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

Diff of /code/trunk/pcre_compile.c

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

revision 579 by ph10, Wed Nov 24 17:39:25 2010 UTC revision 1055 by chpe, Tue Oct 16 15:53:30 2012 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2010 University of Cambridge             Copyright (c) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_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    
69  /* Macro for setting individual bits in class bitmaps. */  /* Macro for setting individual bits in class bitmaps. */
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73  /* Maximum length value to check against when making sure that the integer that  /* Maximum length value to check against when making sure that the integer that
74  holds the compiled pattern length does not overflow. We make it a bit less than  holds the compiled pattern length does not overflow. We make it a bit less than
# Line 73  to check them every time. */ Line 77  to check them every time. */
77    
78  #define OFLOW_MAX (INT_MAX - 20)  #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL,
88        int, int, int *, int *, branch_chain *, compile_data *, int *);
89    
90    
91    
92  /*************************************************  /*************************************************
93  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 88  so this number is very generous. Line 103  so this number is very generous.
103  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
104  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
105  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
106  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
107    filled up by repetitions of forward references, for example patterns like
108    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
109    that the workspace is expanded using malloc() in this situation. The value
110    below is therefore a minimum, and we put a maximum on it for safety. The
111    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
112    kicks in at the same number of forward references in all cases. */
113    
114  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
115    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
116    
117  /* 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
118  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. */
119    
120  #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)  #define WORK_SIZE_SAFETY_MARGIN (100)
121    
122    /* Private flags added to firstchar and reqchar. */
123    
124    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
125    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
126    #define REQ_MASK       (REQ_CASELESS | REQ_VARY)
127    
128    /* Repeated character flags. */
129    
130    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
131    
132  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
133  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 262  static const char posix_names[] =
262    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
263    STRING_word0  STRING_xdigit;    STRING_word0  STRING_xdigit;
264    
265  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
266    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 };
267    
268  /* 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 297  substitutes must be in the order of the
297  both positive and negative cases. NULL means no substitute. */  both positive and negative cases. NULL means no substitute. */
298    
299  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
300  static const uschar *substitutes[] = {  static const pcre_uchar string_PNd[]  = {
301    (uschar *)"\\P{Nd}",    /* \D */    CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
302    (uschar *)"\\p{Nd}",    /* \d */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
303    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */  static const pcre_uchar string_pNd[]  = {
304    (uschar *)"\\p{Xsp}",   /* \s */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
305    (uschar *)"\\P{Xwd}",   /* \W */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    (uschar *)"\\p{Xwd}"    /* \w */  static const pcre_uchar string_PXsp[] = {
307      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
308      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
309    static const pcre_uchar string_pXsp[] = {
310      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
311      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
312    static const pcre_uchar string_PXwd[] = {
313      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
314      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
315    static const pcre_uchar string_pXwd[] = {
316      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
317      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318    
319    static const pcre_uchar *substitutes[] = {
320      string_PNd,           /* \D */
321      string_pNd,           /* \d */
322      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
323      string_pXsp,          /* \s */
324      string_PXwd,          /* \W */
325      string_pXwd           /* \w */
326  };  };
327    
328  static const uschar *posix_substitutes[] = {  static const pcre_uchar string_pL[] =   {
329    (uschar *)"\\p{L}",     /* alpha */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
330    (uschar *)"\\p{Ll}",    /* lower */    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    (uschar *)"\\p{Lu}",    /* upper */  static const pcre_uchar string_pLl[] =  {
332    (uschar *)"\\p{Xan}",   /* alnum */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333    NULL,                   /* ascii */    CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    (uschar *)"\\h",        /* blank */  static const pcre_uchar string_pLu[] =  {
335    NULL,                   /* cntrl */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
336    (uschar *)"\\p{Nd}",    /* digit */    CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    NULL,                   /* graph */  static const pcre_uchar string_pXan[] = {
338    NULL,                   /* print */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339    NULL,                   /* punct */    CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */  static const pcre_uchar string_h[] =    {
341    (uschar *)"\\p{Xwd}",   /* word */    CHAR_BACKSLASH, CHAR_h, '\0' };
342    NULL,                   /* xdigit */  static const pcre_uchar string_pXps[] = {
343      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
344      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
345    static const pcre_uchar string_PL[] =   {
346      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
347      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
348    static const pcre_uchar string_PLl[] =  {
349      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
350      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
351    static const pcre_uchar string_PLu[] =  {
352      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
353      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
354    static const pcre_uchar string_PXan[] = {
355      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
356      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
357    static const pcre_uchar string_H[] =    {
358      CHAR_BACKSLASH, CHAR_H, '\0' };
359    static const pcre_uchar string_PXps[] = {
360      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
361      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
362    
363    static const pcre_uchar *posix_substitutes[] = {
364      string_pL,            /* alpha */
365      string_pLl,           /* lower */
366      string_pLu,           /* upper */
367      string_pXan,          /* alnum */
368      NULL,                 /* ascii */
369      string_h,             /* blank */
370      NULL,                 /* cntrl */
371      string_pNd,           /* digit */
372      NULL,                 /* graph */
373      NULL,                 /* print */
374      NULL,                 /* punct */
375      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
376      string_pXwd,          /* word */
377      NULL,                 /* xdigit */
378    /* Negated cases */    /* Negated cases */
379    (uschar *)"\\P{L}",     /* ^alpha */    string_PL,            /* ^alpha */
380    (uschar *)"\\P{Ll}",    /* ^lower */    string_PLl,           /* ^lower */
381    (uschar *)"\\P{Lu}",    /* ^upper */    string_PLu,           /* ^upper */
382    (uschar *)"\\P{Xan}",   /* ^alnum */    string_PXan,          /* ^alnum */
383    NULL,                   /* ^ascii */    NULL,                 /* ^ascii */
384    (uschar *)"\\H",        /* ^blank */    string_H,             /* ^blank */
385    NULL,                   /* ^cntrl */    NULL,                 /* ^cntrl */
386    (uschar *)"\\P{Nd}",    /* ^digit */    string_PNd,           /* ^digit */
387    NULL,                   /* ^graph */    NULL,                 /* ^graph */
388    NULL,                   /* ^print */    NULL,                 /* ^print */
389    NULL,                   /* ^punct */    NULL,                 /* ^punct */
390    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
391    (uschar *)"\\P{Xwd}",   /* ^word */    string_PXwd,          /* ^word */
392    NULL                    /* ^xdigit */    NULL                  /* ^xdigit */
393  };  };
394  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
395  #endif  #endif
396    
397  #define STRING(a)  # a  #define STRING(a)  # a
# Line 365  static const char error_texts[] = Line 450  static const char error_texts[] =
450    /* 30 */    /* 30 */
451    "unknown POSIX class name\0"    "unknown POSIX class name\0"
452    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
453    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
454    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
455    "character value in \\x{...} sequence is too large\0"    "character value in \\x{...} sequence is too large\0"
456    /* 35 */    /* 35 */
# Line 388  static const char error_texts[] = Line 473  static const char error_texts[] =
473    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
474    /* 50 */    /* 50 */
475    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
476    "octal value is greater than \\377 (not in UTF-8 mode)\0"    "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
477    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
478    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
479    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
480    /* 55 */    /* 55 */
481    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
482    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
483    "\\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"
484    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
# Line 407  static const char error_texts[] = Line 492  static const char error_texts[] =
492    /* 65 */    /* 65 */
493    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
494    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
495    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with Unicode property support\0"
496    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
497      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
498      /* 70 */
499      "internal error: unknown opcode in find_fixedlength()\0"
500      "\\N is not supported in a class\0"
501      "too many forward references\0"
502      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
503      "invalid UTF-16 string\0"
504      /* 75 */
505      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
506      "character value in \\u.... sequence is too large\0"
507      "invalid UTF-32 string\0"
508    ;    ;
509    
510  /* 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 523  For convenience, we use the same bit def
523    
524  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
525    
526    /* Using a simple comparison for decimal numbers rather than a memory read
527    is much faster, and the resulting code is simpler (the compiler turns it
528    into a subtraction and unsigned comparison). */
529    
530    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
531    
532  #ifndef EBCDIC  #ifndef EBCDIC
533    
534  /* 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
535  UTF-8 mode. */  UTF-8 mode. */
536    
537  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
538    {    {
539    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
540    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 573  static const unsigned char digitab[] =
573    
574  /* 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. */
575    
576  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
577    {    {
578    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
579    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 608  static const unsigned char digitab[] =
608    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
609    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
610    
611  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
612    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
613    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
614    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 542  static const unsigned char ebcdic_charta Line 644  static const unsigned char ebcdic_charta
644  #endif  #endif
645    
646    
 /* Definition to allow mutual recursion */  
   
 static BOOL  
   compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,  
     int *, int *, branch_chain *, compile_data *, int *);  
   
647    
648    
649  /*************************************************  /*************************************************
# Line 577  return s; Line 673  return s;
673    
674    
675  /*************************************************  /*************************************************
676    *           Expand the workspace                 *
677    *************************************************/
678    
679    /* This function is called during the second compiling phase, if the number of
680    forward references fills the existing workspace, which is originally a block on
681    the stack. A larger block is obtained from malloc() unless the ultimate limit
682    has been reached or the increase will be rather small.
683    
684    Argument: pointer to the compile data block
685    Returns:  0 if all went well, else an error number
686    */
687    
688    static int
689    expand_workspace(compile_data *cd)
690    {
691    pcre_uchar *newspace;
692    int newsize = cd->workspace_size * 2;
693    
694    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
695    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
696        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
697     return ERR72;
698    
699    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
700    if (newspace == NULL) return ERR21;
701    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
702    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
703    if (cd->workspace_size > COMPILE_WORK_SIZE)
704      (PUBL(free))((void *)cd->start_workspace);
705    cd->start_workspace = newspace;
706    cd->workspace_size = newsize;
707    return 0;
708    }
709    
710    
711    
712    /*************************************************
713    *            Check for counted repeat            *
714    *************************************************/
715    
716    /* This function is called when a '{' is encountered in a place where it might
717    start a quantifier. It looks ahead to see if it really is a quantifier or not.
718    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
719    where the ddds are digits.
720    
721    Arguments:
722      p         pointer to the first char after '{'
723    
724    Returns:    TRUE or FALSE
725    */
726    
727    static BOOL
728    is_counted_repeat(const pcre_uchar *p)
729    {
730    if (!IS_DIGIT(*p)) return FALSE;
731    p++;
732    while (IS_DIGIT(*p)) p++;
733    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
734    
735    if (*p++ != CHAR_COMMA) return FALSE;
736    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
737    
738    if (!IS_DIGIT(*p)) return FALSE;
739    p++;
740    while (IS_DIGIT(*p)) p++;
741    
742    return (*p == CHAR_RIGHT_CURLY_BRACKET);
743    }
744    
745    
746    
747    /*************************************************
748  *            Handle escapes                      *  *            Handle escapes                      *
749  *************************************************/  *************************************************/
750    
# Line 601  Returns:         zero or positive => a d Line 769  Returns:         zero or positive => a d
769  */  */
770    
771  static int  static int
772  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
773    int options, BOOL isclass)    int options, BOOL isclass)
774  {  {
775  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
776  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
777  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
778    pcre_int32 c;
779    int i;
780    
781  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
782  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 790  in a table. A non-zero result is somethi
790  Otherwise further processing may be required. */  Otherwise further processing may be required. */
791    
792  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
793  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */  /* Not alphanumeric */
794    else if (c < CHAR_0 || c > CHAR_z) {}
795  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
796    
797  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
798  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
799    else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
800  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
801  #endif  #endif
802    
# Line 632  else if ((i = escapes[c - 0x48]) != 0) Line 804  else if ((i = escapes[c - 0x48]) != 0)
804    
805  else  else
806    {    {
807    const uschar *oldptr;    const pcre_uchar *oldptr;
808    BOOL braced, negated;    BOOL braced, negated;
809    
810    switch (c)    switch (c)
# Line 642  else Line 814  else
814    
815      case CHAR_l:      case CHAR_l:
816      case CHAR_L:      case CHAR_L:
817        *errorcodeptr = ERR37;
818        break;
819    
820      case CHAR_u:      case CHAR_u:
821        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
822          {
823          /* In JavaScript, \u must be followed by four hexadecimal numbers.
824          Otherwise it is a lowercase u letter. */
825          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
826            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
827            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
828            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
829            {
830            c = 0;
831            for (i = 0; i < 4; ++i)
832              {
833              register int cc = *(++ptr);
834    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
835              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
836              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
837    #else           /* EBCDIC coding */
838              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
839              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
840    #endif
841              }
842    
843    #if defined COMPILE_PCRE8
844            if (c > (utf ? 0x10ffff : 0xff))
845    #elif defined COMPILE_PCRE16
846            if (c > (utf ? 0x10ffff : 0xffff))
847    #elif defined COMPILE_PCRE32
848            if (utf && c > 0x10ffff)
849    #endif
850              {
851              *errorcodeptr = ERR76;
852              }
853            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
854            }
855          }
856        else
857          *errorcodeptr = ERR37;
858        break;
859    
860      case CHAR_U:      case CHAR_U:
861      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
862        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
863      break;      break;
864    
865      /* \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
866        class, \g must be followed by one of a number of specific things:
867    
868      (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
869      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 880  else
880      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
881    
882      case CHAR_g:      case CHAR_g:
883        if (isclass) break;
884      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
885        {        {
886        c = -ESC_g;        c = -ESC_g;
# Line 674  else Line 891  else
891    
892      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
893        {        {
894        const uschar *p;        const pcre_uchar *p;
895        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
896          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
897        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
898          {          {
899          c = -ESC_k;          c = -ESC_k;
# Line 694  else Line 911  else
911        }        }
912      else negated = FALSE;      else negated = FALSE;
913    
914        /* The integer range is limited by the machine's int representation. */
915      c = 0;      c = 0;
916      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
917          {
918          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
919            {
920            c = -1;
921            break;
922            }
923        c = c * 10 + *(++ptr) - CHAR_0;        c = c * 10 + *(++ptr) - CHAR_0;
924          }
925      if (c < 0)   /* Integer overflow */      if (((unsigned int)c) > INT_MAX) /* Integer overflow */
926        {        {
927          while (IS_DIGIT(ptr[1]))
928            ptr++;
929        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
930        break;        break;
931        }        }
# Line 747  else Line 973  else
973      if (!isclass)      if (!isclass)
974        {        {
975        oldptr = ptr;        oldptr = ptr;
976          /* The integer range is limited by the machine's int representation. */
977        c -= CHAR_0;        c -= CHAR_0;
978        while ((digitab[ptr[1]] & ctype_digit) != 0)        while (IS_DIGIT(ptr[1]))
979            {
980            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
981              {
982              c = -1;
983              break;
984              }
985          c = c * 10 + *(++ptr) - CHAR_0;          c = c * 10 + *(++ptr) - CHAR_0;
986        if (c < 0)    /* Integer overflow */          }
987          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
988          {          {
989            while (IS_DIGIT(ptr[1]))
990              ptr++;
991          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
992          break;          break;
993          }          }
# Line 777  else Line 1013  else
1013      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1014      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
1015      significant 8 bits of octal numbers (I think this is what early Perls used      significant 8 bits of octal numbers (I think this is what early Perls used
1016      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1017      than 3 octal digits. */      but no more than 3 octal digits. */
1018    
1019      case CHAR_0:      case CHAR_0:
1020      c -= CHAR_0;      c -= CHAR_0;
1021      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1022          c = c * 8 + *(++ptr) - CHAR_0;          c = c * 8 + *(++ptr) - CHAR_0;
1023      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1024        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1025    #endif
1026      break;      break;
1027    
1028      /* \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
1029      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.
1030      treated as a data character. */      If not, { is treated as a data character. */
1031    
1032      case CHAR_x:      case CHAR_x:
1033        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1034          {
1035          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1036          Otherwise it is a lowercase x letter. */
1037          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1038            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1039            {
1040            c = 0;
1041            for (i = 0; i < 2; ++i)
1042              {
1043              register int cc = *(++ptr);
1044    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1045              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1046              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1047    #else           /* EBCDIC coding */
1048              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1049              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1050    #endif
1051              }
1052            }
1053          break;
1054          }
1055    
1056      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1057        {        {
1058        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1059    
1060        c = 0;        c = 0;
1061        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1062          {          {
1063          register int cc = *pt++;          register int cc = *pt++;
1064          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
1065    
1066  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1067          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
# Line 811  else Line 1070  else
1070          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 */
1071          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1072  #endif  #endif
1073    
1074    #if defined COMPILE_PCRE8
1075            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1076    #elif defined COMPILE_PCRE16
1077            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1078    #elif defined COMPILE_PCRE32
1079            if (utf && c > 0x10ffff) { c = -1; break; }
1080    #endif
1081            }
1082    
1083          if (c < 0)
1084            {
1085            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1086            *errorcodeptr = ERR34;
1087          }          }
1088    
1089        if (*pt == CHAR_RIGHT_CURLY_BRACKET)        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1090          {          {
1091          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1092          ptr = pt;          ptr = pt;
1093          break;          break;
1094          }          }
# Line 827  else Line 1100  else
1100      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1101    
1102      c = 0;      c = 0;
1103      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1104        {        {
1105        int cc;                                  /* Some compilers don't like */        int cc;                                  /* Some compilers don't like */
1106        cc = *(++ptr);                           /* ++ in initializers */        cc = *(++ptr);                           /* ++ in initializers */
# Line 885  else Line 1158  else
1158    }    }
1159    
1160  /* 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
1161  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
1162    quantification such as \N{2,3}. */
1163    
1164  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1165         !is_counted_repeat(ptr+2))
1166    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1167    
1168  /* 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 1198  Returns:         type value from ucp_typ
1198  */  */
1199    
1200  static int  static int
1201  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1202  {  {
1203  int c, i, bot, top;  int c, i, bot, top;
1204  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1205  char name[32];  pcre_uchar name[32];
1206    
1207  c = *(++ptr);  c = *(++ptr);
1208  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 944  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1219  if (c == CHAR_LEFT_CURLY_BRACKET)
1219      *negptr = TRUE;      *negptr = TRUE;
1220      ptr++;      ptr++;
1221      }      }
1222    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1223      {      {
1224      c = *(++ptr);      c = *(++ptr);
1225      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
# Line 968  else Line 1243  else
1243  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1244    
1245  bot = 0;  bot = 0;
1246  top = _pcre_utt_size;  top = PRIV(utt_size);
1247    
1248  while (bot < top)  while (bot < top)
1249    {    {
1250    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1251    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1252    if (c == 0)    if (c == 0)
1253      {      {
1254      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1255      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1256      }      }
1257    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1258    }    }
# Line 997  return -1; Line 1272  return -1;
1272    
1273    
1274  /*************************************************  /*************************************************
 *            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);  
 }  
   
   
   
 /*************************************************  
1275  *         Read repeat counts                     *  *         Read repeat counts                     *
1276  *************************************************/  *************************************************/
1277    
# Line 1048  Returns:         pointer to '}' on succe Line 1290  Returns:         pointer to '}' on succe
1290                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1291  */  */
1292    
1293  static const uschar *  static const pcre_uchar *
1294  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)
1295  {  {
1296  int min = 0;  int min = 0;
1297  int max = -1;  int max = -1;
# Line 1057  int max = -1; Line 1299  int max = -1;
1299  /* 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
1300  an integer overflow. */  an integer overflow. */
1301    
1302  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1303  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1304    {    {
1305    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 1072  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1314  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1314    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1315      {      {
1316      max = 0;      max = 0;
1317      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1318      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1319        {        {
1320        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 1127  Arguments: Line 1369  Arguments:
1369    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1370    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1371    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1372    utf8         TRUE if we are in UTF-8 mode    utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1373    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1374    
1375  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1376  */  */
1377    
1378  static int  static int
1379  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,
1380    BOOL xmode, BOOL utf8, int *count)    BOOL xmode, BOOL utf, int *count)
1381  {  {
1382  uschar *ptr = *ptrptr;  pcre_uchar *ptr = *ptrptr;
1383  int start_count = *count;  int start_count = *count;
1384  int hwm_count = start_count;  int hwm_count = start_count;
1385  BOOL dup_parens = FALSE;  BOOL dup_parens = FALSE;
# Line 1204  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1446  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1446          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1447        {        {
1448        int term;        int term;
1449        const uschar *thisname;        const pcre_uchar *thisname;
1450        *count += 1;        *count += 1;
1451        if (name == NULL && *count == lorn) return *count;        if (name == NULL && *count == lorn) return *count;
1452        term = *ptr++;        term = *ptr++;
# Line 1212  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1454  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1454        thisname = ptr;        thisname = ptr;
1455        while (*ptr != term) ptr++;        while (*ptr != term) ptr++;
1456        if (name != NULL && lorn == ptr - thisname &&        if (name != NULL && lorn == ptr - thisname &&
1457            strncmp((const char *)name, (const char *)thisname, lorn) == 0)            STRNCMP_UC_UC(name, thisname, lorn) == 0)
1458          return *count;          return *count;
1459        term++;        term++;
1460        }        }
# Line 1255  for (; ptr < cd->end_pattern; ptr++) Line 1497  for (; ptr < cd->end_pattern; ptr++)
1497          {          {
1498          if (ptr[2] == CHAR_E)          if (ptr[2] == CHAR_E)
1499            ptr+= 2;            ptr+= 2;
1500          else if (strncmp((const char *)ptr+2,          else if (STRNCMP_UC_C8(ptr + 2,
1501                   STR_Q STR_BACKSLASH STR_E, 3) == 0)                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
1502            ptr += 4;            ptr += 4;
1503          else          else
# Line 1303  for (; ptr < cd->end_pattern; ptr++) Line 1545  for (; ptr < cd->end_pattern; ptr++)
1545        {        {
1546        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1547        ptr++;        ptr++;
1548  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1549        if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;        if (utf) FORWARDCHAR(ptr);
1550  #endif  #endif
1551        }        }
1552      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
# Line 1315  for (; ptr < cd->end_pattern; ptr++) Line 1557  for (; ptr < cd->end_pattern; ptr++)
1557    
1558    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1559      {      {
1560      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1561      if (rc > 0) return rc;      if (rc > 0) return rc;
1562      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1563      }      }
# Line 1361  Arguments: Line 1603  Arguments:
1603    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1604    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1605    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1606    utf8         TRUE if we are in UTF-8 mode    utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1607    
1608  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1609  */  */
1610    
1611  static int  static int
1612  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,
1613    BOOL utf8)    BOOL utf)
1614  {  {
1615  uschar *ptr = (uschar *)cd->start_pattern;  pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1616  int count = 0;  int count = 0;
1617  int rc;  int rc;
1618    
# Line 1381  matching closing parens. That is why we Line 1623  matching closing parens. That is why we
1623    
1624  for (;;)  for (;;)
1625    {    {
1626    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1627    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1628    }    }
1629    
# Line 1397  return rc; Line 1639  return rc;
1639    
1640  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1641  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
1642  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
1643  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
1644  assertions, and also the \b assertion; for others it does not.  does not.
1645    
1646  Arguments:  Arguments:
1647    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  
1648    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1649    
1650  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1651  */  */
1652    
1653  static const uschar*  static const pcre_uchar*
1654  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1655  {  {
1656  for (;;)  for (;;)
1657    {    {
1658    switch ((int)*code)    switch ((int)*code)
1659      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1660      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1661      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1662      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1663      if (!skipassert) return code;      if (!skipassert) return code;
1664      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1665      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1666      break;      break;
1667    
1668      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1444  for (;;) Line 1676  for (;;)
1676      case OP_RREF:      case OP_RREF:
1677      case OP_NRREF:      case OP_NRREF:
1678      case OP_DEF:      case OP_DEF:
1679      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1680      break;      break;
1681    
1682      default:      default:
# Line 1474  and doing the check at the end; a flag s Line 1706  and doing the check at the end; a flag s
1706    
1707  Arguments:  Arguments:
1708    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1709    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1710    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1711    cd       the "compile data" structure    cd       the "compile data" structure
1712    
1713  Returns:   the fixed length,  Returns:   the fixed length,
1714               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1715               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1716               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
1717                 or -4 if an unknown opcode was encountered (internal error)
1718  */  */
1719    
1720  static int  static int
1721  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1722  {  {
1723  int length = -1;  int length = -1;
1724    
1725  register int branchlength = 0;  register int branchlength = 0;
1726  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1727    
1728  /* 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
1729  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1498  branch, check the length against that of Line 1731  branch, check the length against that of
1731  for (;;)  for (;;)
1732    {    {
1733    int d;    int d;
1734    uschar *ce, *cs;    pcre_uchar *ce, *cs;
1735    register int op = *cc;    register int op = *cc;
1736    
1737    switch (op)    switch (op)
1738      {      {
1739        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1740        OP_BRA (normal non-capturing bracket) because the other variants of these
1741        opcodes are all concerned with unlimited repeated groups, which of course
1742        are not of fixed length. */
1743    
1744      case OP_CBRA:      case OP_CBRA:
1745      case OP_BRA:      case OP_BRA:
1746      case OP_ONCE:      case OP_ONCE:
1747        case OP_ONCE_NC:
1748      case OP_COND:      case OP_COND:
1749      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);
1750      if (d < 0) return d;      if (d < 0) return d;
1751      branchlength += d;      branchlength += d;
1752      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1753      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1754      break;      break;
1755    
1756      /* 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.
1757      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
1758      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
1759        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1760        because they all imply an unlimited repeat. */
1761    
1762      case OP_ALT:      case OP_ALT:
1763      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1764      case OP_END:      case OP_END:
1765        case OP_ACCEPT:
1766        case OP_ASSERT_ACCEPT:
1767      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1768        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1769      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1535  for (;;) Line 1777  for (;;)
1777    
1778      case OP_RECURSE:      case OP_RECURSE:
1779      if (!atend) return -3;      if (!atend) return -3;
1780      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1781      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1782      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                    /* Recursion */
1783      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1784      if (d < 0) return d;      if (d < 0) return d;
1785      branchlength += d;      branchlength += d;
1786      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1551  for (;;) Line 1793  for (;;)
1793      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1794      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1795      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1796      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1797        break;
1798    
1799      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1800    
1801      case OP_REVERSE:      case OP_MARK:
1802        case OP_PRUNE_ARG:
1803        case OP_SKIP_ARG:
1804        case OP_THEN_ARG:
1805        cc += cc[1] + PRIV(OP_lengths)[*cc];
1806        break;
1807    
1808        case OP_CALLOUT:
1809        case OP_CIRC:
1810        case OP_CIRCM:
1811        case OP_CLOSE:
1812        case OP_COMMIT:
1813      case OP_CREF:      case OP_CREF:
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
1814      case OP_DEF:      case OP_DEF:
1815      case OP_OPT:      case OP_DOLL:
1816      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
1817      case OP_EOD:      case OP_EOD:
1818      case OP_EODN:      case OP_EODN:
1819      case OP_CIRC:      case OP_FAIL:
1820      case OP_DOLL:      case OP_NCREF:
1821        case OP_NRREF:
1822      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1823        case OP_PRUNE:
1824        case OP_REVERSE:
1825        case OP_RREF:
1826        case OP_SET_SOM:
1827        case OP_SKIP:
1828        case OP_SOD:
1829        case OP_SOM:
1830        case OP_THEN:
1831      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1832      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1833      break;      break;
1834    
1835      /* Handle literal characters */      /* Handle literal characters */
1836    
1837      case OP_CHAR:      case OP_CHAR:
1838      case OP_CHARNC:      case OP_CHARI:
1839      case OP_NOT:      case OP_NOT:
1840        case OP_NOTI:
1841      branchlength++;      branchlength++;
1842      cc += 2;      cc += 2;
1843  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1844      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];  
1845  #endif  #endif
1846      break;      break;
1847    
# Line 1592  for (;;) Line 1849  for (;;)
1849      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1850    
1851      case OP_EXACT:      case OP_EXACT:
1852        case OP_EXACTI:
1853        case OP_NOTEXACT:
1854        case OP_NOTEXACTI:
1855      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1856      cc += 4;      cc += 2 + IMM2_SIZE;
1857  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1858      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];  
1859  #endif  #endif
1860      break;      break;
1861    
1862      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1863      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1864      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1865      cc += 4;        cc += 2;
1866        cc += 1 + IMM2_SIZE + 1;
1867      break;      break;
1868    
1869      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1613  for (;;) Line 1873  for (;;)
1873      cc += 2;      cc += 2;
1874      /* Fall through */      /* Fall through */
1875    
1876        case OP_HSPACE:
1877        case OP_VSPACE:
1878        case OP_NOT_HSPACE:
1879        case OP_NOT_VSPACE:
1880      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1881      case OP_DIGIT:      case OP_DIGIT:
1882      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1625  for (;;) Line 1889  for (;;)
1889      cc++;      cc++;
1890      break;      break;
1891    
1892      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1893        otherwise \C is coded as OP_ALLANY. */
1894    
1895      case OP_ANYBYTE:      case OP_ANYBYTE:
1896      return -2;      return -2;
1897    
1898      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1899    
1900  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1901      case OP_XCLASS:      case OP_XCLASS:
1902      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1903      /* Fall through */      /* Fall through */
1904  #endif  #endif
1905    
1906      case OP_CLASS:      case OP_CLASS:
1907      case OP_NCLASS:      case OP_NCLASS:
1908      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1909    
1910      switch (*cc)      switch (*cc)
1911        {        {
1912          case OP_CRPLUS:
1913          case OP_CRMINPLUS:
1914        case OP_CRSTAR:        case OP_CRSTAR:
1915        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1916        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1652  for (;;) Line 1919  for (;;)
1919    
1920        case OP_CRRANGE:        case OP_CRRANGE:
1921        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1922        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1923        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1924        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1925        break;        break;
1926    
1927        default:        default:
# Line 1664  for (;;) Line 1931  for (;;)
1931    
1932      /* Anything else is variable length */      /* Anything else is variable length */
1933    
1934      default:      case OP_ANYNL:
1935        case OP_BRAMINZERO:
1936        case OP_BRAPOS:
1937        case OP_BRAPOSZERO:
1938        case OP_BRAZERO:
1939        case OP_CBRAPOS:
1940        case OP_EXTUNI:
1941        case OP_KETRMAX:
1942        case OP_KETRMIN:
1943        case OP_KETRPOS:
1944        case OP_MINPLUS:
1945        case OP_MINPLUSI:
1946        case OP_MINQUERY:
1947        case OP_MINQUERYI:
1948        case OP_MINSTAR:
1949        case OP_MINSTARI:
1950        case OP_MINUPTO:
1951        case OP_MINUPTOI:
1952        case OP_NOTMINPLUS:
1953        case OP_NOTMINPLUSI:
1954        case OP_NOTMINQUERY:
1955        case OP_NOTMINQUERYI:
1956        case OP_NOTMINSTAR:
1957        case OP_NOTMINSTARI:
1958        case OP_NOTMINUPTO:
1959        case OP_NOTMINUPTOI:
1960        case OP_NOTPLUS:
1961        case OP_NOTPLUSI:
1962        case OP_NOTPOSPLUS:
1963        case OP_NOTPOSPLUSI:
1964        case OP_NOTPOSQUERY:
1965        case OP_NOTPOSQUERYI:
1966        case OP_NOTPOSSTAR:
1967        case OP_NOTPOSSTARI:
1968        case OP_NOTPOSUPTO:
1969        case OP_NOTPOSUPTOI:
1970        case OP_NOTQUERY:
1971        case OP_NOTQUERYI:
1972        case OP_NOTSTAR:
1973        case OP_NOTSTARI:
1974        case OP_NOTUPTO:
1975        case OP_NOTUPTOI:
1976        case OP_PLUS:
1977        case OP_PLUSI:
1978        case OP_POSPLUS:
1979        case OP_POSPLUSI:
1980        case OP_POSQUERY:
1981        case OP_POSQUERYI:
1982        case OP_POSSTAR:
1983        case OP_POSSTARI:
1984        case OP_POSUPTO:
1985        case OP_POSUPTOI:
1986        case OP_QUERY:
1987        case OP_QUERYI:
1988        case OP_REF:
1989        case OP_REFI:
1990        case OP_SBRA:
1991        case OP_SBRAPOS:
1992        case OP_SCBRA:
1993        case OP_SCBRAPOS:
1994        case OP_SCOND:
1995        case OP_SKIPZERO:
1996        case OP_STAR:
1997        case OP_STARI:
1998        case OP_TYPEMINPLUS:
1999        case OP_TYPEMINQUERY:
2000        case OP_TYPEMINSTAR:
2001        case OP_TYPEMINUPTO:
2002        case OP_TYPEPLUS:
2003        case OP_TYPEPOSPLUS:
2004        case OP_TYPEPOSQUERY:
2005        case OP_TYPEPOSSTAR:
2006        case OP_TYPEPOSUPTO:
2007        case OP_TYPEQUERY:
2008        case OP_TYPESTAR:
2009        case OP_TYPEUPTO:
2010        case OP_UPTO:
2011        case OP_UPTOI:
2012      return -1;      return -1;
2013    
2014        /* Catch unrecognized opcodes so that when new ones are added they
2015        are not forgotten, as has happened in the past. */
2016    
2017        default:
2018        return -4;
2019      }      }
2020    }    }
2021  /* Control never gets here */  /* Control never gets here */
# Line 1686  length. Line 2036  length.
2036    
2037  Arguments:  Arguments:
2038    code        points to start of expression    code        points to start of expression
2039    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2040    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2041    
2042  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
2043  */  */
2044    
2045  const uschar *  const pcre_uchar *
2046  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2047  {  {
2048  for (;;)  for (;;)
2049    {    {
2050    register int c = *code;    register int c = *code;
2051    
2052    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2053    
2054    /* 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 1710  for (;;) Line 2061  for (;;)
2061    
2062    else if (c == OP_REVERSE)    else if (c == OP_REVERSE)
2063      {      {
2064      if (number < 0) return (uschar *)code;      if (number < 0) return (pcre_uchar *)code;
2065      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2066      }      }
2067    
2068    /* Handle capturing bracket */    /* Handle capturing bracket */
2069    
2070    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2071               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2072      {      {
2073      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2074      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2075      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2076      }      }
2077    
2078    /* 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 1748  for (;;) Line 2100  for (;;)
2100        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2101        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2102        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2103        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2104            code += 2;
2105        break;        break;
2106    
2107        case OP_MARK:        case OP_MARK:
# Line 1758  for (;;) Line 2111  for (;;)
2111        break;        break;
2112    
2113        case OP_THEN_ARG:        case OP_THEN_ARG:
2114        code += code[1+LINK_SIZE];        code += code[1];
2115        break;        break;
2116        }        }
2117    
2118      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2119    
2120      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2121    
2122    /* 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
2123    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
2124    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2125    
2126  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2127      if (utf8) switch(c)      if (utf) switch(c)
2128        {        {
2129        case OP_CHAR:        case OP_CHAR:
2130        case OP_CHARNC:        case OP_CHARI:
2131        case OP_EXACT:        case OP_EXACT:
2132          case OP_EXACTI:
2133        case OP_UPTO:        case OP_UPTO:
2134          case OP_UPTOI:
2135        case OP_MINUPTO:        case OP_MINUPTO:
2136          case OP_MINUPTOI:
2137        case OP_POSUPTO:        case OP_POSUPTO:
2138          case OP_POSUPTOI:
2139        case OP_STAR:        case OP_STAR:
2140          case OP_STARI:
2141        case OP_MINSTAR:        case OP_MINSTAR:
2142          case OP_MINSTARI:
2143        case OP_POSSTAR:        case OP_POSSTAR:
2144          case OP_POSSTARI:
2145        case OP_PLUS:        case OP_PLUS:
2146          case OP_PLUSI:
2147        case OP_MINPLUS:        case OP_MINPLUS:
2148          case OP_MINPLUSI:
2149        case OP_POSPLUS:        case OP_POSPLUS:
2150          case OP_POSPLUSI:
2151        case OP_QUERY:        case OP_QUERY:
2152          case OP_QUERYI:
2153        case OP_MINQUERY:        case OP_MINQUERY:
2154          case OP_MINQUERYI:
2155        case OP_POSQUERY:        case OP_POSQUERY:
2156        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2157          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2158        break;        break;
2159        }        }
2160  #else  #else
2161      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2162  #endif  #endif
2163      }      }
2164    }    }
# Line 1809  instance of OP_RECURSE. Line 2175  instance of OP_RECURSE.
2175    
2176  Arguments:  Arguments:
2177    code        points to start of expression    code        points to start of expression
2178    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2179    
2180  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
2181  */  */
2182    
2183  static const uschar *  static const pcre_uchar *
2184  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2185  {  {
2186  for (;;)  for (;;)
2187    {    {
# Line 1854  for (;;) Line 2220  for (;;)
2220        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2221        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2222        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2223        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2224            code += 2;
2225        break;        break;
2226    
2227        case OP_MARK:        case OP_MARK:
# Line 1864  for (;;) Line 2231  for (;;)
2231        break;        break;
2232    
2233        case OP_THEN_ARG:        case OP_THEN_ARG:
2234        code += code[1+LINK_SIZE];        code += code[1];
2235        break;        break;
2236        }        }
2237    
2238      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2239    
2240      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2241    
2242      /* 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
2243      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
2244      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2245    
2246  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2247      if (utf8) switch(c)      if (utf) switch(c)
2248        {        {
2249        case OP_CHAR:        case OP_CHAR:
2250        case OP_CHARNC:        case OP_CHARI:
2251          case OP_NOT:
2252          case OP_NOTI:
2253        case OP_EXACT:        case OP_EXACT:
2254          case OP_EXACTI:
2255          case OP_NOTEXACT:
2256          case OP_NOTEXACTI:
2257        case OP_UPTO:        case OP_UPTO:
2258          case OP_UPTOI:
2259          case OP_NOTUPTO:
2260          case OP_NOTUPTOI:
2261        case OP_MINUPTO:        case OP_MINUPTO:
2262          case OP_MINUPTOI:
2263          case OP_NOTMINUPTO:
2264          case OP_NOTMINUPTOI:
2265        case OP_POSUPTO:        case OP_POSUPTO:
2266          case OP_POSUPTOI:
2267          case OP_NOTPOSUPTO:
2268          case OP_NOTPOSUPTOI:
2269        case OP_STAR:        case OP_STAR:
2270          case OP_STARI:
2271          case OP_NOTSTAR:
2272          case OP_NOTSTARI:
2273        case OP_MINSTAR:        case OP_MINSTAR:
2274          case OP_MINSTARI:
2275          case OP_NOTMINSTAR:
2276          case OP_NOTMINSTARI:
2277        case OP_POSSTAR:        case OP_POSSTAR:
2278          case OP_POSSTARI:
2279          case OP_NOTPOSSTAR:
2280          case OP_NOTPOSSTARI:
2281        case OP_PLUS:        case OP_PLUS:
2282          case OP_PLUSI:
2283          case OP_NOTPLUS:
2284          case OP_NOTPLUSI:
2285        case OP_MINPLUS:        case OP_MINPLUS:
2286          case OP_MINPLUSI:
2287          case OP_NOTMINPLUS:
2288          case OP_NOTMINPLUSI:
2289        case OP_POSPLUS:        case OP_POSPLUS:
2290          case OP_POSPLUSI:
2291          case OP_NOTPOSPLUS:
2292          case OP_NOTPOSPLUSI:
2293        case OP_QUERY:        case OP_QUERY:
2294          case OP_QUERYI:
2295          case OP_NOTQUERY:
2296          case OP_NOTQUERYI:
2297        case OP_MINQUERY:        case OP_MINQUERY:
2298          case OP_MINQUERYI:
2299          case OP_NOTMINQUERY:
2300          case OP_NOTMINQUERYI:
2301        case OP_POSQUERY:        case OP_POSQUERY:
2302        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2303          case OP_NOTPOSQUERY:
2304          case OP_NOTPOSQUERYI:
2305          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2306        break;        break;
2307        }        }
2308  #else  #else
2309      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2310  #endif  #endif
2311      }      }
2312    }    }
# Line 1921  bracket whose current branch will alread Line 2329  bracket whose current branch will alread
2329  Arguments:  Arguments:
2330    code        points to start of search    code        points to start of search
2331    endcode     points to where to stop    endcode     points to where to stop
2332    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2333    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2334    
2335  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2336  */  */
2337    
2338  static BOOL  static BOOL
2339  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2340    compile_data *cd)    BOOL utf, compile_data *cd)
2341  {  {
2342  register int c;  register int c;
2343  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2344       code < endcode;       code < endcode;
2345       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2346    {    {
2347    const uschar *ccode;    const pcre_uchar *ccode;
2348    
2349    c = *code;    c = *code;
2350    
# Line 1950  for (code = first_significant_code(code Line 2358  for (code = first_significant_code(code
2358      continue;      continue;
2359      }      }
2360    
   /* 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;  
     }  
   
2361    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2362    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2363      forward reference subroutine call, we can't. To detect forward reference
2364      we have to scan up the list that is kept in the workspace. This function is
2365      called only when doing the real compile, not during the pre-compile that
2366      measures the size of the compiled pattern. */
2367    
2368    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2369      {      {
2370      BOOL empty_branch = FALSE;      const pcre_uchar *scode;
2371      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
2372    
2373        /* Test for forward reference */
2374    
2375        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2376          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2377    
2378        /* Not a forward reference, test for completed backward reference */
2379    
2380        empty_branch = FALSE;
2381        scode = cd->start_code + GET(code, 1);
2382      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2383    
2384        /* Completed backwards reference */
2385    
2386      do      do
2387        {        {
2388        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf, cd))
2389          {          {
2390          empty_branch = TRUE;          empty_branch = TRUE;
2391          break;          break;
# Line 1978  for (code = first_significant_code(code Line 2393  for (code = first_significant_code(code
2393        scode += GET(scode, 1);        scode += GET(scode, 1);
2394        }        }
2395      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2396    
2397      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2398      continue;      continue;
2399      }      }
2400    
2401      /* Groups with zero repeats can of course be empty; skip them. */
2402    
2403      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2404          c == OP_BRAPOSZERO)
2405        {
2406        code += PRIV(OP_lengths)[c];
2407        do code += GET(code, 1); while (*code == OP_ALT);
2408        c = *code;
2409        continue;
2410        }
2411    
2412      /* A nested group that is already marked as "could be empty" can just be
2413      skipped. */
2414    
2415      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2416          c == OP_SCBRA || c == OP_SCBRAPOS)
2417        {
2418        do code += GET(code, 1); while (*code == OP_ALT);
2419        c = *code;
2420        continue;
2421        }
2422    
2423    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2424    
2425    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2426          c == OP_CBRA || c == OP_CBRAPOS ||
2427          c == OP_ONCE || c == OP_ONCE_NC ||
2428          c == OP_COND)
2429      {      {
2430      BOOL empty_branch;      BOOL empty_branch;
2431      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2000  for (code = first_significant_code(code Line 2441  for (code = first_significant_code(code
2441        empty_branch = FALSE;        empty_branch = FALSE;
2442        do        do
2443          {          {
2444          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2445            empty_branch = TRUE;            empty_branch = TRUE;
2446          code += GET(code, 1);          code += GET(code, 1);
2447          }          }
# Line 2018  for (code = first_significant_code(code Line 2459  for (code = first_significant_code(code
2459      {      {
2460      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2461      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2462      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2463      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"
2464      here. */      here. */
2465    
2466  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2467      case OP_XCLASS:      case OP_XCLASS:
2468      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2469      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 2030  for (code = first_significant_code(code Line 2471  for (code = first_significant_code(code
2471    
2472      case OP_CLASS:      case OP_CLASS:
2473      case OP_NCLASS:      case OP_NCLASS:
2474      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2475    
2476  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2477      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2478  #endif  #endif
2479    
# Line 2071  for (code = first_significant_code(code Line 2512  for (code = first_significant_code(code
2512      case OP_ALLANY:      case OP_ALLANY:
2513      case OP_ANYBYTE:      case OP_ANYBYTE:
2514      case OP_CHAR:      case OP_CHAR:
2515      case OP_CHARNC:      case OP_CHARI:
2516      case OP_NOT:      case OP_NOT:
2517        case OP_NOTI:
2518      case OP_PLUS:      case OP_PLUS:
2519      case OP_MINPLUS:      case OP_MINPLUS:
2520      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2104  for (code = first_significant_code(code Line 2546  for (code = first_significant_code(code
2546      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2547      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2548      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2549      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2550          code += 2;
2551      break;      break;
2552    
2553      /* End of branch */      /* End of branch */
# Line 2112  for (code = first_significant_code(code Line 2555  for (code = first_significant_code(code
2555      case OP_KET:      case OP_KET:
2556      case OP_KETRMAX:      case OP_KETRMAX:
2557      case OP_KETRMIN:      case OP_KETRMIN:
2558        case OP_KETRPOS:
2559      case OP_ALT:      case OP_ALT:
2560      return TRUE;      return TRUE;
2561    
2562      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2563      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2564    
2565  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2566      case OP_STAR:      case OP_STAR:
2567        case OP_STARI:
2568      case OP_MINSTAR:      case OP_MINSTAR:
2569        case OP_MINSTARI:
2570      case OP_POSSTAR:      case OP_POSSTAR:
2571        case OP_POSSTARI:
2572      case OP_QUERY:      case OP_QUERY:
2573        case OP_QUERYI:
2574      case OP_MINQUERY:      case OP_MINQUERY:
2575        case OP_MINQUERYI:
2576      case OP_POSQUERY:      case OP_POSQUERY:
2577      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      case OP_POSQUERYI:
2578        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2579      break;      break;
2580    
2581      case OP_UPTO:      case OP_UPTO:
2582        case OP_UPTOI:
2583      case OP_MINUPTO:      case OP_MINUPTO:
2584        case OP_MINUPTOI:
2585      case OP_POSUPTO:      case OP_POSUPTO:
2586      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      case OP_POSUPTOI:
2587        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2588      break;      break;
2589  #endif  #endif
2590    
# Line 2145  for (code = first_significant_code(code Line 2598  for (code = first_significant_code(code
2598      break;      break;
2599    
2600      case OP_THEN_ARG:      case OP_THEN_ARG:
2601      code += code[1+LINK_SIZE];      code += code[1];
2602      break;      break;
2603    
2604      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
# Line 2168  return TRUE; Line 2621  return TRUE;
2621  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
2622  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,
2623  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.
2624    This function is called only during the real compile, not during the
2625    pre-compile.
2626    
2627  Arguments:  Arguments:
2628    code        points to start of the recursion    code        points to start of the recursion
2629    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2630    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2631    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2632    cd          pointers to tables etc    cd          pointers to tables etc
2633    
2634  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2635  */  */
2636    
2637  static BOOL  static BOOL
2638  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2639    BOOL utf8, compile_data *cd)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2640  {  {
2641  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2642    {    {
2643    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2644      return FALSE;      return FALSE;
2645    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2646    }    }
# Line 2218  where Perl recognizes it as the POSIX cl Line 2673  where Perl recognizes it as the POSIX cl
2673  "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,
2674  I think.  I think.
2675    
2676    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2677    It seems that the appearance of a nested POSIX class supersedes an apparent
2678    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2679    a digit.
2680    
2681    In Perl, unescaped square brackets may also appear as part of class names. For
2682    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2683    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2684    seem right at all. PCRE does not allow closing square brackets in POSIX class
2685    names.
2686    
2687  Arguments:  Arguments:
2688    ptr      pointer to the initial [    ptr      pointer to the initial [
2689    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2226  Returns:   TRUE or FALSE Line 2692  Returns:   TRUE or FALSE
2692  */  */
2693    
2694  static BOOL  static BOOL
2695  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2696  {  {
2697  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2698  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2699  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2700    {    {
2701    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2702        ptr++;
2703      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2704      else
2705      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2706      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2707        {        {
2708        *endptr = ptr;        *endptr = ptr;
2709        return TRUE;        return TRUE;
2710        }        }
2711        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2712             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2713              ptr[1] == CHAR_EQUALS_SIGN) &&
2714            check_posix_syntax(ptr, endptr))
2715          return FALSE;
2716      }      }
2717    }    }
2718  return FALSE;  return FALSE;
# Line 2263  Returns:     a value representing the na Line 2736  Returns:     a value representing the na
2736  */  */
2737    
2738  static int  static int
2739  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2740  {  {
2741  const char *pn = posix_names;  const char *pn = posix_names;
2742  register int yield = 0;  register int yield = 0;
2743  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2744    {    {
2745    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2746      strncmp((const char *)ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2747    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
2748    yield++;    yield++;
2749    }    }
# Line 2302  value in the reference (which is a group Line 2775  value in the reference (which is a group
2775  Arguments:  Arguments:
2776    group      points to the start of the group    group      points to the start of the group
2777    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2778    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
2779    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2780    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
2781    
# Line 2310  Returns:     nothing Line 2783  Returns:     nothing
2783  */  */
2784    
2785  static void  static void
2786  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2787    uschar *save_hwm)    pcre_uchar *save_hwm)
2788  {  {
2789  uschar *ptr = group;  pcre_uchar *ptr = group;
2790    
2791  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2792    {    {
2793    int offset;    int offset;
2794    uschar *hc;    pcre_uchar *hc;
2795    
2796    /* 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
2797    reference. */    reference. */
# Line 2363  Arguments: Line 2836  Arguments:
2836  Returns:         new code pointer  Returns:         new code pointer
2837  */  */
2838    
2839  static uschar *  static pcre_uchar *
2840  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2841  {  {
2842  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2843  *code++ = 255;  *code++ = 255;
2844  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2845  PUT(code, LINK_SIZE, 0);                       /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2846  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2847  }  }
2848    
2849    
# Line 2392  Returns:             nothing Line 2865  Returns:             nothing
2865  */  */
2866    
2867  static void  static void
2868  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2869  {  {
2870  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2871  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
# Line 2406  PUT(previous_callout, 2 + LINK_SIZE, len Line 2879  PUT(previous_callout, 2 + LINK_SIZE, len
2879  *************************************************/  *************************************************/
2880    
2881  /* This function is passed the start and end of a class range, in UTF-8 mode  /* This function is passed the start and end of a class range, in UTF-8 mode
2882  with UCP support. It searches up the characters, looking for internal ranges of  with UCP support. It searches up the characters, looking for ranges of
2883  characters in the "other" case. Each call returns the next one, updating the  characters in the "other" case. Each call returns the next one, updating the
2884  start address.  start address. A character with multiple other cases is returned on its own
2885    with a special return value.
2886    
2887  Arguments:  Arguments:
2888    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 2416  Arguments: Line 2890  Arguments:
2890    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2891    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2892    
2893  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2894                   0 when a range is returned
2895                  >0 the CASESET offset for char with multiple other cases
2896                    in this case, ocptr contains the original
2897  */  */
2898    
2899  static BOOL  static int
2900  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2901    unsigned int *odptr)    unsigned int *odptr)
2902  {  {
2903  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2904    int co;
2905    
2906    /* Find the first character that has an other case. If it has multiple other
2907    cases, return its case offset value. */
2908    
2909  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2910    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }    {
2911      if ((co = UCD_CASESET(c)) != 0)
2912        {
2913        *ocptr = c++;   /* Character that has the set */
2914        *cptr = c;      /* Rest of input range */
2915        return co;
2916        }
2917      if ((othercase = UCD_OTHERCASE(c)) != c) break;
2918      }
2919    
2920  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
2921    
2922  *ocptr = othercase;  *ocptr = othercase;
2923  next = othercase + 1;  next = othercase + 1;
# Line 2439  for (++c; c <= d; c++) Line 2928  for (++c; c <= d; c++)
2928    next++;    next++;
2929    }    }
2930    
2931  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
2932  *cptr = c;  *cptr = c;             /* Rest of input range */
2933    return 0;
 return TRUE;  
2934  }  }
2935    
2936    
# Line 2466  Returns:       TRUE if auto-possessifyin Line 2954  Returns:       TRUE if auto-possessifyin
2954  static BOOL  static BOOL
2955  check_char_prop(int c, int ptype, int pdata, BOOL negated)  check_char_prop(int c, int ptype, int pdata, BOOL negated)
2956  {  {
2957    #ifdef SUPPORT_UCP
2958    const pcre_uint32 *p;
2959    #endif
2960    
2961  const ucd_record *prop = GET_UCD(c);  const ucd_record *prop = GET_UCD(c);
2962    
2963  switch(ptype)  switch(ptype)
2964    {    {
2965    case PT_LAMP:    case PT_LAMP:
# Line 2475  switch(ptype) Line 2968  switch(ptype)
2968            prop->chartype == ucp_Lt) == negated;            prop->chartype == ucp_Lt) == negated;
2969    
2970    case PT_GC:    case PT_GC:
2971    return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2972    
2973    case PT_PC:    case PT_PC:
2974    return (pdata == prop->chartype) == negated;    return (pdata == prop->chartype) == negated;
# Line 2486  switch(ptype) Line 2979  switch(ptype)
2979    /* These are specials */    /* These are specials */
2980    
2981    case PT_ALNUM:    case PT_ALNUM:
2982    return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2983            _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2984    
2985    case PT_SPACE:    /* Perl space */    case PT_SPACE:    /* Perl space */
2986    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2987            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2988            == negated;            == negated;
2989    
2990    case PT_PXSPACE:  /* POSIX space */    case PT_PXSPACE:  /* POSIX space */
2991    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2992            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2993            c == CHAR_FF || c == CHAR_CR)            c == CHAR_FF || c == CHAR_CR)
2994            == negated;            == negated;
2995    
2996    case PT_WORD:    case PT_WORD:
2997    return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2998            _pcre_ucp_gentype[prop->chartype] == ucp_N ||            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2999            c == CHAR_UNDERSCORE) == negated;            c == CHAR_UNDERSCORE) == negated;
3000    
3001    #ifdef SUPPORT_UCP
3002      case PT_CLIST:
3003      p = PRIV(ucd_caseless_sets) + prop->caseset;
3004      for (;;)
3005        {
3006        if ((unsigned int)c < *p) return !negated;
3007        if ((unsigned int)c == *p++) return negated;
3008        }
3009      break;  /* Control never reaches here */
3010    #endif
3011    }    }
3012    
3013  return FALSE;  return FALSE;
3014  }  }
3015  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
# Line 2521  sense to automatically possessify the re Line 3026  sense to automatically possessify the re
3026    
3027  Arguments:  Arguments:
3028    previous      pointer to the repeated opcode    previous      pointer to the repeated opcode
3029    utf8          TRUE in UTF-8 mode    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode
3030    ptr           next character in pattern    ptr           next character in pattern
3031    options       options bits    options       options bits
3032    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2530  Returns:        TRUE if possessifying is Line 3035  Returns:        TRUE if possessifying is
3035  */  */
3036    
3037  static BOOL  static BOOL
3038  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
3039    int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
3040  {  {
3041  int c, next;  pcre_int32 c = NOTACHAR;
3042    pcre_int32 next;
3043  int op_code = *previous++;  int op_code = *previous++;
3044    
3045  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2542  if ((options & PCRE_EXTENDED) != 0) Line 3048  if ((options & PCRE_EXTENDED) != 0)
3048    {    {
3049    for (;;)    for (;;)
3050      {      {
3051      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3052      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
3053        {        {
3054        ptr++;        ptr++;
# Line 2550  if ((options & PCRE_EXTENDED) != 0) Line 3056  if ((options & PCRE_EXTENDED) != 0)
3056          {          {
3057          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3058          ptr++;          ptr++;
3059  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3060          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf) FORWARDCHAR(ptr);
3061  #endif  #endif
3062          }          }
3063        }        }
# Line 2569  if (*ptr == CHAR_BACKSLASH) Line 3075  if (*ptr == CHAR_BACKSLASH)
3075    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
3076    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
3077    }    }
3078    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
3079    {    {
3080  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3081    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
3082  #endif  #endif
3083    next = *ptr++;    next = *ptr++;
3084    }    }
   
3085  else return FALSE;  else return FALSE;
3086    
3087  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2586  if ((options & PCRE_EXTENDED) != 0) Line 3090  if ((options & PCRE_EXTENDED) != 0)
3090    {    {
3091    for (;;)    for (;;)
3092      {      {
3093      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3094      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
3095        {        {
3096        ptr++;        ptr++;
# Line 2594  if ((options & PCRE_EXTENDED) != 0) Line 3098  if ((options & PCRE_EXTENDED) != 0)
3098          {          {
3099          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3100          ptr++;          ptr++;
3101  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3102          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf) FORWARDCHAR(ptr);
3103  #endif  #endif
3104          }          }
3105        }        }
# Line 2606  if ((options & PCRE_EXTENDED) != 0) Line 3110  if ((options & PCRE_EXTENDED) != 0)
3110  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3111    
3112  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3113    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)
3114      return FALSE;      return FALSE;
3115    
3116  /* Now compare the next item with the previous opcode. First, handle cases when  /* If the previous item is a character, get its value. */
 the next item is a character. */  
3117    
3118  if (next >= 0) switch(op_code)  if (op_code == OP_CHAR || op_code == OP_CHARI ||
3119        op_code == OP_NOT || op_code == OP_NOTI)
3120    {    {
3121    case OP_CHAR:  #ifdef SUPPORT_UTF
 #ifdef SUPPORT_UTF8  
3122    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
3123  #else  #else
3124    c = *previous;    c = *previous;
3125  #endif  #endif
3126    return c != next;    }
3127    
3128    /* For CHARNC (caseless character) we must check the other case. If we have  /* Now compare the next item with the previous opcode. First, handle cases when
3129    Unicode property support, we can use it to test the other case of  the next item is a character. */
   high-valued characters. */  
3130    
3131    case OP_CHARNC:  if (next >= 0)
3132  #ifdef SUPPORT_UTF8    {
3133    GETCHARTEST(c, previous);    /* For a caseless UTF match, the next character may have more than one other
3134  #else    case, which maps to the special PT_CLIST property. Check this first. */
3135    c = *previous;  
3136    #ifdef SUPPORT_UCP
3137      if (utf && (unsigned int)c != NOTACHAR && (options & PCRE_CASELESS) != 0)
3138        {
3139        int ocs = UCD_CASESET(next);
3140        if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
3141        }
3142  #endif  #endif
3143    if (c == next) return FALSE;  
3144  #ifdef SUPPORT_UTF8    switch(op_code)
   if (utf8)  
3145      {      {
3146      unsigned int othercase;      case OP_CHAR:
3147      if (next < 128) othercase = cd->fcc[next]; else      return c != next;
3148    
3149        /* For CHARI (caseless character) we must check the other case. If we have
3150        Unicode property support, we can use it to test the other case of
3151        high-valued characters. We know that next can have only one other case,
3152        because multi-other-case characters are dealt with above. */
3153    
3154        case OP_CHARI:
3155        if (c == next) return FALSE;
3156    #ifdef SUPPORT_UTF
3157        if (utf)
3158          {
3159          unsigned int othercase;
3160          if (next < 128) othercase = cd->fcc[next]; else
3161  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3162      othercase = UCD_OTHERCASE((unsigned int)next);        othercase = UCD_OTHERCASE((unsigned int)next);
3163  #else  #else
3164      othercase = NOTACHAR;        othercase = NOTACHAR;
3165  #endif  #endif
3166      return (unsigned int)c != othercase;        return (unsigned int)c != othercase;
3167      }        }
3168    else      else
3169  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3170    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */      return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3171    
3172    /* For OP_NOT, its data is always a single-byte character. */      case OP_NOT:
3173        return c == next;
3174    case OP_NOT:  
3175    if ((c = *previous) == next) return TRUE;      case OP_NOTI:
3176    if ((options & PCRE_CASELESS) == 0) return FALSE;      if (c == next) return TRUE;
3177  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3178    if (utf8)      if (utf)
3179      {        {
3180      unsigned int othercase;        unsigned int othercase;
3181      if (next < 128) othercase = cd->fcc[next]; else        if (next < 128) othercase = cd->fcc[next]; else
3182  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3183      othercase = UCD_OTHERCASE(next);        othercase = UCD_OTHERCASE((unsigned int)next);
3184  #else  #else
3185      othercase = NOTACHAR;        othercase = NOTACHAR;
3186  #endif  #endif
3187      return (unsigned int)c == othercase;        return (unsigned int)c == othercase;
3188      }        }
3189    else      else
3190  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3191    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */      return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3192    
3193    /* 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.
3194    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. */
3195    
3196    case OP_DIGIT:      case OP_DIGIT:
3197    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3198    
3199    case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
3200    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
3201    
3202    case OP_WHITESPACE:      case OP_WHITESPACE:
3203    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
3204    
3205    case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
3206    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
3207    
3208    case OP_WORDCHAR:      case OP_WORDCHAR:
3209    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3210    
3211    case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
3212    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
3213    
3214    case OP_HSPACE:      case OP_HSPACE:
3215    case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
3216    switch(next)      switch(next)
3217      {        {
3218      case 0x09:        HSPACE_CASES:
3219      case 0x20:        return op_code == OP_NOT_HSPACE;
     case 0xa0:  
     case 0x1680:  
     case 0x180e:  
     case 0x2000:  
     case 0x2001:  
     case 0x2002:  
     case 0x2003:  
     case 0x2004:  
     case 0x2005:  
     case 0x2006:  
     case 0x2007:  
     case 0x2008:  
     case 0x2009:  
     case 0x200A:  
     case 0x202f:  
     case 0x205f:  
     case 0x3000:  
     return op_code == OP_NOT_HSPACE;  
     default:  
     return op_code != OP_NOT_HSPACE;  
     }  
3220    
3221    case OP_ANYNL:        default:
3222    case OP_VSPACE:        return op_code != OP_NOT_HSPACE;
3223    case OP_NOT_VSPACE:        }
3224    switch(next)  
3225      {      case OP_ANYNL:
3226      case 0x0a:      case OP_VSPACE:
3227      case 0x0b:      case OP_NOT_VSPACE:
3228      case 0x0c:      switch(next)
3229      case 0x0d:        {
3230      case 0x85:        VSPACE_CASES:
3231      case 0x2028:        return op_code == OP_NOT_VSPACE;
3232      case 0x2029:  
3233      return op_code == OP_NOT_VSPACE;        default:
3234      default:        return op_code != OP_NOT_VSPACE;
3235      return op_code != OP_NOT_VSPACE;        }
     }  
3236    
3237  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3238    case OP_PROP:      case OP_PROP:
3239    return check_char_prop(next, previous[0], previous[1], FALSE);      return check_char_prop(next, previous[0], previous[1], FALSE);
3240    
3241    case OP_NOTPROP:      case OP_NOTPROP:
3242    return check_char_prop(next, previous[0], previous[1], TRUE);      return check_char_prop(next, previous[0], previous[1], TRUE);
3243  #endif  #endif
3244    
3245    default:      default:
3246    return FALSE;      return FALSE;
3247        }
3248    }    }
3249    
   
3250  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3251  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3252  generated only when PCRE_UCP is *not* set, that is, when only ASCII  generated only when PCRE_UCP is *not* set, that is, when only ASCII
# Line 2758  replaced by OP_PROP codes when PCRE_UCP Line 3256  replaced by OP_PROP codes when PCRE_UCP
3256  switch(op_code)  switch(op_code)
3257    {    {
3258    case OP_CHAR:    case OP_CHAR:
3259    case OP_CHARNC:    case OP_CHARI:
 #ifdef SUPPORT_UTF8  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
3260    switch(-next)    switch(-next)
3261      {      {
3262      case ESC_d:      case ESC_d:
3263      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3264    
3265      case ESC_D:      case ESC_D:
3266      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
3267    
3268      case ESC_s:      case ESC_s:
3269      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
3270    
3271      case ESC_S:      case ESC_S:
3272      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
3273    
3274      case ESC_w:      case ESC_w:
3275      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
3276    
3277      case ESC_W:      case ESC_W:
3278      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
3279    
3280      case ESC_h:      case ESC_h:
3281      case ESC_H:      case ESC_H:
3282      switch(c)      switch(c)
3283        {        {
3284        case 0x09:        HSPACE_CASES:
       case 0x20:  
       case 0xa0:  
       case 0x1680:  
       case 0x180e:  
       case 0x2000:  
       case 0x2001:  
       case 0x2002:  
       case 0x2003:  
       case 0x2004:  
       case 0x2005:  
       case 0x2006:  
       case 0x2007:  
       case 0x2008:  
       case 0x2009:  
       case 0x200A:  
       case 0x202f:  
       case 0x205f:  
       case 0x3000:  
3285        return -next != ESC_h;        return -next != ESC_h;
3286    
3287        default:        default:
3288        return -next == ESC_h;        return -next == ESC_h;
3289        }        }
# Line 2816  switch(op_code) Line 3292  switch(op_code)
3292      case ESC_V:      case ESC_V:
3293      switch(c)      switch(c)
3294        {        {
3295        case 0x0a:        VSPACE_CASES:
       case 0x0b:  
       case 0x0c:  
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
3296        return -next != ESC_v;        return -next != ESC_v;
3297    
3298        default:        default:
3299        return -next == ESC_v;        return -next == ESC_v;
3300        }        }
# Line 2864  switch(op_code) Line 3335  switch(op_code)
3335        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. */
3336    
3337        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3338          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)
3339            return FALSE;            return FALSE;
3340    
3341        /* Do the property check. */        /* Do the property check. */
# Line 2891  switch(op_code) Line 3362  switch(op_code)
3362    return next == -ESC_d;    return next == -ESC_d;
3363    
3364    case OP_WHITESPACE:    case OP_WHITESPACE:
3365    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
3366    
3367    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3368    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v || next == -ESC_R;
3369    
3370    case OP_HSPACE:    case OP_HSPACE:
3371    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
# Line 2928  switch(op_code) Line 3399  switch(op_code)
3399    
3400    
3401  /*************************************************  /*************************************************
3402    *        Add a character or range to a class     *
3403    *************************************************/
3404    
3405    /* This function packages up the logic of adding a character or range of
3406    characters to a class. The character values in the arguments will be within the
3407    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3408    mutually recursive with the function immediately below.
3409    
3410    Arguments:
3411      classbits     the bit map for characters < 256
3412      uchardptr     points to the pointer for extra data
3413      options       the options word
3414      cd            contains pointers to tables etc.
3415      start         start of range character
3416      end           end of range character
3417    
3418    Returns:        the number of < 256 characters added
3419                    the pointer to extra data is updated
3420    */
3421    
3422    static int
3423    add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3424      compile_data *cd, unsigned int start, unsigned int end)
3425    {
3426    unsigned int c;
3427    int n8 = 0;
3428    
3429    /* If caseless matching is required, scan the range and process alternate
3430    cases. In Unicode, there are 8-bit characters that have alternate cases that
3431    are greater than 255 and vice-versa. Sometimes we can just extend the original
3432    range. */
3433    
3434    if ((options & PCRE_CASELESS) != 0)
3435      {
3436    #ifdef SUPPORT_UCP
3437      if ((options & PCRE_UTF8) != 0)
3438        {
3439        int rc;
3440        unsigned int oc, od;
3441    
3442        options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3443        c = start;
3444    
3445        while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3446          {
3447          /* Handle a single character that has more than one other case. */
3448    
3449          if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3450            PRIV(ucd_caseless_sets) + rc, oc);
3451    
3452          /* Do nothing if the other case range is within the original range. */
3453    
3454          else if (oc >= start && od <= end) continue;
3455    
3456          /* Extend the original range if there is overlap, noting that if oc < c, we
3457          can't have od > end because a subrange is always shorter than the basic
3458          range. Otherwise, use a recursive call to add the additional range. */
3459    
3460          else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3461          else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3462          else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3463          }
3464        }
3465      else
3466    #endif  /* SUPPORT_UCP */
3467    
3468      /* Not UTF-mode, or no UCP */
3469    
3470      for (c = start; c <= end && c < 256; c++)
3471        {
3472        SETBIT(classbits, cd->fcc[c]);
3473        n8++;
3474        }
3475      }
3476    
3477    /* Now handle the original range. Adjust the final value according to the bit
3478    length - this means that the same lists of (e.g.) horizontal spaces can be used
3479    in all cases. */
3480    
3481    #if defined COMPILE_PCRE8
3482    #ifdef SUPPORT_UTF
3483      if ((options & PCRE_UTF8) == 0)
3484    #endif
3485      if (end > 0xff) end = 0xff;
3486    
3487    #elif defined COMPILE_PCRE16
3488    #ifdef SUPPORT_UTF
3489      if ((options & PCRE_UTF16) == 0)
3490    #endif
3491      if (end > 0xffff) end = 0xffff;
3492    
3493    #elif defined COMPILE_PCRE32
3494    #ifdef SUPPORT_UTF
3495      if ((options & PCRE_UTF32) == 0)
3496        if (end > 0xffffu) end = 0xffffu; // FIXMEchpe rebase fix this
3497    #endif
3498    #endif /* COMPILE_PCRE[8|16|32] */
3499    
3500    /* If all characters are less than 256, use the bit map. Otherwise use extra
3501    data. */
3502    
3503    if (end < 0x100)
3504      {
3505      for (c = start; c <= end; c++)
3506        {
3507        n8++;
3508        SETBIT(classbits, c);
3509        }
3510      }
3511    
3512    else
3513      {
3514      pcre_uchar *uchardata = *uchardptr;
3515    
3516    #ifdef SUPPORT_UTF
3517      if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3518        {
3519        if (start < end)
3520          {
3521          *uchardata++ = XCL_RANGE;
3522          uchardata += PRIV(ord2utf)(start, uchardata);
3523          uchardata += PRIV(ord2utf)(end, uchardata);
3524          }
3525        else if (start == end)
3526          {
3527          *uchardata++ = XCL_SINGLE;
3528          uchardata += PRIV(ord2utf)(start, uchardata);
3529          }
3530        }
3531      else
3532    #endif  /* SUPPORT_UTF */
3533    
3534      /* Without UTF support, character values are constrained by the bit length,
3535      and can only be > 256 for 16-bit and 32-bit libraries. */
3536    
3537    #ifdef COMPILE_PCRE8
3538        {}
3539    #else
3540      if (start < end)
3541        {
3542        *uchardata++ = XCL_RANGE;
3543        *uchardata++ = start;
3544        *uchardata++ = end;
3545        }
3546      else if (start == end)
3547        {
3548        *uchardata++ = XCL_SINGLE;
3549        *uchardata++ = start;
3550        }
3551    #endif
3552    
3553      *uchardptr = uchardata;   /* Updata extra data pointer */
3554      }
3555    
3556    return n8;    /* Number of 8-bit characters */
3557    }
3558    
3559    
3560    
3561    
3562    /*************************************************
3563    *        Add a list of characters to a class     *
3564    *************************************************/
3565    
3566    /* This function is used for adding a list of case-equivalent characters to a
3567    class, and also for adding a list of horizontal or vertical whitespace. If the
3568    list is in order (which it should be), ranges of characters are detected and
3569    handled appropriately. This function is mutually recursive with the function
3570    above.
3571    
3572    Arguments:
3573      classbits     the bit map for characters < 256
3574      uchardptr     points to the pointer for extra data
3575      options       the options word
3576      cd            contains pointers to tables etc.
3577      p             points to row of 32-bit values, terminated by NOTACHAR
3578      except        character to omit; this is used when adding lists of
3579                      case-equivalent characters to avoid including the one we
3580                      already know about
3581    
3582    Returns:        the number of < 256 characters added
3583                    the pointer to extra data is updated
3584    */
3585    
3586    static int
3587    add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3588      compile_data *cd, const pcre_uint32 *p, unsigned int except)
3589    {
3590    int n8 = 0;
3591    while (p[0] < NOTACHAR)
3592      {
3593      int n = 0;
3594      if (p[0] != except)
3595        {
3596        while(p[n+1] == p[0] + n + 1) n++;
3597        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
3598        }
3599      p += n + 1;
3600      }
3601    return n8;
3602    }
3603    
3604    
3605    
3606    /*************************************************
3607    *    Add characters not in a list to a class     *
3608    *************************************************/
3609    
3610    /* This function is used for adding the complement of a list of horizontal or
3611    vertical whitespace to a class. The list must be in order.
3612    
3613    Arguments:
3614      classbits     the bit map for characters < 256
3615      uchardptr     points to the pointer for extra data
3616      options       the options word
3617      cd            contains pointers to tables etc.
3618      p             points to row of 32-bit values, terminated by NOTACHAR
3619    
3620    Returns:        the number of < 256 characters added
3621                    the pointer to extra data is updated
3622    */
3623    
3624    static int
3625    add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
3626      int options, compile_data *cd, const pcre_uint32 *p)
3627    {
3628    int n8 = 0;
3629    if (p[0] > 0)
3630      n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
3631    while (p[0] < NOTACHAR)
3632      {
3633      while (p[1] == p[0] + 1) p++;
3634      n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
3635        (p[1] == NOTACHAR)? 0x10ffff : p[1] - 1);
3636      p++;
3637      }
3638    return n8;
3639    }
3640    
3641    
3642    
3643    /*************************************************
3644  *           Compile one branch                   *  *           Compile one branch                   *
3645  *************************************************/  *************************************************/
3646    
# Line 2942  Arguments: Line 3655  Arguments:
3655    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3656    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3657    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3658    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3659    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3660    bcptr          points to current branch chain    bcptr          points to current branch chain
3661      cond_depth     conditional nesting depth
3662    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3663    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3664                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2954  Returns:         TRUE on success Line 3668  Returns:         TRUE on success
3668  */  */
3669    
3670  static BOOL  static BOOL
3671  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3672    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3673      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3674    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3675  {  {
3676  int repeat_type, op_type;  int repeat_type, op_type;
3677  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3678  int bravalue = 0;  int bravalue = 0;
3679  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3680  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3681  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3682  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3683  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3684  int after_manual_callout = 0;  int after_manual_callout = 0;
3685  int length_prevgroup = 0;  int length_prevgroup = 0;
3686  register int c;  register int c;
3687  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3688  uschar *last_code = code;  pcre_uchar *last_code = code;
3689  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3690  uschar *tempcode;  pcre_uchar *tempcode;
3691  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3692  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3693  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3694  const uschar *tempptr;  const pcre_uchar *tempptr;
3695  const uschar *nestptr = NULL;  const pcre_uchar *nestptr = NULL;
3696  uschar *previous = NULL;  pcre_uchar *previous = NULL;
3697  uschar *previous_callout = NULL;  pcre_uchar *previous_callout = NULL;
3698  uschar *save_hwm = NULL;  pcre_uchar *save_hwm = NULL;
3699  uschar classbits[32];  pcre_uint8 classbits[32];
3700    
3701  #ifdef SUPPORT_UTF8  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3702  BOOL class_utf8;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3703  BOOL utf8 = (options & PCRE_UTF8) != 0;  dynamically as we process the pattern. */
3704  uschar *class_utf8data;  
3705  uschar *class_utf8data_base;  #ifdef SUPPORT_UTF
3706  uschar utf8_char[6];  /* PCRE_UTF(16|32) have the same value as PCRE_UTF8. */
3707    BOOL utf = (options & PCRE_UTF8) != 0;
3708    pcre_uchar utf_chars[6];
3709  #else  #else
3710  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3711  uschar *utf8_char = NULL;  #endif
3712    
3713    /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
3714    class_uchardata always so that it can be passed to add_to_class() always,
3715    though it will not be used in non-UTF 8-bit cases. This avoids having to supply
3716    alternative calls for the different cases. */
3717    
3718    pcre_uchar *class_uchardata;
3719    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3720    BOOL xclass;
3721    pcre_uchar *class_uchardata_base;
3722  #endif  #endif
3723    
3724  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 3005  greedy_non_default = greedy_default ^ 1; Line 3732  greedy_non_default = greedy_default ^ 1;
3732    
3733  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3734  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
3735  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
3736  find one.  find one.
3737    
3738  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
3739  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
3740  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3741  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3742    
3743  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3744    
3745  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3746  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
3747  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
3748  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3749    value. This is used only for ASCII characters. */
3750    
3751  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3752    
3753  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3754    
# Line 3032  for (;; ptr++) Line 3760  for (;; ptr++)
3760    BOOL is_quantifier;    BOOL is_quantifier;
3761    BOOL is_recurse;    BOOL is_recurse;
3762    BOOL reset_bracount;    BOOL reset_bracount;
3763    int class_charcount;    int class_has_8bitchar;
3764    int class_lastchar;    int class_one_char;
3765    int newoptions;    int newoptions;
3766    int recno;    int recno;
3767    int refsign;    int refsign;
3768    int skipbytes;    int skipbytes;
3769    int subreqbyte;    int subreqchar;
3770    int subfirstbyte;    int subfirstchar;
3771    int terminator;    int terminator;
3772    int mclength;    int mclength;
3773    uschar mcbuffer[8];    int tempbracount;
3774      pcre_uchar mcbuffer[8];
3775    
3776    /* Get next byte in the pattern */    /* Get next character in the pattern */
3777    
3778    c = *ptr;    c = *ptr;
3779    
# Line 3066  for (;; ptr++) Line 3795  for (;; ptr++)
3795  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
3796      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3797  #endif  #endif
3798      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3799            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3800        {        {
3801        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3802        goto FAILED;        goto FAILED;
# Line 3089  for (;; ptr++) Line 3819  for (;; ptr++)
3819        }        }
3820    
3821      *lengthptr += (int)(code - last_code);      *lengthptr += (int)(code - last_code);
3822      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3823          (int)(code - last_code), c, c));
3824    
3825      /* 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
3826      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 3099  for (;; ptr++) Line 3830  for (;; ptr++)
3830        {        {
3831        if (previous > orig_code)        if (previous > orig_code)
3832          {          {
3833          memmove(orig_code, previous, code - previous);          memmove(orig_code, previous, IN_UCHARS(code - previous));
3834          code -= previous - orig_code;          code -= previous - orig_code;
3835          previous = orig_code;          previous = orig_code;
3836          }          }
# Line 3115  for (;; ptr++) Line 3846  for (;; ptr++)
3846    /* 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
3847    reference list. */    reference list. */
3848    
3849    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3850               WORK_SIZE_SAFETY_MARGIN)
3851      {      {
3852      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3853      goto FAILED;      goto FAILED;
# Line 3163  for (;; ptr++) Line 3895  for (;; ptr++)
3895      previous_callout = NULL;      previous_callout = NULL;
3896      }      }
3897    
3898    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3899    
3900    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3901      {      {
3902      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
3903      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3904        {        {
3905        ptr++;        ptr++;
# Line 3175  for (;; ptr++) Line 3907  for (;; ptr++)
3907          {          {
3908          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3909          ptr++;          ptr++;
3910  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3911          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf) FORWARDCHAR(ptr);
3912  #endif  #endif
3913          }          }
3914        if (*ptr != 0) continue;        if (*ptr != 0) continue;
# Line 3200  for (;; ptr++) Line 3932  for (;; ptr++)
3932      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3933      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3934      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
3935      *firstbyteptr = firstbyte;      *firstcharptr = firstchar;
3936      *reqbyteptr = reqbyte;      *reqcharptr = reqchar;
3937      *codeptr = code;      *codeptr = code;
3938      *ptrptr = ptr;      *ptrptr = ptr;
3939      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3222  for (;; ptr++) Line 3954  for (;; ptr++)
3954      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3955    
3956      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3957        previous = NULL;
3958      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3959        {        {
3960        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3961          *code++ = OP_CIRCM;
3962        }        }
3963      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3964      break;      break;
3965    
3966      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3967      previous = NULL;      previous = NULL;
3968      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3969      break;      break;
3970    
3971      /* 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
3972      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqchar doesn't change either. */
3973    
3974      case CHAR_DOT:      case CHAR_DOT:
3975      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3976      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
3977      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
3978      previous = code;      previous = code;
3979      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3980      break;      break;
# Line 3296  for (;; ptr++) Line 4029  for (;; ptr++)
4029          {          {
4030          if (ptr[1] == CHAR_E)          if (ptr[1] == CHAR_E)
4031            ptr++;            ptr++;
4032          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)  
4033            ptr += 3;            ptr += 3;
4034          else          else
4035            break;            break;
# Line 3316  for (;; ptr++) Line 4048  for (;; ptr++)
4048          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4049        {        {
4050        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
4051        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4052        zerofirstbyte = firstbyte;        zerofirstchar = firstchar;
4053        break;        break;
4054        }        }
4055    
# Line 3327  for (;; ptr++) Line 4059  for (;; ptr++)
4059    
4060      should_flip_negation = FALSE;      should_flip_negation = FALSE;
4061    
4062      /* 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:
4063      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 <
4064      valued UTF-8 characters, we don't yet do any optimization. */      256 character; class_one_char will be 1 if the class contains just one
4065        character. */
4066    
4067      class_charcount = 0;      class_has_8bitchar = 0;
4068      class_lastchar = -1;      class_one_char = 0;
4069    
4070      /* 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
4071      temporary bit of memory, in case the class contains only 1 character (less      temporary bit of memory, in case the class contains fewer than two
4072      than 256), because in that case the compiled code doesn't use the bit map.      8-bit characters because in that case the compiled code doesn't use the bit
4073      */      map. */
4074    
4075      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
4076    
4077  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4078      class_utf8 = FALSE;                       /* No chars >= 256 */      xclass = FALSE;
4079      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
4080      class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */      class_uchardata_base = class_uchardata;   /* Save the start */
4081  #endif  #endif
4082    
4083      /* 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 3353  for (;; ptr++) Line 4086  for (;; ptr++)
4086    
4087      if (c != 0) do      if (c != 0) do
4088        {        {
4089        const uschar *oldptr;        const pcre_uchar *oldptr;
4090    
4091  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
4092        if (utf8 && c > 127)        if (utf && HAS_EXTRALEN(c))
4093          {                           /* Braces are required because the */          {                           /* Braces are required because the */
4094          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
4095          }          }
4096    #endif
4097    
4098        /* In the pre-compile phase, accumulate the length of any UTF-8 extra  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4099          /* In the pre-compile phase, accumulate the length of any extra
4100        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
4101        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
4102        (which is on the stack). */        (which is on the stack). We have to remember that there was XCLASS data,
4103          however. */
4104        if (lengthptr != NULL)  
4105          {        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4106          *lengthptr += class_utf8data - class_utf8data_base;          {
4107          class_utf8data = class_utf8data_base;          xclass = TRUE;
4108            *lengthptr += class_uchardata - class_uchardata_base;
4109            class_uchardata = class_uchardata_base;
4110          }          }
   
4111  #endif  #endif
4112    
4113        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 3399  for (;; ptr++) Line 4135  for (;; ptr++)
4135          {          {
4136          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
4137          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
4138          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
4139          uschar pbits[32];          pcre_uint8 pbits[32];
4140    
4141          if (ptr[1] != CHAR_COLON)          if (ptr[1] != CHAR_COLON)
4142            {            {
# Line 3427  for (;; ptr++) Line 4163  for (;; ptr++)
4163          alpha. This relies on the fact that the class table starts with          alpha. This relies on the fact that the class table starts with
4164          alpha, lower, upper as the first 3 entries. */          alpha, lower, upper as the first 3 entries. */
4165    
4166          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4167            posix_class = 0;            posix_class = 0;
4168    
4169          /* When PCRE_UCP is set, some of the POSIX classes are converted to          /* When PCRE_UCP is set, some of the POSIX classes are converted to
# Line 3455  for (;; ptr++) Line 4191  for (;; ptr++)
4191          /* Copy in the first table (always present) */          /* Copy in the first table (always present) */
4192    
4193          memcpy(pbits, cbits + posix_class_maps[posix_class],          memcpy(pbits, cbits + posix_class_maps[posix_class],
4194            32 * sizeof(uschar));            32 * sizeof(pcre_uint8));
4195    
4196          /* If there is a second table, add or remove it as required. */          /* If there is a second table, add or remove it as required. */
4197    
# Line 3470  for (;; ptr++) Line 4206  for (;; ptr++)
4206              for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];              for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
4207            }            }
4208    
4209          /* Not see if we need to remove any special characters. An option          /* Now see if we need to remove any special characters. An option
4210          value of 1 removes vertical space and 2 removes underscore. */          value of 1 removes vertical space and 2 removes underscore. */
4211    
4212          if (tabopt < 0) tabopt = -tabopt;          if (tabopt < 0) tabopt = -tabopt;
# Line 3486  for (;; ptr++) Line 4222  for (;; ptr++)
4222            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
4223    
4224          ptr = tempptr + 1;          ptr = tempptr + 1;
4225          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          /* Every class contains at least one < 256 character. */
4226            class_has_8bitchar = 1;
4227            /* Every class contains at least two characters. */
4228            class_one_char = 2;
4229          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
4230          }          }
4231    
4232        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
4233        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
4234        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
4235        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
4236        class_charcount bigger than one. Unrecognized escapes fall through and        speculatively set both class_has_8bitchar and class_one_char bigger
4237        are either treated as literal characters (by default), or are faulted if        than one. Unrecognized escapes fall through and are either treated
4238          as literal characters (by default), or are faulted if
4239        PCRE_EXTRA is set. */        PCRE_EXTRA is set. */
4240    
4241        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
# Line 3504  for (;; ptr++) Line 4244  for (;; ptr++)
4244          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
4245    
4246          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 */
4247            else if (-c == ESC_N)            /* \N is not supported in a class */
4248              {
4249              *errorcodeptr = ERR71;
4250              goto FAILED;
4251              }
4252          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
4253            {            {
4254            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3517  for (;; ptr++) Line 4262  for (;; ptr++)
4262    
4263          if (c < 0)          if (c < 0)
4264            {            {
4265            register const uschar *cbits = cd->cbits;            register const pcre_uint8 *cbits = cd->cbits;
4266            class_charcount += 2;     /* Greater than 1 is what matters */            /* Every class contains at least two < 256 characters. */
4267              class_has_8bitchar++;
4268              /* Every class contains at least two characters. */
4269              class_one_char += 2;
4270    
4271            switch (-c)            switch (-c)
4272              {              {
# Line 3531  for (;; ptr++) Line 4279  for (;; ptr++)
4279              case ESC_SU:              case ESC_SU:
4280              nestptr = ptr;              nestptr = ptr;
4281              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
4282              class_charcount -= 2;                /* Undo! */              class_has_8bitchar--;                /* Undo! */
4283              continue;              continue;
4284  #endif  #endif
4285              case ESC_d:              case ESC_d:
# Line 3554  for (;; ptr++) Line 4302  for (;; ptr++)
4302    
4303              /* 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
4304              if it was previously set by something earlier in the character              if it was previously set by something earlier in the character
4305              class. */              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and
4306                EBCDIC, so we lazily just adjust the appropriate bit. */
4307    
4308              case ESC_s:              case ESC_s:
4309              classbits[0] |= cbits[cbit_space];              classbits[0] |= cbits[cbit_space];
# Line 3567  for (;; ptr++) Line 4316  for (;; ptr++)
4316              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
4317              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
4318              continue;              continue;
4319    
4320                /* The rest apply in both UCP and non-UCP cases. */
4321    
4322              case ESC_h:              case ESC_h:
4323              SETBIT(classbits, 0x09); /* VT */              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4324              SETBIT(classbits, 0x20); /* SPACE */                PRIV(hspace_list), NOTACHAR);
             SETBIT(classbits, 0xa0); /* NSBP */  
 #ifdef SUPPORT_UTF8  
             if (utf8)  
               {  
               class_utf8 = TRUE;  
               *class_utf8data++ = XCL_SINGLE;  
               class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);  
               *class_utf8data++ = XCL_SINGLE;  
               class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);  
               *class_utf8data++ = XCL_RANGE;  
               class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);  
               class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);  
               *class_utf8data++ = XCL_SINGLE;  
               class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);  
               *class_utf8data++ = XCL_SINGLE;  
               class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);  
               *class_utf8data++ = XCL_SINGLE;  
               class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);  
               }  
 #endif  
4325              continue;              continue;
4326    
4327              case ESC_H:              case ESC_H:
4328              for (c = 0; c < 32; c++)              (void)add_not_list_to_class(classbits, &class_uchardata, options,
4329                {                cd, PRIV(hspace_list));
               int x = 0xff;  
               switch (c)  
                 {  
                 case 0x09/8: x ^= 1 << (0x09%8); break;  
                 case 0x20/8: x ^= 1 << (0x20%8); break;  
                 case 0xa0/8: x ^= 1 << (0xa0%8); break;  
                 default: break;  
                 }  
               classbits[c] |= x;  
               }  
   
 #ifdef SUPPORT_UTF8  
             if (utf8)  
               {  
               class_utf8 = TRUE;  
               *class_utf8data++ = XCL_RANGE;  
               class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);  
               class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);  
               *class_utf8data++ = XCL_RANGE;  
               class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);  
               class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);  
               *class_utf8data++ = XCL_RANGE;  
               class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);  
               class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);  
               *class_utf8data++ = XCL_RANGE;  
               class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);  
               class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);  
               *class_utf8data++ = XCL_RANGE;  
               class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);  
               class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);  
               *class_utf8data++ = XCL_RANGE;  
               class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);  
               class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);  
               *class_utf8data++ = XCL_RANGE;  
               class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);  
               class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);  
               }  
 #endif  
4330              continue;              continue;
4331    
4332              case ESC_v:              case ESC_v:
4333              SETBIT(classbits, 0x0a); /* LF */              (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4334              SETBIT(classbits, 0x0b); /* VT */                PRIV(vspace_list), NOTACHAR);
             SETBIT(classbits, 0x0c); /* FF */  
             SETBIT(classbits, 0x0d); /* CR */  
             SETBIT(classbits, 0x85); /* NEL */  
 #ifdef SUPPORT_UTF8  
             if (utf8)  
               {  
               class_utf8 = TRUE;  
               *class_utf8data++ = XCL_RANGE;  
               class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);  
               class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);  
               }  
 #endif  
4335              continue;              continue;
4336    
4337              case ESC_V:              case ESC_V:
4338              for (c = 0; c < 32; c++)              (void)add_not_list_to_class(classbits, &class_uchardata, options,
4339                {                cd, PRIV(vspace_list));
               int x = 0xff;  
               switch (c)  
                 {  
                 case 0x0a/8: x ^= 1 << (0x0a%8);  
                              x ^= 1 << (0x0b%8);  
                              x ^= 1 << (0x0c%8);  
                              x ^= 1 << (0x0d%8);  
                              break;  
                 case 0x85/8: x ^= 1 << (0x85%8); break;  
                 default: break;  
                 }  
               classbits[c] |= x;  
               }  
   
 #ifdef SUPPORT_UTF8  
             if (utf8)  
               {  
               class_utf8 = TRUE;  
               *class_utf8data++ = XCL_RANGE;  
               class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);  
               class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);  
               *class_utf8data++ = XCL_RANGE;  
               class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);  
               class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);  
               }  
 #endif  
4340              continue;              continue;
4341    
4342  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 3692  for (;; ptr++) Line 4347  for (;; ptr++)
4347                int pdata;                int pdata;
4348                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
4349                if (ptype < 0) goto FAILED;                if (ptype < 0) goto FAILED;
4350                class_utf8 = TRUE;                *class_uchardata++ = ((-c == ESC_p) != negated)?
               *class_utf8data++ = ((-c == ESC_p) != negated)?  
4351                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
4352                *class_utf8data++ = ptype;                *class_uchardata++ = ptype;
4353                *class_utf8data++ = pdata;                *class_uchardata++ = pdata;
4354                class_charcount -= 2;   /* Not a < 256 character */                class_has_8bitchar--;                /* Undo! */
4355                continue;                continue;
4356                }                }
4357  #endif  #endif
# Line 3711  for (;; ptr++) Line 4365  for (;; ptr++)
4365                *errorcodeptr = ERR7;                *errorcodeptr = ERR7;
4366                goto FAILED;                goto FAILED;
4367                }                }
4368              class_charcount -= 2;  /* Undo the default count from above */              class_has_8bitchar--;    /* Undo the speculative increase. */
4369              c = *ptr;              /* Get the final character and fall through */              class_one_char -= 2;     /* Undo the speculative increase. */
4370                c = *ptr;                /* Get the final character and fall through */
4371              break;              break;
4372              }              }
4373            }            }
4374    
4375          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if the escape just defined a single character (c >= 0).
4376          greater than 256 in UTF-8 mode. */          This may be greater than 256. */
4377    
4378          }   /* End of backslash handling */          }   /* End of backslash handling */
4379    
4380        /* A single character may be followed by '-' to form a range. However,        /* A character may be followed by '-' to form a range. However, Perl does
4381        Perl does not permit ']' to be the end of the range. A '-' character        not permit ']' to be the end of the range. A '-' character at the end is
4382        at the end is treated as a literal. Perl ignores orphaned \E sequences        treated as a literal. Perl ignores orphaned \E sequences entirely. The
4383        entirely. The code for handling \Q and \E is messy. */        code for handling \Q and \E is messy. */
4384    
4385        CHECK_RANGE:        CHECK_RANGE:
4386        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3733  for (;; ptr++) Line 4388  for (;; ptr++)
4388          inescq = FALSE;          inescq = FALSE;
4389          ptr += 2;          ptr += 2;
4390          }          }
   
4391        oldptr = ptr;        oldptr = ptr;
4392    
4393        /* Remember \r or \n */        /* Remember if \r or \n were explicitly used */
4394    
4395        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4396    
# Line 3759  for (;; ptr++) Line 4413  for (;; ptr++)
4413            inescq = TRUE;            inescq = TRUE;
4414            break;            break;
4415            }            }
4416    
4417            /* Minus (hyphen) at the end of a class is treated as a literal, so put
4418            back the pointer and jump to handle the character that preceded it. */
4419    
4420          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
4421            {            {
4422            ptr = oldptr;            ptr = oldptr;
4423            goto LONE_SINGLE_CHARACTER;            goto CLASS_SINGLE_CHARACTER;
4424            }            }
4425    
4426            /* Otherwise, we have a potential range; pick up the next character */
4427    
4428  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4429          if (utf8)          if (utf)
4430            {                           /* Braces are required because the */            {                           /* Braces are required because the */
4431            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */
4432            }            }
# Line 3784  for (;; ptr++) Line 4443  for (;; ptr++)
4443            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
4444            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
4445    
4446            /* \b is backspace; any other special means the '-' was literal */            /* \b is backspace; any other special means the '-' was literal. */
4447    
4448            if (d < 0)            if (d < 0)
4449              {              {
4450              if (d == -ESC_b) d = CHAR_BS; else              if (d == -ESC_b) d = CHAR_BS; else
4451                {                {
4452                ptr = oldptr;                ptr = oldptr;
4453                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto CLASS_SINGLE_CHARACTER;  /* A few lines below */
4454                }                }
4455              }              }
           }  
   
         /* Check that the two values are in the correct order. Optimize  
         one-character ranges */  
   
         if (d < c)  
           {  
           *errorcodeptr = ERR8;  
           goto FAILED;  
           }  
   
         if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */  
   
         /* Remember \r or \n */  
   
         if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;  
   
         /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless  
         matching, we have to use an XCLASS with extra data items. Caseless  
         matching for characters > 127 is available only if UCP support is  
         available. */  
   
 #ifdef SUPPORT_UTF8  
         if (utf8 && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))  
           {  
           class_utf8 = TRUE;  
   
           /* With UCP support, we can find the other case equivalents of  
           the relevant characters. There may be several ranges. Optimize how  
           they fit with the basic range. */  
   
 #ifdef SUPPORT_UCP  
           if ((options & PCRE_CASELESS) != 0)  
             {  
             unsigned int occ, ocd;  
             unsigned int cc = c;  
             unsigned int origd = d;  
             while (get_othercase_range(&cc, origd, &occ, &ocd))  
               {  
               if (occ >= (unsigned int)c &&  
                   ocd <= (unsigned int)d)  
                 continue;                          /* Skip embedded ranges */  
   
               if (occ < (unsigned int)c  &&  
                   ocd >= (unsigned int)c - 1)      /* Extend the basic range */  
                 {                                  /* if there is overlap,   */  
                 c = occ;                           /* noting that if occ < c */  
                 continue;                          /* we can't have ocd > d  */  
                 }                                  /* because a subrange is  */  
               if (ocd > (unsigned int)d &&  
                   occ <= (unsigned int)d + 1)      /* always shorter than    */  
                 {                                  /* the basic range.       */  
                 d = ocd;  
                 continue;  
                 }  
   
               if (occ == ocd)  
                 {  
                 *class_utf8data++ = XCL_SINGLE;  
                 }  
               else  
                 {  
                 *class_utf8data++ = XCL_RANGE;  
                 class_utf8data += _pcre_ord2utf8(occ, class_utf8data);  
                 }  
               class_utf8data += _pcre_ord2utf8(ocd, class_utf8data);  
               }  
             }  
 #endif  /* SUPPORT_UCP */  
   
           /* Now record the original range, possibly modified for UCP caseless  
           overlapping ranges. */  
   
           *class_utf8data++ = XCL_RANGE;  
           class_utf8data += _pcre_ord2utf8(c, class_utf8data);  
           class_utf8data += _pcre_ord2utf8(d, class_utf8data);  
   
           /* With UCP support, we are done. Without UCP support, there is no  
           caseless matching for UTF-8 characters > 127; we can use the bit map  
           for the smaller ones. */  
   
 #ifdef SUPPORT_UCP  
           continue;    /* With next character in the class */  
 #else  
           if ((options & PCRE_CASELESS) == 0 || c > 127) continue;  
   
           /* Adjust upper limit and fall through to set up the map */  
   
           d = 127;  
   
 #endif  /* SUPPORT_UCP */  
4456            }            }
 #endif  /* SUPPORT_UTF8 */  
   
         /* We use the bit map for all cases when not in UTF-8 mode; else  
         ranges that lie entirely within 0-127 when there is UCP support; else  
         for partial ranges without UCP support. */  
   
         class_charcount += d - c + 1;  
         class_lastchar = d;  
4457    
4458          /* We can save a bit of time by skipping this in the pre-compile. */          /* Check that the two values are in the correct order. Optimize
4459            one-character ranges. */
4460    
4461          if (lengthptr == NULL) for (; c <= d; c++)          if (d < c)
4462            {            {
4463            classbits[c/8] |= (1 << (c&7));            *errorcodeptr = ERR8;
4464            if ((options & PCRE_CASELESS) != 0)            goto FAILED;
             {  
             int uc = cd->fcc[c];           /* flip case */  
             classbits[uc/8] |= (1 << (uc&7));  
             }  
4465            }            }
4466            if (d == c) goto CLASS_SINGLE_CHARACTER;  /* A few lines below */
4467    
4468          continue;   /* Go get the next char in the class */          /* We have found a character range, so single character optimizations
4469          }          cannot be done anymore. Any value greater than 1 indicates that there
4470            is more than one character. */
4471    
4472            class_one_char = 2;
4473    
4474        /* Handle a lone single character - we can get here for a normal          /* Remember an explicit \r or \n, and add the range to the class. */
       non-escape char, or after \ that introduces a single character or for an  
       apparent range that isn't. */  
4475    
4476        LONE_SINGLE_CHARACTER:          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4477    
4478            class_has_8bitchar +=
4479              add_to_class(classbits, &class_uchardata, options, cd, c, d);
4480    
4481            continue;   /* Go get the next char in the class */
4482            }
4483    
4484        /* Handle a character that cannot go in the bit map */        /* Handle a single character - we can get here for a normal non-escape
4485          char, or after \ that introduces a single character or for an apparent
4486          range that isn't. Only the value 1 matters for class_one_char, so don't
4487          increase it if it is already 2 or more ... just in case there's a class
4488          with a zillion characters in it. */
4489    
4490          CLASS_SINGLE_CHARACTER:
4491          if (class_one_char < 2) class_one_char++;
4492    
4493          /* If class_one_char is 1, we have the first single character in the
4494          class, and there have been no prior ranges, or XCLASS items generated by
4495          escapes. If this is the final character in the class, we can optimize by
4496          turning the item into a 1-character OP_CHAR[I] if it's positive, or
4497          OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
4498          to be set. Otherwise, there can be no first char if this item is first,
4499          whatever repeat count may follow. In the case of reqchar, save the
4500          previous value for reinstating. */
4501    
4502  #ifdef SUPPORT_UTF8        if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
       if (utf8 && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))  
4503          {          {
4504          class_utf8 = TRUE;          ptr++;
4505          *class_utf8data++ = XCL_SINGLE;          zeroreqchar = reqchar;
         class_utf8data += _pcre_ord2utf8(c, class_utf8data);  
4506    
4507  #ifdef SUPPORT_UCP          if (negate_class)
         if ((options & PCRE_CASELESS) != 0)  
4508            {            {
4509            unsigned int othercase;  #ifdef SUPPORT_UCP
4510            if ((othercase = UCD_OTHERCASE(c)) != c)              // FIXMEchpe pcreuint32?
4511              int d;
4512    #endif
4513              if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4514              zerofirstchar = firstchar;
4515    
4516              /* For caseless UTF-8 mode when UCP support is available, check
4517              whether this character has more than one other case. If so, generate
4518              a special OP_NOTPROP item instead of OP_NOTI. */
4519    
4520    #ifdef SUPPORT_UCP
4521              if (utf && (options & PCRE_CASELESS) != 0 &&
4522                  (d = UCD_CASESET(c)) != 0)
4523                {
4524                *code++ = OP_NOTPROP;
4525                *code++ = PT_CLIST;
4526                *code++ = d;
4527                }
4528              else
4529    #endif
4530              /* Char has only one other case, or UCP not available */
4531    
4532              {              {
4533              *class_utf8data++ = XCL_SINGLE;              *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4534              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
4535                if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
4536                  code += PRIV(ord2utf)(c, code);
4537                else
4538    #endif
4539                  *code++ = c;
4540              }              }
4541    
4542              /* We are finished with this character class */
4543    
4544              goto END_CLASS;
4545            }            }
 #endif  /* SUPPORT_UCP */  
4546    
4547          }          /* For a single, positive character, get the value into mcbuffer, and
4548        else          then we can handle this with the normal one-character code. */
 #endif  /* SUPPORT_UTF8 */  
4549    
4550        /* Handle a single-byte character */  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
4551          {          if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
4552          classbits[c/8] |= (1 << (c&7));            mclength = PRIV(ord2utf)(c, mcbuffer);
4553          if ((options & PCRE_CASELESS) != 0)          else
4554    #endif
4555            {            {
4556            c = cd->fcc[c];   /* flip case */            mcbuffer[0] = c;
4557            classbits[c/8] |= (1 << (c&7));            mclength = 1;
4558            }            }
4559          class_charcount++;          goto ONE_CHAR;
4560          class_lastchar = c;          }       /* End of 1-char optimization */
4561          }  
4562          /* There is more than one character in the class, or an XCLASS item
4563          has been generated. Add this character to the class. */
4564    
4565          class_has_8bitchar +=
4566            add_to_class(classbits, &class_uchardata, options, cd, c, c);
4567        }        }
4568    
4569      /* Loop until ']' reached. This "while" is the end of the "do" far above.      /* Loop until ']' reached. This "while" is the end of the "do" far above.
# Line 3971  for (;; ptr++) Line 4583  for (;; ptr++)
4583        goto FAILED;        goto FAILED;
4584        }        }
4585    
4586      /* If class_charcount is 1, we saw precisely one character whose value is      /* We will need an XCLASS if data has been placed in class_uchardata. In
4587      less than 256. As long as there were no characters >= 128 and there was no      the second phase this is a sufficient test. However, in the pre-compile
4588      use of \p or \P, in other words, no use of any XCLASS features, we can      phase, class_uchardata gets emptied to prevent workspace overflow, so it
4589      optimize.      only if the very last character in the class needs XCLASS will it contain
4590        anything at this point. For this reason, xclass gets set TRUE above when
4591      In UTF-8 mode, we can optimize the negative case only if there were no      uchar_classdata is emptied, and that's why this code is the way it is here
4592      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      instead of just doing a test on class_uchardata below. */
4593      operate on single-bytes only. This is an historical hangover. Maybe one day  
4594      we can tidy these opcodes to handle multi-byte characters.  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4595        if (class_uchardata > class_uchardata_base) xclass = TRUE;
4596      The optimization throws away the bit map. We turn the item into a  #endif
4597      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note  
4598      that OP_NOT does not support multibyte characters. In the positive case, it      /* If this is the first thing in the branch, there can be no first char
4599      can cause firstbyte to be set. Otherwise, there can be no first char if      setting, whatever the repeat count. Any reqchar setting must remain
4600      this item is first, whatever repeat count may follow. In the case of      unchanged after any kind of repeat. */
4601      reqbyte, save the previous value for reinstating. */  
4602        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4603  #ifdef SUPPORT_UTF8      zerofirstchar = firstchar;
4604      if (class_charcount == 1 && !class_utf8 &&      zeroreqchar = reqchar;
       (!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;  
4605    
4606      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4607      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 4040  for (;; ptr++) Line 4611  for (;; ptr++)
4611      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
4612      actual compiled code. */      actual compiled code. */
4613    
4614  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4615      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))      if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
4616    #elif !defined COMPILE_PCRE8
4617        if (xclass && !should_flip_negation)
4618    #endif
4619    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4620        {        {
4621        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_uchardata++ = XCL_END;    /* Marks the end of extra data */
4622        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
4623        code += LINK_SIZE;        code += LINK_SIZE;
4624        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT:0;
4625    
4626        /* 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;
4627        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. */
4628    
4629        if (class_charcount > 0)        if (class_has_8bitchar > 0)
4630          {          {
4631          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
4632          memmove(code + 32, code, class_utf8data - code);          memmove(code + (32 / sizeof(pcre_uchar)), code,
4633              IN_UCHARS(class_uchardata - code));
4634          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
4635          code = class_utf8data + 32;          code = class_uchardata + (32 / sizeof(pcre_uchar));
4636          }          }
4637        else code = class_utf8data;        else code = class_uchardata;
4638    
4639        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
4640    
4641        PUT(previous, 1, code - previous);        PUT(previous, 1, (int)(code - previous));
4642        break;   /* End of class handling */        break;   /* End of class handling */
4643        }        }
4644  #endif  #endif
# Line 4074  for (;; ptr++) Line 4650  for (;; ptr++)
4650      negating it if necessary. */      negating it if necessary. */
4651    
4652      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4653      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  
4654        {        {
4655          if (negate_class)
4656            for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
4657        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4658        }        }
4659      code += 32;      code += 32 / sizeof(pcre_uchar);
4660    
4661        END_CLASS:
4662      break;      break;
4663    
4664    
# Line 4120  for (;; ptr++) Line 4695  for (;; ptr++)
4695    
4696      if (repeat_min == 0)      if (repeat_min == 0)
4697        {        {
4698        firstbyte = zerofirstbyte;    /* Adjust for zero repeat */        firstchar = zerofirstchar;    /* Adjust for zero repeat */
4699        reqbyte = zeroreqbyte;        /* Ditto */        reqchar = zeroreqchar;        /* Ditto */
4700        }        }
4701    
4702      /* Remember whether this is a variable length repeat */      /* Remember whether this is a variable length repeat */
# Line 4131  for (;; ptr++) Line 4706  for (;; ptr++)
4706      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4707      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4708    
4709      /* 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
4710      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4711    
4712      tempcode = previous;      tempcode = previous;
4713    
# Line 4155  for (;; ptr++) Line 4730  for (;; ptr++)
4730        }        }
4731      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4732    
4733      /* If previous was a character match, abolish the item and generate a      /* If previous was a recursion call, wrap it in atomic brackets so that
4734      repeat item instead. If a char item has a minumum of more than one, ensure      previous becomes the atomic group. All recursions were so wrapped in the
4735      that it is set in reqbyte - it might not be if a sequence such as x{3} is      past, but it no longer happens for non-repeated recursions. In fact, the
4736      the first thing in a branch because the x will have gone into firstbyte      repeated ones could be re-implemented independently so as not to need this,
4737      instead.  */      but for the moment we rely on the code for repeating groups. */
4738    
4739        if (*previous == OP_RECURSE)
4740          {
4741          memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
4742          *previous = OP_ONCE;
4743          PUT(previous, 1, 2 + 2*LINK_SIZE);
4744          previous[2 + 2*LINK_SIZE] = OP_KET;
4745          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4746          code += 2 + 2 * LINK_SIZE;
4747          length_prevgroup = 3 + 3*LINK_SIZE;
4748    
4749          /* When actually compiling, we need to check whether this was a forward
4750          reference, and if so, adjust the offset. */
4751    
4752      if (*previous == OP_CHAR || *previous == OP_CHARNC)        if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4753            {
4754            int offset = GET(cd->hwm, -LINK_SIZE);
4755            if (offset == previous + 1 - cd->start_code)
4756              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4757            }
4758          }
4759    
4760        /* Now handle repetition for the different types of item. */
4761    
4762        /* If previous was a character or negated character match, abolish the item
4763        and generate a repeat item instead. If a char item has a minimum of more
4764        than one, ensure that it is set in reqchar - it might not be if a sequence
4765        such as x{3} is the first thing in a branch because the x will have gone
4766        into firstchar instead.  */
4767    
4768        if (*previous == OP_CHAR || *previous == OP_CHARI
4769            || *previous == OP_NOT || *previous == OP_NOTI)
4770        {        {
4771        /* Deal with UTF-8 characters that take up more than one byte. It's        switch (*previous)
4772            {
4773            default: /* Make compiler happy. */
4774            case OP_CHAR:  op_type = OP_STAR - OP_STAR; break;
4775            case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
4776            case OP_NOT:   op_type = OP_NOTSTAR - OP_STAR; break;
4777            case OP_NOTI:  op_type = OP_NOTSTARI - OP_STAR; break;
4778            }
4779    
4780          /* Deal with UTF characters that take up more than one character. It's
4781        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
4782        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
4783        length rather than a small character. */        it's a length rather than a small character. */
4784    
4785  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
4786        if (utf8 && (code[-1] & 0x80) != 0)        if (utf && NOT_FIRSTCHAR(code[-1]))
4787          {          {
4788          uschar *lastchar = code - 1;          pcre_uchar *lastchar = code - 1;
4789          while((*lastchar & 0xc0) == 0x80) lastchar--;          BACKCHAR(lastchar);
4790          c = code - lastchar;            /* Length of UTF-8 character */          c = (int)(code - lastchar);     /* Length of UTF-8 character */
4791          memcpy(utf8_char, lastchar, c); /* Save the char */          memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
4792          c |= 0x80;                      /* Flag c as a length */          c |= UTF_LENGTH;                /* Flag c as a length */
4793          }          }
4794        else        else
4795  #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. */  
4796    
4797          /* Handle the case of a single charater - either with no UTF support, or
4798          with UTF disabled, or for a single character UTF character. */
4799          {          {
4800          c = code[-1];          c = code[-1];
4801          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (*previous <= OP_CHARI && repeat_min > 1)
4802              reqchar = c | req_caseopt | cd->req_varyopt;
4803          }          }
4804    
4805        /* 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 4195  for (;; ptr++) Line 4809  for (;; ptr++)
4809    
4810        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4811            repeat_max < 0 &&            repeat_max < 0 &&
4812            check_auto_possessive(previous, utf8, ptr + 1, options, cd))            check_auto_possessive(previous, utf, ptr + 1, options, cd))
4813          {          {
4814          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4815          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4204  for (;; ptr++) Line 4818  for (;; ptr++)
4818        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
4819        }        }
4820    
     /* If previous was a single negated character ([^a] or similar), we use  
     one of the special opcodes, replacing it. The code is shared with single-  
     character repeats by setting opt_type to add a suitable offset into  
     repeat_type. We can also test for auto-possessification. OP_NOT is  
     currently used only for single-byte chars. */  
   
     else if (*previous == OP_NOT)  
       {  
       op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */  
       c = previous[1];  
       if (!possessive_quantifier &&  
           repeat_max < 0 &&  
           check_auto_possessive(previous, utf8, ptr + 1, options, cd))  
         {  
         repeat_type = 0;    /* Force greedy */  
         possessive_quantifier = TRUE;  
         }  
       goto OUTPUT_SINGLE_REPEAT;  
       }  
   
4821      /* If previous was a character type match (\d or similar), abolish it and      /* If previous was a character type match (\d or similar), abolish it and
4822      create a suitable repeat item. The code is shared with single-character      create a suitable repeat item. The code is shared with single-character
4823      repeats by setting op_type to add a suitable offset into repeat_type. Note      repeats by setting op_type to add a suitable offset into repeat_type. Note
# Line 4233  for (;; ptr++) Line 4827  for (;; ptr++)
4827    
4828      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
4829        {        {
4830        uschar *oldcode;        pcre_uchar *oldcode;
4831        int prop_type, prop_value;        int prop_type, prop_value;
4832        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
4833        c = *previous;        c = *previous;
4834    
4835        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4836            repeat_max < 0 &&            repeat_max < 0 &&
4837            check_auto_possessive(previous, utf8, ptr + 1, options, cd))            check_auto_possessive(previous, utf, ptr + 1, options, cd))
4838          {          {
4839          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4840          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4320  for (;; ptr++) Line 4914  for (;; ptr++)
4914          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
4915          Unicode property match, there are two extra bytes that define the          Unicode property match, there are two extra bytes that define the
4916          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
4917          c, with the 0x80 bit as a flag. */          c, with the UTF_LENGTH bit as a flag. */
4918    
4919          if (repeat_max < 0)          if (repeat_max < 0)
4920            {            {
4921  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
4922            if (utf8 && c >= 128)            if (utf && (c & UTF_LENGTH) != 0)
4923              {              {
4924              memcpy(code, utf8_char, c & 7);              memcpy(code, utf_chars, IN_UCHARS(c & 7));
4925              code += c & 7;              code += c & 7;
4926              }              }
4927            else            else
# Line 4349  for (;; ptr++) Line 4943  for (;; ptr++)
4943    
4944          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
4945            {            {
4946  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
4947            if (utf8 && c >= 128)            if (utf && (c & UTF_LENGTH) != 0)
4948              {              {
4949              memcpy(code, utf8_char, c & 7);              memcpy(code, utf_chars, IN_UCHARS(c & 7));
4950              code += c & 7;              code += c & 7;
4951              }              }
4952            else            else
# Line 4379  for (;; ptr++) Line 4973  for (;; ptr++)
4973    
4974        /* The character or character type itself comes last in all cases. */        /* The character or character type itself comes last in all cases. */
4975    
4976  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
4977        if (utf8 && c >= 128)        if (utf && (c & UTF_LENGTH) != 0)
4978          {          {
4979          memcpy(code, utf8_char, c & 7);          memcpy(code, utf_chars, IN_UCHARS(c & 7));
4980          code += c & 7;          code += c & 7;
4981          }          }
4982        else        else
# Line 4406  for (;; ptr++) Line 5000  for (;; ptr++)
5000    
5001      else if (*previous == OP_CLASS ||      else if (*previous == OP_CLASS ||
5002               *previous == OP_NCLASS ||               *previous == OP_NCLASS ||
5003  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5004               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
5005  #endif  #endif
5006               *previous == OP_REF)               *previous == OP_REF ||
5007                 *previous == OP_REFI)
5008        {        {
5009        if (repeat_max == 0)        if (repeat_max == 0)
5010          {          {
# Line 4443  for (;; ptr++) Line 5038  for (;; ptr++)
5038        }        }
5039    
5040      /* 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
5041      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
5042        opcodes such as BRA and CBRA, as this is the place where they get converted
5043        into the more special varieties such as BRAPOS and SBRA. A test for >=
5044        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5045        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
5046        repetition of assertions, but now it does, for Perl compatibility. */
5047    
5048      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
5049        {        {
5050        register int i;        register int i;
       int ketoffset = 0;  
5051        int len = (int)(code - previous);        int len = (int)(code - previous);
5052        uschar *bralink = NULL;        pcre_uchar *bralink = NULL;
5053          pcre_uchar *brazeroptr = NULL;
5054    
5055        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5056          we just ignore the repeat. */
5057    
5058        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5059          {          goto END_REPEAT;
5060          *errorcodeptr = ERR55;  
5061          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
5062          }        use of repetition is in cases when the assertion is optional. Therefore,
5063          if the minimum is greater than zero, just ignore the repeat. If the
5064          maximum is not not zero or one, set it to 1. */
5065    
5066        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
5067        by scanning through from the start, and compute the offset back to it          {
5068        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
5069        the final KET, so we can't find the end just by going back from the code          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = (int)(code - ket);  
5070          }          }
5071    
5072        /* 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 4492  for (;; ptr++) Line 5087  for (;; ptr++)
5087          **   goto END_REPEAT;          **   goto END_REPEAT;
5088          **   }          **   }
5089    
5090          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
5091          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
5092          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
5093          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
5094            selectively.