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

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

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

code/trunk/pcre_compile.c revision 606 by ph10, Mon Jun 6 17:46:22 2011 UTC code/branches/pcre16/pcre_compile.c revision 786 by zherczeg, Tue Dec 6 11:33:41 2011 UTC
# Line 97  overrun before it actually does run off Line 97  overrun before it actually does run off
97    
98  #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)  #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    /* Private flags added to firstchar and reqchar. */
101    
102    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
103    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
104    
105    /* Repeated character flags. */
106    
107    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
108    
109  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
110  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 239  static const char posix_names[] =
239    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
240    STRING_word0  STRING_xdigit;    STRING_word0  STRING_xdigit;
241    
242  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
243    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 };
244    
245  /* 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 274  substitutes must be in the order of the
274  both positive and negative cases. NULL means no substitute. */  both positive and negative cases. NULL means no substitute. */
275    
276  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
277  static const uschar *substitutes[] = {  static const pcre_uchar string_PNd[]  = {
278    (uschar *)"\\P{Nd}",    /* \D */    CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
279    (uschar *)"\\p{Nd}",    /* \d */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
280    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */  static const pcre_uchar string_pNd[]  = {
281    (uschar *)"\\p{Xsp}",   /* \s */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
282    (uschar *)"\\P{Xwd}",   /* \W */    CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
283    (uschar *)"\\p{Xwd}"    /* \w */  static const pcre_uchar string_PXsp[] = {
284      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
285      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
286    static const pcre_uchar string_pXsp[] = {
287      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
288      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
289    static const pcre_uchar string_PXwd[] = {
290      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
291      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
292    static const pcre_uchar string_pXwd[] = {
293      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
294      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
295    
296    static const pcre_uchar *substitutes[] = {
297      string_PNd,           /* \D */
298      string_pNd,           /* \d */
299      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
300      string_pXsp,          /* \s */
301      string_PXwd,          /* \W */
302      string_pXwd           /* \w */
303  };  };
304    
305  static const uschar *posix_substitutes[] = {  static const pcre_uchar string_pL[] =   {
306    (uschar *)"\\p{L}",     /* alpha */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
307    (uschar *)"\\p{Ll}",    /* lower */    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
308    (uschar *)"\\p{Lu}",    /* upper */  static const pcre_uchar string_pLl[] =  {
309    (uschar *)"\\p{Xan}",   /* alnum */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
310    NULL,                   /* ascii */    CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
311    (uschar *)"\\h",        /* blank */  static const pcre_uchar string_pLu[] =  {
312    NULL,                   /* cntrl */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
313    (uschar *)"\\p{Nd}",    /* digit */    CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
314    NULL,                   /* graph */  static const pcre_uchar string_pXan[] = {
315    NULL,                   /* print */    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
316    NULL,                   /* punct */    CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
317    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */  static const pcre_uchar string_h[] =    {
318    (uschar *)"\\p{Xwd}",   /* word */    CHAR_BACKSLASH, CHAR_h, '\0' };
319    NULL,                   /* xdigit */  static const pcre_uchar string_pXps[] = {
320      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
321      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
322    static const pcre_uchar string_PL[] =   {
323      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
324      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
325    static const pcre_uchar string_PLl[] =  {
326      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
327      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
328    static const pcre_uchar string_PLu[] =  {
329      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    static const pcre_uchar string_PXan[] = {
332      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
333      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    static const pcre_uchar string_H[] =    {
335      CHAR_BACKSLASH, CHAR_H, '\0' };
336    static const pcre_uchar string_PXps[] = {
337      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
338      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
339    
340    static const pcre_uchar *posix_substitutes[] = {
341      string_pL,            /* alpha */
342      string_pLl,           /* lower */
343      string_pLu,           /* upper */
344      string_pXan,          /* alnum */
345      NULL,                 /* ascii */
346      string_h,             /* blank */
347      NULL,                 /* cntrl */
348      string_pNd,           /* digit */
349      NULL,                 /* graph */
350      NULL,                 /* print */
351      NULL,                 /* punct */
352      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
353      string_pXwd,          /* word */
354      NULL,                 /* xdigit */
355    /* Negated cases */    /* Negated cases */
356    (uschar *)"\\P{L}",     /* ^alpha */    string_PL,            /* ^alpha */
357    (uschar *)"\\P{Ll}",    /* ^lower */    string_PLl,           /* ^lower */
358    (uschar *)"\\P{Lu}",    /* ^upper */    string_PLu,           /* ^upper */
359    (uschar *)"\\P{Xan}",   /* ^alnum */    string_PXan,          /* ^alnum */
360    NULL,                   /* ^ascii */    NULL,                 /* ^ascii */
361    (uschar *)"\\H",        /* ^blank */    string_H,             /* ^blank */
362    NULL,                   /* ^cntrl */    NULL,                 /* ^cntrl */
363    (uschar *)"\\P{Nd}",    /* ^digit */    string_PNd,           /* ^digit */
364    NULL,                   /* ^graph */    NULL,                 /* ^graph */
365    NULL,                   /* ^print */    NULL,                 /* ^print */
366    NULL,                   /* ^punct */    NULL,                 /* ^punct */
367    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
368    (uschar *)"\\P{Xwd}",   /* ^word */    string_PXwd,          /* ^word */
369    NULL                    /* ^xdigit */    NULL                  /* ^xdigit */
370  };  };
371  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
372  #endif  #endif
373    
374  #define STRING(a)  # a  #define STRING(a)  # a
# Line 393  static const char error_texts[] = Line 455  static const char error_texts[] =
455    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
456    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
457    /* 55 */    /* 55 */
458    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
459    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
460    "\\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"
461    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
# Line 409  static const char error_texts[] = Line 471  static const char error_texts[] =
471    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
472    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
473    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
474      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
475      /* 70 */
476      "internal error: unknown opcode in find_fixedlength()\0"
477      "Not allowed UTF-8 / UTF-16 code point (>= 0xd800 && <= 0xdfff)\0"
478    ;    ;
479    
480  /* 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 493  For convenience, we use the same bit def
493    
494  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
495    
496    /* Using a simple comparison for decimal numbers rather than a memory read
497    is much faster, and the resulting code is simpler (the compiler turns it
498    into a subtraction and unsigned comparison). */
499    
500    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
501    
502  #ifndef EBCDIC  #ifndef EBCDIC
503    
504  /* 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
505  UTF-8 mode. */  UTF-8 mode. */
506    
507  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
508    {    {
509    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
510    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 543  static const unsigned char digitab[] =
543    
544  /* 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. */
545    
546  static const unsigned char digitab[] =  static const pcre_uint8 digitab[] =
547    {    {
548    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
549    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 578  static const unsigned char digitab[] =
578    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
579    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
580    
581  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
582    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
583    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
584    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 617  static const unsigned char ebcdic_charta
617  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
618    
619  static BOOL  static BOOL
620    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
621      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
622    
623    
624    
# Line 577  return s; Line 649  return s;
649    
650    
651  /*************************************************  /*************************************************
652    *            Check for counted repeat            *
653    *************************************************/
654    
655    /* This function is called when a '{' is encountered in a place where it might
656    start a quantifier. It looks ahead to see if it really is a quantifier or not.
657    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
658    where the ddds are digits.
659    
660    Arguments:
661      p         pointer to the first char after '{'
662    
663    Returns:    TRUE or FALSE
664    */
665    
666    static BOOL
667    is_counted_repeat(const pcre_uchar *p)
668    {
669    if (!IS_DIGIT(*p)) return FALSE;
670    p++;
671    while (IS_DIGIT(*p)) p++;
672    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
673    
674    if (*p++ != CHAR_COMMA) return FALSE;
675    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
676    
677    if (!IS_DIGIT(*p)) return FALSE;
678    p++;
679    while (IS_DIGIT(*p)) p++;
680    
681    return (*p == CHAR_RIGHT_CURLY_BRACKET);
682    }
683    
684    
685    
686    /*************************************************
687  *            Handle escapes                      *  *            Handle escapes                      *
688  *************************************************/  *************************************************/
689    
# Line 601  Returns:         zero or positive => a d Line 708  Returns:         zero or positive => a d
708  */  */
709    
710  static int  static int
711  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
712    int options, BOOL isclass)    int options, BOOL isclass)
713  {  {
714  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
715  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
716  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
717    pcre_int32 c;
718    int i;
719    
720  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
721  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 729  in a table. A non-zero result is somethi
729  Otherwise further processing may be required. */  Otherwise further processing may be required. */
730    
731  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
732  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */  /* Not alphanumeric */
733    else if (c < CHAR_0 || c > CHAR_z) {}
734  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
735    
736  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
737  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  /* Not alphanumeric */
738    else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
739  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
740  #endif  #endif
741    
# Line 632  else if ((i = escapes[c - 0x48]) != 0) Line 743  else if ((i = escapes[c - 0x48]) != 0)
743    
744  else  else
745    {    {
746    const uschar *oldptr;    const pcre_uchar *oldptr;
747    BOOL braced, negated;    BOOL braced, negated;
748    
749    switch (c)    switch (c)
# Line 642  else Line 753  else
753    
754      case CHAR_l:      case CHAR_l:
755      case CHAR_L:      case CHAR_L:
756        *errorcodeptr = ERR37;
757        break;
758    
759      case CHAR_u:      case CHAR_u:
760        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
761          {
762          /* In JavaScript, \u must be followed by four hexadecimal numbers.
763          Otherwise it is a lowercase u letter. */
764          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
765            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
766            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
767            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
768            {
769            c = 0;
770            for (i = 0; i < 4; ++i)
771              {
772              register int cc = *(++ptr);
773    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
774              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
775              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
776    #else           /* EBCDIC coding */
777              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
778              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
779    #endif
780              }
781            }
782          }
783        else
784          *errorcodeptr = ERR37;
785        break;
786    
787      case CHAR_U:      case CHAR_U:
788      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
789        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
790      break;      break;
791    
792      /* \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
793        class, \g must be followed by one of a number of specific things:
794    
795      (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
796      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 807  else
807      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
808    
809      case CHAR_g:      case CHAR_g:
810        if (isclass) break;
811      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
812        {        {
813        c = -ESC_g;        c = -ESC_g;
# Line 674  else Line 818  else
818    
819      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
820        {        {
821        const uschar *p;        const pcre_uchar *p;
822        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
823          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
824        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
825          {          {
826          c = -ESC_k;          c = -ESC_k;
# Line 694  else Line 838  else
838        }        }
839      else negated = FALSE;      else negated = FALSE;
840    
841        /* The integer range is limited by the machine's int representation. */
842      c = 0;      c = 0;
843      while ((digitab[ptr[1]] & ctype_digit) != 0)      while (IS_DIGIT(ptr[1]))
844          {
845          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
846            {
847            c = -1;
848            break;
849            }
850        c = c * 10 + *(++ptr) - CHAR_0;        c = c * 10 + *(++ptr) - CHAR_0;
851          }
852      if (c < 0)   /* Integer overflow */      if (((unsigned int)c) > INT_MAX) /* Integer overflow */
853        {        {
854          while (IS_DIGIT(ptr[1]))
855            ptr++;
856        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
857        break;        break;
858        }        }
# Line 747  else Line 900  else
900      if (!isclass)      if (!isclass)
901        {        {
902        oldptr = ptr;        oldptr = ptr;
903          /* The integer range is limited by the machine's int representation. */
904        c -= CHAR_0;        c -= CHAR_0;
905        while ((digitab[ptr[1]] & ctype_digit) != 0)        while (IS_DIGIT(ptr[1]))
906            {
907            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
908              {
909              c = -1;
910              break;
911              }
912          c = c * 10 + *(++ptr) - CHAR_0;          c = c * 10 + *(++ptr) - CHAR_0;
913        if (c < 0)    /* Integer overflow */          }
914          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
915          {          {
916            while (IS_DIGIT(ptr[1]))
917              ptr++;
918          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
919          break;          break;
920          }          }
# Line 784  else Line 947  else
947      c -= CHAR_0;      c -= CHAR_0;
948      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
949          c = c * 8 + *(++ptr) - CHAR_0;          c = c * 8 + *(++ptr) - CHAR_0;
950      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf && c > 0xff) *errorcodeptr = ERR51;
951      break;      break;
952    
953      /* \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
954      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.
955      treated as a data character. */      If not, { is treated as a data character. */
956    
957      case CHAR_x:      case CHAR_x:
958        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
959          {
960          /* In JavaScript, \x must be followed by two hexadecimal numbers.
961          Otherwise it is a lowercase x letter. */
962          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
963            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
964            {
965            c = 0;
966            for (i = 0; i < 2; ++i)
967              {
968              register int cc = *(++ptr);
969    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
970              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
971              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
972    #else           /* EBCDIC coding */
973              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
974              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
975    #endif
976              }
977            }
978          break;
979          }
980    
981      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
982        {        {
983        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
       int count = 0;  
984    
985        c = 0;        c = 0;
986        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
987          {          {
988          register int cc = *pt++;          register int cc = *pt++;
989          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
         count++;  
990    
991  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
992          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
# Line 811  else Line 995  else
995          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 */
996          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
997  #endif  #endif
998    
999    #ifdef COMPILE_PCRE8
1000            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1001    #else
1002    #ifdef COMPILE_PCRE16
1003            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1004    #endif
1005    #endif
1006            }
1007    
1008          if (c < 0)
1009            {
1010            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1011            *errorcodeptr = ERR34;
1012          }          }
1013    
1014        if (*pt == CHAR_RIGHT_CURLY_BRACKET)        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1015          {          {
1016          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR71;
1017          ptr = pt;          ptr = pt;
1018          break;          break;
1019          }          }
# Line 827  else Line 1025  else
1025      /* Read just a single-byte hex-defined char */      /* Read just a single-byte hex-defined char */
1026    
1027      c = 0;      c = 0;
1028      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1029        {        {
1030        int cc;                                  /* Some compilers don't like */        int cc;                                  /* Some compilers don't like */
1031        cc = *(++ptr);                           /* ++ in initializers */        cc = *(++ptr);                           /* ++ in initializers */
# Line 885  else Line 1083  else
1083    }    }
1084    
1085  /* 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
1086  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
1087    quantification such as \N{2,3}. */
1088    
1089  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1090         !is_counted_repeat(ptr+2))
1091    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1092    
1093  /* 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 1123  Returns:         type value from ucp_typ
1123  */  */
1124    
1125  static int  static int
1126  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1127  {  {
1128  int c, i, bot, top;  int c, i, bot, top;
1129  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1130  char name[32];  pcre_uchar name[32];
1131    
1132  c = *(++ptr);  c = *(++ptr);
1133  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
# Line 968  else Line 1168  else
1168  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1169    
1170  bot = 0;  bot = 0;
1171  top = _pcre_utt_size;  top = PRIV(utt_size);
1172    
1173  while (bot < top)  while (bot < top)
1174    {    {
1175    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1176    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1177    if (c == 0)    if (c == 0)
1178      {      {
1179      *dptr = _pcre_utt[i].value;      *dptr = PRIV(utt)[i].value;
1180      return _pcre_utt[i].type;      return PRIV(utt)[i].type;
1181      }      }
1182    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1183    }    }
# Line 997  return -1; Line 1197  return -1;
1197    
1198    
1199  /*************************************************  /*************************************************
 *            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);  
 }  
   
   
   
 /*************************************************  
1200  *         Read repeat counts                     *  *         Read repeat counts                     *
1201  *************************************************/  *************************************************/
1202    
# Line 1048  Returns:         pointer to '}' on succe Line 1215  Returns:         pointer to '}' on succe
1215                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1216  */  */
1217    
1218  static const uschar *  static const pcre_uchar *
1219  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)
1220  {  {
1221  int min = 0;  int min = 0;
1222  int max = -1;  int max = -1;
# Line 1057  int max = -1; Line 1224  int max = -1;
1224  /* 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
1225  an integer overflow. */  an integer overflow. */
1226    
1227  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1228  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1229    {    {
1230    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 1072  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1239  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1239    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1240      {      {
1241      max = 0;      max = 0;
1242      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1243      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1244        {        {
1245        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 1127  Arguments: Line 1294  Arguments:
1294    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1295    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1296    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1297    utf8         TRUE if we are in UTF-8 mode    utf          TRUE if we are in UTF-8 / UTF-16 mode
1298    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1299    
1300  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1301  */  */
1302    
1303  static int  static int
1304  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,
1305    BOOL xmode, BOOL utf8, int *count)    BOOL xmode, BOOL utf, int *count)
1306  {  {
1307  uschar *ptr = *ptrptr;  pcre_uchar *ptr = *ptrptr;
1308  int start_count = *count;  int start_count = *count;
1309  int hwm_count = start_count;  int hwm_count = start_count;
1310  BOOL dup_parens = FALSE;  BOOL dup_parens = FALSE;
# Line 1204  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1371  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1371          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1372        {        {
1373        int term;        int term;
1374        const uschar *thisname;        const pcre_uchar *thisname;
1375        *count += 1;        *count += 1;
1376        if (name == NULL && *count == lorn) return *count;        if (name == NULL && *count == lorn) return *count;
1377        term = *ptr++;        term = *ptr++;
# Line 1212  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1379  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1379        thisname = ptr;        thisname = ptr;
1380        while (*ptr != term) ptr++;        while (*ptr != term) ptr++;
1381        if (name != NULL && lorn == ptr - thisname &&        if (name != NULL && lorn == ptr - thisname &&
1382            strncmp((const char *)name, (const char *)thisname, lorn) == 0)            STRNCMP_UC_UC(name, thisname, lorn) == 0)
1383          return *count;          return *count;
1384        term++;        term++;
1385        }        }
# Line 1255  for (; ptr < cd->end_pattern; ptr++) Line 1422  for (; ptr < cd->end_pattern; ptr++)
1422          {          {
1423          if (ptr[2] == CHAR_E)          if (ptr[2] == CHAR_E)
1424            ptr+= 2;            ptr+= 2;
1425          else if (strncmp((const char *)ptr+2,          else if (STRNCMP_UC_C8(ptr + 2,
1426                   STR_Q STR_BACKSLASH STR_E, 3) == 0)                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
1427            ptr += 4;            ptr += 4;
1428          else          else
# Line 1303  for (; ptr < cd->end_pattern; ptr++) Line 1470  for (; ptr < cd->end_pattern; ptr++)
1470        {        {
1471        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1472        ptr++;        ptr++;
1473  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1474        if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;        if (utf) FORWARDCHAR(ptr);
1475  #endif  #endif
1476        }        }
1477      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
# Line 1315  for (; ptr < cd->end_pattern; ptr++) Line 1482  for (; ptr < cd->end_pattern; ptr++)
1482    
1483    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1484      {      {
1485      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1486      if (rc > 0) return rc;      if (rc > 0) return rc;
1487      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1488      }      }
# Line 1361  Arguments: Line 1528  Arguments:
1528    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1529    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1530    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1531    utf8         TRUE if we are in UTF-8 mode    utf          TRUE if we are in UTF-8 / UTF-16 mode
1532    
1533  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1534  */  */
1535    
1536  static int  static int
1537  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,
1538    BOOL utf8)    BOOL utf)
1539  {  {
1540  uschar *ptr = (uschar *)cd->start_pattern;  pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1541  int count = 0;  int count = 0;
1542  int rc;  int rc;
1543    
# Line 1381  matching closing parens. That is why we Line 1548  matching closing parens. That is why we
1548    
1549  for (;;)  for (;;)
1550    {    {
1551    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1552    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1553    }    }
1554    
# Line 1408  Arguments: Line 1575  Arguments:
1575  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1576  */  */
1577    
1578  static const uschar*  static const pcre_uchar*
1579  first_significant_code(const uschar *code, BOOL skipassert)  first_significant_code(const pcre_uchar *code, BOOL skipassert)
1580  {  {
1581  for (;;)  for (;;)
1582    {    {
# Line 1420  for (;;) Line 1587  for (;;)
1587      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1588      if (!skipassert) return code;      if (!skipassert) return code;
1589      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1590      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1591      break;      break;
1592    
1593      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 1434  for (;;) Line 1601  for (;;)
1601      case OP_RREF:      case OP_RREF:
1602      case OP_NRREF:      case OP_NRREF:
1603      case OP_DEF:      case OP_DEF:
1604      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1605      break;      break;
1606    
1607      default:      default:
# Line 1464  and doing the check at the end; a flag s Line 1631  and doing the check at the end; a flag s
1631    
1632  Arguments:  Arguments:
1633    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1634    utf8     TRUE in UTF-8 mode    utf      TRUE in UTF-8 / UTF-16 mode
1635    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1636    cd       the "compile data" structure    cd       the "compile data" structure
1637    
1638  Returns:   the fixed length,  Returns:   the fixed length,
1639               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1640               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1641               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
1642                 or -4 if an unknown opcode was encountered (internal error)
1643  */  */
1644    
1645  static int  static int
1646  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1647  {  {
1648  int length = -1;  int length = -1;
1649    
1650  register int branchlength = 0;  register int branchlength = 0;
1651  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1652    
1653  /* 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
1654  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1488  branch, check the length against that of Line 1656  branch, check the length against that of
1656  for (;;)  for (;;)
1657    {    {
1658    int d;    int d;
1659    uschar *ce, *cs;    pcre_uchar *ce, *cs;
1660    register int op = *cc;    register int op = *cc;
1661    switch (op)    switch (op)
1662      {      {
1663      /* We only need to continue for OP_CBRA (normal capturing bracket) and      /* We only need to continue for OP_CBRA (normal capturing bracket) and
1664      OP_BRA (normal non-capturing bracket) because the other variants of these      OP_BRA (normal non-capturing bracket) because the other variants of these
1665      opcodes are all concerned with unlimited repeated groups, which of course      opcodes are all concerned with unlimited repeated groups, which of course
1666      are not of fixed length. They will cause a -1 response from the default      are not of fixed length. */
     case of this switch. */  
1667    
1668      case OP_CBRA:      case OP_CBRA:
1669      case OP_BRA:      case OP_BRA:
1670      case OP_ONCE:      case OP_ONCE:
1671        case OP_ONCE_NC:
1672      case OP_COND:      case OP_COND:
1673      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1674      if (d < 0) return d;      if (d < 0) return d;
1675      branchlength += d;      branchlength += d;
1676      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1677      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1678      break;      break;
1679    
1680      /* 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.
1681      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
1682      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
1683      Note that we must not include the OP_KETRxxx opcodes here, because they      the same code. Note that we must not include the OP_KETRxxx opcodes here,
1684      all imply an unlimited repeat. */      because they all imply an unlimited repeat. */
1685    
1686      case OP_ALT:      case OP_ALT:
1687      case OP_KET:      case OP_KET:
1688      case OP_END:      case OP_END:
1689        case OP_ACCEPT:
1690        case OP_ASSERT_ACCEPT:
1691      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1692        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1693      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1531  for (;;) Line 1701  for (;;)
1701    
1702      case OP_RECURSE:      case OP_RECURSE:
1703      if (!atend) return -3;      if (!atend) return -3;
1704      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1705      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1706      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                    /* Recursion */
1707      d = find_fixedlength(cs + 2, utf8, atend, cd);      d = find_fixedlength(cs + 2, utf, atend, cd);
1708      if (d < 0) return d;      if (d < 0) return d;
1709      branchlength += d;      branchlength += d;
1710      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1551  for (;;) Line 1721  for (;;)
1721    
1722      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1723    
1724      case OP_REVERSE:      case OP_MARK:
1725      case OP_CREF:      case OP_PRUNE_ARG:
1726      case OP_NCREF:      case OP_SKIP_ARG:
1727      case OP_RREF:      case OP_THEN_ARG:
1728      case OP_NRREF:      cc += cc[1] + PRIV(OP_lengths)[*cc];
1729      case OP_DEF:      break;
1730    
1731      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1732      case OP_CIRC:      case OP_CIRC:
1733      case OP_CIRCM:      case OP_CIRCM:
1734        case OP_CLOSE:
1735        case OP_COMMIT:
1736        case OP_CREF:
1737        case OP_DEF:
1738      case OP_DOLL:      case OP_DOLL:
1739      case OP_DOLLM:      case OP_DOLLM:
1740        case OP_EOD:
1741        case OP_EODN:
1742        case OP_FAIL:
1743        case OP_NCREF:
1744        case OP_NRREF:
1745      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1746        case OP_PRUNE:
1747        case OP_REVERSE:
1748        case OP_RREF:
1749        case OP_SET_SOM:
1750        case OP_SKIP:
1751        case OP_SOD:
1752        case OP_SOM:
1753        case OP_THEN:
1754      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1755      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1756      break;      break;
1757    
1758      /* Handle literal characters */      /* Handle literal characters */
# Line 1580  for (;;) Line 1763  for (;;)
1763      case OP_NOTI:      case OP_NOTI:
1764      branchlength++;      branchlength++;
1765      cc += 2;      cc += 2;
1766  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1767      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1768  #endif  #endif
1769      break;      break;
1770    
# Line 1589  for (;;) Line 1772  for (;;)
1772      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1773    
1774      case OP_EXACT:      case OP_EXACT:
1775        case OP_EXACTI:
1776        case OP_NOTEXACT:
1777        case OP_NOTEXACTI:
1778      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1779      cc += 4;      cc += 2 + IMM2_SIZE;
1780  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1781      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1782  #endif  #endif
1783      break;      break;
1784    
1785      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1786      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1787      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;
1788      cc += 4;      cc += 1 + IMM2_SIZE + 1;
1789      break;      break;
1790    
1791      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1609  for (;;) Line 1795  for (;;)
1795      cc += 2;      cc += 2;
1796      /* Fall through */      /* Fall through */
1797    
1798        case OP_HSPACE:
1799        case OP_VSPACE:
1800        case OP_NOT_HSPACE:
1801        case OP_NOT_VSPACE:
1802      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1803      case OP_DIGIT:      case OP_DIGIT:
1804      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1621  for (;;) Line 1811  for (;;)
1811      cc++;      cc++;
1812      break;      break;
1813    
1814      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1815        otherwise \C is coded as OP_ALLANY. */
1816    
1817      case OP_ANYBYTE:      case OP_ANYBYTE:
1818      return -2;      return -2;
1819    
1820      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1821    
1822  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16
1823      case OP_XCLASS:      case OP_XCLASS:
1824      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1825      /* Fall through */      /* Fall through */
1826  #endif  #endif
1827    
1828      case OP_CLASS:      case OP_CLASS:
1829      case OP_NCLASS:      case OP_NCLASS:
1830      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1831    
1832      switch (*cc)      switch (*cc)
1833        {        {
1834          case OP_CRPLUS:
1835          case OP_CRMINPLUS:
1836        case OP_CRSTAR:        case OP_CRSTAR:
1837        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1838        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1648  for (;;) Line 1841  for (;;)
1841    
1842        case OP_CRRANGE:        case OP_CRRANGE:
1843        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1844        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1845        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1846        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1847        break;        break;
1848    
1849        default:        default:
# Line 1660  for (;;) Line 1853  for (;;)
1853    
1854      /* Anything else is variable length */      /* Anything else is variable length */
1855    
1856      default:      case OP_ANYNL:
1857        case OP_BRAMINZERO:
1858        case OP_BRAPOS:
1859        case OP_BRAPOSZERO:
1860        case OP_BRAZERO:
1861        case OP_CBRAPOS:
1862        case OP_EXTUNI:
1863        case OP_KETRMAX:
1864        case OP_KETRMIN:
1865        case OP_KETRPOS:
1866        case OP_MINPLUS:
1867        case OP_MINPLUSI:
1868        case OP_MINQUERY:
1869        case OP_MINQUERYI:
1870        case OP_MINSTAR:
1871        case OP_MINSTARI:
1872        case OP_MINUPTO:
1873        case OP_MINUPTOI:
1874        case OP_NOTMINPLUS:
1875        case OP_NOTMINPLUSI:
1876        case OP_NOTMINQUERY:
1877        case OP_NOTMINQUERYI:
1878        case OP_NOTMINSTAR:
1879        case OP_NOTMINSTARI:
1880        case OP_NOTMINUPTO:
1881        case OP_NOTMINUPTOI:
1882        case OP_NOTPLUS:
1883        case OP_NOTPLUSI:
1884        case OP_NOTPOSPLUS:
1885        case OP_NOTPOSPLUSI:
1886        case OP_NOTPOSQUERY:
1887        case OP_NOTPOSQUERYI:
1888        case OP_NOTPOSSTAR:
1889        case OP_NOTPOSSTARI:
1890        case OP_NOTPOSUPTO:
1891        case OP_NOTPOSUPTOI:
1892        case OP_NOTQUERY:
1893        case OP_NOTQUERYI:
1894        case OP_NOTSTAR:
1895        case OP_NOTSTARI:
1896        case OP_NOTUPTO:
1897        case OP_NOTUPTOI:
1898        case OP_PLUS:
1899        case OP_PLUSI:
1900        case OP_POSPLUS:
1901        case OP_POSPLUSI:
1902        case OP_POSQUERY:
1903        case OP_POSQUERYI:
1904        case OP_POSSTAR:
1905        case OP_POSSTARI:
1906        case OP_POSUPTO:
1907        case OP_POSUPTOI:
1908        case OP_QUERY:
1909        case OP_QUERYI:
1910        case OP_REF:
1911        case OP_REFI:
1912        case OP_SBRA:
1913        case OP_SBRAPOS:
1914        case OP_SCBRA:
1915        case OP_SCBRAPOS:
1916        case OP_SCOND:
1917        case OP_SKIPZERO:
1918        case OP_STAR:
1919        case OP_STARI:
1920        case OP_TYPEMINPLUS:
1921        case OP_TYPEMINQUERY:
1922        case OP_TYPEMINSTAR:
1923        case OP_TYPEMINUPTO:
1924        case OP_TYPEPLUS:
1925        case OP_TYPEPOSPLUS:
1926        case OP_TYPEPOSQUERY:
1927        case OP_TYPEPOSSTAR:
1928        case OP_TYPEPOSUPTO:
1929        case OP_TYPEQUERY:
1930        case OP_TYPESTAR:
1931        case OP_TYPEUPTO:
1932        case OP_UPTO:
1933        case OP_UPTOI:
1934      return -1;      return -1;
1935    
1936        /* Catch unrecognized opcodes so that when new ones are added they
1937        are not forgotten, as has happened in the past. */
1938    
1939        default:
1940        return -4;
1941      }      }
1942    }    }
1943  /* Control never gets here */  /* Control never gets here */
# Line 1682  length. Line 1958  length.
1958    
1959  Arguments:  Arguments:
1960    code        points to start of expression    code        points to start of expression
1961    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
1962    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
1963    
1964  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
1965  */  */
1966    
1967  const uschar *  const pcre_uchar *
1968  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
1969  {  {
1970  for (;;)  for (;;)
1971    {    {
1972    register int c = *code;    register int c = *code;
1973    
1974    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1975    
1976    /* 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 1706  for (;;) Line 1983  for (;;)
1983    
1984    else if (c == OP_REVERSE)    else if (c == OP_REVERSE)
1985      {      {
1986      if (number < 0) return (uschar *)code;      if (number < 0) return (pcre_uchar *)code;
1987      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
1988      }      }
1989    
1990    /* Handle capturing bracket */    /* Handle capturing bracket */
# Line 1716  for (;;) Line 1993  for (;;)
1993             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
1994      {      {
1995      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1996      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
1997      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
1998      }      }
1999    
2000    /* 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 1745  for (;;) Line 2022  for (;;)
2022        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2023        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2024        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2025        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2026            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2027        break;        break;
2028    
2029        case OP_MARK:        case OP_MARK:
# Line 1755  for (;;) Line 2033  for (;;)
2033        break;        break;
2034    
2035        case OP_THEN_ARG:        case OP_THEN_ARG:
2036        code += code[1+LINK_SIZE];        code += code[1];
2037        break;        break;
2038        }        }
2039    
2040      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2041    
2042      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2043    
2044    /* 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
2045    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
2046    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2047    
2048  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2049      if (utf8) switch(c)      if (utf) switch(c)
2050        {        {
2051        case OP_CHAR:        case OP_CHAR:
2052        case OP_CHARI:        case OP_CHARI:
# Line 1798  for (;;) Line 2076  for (;;)
2076        case OP_MINQUERYI:        case OP_MINQUERYI:
2077        case OP_POSQUERY:        case OP_POSQUERY:
2078        case OP_POSQUERYI:        case OP_POSQUERYI:
2079        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2080        break;        break;
2081        }        }
2082  #else  #else
2083      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2084  #endif  #endif
2085      }      }
2086    }    }
# Line 1819  instance of OP_RECURSE. Line 2097  instance of OP_RECURSE.
2097    
2098  Arguments:  Arguments:
2099    code        points to start of expression    code        points to start of expression
2100    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2101    
2102  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
2103  */  */
2104    
2105  static const uschar *  static const pcre_uchar *
2106  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2107  {  {
2108  for (;;)  for (;;)
2109    {    {
# Line 1864  for (;;) Line 2142  for (;;)
2142        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2143        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2144        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2145        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[1 + IMM2_SIZE] == OP_PROP
2146            || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2147        break;        break;
2148    
2149        case OP_MARK:        case OP_MARK:
# Line 1874  for (;;) Line 2153  for (;;)
2153        break;        break;
2154    
2155        case OP_THEN_ARG:        case OP_THEN_ARG:
2156        code += code[1+LINK_SIZE];        code += code[1];
2157        break;        break;
2158        }        }
2159    
2160      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2161    
2162      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2163    
2164      /* 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
2165      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
2166      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2167    
2168  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2169      if (utf8) switch(c)      if (utf) switch(c)
2170        {        {
2171        case OP_CHAR:        case OP_CHAR:
2172        case OP_CHARI:        case OP_CHARI:
# Line 1917  for (;;) Line 2196  for (;;)
2196        case OP_MINQUERYI:        case OP_MINQUERYI:
2197        case OP_POSQUERY:        case OP_POSQUERY:
2198        case OP_POSQUERYI:        case OP_POSQUERYI:
2199        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2200        break;        break;
2201        }        }
2202  #else  #else
2203      (void)(utf8);  /* Keep compiler happy by referencing function argument */      (void)(utf);  /* Keep compiler happy by referencing function argument */
2204  #endif  #endif
2205      }      }
2206    }    }
# Line 1944  bracket whose current branch will alread Line 2223  bracket whose current branch will alread
2223  Arguments:  Arguments:
2224    code        points to start of search    code        points to start of search
2225    endcode     points to where to stop    endcode     points to where to stop
2226    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2227    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2228    
2229  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2230  */  */
2231    
2232  static BOOL  static BOOL
2233  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2234    compile_data *cd)    BOOL utf, compile_data *cd)
2235  {  {
2236  register int c;  register int c;
2237  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2238       code < endcode;       code < endcode;
2239       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2240    {    {
2241    const uschar *ccode;    const pcre_uchar *ccode;
2242    
2243    c = *code;    c = *code;
2244    
# Line 1974  for (code = first_significant_code(code Line 2253  for (code = first_significant_code(code
2253      }      }
2254    
2255    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2256    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2257      forward reference subroutine call, we can't. To detect forward reference
2258      we have to scan up the list that is kept in the workspace. This function is
2259      called only when doing the real compile, not during the pre-compile that
2260      measures the size of the compiled pattern. */
2261    
2262    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2263      {      {
2264      BOOL empty_branch = FALSE;      const pcre_uchar *scode;
2265      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
2266    
2267        /* Test for forward reference */
2268    
2269        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2270          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2271    
2272        /* Not a forward reference, test for completed backward reference */
2273    
2274        empty_branch = FALSE;
2275        scode = cd->start_code + GET(code, 1);
2276      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2277    
2278        /* Completed backwards reference */
2279    
2280      do      do
2281        {        {
2282        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf, cd))
2283          {          {
2284          empty_branch = TRUE;          empty_branch = TRUE;
2285          break;          break;
# Line 1991  for (code = first_significant_code(code Line 2287  for (code = first_significant_code(code
2287        scode += GET(scode, 1);        scode += GET(scode, 1);
2288        }        }
2289      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2290    
2291      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2292      continue;      continue;
2293      }      }
# Line 2000  for (code = first_significant_code(code Line 2297  for (code = first_significant_code(code
2297    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2298        c == OP_BRAPOSZERO)        c == OP_BRAPOSZERO)
2299      {      {
2300      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2301      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2302      c = *code;      c = *code;
2303      continue;      continue;
# Line 2021  for (code = first_significant_code(code Line 2318  for (code = first_significant_code(code
2318    
2319    if (c == OP_BRA  || c == OP_BRAPOS ||    if (c == OP_BRA  || c == OP_BRAPOS ||
2320        c == OP_CBRA || c == OP_CBRAPOS ||        c == OP_CBRA || c == OP_CBRAPOS ||
2321        c == OP_ONCE || c == OP_COND)        c == OP_ONCE || c == OP_ONCE_NC ||
2322          c == OP_COND)
2323      {      {
2324      BOOL empty_branch;      BOOL empty_branch;
2325      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2037  for (code = first_significant_code(code Line 2335  for (code = first_significant_code(code
2335        empty_branch = FALSE;        empty_branch = FALSE;
2336        do        do
2337          {          {
2338          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2339            empty_branch = TRUE;            empty_branch = TRUE;
2340          code += GET(code, 1);          code += GET(code, 1);
2341          }          }
# Line 2055  for (code = first_significant_code(code Line 2353  for (code = first_significant_code(code
2353      {      {
2354      /* Check for quantifiers after a class. XCLASS is used for classes that      /* Check for quantifiers after a class. XCLASS is used for classes that
2355      cannot be represented just by a bit map. This includes negated single      cannot be represented just by a bit map. This includes negated single
2356      high-valued characters. The length in _pcre_OP_lengths[] is zero; the      high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2357      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"
2358      here. */      here. */
2359    
# Line 2067  for (code = first_significant_code(code Line 2365  for (code = first_significant_code(code
2365    
2366      case OP_CLASS:      case OP_CLASS:
2367      case OP_NCLASS:      case OP_NCLASS:
2368      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2369    
2370  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2371      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
# Line 2142  for (code = first_significant_code(code Line 2440  for (code = first_significant_code(code
2440      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2441      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2442      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2443      if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;      if (code[1 + IMM2_SIZE] == OP_PROP
2444          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;
2445      break;      break;
2446    
2447      /* End of branch */      /* End of branch */
# Line 2157  for (code = first_significant_code(code Line 2456  for (code = first_significant_code(code
2456      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2457      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2458    
2459  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2460      case OP_STAR:      case OP_STAR:
2461      case OP_STARI:      case OP_STARI:
2462      case OP_MINSTAR:      case OP_MINSTAR:
# Line 2170  for (code = first_significant_code(code Line 2469  for (code = first_significant_code(code
2469      case OP_MINQUERYI:      case OP_MINQUERYI:
2470      case OP_POSQUERY:      case OP_POSQUERY:
2471      case OP_POSQUERYI:      case OP_POSQUERYI:
2472      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2473      break;      break;
2474    
2475      case OP_UPTO:      case OP_UPTO:
# Line 2179  for (code = first_significant_code(code Line 2478  for (code = first_significant_code(code
2478      case OP_MINUPTOI:      case OP_MINUPTOI:
2479      case OP_POSUPTO:      case OP_POSUPTO:
2480      case OP_POSUPTOI:      case OP_POSUPTOI:
2481      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2482      break;      break;
2483  #endif  #endif
2484    
# Line 2193  for (code = first_significant_code(code Line 2492  for (code = first_significant_code(code
2492      break;      break;
2493    
2494      case OP_THEN_ARG:      case OP_THEN_ARG:
2495      code += code[1+LINK_SIZE];      code += code[1];
2496      break;      break;
2497    
2498      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
# Line 2216  return TRUE; Line 2515  return TRUE;
2515  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
2516  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,
2517  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.
2518    This function is called only during the real compile, not during the
2519    pre-compile.
2520    
2521  Arguments:  Arguments:
2522    code        points to start of the recursion    code        points to start of the recursion
2523    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2524    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2525    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2526    cd          pointers to tables etc    cd          pointers to tables etc
2527    
2528  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2529  */  */
2530    
2531  static BOOL  static BOOL
2532  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2533    BOOL utf8, compile_data *cd)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2534  {  {
2535  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2536    {    {
2537    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2538      return FALSE;      return FALSE;
2539    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2540    }    }
# Line 2266  where Perl recognizes it as the POSIX cl Line 2567  where Perl recognizes it as the POSIX cl
2567  "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,
2568  I think.  I think.
2569    
2570    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2571    It seems that the appearance of a nested POSIX class supersedes an apparent
2572    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2573    a digit.
2574    
2575    In Perl, unescaped square brackets may also appear as part of class names. For
2576    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2577    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2578    seem right at all. PCRE does not allow closing square brackets in POSIX class
2579    names.
2580    
2581  Arguments:  Arguments:
2582    ptr      pointer to the initial [    ptr      pointer to the initial [
2583    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2274  Returns:   TRUE or FALSE Line 2586  Returns:   TRUE or FALSE
2586  */  */
2587    
2588  static BOOL  static BOOL
2589  check_posix_syntax(const uschar *ptr, const uschar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2590  {  {
2591  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2592  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2593  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2594    {    {
2595    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2596        ptr++;
2597      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2598      else
2599      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2600      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2601        {        {
2602        *endptr = ptr;        *endptr = ptr;
2603        return TRUE;        return TRUE;
2604        }        }
2605        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2606             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2607              ptr[1] == CHAR_EQUALS_SIGN) &&
2608            check_posix_syntax(ptr, endptr))
2609          return FALSE;
2610      }      }
2611    }    }
2612  return FALSE;  return FALSE;
# Line 2311  Returns:     a value representing the na Line 2630  Returns:     a value representing the na
2630  */  */
2631    
2632  static int  static int
2633  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2634  {  {
2635  const char *pn = posix_names;  const char *pn = posix_names;
2636  register int yield = 0;  register int yield = 0;
2637  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2638    {    {
2639    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2640      strncmp((const char *)ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2641    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
2642    yield++;    yield++;
2643    }    }
# Line 2350  value in the reference (which is a group Line 2669  value in the reference (which is a group
2669  Arguments:  Arguments:
2670    group      points to the start of the group    group      points to the start of the group
2671    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2672    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 mode
2673    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2674    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
2675    
# Line 2358  Returns:     nothing Line 2677  Returns:     nothing
2677  */  */
2678    
2679  static void  static void
2680  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2681    uschar *save_hwm)    pcre_uchar *save_hwm)
2682  {  {
2683  uschar *ptr = group;  pcre_uchar *ptr = group;
2684    
2685  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2686    {    {
2687    int offset;    int offset;
2688    uschar *hc;    pcre_uchar *hc;
2689    
2690    /* 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
2691    reference. */    reference. */
# Line 2411  Arguments: Line 2730  Arguments:
2730  Returns:         new code pointer  Returns:         new code pointer
2731  */  */
2732    
2733  static uschar *  static pcre_uchar *
2734  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2735  {  {
2736  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2737  *code++ = 255;  *code++ = 255;
2738  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2739  PUT(code, LINK_SIZE, 0);                       /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2740  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2741  }  }
2742    
2743    
# Line 2440  Returns:             nothing Line 2759  Returns:             nothing
2759  */  */
2760    
2761  static void  static void
2762  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2763  {  {
2764  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2765  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
# Line 2523  switch(ptype) Line 2842  switch(ptype)
2842            prop->chartype == ucp_Lt) == negated;            prop->chartype == ucp_Lt) == negated;
2843    
2844    case PT_GC:    case PT_GC:
2845    return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2846    
2847    case PT_PC:    case PT_PC:
2848    return (pdata == prop->chartype) == negated;    return (pdata == prop->chartype) == negated;
# Line 2534  switch(ptype) Line 2853  switch(ptype)
2853    /* These are specials */    /* These are specials */
2854    
2855    case PT_ALNUM:    case PT_ALNUM:
2856    return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2857            _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2858    
2859    case PT_SPACE:    /* Perl space */    case PT_SPACE:    /* Perl space */
2860    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2861            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2862            == negated;            == negated;
2863    
2864    case PT_PXSPACE:  /* POSIX space */    case PT_PXSPACE:  /* POSIX space */
2865    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2866            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2867            c == CHAR_FF || c == CHAR_CR)            c == CHAR_FF || c == CHAR_CR)
2868            == negated;            == negated;
2869    
2870    case PT_WORD:    case PT_WORD:
2871    return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2872            _pcre_ucp_gentype[prop->chartype] == ucp_N ||            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2873            c == CHAR_UNDERSCORE) == negated;            c == CHAR_UNDERSCORE) == negated;
2874    }    }
2875  return FALSE;  return FALSE;
# Line 2569  sense to automatically possessify the re Line 2888  sense to automatically possessify the re
2888    
2889  Arguments:  Arguments:
2890    previous      pointer to the repeated opcode    previous      pointer to the repeated opcode
2891    utf8          TRUE in UTF-8 mode    utf           TRUE in UTF-8 / UTF-16 mode
2892    ptr           next character in pattern    ptr           next character in pattern
2893    options       options bits    options       options bits
2894    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2578  Returns:        TRUE if possessifying is Line 2897  Returns:        TRUE if possessifying is
2897  */  */
2898    
2899  static BOOL  static BOOL
2900  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
2901    int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2902  {  {
2903  int c, next;  pcre_int32 c, next;
2904  int op_code = *previous++;  int op_code = *previous++;
2905    
2906  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2598  if ((options & PCRE_EXTENDED) != 0) Line 2917  if ((options & PCRE_EXTENDED) != 0)
2917          {          {
2918          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2919          ptr++;          ptr++;
2920  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2921          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf) FORWARDCHAR(ptr);
2922  #endif  #endif
2923          }          }
2924        }        }
# Line 2617  if (*ptr == CHAR_BACKSLASH) Line 2936  if (*ptr == CHAR_BACKSLASH)
2936    if (temperrorcode != 0) return FALSE;    if (temperrorcode != 0) return FALSE;
2937    ptr++;    /* Point after the escape sequence */    ptr++;    /* Point after the escape sequence */
2938    }    }
2939    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
 else if ((cd->ctypes[*ptr] & ctype_meta) == 0)  
2940    {    {
2941  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2942    if (utf8) { GETCHARINC(next, ptr); } else    if (utf) { GETCHARINC(next, ptr); } else
2943  #endif  #endif
2944    next = *ptr++;    next = *ptr++;
2945    }    }
   
2946  else return FALSE;  else return FALSE;
2947    
2948  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
# Line 2642  if ((options & PCRE_EXTENDED) != 0) Line 2959  if ((options & PCRE_EXTENDED) != 0)
2959          {          {
2960          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2961          ptr++;          ptr++;
2962  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2963          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf) FORWARDCHAR(ptr);
2964  #endif  #endif
2965          }          }
2966        }        }
# Line 2654  if ((options & PCRE_EXTENDED) != 0) Line 2971  if ((options & PCRE_EXTENDED) != 0)
2971  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2972    
2973  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2974    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)
2975      return FALSE;      return FALSE;
2976    
2977  /* 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 2682  if (next >= 0) switch(op_code) Line 2999  if (next >= 0) switch(op_code)
2999  #endif  #endif
3000    if (c == next) return FALSE;    if (c == next) return FALSE;
3001  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3002    if (utf8)    if (utf)
3003      {      {
3004      unsigned int othercase;      unsigned int othercase;
3005      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
# Line 2707  if (next >= 0) switch(op_code) Line 3024  if (next >= 0) switch(op_code)
3024    case OP_NOTI:    case OP_NOTI:
3025    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
3026  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3027    if (utf8)    if (utf)
3028      {      {
3029      unsigned int othercase;      unsigned int othercase;
3030      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
# Line 2916  switch(op_code) Line 3233  switch(op_code)
3233        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. */
3234    
3235        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3236          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)
3237            return FALSE;            return FALSE;
3238    
3239        /* Do the property check. */        /* Do the property check. */
# Line 2994  Arguments: Line 3311  Arguments:
3311    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3312    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3313    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3314    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstcharptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3315    reqbyteptr     set to the last literal character required, else < 0    reqcharptr     set to the last literal character required, else < 0
3316    bcptr          points to current branch chain    bcptr          points to current branch chain
3317      cond_depth     conditional nesting depth
3318    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3319    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3320                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 3006  Returns:         TRUE on success Line 3324  Returns:         TRUE on success
3324  */  */
3325    
3326  static BOOL  static BOOL
3327  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, pcre_uchar **codeptr,
3328    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    const pcre_uchar **ptrptr, int *errorcodeptr, pcre_int32 *firstcharptr,
3329      pcre_int32 *reqcharptr, branch_chain *bcptr, int cond_depth,
3330    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
3331  {  {
3332  int repeat_type, op_type;  int repeat_type, op_type;
3333  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3334  int bravalue = 0;  int bravalue = 0;
3335  int greedy_default, greedy_non_default;  int greedy_default, greedy_non_default;
3336  int firstbyte, reqbyte;  pcre_int32 firstchar, reqchar;
3337  int zeroreqbyte, zerofirstbyte;  pcre_int32 zeroreqchar, zerofirstchar;
3338  int req_caseopt, reqvary, tempreqvary;  pcre_int32 req_caseopt, reqvary, tempreqvary;
3339  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3340  int after_manual_callout = 0;  int after_manual_callout = 0;
3341  int length_prevgroup = 0;  int length_prevgroup = 0;
3342  register int c;  register int c;
3343  register uschar *code = *codeptr;  register pcre_uchar *code = *codeptr;
3344  uschar *last_code = code;  pcre_uchar *last_code = code;
3345  uschar *orig_code = code;  pcre_uchar *orig_code = code;
3346  uschar *tempcode;  pcre_uchar *tempcode;
3347  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3348  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstchar = FALSE;
3349  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
3350  const uschar *tempptr;  const pcre_uchar *tempptr;
3351  const uschar *nestptr = NULL;  const pcre_uchar *nestptr = NULL;
3352  uschar *previous = NULL;  pcre_uchar *previous = NULL;
3353  uschar *previous_callout = NULL;  pcre_uchar *previous_callout = NULL;
3354  uschar *save_hwm = NULL;  pcre_uchar *save_hwm = NULL;
3355  uschar classbits[32];  pcre_uint8 classbits[32];
3356    
3357    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3358    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3359    dynamically as we process the pattern. */
3360    
3361  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3362  BOOL class_utf8;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
3363  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
3364  uschar *class_utf8data;  pcre_uchar utf_chars[6];
 uschar *class_utf8data_base;  
 uschar utf8_char[6];  
3365  #else  #else
3366  BOOL utf8 = FALSE;  BOOL utf = FALSE;
3367  uschar *utf8_char = NULL;  #endif
3368    
3369    /* Helper variables for OP_XCLASS opcode (for characters > 255). */
3370    
3371    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3372    BOOL xclass;
3373    pcre_uchar *class_uchardata;
3374    pcre_uchar *class_uchardata_base;
3375  #endif  #endif
3376    
3377  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 3057  greedy_non_default = greedy_default ^ 1; Line 3385  greedy_non_default = greedy_default ^ 1;
3385    
3386  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3387  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
3388  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
3389  find one.  find one.
3390    
3391  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
3392  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
3393  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
3394  item types that can be repeated set these backoff variables appropriately. */  item types that can be repeated set these backoff variables appropriately. */
3395    
3396  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  firstchar = reqchar = zerofirstchar = zeroreqchar = REQ_UNSET;
3397    
3398  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  /* The variable req_caseopt contains either the REQ_CASELESS value
3399  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
3400  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
3401  case status of the value. This is used only for ASCII characters. */  firstchar or reqchar variables to record the case status of the
3402    value. This is used only for ASCII characters. */
3403    
3404  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
3405    
3406  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
3407    
# Line 3090  for (;; ptr++) Line 3419  for (;; ptr++)
3419    int recno;    int recno;
3420    int refsign;    int refsign;
3421    int skipbytes;    int skipbytes;
3422    int subreqbyte;    int subreqchar;
3423    int subfirstbyte;    int subfirstchar;
3424    int terminator;    int terminator;
3425    int mclength;    int mclength;
3426    uschar mcbuffer[8];    int tempbracount;
3427      pcre_uchar mcbuffer[8];
3428    
3429    /* Get next byte in the pattern */    /* Get next character in the pattern */
3430    
3431    c = *ptr;    c = *ptr;
3432    
# Line 3141  for (;; ptr++) Line 3471  for (;; ptr++)
3471        }        }
3472    
3473      *lengthptr += (int)(code - last_code);      *lengthptr += (int)(code - last_code);
3474      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
3475          (int)(code - last_code), c, c));
3476    
3477      /* 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
3478      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 3151  for (;; ptr++) Line 3482  for (;; ptr++)
3482        {        {
3483        if (previous > orig_code)        if (previous > orig_code)
3484          {          {
3485          memmove(orig_code, previous, code - previous);          memmove(orig_code, previous, IN_UCHARS(code - previous));
3486          code -= previous - orig_code;          code -= previous - orig_code;
3487          previous = orig_code;          previous = orig_code;
3488          }          }
# Line 3215  for (;; ptr++) Line 3546  for (;; ptr++)
3546      previous_callout = NULL;      previous_callout = NULL;
3547      }      }
3548    
3549    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3550    
3551    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3552      {      {
# Line 3227  for (;; ptr++) Line 3558  for (;; ptr++)
3558          {          {
3559          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3560          ptr++;          ptr++;
3561  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3562          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf) FORWARDCHAR(ptr);
3563  #endif  #endif
3564          }          }
3565        if (*ptr != 0) continue;        if (*ptr != 0) continue;
# Line 3252  for (;; ptr++) Line 3583  for (;; ptr++)
3583      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3584      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3585      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
3586      *firstbyteptr = firstbyte;      *firstcharptr = firstchar;
3587      *reqbyteptr = reqbyte;      *reqcharptr = reqchar;
3588      *codeptr = code;      *codeptr = code;
3589      *ptrptr = ptr;      *ptrptr = ptr;
3590      if (lengthptr != NULL)      if (lengthptr != NULL)
# Line 3277  for (;; ptr++) Line 3608  for (;; ptr++)
3608      previous = NULL;      previous = NULL;
3609      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3610        {        {
3611        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3612        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
3613        }        }
3614      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 3289  for (;; ptr++) Line 3620  for (;; ptr++)
3620      break;      break;
3621    
3622      /* 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
3623      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqchar doesn't change either. */
3624    
3625      case CHAR_DOT:      case CHAR_DOT:
3626      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3627      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
3628      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
3629      previous = code;      previous = code;
3630      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3631      break;      break;
# Line 3349  for (;; ptr++) Line 3680  for (;; ptr++)
3680          {          {
3681          if (ptr[1] == CHAR_E)          if (ptr[1] == CHAR_E)
3682            ptr++;            ptr++;
3683          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)  
3684            ptr += 3;            ptr += 3;
3685          else          else
3686            break;            break;
# Line 3369  for (;; ptr++) Line 3699  for (;; ptr++)
3699          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3700        {        {
3701        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3702        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
3703        zerofirstbyte = firstbyte;        zerofirstchar = firstchar;
3704        break;        break;
3705        }        }
3706    
# Line 3392  for (;; ptr++) Line 3722  for (;; ptr++)
3722      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.
3723      */      */
3724    
3725      memset(classbits, 0, 32 * sizeof(uschar));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
3726    
3727  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3728      class_utf8 = FALSE;                       /* No chars >= 256 */      xclass = FALSE;                           /* No chars >= 256 */
3729      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_uchardata = code + LINK_SIZE + 2;   /* For UTF-8 items */
3730      class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */      class_uchardata_base = class_uchardata;   /* For resetting in pass 1 */
3731  #endif  #endif
3732    
3733      /* 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 3406  for (;; ptr++) Line 3736  for (;; ptr++)
3736    
3737      if (c != 0) do      if (c != 0) do
3738        {        {
3739        const uschar *oldptr;        const pcre_uchar *oldptr;
3740    
3741  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3742        if (utf8 && c > 127)        if (utf && HAS_EXTRALEN(c))
3743          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3744          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3745          }          }
3746    #endif
3747    
3748        /* In the pre-compile phase, accumulate the length of any UTF-8 extra  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3749          /* In the pre-compile phase, accumulate the length of any extra
3750        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
3751        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion > 255 characters no longer overwrite the work space
3752        (which is on the stack). */        (which is on the stack). */
3753    
3754        if (lengthptr != NULL)        if (lengthptr != NULL)
3755          {          {
3756          *lengthptr += class_utf8data - class_utf8data_base;          *lengthptr += class_uchardata - class_uchardata_base;
3757          class_utf8data = class_utf8data_base;          class_uchardata = class_uchardata_base;
3758          }          }
   
3759  #endif  #endif
3760    
3761        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 3452  for (;; ptr++) Line 3783  for (;; ptr++)
3783          {          {
3784          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3785          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3786          register const uschar *cbits = cd->cbits;          register const pcre_uint8 *cbits = cd->cbits;
3787          uschar pbits[32];          pcre_uint8 pbits[32];
3788    
3789          if (ptr[1] != CHAR_COLON)          if (ptr[1] != CHAR_COLON)
3790            {            {
# Line 3508  for (;; ptr++) Line 3839  for (;; ptr++)
3839          /* Copy in the first table (always present) */          /* Copy in the first table (always present) */
3840    
3841          memcpy(pbits, cbits + posix_class_maps[posix_class],          memcpy(pbits, cbits + posix_class_maps[posix_class],
3842            32 * sizeof(uschar));            32 * sizeof(pcre_uint8));
3843    
3844          /* If there is a second table, add or remove it as required. */          /* If there is a second table, add or remove it as required. */
3845    
# Line 3570  for (;; ptr++) Line 3901  for (;; ptr++)
3901    
3902          if (c < 0)          if (c < 0)
3903            {            {
3904            register const uschar *cbits = cd->cbits;            register const pcre_uint8 *cbits = cd->cbits;
3905            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3906    
3907            switch (-c)            switch (-c)
# Line 3625  for (;; ptr++) Line 3956  for (;; ptr++)
3956              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3957              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3958              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
3959  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3960              if (utf8)              if (utf)
3961                {                {
3962                class_utf8 = TRUE;                xclass = TRUE;
3963                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
3964                class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1680, class_uchardata);
3965                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
3966                class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180e, class_uchardata);
3967                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
3968                class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2000, class_uchardata);
3969                class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x200A, class_uchardata);
3970                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
3971                class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x202f, class_uchardata);
3972                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
3973                class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x205f, class_uchardata);
3974                *class_utf8data++ = XCL_SINGLE;                *class_uchardata++ = XCL_SINGLE;
3975                class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x3000, class_uchardata);
3976                }                }
3977  #endif  #endif
3978              continue;              continue;
# Line 3660  for (;; ptr++) Line 3991  for (;; ptr++)
3991                classbits[c] |= x;                classbits[c] |= x;
3992                }                }
3993    
3994  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3995              if (utf8)              if (utf)
3996                {                {
3997                class_utf8 = TRUE;                xclass = TRUE;
3998                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
3999                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);
4000                class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x167f, class_uchardata);
4001                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4002                class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1681, class_uchardata);
4003                class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180d, class_uchardata);
4004                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4005                class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x180f, class_uchardata);
4006                class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x1fff, class_uchardata);
4007                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4008                class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x200B, class_uchardata);
4009                class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x202e, class_uchardata);
4010                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4011                class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2030, class_uchardata);
4012                class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x205e, class_uchardata);
4013                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4014                class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2060, class_uchardata);
4015                class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2fff, class_uchardata);
4016                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4017                class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x3001, class_uchardata);
4018                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4019                }                }
4020  #endif  #endif
4021              continue;              continue;
# Line 3695  for (;; ptr++) Line 4026  for (;; ptr++)
4026              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
4027              SETBIT(classbits, 0x0d); /* CR */              SETBIT(classbits, 0x0d); /* CR */
4028              SETBIT(classbits, 0x85); /* NEL */              SETBIT(classbits, 0x85); /* NEL */
4029  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4030              if (utf8)              if (utf)
4031                {                {
4032                class_utf8 = TRUE;                xclass = TRUE;
4033                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4034                class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2028, class_uchardata);
4035                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2029, class_uchardata);
4036                }                }
4037  #endif  #endif
4038              continue;              continue;
# Line 3723  for (;; ptr++) Line 4054  for (;; ptr++)
4054                classbits[c] |= x;                classbits[c] |= x;
4055                }                }
4056    
4057  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4058              if (utf8)              if (utf)
4059                {                {
4060                class_utf8 = TRUE;                xclass = TRUE;
4061                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4062                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x0100, class_uchardata);
4063                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2027, class_uchardata);
4064                *class_utf8data++ = XCL_RANGE;                *class_uchardata++ = XCL_RANGE;
4065                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x2029, class_uchardata);
4066                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);                class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
4067                }                }
4068  #endif  #endif
4069              continue;              continue;
# Line 3745  for (;; ptr++) Line 4076  for (;; ptr++)
4076                int pdata;                int pdata;
4077                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
4078                if (ptype < 0) goto FAILED;                if (ptype < 0) goto FAILED;
4079                class_utf8 = TRUE;                xclass = TRUE;
4080                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_uchardata++ = ((-c == ESC_p) != negated)?
4081                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
4082                *class_utf8data++ = ptype;                *class_uchardata++ = ptype;
4083                *class_utf8data++ = pdata;                *class_uchardata++ = pdata;
4084                class_charcount -= 2;   /* Not a < 256 character */                class_charcount -= 2;   /* Not a < 256 character */
4085                continue;                continue;
4086                }                }
# Line 3771  for (;; ptr++) Line 4102  for (;; ptr++)
4102            }            }
4103    
4104          /* 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
4105          greater than 256 in UTF-8 mode. */          greater than 256 mode. */
4106    
4107          }   /* End of backslash handling */          }   /* End of backslash handling */
4108    
# Line 3820  for (;; ptr++) Line 4151  for (;; ptr++)
4151            }            }
4152    
4153  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4154          if (utf8)          if (utf)
4155            {                           /* Braces are required because the */            {                           /* Braces are required because the */
4156            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */            GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */
4157            }            }
# Line 3869  for (;; ptr++) Line 4200  for (;; ptr++)
4200          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
4201          available. */          available. */
4202    
4203  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4204          if (utf8 && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))          if (utf && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))
4205    #endif
4206    #ifndef COMPILE_PCRE8
4207            if (d > 255)
4208    #endif
4209    #if defined SUPPORT_UTF || defined COMPILE_PCRE16
4210            {            {
4211            class_utf8 = TRUE;            xclass = TRUE;
4212    
4213            /* With UCP support, we can find the other case equivalents of            /* With UCP support, we can find the other case equivalents of
4214            the relevant characters. There may be several ranges. Optimize how            the relevant characters. There may be several ranges. Optimize how
# Line 3905  for (;; ptr++) Line 4241  for (;; ptr++)
4241    
4242                if (occ == ocd)                if (occ == ocd)
4243                  {                  {
4244                  *class_utf8data++ = XCL_SINGLE;                  *class_uchardata++ = XCL_SINGLE;
4245                  }                  }
4246                else                else
4247                  {                  {
4248                  *class_utf8data++ = XCL_RANGE;                  *class_uchardata++ = XCL_RANGE;
4249                  class_utf8data += _pcre_ord2utf8(occ, class_utf8data);                  class_uchardata += PRIV(ord2utf)(occ, class_uchardata);
4250                  }                  }
4251                class_utf8data += _pcre_ord2utf8(ocd, class_utf8data);                class_uchardata += PRIV(ord2utf)(ocd, class_uchardata);
4252                }                }
4253              }              }
4254  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
# Line 3920  for (;; ptr++) Line 4256  for (;; ptr++)
4256            /* Now record the original range, possibly modified for UCP caseless            /* Now record the original range, possibly modified for UCP caseless
4257            overlapping ranges. */            overlapping ranges. */
4258    
4259            *class_utf8data++ = XCL_RANGE;            *class_uchardata++ = XCL_RANGE;
4260            class_utf8data += _pcre_ord2utf8(c, class_utf8data);  #ifdef SUPPORT_UTF
4261            class_utf8data += _pcre_ord2utf8(d, class_utf8data);            class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4262              class_uchardata += PRIV(ord2utf)(d, class_uchardata);
4263    #else
4264              *class_uchardata++ = c;
4265              *class_uchardata++ = d;
4266    #endif
4267    
4268            /* With UCP support, we are done. Without UCP support, there is no            /* With UCP support, we are done. Without UCP support, there is no
4269            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
4270            for the smaller ones. */            for the smaller ones. As for 16 bit characters without UTF, we
4271              can still use  */
4272    
4273  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4274            continue;    /* With next character in the class */            continue;    /* With next character in the class */
4275  #else  #else
4276    #ifdef SUPPORT_UTF
4277            if ((options & PCRE_CASELESS) == 0 || c > 127) continue;            if ((options & PCRE_CASELESS) == 0 || c > 127) continue;
   
4278            /* Adjust upper limit and fall through to set up the map */            /* Adjust upper limit and fall through to set up the map */
   
4279            d = 127;            d = 127;
4280    #else
4281              if (c > 255) continue;
4282              /* Adjust upper limit and fall through to set up the map */
4283              d = 255;
4284    #endif  /* SUPPORT_UTF */
4285  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4286            }            }
4287  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 || COMPILE_PCRE16 */
4288    
4289          /* 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
4290          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. */  
4291    
4292          class_charcount += d - c + 1;          class_charcount += d - c + 1;
4293          class_lastchar = d;          class_lastchar = d;
# Line 3971  for (;; ptr++) Line 4315  for (;; ptr++)
4315    
4316        /* Handle a character that cannot go in the bit map */        /* Handle a character that cannot go in the bit map */
4317    
4318  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4319        if (utf8 && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))        if (utf && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))
4320          {  #elif !(defined COMPILE_PCRE8)
4321          class_utf8 = TRUE;        if (c > 255)
4322          *class_utf8data++ = XCL_SINGLE;  #endif
4323          class_utf8data += _pcre_ord2utf8(c, class_utf8data);  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
4324            {
4325            xclass = TRUE;
4326            *class_uchardata++ = XCL_SINGLE;
4327    #ifdef SUPPORT_UTF
4328            class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4329    #else
4330            *class_uchardata++ = c;
4331    #endif
4332    
4333  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4334          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
# Line 3984  for (;; ptr++) Line 4336  for (;; ptr++)
4336            unsigned int othercase;            unsigned int othercase;
4337            if ((othercase = UCD_OTHERCASE(c)) != c)            if ((othercase = UCD_OTHERCASE(c)) != c)
4338              {              {
4339              *class_utf8data++ = XCL_SINGLE;              *class_uchardata++ = XCL_SINGLE;
4340              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_uchardata += PRIV(ord2utf)(othercase, class_uchardata);
4341              }              }
4342            }            }
4343  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4344    
4345          }          }
4346        else        else
4347  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF || COMPILE_PCRE16 */
   
4348        /* Handle a single-byte character */        /* Handle a single-byte character */
4349          {          {
4350          classbits[c/8] |= (1 << (c&7));          classbits[c/8] |= (1 << (c&7));
# Line 4005  for (;; ptr++) Line 4356  for (;; ptr++)
4356          class_charcount++;          class_charcount++;
4357          class_lastchar = c;          class_lastchar = c;
4358          }          }
4359    
4360        }        }
4361    
4362      /* Loop until ']' reached. This "while" is the end of the "do" far above.      /* Loop until ']' reached. This "while" is the end of the "do" far above.
# Line 4037  for (;; ptr++) Line 4389  for (;; ptr++)
4389      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
4390      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4391      Note that OP_NOT[I] does not support multibyte characters. In the positive      Note that OP_NOT[I] does not support multibyte characters. In the positive
4392      case, it can cause firstbyte to be set. Otherwise, there can be no first      case, it can cause firstchar to be set. Otherwise, there can be no first
4393      char if this item is first, whatever repeat count may follow. In the case      char if this item is first, whatever repeat count may follow. In the case
4394      of reqbyte, save the previous value for reinstating. */      of reqchar, save the previous value for reinstating. */
4395    
4396  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4397      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !xclass &&
4398        (!utf8 || !negate_class || class_lastchar < 128))        (!utf || !negate_class || class_lastchar < 128))
4399  #else  #else
4400      if (class_charcount == 1)      if (class_charcount == 1)
4401  #endif  #endif
4402        {        {
4403        zeroreqbyte = reqbyte;        zeroreqchar = reqchar;
4404    
4405        /* The OP_NOT[I] opcodes work on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4406    
4407        if (negate_class)        if (negate_class)
4408          {          {
4409          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4410          zerofirstbyte = firstbyte;          zerofirstchar = firstchar;
4411          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4412          *code++ = class_lastchar;          *code++ = class_lastchar;
4413          break;          break;
# Line 4065  for (;; ptr++) Line 4417  for (;; ptr++)
4417        then we can handle this with the normal one-character code. */        then we can handle this with the normal one-character code. */
4418    
4419  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4420        if (utf8 && class_lastchar > 127)        if (utf && class_lastchar > 127)
4421          mclength = _pcre_ord2utf8(class_lastchar, mcbuffer);          mclength = PRIV(ord2utf)(class_lastchar, mcbuffer);
4422        else        else
4423  #endif  #endif
4424          {          {
# Line 4078  for (;; ptr++) Line 4430  for (;; ptr++)
4430    
4431      /* The general case - not the one-char optimization. If this is the first      /* The general case - not the one-char optimization. If this is the first
4432      thing in the branch, there can be no first char setting, whatever the      thing in the branch, there can be no first char setting, whatever the
4433      repeat count. Any reqbyte setting must remain unchanged after any kind of      repeat count. Any reqchar setting must remain unchanged after any kind of
4434      repeat. */      repeat. */
4435    
4436      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4437      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
4438      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
4439    
4440      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4441      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 4093  for (;; ptr++) Line 4445  for (;; ptr++)
4445      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
4446      actual compiled code. */      actual compiled code. */
4447    
4448  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4449      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))      if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
4450    #endif
4451    #ifndef COMPILE_PCRE8
4452        if (xclass && !should_flip_negation)
4453    #endif
4454    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4455        {        {
4456        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_uchardata++ = XCL_END;    /* Marks the end of extra data */
4457        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
4458        code += LINK_SIZE;        code += LINK_SIZE;
4459        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT:0;
4460    
4461        /* 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;
4462        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. */
# Line 4107  for (;; ptr++) Line 4464  for (;; ptr++)
4464        if (class_charcount > 0)        if (class_charcount > 0)
4465          {          {
4466          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
4467          memmove(code + 32, code, class_utf8data - code);          memmove(code + (32 / sizeof(pcre_uchar)), code,
4468              IN_UCHARS(class_uchardata - code));
4469          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
4470          code = class_utf8data + 32;          code = class_uchardata + (32 / sizeof(pcre_uchar));
4471          }          }
4472        else code = class_utf8data;        else code = class_uchardata;
4473    
4474        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
4475    
# Line 4127  for (;; ptr++) Line 4485  for (;; ptr++)
4485      negating it if necessary. */      negating it if necessary. */
4486    
4487      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4488      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  
4489        {        {
4490          if (negate_class)
4491            for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
4492        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4493        }        }
4494      code += 32;      code += 32 / sizeof(pcre_uchar);
4495      break;      break;
4496    
4497    
# Line 4173  for (;; ptr++) Line 4528  for (;; ptr++)
4528    
4529      if (repeat_min == 0)      if (repeat_min == 0)
4530        {        {
4531        firstbyte = zerofirstbyte;    /* Adjust for zero repeat */        firstchar = zerofirstchar;    /* Adjust for zero repeat */
4532        reqbyte = zeroreqbyte;        /* Ditto */        reqchar = zeroreqchar;        /* Ditto */
4533        }        }
4534    
4535      /* Remember whether this is a variable length repeat */      /* Remember whether this is a variable length repeat */
# Line 4184  for (;; ptr++) Line 4539  for (;; ptr++)
4539      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4540      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4541    
4542      /* 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
4543      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4544    
4545      tempcode = previous;      tempcode = previous;
4546    
# Line 4208  for (;; ptr++) Line 4563  for (;; ptr++)
4563        }        }
4564      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4565    
4566        /* If previous was a recursion call, wrap it in atomic brackets so that
4567        previous becomes the atomic group. All recursions were so wrapped in the
4568        past, but it no longer happens for non-repeated recursions. In fact, the
4569        repeated ones could be re-implemented independently so as not to need this,
4570        but for the moment we rely on the code for repeating groups. */
4571    
4572        if (*previous == OP_RECURSE)
4573          {
4574          memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
4575          *previous = OP_ONCE;
4576          PUT(previous, 1, 2 + 2*LINK_SIZE);
4577          previous[2 + 2*LINK_SIZE] = OP_KET;
4578          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4579          code += 2 + 2 * LINK_SIZE;
4580          length_prevgroup = 3 + 3*LINK_SIZE;
4581    
4582          /* When actually compiling, we need to check whether this was a forward
4583          reference, and if so, adjust the offset. */
4584    
4585          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4586            {
4587            int offset = GET(cd->hwm, -LINK_SIZE);
4588            if (offset == previous + 1 - cd->start_code)
4589              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4590            }
4591          }
4592    
4593        /* Now handle repetition for the different types of item. */
4594    
4595      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4596      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
4597      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
4598      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
4599      instead.  */      instead.  */
4600    
4601      if (*previous == OP_CHAR || *previous == OP_CHARI)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4602        {        {
4603        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4604    
4605        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF characters that take up more than one character. It's
4606        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
4607        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
4608        length rather than a small character. */        it's a length rather than a small character. */
4609    
4610  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4611        if (utf8 && (code[-1] & 0x80) != 0)  #ifdef COMPILE_PCRE8
4612          if (utf && (code[-1] & 0x80) != 0)
4613    #endif /* COMPILE_PCRE8 */
4614    #ifdef COMPILE_PCRE16
4615          if (utf && (code[-1] & 0xfc00) == 0xdc00)
4616    #endif /* COMPILE_PCRE8 */
4617          {          {
4618          uschar *lastchar = code - 1;          pcre_uchar *lastchar = code - 1;
4619          while((*lastchar & 0xc0) == 0x80) lastchar--;          BACKCHAR(lastchar);
4620          c = code - lastchar;            /* Length of UTF-8 character */          c = code - lastchar;            /* Length of UTF-8 character */
4621          memcpy(utf8_char, lastchar, c); /* Save the char */          memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
4622          c |= 0x80;                      /* Flag c as a length */          c |= UTF_LENGTH;                /* Flag c as a length */
4623          }          }
4624        else        else
4625  #endif  #endif /* SUPPORT_UTF */
4626    
4627        /* Handle the case of a single byte - either with no UTF8 support, or        /* Handle the case of a single charater - either with no UTF support, or
4628        with UTF-8 disabled, or for a UTF-8 character < 128. */        with UTF disabled, or for a single character UTF character. */
4629    
4630          {          {
4631          c = code[-1];          c = code[-1];
4632          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqchar = c | req_caseopt | cd->req_varyopt;
4633          }          }
4634    
4635        /* 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 4250  for (;; ptr++) Line 4639  for (;; ptr++)
4639    
4640        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4641            repeat_max < 0 &&            repeat_max < 0 &&
4642            check_auto_possessive(previous, utf8, ptr + 1, options, cd))            check_auto_possessive(previous, utf, ptr + 1, options, cd))
4643          {          {
4644          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4645          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4271  for (;; ptr++) Line 4660  for (;; ptr++)
4660        c = previous[1];        c = previous[1];
4661        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4662            repeat_max < 0 &&            repeat_max < 0 &&
4663            check_auto_possessive(previous, utf8, ptr + 1, options, cd))            check_auto_possessive(previous, utf, ptr + 1, options, cd))
4664          {          {
4665          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4666          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4288  for (;; ptr++) Line 4677  for (;; ptr++)
4677    
4678      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
4679        {        {
4680        uschar *oldcode;        pcre_uchar *oldcode;
4681        int prop_type, prop_value;        int prop_type, prop_value;
4682        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
4683        c = *previous;        c = *previous;
4684    
4685        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4686            repeat_max < 0 &&            repeat_max < 0 &&
4687            check_auto_possessive(previous, utf8, ptr + 1, options, cd))            check_auto_possessive(previous, utf, ptr + 1, options, cd))
4688          {          {
4689          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4690          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4375  for (;; ptr++) Line 4764  for (;; ptr++)
4764          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
4765          Unicode property match, there are two extra bytes that define the          Unicode property match, there are two extra bytes that define the
4766          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
4767          c, with the 0x80 bit as a flag. */          c, with the UTF_LENGTH bit as a flag. */
4768    
4769          if (repeat_max < 0)          if (repeat_max < 0)
4770            {            {
4771  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4772            if (utf8 && c >= 128)            if (utf && (c & UTF_LENGTH) != 0)
4773              {              {
4774              memcpy(code, utf8_char, c & 7);              memcpy(code, utf_chars, IN_UCHARS(c & 7));
4775              code += c & 7;              code += c & 7;
4776              }              }
4777            else            else
# Line 4404  for (;; ptr++) Line 4793  for (;; ptr++)
4793    
4794          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
4795            {            {
4796  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4797            if (utf8 && c >= 128)            if (utf && (c & UTF_LENGTH) != 0)
4798              {              {
4799              memcpy(code, utf8_char, c & 7);              memcpy(code, utf_chars, IN_UCHARS(c & 7));
4800              code += c & 7;              code += c & 7;
4801              }              }
4802            else            else
# Line 4434  for (;; ptr++) Line 4823  for (;; ptr++)
4823    
4824        /* The character or character type itself comes last in all cases. */        /* The character or character type itself comes last in all cases. */
4825    
4826  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4827        if (utf8 && c >= 128)        if (utf && (c & UTF_LENGTH) != 0)
4828          {          {
4829          memcpy(code, utf8_char, c & 7);          memcpy(code, utf_chars, IN_UCHARS(c & 7));
4830          code += c & 7;          code += c & 7;
4831          }          }
4832        else        else
# Line 4461  for (;; ptr++) Line 4850  for (;; ptr++)
4850    
4851      else if (*previous == OP_CLASS ||      else if (*previous == OP_CLASS ||
4852               *previous == OP_NCLASS ||               *previous == OP_NCLASS ||
4853  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF8 || !defined COMPILE_PCRE8
4854               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4855  #endif  #endif
4856               *previous == OP_REF ||               *previous == OP_REF ||
# Line 4499  for (;; ptr++) Line 4888  for (;; ptr++)
4888        }        }
4889    
4890      /* 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
4891      cases. Note that at this point we can encounter only the "basic" BRA and      cases. Note that at this point we can encounter only the "basic" bracket
4892      KET opcodes, as this is the place where they get converted into the more      opcodes such as BRA and CBRA, as this is the place where they get converted
4893      special varieties. */      into the more special varieties such as BRAPOS and SBRA. A test for >=
4894        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4895        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4896        repetition of assertions, but now it does, for Perl compatibility. */
4897    
4898      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4899        {        {
4900        register int i;        register int i;
4901        int len = (int)(code - previous);        int len = (int)(code - previous);
4902        uschar *bralink = NULL;        pcre_uchar *bralink = NULL;
4903        uschar *brazeroptr = NULL;        pcre_uchar *brazeroptr = NULL;
4904    
4905        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4906          we just ignore the repeat. */
4907    
4908        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4909            goto END_REPEAT;
4910    
4911          /* There is no sense in actually repeating assertions. The only potential
4912          use of repetition is in cases when the assertion is optional. Therefore,
4913          if the minimum is greater than zero, just ignore the repeat. If the
4914          maximum is not not zero or one, set it to 1. */
4915    
4916          if (*previous < OP_ONCE)    /* Assertion */
4917          {          {
4918          *errorcodeptr = ERR55;          if (repeat_min > 0) goto END_REPEAT;
4919          goto FAILED;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
4920          }          }
4921    
4922        /* 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 4537  for (;; ptr++) Line 4937  for (;; ptr++)
4937          **   goto END_REPEAT;          **   goto END_REPEAT;
4938          **   }          **   }
4939    
4940          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
4941          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
4942          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
4943          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4944            selectively.
4945    
4946          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
4947          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 4552  for (;; ptr++) Line 4953  for (;; ptr++)
4953          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4954            {            {
4955            *code = OP_END;            *code = OP_END;
4956            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf, cd, save_hwm);
4957            memmove(previous+1, previous, len);            memmove(previous + 1, previous, IN_UCHARS(len));
4958            code++;            code++;
4959            if (repeat_max == 0)            if (repeat_max == 0)
4960              {              {
# Line 4576  for (;; ptr++) Line 4977  for (;; ptr++)
4977            {            {
4978            int offset;            int offset;
4979            *code = OP_END;            *code = OP_END;
4980            adjust_recurse(previous, 2 + LINK_SIZE, utf8, cd, save_hwm);            adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
4981            memmove(previous + 2 + LINK_SIZE, previous, len);            memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
4982            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
4983            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4984            *previous++ = OP_BRA;            *previous++ = OP_BRA;
# Line 4627  for (;; ptr++) Line 5028  for (;; ptr++)
5028    
5029            else            else
5030              {              {
5031              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;              if (groupsetfirstchar && reqchar < 0) reqchar = firstchar;
5032              for (i = 1; i < repeat_min; i++)              for (i = 1; i < repeat_min; i++)
5033                {                {
5034                uschar *hc;                pcre_uchar *hc;
5035                uschar *this_hwm = cd->hwm;                pcre_uchar *this_hwm = cd->hwm;
5036                memcpy(code, previous, len);                memcpy(code, previous, IN_UCHARS(len));
5037                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5038                  {                  {
5039                  PUT(cd->hwm, 0, GET(hc, 0) + len);                  PUT(cd->hwm, 0, GET(hc, 0) + len);
# Line 4682  for (;; ptr++) Line 5083  for (;; ptr++)
5083    
5084          else for (i = repeat_max - 1; i >= 0; i--)          else for (i = repeat_max - 1; i >= 0; i--)
5085            {            {
5086            uschar *hc;            pcre_uchar *hc;
5087            uschar *this_hwm = cd->hwm;            pcre_uchar *this_hwm = cd->hwm;
5088    
5089            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
5090    
# Line 4699  for (;; ptr++) Line 5100  for (;; ptr++)
5100              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
5101              }              }
5102    
5103            memcpy(code, previous, len);            memcpy(code, previous, IN_UCHARS(len));
5104            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5105              {              {
5106              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 4716  for (;; ptr++) Line 5117  for (;; ptr++)
5117            {            {
5118            int oldlinkoffset;            int oldlinkoffset;
5119            int offset = (int)(code - bralink + 1);            int offset = (int)(code - bralink + 1);
5120            uschar *bra = code - offset;            pcre_uchar *bra = code - offset;
5121            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
5122            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
5123            *code++ = OP_KET;            *code++ = OP_KET;
# Line 4726  for (;; ptr++) Line 5127  for (;; ptr++)
5127          }          }
5128    
5129        /* If the maximum is unlimited, set a repeater in the final copy. For        /* If the maximum is unlimited, set a repeater in the final copy. For
5130        ONCE brackets, that's all we need to do.        ONCE brackets, that's all we need to do. However, possessively repeated
5131          ONCE brackets can be converted into non-capturing brackets, as the
5132        Otherwise, if the quantifier was possessive, we convert the BRA code to        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5133        the POS form, and the KET code to KETRPOS. (It turns out to be convenient        deal with possessive ONCEs specially.
5134        at runtime to detect this kind of subpattern at both the start and at the  
5135        end.) The use of special opcodes makes it possible to reduce greatly the        Otherwise, when we are doing the actual compile phase, check to see
5136        stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,        whether this group is one that could match an empty string. If so,
5137        convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
5138        the default action below, of wrapping everything inside atomic brackets,        that runtime checking can be done. [This check is also applied to ONCE
5139        does not happen.        groups at runtime, but in a different way.]
5140    
5141        Then, when we are doing the actual compile phase, check to see whether        Then, if the quantifier was possessive and the bracket is not a
5142        this group is one that could match an empty string. If so, convert the        conditional, we convert the BRA code to the POS form, and the KET code to
5143        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime        KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5144        checking can be done. [This check is also applied to ONCE groups at        subpattern at both the start and at the end.) The use of special opcodes
5145        runtime, but in a different way.] */        makes it possible to reduce greatly the stack usage in pcre_exec(). If
5146          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5147    
5148          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5149          flag so that the default action below, of wrapping everything inside
5150          atomic brackets, does not happen. When the minimum is greater than 1,
5151          there will be earlier copies of the group, and so we still have to wrap
5152          the whole thing. */
5153    
5154        else        else
5155          {          {
5156          uschar *ketcode = code - 1 - LINK_SIZE;          pcre_uchar *ketcode = code - 1 - LINK_SIZE;
5157          uschar *bracode = ketcode - GET(ketcode, 1);          pcre_uchar *bracode = ketcode - GET(ketcode, 1);
5158    
5159          if (*bracode == OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5160    
5161            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5162                possessive_quantifier) *bracode = OP_BRA;
5163    
5164            /* For non-possessive ONCE brackets, all we need to do is to
5165            set the KET. */
5166    
5167            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5168            *ketcode = OP_KETRMAX + repeat_type;            *ketcode = OP_KETRMAX + repeat_type;
5169    
5170            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5171            converted to non-capturing above). */
5172    
5173          else          else
5174            {            {
5175            if (possessive_quantifier)            /* In the compile phase, check for empty string matching. */
5176              {  
             *bracode += 1;                   /* Switch to xxxPOS opcodes */  
             *ketcode = OP_KETRPOS;  
             if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;  
             possessive_quantifier = FALSE;  
             }  
           else *ketcode = OP_KETRMAX + repeat_type;  
   
5177            if (lengthptr == NULL)            if (lengthptr == NULL)
5178              {              {
5179              uschar *scode = bracode;              pcre_uchar *scode = bracode;
5180              do              do
5181                {                {
5182                if (could_be_empty_branch(scode, ketcode, utf8, cd))                if (could_be_empty_branch(scode, ketcode, utf, cd))
5183                  {                  {
5184                  *bracode += OP_SBRA - OP_BRA;                  *bracode += OP_SBRA - OP_BRA;
5185                  break;                  break;
# Line 4775  for (;; ptr++) Line 5188  for (;; ptr++)
5188                }                }
5189              while (*scode == OP_ALT);              while (*scode == OP_ALT);
5190              }              }
5191    
5192              /* Handle possessive quantifiers. */
5193    
5194              if (possessive_quantifier)
5195                {
5196                /* For COND brackets, we wrap the whole thing in a possessively
5197                repeated non-capturing bracket, because we have not invented POS
5198                versions of the COND opcodes. Because we are moving code along, we
5199                must ensure that any pending recursive references are updated. */
5200    
5201                if (*bracode == OP_COND || *bracode == OP_SCOND)
5202                  {
5203                  int nlen = (int)(code - bracode);
5204                  *code = OP_END;
5205                  adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
5206                  memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
5207                  code += 1 + LINK_SIZE;
5208                  nlen += 1 + LINK_SIZE;
5209                  *bracode = OP_BRAPOS;
5210                  *code++ = OP_KETRPOS;
5211                  PUTINC(code, 0, nlen);
5212                  PUT(bracode, 1, nlen);
5213                  }
5214    
5215                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5216    
5217                else
5218                  {
5219                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5220                  *ketcode = OP_KETRPOS;
5221                  }
5222    
5223                /* If the minimum is zero, mark it as possessive, then unset the
5224                possessive flag when the minimum is 0 or 1. */
5225    
5226                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5227                if (repeat_min < 2) possessive_quantifier = FALSE;
5228                }
5229    
5230              /* Non-possessive quantifier */
5231    
5232              else *ketcode = OP_KETRMAX + repeat_type;
5233            }            }
5234          }          }
5235        }        }
# Line 4799  for (;; ptr++) Line 5254  for (;; ptr++)
5254      there are special alternative opcodes for this case. For anything else, we      there are special alternative opcodes for this case. For anything else, we
5255      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5256      notation is just syntactic sugar, taken from Sun's Java package, but the      notation is just syntactic sugar, taken from Sun's Java package, but the
5257      special opcodes can optimize it.      special opcodes can optimize it.
5258    
5259      Possessively repeated subpatterns have already been handled in the code      Some (but not all) possessively repeated subpatterns have already been
5260      just above, so possessive_quantifier is always FALSE for them at this      completely handled in the code just above. For them, possessive_quantifier
5261      stage.      is always FALSE at this stage.
5262    
5263      Note that the repeated item starts at tempcode, not at previous, which      Note that the repeated item starts at tempcode, not at previous, which
5264      might be the first part of a string whose (former) last char we repeated.      might be the first part of a string whose (former) last char we repeated.
5265    
# Line 4817  for (;; ptr++) Line 5272  for (;; ptr++)
5272        int len;        int len;
5273    
5274        if (*tempcode == OP_TYPEEXACT)        if (*tempcode == OP_TYPEEXACT)
5275          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += PRIV(OP_lengths)[*tempcode] +
5276            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);            ((tempcode[1 + IMM2_SIZE] == OP_PROP
5277              || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
5278    
5279        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
5280          {          {
5281          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += PRIV(OP_lengths)[*tempcode];
5282  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
5283          if (utf8 && tempcode[-1] >= 0xc0)          if (utf && HAS_EXTRALEN(tempcode[-1]))
5284            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];            tempcode += GET_EXTRALEN(tempcode[-1]);
5285  #endif  #endif
5286          }          }
5287    
# Line 4862  for (;; ptr++) Line 5318  for (;; ptr++)
5318    
5319          default:          default:
5320          *code = OP_END;          *code = OP_END;
5321          adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);          adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
5322          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
5323          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
5324          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
5325          tempcode[0] = OP_ONCE;          tempcode[0] = OP_ONCE;
# Line 4875  for (;; ptr++) Line 5331  for (;; ptr++)
5331        }        }
5332    
5333      /* 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
5334      "follows varying string" flag for subsequently encountered reqbytes if      "follows varying string" flag for subsequently encountered reqchars if
5335      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. */
5336    
5337      END_REPEAT:      END_REPEAT:
# Line 4904  for (;; ptr++) Line 5360  for (;; ptr++)
5360        int i, namelen;        int i, namelen;
5361        int arglen = 0;        int arglen = 0;
5362        const char *vn = verbnames;        const char *vn = verbnames;
5363        const uschar *name = ptr + 1;        const pcre_uchar *name = ptr + 1;
5364        const uschar *arg = NULL;        const pcre_uchar *arg = NULL;
5365        previous = NULL;        previous = NULL;
5366        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5367        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5368    
5369          /* It appears that Perl allows any characters whatsoever, other than
5370          a closing parenthesis, to appear in arguments, so we no longer insist on
5371          letters, digits, and underscores. */
5372    
5373        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5374          {          {
5375          arg = ++ptr;          arg = ++ptr;
5376          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
           || *ptr == '_') ptr++;  
5377          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5378          }          }
5379    
# Line 4929  for (;; ptr++) Line 5388  for (;; ptr++)
5388        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5389          {          {
5390          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5391              strncmp((char *)name, vn, namelen) == 0)              STRNCMP_UC_C8(name, vn, namelen) == 0)
5392            {            {
5393            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5394              ASSERT_ACCEPT if in an assertion. */
5395    
5396            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5397              {              {
5398              open_capitem *oc;              open_capitem *oc;
5399                if (arglen != 0)
5400                  {
5401                  *errorcodeptr = ERR59;
5402                  goto FAILED;
5403                  }
5404              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5405              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5406                {                {
5407                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5408                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5409                }                }
5410                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5411    
5412                /* Do not set firstchar after *ACCEPT */
5413                if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
5414              }              }
5415    
5416            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5417    
5418            if (arglen == 0)            else if (arglen == 0)
5419              {              {
5420              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5421                {                {
# Line 4954  for (;; ptr++) Line 5423  for (;; ptr++)
5423                goto FAILED;                goto FAILED;
5424                }                }
5425              *code = verbs[i].op;              *code = verbs[i].op;
5426              if (*code++ == OP_THEN)              if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5427              }              }
5428    
5429            else            else
# Line 4969  for (;; ptr++) Line 5434  for (;; ptr++)
5434                goto FAILED;                goto FAILED;
5435                }                }
5436              *code = verbs[i].op_arg;              *code = verbs[i].op_arg;
5437              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;  
               }  
5438              *code++ = arglen;              *code++ = arglen;
5439              memcpy(code, arg, arglen);              memcpy(code, arg, IN_UCHARS(arglen));
5440              code += arglen;              code += arglen;
5441              *code++ = 0;              *code++ = 0;
5442              }              }
# Line 4998  for (;; ptr++) Line 5459  for (;; ptr++)
5459        {        {
5460        int i, set, unset, namelen;        int i, set, unset, namelen;
5461        int *optset;        int *optset;
5462        const uschar *name;        const pcre_uchar *name;
5463        uschar *slot;        pcre_uchar *slot;
5464    
5465        switch (*(++ptr))        switch (*(++ptr))
5466          {          {
# Line 5052  for (;; ptr++) Line 5513  for (;; ptr++)
5513            break;            break;
5514    
5515          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Most other conditions use OP_CREF (a couple change to OP_RREF
5516          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. */
5517    
5518          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
5519          skipbytes = 3;          skipbytes = 1+IMM2_SIZE;
5520          refsign = -1;          refsign = -1;
5521    
5522          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
# Line 5102  for (;; ptr++) Line 5563  for (;; ptr++)
5563          while ((cd->ctypes[*ptr] & ctype_word) != 0)          while ((cd->ctypes[*ptr] & ctype_word) != 0)
5564            {            {
5565            if (recno >= 0)            if (recno >= 0)
5566              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = (IS_DIGIT(*ptr))? recno * 10 + *ptr - CHAR_0 : -1;
               recno * 10 + *ptr - CHAR_0 : -1;  
5567            ptr++;            ptr++;
5568            }            }
5569          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
# Line 5151  for (;; ptr++) Line 5611  for (;; ptr++)
5611          slot = cd->name_table;          slot = cd->name_table;
5612          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
5613            {            {
5614            if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;            if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0) break;
5615            slot += cd->name_entry_size;            slot += cd->name_entry_size;
5616            }            }
5617    
# Line 5167  for (;; ptr++) Line 5627  for (;; ptr++)
5627          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5628    
5629          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5630                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)                          (options & PCRE_EXTENDED) != 0, utf)) > 0)
5631            {            {
5632            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5633            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 5193  for (;; ptr++) Line 5653  for (;; ptr++)
5653            recno = 0;            recno = 0;
5654            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
5655              {              {
5656              if ((digitab[name[i]] & ctype_digit) == 0)              if (!IS_DIGIT(name[i]))
5657                {                {
5658                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
5659                goto FAILED;                goto FAILED;
# Line 5208  for (;; ptr++) Line 5668  for (;; ptr++)
5668          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
5669          false. */          false. */
5670    
5671          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)          else if (namelen == 6 && STRNCMP_UC_C8(name, STRING_DEFINE, 6) == 0)
5672            {            {
5673            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
5674            skipbytes = 1;            skipbytes = 1;
# Line 5235  for (;; ptr++) Line 5695  for (;; ptr++)
5695          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5696          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5697          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5698            cd->assert_depth += 1;
5699          ptr++;          ptr++;
5700          break;          break;
5701    
# Line 5249  for (;; ptr++) Line 5710  for (;; ptr++)
5710            continue;            continue;
5711            }            }
5712          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5713            cd->assert_depth += 1;
5714          break;          break;
5715    
5716    
# Line 5258  for (;; ptr++) Line 5720  for (;; ptr++)
5720            {            {
5721            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5722            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5723              cd->assert_depth += 1;
5724            ptr += 2;            ptr += 2;
5725            break;            break;
5726    
5727            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5728            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5729              cd->assert_depth += 1;
5730            ptr += 2;            ptr += 2;
5731            break;            break;
5732    
# Line 5284  for (;; ptr++) Line 5748  for (;; ptr++)
5748    
5749          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5750          case CHAR_C:                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5751          previous_callout = code;  /* Save for later completion */          previous_callout = code;     /* Save for later completion */
5752          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1;    /* Skip one item before completing */
5753          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5754            {            {
5755            int n = 0;            int n = 0;
5756            while ((digitab[*(++ptr)] & ctype_digit) != 0)            ptr++;
5757              n = n * 10 + *ptr - CHAR_0;            while(IS_DIGIT(*ptr))
5758                n = n * 10 + *ptr++ - CHAR_0;
5759            if (*ptr != CHAR_RIGHT_PARENTHESIS)            if (*ptr != CHAR_RIGHT_PARENTHESIS)
5760              {              {
5761              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
# Line 5383  for (;; ptr++) Line 5848  for (;; ptr++)
5848    
5849              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5850                {                {
5851                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, IN_UCHARS(namelen));
5852                if (crc == 0)                if (crc == 0)
5853                  {                  {
5854                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
# Line 5407  for (;; ptr++) Line 5872  for (;; ptr++)
5872                if (crc < 0)                if (crc < 0)
5873                  {                  {
5874                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5875                    (cd->names_found - i) * cd->name_entry_size);                    IN_UCHARS((cd->names_found - i) * cd->name_entry_size));
5876                  break;                  break;
5877                  }                  }
5878    
# Line 5421  for (;; ptr++) Line 5886  for (;; ptr++)
5886    
5887              if (!dupname)              if (!dupname)
5888                {                {
5889                uschar *cslot = cd->name_table;                pcre_uchar *cslot = cd->name_table;
5890                for (i = 0; i < cd->names_found; i++)                for (i = 0; i < cd->names_found; i++)
5891                  {                  {
5892                  if (cslot != slot)                  if (cslot != slot)
# Line 5438  for (;; ptr++) Line 5903  for (;; ptr++)
5903                }                }
5904    
5905              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5906              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, IN_UCHARS(namelen));
5907              slot[2+namelen] = 0;              slot[2 + namelen] = 0;
5908              }              }
5909            }            }
5910    
# Line 5477  for (;; ptr++) Line 5942  for (;; ptr++)
5942    
5943          if (lengthptr != NULL)          if (lengthptr != NULL)
5944            {            {
5945            const uschar *temp;            const pcre_uchar *temp;
5946    
5947            if (namelen == 0)            if (namelen == 0)
5948              {              {
# Line 5507  for (;; ptr++) Line 5972  for (;; ptr++)
5972            temp = cd->end_pattern;            temp = cd->end_pattern;
5973            cd->end_pattern = ptr;            cd->end_pattern = ptr;
5974            recno = find_parens(cd, name, namelen,            recno = find_parens(cd, name, namelen,
5975              (options & PCRE_EXTENDED) != 0, utf8);              (options & PCRE_EXTENDED) != 0, utf);
5976            cd->end_pattern = temp;            cd->end_pattern = temp;
5977            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5978            }            }
# Line 5522  for (;; ptr++) Line 5987  for (;; ptr++)
5987            slot = cd->name_table;            slot = cd->name_table;
5988            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
5989              {              {
5990              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&              if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0 &&
5991                  slot[2+namelen] == 0)                  slot[2+namelen] == 0)
5992                break;                break;
5993              slot += cd->name_entry_size;              slot += cd->name_entry_size;
# Line 5534  for (;; ptr++) Line 5999  for (;; ptr++)
5999              }              }
6000            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
6001                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
6002                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf)) <= 0)
6003              {              {
6004              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
6005              goto FAILED;              goto FAILED;
# Line 5559  for (;; ptr++) Line 6024  for (;; ptr++)
6024          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
6025          case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:          case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
6026            {            {
6027            const uschar *called;            const pcre_uchar *called;
6028            terminator = CHAR_RIGHT_PARENTHESIS;            terminator = CHAR_RIGHT_PARENTHESIS;
6029    
6030            /* Come here from the \g<...> and \g'...' code (Oniguruma            /* Come here from the \g<...> and \g'...' code (Oniguruma
# Line 5573  for (;; ptr++) Line 6038  for (;; ptr++)
6038            if ((refsign = *ptr) == CHAR_PLUS)            if ((refsign = *ptr) == CHAR_PLUS)
6039              {              {
6040              ptr++;              ptr++;
6041              if ((digitab[*ptr] & ctype_digit) == 0)              if (!IS_DIGIT(*ptr))
6042                {                {
6043                *errorcodeptr = ERR63;                *errorcodeptr = ERR63;
6044                goto FAILED;                goto FAILED;
# Line 5581  for (;; ptr++) Line 6046  for (;; ptr++)
6046              }              }
6047            else if (refsign == CHAR_MINUS)            else if (refsign == CHAR_MINUS)
6048              {              {
6049              if ((digitab[ptr[1]] & ctype_digit) == 0)              if (!IS_DIGIT(ptr[1]))
6050                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
6051              ptr++;              ptr++;
6052              }              }
6053    
6054            recno = 0;            recno = 0;
6055            while((digitab[*ptr] & ctype_digit) != 0)            while(IS_DIGIT(*ptr))
6056              recno = recno * 10 + *ptr++ - CHAR_0;              recno = recno * 10 + *ptr++ - CHAR_0;
6057    
6058            if (*ptr != terminator)            if (*ptr != terminator)
# Line 5638  for (;; ptr++) Line 6103  for (;; ptr++)
6103              {              {
6104              *code = OP_END;              *code = OP_END;
6105              if (recno != 0)              if (recno != 0)
6106                called = _pcre_find_bracket(cd->start_code, utf8, recno);                called = PRIV(find_bracket)(cd->start_code, utf, recno);
6107    
6108              /* Forward reference */              /* Forward reference */
6109    
6110              if (called == NULL)              if (called == NULL)
6111                {                {
6112                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
6113                      (options & PCRE_EXTENDED) != 0, utf8) < 0)                      (options & PCRE_EXTENDED) != 0, utf) < 0)
6114                  {                  {
6115                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
6116                  goto FAILED;                  goto FAILED;
# Line 5653  for (;; ptr++) Line 6118  for (;; ptr++)
6118    
6119                /* Fudge the value of "called" so that when it is inserted as an                /* Fudge the value of "called" so that when it is inserted as an
6120                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
6121                of the group. */                of the group. Then remember the forward reference. */
6122    
6123                called = cd->start_code + recno;                called = cd->start_code + recno;
6124                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
6125                }                }
6126    
6127              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
6128              this is a recursive call. We check to see if this is a left              this is a recursive call. We check to see if this is a left
6129              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
6130                must not, however, do this check if we are in a conditional
6131                subpattern because the condition might be testing for recursion in
6132                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
6133                Forever loops are also detected at runtime, so those that occur in
6134                conditional subpatterns will be picked up then. */
6135    
6136              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
6137                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf, cd))
6138                {                {
6139                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
6140                goto FAILED;                goto FAILED;
6141                }                }
6142              }              }
6143    
6144            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
           "once" brackets. Set up a "previous group" length so that a  
           subsequent quantifier will work. */  
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
6145    
6146            *code = OP_RECURSE;            *code = OP_RECURSE;
6147            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
6148            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
   
           *code = OP_KET;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
6149            }            }
6150    
6151          /* Can't determine a first byte now */          /* Can't determine a first byte now */
6152    
6153          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
6154          continue;          continue;
6155    
6156    
# Line 5749  for (;; ptr++) Line 6207  for (;; ptr++)
6207          both phases.          both phases.
6208    
6209          If we are not at the pattern start, reset the greedy defaults and the          If we are not at the pattern start, reset the greedy defaults and the
6210          case value for firstbyte and reqbyte. */          case value for firstchar and reqchar. */
6211    
6212          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
6213            {            {
# Line 5762  for (;; ptr++) Line 6220  for (;; ptr++)
6220              {              {
6221              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
6222              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
6223              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
6224              }              }
6225    
6226            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
# Line 5799  for (;; ptr++) Line 6257  for (;; ptr++)
6257        NUMBERED_GROUP:        NUMBERED_GROUP:
6258        cd->bracount += 1;        cd->bracount += 1;
6259        PUT2(code, 1+LINK_SIZE, cd->bracount);        PUT2(code, 1+LINK_SIZE, cd->bracount);
6260        skipbytes = 2;        skipbytes = IMM2_SIZE;
6261        }        }
6262    
6263      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
6264      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a      but this was changed for Perl compatibility, so all kinds can now be
6265      non-register variable (tempcode) in order to be able to pass its address      repeated. We copy code into a non-register variable (tempcode) in order to
6266      because some compilers complain otherwise. */      be able to pass its address because some compilers complain otherwise. */
6267    
6268      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
6269      *code = bravalue;      *code = bravalue;
6270      tempcode = code;      tempcode = code;
6271      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
6272      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
6273        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
6274    
6275      if (!compile_regex(      if (!compile_regex(
6276           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
6277           &tempcode,                    /* Where to put code (updated) */           &tempcode,                       /* Where to put code (updated) */
6278           &ptr,                         /* Input pointer (updated) */           &ptr,                            /* Input pointer (updated) */
6279           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                    /* Where to put an error message */
6280           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
6281            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
6282           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
6283           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
6284           &subfirstbyte,                /* For possible first char */           cond_depth +
6285           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
6286           bcptr,                        /* Current branch chain */           &subfirstchar,                   /* For possible first char */
6287           cd,                           /* Tables block */           &subreqchar,                     /* For possible last char */
6288           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
6289             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
6290             (lengthptr == NULL)? NULL :      /* Actual compile phase */
6291               &length_prevgroup              /* Pre-compile phase */
6292           ))           ))
6293        goto FAILED;        goto FAILED;
6294    
6295        /* If this was an atomic group and there are no capturing groups within it,
6296        generate OP_ONCE_NC instead of OP_ONCE. */
6297    
6298        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
6299          *code = OP_ONCE_NC;
6300    
6301        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
6302          cd->assert_depth -= 1;
6303    
6304      /* At the end of compiling, code is still pointing to the start of the      /* At the end of compiling, code is still pointing to the start of the
6305      group, while tempcode has been updated to point past the end of the group      group, while tempcode has been updated to point past the end of the group.
6306      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6307    
6308      /* If this is a conditional bracket, check that there are no more than      If this is a conditional bracket, check that there are no more than
6309      two branches in the group, or just one if it's a DEFINE group. We do this      two branches in the group, or just one if it's a DEFINE group. We do this
6310      in the real compile phase, not in the pre-pass, where the whole group may      in the real compile phase, not in the pre-pass, where the whole group may
6311      not be available. */      not be available. */
6312    
6313      if (bravalue == OP_COND && lengthptr == NULL)      if (bravalue == OP_COND && lengthptr == NULL)
6314        {        {
6315        uschar *tc = code;        pcre_uchar *tc = code;
6316        int condcount = 0;        int condcount = 0;
6317    
6318        do {        do {
# Line 5866  for (;; ptr++) Line 6335  for (;; ptr++)
6335          }          }
6336    
6337        /* A "normal" conditional group. If there is just one branch, we must not        /* A "normal" conditional group. If there is just one branch, we must not
6338        make use of its firstbyte or reqbyte, because this is equivalent to an        make use of its firstchar or reqchar, because this is equivalent to an
6339        empty second branch. */        empty second branch. */
6340    
6341        else        else
# Line 5876  for (;; ptr++) Line 6345  for (;; ptr++)
6345            *errorcodeptr = ERR27;            *errorcodeptr = ERR27;
6346            goto FAILED;            goto FAILED;
6347            }            }
6348          if (condcount == 1) subfirstbyte = subreqbyte = REQ_NONE;          if (condcount == 1) subfirstchar = subreqchar = REQ_NONE;
6349          }          }
6350        }        }
6351    
# Line 5920  for (;; ptr++) Line 6389  for (;; ptr++)
6389      /* Handle updating of the required and first characters for other types of      /* Handle updating of the required and first characters for other types of
6390      group. Update for normal brackets of all kinds, and conditions with two      group. Update for normal brackets of all kinds, and conditions with two
6391      branches (see code above). If the bracket is followed by a quantifier with      branches (see code above). If the bracket is followed by a quantifier with
6392      zero repeat, we have to back off. Hence the definition of zeroreqbyte and      zero repeat, we have to back off. Hence the definition of zeroreqchar and
6393      zerofirstbyte outside the main loop so that they can be accessed for the      zerofirstchar outside the main loop so that they can be accessed for the
6394      back off. */      back off. */
6395    
6396      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
6397      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
6398      groupsetfirstbyte = FALSE;      groupsetfirstchar = FALSE;
6399    
6400      if (bravalue >= OP_ONCE)      if (bravalue >= OP_ONCE)
6401        {        {
6402        /* If we have not yet set a firstbyte in this branch, take it from the        /* If we have not yet set a firstchar in this branch, take it from the
6403        subpattern, remembering that it was set here so that a repeat of more        subpattern, remembering that it was set here so that a repeat of more
6404        than one can replicate it as reqbyte if necessary. If the subpattern has        than one can replicate it as reqchar if necessary. If the subpattern has
6405        no firstbyte, set "none" for the whole branch. In both cases, a zero        no firstchar, set "none" for the whole branch. In both cases, a zero
6406        repeat forces firstbyte to "none". */        repeat forces firstchar to "none". */
6407    
6408        if (firstbyte == REQ_UNSET)        if (firstchar == REQ_UNSET)
6409          {          {
6410          if (subfirstbyte >= 0)          if (subfirstchar >= 0)
6411            {            {
6412            firstbyte = subfirstbyte;            firstchar = subfirstchar;
6413            groupsetfirstbyte = TRUE;            groupsetfirstchar = TRUE;
6414            }            }
6415          else firstbyte = REQ_NONE;          else firstchar = REQ_NONE;
6416          zerofirstbyte = REQ_NONE;          zerofirstchar = REQ_NONE;
6417          }          }
6418    
6419        /* If firstbyte was previously set, convert the subpattern's firstbyte        /* If firstchar was previously set, convert the subpattern's firstchar
6420        into reqbyte if there wasn't one, using the vary flag that was in        into reqchar if there wasn't one, using the vary flag that was in
6421        existence beforehand. */        existence beforehand. */
6422    
6423        else if (subfirstbyte >= 0 && subreqbyte < 0)        else if (subfirstchar >= 0 && subreqchar < 0)
6424          subreqbyte = subfirstbyte | tempreqvary;          subreqchar = subfirstchar | tempreqvary;
6425    
6426        /* If the subpattern set a required byte (or set a first byte that isn't        /* If the subpattern set a required byte (or set a first byte that isn't
6427        really the first byte - see above), set it. */        really the first byte - see above), set it. */
6428    
6429        if (subreqbyte >= 0) reqbyte = subreqbyte;        if (subreqchar >= 0) reqchar = subreqchar;
6430        }        }
6431    
6432      /* For a forward assertion, we take the reqbyte, if set. This can be      /* For a forward assertion, we take the reqchar, if set. This can be
6433      helpful if the pattern that follows the assertion doesn't set a different      helpful if the pattern that follows the assertion doesn't set a different
6434      char. For example, it's useful for /(?=abcde).+/. We can't set firstbyte      char. For example, it's useful for /(?=abcde).+/. We can't set firstchar
6435      for an assertion, however because it leads to incorrect effect for patterns      for an assertion, however because it leads to incorrect effect for patterns
6436      such as /(?=a)a.+/ when the "real" "a" would then become a reqbyte instead      such as /(?=a)a.+/ when the "real" "a" would then become a reqchar instead
6437      of a firstbyte. This is overcome by a scan at the end if there's no      of a firstchar. This is overcome by a scan at the end if there's no
6438      firstbyte, looking for an asserted first char. */      firstchar, looking for an asserted first char. */
6439    
6440      else if (bravalue == OP_ASSERT && subreqbyte >= 0) reqbyte = subreqbyte;      else if (bravalue == OP_ASSERT && subreqchar >= 0) reqchar = subreqchar;
6441      break;     /* End of processing '(' */      break;     /* End of processing '(' */
6442    
6443    
# Line 6001  for (;; ptr++) Line 6470  for (;; ptr++)
6470        /* For metasequences that actually match a character, we disable the        /* For metasequences that actually match a character, we disable the
6471        setting of a first character if it hasn't already been set. */        setting of a first character if it hasn't already been set. */
6472    
6473        if (firstbyte == REQ_UNSET && -c > ESC_b && -c < ESC_Z)        if (firstchar == REQ_UNSET && -c > ESC_b && -c < ESC_Z)
6474          firstbyte = REQ_NONE;          firstchar = REQ_NONE;
6475    
6476        /* Set values to reset to if this is followed by a zero repeat. */        /* Set values to reset to if this is followed by a zero repeat. */
6477    
6478        zerofirstbyte = firstbyte;        zerofirstchar = firstchar;
6479        zeroreqbyte = reqbyte;        zeroreqchar = reqchar;
6480    
6481        /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'        /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
6482        is a subroutine call by number (Oniguruma syntax). In fact, the value        is a subroutine call by number (Oniguruma syntax). In fact, the value
# Line 6018  for (;; ptr++) Line 6487  for (;; ptr++)
6487    
6488        if (-c == ESC_g)        if (-c == ESC_g)
6489          {          {
6490          const uschar *p;          const pcre_uchar *p;
6491          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
6492          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?