/[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 598 by ph10, Sat May 7 15:37:31 2011 UTC revision 848 by ph10, Wed Jan 4 16:34:01 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) 1997-2011 University of Cambridge             Copyright (c) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which
57  also used by pcretest. PCRE_DEBUG is not defined when building a production  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58  library. */  library. We do not need to select pcre16_printint.c specially, because the
59    COMPILE_PCREx macro will already be appropriately set. */
60    
61  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
62  #include "pcre_printint.src"  /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66  #endif  #endif
67    
68    
# Line 88  so this number is very generous. Line 92  so this number is very generous.
92  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
93  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
94  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
95  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
96    filled up by repetitions of forward references, for example patterns like
97    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
98    that the workspace is expanded using malloc() in this situation. The value
99    below is therefore a minimum, and we put a maximum on it for safety. The
100    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
101    kicks in at the same number of forward references in all cases. */
102    
103  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
104    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
105    
106  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
107  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
108    
109  #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)  #define WORK_SIZE_SAFETY_MARGIN (100)
110    
111    /* Private flags added to firstchar and reqchar. */
112    
113    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
114    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
115    
116    /* Repeated character flags. */
117    
118    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
119    
120  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
121  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
# Line 231  static const char posix_names[] = Line 250  static const char posix_names[] =
250    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
251    STRING_word0  STRING_xdigit;    STRING_word0  STRING_xdigit;
252    
253  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
254    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
255    
256  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
# Line 266  substitutes must be in the order of the Line 285  substitutes must be in the order of the
285  both positive and negative cases. NULL means no substitute. */  both positive and negative cases. NULL means no substitute. */
286    
287  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
288  static const uschar *substitutes[] = {  static const pcre_uchar string_PNd[]  = {
289    (uschar *)"\\P{Nd}",    /* \D */    CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
290    (uschar *)"\\p{Nd}",    /* \d */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
291    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */  static const pcre_uchar string_pNd[]  = {
292    (uschar *)"\\p{Xsp}",   /* \s */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
293    (uschar *)"\\P{Xwd}",   /* \W */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
294    (uschar *)"\\p{Xwd}"    /* \w */  static const pcre_uchar string_PXsp[] = {
295      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
296      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
297    static const pcre_uchar string_pXsp[] = {
298      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
299      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
300    static const pcre_uchar string_PXwd[] = {
301      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
302      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
303    static const pcre_uchar string_pXwd[] = {
304      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
305      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    
307    static const pcre_uchar *substitutes[] = {
308      string_PNd,           /* \D */
309      string_pNd,           /* \d */
310      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
311      string_pXsp,          /* \s */
312      string_PXwd,          /* \W */
313      string_pXwd           /* \w */
314  };  };
315    
316  static const uschar *posix_substitutes[] = {  static const pcre_uchar string_pL[] =   {
317    (uschar *)"\\p{L}",     /* alpha */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
318    (uschar *)"\\p{Ll}",    /* lower */    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
319    (uschar *)"\\p{Lu}",    /* upper */  static const pcre_uchar string_pLl[] =  {
320    (uschar *)"\\p{Xan}",   /* alnum */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
321    NULL,                   /* ascii */    CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
322    (uschar *)"\\h",        /* blank */  static const pcre_uchar string_pLu[] =  {
323    NULL,                   /* cntrl */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
324    (uschar *)"\\p{Nd}",    /* digit */    CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
325    NULL,                   /* graph */  static const pcre_uchar string_pXan[] = {
326    NULL,                   /* print */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
327    NULL,                   /* punct */    CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
328    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */  static const pcre_uchar string_h[] =    {
329    (uschar *)"\\p{Xwd}",   /* word */    CHAR_BACKSLASH, CHAR_h, '\0' };
330    NULL,                   /* xdigit */  static const pcre_uchar string_pXps[] = {
331      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
332      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
333    static const pcre_uchar string_PL[] =   {
334      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
335      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
336    static const pcre_uchar string_PLl[] =  {
337      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
338      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
339    static const pcre_uchar string_PLu[] =  {
340      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
341      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
342    static const pcre_uchar string_PXan[] = {
343      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
344      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
345    static const pcre_uchar string_H[] =    {
346      CHAR_BACKSLASH, CHAR_H, '\0' };
347    static const pcre_uchar string_PXps[] = {
348      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
349      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
350    
351    static const pcre_uchar *posix_substitutes[] = {
352      string_pL,            /* alpha */
353      string_pLl,           /* lower */
354      string_pLu,           /* upper */
355      string_pXan,          /* alnum */
356      NULL,                 /* ascii */
357      string_h,             /* blank */
358      NULL,                 /* cntrl */
359      string_pNd,           /* digit */
360      NULL,                 /* graph */
361      NULL,                 /* print */
362      NULL,                 /* punct */
363      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
364      string_pXwd,          /* word */
365      NULL,                 /* xdigit */
366    /* Negated cases */    /* Negated cases */
367    (uschar *)"\\P{L}",     /* ^alpha */    string_PL,            /* ^alpha */
368    (uschar *)"\\P{Ll}",    /* ^lower */    string_PLl,           /* ^lower */
369    (uschar *)"\\P{Lu}",    /* ^upper */    string_PLu,           /* ^upper */
370    (uschar *)"\\P{Xan}",   /* ^alnum */    string_PXan,          /* ^alnum */
371    NULL,                   /* ^ascii */    NULL,                 /* ^ascii */
372    (uschar *)"\\H",        /* ^blank */    string_H,             /* ^blank */
373    NULL,                   /* ^cntrl */    NULL,                 /* ^cntrl */
374    (uschar *)"\\P{Nd}",    /* ^digit */    string_PNd,           /* ^digit */
375    NULL,                   /* ^graph */    NULL,                 /* ^graph */
376    NULL,                   /* ^print */    NULL,                 /* ^print */
377    NULL,                   /* ^punct */    NULL,                 /* ^punct */
378    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
379    (uschar *)"\\P{Xwd}",   /* ^word */    string_PXwd,          /* ^word */
380    NULL                    /* ^xdigit */    NULL                  /* ^xdigit */
381  };  };
382  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
383  #endif  #endif
384    
385  #define STRING(a)  # a  #define STRING(a)  # a
# Line 365  static const char error_texts[] = Line 438  static const char error_texts[] =
438    /* 30 */    /* 30 */
439    "unknown POSIX class name\0"    "unknown POSIX class name\0"
440    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
441    "this version of PCRE is not compiled with PCRE_UTF8 support\0"    "this version of PCRE is compiled without UTF support\0"
442    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
443    "character value in \\x{...} sequence is too large\0"    "character value in \\x{...} sequence is too large\0"
444    /* 35 */    /* 35 */
# Line 388  static const char error_texts[] = Line 461  static const char error_texts[] =
461    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
462    /* 50 */    /* 50 */
463    "repeated subpattern is too long\0"    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
464    "octal value is greater than \\377 (not in UTF-8 mode)\0"    "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
465    "internal error: overran compiling workspace\0"    "internal error: overran compiling workspace\0"
466    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
467    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
468    /* 55 */    /* 55 */
469    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
470    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
471    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
472    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
# Line 407  static const char error_texts[] = Line 480  static const char error_texts[] =
480    /* 65 */    /* 65 */
481    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
482    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
483    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with Unicode property support\0"
484    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
485      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
486      /* 70 */
487      "internal error: unknown opcode in find_fixedlength()\0"
488      "\\N is not supported in a class\0"
489      "too many forward references\0"
490      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
491      "invalid UTF-16 string\0"
492    ;    ;
493    
494  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 427  For convenience, we use the same bit def Line 507  For convenience, we use the same bit def
507    
508  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
509    
510    /* Using a simple comparison for decimal numbers rather than a memory read
511    is much faster, and the resulting code is simpler (the compiler turns it
512    into a subtraction and unsigned comparison). */
513    
514    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
515    
516  #ifndef EBCDIC  #ifndef EBCDIC
517    
518  /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in  /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
519  UTF-8 mode. */  UTF-8 mode. */
520    
521  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
522    {    {
523    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
524    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 471  static const unsigned char digitab[] = Line 557  static const unsigned char digitab[] =
557    
558  /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */  /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
559    
560  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
561    {    {
562    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
563    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 506  static const unsigned char digitab[] = Line 592  static const unsigned char digitab[] =
592    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
593    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
594    
595  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
596    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
597    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
598    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 545  static const unsigned char ebcdic_charta Line 631  static const unsigned char ebcdic_charta
631  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
632    
633  static BOOL  static BOOL
634    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
635      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
636    
637    
# Line 577  return s; Line 663  return s;
663    
664    
665  /*************************************************  /*************************************************
666    *           Expand the workspace                 *
667    *************************************************/
668    
669    /* This function is called during the second compiling phase, if the number of
670    forward references fills the existing workspace, which is originally a block on
671    the stack. A larger block is obtained from malloc() unless the ultimate limit
672    has been reached or the increase will be rather small.
673    
674    Argument: pointer to the compile data block
675    Returns:  0 if all went well, else an error number
676    */
677    
678    static int
679    expand_workspace(compile_data *cd)
680    {
681    pcre_uchar *newspace;
682    int newsize = cd->workspace_size * 2;
683    
684    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
685    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
686        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
687     return ERR72;
688    
689    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
690    if (newspace == NULL) return ERR21;
691    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
692    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
693    if (cd->workspace_size > COMPILE_WORK_SIZE)
694      (PUBL(free))((void *)cd->start_workspace);
695    cd->start_workspace = newspace;
696    cd->workspace_size = newsize;
697    return 0;
698    }
699    
700    
701    
702    /*************************************************
703    *            Check for counted repeat            *
704    *************************************************/
705    
706    /* This function is called when a '{' is encountered in a place where it might
707    start a quantifier. It looks ahead to see if it really is a quantifier or not.
708    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
709    where the ddds are digits.
710    
711    Arguments:
712      p         pointer to the first char after '{'
713    
714    Returns:    TRUE or FALSE
715    */
716    
717    static BOOL
718    is_counted_repeat(const pcre_uchar *p)
719    {
720    if (!IS_DIGIT(*p)) return FALSE;
721    p++;
722    while (IS_DIGIT(*p)) p++;
723    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
724    
725    if (*p++ != CHAR_COMMA) return FALSE;
726    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
727    
728    if (!IS_DIGIT(*p)) return FALSE;
729    p++;
730    while (IS_DIGIT(*p)) p++;
731    
732    return (*p == CHAR_RIGHT_CURLY_BRACKET);
733    }
734    
735    
736    
737    /*************************************************
738  *            Handle escapes                      *  *            Handle escapes                      *
739  *************************************************/  *************************************************/
740    
# Line 601  Returns:         zero or positive => a d Line 759  Returns:         zero or positive => a d
759  */  */
760    
761  static int  static int
762  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
763    int options, BOOL isclass)    int options, BOOL isclass)
764  {  {
765  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
766  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
767  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
768    pcre_int32 c;
769    int i;
770    
771  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
772  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
# Line 620  in a table. A non-zero result is somethi Line 780  in a table. A non-zero result is somethi
780  Otherwise further processing may be required. */  Otherwise further processing may be required. */
781    
782  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
783  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */  /* Not alphanumeric */
784    else if (c < CHAR_0 || c > CHAR_z) {}
785  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
786    
787  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
788  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
789    else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
790  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
791  #endif  #endif
792    
# Line 632  else if ((i = escapes[c - 0x48]) != 0) Line 794  else if ((i = escapes[c - 0x48]) != 0)
794    
795  else  else
796    {    {
797    const uschar *oldptr;    const pcre_uchar *oldptr;
798    BOOL braced, negated;    BOOL braced, negated;
799    
800    switch (c)    switch (c)
# Line 642  else Line 804  else
804    
805      case CHAR_l:      case CHAR_l:
806      case CHAR_L:      case CHAR_L:
807        *errorcodeptr = ERR37;
808        break;
809    
810      case CHAR_u:      case CHAR_u:
811        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
812          {
813          /* In JavaScript, \u must be followed by four hexadecimal numbers.
814          Otherwise it is a lowercase u letter. */
815          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
816            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
817            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
818            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
819            {
820            c = 0;
821            for (i = 0; i < 4; ++i)
822              {
823              register int cc = *(++ptr);
824    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
825              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
826              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
827    #else           /* EBCDIC coding */
828              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
829              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
830    #endif
831              }
832            }
833          }
834        else
835          *errorcodeptr = ERR37;
836        break;
837    
838      case CHAR_U:      case CHAR_U:
839      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
840        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
841      break;      break;
842    
843      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
844        class, \g must be followed by one of a number of specific things:
845    
846      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
847      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 664  else Line 858  else
858      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
859    
860      case CHAR_g:      case CHAR_g:
861        if (isclass) break;
862      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
863        {        {
864        c = -ESC_g;        c = -ESC_g;
# Line 674  else Line 869  else
869    
870      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
871        {        {
872        const uschar *p;        const pcre_uchar *p;
873        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
874          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
875        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
876          {          {
877          c = -ESC_k;          c = -ESC_k;
# Line 694  else Line 889  else
889        }        }
890      else negated = FALSE;      else negated = FALSE;
891    
892        /* The integer range is limited by the machine's int representation. */
893      c = 0;      c = 0;
894      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
895          {
896          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
897            {
898            c = -1;
899            break;
900            }
901        c = c * 10 + *(++ptr) - CHAR_0;        c = c * 10 + *(++ptr) - CHAR_0;
902          }
903      if (c < 0)   /* Integer overflow */      if (((unsigned int)c) > INT_MAX) /* Integer overflow */
904        {        {
905          while (IS_DIGIT(ptr[1]))
906            ptr++;
907        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
908        break;        break;
909        }        }
# Line 747  else Line 951  else
951      if (!isclass)      if (!isclass)
952        {        {
953        oldptr = ptr;        oldptr = ptr;
954          /* The integer range is limited by the machine's int representation. */
955        c -= CHAR_0;        c -= CHAR_0;
956        while ((digitab[ptr[1]] & ctype_digit) != 0)        while (IS_DIGIT(ptr[1]))
957            {
958            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
959              {
960              c = -1;
961              break;
962              }
963          c = c * 10 + *(++ptr) - CHAR_0;          c = c * 10 + *(++ptr) - CHAR_0;
964        if (c < 0)    /* Integer overflow */          }
965          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
966          {          {
967            while (IS_DIGIT(ptr[1]))
968              ptr++;
969          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
970          break;          break;
971          }          }
# Line 784  else Line 998  else
998      c -= CHAR_0;      c -= CHAR_0;
999      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1000          c = c * 8 + *(++ptr) - CHAR_0;          c = c * 8 + *(++ptr) - CHAR_0;
1001      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf && c > 0xff) *errorcodeptr = ERR51;
1002      break;      break;
1003    
1004      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \x is complicated. \x{ddd} is a character number which can be greater
1005      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.
1006      treated as a data character. */      If not, { is treated as a data character. */
1007    
1008      case CHAR_x:      case CHAR_x:
1009        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1010          {
1011          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1012          Otherwise it is a lowercase x letter. */
1013          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1014            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1015            {
1016            c = 0;
1017            for (i = 0; i < 2; ++i)
1018              {
1019              register int cc = *(++ptr);
1020    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1021              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1022              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1023    #else           /* EBCDIC coding */
1024              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1025              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1026    #endif
1027              }
1028            }
1029          break;
1030          }
1031    
1032      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1033        {        {
1034        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
1035    
1036        c = 0;        c = 0;
1037        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1038          {          {
1039          register int cc = *pt++;          register int cc = *pt++;
1040          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
1041    
1042  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1043          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
# Line 811  else Line 1046  else
1046          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1047          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1048  #endif  #endif
1049    
1050    #ifdef COMPILE_PCRE8
1051            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1052    #else
1053    #ifdef COMPILE_PCRE16
1054            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1055    #endif
1056    #endif
1057            }
1058    
1059          if (c < 0)
1060            {
1061            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1062            *errorcodeptr = ERR34;
1063          }          }
1064    
1065        if (*pt == CHAR_RIGHT_CURLY_BRACKET)        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1066          {          {
1067          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1068          ptr = pt;          ptr = pt;
1069          break;          break;
1070          }          }
# Line 827  else Line 1076  else
1076      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1077    
1078      c = 0;      c = 0;
1079      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1080        {        {
1081        int cc;                                  /* Some compilers don't like */        int cc;                                  /* Some compilers don't like */
1082        cc = *(++ptr);                           /* ++ in initializers */        cc = *(++ptr);                           /* ++ in initializers */
# Line 885  else Line 1134  else
1134    }    }
1135    
1136  /* Perl supports \N{name} for character names, as well as plain \N for "not  /* Perl supports \N{name} for character names, as well as plain \N for "not
1137  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
1138    quantification such as \N{2,3}. */
1139    
1140  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1141         !is_counted_repeat(ptr+2))
1142    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1143    
1144  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
# Line 923  Returns:         type value from ucp_typ Line 1174  Returns:         type value from ucp_typ
1174  */  */
1175    
1176  static int  static int
1177  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1178  {  {
1179  int c, i, bot, top;  int c, i, bot, top;
1180  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1181  char name[32];  pcre_uchar name[32];
1182    
1183  c = *(++ptr);  c = *(++ptr);
1184  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 944  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1195  if (c == CHAR_LEFT_CURLY_BRACKET)
1195      *negptr = TRUE;      *negptr = TRUE;
1196      ptr++;      ptr++;
1197      }      }
1198    for (i = 0; i < (int)sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1199      {      {
1200      c = *(++ptr);      c = *(++ptr);
1201      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
# Line 968  else Line 1219  else
1219  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1220    
1221  bot = 0;  bot = 0;
1222  top = _pcre_utt_size;  top = PRIV(utt_size);
1223    
1224  while (bot < top)  while (bot < top)
1225    {    {
1226    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1227    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1228    if (c == 0)    if (c == 0)
1229      {      {
1230      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1231      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1232      }      }
1233    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1234    }    }
# Line 997  return -1; Line 1248  return -1;
1248    
1249    
1250  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
 /*************************************************  
1251  *         Read repeat counts                     *  *         Read repeat counts                     *
1252  *************************************************/  *************************************************/
1253    
# Line 1048  Returns:         pointer to '}' on succe Line 1266  Returns:         pointer to '}' on succe
1266                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1267  */  */
1268    
1269  static const uschar *  static const pcre_uchar *
1270  read_repeat_counts(const uschar *p, int *minp, int *maxp, int *errorcodeptr)  read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1271  {  {
1272  int min = 0;  int min = 0;
1273  int max = -1;  int max = -1;
# Line 1057  int max = -1; Line 1275  int max = -1;
1275  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1276  an integer overflow. */  an integer overflow. */
1277    
1278  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1279  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1280    {    {
1281    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 1072  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1290  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1290    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1291      {      {
1292      max = 0;      max = 0;
1293      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1294      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1295        {        {
1296        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 1127  Arguments: Line 1345  Arguments:
1345    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1346    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1347    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1348    utf8         TRUE if we are in UTF-8 mode    utf          TRUE if we are in UTF-8 / UTF-16 mode
1349    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1350    
1351  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1352  */  */
1353    
1354  static int  static int
1355  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,
1356    BOOL xmode, BOOL utf8, int *count)    BOOL xmode, BOOL utf, int *count)
1357  {  {
1358  uschar *ptr = *ptrptr;  pcre_uchar *ptr = *ptrptr;
1359  int start_count = *count;  int start_count = *count;
1360  int hwm_count = start_count;  int hwm_count = start_count;
1361  BOOL dup_parens = FALSE;  BOOL dup_parens = FALSE;
# Line 1204  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1422  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1422          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1423        {        {
1424        int term;        int term;
1425        const uschar *thisname;        const pcre_uchar *thisname;
1426        *count += 1;        *count += 1;
1427        if (name == NULL && *count == lorn) return *count;        if (name == NULL && *count == lorn) return *count;
1428        term = *ptr++;        term = *ptr++;
# Line 1212  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1430  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1430        thisname = ptr;        thisname = ptr;
1431        while (*ptr != term) ptr++;        while (*ptr != term) ptr++;
1432        if (name != NULL && lorn == ptr - thisname &&        if (name != NULL && lorn == ptr - thisname &&
1433            strncmp((const char *)name, (const char *)thisname, lorn) == 0)            STRNCMP_UC_UC(name, thisname, lorn) == 0)
1434          return *count;          return *count;
1435        term++;        term++;
1436        }        }
# Line 1255  for (; ptr < cd->end_pattern; ptr++) Line 1473  for (; ptr < cd->end_pattern; ptr++)
1473          {          {
1474          if (ptr[2] == CHAR_E)          if (ptr[2] == CHAR_E)
1475            ptr+= 2;            ptr+= 2;
1476          else if (strncmp((const char *)ptr+2,          else if (STRNCMP_UC_C8(ptr + 2,
1477                   STR_Q STR_BACKSLASH STR_E, 3) == 0)                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
1478            ptr += 4;            ptr += 4;
1479          else          else
# Line 1303  for (; ptr < cd->end_pattern; ptr++) Line 1521  for (; ptr < cd->end_pattern; ptr++)
1521        {        {
1522        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1523        ptr++;        ptr++;
1524  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1525        if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;        if (utf) FORWARDCHAR(ptr);
1526  #endif  #endif
1527        }        }
1528      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
# Line 1315  for (; ptr < cd->end_pattern; ptr++) Line 1533  for (; ptr < cd->end_pattern; ptr++)
1533    
1534    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1535      {      {
1536      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1537      if (rc > 0) return rc;      if (rc > 0) return rc;
1538      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1539      }      }
# Line 1361  Arguments: Line 1579  Arguments:
1579    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1580    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1581    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1582    utf8         TRUE if we are in UTF-8 mode    utf          TRUE if we are in UTF-8 / UTF-16 mode
1583    
1584  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1585  */  */
1586    
1587  static int  static int
1588  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,  find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1589    BOOL utf8)    BOOL utf)
1590  {  {
1591  uschar *ptr = (uschar *)cd->start_pattern;  pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1592  int count = 0;  int count = 0;
1593  int rc;  int rc;
1594    
# Line 1381  matching closing parens. That is why we Line 1599  matching closing parens. That is why we
1599    
1600  for (;;)  for (;;)
1601    {    {
1602    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1603    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1604    }    }
1605    
# Line 1397  return rc; Line 1615  return rc;
1615    
1616  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1617  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1618  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1619  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1620  assertions, and also the \b assertion; for others it does not.  does not.
1621    
1622  Arguments:  Arguments:
1623    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1624    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1625    
1626  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1627  */  */
1628    
1629  static const uschar*  static const pcre_uchar*
1630  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1631  {  {
1632  for (;;)  for (;;)
1633    {    {
1634    switch ((int)*code)    switch ((int)*code)
1635      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1636      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1637      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1638      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1639      if (!skipassert) return code;      if (!skipassert) return code;
1640      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1641      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1642      break;      break;
1643    
1644      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1444  for (;;) Line 1652  for (;;)
1652      case OP_RREF:      case OP_RREF:
1653      case OP_NRREF:      case OP_NRREF:
1654      case OP_DEF:      case OP_DEF:
1655      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1656      break;      break;
1657    
1658      default:      default:
# Line 1474  and doing the check at the end; a flag s Line 1682  and doing the check at the end; a flag s
1682    
1683  Arguments:  Arguments:
1684    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1685    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 mode
1686    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1687    cd       the "compile data" structure    cd       the "compile data" structure
1688    
1689  Returns:   the fixed length,  Returns:   the fixed length,
1690               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1691               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1692               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1693                 or -4 if an unknown opcode was encountered (internal error)
1694  */  */
1695    
1696  static int  static int
1697  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1698  {  {
1699  int length = -1;  int length = -1;
1700    
1701  register int branchlength = 0;  register int branchlength = 0;
1702  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1703    
1704  /* Scan along the opcodes for this branch. If we get to the end of the  /* Scan along the opcodes for this branch. If we get to the end of the
1705  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1498  branch, check the length against that of Line 1707  branch, check the length against that of
1707  for (;;)  for (;;)
1708    {    {
1709    int d;    int d;
1710    uschar *ce, *cs;    pcre_uchar *ce, *cs;
1711    register int op = *cc;    register int op = *cc;
1712    
1713    switch (op)    switch (op)
1714      {      {
1715        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1716        OP_BRA (normal non-capturing bracket) because the other variants of these
1717        opcodes are all concerned with unlimited repeated groups, which of course
1718        are not of fixed length. */
1719    
1720      case OP_CBRA:      case OP_CBRA:
1721      case OP_BRA:      case OP_BRA:
1722      case OP_ONCE:      case OP_ONCE:
1723        case OP_ONCE_NC:
1724      case OP_COND:      case OP_COND:
1725      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);
1726      if (d < 0) return d;      if (d < 0) return d;
1727      branchlength += d;      branchlength += d;
1728      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1729      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1730      break;      break;
1731    
1732      /* 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.
1733      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
1734      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
1735        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1736        because they all imply an unlimited repeat. */
1737    
1738      case OP_ALT:      case OP_ALT:
1739      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1740      case OP_END:      case OP_END:
1741        case OP_ACCEPT:
1742        case OP_ASSERT_ACCEPT:
1743      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1744        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1745      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1535  for (;;) Line 1753  for (;;)
1753    
1754      case OP_RECURSE:      case OP_RECURSE:
1755      if (!atend) return -3;      if (!atend) return -3;
1756      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1757      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1758      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                    /* Recursion */
1759      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1760      if (d < 0) return d;      if (d < 0) return d;
1761      branchlength += d;      branchlength += d;
1762      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1551  for (;;) Line 1769  for (;;)
1769      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1770      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1771      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1772      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1773        break;
1774    
1775      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1776    
1777      case OP_REVERSE:      case OP_MARK:
1778        case OP_PRUNE_ARG:
1779        case OP_SKIP_ARG:
1780        case OP_THEN_ARG:
1781        cc += cc[1] + PRIV(OP_lengths)[*cc];
1782        break;
1783    
1784        case OP_CALLOUT:
1785        case OP_CIRC:
1786        case OP_CIRCM:
1787        case OP_CLOSE:
1788        case OP_COMMIT:
1789      case OP_CREF:      case OP_CREF:
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
1790      case OP_DEF:      case OP_DEF:
1791      case OP_OPT:      case OP_DOLL:
1792      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
1793      case OP_EOD:      case OP_EOD:
1794      case OP_EODN:      case OP_EODN:
1795      case OP_CIRC:      case OP_FAIL:
1796      case OP_DOLL:      case OP_NCREF:
1797        case OP_NRREF:
1798      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1799        case OP_PRUNE:
1800        case OP_REVERSE:
1801        case OP_RREF:
1802        case OP_SET_SOM:
1803        case OP_SKIP:
1804        case OP_SOD:
1805        case OP_SOM:
1806        case OP_THEN:
1807      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1808      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1809      break;      break;
1810    
1811      /* Handle literal characters */      /* Handle literal characters */
1812    
1813      case OP_CHAR:      case OP_CHAR:
1814      case OP_CHARNC:      case OP_CHARI:
1815      case OP_NOT:      case OP_NOT:
1816        case OP_NOTI:
1817      branchlength++;      branchlength++;
1818      cc += 2;      cc += 2;
1819  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1820      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];  
1821  #endif  #endif
1822      break;      break;
1823    
# Line 1592  for (;;) Line 1825  for (;;)
1825      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1826    
1827      case OP_EXACT:      case OP_EXACT:
1828        case OP_EXACTI:
1829        case OP_NOTEXACT:
1830        case OP_NOTEXACTI:
1831      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1832      cc += 4;      cc += 2 + IMM2_SIZE;
1833  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1834      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];  
1835  #endif  #endif
1836      break;      break;
1837    
1838      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1839      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1840      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;
1841      cc += 4;      cc += 1 + IMM2_SIZE + 1;
1842      break;      break;
1843    
1844      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1613  for (;;) Line 1848  for (;;)
1848      cc += 2;      cc += 2;
1849      /* Fall through */      /* Fall through */
1850    
1851        case OP_HSPACE:
1852        case OP_VSPACE:
1853        case OP_NOT_HSPACE:
1854        case OP_NOT_VSPACE:
1855      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1856      case OP_DIGIT:      case OP_DIGIT:
1857      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1625  for (;;) Line 1864  for (;;)
1864      cc++;      cc++;
1865      break;      break;
1866    
1867      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1868        otherwise \C is coded as OP_ALLANY. */
1869    
1870      case OP_ANYBYTE:      case OP_ANYBYTE:
1871      return -2;      return -2;
1872    
1873      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1874    
1875  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16
1876      case OP_XCLASS:      case OP_XCLASS:
1877      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1878      /* Fall through */      /* Fall through */
1879  #endif  #endif
1880    
1881      case OP_CLASS:      case OP_CLASS:
1882      case OP_NCLASS:      case OP_NCLASS:
1883      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1884    
1885      switch (*cc)      switch (*cc)
1886        {        {
1887          case OP_CRPLUS:
1888          case OP_CRMINPLUS:
1889        case OP_CRSTAR:        case OP_CRSTAR:
1890        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1891        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1652  for (;;) Line 1894  for (;;)
1894    
1895        case OP_CRRANGE:        case OP_CRRANGE:
1896        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1897        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1898        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1899        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1900        break;        break;
1901    
1902        default:        default:
# Line 1664  for (;;) Line 1906  for (;;)
1906    
1907      /* Anything else is variable length */      /* Anything else is variable length */
1908    
1909      default:      case OP_ANYNL:
1910        case OP_BRAMINZERO:
1911        case OP_BRAPOS:
1912        case OP_BRAPOSZERO:
1913        case OP_BRAZERO:
1914        case OP_CBRAPOS:
1915        case OP_EXTUNI:
1916        case OP_KETRMAX:
1917        case OP_KETRMIN:
1918        case OP_KETRPOS:
1919        case OP_MINPLUS:
1920        case OP_MINPLUSI:
1921        case OP_MINQUERY:
1922        case OP_MINQUERYI:
1923        case OP_MINSTAR:
1924        case OP_MINSTARI:
1925        case OP_MINUPTO:
1926        case OP_MINUPTOI:
1927        case OP_NOTMINPLUS:
1928        case OP_NOTMINPLUSI:
1929        case OP_NOTMINQUERY:
1930        case OP_NOTMINQUERYI:
1931        case OP_NOTMINSTAR:
1932        case OP_NOTMINSTARI:
1933        case OP_NOTMINUPTO:
1934        case OP_NOTMINUPTOI:
1935        case OP_NOTPLUS:
1936        case OP_NOTPLUSI:
1937        case OP_NOTPOSPLUS:
1938        case OP_NOTPOSPLUSI:
1939        case OP_NOTPOSQUERY:
1940        case OP_NOTPOSQUERYI:
1941        case OP_NOTPOSSTAR:
1942        case OP_NOTPOSSTARI:
1943        case OP_NOTPOSUPTO:
1944        case OP_NOTPOSUPTOI:
1945        case OP_NOTQUERY:
1946        case OP_NOTQUERYI:
1947        case OP_NOTSTAR:
1948        case OP_NOTSTARI:
1949        case OP_NOTUPTO:
1950        case OP_NOTUPTOI:
1951        case OP_PLUS:
1952        case OP_PLUSI:
1953        case OP_POSPLUS:
1954        case OP_POSPLUSI:
1955        case OP_POSQUERY:
1956        case OP_POSQUERYI:
1957        case OP_POSSTAR:
1958        case OP_POSSTARI:
1959        case OP_POSUPTO:
1960        case OP_POSUPTOI:
1961        case OP_QUERY:
1962        case OP_QUERYI:
1963        case OP_REF:
1964        case OP_REFI:
1965        case OP_SBRA:
1966        case OP_SBRAPOS:
1967        case OP_SCBRA:
1968        case OP_SCBRAPOS:
1969        case OP_SCOND:
1970        case OP_SKIPZERO:
1971        case OP_STAR:
1972        case OP_STARI:
1973        case OP_TYPEMINPLUS:
1974        case OP_TYPEMINQUERY:
1975        case OP_TYPEMINSTAR:
1976        case OP_TYPEMINUPTO:
1977        case OP_TYPEPLUS:
1978        case OP_TYPEPOSPLUS:
1979        case OP_TYPEPOSQUERY:
1980        case OP_TYPEPOSSTAR:
1981        case OP_TYPEPOSUPTO:
1982        case OP_TYPEQUERY:
1983        case OP_TYPESTAR:
1984        case OP_TYPEUPTO:
1985        case OP_UPTO:
1986        case OP_UPTOI:
1987      return -1;      return -1;
1988    
1989        /* Catch unrecognized opcodes so that when new ones are added they
1990        are not forgotten, as has happened in the past. */
1991    
1992        default:
1993        return -4;
1994      }      }
1995    }    }
1996  /* Control never gets here */  /* Control never gets here */
# Line 1686  length. Line 2011  length.
2011    
2012  Arguments:  Arguments:
2013    code        points to start of expression    code        points to start of expression
2014    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2015    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2016    
2017  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
2018  */  */
2019    
2020  const uschar *  const pcre_uchar *
2021  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2022  {  {
2023  for (;;)  for (;;)
2024    {    {
2025    register int c = *code;    register int c = *code;
2026    
2027    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2028    
2029    /* 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 2036  for (;;)
2036    
2037    else if (c == OP_REVERSE)    else if (c == OP_REVERSE)
2038      {      {
2039      if (number < 0) return (uschar *)code;      if (number < 0) return (pcre_uchar *)code;
2040      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2041      }      }
2042    
2043    /* Handle capturing bracket */    /* Handle capturing bracket */
2044    
2045    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2046               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2047      {      {
2048      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
2049      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2050      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2051      }      }
2052    
2053    /* 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 2075  for (;;)
2075        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2076        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2077        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2078        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2079            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2080        break;        break;
2081    
2082        case OP_MARK:        case OP_MARK:
# Line 1758  for (;;) Line 2086  for (;;)
2086        break;        break;
2087    
2088        case OP_THEN_ARG:        case OP_THEN_ARG:
2089        code += code[1+LINK_SIZE];        code += code[1];
2090        break;        break;
2091        }        }
2092    
2093      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2094    
2095      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2096    
2097    /* 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
2098    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
2099    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2100    
2101  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2102      if (utf8) switch(c)      if (utf) switch(c)
2103        {        {
2104        case OP_CHAR:        case OP_CHAR:
2105        case OP_CHARNC:        case OP_CHARI:
2106        case OP_EXACT:        case OP_EXACT:
2107          case OP_EXACTI:
2108        case OP_UPTO:        case OP_UPTO:
2109          case OP_UPTOI:
2110        case OP_MINUPTO:        case OP_MINUPTO:
2111          case OP_MINUPTOI:
2112        case OP_POSUPTO:        case OP_POSUPTO:
2113          case OP_POSUPTOI:
2114        case OP_STAR:        case OP_STAR:
2115          case OP_STARI:
2116        case OP_MINSTAR:        case OP_MINSTAR:
2117          case OP_MINSTARI:
2118        case OP_POSSTAR:        case OP_POSSTAR:
2119          case OP_POSSTARI:
2120        case OP_PLUS:        case OP_PLUS:
2121          case OP_PLUSI:
2122        case OP_MINPLUS:        case OP_MINPLUS:
2123          case OP_MINPLUSI:
2124        case OP_POSPLUS:        case OP_POSPLUS:
2125          case OP_POSPLUSI:
2126        case OP_QUERY:        case OP_QUERY:
2127          case OP_QUERYI:
2128        case OP_MINQUERY:        case OP_MINQUERY:
2129          case OP_MINQUERYI:
2130        case OP_POSQUERY:        case OP_POSQUERY:
2131        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2132          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2133        break;        break;
2134        }        }
2135  #else  #else
2136      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2137  #endif  #endif
2138      }      }
2139    }    }
# Line 1809  instance of OP_RECURSE. Line 2150  instance of OP_RECURSE.
2150    
2151  Arguments:  Arguments:
2152    code        points to start of expression    code        points to start of expression
2153    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2154    
2155  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
2156  */  */
2157    
2158  static const uschar *  static const pcre_uchar *
2159  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2160  {  {
2161  for (;;)  for (;;)
2162    {    {
# Line 1854  for (;;) Line 2195  for (;;)
2195        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2196        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2197        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2198        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2199            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2200        break;        break;
2201    
2202        case OP_MARK:        case OP_MARK:
# Line 1864  for (;;) Line 2206  for (;;)
2206        break;        break;
2207    
2208        case OP_THEN_ARG:        case OP_THEN_ARG:
2209        code += code[1+LINK_SIZE];        code += code[1];
2210        break;        break;
2211        }        }
2212    
2213      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2214    
2215      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2216    
2217      /* 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
2218      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
2219      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2220    
2221  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2222      if (utf8) switch(c)      if (utf) switch(c)
2223        {        {
2224        case OP_CHAR:        case OP_CHAR:
2225        case OP_CHARNC:        case OP_CHARI:
2226        case OP_EXACT:        case OP_EXACT:
2227          case OP_EXACTI:
2228        case OP_UPTO:        case OP_UPTO:
2229          case OP_UPTOI:
2230        case OP_MINUPTO:        case OP_MINUPTO:
2231          case OP_MINUPTOI:
2232        case OP_POSUPTO:        case OP_POSUPTO:
2233          case OP_POSUPTOI:
2234        case OP_STAR:        case OP_STAR:
2235          case OP_STARI:
2236        case OP_MINSTAR:        case OP_MINSTAR:
2237          case OP_MINSTARI:
2238        case OP_POSSTAR:        case OP_POSSTAR:
2239          case OP_POSSTARI:
2240        case OP_PLUS:        case OP_PLUS:
2241          case OP_PLUSI:
2242        case OP_MINPLUS:        case OP_MINPLUS:
2243          case OP_MINPLUSI:
2244        case OP_POSPLUS:        case OP_POSPLUS:
2245          case OP_POSPLUSI:
2246        case OP_QUERY:        case OP_QUERY:
2247          case OP_QUERYI:
2248        case OP_MINQUERY:        case OP_MINQUERY:
2249          case OP_MINQUERYI:
2250        case OP_POSQUERY:        case OP_POSQUERY:
2251        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2252          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2253        break;        break;
2254        }        }
2255  #else  #else
2256      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2257  #endif  #endif
2258      }      }
2259    }    }
# Line 1921  bracket whose current branch will alread Line 2276  bracket whose current branch will alread
2276  Arguments:  Arguments:
2277    code        points to start of search    code        points to start of search
2278    endcode     points to where to stop    endcode     points to where to stop
2279    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2280    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2281    
2282  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2283  */  */
2284    
2285  static BOOL  static BOOL
2286  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2287    compile_data *cd)    BOOL utf, compile_data *cd)
2288  {  {
2289  register int c;  register int c;
2290  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2291       code < endcode;       code < endcode;
2292       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2293    {    {
2294    const uschar *ccode;    const pcre_uchar *ccode;
2295    
2296    c = *code;    c = *code;
2297    
# Line 1950  for (code = first_significant_code(code Line 2305  for (code = first_significant_code(code
2305      continue;      continue;
2306      }      }
2307    
   /* 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;  
     }  
   
2308    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2309    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2310      forward reference subroutine call, we can't. To detect forward reference
2311      we have to scan up the list that is kept in the workspace. This function is
2312      called only when doing the real compile, not during the pre-compile that
2313      measures the size of the compiled pattern. */
2314    
2315    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2316      {      {
2317      BOOL empty_branch = FALSE;      const pcre_uchar *scode;
2318      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
2319    
2320        /* Test for forward reference */
2321    
2322        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2323          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2324    
2325        /* Not a forward reference, test for completed backward reference */
2326    
2327        empty_branch = FALSE;
2328        scode = cd->start_code + GET(code, 1);
2329      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2330    
2331        /* Completed backwards reference */
2332    
2333      do      do
2334        {        {
2335        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf, cd))
2336          {          {
2337          empty_branch = TRUE;          empty_branch = TRUE;
2338          break;          break;
# Line 1978  for (code = first_significant_code(code Line 2340  for (code = first_significant_code(code
2340        scode += GET(scode, 1);        scode += GET(scode, 1);
2341        }        }
2342      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2343    
2344      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2345      continue;      continue;
2346      }      }
2347    
2348      /* Groups with zero repeats can of course be empty; skip them. */
2349    
2350      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2351          c == OP_BRAPOSZERO)
2352        {
2353        code += PRIV(OP_lengths)[c];
2354        do code += GET(code, 1); while (*code == OP_ALT);
2355        c = *code;
2356        continue;
2357        }
2358    
2359      /* A nested group that is already marked as "could be empty" can just be
2360      skipped. */
2361    
2362      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2363          c == OP_SCBRA || c == OP_SCBRAPOS)
2364        {
2365        do code += GET(code, 1); while (*code == OP_ALT);
2366        c = *code;
2367        continue;
2368        }
2369    
2370    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2371    
2372    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2373          c == OP_CBRA || c == OP_CBRAPOS ||
2374          c == OP_ONCE || c == OP_ONCE_NC ||
2375          c == OP_COND)
2376      {      {
2377      BOOL empty_branch;      BOOL empty_branch;
2378      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 2388  for (code = first_significant_code(code
2388        empty_branch = FALSE;        empty_branch = FALSE;
2389        do        do
2390          {          {
2391          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2392            empty_branch = TRUE;            empty_branch = TRUE;
2393          code += GET(code, 1);          code += GET(code, 1);
2394          }          }
# Line 2018  for (code = first_significant_code(code Line 2406  for (code = first_significant_code(code
2406      {      {
2407      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2408      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2409      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2410      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"
2411      here. */      here. */
2412    
2413  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2414      case OP_XCLASS:      case OP_XCLASS:
2415      ccode = code += GET(code, 1);      ccode = code += GET(code, 1);
2416      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
# Line 2030  for (code = first_significant_code(code Line 2418  for (code = first_significant_code(code
2418    
2419      case OP_CLASS:      case OP_CLASS:
2420      case OP_NCLASS:      case OP_NCLASS:
2421      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2422    
2423  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2424      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2425  #endif  #endif
2426    
# Line 2071  for (code = first_significant_code(code Line 2459  for (code = first_significant_code(code
2459      case OP_ALLANY:      case OP_ALLANY:
2460      case OP_ANYBYTE:      case OP_ANYBYTE:
2461      case OP_CHAR:      case OP_CHAR:
2462      case OP_CHARNC:      case OP_CHARI:
2463      case OP_NOT:      case OP_NOT:
2464        case OP_NOTI:
2465      case OP_PLUS:      case OP_PLUS:
2466      case OP_MINPLUS:      case OP_MINPLUS:
2467      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2104  for (code = first_significant_code(code Line 2493  for (code = first_significant_code(code
2493      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2494      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2495      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2496      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP
2497          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2498      break;      break;
2499    
2500      /* End of branch */      /* End of branch */
# Line 2112  for (code = first_significant_code(code Line 2502  for (code = first_significant_code(code
2502      case OP_KET:      case OP_KET:
2503      case OP_KETRMAX:      case OP_KETRMAX:
2504      case OP_KETRMIN:      case OP_KETRMIN:
2505        case OP_KETRPOS:
2506      case OP_ALT:      case OP_ALT:
2507      return TRUE;      return TRUE;
2508    
2509      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2510      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2511    
2512  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2513      case OP_STAR:      case OP_STAR:
2514        case OP_STARI:
2515      case OP_MINSTAR:      case OP_MINSTAR:
2516        case OP_MINSTARI:
2517      case OP_POSSTAR:      case OP_POSSTAR:
2518        case OP_POSSTARI:
2519      case OP_QUERY:      case OP_QUERY:
2520        case OP_QUERYI:
2521      case OP_MINQUERY:      case OP_MINQUERY:
2522        case OP_MINQUERYI:
2523      case OP_POSQUERY:      case OP_POSQUERY:
2524      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      case OP_POSQUERYI:
2525        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2526      break;      break;
2527    
2528      case OP_UPTO:      case OP_UPTO:
2529        case OP_UPTOI:
2530      case OP_MINUPTO:      case OP_MINUPTO:
2531        case OP_MINUPTOI:
2532      case OP_POSUPTO:      case OP_POSUPTO:
2533      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      case OP_POSUPTOI:
2534        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2535      break;      break;
2536  #endif  #endif
2537    
# Line 2145  for (code = first_significant_code(code Line 2545  for (code = first_significant_code(code
2545      break;      break;
2546    
2547      case OP_THEN_ARG:      case OP_THEN_ARG:
2548      code += code[1+LINK_SIZE];      code += code[1];
2549      break;      break;
2550    
2551      /* 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 2568  return TRUE;
2568  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
2569  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,
2570  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.
2571    This function is called only during the real compile, not during the
2572    pre-compile.
2573    
2574  Arguments:  Arguments:
2575    code        points to start of the recursion    code        points to start of the recursion
2576    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2577    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2578    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2579    cd          pointers to tables etc    cd          pointers to tables etc
2580    
2581  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2582  */  */
2583    
2584  static BOOL  static BOOL
2585  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2586    BOOL utf8, compile_data *cd)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2587  {  {
2588  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2589    {    {
2590    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2591      return FALSE;      return FALSE;
2592    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2593    }    }
# Line 2218  where Perl recognizes it as the POSIX cl Line 2620  where Perl recognizes it as the POSIX cl
2620  "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,
2621  I think.  I think.
2622    
2623    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2624    It seems that the appearance of a nested POSIX class supersedes an apparent
2625    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2626    a digit.
2627    
2628    In Perl, unescaped square brackets may also appear as part of class names. For
2629    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2630    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2631    seem right at all. PCRE does not allow closing square brackets in POSIX class
2632    names.
2633    
2634  Arguments:  Arguments:
2635    ptr      pointer to the initial [    ptr      pointer to the initial [
2636    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2226  Returns:   TRUE or FALSE Line 2639  Returns:   TRUE or FALSE
2639  */  */
2640    
2641  static BOOL  static BOOL
2642  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2643  {  {
2644  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2645  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2646  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2647    {    {
2648    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2649        ptr++;
2650      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2651      else
2652      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2653      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2654        {        {
2655        *endptr = ptr;        *endptr = ptr;
2656        return TRUE;        return TRUE;
2657        }        }
2658        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2659             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2660              ptr[1] == CHAR_EQUALS_SIGN) &&
2661            check_posix_syntax(ptr, endptr))
2662          return FALSE;
2663      }      }
2664    }    }
2665  return FALSE;  return FALSE;
# Line 2263  Returns:     a value representing the na Line 2683  Returns:     a value representing the na
2683  */  */
2684    
2685  static int  static int
2686  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2687  {  {
2688  const char *pn = posix_names;  const char *pn = posix_names;
2689  register int yield = 0;  register int yield = 0;
2690  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2691    {    {
2692    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2693      strncmp((const char *)ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2694    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
2695    yield++;    yield++;
2696    }    }
# Line 2302  value in the reference (which is a group Line 2722  value in the reference (which is a group
2722  Arguments:  Arguments:
2723    group      points to the start of the group    group      points to the start of the group
2724    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2725    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 mode
2726    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2727    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
2728    
# Line 2310  Returns:     nothing Line 2730  Returns:     nothing
2730  */  */
2731    
2732  static void  static void
2733  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2734    uschar *save_hwm)    pcre_uchar *save_hwm)
2735  {  {
2736  uschar *ptr = group;  pcre_uchar *ptr = group;
2737    
2738  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2739    {    {
2740    int offset;    int offset;
2741    uschar *hc;    pcre_uchar *hc;
2742    
2743    /* 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
2744    reference. */    reference. */
# Line 2363  Arguments: Line 2783  Arguments:
2783  Returns:         new code pointer  Returns:         new code pointer
2784  */  */
2785    
2786  static uschar *  static pcre_uchar *
2787  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2788  {  {
2789  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2790  *code++ = 255;  *code++ = 255;
2791  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2792  PUT(code, LINK_SIZE, 0);                       /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2793  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2794  }  }
2795    
2796    
# Line 2392  Returns:             nothing Line 2812  Returns:             nothing
2812  */  */
2813    
2814  static void  static void
2815  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2816  {  {
2817  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2818  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
# Line 2475  switch(ptype) Line 2895  switch(ptype)
2895            prop->chartype == ucp_Lt) == negated;            prop->chartype == ucp_Lt) == negated;
2896    
2897    case PT_GC:    case PT_GC:
2898    return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2899    
2900    case PT_PC:    case PT_PC:
2901    return (pdata == prop->chartype) == negated;    return (pdata == prop->chartype) == negated;
# Line 2486  switch(ptype) Line 2906  switch(ptype)
2906    /* These are specials */    /* These are specials */
2907    
2908    case PT_ALNUM:    case PT_ALNUM:
2909    return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2910            _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2911    
2912    case PT_SPACE:    /* Perl space */    case PT_SPACE:    /* Perl space */
2913    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2914            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2915            == negated;            == negated;
2916    
2917    case PT_PXSPACE:  /* POSIX space */    case PT_PXSPACE:  /* POSIX space */
2918    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2919            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2920            c == CHAR_FF || c == CHAR_CR)            c == CHAR_FF || c == CHAR_CR)
2921            == negated;            == negated;
2922    
2923    case PT_WORD:    case PT_WORD:
2924    return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2925            _pcre_ucp_gentype[prop->chartype] == ucp_N ||            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2926            c == CHAR_UNDERSCORE) == negated;            c == CHAR_UNDERSCORE) == negated;
2927    }    }
2928  return FALSE;  return FALSE;
# Line 2521  sense to automatically possessify the re Line 2941  sense to automatically possessify the re
2941    
2942  Arguments:  Arguments:
2943    previous      pointer to the repeated opcode    previous      pointer to the repeated opcode
2944    utf8          TRUE in UTF-8 mode    utf           TRUE in UTF-8 / UTF-16 mode
2945    ptr           next character in pattern    ptr           next character in pattern
2946    options       options bits    options       options bits
2947    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2530  Returns:        TRUE if possessifying is Line 2950  Returns:        TRUE if possessifying is
2950  */  */
2951    
2952  static BOOL  static BOOL
2953  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
2954    int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2955  {  {
2956  int c, next;  pcre_int32 c, next;
2957  int op_code = *previous++;  int op_code = *previous++;
2958    
2959  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2542  if ((options & PCRE_EXTENDED) != 0) Line 2962  if ((options & PCRE_EXTENDED) != 0)
2962    {    {
2963    for (;;)    for (;;)
2964      {      {
2965      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2966      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2967        {        {
2968        ptr++;        ptr++;
# Line 2550  if ((options & PCRE_EXTENDED) != 0) Line 2970  if ((options & PCRE_EXTENDED) != 0)
2970          {          {
2971          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2972          ptr++;          ptr++;
2973  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2974          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf) FORWARDCHAR(ptr);
2975  #endif  #endif
2976          }          }
2977        }        }
# Line 2569  if (*ptr == CHAR_BACKSLASH) Line 2989  if (*ptr == CHAR_BACKSLASH)
2989    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
2990    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
2991    }    }
2992    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
2993    {    {
2994  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2995    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
2996  #endif  #endif
2997    next = *ptr++;    next = *ptr++;
2998    }    }
   
2999  else return FALSE;  else return FALSE;
3000    
3001  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2586  if ((options & PCRE_EXTENDED) != 0) Line 3004  if ((options & PCRE_EXTENDED) != 0)
3004    {    {
3005    for (;;)    for (;;)
3006      {      {
3007      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3008      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
3009        {        {
3010        ptr++;        ptr++;
# Line 2594  if ((options & PCRE_EXTENDED) != 0) Line 3012  if ((options & PCRE_EXTENDED) != 0)
3012          {          {
3013          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3014          ptr++;          ptr++;
3015  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3016          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf) FORWARDCHAR(ptr);
3017  #endif  #endif
3018          }          }
3019        }        }
# Line 2606  if ((options & PCRE_EXTENDED) != 0) Line 3024  if ((options & PCRE_EXTENDED) != 0)
3024  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
3025    
3026  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3027    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)
3028      return FALSE;      return FALSE;
3029    
3030  /* Now compare the next item with the previous opcode. First, handle cases when  /* Now compare the next item with the previous opcode. First, handle cases when
# Line 2615  the next item is a character. */ Line 3033  the next item is a character. */
3033  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
3034    {    {
3035    case OP_CHAR:    case OP_CHAR:
3036  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3037    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
3038  #else  #else
3039    c = *previous;    c = *previous;
3040  #endif  #endif
3041    return c != next;    return c != next;
3042    
3043    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
3044    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
3045    high-valued characters. */    high-valued characters. */
3046    
3047    case OP_CHARNC:    case OP_CHARI:
3048  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3049    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
3050  #else  #else
3051    c = *previous;    c = *previous;
3052  #endif  #endif
3053    if (c == next) return FALSE;    if (c == next) return FALSE;
3054  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3055    if (utf8)    if (utf)
3056      {      {
3057      unsigned int othercase;      unsigned int othercase;
3058      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
# Line 2646  if (next >= 0) switch(op_code) Line 3064  if (next >= 0) switch(op_code)
3064      return (unsigned int)c != othercase;      return (unsigned int)c != othercase;
3065      }      }
3066    else    else
3067  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3068    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */
3069    
3070    /* For OP_NOT, its data is always a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
3071      opcodes are not used for multi-byte characters, because they are coded using
3072      an XCLASS instead. */
3073    
3074    case OP_NOT:    case OP_NOT:
3075      return (c = *previous) == next;
3076    
3077      case OP_NOTI:
3078    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
3079    if ((options & PCRE_CASELESS) == 0) return FALSE;  #ifdef SUPPORT_UTF
3080  #ifdef SUPPORT_UTF8    if (utf)
   if (utf8)  
3081      {      {
3082      unsigned int othercase;      unsigned int othercase;
3083      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
# Line 2667  if (next >= 0) switch(op_code) Line 3089  if (next >= 0) switch(op_code)
3089      return (unsigned int)c == othercase;      return (unsigned int)c == othercase;
3090      }      }
3091    else    else
3092  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3093    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == TABLE_GET(next, cd->fcc, next));  /* Non-UTF-8 mode */
3094    
3095    /* 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.
3096    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
# Line 2758  replaced by OP_PROP codes when PCRE_UCP Line 3180  replaced by OP_PROP codes when PCRE_UCP
3180  switch(op_code)  switch(op_code)
3181    {    {
3182    case OP_CHAR:    case OP_CHAR:
3183    case OP_CHARNC:    case OP_CHARI:
3184  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3185    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
3186  #else  #else
3187    c = *previous;    c = *previous;
# Line 2864  switch(op_code) Line 3286  switch(op_code)
3286        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. */
3287    
3288        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3289          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)
3290            return FALSE;            return FALSE;
3291    
3292        /* Do the property check. */        /* Do the property check. */
# Line 2942  Arguments: Line 3364  Arguments:
3364    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3365    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3366    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3367    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3368    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3369    bcptr          points to current branch chain    bcptr          points to current branch chain
3370      cond_depth     conditional nesting depth
3371    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3372    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3373                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2954  Returns:         TRUE on success Line 3377  Returns:         TRUE on success
3377  */  */
3378    
3379  static BOOL  static BOOL
3380  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3381    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3382      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3383    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3384  {  {
3385  int repeat_type, op_type;  int repeat_type, op_type;
3386  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3387  int bravalue = 0;  int bravalue = 0;
3388  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3389  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3390  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3391  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3392  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3393  int after_manual_callout = 0;  int after_manual_callout = 0;
3394  int length_prevgroup = 0;  int length_prevgroup = 0;
3395  register int c;  register int c;
3396  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3397  uschar *last_code = code;  pcre_uchar *last_code = code;
3398  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3399  uschar *tempcode;  pcre_uchar *tempcode;
3400  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3401  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3402  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3403  const uschar *tempptr;  const pcre_uchar *tempptr;
3404  const uschar *nestptr = NULL;  const pcre_uchar *nestptr = NULL;
3405  uschar *previous = NULL;  pcre_uchar *previous = NULL;
3406  uschar *previous_callout = NULL;  pcre_uchar *previous_callout = NULL;
3407  uschar *save_hwm = NULL;  pcre_uchar *save_hwm = NULL;
3408  uschar classbits[32];  pcre_uint8 classbits[32];
3409    
3410  #ifdef SUPPORT_UTF8  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3411  BOOL class_utf8;  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3412  BOOL utf8 = (options & PCRE_UTF8) != 0;  dynamically as we process the pattern. */
3413  uschar *class_utf8data;  
3414  uschar *class_utf8data_base;  #ifdef SUPPORT_UTF
3415  uschar utf8_char[6];  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
3416    BOOL utf = (options & PCRE_UTF8) != 0;
3417    pcre_uchar utf_chars[6];
3418  #else  #else
3419  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3420  uschar *utf8_char = NULL;  #endif
3421    
3422    /* Helper variables for OP_XCLASS opcode (for characters > 255). */
3423    
3424    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3425    BOOL xclass;
3426    pcre_uchar *class_uchardata;
3427    pcre_uchar *class_uchardata_base;
3428  #endif  #endif
3429    
3430  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 3005  greedy_non_default = greedy_default ^ 1; Line 3438  greedy_non_default = greedy_default ^ 1;
3438    
3439  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3440  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
3441  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
3442  find one.  find one.
3443    
3444  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
3445  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
3446  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3447  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3448    
3449  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3450    
3451  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3452  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
3453  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
3454  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3455    value. This is used only for ASCII characters. */
3456    
3457  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3458    
3459  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3460    
# Line 3032  for (;; ptr++) Line 3466  for (;; ptr++)
3466    BOOL is_quantifier;    BOOL is_quantifier;
3467    BOOL is_recurse;    BOOL is_recurse;
3468    BOOL reset_bracount;    BOOL reset_bracount;
3469    int class_charcount;    int class_has_8bitchar;
3470    int class_lastchar;    int class_single_char;
3471    int newoptions;    int newoptions;
3472    int recno;    int recno;
3473    int refsign;    int refsign;
3474    int skipbytes;    int skipbytes;
3475    int subreqbyte;    int subreqchar;
3476    int subfirstbyte;    int subfirstchar;
3477    int terminator;    int terminator;
3478    int mclength;    int mclength;
3479    uschar mcbuffer[8];    int tempbracount;
3480      pcre_uchar mcbuffer[8];
3481    
3482    /* Get next byte in the pattern */    /* Get next character in the pattern */
3483    
3484    c = *ptr;    c = *ptr;
3485    
# Line 3066  for (;; ptr++) Line 3501  for (;; ptr++)
3501  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
3502      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3503  #endif  #endif
3504      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3505            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3506        {        {
3507        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3508        goto FAILED;        goto FAILED;
# Line 3089  for (;; ptr++) Line 3525  for (;; ptr++)
3525        }        }
3526    
3527      *lengthptr += (int)(code - last_code);      *lengthptr += (int)(code - last_code);
3528      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3529          (int)(code - last_code), c, c));
3530    
3531      /* 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
3532      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 3536  for (;; ptr++)
3536        {        {
3537        if (previous > orig_code)        if (previous > orig_code)
3538          {          {
3539          memmove(orig_code, previous, code - previous);          memmove(orig_code, previous, IN_UCHARS(code - previous));
3540          code -= previous - orig_code;          code -= previous - orig_code;
3541          previous = orig_code;          previous = orig_code;
3542          }          }
# Line 3115  for (;; ptr++) Line 3552  for (;; ptr++)
3552    /* 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
3553    reference list. */    reference list. */
3554    
3555    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3556               WORK_SIZE_SAFETY_MARGIN)
3557      {      {
3558      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3559      goto FAILED;      goto FAILED;
# Line 3163  for (;; ptr++) Line 3601  for (;; ptr++)
3601      previous_callout = NULL;      previous_callout = NULL;
3602      }      }
3603    
3604    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3605    
3606    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3607      {      {
3608      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
3609      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3610        {        {
3611        ptr++;        ptr++;
# Line 3175  for (;; ptr++) Line 3613  for (;; ptr++)
3613          {          {
3614          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3615          ptr++;          ptr++;
3616  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3617          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf) FORWARDCHAR(ptr);
3618  #endif  #endif
3619          }          }
3620        if (*ptr != 0) continue;        if (*ptr != 0) continue;
# Line 3200  for (;; ptr++) Line 3638  for (;; ptr++)
3638      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3639      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3640      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
3641      *firstbyteptr = firstbyte;      *firstcharptr = firstchar;
3642      *reqbyteptr = reqbyte;      *reqcharptr = reqchar;
3643      *codeptr = code;      *codeptr = code;
3644      *ptrptr = ptr;      *ptrptr = ptr;
3645      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3222  for (;; ptr++) Line 3660  for (;; ptr++)
3660      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3661    
3662      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3663        previous = NULL;
3664      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3665        {        {
3666        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3667          *code++ = OP_CIRCM;
3668        }        }
3669      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3670      break;      break;
3671    
3672      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3673      previous = NULL;      previous = NULL;
3674      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3675      break;      break;
3676    
3677      /* 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
3678      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqchar doesn't change either. */
3679    
3680      case CHAR_DOT:      case CHAR_DOT:
3681      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3682      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
3683      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
3684      previous = code;      previous = code;
3685      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3686      break;      break;
# Line 3296  for (;; ptr++) Line 3735  for (;; ptr++)
3735          {          {
3736          if (ptr[1] == CHAR_E)          if (ptr[1] == CHAR_E)
3737            ptr++;            ptr++;
3738          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)  
3739            ptr += 3;            ptr += 3;
3740          else          else
3741            break;            break;
# Line 3316  for (;; ptr++) Line 3754  for (;; ptr++)
3754          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3755        {        {
3756        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3757        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3758        zerofirstbyte = firstbyte;        zerofirstchar = firstchar;
3759        break;        break;
3760        }        }
3761    
# Line 3327  for (;; ptr++) Line 3765  for (;; ptr++)
3765    
3766      should_flip_negation = FALSE;      should_flip_negation = FALSE;
3767    
3768      /* 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.
3769      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
3770      valued UTF-8 characters, we don't yet do any optimization. */      < 256 character. class_single_char will be 1 if the class contains only
3771        a single character. */
3772    
3773      class_charcount = 0;      class_has_8bitchar = 0;
3774      class_lastchar = -1;      class_single_char = 0;
3775    
3776      /* 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
3777      temporary bit of memory, in case the class contains only 1 character (less      temporary bit of memory, in case the class contains only 1 character (less
3778      than 256), because in that case the compiled code doesn't use the bit map.      than 256), because in that case the compiled code doesn't use the bit map.
3779      */      */
3780    
3781      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
3782    
3783  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3784      class_utf8 = FALSE;                       /* No chars >= 256 */      xclass = FALSE;                           /* No chars >= 256 */
3785      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_uchardata = code + LINK_SIZE + 2;   /* For UTF-8 items */
3786      class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */      class_uchardata_base = class_uchardata;   /* For resetting in pass 1 */
3787  #endif  #endif
3788    
3789      /* 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 3792  for (;; ptr++)
3792    
3793      if (c != 0) do      if (c != 0) do
3794        {        {
3795        const uschar *oldptr;        const pcre_uchar *oldptr;
3796    
3797  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3798        if (utf8 && c > 127)        if (utf && HAS_EXTRALEN(c))
3799          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3800          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3801          }          }
3802    #endif
3803    
3804        /* In the pre-compile phase, accumulate the length of any UTF-8 extra  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3805          /* In the pre-compile phase, accumulate the length of any extra
3806        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
3807        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
3808        (which is on the stack). */        (which is on the stack). */
3809    
3810        if (lengthptr != NULL)        if (lengthptr != NULL)
3811          {          {
3812          *lengthptr += class_utf8data - class_utf8data_base;          *lengthptr += class_uchardata - class_uchardata_base;
3813          class_utf8data = class_utf8data_base;          class_uchardata = class_uchardata_base;
3814          }          }
   
3815  #endif  #endif
3816    
3817        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 3399  for (;; ptr++) Line 3839  for (;; ptr++)
3839          {          {
3840          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3841          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3842          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
3843          uschar pbits[32];          pcre_uint8 pbits[32];
3844    
3845          if (ptr[1] != CHAR_COLON)          if (ptr[1] != CHAR_COLON)
3846            {            {
# Line 3455  for (;; ptr++) Line 3895  for (;; ptr++)
3895          /* Copy in the first table (always present) */          /* Copy in the first table (always present) */
3896    
3897          memcpy(pbits, cbits + posix_class_maps[posix_class],          memcpy(pbits, cbits + posix_class_maps[posix_class],
3898            32 * sizeof(uschar));            32 * sizeof(pcre_uint8));
3899    
3900          /* If there is a second table, add or remove it as required. */          /* If there is a second table, add or remove it as required. */
3901    
# Line 3486  for (;; ptr++) Line 3926  for (;; ptr++)
3926            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];            for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3927    
3928          ptr = tempptr + 1;          ptr = tempptr + 1;
3929          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          /* Every class contains at least one < 256 characters. */
3930            class_has_8bitchar = 1;
3931            /* Every class contains at least two characters. */
3932            class_single_char = 2;
3933          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3934          }          }
3935    
3936        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3937        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
3938        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
3939        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
3940        class_charcount bigger than one. Unrecognized escapes fall through and        speculatively set both class_has_8bitchar and class_single_char bigger
3941        are either treated as literal characters (by default), or are faulted if        than one. Unrecognized escapes fall through and are either treated
3942          as literal characters (by default), or are faulted if
3943        PCRE_EXTRA is set. */        PCRE_EXTRA is set. */
3944    
3945        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
# Line 3504  for (;; ptr++) Line 3948  for (;; ptr++)
3948          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3949    
3950          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 */
3951            else if (-c == ESC_N)            /* \N is not supported in a class */
3952              {
3953              *errorcodeptr = ERR71;
3954              goto FAILED;
3955              }
3956          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3957            {            {
3958            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3517  for (;; ptr++) Line 3966  for (;; ptr++)
3966    
3967          if (c < 0)          if (c < 0)
3968            {            {
3969            register const uschar *cbits = cd->cbits;            register const pcre_uint8 *cbits = cd->cbits;
3970            class_charcount += 2;     /* Greater than 1 is what matters */            /* Every class contains at least two < 256 characters. */
3971              class_has_8bitchar++;
3972              /* Every class contains at least two characters. */
3973              class_single_char += 2;
3974    
3975            switch (-c)            switch (-c)
3976              {              {
# Line 3531  for (;; ptr++) Line 3983  for (;; ptr++)
3983              case ESC_SU:              case ESC_SU:
3984              nestptr = ptr;              nestptr = ptr;
3985              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3986              class_charcount -= 2;                /* Undo! */              class_has_8bitchar--;                /* Undo! */
3987              continue;              continue;
3988  #endif  #endif
3989              case ESC_d:              case ESC_d:
# Line 3572  for (;; ptr++) Line 4024  for (;; ptr++)
4024              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
4025              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
4026              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
4027  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
4028              if (utf8)              xclass = TRUE;
4029                *class_uchardata++ = XCL_SINGLE;
4030                *class_uchardata++ = 0x1680;
4031                *class_uchardata++ = XCL_SINGLE;
4032                *class_uchardata++ = 0x180e;
4033                *class_uchardata++ = XCL_RANGE;
4034                *class_uchardata++ = 0x2000;
4035                *class_uchardata++ = 0x200a;
4036                *class_uchardata++ = XCL_SINGLE;
4037                *class_uchardata++ = 0x202f;
4038                *class_uchardata++ = XCL_SINGLE;
4039                *class_uchardata++ = 0x205f;
4040                *class_uchardata++ = XCL_SINGLE;
4041                *class_uchardata++ = 0x3000;
4042    #elif defined SUPPORT_UTF
4043                if (utf)
4044                {                {
4045                class_utf8 = TRUE;                xclass = TRUE;
4046                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4047                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1680, class_uchardata);
4048                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4049                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180e, class_uchardata);
4050                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4051                class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2000, class_uchardata);
4052                class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x200a, class_uchardata);
4053                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4054                class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x202f, class_uchardata);
4055                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4056                class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x205f, class_uchardata);
4057                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
4058                class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x3000, class_uchardata);
4059                }                }
4060  #endif  #endif
4061              continue;              continue;
# Line 3606  for (;; ptr++) Line 4073  for (;; ptr++)
4073                  }                  }
4074                classbits[c] |= x;                classbits[c] |= x;
4075                }                }
4076    #ifndef COMPILE_PCRE8
4077  #ifdef SUPPORT_UTF8              xclass = TRUE;
4078              if (utf8)              *class_uchardata++ = XCL_RANGE;
4079                *class_uchardata++ = 0x0100;
4080                *class_uchardata++ = 0x167f;
4081                *class_uchardata++ = XCL_RANGE;
4082                *class_uchardata++ = 0x1681;
4083                *class_uchardata++ = 0x180d;
4084                *class_uchardata++ = XCL_RANGE;
4085                *class_uchardata++ = 0x180f;
4086                *class_uchardata++ = 0x1fff;
4087                *class_uchardata++ = XCL_RANGE;
4088                *class_uchardata++ = 0x200b;
4089                *class_uchardata++ = 0x202e;
4090                *class_uchardata++ = XCL_RANGE;
4091                *class_uchardata++ = 0x2030;
4092                *class_uchardata++ = 0x205e;
4093                *class_uchardata++ = XCL_RANGE;
4094                *class_uchardata++ = 0x2060;
4095                *class_uchardata++ = 0x2fff;
4096                *class_uchardata++ = XCL_RANGE;
4097                *class_uchardata++ = 0x3001;
4098    #ifdef SUPPORT_UTF
4099                if (utf)
4100                  class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4101                else
4102    #endif
4103                  *class_uchardata++ = 0xffff;
4104    #elif defined SUPPORT_UTF
4105                if (utf)
4106                {                {
4107                class_utf8 = TRUE;                xclass = TRUE;
4108                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4109                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);
4110                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x167f, class_uchardata);
4111                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4112                class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1681, class_uchardata);
4113                class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180d, class_uchardata);
4114                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4115                class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180f, class_uchardata);
4116                class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1fff, class_uchardata);
4117                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4118                class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x200b, class_uchardata);
4119                class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x202e, class_uchardata);
4120                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4121                class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2030, class_uchardata);
4122                class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x205e, class_uchardata);
4123                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4124                class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2060, class_uchardata);
4125                class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2fff, class_uchardata);
4126                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4127                class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x3001, class_uchardata);
4128                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4129                }                }
4130  #endif  #endif
4131              continue;              continue;
# Line 3642  for (;; ptr++) Line 4136  for (;; ptr++)
4136              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
4137              SETBIT(classbits, 0x0d); /* CR */              SETBIT(classbits, 0x0d); /* CR */
4138              SETBIT(classbits, 0x85); /* NEL */              SETBIT(classbits, 0x85); /* NEL */
4139  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
4140              if (utf8)              xclass = TRUE;
4141                *class_uchardata++ = XCL_RANGE;
4142                *class_uchardata++ = 0x2028;
4143                *class_uchardata++ = 0x2029;
4144    #elif defined SUPPORT_UTF
4145                if (utf)
4146                {                {
4147                class_utf8 = TRUE;                xclass = TRUE;
4148                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4149                class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2028, class_uchardata);
4150                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2029, class_uchardata);
4151                }                }
4152  #endif  #endif
4153              continue;              continue;
# Line 3670  for (;; ptr++) Line 4169  for (;; ptr++)
4169                classbits[c] |= x;                classbits[c] |= x;
4170                }                }
4171    
4172  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
4173              if (utf8)              xclass = TRUE;
4174                *class_uchardata++ = XCL_RANGE;
4175                *class_uchardata++ = 0x0100;
4176                *class_uchardata++ = 0x2027;
4177                *class_uchardata++ = XCL_RANGE;
4178                *class_uchardata++ = 0x202a;
4179    #ifdef SUPPORT_UTF
4180                if (utf)
4181                  class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4182                else
4183    #endif
4184                  *class_uchardata++ = 0xffff;
4185    #elif defined SUPPORT_UTF
4186                if (utf)
4187                {                {
4188                class_utf8 = TRUE;                xclass = TRUE;
4189                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4190                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);
4191                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2027, class_uchardata);
4192                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4193                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x202a, class_uchardata);
4194                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4195                }                }
4196  #endif  #endif
4197              continue;              continue;
# Line 3692  for (;; ptr++) Line 4204  for (;; ptr++)
4204                int pdata;                int pdata;
4205                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
4206                if (ptype < 0) goto FAILED;                if (ptype < 0) goto FAILED;
4207                class_utf8 = TRUE;                xclass = TRUE;
4208                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_uchardata++ = ((-c == ESC_p) != negated)?
4209                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
4210                *class_utf8data++ = ptype;                *class_uchardata++ = ptype;
4211                *class_utf8data++ = pdata;                *class_uchardata++ = pdata;
4212                class_charcount -= 2;   /* Not a < 256 character */                class_has_8bitchar--;                /* Undo! */
4213                continue;                continue;
4214                }                }
4215  #endif  #endif
# Line 3711  for (;; ptr++) Line 4223  for (;; ptr++)
4223                *errorcodeptr = ERR7;                *errorcodeptr = ERR7;
4224                goto FAILED;                goto FAILED;
4225                }                }
4226              class_charcount -= 2;  /* Undo the default count from above */              class_has_8bitchar--;    /* Undo the speculative increase. */
4227              c = *ptr;              /* Get the final character and fall through */              class_single_char -= 2;  /* Undo the speculative increase. */
4228                c = *ptr;                /* Get the final character and fall through */
4229              break;              break;
4230              }              }
4231            }            }
4232    
4233          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
4234          greater than 256 in UTF-8 mode. */          greater than 256. */
4235    
4236          }   /* End of backslash handling */          }   /* End of backslash handling */
4237    
# Line 3766  for (;; ptr++) Line 4279  for (;; ptr++)
4279            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
4280            }            }
4281    
4282  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4283          if (utf8)          if (utf)
4284            {                           /* Braces are required because the */            {                           /* Braces are required because the */
4285            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */
4286            }            }
# Line 3811  for (;; ptr++) Line 4324  for (;; ptr++)
4324    
4325          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4326    
4327            /* Since we found a character range, single character optimizations
4328            cannot be done anymore. */
4329            class_single_char = 2;
4330    
4331          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
4332          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
4333          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
4334          available. */          available. */
4335    
4336  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !(defined COMPILE_PCRE8)
4337          if (utf8 && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))          if ((d > 255) || (utf && ((options & PCRE_CASELESS) != 0 && d > 127)))
4338    #elif defined  SUPPORT_UTF
4339            if (utf && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))
4340    #elif !(defined COMPILE_PCRE8)
4341            if (d > 255)
4342    #endif
4343    #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
4344            {            {
4345            class_utf8 = TRUE;            xclass = TRUE;
4346    
4347            /* With UCP support, we can find the other case equivalents of            /* With UCP support, we can find the other case equivalents of
4348            the relevant characters. There may be several ranges. Optimize how            the relevant characters. There may be several ranges. Optimize how
4349            they fit with the basic range. */            they fit with the basic range. */
4350    
4351  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4352    #ifndef COMPILE_PCRE8
4353              if (utf && (options & PCRE_CASELESS) != 0)
4354    #else
4355            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
4356    #endif
4357              {              {
4358              unsigned int occ, ocd;              unsigned int occ, ocd;
4359              unsigned int cc = c;              unsigned int cc = c;
# Line 3852  for (;; ptr++) Line 4379  for (;; ptr++)
4379    
4380                if (occ == ocd)                if (occ == ocd)
4381                  {                  {
4382                  *class_utf8data++ = XCL_SINGLE;                  *class_uchardata++ = XCL_SINGLE;
4383                  }                  }
4384                else                else
4385                  {                  {
4386                  *class_utf8data++ = XCL_RANGE;                  *class_uchardata++ = XCL_RANGE;
4387                  class_utf8data += _pcre_ord2utf8(occ, class_utf8data);                  class_uchardata += PRIV(ord2utf)(occ, class_uchardata);
4388                  }                  }
4389                class_utf8data += _pcre_ord2utf8(ocd, class_utf8data);                class_uchardata += PRIV(ord2utf)(ocd, class_uchardata);
4390                }                }
4391              }              }
4392  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
# Line 3867  for (;; ptr++) Line 4394  for (;; ptr++)
4394            /* Now record the original range, possibly modified for UCP caseless            /* Now record the original range, possibly modified for UCP caseless
4395            overlapping ranges. */            overlapping ranges. */
4396    
4397            *class_utf8data++ = XCL_RANGE;            *class_uchardata++ = XCL_RANGE;
4398            class_utf8data += _pcre_ord2utf8(c, class_utf8data);  #ifdef SUPPORT_UTF
4399            class_utf8data += _pcre_ord2utf8(d, class_utf8data);  #ifndef COMPILE_PCRE8
4400              if (utf)
4401                {
4402                class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4403                class_uchardata += PRIV(ord2utf)(d, class_uchardata);
4404                }
4405              else
4406                {
4407                *class_uchardata++ = c;
4408                *class_uchardata++ = d;
4409                }
4410    #else
4411              class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4412              class_uchardata += PRIV(ord2utf)(d, class_uchardata);
4413    #endif
4414    #else /* SUPPORT_UTF */
4415              *class_uchardata++ = c;
4416              *class_uchardata++ = d;
4417    #endif /* SUPPORT_UTF */
4418    
4419            /* With UCP support, we are done. Without UCP support, there is no            /* With UCP support, we are done. Without UCP support, there is no
4420            caseless matching for UTF-8 characters > 127; we can use the bit map            caseless matching for UTF characters > 127; we can use the bit map
4421            for the smaller ones. */            for the smaller ones. As for 16 bit characters without UTF, we
4422              can still use  */
4423    
4424  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4425            continue;    /* With next character in the class */  #ifndef COMPILE_PCRE8
4426  #else            if (utf)
4427            if ((options & PCRE_CASELESS) == 0 || c > 127) continue;  #endif
4428                continue;    /* With next character in the class */
4429    #endif  /* SUPPORT_UCP */
4430    
4431    #if defined SUPPORT_UTF && !defined(SUPPORT_UCP) && !(defined COMPILE_PCRE8)
4432              if (utf)
4433                {
4434                if ((options & PCRE_CASELESS) == 0 || c > 127) continue;
4435                /* Adjust upper limit and fall through to set up the map */
4436                d = 127;
4437                }
4438              else
4439                {
4440                if (c > 255) continue;
4441                /* Adjust upper limit and fall through to set up the map */
4442                d = 255;
4443                }
4444    #elif defined SUPPORT_UTF && !defined(SUPPORT_UCP)
4445              if ((options & PCRE_CASELESS) == 0 || c > 127) continue;
4446            /* Adjust upper limit and fall through to set up the map */            /* Adjust upper limit and fall through to set up the map */
   
4447            d = 127;            d = 127;
4448    #else
4449  #endif  /* SUPPORT_UCP */            if (c > 255) continue;
4450              /* Adjust upper limit and fall through to set up the map */
4451              d = 255;
4452    #endif  /* SUPPORT_UTF && !SUPPORT_UCP && !COMPILE_PCRE8 */
4453            }            }
4454  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF || !COMPILE_PCRE8 */
4455    
4456          /* We use the bit map for all cases when not in UTF-8 mode; else          /* We use the bit map for 8 bit mode, or when the characters fall
4457          ranges that lie entirely within 0-127 when there is UCP support; else          partially or entirely to [0-255] ([0-127] for UCP) ranges. */
         for partial ranges without UCP support. */  
4458    
4459          class_charcount += d - c + 1;          class_has_8bitchar = 1;
         class_lastchar = d;  
4460    
4461          /* We can save a bit of time by skipping this in the pre-compile. */          /* We can save a bit of time by skipping this in the pre-compile. */
4462    
# Line 3902  for (;; ptr++) Line 4465  for (;; ptr++)
4465            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
4466            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
4467              {              {
4468              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c]; /* flip case */
4469              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
4470              }              }
4471            }            }
# Line 3916  for (;; ptr++) Line 4479  for (;; ptr++)
4479    
4480        LONE_SINGLE_CHARACTER:        LONE_SINGLE_CHARACTER:
4481    
4482        /* Handle a character that cannot go in the bit map */        /* Only the value of 1 matters for class_single_char. */
4483          if (class_single_char < 2) class_single_char++;
4484    
4485  #ifdef SUPPORT_UTF8        /* If class_charcount is 1, we saw precisely one character. As long as
4486        if (utf8 && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))        there were no negated characters >= 128 and there was no use of \p or \P,
4487          in other words, no use of any XCLASS features, we can optimize.
4488    
4489          In UTF-8 mode, we can optimize the negative case only if there were no
4490          characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4491          operate on single-bytes characters only. This is an historical hangover.
4492          Maybe one day we can tidy these opcodes to handle multi-byte characters.
4493    
4494          The optimization throws away the bit map. We turn the item into a
4495          1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4496          Note that OP_NOT[I] does not support multibyte characters. In the positive
4497          case, it can cause firstchar to be set. Otherwise, there can be no first
4498          char if this item is first, whatever repeat count may follow. In the case
4499          of reqchar, save the previous value for reinstating. */
4500    
4501    #ifdef SUPPORT_UTF
4502          if (class_single_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET
4503            && (!utf || !negate_class || c < (MAX_VALUE_FOR_SINGLE_CHAR + 1)))
4504    #else
4505          if (class_single_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
4506    #endif
4507          {          {
4508          class_utf8 = TRUE;          ptr++;
4509          *class_utf8data++ = XCL_SINGLE;          zeroreqchar = reqchar;
4510          class_utf8data += _pcre_ord2utf8(c, class_utf8data);  
4511            /* The OP_NOT[I] opcodes work on single characters only. */
4512    
4513            if (negate_class)
4514              {
4515              if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4516              zerofirstchar = firstchar;
4517              *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4518              *code++ = c;
4519              goto NOT_CHAR;
4520              }
4521    
4522            /* For a single, positive character, get the value into mcbuffer, and
4523            then we can handle this with the normal one-character code. */
4524    
4525    #ifdef SUPPORT_UTF
4526            if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
4527              mclength = PRIV(ord2utf)(c, mcbuffer);
4528            else
4529    #endif
4530              {
4531              mcbuffer[0] = c;
4532              mclength = 1;
4533              }
4534            goto ONE_CHAR;
4535            }       /* End of 1-char optimization */
4536    
4537          /* Handle a character that cannot go in the bit map. */
4538    
4539    #if defined SUPPORT_UTF && !(defined COMPILE_PCRE8)
4540          if ((c > 255) || (utf && ((options & PCRE_CASELESS) != 0 && c > 127)))
4541    #elif defined SUPPORT_UTF
4542          if (utf && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))
4543    #elif !(defined COMPILE_PCRE8)
4544          if (c > 255)
4545    #endif
4546    
4547    #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
4548            {
4549            xclass = TRUE;
4550            *class_uchardata++ = XCL_SINGLE;
4551    #ifdef SUPPORT_UTF
4552    #ifndef COMPILE_PCRE8
4553            /* In non 8 bit mode, we can get here even if we are not in UTF mode. */
4554            if (!utf)
4555              *class_uchardata++ = c;
4556            else
4557    #endif
4558              class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4559    #else /* SUPPORT_UTF */
4560            *class_uchardata++ = c;
4561    #endif /* SUPPORT_UTF */
4562    
4563  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4564    #ifdef COMPILE_PCRE8
4565          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4566    #else
4567            /* In non 8 bit mode, we can get here even if we are not in UTF mode. */
4568            if (utf && (options & PCRE_CASELESS) != 0)
4569    #endif
4570            {            {
4571            unsigned int othercase;            unsigned int othercase;
4572            if ((othercase = UCD_OTHERCASE(c)) != c)            if ((othercase = UCD_OTHERCASE(c)) != c)
4573              {              {
4574              *class_utf8data++ = XCL_SINGLE;              *class_uchardata++ = XCL_SINGLE;
4575              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_uchardata += PRIV(ord2utf)(othercase, class_uchardata);
4576              }              }
4577            }            }
4578  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4579    
4580          }          }
4581        else        else
4582  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF || COMPILE_PCRE16 */
4583    
4584        /* Handle a single-byte character */        /* Handle a single-byte character */
4585          {          {
4586            class_has_8bitchar = 1;
4587          classbits[c/8] |= (1 << (c&7));          classbits[c/8] |= (1 << (c&7));
4588          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4589            {            {
4590            c = cd->fcc[c];   /* flip case */            c = cd->fcc[c]; /* flip case */
4591            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
4592            }            }
         class_charcount++;  
         class_lastchar = c;  
4593          }          }
4594        }        }
4595    
# Line 3971  for (;; ptr++) Line 4610  for (;; ptr++)
4610        goto FAILED;        goto FAILED;
4611        }        }
4612    
4613      /* If class_charcount is 1, we saw precisely one character whose value is      /* If this is the first thing in the branch, there can be no first char
4614      less than 256. As long as there were no characters >= 128 and there was no      setting, whatever the repeat count. Any reqchar setting must remain
4615      use of \p or \P, in other words, no use of any XCLASS features, we can      unchanged after any kind of repeat. */
4616      optimize.  
4617        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4618      In UTF-8 mode, we can optimize the negative case only if there were no      zerofirstchar = firstchar;
4619      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      zeroreqchar = reqchar;
     operate on single-bytes only. This is an historical hangover. Maybe one day  
     we can tidy these opcodes to handle multi-byte characters.  
   
     The optimization throws away the bit map. We turn the item into a  
     1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note  
     that OP_NOT does not support multibyte characters. In the positive case, it  
     can cause firstbyte to be set. Otherwise, there can be no first char if  
     this item is first, whatever repeat count may follow. In the case of  
     reqbyte, save the previous value for reinstating. */  
   
 #ifdef SUPPORT_UTF8  
     if (class_charcount == 1 && !class_utf8 &&  
       (!utf8 || !negate_class || class_lastchar < 128))  
 #else  
     if (class_charcount == 1)  
 #endif  
       {  
       zeroreqbyte = reqbyte;  
   
       /* The OP_NOT opcode works on one-byte characters only. */  
   
       if (negate_class)  
         {  
         if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  
         zerofirstbyte = firstbyte;  
         *code++ = OP_NOT;  
         *code++ = class_lastchar;  
         break;  
         }  
   
       /* For a single, positive character, get the value into mcbuffer, and  
       then we can handle this with the normal one-character code. */  
   
 #ifdef SUPPORT_UTF8  
       if (utf8 && class_lastchar > 127)  
         mclength = _pcre_ord2utf8(class_lastchar, mcbuffer);  
       else  
 #endif  
         {  
         mcbuffer[0] = class_lastchar;  
         mclength = 1;  
         }  
       goto ONE_CHAR;  
       }       /* End of 1-char optimization */  
   
     /* The general case - not the one-char optimization. If this is the first  
     thing in the branch, there can be no first char setting, whatever the  
     repeat count. Any reqbyte setting must remain unchanged after any kind of  
     repeat. */  
   
     if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  
     zerofirstbyte = firstbyte;  
     zeroreqbyte = reqbyte;  
4620    
4621      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4622      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 4626  for (;; ptr++)
4626      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
4627      actual compiled code. */      actual compiled code. */
4628    
4629  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4630      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))      if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
4631    #elif !defined COMPILE_PCRE8
4632        if (xclass && !should_flip_negation)
4633    #endif
4634    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4635        {        {
4636        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_uchardata++ = XCL_END;    /* Marks the end of extra data */
4637        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
4638        code += LINK_SIZE;        code += LINK_SIZE;
4639        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT:0;
4640    
4641        /* 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;
4642        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. */
4643    
4644        if (class_charcount > 0)        if (class_has_8bitchar > 0)
4645          {          {
4646          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
4647          memmove(code + 32, code, class_utf8data - code);          memmove(code + (32 / sizeof(pcre_uchar)), code,
4648              IN_UCHARS(class_uchardata - code));
4649          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
4650          code = class_utf8data + 32;          code = class_uchardata + (32 / sizeof(pcre_uchar));
4651          }          }
4652        else code = class_utf8data;        else code = class_uchardata;
4653    
4654        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
4655    
4656        PUT(previous, 1, code - previous);        PUT(previous, 1, (int)(code - previous));
4657        break;   /* End of class handling */        break;   /* End of class handling */
4658        }        }
4659  #endif  #endif
# Line 4074  for (;; ptr++) Line 4665  for (;; ptr++)
4665      negating it if necessary. */      negating it if necessary. */
4666    
4667      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4668      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  
4669        {        {
4670          if (negate_class)
4671            for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
4672        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4673        }        }
4674      code += 32;      code += 32 / sizeof(pcre_uchar);
4675        NOT_CHAR:
4676      break;      break;
4677    
4678    
# Line 4120  for (;; ptr++) Line 4709  for (;; ptr++)
4709    
4710      if (repeat_min == 0)      if (repeat_min == 0)
4711        {        {
4712        firstbyte = zerofirstbyte;    /* Adjust for zero repeat */        firstchar = zerofirstchar;    /* Adjust for zero repeat */
4713        reqbyte = zeroreqbyte;        /* Ditto */        reqchar = zeroreqchar;        /* Ditto */
4714        }        }
4715    
4716      /* Remember whether this is a variable length repeat */      /* Remember whether this is a variable length repeat */
# Line 4131  for (;; ptr++) Line 4720  for (;; ptr++)
4720      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4721      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4722    
4723      /* 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
4724      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4725    
4726      tempcode = previous;      tempcode = previous;
4727    
# Line 4155  for (;; ptr++) Line 4744  for (;; ptr++)
4744        }        }
4745      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4746    
4747        /* If previous was a recursion call, wrap it in atomic brackets so that
4748        previous becomes the atomic group. All recursions were so wrapped in the
4749        past, but it no longer happens for non-repeated recursions. In fact, the
4750        repeated ones could be re-implemented independently so as not to need this,
4751        but for the moment we rely on the code for repeating groups. */
4752    
4753        if (*previous == OP_RECURSE)
4754          {
4755          memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
4756          *previous = OP_ONCE;
4757          PUT(previous, 1, 2 + 2*LINK_SIZE);
4758          previous[2 + 2*LINK_SIZE] = OP_KET;
4759          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4760          code += 2 + 2 * LINK_SIZE;
4761          length_prevgroup = 3 + 3*LINK_SIZE;
4762    
4763          /* When actually compiling, we need to check whether this was a forward
4764          reference, and if so, adjust the offset. */
4765    
4766          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4767            {
4768            int offset = GET(cd->hwm, -LINK_SIZE);
4769            if (offset == previous + 1 - cd->start_code)
4770              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4771            }
4772          }
4773    
4774        /* Now handle repetition for the different types of item. */
4775    
4776      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4777      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
4778      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqchar - it might not be if a sequence such as x{3} is
4779      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstchar
4780      instead.  */      instead.  */
4781    
4782      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4783        {        {
4784        /* Deal with UTF-8 characters that take up more than one byte. It's        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4785    
4786          /* Deal with UTF characters that take up more than one character. It's
4787        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
4788        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
4789        length rather than a small character. */        it's a length rather than a small character. */
4790    
4791  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4792        if (utf8 && (code[-1] & 0x80) != 0)        if (utf && NOT_FIRSTCHAR(code[-1]))
4793          {          {
4794          uschar *lastchar = code - 1;          pcre_uchar *lastchar = code - 1;
4795          while((*lastchar & 0xc0) == 0x80) lastchar--;          BACKCHAR(lastchar);
4796          c = code - lastchar;            /* Length of UTF-8 character */          c = (int)(code - lastchar);     /* Length of UTF-8 character */
4797          memcpy(utf8_char, lastchar, c); /* Save the char */          memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
4798          c |= 0x80;                      /* Flag c as a length */          c |= UTF_LENGTH;                /* Flag c as a length */
4799          }          }
4800        else        else
4801  #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. */  
4802    
4803          /* Handle the case of a single charater - either with no UTF support, or
4804          with UTF disabled, or for a single character UTF character. */
4805          {          {
4806          c = code[-1];          c = code[-1];
4807          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqchar = c | req_caseopt | cd->req_varyopt;
4808          }          }
4809    
4810        /* 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 4814  for (;; ptr++)
4814    
4815        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4816            repeat_max < 0 &&            repeat_max < 0 &&
4817            check_auto_possessive(previous, utf8, ptr + 1, options, cd))            check_auto_possessive(previous, utf, ptr + 1, options, cd))
4818          {          {
4819          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4820          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4207  for (;; ptr++) Line 4826  for (;; ptr++)
4826      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4827      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4828      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4829      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4830      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4831    
4832      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4833        {        {
4834        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4835        c = previous[1];        c = previous[1];
4836        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4837            repeat_max < 0 &&            repeat_max < 0 &&
4838            check_auto_possessive(previous, utf8, ptr + 1, options, cd))            check_auto_possessive(previous, utf, ptr + 1, options, cd))
4839          {          {
4840          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4841          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4233  for (;; ptr++) Line 4852  for (;; ptr++)
4852    
4853      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
4854        {        {
4855        uschar *oldcode;        pcre_uchar *oldcode;
4856        int prop_type, prop_value;        int prop_type, prop_value;
4857        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
4858        c = *previous;        c = *previous;
4859    
4860        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4861            repeat_max < 0 &&            repeat_max < 0 &&
4862            check_auto_possessive(previous, utf8, ptr + 1, options, cd))            check_auto_possessive(previous, utf, ptr + 1, options, cd))
4863          {          {
4864          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4865          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4320  for (;; ptr++) Line 4939  for (;; ptr++)
4939          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
4940          Unicode property match, there are two extra bytes that define the          Unicode property match, there are two extra bytes that define the
4941          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
4942          c, with the 0x80 bit as a flag. */          c, with the UTF_LENGTH bit as a flag. */
4943    
4944          if (repeat_max < 0)          if (repeat_max < 0)
4945            {            {
4946  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
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 4349  for (;; ptr++) Line 4968  for (;; ptr++)
4968    
4969          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
4970            {            {
4971  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4972            if (utf8 && c >= 128)            if (utf && (c & UTF_LENGTH) != 0)
4973              {              {
4974              memcpy(code, utf8_char, c & 7);              memcpy(code, utf_chars, IN_UCHARS(c & 7));
4975              code += c & 7;              code += c & 7;
4976              }              }
4977            else            else
# Line 4379  for (;; ptr++) Line 4998  for (;; ptr++)
4998    
4999        /* The character or character type itself comes last in all cases. */        /* The character or character type itself comes last in all cases. */
5000    
5001  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
5002        if (utf8 && c >= 128)        if (utf && (c & UTF_LENGTH) != 0)
5003          {          {
5004          memcpy(code, utf8_char, c & 7);          memcpy(code, utf_chars, IN_UCHARS(c & 7));
5005          code += c & 7;          code += c & 7;
5006          }          }
5007        else        else
# Line 4406  for (;; ptr++) Line 5025  for (;; ptr++)
5025    
5026      else if (*previous == OP_CLASS ||      else if (*previous == OP_CLASS ||
5027               *previous == OP_NCLASS ||               *previous == OP_NCLASS ||
5028  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5029               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
5030  #endif  #endif
5031               *previous == OP_REF)               *previous == OP_REF ||
5032                 *previous == OP_REFI)
5033        {        {
5034        if (repeat_max == 0)        if (repeat_max == 0)
5035          {          {
# Line 4443  for (;; ptr++) Line 5063  for (;; ptr++)
5063        }        }
5064    
5065      /* 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
5066      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
5067        opcodes such as BRA and CBRA, as this is the place where they get converted
5068        into the more special varieties such as BRAPOS and SBRA. A test for >=
5069        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5070        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
5071        repetition of assertions, but now it does, for Perl compatibility. */
5072    
5073      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
5074        {        {
5075        register int i;        register int i;
       int ketoffset = 0;  
5076        int len = (int)(code - previous);        int len = (int)(code - previous);
5077        uschar *bralink = NULL;        pcre_uchar *bralink = NULL;
5078          pcre_uchar *brazeroptr = NULL;
5079    
5080        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5081          we just ignore the repeat. */
5082    
5083        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5084          {          goto END_REPEAT;
         *errorcodeptr = ERR55;  
         goto FAILED;  
         }  
5085    
5086        /* If the maximum repeat count is unlimited, find the end of the bracket        /* There is no sense in actually repeating assertions. The only potential
5087        by scanning through from the start, and compute the offset back to it        use of repetition is in cases when the assertion is optional. Therefore,
5088        from the current code pointer. There may be an OP_OPT setting following        if the minimum is greater than zero, just ignore the repeat. If the
5089        the final KET, so we can't find the end just by going back from the code        maximum is not not zero or one, set it to 1. */
5090        pointer. */  
5091          if (*previous < OP_ONCE)    /* Assertion */
5092        if (repeat_max == -1)          {
5093          {          if (repeat_min > 0) goto END_REPEAT;
5094          register uschar *ket = previous;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = (int)(code - ket);  
5095          }          }
5096    
5097        /* 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 5112  for (;; ptr++)
5112          **   goto END_REPEAT;          **   goto END_REPEAT;
5113          **   }          **   }
5114    
5115          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
5116          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
5117          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
5118          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
5119            selectively.
5120    
5121          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5122          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 4507  for (;; ptr++) Line 5128  for (;; ptr++)
5128          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
5129            {            {
5130            *code = OP_END;            *code = OP_END;
5131            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf, cd, save_hwm);
5132            memmove(previous+1, previous, len);            memmove(previous + 1, previous, IN_UCHARS(len));
5133            code++;            code++;
5134            if (repeat_max == 0)            if (repeat_max == 0)
5135              {              {
5136              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
5137              goto END_REPEAT;              goto END_REPEAT;
5138              }              }
5139              brazeroptr = previous;    /* Save for possessive optimizing */
5140            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
5141            }            }
5142    
# Line 4530  for (;; ptr++) Line 5152  for (;; ptr++)
5152            {            {
5153            int offset;            int offset;
5154            *code = OP_END;            *code = OP_END;
5155            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);            adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
5156            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5157            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
5158            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
5159            *previous++ = OP_BRA;            *previous++ = OP_BRA;
# Line 4577  for (;; ptr++) Line 5199  for (;; ptr++)
5199              *lengthptr += delta;              *lengthptr += delta;
5200              }              }
5201    
5202            /* This is compiling for real */            /* This is compiling for real. If there is a set first byte for
5203              the group, and we have not yet set a "required byte", set it. Make
5204              sure there is enough workspace for copying forward references before
5205              doing the copy. */
5206    
5207            else            else
5208              {              {
5209              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;              if (groupsetfirstchar && reqchar < 0) reqchar = firstchar;
5210    
5211              for (i = 1; i < repeat_min; i++)              for (i = 1; i < repeat_min; i++)
5212                {                {
5213                uschar *hc;                pcre_uchar *hc;
5214                uschar *this_hwm = cd->hwm;                pcre_uchar *this_hwm = cd->hwm;
5215                memcpy(code, previous, len);                memcpy(code, previous, IN_UCHARS(len));
5216    
5217                  while (cd->hwm > cd->start_workspace + cd->workspace_size -
5218                         WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5219                    {
5220                    int save_offset = save_hwm - cd->start_workspace;
5221                    int this_offset = this_hwm - cd->start_workspace;
5222                    *errorcodeptr = expand_workspace(cd);
5223                    if (*errorcodeptr != 0) goto FAILED;
5224                    save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5225                    this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5226                    }
5227    
5228                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5229                  {                  {
5230                  PUT(cd->hwm, 0, GET(hc, 0) + len);                  PUT(cd->hwm, 0, GET(hc, 0) + len);
# Line 4636  for (;; ptr++) Line 5274  for (;; ptr++)
5274    
5275          else for (i = repeat_max - 1; i >= 0; i--)          else for (i = repeat_max - 1; i >= 0; i--)
5276            {            {
5277            uschar *hc;            pcre_uchar *hc;
5278            uschar *this_hwm = cd->hwm;            pcre_uchar *this_hwm = cd->hwm;
5279    
5280            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
5281    
# Line 4653  for (;; ptr++) Line 5291  for (;; ptr++)
5291              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
5292              }              }
5293    
5294            memcpy(code, previous, len);            memcpy(code, previous, IN_UCHARS(len));
5295    
5296              /* Ensure there is enough workspace for forward references before
5297              copying them. */
5298    
5299              while (cd->hwm > cd->start_workspace + cd->workspace_size -
5300                     WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5301                {
5302                int save_offset = save_hwm - cd->start_workspace;
5303                int this_offset = this_hwm - cd->start_workspace;
5304                *errorcodeptr = expand_workspace(cd);
5305                if (*errorcodeptr != 0) goto FAILED;
5306                save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5307                this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5308                }
5309    
5310            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5311              {              {
5312              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
# Line 4670  for (;; ptr++) Line 5323  for (;; ptr++)
5323            {            {
5324            int oldlinkoffset;            int oldlinkoffset;
5325            int offset = (int)(code - bralink + 1);            int offset = (int)(code - bralink + 1);
5326            uschar *bra = code - offset;            pcre_uchar *bra = code - offset;
5327            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
5328            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
5329            *code++ = OP_KET;            *code++ = OP_KET;
# Line 4679  for (;; ptr++) Line 5332  for (;; ptr++)
5332            }            }
5333          }          }
5334    
5335        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
5336        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
5337        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
5338        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5339          deal with possessive ONCEs specially.
5340    
5341        Then, when we are doing the actual compile phase, check to see whether        Otherwise, when we are doing the actual compile phase, check to see
5342        this group is a non-atomic one that could match an empty string. If so,        whether this group is one that could match an empty string. If so,
5343        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
5344        that runtime checking can be done. [This check is also applied to        that runtime checking can be done. [This check is also applied to ONCE
5345        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
5346    
5347          Then, if the quantifier was possessive and the bracket is not a
5348          conditional, we convert the BRA code to the POS form, and the KET code to
5349          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5350          subpattern at both the start and at the end.) The use of special opcodes
5351          makes it possible to reduce greatly the stack usage in pcre_exec(). If
5352          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5353    
5354          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5355          flag so that the default action below, of wrapping everything inside
5356          atomic brackets, does not happen. When the minimum is greater than 1,
5357          there will be earlier copies of the group, and so we still have to wrap
5358          the whole thing. */
5359    
5360        else        else
5361          {          {
5362          uschar *ketcode = code - ketoffset;          pcre_uchar *ketcode = code - 1 - LINK_SIZE;
5363          uschar *bracode = ketcode - GET(ketcode, 1);          pcre_uchar *bracode = ketcode - GET(ketcode, 1);
5364          *ketcode = OP_KETRMAX + repeat_type;  
5365          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5366    
5367            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5368                possessive_quantifier) *bracode = OP_BRA;
5369    
5370            /* For non-possessive ONCE brackets, all we need to do is to
5371            set the KET. */
5372    
5373            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5374              *ketcode = OP_KETRMAX + repeat_type;
5375    
5376            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5377            converted to non-capturing above). */
5378    
5379            else
5380            {            {
5381            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5382            do  
5383              if (lengthptr == NULL)
5384              {              {
5385              if (could_be_empty_branch(scode, ketcode, utf8, cd))              pcre_uchar *scode = bracode;
5386                do
5387                {                {
5388                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf, cd))
5389                break;                  {
5390                    *bracode += OP_SBRA - OP_BRA;
5391                    break;
5392                    }
5393                  scode += GET(scode, 1);
5394                }                }
5395              scode += GET(scode, 1);              while (*scode == OP_ALT);
5396              }              }
5397            while (*scode == OP_ALT);  
5398              /* Handle possessive quantifiers. */
5399    
5400              if (possessive_quantifier)
5401                {
5402                /* For COND brackets, we wrap the whole thing in a possessively
5403                repeated non-capturing bracket, because we have not invented POS
5404                versions of the COND opcodes. Because we are moving code along, we
5405                must ensure that any pending recursive references are updated. */
5406    
5407                if (*bracode == OP_COND || *bracode == OP_SCOND)
5408                  {
5409                  int nlen = (int)(code - bracode);
5410                  *code = OP_END;
5411                  adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
5412                  memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
5413                  code += 1 + LINK_SIZE;
5414                  nlen += 1 + LINK_SIZE;
5415                  *bracode = OP_BRAPOS;
5416                  *code++ = OP_KETRPOS;
5417                  PUTINC(code, 0, nlen);
5418                  PUT(bracode, 1, nlen);
5419                  }
5420    
5421                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5422    
5423                else
5424                  {
5425                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5426                  *ketcode = OP_KETRPOS;
5427                  }
5428    
5429                /* If the minimum is zero, mark it as possessive, then unset the
5430                possessive flag when the minimum is 0 or 1. */
5431    
5432                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5433                if (repeat_min < 2) possessive_quantifier = FALSE;
5434                }
5435    
5436              /* Non-possessive quantifier */
5437    
5438              else *ketcode = OP_KETRMAX + repeat_type;
5439            }            }
5440          }          }
5441        }        }
# Line 4728  for (;; ptr++) Line 5456  for (;; ptr++)
5456        }        }
5457    
5458      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
5459      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5460      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
5461      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5462      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
5463      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
5464      tempcode, not at previous, which might be the first part of a string whose  
5465      (former) last char we repeated.      Some (but not all) possessively repeated subpatterns have already been
5466        completely handled in the code just above. For them, possessive_quantifier
5467        is always FALSE at this stage.
5468    
5469        Note that the repeated item starts at tempcode, not at previous, which
5470        might be the first part of a string whose (former) last char we repeated.
5471    
5472      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
5473      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4745  for (;; ptr++) Line 5478  for (;; ptr++)
5478        int len;        int len;
5479    
5480        if (*tempcode == OP_TYPEEXACT)        if (*tempcode == OP_TYPEEXACT)
5481          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += PRIV(OP_lengths)[*tempcode] +
5482            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);            ((tempcode[1 + IMM2_SIZE] == OP_PROP
5483              || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
5484    
5485        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
5486          {          {
5487          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += PRIV(OP_lengths)[*tempcode];
5488  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
5489          if (utf8 && tempcode[-1] >= 0xc0)          if (utf && HAS_EXTRALEN(tempcode[-1]))
5490            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];            tempcode += GET_EXTRALEN(tempcode[-1]);
5491  #endif  #endif
5492          }          }
5493    
# Line 4765  for (;; ptr++) Line 5499  for (;; ptr++)
5499          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
5500          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
5501    
5502          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
5503          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
5504          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
5505          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
5506    
5507          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
5508          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
5509          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
5510          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
5511    
5512            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
5513            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
5514            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5515            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
5516    
5517            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
5518            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
5519            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5520            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
5521    
5522          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
5523          pending recursive references are updated. */          pending recursive references are updated. */
5524    
5525          default:          default:
5526          *code = OP_END;          *code = OP_END;
5527          adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);          adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
5528          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
5529          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
5530          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
5531          tempcode[0] = OP_ONCE;          tempcode[0] = OP_ONCE;
# Line 4793  for (;; ptr++) Line 5537  for (;; ptr++)
5537        }        }
5538    
5539      /* In all case we no longer have a previous item. We also set the      /* In all case we no longer have a previous item. We also set the
5540      "follows varying string" flag for subsequently encountered reqbytes if      "follows varying string" flag for subsequently encountered reqchars if
5541      it isn't already set and we have just passed a varying length item. */      it isn't already set and we have just passed a varying length item. */
5542    
5543      END_REPEAT:      END_REPEAT:
# Line 4816  for (;; ptr++) Line 5560  for (;; ptr++)
5560    
5561      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
5562    
5563      if (*(++ptr) == CHAR_ASTERISK &&      ptr++;
5564           ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))      if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
5565             || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
5566        {        {
5567        int i, namelen;        int i, namelen;
5568        int arglen = 0;        int arglen = 0;
5569        const char *vn = verbnames;        const char *vn = verbnames;
5570        const uschar *name = ptr + 1;        const pcre_uchar *name = ptr + 1;
5571        const uschar *arg = NULL;        const pcre_uchar *arg = NULL;
5572        previous = NULL;        previous = NULL;
5573        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        ptr++;
5574          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
5575        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5576    
5577          /* It appears that Perl allows any characters whatsoever, other than
5578          a closing parenthesis, to appear in arguments, so we no longer insist on
5579          letters, digits, and underscores. */
5580    
5581        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5582          {          {
5583          arg = ++ptr;          arg = ++ptr;
5584          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
           || *ptr == '_') ptr++;  
5585          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5586          }          }
5587    
# Line 4847  for (;; ptr++) Line 5596  for (;; ptr++)
5596        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5597          {          {
5598          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5599              strncmp((char *)name, vn, namelen) == 0)              STRNCMP_UC_C8(name, vn, namelen) == 0)
5600            {            {
5601            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5602              ASSERT_ACCEPT if in an assertion. */
5603    
5604            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5605              {              {
5606              open_capitem *oc;              open_capitem *oc;
5607                if (arglen != 0)
5608                  {
5609                  *errorcodeptr = ERR59;
5610                  goto FAILED;
5611                  }
5612              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5613              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5614                {                {
5615                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5616                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5617                }                }
5618                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5619    
5620                /* Do not set firstchar after *ACCEPT */
5621                if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
5622              }              }
5623    
5624            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5625    
5626            if (arglen == 0)            else if (arglen == 0)
5627              {              {
5628              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5629                {                {
# Line 4872  for (;; ptr++) Line 5631  for (;; ptr++)
5631                goto FAILED;                goto FAILED;
5632                }                }
5633              *code = verbs[i].op;              *code = verbs[i].op;
5634              if (*code++ == OP_THEN)              if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5635              }              }
5636    
5637            else            else
# Line 4887  for (;; ptr++) Line 5642  for (;; ptr++)
5642                goto FAILED;                goto FAILED;
5643                }                }
5644              *code = verbs[i].op_arg;              *code = verbs[i].op_arg;
5645              if (*code++ == OP_THEN_ARG)              if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5646              *code++ = arglen;              *code++ = arglen;
5647              memcpy(code, arg, arglen);              memcpy(code, arg, IN_UCHARS(arglen));
5648              code += arglen;              code += arglen;
5649              *code++ = 0;              *code++ = 0;
5650              }              }
# Line 4916  for (;; ptr++) Line 5667  for (;; ptr++)
5667        {        {
5668        int i, set, unset, namelen;        int i, set, unset, namelen;
5669        int *optset;        int *optset;
5670        const uschar *name;        const pcre_uchar *name;
5671        uschar *slot;        pcre_uchar *slot;
5672    
5673        switch (*(++ptr))        switch (*(++ptr))
5674          {          {
# Line 4970  for (;; ptr++) Line 5721  for (;; ptr++)
5721            break;            break;
5722    
5723          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Most other conditions use OP_CREF (a couple change to OP_RREF
5724          below), and all need to skip 3 bytes at the start of the group. */          below), and all need to skip 1+IMM2_SIZE bytes at the start of the group. */
5725    
5726          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
5727          skipbytes = 3;          skipbytes = 1+IMM2_SIZE;
5728          refsign = -1;          refsign = -1;
5729    
5730          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
# Line 5006  for (;; ptr++) Line 5757  for (;; ptr++)
5757    
5758          /* We now expect to read a name; any thing else is an error */          /* We now expect to read a name; any thing else is an error */
5759    
5760          if ((cd->ctypes[ptr[1]] & ctype_word) == 0)          if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)
5761            {            {
5762            ptr += 1;  /* To get the right offset */            ptr += 1;  /* To get the right offset */
5763            *errorcodeptr = ERR28;            *errorcodeptr = ERR28;
# Line 5017  for (;; ptr++) Line 5768  for (;; ptr++)
5768    
5769          recno = 0;          recno = 0;
5770          name = ++ptr;          name = ++ptr;
5771          while ((cd->ctypes[*ptr] & ctype_word) != 0)          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
5772            {            {
5773            if (recno >= 0)            if (recno >= 0)
5774              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = (IS_DIGIT(*ptr))? recno * 10 + *ptr - CHAR_0 : -1;
               recno * 10 + *ptr - CHAR_0 : -1;  
5775            ptr++;            ptr++;
5776            }            }
5777          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
# Line 5069  for (;; ptr++) Line 5819  for (;; ptr++)
5819          slot = cd->name_table;          slot = cd->name_table;
5820          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
5821            {            {
5822            if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;            if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0) break;
5823            slot += cd->name_entry_size;            slot += cd->name_entry_size;
5824            }            }
5825    
# Line 5085  for (;; ptr++) Line 5835  for (;; ptr++)
5835          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5836    
5837          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5838                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)                          (options & PCRE_EXTENDED) != 0, utf)) > 0)
5839            {            {
5840            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5841            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 5111  for (;; ptr++) Line 5861  for (;; ptr++)
5861            recno = 0;            recno = 0;
5862            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
5863              {              {
5864              if ((digitab[name[i]] & ctype_digit) == 0)              if (!IS_DIGIT(name[i]))
5865                {                {
5866                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
5867                goto FAILED;                goto FAILED;
# Line 5126  for (;; ptr++) Line 5876  for (;; ptr++)