/[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 97 by ph10, Mon Mar 5 12:36:47 2007 UTC revision 340 by ph10, Fri Apr 18 20:00:21 2008 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) 1997-2006 University of Cambridge             Copyright (c) 1997-2008 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 87  static const short int escapes[] = { Line 102  static const short int escapes[] = {
102       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */
103       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */
104     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */
105       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */
106  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */
107  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */
108     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */
109       0,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */
110  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */
111       0,      0, -ESC_z                                            /* x - z */       0,      0, -ESC_z                                            /* x - z */
112  };  };
113    
# Line 106  static const short int escapes[] = { Line 121  static const short int escapes[] = {
121  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
122  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
123  /*  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,
124  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
125  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
126  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
127  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
128  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
129  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
130  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
131  /*  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,
132  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
133  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
134  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
135  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
136  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
137  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
138  /*  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 140  static const short int escapes[] = {
140  #endif  #endif
141    
142    
143  /* 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
144  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
145  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. */
146    
147  static const char *const posix_names[] = {  typedef struct verbitem {
148    "alpha", "lower", "upper",    int   len;
149    "alnum", "ascii", "blank", "cntrl", "digit", "graph",    int   op;
150    "print", "punct", "space", "word",  "xdigit" };  } verbitem;
151    
152    static const char verbnames[] =
153      "ACCEPT\0"
154      "COMMIT\0"
155      "F\0"
156      "FAIL\0"
157      "PRUNE\0"
158      "SKIP\0"
159      "THEN";
160    
161    static const verbitem verbs[] = {
162      { 6, OP_ACCEPT },
163      { 6, OP_COMMIT },
164      { 1, OP_FAIL },
165      { 4, OP_FAIL },
166      { 5, OP_PRUNE },
167      { 4, OP_SKIP  },
168      { 4, OP_THEN  }
169    };
170    
171    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
172    
173    
174    /* Tables of names of POSIX character classes and their lengths. The names are
175    now all in a single string, to reduce the number of relocations when a shared
176    library is dynamically loaded. The list of lengths is terminated by a zero
177    length entry. The first three must be alpha, lower, upper, as this is assumed
178    for handling case independence. */
179    
180    static const char posix_names[] =
181      "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"
182      "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"
183      "word\0"   "xdigit";
184    
185  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
186    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 219  static const int posix_class_maps[] = {
219  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
220  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
221  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
222  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
223    the number of relocations needed when a shared library is loaded dynamically,
224  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
225    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
226    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
227    "\\c at end of pattern",  because these strings are used only when there is a compilation error. */
228    "unrecognized character follows \\",  
229    "numbers out of order in {} quantifier",  static const char error_texts[] =
230      "no error\0"
231      "\\ at end of pattern\0"
232      "\\c at end of pattern\0"
233      "unrecognized character follows \\\0"
234      "numbers out of order in {} quantifier\0"
235    /* 5 */    /* 5 */
236    "number too big in {} quantifier",    "number too big in {} quantifier\0"
237    "missing terminating ] for character class",    "missing terminating ] for character class\0"
238    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
239    "range out of order in character class",    "range out of order in character class\0"
240    "nothing to repeat",    "nothing to repeat\0"
241    /* 10 */    /* 10 */
242    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
243    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
244    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
245    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
246    "missing )",    "missing )\0"
247    /* 15 */    /* 15 */
248    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
249    "erroffset passed as NULL",    "erroffset passed as NULL\0"
250    "unknown option bit(s) set",    "unknown option bit(s) set\0"
251    "missing ) after comment",    "missing ) after comment\0"
252    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
253    /* 20 */    /* 20 */
254    "regular expression too large",    "regular expression is too large\0"
255    "failed to get memory",    "failed to get memory\0"
256    "unmatched parentheses",    "unmatched parentheses\0"
257    "internal error: code overflow",    "internal error: code overflow\0"
258    "unrecognized character after (?<",    "unrecognized character after (?<\0"
259    /* 25 */    /* 25 */
260    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
261    "malformed number or name after (?(",    "malformed number or name after (?(\0"
262    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
263    "assertion expected after (?(",    "assertion expected after (?(\0"
264    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
265    /* 30 */    /* 30 */
266    "unknown POSIX class name",    "unknown POSIX class name\0"
267    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
268    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
269    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
270    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
271    /* 35 */    /* 35 */
272    "invalid condition (?(0)",    "invalid condition (?(0)\0"
273    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
274    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
275    "number after (?C is > 255",    "number after (?C is > 255\0"
276    "closing ) for (?C expected",    "closing ) for (?C expected\0"
277    /* 40 */    /* 40 */
278    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
279    "unrecognized character after (?P",    "unrecognized character after (?P\0"
280    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
281    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
282    "invalid UTF-8 string",    "invalid UTF-8 string\0"
283    /* 45 */    /* 45 */
284    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
285    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
286    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
287    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
288    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
289    /* 50 */    /* 50 */
290    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
291    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
292    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
293    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
294    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
295    /* 55 */    /* 55 */
296    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
297    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
298    "\\g is not followed by 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"
299  };    "a numbered reference must not be zero\0"
300      "(*VERB) with an argument is not supported\0"
301      /* 60 */
302      "(*VERB) not recognized\0"
303      "number is too big\0"
304      "subpattern name expected\0"
305      "digit expected after (?+\0"
306      "] is an invalid data character in JavaScript compatibility mode";
307    
308    
309  /* 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 373  static const unsigned char ebcdic_charta Line 433  static const unsigned char ebcdic_charta
433  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
434    
435  static BOOL  static BOOL
436    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
437      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
438    
439    
440    
441    /*************************************************
442    *            Find an error text                  *
443    *************************************************/
444    
445    /* The error texts are now all in one long string, to save on relocations. As
446    some of the text is of unknown length, we can't use a table of offsets.
447    Instead, just count through the strings. This is not a performance issue
448    because it happens only when there has been a compilation error.
449    
450    Argument:   the error number
451    Returns:    pointer to the error string
452    */
453    
454    static const char *
455    find_error_text(int n)
456    {
457    const char *s = error_texts;
458    for (; n > 0; n--) while (*s++ != 0);
459    return s;
460    }
461    
462    
463  /*************************************************  /*************************************************
464  *            Handle escapes                      *  *            Handle escapes                      *
# Line 399  Arguments: Line 481  Arguments:
481    
482  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
483                   negative => a special escape sequence                   negative => a special escape sequence
484                   on error, errorptr is set                   on error, errorcodeptr is set
485  */  */
486    
487  static int  static int
# Line 417  ptr--;                            /* Set Line 499  ptr--;                            /* Set
499    
500  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
501    
502  /* 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
503  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.
504  Otherwise further processing may be required. */  Otherwise further processing may be required. */
505    
506  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
507  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */
508  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
509    
510  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
511  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
512  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
513  #endif  #endif
514    
# Line 450  else Line 532  else
532      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
533      break;      break;
534    
535      /* \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:
536      is an absolute backreference. If negative, it is a relative backreference.  
537      This is a Perl 5.10 feature. */      (1) A number, either plain or braced. If positive, it is an absolute
538        backreference. If negative, it is a relative backreference. This is a Perl
539        5.10 feature.
540    
541        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
542        is part of Perl's movement towards a unified syntax for back references. As
543        this is synonymous with \k{name}, we fudge it up by pretending it really
544        was \k.
545    
546        (3) For Oniguruma compatibility we also support \g followed by a name or a
547        number either in angle brackets or in single quotes. However, these are
548        (possibly recursive) subroutine calls, _not_ backreferences. Just return
549        the -ESC_g code (cf \k). */
550    
551      case 'g':      case 'g':
552        if (ptr[1] == '<' || ptr[1] == '\'')
553          {
554          c = -ESC_g;
555          break;
556          }
557    
558        /* Handle the Perl-compatible cases */
559    
560      if (ptr[1] == '{')      if (ptr[1] == '{')
561        {        {
562          const uschar *p;
563          for (p = ptr+2; *p != 0 && *p != '}'; p++)
564            if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;
565          if (*p != 0 && *p != '}')
566            {
567            c = -ESC_k;
568            break;
569            }
570        braced = TRUE;        braced = TRUE;
571        ptr++;        ptr++;
572        }        }
# Line 473  else Line 583  else
583      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
584        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - '0';
585    
586      if (c == 0 || (braced && *(++ptr) != '}'))      if (c < 0)   /* Integer overflow */
587          {
588          *errorcodeptr = ERR61;
589          break;
590          }
591    
592        if (braced && *(++ptr) != '}')
593        {        {
594        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
595        return 0;        break;
596        }        }
597    
598        if (c == 0)
599          {
600          *errorcodeptr = ERR58;
601          break;
602          }
603    
604      if (negated)      if (negated)
605        {        {
606        if (c > bracount)        if (c > bracount)
607          {          {
608          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
609          return 0;          break;
610          }          }
611        c = bracount - (c - 1);        c = bracount - (c - 1);
612        }        }
# Line 513  else Line 635  else
635        c -= '0';        c -= '0';
636        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
637          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
638          if (c < 0)    /* Integer overflow */
639            {
640            *errorcodeptr = ERR61;
641            break;
642            }
643        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
644          {          {
645          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 608  else Line 735  else
735      if (c == 0)      if (c == 0)
736        {        {
737        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
738        return 0;        break;
739        }        }
740    
741  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
# Line 621  else Line 748  else
748      break;      break;
749    
750      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
751      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
752      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
753      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
754      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
755    
756      default:      default:
757      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 684  if (c == '{') Line 811  if (c == '{')
811      *negptr = TRUE;      *negptr = TRUE;
812      ptr++;      ptr++;
813      }      }
814    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
815      {      {
816      c = *(++ptr);      c = *(++ptr);
817      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
# Line 713  top = _pcre_utt_size; Line 840  top = _pcre_utt_size;
840  while (bot < top)  while (bot < top)
841    {    {
842    i = (bot + top) >> 1;    i = (bot + top) >> 1;
843    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
844    if (c == 0)    if (c == 0)
845      {      {
846      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 881  for (; *ptr != 0; ptr++) Line 1008  for (; *ptr != 0; ptr++)
1008      continue;      continue;
1009      }      }
1010    
1011    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1012      are handled for real. If the first character is '^', skip it. Also, if the
1013      first few characters (either before or after ^) are \Q\E or \E we skip them
1014      too. This makes for compatibility with Perl. */
1015    
1016    if (*ptr == '[')    if (*ptr == '[')
1017      {      {
1018        BOOL negate_class = FALSE;
1019        for (;;)
1020          {
1021          int c = *(++ptr);
1022          if (c == '\\')
1023            {
1024            if (ptr[1] == 'E') ptr++;
1025              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
1026                else break;
1027            }
1028          else if (!negate_class && c == '^')
1029            negate_class = TRUE;
1030          else break;
1031          }
1032    
1033        /* If the next character is ']', it is a data character that must be
1034        skipped. */
1035    
1036        if (ptr[1] == ']') ptr++;
1037    
1038      while (*(++ptr) != ']')      while (*(++ptr) != ']')
1039        {        {
1040          if (*ptr == 0) return -1;
1041        if (*ptr == '\\')        if (*ptr == '\\')
1042          {          {
1043          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) return -1;
# Line 914  for (; *ptr != 0; ptr++) Line 1065  for (; *ptr != 0; ptr++)
1065    /* An opening parens must now be a real metacharacter */    /* An opening parens must now be a real metacharacter */
1066    
1067    if (*ptr != '(') continue;    if (*ptr != '(') continue;
1068    if (ptr[1] != '?')    if (ptr[1] != '?' && ptr[1] != '*')
1069      {      {
1070      count++;      count++;
1071      if (name == NULL && count == lorn) return count;      if (name == NULL && count == lorn) return count;
# Line 1042  for (;;) Line 1193  for (;;)
1193    {    {
1194    int d;    int d;
1195    register int op = *cc;    register int op = *cc;
   
1196    switch (op)    switch (op)
1197      {      {
1198      case OP_CBRA:      case OP_CBRA:
# Line 1131  for (;;) Line 1281  for (;;)
1281    
1282      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1283      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1284        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1285      cc += 4;      cc += 4;
1286      break;      break;
1287    
# Line 1239  for (;;) Line 1390  for (;;)
1390      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1391      }      }
1392    
1393    /* 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
1394    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
1395    arrange to skip the extra bytes. */    two bytes of parameters. */
1396    
1397    else    else
1398      {      {
1399        switch(c)
1400          {
1401          case OP_TYPESTAR:
1402          case OP_TYPEMINSTAR:
1403          case OP_TYPEPLUS:
1404          case OP_TYPEMINPLUS:
1405          case OP_TYPEQUERY:
1406          case OP_TYPEMINQUERY:
1407          case OP_TYPEPOSSTAR:
1408          case OP_TYPEPOSPLUS:
1409          case OP_TYPEPOSQUERY:
1410          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1411          break;
1412    
1413          case OP_TYPEUPTO:
1414          case OP_TYPEMINUPTO:
1415          case OP_TYPEEXACT:
1416          case OP_TYPEPOSUPTO:
1417          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1418          break;
1419          }
1420    
1421        /* Add in the fixed length from the table */
1422    
1423      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1424    
1425      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1426      a multi-byte character. The length in the table is a minimum, so we have to
1427      arrange to skip the extra bytes. */
1428    
1429    #ifdef SUPPORT_UTF8
1430      if (utf8) switch(c)      if (utf8) switch(c)
1431        {        {
1432        case OP_CHAR:        case OP_CHAR:
# Line 1266  for (;;) Line 1447  for (;;)
1447        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1448        break;        break;
1449        }        }
1450    #endif
1451      }      }
1452    }    }
1453  }  }
# Line 1301  for (;;) Line 1483  for (;;)
1483    
1484    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1485    
1486    /* 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
1487    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
1488    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1489    
1490    else    else
1491      {      {
1492        switch(c)
1493          {
1494          case OP_TYPESTAR:
1495          case OP_TYPEMINSTAR:
1496          case OP_TYPEPLUS:
1497          case OP_TYPEMINPLUS:
1498          case OP_TYPEQUERY:
1499          case OP_TYPEMINQUERY:
1500          case OP_TYPEPOSSTAR:
1501          case OP_TYPEPOSPLUS:
1502          case OP_TYPEPOSQUERY:
1503          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1504          break;
1505    
1506          case OP_TYPEPOSUPTO:
1507          case OP_TYPEUPTO:
1508          case OP_TYPEMINUPTO:
1509          case OP_TYPEEXACT:
1510          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1511          break;
1512          }
1513    
1514        /* Add in the fixed length from the table */
1515    
1516      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1517    
1518        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1519        by a multi-byte character. The length in the table is a minimum, so we have
1520        to arrange to skip the extra bytes. */
1521    
1522    #ifdef SUPPORT_UTF8
1523      if (utf8) switch(c)      if (utf8) switch(c)
1524        {        {
1525        case OP_CHAR:        case OP_CHAR:
# Line 1329  for (;;) Line 1540  for (;;)
1540        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1541        break;        break;
1542        }        }
1543    #endif
1544      }      }
1545    }    }
1546  }  }
# Line 1343  for (;;) Line 1555  for (;;)
1555  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()
1556  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
1557  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1558  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
1559  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
1560    bracket whose current branch will already have been scanned.
1561    
1562  Arguments:  Arguments:
1563    code        points to start of search    code        points to start of search
# Line 1366  for (code = first_significant_code(code Line 1579  for (code = first_significant_code(code
1579    
1580    c = *code;    c = *code;
1581    
1582    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
1583      first_significant_code() with a TRUE final argument. */
1584    
1585      if (c == OP_ASSERT)
1586        {
1587        do code += GET(code, 1); while (*code == OP_ALT);
1588        c = *code;
1589        continue;
1590        }
1591    
1592      /* Groups with zero repeats can of course be empty; skip them. */
1593    
1594      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1595        {
1596        code += _pcre_OP_lengths[c];
1597        do code += GET(code, 1); while (*code == OP_ALT);
1598        c = *code;
1599        continue;
1600        }
1601    
1602      /* For other groups, scan the branches. */
1603    
1604      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1605      {      {
1606      BOOL empty_branch;      BOOL empty_branch;
1607      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1382  for (code = first_significant_code(code Line 1617  for (code = first_significant_code(code
1617        }        }
1618      while (*code == OP_ALT);      while (*code == OP_ALT);
1619      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
1620        c = *code;
     /* Move past the KET and fudge things so that the increment in the "for"  
     above has no effect. */  
   
     c = OP_END;  
     code += 1 + LINK_SIZE - _pcre_OP_lengths[c];  
1621      continue;      continue;
1622      }      }
1623    
# Line 1395  for (code = first_significant_code(code Line 1625  for (code = first_significant_code(code
1625    
1626    switch (c)    switch (c)
1627      {      {
1628      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1629        cannot be represented just by a bit map. This includes negated single
1630        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1631        actual length is stored in the compiled code, so we must update "code"
1632        here. */
1633    
1634  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1635      case OP_XCLASS:      case OP_XCLASS:
1636      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1637      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1638  #endif  #endif
1639    
# Line 1461  for (code = first_significant_code(code Line 1695  for (code = first_significant_code(code
1695      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1696      return FALSE;      return FALSE;
1697    
1698        /* These are going to continue, as they may be empty, but we have to
1699        fudge the length for the \p and \P cases. */
1700    
1701        case OP_TYPESTAR:
1702        case OP_TYPEMINSTAR:
1703        case OP_TYPEPOSSTAR:
1704        case OP_TYPEQUERY:
1705        case OP_TYPEMINQUERY:
1706        case OP_TYPEPOSQUERY:
1707        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1708        break;
1709    
1710        /* Same for these */
1711    
1712        case OP_TYPEUPTO:
1713        case OP_TYPEMINUPTO:
1714        case OP_TYPEPOSUPTO:
1715        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1716        break;
1717    
1718      /* End of branch */      /* End of branch */
1719    
1720      case OP_KET:      case OP_KET:
# Line 1530  return TRUE; Line 1784  return TRUE;
1784  *************************************************/  *************************************************/
1785    
1786  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1787  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
1788  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1789  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1790    
1791    Originally, this function only recognized a sequence of letters between the
1792    terminators, but it seems that Perl recognizes any sequence of characters,
1793    though of course unknown POSIX names are subsequently rejected. Perl gives an
1794    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1795    didn't consider this to be a POSIX class. Likewise for [:1234:].
1796    
1797    The problem in trying to be exactly like Perl is in the handling of escapes. We
1798    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1799    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1800    below handles the special case of \], but does not try to do any other escape
1801    processing. This makes it different from Perl for cases such as [:l\ower:]
1802    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1803    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1804    I think.
1805    
1806  Argument:  Arguments:
1807    ptr      pointer to the initial [    ptr      pointer to the initial [
1808    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1809    
1810  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1811  */  */
1812    
1813  static BOOL  static BOOL
1814  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1815  {  {
1816  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1817  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1818  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1819    {    {
1820    *endptr = ptr;    if (*ptr == '\\' && ptr[1] == ']') ptr++; else
1821    return TRUE;      {
1822        if (*ptr == ']') return FALSE;
1823        if (*ptr == terminator && ptr[1] == ']')
1824          {
1825          *endptr = ptr;
1826          return TRUE;
1827          }
1828        }
1829    }    }
1830  return FALSE;  return FALSE;
1831  }  }
# Line 1577  Returns:     a value representing the na Line 1850  Returns:     a value representing the na
1850  static int  static int
1851  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1852  {  {
1853    const char *pn = posix_names;
1854  register int yield = 0;  register int yield = 0;
1855  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1856    {    {
1857    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1858      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1859      pn += posix_name_lengths[yield] + 1;
1860    yield++;    yield++;
1861    }    }
1862  return -1;  return -1;
# Line 1596  return -1; Line 1871  return -1;
1871  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
1872  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
1873  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
1874  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
1875  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
1876  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
1877  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
1878  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
1879    OP_END.
1880    
1881  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
1882  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 1623  adjust_recurse(uschar *group, int adjust Line 1899  adjust_recurse(uschar *group, int adjust
1899    uschar *save_hwm)    uschar *save_hwm)
1900  {  {
1901  uschar *ptr = group;  uschar *ptr = group;
1902    
1903  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1904    {    {
1905    int offset;    int offset;
# Line 1630  while ((ptr = (uschar *)find_recurse(ptr Line 1907  while ((ptr = (uschar *)find_recurse(ptr
1907    
1908    /* See if this recursion is on the forward reference list. If so, adjust the    /* See if this recursion is on the forward reference list. If so, adjust the
1909    reference. */    reference. */
1910    
1911    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1912      {      {
1913      offset = GET(hc, 0);      offset = GET(hc, 0);
# Line 1884  if (next >= 0) switch(op_code) Line 2161  if (next >= 0) switch(op_code)
2161    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2162    
2163    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2164    if (item == next) return TRUE;    if (item == next) return TRUE;
2165    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2166  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1921  if (next >= 0) switch(op_code) Line 2197  if (next >= 0) switch(op_code)
2197    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2198    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2199    
2200      case OP_HSPACE:
2201      case OP_NOT_HSPACE:
2202      switch(next)
2203        {
2204        case 0x09:
2205        case 0x20:
2206        case 0xa0:
2207        case 0x1680:
2208        case 0x180e:
2209        case 0x2000:
2210        case 0x2001:
2211        case 0x2002:
2212        case 0x2003:
2213        case 0x2004:
2214        case 0x2005:
2215        case 0x2006:
2216        case 0x2007:
2217        case 0x2008:
2218        case 0x2009:
2219        case 0x200A:
2220        case 0x202f:
2221        case 0x205f:
2222        case 0x3000:
2223        return op_code != OP_HSPACE;
2224        default:
2225        return op_code == OP_HSPACE;
2226        }
2227    
2228      case OP_VSPACE:
2229      case OP_NOT_VSPACE:
2230      switch(next)
2231        {
2232        case 0x0a:
2233        case 0x0b:
2234        case 0x0c:
2235        case 0x0d:
2236        case 0x85:
2237        case 0x2028:
2238        case 0x2029:
2239        return op_code != OP_VSPACE;
2240        default:
2241        return op_code == OP_VSPACE;
2242        }
2243    
2244    default:    default:
2245    return FALSE;    return FALSE;
2246    }    }
# Line 1955  switch(op_code) Line 2275  switch(op_code)
2275      case ESC_W:      case ESC_W:
2276      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2277    
2278        case ESC_h:
2279        case ESC_H:
2280        switch(item)
2281          {
2282          case 0x09:
2283          case 0x20:
2284          case 0xa0:
2285          case 0x1680:
2286          case 0x180e:
2287          case 0x2000:
2288          case 0x2001:
2289          case 0x2002:
2290          case 0x2003:
2291          case 0x2004:
2292          case 0x2005:
2293          case 0x2006:
2294          case 0x2007:
2295          case 0x2008:
2296          case 0x2009:
2297          case 0x200A:
2298          case 0x202f:
2299          case 0x205f:
2300          case 0x3000:
2301          return -next != ESC_h;
2302          default:
2303          return -next == ESC_h;
2304          }
2305    
2306        case ESC_v:
2307        case ESC_V:
2308        switch(item)
2309          {
2310          case 0x0a:
2311          case 0x0b:
2312          case 0x0c:
2313          case 0x0d:
2314          case 0x85:
2315          case 0x2028:
2316          case 0x2029:
2317          return -next != ESC_v;
2318          default:
2319          return -next == ESC_v;
2320          }
2321    
2322      default:      default:
2323      return FALSE;      return FALSE;
2324      }      }
2325    
2326    case OP_DIGIT:    case OP_DIGIT:
2327    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2328             next == -ESC_h || next == -ESC_v;
2329    
2330    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2331    return next == -ESC_d;    return next == -ESC_d;
# Line 1969  switch(op_code) Line 2334  switch(op_code)
2334    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2335    
2336    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2337    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2338    
2339      case OP_HSPACE:
2340      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2341    
2342      case OP_NOT_HSPACE:
2343      return next == -ESC_h;
2344    
2345      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2346      case OP_VSPACE:
2347      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2348    
2349      case OP_NOT_VSPACE:
2350      return next == -ESC_v;
2351    
2352    case OP_WORDCHAR:    case OP_WORDCHAR:
2353    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2354    
2355    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2356    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2045  uschar classbits[32]; Line 2423  uschar classbits[32];
2423  BOOL class_utf8;  BOOL class_utf8;
2424  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2425  uschar *class_utf8data;  uschar *class_utf8data;
2426    uschar *class_utf8data_base;
2427  uschar utf8_char[6];  uschar utf8_char[6];
2428  #else  #else
2429  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2084  req_caseopt = ((options & PCRE_CASELESS) Line 2463  req_caseopt = ((options & PCRE_CASELESS)
2463  for (;; ptr++)  for (;; ptr++)
2464    {    {
2465    BOOL negate_class;    BOOL negate_class;
2466      BOOL should_flip_negation;
2467    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2468    BOOL is_quantifier;    BOOL is_quantifier;
2469    BOOL is_recurse;    BOOL is_recurse;
2470      BOOL reset_bracount;
2471    int class_charcount;    int class_charcount;
2472    int class_lastchar;    int class_lastchar;
2473    int newoptions;    int newoptions;
2474    int recno;    int recno;
2475      int refsign;
2476    int skipbytes;    int skipbytes;
2477    int subreqbyte;    int subreqbyte;
2478    int subfirstbyte;    int subfirstbyte;
# Line 2101  for (;; ptr++) Line 2483  for (;; ptr++)
2483    /* Get next byte in the pattern */    /* Get next byte in the pattern */
2484    
2485    c = *ptr;    c = *ptr;
2486    
2487    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
2488    previous cycle of this loop. */    previous cycle of this loop. */
2489    
# Line 2123  for (;; ptr++) Line 2505  for (;; ptr++)
2505      */      */
2506    
2507      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2508    
2509        /* Paranoid check for integer overflow */
2510    
2511        if (OFLOW_MAX - *lengthptr < code - last_code)
2512          {
2513          *errorcodeptr = ERR20;
2514          goto FAILED;
2515          }
2516    
2517      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2518      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));
2519    
# Line 2235  for (;; ptr++) Line 2626  for (;; ptr++)
2626      *ptrptr = ptr;      *ptrptr = ptr;
2627      if (lengthptr != NULL)      if (lengthptr != NULL)
2628        {        {
2629          if (OFLOW_MAX - *lengthptr < code - last_code)
2630            {
2631            *errorcodeptr = ERR20;
2632            goto FAILED;
2633            }
2634        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2635        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2636        }        }
# Line 2282  for (;; ptr++) Line 2678  for (;; ptr++)
2678      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,
2679      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2680      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.
2681      */  
2682        In JavaScript compatibility mode, an isolated ']' causes an error. In
2683        default (Perl) mode, it is treated as a data character. */
2684    
2685        case ']':
2686        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2687          {
2688          *errorcodeptr = ERR64;
2689          goto FAILED;
2690          }
2691        goto NORMAL_CHAR;
2692    
2693      case '[':      case '[':
2694      previous = code;      previous = code;
# Line 2291  for (;; ptr++) Line 2697  for (;; ptr++)
2697      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. */
2698    
2699      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2700          check_posix_syntax(ptr, &tempptr, cd))          check_posix_syntax(ptr, &tempptr))
2701        {        {
2702        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;
2703        goto FAILED;        goto FAILED;
2704        }        }
2705    
2706      /* 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,
2707        if the first few characters (either before or after ^) are \Q\E or \E we
2708        skip them too. This makes for compatibility with Perl. */
2709    
2710      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2711        for (;;)
2712        {        {
       negate_class = TRUE;  
2713        c = *(++ptr);        c = *(++ptr);
2714        }        if (c == '\\')
2715      else          {
2716        {          if (ptr[1] == 'E') ptr++;
2717        negate_class = FALSE;            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2718                else break;
2719            }
2720          else if (!negate_class && c == '^')
2721            negate_class = TRUE;
2722          else break;
2723        }        }
2724    
2725        /* If a class contains a negative special such as \S, we need to flip the
2726        negation flag at the end, so that support for characters > 255 works
2727        correctly (they are all included in the class). */
2728    
2729        should_flip_negation = FALSE;
2730    
2731      /* 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
2732      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
2733      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2326  for (;; ptr++) Line 2745  for (;; ptr++)
2745  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2746      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2747      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2748        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2749  #endif  #endif
2750    
2751      /* 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 2341  for (;; ptr++) Line 2761  for (;; ptr++)
2761          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2762          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2763          }          }
2764    
2765          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2766          data and reset the pointer. This is so that very large classes that
2767          contain a zillion UTF-8 characters no longer overwrite the work space
2768          (which is on the stack). */
2769    
2770          if (lengthptr != NULL)
2771            {
2772            *lengthptr += class_utf8data - class_utf8data_base;
2773            class_utf8data = class_utf8data_base;
2774            }
2775    
2776  #endif  #endif
2777    
2778        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 2364  for (;; ptr++) Line 2796  for (;; ptr++)
2796    
2797        if (c == '[' &&        if (c == '[' &&
2798            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2799            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr))
2800          {          {
2801          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2802          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
# Line 2381  for (;; ptr++) Line 2813  for (;; ptr++)
2813          if (*ptr == '^')          if (*ptr == '^')
2814            {            {
2815            local_negate = TRUE;            local_negate = TRUE;
2816              should_flip_negation = TRUE;  /* Note negative special */
2817            ptr++;            ptr++;
2818            }            }
2819    
# Line 2447  for (;; ptr++) Line 2880  for (;; ptr++)
2880        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
2881        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
2882        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2883        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
2884        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2885    
2886        if (c == '\\')        if (c == '\\')
# Line 2455  for (;; ptr++) Line 2888  for (;; ptr++)
2888          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2889          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
2890    
2891          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
2892          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */
2893          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2894          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
# Line 2467  for (;; ptr++) Line 2900  for (;; ptr++)
2900            else inescq = TRUE;            else inescq = TRUE;
2901            continue;            continue;
2902            }            }
2903            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2904    
2905          if (c < 0)          if (c < 0)
2906            {            {
# Line 2482  for (;; ptr++) Line 2916  for (;; ptr++)
2916              continue;              continue;
2917    
2918              case ESC_D:              case ESC_D:
2919                should_flip_negation = TRUE;
2920              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2921              continue;              continue;
2922    
# Line 2490  for (;; ptr++) Line 2925  for (;; ptr++)
2925              continue;              continue;
2926    
2927              case ESC_W:              case ESC_W:
2928                should_flip_negation = TRUE;
2929              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2930              continue;              continue;
2931    
# Line 2499  for (;; ptr++) Line 2935  for (;; ptr++)
2935              continue;              continue;
2936    
2937              case ESC_S:              case ESC_S:
2938                should_flip_negation = TRUE;
2939              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2940              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2941              continue;              continue;
2942    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
2943              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
2944              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
2945              }              }
# Line 2515  for (;; ptr++) Line 2949  for (;; ptr++)
2949            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2950                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
2951    
2952              /* We need to deal with \H, \h, \V, and \v in both phases because
2953              they use extra memory. */
2954    
2955              if (-c == ESC_h)
2956                {
2957                SETBIT(classbits, 0x09); /* VT */
2958                SETBIT(classbits, 0x20); /* SPACE */
2959                SETBIT(classbits, 0xa0); /* NSBP */
2960    #ifdef SUPPORT_UTF8
2961                if (utf8)
2962                  {
2963                  class_utf8 = TRUE;
2964                  *class_utf8data++ = XCL_SINGLE;
2965                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
2966                  *class_utf8data++ = XCL_SINGLE;
2967                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
2968                  *class_utf8data++ = XCL_RANGE;
2969                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
2970                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
2971                  *class_utf8data++ = XCL_SINGLE;
2972                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
2973                  *class_utf8data++ = XCL_SINGLE;
2974                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
2975                  *class_utf8data++ = XCL_SINGLE;
2976                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
2977                  }
2978    #endif
2979                continue;
2980                }
2981    
2982              if (-c == ESC_H)
2983                {
2984                for (c = 0; c < 32; c++)
2985                  {
2986                  int x = 0xff;
2987                  switch (c)
2988                    {
2989                    case 0x09/8: x ^= 1 << (0x09%8); break;
2990                    case 0x20/8: x ^= 1 << (0x20%8); break;
2991                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
2992                    default: break;
2993                    }
2994                  classbits[c] |= x;
2995                  }
2996    
2997    #ifdef SUPPORT_UTF8
2998                if (utf8)
2999                  {
3000                  class_utf8 = TRUE;
3001                  *class_utf8data++ = XCL_RANGE;
3002                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3003                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3004                  *class_utf8data++ = XCL_RANGE;
3005                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3006                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3007                  *class_utf8data++ = XCL_RANGE;
3008                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3009                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3010                  *class_utf8data++ = XCL_RANGE;
3011                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3012                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3013                  *class_utf8data++ = XCL_RANGE;
3014                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3015                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3016                  *class_utf8data++ = XCL_RANGE;
3017                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3018                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3019                  *class_utf8data++ = XCL_RANGE;
3020                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3021                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3022                  }
3023    #endif
3024                continue;
3025                }
3026    
3027              if (-c == ESC_v)
3028                {
3029                SETBIT(classbits, 0x0a); /* LF */
3030                SETBIT(classbits, 0x0b); /* VT */
3031                SETBIT(classbits, 0x0c); /* FF */
3032                SETBIT(classbits, 0x0d); /* CR */
3033                SETBIT(classbits, 0x85); /* NEL */
3034    #ifdef SUPPORT_UTF8
3035                if (utf8)
3036                  {
3037                  class_utf8 = TRUE;
3038                  *class_utf8data++ = XCL_RANGE;
3039                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3040                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3041                  }
3042    #endif
3043                continue;
3044                }
3045    
3046              if (-c == ESC_V)
3047                {
3048                for (c = 0; c < 32; c++)
3049                  {
3050                  int x = 0xff;
3051                  switch (c)
3052                    {
3053                    case 0x0a/8: x ^= 1 << (0x0a%8);
3054                                 x ^= 1 << (0x0b%8);
3055                                 x ^= 1 << (0x0c%8);
3056                                 x ^= 1 << (0x0d%8);
3057                                 break;
3058                    case 0x85/8: x ^= 1 << (0x85%8); break;
3059                    default: break;
3060                    }
3061                  classbits[c] |= x;
3062                  }
3063    
3064    #ifdef SUPPORT_UTF8
3065                if (utf8)
3066                  {
3067                  class_utf8 = TRUE;
3068                  *class_utf8data++ = XCL_RANGE;
3069                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3070                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3071                  *class_utf8data++ = XCL_RANGE;
3072                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3073                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3074                  }
3075    #endif
3076                continue;
3077                }
3078    
3079            /* We need to deal with \P and \p in both phases. */            /* We need to deal with \P and \p in both phases. */
3080    
3081  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 2566  for (;; ptr++) Line 3127  for (;; ptr++)
3127    
3128        oldptr = ptr;        oldptr = ptr;
3129    
3130          /* Remember \r or \n */
3131    
3132          if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3133    
3134          /* Check for range */
3135    
3136        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == '-')
3137          {          {
3138          int d;          int d;
# Line 2607  for (;; ptr++) Line 3174  for (;; ptr++)
3174            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3175            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3176    
3177            /* \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
3178            special means the '-' was literal */            special means the '-' was literal */
3179    
3180            if (d < 0)            if (d < 0)
# Line 2633  for (;; ptr++) Line 3200  for (;; ptr++)
3200    
3201          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3202    
3203            /* Remember \r or \n */
3204    
3205            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3206    
3207          /* 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
3208          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3209          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2655  for (;; ptr++) Line 3226  for (;; ptr++)
3226              unsigned int origd = d;              unsigned int origd = d;
3227              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3228                {                {
3229                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3230                      ocd <= (unsigned int)d)
3231                    continue;                          /* Skip embedded ranges */
3232    
3233                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3234                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3235                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3236                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3237                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3238                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3239                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3240                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3241                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3242                  d = ocd;                  d = ocd;
3243                  continue;                  continue;
# Line 2782  for (;; ptr++) Line 3357  for (;; ptr++)
3357        goto FAILED;        goto FAILED;
3358        }        }
3359    
3360    
3361    /* This code has been disabled because it would mean that \s counts as
3362    an explicit \r or \n reference, and that's not really what is wanted. Now
3363    we set the flag only if there is a literal "\r" or "\n" in the class. */
3364    
3365    #if 0
3366        /* Remember whether \r or \n are in this class */
3367    
3368        if (negate_class)
3369          {
3370          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3371          }
3372        else
3373          {
3374          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3375          }
3376    #endif
3377    
3378    
3379      /* 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
3380      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
3381      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
3382      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3383      single-bytes only. This is an historical hangover. Maybe one day we can  
3384      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3385        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3386        operate on single-bytes only. This is an historical hangover. Maybe one day
3387        we can tidy these opcodes to handle multi-byte characters.
3388    
3389      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
3390      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 2797  for (;; ptr++) Line 3394  for (;; ptr++)
3394      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3395    
3396  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3397      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3398            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3399  #else  #else
3400      if (class_charcount == 1)      if (class_charcount == 1)
3401  #endif  #endif
# Line 2843  for (;; ptr++) Line 3438  for (;; ptr++)
3438      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3439    
3440      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3441      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3442      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
3443        the class, so any that were explicitly given as well can be ignored. If
3444        (when there are explicit characters > 255 that must be listed) there are no
3445        characters < 256, we can omit the bitmap in the actual compiled code. */
3446    
3447  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3448      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3449        {        {
3450        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3451        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2873  for (;; ptr++) Line 3471  for (;; ptr++)
3471        }        }
3472  #endif  #endif
3473    
3474      /* 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
3475      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
3476      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
3477      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3478    
3479        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3480      if (negate_class)      if (negate_class)
3481        {        {
       *code++ = OP_NCLASS;  
3482        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3483          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3484        }        }
3485      else      else
3486        {        {
       *code++ = OP_CLASS;  
3487        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3488        }        }
3489      code += 32;      code += 32;
# Line 3072  for (;; ptr++) Line 3669  for (;; ptr++)
3669        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3670        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3671    
3672        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3673    
3674        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3675    
# Line 3222  for (;; ptr++) Line 3819  for (;; ptr++)
3819        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3820        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3821    
3822        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3823    
3824        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3825          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3258  for (;; ptr++) Line 3855  for (;; ptr++)
3855          goto FAILED;          goto FAILED;
3856          }          }
3857    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
3858        /* 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
3859        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
3860        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 3288  for (;; ptr++) Line 3877  for (;; ptr++)
3877    
3878        if (repeat_min == 0)        if (repeat_min == 0)
3879          {          {
3880          /* 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
3881          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
3882    
3883          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
3884          BRAZERO and do no more at this point. However, we do need to adjust          **   {
3885          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
3886          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
3887          the start of the whole regex. Temporarily terminate the pattern while          **   }
3888          doing this. */  
3889            However, that fails when a group is referenced as a subroutine from
3890            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
3891            so that it is skipped on execution. As we don't have a list of which
3892            groups are referenced, we cannot do this selectively.
3893    
3894            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
3895            and do no more at this point. However, we do need to adjust any
3896            OP_RECURSE calls inside the group that refer to the group itself or any
3897            internal or forward referenced group, because the offset is from the
3898            start of the whole regex. Temporarily terminate the pattern while doing
3899            this. */
3900    
3901          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
3902            {            {
3903            *code = OP_END;            *code = OP_END;
3904            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3905            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3906            code++;            code++;
3907              if (repeat_max == 0)
3908                {
3909                *previous++ = OP_SKIPZERO;
3910                goto END_REPEAT;
3911                }
3912            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
3913            }            }
3914    
# Line 3354  for (;; ptr++) Line 3953  for (;; ptr++)
3953          if (repeat_min > 1)          if (repeat_min > 1)
3954            {            {
3955            /* 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
3956            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
3957              potential integer overflow. */
3958    
3959            if (lengthptr != NULL)            if (lengthptr != NULL)
3960              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
3961                int delta = (repeat_min - 1)*length_prevgroup;
3962                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3963                                                                (double)INT_MAX ||
3964                    OFLOW_MAX - *lengthptr < delta)
3965                  {
3966                  *errorcodeptr = ERR20;
3967                  goto FAILED;
3968                  }
3969                *lengthptr += delta;
3970                }
3971    
3972            /* This is compiling for real */            /* This is compiling for real */
3973    
# Line 3395  for (;; ptr++) Line 4005  for (;; ptr++)
4005          /* 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
4006          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
4007          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
4008          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4009            paranoid checks to avoid integer overflow. */
4010    
4011          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4012            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4013              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4014                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4015              if ((double)repeat_max *
4016                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4017                      > (double)INT_MAX ||
4018                  OFLOW_MAX - *lengthptr < delta)
4019                {
4020                *errorcodeptr = ERR20;
4021                goto FAILED;
4022                }
4023              *lengthptr += delta;
4024              }
4025    
4026          /* This is compiling for real */          /* This is compiling for real */
4027    
# Line 3507  for (;; ptr++) Line 4129  for (;; ptr++)
4129        int len;        int len;
4130        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4131            *tempcode == OP_NOTEXACT)            *tempcode == OP_NOTEXACT)
4132          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode] +
4133              ((*tempcode == OP_TYPEEXACT &&
4134                 (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4135        len = code - tempcode;        len = code - tempcode;
4136        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4137          {          {
# Line 3551  for (;; ptr++) Line 4175  for (;; ptr++)
4175      /* ===================================================================*/      /* ===================================================================*/
4176      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4177      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4178      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. */  
4179    
4180      case '(':      case '(':
4181      newoptions = options;      newoptions = options;
4182      skipbytes = 0;      skipbytes = 0;
4183      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4184      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4185        reset_bracount = FALSE;
4186    
4187        /* First deal with various "verbs" that can be introduced by '*'. */
4188    
4189        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4190          {
4191          int i, namelen;
4192          const char *vn = verbnames;
4193          const uschar *name = ++ptr;
4194          previous = NULL;
4195          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4196          if (*ptr == ':')
4197            {
4198            *errorcodeptr = ERR59;   /* Not supported */
4199            goto FAILED;
4200            }
4201          if (*ptr != ')')
4202            {
4203            *errorcodeptr = ERR60;
4204            goto FAILED;
4205            }
4206          namelen = ptr - name;
4207          for (i = 0; i < verbcount; i++)
4208            {
4209            if (namelen == verbs[i].len &&
4210                strncmp((char *)name, vn, namelen) == 0)
4211              {
4212              *code = verbs[i].op;
4213              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4214              break;
4215              }
4216            vn += verbs[i].len + 1;
4217            }
4218          if (i < verbcount) continue;
4219          *errorcodeptr = ERR60;
4220          goto FAILED;
4221          }
4222    
4223        /* Deal with the extended parentheses; all are introduced by '?', and the
4224        appearance of any of them means that this is not a capturing group. */
4225    
4226      if (*(++ptr) == '?')      else if (*ptr == '?')
4227        {        {
4228        int i, set, unset, namelen;        int i, set, unset, namelen;
4229        int *optset;        int *optset;
# Line 3582  for (;; ptr++) Line 4244  for (;; ptr++)
4244    
4245    
4246          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4247            case '|':                 /* Reset capture count for each branch */
4248            reset_bracount = TRUE;
4249            /* Fall through */
4250    
4251            /* ------------------------------------------------------------ */
4252          case ':':                 /* Non-capturing bracket */          case ':':                 /* Non-capturing bracket */
4253          bravalue = OP_BRA;          bravalue = OP_BRA;
4254          ptr++;          ptr++;
# Line 3617  for (;; ptr++) Line 4284  for (;; ptr++)
4284    
4285          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
4286          skipbytes = 3;          skipbytes = 3;
4287            refsign = -1;
4288    
4289          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4290    
# Line 3640  for (;; ptr++) Line 4308  for (;; ptr++)
4308            terminator = '\'';            terminator = '\'';
4309            ptr++;            ptr++;
4310            }            }
4311          else terminator = 0;          else
4312              {
4313              terminator = 0;
4314              if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);
4315              }
4316    
4317          /* 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 */
4318    
# Line 3676  for (;; ptr++) Line 4348  for (;; ptr++)
4348          if (lengthptr != NULL) break;          if (lengthptr != NULL) break;
4349    
4350          /* 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
4351          reference. */          reference. If the string started with "+" or "-" we require the rest to
4352            be digits, in which case recno will be set. */
4353    
4354            if (refsign > 0)
4355              {
4356              if (recno <= 0)
4357                {
4358                *errorcodeptr = ERR58;
4359                goto FAILED;
4360                }
4361              recno = (refsign == '-')?
4362                cd->bracount - recno + 1 : recno +cd->bracount;
4363              if (recno <= 0 || recno > cd->final_bracount)
4364                {
4365                *errorcodeptr = ERR15;
4366                goto FAILED;
4367                }
4368              PUT2(code, 2+LINK_SIZE, recno);
4369              break;
4370              }
4371    
4372            /* Otherwise (did not start with "+" or "-"), start by looking for the
4373            name. */
4374    
4375          slot = cd->name_table;          slot = cd->name_table;
4376          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 3742  for (;; ptr++) Line 4436  for (;; ptr++)
4436            skipbytes = 1;            skipbytes = 1;
4437            }            }
4438    
4439          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4440            in the second pass here, so final_bracount is set. */
4441    
4442          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4443            {            {
4444            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4445            }            }
# Line 3768  for (;; ptr++) Line 4463  for (;; ptr++)
4463    
4464          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4465          case '!':                 /* Negative lookahead */          case '!':                 /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4466          ptr++;          ptr++;
4467            if (*ptr == ')')          /* Optimize (?!) */
4468              {
4469              *code++ = OP_FAIL;
4470              previous = NULL;
4471              continue;
4472              }
4473            bravalue = OP_ASSERT_NOT;
4474          break;          break;
4475    
4476    
# Line 3932  for (;; ptr++) Line 4633  for (;; ptr++)
4633    
4634          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4635          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4636          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4637            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4638            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4639    
4640          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4641          name = ++ptr;          name = ++ptr;
# Line 3944  for (;; ptr++) Line 4647  for (;; ptr++)
4647    
4648          if (lengthptr != NULL)          if (lengthptr != NULL)
4649            {            {
4650              if (namelen == 0)
4651                {
4652                *errorcodeptr = ERR62;
4653                goto FAILED;
4654                }
4655            if (*ptr != terminator)            if (*ptr != terminator)
4656              {              {
4657              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 3957  for (;; ptr++) Line 4665  for (;; ptr++)
4665            recno = 0;            recno = 0;
4666            }            }
4667    
4668          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4669            first, and then check that we have reached the end of the name in the
4670            table. That way, if the name that is longer than any in the table,
4671            the comparison will fail without reading beyond the table entry. */
4672    
4673          else          else
4674            {            {
4675            slot = cd->name_table;            slot = cd->name_table;
4676            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4677              {              {
4678              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4679                    slot[2+namelen] == 0)
4680                  break;
4681              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4682              }              }
4683    
# Line 3995  for (;; ptr++) Line 4708  for (;; ptr++)
4708    
4709    
4710          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4711            case '-': case '+':
4712          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */
4713          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case '5': case '6': case '7': case '8': case '9':   /* subroutine */
4714            {            {
4715            const uschar *called;            const uschar *called;
4716              terminator = ')';
4717    
4718              /* Come here from the \g<...> and \g'...' code (Oniguruma
4719              compatibility). However, the syntax has been checked to ensure that
4720              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
4721              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
4722              ever be taken. */
4723    
4724              HANDLE_NUMERICAL_RECURSION:
4725    
4726              if ((refsign = *ptr) == '+')
4727                {
4728                ptr++;
4729                if ((digitab[*ptr] & ctype_digit) == 0)
4730                  {
4731                  *errorcodeptr = ERR63;
4732                  goto FAILED;
4733                  }
4734                }
4735              else if (refsign == '-')
4736                {
4737                if ((digitab[ptr[1]] & ctype_digit) == 0)
4738                  goto OTHER_CHAR_AFTER_QUERY;
4739                ptr++;
4740                }
4741    
4742            recno = 0;            recno = 0;
4743            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4744              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - '0';
4745            if (*ptr != ')')  
4746              if (*ptr != terminator)
4747              {              {
4748              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
4749              goto FAILED;              goto FAILED;
4750              }              }
4751    
4752              if (refsign == '-')
4753                {
4754                if (recno == 0)
4755                  {
4756                  *errorcodeptr = ERR58;
4757                  goto FAILED;
4758                  }
4759                recno = cd->bracount - recno + 1;
4760                if (recno <= 0)
4761                  {
4762                  *errorcodeptr = ERR15;
4763                  goto FAILED;
4764                  }
4765                }
4766              else if (refsign == '+')
4767                {
4768                if (recno == 0)
4769                  {
4770                  *errorcodeptr = ERR58;
4771                  goto FAILED;
4772                  }
4773                recno += cd->bracount;
4774                }
4775    
4776            /* Come here from code above that handles a named recursion */            /* Come here from code above that handles a named recursion */
4777    
4778            HANDLE_RECURSION:            HANDLE_RECURSION:
# Line 4080  for (;; ptr++) Line 4845  for (;; ptr++)
4845    
4846          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4847          default:              /* Other characters: check option setting */          default:              /* Other characters: check option setting */
4848            OTHER_CHAR_AFTER_QUERY:
4849          set = unset = 0;          set = unset = 0;
4850          optset = &set;          optset = &set;
4851    
# Line 4091  for (;; ptr++) Line 4857  for (;; ptr++)
4857    
4858              case 'J':    /* Record that it changed in the external options */              case 'J':    /* Record that it changed in the external options */
4859              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
4860              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
4861              break;              break;
4862    
4863              case 'i': *optset |= PCRE_CASELESS; break;              case 'i': *optset |= PCRE_CASELESS; break;
# Line 4214  for (;; ptr++) Line 4980  for (;; ptr++)
4980           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
4981           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
4982            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
4983             reset_bracount,               /* True if (?| group */
4984           skipbytes,                    /* Skip over bracket number */           skipbytes,                    /* Skip over bracket number */
4985           &subfirstbyte,                /* For possible first char */           &subfirstbyte,                /* For possible first char */
4986           &subreqbyte,                  /* For possible last char */           &subreqbyte,                  /* For possible last char */
# Line 4230  for (;; ptr++) Line 4997  for (;; ptr++)
4997      is on the bracket. */      is on the bracket. */
4998    
4999      /* 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
5000      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
5001        in the real compile phase, not in the pre-pass, where the whole group may
5002        not be available. */
5003    
5004      if (bravalue == OP_COND)      if (bravalue == OP_COND && lengthptr == NULL)
5005        {        {
5006        uschar *tc = code;        uschar *tc = code;
5007        int condcount = 0;        int condcount = 0;
# Line 4279  for (;; ptr++) Line 5048  for (;; ptr++)
5048        goto FAILED;        goto FAILED;
5049        }        }
5050    
5051      /* 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,
5052      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
5053      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
5054      a quantifier. */      duplicated by a quantifier.*/
5055    
5056      if (lengthptr != NULL)      if (lengthptr != NULL)
5057        {        {
5058          if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
5059            {
5060            *errorcodeptr = ERR20;
5061            goto FAILED;
5062            }
5063        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5064        code++;        *code++ = OP_BRA;
5065        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5066        *code++ = OP_KET;        *code++ = OP_KET;
5067        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5068          break;    /* No need to waste time with special character handling */
5069        }        }
5070    
5071      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
5072    
5073      else code = tempcode;      code = tempcode;
5074    
5075      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
5076      relevant. */      relevant. */
# Line 4364  for (;; ptr++) Line 5139  for (;; ptr++)
5139      back references and those types that consume a character may be repeated.      back references and those types that consume a character may be repeated.
5140      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
5141      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5142    
5143      case '\\':      case '\\':
5144      tempptr = ptr;      tempptr = ptr;
5145      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
# Line 4391  for (;; ptr++) Line 5166  for (;; ptr++)
5166    
5167        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5168        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5169    
5170          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5171          is a subroutine call by number (Oniguruma syntax). In fact, the value
5172          -ESC_g is returned only for these cases. So we don't need to check for <
5173          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5174          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5175          that is a synonym for a named back reference). */
5176    
5177          if (-c == ESC_g)
5178            {
5179            const uschar *p;
5180            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5181            terminator = (*(++ptr) == '<')? '>' : '\'';
5182    
5183            /* These two statements stop the compiler for warning about possibly
5184            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5185            fact, because we actually check for a number below, the paths that
5186            would actually be in error are never taken. */
5187    
5188            skipbytes = 0;
5189            reset_bracount = FALSE;
5190    
5191            /* Test for a name */
5192    
5193            if (ptr[1] != '+' && ptr[1] != '-')
5194              {
5195              BOOL isnumber = TRUE;
5196              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5197                {
5198                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5199                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5200                }
5201              if (*p != terminator)
5202                {
5203                *errorcodeptr = ERR57;
5204                break;
5205                }
5206              if (isnumber)
5207                {
5208                ptr++;
5209                goto HANDLE_NUMERICAL_RECURSION;
5210                }
5211              is_recurse = TRUE;
5212              goto NAMED_REF_OR_RECURSE;
5213              }
5214    
5215            /* Test a signed number in angle brackets or quotes. */
5216    
5217            p = ptr + 2;
5218            while ((digitab[*p] & ctype_digit) != 0) p++;
5219            if (*p != terminator)
5220              {
5221              *errorcodeptr = ERR57;
5222              break;
5223              }
5224            ptr++;
5225            goto HANDLE_NUMERICAL_RECURSION;
5226            }
5227    
5228        /* \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).
5229          We also support \k{name} (.NET syntax) */
5230    
5231        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\''))        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))
5232          {          {
5233          is_recurse = FALSE;          is_recurse = FALSE;
5234          terminator = (*(++ptr) == '<')? '>' : '\'';          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';
5235          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5236          }          }
5237    
# Line 4498  for (;; ptr++) Line 5332  for (;; ptr++)
5332      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5333      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5334    
5335        /* Remember if \r or \n were seen */
5336    
5337        if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')
5338          cd->external_flags |= PCRE_HASCRORLF;
5339    
5340      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5341      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.
5342      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 4563  This function is used during the pre-com Line 5402  This function is used during the pre-com
5402  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
5403  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
5404    
5405  Argument:  Arguments:
5406    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
5407    oldims         previous settings of ims option bits    oldims         previous settings of ims option bits
5408    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
5409    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
5410    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
5411    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
5412      reset_bracount TRUE to reset the count for each branch
5413    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
5414    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
5415    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 4583  Returns:         TRUE on success Line 5423  Returns:         TRUE on success
5423    
5424  static BOOL  static BOOL
5425  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,
5426    int *errorcodeptr, BOOL lookbehind, int skipbytes, int *firstbyteptr,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
5427    int *reqbyteptr, branch_chain *bcptr, compile_data *cd, int *lengthptr)    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
5428      int *lengthptr)
5429  {  {
5430  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
5431  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 4594  uschar *reverse_count = NULL; Line 5435  uschar *reverse_count = NULL;
5435  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5436  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5437  int length;  int length;
5438    int orig_bracount;
5439    int max_bracount;
5440  branch_chain bc;  branch_chain bc;
5441    
5442  bc.outer = bcptr;  bc.outer = bcptr;
# Line 4622  code += 1 + LINK_SIZE + skipbytes; Line 5465  code += 1 + LINK_SIZE + skipbytes;
5465    
5466  /* Loop for each alternative branch */  /* Loop for each alternative branch */
5467    
5468    orig_bracount = max_bracount = cd->bracount;
5469  for (;;)  for (;;)
5470    {    {
5471      /* For a (?| group, reset the capturing bracket count so that each branch
5472      uses the same numbers. */
5473    
5474      if (reset_bracount) cd->bracount = orig_bracount;
5475    
5476    /* Handle a change of ims options at the start of the branch */    /* Handle a change of ims options at the start of the branch */
5477    
5478    if ((options & PCRE_IMS) != oldims)    if ((options & PCRE_IMS) != oldims)
# Line 4653  for (;;) Line 5502  for (;;)
5502      return FALSE;      return FALSE;
5503      }      }
5504    
5505      /* Keep the highest bracket count in case (?| was used and some branch
5506      has fewer than the rest. */
5507    
5508      if (cd->bracount > max_bracount) max_bracount = cd->bracount;
5509    
5510    /* In the real compile phase, there is some post-processing to be done. */    /* In the real compile phase, there is some post-processing to be done. */
5511    
5512    if (lengthptr == NULL)    if (lengthptr == NULL)
# Line 4716  for (;;) Line 5570  for (;;)
5570        }        }
5571      }      }
5572    
5573    /* Reached end of expression, either ')' or end of pattern. Go back through    /* Reached end of expression, either ')' or end of pattern. In the real
5574    the alternative branches and reverse the chain of offsets, with the field in    compile phase, go back through the alternative branches and reverse the chain
5575    the BRA item now becoming an offset to the first alternative. If there are    of offsets, with the field in the BRA item now becoming an offset to the
5576    no alternatives, it points to the end of the group. The length in the    first alternative. If there are no alternatives, it points to the end of the
5577    terminating ket is always the length of the whole bracketed item. If any of    group. The length in the terminating ket is always the length of the whole
5578    the ims options were changed inside the group, compile a resetting op-code    bracketed item. If any of the ims options were changed inside the group,
5579    following, except at the very end of the pattern. Return leaving the pointer    compile a resetting op-code following, except at the very end of the pattern.
5580    at the terminating char. */    Return leaving the pointer at the terminating char. */
5581    
5582    if (*ptr != '|')    if (*ptr != '|')
5583      {      {
5584      int branch_length = code - last_branch;      if (lengthptr == NULL)
     do  
5585        {        {
5586        int prev_length = GET(last_branch, 1);        int branch_length = code - last_branch;
5587        PUT(last_branch, 1, branch_length);        do
5588        branch_length = prev_length;          {
5589        last_branch -= branch_length;          int prev_length = GET(last_branch, 1);
5590            PUT(last_branch, 1, branch_length);
5591            branch_length = prev_length;
5592            last_branch -= branch_length;
5593            }
5594          while (branch_length > 0);
5595        }        }
     while (branch_length > 0);  
5596    
5597      /* Fill in the ket */      /* Fill in the ket */
5598    
# Line 4752  for (;;) Line 5609  for (;;)
5609        length += 2;        length += 2;
5610        }        }
5611    
5612        /* Retain the highest bracket number, in case resetting was used. */
5613    
5614        cd->bracount = max_bracount;
5615    
5616      /* Set values to pass back */      /* Set values to pass back */
5617    
5618      *codeptr = code;      *codeptr = code;
5619      *ptrptr = ptr;      *ptrptr = ptr;
5620      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
5621      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
5622      if (lengthptr != NULL) *lengthptr += length;      if (lengthptr != NULL)
5623          {
5624          if (OFLOW_MAX - *lengthptr < length)
5625            {
5626            *errorcodeptr = ERR20;
5627            return FALSE;
5628            }
5629          *lengthptr += length;
5630          }
5631      return TRUE;      return TRUE;
5632      }      }
5633    
5634    /* Another branch follows; insert an "or" node. Its length field points back    /* Another branch follows. In the pre-compile phase, we can move the code
5635      pointer back to where it was for the start of the first branch. (That is,
5636      pretend that each branch is the only one.)
5637    
5638      In the real compile phase, insert an ALT node. Its length field points back
5639    to the previous branch while the bracket remains open. At the end the chain    to the previous branch while the bracket remains open. At the end the chain
5640    is reversed. It's done like this so that the start of the bracket has a    is reversed. It's done like this so that the start of the bracket has a
5641    zero offset until it is closed, making it possible to detect recursion. */    zero offset until it is closed, making it possible to detect recursion. */
5642    
5643    *code = OP_ALT;    if (lengthptr != NULL)
5644    PUT(code, 1, code - last_branch);      {
5645    bc.current = last_branch = code;      code = *codeptr + 1 + LINK_SIZE + skipbytes;
5646    code += 1 + LINK_SIZE;      length += 1 + LINK_SIZE;
5647        }
5648      else
5649        {
5650        *code = OP_ALT;
5651        PUT(code, 1, code - last_branch);
5652        bc.current = last_branch = code;
5653        code += 1 + LINK_SIZE;
5654        }
5655    
5656    ptr++;    ptr++;
   length += 1 + LINK_SIZE;  
5657    }    }
5658  /* Control never reaches here */  /* Control never reaches here */
5659  }  }
# Line 5039  Returns:        pointer to compiled data Line 5920  Returns:        pointer to compiled data
5920                  with errorptr and erroroffset set                  with errorptr and erroroffset set
5921  */  */
5922    
5923  PCRE_DATA_SCOPE pcre *  PCRE_EXP_DEFN pcre *
5924  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
5925    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
5926  {  {
# Line 5047  return pcre_compile2(pattern, options, N Line 5928  return pcre_compile2(pattern, options, N
5928  }  }
5929    
5930    
5931  PCRE_DATA_SCOPE pcre *  PCRE_EXP_DEFN pcre *
5932  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
5933    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
5934  {  {
# Line 5055  real_pcre *re; Line 5936  real_pcre *re;
5936  int length = 1;  /* For final END opcode */  int length = 1;  /* For final END opcode */
5937  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
5938  int errorcode = 0;  int errorcode = 0;
5939    int skipatstart = 0;
5940  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5941  BOOL utf8;  BOOL utf8;
5942  #endif  #endif
# Line 5073  to fill in forward references to subpatt Line 5955  to fill in forward references to subpatt
5955    
5956  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
5957    
   
5958  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
5959    
5960  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;
# Line 5096  if (errorcodeptr != NULL) *errorcodeptr Line 5977  if (errorcodeptr != NULL) *errorcodeptr
5977  if (erroroffset == NULL)  if (erroroffset == NULL)
5978    {    {
5979    errorcode = ERR16;    errorcode = ERR16;
5980    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN2;
5981    }    }
5982    
5983  *erroroffset = 0;  *erroroffset = 0;
# Line 5109  if (utf8 && (options & PCRE_NO_UTF8_CHEC Line 5990  if (utf8 && (options & PCRE_NO_UTF8_CHEC
5990       (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)       (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)
5991    {    {
5992    errorcode = ERR44;    errorcode = ERR44;
5993    goto PCRE_UTF8_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN2;
5994    }    }
5995  #else  #else
5996  if ((options & PCRE_UTF8) != 0)  if ((options & PCRE_UTF8) != 0)
# Line 5133  cd->fcc = tables + fcc_offset; Line 6014  cd->fcc = tables + fcc_offset;
6014  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6015  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6016    
6017    /* Check for global one-time settings at the start of the pattern, and remember
6018    the offset for later. */
6019    
6020    while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')
6021      {
6022      int newnl = 0;
6023      int newbsr = 0;
6024    
6025      if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)
6026        { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6027      else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)
6028        { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6029      else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)
6030        { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6031      else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)
6032        { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6033      else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)
6034        { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6035    
6036      else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)
6037        { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6038      else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)
6039        { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6040    
6041      if (newnl != 0)
6042        options = (options & ~PCRE_NEWLINE_BITS) | newnl;
6043      else if (newbsr != 0)
6044        options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
6045      else break;
6046      }
6047    
6048    /* Check validity of \R options. */
6049    
6050    switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6051      {
6052      case 0:
6053      case PCRE_BSR_ANYCRLF:
6054      case PCRE_BSR_UNICODE:
6055      break;
6056      default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
6057      }
6058    
6059  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
6060  current code allows for fixed one- or two-byte sequences, plus "any". */  current code allows for fixed one- or two-byte sequences, plus "any" and
6061    "anycrlf". */
6062    
6063  switch (options & (PCRE_NEWLINE_CRLF | PCRE_NEWLINE_ANY))  switch (options & PCRE_NEWLINE_BITS)
6064    {    {
6065    case 0: newline = NEWLINE; break;   /* Compile-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6066    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = '\r'; break;
6067    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = '\n'; break;
6068    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6069         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;
6070    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6071      case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6072    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
6073    }    }
6074    
6075  if (newline < 0)  if (newline == -2)
6076      {
6077      cd->nltype = NLTYPE_ANYCRLF;
6078      }
6079    else if (newline < 0)
6080    {    {
6081    cd->nltype = NLTYPE_ANY;    cd->nltype = NLTYPE_ANY;
6082    }    }
# Line 5186  to compile parts of the pattern into; th Line 6115  to compile parts of the pattern into; th
6115  no longer needed, so hopefully this workspace will never overflow, though there  no longer needed, so hopefully this workspace will never overflow, though there
6116  is a test for its doing so. */  is a test for its doing so. */
6117    
6118  cd->bracount = 0;  cd->bracount = cd->final_bracount = 0;
6119  cd->names_found = 0;  cd->names_found = 0;
6120  cd->name_entry_size = 0;  cd->name_entry_size = 0;
6121  cd->name_table = NULL;  cd->name_table = NULL;
# Line 5196  cd->hwm = cworkspace; Line 6125  cd->hwm = cworkspace;
6125  cd->start_pattern = (const uschar *)pattern;  cd->start_pattern = (const uschar *)pattern;
6126  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));
6127  cd->req_varyopt = 0;  cd->req_varyopt = 0;
 cd->nopartial = FALSE;  
6128  cd->external_options = options;  cd->external_options = options;
6129    cd->external_flags = 0;
6130    
6131  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we
6132  don't need to look at the result of the function here. The initial options have  don't need to look at the result of the function here. The initial options have
# Line 5205  been put into the cd block so that they Line 6134  been put into the cd block so that they
6134  found within the regex right at the beginning. Bringing initial option settings  found within the regex right at the beginning. Bringing initial option settings
6135  outside can help speed up starting point checks. */  outside can help speed up starting point checks. */
6136    
6137    ptr += skipatstart;
6138  code = cworkspace;  code = cworkspace;
6139  *code = OP_BRA;  *code = OP_BRA;
6140  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,
6141    &code, &ptr, &errorcode, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,
6142      &length);
6143  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
6144    
6145  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 5234  if (re == NULL) Line 6165  if (re == NULL)
6165    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
6166    }    }
6167    
6168  /* Put in the magic number, and save the sizes, initial options, and character  /* Put in the magic number, and save the sizes, initial options, internal
6169  table pointer. NULL is used for the default character tables. The nullpad field  flags, and character table pointer. NULL is used for the default character
6170  is at the end; it's there to help in the case when a regex compiled on a system  tables. The nullpad field is at the end; it's there to help in the case when a
6171  with 4-byte pointers is run on another with 8-byte pointers. */  regex compiled on a system with 4-byte pointers is run on another with 8-byte
6172    pointers. */
6173    
6174  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
6175  re->size = size;  re->size = size;
6176  re->options = cd->external_options;  re->options = cd->external_options;
6177    re->flags = cd->external_flags;
6178  re->dummy1 = 0;  re->dummy1 = 0;
6179  re->first_byte = 0;  re->first_byte = 0;
6180  re->req_byte = 0;  re->req_byte = 0;
# Line 5259  field. Reset the bracket count and the n Line 6192  field. Reset the bracket count and the n
6192  field; this time it's used for remembering forward references to subpatterns.  field; this time it's used for remembering forward references to subpatterns.
6193  */  */
6194    
6195    cd->final_bracount = cd->bracount;  /* Save for checking forward references */
6196  cd->bracount = 0;  cd->bracount = 0;
6197  cd->names_found = 0;  cd->names_found = 0;
6198  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 5266  codestart = cd->name_table + re->name_en Line 6200  codestart = cd->name_table + re->name_en
6200  cd->start_code = codestart;  cd->start_code = codestart;
6201  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6202  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6203  cd->nopartial = FALSE;  cd->had_accept = FALSE;
6204    
6205  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6206  error, errorcode will be set non-zero, so we don't need to look at the result  error, errorcode will be set non-zero, so we don't need to look at the result
6207  of the function here. */  of the function here. */
6208    
6209  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern + skipatstart;
6210  code = (uschar *)codestart;  code = (uschar *)codestart;
6211  *code = OP_BRA;  *code = OP_BRA;
6212  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,
6213    &errorcode, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);
6214  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
6215  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
6216    re->flags = cd->external_flags;
6217    
6218  if (cd->nopartial) re->options |= PCRE_NOPARTIAL;  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */
6219    
6220  /* If not reached end of pattern on success, there's an excess bracket. */  /* If not reached end of pattern on success, there's an excess bracket. */
6221    
# Line 5308  while (errorcode == 0 && cd->hwm > cwork Line 6243  while (errorcode == 0 && cd->hwm > cwork
6243    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6244      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6245    }    }
6246    
6247  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
6248  subpattern. */  subpattern. */
6249    
# Line 5321  if (errorcode != 0) Line 6256  if (errorcode != 0)
6256    (pcre_free)(re);    (pcre_free)(re);
6257    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
6258    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = ptr - (const uschar *)pattern;
6259  #ifdef SUPPORT_UTF8    PCRE_EARLY_ERROR_RETURN2:
6260    PCRE_UTF8_ERROR_RETURN:    *errorptr = find_error_text(errorcode);
 #endif  
   *errorptr = error_texts[errorcode];  
6261    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
6262    return NULL;    return NULL;
6263    }    }
# Line 5353  if ((re->options & PCRE_ANCHORED) == 0) Line 6286  if ((re->options & PCRE_ANCHORED) == 0)
6286        int ch = firstbyte & 255;        int ch = firstbyte & 255;
6287        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&
6288           cd->fcc[ch] == ch)? ch : firstbyte;           cd->fcc[ch] == ch)? ch : firstbyte;
6289        re->options |= PCRE_FIRSTSET;        re->flags |= PCRE_FIRSTSET;
6290        }        }
6291      else if (is_startline(codestart, 0, cd->backref_map))      else if (is_startline(codestart, 0, cd->backref_map))
6292        re->options |= PCRE_STARTLINE;        re->flags |= PCRE_STARTLINE;
6293      }      }
6294    }    }
6295    
# Line 5370  if (reqbyte >= 0 && Line 6303  if (reqbyte >= 0 &&
6303    int ch = reqbyte & 255;    int ch = reqbyte & 255;
6304    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&
6305      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;
6306    re->options |= PCRE_REQCHSET;    re->flags |= PCRE_REQCHSET;
6307    }    }
6308    
6309  /* Print out the compiled data if debugging is enabled. This is never the  /* Print out the compiled data if debugging is enabled. This is never the
# Line 5381  case when building a production library. Line 6314  case when building a production library.
6314  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6315    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6316    
6317  if (re->options != 0)  printf("Options=%08x\n", re->options);
   {  
   printf("%s%s%s%s%s%s%s%s%s\n",  
     ((re->options & PCRE_NOPARTIAL) != 0)? "nopartial " : "",  
     ((re->options & PCRE_ANCHORED) != 0)? "anchored " : "",  
     ((re->options & PCRE_CASELESS) != 0)? "caseless " : "",  
     ((re->options & PCRE_EXTENDED) != 0)? "extended " : "",  
     ((re->options & PCRE_MULTILINE) != 0)? "multiline " : "",  
     ((re->options & PCRE_DOTALL) != 0)? "dotall " : "",  
     ((re->options & PCRE_DOLLAR_ENDONLY) != 0)? "endonly " : "",  
     ((re->options & PCRE_EXTRA) != 0)? "extra " : "",  
     ((re->options & PCRE_UNGREEDY) != 0)? "ungreedy " : "");  
   }  
6318    
6319  if ((re->options & PCRE_FIRSTSET) != 0)  if ((re->flags & PCRE_FIRSTSET) != 0)
6320    {    {
6321    int ch = re->first_byte & 255;    int ch = re->first_byte & 255;
6322    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?
# Line 5404  if ((re->options & PCRE_FIRSTSET) != 0) Line 6325  if ((re->options & PCRE_FIRSTSET) != 0)
6325      else printf("First char = \\x%02x%s\n", ch, caseless);      else printf("First char = \\x%02x%s\n", ch, caseless);
6326    }    }
6327    
6328  if ((re->options & PCRE_REQCHSET) != 0)  if ((re->flags & PCRE_REQCHSET) != 0)
6329    {    {
6330    int ch = re->req_byte & 255;    int ch = re->req_byte & 255;
6331    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?
# Line 5413  if ((re->options & PCRE_REQCHSET) != 0) Line 6334  if ((re->options & PCRE_REQCHSET) != 0)
6334      else printf("Req char = \\x%02x%s\n", ch, caseless);      else printf("Req char = \\x%02x%s\n", ch, caseless);
6335    }    }
6336    
6337  pcre_printint(re, stdout);  pcre_printint(re, stdout, TRUE);
6338    
6339  /* This check is done here in the debugging case so that the code that  /* This check is done here in the debugging case so that the code that
6340  was compiled can be seen. */  was compiled can be seen. */
# Line 5421  was compiled can be seen. */ Line 6342  was compiled can be seen. */
6342  if (code - codestart > length)  if (code - codestart > length)
6343    {    {
6344    (pcre_free)(re);    (pcre_free)(re);
6345    *errorptr = error_texts[ERR23];    *errorptr = find_error_text(ERR23);
6346    *erroroffset = ptr - (uschar *)pattern;    *erroroffset = ptr - (uschar *)pattern;
6347    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6348    return NULL;    return NULL;

Legend:
Removed from v.97  
changed lines
  Added in v.340

  ViewVC Help
Powered by ViewVC 1.1.5