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

Diff of /code/trunk/pcre_compile.c

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

revision 514 by ph10, Mon May 3 12:54:22 2010 UTC revision 924 by zherczeg, Wed Feb 22 10:23:56 2012 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) 1997-2010 University of Cambridge             Copyright (c) 1997-2012 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 PCRE_DEBUG is defined, we need the pcre_printint() function, which is  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which
57  also used by pcretest. PCRE_DEBUG is not defined when building a production  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58  library. */  library. We do not need to select pcre16_printint.c specially, because the
59    COMPILE_PCREx macro will already be appropriately set. */
60    
61  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
62  #include "pcre_printint.src"  /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66  #endif  #endif
67    
68    
# Line 88  so this number is very generous. Line 92  so this number is very generous.
92  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
93  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
94  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
95  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
96    filled up by repetitions of forward references, for example patterns like
97    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
98    that the workspace is expanded using malloc() in this situation. The value
99    below is therefore a minimum, and we put a maximum on it for safety. The
100    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
101    kicks in at the same number of forward references in all cases. */
102    
103  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
104    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
105    
106  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
107  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
108    
109  #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)  #define WORK_SIZE_SAFETY_MARGIN (100)
110    
111    /* Private flags added to firstchar and reqchar. */
112    
113    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
114    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
115    
116    /* Repeated character flags. */
117    
118    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
119    
120  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
121  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 231  static const char posix_names[] = Line 250  static const char posix_names[] =
250    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
251    STRING_word0  STRING_xdigit;    STRING_word0  STRING_xdigit;
252    
253  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
254    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 };
255    
256  /* 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 261  static const int posix_class_maps[] = { Line 280  static const int posix_class_maps[] = {
280    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
281  };  };
282    
283    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
284    substitutes must be in the order of the names, defined above, and there are
285    both positive and negative cases. NULL means no substitute. */
286    
287    #ifdef SUPPORT_UCP
288    static const pcre_uchar string_PNd[]  = {
289      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
290      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
291    static const pcre_uchar string_pNd[]  = {
292      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
293      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
294    static const pcre_uchar string_PXsp[] = {
295      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
296      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
297    static const pcre_uchar string_pXsp[] = {
298      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
299      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
300    static const pcre_uchar string_PXwd[] = {
301      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
302      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
303    static const pcre_uchar string_pXwd[] = {
304      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
305      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    
307    static const pcre_uchar *substitutes[] = {
308      string_PNd,           /* \D */
309      string_pNd,           /* \d */
310      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
311      string_pXsp,          /* \s */
312      string_PXwd,          /* \W */
313      string_pXwd           /* \w */
314    };
315    
316    static const pcre_uchar string_pL[] =   {
317      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
318      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
319    static const pcre_uchar string_pLl[] =  {
320      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
321      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
322    static const pcre_uchar string_pLu[] =  {
323      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
324      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
325    static const pcre_uchar string_pXan[] = {
326      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
327      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
328    static const pcre_uchar string_h[] =    {
329      CHAR_BACKSLASH, CHAR_h, '\0' };
330    static const pcre_uchar string_pXps[] = {
331      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
332      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
333    static const pcre_uchar string_PL[] =   {
334      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
335      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
336    static const pcre_uchar string_PLl[] =  {
337      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
338      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
339    static const pcre_uchar string_PLu[] =  {
340      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
341      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
342    static const pcre_uchar string_PXan[] = {
343      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
344      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
345    static const pcre_uchar string_H[] =    {
346      CHAR_BACKSLASH, CHAR_H, '\0' };
347    static const pcre_uchar string_PXps[] = {
348      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
349      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
350    
351    static const pcre_uchar *posix_substitutes[] = {
352      string_pL,            /* alpha */
353      string_pLl,           /* lower */
354      string_pLu,           /* upper */
355      string_pXan,          /* alnum */
356      NULL,                 /* ascii */
357      string_h,             /* blank */
358      NULL,                 /* cntrl */
359      string_pNd,           /* digit */
360      NULL,                 /* graph */
361      NULL,                 /* print */
362      NULL,                 /* punct */
363      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
364      string_pXwd,          /* word */
365      NULL,                 /* xdigit */
366      /* Negated cases */
367      string_PL,            /* ^alpha */
368      string_PLl,           /* ^lower */
369      string_PLu,           /* ^upper */
370      string_PXan,          /* ^alnum */
371      NULL,                 /* ^ascii */
372      string_H,             /* ^blank */
373      NULL,                 /* ^cntrl */
374      string_PNd,           /* ^digit */
375      NULL,                 /* ^graph */
376      NULL,                 /* ^print */
377      NULL,                 /* ^punct */
378      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
379      string_PXwd,          /* ^word */
380      NULL                  /* ^xdigit */
381    };
382    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
383    #endif
384    
385  #define STRING(a)  # a  #define STRING(a)  # a
386  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 318  static const char error_texts[] = Line 438  static const char error_texts[] =
438    /* 30 */    /* 30 */
439    "unknown POSIX class name\0"    "unknown POSIX class name\0"
440    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
441    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
442    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
443    "character value in \\x{...} sequence is too large\0"    "character value in \\x{...} sequence is too large\0"
444    /* 35 */    /* 35 */
# Line 341  static const char error_texts[] = Line 461  static const char error_texts[] =
461    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
462    /* 50 */    /* 50 */
463    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
464    "octal value is greater than \\377 (not in UTF-8 mode)\0"    "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
465    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
466    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
467    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
468    /* 55 */    /* 55 */
469    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
470    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
471    "\\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"
472    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
# Line 360  static const char error_texts[] = Line 480  static const char error_texts[] =
480    /* 65 */    /* 65 */
481    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
482    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
483      "this version of PCRE is not compiled with Unicode property support\0"
484      "\\c must be followed by an ASCII character\0"
485      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
486      /* 70 */
487      "internal error: unknown opcode in find_fixedlength()\0"
488      "\\N is not supported in a class\0"
489      "too many forward references\0"
490      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
491      "invalid UTF-16 string\0"
492    ;    ;
493    
494  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 378  For convenience, we use the same bit def Line 507  For convenience, we use the same bit def
507    
508  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
509    
510    /* Using a simple comparison for decimal numbers rather than a memory read
511    is much faster, and the resulting code is simpler (the compiler turns it
512    into a subtraction and unsigned comparison). */
513    
514    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
515    
516  #ifndef EBCDIC  #ifndef EBCDIC
517    
518  /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in  /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
519  UTF-8 mode. */  UTF-8 mode. */
520    
521  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
522    {    {
523    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
524    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 422  static const unsigned char digitab[] = Line 557  static const unsigned char digitab[] =
557    
558  /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */  /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
559    
560  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
561    {    {
562    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
563    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 457  static const unsigned char digitab[] = Line 592  static const unsigned char digitab[] =
592    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
593    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
594    
595  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
596    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
597    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
598    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 496  static const unsigned char ebcdic_charta Line 631  static const unsigned char ebcdic_charta
631  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
632    
633  static BOOL  static BOOL
634    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
635      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
636    
637    
# Line 528  return s; Line 663  return s;
663    
664    
665  /*************************************************  /*************************************************
666    *           Expand the workspace                 *
667    *************************************************/
668    
669    /* This function is called during the second compiling phase, if the number of
670    forward references fills the existing workspace, which is originally a block on
671    the stack. A larger block is obtained from malloc() unless the ultimate limit
672    has been reached or the increase will be rather small.
673    
674    Argument: pointer to the compile data block
675    Returns:  0 if all went well, else an error number
676    */
677    
678    static int
679    expand_workspace(compile_data *cd)
680    {
681    pcre_uchar *newspace;
682    int newsize = cd->workspace_size * 2;
683    
684    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
685    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
686        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
687     return ERR72;
688    
689    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
690    if (newspace == NULL) return ERR21;
691    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
692    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
693    if (cd->workspace_size > COMPILE_WORK_SIZE)
694      (PUBL(free))((void *)cd->start_workspace);
695    cd->start_workspace = newspace;
696    cd->workspace_size = newsize;
697    return 0;
698    }
699    
700    
701    
702    /*************************************************
703    *            Check for counted repeat            *
704    *************************************************/
705    
706    /* This function is called when a '{' is encountered in a place where it might
707    start a quantifier. It looks ahead to see if it really is a quantifier or not.
708    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
709    where the ddds are digits.
710    
711    Arguments:
712      p         pointer to the first char after '{'
713    
714    Returns:    TRUE or FALSE
715    */
716    
717    static BOOL
718    is_counted_repeat(const pcre_uchar *p)
719    {
720    if (!IS_DIGIT(*p)) return FALSE;
721    p++;
722    while (IS_DIGIT(*p)) p++;
723    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
724    
725    if (*p++ != CHAR_COMMA) return FALSE;
726    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
727    
728    if (!IS_DIGIT(*p)) return FALSE;
729    p++;
730    while (IS_DIGIT(*p)) p++;
731    
732    return (*p == CHAR_RIGHT_CURLY_BRACKET);
733    }
734    
735    
736    
737    /*************************************************
738  *            Handle escapes                      *  *            Handle escapes                      *
739  *************************************************/  *************************************************/
740    
# Line 552  Returns: zero or positive => a d Line 759  Returns: zero or positive => a d
759  */  */
760    
761  static int  static int
762  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
763    int options, BOOL isclass)    int options, BOOL isclass)
764  {  {
765  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
766  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
767  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
768    pcre_int32 c;
769    int i;
770    
771  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
772  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
# Line 571  in a table. A non-zero result is somethi Line 780  in a table. A non-zero result is somethi
780  Otherwise further processing may be required. */  Otherwise further processing may be required. */
781    
782  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
783  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */  /* Not alphanumeric */
784    else if (c < CHAR_0 || c > CHAR_z) {}
785  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
786    
787  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
788  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
789    else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
790  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
791  #endif  #endif
792    
# Line 583  else if ((i = escapes[c - 0x48]) != 0) Line 794  else if ((i = escapes[c - 0x48]) != 0)
794    
795  else  else
796    {    {
797    const uschar *oldptr;    const pcre_uchar *oldptr;
798    BOOL braced, negated;    BOOL braced, negated;
799    
800    switch (c)    switch (c)
# Line 593  else Line 804  else
804    
805      case CHAR_l:      case CHAR_l:
806      case CHAR_L:      case CHAR_L:
807        *errorcodeptr = ERR37;
808        break;
809    
810      case CHAR_u:      case CHAR_u:
811        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
812          {
813          /* In JavaScript, \u must be followed by four hexadecimal numbers.
814          Otherwise it is a lowercase u letter. */
815          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
816            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
817            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
818            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
819            {
820            c = 0;
821            for (i = 0; i < 4; ++i)
822              {
823              register int cc = *(++ptr);
824    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
825              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
826              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
827    #else           /* EBCDIC coding */
828              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
829              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
830    #endif
831              }
832            }
833          }
834        else
835          *errorcodeptr = ERR37;
836        break;
837    
838      case CHAR_U:      case CHAR_U:
839      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
840        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
841      break;      break;
842    
843      /* \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
844        class, \g must be followed by one of a number of specific things:
845    
846      (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
847      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 615  else Line 858  else
858      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
859    
860      case CHAR_g:      case CHAR_g:
861        if (isclass) break;
862      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
863        {        {
864        c = -ESC_g;        c = -ESC_g;
# Line 625  else Line 869  else
869    
870      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
871        {        {
872        const uschar *p;        const pcre_uchar *p;
873        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
874          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
875        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
876          {          {
877          c = -ESC_k;          c = -ESC_k;
# Line 645  else Line 889  else
889        }        }
890      else negated = FALSE;      else negated = FALSE;
891    
892        /* The integer range is limited by the machine's int representation. */
893      c = 0;      c = 0;
894      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
895          {
896          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
897            {
898            c = -1;
899            break;
900            }
901        c = c * 10 + *(++ptr) - CHAR_0;        c = c * 10 + *(++ptr) - CHAR_0;
902          }
903      if (c < 0)   /* Integer overflow */      if (((unsigned int)c) > INT_MAX) /* Integer overflow */
904        {        {
905          while (IS_DIGIT(ptr[1]))
906            ptr++;
907        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
908        break;        break;
909        }        }
# Line 698  else Line 951  else
951      if (!isclass)      if (!isclass)
952        {        {
953        oldptr = ptr;        oldptr = ptr;
954          /* The integer range is limited by the machine's int representation. */
955        c -= CHAR_0;        c -= CHAR_0;
956        while ((digitab[ptr[1]] & ctype_digit) != 0)        while (IS_DIGIT(ptr[1]))
957            {
958            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
959              {
960              c = -1;
961              break;
962              }
963          c = c * 10 + *(++ptr) - CHAR_0;          c = c * 10 + *(++ptr) - CHAR_0;
964        if (c < 0)    /* Integer overflow */          }
965          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
966          {          {
967            while (IS_DIGIT(ptr[1]))
968              ptr++;
969          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
970          break;          break;
971          }          }
# Line 728  else Line 991  else
991      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
992      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
993      significant 8 bits of octal numbers (I think this is what early Perls used      significant 8 bits of octal numbers (I think this is what early Perls used
994      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
995      than 3 octal digits. */      but no more than 3 octal digits. */
996    
997      case CHAR_0:      case CHAR_0:
998      c -= CHAR_0;      c -= CHAR_0;
999      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1000          c = c * 8 + *(++ptr) - CHAR_0;          c = c * 8 + *(++ptr) - CHAR_0;
1001      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1002        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1003    #endif
1004      break;      break;
1005    
1006      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \x is complicated. \x{ddd} is a character number which can be greater
1007      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.
1008      treated as a data character. */      If not, { is treated as a data character. */
1009    
1010      case CHAR_x:      case CHAR_x:
1011        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1012          {
1013          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1014          Otherwise it is a lowercase x letter. */
1015          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1016            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1017            {
1018            c = 0;
1019            for (i = 0; i < 2; ++i)
1020              {
1021              register int cc = *(++ptr);
1022    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1023              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1024              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1025    #else           /* EBCDIC coding */
1026              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1027              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1028    #endif
1029              }
1030            }
1031          break;
1032          }
1033    
1034      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1035        {        {
1036        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1037    
1038        c = 0;        c = 0;
1039        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1040          {          {
1041          register int cc = *pt++;          register int cc = *pt++;
1042          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
1043    
1044  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1045          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
# Line 762  else Line 1048  else
1048          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1049          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1050  #endif  #endif
1051    
1052    #ifdef COMPILE_PCRE8
1053            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1054    #else
1055    #ifdef COMPILE_PCRE16
1056            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1057    #endif
1058    #endif
1059            }
1060    
1061          if (c < 0)
1062            {
1063            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1064            *errorcodeptr = ERR34;
1065          }          }
1066    
1067        if (*pt == CHAR_RIGHT_CURLY_BRACKET)        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1068          {          {
1069          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1070          ptr = pt;          ptr = pt;
1071          break;          break;
1072          }          }
# Line 778  else Line 1078  else
1078      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1079    
1080      c = 0;      c = 0;
1081      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1082        {        {
1083        int cc;                                  /* Some compilers don't like */        int cc;                                  /* Some compilers don't like */
1084        cc = *(++ptr);                           /* ++ in initializers */        cc = *(++ptr);                           /* ++ in initializers */
# Line 793  else Line 1093  else
1093      break;      break;
1094    
1095      /* 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.
1096      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
1097        coding is ASCII-specific, but then the whole concept of \cx is
1098      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1099    
1100      case CHAR_c:      case CHAR_c:
# Line 803  else Line 1104  else
1104        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1105        break;        break;
1106        }        }
1107    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1108  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1109          {
1110          *errorcodeptr = ERR68;
1111          break;
1112          }
1113      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1114      c ^= 0x40;      c ^= 0x40;
1115  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1116      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1117      c ^= 0xC0;      c ^= 0xC0;
1118  #endif  #endif
# Line 829  else Line 1134  else
1134      break;      break;
1135      }      }
1136    }    }
   
 /* Perl supports \N{name} for character names, as well as plain \N for "not  
 newline". PCRE does not support \N{name}. */  
1137    
1138  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  /* Perl supports \N{name} for character names, as well as plain \N for "not
1139    *errorcodeptr = ERR37;  newline". PCRE does not support \N{name}. However, it does support
1140    quantification such as \N{2,3}. */
1141    
1142    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1143         !is_counted_repeat(ptr+2))
1144      *errorcodeptr = ERR37;
1145    
1146    /* If PCRE_UCP is set, we change the values for \d etc. */
1147    
1148    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1149      c -= (ESC_DU - ESC_D);
1150    
1151    /* Set the pointer to the final character before returning. */
1152    
1153  *ptrptr = ptr;  *ptrptr = ptr;
1154  return c;  return c;
# Line 862  Returns: type value from ucp_typ Line 1176  Returns: type value from ucp_typ
1176  */  */
1177    
1178  static int  static int
1179  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1180  {  {
1181  int c, i, bot, top;  int c, i, bot, top;
1182  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1183  char name[32];  pcre_uchar name[32];
1184    
1185  c = *(++ptr);  c = *(++ptr);
1186  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 883  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1197  if (c == CHAR_LEFT_CURLY_BRACKET)
1197      *negptr = TRUE;      *negptr = TRUE;
1198      ptr++;      ptr++;
1199      }      }
1200    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1201      {      {
1202      c = *(++ptr);      c = *(++ptr);
1203      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
# Line 907  else Line 1221  else
1221  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1222    
1223  bot = 0;  bot = 0;
1224  top = _pcre_utt_size;  top = PRIV(utt_size);
1225    
1226  while (bot < top)  while (bot < top)
1227    {    {
1228    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1229    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1230    if (c == 0)    if (c == 0)
1231      {      {
1232      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1233      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1234      }      }
1235    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1236    }    }
# Line 936  return -1; Line 1250  return -1;
1250    
1251    
1252  /*************************************************  /*************************************************
 *            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);  
 }  
   
   
   
 /*************************************************  
1253  *         Read repeat counts                     *  *         Read repeat counts                     *
1254  *************************************************/  *************************************************/
1255    
# Line 987  Returns: pointer to '}' on succe Line 1268  Returns: pointer to '}' on succe
1268                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1269  */  */
1270    
1271  static const uschar *  static const pcre_uchar *
1272  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)
1273  {  {
1274  int min = 0;  int min = 0;
1275  int max = -1;  int max = -1;
# Line 996  int max = -1; Line 1277  int max = -1;
1277  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1278  an integer overflow. */  an integer overflow. */
1279    
1280  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1281  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1282    {    {
1283    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 1011  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1292  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1292    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1293      {      {
1294      max = 0;      max = 0;
1295      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1296      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1297        {        {
1298        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 1044  top-level call starts at the beginning o Line 1325  top-level call starts at the beginning o
1325  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
1326  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
1327  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
1328  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
1329  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1330  first pass. Recursion is used to keep track of subpatterns that reset the  
1331  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1332    that if (?< or (?' or (?P< is encountered, the name will be correctly
1333    terminated because that is checked in the first pass. There is now one call to
1334    this function in the first pass, to check for a recursive back reference by
1335    name (so that we can make the whole group atomic). In this case, we need check
1336    only up to the current position in the pattern, and that is still OK because
1337    and previous occurrences will have been checked. To make this work, the test
1338    for "end of pattern" is a check against cd->end_pattern in the main loop,
1339    instead of looking for a binary zero. This means that the special first-pass
1340    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1341    processing items within the loop are OK, because afterwards the main loop will
1342    terminate.)
1343    
1344  Arguments:  Arguments:
1345    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1055  Arguments: Line 1347  Arguments:
1347    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1348    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1349    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1350      utf          TRUE if we are in UTF-8 / UTF-16 mode
1351    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1352    
1353  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1354  */  */
1355    
1356  static int  static int
1357  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,
1358    BOOL xmode, int *count)    BOOL xmode, BOOL utf, int *count)
1359  {  {
1360  uschar *ptr = *ptrptr;  pcre_uchar *ptr = *ptrptr;
1361  int start_count = *count;  int start_count = *count;
1362  int hwm_count = start_count;  int hwm_count = start_count;
1363  BOOL dup_parens = FALSE;  BOOL dup_parens = FALSE;
# Line 1074  dealing with. The very first call may no Line 1367  dealing with. The very first call may no
1367    
1368  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1369    {    {
1370    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1371        ptr[2] == CHAR_VERTICAL_LINE)  
1372      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1373    
1374      /* Handle a normal, unnamed capturing parenthesis. */
1375    
1376      else if (ptr[1] != CHAR_QUESTION_MARK)
1377        {
1378        *count += 1;
1379        if (name == NULL && *count == lorn) return *count;
1380        ptr++;
1381        }
1382    
1383      /* All cases now have (? at the start. Remember when we are in a group
1384      where the parenthesis numbers are duplicated. */
1385    
1386      else if (ptr[2] == CHAR_VERTICAL_LINE)
1387      {      {
1388      ptr += 3;      ptr += 3;
1389      dup_parens = TRUE;      dup_parens = TRUE;
1390      }      }
1391    
1392    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1393    
1394    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1395      {      {
1396      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1397      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1398      }      }
1399    
1400    /* 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
1401    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
1402    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1403    
1404    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1405      {      {
# Line 1104  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1411  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1411        }        }
1412      }      }
1413    
1414    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1415    
1416    else    else
1417      {      {
# Line 1117  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1424  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1424          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1425        {        {
1426        int term;        int term;
1427        const uschar *thisname;        const pcre_uchar *thisname;
1428        *count += 1;        *count += 1;
1429        if (name == NULL && *count == lorn) return *count;        if (name == NULL && *count == lorn) return *count;
1430        term = *ptr++;        term = *ptr++;
# Line 1125  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1432  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1432        thisname = ptr;        thisname = ptr;
1433        while (*ptr != term) ptr++;        while (*ptr != term) ptr++;
1434        if (name != NULL && lorn == ptr - thisname &&        if (name != NULL && lorn == ptr - thisname &&
1435            strncmp((const char *)name, (const char *)thisname, lorn) == 0)            STRNCMP_UC_UC(name, thisname, lorn) == 0)
1436          return *count;          return *count;
1437        term++;        term++;
1438        }        }
# Line 1133  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1440  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1440    }    }
1441    
1442  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1443  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1444    first-pass call when this value is temporarily adjusted to stop at the current
1445    position. So DO NOT change this to a test for binary zero. */
1446    
1447  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1448    {    {
1449    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1450    
# Line 1166  for (; *ptr != 0; ptr++) Line 1475  for (; *ptr != 0; ptr++)
1475          {          {
1476          if (ptr[2] == CHAR_E)          if (ptr[2] == CHAR_E)
1477            ptr+= 2;            ptr+= 2;
1478          else if (strncmp((const char *)ptr+2,          else if (STRNCMP_UC_C8(ptr + 2,
1479                   STR_Q STR_BACKSLASH STR_E, 3) == 0)                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
1480            ptr += 4;            ptr += 4;
1481          else          else
# Line 1209  for (; *ptr != 0; ptr++) Line 1518  for (; *ptr != 0; ptr++)
1518    
1519    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1520      {      {
1521      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1522        while (*ptr != 0)
1523          {
1524          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1525          ptr++;
1526    #ifdef SUPPORT_UTF
1527          if (utf) FORWARDCHAR(ptr);
1528    #endif
1529          }
1530      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1531      continue;      continue;
1532      }      }
# Line 1218  for (; *ptr != 0; ptr++) Line 1535  for (; *ptr != 0; ptr++)
1535    
1536    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1537      {      {
1538      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1539      if (rc > 0) return rc;      if (rc > 0) return rc;
1540      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1541      }      }
# Line 1226  for (; *ptr != 0; ptr++) Line 1543  for (; *ptr != 0; ptr++)
1543    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1544      {      {
1545      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1546      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1547      }      }
1548    
1549    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1265  Arguments: Line 1581  Arguments:
1581    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1582    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1583    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1584      utf          TRUE if we are in UTF-8 / UTF-16 mode
1585    
1586  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1587  */  */
1588    
1589  static int  static int
1590  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,
1591      BOOL utf)
1592  {  {
1593  uschar *ptr = (uschar *)cd->start_pattern;  pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1594  int count = 0;  int count = 0;
1595  int rc;  int rc;
1596    
# Line 1283  matching closing parens. That is why we Line 1601  matching closing parens. That is why we
1601    
1602  for (;;)  for (;;)
1603    {    {
1604    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1605    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1606    }    }
1607    
# Line 1299  return rc; Line 1617  return rc;
1617    
1618  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1619  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
1620  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
1621  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
1622  assertions, and also the \b assertion; for others it does not.  does not.
1623    
1624  Arguments:  Arguments:
1625    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  
1626    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1627    
1628  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1629  */  */
1630    
1631  static const uschar*  static const pcre_uchar*
1632  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1633  {  {
1634  for (;;)  for (;;)
1635    {    {
1636    switch ((int)*code)    switch ((int)*code)
1637      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1638      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1639      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1640      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1641      if (!skipassert) return code;      if (!skipassert) return code;
1642      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1643      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1644      break;      break;
1645    
1646      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1346  for (;;) Line 1654  for (;;)
1654      case OP_RREF:      case OP_RREF:
1655      case OP_NRREF:      case OP_NRREF:
1656      case OP_DEF:      case OP_DEF:
1657      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1658      break;      break;
1659    
1660      default:      default:
# Line 1376  and doing the check at the end; a flag s Line 1684  and doing the check at the end; a flag s
1684    
1685  Arguments:  Arguments:
1686    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1687    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 mode
1688    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1689    cd       the "compile data" structure    cd       the "compile data" structure
1690    
1691  Returns:   the fixed length,  Returns:   the fixed length,
1692               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1693               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1694               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
1695                 or -4 if an unknown opcode was encountered (internal error)
1696  */  */
1697    
1698  static int  static int
1699  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1700  {  {
1701  int length = -1;  int length = -1;
1702    
1703  register int branchlength = 0;  register int branchlength = 0;
1704  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1705    
1706  /* 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
1707  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1400  branch, check the length against that of Line 1709  branch, check the length against that of
1709  for (;;)  for (;;)
1710    {    {
1711    int d;    int d;
1712    uschar *ce, *cs;    pcre_uchar *ce, *cs;
1713    register int op = *cc;    register int op = *cc;
1714    
1715    switch (op)    switch (op)
1716      {      {
1717        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1718        OP_BRA (normal non-capturing bracket) because the other variants of these
1719        opcodes are all concerned with unlimited repeated groups, which of course
1720        are not of fixed length. */
1721    
1722      case OP_CBRA:      case OP_CBRA:
1723      case OP_BRA:      case OP_BRA:
1724      case OP_ONCE:      case OP_ONCE:
1725        case OP_ONCE_NC:
1726      case OP_COND:      case OP_COND:
1727      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1728      if (d < 0) return d;      if (d < 0) return d;
1729      branchlength += d;      branchlength += d;
1730      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1731      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1732      break;      break;
1733    
1734      /* 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.
1735      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
1736      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
1737        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1738        because they all imply an unlimited repeat. */
1739    
1740      case OP_ALT:      case OP_ALT:
1741      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1742      case OP_END:      case OP_END:
1743        case OP_ACCEPT:
1744        case OP_ASSERT_ACCEPT:
1745      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1746        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1747      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1437  for (;;) Line 1755  for (;;)
1755    
1756      case OP_RECURSE:      case OP_RECURSE:
1757      if (!atend) return -3;      if (!atend) return -3;
1758      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1759      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1760      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                    /* Recursion */
1761      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1762      if (d < 0) return d;      if (d < 0) return d;
1763      branchlength += d;      branchlength += d;
1764      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1453  for (;;) Line 1771  for (;;)
1771      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1772      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1773      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1774      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1775        break;
1776    
1777      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1778    
1779      case OP_REVERSE:      case OP_MARK:
1780        case OP_PRUNE_ARG:
1781        case OP_SKIP_ARG:
1782        case OP_THEN_ARG:
1783        cc += cc[1] + PRIV(OP_lengths)[*cc];
1784        break;
1785    
1786        case OP_CALLOUT:
1787        case OP_CIRC:
1788        case OP_CIRCM:
1789        case OP_CLOSE:
1790        case OP_COMMIT:
1791      case OP_CREF:      case OP_CREF:
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
1792      case OP_DEF:      case OP_DEF:
1793      case OP_OPT:      case OP_DOLL:
1794      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
1795      case OP_EOD:      case OP_EOD:
1796      case OP_EODN:      case OP_EODN:
1797      case OP_CIRC:      case OP_FAIL:
1798      case OP_DOLL:      case OP_NCREF:
1799        case OP_NRREF:
1800      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1801        case OP_PRUNE:
1802        case OP_REVERSE:
1803        case OP_RREF:
1804        case OP_SET_SOM:
1805        case OP_SKIP:
1806        case OP_SOD:
1807        case OP_SOM:
1808        case OP_THEN:
1809      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1810      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1811      break;      break;
1812    
1813      /* Handle literal characters */      /* Handle literal characters */
1814    
1815      case OP_CHAR:      case OP_CHAR:
1816      case OP_CHARNC:      case OP_CHARI:
1817      case OP_NOT:      case OP_NOT:
1818        case OP_NOTI:
1819      branchlength++;      branchlength++;
1820      cc += 2;      cc += 2;
1821  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1822      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1823  #endif  #endif
1824      break;      break;
1825    
# Line 1494  for (;;) Line 1827  for (;;)
1827      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1828    
1829      case OP_EXACT:      case OP_EXACT:
1830        case OP_EXACTI:
1831        case OP_NOTEXACT:
1832        case OP_NOTEXACTI:
1833      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1834      cc += 4;      cc += 2 + IMM2_SIZE;
1835  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1836      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1837  #endif  #endif
1838      break;      break;
1839    
1840      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1841      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1842      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;
1843      cc += 4;      cc += 1 + IMM2_SIZE + 1;
1844      break;      break;
1845    
1846      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1515  for (;;) Line 1850  for (;;)
1850      cc += 2;      cc += 2;
1851      /* Fall through */      /* Fall through */
1852    
1853        case OP_HSPACE:
1854        case OP_VSPACE:
1855        case OP_NOT_HSPACE:
1856        case OP_NOT_VSPACE:
1857      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1858      case OP_DIGIT:      case OP_DIGIT:
1859      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1527  for (;;) Line 1866  for (;;)
1866      cc++;      cc++;
1867      break;      break;
1868    
1869      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1870        otherwise \C is coded as OP_ALLANY. */
1871    
1872      case OP_ANYBYTE:      case OP_ANYBYTE:
1873      return -2;      return -2;
1874    
1875      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1876    
1877  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16
1878      case OP_XCLASS:      case OP_XCLASS:
1879      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1880      /* Fall through */      /* Fall through */
1881  #endif  #endif
1882    
1883      case OP_CLASS:      case OP_CLASS:
1884      case OP_NCLASS:      case OP_NCLASS:
1885      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1886    
1887      switch (*cc)      switch (*cc)
1888        {        {
1889          case OP_CRPLUS:
1890          case OP_CRMINPLUS:
1891        case OP_CRSTAR:        case OP_CRSTAR:
1892        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1893        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1554  for (;;) Line 1896  for (;;)
1896    
1897        case OP_CRRANGE:        case OP_CRRANGE:
1898        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1899        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1900        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1901        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1902        break;        break;
1903    
1904        default:        default:
# Line 1566  for (;;) Line 1908  for (;;)
1908    
1909      /* Anything else is variable length */      /* Anything else is variable length */
1910    
1911      default:      case OP_ANYNL:
1912        case OP_BRAMINZERO:
1913        case OP_BRAPOS:
1914        case OP_BRAPOSZERO:
1915        case OP_BRAZERO:
1916        case OP_CBRAPOS:
1917        case OP_EXTUNI:
1918        case OP_KETRMAX:
1919        case OP_KETRMIN:
1920        case OP_KETRPOS:
1921        case OP_MINPLUS:
1922        case OP_MINPLUSI:
1923        case OP_MINQUERY:
1924        case OP_MINQUERYI:
1925        case OP_MINSTAR:
1926        case OP_MINSTARI:
1927        case OP_MINUPTO:
1928        case OP_MINUPTOI:
1929        case OP_NOTMINPLUS:
1930        case OP_NOTMINPLUSI:
1931        case OP_NOTMINQUERY:
1932        case OP_NOTMINQUERYI:
1933        case OP_NOTMINSTAR:
1934        case OP_NOTMINSTARI:
1935        case OP_NOTMINUPTO:
1936        case OP_NOTMINUPTOI:
1937        case OP_NOTPLUS:
1938        case OP_NOTPLUSI:
1939        case OP_NOTPOSPLUS:
1940        case OP_NOTPOSPLUSI:
1941        case OP_NOTPOSQUERY:
1942        case OP_NOTPOSQUERYI:
1943        case OP_NOTPOSSTAR:
1944        case OP_NOTPOSSTARI:
1945        case OP_NOTPOSUPTO:
1946        case OP_NOTPOSUPTOI:
1947        case OP_NOTQUERY:
1948        case OP_NOTQUERYI:
1949        case OP_NOTSTAR:
1950        case OP_NOTSTARI:
1951        case OP_NOTUPTO:
1952        case OP_NOTUPTOI:
1953        case OP_PLUS:
1954        case OP_PLUSI:
1955        case OP_POSPLUS:
1956        case OP_POSPLUSI:
1957        case OP_POSQUERY:
1958        case OP_POSQUERYI:
1959        case OP_POSSTAR:
1960        case OP_POSSTARI:
1961        case OP_POSUPTO:
1962        case OP_POSUPTOI:
1963        case OP_QUERY:
1964        case OP_QUERYI:
1965        case OP_REF:
1966        case OP_REFI:
1967        case OP_SBRA:
1968        case OP_SBRAPOS:
1969        case OP_SCBRA:
1970        case OP_SCBRAPOS:
1971        case OP_SCOND:
1972        case OP_SKIPZERO:
1973        case OP_STAR:
1974        case OP_STARI:
1975        case OP_TYPEMINPLUS:
1976        case OP_TYPEMINQUERY:
1977        case OP_TYPEMINSTAR:
1978        case OP_TYPEMINUPTO:
1979        case OP_TYPEPLUS:
1980        case OP_TYPEPOSPLUS:
1981        case OP_TYPEPOSQUERY:
1982        case OP_TYPEPOSSTAR:
1983        case OP_TYPEPOSUPTO:
1984        case OP_TYPEQUERY:
1985        case OP_TYPESTAR:
1986        case OP_TYPEUPTO:
1987        case OP_UPTO:
1988        case OP_UPTOI:
1989      return -1;      return -1;
1990    
1991        /* Catch unrecognized opcodes so that when new ones are added they
1992        are not forgotten, as has happened in the past. */
1993    
1994        default:
1995        return -4;
1996      }      }
1997    }    }
1998  /* Control never gets here */  /* Control never gets here */
# Line 1588  length. Line 2013  length.
2013    
2014  Arguments:  Arguments:
2015    code        points to start of expression    code        points to start of expression
2016    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2017    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2018    
2019  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
2020  */  */
2021    
2022  const uschar *  const pcre_uchar *
2023  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2024  {  {
2025  for (;;)  for (;;)
2026    {    {
2027    register int c = *code;    register int c = *code;
2028    
2029    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2030    
2031    /* 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 1612  for (;;) Line 2038  for (;;)
2038    
2039    else if (c == OP_REVERSE)    else if (c == OP_REVERSE)
2040      {      {
2041      if (number < 0) return (uschar *)code;      if (number < 0) return (pcre_uchar *)code;
2042      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2043      }      }
2044    
2045    /* Handle capturing bracket */    /* Handle capturing bracket */
2046    
2047    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2048               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2049      {      {
2050      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2051      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2052      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2053      }      }
2054    
2055    /* 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
# Line 1650  for (;;) Line 2077  for (;;)
2077        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2078        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2079        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2080        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2081            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2082        break;        break;
2083    
2084        case OP_MARK:        case OP_MARK:
2085        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2086        case OP_SKIP_ARG:        case OP_SKIP_ARG:
2087          code += code[1];
2088          break;
2089    
2090        case OP_THEN_ARG:        case OP_THEN_ARG:
2091        code += code[1];        code += code[1];
2092        break;        break;
# Line 1663  for (;;) Line 2094  for (;;)
2094    
2095      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2096    
2097      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2098    
2099    /* 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
2100    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
2101    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2102    
2103  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2104      if (utf8) switch(c)      if (utf) switch(c)
2105        {        {
2106        case OP_CHAR:        case OP_CHAR:
2107        case OP_CHARNC:        case OP_CHARI:
2108        case OP_EXACT:        case OP_EXACT:
2109          case OP_EXACTI:
2110        case OP_UPTO:        case OP_UPTO:
2111          case OP_UPTOI:
2112        case OP_MINUPTO:        case OP_MINUPTO:
2113          case OP_MINUPTOI:
2114        case OP_POSUPTO:        case OP_POSUPTO:
2115          case OP_POSUPTOI:
2116        case OP_STAR:        case OP_STAR:
2117          case OP_STARI:
2118        case OP_MINSTAR:        case OP_MINSTAR:
2119          case OP_MINSTARI:
2120        case OP_POSSTAR:        case OP_POSSTAR:
2121          case OP_POSSTARI:
2122        case OP_PLUS:        case OP_PLUS:
2123          case OP_PLUSI:
2124        case OP_MINPLUS:        case OP_MINPLUS:
2125          case OP_MINPLUSI:
2126        case OP_POSPLUS:        case OP_POSPLUS:
2127          case OP_POSPLUSI:
2128        case OP_QUERY:        case OP_QUERY:
2129          case OP_QUERYI:
2130        case OP_MINQUERY:        case OP_MINQUERY:
2131          case OP_MINQUERYI:
2132        case OP_POSQUERY:        case OP_POSQUERY:
2133        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2134          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2135        break;        break;
2136        }        }
2137  #else  #else
2138      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2139  #endif  #endif
2140      }      }
2141    }    }
# Line 1708  instance of OP_RECURSE. Line 2152  instance of OP_RECURSE.
2152    
2153  Arguments:  Arguments:
2154    code        points to start of expression    code        points to start of expression
2155    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2156    
2157  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
2158  */  */
2159    
2160  static const uschar *  static const pcre_uchar *
2161  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2162  {  {
2163  for (;;)  for (;;)
2164    {    {
# Line 1753  for (;;) Line 2197  for (;;)
2197        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2198        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2199        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2200        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2201            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2202        break;        break;
2203    
2204        case OP_MARK:        case OP_MARK:
2205        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2206        case OP_SKIP_ARG:        case OP_SKIP_ARG:
2207          code += code[1];
2208          break;
2209    
2210        case OP_THEN_ARG:        case OP_THEN_ARG:
2211        code += code[1];        code += code[1];
2212        break;        break;
# Line 1766  for (;;) Line 2214  for (;;)
2214    
2215      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2216    
2217      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2218    
2219      /* 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
2220      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
2221      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2222    
2223  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2224      if (utf8) switch(c)      if (utf) switch(c)
2225        {        {
2226        case OP_CHAR:        case OP_CHAR:
2227        case OP_CHARNC:        case OP_CHARI:
2228          case OP_NOT:
2229          case OP_NOTI:
2230        case OP_EXACT:        case OP_EXACT:
2231          case OP_EXACTI:
2232          case OP_NOTEXACT:
2233          case OP_NOTEXACTI:
2234        case OP_UPTO:        case OP_UPTO:
2235          case OP_UPTOI:
2236          case OP_NOTUPTO:
2237          case OP_NOTUPTOI:
2238        case OP_MINUPTO:        case OP_MINUPTO:
2239          case OP_MINUPTOI:
2240          case OP_NOTMINUPTO:
2241          case OP_NOTMINUPTOI:
2242        case OP_POSUPTO:        case OP_POSUPTO:
2243          case OP_POSUPTOI:
2244          case OP_NOTPOSUPTO:
2245          case OP_NOTPOSUPTOI:
2246        case OP_STAR:        case OP_STAR:
2247          case OP_STARI:
2248          case OP_NOTSTAR:
2249          case OP_NOTSTARI:
2250        case OP_MINSTAR:        case OP_MINSTAR:
2251          case OP_MINSTARI:
2252          case OP_NOTMINSTAR:
2253          case OP_NOTMINSTARI:
2254        case OP_POSSTAR:        case OP_POSSTAR:
2255          case OP_POSSTARI:
2256          case OP_NOTPOSSTAR:
2257          case OP_NOTPOSSTARI:
2258        case OP_PLUS:        case OP_PLUS:
2259          case OP_PLUSI:
2260          case OP_NOTPLUS:
2261          case OP_NOTPLUSI:
2262        case OP_MINPLUS:        case OP_MINPLUS:
2263          case OP_MINPLUSI:
2264          case OP_NOTMINPLUS:
2265          case OP_NOTMINPLUSI:
2266        case OP_POSPLUS:        case OP_POSPLUS:
2267          case OP_POSPLUSI:
2268          case OP_NOTPOSPLUS:
2269          case OP_NOTPOSPLUSI:
2270        case OP_QUERY:        case OP_QUERY:
2271          case OP_QUERYI:
2272          case OP_NOTQUERY:
2273          case OP_NOTQUERYI:
2274        case OP_MINQUERY:        case OP_MINQUERY:
2275          case OP_MINQUERYI:
2276          case OP_NOTMINQUERY:
2277          case OP_NOTMINQUERYI:
2278        case OP_POSQUERY:        case OP_POSQUERY:
2279        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2280          case OP_NOTPOSQUERY:
2281          case OP_NOTPOSQUERYI:
2282          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2283        break;        break;
2284        }        }
2285  #else  #else
2286      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2287  #endif  #endif
2288      }      }
2289    }    }
# Line 1817  bracket whose current branch will alread Line 2306  bracket whose current branch will alread
2306  Arguments:  Arguments:
2307    code        points to start of search    code        points to start of search
2308    endcode     points to where to stop    endcode     points to where to stop
2309    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2310    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2311    
2312  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2313  */  */
2314    
2315  static BOOL  static BOOL
2316  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2317    compile_data *cd)    BOOL utf, compile_data *cd)
2318  {  {
2319  register int c;  register int c;
2320  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2321       code < endcode;       code < endcode;
2322       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2323    {    {
2324    const uschar *ccode;    const pcre_uchar *ccode;
2325    
2326    c = *code;    c = *code;
2327    
# Line 1846  for (code = first_significant_code(code Line 2335  for (code = first_significant_code(code
2335      continue;      continue;
2336      }      }
2337    
   /* Groups with zero repeats can of course be empty; skip them. */  
   
   if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)  
     {  
     code += _pcre_OP_lengths[c];  
     do code += GET(code, 1); while (*code == OP_ALT);  
     c = *code;  
     continue;  
     }  
   
2338    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2339    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2340      forward reference subroutine call, we can't. To detect forward reference
2341      we have to scan up the list that is kept in the workspace. This function is
2342      called only when doing the real compile, not during the pre-compile that
2343      measures the size of the compiled pattern. */
2344    
2345    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2346      {      {
2347      BOOL empty_branch = FALSE;      const pcre_uchar *scode;
2348      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
2349    
2350        /* Test for forward reference */
2351    
2352        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2353          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2354    
2355        /* Not a forward reference, test for completed backward reference */
2356    
2357        empty_branch = FALSE;
2358        scode = cd->start_code + GET(code, 1);
2359      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2360    
2361        /* Completed backwards reference */
2362    
2363      do      do
2364        {        {
2365        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf, cd))
2366          {          {
2367          empty_branch = TRUE;          empty_branch = TRUE;
2368          break;          break;
# Line 1874  for (code = first_significant_code(code Line 2370  for (code = first_significant_code(code
2370        scode += GET(scode, 1);        scode += GET(scode, 1);
2371        }        }
2372      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2373    
2374      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2375      continue;      continue;
2376      }      }
2377    
2378      /* Groups with zero repeats can of course be empty; skip them. */
2379    
2380      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2381          c == OP_BRAPOSZERO)
2382        {
2383        code += PRIV(OP_lengths)[c];
2384        do code += GET(code, 1); while (*code == OP_ALT);
2385        c = *code;
2386        continue;
2387        }
2388    
2389      /* A nested group that is already marked as "could be empty" can just be
2390      skipped. */
2391    
2392      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2393          c == OP_SCBRA || c == OP_SCBRAPOS)
2394        {
2395        do code += GET(code, 1); while (*code == OP_ALT);
2396        c = *code;
2397        continue;
2398        }
2399    
2400    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2401    
2402    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2403          c == OP_CBRA || c == OP_CBRAPOS ||
2404          c == OP_ONCE || c == OP_ONCE_NC ||
2405          c == OP_COND)
2406      {      {
2407      BOOL empty_branch;      BOOL empty_branch;
2408      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1896  for (code = first_significant_code(code Line 2418  for (code = first_significant_code(code
2418        empty_branch = FALSE;        empty_branch = FALSE;
2419        do        do
2420          {          {
2421          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2422            empty_branch = TRUE;            empty_branch = TRUE;
2423          code += GET(code, 1);          code += GET(code, 1);
2424          }          }
# Line 1914  for (code = first_significant_code(code Line 2436  for (code = first_significant_code(code
2436      {      {
2437      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2438      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2439      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2440      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"
2441      here. */      here. */
2442    
2443  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2444      case OP_XCLASS:      case OP_XCLASS:
2445      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2446      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 1926  for (code = first_significant_code(code Line 2448  for (code = first_significant_code(code
2448    
2449      case OP_CLASS:      case OP_CLASS:
2450      case OP_NCLASS:      case OP_NCLASS:
2451      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2452    
2453  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2454      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2455  #endif  #endif
2456    
# Line 1967  for (code = first_significant_code(code Line 2489  for (code = first_significant_code(code
2489      case OP_ALLANY:      case OP_ALLANY:
2490      case OP_ANYBYTE:      case OP_ANYBYTE:
2491      case OP_CHAR:      case OP_CHAR:
2492      case OP_CHARNC:      case OP_CHARI:
2493      case OP_NOT:      case OP_NOT:
2494        case OP_NOTI:
2495      case OP_PLUS:      case OP_PLUS:
2496      case OP_MINPLUS:      case OP_MINPLUS:
2497      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2000  for (code = first_significant_code(code Line 2523  for (code = first_significant_code(code
2523      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2524      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2525      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2526      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP
2527          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2528      break;      break;
2529    
2530      /* End of branch */      /* End of branch */
# Line 2008  for (code = first_significant_code(code Line 2532  for (code = first_significant_code(code
2532      case OP_KET:      case OP_KET:
2533      case OP_KETRMAX:      case OP_KETRMAX:
2534      case OP_KETRMIN:      case OP_KETRMIN:
2535        case OP_KETRPOS:
2536      case OP_ALT:      case OP_ALT:
2537      return TRUE;      return TRUE;
2538    
2539      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2540      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2541    
2542  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2543      case OP_STAR:      case OP_STAR:
2544        case OP_STARI:
2545      case OP_MINSTAR:      case OP_MINSTAR:
2546        case OP_MINSTARI:
2547      case OP_POSSTAR:      case OP_POSSTAR:
2548        case OP_POSSTARI:
2549      case OP_QUERY:      case OP_QUERY:
2550        case OP_QUERYI:
2551      case OP_MINQUERY:      case OP_MINQUERY:
2552        case OP_MINQUERYI:
2553      case OP_POSQUERY:      case OP_POSQUERY:
2554      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      case OP_POSQUERYI:
2555        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2556      break;      break;
2557    
2558      case OP_UPTO:      case OP_UPTO:
2559        case OP_UPTOI:
2560      case OP_MINUPTO:      case OP_MINUPTO:
2561        case OP_MINUPTOI:
2562      case OP_POSUPTO:      case OP_POSUPTO:
2563      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      case OP_POSUPTOI:
2564        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2565      break;      break;
2566  #endif  #endif
2567    
# Line 2037  for (code = first_significant_code(code Line 2571  for (code = first_significant_code(code
2571      case OP_MARK:      case OP_MARK:
2572      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2573      case OP_SKIP_ARG:      case OP_SKIP_ARG:
2574        code += code[1];
2575        break;
2576    
2577      case OP_THEN_ARG:      case OP_THEN_ARG:
2578      code += code[1];      code += code[1];
2579      break;      break;
# Line 2061  return TRUE; Line 2598  return TRUE;
2598  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
2599  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,
2600  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.
2601    This function is called only during the real compile, not during the
2602    pre-compile.
2603    
2604  Arguments:  Arguments:
2605    code        points to start of the recursion    code        points to start of the recursion
2606    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2607    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2608    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2609    cd          pointers to tables etc    cd          pointers to tables etc
2610    
2611  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2612  */  */
2613    
2614  static BOOL  static BOOL
2615  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2616    BOOL utf8, compile_data *cd)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2617  {  {
2618  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2619    {    {
2620    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2621      return FALSE;      return FALSE;
2622    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2623    }    }
# Line 2111  where Perl recognizes it as the POSIX cl Line 2650  where Perl recognizes it as the POSIX cl
2650  "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,
2651  I think.  I think.
2652    
2653    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2654    It seems that the appearance of a nested POSIX class supersedes an apparent
2655    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2656    a digit.
2657    
2658    In Perl, unescaped square brackets may also appear as part of class names. For
2659    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2660    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2661    seem right at all. PCRE does not allow closing square brackets in POSIX class
2662    names.
2663    
2664  Arguments:  Arguments:
2665    ptr      pointer to the initial [    ptr      pointer to the initial [
2666    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2119  Returns: TRUE or FALSE Line 2669  Returns: TRUE or FALSE
2669  */  */
2670    
2671  static BOOL  static BOOL
2672  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2673  {  {
2674  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2675  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2676  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2677    {    {
2678    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2679        ptr++;
2680      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2681      else
2682      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2683      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2684        {        {
2685        *endptr = ptr;        *endptr = ptr;
2686        return TRUE;        return TRUE;
2687        }        }
2688        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2689             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2690              ptr[1] == CHAR_EQUALS_SIGN) &&
2691            check_posix_syntax(ptr, endptr))
2692          return FALSE;
2693      }      }
2694    }    }
2695  return FALSE;  return FALSE;
# Line 2156  Returns: a value representing the na Line 2713  Returns: a value representing the na
2713  */  */
2714    
2715  static int  static int
2716  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2717  {  {
2718  const char *pn = posix_names;  const char *pn = posix_names;
2719  register int yield = 0;  register int yield = 0;
2720  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2721    {    {
2722    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2723      strncmp((const char *)ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2724    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
2725    yield++;    yield++;
2726    }    }
# Line 2195  value in the reference (which is a group Line 2752  value in the reference (which is a group
2752  Arguments:  Arguments:
2753    group      points to the start of the group    group      points to the start of the group
2754    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2755    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 mode
2756    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2757    save_hwm   the hwm forward reference pointer at the start of the group    save_hwm   the hwm forward reference pointer at the start of the group
2758    
# Line 2203  Returns: nothing Line 2760  Returns: nothing
2760  */  */
2761    
2762  static void  static void
2763  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2764    uschar *save_hwm)    pcre_uchar *save_hwm)
2765  {  {
2766  uschar *ptr = group;  pcre_uchar *ptr = group;
2767    
2768  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2769    {    {
2770    int offset;    int offset;
2771    uschar *hc;    pcre_uchar *hc;
2772    
2773    /* 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
2774    reference. */    reference. */
# Line 2256  Arguments: Line 2813  Arguments:
2813  Returns:         new code pointer  Returns:         new code pointer
2814  */  */
2815    
2816  static uschar *  static pcre_uchar *
2817  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2818  {  {
2819  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2820  *code++ = 255;  *code++ = 255;
2821  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2822  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2823  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2824  }  }
2825    
2826    
# Line 2285  Returns: nothing Line 2842  Returns: nothing
2842  */  */
2843    
2844  static void  static void
2845  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2846  {  {
2847  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2848  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2849  }  }
2850    
# Line 2337  for (++c; c <= d; c++) Line 2894  for (++c; c <= d; c++)
2894    
2895  return TRUE;  return TRUE;
2896  }  }
2897    
2898    
2899    
2900    /*************************************************
2901    *        Check a character and a property        *
2902    *************************************************/
2903    
2904    /* This function is called by check_auto_possessive() when a property item
2905    is adjacent to a fixed character.
2906    
2907    Arguments:
2908      c            the character
2909      ptype        the property type
2910      pdata        the data for the type
2911      negated      TRUE if it's a negated property (\P or \p{^)
2912    
2913    Returns:       TRUE if auto-possessifying is OK
2914    */
2915    
2916    static BOOL
2917    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2918    {
2919    const ucd_record *prop = GET_UCD(c);
2920    switch(ptype)
2921      {
2922      case PT_LAMP:
2923      return (prop->chartype == ucp_Lu ||
2924              prop->chartype == ucp_Ll ||
2925              prop->chartype == ucp_Lt) == negated;
2926    
2927      case PT_GC:
2928      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2929    
2930      case PT_PC:
2931      return (pdata == prop->chartype) == negated;
2932    
2933      case PT_SC:
2934      return (pdata == prop->script) == negated;
2935    
2936      /* These are specials */
2937    
2938      case PT_ALNUM:
2939      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2940              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2941    
2942      case PT_SPACE:    /* Perl space */
2943      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2944              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2945              == negated;
2946    
2947      case PT_PXSPACE:  /* POSIX space */
2948      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2949              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2950              c == CHAR_FF || c == CHAR_CR)
2951              == negated;
2952    
2953      case PT_WORD:
2954      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2955              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2956              c == CHAR_UNDERSCORE) == negated;
2957      }
2958    return FALSE;
2959    }
2960  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2961    
2962    
# Line 2350  whether the next thing could possibly ma Line 2970  whether the next thing could possibly ma
2970  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2971    
2972  Arguments:  Arguments:
2973    op_code       the repeated op code    previous      pointer to the repeated opcode
2974    this          data for this item, depends on the opcode    utf           TRUE in UTF-8 / UTF-16 mode
   utf8          TRUE in UTF-8 mode  
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2975    ptr           next character in pattern    ptr           next character in pattern
2976    options       options bits    options       options bits
2977    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2362  Returns: TRUE if possessifying is Line 2980  Returns: TRUE if possessifying is
2980  */  */
2981    
2982  static BOOL  static BOOL
2983  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
2984    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2985  {  {
2986  int next;  pcre_int32 c, next;
2987    int op_code = *previous++;
2988    
2989  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2990    
# Line 2373  if ((options & PCRE_EXTENDED) != 0) Line 2992  if ((options & PCRE_EXTENDED) != 0)
2992    {    {
2993    for (;;)    for (;;)
2994      {      {
2995      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2996      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2997        {        {
2998        while (*(++ptr) != 0)        ptr++;
2999          while (*ptr != 0)
3000            {
3001          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3002            ptr++;
3003    #ifdef SUPPORT_UTF
3004            if (utf) FORWARDCHAR(ptr);
3005    #endif
3006            }
3007        }        }
3008      else break;      else break;
3009      }      }
# Line 2393  if (*ptr == CHAR_BACKSLASH) Line 3019  if (*ptr == CHAR_BACKSLASH)
3019    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
3020    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
3021    }    }
3022    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
3023    {    {
3024  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3025    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
3026  #endif  #endif
3027    next = *ptr++;    next = *ptr++;
3028    }    }
   
3029  else return FALSE;  else return FALSE;
3030    
3031  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2410  if ((options & PCRE_EXTENDED) != 0) Line 3034  if ((options & PCRE_EXTENDED) != 0)
3034    {    {
3035    for (;;)    for (;;)
3036      {      {
3037      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3038      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
3039        {        {
3040        while (*(++ptr) != 0)        ptr++;
3041          while (*ptr != 0)
3042            {
3043          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3044            ptr++;
3045    #ifdef SUPPORT_UTF
3046            if (utf) FORWARDCHAR(ptr);
3047    #endif
3048            }
3049        }        }
3050      else break;      else break;
3051      }      }
# Line 2423  if ((options & PCRE_EXTENDED) != 0) Line 3054  if ((options & PCRE_EXTENDED) != 0)
3054  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3055    
3056  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3057    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)
3058      return FALSE;      return FALSE;
3059    
3060  /* 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
3061  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. */  
3062    
3063  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
3064    {    {
3065    case OP_CHAR:    case OP_CHAR:
3066  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3067    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3068  #else  #else
3069    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
3070  #endif  #endif
3071    return item != next;    return c != next;
3072    
3073    /* 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
3074    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
3075    high-valued characters. */    high-valued characters. */
3076    
3077    case OP_CHARNC:    case OP_CHARI:
3078  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3079    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3080  #endif  #else
3081    if (item == next) return FALSE;    c = *previous;
3082  #ifdef SUPPORT_UTF8  #endif
3083    if (utf8)    if (c == next) return FALSE;
3084    #ifdef SUPPORT_UTF
3085      if (utf)
3086      {      {
3087      unsigned int othercase;      unsigned int othercase;
3088      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
# Line 2463  if (next >= 0) switch(op_code) Line 3091  if (next >= 0) switch(op_code)
3091  #else  #else
3092      othercase = NOTACHAR;      othercase = NOTACHAR;
3093  #endif  #endif
3094      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
3095      }      }
3096    else    else
3097  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3098    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */
   
   /* For OP_NOT, "item" must be a single-byte character. */  
3099    
3100    case OP_NOT:    case OP_NOT:
3101    if (item == next) return TRUE;  #ifdef SUPPORT_UTF
3102    if ((options & PCRE_CASELESS) == 0) return FALSE;    GETCHARTEST(c, previous);
3103  #ifdef SUPPORT_UTF8  #else
3104    if (utf8)    c = *previous;
3105    #endif
3106      return c == next;
3107    
3108      case OP_NOTI:
3109    #ifdef SUPPORT_UTF
3110      GETCHARTEST(c, previous);
3111    #else
3112      c = *previous;
3113    #endif
3114      if (c == next) return TRUE;
3115    #ifdef SUPPORT_UTF
3116      if (utf)
3117      {      {
3118      unsigned int othercase;      unsigned int othercase;
3119      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
3120  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3121      othercase = UCD_OTHERCASE(next);      othercase = UCD_OTHERCASE((unsigned int)next);
3122  #else  #else
3123      othercase = NOTACHAR;      othercase = NOTACHAR;
3124  #endif  #endif
3125      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
3126      }      }
3127    else    else
3128  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3129    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */
3130    
3131      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3132      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3133    
3134    case OP_DIGIT:    case OP_DIGIT:
3135    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2531  if (next >= 0) switch(op_code) Line 3172  if (next >= 0) switch(op_code)
3172      case 0x202f:      case 0x202f:
3173      case 0x205f:      case 0x205f:
3174      case 0x3000:      case 0x3000:
3175      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
3176      default:      default:
3177      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
3178      }      }
3179    
3180      case OP_ANYNL:
3181    case OP_VSPACE:    case OP_VSPACE:
3182    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3183    switch(next)    switch(next)
# Line 2547  if (next >= 0) switch(op_code) Line 3189  if (next >= 0) switch(op_code)
3189      case 0x85:      case 0x85:
3190      case 0x2028:      case 0x2028:
3191      case 0x2029:      case 0x2029:
3192      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
3193      default:      default:
3194      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
3195      }      }
3196    
3197    #ifdef SUPPORT_UCP
3198      case OP_PROP:
3199      return check_char_prop(next, previous[0], previous[1], FALSE);
3200    
3201      case OP_NOTPROP:
3202      return check_char_prop(next, previous[0], previous[1], TRUE);
3203    #endif
3204    
3205    default:    default:
3206    return FALSE;    return FALSE;
3207    }    }
3208    
3209    
3210  /* 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
3211    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3212    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3213    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3214    replaced by OP_PROP codes when PCRE_UCP is set. */
3215    
3216  switch(op_code)  switch(op_code)
3217    {    {
3218    case OP_CHAR:    case OP_CHAR:
3219    case OP_CHARNC:    case OP_CHARI:
3220  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3221    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3222    #else
3223      c = *previous;
3224  #endif  #endif
3225    switch(-next)    switch(-next)
3226      {      {
3227      case ESC_d:      case ESC_d:
3228      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3229    
3230      case ESC_D:      case ESC_D:
3231      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3232    
3233      case ESC_s:      case ESC_s:
3234      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3235    
3236      case ESC_S:      case ESC_S:
3237      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3238    
3239      case ESC_w:      case ESC_w:
3240      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3241    
3242      case ESC_W:      case ESC_W:
3243      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3244    
3245      case ESC_h:      case ESC_h:
3246      case ESC_H:      case ESC_H:
3247      switch(item)      switch(c)
3248        {        {
3249        case 0x09:        case 0x09:
3250        case 0x20:        case 0x20:
# Line 2616  switch(op_code) Line 3272  switch(op_code)
3272    
3273      case ESC_v:      case ESC_v:
3274      case ESC_V:      case ESC_V:
3275      switch(item)      switch(c)
3276        {        {
3277        case 0x0a:        case 0x0a:
3278        case 0x0b:        case 0x0b:
# Line 2630  switch(op_code) Line 3286  switch(op_code)
3286        return -next == ESC_v;        return -next == ESC_v;
3287        }        }
3288    
3289        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3290        their substitutions and process them. The result will always be either
3291        -ESC_p or -ESC_P. Then fall through to process those values. */
3292    
3293    #ifdef SUPPORT_UCP
3294        case ESC_du:
3295        case ESC_DU:
3296        case ESC_wu:
3297        case ESC_WU:
3298        case ESC_su:
3299        case ESC_SU:
3300          {
3301          int temperrorcode = 0;
3302          ptr = substitutes[-next - ESC_DU];
3303          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3304          if (temperrorcode != 0) return FALSE;
3305          ptr++;    /* For compatibility */
3306          }
3307        /* Fall through */
3308    
3309        case ESC_p:
3310        case ESC_P:
3311          {
3312          int ptype, pdata, errorcodeptr;
3313          BOOL negated;
3314    
3315          ptr--;      /* Make ptr point at the p or P */
3316          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3317          if (ptype < 0) return FALSE;
3318          ptr++;      /* Point past the final curly ket */
3319    
3320          /* If the property item is optional, we have to give up. (When generated
3321          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3322          to the original \d etc. At this point, ptr will point to a zero byte. */
3323    
3324          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3325            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3326              return FALSE;
3327    
3328          /* Do the property check. */
3329    
3330          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3331          }
3332    #endif
3333    
3334      default:      default:
3335      return FALSE;      return FALSE;
3336      }      }
3337    
3338      /* In principle, support for Unicode properties should be integrated here as
3339      well. It means re-organizing the above code so as to get hold of the property
3340      values before switching on the op-code. However, I wonder how many patterns
3341      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3342      these op-codes are never generated.) */
3343    
3344    case OP_DIGIT:    case OP_DIGIT:
3345    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3346           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3347    
3348    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3349    return next == -ESC_d;    return next == -ESC_d;
3350    
3351    case OP_WHITESPACE:    case OP_WHITESPACE:
3352    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3353    
3354    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3355    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3356    
3357    case OP_HSPACE:    case OP_HSPACE:
3358    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3359             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3360    
3361    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3362    return next == -ESC_h;    return next == -ESC_h;
3363    
3364    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3365      case OP_ANYNL:
3366    case OP_VSPACE:    case OP_VSPACE:
3367    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3368    
3369    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3370    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3371    
3372    case OP_WORDCHAR:    case OP_WORDCHAR:
3373    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3374             next == -ESC_v || next == -ESC_R;
3375    
3376    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3377    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2690  Arguments: Line 3400  Arguments:
3400    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3401    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3402    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3403    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3404    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3405    bcptr          points to current branch chain    bcptr          points to current branch chain
3406      cond_depth     conditional nesting depth
3407    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3408    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3409                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2702  Returns: TRUE on success Line 3413  Returns: TRUE on success
3413  */  */
3414    
3415  static BOOL  static BOOL
3416  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3417    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3418      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3419    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3420  {  {
3421  int repeat_type, op_type;  int repeat_type, op_type;
3422  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3423  int bravalue = 0;  int bravalue = 0;
3424  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3425  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3426  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3427  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3428  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3429  int after_manual_callout = 0;  int after_manual_callout = 0;
3430  int length_prevgroup = 0;  int length_prevgroup = 0;
3431  register int c;  register int c;
3432  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3433  uschar *last_code = code;  pcre_uchar *last_code = code;
3434  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3435  uschar *tempcode;  pcre_uchar *tempcode;
3436  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3437  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3438  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3439  const uschar *tempptr;  const pcre_uchar *tempptr;
3440  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3441  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3442  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3443  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3444    pcre_uint8 classbits[32];
3445  #ifdef SUPPORT_UTF8  
3446  BOOL class_utf8;  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3447  BOOL utf8 = (options & PCRE_UTF8) != 0;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3448  uschar *class_utf8data;  dynamically as we process the pattern. */
3449  uschar *class_utf8data_base;  
3450  uschar utf8_char[6];  #ifdef SUPPORT_UTF
3451    /* PCRE_UTF16 has the same value as PCRE_UTF8. */
3452    BOOL utf = (options & PCRE_UTF8) != 0;
3453    pcre_uchar utf_chars[6];
3454  #else  #else
3455  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3456  uschar *utf8_char = NULL;  #endif
3457    
3458    /* Helper variables for OP_XCLASS opcode (for characters > 255). */
3459    
3460    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3461    BOOL xclass;
3462    pcre_uchar *class_uchardata;
3463    pcre_uchar *class_uchardata_base;
3464  #endif  #endif
3465    
3466  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 2752  greedy_non_default = greedy_default ^ 1; Line 3474  greedy_non_default = greedy_default ^ 1;
3474    
3475  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3476  matching encountered yet". It gets changed to REQ_NONE if we hit something that  matching encountered yet". It gets changed to REQ_NONE if we hit something that
3477  matches a non-fixed char first char; reqbyte just remains unset if we never  matches a non-fixed char first char; reqchar just remains unset if we never
3478  find one.  find one.
3479    
3480  When we hit a repeat whose minimum is zero, we may have to adjust these values  When we hit a repeat whose minimum is zero, we may have to adjust these values
3481  to take the zero repeat into account. This is implemented by setting them to  to take the zero repeat into account. This is implemented by setting them to
3482  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3483  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3484    
3485  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3486    
3487  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3488  according to the current setting of the caseless flag. REQ_CASELESS is a bit  or zero, according to the current setting of the caseless flag. The
3489  value > 255. It is added into the firstbyte or reqbyte variables to record the  REQ_CASELESS leaves the lower 28 bit empty. It is added into the
3490  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3491    value. This is used only for ASCII characters. */
3492    
3493  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3494    
3495  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3496    
# Line 2779  for (;; ptr++) Line 3502  for (;; ptr++)
3502    BOOL is_quantifier;    BOOL is_quantifier;
3503    BOOL is_recurse;    BOOL is_recurse;
3504    BOOL reset_bracount;    BOOL reset_bracount;
3505    int class_charcount;    int class_has_8bitchar;
3506    int class_lastchar;    int class_single_char;
3507    int newoptions;    int newoptions;
3508    int recno;    int recno;
3509    int refsign;    int refsign;
3510    int skipbytes;    int skipbytes;
3511    int subreqbyte;    int subreqchar;
3512    int subfirstbyte;    int subfirstchar;
3513    int terminator;    int terminator;
3514    int mclength;    int mclength;
3515    uschar mcbuffer[8];    int tempbracount;
3516      pcre_uchar mcbuffer[8];
3517    
3518    /* Get next byte in the pattern */    /* Get next character in the pattern */
3519    
3520    c = *ptr;    c = *ptr;
3521    
3522      /* If we are at the end of a nested substitution, revert to the outer level
3523      string. Nesting only happens one level deep. */
3524    
3525      if (c == 0 && nestptr != NULL)
3526        {
3527        ptr = nestptr;
3528        nestptr = NULL;
3529        c = *ptr;
3530        }
3531    
3532    /* 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
3533    previous cycle of this loop. */    previous cycle of this loop. */
3534    
# Line 2803  for (;; ptr++) Line 3537  for (;; ptr++)
3537  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
3538      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3539  #endif  #endif
3540      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3541            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3542        {        {
3543        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3544        goto FAILED;        goto FAILED;
# Line 2825  for (;; ptr++) Line 3560  for (;; ptr++)
3560        goto FAILED;        goto FAILED;
3561        }        }
3562    
3563      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3564      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3565          (int)(code - last_code), c, c));
3566    
3567      /* 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
3568      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 2836  for (;; ptr++) Line 3572  for (;; ptr++)
3572        {        {
3573        if (previous > orig_code)        if (previous > orig_code)
3574          {          {
3575          memmove(orig_code, previous, code - previous);          memmove(orig_code, previous, IN_UCHARS(code - previous));
3576          code -= previous - orig_code;          code -= previous - orig_code;
3577          previous = orig_code;          previous = orig_code;
3578          }          }
# Line 2852  for (;; ptr++) Line 3588  for (;; ptr++)
3588    /* 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
3589    reference list. */    reference list. */
3590    
3591    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3592               WORK_SIZE_SAFETY_MARGIN)
3593      {      {
3594      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3595      goto FAILED;      goto FAILED;
# Line 2900  for (;; ptr++) Line 3637  for (;; ptr++)
3637      previous_callout = NULL;      previous_callout = NULL;
3638      }      }
3639    
3640    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3641    
3642    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3643      {      {
3644      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
3645      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3646        {        {
3647        while (*(++ptr) != 0)        ptr++;
3648          while (*ptr != 0)
3649          {          {
3650          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3651            ptr++;
3652    #ifdef SUPPORT_UTF
3653            if (utf) FORWARDCHAR(ptr);
3654    #endif
3655          }          }
3656        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3657    
# Line 2932  for (;; ptr++) Line 3674  for (;; ptr++)
3674      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3675      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3676      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
3677      *firstbyteptr = firstbyte;      *firstcharptr = firstchar;
3678      *reqbyteptr = reqbyte;      *reqcharptr = reqchar;
3679      *codeptr = code;      *codeptr = code;
3680      *ptrptr = ptr;      *ptrptr = ptr;
3681      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 2943  for (;; ptr++) Line 3685  for (;; ptr++)
3685          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3686          goto FAILED;          goto FAILED;
3687          }          }
3688        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3689        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3690        }        }
3691      return TRUE;      return TRUE;
# Line 2954  for (;; ptr++) Line 3696  for (;; ptr++)
3696      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3697    
3698      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3699        previous = NULL;
3700      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3701        {        {
3702        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3703          *code++ = OP_CIRCM;
3704        }        }
3705      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3706      break;      break;
3707    
3708      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3709      previous = NULL;      previous = NULL;
3710      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3711      break;      break;
3712    
3713      /* 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
3714      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqchar doesn't change either. */
3715    
3716      case CHAR_DOT:      case CHAR_DOT:
3717      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3718      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
3719      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
3720      previous = code;      previous = code;
3721      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3722      break;      break;
# Line 3028  for (;; ptr++) Line 3771  for (;; ptr++)
3771          {          {
3772          if (ptr[1] == CHAR_E)          if (ptr[1] == CHAR_E)
3773            ptr++;            ptr++;
3774          else if (strncmp((const char *)ptr+1,          else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
                           STR_Q STR_BACKSLASH STR_E, 3) == 0)  
3775            ptr += 3;            ptr += 3;
3776          else          else
3777            break;            break;
# Line 3048  for (;; ptr++) Line 3790  for (;; ptr++)
3790          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3791        {        {
3792        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3793        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3794        zerofirstbyte = firstbyte;        zerofirstchar = firstchar;
3795        break;        break;
3796        }        }
3797    
# Line 3059  for (;; ptr++) Line 3801  for (;; ptr++)
3801    
3802      should_flip_negation = FALSE;      should_flip_negation = FALSE;
3803    
3804      /* Keep a count of chars with values < 256 so that we can optimize the case      /* For optimization purposes, we track some properties of the class.
3805      of just a single character (as long as it's < 256). However, For higher      class_has_8bitchar will be non-zero, if the class contains at least one
3806      valued UTF-8 characters, we don't yet do any optimization. */      < 256 character. class_single_char will be 1 if the class contains only
3807        a single character. */
3808    
3809      class_charcount = 0;      class_has_8bitchar = 0;
3810      class_lastchar = -1;      class_single_char = 0;
3811    
3812      /* Initialize the 32-char bit map to all zeros. We build the map in a      /* Initialize the 32-char bit map to all zeros. We build the map in a
3813      temporary bit of memory, in case the class contains only 1 character (less      temporary bit of memory, in case the class contains only 1 character (less
3814      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.
3815      */      */
3816    
3817      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
3818    
3819  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3820      class_utf8 = FALSE;                       /* No chars >= 256 */      xclass = FALSE;                           /* No chars >= 256 */
3821      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_uchardata = code + LINK_SIZE + 2;   /* For UTF-8 items */
3822      class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */      class_uchardata_base = class_uchardata;   /* For resetting in pass 1 */
3823  #endif  #endif
3824    
3825      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 3085  for (;; ptr++) Line 3828  for (;; ptr++)
3828    
3829      if (c != 0) do      if (c != 0) do
3830        {        {
3831        const uschar *oldptr;        const pcre_uchar *oldptr;
3832    
3833  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3834        if (utf8 && c > 127)        if (utf && HAS_EXTRALEN(c))
3835          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3836          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3837          }          }
3838    #endif
3839    
3840        /* In the pre-compile phase, accumulate the length of any UTF-8 extra  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3841          /* In the pre-compile phase, accumulate the length of any extra
3842        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
3843        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
3844        (which is on the stack). */        (which is on the stack). */
3845    
3846        if (lengthptr != NULL)        if (lengthptr != NULL)
3847          {          {
3848          *lengthptr += class_utf8data - class_utf8data_base;          *lengthptr += class_uchardata - class_uchardata_base;
3849          class_utf8data = class_utf8data_base;          class_uchardata = class_uchardata_base;
3850          }          }
   
3851  #endif  #endif
3852    
3853        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 3131  for (;; ptr++) Line 3875  for (;; ptr++)
3875          {          {
3876          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3877          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3878          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
3879          uschar pbits[32];          pcre_uint8 pbits[32];
3880    
3881          if (ptr[1] != CHAR_COLON)          if (ptr[1] != CHAR_COLON)
3882            {            {
# Line 3148  for (;; ptr++) Line 3892  for (;; ptr++)
3892            ptr++;            ptr++;
3893            }            }
3894    
3895          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3896          if (posix_class < 0)          if (posix_class < 0)
3897            {            {
3898            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3162  for (;; ptr++) Line 3906  for (;; ptr++)
3906          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3907            posix_class = 0;            posix_class = 0;
3908    
3909          /* 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
3910          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3911          subtract bits that may be in the main map already. At the end we or the  
3912          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3913            if ((options & PCRE_UCP) != 0)
3914              {
3915              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3916              if (posix_substitutes[pc] != NULL)
3917                {
3918                nestptr = tempptr + 1;
3919                ptr = posix_substitutes[pc] - 1;
3920                continue;
3921                }
3922              }
3923    #endif
3924            /* In the non-UCP case, we build the bit map for the POSIX class in a
3925            chunk of local store because we may be adding and subtracting from it,
3926            and we don't want to subtract bits that may be in the main map already.
3927            At the end we or the result into the bit map that is being built. */
3928    
3929          posix_class *= 3;          posix_class *= 3;
3930    
3931          /* Copy in the first table (always present) */          /* Copy in the first table (always present) */
3932    
3933          memcpy(pbits, cbits + posix_class_maps[posix_class],          memcpy(pbits, cbits + posix_class_maps[posix_class],
3934            32 * sizeof(uschar));            32 * sizeof(pcre_uint8));
3935    
3936          /* If there is a second table, add or remove it as required. */          /* If there is a second table, add or remove it as required. */
3937    
# Line 3203  for (;; ptr++) Line 3962  for (;; ptr++)
3962            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3963    
3964          ptr = tempptr + 1;          ptr = tempptr + 1;
3965          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          /* Every class contains at least one < 256 characters. */
3966            class_has_8bitchar = 1;
3967            /* Every class contains at least two characters. */
3968            class_single_char = 2;
3969          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3970          }          }
3971    
3972        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3973        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
3974        case. Inside a class (and only there) it is treated as backspace. We        case. Inside a class (and only there) it is treated as backspace. We
3975        assume that other escapes have more than one character in them, so set        assume that other escapes have more than one character in them, so
3976        class_charcount bigger than one. Unrecognized escapes fall through and        speculatively set both class_has_8bitchar and class_single_char bigger
3977        are either treated as literal characters (by default), or are faulted if        than one. Unrecognized escapes fall through and are either treated
3978          as literal characters (by default), or are faulted if
3979        PCRE_EXTRA is set. */        PCRE_EXTRA is set. */
3980    
3981        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
# Line 3221  for (;; ptr++) Line 3984  for (;; ptr++)
3984          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3985    
3986          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 */
3987            else if (-c == ESC_N)            /* \N is not supported in a class */
3988              {
3989              *errorcodeptr = ERR71;
3990              goto FAILED;
3991              }
3992          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3993            {            {
3994            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3234  for (;; ptr++) Line 4002  for (;; ptr++)
4002    
4003          if (c < 0)          if (c < 0)
4004            {            {
4005            register const uschar *cbits = cd->cbits;            register const pcre_uint8 *cbits = cd->cbits;
4006            class_charcount += 2;     /* Greater than 1 is what matters */            /* Every class contains at least two < 256 characters. */
4007              class_has_8bitchar++;
4008            /* Save time by not doing this in the pre-compile phase. */            /* Every class contains at least two characters. */
4009              class_single_char += 2;
4010    
4011            if (lengthptr == NULL) switch (-c)            switch (-c)
4012              {              {
4013    #ifdef SUPPORT_UCP
4014                case ESC_du:     /* These are the values given for \d etc */
4015                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
4016                case ESC_wu:     /* escape sequence with an appropriate \p */
4017                case ESC_WU:     /* or \P to test Unicode properties instead */
4018                case ESC_su:     /* of the default ASCII testing. */
4019                case ESC_SU:
4020                nestptr = ptr;
4021                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
4022                class_has_8bitchar--;                /* Undo! */
4023                continue;
4024    #endif
4025              case ESC_d:              case ESC_d:
4026              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
4027              continue;              continue;
# Line 3259  for (;; ptr++) Line 4040  for (;; ptr++)
4040              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
4041              continue;              continue;
4042    
4043                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
4044                if it was previously set by something earlier in the character
4045                class. */
4046    
4047              case ESC_s:              case ESC_s:
4048              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
4049              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
4050                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
4051              continue;              continue;
4052    
4053              case ESC_S:              case ESC_S:
# Line 3270  for (;; ptr++) Line 4056  for (;; ptr++)
4056              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
4057              continue;              continue;
4058    
4059              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)  
             {  
4060              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
4061              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
4062              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
4063  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
4064              if (utf8)              xclass = TRUE;
4065                *class_uchardata++ = XCL_SINGLE;
4066                *class_uchardata++ = 0x1680;
4067                *class_uchardata++ = XCL_SINGLE;
4068                *class_uchardata++ = 0x180e;
4069                *class_uchardata++ = XCL_RANGE;
4070                *class_uchardata++ = 0x2000;
4071                *class_uchardata++ = 0x200a;
4072                *class_uchardata++ = XCL_SINGLE;
4073                *class_uchardata++ = 0x202f;
4074                *class_uchardata++ = XCL_SINGLE;
4075                *class_uchardata++ = 0x205f;
4076                *class_uchardata++ = XCL_SINGLE;
4077                *class_uchardata++ = 0x3000;
4078    #elif defined SUPPORT_UTF
4079                if (utf)
4080                {                {
4081                class_utf8 = TRUE;                xclass = TRUE;
4082                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4083                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1680, class_uchardata);
4084                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4085                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180e, class_uchardata);
4086                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4087                class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2000, class_uchardata);
4088                class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x200a, class_uchardata);
4089                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4090                class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x202f, class_uchardata);
4091                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4092                class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x205f, class_uchardata);
4093                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4094                class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x3000, class_uchardata);
4095                }                }
4096  #endif  #endif
4097              continue;              continue;
             }  
4098    
4099            if (-c == ESC_H)              case ESC_H:
             {  
4100              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
4101                {                {
4102                int x = 0xff;                int x = 0xff;
# Line 3323  for (;; ptr++) Line 4109  for (;; ptr++)
4109                  }                  }
4110                classbits[c] |= x;                classbits[c] |= x;
4111                }                }
4112    #ifndef COMPILE_PCRE8
4113  #ifdef SUPPORT_UTF8              xclass = TRUE;
4114              if (utf8)              *class_uchardata++ = XCL_RANGE;
4115                *class_uchardata++ = 0x0100;
4116                *class_uchardata++ = 0x167f;
4117                *class_uchardata++ = XCL_RANGE;
4118                *class_uchardata++ = 0x1681;
4119                *class_uchardata++ = 0x180d;
4120                *class_uchardata++ = XCL_RANGE;
4121                *class_uchardata++ = 0x180f;
4122                *class_uchardata++ = 0x1fff;
4123                *class_uchardata++ = XCL_RANGE;
4124                *class_uchardata++ = 0x200b;
4125                *class_uchardata++ = 0x202e;
4126                *class_uchardata++ = XCL_RANGE;
4127                *class_uchardata++ = 0x2030;
4128                *class_uchardata++ = 0x205e;
4129                *class_uchardata++ = XCL_RANGE;
4130                *class_uchardata++ = 0x2060;
4131                *class_uchardata++ = 0x2fff;
4132                *class_uchardata++ = XCL_RANGE;
4133                *class_uchardata++ = 0x3001;
4134    #ifdef SUPPORT_UTF
4135                if (utf)
4136                  class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4137                else
4138    #endif
4139                  *class_uchardata++ = 0xffff;
4140    #elif defined SUPPORT_UTF
4141                if (utf)
4142                {                {
4143                class_utf8 = TRUE;                xclass = TRUE;
4144                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4145                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);
4146                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x167f, class_uchardata);
4147                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4148                class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1681, class_uchardata);
4149                class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180d, class_uchardata);
4150                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4151                class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180f, class_uchardata);
4152                class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1fff, class_uchardata);
4153                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4154                class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x200b, class_uchardata);
4155                class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x202e, class_uchardata);
4156                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4157                class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2030, class_uchardata);
4158                class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x205e, class_uchardata);
4159                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4160                class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2060, class_uchardata);
4161                class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2fff, class_uchardata);
4162                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4163                class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x3001, class_uchardata);
4164                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4165                }                }
4166  #endif  #endif
4167              continue;              continue;
             }  
4168    
4169            if (-c == ESC_v)              case ESC_v:
             {  
4170              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
4171              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
4172              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
4173              SETBIT(classbits, 0x0d); /* CR */              SETBIT(classbits, 0x0d); /* CR */
4174              SETBIT(classbits, 0x85); /* NEL */              SETBIT(classbits, 0x85); /* NEL */
4175  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
4176              if (utf8)              xclass = TRUE;
4177                *class_uchardata++ = XCL_RANGE;
4178                *class_uchardata++ = 0x2028;
4179                *class_uchardata++ = 0x2029;
4180    #elif defined SUPPORT_UTF
4181                if (utf)
4182                {                {
4183                class_utf8 = TRUE;                xclass = TRUE;
4184                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4185                class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2028, class_uchardata);
4186                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2029, class_uchardata);
4187                }                }
4188  #endif  #endif
4189              continue;              continue;
             }  
4190    
4191            if (-c == ESC_V)              case ESC_V:
             {  
4192              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
4193                {                {
4194                int x = 0xff;                int x = 0xff;
# Line 3391  for (;; ptr++) Line 4205  for (;; ptr++)
4205                classbits[c] |= x;                classbits[c] |= x;
4206                }                }
4207    
4208  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
4209              if (utf8)              xclass = TRUE;
4210                *class_uchardata++ = XCL_RANGE;
4211                *class_uchardata++ = 0x0100;
4212                *class_uchardata++ = 0x2027;
4213                *class_uchardata++ = XCL_RANGE;
4214                *class_uchardata++ = 0x202a;
4215    #ifdef SUPPORT_UTF
4216                if (utf)
4217                  class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4218                else
4219    #endif
4220                  *class_uchardata++ = 0xffff;
4221    #elif defined SUPPORT_UTF
4222                if (utf)
4223                {                {
4224                class_utf8 = TRUE;                xclass = TRUE;
4225                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4226                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);
4227                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2027, class_uchardata);
4228                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4229                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x202a, class_uchardata);
4230                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4231                }                }
4232  #endif  #endif
4233              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
4234    
4235  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4236            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
4237              {              case ESC_P:
4238              BOOL negated;                {
4239              int pdata;                BOOL negated;
4240              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
4241              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
4242              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
4243              *class_utf8data++ = ((-c == ESC_p) != negated)?                xclass = TRUE;
4244                XCL_PROP : XCL_NOTPROP;                *class_uchardata++ = ((-c == ESC_p) != negated)?
4245              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
4246              *class_utf8data++ = pdata;                *class_uchardata++ = ptype;
4247              class_charcount -= 2;   /* Not a < 256 character */                *class_uchardata++ = pdata;
4248              continue;                class_has_8bitchar--;                /* Undo! */
4249              }                continue;
4250                  }
4251  #endif  #endif
4252            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
4253            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
4254            treated as literals. */              treated as literals. */
4255    
4256            if ((options & PCRE_EXTRA) != 0)              default:
4257              {              if ((options & PCRE_EXTRA) != 0)
4258              *errorcodeptr = ERR7;                {
4259              goto FAILED;                *errorcodeptr = ERR7;
4260                  goto FAILED;
4261                  }
4262                class_has_8bitchar--;    /* Undo the speculative increase. */
4263                class_single_char -= 2;  /* Undo the speculative increase. */
4264                c = *ptr;                /* Get the final character and fall through */
4265                break;
4266              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
4267            }            }
4268    
4269          /* 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
4270          greater than 256 in UTF-8 mode. */          greater than 256. */
4271    
4272          }   /* End of backslash handling */          }   /* End of backslash handling */
4273    
# Line 3487  for (;; ptr++) Line 4315  for (;; ptr++)
4315            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
4316            }            }
4317    
4318  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4319          if (utf8)          if (utf)
4320            {                           /* Braces are required because the */            {                           /* Braces are required because the */
4321            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */
4322            }            }
# Line 3532  for (;; ptr++) Line 4360  for (;; ptr++)
4360    
4361          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4362    
4363            /* Since we found a character range, single character optimizations
4364            cannot be done anymore. */
4365            class_single_char = 2;
4366    
4367          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
4368          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
4369          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
4370          available. */          available. */
4371    
4372  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !(defined COMPILE_PCRE8)
4373          if (utf8 && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))          if ((d > 255) || (utf && ((options & PCRE_CASELESS) != 0 && d > 127)))
4374    #elif defined  SUPPORT_UTF
4375            if (utf && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))
4376    #elif !(defined COMPILE_PCRE8)
4377            if (d > 255)
4378    #endif
4379    #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
4380            {            {
4381            class_utf8 = TRUE;            xclass = TRUE;
4382    
4383            /* With UCP support, we can find the other case equivalents of            /* With UCP support, we can find the other case equivalents of
4384            the relevant characters. There may be several ranges. Optimize how            the relevant characters. There may be several ranges. Optimize how
4385            they fit with the basic range. */            they fit with the basic range. */
4386    
4387  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4388    #ifndef COMPILE_PCRE8
4389              if (utf && (options & PCRE_CASELESS) != 0)
4390    #else
4391            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
4392    #endif
4393              {              {
4394              unsigned int occ, ocd;              unsigned int occ, ocd;
4395              unsigned int cc = c;              unsigned int cc = c;
# Line 3573  for (;; ptr++) Line 4415  for (;; ptr++)
4415    
4416                if (occ == ocd)                if (occ == ocd)
4417                  {                  {
4418                  *class_utf8data++ = XCL_SINGLE;                  *class_uchardata++ = XCL_SINGLE;
4419                  }                  }
4420                else                else
4421                  {                  {
4422                  *class_utf8data++ = XCL_RANGE;                  *class_uchardata++ = XCL_RANGE;
4423                  class_utf8data += _pcre_ord2utf8(occ, class_utf8data);                  class_uchardata += PRIV(ord2utf)(occ, class_uchardata);
4424                  }                  }
4425                class_utf8data += _pcre_ord2utf8(ocd, class_utf8data);                class_uchardata += PRIV(ord2utf)(ocd, class_uchardata);
4426                }                }
4427              }              }
4428  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
# Line 3588  for (;; ptr++) Line 4430  for (;; ptr++)
4430            /* Now record the original range, possibly modified for UCP caseless            /* Now record the original range, possibly modified for UCP caseless
4431            overlapping ranges. */            overlapping ranges. */
4432    
4433            *class_utf8data++ = XCL_RANGE;            *class_uchardata++ = XCL_RANGE;
4434            class_utf8data += _pcre_ord2utf8(c, class_utf8data);  #ifdef SUPPORT_UTF
4435            class_utf8data += _pcre_ord2utf8(d, class_utf8data);  #ifndef COMPILE_PCRE8
4436              if (utf)
4437                {
4438                class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4439                class_uchardata += PRIV(ord2utf)(d, class_uchardata);
4440                }
4441              else
4442                {
4443                *class_uchardata++ = c;
4444                *class_uchardata++ = d;
4445                }
4446    #else
4447              class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4448              class_uchardata += PRIV(ord2utf)(d, class_uchardata);
4449    #endif
4450    #else /* SUPPORT_UTF */
4451              *class_uchardata++ = c;
4452              *class_uchardata++ = d;
4453    #endif /* SUPPORT_UTF */
4454    
4455            /* With UCP support, we are done. Without UCP support, there is no            /* With UCP support, we are done. Without UCP support, there is no
4456            caseless matching for UTF-8 characters > 127; we can use the bit map            caseless matching for UTF characters > 127; we can use the bit map
4457            for the smaller ones. */            for the smaller ones. As for 16 bit characters without UTF, we
4458              can still use  */
4459    
4460  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4461            continue;    /* With next character in the class */  #ifndef COMPILE_PCRE8
4462  #else            if (utf)
4463            if ((options & PCRE_CASELESS) == 0 || c > 127) continue;  #endif
4464                continue;    /* With next character in the class */
4465    #endif  /* SUPPORT_UCP */
4466    
4467    #if defined SUPPORT_UTF && !defined(SUPPORT_UCP) && !(defined COMPILE_PCRE8)
4468              if (utf)
4469                {
4470                if ((options & PCRE_CASELESS) == 0 || c > 127) continue;
4471                /* Adjust upper limit and fall through to set up the map */
4472                d = 127;
4473                }
4474              else
4475                {
4476                if (c > 255) continue;
4477                /* Adjust upper limit and fall through to set up the map */
4478                d = 255;
4479                }
4480    #elif defined SUPPORT_UTF && !defined(SUPPORT_UCP)
4481              if ((options & PCRE_CASELESS) == 0 || c > 127) continue;
4482            /* Adjust upper limit and fall through to set up the map */            /* Adjust upper limit and fall through to set up the map */
   
4483            d = 127;            d = 127;
4484    #else
4485  #endif  /* SUPPORT_UCP */            if (c > 255) continue;
4486              /* Adjust upper limit and fall through to set up the map */
4487              d = 255;
4488    #endif  /* SUPPORT_UTF && !SUPPORT_UCP && !COMPILE_PCRE8 */
4489            }            }
4490  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF || !COMPILE_PCRE8 */
4491    
4492          /* We use the bit map for all cases when not in UTF-8 mode; else          /* We use the bit map for 8 bit mode, or when the characters fall
4493          ranges that lie entirely within 0-127 when there is UCP support; else          partially or entirely to [0-255] ([0-127] for UCP) ranges. */
         for partial ranges without UCP support. */  
4494    
4495          class_charcount += d - c + 1;          class_has_8bitchar = 1;
         class_lastchar = d;  
4496    
4497          /* We can save a bit of time by skipping this in the pre-compile. */          /* We can save a bit of time by skipping this in the pre-compile. */
4498    
# Line 3623  for (;; ptr++) Line 4501  for (;; ptr++)
4501            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
4502            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
4503              {              {
4504              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c]; /* flip case */
4505              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
4506              }              }
4507            }            }
# Line 3637  for (;; ptr++) Line 4515  for (;; ptr++)
4515    
4516        LONE_SINGLE_CHARACTER:        LONE_SINGLE_CHARACTER:
4517    
4518        /* Handle a character that cannot go in the bit map */        /* Only the value of 1 matters for class_single_char. */
4519          if (class_single_char < 2) class_single_char++;
4520    
4521          /* If class_charcount is 1, we saw precisely one character. As long as
4522          there was no use of \p or \P, in other words, no use of any XCLASS features,
4523          we can optimize.
4524    
4525          The optimization throws away the bit map. We turn the item into a
4526          1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4527          In the positive case, it can cause firstchar to be set. Otherwise, there
4528          can be no first char if this item is first, whatever repeat count may
4529          follow. In the case of reqchar, save the previous value for reinstating. */
4530    
4531  #ifdef SUPPORT_UTF8        if (class_single_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
       if (utf8 && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))  
4532          {          {
4533          class_utf8 = TRUE;          ptr++;
4534          *class_utf8data++ = XCL_SINGLE;          zeroreqchar = reqchar;
4535          class_utf8data += _pcre_ord2utf8(c, class_utf8data);  
4536            /* The OP_NOT[I] opcodes work on single characters only. */
4537    
4538            if (negate_class)
4539              {
4540              if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4541              zerofirstchar = firstchar;
4542              *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4543    #ifdef SUPPORT_UTF
4544              if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
4545                code += PRIV(ord2utf)(c, code);
4546              else
4547    #endif
4548                *code++ = c;
4549              goto NOT_CHAR;
4550              }
4551    
4552            /* For a single, positive character, get the value into mcbuffer, and
4553            then we can handle this with the normal one-character code. */
4554    
4555    #ifdef SUPPORT_UTF
4556            if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
4557              mclength = PRIV(ord2utf)(c, mcbuffer);
4558            else
4559    #endif
4560              {
4561              mcbuffer[0] = c;
4562              mclength = 1;
4563              }
4564            goto ONE_CHAR;
4565            }       /* End of 1-char optimization */
4566    
4567          /* Handle a character that cannot go in the bit map. */
4568    
4569    #if defined SUPPORT_UTF && !(defined COMPILE_PCRE8)
4570          if ((c > 255) || (utf && ((options & PCRE_CASELESS) != 0 && c > 127)))
4571    #elif defined SUPPORT_UTF
4572          if (utf && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))
4573    #elif !(defined COMPILE_PCRE8)
4574          if (c > 255)
4575    #endif
4576    
4577    #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
4578            {
4579            xclass = TRUE;
4580            *class_uchardata++ = XCL_SINGLE;
4581    #ifdef SUPPORT_UTF
4582    #ifndef COMPILE_PCRE8
4583            /* In non 8 bit mode, we can get here even if we are not in UTF mode. */
4584            if (!utf)
4585              *class_uchardata++ = c;
4586            else
4587    #endif
4588              class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4589    #else /* SUPPORT_UTF */
4590            *class_uchardata++ = c;
4591    #endif /* SUPPORT_UTF */
4592    
4593  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4594    #ifdef COMPILE_PCRE8
4595          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4596    #else
4597            /* In non 8 bit mode, we can get here even if we are not in UTF mode. */
4598            if (utf && (options & PCRE_CASELESS) != 0)
4599    #endif
4600            {            {
4601            unsigned int othercase;            unsigned int othercase;
4602            if ((othercase = UCD_OTHERCASE(c)) != c)            if ((int)(othercase = UCD_OTHERCASE(c)) != c)
4603              {              {
4604              *class_utf8data++ = XCL_SINGLE;              *class_uchardata++ = XCL_SINGLE;
4605              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_uchardata += PRIV(ord2utf)(othercase, class_uchardata);
4606              }              }
4607            }            }
4608  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4609    
4610          }          }
4611        else        else
4612  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF || COMPILE_PCRE16 */
4613    
4614        /* Handle a single-byte character */        /* Handle a single-byte character */
4615          {          {
4616            class_has_8bitchar = 1;
4617          classbits[c/8] |= (1 << (c&7));          classbits[c/8] |= (1 << (c&7));
4618          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4619            {            {
4620            c = cd->fcc[c];   /* flip case */            c = cd->fcc[c]; /* flip case */
4621            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
4622            }            }
         class_charcount++;  
         class_lastchar = c;  
4623          }          }
4624        }        }
4625    
4626      /* 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.
4627        If we are at the end of an internal nested string, revert to the outer
4628        string. */
4629    
4630        while (((c = *(++ptr)) != 0 ||
4631               (nestptr != NULL &&
4632                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4633               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4634    
4635      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4636    
4637      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4638        {        {
4639        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4640        goto FAILED;        goto FAILED;
4641        }        }
4642    
4643        /* If this is the first thing in the branch, there can be no first char
4644  /* This code has been disabled because it would mean that \s counts as      setting, whatever the repeat count. Any reqchar setting must remain
4645  an explicit \r or \n reference, and that's not really what is wanted. Now      unchanged after any kind of repeat. */
4646  we set the flag only if there is a literal "\r" or "\n" in the class. */  
4647        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4648  #if 0      zerofirstchar = firstchar;
4649      /* Remember whether \r or \n are in this class */      zeroreqchar = reqchar;
   
     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  
   
   
     /* If class_charcount is 1, we saw precisely one character whose value is  
     less than 256. As long as there were no characters >= 128 and there was no  
     use of \p or \P, in other words, no use of any XCLASS features, we can  
     optimize.  
   
     In UTF-8 mode, we can optimize the negative case only if there were no  
     characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR  
     operate on single-bytes only. This is an historical hangover. Maybe one day  
     we can tidy these opcodes to handle multi-byte characters.  
   
     The optimization throws away the bit map. We turn the item into a  
     1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note  
     that OP_NOT does not support multibyte characters. In the positive case, it  
     can cause firstbyte to be set. Otherwise, there can be no first char if  
     this item is first, whatever repeat count may follow. In the case of  
     reqbyte, save the previous value for reinstating. */  
   
 #ifdef SUPPORT_UTF8  
     if (class_charcount == 1 && !class_utf8 &&  
       (!utf8 || !negate_class || class_lastchar < 128))  
 #else  
     if (class_charcount == 1)  
 #endif  
       {  
       zeroreqbyte = reqbyte;  
   
       /* The OP_NOT opcode works on one-byte characters only. */  
   
       if (negate_class)  
         {  
         if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  
         zerofirstbyte = firstbyte;  
         *code++ = OP_NOT;  
         *code++ = class_lastchar;  
         break;  
         }  
   
       /* For a single, positive character, get the value into mcbuffer, and  
       then we can handle this with the normal one-character code. */  
   
 #ifdef SUPPORT_UTF8  
       if (utf8 && class_lastchar > 127)  
         mclength = _pcre_ord2utf8(class_lastchar, mcbuffer);  
       else  
 #endif  
         {  
         mcbuffer[0] = class_lastchar;  
         mclength = 1;  
         }  
       goto ONE_CHAR;  
       }       /* End of 1-char optimization */  
   
     /* The general case - not the one-char optimization. If this is the first  
     thing in the branch, there can be no first char setting, whatever the  
     repeat count. Any reqbyte setting must remain unchanged after any kind of  
     repeat. */  
   
     if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  
     zerofirstbyte = firstbyte;  
     zeroreqbyte = reqbyte;  
4650    
4651      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4652      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4653      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
4654      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
4655      (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
4656      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
4657        actual compiled code. */
4658  #ifdef SUPPORT_UTF8  
4659      if (class_utf8 && !should_flip_negation)  #ifdef SUPPORT_UTF
4660        if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
4661    #elif !defined COMPILE_PCRE8
4662        if (xclass && !should_flip_negation)
4663    #endif
4664    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4665        {        {
4666        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_uchardata++ = XCL_END;    /* Marks the end of extra data */
4667        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
4668        code += LINK_SIZE;        code += LINK_SIZE;
4669        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT:0;
4670    
4671        /* If the map is required, move up the extra data to make room for it;        /* If the map is required, move up the extra data to make room for it;
4672        otherwise just move the code pointer to the end of the extra data. */        otherwise just move the code pointer to the end of the extra data. */
4673    
4674        if (class_charcount > 0)        if (class_has_8bitchar > 0)
4675          {          {
4676          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
4677          memmove(code + 32, code, class_utf8data - code);          memmove(code + (32 / sizeof(pcre_uchar)), code,
4678              IN_UCHARS(class_uchardata - code));
4679          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
4680          code = class_utf8data + 32;          code = class_uchardata + (32 / sizeof(pcre_uchar));
4681          }          }