/[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

code/trunk/pcre_compile.c revision 459 by ph10, Sun Oct 4 09:21:39 2009 UTC code/branches/pcre16/pcre_compile.c revision 764 by zherczeg, Wed Nov 23 17:23:20 2011 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) 1997-2009 University of Cambridge             Copyright (c) 1997-2011 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 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
# Line 118  static const short int escapes[] = { Line 124  static const short int escapes[] = {
124       -ESC_H,                  0,       -ESC_H,                  0,
125       0,                       -ESC_K,       0,                       -ESC_K,
126       0,                       0,       0,                       0,
127       0,                       0,       -ESC_N,                  0,
128       -ESC_P,                  -ESC_Q,       -ESC_P,                  -ESC_Q,
129       -ESC_R,                  -ESC_S,       -ESC_R,                  -ESC_S,
130       0,                       0,       0,                       0,
# Line 165  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 182  string is built from string macros so th Line 188  string is built from string macros so th
188  platforms. */  platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199    STRING_ACCEPT0    STRING_ACCEPT0
200    STRING_COMMIT0    STRING_COMMIT0
201    STRING_F0    STRING_F0
# Line 196  static const char verbnames[] = Line 205  static const char verbnames[] =
205    STRING_THEN;    STRING_THEN;
206    
207  static const verbitem verbs[] = {  static const verbitem verbs[] = {
208    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
209    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
210    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
212    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
213    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
214    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 220  static const char posix_names[] = Line 231  static const char posix_names[] =
231    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232    STRING_word0  STRING_xdigit;    STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
236    
237  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
# Line 250  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const pcre_uchar string_PNd[]  = {
270      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
271      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
272    static const pcre_uchar string_pNd[]  = {
273      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
274      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
275    static const pcre_uchar string_PXsp[] = {
276      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
277      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
278    static const pcre_uchar string_pXsp[] = {
279      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
280      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
281    static const pcre_uchar string_PXwd[] = {
282      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
283      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
284    static const pcre_uchar string_pXwd[] = {
285      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
286      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
287    
288    static const pcre_uchar *substitutes[] = {
289      string_PNd,           /* \D */
290      string_pNd,           /* \d */
291      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
292      string_pXsp,          /* \s */
293      string_PXwd,          /* \W */
294      string_pXwd           /* \w */
295    };
296    
297    static const pcre_uchar string_pL[] =   {
298      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
299      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
300    static const pcre_uchar string_pLl[] =  {
301      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
302      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
303    static const pcre_uchar string_pLu[] =  {
304      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
305      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    static const pcre_uchar string_pXan[] = {
307      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
308      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
309    static const pcre_uchar string_h[] =    {
310      CHAR_BACKSLASH, CHAR_h, '\0' };
311    static const pcre_uchar string_pXps[] = {
312      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
313      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
314    static const pcre_uchar string_PL[] =   {
315      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
316      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
317    static const pcre_uchar string_PLl[] =  {
318      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
319      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
320    static const pcre_uchar string_PLu[] =  {
321      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
322      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
323    static const pcre_uchar string_PXan[] = {
324      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
325      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
326    static const pcre_uchar string_H[] =    {
327      CHAR_BACKSLASH, CHAR_H, '\0' };
328    static const pcre_uchar string_PXps[] = {
329      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    
332    static const pcre_uchar *posix_substitutes[] = {
333      string_pL,            /* alpha */
334      string_pLl,           /* lower */
335      string_pLu,           /* upper */
336      string_pXan,          /* alnum */
337      NULL,                 /* ascii */
338      string_h,             /* blank */
339      NULL,                 /* cntrl */
340      string_pNd,           /* digit */
341      NULL,                 /* graph */
342      NULL,                 /* print */
343      NULL,                 /* punct */
344      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
345      string_pXwd,          /* word */
346      NULL,                 /* xdigit */
347      /* Negated cases */
348      string_PL,            /* ^alpha */
349      string_PLl,           /* ^lower */
350      string_PLu,           /* ^upper */
351      string_PXan,          /* ^alnum */
352      NULL,                 /* ^ascii */
353      string_H,             /* ^blank */
354      NULL,                 /* ^cntrl */
355      string_PNd,           /* ^digit */
356      NULL,                 /* ^graph */
357      NULL,                 /* ^print */
358      NULL,                 /* ^punct */
359      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
360      string_PXwd,          /* ^word */
361      NULL                  /* ^xdigit */
362    };
363    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
364    #endif
365    
366  #define STRING(a)  # a  #define STRING(a)  # a
367  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 262  the number of relocations needed when a Line 374  the number of relocations needed when a
374  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
375  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
376  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
377  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
378    
379    Each substring ends with \0 to insert a null character. This includes the final
380    substring, so that the whole string ends with \0\0, which can be detected when
381    counting through. */
382    
383  static const char error_texts[] =  static const char error_texts[] =
384    "no error\0"    "no error\0"
# Line 309  static const char error_texts[] = Line 425  static const char error_texts[] =
425    /* 35 */    /* 35 */
426    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
427    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
428    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
429    "number after (?C is > 255\0"    "number after (?C is > 255\0"
430    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
431    /* 40 */    /* 40 */
# Line 331  static const char error_texts[] = Line 447  static const char error_texts[] =
447    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
448    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
449    /* 55 */    /* 55 */
450    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
451    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
452    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
453    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
454    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
455    /* 60 */    /* 60 */
456    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
457    "number is too big\0"    "number is too big\0"
# Line 343  static const char error_texts[] = Line 459  static const char error_texts[] =
459    "digit expected after (?+\0"    "digit expected after (?+\0"
460    "] is an invalid data character in JavaScript compatibility mode\0"    "] is an invalid data character in JavaScript compatibility mode\0"
461    /* 65 */    /* 65 */
462    "different names for subpatterns of the same number are not allowed";    "different names for subpatterns of the same number are not allowed\0"
463      "(*MARK) must have an argument\0"
464      "this version of PCRE is not compiled with PCRE_UCP support\0"
465      "\\c must be followed by an ASCII character\0"
466      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
467      /* 70 */
468      "internal error: unknown opcode in find_fixedlength()\0"
469      ;
470    
471  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
472  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 480  static const unsigned char ebcdic_charta Line 602  static const unsigned char ebcdic_charta
602  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
603    
604  static BOOL  static BOOL
605    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
606      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
607    
608    
# Line 502  static const char * Line 624  static const char *
624  find_error_text(int n)  find_error_text(int n)
625  {  {
626  const char *s = error_texts;  const char *s = error_texts;
627  for (; n > 0; n--) while (*s++ != 0) {};  for (; n > 0; n--)
628      {
629      while (*s++ != 0) {};
630      if (*s == 0) return "Error text not found (please report)";
631      }
632  return s;  return s;
633  }  }
634    
635    
636  /*************************************************  /*************************************************
637    *            Check for counted repeat            *
638    *************************************************/
639    
640    /* This function is called when a '{' is encountered in a place where it might
641    start a quantifier. It looks ahead to see if it really is a quantifier or not.
642    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
643    where the ddds are digits.
644    
645    Arguments:
646      p         pointer to the first char after '{'
647    
648    Returns:    TRUE or FALSE
649    */
650    
651    static BOOL
652    is_counted_repeat(const pcre_uchar *p)
653    {
654    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
655    while ((digitab[*p] & ctype_digit) != 0) p++;
656    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
657    
658    if (*p++ != CHAR_COMMA) return FALSE;
659    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
660    
661    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
662    while ((digitab[*p] & ctype_digit) != 0) p++;
663    
664    return (*p == CHAR_RIGHT_CURLY_BRACKET);
665    }
666    
667    
668    
669    /*************************************************
670  *            Handle escapes                      *  *            Handle escapes                      *
671  *************************************************/  *************************************************/
672    
# Line 532  Returns:         zero or positive => a d Line 691  Returns:         zero or positive => a d
691  */  */
692    
693  static int  static int
694  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
695    int options, BOOL isclass)    int options, BOOL isclass)
696  {  {
697  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
698  const uschar *ptr = *ptrptr + 1;  const pcre_uchar *ptr = *ptrptr + 1;
699  int c, i;  int c, i;
700    
701  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
# Line 563  else if ((i = escapes[c - 0x48]) != 0) Line 722  else if ((i = escapes[c - 0x48]) != 0)
722    
723  else  else
724    {    {
725    const uschar *oldptr;    const pcre_uchar *oldptr;
726    BOOL braced, negated;    BOOL braced, negated;
727    
728    switch (c)    switch (c)
# Line 573  else Line 732  else
732    
733      case CHAR_l:      case CHAR_l:
734      case CHAR_L:      case CHAR_L:
735      case CHAR_N:      *errorcodeptr = ERR37;
736        break;
737    
738      case CHAR_u:      case CHAR_u:
739        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
740          {
741          /* In JavaScript, \u must be followed by four hexadecimal numbers.
742          Otherwise it is a lowercase u letter. */
743          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
744               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
745            {
746            c = 0;
747            for (i = 0; i < 4; ++i)
748              {
749              register int cc = *(++ptr);
750    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
751              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
752              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
753    #else           /* EBCDIC coding */
754              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
755              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
756    #endif
757              }
758            }
759          }
760        else
761          *errorcodeptr = ERR37;
762        break;
763    
764      case CHAR_U:      case CHAR_U:
765      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
766        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
767      break;      break;
768    
769      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
770        class, \g must be followed by one of a number of specific things:
771    
772      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
773      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 596  else Line 784  else
784      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
785    
786      case CHAR_g:      case CHAR_g:
787        if (isclass) break;
788      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
789        {        {
790        c = -ESC_g;        c = -ESC_g;
# Line 606  else Line 795  else
795    
796      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
797        {        {
798        const uschar *p;        const pcre_uchar *p;
799        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
800          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
801        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
# Line 724  else Line 913  else
913      treated as a data character. */      treated as a data character. */
914    
915      case CHAR_x:      case CHAR_x:
916        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
917          {
918          /* In JavaScript, \x must be followed by two hexadecimal numbers.
919          Otherwise it is a lowercase x letter. */
920          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
921            {
922            c = 0;
923            for (i = 0; i < 2; ++i)
924              {
925              register int cc = *(++ptr);
926    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
927              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
928              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
929    #else           /* EBCDIC coding */
930              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
931              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
932    #endif
933              }
934            }
935          break;
936          }
937    
938      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
939        {        {
940        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
941        int count = 0;        int count = 0;
942    
943        c = 0;        c = 0;
# Line 774  else Line 985  else
985      break;      break;
986    
987      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
988      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
989        coding is ASCII-specific, but then the whole concept of \cx is
990      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
991    
992      case CHAR_c:      case CHAR_c:
# Line 784  else Line 996  else
996        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
997        break;        break;
998        }        }
999    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1000  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1001          {
1002          *errorcodeptr = ERR68;
1003          break;
1004          }
1005      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1006      c ^= 0x40;      c ^= 0x40;
1007  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1008      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1009      c ^= 0xC0;      c ^= 0xC0;
1010  #endif  #endif
# Line 811  else Line 1027  else
1027      }      }
1028    }    }
1029    
1030    /* Perl supports \N{name} for character names, as well as plain \N for "not
1031    newline". PCRE does not support \N{name}. However, it does support
1032    quantification such as \N{2,3}. */
1033    
1034    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1035         !is_counted_repeat(ptr+2))
1036      *errorcodeptr = ERR37;
1037    
1038    /* If PCRE_UCP is set, we change the values for \d etc. */
1039    
1040    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1041      c -= (ESC_DU - ESC_D);
1042    
1043    /* Set the pointer to the final character before returning. */
1044    
1045  *ptrptr = ptr;  *ptrptr = ptr;
1046  return c;  return c;
1047  }  }
# Line 837  Returns:         type value from ucp_typ Line 1068  Returns:         type value from ucp_typ
1068  */  */
1069    
1070  static int  static int
1071  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1072  {  {
1073  int c, i, bot, top;  int c, i, bot, top;
1074  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1075  char name[32];  pcre_uchar name[32];
1076    
1077  c = *(++ptr);  c = *(++ptr);
1078  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 882  else Line 1113  else
1113  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1114    
1115  bot = 0;  bot = 0;
1116  top = _pcre_utt_size;  top = PRIV(utt_size);
1117    
1118  while (bot < top)  while (bot < top)
1119    {    {
1120    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1121    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1122    if (c == 0)    if (c == 0)
1123      {      {
1124      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1125      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1126      }      }
1127    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1128    }    }
# Line 911  return -1; Line 1142  return -1;
1142    
1143    
1144  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
 /*************************************************  
1145  *         Read repeat counts                     *  *         Read repeat counts                     *
1146  *************************************************/  *************************************************/
1147    
# Line 962  Returns:         pointer to '}' on succe Line 1160  Returns:         pointer to '}' on succe
1160                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1161  */  */
1162    
1163  static const uschar *  static const pcre_uchar *
1164  read_repeat_counts(const uschar *p, int *minp, int *maxp, int *errorcodeptr)  read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1165  {  {
1166  int min = 0;  int min = 0;
1167  int max = -1;  int max = -1;
# Line 1019  top-level call starts at the beginning o Line 1217  top-level call starts at the beginning o
1217  start at a parenthesis. It scans along a pattern's text looking for capturing  start at a parenthesis. It scans along a pattern's text looking for capturing
1218  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1219  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1220  returns when it reaches a given numbered subpattern. We know that if (?P< is  returns when it reaches a given numbered subpattern. Recursion is used to keep
1221  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1222  first pass. Recursion is used to keep track of subpatterns that reset the  
1223  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1224    that if (?< or (?' or (?P< is encountered, the name will be correctly
1225    terminated because that is checked in the first pass. There is now one call to
1226    this function in the first pass, to check for a recursive back reference by
1227    name (so that we can make the whole group atomic). In this case, we need check
1228    only up to the current position in the pattern, and that is still OK because
1229    and previous occurrences will have been checked. To make this work, the test
1230    for "end of pattern" is a check against cd->end_pattern in the main loop,
1231    instead of looking for a binary zero. This means that the special first-pass
1232    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1233    processing items within the loop are OK, because afterwards the main loop will
1234    terminate.)
1235    
1236  Arguments:  Arguments:
1237    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1030  Arguments: Line 1239  Arguments:
1239    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1240    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1241    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1242      utf8         TRUE if we are in UTF-8 mode
1243    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1244    
1245  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1246  */  */
1247    
1248  static int  static int
1249  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,
1250    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1251  {  {
1252  uschar *ptr = *ptrptr;  pcre_uchar *ptr = *ptrptr;
1253  int start_count = *count;  int start_count = *count;
1254  int hwm_count = start_count;  int hwm_count = start_count;
1255  BOOL dup_parens = FALSE;  BOOL dup_parens = FALSE;
# Line 1049  dealing with. The very first call may no Line 1259  dealing with. The very first call may no
1259    
1260  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1261    {    {
1262    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1263        ptr[2] == CHAR_VERTICAL_LINE)  
1264      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1265    
1266      /* Handle a normal, unnamed capturing parenthesis. */
1267    
1268      else if (ptr[1] != CHAR_QUESTION_MARK)
1269        {
1270        *count += 1;
1271        if (name == NULL && *count == lorn) return *count;
1272        ptr++;
1273        }
1274    
1275      /* All cases now have (? at the start. Remember when we are in a group
1276      where the parenthesis numbers are duplicated. */
1277    
1278      else if (ptr[2] == CHAR_VERTICAL_LINE)
1279      {      {
1280      ptr += 3;      ptr += 3;
1281      dup_parens = TRUE;      dup_parens = TRUE;
1282      }      }
1283    
1284    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1285    
1286    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1287      {      {
1288      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1289      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1290      }      }
1291    
1292    /* Handle a condition. If it is an assertion, just carry on so that it    /* Handle a condition. If it is an assertion, just carry on so that it
1293    is processed as normal. If not, skip to the closing parenthesis of the    is processed as normal. If not, skip to the closing parenthesis of the
1294    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1295    
1296    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1297      {      {
# Line 1079  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1303  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1303        }        }
1304      }      }
1305    
1306    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1307    
1308    else    else
1309      {      {
# Line 1092  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1316  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1316          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1317        {        {
1318        int term;        int term;
1319        const uschar *thisname;        const pcre_uchar *thisname;
1320        *count += 1;        *count += 1;
1321        if (name == NULL && *count == lorn) return *count;        if (name == NULL && *count == lorn) return *count;
1322        term = *ptr++;        term = *ptr++;
# Line 1100  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1324  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1324        thisname = ptr;        thisname = ptr;
1325        while (*ptr != term) ptr++;        while (*ptr != term) ptr++;
1326        if (name != NULL && lorn == ptr - thisname &&        if (name != NULL && lorn == ptr - thisname &&
1327            strncmp((const char *)name, (const char *)thisname, lorn) == 0)            STRNCMP_UC_UC(name, thisname, lorn) == 0)
1328          return *count;          return *count;
1329        term++;        term++;
1330        }        }
1331      }      }
1332    }    }
1333    
1334  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1335  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1336    first-pass call when this value is temporarily adjusted to stop at the current
1337    position. So DO NOT change this to a test for binary zero. */
1338    
1339  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1340    {    {
1341    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1342    
# Line 1141  for (; *ptr != 0; ptr++) Line 1367  for (; *ptr != 0; ptr++)
1367          {          {
1368          if (ptr[2] == CHAR_E)          if (ptr[2] == CHAR_E)
1369            ptr+= 2;            ptr+= 2;
1370          else if (strncmp((const char *)ptr+2,          else if (STRNCMP_UC_C8(ptr + 2,
1371                   STR_Q STR_BACKSLASH STR_E, 3) == 0)                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
1372            ptr += 4;            ptr += 4;
1373          else          else
1374            break;            break;
1375          }          }
1376        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1377          {          {
1378          negate_class = TRUE;          negate_class = TRUE;
1379          ptr++;          ptr++;
1380          }          }
1381        else break;        else break;
1382        }        }
1383    
# Line 1184  for (; *ptr != 0; ptr++) Line 1410  for (; *ptr != 0; ptr++)
1410    
1411    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1412      {      {
1413      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1414        while (*ptr != 0)
1415          {
1416          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1417          ptr++;
1418    #ifdef SUPPORT_UTF8
1419          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1420    #endif
1421          }
1422      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1423      continue;      continue;
1424      }      }
# Line 1193  for (; *ptr != 0; ptr++) Line 1427  for (; *ptr != 0; ptr++)
1427    
1428    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1429      {      {
1430      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1431      if (rc > 0) return rc;      if (rc > 0) return rc;
1432      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1433      }      }
# Line 1201  for (; *ptr != 0; ptr++) Line 1435  for (; *ptr != 0; ptr++)
1435    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1436      {      {
1437      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1438      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1439      }      }
1440    
1441    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1240  Arguments: Line 1473  Arguments:
1473    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1474    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1475    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1476      utf8         TRUE if we are in UTF-8 mode
1477    
1478  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1479  */  */
1480    
1481  static int  static int
1482  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1483      BOOL utf8)
1484  {  {
1485  uschar *ptr = (uschar *)cd->start_pattern;  pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1486  int count = 0;  int count = 0;
1487  int rc;  int rc;
1488    
# Line 1258  matching closing parens. That is why we Line 1493  matching closing parens. That is why we
1493    
1494  for (;;)  for (;;)
1495    {    {
1496    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1497    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1498    }    }
1499    
# Line 1274  return rc; Line 1509  return rc;
1509    
1510  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1511  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1512  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1513  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1514  assertions, and also the \b assertion; for others it does not.  does not.
1515    
1516  Arguments:  Arguments:
1517    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1518    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1519    
1520  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1521  */  */
1522    
1523  static const uschar*  static const pcre_uchar*
1524  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1525  {  {
1526  for (;;)  for (;;)
1527    {    {
1528    switch ((int)*code)    switch ((int)*code)
1529      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1530      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1531      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1532      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1533      if (!skipassert) return code;      if (!skipassert) return code;
1534      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1535      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1536      break;      break;
1537    
1538      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1321  for (;;) Line 1546  for (;;)
1546      case OP_RREF:      case OP_RREF:
1547      case OP_NRREF:      case OP_NRREF:
1548      case OP_DEF:      case OP_DEF:
1549      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1550      break;      break;
1551    
1552      default:      default:
# Line 1340  for (;;) Line 1565  for (;;)
1565    
1566  /* Scan a branch and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1567  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1568  In UTF8 mode, the result is in characters rather than bytes. The branch is  In UTF8 mode, the result is in characters rather than bytes. The branch is
1569  temporarily terminated with OP_END when this function is called.  temporarily terminated with OP_END when this function is called.
1570    
1571  This function is called when a backward assertion is encountered, so that if it  This function is called when a backward assertion is encountered, so that if it
1572  fails, the error message can point to the correct place in the pattern.  fails, the error message can point to the correct place in the pattern.
1573  However, we cannot do this when the assertion contains subroutine calls,  However, we cannot do this when the assertion contains subroutine calls,
1574  because they can be forward references. We solve this by remembering this case  because they can be forward references. We solve this by remembering this case
1575  and doing the check at the end; a flag specifies which mode we are running in.  and doing the check at the end; a flag specifies which mode we are running in.
1576    
1577  Arguments:  Arguments:
1578    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1579    options  the compiling options    utf8     TRUE in UTF-8 mode
1580    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1581    cd       the "compile data" structure    cd       the "compile data" structure
1582    
1583  Returns:   the fixed length,  Returns:   the fixed length,
1584               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1585               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1586               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1587                 or -4 if an unknown opcode was encountered (internal error)
1588  */  */
1589    
1590  static int  static int
1591  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(pcre_uchar *code, BOOL utf8, BOOL atend, compile_data *cd)
1592  {  {
1593  int length = -1;  int length = -1;
1594    
1595  register int branchlength = 0;  register int branchlength = 0;
1596  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1597    
1598  /* Scan along the opcodes for this branch. If we get to the end of the  /* Scan along the opcodes for this branch. If we get to the end of the
1599  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1375  branch, check the length against that of Line 1601  branch, check the length against that of
1601  for (;;)  for (;;)
1602    {    {
1603    int d;    int d;
1604    uschar *ce, *cs;    pcre_uchar *ce, *cs;
1605    register int op = *cc;    register int op = *cc;
1606    switch (op)    switch (op)
1607      {      {
1608        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1609        OP_BRA (normal non-capturing bracket) because the other variants of these
1610        opcodes are all concerned with unlimited repeated groups, which of course
1611        are not of fixed length. */
1612    
1613      case OP_CBRA:      case OP_CBRA:
1614      case OP_BRA:      case OP_BRA:
1615      case OP_ONCE:      case OP_ONCE:
1616        case OP_ONCE_NC:
1617      case OP_COND:      case OP_COND:
1618      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1619      if (d < 0) return d;      if (d < 0) return d;
1620      branchlength += d;      branchlength += d;
1621      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1622      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1623      break;      break;
1624    
1625      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1626      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1627      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1628        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1629        because they all imply an unlimited repeat. */
1630    
1631      case OP_ALT:      case OP_ALT:
1632      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1633      case OP_END:      case OP_END:
1634        case OP_ACCEPT:
1635        case OP_ASSERT_ACCEPT:
1636      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1637        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1638      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
1639      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1640      branchlength = 0;      branchlength = 0;
1641      break;      break;
1642    
1643      /* A true recursion implies not fixed length, but a subroutine call may      /* A true recursion implies not fixed length, but a subroutine call may
1644      be OK. If the subroutine is a forward reference, we can't deal with      be OK. If the subroutine is a forward reference, we can't deal with
1645      it until the end of the pattern, so return -3. */      it until the end of the pattern, so return -3. */
1646    
1647      case OP_RECURSE:      case OP_RECURSE:
1648      if (!atend) return -3;      if (!atend) return -3;
1649      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1650      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1651      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                    /* Recursion */
1652      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1653      if (d < 0) return d;      if (d < 0) return d;
1654      branchlength += d;      branchlength += d;
1655      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1656      break;      break;
1657    
1658      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1659    
# Line 1432  for (;;) Line 1666  for (;;)
1666    
1667      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1668    
1669      case OP_REVERSE:      case OP_MARK:
1670        case OP_PRUNE_ARG:
1671        case OP_SKIP_ARG:
1672        case OP_THEN_ARG:
1673        cc += cc[1] + PRIV(OP_lengths)[*cc];
1674        break;
1675    
1676        case OP_CALLOUT:
1677        case OP_CIRC:
1678        case OP_CIRCM:
1679        case OP_CLOSE:
1680        case OP_COMMIT:
1681      case OP_CREF:      case OP_CREF:
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
1682      case OP_DEF:      case OP_DEF:
1683      case OP_OPT:      case OP_DOLL:
1684      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1685      case OP_EOD:      case OP_EOD:
1686      case OP_EODN:      case OP_EODN:
1687      case OP_CIRC:      case OP_FAIL:
1688      case OP_DOLL:      case OP_NCREF:
1689        case OP_NRREF:
1690      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1691        case OP_PRUNE:
1692        case OP_REVERSE:
1693        case OP_RREF:
1694        case OP_SET_SOM:
1695        case OP_SKIP:
1696        case OP_SOD:
1697        case OP_SOM:
1698        case OP_THEN:
1699      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1700      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1701      break;      break;
1702    
1703      /* Handle literal characters */      /* Handle literal characters */
1704    
1705      case OP_CHAR:      case OP_CHAR:
1706      case OP_CHARNC:      case OP_CHARI:
1707      case OP_NOT:      case OP_NOT:
1708        case OP_NOTI:
1709      branchlength++;      branchlength++;
1710      cc += 2;      cc += 2;
1711  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1712      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += PRIV(utf8_table4)[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1713  #endif  #endif
1714      break;      break;
1715    
# Line 1468  for (;;) Line 1717  for (;;)
1717      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1718    
1719      case OP_EXACT:      case OP_EXACT:
1720        case OP_EXACTI:
1721        case OP_NOTEXACT:
1722        case OP_NOTEXACTI:
1723      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1724      cc += 4;      cc += 4;
1725  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1726      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += PRIV(utf8_table4)[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1727  #endif  #endif
1728      break;      break;
1729    
# Line 1489  for (;;) Line 1740  for (;;)
1740      cc += 2;      cc += 2;
1741      /* Fall through */      /* Fall through */
1742    
1743        case OP_HSPACE:
1744        case OP_VSPACE:
1745        case OP_NOT_HSPACE:
1746        case OP_NOT_VSPACE:
1747      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1748      case OP_DIGIT:      case OP_DIGIT:
1749      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1501  for (;;) Line 1756  for (;;)
1756      cc++;      cc++;
1757      break;      break;
1758    
1759      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1760        otherwise \C is coded as OP_ALLANY. */
1761    
1762      case OP_ANYBYTE:      case OP_ANYBYTE:
1763      return -2;      return -2;
# Line 1520  for (;;) Line 1776  for (;;)
1776    
1777      switch (*cc)      switch (*cc)
1778        {        {
1779          case OP_CRPLUS:
1780          case OP_CRMINPLUS:
1781        case OP_CRSTAR:        case OP_CRSTAR:
1782        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1783        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1540  for (;;) Line 1798  for (;;)
1798    
1799      /* Anything else is variable length */      /* Anything else is variable length */
1800    
1801      default:      case OP_ANYNL:
1802        case OP_BRAMINZERO:
1803        case OP_BRAPOS:
1804        case OP_BRAPOSZERO:
1805        case OP_BRAZERO:
1806        case OP_CBRAPOS:
1807        case OP_EXTUNI:
1808        case OP_KETRMAX:
1809        case OP_KETRMIN:
1810        case OP_KETRPOS:
1811        case OP_MINPLUS:
1812        case OP_MINPLUSI:
1813        case OP_MINQUERY:
1814        case OP_MINQUERYI:
1815        case OP_MINSTAR:
1816        case OP_MINSTARI:
1817        case OP_MINUPTO:
1818        case OP_MINUPTOI:
1819        case OP_NOTMINPLUS:
1820        case OP_NOTMINPLUSI:
1821        case OP_NOTMINQUERY:
1822        case OP_NOTMINQUERYI:
1823        case OP_NOTMINSTAR:
1824        case OP_NOTMINSTARI:
1825        case OP_NOTMINUPTO:
1826        case OP_NOTMINUPTOI:
1827        case OP_NOTPLUS:
1828        case OP_NOTPLUSI:
1829        case OP_NOTPOSPLUS:
1830        case OP_NOTPOSPLUSI:
1831        case OP_NOTPOSQUERY:
1832        case OP_NOTPOSQUERYI:
1833        case OP_NOTPOSSTAR:
1834        case OP_NOTPOSSTARI:
1835        case OP_NOTPOSUPTO:
1836        case OP_NOTPOSUPTOI:
1837        case OP_NOTQUERY:
1838        case OP_NOTQUERYI:
1839        case OP_NOTSTAR:
1840        case OP_NOTSTARI:
1841        case OP_NOTUPTO:
1842        case OP_NOTUPTOI:
1843        case OP_PLUS:
1844        case OP_PLUSI:
1845        case OP_POSPLUS:
1846        case OP_POSPLUSI:
1847        case OP_POSQUERY:
1848        case OP_POSQUERYI:
1849        case OP_POSSTAR:
1850        case OP_POSSTARI:
1851        case OP_POSUPTO:
1852        case OP_POSUPTOI:
1853        case OP_QUERY:
1854        case OP_QUERYI:
1855        case OP_REF:
1856        case OP_REFI:
1857        case OP_SBRA:
1858        case OP_SBRAPOS:
1859        case OP_SCBRA:
1860        case OP_SCBRAPOS:
1861        case OP_SCOND:
1862        case OP_SKIPZERO:
1863        case OP_STAR:
1864        case OP_STARI:
1865        case OP_TYPEMINPLUS:
1866        case OP_TYPEMINQUERY:
1867        case OP_TYPEMINSTAR:
1868        case OP_TYPEMINUPTO:
1869        case OP_TYPEPLUS:
1870        case OP_TYPEPOSPLUS:
1871        case OP_TYPEPOSQUERY:
1872        case OP_TYPEPOSSTAR:
1873        case OP_TYPEPOSUPTO:
1874        case OP_TYPEQUERY:
1875        case OP_TYPESTAR:
1876        case OP_TYPEUPTO:
1877        case OP_UPTO:
1878        case OP_UPTOI:
1879      return -1;      return -1;
1880    
1881        /* Catch unrecognized opcodes so that when new ones are added they
1882        are not forgotten, as has happened in the past. */
1883    
1884        default:
1885        return -4;
1886      }      }
1887    }    }
1888  /* Control never gets here */  /* Control never gets here */
# Line 1556  for (;;) Line 1897  for (;;)
1897    
1898  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1899  capturing bracket with the given number, or, if the number is negative, an  capturing bracket with the given number, or, if the number is negative, an
1900  instance of OP_REVERSE for a lookbehind. The function is global in the C sense  instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1901  so that it can be called from pcre_study() when finding the minimum matching  so that it can be called from pcre_study() when finding the minimum matching
1902  length.  length.
1903    
1904  Arguments:  Arguments:
# Line 1568  Arguments: Line 1909  Arguments:
1909  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
1910  */  */
1911    
1912  const uschar *  const pcre_uchar *
1913  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf8, int number)
1914  {  {
1915  for (;;)  for (;;)
1916    {    {
1917    register int c = *code;    register int c = *code;
1918    
1919    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1920    
1921    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1581  for (;;) Line 1923  for (;;)
1923    the table is zero; the actual length is stored in the compiled code. */    the table is zero; the actual length is stored in the compiled code. */
1924    
1925    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1926    
1927    /* Handle recursion */    /* Handle recursion */
1928    
1929    else if (c == OP_REVERSE)    else if (c == OP_REVERSE)
1930      {      {
1931      if (number < 0) return (uschar *)code;      if (number < 0) return (pcre_uchar *)code;
1932      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
1933      }      }
1934    
1935    /* Handle capturing bracket */    /* Handle capturing bracket */
1936    
1937    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1938               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1939      {      {
1940      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1941      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
1942      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
1943      }      }
1944    
1945    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1946    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1947    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1948      must add in its length. */
1949    
1950    else    else
1951      {      {
# Line 1625  for (;;) Line 1969  for (;;)
1969        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1970        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1971        break;        break;
1972    
1973          case OP_MARK:
1974          case OP_PRUNE_ARG:
1975          case OP_SKIP_ARG:
1976          code += code[1];
1977          break;
1978    
1979          case OP_THEN_ARG:
1980          code += code[1];
1981          break;
1982        }        }
1983    
1984      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
1985    
1986      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
1987    
1988    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1989    a multi-byte character. The length in the table is a minimum, so we have to    a multi-byte character. The length in the table is a minimum, so we have to
# Line 1639  for (;;) Line 1993  for (;;)
1993      if (utf8) switch(c)      if (utf8) switch(c)
1994        {        {
1995        case OP_CHAR:        case OP_CHAR:
1996        case OP_CHARNC:        case OP_CHARI:
1997        case OP_EXACT:        case OP_EXACT:
1998          case OP_EXACTI:
1999        case OP_UPTO:        case OP_UPTO:
2000          case OP_UPTOI:
2001        case OP_MINUPTO:        case OP_MINUPTO:
2002          case OP_MINUPTOI:
2003        case OP_POSUPTO:        case OP_POSUPTO:
2004          case OP_POSUPTOI:
2005        case OP_STAR:        case OP_STAR:
2006          case OP_STARI:
2007        case OP_MINSTAR:        case OP_MINSTAR:
2008          case OP_MINSTARI:
2009        case OP_POSSTAR:        case OP_POSSTAR:
2010          case OP_POSSTARI:
2011        case OP_PLUS:        case OP_PLUS:
2012          case OP_PLUSI:
2013        case OP_MINPLUS:        case OP_MINPLUS:
2014          case OP_MINPLUSI:
2015        case OP_POSPLUS:        case OP_POSPLUS:
2016          case OP_POSPLUSI:
2017        case OP_QUERY:        case OP_QUERY:
2018          case OP_QUERYI:
2019        case OP_MINQUERY:        case OP_MINQUERY:
2020          case OP_MINQUERYI:
2021        case OP_POSQUERY:        case OP_POSQUERY:
2022        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2023          if (code[-1] >= 0xc0) code += PRIV(utf8_table4)[code[-1] & 0x3f];
2024        break;        break;
2025        }        }
2026  #else  #else
# Line 1679  Arguments: Line 2046  Arguments:
2046  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found
2047  */  */
2048    
2049  static const uschar *  static const pcre_uchar *
2050  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf8)
2051  {  {
2052  for (;;)  for (;;)
2053    {    {
# Line 1696  for (;;) Line 2063  for (;;)
2063    
2064    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
2065    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
2066    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2067      must add in its length. */
2068    
2069    else    else
2070      {      {
# Line 1720  for (;;) Line 2088  for (;;)
2088        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2089        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2090        break;        break;
2091    
2092          case OP_MARK:
2093          case OP_PRUNE_ARG:
2094          case OP_SKIP_ARG:
2095          code += code[1];
2096          break;
2097    
2098          case OP_THEN_ARG:
2099          code += code[1];
2100          break;
2101        }        }
2102    
2103      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2104    
2105      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2106    
2107      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
2108      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
# Line 1734  for (;;) Line 2112  for (;;)
2112      if (utf8) switch(c)      if (utf8) switch(c)
2113        {        {
2114        case OP_CHAR:        case OP_CHAR:
2115        case OP_CHARNC:        case OP_CHARI:
2116        case OP_EXACT:        case OP_EXACT:
2117          case OP_EXACTI:
2118        case OP_UPTO:        case OP_UPTO:
2119          case OP_UPTOI:
2120        case OP_MINUPTO:        case OP_MINUPTO:
2121          case OP_MINUPTOI:
2122        case OP_POSUPTO:        case OP_POSUPTO:
2123          case OP_POSUPTOI:
2124        case OP_STAR:        case OP_STAR:
2125          case OP_STARI:
2126        case OP_MINSTAR:        case OP_MINSTAR:
2127          case OP_MINSTARI:
2128        case OP_POSSTAR:        case OP_POSSTAR:
2129          case OP_POSSTARI:
2130        case OP_PLUS:        case OP_PLUS:
2131          case OP_PLUSI:
2132        case OP_MINPLUS:        case OP_MINPLUS:
2133          case OP_MINPLUSI:
2134        case OP_POSPLUS:        case OP_POSPLUS:
2135          case OP_POSPLUSI:
2136        case OP_QUERY:        case OP_QUERY:
2137          case OP_QUERYI:
2138        case OP_MINQUERY:        case OP_MINQUERY:
2139          case OP_MINQUERYI:
2140        case OP_POSQUERY:        case OP_POSQUERY:
2141        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2142          if (code[-1] >= 0xc0) code += PRIV(utf8_table4)[code[-1] & 0x3f];
2143        break;        break;
2144        }        }
2145  #else  #else
# Line 1776  Arguments: Line 2167  Arguments:
2167    code        points to start of search    code        points to start of search
2168    endcode     points to where to stop    endcode     points to where to stop
2169    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2170      cd          contains pointers to tables etc.
2171    
2172  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2173  */  */
2174    
2175  static BOOL  static BOOL
2176  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2177      BOOL utf8, compile_data *cd)
2178  {  {
2179  register int c;  register int c;
2180  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2181       code < endcode;       code < endcode;
2182       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2183    {    {
2184    const uschar *ccode;    const pcre_uchar *ccode;
2185    
2186    c = *code;    c = *code;
2187    
# Line 1802  for (code = first_significant_code(code Line 2195  for (code = first_significant_code(code
2195      continue;      continue;
2196      }      }
2197    
2198      /* For a recursion/subroutine call, if its end has been reached, which
2199      implies a backward reference subroutine call, we can scan it. If it's a
2200      forward reference subroutine call, we can't. To detect forward reference
2201      we have to scan up the list that is kept in the workspace. This function is
2202      called only when doing the real compile, not during the pre-compile that
2203      measures the size of the compiled pattern. */
2204    
2205      if (c == OP_RECURSE)
2206        {
2207        const pcre_uchar *scode;
2208        BOOL empty_branch;
2209    
2210        /* Test for forward reference */
2211    
2212        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2213          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2214    
2215        /* Not a forward reference, test for completed backward reference */
2216    
2217        empty_branch = FALSE;
2218        scode = cd->start_code + GET(code, 1);
2219        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2220    
2221        /* Completed backwards reference */
2222    
2223        do
2224          {
2225          if (could_be_empty_branch(scode, endcode, utf8, cd))
2226            {
2227            empty_branch = TRUE;
2228            break;
2229            }
2230          scode += GET(scode, 1);
2231          }
2232        while (*scode == OP_ALT);
2233    
2234        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2235        continue;
2236        }
2237    
2238    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2239    
2240    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2241          c == OP_BRAPOSZERO)
2242        {
2243        code += PRIV(OP_lengths)[c];
2244        do code += GET(code, 1); while (*code == OP_ALT);
2245        c = *code;
2246        continue;
2247        }
2248    
2249      /* A nested group that is already marked as "could be empty" can just be
2250      skipped. */
2251    
2252      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2253          c == OP_SCBRA || c == OP_SCBRAPOS)
2254      {      {
     code += _pcre_OP_lengths[c];  
2255      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2256      c = *code;      c = *code;
2257      continue;      continue;
# Line 1814  for (code = first_significant_code(code Line 2259  for (code = first_significant_code(code
2259    
2260    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2261    
2262    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2263          c == OP_CBRA || c == OP_CBRAPOS ||
2264          c == OP_ONCE || c == OP_ONCE_NC ||
2265          c == OP_COND)
2266      {      {
2267      BOOL empty_branch;      BOOL empty_branch;
2268      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1830  for (code = first_significant_code(code Line 2278  for (code = first_significant_code(code
2278        empty_branch = FALSE;        empty_branch = FALSE;
2279        do        do
2280          {          {
2281          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2282            empty_branch = TRUE;            empty_branch = TRUE;
2283          code += GET(code, 1);          code += GET(code, 1);
2284          }          }
# Line 1848  for (code = first_significant_code(code Line 2296  for (code = first_significant_code(code
2296      {      {
2297      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2298      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2299      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2300      actual length is stored in the compiled code, so we must update "code"      actual length is stored in the compiled code, so we must update "code"
2301      here. */      here. */
2302    
# Line 1901  for (code = first_significant_code(code Line 2349  for (code = first_significant_code(code
2349      case OP_ALLANY:      case OP_ALLANY:
2350      case OP_ANYBYTE:      case OP_ANYBYTE:
2351      case OP_CHAR:      case OP_CHAR:
2352      case OP_CHARNC:      case OP_CHARI:
2353      case OP_NOT:      case OP_NOT:
2354        case OP_NOTI:
2355      case OP_PLUS:      case OP_PLUS:
2356      case OP_MINPLUS:      case OP_MINPLUS:
2357      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1942  for (code = first_significant_code(code Line 2391  for (code = first_significant_code(code
2391      case OP_KET:      case OP_KET:
2392      case OP_KETRMAX:      case OP_KETRMAX:
2393      case OP_KETRMIN:      case OP_KETRMIN:
2394        case OP_KETRPOS:
2395      case OP_ALT:      case OP_ALT:
2396      return TRUE;      return TRUE;
2397    
# Line 1950  for (code = first_significant_code(code Line 2400  for (code = first_significant_code(code
2400    
2401  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2402      case OP_STAR:      case OP_STAR:
2403        case OP_STARI:
2404      case OP_MINSTAR:      case OP_MINSTAR:
2405        case OP_MINSTARI:
2406      case OP_POSSTAR:      case OP_POSSTAR:
2407        case OP_POSSTARI:
2408      case OP_QUERY:      case OP_QUERY:
2409        case OP_QUERYI:
2410      case OP_MINQUERY:      case OP_MINQUERY:
2411        case OP_MINQUERYI:
2412      case OP_POSQUERY:      case OP_POSQUERY:
2413      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      case OP_POSQUERYI:
2414        if (utf8 && code[1] >= 0xc0) code += PRIV(utf8_table4)[code[1] & 0x3f];
2415      break;      break;
2416    
2417      case OP_UPTO:      case OP_UPTO:
2418        case OP_UPTOI:
2419      case OP_MINUPTO:      case OP_MINUPTO:
2420        case OP_MINUPTOI:
2421      case OP_POSUPTO:      case OP_POSUPTO:
2422      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      case OP_POSUPTOI:
2423        if (utf8 && code[3] >= 0xc0) code += PRIV(utf8_table4)[code[3] & 0x3f];
2424      break;      break;
2425  #endif  #endif
2426    
2427        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2428        string. */
2429    
2430        case OP_MARK:
2431        case OP_PRUNE_ARG:
2432        case OP_SKIP_ARG:
2433        code += code[1];
2434        break;
2435    
2436        case OP_THEN_ARG:
2437        code += code[1];
2438        break;
2439    
2440        /* None of the remaining opcodes are required to match a character. */
2441    
2442        default:
2443        break;
2444      }      }
2445    }    }
2446    
# Line 1980  return TRUE; Line 2457  return TRUE;
2457  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2458  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2459  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2460    This function is called only during the real compile, not during the
2461    pre-compile.
2462    
2463  Arguments:  Arguments:
2464    code        points to start of the recursion    code        points to start of the recursion
2465    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2466    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2467    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2468      cd          pointers to tables etc
2469    
2470  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2471  */  */
2472    
2473  static BOOL  static BOOL
2474  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2475    BOOL utf8)    branch_chain *bcptr, BOOL utf8, compile_data *cd)
2476  {  {
2477  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2478    {    {
2479    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2480        return FALSE;
2481    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2482    }    }
2483  return TRUE;  return TRUE;
# Line 2028  where Perl recognizes it as the POSIX cl Line 2509  where Perl recognizes it as the POSIX cl
2509  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2510  I think.  I think.
2511    
2512    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2513    It seems that the appearance of a nested POSIX class supersedes an apparent
2514    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2515    a digit.
2516    
2517    In Perl, unescaped square brackets may also appear as part of class names. For
2518    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2519    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2520    seem right at all. PCRE does not allow closing square brackets in POSIX class
2521    names.
2522    
2523  Arguments:  Arguments:
2524    ptr      pointer to the initial [    ptr      pointer to the initial [
2525    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2036  Returns:   TRUE or FALSE Line 2528  Returns:   TRUE or FALSE
2528  */  */
2529    
2530  static BOOL  static BOOL
2531  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2532  {  {
2533  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2534  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2535  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2536    {    {
2537    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2538        ptr++;
2539      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2540      else
2541      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2542      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2543        {        {
2544        *endptr = ptr;        *endptr = ptr;
2545        return TRUE;        return TRUE;
2546        }        }
2547        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2548             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2549              ptr[1] == CHAR_EQUALS_SIGN) &&
2550            check_posix_syntax(ptr, endptr))
2551          return FALSE;
2552      }      }
2553    }    }
2554  return FALSE;  return FALSE;
# Line 2073  Returns:     a value representing the na Line 2572  Returns:     a value representing the na
2572  */  */
2573    
2574  static int  static int
2575  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2576  {  {
2577  const char *pn = posix_names;  const char *pn = posix_names;
2578  register int yield = 0;  register int yield = 0;
2579  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2580    {    {
2581    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2582      strncmp((const char *)ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2583    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
2584    yield++;    yield++;
2585    }    }
# Line 2120  Returns:     nothing Line 2619  Returns:     nothing
2619  */  */
2620    
2621  static void  static void
2622  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf8, compile_data *cd,
2623    uschar *save_hwm)    pcre_uchar *save_hwm)
2624  {  {
2625  uschar *ptr = group;  pcre_uchar *ptr = group;
2626    
2627  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf8)) != NULL)
2628    {    {
2629    int offset;    int offset;
2630    uschar *hc;    pcre_uchar *hc;
2631    
2632    /* See if this recursion is on the forward reference list. If so, adjust the    /* See if this recursion is on the forward reference list. If so, adjust the
2633    reference. */    reference. */
# Line 2173  Arguments: Line 2672  Arguments:
2672  Returns:         new code pointer  Returns:         new code pointer
2673  */  */
2674    
2675  static uschar *  static pcre_uchar *
2676  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2677  {  {
2678  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2679  *code++ = 255;  *code++ = 255;
2680  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2681  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2682  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2683  }  }
2684    
2685    
# Line 2202  Returns:             nothing Line 2701  Returns:             nothing
2701  */  */
2702    
2703  static void  static void
2704  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2705  {  {
2706  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2707  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2708  }  }
2709    
# Line 2254  for (++c; c <= d; c++) Line 2753  for (++c; c <= d; c++)
2753    
2754  return TRUE;  return TRUE;
2755  }  }
2756    
2757    
2758    
2759    /*************************************************
2760    *        Check a character and a property        *
2761    *************************************************/
2762    
2763    /* This function is called by check_auto_possessive() when a property item
2764    is adjacent to a fixed character.
2765    
2766    Arguments:
2767      c            the character
2768      ptype        the property type
2769      pdata        the data for the type
2770      negated      TRUE if it's a negated property (\P or \p{^)
2771    
2772    Returns:       TRUE if auto-possessifying is OK
2773    */
2774    
2775    static BOOL
2776    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2777    {
2778    const ucd_record *prop = GET_UCD(c);
2779    switch(ptype)
2780      {
2781      case PT_LAMP:
2782      return (prop->chartype == ucp_Lu ||
2783              prop->chartype == ucp_Ll ||
2784              prop->chartype == ucp_Lt) == negated;
2785    
2786      case PT_GC:
2787      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2788    
2789      case PT_PC:
2790      return (pdata == prop->chartype) == negated;
2791    
2792      case PT_SC:
2793      return (pdata == prop->script) == negated;
2794    
2795      /* These are specials */
2796    
2797      case PT_ALNUM:
2798      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2799              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2800    
2801      case PT_SPACE:    /* Perl space */
2802      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2803              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2804              == negated;
2805    
2806      case PT_PXSPACE:  /* POSIX space */
2807      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2808              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2809              c == CHAR_FF || c == CHAR_CR)
2810              == negated;
2811    
2812      case PT_WORD:
2813      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2814              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2815              c == CHAR_UNDERSCORE) == negated;
2816      }
2817    return FALSE;
2818    }
2819  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2820    
2821    
# Line 2267  whether the next thing could possibly ma Line 2829  whether the next thing could possibly ma
2829  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2830    
2831  Arguments:  Arguments:
2832    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2833    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2834    ptr           next character in pattern    ptr           next character in pattern
2835    options       options bits    options       options bits
2836    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2279  Returns:        TRUE if possessifying is Line 2839  Returns:        TRUE if possessifying is
2839  */  */
2840    
2841  static BOOL  static BOOL
2842  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf8,
2843    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2844  {  {
2845  int next;  int c, next;
2846    int op_code = *previous++;
2847    
2848  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2849    
# Line 2293  if ((options & PCRE_EXTENDED) != 0) Line 2854  if ((options & PCRE_EXTENDED) != 0)
2854      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2855      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2856        {        {
2857        while (*(++ptr) != 0)        ptr++;
2858          while (*ptr != 0)
2859            {
2860          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2861            ptr++;
2862    #ifdef SUPPORT_UTF8
2863            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2864    #endif
2865            }
2866        }        }
2867      else break;      else break;
2868      }      }
# Line 2330  if ((options & PCRE_EXTENDED) != 0) Line 2898  if ((options & PCRE_EXTENDED) != 0)
2898      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2899      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2900        {        {
2901        while (*(++ptr) != 0)        ptr++;
2902          while (*ptr != 0)
2903            {
2904          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2905            ptr++;
2906    #ifdef SUPPORT_UTF8
2907            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2908    #endif
2909            }
2910        }        }
2911      else break;      else break;
2912      }      }
# Line 2340  if ((options & PCRE_EXTENDED) != 0) Line 2915  if ((options & PCRE_EXTENDED) != 0)
2915  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2916    
2917  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2918    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2919      return FALSE;      return FALSE;
2920    
2921  /* Now compare the next item with the previous opcode. If the previous is a  /* Now compare the next item with the previous opcode. First, handle cases when
2922  positive single character match, "item" either contains the character or, if  the next item is a character. */
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
   
 /* Handle cases when the next item is a character. */  
2923    
2924  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2925    {    {
2926    case OP_CHAR:    case OP_CHAR:
2927  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2928    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2929  #else  #else
2930    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2931  #endif  #endif
2932    return item != next;    return c != next;
2933    
2934    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2935    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2936    high-valued characters. */    high-valued characters. */
2937    
2938    case OP_CHARNC:    case OP_CHARI:
2939  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2940    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2941    #else
2942      c = *previous;
2943  #endif  #endif
2944    if (item == next) return FALSE;    if (c == next) return FALSE;
2945  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2946    if (utf8)    if (utf8)
2947      {      {
# Line 2380  if (next >= 0) switch(op_code) Line 2952  if (next >= 0) switch(op_code)
2952  #else  #else
2953      othercase = NOTACHAR;      othercase = NOTACHAR;
2954  #endif  #endif
2955      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2956      }      }
2957    else    else
2958  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2959    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2960    
2961    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2962      opcodes are not used for multi-byte characters, because they are coded using
2963      an XCLASS instead. */
2964    
2965    case OP_NOT:    case OP_NOT:
2966    if (item == next) return TRUE;    return (c = *previous) == next;
2967    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2968      case OP_NOTI:
2969      if ((c = *previous) == next) return TRUE;
2970  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2971    if (utf8)    if (utf8)
2972      {      {
# Line 2401  if (next >= 0) switch(op_code) Line 2977  if (next >= 0) switch(op_code)
2977  #else  #else
2978      othercase = NOTACHAR;      othercase = NOTACHAR;
2979  #endif  #endif
2980      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2981      }      }
2982    else    else
2983  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2984    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2985    
2986      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2987      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2988    
2989    case OP_DIGIT:    case OP_DIGIT:
2990    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2448  if (next >= 0) switch(op_code) Line 3027  if (next >= 0) switch(op_code)
3027      case 0x202f:      case 0x202f:
3028      case 0x205f:      case 0x205f:
3029      case 0x3000:      case 0x3000:
3030      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
3031      default:      default:
3032      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
3033      }      }
3034    
3035      case OP_ANYNL:
3036    case OP_VSPACE:    case OP_VSPACE:
3037    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3038    switch(next)    switch(next)
# Line 2464  if (next >= 0) switch(op_code) Line 3044  if (next >= 0) switch(op_code)
3044      case 0x85:      case 0x85:
3045      case 0x2028:      case 0x2028:
3046      case 0x2029:      case 0x2029:
3047      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
3048      default:      default:
3049      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
3050      }      }
3051    
3052    #ifdef SUPPORT_UCP
3053      case OP_PROP:
3054      return check_char_prop(next, previous[0], previous[1], FALSE);
3055    
3056      case OP_NOTPROP:
3057      return check_char_prop(next, previous[0], previous[1], TRUE);
3058    #endif
3059    
3060    default:    default:
3061    return FALSE;    return FALSE;
3062    }    }
3063    
3064    
3065  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3066    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3067    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3068    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3069    replaced by OP_PROP codes when PCRE_UCP is set. */
3070    
3071  switch(op_code)  switch(op_code)
3072    {    {
3073    case OP_CHAR:    case OP_CHAR:
3074    case OP_CHARNC:    case OP_CHARI:
3075  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3076    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3077    #else
3078      c = *previous;
3079  #endif  #endif
3080    switch(-next)    switch(-next)
3081      {      {
3082      case ESC_d:      case ESC_d:
3083      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3084    
3085      case ESC_D:      case ESC_D:
3086      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3087    
3088      case ESC_s:      case ESC_s:
3089      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3090    
3091      case ESC_S:      case ESC_S:
3092      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3093    
3094      case ESC_w:      case ESC_w:
3095      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3096    
3097      case ESC_W:      case ESC_W:
3098      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3099    
3100      case ESC_h:      case ESC_h:
3101      case ESC_H:      case ESC_H:
3102      switch(item)      switch(c)
3103        {        {
3104        case 0x09:        case 0x09:
3105        case 0x20:        case 0x20:
# Line 2533  switch(op_code) Line 3127  switch(op_code)
3127    
3128      case ESC_v:      case ESC_v:
3129      case ESC_V:      case ESC_V:
3130      switch(item)      switch(c)
3131        {        {
3132        case 0x0a:        case 0x0a:
3133        case 0x0b:        case 0x0b:
# Line 2547  switch(op_code) Line 3141  switch(op_code)
3141        return -next == ESC_v;        return -next == ESC_v;
3142        }        }
3143    
3144        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3145        their substitutions and process them. The result will always be either
3146        -ESC_p or -ESC_P. Then fall through to process those values. */
3147    
3148    #ifdef SUPPORT_UCP
3149        case ESC_du:
3150        case ESC_DU:
3151        case ESC_wu:
3152        case ESC_WU:
3153        case ESC_su:
3154        case ESC_SU:
3155          {
3156          int temperrorcode = 0;
3157          ptr = substitutes[-next - ESC_DU];
3158          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3159          if (temperrorcode != 0) return FALSE;
3160          ptr++;    /* For compatibility */
3161          }
3162        /* Fall through */
3163    
3164        case ESC_p:
3165        case ESC_P:
3166          {
3167          int ptype, pdata, errorcodeptr;
3168          BOOL negated;
3169    
3170          ptr--;      /* Make ptr point at the p or P */
3171          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3172          if (ptype < 0) return FALSE;
3173          ptr++;      /* Point past the final curly ket */
3174    
3175          /* If the property item is optional, we have to give up. (When generated
3176          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3177          to the original \d etc. At this point, ptr will point to a zero byte. */
3178    
3179          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3180            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3181              return FALSE;
3182    
3183          /* Do the property check. */
3184    
3185          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3186          }
3187    #endif
3188    
3189      default:      default:
3190      return FALSE;      return FALSE;
3191      }      }
3192    
3193      /* In principle, support for Unicode properties should be integrated here as
3194      well. It means re-organizing the above code so as to get hold of the property
3195      values before switching on the op-code. However, I wonder how many patterns
3196      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3197      these op-codes are never generated.) */
3198    
3199    case OP_DIGIT:    case OP_DIGIT:
3200    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3201           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3202    
3203    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3204    return next == -ESC_d;    return next == -ESC_d;
3205    
3206    case OP_WHITESPACE:    case OP_WHITESPACE:
3207    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3208    
3209    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3210    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3211    
3212    case OP_HSPACE:    case OP_HSPACE:
3213    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3214             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3215    
3216    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3217    return next == -ESC_h;    return next == -ESC_h;
3218    
3219    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3220      case OP_ANYNL:
3221    case OP_VSPACE:    case OP_VSPACE:
3222    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3223    
3224    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3225    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3226    
3227    case OP_WORDCHAR:    case OP_WORDCHAR:
3228    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3229             next == -ESC_v || next == -ESC_R;
3230    
3231    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3232    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2610  Arguments: Line 3258  Arguments:
3258    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3259    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3260    bcptr          points to current branch chain    bcptr          points to current branch chain
3261      cond_depth     conditional nesting depth
3262    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3263    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3264                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2619  Returns:         TRUE on success Line 3268  Returns:         TRUE on success
3268  */  */
3269    
3270  static BOOL  static BOOL
3271  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3272    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, int *firstbyteptr,
3273    compile_data *cd, int *lengthptr)    int *reqbyteptr, branch_chain *bcptr, int cond_depth, compile_data *cd,
3274      int *lengthptr)
3275  {  {
3276  int repeat_type, op_type;  int repeat_type, op_type;
3277  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2630  int greedy_default, greedy_non_default; Line 3280  int greedy_default, greedy_non_default;
3280  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3281  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3282  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3283  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3284  int after_manual_callout = 0;  int after_manual_callout = 0;
3285  int length_prevgroup = 0;  int length_prevgroup = 0;
3286  register int c;  register int c;
3287  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3288  uschar *last_code = code;  pcre_uchar *last_code = code;
3289  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3290  uschar *tempcode;  pcre_uchar *tempcode;
3291  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3292  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3293  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3294  const uschar *tempptr;  const pcre_uchar *tempptr;
3295  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3296  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3297  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3298  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3299    pcre_uchar classbits[32];
3300    
3301    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3302    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3303    dynamically as we process the pattern. */
3304    
3305  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3306  BOOL class_utf8;  BOOL class_utf8;
3307  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3308  uschar *class_utf8data;  pcre_uint8 *class_utf8data;
3309  uschar *class_utf8data_base;  pcre_uint8 *class_utf8data_base;
3310  uschar utf8_char[6];  pcre_uint8 utf8_char[6];
3311  #else  #else
3312  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3313  #endif  #endif
3314    
3315  #ifdef DEBUG  #ifdef PCRE_DEBUG
3316  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3317  #endif  #endif
3318    
# Line 2706  for (;; ptr++) Line 3360  for (;; ptr++)
3360    int subfirstbyte;    int subfirstbyte;
3361    int terminator;    int terminator;
3362    int mclength;    int mclength;
3363    uschar mcbuffer[8];    int tempbracount;
3364      pcre_uchar mcbuffer[8];
3365    
3366    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3367    
3368    c = *ptr;    c = *ptr;
3369    
3370      /* If we are at the end of a nested substitution, revert to the outer level
3371      string. Nesting only happens one level deep. */
3372    
3373      if (c == 0 && nestptr != NULL)
3374        {
3375        ptr = nestptr;
3376        nestptr = NULL;
3377        c = *ptr;
3378        }
3379    
3380    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
3381    previous cycle of this loop. */    previous cycle of this loop. */
3382    
3383    if (lengthptr != NULL)    if (lengthptr != NULL)
3384      {      {
3385  #ifdef DEBUG  #ifdef PCRE_DEBUG
3386      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3387  #endif  #endif
3388      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3389        {        {
3390        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3391        goto FAILED;        goto FAILED;
# Line 2742  for (;; ptr++) Line 3407  for (;; ptr++)
3407        goto FAILED;        goto FAILED;
3408        }        }
3409    
3410      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3411      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
3412          c));
3413    
3414      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3415      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 2769  for (;; ptr++) Line 3435  for (;; ptr++)
3435    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3436    reference list. */    reference list. */
3437    
3438    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3439      {      {
3440      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3441      goto FAILED;      goto FAILED;
# Line 2817  for (;; ptr++) Line 3483  for (;; ptr++)
3483      previous_callout = NULL;      previous_callout = NULL;
3484      }      }
3485    
3486    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3487    
3488    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3489      {      {
3490      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3491      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3492        {        {
3493        while (*(++ptr) != 0)        ptr++;
3494          while (*ptr != 0)
3495          {          {
3496          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3497            ptr++;
3498    #ifdef SUPPORT_UTF8
3499            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3500    #endif
3501          }          }
3502        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3503    
# Line 2860  for (;; ptr++) Line 3531  for (;; ptr++)
3531          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3532          goto FAILED;          goto FAILED;
3533          }          }
3534        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3535        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3536        }        }
3537      return TRUE;      return TRUE;
# Line 2871  for (;; ptr++) Line 3542  for (;; ptr++)
3542      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3543    
3544      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3545        previous = NULL;
3546      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3547        {        {
3548        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3549          *code++ = OP_CIRCM;
3550        }        }
3551      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3552      break;      break;
3553    
3554      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3555      previous = NULL;      previous = NULL;
3556      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3557      break;      break;
3558    
3559      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
# Line 2945  for (;; ptr++) Line 3617  for (;; ptr++)
3617          {          {
3618          if (ptr[1] == CHAR_E)          if (ptr[1] == CHAR_E)
3619            ptr++;            ptr++;
3620          else if (strncmp((const char *)ptr+1,          else if (STRNCMP_UC_C8(ptr + 1,
3621                            STR_Q STR_BACKSLASH STR_E, 3) == 0)                            STR_Q STR_BACKSLASH STR_E, 3) == 0)
3622            ptr += 3;            ptr += 3;
3623          else          else
# Line 2988  for (;; ptr++) Line 3660  for (;; ptr++)
3660      than 256), because in that case the compiled code doesn't use the bit map.      than 256), because in that case the compiled code doesn't use the bit map.
3661      */      */
3662    
3663      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
3664    
3665  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3666      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
# Line 3002  for (;; ptr++) Line 3674  for (;; ptr++)
3674    
3675      if (c != 0) do      if (c != 0) do
3676        {        {
3677        const uschar *oldptr;        const pcre_uchar *oldptr;
3678    
3679  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3680        if (utf8 && c > 127)        if (utf8 && c > 127)
# Line 3048  for (;; ptr++) Line 3720  for (;; ptr++)
3720          {          {
3721          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3722          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3723          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
3724          uschar pbits[32];          pcre_uint8 pbits[32];
3725    
3726          if (ptr[1] != CHAR_COLON)          if (ptr[1] != CHAR_COLON)
3727            {            {
# Line 3065  for (;; ptr++) Line 3737  for (;; ptr++)
3737            ptr++;            ptr++;
3738            }            }
3739    
3740          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3741          if (posix_class < 0)          if (posix_class < 0)
3742            {            {
3743            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3079  for (;; ptr++) Line 3751  for (;; ptr++)
3751          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3752            posix_class = 0;            posix_class = 0;
3753    
3754          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3755          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3756          subtract bits that may be in the main map already. At the end we or the  
3757          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3758            if ((options & PCRE_UCP) != 0)
3759              {
3760              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3761              if (posix_substitutes[pc] != NULL)
3762                {
3763                nestptr = tempptr + 1;
3764                ptr = posix_substitutes[pc] - 1;
3765                continue;
3766                }
3767              }
3768    #endif
3769            /* In the non-UCP case, we build the bit map for the POSIX class in a
3770            chunk of local store because we may be adding and subtracting from it,
3771            and we don't want to subtract bits that may be in the main map already.
3772            At the end we or the result into the bit map that is being built. */
3773    
3774          posix_class *= 3;          posix_class *= 3;
3775    
3776          /* Copy in the first table (always present) */          /* Copy in the first table (always present) */
3777    
3778          memcpy(pbits, cbits + posix_class_maps[posix_class],          memcpy(pbits, cbits + posix_class_maps[posix_class],
3779            32 * sizeof(uschar));            32 * sizeof(pcre_uint8));
3780    
3781          /* If there is a second table, add or remove it as required. */          /* If there is a second table, add or remove it as required. */
3782    
# Line 3126  for (;; ptr++) Line 3813  for (;; ptr++)
3813    
3814        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3815        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
3816        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3817        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so set
3818        to 'or' into the one we are building. We assume they have more than one        class_charcount bigger than one. Unrecognized escapes fall through and
3819        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3820          PCRE_EXTRA is set. */
3821    
3822        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3823          {          {
3824          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3825          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3826    
3827          if (-c == ESC_b) c = CHAR_BS;       /* \b is backspace in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
         else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */  
3828          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3829            {            {
3830            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3152  for (;; ptr++) Line 3838  for (;; ptr++)
3838    
3839          if (c < 0)          if (c < 0)
3840            {            {
3841            register const uschar *cbits = cd->cbits;            register const pcre_uint8 *cbits = cd->cbits;
3842            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3843    
3844            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3845              {              {
3846    #ifdef SUPPORT_UCP
3847                case ESC_du:     /* These are the values given for \d etc */
3848                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3849                case ESC_wu:     /* escape sequence with an appropriate \p */
3850                case ESC_WU:     /* or \P to test Unicode properties instead */
3851                case ESC_su:     /* of the default ASCII testing. */
3852                case ESC_SU:
3853                nestptr = ptr;
3854                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3855                class_charcount -= 2;                /* Undo! */
3856                continue;
3857    #endif
3858              case ESC_d:              case ESC_d:
3859              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3860              continue;              continue;
# Line 3177  for (;; ptr++) Line 3873  for (;; ptr++)
3873              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3874              continue;              continue;
3875    
3876                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3877                if it was previously set by something earlier in the character
3878                class. */
3879    
3880              case ESC_s:              case ESC_s:
3881              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3882              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3883                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3884              continue;              continue;
3885    
3886              case ESC_S:              case ESC_S:
# Line 3188  for (;; ptr++) Line 3889  for (;; ptr++)
3889              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3890              continue;              continue;
3891    
3892              default:    /* Not recognized; fall through */              case ESC_h:
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
           /* We need to deal with \H, \h, \V, and \v in both phases because  
           they use extra memory. */  
   
           if (-c == ESC_h)  
             {  
3893              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3894              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3895              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3210  for (;; ptr++) Line 3898  for (;; ptr++)
3898                {                {
3899                class_utf8 = TRUE;                class_utf8 = TRUE;
3900                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3901                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x1680, class_utf8data);
3902                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3903                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x180e, class_utf8data);
3904                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3905                class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2000, class_utf8data);
3906                class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x200A, class_utf8data);
3907                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3908                class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x202f, class_utf8data);
3909                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3910                class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x205f, class_utf8data);
3911                *class_utf8data++ = XCL_SINGLE;                *class_utf8data++ = XCL_SINGLE;
3912                class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x3000, class_utf8data);
3913                }                }
3914  #endif  #endif
3915              continue;              continue;
             }  
3916    
3917            if (-c == ESC_H)              case ESC_H:
             {  
3918              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3919                {                {
3920                int x = 0xff;                int x = 0xff;
# Line 3247  for (;; ptr++) Line 3933  for (;; ptr++)
3933                {                {
3934                class_utf8 = TRUE;                class_utf8 = TRUE;
3935                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3936                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x0100, class_utf8data);
3937                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x167f, class_utf8data);
3938                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3939                class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x1681, class_utf8data);
3940                class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x180d, class_utf8data);
3941                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3942                class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x180f, class_utf8data);
3943                class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x1fff, class_utf8data);
3944                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3945                class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x200B, class_utf8data);
3946                class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x202e, class_utf8data);
3947                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3948                class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2030, class_utf8data);
3949                class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x205e, class_utf8data);
3950                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3951                class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2060, class_utf8data);
3952                class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2fff, class_utf8data);
3953                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3954                class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x3001, class_utf8data);
3955                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x7fffffff, class_utf8data);
3956                }                }
3957  #endif  #endif
3958              continue;              continue;
             }  
3959    
3960            if (-c == ESC_v)              case ESC_v:
             {  
3961              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3962              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3963              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3284  for (;; ptr++) Line 3968  for (;; ptr++)
3968                {                {
3969                class_utf8 = TRUE;                class_utf8 = TRUE;
3970                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3971                class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2028, class_utf8data);
3972                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2029, class_utf8data);
3973                }                }
3974  #endif  #endif
3975              continue;              continue;
             }  
3976    
3977            if (-c == ESC_V)              case ESC_V:
             {  
3978              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3979                {                {
3980                int x = 0xff;                int x = 0xff;
# Line 3314  for (;; ptr++) Line 3996  for (;; ptr++)
3996                {                {
3997                class_utf8 = TRUE;                class_utf8 = TRUE;
3998                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
3999                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x0100, class_utf8data);
4000                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2027, class_utf8data);
4001                *class_utf8data++ = XCL_RANGE;                *class_utf8data++ = XCL_RANGE;
4002                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x2029, class_utf8data);
4003                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_utf8data += PRIV(ord2utf8)(0x7fffffff, class_utf8data);
4004                }                }
4005  #endif  #endif
4006              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
4007    
4008  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4009            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
4010              {              case ESC_P:
4011              BOOL negated;                {
4012              int pdata;                BOOL negated;
4013              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
4014              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
4015              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
4016              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
4017                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
4018              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
4019              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
4020              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
4021              continue;                class_charcount -= 2;   /* Not a < 256 character */
4022              }                continue;
4023                  }
4024  #endif  #endif
4025            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
4026            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
4027            treated as literals. */              treated as literals. */
4028    
4029            if ((options & PCRE_EXTRA) != 0)              default:
4030              {              if ((options & PCRE_EXTRA) != 0)
4031              *errorcodeptr = ERR7;                {
4032              goto FAILED;                *errorcodeptr = ERR7;
4033                  goto FAILED;
4034                  }
4035                class_charcount -= 2;  /* Undo the default count from above */
4036                c = *ptr;              /* Get the final character and fall through */
4037                break;
4038              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
4039            }            }
4040    
4041          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
# Line 3423  for (;; ptr++) Line 4105  for (;; ptr++)
4105            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
4106            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
4107    
4108            /* \b is backspace; \X is literal X; \R is literal R; any other            /* \b is backspace; any other special means the '-' was literal */
           special means the '-' was literal */  
4109    
4110            if (d < 0)            if (d < 0)
4111              {              {
4112              if (d == -ESC_b) d = CHAR_BS;              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = CHAR_X;  
             else if (d == -ESC_R) d = CHAR_R; else  
4113                {                {
4114                ptr = oldptr;                ptr = oldptr;
4115                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3499  for (;; ptr++) Line 4178  for (;; ptr++)
4178                else                else
4179                  {                  {
4180                  *class_utf8data++ = XCL_RANGE;                  *class_utf8data++ = XCL_RANGE;
4181                  class_utf8data += _pcre_ord2utf8(occ, class_utf8data);                  class_utf8data += PRIV(ord2utf8)(occ, class_utf8data);
4182                  }                  }
4183                class_utf8data += _pcre_ord2utf8(ocd, class_utf8data);                class_utf8data += PRIV(ord2utf8)(ocd, class_utf8data);
4184                }                }
4185              }              }
4186  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
# Line 3510  for (;; ptr++) Line 4189  for (;; ptr++)
4189            overlapping ranges. */            overlapping ranges. */
4190    
4191            *class_utf8data++ = XCL_RANGE;            *class_utf8data++ = XCL_RANGE;
4192            class_utf8data += _pcre_ord2utf8(c, class_utf8data);            class_utf8data += PRIV(ord2utf8)(c, class_utf8data);
4193            class_utf8data += _pcre_ord2utf8(d, class_utf8data);            class_utf8data += PRIV(ord2utf8)(d, class_utf8data);
4194    
4195            /* With UCP support, we are done. Without UCP support, there is no            /* With UCP support, we are done. Without UCP support, there is no
4196            caseless matching for UTF-8 characters > 127; we can use the bit map            caseless matching for UTF-8 characters > 127; we can use the bit map
# Line 3565  for (;; ptr++) Line 4244  for (;; ptr++)
4244          {          {
4245          class_utf8 = TRUE;          class_utf8 = TRUE;
4246          *class_utf8data++ = XCL_SINGLE;          *class_utf8data++ = XCL_SINGLE;
4247          class_utf8data += _pcre_ord2utf8(c, class_utf8data);          class_utf8data += PRIV(ord2utf8)(c, class_utf8data);
4248    
4249  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4250          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
# Line 3574  for (;; ptr++) Line 4253  for (;; ptr++)
4253            if ((othercase = UCD_OTHERCASE(c)) != c)            if ((othercase = UCD_OTHERCASE(c)) != c)
4254              {              {
4255              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4256              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += PRIV(ord2utf8)(othercase, class_utf8data);
4257              }              }
4258            }            }
4259  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
# Line 3596  for (;; ptr++) Line 4275  for (;; ptr++)
4275          }          }
4276        }        }
4277    
4278      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" far above.
4279        If we are at the end of an internal nested string, revert to the outer
4280        string. */
4281    
4282        while (((c = *(++ptr)) != 0 ||
4283               (nestptr != NULL &&
4284                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4285               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4286    
4287      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4288    
4289      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4290        {        {
4291        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4292        goto FAILED;        goto FAILED;
4293        }        }
4294    
   
 /* This code has been disabled because it would mean that \s counts as  
 an explicit \r or \n reference, and that's not really what is wanted. Now  
 we set the flag only if there is a literal "\r" or "\n" in the class. */  
   
 #if 0  
     /* Remember whether \r or \n are in this class */  
   
     if (negate_class)  
       {  
       if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;  
       }  
     else  
       {  
       if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;  
       }  
 #endif  
   
   
4295      /* 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
4296      less than 256. As long as there were no characters >= 128 and there was no      less than 256. As long as there were no characters >= 128 and there was no
4297      use of \p or \P, in other words, no use of any XCLASS features, we can      use of \p or \P, in other words, no use of any XCLASS features, we can
# Line 3632  we set the flag only if there is a liter Line 4299  we set the flag only if there is a liter
4299    
4300      In UTF-8 mode, we can optimize the negative case only if there were no      In UTF-8 mode, we can optimize the negative case only if there were no
4301      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4302      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4303      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4304    
4305      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
4306      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4307      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4308      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4309      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4310      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4311    
4312  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4313      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3651  we set the flag only if there is a liter Line 4318  we set the flag only if there is a liter
4318        {        {
4319        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4320    
4321        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4322    
4323        if (negate_class)        if (negate_class)
4324          {          {
4325          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4326          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4327          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4328          *code++ = class_lastchar;          *code++ = class_lastchar;
4329          break;          break;
4330          }          }
# Line 3667  we set the flag only if there is a liter Line 4334  we set the flag only if there is a liter
4334    
4335  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4336        if (utf8 && class_lastchar > 127)        if (utf8 && class_lastchar > 127)
4337          mclength = _pcre_ord2utf8(class_lastchar, mcbuffer);          mclength = PRIV(ord2utf8)(class_lastchar, mcbuffer);
4338        else        else
4339  #endif  #endif
4340          {          {
# Line 3688  we set the flag only if there is a liter Line 4355  we set the flag only if there is a liter
4355    
4356      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4357      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4358      such as \S in the class, because in that case all characters > 255 are in      such as \S in the class, and PCRE_UCP is not set, because in that case all
4359      the class, so any that were explicitly given as well can be ignored. If      characters > 255 are in the class, so any that were explicitly given as
4360      (when there are explicit characters > 255 that must be listed) there are no      well can be ignored. If (when there are explicit characters > 255 that must
4361      characters < 256, we can omit the bitmap in the actual compiled code. */      be listed) there are no characters < 256, we can omit the bitmap in the
4362        actual compiled code. */
4363    
4364  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4365      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4366        {        {
4367        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4368        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3720  we set the flag only if there is a liter Line 4388  we set the flag only if there is a liter
4388        }        }
4389  #endif  #endif
4390    
4391      /* If there are no characters > 255, set the opcode to OP_CLASS or      /* If there are no characters > 255, or they are all to be included or
4392      OP_NCLASS, depending on whether the whole class was negated and whether      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4393      there were negative specials such as \S in the class. Then copy the 32-byte      whole class was negated and whether there were negative specials such as \S
4394      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4395        negating it if necessary. */
4396    
4397      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4398      if (negate_class)      if (negate_class)
# Line 3783  we set the flag only if there is a liter Line 4452  we set the flag only if there is a liter
4452      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4453      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4454    
4455      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4456      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4457    
4458      tempcode = previous;      tempcode = previous;
4459    
# Line 3807  we set the flag only if there is a liter Line 4476  we set the flag only if there is a liter
4476        }        }
4477      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4478    
4479        /* If previous was a recursion call, wrap it in atomic brackets so that
4480        previous becomes the atomic group. All recursions were so wrapped in the
4481        past, but it no longer happens for non-repeated recursions. In fact, the
4482        repeated ones could be re-implemented independently so as not to need this,
4483        but for the moment we rely on the code for repeating groups. */
4484    
4485        if (*previous == OP_RECURSE)
4486          {
4487          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4488          *previous = OP_ONCE;
4489          PUT(previous, 1, 2 + 2*LINK_SIZE);
4490          previous[2 + 2*LINK_SIZE] = OP_KET;
4491          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4492          code += 2 + 2 * LINK_SIZE;
4493          length_prevgroup = 3 + 3*LINK_SIZE;
4494    
4495          /* When actually compiling, we need to check whether this was a forward
4496          reference, and if so, adjust the offset. */
4497    
4498          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4499            {
4500            int offset = GET(cd->hwm, -LINK_SIZE);
4501            if (offset == previous + 1 - cd->start_code)
4502              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4503            }
4504          }
4505    
4506        /* Now handle repetition for the different types of item. */
4507    
4508      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4509      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
4510      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
4511      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstbyte
4512      instead.  */      instead.  */
4513    
4514      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4515        {        {
4516          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4517    
4518        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4519        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4520        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 3823  we set the flag only if there is a liter Line 4523  we set the flag only if there is a liter
4523  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4524        if (utf8 && (code[-1] & 0x80) != 0)        if (utf8 && (code[-1] & 0x80) != 0)
4525          {          {
4526          uschar *lastchar = code - 1;          pcre_uchar *lastchar = code - 1;
4527          while((*lastchar & 0xc0) == 0x80) lastchar--;          while((*lastchar & 0xc0) == 0x80) lastchar--;
4528          c = code - lastchar;            /* Length of UTF-8 character */          c = code - lastchar;            /* Length of UTF-8 character */
4529          memcpy(utf8_char, lastchar, c); /* Save the char */          memcpy(utf8_char, lastchar, c); /* Save the char */
# Line 3847  we set the flag only if there is a liter Line 4547  we set the flag only if there is a liter
4547    
4548        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4549            repeat_max < 0 &&            repeat_max < 0 &&
4550            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4551          {          {
4552          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4553          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3860  we set the flag only if there is a liter Line 4559  we set the flag only if there is a liter
4559      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4560      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4561      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4562      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4563      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4564    
4565      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4566        {        {
4567        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4568        c = previous[1];        c = previous[1];
4569        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4570            repeat_max < 0 &&            repeat_max < 0 &&
4571            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4572          {          {
4573          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4574          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3886  we set the flag only if there is a liter Line 4585  we set the flag only if there is a liter
4585    
4586      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
4587        {        {
4588        uschar *oldcode;        pcre_uchar *oldcode;
4589        int prop_type, prop_value;        int prop_type, prop_value;
4590        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
4591        c = *previous;        c = *previous;
4592    
4593        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4594            repeat_max < 0 &&            repeat_max < 0 &&
4595            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4596          {          {
4597          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4598          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3915  we set the flag only if there is a liter Line 4614  we set the flag only if there is a liter
4614    
4615        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4616    
4617        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4618        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4619        removed: */        removed: */
4620    
4621        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4622        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4623    
4624        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4625        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4626    
4627        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4628    
# Line 4062  we set the flag only if there is a liter Line 4761  we set the flag only if there is a liter
4761  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4762               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4763  #endif  #endif
4764               *previous == OP_REF)               *previous == OP_REF ||
4765                 *previous == OP_REFI)
4766        {        {
4767        if (repeat_max == 0)        if (repeat_max == 0)
4768          {          {
# Line 4070  we set the flag only if there is a liter Line 4770  we set the flag only if there is a liter
4770          goto END_REPEAT;          goto END_REPEAT;
4771          }          }
4772    
4773        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4774        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4775        removed: */        removed: */
4776    
# Line 4078  we set the flag only if there is a liter Line 4778  we set the flag only if there is a liter
4778        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4779    
4780        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4781        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4782    
4783        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4784          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 4096  we set the flag only if there is a liter Line 4796  we set the flag only if there is a liter
4796        }        }
4797    
4798      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4799      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4800        opcodes such as BRA and CBRA, as this is the place where they get converted
4801        into the more special varieties such as BRAPOS and SBRA. A test for >=
4802        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4803        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4804        repetition of assertions, but now it does, for Perl compatibility. */
4805    
4806      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4807        {        {
4808        register int i;        register int i;
4809        int ketoffset = 0;        int len = (int)(code - previous);
4810        int len = code - previous;        pcre_uchar *bralink = NULL;
4811        uschar *bralink = NULL;        pcre_uchar *brazeroptr = NULL;
4812    
4813        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4814          we just ignore the repeat. */
4815    
4816        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4817          {          goto END_REPEAT;
4818          *errorcodeptr = ERR55;  
4819          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4820          }        use of repetition is in cases when the assertion is optional. Therefore,
4821          if the minimum is greater than zero, just ignore the repeat. If the
4822          maximum is not not zero or one, set it to 1. */
4823    
4824        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4825        by scanning through from the start, and compute the offset back to it          {
4826        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4827        the final KET, so we can't find the end just by going back from the code          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
4828          }          }
4829    
4830        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 4145  we set the flag only if there is a liter Line 4845  we set the flag only if there is a liter
4845          **   goto END_REPEAT;          **   goto END_REPEAT;
4846          **   }          **   }
4847    
4848          However, that fails when a group is referenced as a subroutine from          However, that fails when a group or a subgroup within it is referenced
4849          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it          as a subroutine from elsewhere in the pattern, so now we stick in
4850          so that it is skipped on execution. As we don't have a list of which          OP_SKIPZERO in front of it so that it is skipped on execution. As we
4851          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4852            selectively.
4853    
4854          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4855          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 4168  we set the flag only if there is a liter Line 4869  we set the flag only if there is a liter
4869              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4870              goto END_REPEAT;              goto END_REPEAT;
4871              }              }
4872              brazeroptr = previous;    /* Save for possessive optimizing */
4873            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4874            }            }
4875    
# Line 4192  we set the flag only if there is a liter Line 4894  we set the flag only if there is a liter
4894            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4895            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4896    
4897            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4898            bralink = previous;            bralink = previous;
4899            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4900            }            }
# Line 4213  we set the flag only if there is a liter Line 4915  we set the flag only if there is a liter
4915            {            {
4916            /* 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
4917            just adjust the length as if we had. Do some paranoid checks for            just adjust the length as if we had. Do some paranoid checks for
4918            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4919              integer type when available, otherwise double. */
4920    
4921            if (lengthptr != NULL)            if (lengthptr != NULL)
4922              {              {
4923              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4924              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4925                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4926                        (INT64_OR_DOUBLE)INT_MAX ||
4927                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4928                {                {
4929                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4235  we set the flag only if there is a liter Line 4939  we set the flag only if there is a liter
4939              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
4940              for (i = 1; i < repeat_min; i++)              for (i = 1; i < repeat_min; i++)
4941                {                {
4942                uschar *hc;                pcre_uchar *hc;
4943                uschar *this_hwm = cd->hwm;                pcre_uchar *this_hwm = cd->hwm;
4944                memcpy(code, previous, len);                memcpy(code, previous, len);
4945                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4946                  {                  {
# Line 4265  we set the flag only if there is a liter Line 4969  we set the flag only if there is a liter
4969          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
4970          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
4971          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4972          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4973            a 64-bit integer type when available, otherwise double. */
4974    
4975          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4976            {            {
4977            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4978                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4979            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4980                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4981                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4982                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4983              {              {
4984              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4286  we set the flag only if there is a liter Line 4991  we set the flag only if there is a liter
4991    
4992          else for (i = repeat_max - 1; i >= 0; i--)          else for (i = repeat_max - 1; i >= 0; i--)
4993            {            {
4994            uschar *hc;            pcre_uchar *hc;
4995            uschar *this_hwm = cd->hwm;            pcre_uchar *this_hwm = cd->hwm;
4996    
4997            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
4998    
# Line 4298  we set the flag only if there is a liter Line 5003  we set the flag only if there is a liter
5003              {              {
5004              int offset;              int offset;
5005              *code++ = OP_BRA;              *code++ = OP_BRA;
5006              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
5007              bralink = code;              bralink = code;
5008              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
5009              }              }
# Line 4319  we set the flag only if there is a liter Line 5024  we set the flag only if there is a liter
5024          while (bralink != NULL)          while (bralink != NULL)
5025            {            {
5026            int oldlinkoffset;            int oldlinkoffset;
5027            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
5028            uschar *bra = code - offset;            pcre_uchar *bra = code - offset;
5029            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
5030            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
5031            *code++ = OP_KET;            *code++ = OP_KET;
# Line 4329  we set the flag only if there is a liter Line 5034  we set the flag only if there is a liter
5034            }            }
5035          }          }
5036    
5037        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
5038        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
5039        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
5040        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5041          deal with possessive ONCEs specially.
5042        Then, when we are doing the actual compile phase, check to see whether  
5043        this group is a non-atomic one that could match an empty string. If so,        Otherwise, when we are doing the actual compile phase, check to see
5044          whether this group is one that could match an empty string. If so,
5045        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
5046        that runtime checking can be done. [This check is also applied to        that runtime checking can be done. [This check is also applied to ONCE
5047        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
5048    
5049          Then, if the quantifier was possessive and the bracket is not a
5050          conditional, we convert the BRA code to the POS form, and the KET code to
5051          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5052          subpattern at both the start and at the end.) The use of special opcodes
5053          makes it possible to reduce greatly the stack usage in pcre_exec(). If
5054          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5055    
5056          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5057          flag so that the default action below, of wrapping everything inside
5058          atomic brackets, does not happen. When the minimum is greater than 1,
5059          there will be earlier copies of the group, and so we still have to wrap
5060          the whole thing. */
5061    
5062        else        else
5063          {          {
5064          uschar *ketcode = code - ketoffset;          pcre_uchar *ketcode = code - 1 - LINK_SIZE;
5065          uschar *bracode = ketcode - GET(ketcode, 1);          pcre_uchar *bracode = ketcode - GET(ketcode, 1);
5066          *ketcode = OP_KETRMAX + repeat_type;  
5067          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5068    
5069            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5070                possessive_quantifier) *bracode = OP_BRA;
5071    
5072            /* For non-possessive ONCE brackets, all we need to do is to
5073            set the KET. */
5074    
5075            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5076              *ketcode = OP_KETRMAX + repeat_type;
5077    
5078            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5079            converted to non-capturing above). */
5080    
5081            else
5082            {            {
5083            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5084            do  
5085              if (lengthptr == NULL)
5086              {              {
5087              if (could_be_empty_branch(scode, ketcode, utf8))              pcre_uchar *scode = bracode;
5088                do
5089                {                {
5090                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
5091                break;                  {
5092                    *bracode += OP_SBRA - OP_BRA;
5093                    break;
5094                    }
5095                  scode += GET(scode, 1);
5096                  }
5097                while (*scode == OP_ALT);
5098                }
5099    
5100              /* Handle possessive quantifiers. */
5101    
5102              if (possessive_quantifier)
5103                {
5104                /* For COND brackets, we wrap the whole thing in a possessively
5105                repeated non-capturing bracket, because we have not invented POS
5106                versions of the COND opcodes. Because we are moving code along, we
5107                must ensure that any pending recursive references are updated. */
5108    
5109                if (*bracode == OP_COND || *bracode == OP_SCOND)
5110                  {
5111                  int nlen = (int)(code - bracode);
5112                  *code = OP_END;
5113                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5114                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5115                  code += 1 + LINK_SIZE;
5116                  nlen += 1 + LINK_SIZE;
5117                  *bracode = OP_BRAPOS;
5118                  *code++ = OP_KETRPOS;
5119                  PUTINC(code, 0, nlen);
5120                  PUT(bracode, 1, nlen);
5121                  }
5122    
5123                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5124    
5125                else
5126                  {
5127                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5128                  *ketcode = OP_KETRPOS;
5129                }                }
5130              scode += GET(scode, 1);  
5131                /* If the minimum is zero, mark it as possessive, then unset the
5132                possessive flag when the minimum is 0 or 1. */
5133    
5134                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5135                if (repeat_min < 2) possessive_quantifier = FALSE;
5136              }              }
5137            while (*scode == OP_ALT);  
5138              /* Non-possessive quantifier */
5139    
5140              else *ketcode = OP_KETRMAX + repeat_type;
5141            }            }
5142          }          }
5143        }        }
# Line 4378  we set the flag only if there is a liter Line 5158  we set the flag only if there is a liter
5158        }        }
5159    
5160      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
5161      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5162      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
5163      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5164      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
5165      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
5166      tempcode, not at previous, which might be the first part of a string whose  
5167      (former) last char we repeated.      Some (but not all) possessively repeated subpatterns have already been
5168        completely handled in the code just above. For them, possessive_quantifier
5169        is always FALSE at this stage.
5170    
5171        Note that the repeated item starts at tempcode, not at previous, which
5172        might be the first part of a string whose (former) last char we repeated.
5173    
5174      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
5175      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4393  we set the flag only if there is a liter Line 5178  we set the flag only if there is a liter
5178      if (possessive_quantifier)      if (possessive_quantifier)
5179        {        {
5180        int len;        int len;
5181    
5182        if (*tempcode == OP_TYPEEXACT)        if (*tempcode == OP_TYPEEXACT)
5183          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += PRIV(OP_lengths)[*tempcode] +
5184            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
5185    
5186        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
5187          {          {
5188          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += PRIV(OP_lengths)[*tempcode];
5189  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5190          if (utf8 && tempcode[-1] >= 0xc0)          if (utf8 && tempcode[-1] >= 0xc0)
5191            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];            tempcode += PRIV(utf8_table4)[tempcode[-1] & 0x3f];
5192  #endif  #endif
5193          }          }
5194    
5195        len = code - tempcode;        len = (int)(code - tempcode);
5196        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
5197          {          {
5198          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4415  we set the flag only if there is a liter Line 5200  we set the flag only if there is a liter
5200          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
5201          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
5202    
5203          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
5204          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
5205          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
5206          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
5207    
5208          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
5209          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
5210          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
5211          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
5212    
5213            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
5214            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
5215            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5216            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
5217    
5218            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
5219            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
5220            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5221            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
5222    
5223            /* Because we are moving code along, we must ensure that any
5224            pending recursive references are updated. */
5225    
5226          default:          default:
5227            *code = OP_END;
5228            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5229          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
5230          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
5231          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4461  we set the flag only if there is a liter Line 5261  we set the flag only if there is a liter
5261    
5262      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
5263    
5264      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
5265             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
5266        {        {
5267        int i, namelen;        int i, namelen;
5268          int arglen = 0;
5269        const char *vn = verbnames;        const char *vn = verbnames;
5270        const uschar *name = ++ptr;        const pcre_uchar *name = ptr + 1;
5271          const pcre_uchar *arg = NULL;
5272        previous = NULL;        previous = NULL;
5273        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5274          namelen = (int)(ptr - name);
5275    
5276          /* It appears that Perl allows any characters whatsoever, other than
5277          a closing parenthesis, to appear in arguments, so we no longer insist on
5278          letters, digits, and underscores. */
5279    
5280        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5281          {          {
5282          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
5283          goto FAILED;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5284            arglen = (int)(ptr - arg);
5285          }          }
5286    
5287        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
5288          {          {
5289          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
5290          goto FAILED;          goto FAILED;
5291          }          }
5292        namelen = ptr - name;  
5293          /* Scan the table of verb names */
5294    
5295        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5296          {          {
5297          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5298              strncmp((char *)name, vn, namelen) == 0)              STRNCMP_UC_C8(name, vn, namelen) == 0)
5299            {            {
5300            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5301              ASSERT_ACCEPT if in an assertion. */
5302    
5303            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5304              {              {
5305              open_capitem *oc;              open_capitem *oc;
5306              cd->had_accept = TRUE;              if (arglen != 0)
5307                  {
5308                  *errorcodeptr = ERR59;
5309                  goto FAILED;
5310                  }
5311                cd->had_accept = TRUE;
5312              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5313                {                {
5314                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5315                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5316                }                }
5317              }              *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5318            *code++ = verbs[i].op;  
5319            break;              /* Do not set firstbyte after *ACCEPT */
5320                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5321                }
5322    
5323              /* Handle other cases with/without an argument */
5324    
5325              else if (arglen == 0)
5326                {
5327                if (verbs[i].op < 0)   /* Argument is mandatory */
5328                  {
5329                  *errorcodeptr = ERR66;
5330                  goto FAILED;
5331                  }
5332                *code = verbs[i].op;
5333                if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
5334                }
5335    
5336              else
5337                {
5338                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
5339                  {
5340                  *errorcodeptr = ERR59;
5341                  goto FAILED;
5342                  }
5343                *code = verbs[i].op_arg;
5344                if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
5345                *code++ = arglen;
5346                memcpy(code, arg, arglen);
5347                code += arglen;
5348                *code++ = 0;
5349                }
5350    
5351              break;  /* Found verb, exit loop */
5352            }            }
5353    
5354          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
5355          }          }
5356        if (i < verbcount) continue;  
5357        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
5358          *errorcodeptr = ERR60;          /* Verb not recognized */
5359        goto FAILED;        goto FAILED;
5360        }        }
5361    
# Line 4513  we set the flag only if there is a liter Line 5366  we set the flag only if there is a liter
5366        {        {
5367        int i, set, unset, namelen;        int i, set, unset, namelen;
5368        int *optset;        int *optset;
5369        const uschar *name;        const pcre_uchar *name;
5370        uschar *slot;        pcre_uchar *slot;
5371    
5372        switch (*(++ptr))        switch (*(++ptr))
5373          {          {
# Line 4621  we set the flag only if there is a liter Line 5474  we set the flag only if there is a liter
5474                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5475            ptr++;            ptr++;
5476            }            }
5477          namelen = ptr - name;          namelen = (int)(ptr - name);
5478    
5479          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5480              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4658  we set the flag only if there is a liter Line 5511  we set the flag only if there is a liter
5511            }            }
5512    
5513          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5514          name. If we find a name, add one to the opcode to change OP_CREF or          name. If we find a name, add one to the opcode to change OP_CREF or
5515          OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,          OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5516          except they record that the reference was originally to a name. The          except they record that the reference was originally to a name. The
5517          information is used to check duplicate names. */          information is used to check duplicate names. */
5518    
5519          slot = cd->name_table;          slot = cd->name_table;
5520          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
5521            {            {
5522            if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;            if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0) break;
5523            slot += cd->name_entry_size;            slot += cd->name_entry_size;
5524            }            }
5525    
# Line 4682  we set the flag only if there is a liter Line 5535  we set the flag only if there is a liter
5535          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5536    
5537          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5538                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5539            {            {
5540            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5541            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 4723  we set the flag only if there is a liter