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

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

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

revision 145 by ph10, Wed Apr 4 14:06:52 2007 UTC revision 333 by ph10, Thu Apr 10 19:55:57 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) 1997-2007 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 (?+";
306    
307    
308  /* 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 432  static const unsigned char ebcdic_charta
432  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
433    
434  static BOOL  static BOOL
435    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
436      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
437    
438    
439    
440  /*************************************************  /*************************************************
441    *            Find an error text                  *
442    *************************************************/
443    
444    /* The error texts are now all in one long string, to save on relocations. As
445    some of the text is of unknown length, we can't use a table of offsets.
446    Instead, just count through the strings. This is not a performance issue
447    because it happens only when there has been a compilation error.
448    
449    Argument:   the error number
450    Returns:    pointer to the error string
451    */
452    
453    static const char *
454    find_error_text(int n)
455    {
456    const char *s = error_texts;
457    for (; n > 0; n--) while (*s++ != 0);
458    return s;
459    }
460    
461    
462    /*************************************************
463  *            Handle escapes                      *  *            Handle escapes                      *
464  *************************************************/  *************************************************/
465    
# Line 399  Arguments: Line 480  Arguments:
480    
481  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
482                   negative => a special escape sequence                   negative => a special escape sequence
483                   on error, errorptr is set                   on error, errorcodeptr is set
484  */  */
485    
486  static int  static int
# Line 417  ptr--;                            /* Set Line 498  ptr--;                            /* Set
498    
499  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
500    
501  /* 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
502  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.
503  Otherwise further processing may be required. */  Otherwise further processing may be required. */
504    
505  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
506  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */
507  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
508    
509  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
510  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
511  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
512  #endif  #endif
513    
# Line 450  else Line 531  else
531      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
532      break;      break;
533    
534      /* \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:
535      is an absolute backreference. If negative, it is a relative backreference.  
536      This is a Perl 5.10 feature. */      (1) A number, either plain or braced. If positive, it is an absolute
537        backreference. If negative, it is a relative backreference. This is a Perl
538        5.10 feature.
539    
540        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
541        is part of Perl's movement towards a unified syntax for back references. As
542        this is synonymous with \k{name}, we fudge it up by pretending it really
543        was \k.
544    
545        (3) For Oniguruma compatibility we also support \g followed by a name or a
546        number either in angle brackets or in single quotes. However, these are
547        (possibly recursive) subroutine calls, _not_ backreferences. Just return
548        the -ESC_g code (cf \k). */
549    
550      case 'g':      case 'g':
551        if (ptr[1] == '<' || ptr[1] == '\'')
552          {
553          c = -ESC_g;
554          break;
555          }
556    
557        /* Handle the Perl-compatible cases */
558    
559      if (ptr[1] == '{')      if (ptr[1] == '{')
560        {        {
561          const uschar *p;
562          for (p = ptr+2; *p != 0 && *p != '}'; p++)
563            if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;
564          if (*p != 0 && *p != '}')
565            {
566            c = -ESC_k;
567            break;
568            }
569        braced = TRUE;        braced = TRUE;
570        ptr++;        ptr++;
571        }        }
# Line 473  else Line 582  else
582      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
583        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - '0';
584    
585      if (c == 0 || (braced && *(++ptr) != '}'))      if (c < 0)   /* Integer overflow */
586          {
587          *errorcodeptr = ERR61;
588          break;
589          }
590    
591        if (braced && *(++ptr) != '}')
592        {        {
593        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
594        return 0;        break;
595        }        }
596    
597        if (c == 0)
598          {
599          *errorcodeptr = ERR58;
600          break;
601          }
602    
603      if (negated)      if (negated)
604        {        {
605        if (c > bracount)        if (c > bracount)
606          {          {
607          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
608          return 0;          break;
609          }          }
610        c = bracount - (c - 1);        c = bracount - (c - 1);
611        }        }
# Line 513  else Line 634  else
634        c -= '0';        c -= '0';
635        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
636          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
637          if (c < 0)    /* Integer overflow */
638            {
639            *errorcodeptr = ERR61;
640            break;
641            }
642        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
643          {          {
644          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 608  else Line 734  else
734      if (c == 0)      if (c == 0)
735        {        {
736        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
737        return 0;        break;
738        }        }
739    
740  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
# Line 621  else Line 747  else
747      break;      break;
748    
749      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
750      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
751      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
752      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
753      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
754    
755      default:      default:
756      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 684  if (c == '{') Line 810  if (c == '{')
810      *negptr = TRUE;      *negptr = TRUE;
811      ptr++;      ptr++;
812      }      }
813    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
814      {      {
815      c = *(++ptr);      c = *(++ptr);
816      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
# Line 713  top = _pcre_utt_size; Line 839  top = _pcre_utt_size;
839  while (bot < top)  while (bot < top)
840    {    {
841    i = (bot + top) >> 1;    i = (bot + top) >> 1;
842    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
843    if (c == 0)    if (c == 0)
844      {      {
845      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 887  for (; *ptr != 0; ptr++) Line 1013  for (; *ptr != 0; ptr++)
1013      {      {
1014      while (*(++ptr) != ']')      while (*(++ptr) != ']')
1015        {        {
1016          if (*ptr == 0) return -1;
1017        if (*ptr == '\\')        if (*ptr == '\\')
1018          {          {
1019          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) return -1;
# Line 914  for (; *ptr != 0; ptr++) Line 1041  for (; *ptr != 0; ptr++)
1041    /* An opening parens must now be a real metacharacter */    /* An opening parens must now be a real metacharacter */
1042    
1043    if (*ptr != '(') continue;    if (*ptr != '(') continue;
1044    if (ptr[1] != '?')    if (ptr[1] != '?' && ptr[1] != '*')
1045      {      {
1046      count++;      count++;
1047      if (name == NULL && count == lorn) return count;      if (name == NULL && count == lorn) return count;
# Line 1042  for (;;) Line 1169  for (;;)
1169    {    {
1170    int d;    int d;
1171    register int op = *cc;    register int op = *cc;
   
1172    switch (op)    switch (op)
1173      {      {
1174      case OP_CBRA:      case OP_CBRA:
# Line 1131  for (;;) Line 1257  for (;;)
1257    
1258      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1259      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1260        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1261      cc += 4;      cc += 4;
1262      break;      break;
1263    
# Line 1239  for (;;) Line 1366  for (;;)
1366      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1367      }      }
1368    
1369    /* 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
1370    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
1371    arrange to skip the extra bytes. */    two bytes of parameters. */
1372    
1373    else    else
1374      {      {
1375        switch(c)
1376          {
1377          case OP_TYPESTAR:
1378          case OP_TYPEMINSTAR:
1379          case OP_TYPEPLUS:
1380          case OP_TYPEMINPLUS:
1381          case OP_TYPEQUERY:
1382          case OP_TYPEMINQUERY:
1383          case OP_TYPEPOSSTAR:
1384          case OP_TYPEPOSPLUS:
1385          case OP_TYPEPOSQUERY:
1386          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1387          break;
1388    
1389          case OP_TYPEUPTO:
1390          case OP_TYPEMINUPTO:
1391          case OP_TYPEEXACT:
1392          case OP_TYPEPOSUPTO:
1393          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1394          break;
1395          }
1396    
1397        /* Add in the fixed length from the table */
1398    
1399      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1400    
1401      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1402      a multi-byte character. The length in the table is a minimum, so we have to
1403      arrange to skip the extra bytes. */
1404    
1405  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1406      if (utf8) switch(c)      if (utf8) switch(c)
1407        {        {
# Line 1303  for (;;) Line 1459  for (;;)
1459    
1460    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1461    
1462    /* 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
1463    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
1464    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1465    
1466    else    else
1467      {      {
1468        switch(c)
1469          {
1470          case OP_TYPESTAR:
1471          case OP_TYPEMINSTAR:
1472          case OP_TYPEPLUS:
1473          case OP_TYPEMINPLUS:
1474          case OP_TYPEQUERY:
1475          case OP_TYPEMINQUERY:
1476          case OP_TYPEPOSSTAR:
1477          case OP_TYPEPOSPLUS:
1478          case OP_TYPEPOSQUERY:
1479          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1480          break;
1481    
1482          case OP_TYPEPOSUPTO:
1483          case OP_TYPEUPTO:
1484          case OP_TYPEMINUPTO:
1485          case OP_TYPEEXACT:
1486          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1487          break;
1488          }
1489    
1490        /* Add in the fixed length from the table */
1491    
1492      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1493    
1494        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1495        by a multi-byte character. The length in the table is a minimum, so we have
1496        to arrange to skip the extra bytes. */
1497    
1498  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1499      if (utf8) switch(c)      if (utf8) switch(c)
1500        {        {
# Line 1347  for (;;) Line 1531  for (;;)
1531  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()
1532  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
1533  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1534  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
1535  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
1536    bracket whose current branch will already have been scanned.
1537    
1538  Arguments:  Arguments:
1539    code        points to start of search    code        points to start of search
# Line 1370  for (code = first_significant_code(code Line 1555  for (code = first_significant_code(code
1555    
1556    c = *code;    c = *code;
1557    
1558    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
1559      first_significant_code() with a TRUE final argument. */
1560    
1561      if (c == OP_ASSERT)
1562        {
1563        do code += GET(code, 1); while (*code == OP_ALT);
1564        c = *code;
1565        continue;
1566        }
1567    
1568      /* Groups with zero repeats can of course be empty; skip them. */
1569    
1570      if (c == OP_BRAZERO || c == OP_BRAMINZERO)
1571        {
1572        code += _pcre_OP_lengths[c];
1573        do code += GET(code, 1); while (*code == OP_ALT);
1574        c = *code;
1575        continue;
1576        }
1577    
1578      /* For other groups, scan the branches. */
1579    
1580      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1581      {      {
1582      BOOL empty_branch;      BOOL empty_branch;
1583      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1386  for (code = first_significant_code(code Line 1593  for (code = first_significant_code(code
1593        }        }
1594      while (*code == OP_ALT);      while (*code == OP_ALT);
1595      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;   /* All branches are non-empty */
1596        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];  
1597      continue;      continue;
1598      }      }
1599    
# Line 1399  for (code = first_significant_code(code Line 1601  for (code = first_significant_code(code
1601    
1602    switch (c)    switch (c)
1603      {      {
1604      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1605        cannot be represented just by a bit map. This includes negated single
1606        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1607        actual length is stored in the compiled code, so we must update "code"
1608        here. */
1609    
1610  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1611      case OP_XCLASS:      case OP_XCLASS:
1612      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1613      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1614  #endif  #endif
1615    
# Line 1465  for (code = first_significant_code(code Line 1671  for (code = first_significant_code(code
1671      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1672      return FALSE;      return FALSE;
1673    
1674        /* These are going to continue, as they may be empty, but we have to
1675        fudge the length for the \p and \P cases. */
1676    
1677        case OP_TYPESTAR:
1678        case OP_TYPEMINSTAR:
1679        case OP_TYPEPOSSTAR:
1680        case OP_TYPEQUERY:
1681        case OP_TYPEMINQUERY:
1682        case OP_TYPEPOSQUERY:
1683        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1684        break;
1685    
1686        /* Same for these */
1687    
1688        case OP_TYPEUPTO:
1689        case OP_TYPEMINUPTO:
1690        case OP_TYPEPOSUPTO:
1691        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1692        break;
1693    
1694      /* End of branch */      /* End of branch */
1695    
1696      case OP_KET:      case OP_KET:
# Line 1534  return TRUE; Line 1760  return TRUE;
1760  *************************************************/  *************************************************/
1761    
1762  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1763  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
1764  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1765  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1766    
1767    Originally, this function only recognized a sequence of letters between the
1768    terminators, but it seems that Perl recognizes any sequence of characters,
1769    though of course unknown POSIX names are subsequently rejected. Perl gives an
1770    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1771    didn't consider this to be a POSIX class. Likewise for [:1234:].
1772    
1773    The problem in trying to be exactly like Perl is in the handling of escapes. We
1774    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1775    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1776    below handles the special case of \], but does not try to do any other escape
1777    processing. This makes it different from Perl for cases such as [:l\ower:]
1778    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1779    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1780    I think.
1781    
1782  Argument:  Arguments:
1783    ptr      pointer to the initial [    ptr      pointer to the initial [
1784    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1785    
1786  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1787  */  */
1788    
1789  static BOOL  static BOOL
1790  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1791  {  {
1792  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1793  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1794  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1795    {    {
1796    *endptr = ptr;    if (*ptr == '\\' && ptr[1] == ']') ptr++; else
1797    return TRUE;      {
1798        if (*ptr == ']') return FALSE;
1799        if (*ptr == terminator && ptr[1] == ']')
1800          {
1801          *endptr = ptr;
1802          return TRUE;
1803          }
1804        }
1805    }    }
1806  return FALSE;  return FALSE;
1807  }  }
# Line 1581  Returns:     a value representing the na Line 1826  Returns:     a value representing the na
1826  static int  static int
1827  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1828  {  {
1829    const char *pn = posix_names;
1830  register int yield = 0;  register int yield = 0;
1831  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1832    {    {
1833    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1834      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1835      pn += posix_name_lengths[yield] + 1;
1836    yield++;    yield++;
1837    }    }
1838  return -1;  return -1;
# Line 1627  adjust_recurse(uschar *group, int adjust Line 1874  adjust_recurse(uschar *group, int adjust
1874    uschar *save_hwm)    uschar *save_hwm)
1875  {  {
1876  uschar *ptr = group;  uschar *ptr = group;
1877    
1878  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1879    {    {
1880    int offset;    int offset;
# Line 1888  if (next >= 0) switch(op_code) Line 2136  if (next >= 0) switch(op_code)
2136    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2137    
2138    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2139    if (item == next) return TRUE;    if (item == next) return TRUE;
2140    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2141  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1925  if (next >= 0) switch(op_code) Line 2172  if (next >= 0) switch(op_code)
2172    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2173    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2174    
2175      case OP_HSPACE:
2176      case OP_NOT_HSPACE:
2177      switch(next)
2178        {
2179        case 0x09:
2180        case 0x20:
2181        case 0xa0:
2182        case 0x1680:
2183        case 0x180e:
2184        case 0x2000:
2185        case 0x2001:
2186        case 0x2002:
2187        case 0x2003:
2188        case 0x2004:
2189        case 0x2005:
2190        case 0x2006:
2191        case 0x2007:
2192        case 0x2008:
2193        case 0x2009:
2194        case 0x200A:
2195        case 0x202f:
2196        case 0x205f:
2197        case 0x3000:
2198        return op_code != OP_HSPACE;
2199        default:
2200        return op_code == OP_HSPACE;
2201        }
2202    
2203      case OP_VSPACE:
2204      case OP_NOT_VSPACE:
2205      switch(next)
2206        {
2207        case 0x0a:
2208        case 0x0b:
2209        case 0x0c:
2210        case 0x0d:
2211        case 0x85:
2212        case 0x2028:
2213        case 0x2029:
2214        return op_code != OP_VSPACE;
2215        default:
2216        return op_code == OP_VSPACE;
2217        }
2218    
2219    default:    default:
2220    return FALSE;    return FALSE;
2221    }    }
# Line 1959  switch(op_code) Line 2250  switch(op_code)
2250      case ESC_W:      case ESC_W:
2251      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2252    
2253        case ESC_h:
2254        case ESC_H:
2255        switch(item)
2256          {
2257          case 0x09:
2258          case 0x20:
2259          case 0xa0:
2260          case 0x1680:
2261          case 0x180e:
2262          case 0x2000:
2263          case 0x2001:
2264          case 0x2002:
2265          case 0x2003:
2266          case 0x2004:
2267          case 0x2005:
2268          case 0x2006:
2269          case 0x2007:
2270          case 0x2008:
2271          case 0x2009:
2272          case 0x200A:
2273          case 0x202f:
2274          case 0x205f:
2275          case 0x3000:
2276          return -next != ESC_h;
2277          default:
2278          return -next == ESC_h;
2279          }
2280    
2281        case ESC_v:
2282        case ESC_V:
2283        switch(item)
2284          {
2285          case 0x0a:
2286          case 0x0b:
2287          case 0x0c:
2288          case 0x0d:
2289          case 0x85:
2290          case 0x2028:
2291          case 0x2029:
2292          return -next != ESC_v;
2293          default:
2294          return -next == ESC_v;
2295          }
2296    
2297      default:      default:
2298      return FALSE;      return FALSE;
2299      }      }
2300    
2301    case OP_DIGIT:    case OP_DIGIT:
2302    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2303             next == -ESC_h || next == -ESC_v;
2304    
2305    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2306    return next == -ESC_d;    return next == -ESC_d;
# Line 1973  switch(op_code) Line 2309  switch(op_code)
2309    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2310    
2311    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2312    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2313    
2314      case OP_HSPACE:
2315      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2316    
2317      case OP_NOT_HSPACE:
2318      return next == -ESC_h;
2319    
2320      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2321      case OP_VSPACE:
2322      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2323    
2324      case OP_NOT_VSPACE:
2325      return next == -ESC_v;
2326    
2327    case OP_WORDCHAR:    case OP_WORDCHAR:
2328    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2329    
2330    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2331    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2049  uschar classbits[32]; Line 2398  uschar classbits[32];
2398  BOOL class_utf8;  BOOL class_utf8;
2399  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2400  uschar *class_utf8data;  uschar *class_utf8data;
2401    uschar *class_utf8data_base;
2402  uschar utf8_char[6];  uschar utf8_char[6];
2403  #else  #else
2404  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2088  req_caseopt = ((options & PCRE_CASELESS) Line 2438  req_caseopt = ((options & PCRE_CASELESS)
2438  for (;; ptr++)  for (;; ptr++)
2439    {    {
2440    BOOL negate_class;    BOOL negate_class;
2441      BOOL should_flip_negation;
2442    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2443    BOOL is_quantifier;    BOOL is_quantifier;
2444    BOOL is_recurse;    BOOL is_recurse;
2445      BOOL reset_bracount;
2446    int class_charcount;    int class_charcount;
2447    int class_lastchar;    int class_lastchar;
2448    int newoptions;    int newoptions;
2449    int recno;    int recno;
2450      int refsign;
2451    int skipbytes;    int skipbytes;
2452    int subreqbyte;    int subreqbyte;
2453    int subfirstbyte;    int subfirstbyte;
# Line 2127  for (;; ptr++) Line 2480  for (;; ptr++)
2480      */      */
2481    
2482      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2483    
2484        /* Paranoid check for integer overflow */
2485    
2486        if (OFLOW_MAX - *lengthptr < code - last_code)
2487          {
2488          *errorcodeptr = ERR20;
2489          goto FAILED;
2490          }
2491    
2492      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2493      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));
2494    
# Line 2239  for (;; ptr++) Line 2601  for (;; ptr++)
2601      *ptrptr = ptr;      *ptrptr = ptr;
2602      if (lengthptr != NULL)      if (lengthptr != NULL)
2603        {        {
2604          if (OFLOW_MAX - *lengthptr < code - last_code)
2605            {
2606            *errorcodeptr = ERR20;
2607            goto FAILED;
2608            }
2609        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2610        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2611        }        }
# Line 2295  for (;; ptr++) Line 2662  for (;; ptr++)
2662      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. */
2663    
2664      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2665          check_posix_syntax(ptr, &tempptr, cd))          check_posix_syntax(ptr, &tempptr))
2666        {        {
2667        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;
2668        goto FAILED;        goto FAILED;
2669        }        }
2670    
2671      /* 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,
2672        if the first few characters (either before or after ^) are \Q\E or \E we
2673        skip them too. This makes for compatibility with Perl. */
2674    
2675      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2676        for (;;)
2677        {        {
       negate_class = TRUE;  
2678        c = *(++ptr);        c = *(++ptr);
2679        }        if (c == '\\')
2680      else          {
2681        {          if (ptr[1] == 'E') ptr++;
2682        negate_class = FALSE;            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2683                else break;
2684            }
2685          else if (!negate_class && c == '^')
2686            negate_class = TRUE;
2687          else break;
2688        }        }
2689    
2690        /* If a class contains a negative special such as \S, we need to flip the
2691        negation flag at the end, so that support for characters > 255 works
2692        correctly (they are all included in the class). */
2693    
2694        should_flip_negation = FALSE;
2695    
2696      /* 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
2697      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
2698      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2330  for (;; ptr++) Line 2710  for (;; ptr++)
2710  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2711      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2712      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2713        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2714  #endif  #endif
2715    
2716      /* 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 2345  for (;; ptr++) Line 2726  for (;; ptr++)
2726          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2727          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2728          }          }
2729    
2730          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2731          data and reset the pointer. This is so that very large classes that
2732          contain a zillion UTF-8 characters no longer overwrite the work space
2733          (which is on the stack). */
2734    
2735          if (lengthptr != NULL)
2736            {
2737            *lengthptr += class_utf8data - class_utf8data_base;
2738            class_utf8data = class_utf8data_base;
2739            }
2740    
2741  #endif  #endif
2742    
2743        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 2368  for (;; ptr++) Line 2761  for (;; ptr++)
2761    
2762        if (c == '[' &&        if (c == '[' &&
2763            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2764            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr))
2765          {          {
2766          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2767          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
# Line 2385  for (;; ptr++) Line 2778  for (;; ptr++)
2778          if (*ptr == '^')          if (*ptr == '^')
2779            {            {
2780            local_negate = TRUE;            local_negate = TRUE;
2781              should_flip_negation = TRUE;  /* Note negative special */
2782            ptr++;            ptr++;
2783            }            }
2784    
# Line 2451  for (;; ptr++) Line 2845  for (;; ptr++)
2845        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
2846        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
2847        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2848        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
2849        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2850    
2851        if (c == '\\')        if (c == '\\')
# Line 2459  for (;; ptr++) Line 2853  for (;; ptr++)
2853          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2854          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
2855    
2856          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
2857          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 */
2858          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 */
2859          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
# Line 2471  for (;; ptr++) Line 2865  for (;; ptr++)
2865            else inescq = TRUE;            else inescq = TRUE;
2866            continue;            continue;
2867            }            }
2868            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2869    
2870          if (c < 0)          if (c < 0)
2871            {            {
# Line 2486  for (;; ptr++) Line 2881  for (;; ptr++)
2881              continue;              continue;
2882    
2883              case ESC_D:              case ESC_D:
2884                should_flip_negation = TRUE;
2885              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2886              continue;              continue;
2887    
# Line 2494  for (;; ptr++) Line 2890  for (;; ptr++)
2890              continue;              continue;
2891    
2892              case ESC_W:              case ESC_W:
2893                should_flip_negation = TRUE;
2894              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2895              continue;              continue;
2896    
# Line 2503  for (;; ptr++) Line 2900  for (;; ptr++)
2900              continue;              continue;
2901    
2902              case ESC_S:              case ESC_S:
2903                should_flip_negation = TRUE;
2904              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2905              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2906              continue;              continue;
2907    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
2908              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
2909              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
2910              }              }
# Line 2519  for (;; ptr++) Line 2914  for (;; ptr++)
2914            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
2915                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
2916    
2917              /* We need to deal with \H, \h, \V, and \v in both phases because
2918              they use extra memory. */
2919    
2920              if (-c == ESC_h)
2921                {
2922                SETBIT(classbits, 0x09); /* VT */
2923                SETBIT(classbits, 0x20); /* SPACE */
2924                SETBIT(classbits, 0xa0); /* NSBP */
2925    #ifdef SUPPORT_UTF8
2926                if (utf8)
2927                  {
2928                  class_utf8 = TRUE;
2929                  *class_utf8data++ = XCL_SINGLE;
2930                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
2931                  *class_utf8data++ = XCL_SINGLE;
2932                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
2933                  *class_utf8data++ = XCL_RANGE;
2934                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
2935                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
2936                  *class_utf8data++ = XCL_SINGLE;
2937                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
2938                  *class_utf8data++ = XCL_SINGLE;
2939                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
2940                  *class_utf8data++ = XCL_SINGLE;
2941                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
2942                  }
2943    #endif
2944                continue;
2945                }
2946    
2947              if (-c == ESC_H)
2948                {
2949                for (c = 0; c < 32; c++)
2950                  {
2951                  int x = 0xff;
2952                  switch (c)
2953                    {
2954                    case 0x09/8: x ^= 1 << (0x09%8); break;
2955                    case 0x20/8: x ^= 1 << (0x20%8); break;
2956                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
2957                    default: break;
2958                    }
2959                  classbits[c] |= x;
2960                  }
2961    
2962    #ifdef SUPPORT_UTF8
2963                if (utf8)
2964                  {
2965                  class_utf8 = TRUE;
2966                  *class_utf8data++ = XCL_RANGE;
2967                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
2968                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
2969                  *class_utf8data++ = XCL_RANGE;
2970                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
2971                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
2972                  *class_utf8data++ = XCL_RANGE;
2973                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
2974                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
2975                  *class_utf8data++ = XCL_RANGE;
2976                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
2977                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
2978                  *class_utf8data++ = XCL_RANGE;
2979                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
2980                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
2981                  *class_utf8data++ = XCL_RANGE;
2982                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
2983                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
2984                  *class_utf8data++ = XCL_RANGE;
2985                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
2986                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
2987                  }
2988    #endif
2989                continue;
2990                }
2991    
2992              if (-c == ESC_v)
2993                {
2994                SETBIT(classbits, 0x0a); /* LF */
2995                SETBIT(classbits, 0x0b); /* VT */
2996                SETBIT(classbits, 0x0c); /* FF */
2997                SETBIT(classbits, 0x0d); /* CR */
2998                SETBIT(classbits, 0x85); /* NEL */
2999    #ifdef SUPPORT_UTF8
3000                if (utf8)
3001                  {
3002                  class_utf8 = TRUE;
3003                  *class_utf8data++ = XCL_RANGE;
3004                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3005                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3006                  }
3007    #endif
3008                continue;
3009                }
3010    
3011              if (-c == ESC_V)
3012                {
3013                for (c = 0; c < 32; c++)
3014                  {
3015                  int x = 0xff;
3016                  switch (c)
3017                    {
3018                    case 0x0a/8: x ^= 1 << (0x0a%8);
3019                                 x ^= 1 << (0x0b%8);
3020                                 x ^= 1 << (0x0c%8);
3021                                 x ^= 1 << (0x0d%8);
3022                                 break;
3023                    case 0x85/8: x ^= 1 << (0x85%8); break;
3024                    default: break;
3025                    }
3026                  classbits[c] |= x;
3027                  }
3028    
3029    #ifdef SUPPORT_UTF8
3030                if (utf8)
3031                  {
3032                  class_utf8 = TRUE;
3033                  *class_utf8data++ = XCL_RANGE;
3034                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3035                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3036                  *class_utf8data++ = XCL_RANGE;
3037                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3038                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3039                  }
3040    #endif
3041                continue;
3042                }
3043    
3044            /* We need to deal with \P and \p in both phases. */            /* We need to deal with \P and \p in both phases. */
3045    
3046  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 2570  for (;; ptr++) Line 3092  for (;; ptr++)
3092    
3093        oldptr = ptr;        oldptr = ptr;
3094    
3095          /* Remember \r or \n */
3096    
3097          if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3098    
3099          /* Check for range */
3100    
3101        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == '-')
3102          {          {
3103          int d;          int d;
# Line 2611  for (;; ptr++) Line 3139  for (;; ptr++)
3139            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3140            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3141    
3142            /* \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
3143            special means the '-' was literal */            special means the '-' was literal */
3144    
3145            if (d < 0)            if (d < 0)
# Line 2637  for (;; ptr++) Line 3165  for (;; ptr++)
3165    
3166          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3167    
3168            /* Remember \r or \n */
3169    
3170            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3171    
3172          /* 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
3173          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3174          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2659  for (;; ptr++) Line 3191  for (;; ptr++)
3191              unsigned int origd = d;              unsigned int origd = d;
3192              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3193                {                {
3194                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3195                      ocd <= (unsigned int)d)
3196                    continue;                          /* Skip embedded ranges */
3197    
3198                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3199                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3200                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3201                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3202                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3203                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3204                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3205                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3206                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3207                  d = ocd;                  d = ocd;
3208                  continue;                  continue;
# Line 2786  for (;; ptr++) Line 3322  for (;; ptr++)
3322        goto FAILED;        goto FAILED;
3323        }        }
3324    
3325    
3326    /* This code has been disabled because it would mean that \s counts as
3327    an explicit \r or \n reference, and that's not really what is wanted. Now
3328    we set the flag only if there is a literal "\r" or "\n" in the class. */
3329    
3330    #if 0
3331        /* Remember whether \r or \n are in this class */
3332    
3333        if (negate_class)
3334          {
3335          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3336          }
3337        else
3338          {
3339          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3340          }
3341    #endif
3342    
3343    
3344      /* 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
3345      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
3346      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
3347      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3348      single-bytes only. This is an historical hangover. Maybe one day we can  
3349      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3350        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3351        operate on single-bytes only. This is an historical hangover. Maybe one day
3352        we can tidy these opcodes to handle multi-byte characters.
3353    
3354      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
3355      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 2801  for (;; ptr++) Line 3359  for (;; ptr++)
3359      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3360    
3361  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3362      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3363            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3364  #else  #else
3365      if (class_charcount == 1)      if (class_charcount == 1)
3366  #endif  #endif
# Line 2847  for (;; ptr++) Line 3403  for (;; ptr++)
3403      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3404    
3405      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3406      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3407      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
3408        the class, so any that were explicitly given as well can be ignored. If
3409        (when there are explicit characters > 255 that must be listed) there are no
3410        characters < 256, we can omit the bitmap in the actual compiled code. */
3411    
3412  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3413      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3414        {        {
3415        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3416        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2877  for (;; ptr++) Line 3436  for (;; ptr++)
3436        }        }
3437  #endif  #endif
3438    
3439      /* 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
3440      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
3441      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
3442      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3443    
3444        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3445      if (negate_class)      if (negate_class)
3446        {        {
       *code++ = OP_NCLASS;  
3447        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3448          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3449        }        }
3450      else      else
3451        {        {
       *code++ = OP_CLASS;  
3452        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3453        }        }
3454      code += 32;      code += 32;
# Line 3076  for (;; ptr++) Line 3634  for (;; ptr++)
3634        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3635        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3636    
3637        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3638    
3639        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3640    
# Line 3226  for (;; ptr++) Line 3784  for (;; ptr++)
3784        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3785        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3786    
3787        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3788    
3789        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3790          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3262  for (;; ptr++) Line 3820  for (;; ptr++)
3820          goto FAILED;          goto FAILED;
3821          }          }
3822    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
3823        /* 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
3824        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
3825        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 3358  for (;; ptr++) Line 3908  for (;; ptr++)
3908          if (repeat_min > 1)          if (repeat_min > 1)
3909            {            {
3910            /* 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
3911            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
3912              potential integer overflow. */
3913    
3914            if (lengthptr != NULL)            if (lengthptr != NULL)
3915              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
3916                int delta = (repeat_min - 1)*length_prevgroup;
3917                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3918                                                                (double)INT_MAX ||
3919                    OFLOW_MAX - *lengthptr < delta)
3920                  {
3921                  *errorcodeptr = ERR20;
3922                  goto FAILED;
3923                  }
3924                *lengthptr += delta;
3925                }
3926    
3927            /* This is compiling for real */            /* This is compiling for real */
3928    
# Line 3399  for (;; ptr++) Line 3960  for (;; ptr++)
3960          /* 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
3961          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
3962          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
3963          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3964            paranoid checks to avoid integer overflow. */
3965    
3966          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
3967            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
3968              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3969                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3970              if ((double)repeat_max *
3971                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3972                      > (double)INT_MAX ||
3973                  OFLOW_MAX - *lengthptr < delta)
3974                {
3975                *errorcodeptr = ERR20;
3976                goto FAILED;
3977                }
3978              *lengthptr += delta;
3979              }
3980    
3981          /* This is compiling for real */          /* This is compiling for real */
3982    
# Line 3511  for (;; ptr++) Line 4084  for (;; ptr++)
4084        int len;        int len;
4085        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4086            *tempcode == OP_NOTEXACT)            *tempcode == OP_NOTEXACT)
4087          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode] +
4088              ((*tempcode == OP_TYPEEXACT &&
4089                 (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4090        len = code - tempcode;        len = code - tempcode;
4091        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4092          {          {
# Line 3555  for (;; ptr++) Line 4130  for (;; ptr++)
4130      /* ===================================================================*/      /* ===================================================================*/
4131      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4132      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4133      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. */  
4134    
4135      case '(':      case '(':
4136      newoptions = options;      newoptions = options;
4137      skipbytes = 0;      skipbytes = 0;
4138      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4139      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4140        reset_bracount = FALSE;
4141    
4142        /* First deal with various "verbs" that can be introduced by '*'. */
4143    
4144      if (*(++ptr) == '?')      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4145          {
4146          int i, namelen;
4147          const char *vn = verbnames;
4148          const uschar *name = ++ptr;
4149          previous = NULL;
4150          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4151          if (*ptr == ':')
4152            {
4153            *errorcodeptr = ERR59;   /* Not supported */
4154            goto FAILED;
4155            }
4156          if (*ptr != ')')
4157            {
4158            *errorcodeptr = ERR60;
4159            goto FAILED;
4160            }
4161          namelen = ptr - name;
4162          for (i = 0; i < verbcount; i++)
4163            {
4164            if (namelen == verbs[i].len &&
4165                strncmp((char *)name, vn, namelen) == 0)
4166              {
4167              *code = verbs[i].op;
4168              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4169              break;
4170              }
4171            vn += verbs[i].len + 1;
4172            }
4173          if (i < verbcount) continue;
4174          *errorcodeptr = ERR60;
4175          goto FAILED;
4176          }
4177    
4178        /* Deal with the extended parentheses; all are introduced by '?', and the
4179        appearance of any of them means that this is not a capturing group. */
4180    
4181        else if (*ptr == '?')
4182        {        {
4183        int i, set, unset, namelen;        int i, set, unset, namelen;
4184        int *optset;        int *optset;
# Line 3586  for (;; ptr++) Line 4199  for (;; ptr++)
4199    
4200    
4201          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4202            case '|':                 /* Reset capture count for each branch */
4203            reset_bracount = TRUE;
4204            /* Fall through */
4205    
4206            /* ------------------------------------------------------------ */
4207          case ':':                 /* Non-capturing bracket */          case ':':                 /* Non-capturing bracket */
4208          bravalue = OP_BRA;          bravalue = OP_BRA;
4209          ptr++;          ptr++;
# Line 3621  for (;; ptr++) Line 4239  for (;; ptr++)
4239    
4240          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
4241          skipbytes = 3;          skipbytes = 3;
4242            refsign = -1;
4243    
4244          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4245    
# Line 3644  for (;; ptr++) Line 4263  for (;; ptr++)
4263            terminator = '\'';            terminator = '\'';
4264            ptr++;            ptr++;
4265            }            }
4266          else terminator = 0;          else
4267              {
4268              terminator = 0;
4269              if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);
4270              }
4271    
4272          /* 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 */
4273    
# Line 3680  for (;; ptr++) Line 4303  for (;; ptr++)
4303          if (lengthptr != NULL) break;          if (lengthptr != NULL) break;
4304    
4305          /* 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
4306          reference. */          reference. If the string started with "+" or "-" we require the rest to
4307            be digits, in which case recno will be set. */
4308    
4309            if (refsign > 0)
4310              {
4311              if (recno <= 0)
4312                {
4313                *errorcodeptr = ERR58;
4314                goto FAILED;
4315                }
4316              recno = (refsign == '-')?
4317                cd->bracount - recno + 1 : recno +cd->bracount;
4318              if (recno <= 0 || recno > cd->final_bracount)
4319                {
4320                *errorcodeptr = ERR15;
4321                goto FAILED;
4322                }
4323              PUT2(code, 2+LINK_SIZE, recno);
4324              break;
4325              }
4326    
4327            /* Otherwise (did not start with "+" or "-"), start by looking for the
4328            name. */
4329    
4330          slot = cd->name_table;          slot = cd->name_table;
4331          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 3746  for (;; ptr++) Line 4391  for (;; ptr++)
4391            skipbytes = 1;            skipbytes = 1;
4392            }            }
4393    
4394          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4395            in the second pass here, so final_bracount is set. */
4396    
4397          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4398            {            {
4399            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4400            }            }
# Line 3772  for (;; ptr++) Line 4418  for (;; ptr++)
4418    
4419          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4420          case '!':                 /* Negative lookahead */          case '!':                 /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4421          ptr++;          ptr++;
4422            if (*ptr == ')')          /* Optimize (?!) */
4423              {
4424              *code++ = OP_FAIL;
4425              previous = NULL;
4426              continue;
4427              }
4428            bravalue = OP_ASSERT_NOT;
4429          break;          break;
4430    
4431    
# Line 3936  for (;; ptr++) Line 4588  for (;; ptr++)
4588    
4589          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4590          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4591          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4592            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4593            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4594    
4595          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4596          name = ++ptr;          name = ++ptr;
# Line 3948  for (;; ptr++) Line 4602  for (;; ptr++)
4602    
4603          if (lengthptr != NULL)          if (lengthptr != NULL)
4604            {            {
4605              if (namelen == 0)
4606                {
4607                *errorcodeptr = ERR62;
4608                goto FAILED;
4609                }
4610            if (*ptr != terminator)            if (*ptr != terminator)
4611              {              {
4612              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 3961  for (;; ptr++) Line 4620  for (;; ptr++)
4620            recno = 0;            recno = 0;
4621            }            }
4622    
4623          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4624            first, and then check that we have reached the end of the name in the
4625            table. That way, if the name that is longer than any in the table,
4626            the comparison will fail without reading beyond the table entry. */
4627    
4628          else          else
4629            {            {
4630            slot = cd->name_table;            slot = cd->name_table;
4631            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4632              {              {
4633              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4634                    slot[2+namelen] == 0)
4635                  break;
4636              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4637              }              }
4638    
# Line 3999  for (;; ptr++) Line 4663  for (;; ptr++)
4663    
4664    
4665          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4666            case '-': case '+':
4667          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */
4668          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case '5': case '6': case '7': case '8': case '9':   /* subroutine */
4669            {            {
4670            const uschar *called;            const uschar *called;
4671              terminator = ')';
4672    
4673              /* Come here from the \g<...> and \g'...' code (Oniguruma
4674              compatibility). However, the syntax has been checked to ensure that
4675              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
4676              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
4677              ever be taken. */
4678    
4679              HANDLE_NUMERICAL_RECURSION:
4680    
4681              if ((refsign = *ptr) == '+')
4682                {
4683                ptr++;
4684                if ((digitab[*ptr] & ctype_digit) == 0)
4685                  {
4686                  *errorcodeptr = ERR63;
4687                  goto FAILED;
4688                  }
4689                }
4690              else if (refsign == '-')
4691                {
4692                if ((digitab[ptr[1]] & ctype_digit) == 0)
4693                  goto OTHER_CHAR_AFTER_QUERY;
4694                ptr++;
4695                }
4696    
4697            recno = 0;            recno = 0;
4698            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4699              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - '0';
4700            if (*ptr != ')')  
4701              if (*ptr != terminator)
4702              {              {
4703              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
4704              goto FAILED;              goto FAILED;
4705              }              }
4706    
4707              if (refsign == '-')
4708                {
4709                if (recno == 0)
4710                  {
4711                  *errorcodeptr = ERR58;
4712                  goto FAILED;
4713                  }
4714                recno = cd->bracount - recno + 1;
4715                if (recno <= 0)
4716                  {
4717                  *errorcodeptr = ERR15;
4718                  goto FAILED;
4719                  }
4720                }
4721              else if (refsign == '+')
4722                {
4723                if (recno == 0)
4724                  {
4725                  *errorcodeptr = ERR58;
4726                  goto FAILED;
4727                  }
4728                recno += cd->bracount;
4729                }
4730    
4731            /* Come here from code above that handles a named recursion */            /* Come here from code above that handles a named recursion */
4732    
4733            HANDLE_RECURSION:            HANDLE_RECURSION:
# Line 4084  for (;; ptr++) Line 4800  for (;; ptr++)
4800    
4801          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4802          default:              /* Other characters: check option setting */          default:              /* Other characters: check option setting */
4803            OTHER_CHAR_AFTER_QUERY:
4804          set = unset = 0;          set = unset = 0;
4805          optset = &set;          optset = &set;
4806    
# Line 4095  for (;; ptr++) Line 4812  for (;; ptr++)
4812    
4813              case 'J':    /* Record that it changed in the external options */              case 'J':    /* Record that it changed in the external options */
4814              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
4815              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
4816              break;              break;
4817    
4818              case 'i': *optset |= PCRE_CASELESS; break;              case 'i': *optset |= PCRE_CASELESS; break;
# Line 4218  for (;; ptr++) Line 4935  for (;; ptr++)
4935           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
4936           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
4937            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
4938             reset_bracount,               /* True if (?| group */
4939           skipbytes,                    /* Skip over bracket number */           skipbytes,                    /* Skip over bracket number */
4940           &subfirstbyte,                /* For possible first char */           &subfirstbyte,                /* For possible first char */
4941           &subreqbyte,                  /* For possible last char */           &subreqbyte,                  /* For possible last char */
# Line 4234  for (;; ptr++) Line 4952  for (;; ptr++)
4952      is on the bracket. */      is on the bracket. */
4953    
4954      /* 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
4955      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
4956        in the real compile phase, not in the pre-pass, where the whole group may
4957        not be available. */
4958    
4959      if (bravalue == OP_COND)      if (bravalue == OP_COND && lengthptr == NULL)
4960        {        {
4961        uschar *tc = code;        uschar *tc = code;
4962        int condcount = 0;        int condcount = 0;
# Line 4283  for (;; ptr++) Line 5003  for (;; ptr++)
5003        goto FAILED;        goto FAILED;
5004        }        }
5005    
5006      /* 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,
5007      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
5008      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
5009      a quantifier. */      duplicated by a quantifier.*/
5010    
5011      if (lengthptr != NULL)      if (lengthptr != NULL)
5012        {        {
5013          if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
5014            {
5015            *errorcodeptr = ERR20;
5016            goto FAILED;
5017            }
5018        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5019        code++;        *code++ = OP_BRA;
5020        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5021        *code++ = OP_KET;        *code++ = OP_KET;
5022        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5023          break;    /* No need to waste time with special character handling */
5024        }        }
5025    
5026      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
5027    
5028      else code = tempcode;      code = tempcode;
5029    
5030      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
5031      relevant. */      relevant. */
# Line 4368  for (;; ptr++) Line 5094  for (;; ptr++)
5094      back references and those types that consume a character may be repeated.      back references and those types that consume a character may be repeated.
5095      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
5096      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5097    
5098      case '\\':      case '\\':
5099      tempptr = ptr;      tempptr = ptr;
5100      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
# Line 4395  for (;; ptr++) Line 5121  for (;; ptr++)
5121    
5122        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5123        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5124    
5125          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5126          is a subroutine call by number (Oniguruma syntax). In fact, the value
5127          -ESC_g is returned only for these cases. So we don't need to check for <
5128          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5129          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5130          that is a synonym). */
5131    
5132          if (-c == ESC_g)
5133            {
5134            const uschar *p;
5135            terminator = (*(++ptr) == '<')? '>' : '\'';
5136    
5137            /* These two statements stop the compiler for warning about possibly
5138            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5139            fact, because we actually check for a number below, the paths that
5140            would actually be in error are never taken. */
5141    
5142            skipbytes = 0;
5143            reset_bracount = FALSE;
5144    
5145            /* Test for a name */
5146    
5147            if (ptr[1] != '+' && ptr[1] != '-')
5148              {
5149              BOOL isnumber = TRUE;
5150              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5151                {
5152                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5153                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5154                }
5155              if (*p != terminator)
5156                {
5157                *errorcodeptr = ERR57;
5158                break;
5159                }
5160              if (isnumber)
5161                {
5162                ptr++;
5163                goto HANDLE_NUMERICAL_RECURSION;
5164                }
5165              is_recurse = TRUE;
5166              goto NAMED_REF_OR_RECURSE;
5167              }
5168    
5169            /* Test a signed number in angle brackets or quotes. */
5170    
5171            p = ptr + 2;
5172            while ((digitab[*p] & ctype_digit) != 0) p++;
5173            if (*p != terminator)
5174              {
5175              *errorcodeptr = ERR57;
5176              break;
5177              }
5178            ptr++;
5179            goto HANDLE_NUMERICAL_RECURSION;
5180            }
5181    
5182        /* \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).
5183          We also support \k{name} (.NET syntax) */
5184    
5185        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\''))        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))
5186          {          {
5187          is_recurse = FALSE;          is_recurse = FALSE;
5188          terminator = (*(++ptr) == '<')? '>' : '\'';          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';
5189          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5190          }          }
5191    
# Line 4502  for (;; ptr++) Line 5286  for (;; ptr++)
5286      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5287      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5288    
5289        /* Remember if \r or \n were seen */
5290    
5291        if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')
5292          cd->external_flags |= PCRE_HASCRORLF;
5293    
5294      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5295      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.
5296      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 4567  This function is used during the pre-com Line 5356  This function is used during the pre-com
5356  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
5357  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
5358    
5359  Argument:  Arguments:
5360    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
5361    oldims         previous settings of ims option bits    oldims         previous settings of ims option bits
5362    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
5363    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
5364    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
5365    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
5366      reset_bracount TRUE to reset the count for each branch
5367    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
5368    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
5369    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 4587  Returns:         TRUE on success Line 5377  Returns:         TRUE on success
5377    
5378  static BOOL  static BOOL
5379  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,
5380    int *errorcodeptr, BOOL lookbehind, int skipbytes, int *firstbyteptr,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
5381    int *reqbyteptr, branch_chain *bcptr, compile_data *cd, int *lengthptr)    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
5382      int *lengthptr)
5383  {  {
5384  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
5385  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 4598  uschar *reverse_count = NULL; Line 5389  uschar *reverse_count = NULL;
5389  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5390  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5391  int length;  int length;
5392    int orig_bracount;
5393    int max_bracount;
5394  branch_chain bc;  branch_chain bc;
5395    
5396  bc.outer = bcptr;  bc.outer = bcptr;
# Line 4626  code += 1 + LINK_SIZE + skipbytes; Line 5419  code += 1 + LINK_SIZE + skipbytes;
5419    
5420  /* Loop for each alternative branch */  /* Loop for each alternative branch */
5421    
5422    orig_bracount = max_bracount = cd->bracount;
5423  for (;;)  for (;;)
5424    {    {
5425      /* For a (?| group, reset the capturing bracket count so that each branch
5426      uses the same numbers. */
5427    
5428      if (reset_bracount) cd->bracount = orig_bracount;
5429    
5430    /* Handle a change of ims options at the start of the branch */    /* Handle a change of ims options at the start of the branch */
5431    
5432    if ((options & PCRE_IMS) != oldims)    if ((options & PCRE_IMS) != oldims)
# Line 4657  for (;;) Line 5456  for (;;)
5456      return FALSE;      return FALSE;
5457      }      }
5458    
5459      /* Keep the highest bracket count in case (?| was used and some branch
5460      has fewer than the rest. */
5461    
5462      if (cd->bracount > max_bracount) max_bracount = cd->bracount;
5463    
5464    /* 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. */
5465    
5466    if (lengthptr == NULL)    if (lengthptr == NULL)
# Line 4720  for (;;) Line 5524  for (;;)
5524        }        }
5525      }      }
5526    
5527    /* Reached end of expression, either ')' or end of pattern. Go back through    /* Reached end of expression, either ')' or end of pattern. In the real
5528    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
5529    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
5530    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
5531    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
5532    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,
5533    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.
5534    at the terminating char. */    Return leaving the pointer at the terminating char. */
5535    
5536    if (*ptr != '|')    if (*ptr != '|')
5537      {      {
5538      int branch_length = code - last_branch;      if (lengthptr == NULL)
     do  
5539        {        {
5540        int prev_length = GET(last_branch, 1);        int branch_length = code - last_branch;
5541        PUT(last_branch, 1, branch_length);        do
5542        branch_length = prev_length;          {
5543        last_branch -= branch_length;          int prev_length = GET(last_branch, 1);
5544            PUT(last_branch, 1, branch_length);
5545            branch_length = prev_length;
5546            last_branch -= branch_length;
5547            }
5548          while (branch_length > 0);
5549        }        }
     while (branch_length > 0);  
5550    
5551      /* Fill in the ket */      /* Fill in the ket */
5552    
# Line 4756  for (;;) Line 5563  for (;;)
5563        length += 2;        length += 2;
5564        }        }
5565    
5566        /* Retain the highest bracket number, in case resetting was used. */
5567    
5568        cd->bracount = max_bracount;
5569    
5570      /* Set values to pass back */      /* Set values to pass back */
5571    
5572      *codeptr = code;      *codeptr = code;
5573      *ptrptr = ptr;      *ptrptr = ptr;
5574      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
5575      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
5576      if (lengthptr != NULL) *lengthptr += length;      if (lengthptr != NULL)
5577          {
5578          if (OFLOW_MAX - *lengthptr < length)
5579            {
5580            *errorcodeptr = ERR20;
5581            return FALSE;
5582            }
5583          *lengthptr += length;
5584          }
5585      return TRUE;      return TRUE;
5586      }      }
5587    
5588    /* 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
5589      pointer back to where it was for the start of the first branch. (That is,
5590      pretend that each branch is the only one.)
5591    
5592      In the real compile phase, insert an ALT node. Its length field points back
5593    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
5594    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
5595    zero offset until it is closed, making it possible to detect recursion. */    zero offset until it is closed, making it possible to detect recursion. */
5596    
5597    *code = OP_ALT;    if (lengthptr != NULL)
5598    PUT(code, 1, code - last_branch);      {
5599    bc.current = last_branch = code;      code = *codeptr + 1 + LINK_SIZE + skipbytes;
5600    code += 1 + LINK_SIZE;      length += 1 + LINK_SIZE;
5601        }
5602      else
5603        {
5604        *code = OP_ALT;
5605        PUT(code, 1, code - last_branch);
5606        bc.current = last_branch = code;
5607        code += 1 + LINK_SIZE;
5608        }
5609    
5610    ptr++;    ptr++;
   length += 1 + LINK_SIZE;  
5611    }    }
5612  /* Control never reaches here */  /* Control never reaches here */
5613  }  }
# Line 5059  real_pcre *re; Line 5890  real_pcre *re;
5890  int length = 1;  /* For final END opcode */  int length = 1;  /* For final END opcode */
5891  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
5892  int errorcode = 0;  int errorcode = 0;
5893    int skipatstart = 0;
5894  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5895  BOOL utf8;  BOOL utf8;
5896  #endif  #endif
# Line 5077  to fill in forward references to subpatt Line 5909  to fill in forward references to subpatt
5909    
5910  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
5911    
   
5912  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
5913    
5914  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;
# Line 5137  cd->fcc = tables + fcc_offset; Line 5968  cd->fcc = tables + fcc_offset;
5968  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
5969  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
5970    
5971    /* Check for global one-time settings at the start of the pattern, and remember
5972    the offset for later. */
5973    
5974    while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')
5975      {
5976      int newnl = 0;
5977      int newbsr = 0;
5978    
5979      if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)
5980        { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
5981      else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)
5982        { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
5983      else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)
5984        { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
5985      else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)
5986        { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
5987      else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)
5988        { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
5989    
5990      else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)
5991        { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
5992      else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)
5993        { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
5994    
5995      if (newnl != 0)
5996        options = (options & ~PCRE_NEWLINE_BITS) | newnl;
5997      else if (newbsr != 0)
5998        options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
5999      else break;
6000      }
6001    
6002    /* Check validity of \R options. */
6003    
6004    switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
6005      {
6006      case 0:
6007      case PCRE_BSR_ANYCRLF:
6008      case PCRE_BSR_UNICODE:
6009      break;
6010      default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
6011      }
6012    
6013  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
6014  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
6015    "anycrlf". */
6016    
6017  switch (options & (PCRE_NEWLINE_CRLF | PCRE_NEWLINE_ANY))  switch (options & PCRE_NEWLINE_BITS)
6018    {    {
6019    case 0: newline = NEWLINE; break;   /* Compile-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6020    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = '\r'; break;
6021    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = '\n'; break;
6022    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6023         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;
6024    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6025      case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6026    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
6027    }    }
6028    
6029  if (newline < 0)  if (newline == -2)
6030      {
6031      cd->nltype = NLTYPE_ANYCRLF;
6032      }
6033    else if (newline < 0)
6034    {    {
6035    cd->nltype = NLTYPE_ANY;    cd->nltype = NLTYPE_ANY;
6036    }    }
# Line 5190  to compile parts of the pattern into; th Line 6069  to compile parts of the pattern into; th
6069  no longer needed, so hopefully this workspace will never overflow, though there  no longer needed, so hopefully this workspace will never overflow, though there
6070  is a test for its doing so. */  is a test for its doing so. */
6071    
6072  cd->bracount = 0;  cd->bracount = cd->final_bracount = 0;
6073  cd->names_found = 0;  cd->names_found = 0;
6074  cd->name_entry_size = 0;  cd->name_entry_size = 0;
6075  cd->name_table = NULL;  cd->name_table = NULL;
# Line 5200  cd->hwm = cworkspace; Line 6079  cd->hwm = cworkspace;
6079  cd->start_pattern = (const uschar *)pattern;  cd->start_pattern = (const uschar *)pattern;
6080  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));
6081  cd->req_varyopt = 0;  cd->req_varyopt = 0;
 cd->nopartial = FALSE;  
6082  cd->external_options = options;  cd->external_options = options;
6083    cd->external_flags = 0;
6084    
6085  /* 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
6086  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 5209  been put into the cd block so that they Line 6088  been put into the cd block so that they
6088  found within the regex right at the beginning. Bringing initial option settings  found within the regex right at the beginning. Bringing initial option settings
6089  outside can help speed up starting point checks. */  outside can help speed up starting point checks. */
6090    
6091    ptr += skipatstart;
6092  code = cworkspace;  code = cworkspace;
6093  *code = OP_BRA;  *code = OP_BRA;
6094  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,
6095    &code, &ptr, &errorcode, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,
6096      &length);
6097  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
6098    
6099  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 5238  if (re == NULL) Line 6119  if (re == NULL)
6119    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
6120    }    }
6121    
6122  /* 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
6123  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
6124  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
6125  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
6126    pointers. */
6127    
6128  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
6129  re->size = size;  re->size = size;
6130  re->options = cd->external_options;  re->options = cd->external_options;
6131    re->flags = cd->external_flags;
6132  re->dummy1 = 0;  re->dummy1 = 0;
6133  re->first_byte = 0;  re->first_byte = 0;
6134  re->req_byte = 0;  re->req_byte = 0;
# Line 5263  field. Reset the bracket count and the n Line 6146  field. Reset the bracket count and the n
6146  field; this time it's used for remembering forward references to subpatterns.  field; this time it's used for remembering forward references to subpatterns.
6147  */  */
6148    
6149    cd->final_bracount = cd->bracount;  /* Save for checking forward references */
6150  cd->bracount = 0;  cd->bracount = 0;
6151  cd->names_found = 0;  cd->names_found = 0;
6152  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 5270  codestart = cd->name_table + re->name_en Line 6154  codestart = cd->name_table + re->name_en
6154  cd->start_code = codestart;  cd->start_code = codestart;
6155  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6156  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6157  cd->nopartial = FALSE;  cd->had_accept = FALSE;
6158    
6159  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6160  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
6161  of the function here. */  of the function here. */
6162    
6163  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern + skipatstart;
6164  code = (uschar *)codestart;  code = (uschar *)codestart;
6165  *code = OP_BRA;  *code = OP_BRA;
6166  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,
6167    &errorcode, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);
6168  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
6169  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
6170    re->flags = cd->external_flags;
6171    
6172  if (cd->nopartial) re->options |= PCRE_NOPARTIAL;  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */
6173    
6174  /* 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. */
6175    
# Line 5326  if (errorcode != 0) Line 6211  if (errorcode != 0)
6211    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
6212    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = ptr - (const uschar *)pattern;
6213    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
6214    *errorptr = error_texts[errorcode];    *errorptr = find_error_text(errorcode);
6215    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
6216    return NULL;    return NULL;
6217    }    }
# Line 5355  if ((re->options & PCRE_ANCHORED) == 0) Line 6240  if ((re->options & PCRE_ANCHORED) == 0)
6240        int ch = firstbyte & 255;        int ch = firstbyte & 255;
6241        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&
6242           cd->fcc[ch] == ch)? ch : firstbyte;           cd->fcc[ch] == ch)? ch : firstbyte;
6243        re->options |= PCRE_FIRSTSET;        re->flags |= PCRE_FIRSTSET;
6244        }        }
6245      else if (is_startline(codestart, 0, cd->backref_map))      else if (is_startline(codestart, 0, cd->backref_map))
6246        re->options |= PCRE_STARTLINE;        re->flags |= PCRE_STARTLINE;
6247      }      }
6248    }    }
6249    
# Line 5372  if (reqbyte >= 0 && Line 6257  if (reqbyte >= 0 &&
6257    int ch = reqbyte & 255;    int ch = reqbyte & 255;
6258    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&
6259      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;
6260    re->options |= PCRE_REQCHSET;    re->flags |= PCRE_REQCHSET;
6261    }    }
6262    
6263  /* 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 5383  case when building a production library. Line 6268  case when building a production library.
6268  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6269    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6270    
6271  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 " : "");  
   }  
6272    
6273  if ((re->options & PCRE_FIRSTSET) != 0)  if ((re->flags & PCRE_FIRSTSET) != 0)
6274    {    {
6275    int ch = re->first_byte & 255;    int ch = re->first_byte & 255;
6276    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?
# Line 5406  if ((re->options & PCRE_FIRSTSET) != 0) Line 6279  if ((re->options & PCRE_FIRSTSET) != 0)
6279      else printf("First char = \\x%02x%s\n", ch, caseless);      else printf("First char = \\x%02x%s\n", ch, caseless);
6280    }    }
6281    
6282  if ((re->options & PCRE_REQCHSET) != 0)  if ((re->flags & PCRE_REQCHSET) != 0)
6283    {    {
6284    int ch = re->req_byte & 255;    int ch = re->req_byte & 255;
6285    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?
# Line 5423  was compiled can be seen. */ Line 6296  was compiled can be seen. */
6296  if (code - codestart > length)  if (code - codestart > length)
6297    {    {
6298    (pcre_free)(re);    (pcre_free)(re);
6299    *errorptr = error_texts[ERR23];    *errorptr = find_error_text(ERR23);
6300    *erroroffset = ptr - (uschar *)pattern;    *erroroffset = ptr - (uschar *)pattern;
6301    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6302    return NULL;    return NULL;

Legend:
Removed from v.145  
changed lines
  Added in v.333

  ViewVC Help
Powered by ViewVC 1.1.5