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

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

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

code/trunk/pcre_compile.c revision 574 by ph10, Sat Nov 20 17:47:27 2010 UTC code/branches/pcre16/pcre_compile.c revision 806 by zherczeg, Thu Dec 15 11:57:39 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2010 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When 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 266  substitutes must be in the order of the Line 285  substitutes must be in the order of the
285  both positive and negative cases. NULL means no substitute. */  both positive and negative cases. NULL means no substitute. */
286    
287  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
288  static const uschar *substitutes[] = {  static const pcre_uchar string_PNd[]  = {
289    (uschar *)"\\P{Nd}",    /* \D */    CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
290    (uschar *)"\\p{Nd}",    /* \d */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
291    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */  static const pcre_uchar string_pNd[]  = {
292    (uschar *)"\\p{Xsp}",   /* \s */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
293    (uschar *)"\\P{Xwd}",   /* \W */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
294    (uschar *)"\\p{Xwd}"    /* \w */  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 uschar *posix_substitutes[] = {  static const pcre_uchar string_pL[] =   {
317    (uschar *)"\\p{L}",     /* alpha */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
318    (uschar *)"\\p{Ll}",    /* lower */    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
319    (uschar *)"\\p{Lu}",    /* upper */  static const pcre_uchar string_pLl[] =  {
320    (uschar *)"\\p{Xan}",   /* alnum */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
321    NULL,                   /* ascii */    CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
322    (uschar *)"\\h",        /* blank */  static const pcre_uchar string_pLu[] =  {
323    NULL,                   /* cntrl */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
324    (uschar *)"\\p{Nd}",    /* digit */    CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
325    NULL,                   /* graph */  static const pcre_uchar string_pXan[] = {
326    NULL,                   /* print */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
327    NULL,                   /* punct */    CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
328    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */  static const pcre_uchar string_h[] =    {
329    (uschar *)"\\p{Xwd}",   /* word */    CHAR_BACKSLASH, CHAR_h, '\0' };
330    NULL,                   /* xdigit */  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 */    /* Negated cases */
367    (uschar *)"\\P{L}",     /* ^alpha */    string_PL,            /* ^alpha */
368    (uschar *)"\\P{Ll}",    /* ^lower */    string_PLl,           /* ^lower */
369    (uschar *)"\\P{Lu}",    /* ^upper */    string_PLu,           /* ^upper */
370    (uschar *)"\\P{Xan}",   /* ^alnum */    string_PXan,          /* ^alnum */
371    NULL,                   /* ^ascii */    NULL,                 /* ^ascii */
372    (uschar *)"\\H",        /* ^blank */    string_H,             /* ^blank */
373    NULL,                   /* ^cntrl */    NULL,                 /* ^cntrl */
374    (uschar *)"\\P{Nd}",    /* ^digit */    string_PNd,           /* ^digit */
375    NULL,                   /* ^graph */    NULL,                 /* ^graph */
376    NULL,                   /* ^print */    NULL,                 /* ^print */
377    NULL,                   /* ^punct */    NULL,                 /* ^punct */
378    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
379    (uschar *)"\\P{Xwd}",   /* ^word */    string_PXwd,          /* ^word */
380    NULL                    /* ^xdigit */    NULL                  /* ^xdigit */
381  };  };
382  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
383  #endif  #endif
384    
385  #define STRING(a)  # a  #define STRING(a)  # a
# Line 393  static const char error_texts[] = Line 466  static const char error_texts[] =
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 408  static const char error_texts[] = Line 481  static const char error_texts[] =
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 PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
484    "\\c must be followed by an ASCII character\0"    "\\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 UTF-8/16 code point (>= 0xd800 && <= 0xdfff)\0"
491    ;    ;
492    
493  /* 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 427  For convenience, we use the same bit def Line 506  For convenience, we use the same bit def
506    
507  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
508    
509    /* Using a simple comparison for decimal numbers rather than a memory read
510    is much faster, and the resulting code is simpler (the compiler turns it
511    into a subtraction and unsigned comparison). */
512    
513    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
514    
515  #ifndef EBCDIC  #ifndef EBCDIC
516    
517  /* 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
518  UTF-8 mode. */  UTF-8 mode. */
519    
520  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
521    {    {
522    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
523    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 471  static const unsigned char digitab[] = Line 556  static const unsigned char digitab[] =
556    
557  /* 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. */
558    
559  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
560    {    {
561    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
562    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 506  static const unsigned char digitab[] = Line 591  static const unsigned char digitab[] =
591    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
592    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
593    
594  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
595    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
596    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
597    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 545  static const unsigned char ebcdic_charta Line 630  static const unsigned char ebcdic_charta
630  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
631    
632  static BOOL  static BOOL
633    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
634      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
635    
636    
# Line 577  return s; Line 662  return s;
662    
663    
664  /*************************************************  /*************************************************
665    *           Expand the workspace                 *
666    *************************************************/
667    
668    /* This function is called during the second compiling phase, if the number of
669    forward references fills the existing workspace, which is originally a block on
670    the stack. A larger block is obtained from malloc() unless the ultimate limit
671    has been reached or the increase will be rather small.
672    
673    Argument: pointer to the compile data block
674    Returns:  0 if all went well, else an error number
675    */
676    
677    static int
678    expand_workspace(compile_data *cd)
679    {
680    pcre_uchar *newspace;
681    int newsize = cd->workspace_size * 2;
682    
683    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
684    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
685        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
686     return ERR72;
687    
688    newspace = (PUBL(malloc))(newsize);
689    if (newspace == NULL) return ERR21;
690    
691    memcpy(newspace, cd->start_workspace, cd->workspace_size);
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 601  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 620  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 632  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 642  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 664  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 674  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 694  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 747  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 784  else Line 998  else
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;      if (!utf && c > 0xff) *errorcodeptr = ERR51;
1002      break;      break;
1003    
1004      /* \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
1005      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.
1006      treated as a data character. */      If not, { is treated as a data character. */
1007    
1008      case CHAR_x:      case CHAR_x:
1009        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1010          {
1011          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1012          Otherwise it is a lowercase x letter. */
1013          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1014            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1015            {
1016            c = 0;
1017            for (i = 0; i < 2; ++i)
1018              {
1019              register int cc = *(++ptr);
1020    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1021              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1022              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1023    #else           /* EBCDIC coding */
1024              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1025              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1026    #endif
1027              }
1028            }
1029          break;
1030          }
1031    
1032      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1033        {        {
1034        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1035    
1036        c = 0;        c = 0;
1037        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1038          {          {
1039          register int cc = *pt++;          register int cc = *pt++;
1040          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
1041    
1042  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1043          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
# Line 811  else Line 1046  else
1046          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 */
1047          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1048  #endif  #endif
1049    
1050    #ifdef COMPILE_PCRE8
1051            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1052    #else
1053    #ifdef COMPILE_PCRE16
1054            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1055    #endif
1056    #endif
1057            }
1058    
1059          if (c < 0)
1060            {
1061            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1062            *errorcodeptr = ERR34;
1063          }          }
1064    
1065        if (*pt == CHAR_RIGHT_CURLY_BRACKET)        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1066          {          {
1067          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1068          ptr = pt;          ptr = pt;
1069          break;          break;
1070          }          }
# Line 827  else Line 1076  else
1076      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1077    
1078      c = 0;      c = 0;
1079      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1080        {        {
1081        int cc;                                  /* Some compilers don't like */        int cc;                                  /* Some compilers don't like */
1082        cc = *(++ptr);                           /* ++ in initializers */        cc = *(++ptr);                           /* ++ in initializers */
# Line 857  else Line 1106  else
1106      if (c > 127)  /* Excludes all non-ASCII in either mode */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1107        {        {
1108        *errorcodeptr = ERR68;        *errorcodeptr = ERR68;
1109        break;        break;
1110        }        }
1111      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1112      c ^= 0x40;      c ^= 0x40;
1113  #else             /* EBCDIC coding */  #else             /* EBCDIC coding */
# Line 885  else Line 1134  else
1134    }    }
1135    
1136  /* Perl supports \N{name} for character names, as well as plain \N for "not  /* Perl supports \N{name} for character names, as well as plain \N for "not
1137  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
1138    quantification such as \N{2,3}. */
1139    
1140  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1141         !is_counted_repeat(ptr+2))
1142    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1143    
1144  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
# Line 923  Returns:         type value from ucp_typ Line 1174  Returns:         type value from ucp_typ
1174  */  */
1175    
1176  static int  static int
1177  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1178  {  {
1179  int c, i, bot, top;  int c, i, bot, top;
1180  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1181  char name[32];  pcre_uchar name[32];
1182    
1183  c = *(++ptr);  c = *(++ptr);
1184  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 968  else Line 1219  else
1219  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1220    
1221  bot = 0;  bot = 0;
1222  top = _pcre_utt_size;  top = PRIV(utt_size);
1223    
1224  while (bot < top)  while (bot < top)
1225    {    {
1226    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1227    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1228    if (c == 0)    if (c == 0)
1229      {      {
1230      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1231      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1232      }      }
1233    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1234    }    }
# Line 997  return -1; Line 1248  return -1;
1248    
1249    
1250  /*************************************************  /*************************************************
 *            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);  
 }  
   
   
   
 /*************************************************  
1251  *         Read repeat counts                     *  *         Read repeat counts                     *
1252  *************************************************/  *************************************************/
1253    
# Line 1048  Returns:         pointer to '}' on succe Line 1266  Returns:         pointer to '}' on succe
1266                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1267  */  */
1268    
1269  static const uschar *  static const pcre_uchar *
1270  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)
1271  {  {
1272  int min = 0;  int min = 0;
1273  int max = -1;  int max = -1;
# Line 1057  int max = -1; Line 1275  int max = -1;
1275  /* 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
1276  an integer overflow. */  an integer overflow. */
1277    
1278  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1279  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1280    {    {
1281    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 1072  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1290  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1290    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1291      {      {
1292      max = 0;      max = 0;
1293      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1294      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1295        {        {
1296        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 1105  top-level call starts at the beginning o Line 1323  top-level call starts at the beginning o
1323  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
1324  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
1325  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
1326  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
1327  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1328  first pass. Recursion is used to keep track of subpatterns that reset the  
1329  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1330    that if (?< or (?' or (?P< is encountered, the name will be correctly
1331    terminated because that is checked in the first pass. There is now one call to
1332    this function in the first pass, to check for a recursive back reference by
1333    name (so that we can make the whole group atomic). In this case, we need check
1334    only up to the current position in the pattern, and that is still OK because
1335    and previous occurrences will have been checked. To make this work, the test
1336    for "end of pattern" is a check against cd->end_pattern in the main loop,
1337    instead of looking for a binary zero. This means that the special first-pass
1338    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1339    processing items within the loop are OK, because afterwards the main loop will
1340    terminate.)
1341    
1342  Arguments:  Arguments:
1343    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1116  Arguments: Line 1345  Arguments:
1345    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1346    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1347    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1348    utf8         TRUE if we are in UTF-8 mode    utf          TRUE if we are in UTF-8 / UTF-16 mode
1349    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1350    
1351  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1352  */  */
1353    
1354  static int  static int
1355  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,
1356    BOOL xmode, BOOL utf8, int *count)    BOOL xmode, BOOL utf, int *count)
1357  {  {
1358  uschar *ptr = *ptrptr;  pcre_uchar *ptr = *ptrptr;
1359  int start_count = *count;  int start_count = *count;
1360  int hwm_count = start_count;  int hwm_count = start_count;
1361  BOOL dup_parens = FALSE;  BOOL dup_parens = FALSE;
# Line 1193  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1422  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1422          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1423        {        {
1424        int term;        int term;
1425        const uschar *thisname;        const pcre_uchar *thisname;
1426        *count += 1;        *count += 1;
1427        if (name == NULL && *count == lorn) return *count;        if (name == NULL && *count == lorn) return *count;
1428        term = *ptr++;        term = *ptr++;
# Line 1201  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1430  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1430        thisname = ptr;        thisname = ptr;
1431        while (*ptr != term) ptr++;        while (*ptr != term) ptr++;
1432        if (name != NULL && lorn == ptr - thisname &&        if (name != NULL && lorn == ptr - thisname &&
1433            strncmp((const char *)name, (const char *)thisname, lorn) == 0)            STRNCMP_UC_UC(name, thisname, lorn) == 0)
1434          return *count;          return *count;
1435        term++;        term++;
1436        }        }
# Line 1209  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1438  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1438    }    }
1439    
1440  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1441  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1442    first-pass call when this value is temporarily adjusted to stop at the current
1443    position. So DO NOT change this to a test for binary zero. */
1444    
1445  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1446    {    {
1447    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1448    
# Line 1242  for (; *ptr != 0; ptr++) Line 1473  for (; *ptr != 0; ptr++)
1473          {          {
1474          if (ptr[2] == CHAR_E)          if (ptr[2] == CHAR_E)
1475            ptr+= 2;            ptr+= 2;
1476          else if (strncmp((const char *)ptr+2,          else if (STRNCMP_UC_C8(ptr + 2,
1477                   STR_Q STR_BACKSLASH STR_E, 3) == 0)                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
1478            ptr += 4;            ptr += 4;
1479          else          else
# Line 1285  for (; *ptr != 0; ptr++) Line 1516  for (; *ptr != 0; ptr++)
1516    
1517    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1518      {      {
1519      ptr++;      ptr++;
1520      while (*ptr != 0)      while (*ptr != 0)
1521        {        {
1522        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1523        ptr++;        ptr++;
1524  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1525        if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;        if (utf) FORWARDCHAR(ptr);
1526  #endif  #endif
1527        }        }
1528      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
# Line 1302  for (; *ptr != 0; ptr++) Line 1533  for (; *ptr != 0; ptr++)
1533    
1534    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1535      {      {
1536      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1537      if (rc > 0) return rc;      if (rc > 0) return rc;
1538      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1539      }      }
# Line 1348  Arguments: Line 1579  Arguments:
1579    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1580    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1581    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1582    utf8         TRUE if we are in UTF-8 mode    utf          TRUE if we are in UTF-8 / UTF-16 mode
1583    
1584  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1585  */  */
1586    
1587  static int  static int
1588  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,
1589    BOOL utf8)    BOOL utf)
1590  {  {
1591  uschar *ptr = (uschar *)cd->start_pattern;  pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1592  int count = 0;  int count = 0;
1593  int rc;  int rc;
1594    
# Line 1368  matching closing parens. That is why we Line 1599  matching closing parens. That is why we
1599    
1600  for (;;)  for (;;)
1601    {    {
1602    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1603    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1604    }    }
1605    
# Line 1384  return rc; Line 1615  return rc;
1615    
1616  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1617  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
1618  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
1619  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
1620  assertions, and also the \b assertion; for others it does not.  does not.
1621    
1622  Arguments:  Arguments:
1623    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  
1624    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1625    
1626  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1627  */  */
1628    
1629  static const uschar*  static const pcre_uchar*
1630  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1631  {  {
1632  for (;;)  for (;;)
1633    {    {
1634    switch ((int)*code)    switch ((int)*code)
1635      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1636      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1637      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1638      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1639      if (!skipassert) return code;      if (!skipassert) return code;
1640      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1641      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1642      break;      break;
1643    
1644      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1431  for (;;) Line 1652  for (;;)
1652      case OP_RREF:      case OP_RREF:
1653      case OP_NRREF:      case OP_NRREF:
1654      case OP_DEF:      case OP_DEF:
1655      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1656      break;      break;
1657    
1658      default:      default:
# Line 1461  and doing the check at the end; a flag s Line 1682  and doing the check at the end; a flag s
1682    
1683  Arguments:  Arguments:
1684    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1685    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 mode
1686    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1687    cd       the "compile data" structure    cd       the "compile data" structure
1688    
1689  Returns:   the fixed length,  Returns:   the fixed length,
1690               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1691               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1692               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
1693                 or -4 if an unknown opcode was encountered (internal error)
1694  */  */
1695    
1696  static int  static int
1697  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1698  {  {
1699  int length = -1;  int length = -1;
1700    
1701  register int branchlength = 0;  register int branchlength = 0;
1702  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1703    
1704  /* 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
1705  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1485  branch, check the length against that of Line 1707  branch, check the length against that of
1707  for (;;)  for (;;)
1708    {    {
1709    int d;    int d;
1710    uschar *ce, *cs;    pcre_uchar *ce, *cs;
1711    register int op = *cc;    register int op = *cc;
1712    switch (op)    switch (op)
1713      {      {
1714        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1715        OP_BRA (normal non-capturing bracket) because the other variants of these
1716        opcodes are all concerned with unlimited repeated groups, which of course
1717        are not of fixed length. */
1718    
1719      case OP_CBRA:      case OP_CBRA:
1720      case OP_BRA:      case OP_BRA:
1721      case OP_ONCE:      case OP_ONCE:
1722        case OP_ONCE_NC:
1723      case OP_COND:      case OP_COND:
1724      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);
1725      if (d < 0) return d;      if (d < 0) return d;
1726      branchlength += d;      branchlength += d;
1727      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1728      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1729      break;      break;
1730    
1731      /* 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.
1732      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
1733      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
1734        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1735        because they all imply an unlimited repeat. */
1736    
1737      case OP_ALT:      case OP_ALT:
1738      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1739      case OP_END:      case OP_END:
1740        case OP_ACCEPT:
1741        case OP_ASSERT_ACCEPT:
1742      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1743        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1744      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1522  for (;;) Line 1752  for (;;)
1752    
1753      case OP_RECURSE:      case OP_RECURSE:
1754      if (!atend) return -3;      if (!atend) return -3;
1755      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1756      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1757      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                    /* Recursion */
1758      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf, atend, cd);
1759      if (d < 0) return d;      if (d < 0) return d;
1760      branchlength += d;      branchlength += d;
1761      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1542  for (;;) Line 1772  for (;;)
1772    
1773      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1774    
1775      case OP_REVERSE:      case OP_MARK:
1776        case OP_PRUNE_ARG:
1777        case OP_SKIP_ARG:
1778        case OP_THEN_ARG:
1779        cc += cc[1] + PRIV(OP_lengths)[*cc];
1780        break;
1781    
1782        case OP_CALLOUT:
1783        case OP_CIRC:
1784        case OP_CIRCM:
1785        case OP_CLOSE:
1786        case OP_COMMIT:
1787      case OP_CREF:      case OP_CREF:
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
1788      case OP_DEF:      case OP_DEF:
1789      case OP_OPT:      case OP_DOLL:
1790      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
1791      case OP_EOD:      case OP_EOD:
1792      case OP_EODN:      case OP_EODN:
1793      case OP_CIRC:      case OP_FAIL:
1794      case OP_DOLL:      case OP_NCREF:
1795        case OP_NRREF:
1796      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1797        case OP_PRUNE:
1798        case OP_REVERSE:
1799        case OP_RREF:
1800        case OP_SET_SOM:
1801        case OP_SKIP:
1802        case OP_SOD:
1803        case OP_SOM:
1804        case OP_THEN:
1805      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1806      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1807      break;      break;
1808    
1809      /* Handle literal characters */      /* Handle literal characters */
1810    
1811      case OP_CHAR:      case OP_CHAR:
1812      case OP_CHARNC:      case OP_CHARI:
1813      case OP_NOT:      case OP_NOT:
1814        case OP_NOTI:
1815      branchlength++;      branchlength++;
1816      cc += 2;      cc += 2;
1817  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1818      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];  
1819  #endif  #endif
1820      break;      break;
1821    
# Line 1579  for (;;) Line 1823  for (;;)
1823      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1824    
1825      case OP_EXACT:      case OP_EXACT:
1826        case OP_EXACTI:
1827        case OP_NOTEXACT:
1828        case OP_NOTEXACTI:
1829      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1830      cc += 4;      cc += 2 + IMM2_SIZE;
1831  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1832      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];  
1833  #endif  #endif
1834      break;      break;
1835    
1836      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1837      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1838      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;
1839      cc += 4;      cc += 1 + IMM2_SIZE + 1;
1840      break;      break;
1841    
1842      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1600  for (;;) Line 1846  for (;;)
1846      cc += 2;      cc += 2;
1847      /* Fall through */      /* Fall through */
1848    
1849        case OP_HSPACE:
1850        case OP_VSPACE:
1851        case OP_NOT_HSPACE:
1852        case OP_NOT_VSPACE:
1853      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1854      case OP_DIGIT:      case OP_DIGIT:
1855      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1612  for (;;) Line 1862  for (;;)
1862      cc++;      cc++;
1863      break;      break;
1864    
1865      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1866        otherwise \C is coded as OP_ALLANY. */
1867    
1868      case OP_ANYBYTE:      case OP_ANYBYTE:
1869      return -2;      return -2;
1870    
1871      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1872    
1873  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16
1874      case OP_XCLASS:      case OP_XCLASS:
1875      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1876      /* Fall through */      /* Fall through */
1877  #endif  #endif
1878    
1879      case OP_CLASS:      case OP_CLASS:
1880      case OP_NCLASS:      case OP_NCLASS:
1881      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1882    
1883      switch (*cc)      switch (*cc)
1884        {        {
1885          case OP_CRPLUS:
1886          case OP_CRMINPLUS:
1887        case OP_CRSTAR:        case OP_CRSTAR:
1888        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1889        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1639  for (;;) Line 1892  for (;;)
1892    
1893        case OP_CRRANGE:        case OP_CRRANGE:
1894        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1895        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1896        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1897        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1898        break;        break;
1899    
1900        default:        default:
# Line 1651  for (;;) Line 1904  for (;;)
1904    
1905      /* Anything else is variable length */      /* Anything else is variable length */
1906    
1907      default:      case OP_ANYNL:
1908        case OP_BRAMINZERO:
1909        case OP_BRAPOS:
1910        case OP_BRAPOSZERO:
1911        case OP_BRAZERO:
1912        case OP_CBRAPOS:
1913        case OP_EXTUNI:
1914        case OP_KETRMAX:
1915        case OP_KETRMIN:
1916        case OP_KETRPOS:
1917        case OP_MINPLUS:
1918        case OP_MINPLUSI:
1919        case OP_MINQUERY:
1920        case OP_MINQUERYI:
1921        case OP_MINSTAR:
1922        case OP_MINSTARI:
1923        case OP_MINUPTO:
1924        case OP_MINUPTOI:
1925        case OP_NOTMINPLUS:
1926        case OP_NOTMINPLUSI:
1927        case OP_NOTMINQUERY:
1928        case OP_NOTMINQUERYI:
1929        case OP_NOTMINSTAR:
1930        case OP_NOTMINSTARI:
1931        case OP_NOTMINUPTO:
1932        case OP_NOTMINUPTOI:
1933        case OP_NOTPLUS:
1934        case OP_NOTPLUSI:
1935        case OP_NOTPOSPLUS:
1936        case OP_NOTPOSPLUSI:
1937        case OP_NOTPOSQUERY:
1938        case OP_NOTPOSQUERYI:
1939        case OP_NOTPOSSTAR:
1940        case OP_NOTPOSSTARI:
1941        case OP_NOTPOSUPTO:
1942        case OP_NOTPOSUPTOI:
1943        case OP_NOTQUERY:
1944        case OP_NOTQUERYI:
1945        case OP_NOTSTAR:
1946        case OP_NOTSTARI:
1947        case OP_NOTUPTO:
1948        case OP_NOTUPTOI:
1949        case OP_PLUS:
1950        case OP_PLUSI:
1951        case OP_POSPLUS:
1952        case OP_POSPLUSI:
1953        case OP_POSQUERY:
1954        case OP_POSQUERYI:
1955        case OP_POSSTAR:
1956        case OP_POSSTARI:
1957        case OP_POSUPTO:
1958        case OP_POSUPTOI:
1959        case OP_QUERY:
1960        case OP_QUERYI:
1961        case OP_REF:
1962        case OP_REFI:
1963        case OP_SBRA:
1964        case OP_SBRAPOS:
1965        case OP_SCBRA:
1966        case OP_SCBRAPOS:
1967        case OP_SCOND:
1968        case OP_SKIPZERO:
1969        case OP_STAR:
1970        case OP_STARI:
1971        case OP_TYPEMINPLUS:
1972        case OP_TYPEMINQUERY:
1973        case OP_TYPEMINSTAR:
1974        case OP_TYPEMINUPTO:
1975        case OP_TYPEPLUS:
1976        case OP_TYPEPOSPLUS:
1977        case OP_TYPEPOSQUERY:
1978        case OP_TYPEPOSSTAR:
1979        case OP_TYPEPOSUPTO:
1980        case OP_TYPEQUERY:
1981        case OP_TYPESTAR:
1982        case OP_TYPEUPTO:
1983        case OP_UPTO:
1984        case OP_UPTOI:
1985      return -1;      return -1;
1986    
1987        /* Catch unrecognized opcodes so that when new ones are added they
1988        are not forgotten, as has happened in the past. */
1989    
1990        default:
1991        return -4;
1992      }      }
1993    }    }
1994  /* Control never gets here */  /* Control never gets here */
# Line 1673  length. Line 2009  length.
2009    
2010  Arguments:  Arguments:
2011    code        points to start of expression    code        points to start of expression
2012    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2013    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2014    
2015  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
2016  */  */
2017    
2018  const uschar *  const pcre_uchar *
2019  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2020  {  {
2021  for (;;)  for (;;)
2022    {    {
2023    register int c = *code;    register int c = *code;
2024    
2025    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2026    
2027    /* 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 1697  for (;;) Line 2034  for (;;)
2034    
2035    else if (c == OP_REVERSE)    else if (c == OP_REVERSE)
2036      {      {
2037      if (number < 0) return (uschar *)code;      if (number < 0) return (pcre_uchar *)code;
2038      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2039      }      }
2040    
2041    /* Handle capturing bracket */    /* Handle capturing bracket */
2042    
2043    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2044               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2045      {      {
2046      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2047      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2048      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2049      }      }
2050    
2051    /* 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 1735  for (;;) Line 2073  for (;;)
2073        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2074        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2075        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2076        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2077            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2078        break;        break;
2079    
2080        case OP_MARK:        case OP_MARK:
# Line 1745  for (;;) Line 2084  for (;;)
2084        break;        break;
2085    
2086        case OP_THEN_ARG:        case OP_THEN_ARG:
2087        code += code[1+LINK_SIZE];        code += code[1];
2088        break;        break;
2089        }        }
2090    
2091      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2092    
2093      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2094    
2095    /* 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
2096    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
2097    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2098    
2099  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2100      if (utf8) switch(c)      if (utf) switch(c)
2101        {        {
2102        case OP_CHAR:        case OP_CHAR:
2103        case OP_CHARNC:        case OP_CHARI:
2104        case OP_EXACT:        case OP_EXACT:
2105          case OP_EXACTI:
2106        case OP_UPTO:        case OP_UPTO:
2107          case OP_UPTOI:
2108        case OP_MINUPTO:        case OP_MINUPTO:
2109          case OP_MINUPTOI:
2110        case OP_POSUPTO:        case OP_POSUPTO:
2111          case OP_POSUPTOI:
2112        case OP_STAR:        case OP_STAR:
2113          case OP_STARI:
2114        case OP_MINSTAR:        case OP_MINSTAR:
2115          case OP_MINSTARI:
2116        case OP_POSSTAR:        case OP_POSSTAR:
2117          case OP_POSSTARI:
2118        case OP_PLUS:        case OP_PLUS:
2119          case OP_PLUSI:
2120        case OP_MINPLUS:        case OP_MINPLUS:
2121          case OP_MINPLUSI:
2122        case OP_POSPLUS:        case OP_POSPLUS:
2123          case OP_POSPLUSI:
2124        case OP_QUERY:        case OP_QUERY:
2125          case OP_QUERYI:
2126        case OP_MINQUERY:        case OP_MINQUERY:
2127          case OP_MINQUERYI:
2128        case OP_POSQUERY:        case OP_POSQUERY:
2129        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2130          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2131        break;        break;
2132        }        }
2133  #else  #else
2134      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2135  #endif  #endif
2136      }      }
2137    }    }
# Line 1796  instance of OP_RECURSE. Line 2148  instance of OP_RECURSE.
2148    
2149  Arguments:  Arguments:
2150    code        points to start of expression    code        points to start of expression
2151    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2152    
2153  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
2154  */  */
2155    
2156  static const uschar *  static const pcre_uchar *
2157  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2158  {  {
2159  for (;;)  for (;;)
2160    {    {
# Line 1841  for (;;) Line 2193  for (;;)
2193        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2194        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2195        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2196        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2197            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2198        break;        break;
2199    
2200        case OP_MARK:        case OP_MARK:
# Line 1851  for (;;) Line 2204  for (;;)
2204        break;        break;
2205    
2206        case OP_THEN_ARG:        case OP_THEN_ARG:
2207        code += code[1+LINK_SIZE];        code += code[1];
2208        break;        break;
2209        }        }
2210    
2211      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2212    
2213      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2214    
2215      /* 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
2216      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
2217      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2218    
2219  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2220      if (utf8) switch(c)      if (utf) switch(c)
2221        {        {
2222        case OP_CHAR:        case OP_CHAR:
2223        case OP_CHARNC:        case OP_CHARI:
2224        case OP_EXACT:        case OP_EXACT:
2225          case OP_EXACTI:
2226        case OP_UPTO:        case OP_UPTO:
2227          case OP_UPTOI:
2228        case OP_MINUPTO:        case OP_MINUPTO:
2229          case OP_MINUPTOI:
2230        case OP_POSUPTO:        case OP_POSUPTO:
2231          case OP_POSUPTOI:
2232        case OP_STAR:        case OP_STAR:
2233          case OP_STARI:
2234        case OP_MINSTAR:        case OP_MINSTAR:
2235          case OP_MINSTARI:
2236        case OP_POSSTAR:        case OP_POSSTAR:
2237          case OP_POSSTARI:
2238        case OP_PLUS:        case OP_PLUS:
2239          case OP_PLUSI:
2240        case OP_MINPLUS:        case OP_MINPLUS:
2241          case OP_MINPLUSI:
2242        case OP_POSPLUS:        case OP_POSPLUS:
2243          case OP_POSPLUSI:
2244        case OP_QUERY:        case OP_QUERY:
2245          case OP_QUERYI:
2246        case OP_MINQUERY:        case OP_MINQUERY:
2247          case OP_MINQUERYI:
2248        case OP_POSQUERY:        case OP_POSQUERY:
2249        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2250          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2251        break;        break;
2252        }        }
2253  #else  #else
2254      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2255  #endif  #endif
2256      }      }
2257    }    }
# Line 1908  bracket whose current branch will alread Line 2274  bracket whose current branch will alread
2274  Arguments:  Arguments:
2275    code        points to start of search    code        points to start of search
2276    endcode     points to where to stop    endcode     points to where to stop
2277    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2278    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2279    
2280  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2281  */  */
2282    
2283  static BOOL  static BOOL
2284  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2285    compile_data *cd)    BOOL utf, compile_data *cd)
2286  {  {
2287  register int c;  register int c;
2288  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2289       code < endcode;       code < endcode;
2290       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2291    {    {
2292    const uschar *ccode;    const pcre_uchar *ccode;
2293    
2294    c = *code;    c = *code;
2295    
# Line 1937  for (code = first_significant_code(code Line 2303  for (code = first_significant_code(code
2303      continue;      continue;
2304      }      }
2305    
   /* 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;  
     }  
   
2306    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2307    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2308      forward reference subroutine call, we can't. To detect forward reference
2309      we have to scan up the list that is kept in the workspace. This function is
2310      called only when doing the real compile, not during the pre-compile that
2311      measures the size of the compiled pattern. */
2312    
2313    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2314      {      {
2315      BOOL empty_branch = FALSE;      const pcre_uchar *scode;
2316      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
2317    
2318        /* Test for forward reference */
2319    
2320        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2321          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2322    
2323        /* Not a forward reference, test for completed backward reference */
2324    
2325        empty_branch = FALSE;
2326        scode = cd->start_code + GET(code, 1);
2327      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2328    
2329        /* Completed backwards reference */
2330    
2331      do      do
2332        {        {
2333        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf, cd))
2334          {          {
2335          empty_branch = TRUE;          empty_branch = TRUE;
2336          break;          break;
# Line 1965  for (code = first_significant_code(code Line 2338  for (code = first_significant_code(code
2338        scode += GET(scode, 1);        scode += GET(scode, 1);
2339        }        }
2340      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2341    
2342      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2343      continue;      continue;
2344      }      }
2345    
2346      /* Groups with zero repeats can of course be empty; skip them. */
2347    
2348      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2349          c == OP_BRAPOSZERO)
2350        {
2351        code += PRIV(OP_lengths)[c];
2352        do code += GET(code, 1); while (*code == OP_ALT);
2353        c = *code;
2354        continue;
2355        }
2356    
2357      /* A nested group that is already marked as "could be empty" can just be
2358      skipped. */
2359    
2360      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2361          c == OP_SCBRA || c == OP_SCBRAPOS)
2362        {
2363        do code += GET(code, 1); while (*code == OP_ALT);
2364        c = *code;
2365        continue;
2366        }
2367    
2368    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2369    
2370    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2371          c == OP_CBRA || c == OP_CBRAPOS ||
2372          c == OP_ONCE || c == OP_ONCE_NC ||
2373          c == OP_COND)
2374      {      {
2375      BOOL empty_branch;      BOOL empty_branch;
2376      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1987  for (code = first_significant_code(code Line 2386  for (code = first_significant_code(code
2386        empty_branch = FALSE;        empty_branch = FALSE;
2387        do        do
2388          {          {
2389          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2390            empty_branch = TRUE;            empty_branch = TRUE;
2391          code += GET(code, 1);          code += GET(code, 1);
2392          }          }
# Line 2005  for (code = first_significant_code(code Line 2404  for (code = first_significant_code(code
2404      {      {
2405      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2406      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2407      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2408      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"
2409      here. */      here. */
2410    
2411  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2412      case OP_XCLASS:      case OP_XCLASS:
2413      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2414      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 2017  for (code = first_significant_code(code Line 2416  for (code = first_significant_code(code
2416    
2417      case OP_CLASS:      case OP_CLASS:
2418      case OP_NCLASS:      case OP_NCLASS:
2419      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2420    
2421  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2422      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2423  #endif  #endif
2424    
# Line 2058  for (code = first_significant_code(code Line 2457  for (code = first_significant_code(code
2457      case OP_ALLANY:      case OP_ALLANY:
2458      case OP_ANYBYTE:      case OP_ANYBYTE:
2459      case OP_CHAR:      case OP_CHAR:
2460      case OP_CHARNC:      case OP_CHARI:
2461      case OP_NOT:      case OP_NOT:
2462        case OP_NOTI:
2463      case OP_PLUS:      case OP_PLUS:
2464      case OP_MINPLUS:      case OP_MINPLUS:
2465      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2091  for (code = first_significant_code(code Line 2491  for (code = first_significant_code(code
2491      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2492      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2493      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2494      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP
2495          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2496      break;      break;
2497    
2498      /* End of branch */      /* End of branch */
# Line 2099  for (code = first_significant_code(code Line 2500  for (code = first_significant_code(code
2500      case OP_KET:      case OP_KET:
2501      case OP_KETRMAX:      case OP_KETRMAX:
2502      case OP_KETRMIN:      case OP_KETRMIN:
2503        case OP_KETRPOS:
2504      case OP_ALT:      case OP_ALT:
2505      return TRUE;      return TRUE;
2506    
2507      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2508      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2509    
2510  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2511      case OP_STAR:      case OP_STAR:
2512        case OP_STARI:
2513      case OP_MINSTAR:      case OP_MINSTAR:
2514        case OP_MINSTARI:
2515      case OP_POSSTAR:      case OP_POSSTAR:
2516        case OP_POSSTARI:
2517      case OP_QUERY:      case OP_QUERY:
2518        case OP_QUERYI:
2519      case OP_MINQUERY:      case OP_MINQUERY:
2520        case OP_MINQUERYI:
2521      case OP_POSQUERY:      case OP_POSQUERY:
2522      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      case OP_POSQUERYI:
2523        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2524      break;      break;
2525    
2526      case OP_UPTO:      case OP_UPTO:
2527        case OP_UPTOI:
2528      case OP_MINUPTO:      case OP_MINUPTO:
2529        case OP_MINUPTOI:
2530      case OP_POSUPTO:      case OP_POSUPTO:
2531      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      case OP_POSUPTOI:
2532        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2533      break;      break;
2534  #endif  #endif
2535    
# Line 2132  for (code = first_significant_code(code Line 2543  for (code = first_significant_code(code
2543      break;      break;
2544    
2545      case OP_THEN_ARG:      case OP_THEN_ARG:
2546      code += code[1+LINK_SIZE];      code += code[1];
2547      break;      break;
2548    
2549      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
# Line 2155  return TRUE; Line 2566  return TRUE;
2566  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
2567  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,
2568  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.
2569    This function is called only during the real compile, not during the
2570    pre-compile.
2571    
2572  Arguments:  Arguments:
2573    code        points to start of the recursion    code        points to start of the recursion
2574    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2575    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2576    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2577    cd          pointers to tables etc    cd          pointers to tables etc
2578    
2579  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2580  */  */
2581    
2582  static BOOL  static BOOL
2583  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2584    BOOL utf8, compile_data *cd)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2585  {  {
2586  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2587    {    {
2588    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2589      return FALSE;      return FALSE;
2590    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2591    }    }
# Line 2205  where Perl recognizes it as the POSIX cl Line 2618  where Perl recognizes it as the POSIX cl
2618  "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,
2619  I think.  I think.
2620    
2621    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2622    It seems that the appearance of a nested POSIX class supersedes an apparent
2623    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2624    a digit.
2625    
2626    In Perl, unescaped square brackets may also appear as part of class names. For
2627    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2628    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2629    seem right at all. PCRE does not allow closing square brackets in POSIX class
2630    names.
2631    
2632  Arguments:  Arguments:
2633    ptr      pointer to the initial [    ptr      pointer to the initial [
2634    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2213  Returns:   TRUE or FALSE Line 2637  Returns:   TRUE or FALSE
2637  */  */
2638    
2639  static BOOL  static BOOL
2640  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2641  {  {
2642  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2643  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2644  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2645    {    {
2646    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2647        ptr++;
2648      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2649      else
2650      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2651      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2652        {        {
2653        *endptr = ptr;        *endptr = ptr;
2654        return TRUE;        return TRUE;
2655        }        }
2656        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2657             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2658              ptr[1] == CHAR_EQUALS_SIGN) &&
2659            check_posix_syntax(ptr, endptr))
2660          return FALSE;
2661      }      }
2662    }    }
2663  return FALSE;  return FALSE;
# Line 2250  Returns:     a value representing the na Line 2681  Returns:     a value representing the na
2681  */  */
2682    
2683  static int  static int
2684  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2685  {  {
2686  const char *pn = posix_names;  const char *pn = posix_names;
2687  register int yield = 0;  register int yield = 0;
2688  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2689    {    {
2690    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2691      strncmp((const char *)ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2692    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
2693    yield++;    yield++;
2694    }    }
# Line 2289  value in the reference (which is a group Line 2720  value in the reference (which is a group
2720  Arguments:  Arguments:
2721    group      points to the start of the group    group      points to the start of the group
2722    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2723    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 mode
2724    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2725    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
2726    
# Line 2297  Returns:     nothing Line 2728  Returns:     nothing
2728  */  */
2729    
2730  static void  static void
2731  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2732    uschar *save_hwm)    pcre_uchar *save_hwm)
2733  {  {
2734  uschar *ptr = group;  pcre_uchar *ptr = group;
2735    
2736  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2737    {    {
2738    int offset;    int offset;
2739    uschar *hc;    pcre_uchar *hc;
2740    
2741    /* 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
2742    reference. */    reference. */
# Line 2350  Arguments: Line 2781  Arguments:
2781  Returns:         new code pointer  Returns:         new code pointer
2782  */  */
2783    
2784  static uschar *  static pcre_uchar *
2785  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2786  {  {
2787  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2788  *code++ = 255;  *code++ = 255;
2789  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2790  PUT(code, LINK_SIZE, 0);                       /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2791  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2792  }  }
2793    
2794    
# Line 2379  Returns:             nothing Line 2810  Returns:             nothing
2810  */  */
2811    
2812  static void  static void
2813  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2814  {  {
2815  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2816  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
# Line 2462  switch(ptype) Line 2893  switch(ptype)
2893            prop->chartype == ucp_Lt) == negated;            prop->chartype == ucp_Lt) == negated;
2894    
2895    case PT_GC:    case PT_GC:
2896    return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2897    
2898    case PT_PC:    case PT_PC:
2899    return (pdata == prop->chartype) == negated;    return (pdata == prop->chartype) == negated;
# Line 2473  switch(ptype) Line 2904  switch(ptype)
2904    /* These are specials */    /* These are specials */
2905    
2906    case PT_ALNUM:    case PT_ALNUM:
2907    return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2908            _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2909    
2910    case PT_SPACE:    /* Perl space */    case PT_SPACE:    /* Perl space */
2911    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2912            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2913            == negated;            == negated;
2914    
2915    case PT_PXSPACE:  /* POSIX space */    case PT_PXSPACE:  /* POSIX space */
2916    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2917            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2918            c == CHAR_FF || c == CHAR_CR)            c == CHAR_FF || c == CHAR_CR)
2919            == negated;            == negated;
2920    
2921    case PT_WORD:    case PT_WORD:
2922    return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2923            _pcre_ucp_gentype[prop->chartype] == ucp_N ||            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2924            c == CHAR_UNDERSCORE) == negated;            c == CHAR_UNDERSCORE) == negated;
2925    }    }
2926  return FALSE;  return FALSE;
# Line 2508  sense to automatically possessify the re Line 2939  sense to automatically possessify the re
2939    
2940  Arguments:  Arguments:
2941    previous      pointer to the repeated opcode    previous      pointer to the repeated opcode
2942    utf8          TRUE in UTF-8 mode    utf           TRUE in UTF-8 / UTF-16 mode
2943    ptr           next character in pattern    ptr           next character in pattern
2944    options       options bits    options       options bits
2945    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2517  Returns:        TRUE if possessifying is Line 2948  Returns:        TRUE if possessifying is
2948  */  */
2949    
2950  static BOOL  static BOOL
2951  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
2952    int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2953  {  {
2954  int c, next;  pcre_int32 c, next;
2955  int op_code = *previous++;  int op_code = *previous++;
2956    
2957  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2529  if ((options & PCRE_EXTENDED) != 0) Line 2960  if ((options & PCRE_EXTENDED) != 0)
2960    {    {
2961    for (;;)    for (;;)
2962      {      {
2963      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2964      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2965        {        {
2966        ptr++;        ptr++;
2967        while (*ptr != 0)        while (*ptr != 0)
2968          {          {
2969          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2970          ptr++;          ptr++;
2971  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2972          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf) FORWARDCHAR(ptr);
2973  #endif  #endif
2974          }          }
2975        }        }
# Line 2556  if (*ptr == CHAR_BACKSLASH) Line 2987  if (*ptr == CHAR_BACKSLASH)
2987    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
2988    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
2989    }    }
2990    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
2991    {    {
2992  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2993    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
2994  #endif  #endif
2995    next = *ptr++;    next = *ptr++;
2996    }    }
   
2997  else return FALSE;  else return FALSE;
2998    
2999  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2573  if ((options & PCRE_EXTENDED) != 0) Line 3002  if ((options & PCRE_EXTENDED) != 0)
3002    {    {
3003    for (;;)    for (;;)
3004      {      {
3005      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3006      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
3007        {        {
3008        ptr++;        ptr++;
3009        while (*ptr != 0)        while (*ptr != 0)
3010          {          {
3011          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3012          ptr++;          ptr++;
3013  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3014          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf) FORWARDCHAR(ptr);
3015  #endif  #endif
3016          }          }
3017        }        }
# Line 2593  if ((options & PCRE_EXTENDED) != 0) Line 3022  if ((options & PCRE_EXTENDED) != 0)
3022  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3023    
3024  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3025    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)
3026      return FALSE;      return FALSE;
3027    
3028  /* Now compare the next item with the previous opcode. First, handle cases when  /* Now compare the next item with the previous opcode. First, handle cases when
# Line 2602  the next item is a character. */ Line 3031  the next item is a character. */
3031  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
3032    {    {
3033    case OP_CHAR:    case OP_CHAR:
3034  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3035    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
3036  #else  #else
3037    c = *previous;    c = *previous;
3038  #endif  #endif
3039    return c != next;    return c != next;
3040    
3041    /* 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
3042    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
3043    high-valued characters. */    high-valued characters. */
3044    
3045    case OP_CHARNC:    case OP_CHARI:
3046  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3047    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
3048  #else  #else
3049    c = *previous;    c = *previous;
3050  #endif  #endif
3051    if (c == next) return FALSE;    if (c == next) return FALSE;
3052  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3053    if (utf8)    if (utf)
3054      {      {
3055      unsigned int othercase;      unsigned int othercase;
3056      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
# Line 2633  if (next >= 0) switch(op_code) Line 3062  if (next >= 0) switch(op_code)
3062      return (unsigned int)c != othercase;      return (unsigned int)c != othercase;
3063      }      }
3064    else    else
3065  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3066    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */
3067    
3068    /* For OP_NOT, its data is always a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
3069      opcodes are not used for multi-byte characters, because they are coded using
3070      an XCLASS instead. */
3071    
3072    case OP_NOT:    case OP_NOT:
3073      return (c = *previous) == next;
3074    
3075      case OP_NOTI:
3076    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
3077    if ((options & PCRE_CASELESS) == 0) return FALSE;  #ifdef SUPPORT_UTF
3078  #ifdef SUPPORT_UTF8    if (utf)
   if (utf8)  
3079      {      {
3080      unsigned int othercase;      unsigned int othercase;
3081      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
# Line 2654  if (next >= 0) switch(op_code) Line 3087  if (next >= 0) switch(op_code)
3087      return (unsigned int)c == othercase;      return (unsigned int)c == othercase;
3088      }      }
3089    else    else
3090  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3091    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */
3092    
3093    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3094    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
# Line 2745  replaced by OP_PROP codes when PCRE_UCP Line 3178  replaced by OP_PROP codes when PCRE_UCP
3178  switch(op_code)  switch(op_code)
3179    {    {
3180    case OP_CHAR:    case OP_CHAR:
3181    case OP_CHARNC:    case OP_CHARI:
3182  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3183    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
3184  #else  #else
3185    c = *previous;    c = *previous;
# Line 2851  switch(op_code) Line 3284  switch(op_code)
3284        to the original \d etc. At this point, ptr will point to a zero byte. */        to the original \d etc. At this point, ptr will point to a zero byte. */
3285    
3286        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3287          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)
3288            return FALSE;            return FALSE;
3289    
3290        /* Do the property check. */        /* Do the property check. */
# Line 2929  Arguments: Line 3362  Arguments:
3362    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3363    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3364    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3365    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3366    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3367    bcptr          points to current branch chain    bcptr          points to current branch chain
3368      cond_depth     conditional nesting depth
3369    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3370    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3371                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2941  Returns:         TRUE on success Line 3375  Returns:         TRUE on success
3375  */  */
3376    
3377  static BOOL  static BOOL
3378  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3379    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3380      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3381    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3382  {  {
3383  int repeat_type, op_type;  int repeat_type, op_type;
3384  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3385  int bravalue = 0;  int bravalue = 0;
3386  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3387  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3388  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3389  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3390  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3391  int after_manual_callout = 0;  int after_manual_callout = 0;
3392  int length_prevgroup = 0;  int length_prevgroup = 0;
3393  register int c;  register int c;
3394  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3395  uschar *last_code = code;  pcre_uchar *last_code = code;
3396  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3397  uschar *tempcode;  pcre_uchar *tempcode;
3398  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3399  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3400  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3401  const uschar *tempptr;  const pcre_uchar *tempptr;
3402  const uschar *nestptr = NULL;  const pcre_uchar *nestptr = NULL;
3403  uschar *previous = NULL;  pcre_uchar *previous = NULL;
3404  uschar *previous_callout = NULL;  pcre_uchar *previous_callout = NULL;
3405  uschar *save_hwm = NULL;  pcre_uchar *save_hwm = NULL;
3406  uschar classbits[32];  pcre_uint8 classbits[32];
3407    
3408  #ifdef SUPPORT_UTF8  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3409  BOOL class_utf8;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3410  BOOL utf8 = (options & PCRE_UTF8) != 0;  dynamically as we process the pattern. */
3411  uschar *class_utf8data;  
3412  uschar *class_utf8data_base;  #ifdef SUPPORT_UTF
3413  uschar utf8_char[6];  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
3414    BOOL utf = (options & PCRE_UTF8) != 0;
3415    pcre_uchar utf_chars[6];
3416  #else  #else
3417  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3418  uschar *utf8_char = NULL;  #endif
3419    
3420    /* Helper variables for OP_XCLASS opcode (for characters > 255). */
3421    
3422    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3423    BOOL xclass;
3424    pcre_uchar *class_uchardata;
3425    pcre_uchar *class_uchardata_base;
3426  #endif  #endif
3427    
3428  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 2992  greedy_non_default = greedy_default ^ 1; Line 3436  greedy_non_default = greedy_default ^ 1;
3436    
3437  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3438  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
3439  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
3440  find one.  find one.
3441    
3442  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
3443  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
3444  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3445  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3446    
3447  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3448    
3449  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3450  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
3451  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
3452  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3453    value. This is used only for ASCII characters. */
3454    
3455  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3456    
3457  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3458    
# Line 3019  for (;; ptr++) Line 3464  for (;; ptr++)
3464    BOOL is_quantifier;    BOOL is_quantifier;
3465    BOOL is_recurse;    BOOL is_recurse;
3466    BOOL reset_bracount;    BOOL reset_bracount;
3467    int class_charcount;    int class_has_8bitchar;
3468    int class_lastchar;    int class_single_char;
3469    int newoptions;    int newoptions;
3470    int recno;    int recno;
3471    int refsign;    int refsign;
3472    int skipbytes;    int skipbytes;
3473    int subreqbyte;    int subreqchar;
3474    int subfirstbyte;    int subfirstchar;
3475    int terminator;    int terminator;
3476    int mclength;    int mclength;
3477    uschar mcbuffer[8];    int tempbracount;
3478      pcre_uchar mcbuffer[8];
3479    
3480    /* Get next byte in the pattern */    /* Get next character in the pattern */
3481    
3482    c = *ptr;    c = *ptr;
3483    
# Line 3053  for (;; ptr++) Line 3499  for (;; ptr++)
3499  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
3500      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3501  #endif  #endif
3502      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3503            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3504        {        {
3505        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3506        goto FAILED;        goto FAILED;
# Line 3076  for (;; ptr++) Line 3523  for (;; ptr++)
3523        }        }
3524    
3525      *lengthptr += (int)(code - last_code);      *lengthptr += (int)(code - last_code);
3526      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3527          (int)(code - last_code), c, c));
3528    
3529      /* 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
3530      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,
3531      if "previous" is NULL, reset the current code pointer to the start. */      if "previous" is NULL, reset the current code pointer to the start. */
# Line 3086  for (;; ptr++) Line 3534  for (;; ptr++)
3534        {        {
3535        if (previous > orig_code)        if (previous > orig_code)
3536          {          {
3537          memmove(orig_code, previous, code - previous);          memmove(orig_code, previous, IN_UCHARS(code - previous));
3538          code -= previous - orig_code;          code -= previous - orig_code;
3539          previous = orig_code;          previous = orig_code;
3540          }          }
# Line 3102  for (;; ptr++) Line 3550  for (;; ptr++)
3550    /* 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
3551    reference list. */    reference list. */
3552    
3553    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3554               WORK_SIZE_SAFETY_MARGIN)
3555      {      {
3556      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3557      goto FAILED;      goto FAILED;
# Line 3150  for (;; ptr++) Line 3599  for (;; ptr++)
3599      previous_callout = NULL;      previous_callout = NULL;
3600      }      }
3601    
3602    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3603    
3604    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3605      {      {
3606      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
3607      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3608        {        {
3609        ptr++;        ptr++;
3610        while (*ptr != 0)        while (*ptr != 0)
3611          {          {
3612          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3613          ptr++;          ptr++;
3614  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3615          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf) FORWARDCHAR(ptr);
3616  #endif  #endif
3617          }          }
3618        if (*ptr != 0) continue;        if (*ptr != 0) continue;
# Line 3187  for (;; ptr++) Line 3636  for (;; ptr++)
3636      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3637      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3638      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
3639      *firstbyteptr = firstbyte;      *firstcharptr = firstchar;
3640      *reqbyteptr = reqbyte;      *reqcharptr = reqchar;
3641      *codeptr = code;      *codeptr = code;
3642      *ptrptr = ptr;      *ptrptr = ptr;
3643      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3209  for (;; ptr++) Line 3658  for (;; ptr++)
3658      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3659    
3660      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3661        previous = NULL;
3662      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3663        {        {
3664        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3665          *code++ = OP_CIRCM;
3666        }        }
3667      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3668      break;      break;
3669    
3670      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3671      previous = NULL;      previous = NULL;
3672      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3673      break;      break;
3674    
3675      /* 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
3676      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqchar doesn't change either. */
3677    
3678      case CHAR_DOT:      case CHAR_DOT:
3679      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3680      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
3681      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
3682      previous = code;      previous = code;
3683      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3684      break;      break;
# Line 3283  for (;; ptr++) Line 3733  for (;; ptr++)
3733          {          {
3734          if (ptr[1] == CHAR_E)          if (ptr[1] == CHAR_E)
3735            ptr++;            ptr++;
3736          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)  
3737            ptr += 3;            ptr += 3;
3738          else          else
3739            break;            break;
# Line 3303  for (;; ptr++) Line 3752  for (;; ptr++)
3752          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3753        {        {
3754        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3755        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3756        zerofirstbyte = firstbyte;        zerofirstchar = firstchar;
3757        break;        break;
3758        }        }
3759    
# Line 3314  for (;; ptr++) Line 3763  for (;; ptr++)
3763    
3764      should_flip_negation = FALSE;      should_flip_negation = FALSE;
3765    
3766      /* 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.
3767      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
3768      valued UTF-8 characters, we don't yet do any optimization. */      < 256 character. class_single_char will be 1 if the class contains only
3769        a single character. */
3770    
3771      class_charcount = 0;      class_has_8bitchar = 0;
3772      class_lastchar = -1;      class_single_char = 0;
3773    
3774      /* 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
3775      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
3776      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.
3777      */      */
3778    
3779      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
3780    
3781  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3782      class_utf8 = FALSE;                       /* No chars >= 256 */      xclass = FALSE;                           /* No chars >= 256 */
3783      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_uchardata = code + LINK_SIZE + 2;   /* For UTF-8 items */
3784      class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */      class_uchardata_base = class_uchardata;   /* For resetting in pass 1 */
3785  #endif  #endif
3786    
3787      /* 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 3340  for (;; ptr++) Line 3790  for (;; ptr++)
3790    
3791      if (c != 0) do      if (c != 0) do
3792        {        {
3793        const uschar *oldptr;        const pcre_uchar *oldptr;
3794    
3795  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3796        if (utf8 && c > 127)        if (utf && HAS_EXTRALEN(c))
3797          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3798          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3799          }          }
3800    #endif
3801    
3802        /* In the pre-compile phase, accumulate the length of any UTF-8 extra  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3803          /* In the pre-compile phase, accumulate the length of any extra
3804        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
3805        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
3806        (which is on the stack). */        (which is on the stack). */
3807    
3808        if (lengthptr != NULL)        if (lengthptr != NULL)
3809          {          {
3810          *lengthptr += class_utf8data - class_utf8data_base;          *lengthptr += class_uchardata - class_uchardata_base;
3811          class_utf8data = class_utf8data_base;          class_uchardata = class_uchardata_base;
3812          }          }
   
3813  #endif  #endif
3814    
3815        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 3386  for (;; ptr++) Line 3837  for (;; ptr++)
3837          {          {
3838          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3839          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3840          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
3841          uschar pbits[32];          pcre_uint8 pbits[32];
3842    
3843          if (ptr[1] != CHAR_COLON)          if (ptr[1] != CHAR_COLON)
3844            {            {
# Line 3442  for (;; ptr++) Line 3893  for (;; ptr++)
3893          /* Copy in the first table (always present) */          /* Copy in the first table (always present) */
3894    
3895          memcpy(pbits, cbits + posix_class_maps[posix_class],          memcpy(pbits, cbits + posix_class_maps[posix_class],
3896            32 * sizeof(uschar));            32 * sizeof(pcre_uint8));
3897    
3898          /* If there is a second table, add or remove it as required. */          /* If there is a second table, add or remove it as required. */
3899    
# Line 3473  for (;; ptr++) Line 3924  for (;; ptr++)
3924            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3925    
3926          ptr = tempptr + 1;          ptr = tempptr + 1;
3927          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          /* Every class contains at least one < 256 characters. */
3928            class_has_8bitchar = 1;
3929            /* Every class contains at least two characters. */
3930            class_single_char = 2;
3931          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3932          }          }
3933    
3934        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3935        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
3936        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
3937        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
3938        class_charcount bigger than one. Unrecognized escapes fall through and        speculatively set both class_has_8bitchar and class_single_char bigger
3939        are either treated as literal characters (by default), or are faulted if        than one. Unrecognized escapes fall through and are either treated
3940          as literal characters (by default), or are faulted if
3941        PCRE_EXTRA is set. */        PCRE_EXTRA is set. */
3942    
3943        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
# Line 3491  for (;; ptr++) Line 3946  for (;; ptr++)
3946          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3947    
3948          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 */
3949            else if (-c == ESC_N)            /* \N is not supported in a class */
3950              {
3951              *errorcodeptr = ERR71;
3952              goto FAILED;
3953              }
3954          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3955            {            {
3956            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3504  for (;; ptr++) Line 3964  for (;; ptr++)
3964    
3965          if (c < 0)          if (c < 0)
3966            {            {
3967            register const uschar *cbits = cd->cbits;            register const pcre_uint8 *cbits = cd->cbits;
3968            class_charcount += 2;     /* Greater than 1 is what matters */            /* Every class contains at least two < 256 characters. */
3969              class_has_8bitchar++;
3970              /* Every class contains at least two characters. */
3971              class_single_char += 2;
3972    
3973            switch (-c)            switch (-c)
3974              {              {
# Line 3518  for (;; ptr++) Line 3981  for (;; ptr++)
3981              case ESC_SU:              case ESC_SU:
3982              nestptr = ptr;              nestptr = ptr;
3983              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3984              class_charcount -= 2;                /* Undo! */              class_has_8bitchar--;                /* Undo! */
3985              continue;              continue;
3986  #endif  #endif
3987              case ESC_d:              case ESC_d:
# Line 3540  for (;; ptr++) Line 4003  for (;; ptr++)
4003              continue;              continue;
4004    
4005              /* Perl 5.004 onwards omits VT from \s, but we must preserve it              /* Perl 5.004 onwards omits VT from \s, but we must preserve it
4006              if it was previously set by something earlier in the character              if it was previously set by something earlier in the character
4007              class. */              class. */
4008    
4009              case ESC_s:              case ESC_s:
4010              classbits[0] |= cbits[cbit_space];              classbits[0] |= cbits[cbit_space];
4011              classbits[1] |= cbits[cbit_space+1] & ~0x08;              classbits[1] |= cbits[cbit_space+1] & ~0x08;
4012              for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
4013              continue;              continue;
4014    
# Line 3559  for (;; ptr++) Line 4022  for (;; ptr++)
4022              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
4023              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
4024              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
4025  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4026              if (utf8)              if (utf)
4027                {                {
4028                class_utf8 = TRUE;                xclass = TRUE;
4029                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4030                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1680, class_uchardata);
4031                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4032                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180e, class_uchardata);
4033                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4034                class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2000, class_uchardata);
4035                class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x200A, class_uchardata);
4036                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4037                class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x202f, class_uchardata);
4038                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4039                class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x205f, class_uchardata);
4040                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4041                class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x3000, class_uchardata);
4042                }                }
4043  #endif  #endif
4044              continue;              continue;
# Line 3594  for (;; ptr++) Line 4057  for (;; ptr++)
4057                classbits[c] |= x;                classbits[c] |= x;
4058                }                }
4059    
4060  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4061              if (utf8)              if (utf)
4062                {                {
4063                class_utf8 = TRUE;                xclass = TRUE;
4064                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4065                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);
4066                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x167f, class_uchardata);
4067                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4068                class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1681, class_uchardata);
4069                class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180d, class_uchardata);
4070                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4071                class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180f, class_uchardata);
4072                class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1fff, class_uchardata);
4073                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4074                class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x200B, class_uchardata);
4075                class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x202e, class_uchardata);
4076                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4077                class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2030, class_uchardata);
4078                class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x205e, class_uchardata);
4079                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4080                class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2060, class_uchardata);
4081                class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2fff, class_uchardata);
4082                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4083                class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x3001, class_uchardata);
4084                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4085                }                }
4086  #endif  #endif
4087              continue;              continue;
# Line 3629  for (;; ptr++) Line 4092  for (;; ptr++)
4092              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
4093              SETBIT(classbits, 0x0d); /* CR */              SETBIT(classbits, 0x0d); /* CR */
4094              SETBIT(classbits, 0x85); /* NEL */              SETBIT(classbits, 0x85); /* NEL */
4095  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4096              if (utf8)              if (utf)
4097                {                {
4098                class_utf8 = TRUE;                xclass = TRUE;
4099                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4100                class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2028, class_uchardata);
4101                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2029, class_uchardata);
4102                }                }
4103  #endif  #endif
4104              continue;              continue;
# Line 3657  for (;; ptr++) Line 4120  for (;; ptr++)
4120                classbits[c] |= x;                classbits[c] |= x;
4121                }                }
4122    
4123  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4124              if (utf8)              if (utf)
4125                {                {
4126                class_utf8 = TRUE;                xclass = TRUE;
4127                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4128                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);
4129                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2027, class_uchardata);
4130                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4131                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2029, class_uchardata);
4132                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4133                }                }
4134  #endif  #endif
4135              continue;              continue;
# Line 3679  for (;; ptr++) Line 4142  for (;; ptr++)
4142                int pdata;                int pdata;
4143                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
4144                if (ptype < 0) goto FAILED;                if (ptype < 0) goto FAILED;
4145                class_utf8 = TRUE;                xclass = TRUE;
4146                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_uchardata++ = ((-c == ESC_p) != negated)?
4147                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
4148                *class_utf8data++ = ptype;                *class_uchardata++ = ptype;
4149                *class_utf8data++ = pdata;                *class_uchardata++ = pdata;
4150                class_charcount -= 2;   /* Not a < 256 character */                class_has_8bitchar--;                /* Undo! */
4151                continue;                continue;
4152                }                }
4153  #endif  #endif
# Line 3698  for (;; ptr++) Line 4161  for (;; ptr++)
4161                *errorcodeptr = ERR7;                *errorcodeptr = ERR7;
4162                goto FAILED;                goto FAILED;
4163                }                }
4164              class_charcount -= 2;  /* Undo the default count from above */              class_has_8bitchar--;    /* Undo the speculative increase. */
4165              c = *ptr;              /* Get the final character and fall through */              class_single_char -= 2;  /* Undo the speculative increase. */
4166                c = *ptr;                /* Get the final character and fall through */
4167              break;              break;
4168              }              }
4169            }            }
4170    
4171          /* 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
4172          greater than 256 in UTF-8 mode. */          greater than 256. */
4173    
4174          }   /* End of backslash handling */          }   /* End of backslash handling */
4175    
# Line 3753  for (;; ptr++) Line 4217  for (;; ptr++)
4217            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
4218            }            }
4219    
4220  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4221          if (utf8)          if (utf)
4222            {                           /* Braces are required because the */            {                           /* Braces are required because the */
4223            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */
4224            }            }
# Line 3798  for (;; ptr++) Line 4262  for (;; ptr++)
4262    
4263          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4264    
4265            /* Since we found a character range, single character optimizations
4266            cannot be done anymore. */
4267            class_single_char = 2;
4268    
4269          /* 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
4270          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
4271          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
4272          available. */          available. */
4273    
4274  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !(defined COMPILE_PCRE8)
4275          if (utf8 && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))          if ((d > 255) || (utf && ((options & PCRE_CASELESS) != 0 && d > 127)))
4276    #elif defined  SUPPORT_UTF
4277            if (utf && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))
4278    #elif !(defined COMPILE_PCRE8)
4279            if (d > 255)
4280    #endif
4281    #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
4282            {            {
4283            class_utf8 = TRUE;            xclass = TRUE;
4284    
4285            /* With UCP support, we can find the other case equivalents of            /* With UCP support, we can find the other case equivalents of
4286            the relevant characters. There may be several ranges. Optimize how            the relevant characters. There may be several ranges. Optimize how
4287            they fit with the basic range. */            they fit with the basic range. */
4288    
4289  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4290    #ifndef COMPILE_PCRE8
4291              if (utf && (options & PCRE_CASELESS) != 0)
4292    #else
4293            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
4294    #endif
4295              {              {
4296              unsigned int occ, ocd;              unsigned int occ, ocd;
4297              unsigned int cc = c;              unsigned int cc = c;
# Line 3839  for (;; ptr++) Line 4317  for (;; ptr++)
4317    
4318                if (occ == ocd)                if (occ == ocd)
4319                  {                  {
4320                  *class_utf8data++ = XCL_SINGLE;                  *class_uchardata++ = XCL_SINGLE;
4321                  }                  }
4322                else                else
4323                  {                  {
4324                  *class_utf8data++ = XCL_RANGE;                  *class_uchardata++ = XCL_RANGE;
4325                  class_utf8data += _pcre_ord2utf8(occ, class_utf8data);                  class_uchardata += PRIV(ord2utf)(occ, class_uchardata);
4326                  }                  }
4327                class_utf8data += _pcre_ord2utf8(ocd, class_utf8data);                class_uchardata += PRIV(ord2utf)(ocd, class_uchardata);
4328                }                }
4329              }              }
4330  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
# Line 3854  for (;; ptr++) Line 4332  for (;; ptr++)
4332            /* Now record the original range, possibly modified for UCP caseless            /* Now record the original range, possibly modified for UCP caseless
4333            overlapping ranges. */            overlapping ranges. */
4334    
4335            *class_utf8data++ = XCL_RANGE;            *class_uchardata++ = XCL_RANGE;
4336            class_utf8data += _pcre_ord2utf8(c, class_utf8data);  #ifdef SUPPORT_UTF
4337            class_utf8data += _pcre_ord2utf8(d, class_utf8data);  #ifndef COMPILE_PCRE8
4338              if (utf)
4339                {
4340                class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4341                class_uchardata += PRIV(ord2utf)(d, class_uchardata);
4342                }
4343              else
4344                {
4345                *class_uchardata++ = c;
4346                *class_uchardata++ = d;
4347                }
4348    #else
4349              class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4350              class_uchardata += PRIV(ord2utf)(d, class_uchardata);
4351    #endif
4352    #else /* SUPPORT_UTF */
4353              *class_uchardata++ = c;
4354              *class_uchardata++ = d;
4355    #endif /* SUPPORT_UTF */
4356    
4357            /* With UCP support, we are done. Without UCP support, there is no            /* With UCP support, we are done. Without UCP support, there is no
4358            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
4359            for the smaller ones. */            for the smaller ones. As for 16 bit characters without UTF, we
4360              can still use  */
4361    
4362  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4363            continue;    /* With next character in the class */  #ifndef COMPILE_PCRE8
4364  #else            if (utf)
4365            if ((options & PCRE_CASELESS) == 0 || c > 127) continue;  #endif
4366                continue;    /* With next character in the class */
4367    #endif  /* SUPPORT_UCP */
4368    
4369    #if defined SUPPORT_UTF && !defined(SUPPORT_UCP) && !(defined COMPILE_PCRE8)
4370              if (utf)
4371                {
4372                if ((options & PCRE_CASELESS) == 0 || c > 127) continue;
4373                /* Adjust upper limit and fall through to set up the map */
4374                d = 127;
4375                }
4376              else
4377                {
4378                if (c > 255) continue;
4379                /* Adjust upper limit and fall through to set up the map */
4380                d = 255;
4381                }
4382    #elif defined SUPPORT_UTF && !defined(SUPPORT_UCP)
4383              if ((options & PCRE_CASELESS) == 0 || c > 127) continue;
4384            /* Adjust upper limit and fall through to set up the map */            /* Adjust upper limit and fall through to set up the map */
   
4385            d = 127;            d = 127;
4386    #else
4387  #endif  /* SUPPORT_UCP */            if (c > 255) continue;
4388              /* Adjust upper limit and fall through to set up the map */
4389              d = 255;
4390    #endif  /* SUPPORT_UTF && !SUPPORT_UCP && !COMPILE_PCRE8 */
4391            }            }
4392  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF || !COMPILE_PCRE8 */
4393    
4394          /* 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
4395          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. */  
4396    
4397          class_charcount += d - c + 1;          class_has_8bitchar = 1;
         class_lastchar = d;  
4398    
4399          /* 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. */
4400    
# Line 3889  for (;; ptr++) Line 4403  for (;; ptr++)
4403            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
4404            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
4405              {              {
4406              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c]; /* flip case */
4407              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
4408              }              }
4409            }            }
# Line 3903  for (;; ptr++) Line 4417  for (;; ptr++)
4417    
4418        LONE_SINGLE_CHARACTER:        LONE_SINGLE_CHARACTER:
4419    
4420        /* Handle a character that cannot go in the bit map */        /* Only the value of 1 matters for class_single_char. */
4421          if (class_single_char < 2) class_single_char++;
4422    
4423  #ifdef SUPPORT_UTF8        /* If class_charcount is 1, we saw precisely one character. As long as
4424        if (utf8 && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))        there were no negated characters >= 128 and there was no use of \p or \P,
4425          in other words, no use of any XCLASS features, we can optimize.
4426    
4427          In UTF-8 mode, we can optimize the negative case only if there were no
4428          characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4429          operate on single-bytes characters only. This is an historical hangover.
4430          Maybe one day we can tidy these opcodes to handle multi-byte characters.
4431    
4432          The optimization throws away the bit map. We turn the item into a
4433          1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4434          Note that OP_NOT[I] does not support multibyte characters. In the positive
4435          case, it can cause firstchar to be set. Otherwise, there can be no first
4436          char if this item is first, whatever repeat count may follow. In the case
4437          of reqchar, save the previous value for reinstating. */
4438    
4439    #ifdef SUPPORT_UTF
4440          if (class_single_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET
4441            && (!utf || !negate_class || c < (MAX_VALUE_FOR_SINGLE_CHAR + 1)))
4442    #else
4443          if (class_single_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
4444    #endif
4445          {          {
4446          class_utf8 = TRUE;          ptr++;
4447          *class_utf8data++ = XCL_SINGLE;          zeroreqchar = reqchar;
4448          class_utf8data += _pcre_ord2utf8(c, class_utf8data);  
4449            /* The OP_NOT[I] opcodes work on single characters only. */
4450    
4451            if (negate_class)
4452              {
4453              if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4454              zerofirstchar = firstchar;
4455              *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4456              *code++ = c;
4457              goto NOT_CHAR;
4458              }
4459    
4460            /* For a single, positive character, get the value into mcbuffer, and
4461            then we can handle this with the normal one-character code. */
4462    
4463    #ifdef SUPPORT_UTF
4464            if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
4465              mclength = PRIV(ord2utf)(c, mcbuffer);
4466            else
4467    #endif
4468              {
4469              mcbuffer[0] = c;
4470              mclength = 1;
4471              }
4472            goto ONE_CHAR;
4473            }       /* End of 1-char optimization */
4474    
4475          /* Handle a character that cannot go in the bit map. */
4476    
4477    #if defined SUPPORT_UTF && !(defined COMPILE_PCRE8)
4478          if ((c > 255) || (utf && ((options & PCRE_CASELESS) != 0 && c > 127)))
4479    #elif defined SUPPORT_UTF
4480          if (utf && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))
4481    #elif !(defined COMPILE_PCRE8)
4482          if (c > 255)
4483    #endif
4484    
4485    #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
4486            {
4487            xclass = TRUE;
4488            *class_uchardata++ = XCL_SINGLE;
4489    #ifdef SUPPORT_UTF
4490    #ifndef COMPILE_PCRE8
4491            /* In non 8 bit mode, we can get here even if we are not in UTF mode. */
4492            if (!utf)
4493              *class_uchardata++ = c;
4494            else
4495    #endif
4496              class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4497    #else /* SUPPORT_UTF */
4498            *class_uchardata++ = c;
4499    #endif /* SUPPORT_UTF */
4500    
4501  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4502    #ifdef COMPILE_PCRE8
4503          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4504    #else
4505            /* In non 8 bit mode, we can get here even if we are not in UTF mode. */
4506            if (utf && (options & PCRE_CASELESS) != 0)
4507    #endif
4508            {            {
4509            unsigned int othercase;            unsigned int othercase;
4510            if ((othercase = UCD_OTHERCASE(c)) != c)            if ((othercase = UCD_OTHERCASE(c)) != c)
4511              {              {
4512              *class_utf8data++ = XCL_SINGLE;              *class_uchardata++ = XCL_SINGLE;
4513              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_uchardata += PRIV(ord2utf)(othercase, class_uchardata);
4514              }              }
4515            }            }
4516  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4517    
4518          }          }
4519        else        else
4520  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF || COMPILE_PCRE16 */
4521    
4522        /* Handle a single-byte character */        /* Handle a single-byte character */
4523          {          {
4524            class_has_8bitchar = 1;
4525          classbits[c/8] |= (1 << (c&7));          classbits[c/8] |= (1 << (c&7));
4526          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4527            {            {
4528            c = cd->fcc[c];   /* flip case */            c = cd->fcc[c]; /* flip case */
4529            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
4530            }            }
         class_charcount++;  
         class_lastchar = c;  
4531          }          }
4532        }        }
4533    
# Line 3958  for (;; ptr++) Line 4548  for (;; ptr++)
4548        goto FAILED;        goto FAILED;
4549        }        }
4550    
4551      /* If class_charcount is 1, we saw precisely one character whose value is      /* If this is the first thing in the branch, there can be no first char
4552      less than 256. As long as there were no characters >= 128 and there was no      setting, whatever the repeat count. Any reqchar setting must remain
4553      use of \p or \P, in other words, no use of any XCLASS features, we can      unchanged after any kind of repeat. */
4554      optimize.  
4555        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4556      In UTF-8 mode, we can optimize the negative case only if there were no      zerofirstchar = firstchar;
4557      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      zeroreqchar = reqchar;
     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;  
4558    
4559      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4560      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
# Line 4027  for (;; ptr++) Line 4564  for (;; ptr++)
4564      be listed) there are no characters < 256, we can omit the bitmap in the      be listed) there are no characters < 256, we can omit the bitmap in the
4565      actual compiled code. */      actual compiled code. */
4566    
4567  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4568      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))      if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
4569    #endif
4570    #ifndef COMPILE_PCRE8
4571        if (xclass && !should_flip_negation)
4572    #endif
4573    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4574        {        {
4575        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_uchardata++ = XCL_END;    /* Marks the end of extra data */
4576        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
4577        code += LINK_SIZE;        code += LINK_SIZE;
4578        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT:0;
4579    
4580        /* 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;
4581        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. */
4582    
4583        if (class_charcount > 0)        if (class_has_8bitchar > 0)
4584          {          {
4585          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
4586          memmove(code + 32, code, class_utf8data - code);          memmove(code + (32 / sizeof(pcre_uchar)), code,
4587              IN_UCHARS(class_uchardata - code));
4588          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
4589          code = class_utf8data + 32;          code = class_uchardata + (32 / sizeof(pcre_uchar));
4590          }          }
4591        else code = class_utf8data;        else code = class_uchardata;
4592    
4593        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
4594    
4595        PUT(previous, 1, code - previous);        PUT(previous, 1, (int)(code - previous));
4596        break;   /* End of class handling */        break;   /* End of class handling */
4597        }        }
4598  #endif  #endif
# Line 4061  for (;; ptr++) Line 4604  for (;; ptr++)
4604      negating it if necessary. */      negating it if necessary. */
4605    
4606      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4607      if (negate_class)      if (lengthptr == NULL)    /* Save time in the pre-compile phase */
       {  
       if (lengthptr == NULL)    /* Save time in the pre-compile phase */  
         for (c = 0; c < 32; c++) code[c] = ~classbits[c];  
       }  
     else  
4608        {        {
4609          if (negate_class)
4610            for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
4611        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4612        }        }
4613      code += 32;      code += 32 / sizeof(pcre_uchar);
4614        NOT_CHAR:
4615      break;      break;
4616    
4617    
# Line 4107  for (;; ptr++) Line 4648  for (;; ptr++)
4648    
4649      if (repeat_min == 0)      if (repeat_min == 0)
4650        {        {
4651        firstbyte = zerofirstbyte;    /* Adjust for zero repeat */        firstchar = zerofirstchar;    /* Adjust for zero repeat */
4652        reqbyte = zeroreqbyte;        /* Ditto */        reqchar = zeroreqchar;        /* Ditto */
4653        }        }
4654    
4655      /* Remember whether this is a variable length repeat */      /* Remember whether this is a variable length repeat */
# Line 4118  for (;; ptr++) Line 4659  for (;; ptr++)
4659      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4660      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4661    
4662      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4663      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4664    
4665      tempcode = previous;      tempcode = previous;
4666    
# Line 4142  for (;; ptr++) Line 4683  for (;; ptr++)
4683        }        }
4684      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4685    
4686        /* If previous was a recursion call, wrap it in atomic brackets so that
4687        previous becomes the atomic group. All recursions were so wrapped in the
4688        past, but it no longer happens for non-repeated recursions. In fact, the
4689        repeated ones could be re-implemented independently so as not to need this,
4690        but for the moment we rely on the code for repeating groups. */
4691    
4692        if (*previous == OP_RECURSE)
4693          {
4694          memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
4695          *previous = OP_ONCE;
4696          PUT(previous, 1, 2 + 2*LINK_SIZE);
4697          previous[2 + 2*LINK_SIZE] = OP_KET;
4698          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4699          code += 2 + 2 * LINK_SIZE;
4700          length_prevgroup = 3 + 3*LINK_SIZE;
4701    
4702          /* When actually compiling, we need to check whether this was a forward
4703          reference, and if so, adjust the offset. */
4704    
4705          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4706            {
4707            int offset = GET(cd->hwm, -LINK_SIZE);
4708            if (offset == previous + 1 - cd->start_code)
4709              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4710            }
4711          }
4712    
4713        /* Now handle repetition for the different types of item. */
4714    
4715      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4716      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
4717      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqchar - it might not be if a sequence such as x{3} is
4718      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstchar
4719      instead.  */      instead.  */
4720    
4721      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4722        {        {
4723        /* Deal with UTF-8 characters that take up more than one byte. It's        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4724    
4725          /* Deal with UTF characters that take up more than one character. It's
4726        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4727        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus UTF_LENGTH to flag that
4728        length rather than a small character. */        it's a length rather than a small character. */
4729    
4730  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4731        if (utf8 && (code[-1] & 0x80) != 0)        if (utf && NOT_FIRSTCHAR(code[-1]))
4732          {          {
4733          uschar *lastchar = code - 1;          pcre_uchar *lastchar = code - 1;
4734          while((*lastchar & 0xc0) == 0x80) lastchar--;          BACKCHAR(lastchar);
4735          c = code - lastchar;            /* Length of UTF-8 character */          c = (int)(code - lastchar);     /* Length of UTF-8 character */
4736          memcpy(utf8_char, lastchar, c); /* Save the char */          memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
4737          c |= 0x80;                      /* Flag c as a length */          c |= UTF_LENGTH;                /* Flag c as a length */
4738          }          }
4739        else        else
4740  #endif  #endif /* SUPPORT_UTF */
   
       /* Handle the case of a single byte - either with no UTF8 support, or  
       with UTF-8 disabled, or for a UTF-8 character < 128. */  
4741    
4742          /* Handle the case of a single charater - either with no UTF support, or
4743          with UTF disabled, or for a single character UTF character. */
4744          {          {
4745          c = code[-1];          c = code[-1];
4746          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqchar = c | req_caseopt | cd->req_varyopt;
4747          }          }
4748    
4749        /* If the repetition is unlimited, it pays to see if the next thing on        /* If the repetition is unlimited, it pays to see if the next thing on
# Line 4182  for (;; ptr++) Line 4753  for (;; ptr++)
4753    
4754        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4755            repeat_max < 0 &&            repeat_max < 0 &&
4756            check_auto_possessive(previous, utf8, ptr + 1, options, cd))            check_auto_possessive(previous, utf, ptr + 1, options, cd))
4757          {          {
4758          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4759          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4194  for (;; ptr++) Line 4765  for (;; ptr++)
4765      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4766      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4767      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4768      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4769      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4770    
4771      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4772        {        {
4773        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4774        c = previous[1];        c = previous[1];
4775        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4776            repeat_max < 0 &&            repeat_max < 0 &&
4777            check_auto_possessive(previous, utf8, ptr + 1, options, cd))            check_auto_possessive(previous, utf, ptr + 1, options, cd))
4778          {          {
4779          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4780          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4220  for (;; ptr++) Line 4791  for (;; ptr++)
4791    
4792      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
4793        {        {
4794        uschar *oldcode;        pcre_uchar *oldcode;
4795        int prop_type, prop_value;        int prop_type, prop_value;
4796        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
4797        c = *previous;        c = *previous;
4798    
4799        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4800            repeat_max < 0 &&            repeat_max < 0 &&
4801            check_auto_possessive(previous, utf8, ptr + 1, options, cd))            check_auto_possessive(previous, utf, ptr + 1, options, cd))
4802          {          {
4803          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4804          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4307  for (;; ptr++) Line 4878  for (;; ptr++)
4878          we have to insert the character for the previous code. For a repeated          we have to insert the character for the previous code. For a repeated
4879          Unicode property match, there are two extra bytes that define the          Unicode property match, there are two extra bytes that define the
4880          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
4881          c, with the 0x80 bit as a flag. */          c, with the UTF_LENGTH bit as a flag. */
4882    
4883          if (repeat_max < 0)          if (repeat_max < 0)
4884            {            {
4885  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4886            if (utf8 && c >= 128)            if (utf && (c & UTF_LENGTH) != 0)
4887              {              {
4888              memcpy(code, utf8_char, c & 7);              memcpy(code, utf_chars, IN_UCHARS(c & 7));
4889              code += c & 7;              code += c & 7;
4890              }              }
4891            else            else
# Line 4336  for (;; ptr++) Line 4907  for (;; ptr++)
4907    
4908          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
4909            {            {
4910  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4911            if (utf8 && c >= 128)            if (utf && (c & UTF_LENGTH) != 0)
4912              {              {
4913              memcpy(code, utf8_char, c & 7);              memcpy(code, utf_chars, IN_UCHARS(c & 7));
4914              code += c & 7;              code += c & 7;
4915              }              }
4916            else            else
# Line 4366  for (;; ptr++) Line 4937  for (;; ptr++)
4937    
4938        /* The character or character type itself comes last in all cases. */        /* The character or character type itself comes last in all cases. */
4939    
4940  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4941        if (utf8 && c >= 128)        if (utf && (c & UTF_LENGTH) != 0)
4942          {          {
4943          memcpy(code, utf8_char, c & 7);          memcpy(code, utf_chars, IN_UCHARS(c & 7));
4944          code += c & 7;          code += c & 7;
4945          }          }
4946        else        else
# Line 4393  for (;; ptr++) Line 4964  for (;; ptr++)
4964    
4965      else if (*previous == OP_CLASS ||      else if (*previous == OP_CLASS ||
4966               *previous == OP_NCLASS ||               *previous == OP_NCLASS ||
4967  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4968               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4969  #endif  #endif
4970               *previous == OP_REF)               *previous == OP_REF ||
4971                 *previous == OP_REFI)
4972        {        {
4973        if (repeat_max == 0)        if (repeat_max == 0)
4974          {          {
# Line 4430  for (;; ptr++) Line 5002  for (;; ptr++)
5002        }        }
5003    
5004      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
5005      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
5006        opcodes such as BRA and CBRA, as this is the place where they get converted
5007        into the more special varieties such as BRAPOS and SBRA. A test for >=
5008        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5009        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
5010        repetition of assertions, but now it does, for Perl compatibility. */
5011    
5012      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
5013        {        {
5014        register int i;        register int i;
       int ketoffset = 0;  
5015        int len = (int)(code - previous);        int len = (int)(code - previous);
5016        uschar *bralink = NULL;        pcre_uchar *bralink = NULL;
5017          pcre_uchar *brazeroptr = NULL;
5018    
5019        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5020          we just ignore the repeat. */
5021    
5022        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5023          {          goto END_REPEAT;
         *errorcodeptr = ERR55;  
         goto FAILED;  
         }  
5024    
5025        /* If the maximum repeat count is unlimited, find the end of the bracket        /* There is no sense in actually repeating assertions. The only potential
5026        by scanning through from the start, and compute the offset back to it        use of repetition is in cases when the assertion is optional. Therefore,
5027        from the current code pointer. There may be an OP_OPT setting following        if the minimum is greater than zero, just ignore the repeat. If the
5028        the final KET, so we can't find the end just by going back from the code        maximum is not not zero or one, set it to 1. */
5029        pointer. */  
5030          if (*previous < OP_ONCE)    /* Assertion */
5031        if (repeat_max == -1)          {
5032          {          if (repeat_min > 0) goto END_REPEAT;
5033          register uschar *ket = previous;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = (int)(code - ket);  
5034          }          }
5035    
5036        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 4479  for (;; ptr++) Line 5051  for (;; ptr++)
5051          **   goto END_REPEAT;          **   goto END_REPEAT;
5052          **   }          **   }
5053    
5054          However, that fails when a group is referenced as a subroutine from          However, that fails when a group or a subgroup within it is referenced
5055          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it          as a subroutine from elsewhere in the pattern, so now we stick in
5056          so that it is skipped on execution. As we don't have a list of which          OP_SKIPZERO in front of it so that it is skipped on execution. As we
5057          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
5058            selectively.
5059    
5060          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5061          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 4494  for (;; ptr++) Line 5067  for (;; ptr++)
5067          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
5068            {            {
5069            *code = OP_END;            *code = OP_END;
5070            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf, cd, save_hwm);
5071            memmove(previous+1, previous, len);            memmove(previous + 1, previous, IN_UCHARS(len));
5072            code++;            code++;
5073            if (repeat_max == 0)            if (repeat_max == 0)
5074              {              {
5075              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
5076              goto END_REPEAT;              goto END_REPEAT;
5077              }              }
5078              brazeroptr = previous;    /* Save for possessive optimizing */
5079            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
5080            }            }
5081    
# Line 4517  for (;; ptr++) Line 5091  for (;; ptr++)
5091            {            {
5092            int offset;            int offset;
5093            *code = OP_END;            *code = OP_END;
5094            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);            adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
5095            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5096            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
5097            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
5098            *previous++ = OP_BRA;            *previous++ = OP_BRA;
# Line 4564  for (;; ptr++) Line 5138  for (;; ptr++)
5138              *lengthptr += delta;              *lengthptr += delta;
5139              }              }
5140    
5141            /* This is compiling for real */            /* This is compiling for real. If there is a set first byte for
5142              the group, and we have not yet set a "required byte", set it. Make
5143              sure there is enough workspace for copying forward references before
5144              doing the copy. */
5145    
5146            else            else
5147              {              {
5148              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;              if (groupsetfirstchar && reqchar < 0) reqchar = firstchar;
5149    
5150              for (i = 1; i < repeat_min; i++)              for (i = 1; i < repeat_min; i++)
5151                {                {
5152                uschar *hc;                pcre_uchar *hc;
5153                uschar *this_hwm = cd->hwm;                pcre_uchar *this_hwm = cd->hwm;
5154                memcpy(code, previous, len);                memcpy(code, previous, IN_UCHARS(len));
5155    
5156                  while (cd->hwm > cd->start_workspace + cd->workspace_size -
5157                         WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5158                    {
5159                    int save_offset = save_hwm - cd->start_workspace;
5160                    int this_offset = this_hwm - cd->start_workspace;
5161                    *errorcodeptr = expand_workspace(cd);
5162                    if (*errorcodeptr != 0) goto FAILED;
5163                    save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5164                    this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5165                    }
5166    
5167                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5168                  {                  {
5169                  PUT(cd->hwm, 0, GET(hc, 0) + len);                  PUT(cd->hwm, 0, GET(hc, 0) + len);
# Line 4623  for (;; ptr++) Line 5213  for (;; ptr++)
5213    
5214          else for (i = repeat_max - 1; i >= 0; i--)          else for (i = repeat_max - 1; i >= 0; i--)
5215            {            {
5216            uschar *hc;            pcre_uchar *hc;
5217            uschar *this_hwm = cd->hwm;            pcre_uchar *this_hwm = cd->hwm;
5218    
5219            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
5220    
# Line 4640  for (;; ptr++) Line 5230  for (;; ptr++)
5230              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
5231              }              }
5232    
5233            memcpy(code, previous, len);            memcpy(code, previous, IN_UCHARS(len));
5234    
5235              /* Ensure there is enough workspace for forward references before
5236              copying them. */
5237    
5238              while (cd->hwm > cd->start_workspace + cd->workspace_size -
5239                     WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5240                {
5241                int save_offset = save_hwm - cd->start_workspace;
5242                int this_offset = this_hwm - cd->start_workspace;
5243                *errorcodeptr = expand_workspace(cd);
5244                if (*errorcodeptr != 0) goto FAILED;
5245                save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5246                this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5247                }
5248    
5249            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5250              {              {
5251              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
# Line 4657  for (;; ptr++) Line 5262  for (;; ptr++)
5262            {            {
5263            int oldlinkoffset;            int oldlinkoffset;
5264            int offset = (int)(code - bralink + 1);            int offset = (int)(code - bralink + 1);
5265            uschar *bra = code - offset;            pcre_uchar *bra = code - offset;
5266            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
5267            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
5268            *code++ = OP_KET;            *code++ = OP_KET;
# Line 4666  for (;; ptr++) Line 5271  for (;; ptr++)
5271            }            }
5272          }          }
5273    
5274        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
5275        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
5276        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
5277        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5278          deal with possessive ONCEs specially.
5279    
5280        Then, when we are doing the actual compile phase, check to see whether        Otherwise, when we are doing the actual compile phase, check to see
5281        this group is a non-atomic one that could match an empty string. If so,        whether this group is one that could match an empty string. If so,
5282        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
5283        that runtime checking can be done. [This check is also applied to        that runtime checking can be done. [This check is also applied to ONCE
5284        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
5285    
5286          Then, if the quantifier was possessive and the bracket is not a
5287          conditional, we convert the BRA code to the POS form, and the KET code to
5288          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5289          subpattern at both the start and at the end.) The use of special opcodes
5290          makes it possible to reduce greatly the stack usage in pcre_exec(). If
5291          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5292    
5293          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5294          flag so that the default action below, of wrapping everything inside
5295          atomic brackets, does not happen. When the minimum is greater than 1,
5296          there will be earlier copies of the group, and so we still have to wrap
5297          the whole thing. */
5298    
5299        else        else
5300          {          {
5301          uschar *ketcode = code - ketoffset;          pcre_uchar *ketcode = code - 1 - LINK_SIZE;
5302          uschar *bracode = ketcode - GET(ketcode, 1);          pcre_uchar *bracode = ketcode - GET(ketcode, 1);
5303          *ketcode = OP_KETRMAX + repeat_type;  
5304          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5305    
5306            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5307                possessive_quantifier) *bracode = OP_BRA;
5308    
5309            /* For non-possessive ONCE brackets, all we need to do is to
5310            set the KET. */
5311    
5312            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5313              *ketcode = OP_KETRMAX + repeat_type;
5314    
5315            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5316            converted to non-capturing above). */
5317    
5318            else
5319            {            {
5320            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5321            do  
5322              if (lengthptr == NULL)
5323              {              {
5324              if (could_be_empty_branch(scode, ketcode, utf8, cd))              pcre_uchar *scode = bracode;
5325                do
5326                {                {
5327                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf, cd))
5328                break;                  {
5329                    *bracode += OP_SBRA - OP_BRA;
5330                    break;
5331                    }
5332                  scode += GET(scode, 1);
5333                  }
5334                while (*scode == OP_ALT);
5335                }
5336    
5337              /* Handle possessive quantifiers. */
5338    
5339              if (possessive_quantifier)
5340                {
5341                /* For COND brackets, we wrap the whole thing in a possessively
5342                repeated non-capturing bracket, because we have not invented POS
5343                versions of the COND opcodes. Because we are moving code along, we
5344                must ensure that any pending recursive references are updated. */
5345    
5346                if (*bracode == OP_COND || *bracode == OP_SCOND)
5347                  {
5348                  int nlen = (int)(code - bracode);
5349                  *code = OP_END;
5350                  adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
5351                  memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
5352                  code += 1 + LINK_SIZE;
5353                  nlen += 1 + LINK_SIZE;
5354                  *bracode = OP_BRAPOS;
5355                  *code++ = OP_KETRPOS;
5356                  PUTINC(code, 0, nlen);
5357                  PUT(bracode, 1, nlen);
5358                }                }
5359              scode += GET(scode, 1);  
5360                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5361    
5362                else
5363                  {
5364                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5365                  *ketcode = OP_KETRPOS;
5366                  }
5367    
5368                /* If the minimum is zero, mark it as possessive, then unset the
5369                possessive flag when the minimum is 0 or 1. */
5370    
5371                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5372                if (repeat_min < 2) possessive_quantifier = FALSE;
5373              }              }
5374            while (*scode == OP_ALT);  
5375              /* Non-possessive quantifier */
5376    
5377              else *ketcode = OP_KETRMAX + repeat_type;
5378            }            }
5379          }          }
5380        }        }
# Line 4715  for (;; ptr++) Line 5395  for (;; ptr++)
5395        }        }
5396    
5397      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
5398      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5399      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
5400      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5401      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
5402      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
5403      tempcode, not at previous, which might be the first part of a string whose  
5404      (former) last char we repeated.      Some (but not all) possessively repeated subpatterns have already been
5405        completely handled in the code just above. For them, possessive_quantifier