line 3369  for (;; ptr++)
3369    
3370        oldptr = ptr;        oldptr = ptr;
3371    
3372        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3373    
3374          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3375    
3376          /* Check for range */
3377    
3378          if (!inescq && ptr[1] == CHAR_MINUS)
3379          {          {
3380          int d;          int d;
3381          ptr += 2;          ptr += 2;
3382          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3383    
3384          /* 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
3385          mode. */          mode. */
3386    
3387          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3388            {            {
3389            ptr += 2;            ptr += 2;
3390            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3391                { ptr += 2; continue; }
3392            inescq = TRUE;            inescq = TRUE;
3393            break;            break;
3394            }            }
3395    
3396          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3397            {            {
3398            ptr = oldptr;            ptr = oldptr;
3399            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2625  for (;; ptr++) Line 3412  for (;; ptr++)
3412          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
3413          in such circumstances. */          in such circumstances. */
3414    
3415          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3416            {            {
3417            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3418            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3419    
3420            /* \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
3421            special means the '-' was literal */            special means the '-' was literal */
3422    
3423            if (d < 0)            if (d < 0)
3424              {              {
3425              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3426              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3427              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3428                {                {
3429                ptr = oldptr;                ptr = oldptr;
3430                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2656  for (;; ptr++) Line 3443  for (;; ptr++)
3443    
3444          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3445    
3446            /* Remember \r or \n */
3447    
3448            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3449    
3450          /* 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
3451          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3452          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2678  for (;; ptr++) Line 3469  for (;; ptr++)
3469              unsigned int origd = d;              unsigned int origd = d;
3470              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3471                {                {
3472                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3473                      ocd <= (unsigned int)d)
3474                    continue;                          /* Skip embedded ranges */
3475    
3476                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3477                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3478                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3479                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3480                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3481                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3482                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3483                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3484                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3485                  d = ocd;                  d = ocd;
3486                  continue;                  continue;
# Line 2770  for (;; ptr++) Line 3565  for (;; ptr++)
3565          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3566            {            {
3567            unsigned int othercase;            unsigned int othercase;
3568            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3569              {              {
3570              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3571              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2797  for (;; ptr++) Line 3592  for (;; ptr++)
3592    
3593      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3594    
3595      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3596    
3597      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3598        {        {
# Line 2805  for (;; ptr++) Line 3600  for (;; ptr++)
3600        goto FAILED;        goto FAILED;
3601        }        }
3602    
3603    
3604    /* This code has been disabled because it would mean that \s counts as
3605    an explicit \r or \n reference, and that's not really what is wanted. Now
3606    we set the flag only if there is a literal "\r" or "\n" in the class. */
3607    
3608    #if 0
3609        /* Remember whether \r or \n are in this class */
3610    
3611        if (negate_class)
3612          {
3613          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3614          }
3615        else
3616          {
3617          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3618          }
3619    #endif
3620    
3621    
3622      /* 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
3623      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
3624      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
3625      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3626      single-bytes only. This is an historical hangover. Maybe one day we can  
3627      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3628        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3629        operate on single-bytes only. This is an historical hangover. Maybe one day
3630        we can tidy these opcodes to handle multi-byte characters.
3631    
3632      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
3633      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 2820  for (;; ptr++) Line 3637  for (;; ptr++)
3637      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3638    
3639  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3640      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3641            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3642  #else  #else
3643      if (class_charcount == 1)      if (class_charcount == 1)
3644  #endif  #endif
# Line 2866  for (;; ptr++) Line 3681  for (;; ptr++)
3681      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3682    
3683      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3684      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3685      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
3686        the class, so any that were explicitly given as well can be ignored. If
3687        (when there are explicit characters > 255 that must be listed) there are no
3688        characters < 256, we can omit the bitmap in the actual compiled code. */
3689    
3690  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3691      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3692        {        {
3693        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3694        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2896  for (;; ptr++) Line 3714  for (;; ptr++)
3714        }        }
3715  #endif  #endif
3716    
3717      /* 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
3718      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
3719      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
3720      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3721    
3722        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3723      if (negate_class)      if (negate_class)
3724        {        {
       *code++ = OP_NCLASS;  
3725        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3726          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3727        }        }
3728      else      else
3729        {        {
       *code++ = OP_CLASS;  
3730        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3731        }        }
3732      code += 32;      code += 32;
# Line 2920  for (;; ptr++) Line 3737  for (;; ptr++)
3737      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3738      has been tested above. */      has been tested above. */
3739    
3740      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3741      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3742      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3743      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3744      goto REPEAT;      goto REPEAT;
3745    
3746      case '*':      case CHAR_ASTERISK:
3747      repeat_min = 0;      repeat_min = 0;
3748      repeat_max = -1;      repeat_max = -1;
3749      goto REPEAT;      goto REPEAT;
3750    
3751      case '+':      case CHAR_PLUS:
3752      repeat_min = 1;      repeat_min = 1;
3753      repeat_max = -1;      repeat_max = -1;
3754      goto REPEAT;      goto REPEAT;
3755    
3756      case '?':      case CHAR_QUESTION_MARK:
3757      repeat_min = 0;      repeat_min = 0;
3758      repeat_max = 1;      repeat_max = 1;
3759    
# Line 2971  for (;; ptr++) Line 3788  for (;; ptr++)
3788      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
3789      repeat type to the non-default. */      repeat type to the non-default. */
3790    
3791      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3792        {        {
3793        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3794        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3795        ptr++;        ptr++;
3796        }        }
3797      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3798        {        {
3799        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3800        ptr++;        ptr++;
# Line 3092  for (;; ptr++) Line 3909  for (;; ptr++)
3909    
3910        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3911    
3912          /*--------------------------------------------------------------------*/
3913          /* This code is obsolete from release 8.00; the restriction was finally
3914          removed: */
3915    
3916        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3917        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3918    
3919        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3920          /*--------------------------------------------------------------------*/
3921    
3922        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3923    
# Line 3242  for (;; ptr++) Line 4064  for (;; ptr++)
4064          goto END_REPEAT;          goto END_REPEAT;
4065          }          }
4066    
4067          /*--------------------------------------------------------------------*/
4068          /* This code is obsolete from release 8.00; the restriction was finally
4069          removed: */
4070    
4071        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4072        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4073    
4074        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4075          /*--------------------------------------------------------------------*/
4076    
4077        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4078          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3281  for (;; ptr++) Line 4108  for (;; ptr++)
4108          goto FAILED;          goto FAILED;
4109          }          }
4110    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
4111        /* 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
4112        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
4113        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 3311  for (;; ptr++) Line 4130  for (;; ptr++)
4130    
4131        if (repeat_min == 0)        if (repeat_min == 0)
4132          {          {
4133          /* 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
4134          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4135    
4136          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4137          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4138          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4139          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4140          the start of the whole regex. Temporarily terminate the pattern while          **   }
4141          doing this. */  
4142            However, that fails when a group is referenced as a subroutine from
4143            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4144            so that it is skipped on execution. As we don't have a list of which
4145            groups are referenced, we cannot do this selectively.
4146    
4147            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4148            and do no more at this point. However, we do need to adjust any
4149            OP_RECURSE calls inside the group that refer to the group itself or any
4150            internal or forward referenced group, because the offset is from the
4151            start of the whole regex. Temporarily terminate the pattern while doing
4152            this. */
4153    
4154          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4155            {            {
4156            *code = OP_END;            *code = OP_END;
4157            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4158            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4159            code++;            code++;
4160              if (repeat_max == 0)
4161                {
4162                *previous++ = OP_SKIPZERO;
4163                goto END_REPEAT;
4164                }
4165            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4166            }            }
4167    
# Line 3377  for (;; ptr++) Line 4206  for (;; ptr++)
4206          if (repeat_min > 1)          if (repeat_min > 1)
4207            {            {
4208            /* 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
4209            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4210              potential integer overflow. */
4211    
4212            if (lengthptr != NULL)            if (lengthptr != NULL)
4213              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4214                int delta = (repeat_min - 1)*length_prevgroup;
4215                if ((double)(repeat_min - 1)*(double)length_prevgroup >
4216                                                                (double)INT_MAX ||
4217                    OFLOW_MAX - *lengthptr < delta)
4218                  {
4219                  *errorcodeptr = ERR20;
4220                  goto FAILED;
4221                  }
4222                *lengthptr += delta;
4223                }
4224    
4225            /* This is compiling for real */            /* This is compiling for real */
4226    
# Line 3418  for (;; ptr++) Line 4258  for (;; ptr++)
4258          /* 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
4259          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
4260          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
4261          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4262            paranoid checks to avoid integer overflow. */
4263    
4264          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4265            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4266              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4267                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4268              if ((double)repeat_max *
4269                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4270                      > (double)INT_MAX ||
4271                  OFLOW_MAX - *lengthptr < delta)
4272                {
4273                *errorcodeptr = ERR20;
4274                goto FAILED;
4275                }
4276              *lengthptr += delta;
4277              }
4278    
4279          /* This is compiling for real */          /* This is compiling for real */
4280    
# Line 3504  for (;; ptr++) Line 4356  for (;; ptr++)
4356          }          }
4357        }        }
4358    
4359        /* If previous is OP_FAIL, it was generated by an empty class [] in
4360        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4361        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4362        error above. We can just ignore the repeat in JS case. */
4363    
4364        else if (*previous == OP_FAIL) goto END_REPEAT;
4365    
4366      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4367    
4368      else      else
# Line 3528  for (;; ptr++) Line 4387  for (;; ptr++)
4387      if (possessive_quantifier)      if (possessive_quantifier)
4388        {        {
4389        int len;        int len;
4390        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4391            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4392            tempcode += _pcre_OP_lengths[*tempcode] +
4393              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4394    
4395          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4396            {
4397          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4398    #ifdef SUPPORT_UTF8
4399            if (utf8 && tempcode[-1] >= 0xc0)
4400              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4401    #endif
4402            }
4403    
4404        len = code - tempcode;        len = code - tempcode;
4405        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4406          {          {
# Line 3574  for (;; ptr++) Line 4444  for (;; ptr++)
4444      /* ===================================================================*/      /* ===================================================================*/
4445      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4446      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4447      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. */  
4448    
4449      case '(':      case CHAR_LEFT_PARENTHESIS:
4450      newoptions = options;      newoptions = options;
4451      skipbytes = 0;      skipbytes = 0;
4452      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4453      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4454        reset_bracount = FALSE;
4455    
4456        /* First deal with various "verbs" that can be introduced by '*'. */
4457    
4458        if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4459          {
4460          int i, namelen;
4461          const char *vn = verbnames;
4462          const uschar *name = ++ptr;
4463          previous = NULL;
4464          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4465          if (*ptr == CHAR_COLON)
4466            {
4467            *errorcodeptr = ERR59;   /* Not supported */
4468            goto FAILED;
4469            }
4470          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4471            {
4472            *errorcodeptr = ERR60;
4473            goto FAILED;
4474            }
4475          namelen = ptr - name;
4476          for (i = 0; i < verbcount; i++)
4477            {
4478            if (namelen == verbs[i].len &&
4479                strncmp((char *)name, vn, namelen) == 0)
4480              {
4481              /* Check for open captures before ACCEPT */
4482    
4483              if (verbs[i].op == OP_ACCEPT)
4484                {
4485                open_capitem *oc;
4486                cd->had_accept = TRUE;
4487                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4488                  {
4489                  *code++ = OP_CLOSE;
4490                  PUT2INC(code, 0, oc->number);
4491                  }
4492                }
4493              *code++ = verbs[i].op;
4494              break;
4495              }
4496            vn += verbs[i].len + 1;
4497            }
4498          if (i < verbcount) continue;
4499          *errorcodeptr = ERR60;
4500          goto FAILED;
4501          }
4502    
4503        /* Deal with the extended parentheses; all are introduced by '?', and the
4504        appearance of any of them means that this is not a capturing group. */
4505    
4506      if (*(++ptr) == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4507        {        {
4508        int i, set, unset, namelen;        int i, set, unset, namelen;
4509        int *optset;        int *optset;
# Line 3593  for (;; ptr++) Line 4512  for (;; ptr++)
4512    
4513        switch (*(++ptr))        switch (*(++ptr))
4514          {          {
4515          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4516          ptr++;          ptr++;
4517          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4518          if (*ptr == 0)          if (*ptr == 0)
4519            {            {
4520            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3605  for (;; ptr++) Line 4524  for (;; ptr++)
4524    
4525    
4526          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4527          case ':':                 /* Non-capturing bracket */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4528            reset_bracount = TRUE;
4529            /* Fall through */
4530    
4531            /* ------------------------------------------------------------ */
4532            case CHAR_COLON:          /* Non-capturing bracket */
4533          bravalue = OP_BRA;          bravalue = OP_BRA;
4534          ptr++;          ptr++;
4535          break;          break;
4536    
4537    
4538          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4539          case '(':          case CHAR_LEFT_PARENTHESIS:
4540          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4541    
4542          /* 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 3632  for (;; ptr++) Line 4556  for (;; ptr++)
4556          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4557          including assertions, are processed. */          including assertions, are processed. */
4558    
4559          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4560                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4561            break;            break;
4562    
4563          /* 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 3640  for (;; ptr++) Line 4565  for (;; ptr++)
4565    
4566          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
4567          skipbytes = 3;          skipbytes = 3;
4568          refsign = -1;          refsign = -1;
4569    
4570          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4571    
4572          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4573            {            {
4574            terminator = -1;            terminator = -1;
4575            ptr += 2;            ptr += 2;
# Line 3654  for (;; ptr++) Line 4579  for (;; ptr++)
4579          /* 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
4580          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4581    
4582          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4583            {            {
4584            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4585            ptr++;            ptr++;
4586            }            }
4587          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4588            {            {
4589            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4590            ptr++;            ptr++;
4591            }            }
4592          else          else
4593            {            {
4594            terminator = 0;            terminator = 0;
4595            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4596            }            }
4597    
4598          /* 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 */
4599    
# Line 3687  for (;; ptr++) Line 4612  for (;; ptr++)
4612            {            {
4613            if (recno >= 0)            if (recno >= 0)
4614              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4615                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4616            ptr++;            ptr++;
4617            }            }
4618          namelen = ptr - name;          namelen = ptr - name;
4619    
4620          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4621                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4622            {            {
4623            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4624            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 3706  for (;; ptr++) Line 4632  for (;; ptr++)
4632          /* In the real compile we do the work of looking for the actual          /* In the real compile we do the work of looking for the actual
4633          reference. If the string started with "+" or "-" we require the rest to          reference. If the string started with "+" or "-" we require the rest to
4634          be digits, in which case recno will be set. */          be digits, in which case recno will be set. */
4635    
4636          if (refsign > 0)          if (refsign > 0)
4637            {            {
4638            if (recno <= 0)            if (recno <= 0)
4639              {              {
4640              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4641              goto FAILED;              goto FAILED;
4642              }              }
4643            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
4644                cd->bracount - recno + 1 : recno +cd->bracount;
4645              if (recno <= 0 || recno > cd->final_bracount)
4646              {              {
4647              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4648              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4649              }              }
           else recno += cd->bracount;  
4650            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4651            break;            break;
4652            }            }
4653    
4654          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4655          name. */          name. */
4656    
4657          slot = cd->name_table;          slot = cd->name_table;
4658          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
4659            {            {
# Line 3748  for (;; ptr++) Line 4671  for (;; ptr++)
4671    
4672          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4673    
4674          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4675                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4676            {            {
4677            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
# Line 3769  for (;; ptr++) Line 4692  for (;; ptr++)
4692          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4693          specific group number. */          specific group number. */
4694    
4695          else if (*name == 'R')          else if (*name == CHAR_R)
4696            {            {
4697            recno = 0;            recno = 0;
4698            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 3779  for (;; ptr++) Line 4702  for (;; ptr++)
4702                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4703                goto FAILED;                goto FAILED;
4704                }                }
4705              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4706              }              }
4707            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4708            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 3789  for (;; ptr++) Line 4712  for (;; ptr++)
4712          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4713          false. */          false. */
4714    
4715          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4716            {            {
4717            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4718            skipbytes = 1;            skipbytes = 1;
4719            }            }
4720    
4721          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4722            in the second pass here, so final_bracount is set. */
4723    
4724          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4725            {            {
4726            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4727            }            }
# Line 3813  for (;; ptr++) Line 4737  for (;; ptr++)
4737    
4738    
4739          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4740          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4741          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4742          ptr++;          ptr++;
4743          break;          break;
4744    
4745    
4746          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4747          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4748          ptr++;          ptr++;
4749            if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4750              {
4751              *code++ = OP_FAIL;
4752              previous = NULL;
4753              continue;
4754              }
4755            bravalue = OP_ASSERT_NOT;
4756          break;          break;
4757    
4758    
4759          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4760          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4761          switch (ptr[1])          switch (ptr[1])
4762            {            {
4763            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4764            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4765            ptr += 2;            ptr += 2;
4766            break;            break;
4767    
4768            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4769            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4770            ptr += 2;            ptr += 2;
4771            break;            break;
# Line 3850  for (;; ptr++) Line 4780  for (;; ptr++)
4780    
4781    
4782          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4783          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4784          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4785          ptr++;          ptr++;
4786          break;          break;
4787    
4788    
4789          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4790          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4791          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4792          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4793          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4794            {            {
4795            int n = 0;            int n = 0;
4796            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4797              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4798            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4799              {              {
4800              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4801              goto FAILED;              goto FAILED;
# Line 3885  for (;; ptr++) Line 4815  for (;; ptr++)
4815    
4816    
4817          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4818          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4819          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4820                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4821            {            {
4822            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4823            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4824            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4825            }            }
4826          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4827            {            {
4828            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4829            goto FAILED;            goto FAILED;
# Line 3902  for (;; ptr++) Line 4833  for (;; ptr++)
4833    
4834          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4835          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4836          case '\'':          case CHAR_APOSTROPHE:
4837            {            {
4838            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4839                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4840            name = ++ptr;            name = ++ptr;
4841    
4842            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 3978  for (;; ptr++) Line 4910  for (;; ptr++)
4910    
4911    
4912          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4913          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4914          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4915          is_recurse = TRUE;          is_recurse = TRUE;
4916          /* Fall through */          /* Fall through */
4917    
4918          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4919          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4920          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4921            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4922            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4923    
4924          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4925          name = ++ptr;          name = ++ptr;
# Line 3997  for (;; ptr++) Line 4931  for (;; ptr++)
4931    
4932          if (lengthptr != NULL)          if (lengthptr != NULL)
4933            {            {
4934              if (namelen == 0)
4935                {
4936                *errorcodeptr = ERR62;
4937                goto FAILED;
4938                }
4939            if (*ptr != terminator)            if (*ptr != terminator)
4940              {              {
4941              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4010  for (;; ptr++) Line 4949  for (;; ptr++)
4949            recno = 0;            recno = 0;
4950            }            }
4951    
4952          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4953            first, and then check that we have reached the end of the name in the
4954            table. That way, if the name that is longer than any in the table,
4955            the comparison will fail without reading beyond the table entry. */
4956    
4957          else          else
4958            {            {
4959            slot = cd->name_table;            slot = cd->name_table;
4960            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4961              {              {
4962              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4963                    slot[2+namelen] == 0)
4964                  break;
4965              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4966              }              }
4967    
# Line 4026  for (;; ptr++) Line 4970  for (;; ptr++)
4970              recno = GET2(slot, 0);              recno = GET2(slot, 0);
4971              }              }
4972            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
4973                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
4974                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
4975              {              {
4976              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4042  for (;; ptr++) Line 4986  for (;; ptr++)
4986    
4987    
4988          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4989          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
4990          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
4991          /* Fall through */          /* Fall through */
4992    
4993    
4994          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4995          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
4996          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:
4997          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:
4998            {            {
4999            const uschar *called;            const uschar *called;
5000              terminator = CHAR_RIGHT_PARENTHESIS;
5001    
5002              /* Come here from the \g<...> and \g'...' code (Oniguruma
5003              compatibility). However, the syntax has been checked to ensure that
5004              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5005              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5006              ever be taken. */
5007    
5008              HANDLE_NUMERICAL_RECURSION:
5009    
5010            if ((refsign = *ptr) == '+') ptr++;            if ((refsign = *ptr) == CHAR_PLUS)
5011            else if (refsign == '-')              {
5012                ptr++;
5013                if ((digitab[*ptr] & ctype_digit) == 0)
5014                  {
5015                  *errorcodeptr = ERR63;
5016                  goto FAILED;
5017                  }
5018                }
5019              else if (refsign == CHAR_MINUS)
5020              {              {
5021              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5022                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
5023              ptr++;              ptr++;
5024              }              }
5025    
5026            recno = 0;            recno = 0;
5027            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5028              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5029    
5030            if (*ptr != ')')            if (*ptr != terminator)
5031              {              {
5032              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5033              goto FAILED;              goto FAILED;
5034              }              }
5035    
5036            if (refsign == '-')            if (refsign == CHAR_MINUS)
5037              {              {
5038              if (recno == 0)              if (recno == 0)
5039                {                {
5040                *errorcodeptr = ERR58;                *errorcodeptr = ERR58;
5041                goto FAILED;                goto FAILED;
5042                }                }
5043              recno = cd->bracount - recno + 1;              recno = cd->bracount - recno + 1;
5044              if (recno <= 0)              if (recno <= 0)
5045                {                {
5046                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
5047                goto FAILED;                goto FAILED;
5048                }                }
5049              }              }
5050            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5051              {              {
5052              if (recno == 0)              if (recno == 0)
5053                {                {
5054                *errorcodeptr = ERR58;                *errorcodeptr = ERR58;
5055                goto FAILED;                goto FAILED;
5056                }                }
5057              recno += cd->bracount;              recno += cd->bracount;
5058              }              }
5059    
5060            /* Come here from code above that handles a named recursion */            /* Come here from code above that handles a named recursion */
5061    
# Line 4113  for (;; ptr++) Line 5074  for (;; ptr++)
5074            if (lengthptr == NULL)            if (lengthptr == NULL)
5075              {              {
5076              *code = OP_END;              *code = OP_END;
5077              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5078                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5079    
5080              /* Forward reference */              /* Forward reference */
5081    
5082              if (called == NULL)              if (called == NULL)
5083                {                {
5084                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5085                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5086                  {                  {
5087                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5088                  goto FAILED;                  goto FAILED;
# Line 4168  for (;; ptr++) Line 5130  for (;; ptr++)
5130    
5131          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5132          default:              /* Other characters: check option setting */          default:              /* Other characters: check option setting */
5133          OTHER_CHAR_AFTER_QUERY:          OTHER_CHAR_AFTER_QUERY:
5134          set = unset = 0;          set = unset = 0;
5135          optset = &set;          optset = &set;
5136    
5137          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5138            {            {
5139            switch (*ptr++)            switch (*ptr++)
5140              {              {
5141              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5142    
5143              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5144              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5145              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5146              break;              break;
5147    
5148              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5149              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5150              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5151              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5152              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5153              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5154    
5155              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5156                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4219  for (;; ptr++) Line 5181  for (;; ptr++)
5181          both phases.          both phases.
5182    
5183          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
5184          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
5185          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. */  
5186    
5187          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5188            {            {
5189            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5190                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5191              {              {
5192              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5193              }              }
5194           else           else
5195              {              {
# Line 4239  for (;; ptr++) Line 5198  for (;; ptr++)
5198                *code++ = OP_OPT;                *code++ = OP_OPT;
5199                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5200                }                }
   
             /* 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;  
5201              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5202              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5203              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5204              }              }
5205    
5206              /* Change options at this level, and pass them back for use
5207              in subsequent branches. When not at the start of the pattern, this
5208              information is also necessary so that a resetting item can be
5209              compiled at the end of a group (if we are in a group). */
5210    
5211              *optionsptr = options = newoptions;
5212            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5213            continue;              /* It is complete */            continue;              /* It is complete */
5214            }            }
# Line 4303  for (;; ptr++) Line 5262  for (;; ptr++)
5262           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
5263           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5264            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5265             reset_bracount,               /* True if (?| group */
5266           skipbytes,                    /* Skip over bracket number */           skipbytes,                    /* Skip over bracket number */
5267           &subfirstbyte,                /* For possible first char */           &subfirstbyte,                /* For possible first char */
5268           &subreqbyte,                  /* For possible last char */           &subreqbyte,                  /* For possible last char */
# Line 4319  for (;; ptr++) Line 5279  for (;; ptr++)
5279      is on the bracket. */      is on the bracket. */
5280    
5281      /* If this is a conditional bracket, check that there are no more than      /* If this is a conditional bracket, check that there are no more than
5282      two branches in the group, or just one if it's a DEFINE group. */      two branches in the group, or just one if it's a DEFINE group. We do this
5283        in the real compile phase, not in the pre-pass, where the whole group may
5284        not be available. */
5285    
5286      if (bravalue == OP_COND)      if (bravalue == OP_COND && lengthptr == NULL)
5287        {        {
5288        uschar *tc = code;        uschar *tc = code;
5289        int condcount = 0;        int condcount = 0;
# Line 4362  for (;; ptr++) Line 5324  for (;; ptr++)
5324    
5325      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5326    
5327      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5328        {        {
5329        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5330        goto FAILED;        goto FAILED;
5331        }        }
5332    
5333      /* 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,
5334      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
5335      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
5336      a quantifier. */      duplicated by a quantifier.*/
5337    
5338      if (lengthptr != NULL)      if (lengthptr != NULL)
5339        {        {
5340          if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
5341            {
5342            *errorcodeptr = ERR20;
5343            goto FAILED;
5344            }
5345        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5346        code++;        *code++ = OP_BRA;
5347        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5348        *code++ = OP_KET;        *code++ = OP_KET;
5349        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5350          break;    /* No need to waste time with special character handling */
5351        }        }
5352    
5353      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
5354    
5355      else code = tempcode;      code = tempcode;
5356    
5357      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
5358      relevant. */      relevant. */
# Line 4454  for (;; ptr++) Line 5422  for (;; ptr++)
5422      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
5423      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5424    
5425      case '\\':      case CHAR_BACKSLASH:
5426      tempptr = ptr;      tempptr = ptr;
5427      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5428      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 4463  for (;; ptr++) Line 5431  for (;; ptr++)
5431        {        {
5432        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5433          {          {
5434          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5435            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5436                else inescq = TRUE;
5437          continue;          continue;
5438          }          }
5439    
# Line 4481  for (;; ptr++) Line 5450  for (;; ptr++)
5450        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5451        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5452    
5453          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5454          is a subroutine call by number (Oniguruma syntax). In fact, the value
5455          -ESC_g is returned only for these cases. So we don't need to check for <
5456          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5457          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5458          that is a synonym for a named back reference). */
5459    
5460          if (-c == ESC_g)
5461            {
5462            const uschar *p;
5463            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5464            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5465              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5466    
5467            /* These two statements stop the compiler for warning about possibly
5468            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5469            fact, because we actually check for a number below, the paths that
5470            would actually be in error are never taken. */
5471    
5472            skipbytes = 0;
5473            reset_bracount = FALSE;
5474    
5475            /* Test for a name */
5476    
5477            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5478              {
5479              BOOL isnumber = TRUE;
5480              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5481                {
5482                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5483                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5484                }
5485              if (*p != terminator)
5486                {
5487                *errorcodeptr = ERR57;
5488                break;
5489                }
5490              if (isnumber)
5491                {
5492                ptr++;
5493                goto HANDLE_NUMERICAL_RECURSION;
5494                }
5495              is_recurse = TRUE;
5496              goto NAMED_REF_OR_RECURSE;
5497              }
5498    
5499            /* Test a signed number in angle brackets or quotes. */
5500    
5501            p = ptr + 2;
5502            while ((digitab[*p] & ctype_digit) != 0) p++;
5503            if (*p != terminator)
5504              {
5505              *errorcodeptr = ERR57;
5506              break;
5507              }
5508            ptr++;
5509            goto HANDLE_NUMERICAL_RECURSION;
5510            }
5511    
5512        /* \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).
5513        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5514    
5515        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5516              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5517          {          {
5518          is_recurse = FALSE;          is_recurse = FALSE;
5519          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5520              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5521              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5522          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5523          }          }
5524    
# Line 4588  for (;; ptr++) Line 5619  for (;; ptr++)
5619      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5620      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5621    
5622        /* Remember if \r or \n were seen */
5623    
5624        if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5625          cd->external_flags |= PCRE_HASCRORLF;
5626    
5627      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5628      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.
5629      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 4653  This function is used during the pre-com Line 5689  This function is used during the pre-com
5689  out the amount of memory needed, as well as during the real compile phase. The  out the amount of memory needed, as well as during the real compile phase. The
5690  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
5691    
5692  Argument:  Arguments:
5693    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
5694    oldims         previous settings of ims option bits    oldims         previous settings of ims option bits
5695    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
5696    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
5697    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
5698    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
5699      reset_bracount TRUE to reset the count for each branch
5700    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
5701    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
5702    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 4673  Returns:         TRUE on success Line 5710  Returns:         TRUE on success
5710    
5711  static BOOL  static BOOL
5712  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,
5713    int *errorcodeptr, BOOL lookbehind, int skipbytes, int *firstbyteptr,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
5714    int *reqbyteptr, branch_chain *bcptr, compile_data *cd, int *lengthptr)    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
5715      int *lengthptr)
5716  {  {
5717  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
5718  uschar *code = *codeptr;  uschar *code = *codeptr;
5719  uschar *last_branch = code;  uschar *last_branch = code;
5720  uschar *start_bracket = code;  uschar *start_bracket = code;
5721  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5722    open_capitem capitem;
5723    int capnumber = 0;
5724  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5725  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5726  int length;  int length;
5727    int orig_bracount;
5728    int max_bracount;
5729  branch_chain bc;  branch_chain bc;
5730    
5731  bc.outer = bcptr;  bc.outer = bcptr;
# Line 4705  the code that abstracts option settings Line 5747  the code that abstracts option settings
5747  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
5748  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. */
5749    
5750    /* If this is a capturing subpattern, add to the chain of open capturing items
5751    so that we can detect them if (*ACCEPT) is encountered. */
5752    
5753    if (*code == OP_CBRA)
5754      {
5755      capnumber = GET2(code, 1 + LINK_SIZE);
5756      capitem.number = capnumber;
5757      capitem.next = cd->open_caps;
5758      cd->open_caps = &capitem;
5759      }
5760    
5761  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5762    
5763  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 4712  code += 1 + LINK_SIZE + skipbytes; Line 5765  code += 1 + LINK_SIZE + skipbytes;
5765    
5766  /* Loop for each alternative branch */  /* Loop for each alternative branch */
5767    
5768    orig_bracount = max_bracount = cd->bracount;
5769  for (;;)  for (;;)
5770    {    {
5771      /* For a (?| group, reset the capturing bracket count so that each branch
5772      uses the same numbers. */
5773    
5774      if (reset_bracount) cd->bracount = orig_bracount;
5775  </