/[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 459 by ph10, Sun Oct 4 09:21:39 2009 UTC code/branches/pcre16/pcre_compile.c revision 789 by zherczeg, Wed Dec 7 14:36:26 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2009 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #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 118  static const short int escapes[] = { Line 132  static const short int escapes[] = {
132       -ESC_H,                  0,       -ESC_H,                  0,
133       0,                       -ESC_K,       0,                       -ESC_K,
134       0,                       0,       0,                       0,
135       0,                       0,       -ESC_N,                  0,
136       -ESC_P,                  -ESC_Q,       -ESC_P,                  -ESC_Q,
137       -ESC_R,                  -ESC_S,       -ESC_R,                  -ESC_S,
138       0,                       0,       0,                       0,
# Line 165  static const short int escapes[] = { Line 179  static const short int escapes[] = {
179  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
180  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
181  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
182  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
183  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
184  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
185  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 182  string is built from string macros so th Line 196  string is built from string macros so th
196  platforms. */  platforms. */
197    
198  typedef struct verbitem {  typedef struct verbitem {
199    int   len;    int   len;                 /* Length of verb name */
200    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
201      int   op_arg;              /* Op when arg present, or -1 if not allowed */
202  } verbitem;  } verbitem;
203    
204  static const char verbnames[] =  static const char verbnames[] =
205      "\0"                       /* Empty name is a shorthand for MARK */
206      STRING_MARK0
207    STRING_ACCEPT0    STRING_ACCEPT0
208    STRING_COMMIT0    STRING_COMMIT0
209    STRING_F0    STRING_F0
# Line 196  static const char verbnames[] = Line 213  static const char verbnames[] =
213    STRING_THEN;    STRING_THEN;
214    
215  static const verbitem verbs[] = {  static const verbitem verbs[] = {
216    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
217    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
218    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
219    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
220    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
221    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
222    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
223      { 4, OP_SKIP,   OP_SKIP_ARG  },
224      { 4, OP_THEN,   OP_THEN_ARG  }
225  };  };
226    
227  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 220  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 250  static const int posix_class_maps[] = { Line 269  static const int posix_class_maps[] = {
269    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
270  };  };
271    
272    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
273    substitutes must be in the order of the names, defined above, and there are
274    both positive and negative cases. NULL means no substitute. */
275    
276    #ifdef SUPPORT_UCP
277    static const pcre_uchar string_PNd[]  = {
278      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
279      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
280    static const pcre_uchar string_pNd[]  = {
281      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
282      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
283    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 pcre_uchar string_pL[] =   {
306      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
307      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
308    static const pcre_uchar string_pLl[] =  {
309      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
310      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
311    static const pcre_uchar string_pLu[] =  {
312      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
313      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
314    static const pcre_uchar string_pXan[] = {
315      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
316      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
317    static const pcre_uchar string_h[] =    {
318      CHAR_BACKSLASH, CHAR_h, '\0' };
319    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 */
356      string_PL,            /* ^alpha */
357      string_PLl,           /* ^lower */
358      string_PLu,           /* ^upper */
359      string_PXan,          /* ^alnum */
360      NULL,                 /* ^ascii */
361      string_H,             /* ^blank */
362      NULL,                 /* ^cntrl */
363      string_PNd,           /* ^digit */
364      NULL,                 /* ^graph */
365      NULL,                 /* ^print */
366      NULL,                 /* ^punct */
367      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
368      string_PXwd,          /* ^word */
369      NULL                  /* ^xdigit */
370    };
371    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
372    #endif
373    
374  #define STRING(a)  # a  #define STRING(a)  # a
375  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 262  the number of relocations needed when a Line 382  the number of relocations needed when a
382  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
383  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
384  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
385  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
386    
387    Each substring ends with \0 to insert a null character. This includes the final
388    substring, so that the whole string ends with \0\0, which can be detected when
389    counting through. */
390    
391  static const char error_texts[] =  static const char error_texts[] =
392    "no error\0"    "no error\0"
# Line 309  static const char error_texts[] = Line 433  static const char error_texts[] =
433    /* 35 */    /* 35 */
434    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
435    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
436    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
437    "number after (?C is > 255\0"    "number after (?C is > 255\0"
438    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
439    /* 40 */    /* 40 */
# Line 331  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"
462    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
463    /* 60 */    /* 60 */
464    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
465    "number is too big\0"    "number is too big\0"
# Line 343  static const char error_texts[] = Line 467  static const char error_texts[] =
467    "digit expected after (?+\0"    "digit expected after (?+\0"
468    "] is an invalid data character in JavaScript compatibility mode\0"    "] is an invalid data character in JavaScript compatibility mode\0"
469    /* 65 */    /* 65 */
470    "different names for subpatterns of the same number are not allowed";    "different names for subpatterns of the same number are not allowed\0"
471      "(*MARK) must have an argument\0"
472      "this version of PCRE is not compiled with PCRE_UCP support\0"
473      "\\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
481  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 362  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 406  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 441  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 480  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, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
621      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
622    
623    
# Line 502  static const char * Line 639  static const char *
639  find_error_text(int n)  find_error_text(int n)
640  {  {
641  const char *s = error_texts;  const char *s = error_texts;
642  for (; n > 0; n--) while (*s++ != 0) {};  for (; n > 0; n--)
643      {
644      while (*s++ != 0) {};
645      if (*s == 0) return "Error text not found (please report)";
646      }
647  return s;  return s;
648  }  }
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 532  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 551  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 563  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 573  else Line 753  else
753    
754      case CHAR_l:      case CHAR_l:
755      case CHAR_L:      case CHAR_L:
756      case CHAR_N:      *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 596  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 606  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 626  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 679  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 716  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 743  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 759  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 774  else Line 1040  else
1040      break;      break;
1041    
1042      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1043      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
1044        coding is ASCII-specific, but then the whole concept of \cx is
1045      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1046    
1047      case CHAR_c:      case CHAR_c:
# Line 784  else Line 1051  else
1051        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1052        break;        break;
1053        }        }
1054    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1055  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1056          {
1057          *errorcodeptr = ERR68;
1058          break;
1059          }
1060      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1061      c ^= 0x40;      c ^= 0x40;
1062  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1063      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1064      c ^= 0xC0;      c ^= 0xC0;
1065  #endif  #endif
# Line 811  else Line 1082  else
1082      }      }
1083    }    }
1084    
1085    /* Perl supports \N{name} for character names, as well as plain \N for "not
1086    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 &&
1090         !is_counted_repeat(ptr+2))
1091      *errorcodeptr = ERR37;
1092    
1093    /* If PCRE_UCP is set, we change the values for \d etc. */
1094    
1095    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1096      c -= (ESC_DU - ESC_D);
1097    
1098    /* Set the pointer to the final character before returning. */
1099    
1100  *ptrptr = ptr;  *ptrptr = ptr;
1101  return c;  return c;
1102  }  }
# Line 837  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 882  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 911  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 962  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 971  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 986  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 1019  top-level call starts at the beginning o Line 1272  top-level call starts at the beginning o
1272  start at a parenthesis. It scans along a pattern's text looking for capturing  start at a parenthesis. It scans along a pattern's text looking for capturing
1273  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1274  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1275  returns when it reaches a given numbered subpattern. We know that if (?P< is  returns when it reaches a given numbered subpattern. Recursion is used to keep
1276  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1277  first pass. Recursion is used to keep track of subpatterns that reset the  
1278  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1279    that if (?< or (?' or (?P< is encountered, the name will be correctly
1280    terminated because that is checked in the first pass. There is now one call to
1281    this function in the first pass, to check for a recursive back reference by
1282    name (so that we can make the whole group atomic). In this case, we need check
1283    only up to the current position in the pattern, and that is still OK because
1284    and previous occurrences will have been checked. To make this work, the test
1285    for "end of pattern" is a check against cd->end_pattern in the main loop,
1286    instead of looking for a binary zero. This means that the special first-pass
1287    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1288    processing items within the loop are OK, because afterwards the main loop will
1289    terminate.)
1290    
1291  Arguments:  Arguments:
1292    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1030  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      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, 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 1049  dealing with. The very first call may no Line 1314  dealing with. The very first call may no
1314    
1315  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1316    {    {
1317    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1318        ptr[2] == CHAR_VERTICAL_LINE)  
1319      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1320    
1321      /* Handle a normal, unnamed capturing parenthesis. */
1322    
1323      else if (ptr[1] != CHAR_QUESTION_MARK)
1324        {
1325        *count += 1;
1326        if (name == NULL && *count == lorn) return *count;
1327        ptr++;
1328        }
1329    
1330      /* All cases now have (? at the start. Remember when we are in a group
1331      where the parenthesis numbers are duplicated. */
1332    
1333      else if (ptr[2] == CHAR_VERTICAL_LINE)
1334      {      {
1335      ptr += 3;      ptr += 3;
1336      dup_parens = TRUE;      dup_parens = TRUE;
1337      }      }
1338    
1339    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1340    
1341    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1342      {      {
1343      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1344      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1345      }      }
1346    
1347    /* Handle a condition. If it is an assertion, just carry on so that it    /* Handle a condition. If it is an assertion, just carry on so that it
1348    is processed as normal. If not, skip to the closing parenthesis of the    is processed as normal. If not, skip to the closing parenthesis of the
1349    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1350    
1351    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1352      {      {
# Line 1079  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1358  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1358        }        }
1359      }      }
1360    
1361    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1362    
1363    else    else
1364      {      {
# Line 1092  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 1100  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        }        }
1386      }      }
1387    }    }
1388    
1389  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1390  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1391    first-pass call when this value is temporarily adjusted to stop at the current
1392    position. So DO NOT change this to a test for binary zero. */
1393    
1394  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1395    {    {
1396    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1397    
# Line 1141  for (; *ptr != 0; ptr++) Line 1422  for (; *ptr != 0; 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
1429            break;            break;
1430          }          }
1431        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1432          {          {
1433          negate_class = TRUE;          negate_class = TRUE;
1434          ptr++;          ptr++;
1435          }          }
1436        else break;        else break;
1437        }        }
1438    
# Line 1184  for (; *ptr != 0; ptr++) Line 1465  for (; *ptr != 0; ptr++)
1465    
1466    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1467      {      {
1468      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1469        while (*ptr != 0)
1470          {
1471          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1472          ptr++;
1473    #ifdef SUPPORT_UTF
1474          if (utf) FORWARDCHAR(ptr);
1475    #endif
1476          }
1477      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1478      continue;      continue;
1479      }      }
# Line 1193  for (; *ptr != 0; ptr++) Line 1482  for (; *ptr != 0; 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, 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 1201  for (; *ptr != 0; ptr++) Line 1490  for (; *ptr != 0; ptr++)
1490    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1491      {      {
1492      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1493      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1494      }      }
1495    
1496    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1240  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      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 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 1258  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, &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 1274  return rc; Line 1564  return rc;
1564    
1565  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1566  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1567  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1568  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1569  assertions, and also the \b assertion; for others it does not.  does not.
1570    
1571  Arguments:  Arguments:
1572    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1573    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1574    
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, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1580  {  {
1581  for (;;)  for (;;)
1582    {    {
1583    switch ((int)*code)    switch ((int)*code)
1584      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1585      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1586      case OP_ASSERTBACK:      case OP_ASSERTBACK:
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 1321  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 1340  for (;;) Line 1620  for (;;)
1620    
1621  /* Scan a branch and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1622  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1623  In UTF8 mode, the result is in characters rather than bytes. The branch is  In UTF8 mode, the result is in characters rather than bytes. The branch is
1624  temporarily terminated with OP_END when this function is called.  temporarily terminated with OP_END when this function is called.
1625    
1626  This function is called when a backward assertion is encountered, so that if it  This function is called when a backward assertion is encountered, so that if it
1627  fails, the error message can point to the correct place in the pattern.  fails, the error message can point to the correct place in the pattern.
1628  However, we cannot do this when the assertion contains subroutine calls,  However, we cannot do this when the assertion contains subroutine calls,
1629  because they can be forward references. We solve this by remembering this case  because they can be forward references. We solve this by remembering this case
1630  and doing the check at the end; a flag specifies which mode we are running in.  and doing the check at the end; a flag specifies which mode we are running in.
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    options  the compiling options    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, int options, 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 1375  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
1664        OP_BRA (normal non-capturing bracket) because the other variants of these
1665        opcodes are all concerned with unlimited repeated groups, which of course
1666        are not of fixed length. */
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), options, 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        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1684        because they all imply an unlimited repeat. */
1685    
1686      case OP_ALT:      case OP_ALT:
1687      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
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;
1694      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1695      branchlength = 0;      branchlength = 0;
1696      break;      break;
1697    
1698      /* A true recursion implies not fixed length, but a subroutine call may      /* A true recursion implies not fixed length, but a subroutine call may
1699      be OK. If the subroutine is a forward reference, we can't deal with      be OK. If the subroutine is a forward reference, we can't deal with
1700      it until the end of the pattern, so return -3. */      it until the end of the pattern, so return -3. */
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, options, 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;
1711      break;      break;
1712    
1713      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1714    
# Line 1432  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_PRUNE_ARG:
1726        case OP_SKIP_ARG:
1727        case OP_THEN_ARG:
1728        cc += cc[1] + PRIV(OP_lengths)[*cc];
1729        break;
1730    
1731        case OP_CALLOUT:
1732        case OP_CIRC:
1733        case OP_CIRCM:
1734        case OP_CLOSE:
1735        case OP_COMMIT:
1736      case OP_CREF:      case OP_CREF:
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
1737      case OP_DEF:      case OP_DEF:
1738      case OP_OPT:      case OP_DOLL:
1739      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1740      case OP_EOD:      case OP_EOD:
1741      case OP_EODN:      case OP_EODN:
1742      case OP_CIRC:      case OP_FAIL:
1743      case OP_DOLL:      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 */
1759    
1760      case OP_CHAR:      case OP_CHAR:
1761      case OP_CHARNC:      case OP_CHARI:
1762      case OP_NOT:      case OP_NOT:
1763        case OP_NOTI:
1764      branchlength++;      branchlength++;
1765      cc += 2;      cc += 2;
1766  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1767      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1768  #endif  #endif
1769      break;      break;
1770    
# Line 1468  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 ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
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 1489  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 1501  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 1528  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 1540  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 1556  for (;;) Line 1952  for (;;)
1952    
1953  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1954  capturing bracket with the given number, or, if the number is negative, an  capturing bracket with the given number, or, if the number is negative, an
1955  instance of OP_REVERSE for a lookbehind. The function is global in the C sense  instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1956  so that it can be called from pcre_study() when finding the minimum matching  so that it can be called from pcre_study() when finding the minimum matching
1957  length.  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 1581  for (;;) Line 1978  for (;;)
1978    the table is zero; the actual length is stored in the compiled code. */    the table is zero; the actual length is stored in the compiled code. */
1979    
1980    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1981    
1982    /* Handle recursion */    /* Handle recursion */
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 */
1991    
1992    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1993               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
2001    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
2002    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2003      must add in its length. */
2004    
2005    else    else
2006      {      {
# Line 1623  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;
2028    
2029          case OP_MARK:
2030          case OP_PRUNE_ARG:
2031          case OP_SKIP_ARG:
2032          code += code[1];
2033          break;
2034    
2035          case OP_THEN_ARG:
2036          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_CHARNC:        case OP_CHARI:
2053        case OP_EXACT:        case OP_EXACT:
2054          case OP_EXACTI:
2055        case OP_UPTO:        case OP_UPTO:
2056          case OP_UPTOI:
2057        case OP_MINUPTO:        case OP_MINUPTO:
2058          case OP_MINUPTOI:
2059        case OP_POSUPTO:        case OP_POSUPTO:
2060          case OP_POSUPTOI:
2061        case OP_STAR:        case OP_STAR:
2062          case OP_STARI:
2063        case OP_MINSTAR:        case OP_MINSTAR:
2064          case OP_MINSTARI:
2065        case OP_POSSTAR:        case OP_POSSTAR:
2066          case OP_POSSTARI:
2067        case OP_PLUS:        case OP_PLUS:
2068          case OP_PLUSI:
2069        case OP_MINPLUS:        case OP_MINPLUS:
2070          case OP_MINPLUSI:
2071        case OP_POSPLUS:        case OP_POSPLUS:
2072          case OP_POSPLUSI:
2073        case OP_QUERY:        case OP_QUERY:
2074          case OP_QUERYI:
2075        case OP_MINQUERY:        case OP_MINQUERY:
2076          case OP_MINQUERYI:
2077        case OP_POSQUERY:        case OP_POSQUERY:
2078        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2079          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 1674  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 1696  for (;;) Line 2119  for (;;)
2119    
2120    /* 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
2121    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
2122    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2123      must add in its length. */
2124    
2125    else    else
2126      {      {
# Line 1718  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;
2148    
2149          case OP_MARK:
2150          case OP_PRUNE_ARG:
2151          case OP_SKIP_ARG:
2152          code += code[1];
2153          break;
2154    
2155          case OP_THEN_ARG:
2156          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_CHARNC:        case OP_CHARI:
2173        case OP_EXACT:        case OP_EXACT:
2174          case OP_EXACTI:
2175        case OP_UPTO:        case OP_UPTO:
2176          case OP_UPTOI:
2177        case OP_MINUPTO:        case OP_MINUPTO:
2178          case OP_MINUPTOI:
2179        case OP_POSUPTO:        case OP_POSUPTO:
2180          case OP_POSUPTOI:
2181        case OP_STAR:        case OP_STAR:
2182          case OP_STARI:
2183        case OP_MINSTAR:        case OP_MINSTAR:
2184          case OP_MINSTARI:
2185        case OP_POSSTAR:        case OP_POSSTAR:
2186          case OP_POSSTARI:
2187        case OP_PLUS:        case OP_PLUS:
2188          case OP_PLUSI:
2189        case OP_MINPLUS:        case OP_MINPLUS:
2190          case OP_MINPLUSI:
2191        case OP_POSPLUS:        case OP_POSPLUS:
2192          case OP_POSPLUSI:
2193        case OP_QUERY:        case OP_QUERY:
2194          case OP_QUERYI:
2195        case OP_MINQUERY:        case OP_MINQUERY:
2196          case OP_MINQUERYI:
2197        case OP_POSQUERY:        case OP_POSQUERY:
2198        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2199          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 1775  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.
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      BOOL utf, compile_data *cd)
2235  {  {
2236  register int c;  register int c;
2237  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, 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], NULL, 0, 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 1802  for (code = first_significant_code(code Line 2252  for (code = first_significant_code(code
2252      continue;      continue;
2253      }      }
2254    
2255      /* For a recursion/subroutine call, if its end has been reached, which
2256      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)
2263        {
2264        const pcre_uchar *scode;
2265        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 */
2277    
2278        /* Completed backwards reference */
2279    
2280        do
2281          {
2282          if (could_be_empty_branch(scode, endcode, utf, cd))
2283            {
2284            empty_branch = TRUE;
2285            break;
2286            }
2287          scode += GET(scode, 1);
2288          }
2289        while (*scode == OP_ALT);
2290    
2291        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2292        continue;
2293        }
2294    
2295    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2296    
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)
2299        {
2300        code += PRIV(OP_lengths)[c];
2301        do code += GET(code, 1); while (*code == OP_ALT);
2302        c = *code;
2303        continue;
2304        }
2305    
2306      /* A nested group that is already marked as "could be empty" can just be
2307      skipped. */
2308    
2309      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2310          c == OP_SCBRA || c == OP_SCBRAPOS)
2311      {      {
     code += _pcre_OP_lengths[c];  
2312      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2313      c = *code;      c = *code;
2314      continue;      continue;
# Line 1814  for (code = first_significant_code(code Line 2316  for (code = first_significant_code(code
2316    
2317    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2318    
2319    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2320          c == OP_CBRA || c == OP_CBRAPOS ||
2321          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 1830  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))          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 1848  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 1860  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 1901  for (code = first_significant_code(code Line 2406  for (code = first_significant_code(code
2406      case OP_ALLANY:      case OP_ALLANY:
2407      case OP_ANYBYTE:      case OP_ANYBYTE:
2408      case OP_CHAR:      case OP_CHAR:
2409      case OP_CHARNC:      case OP_CHARI:
2410      case OP_NOT:      case OP_NOT:
2411        case OP_NOTI:
2412      case OP_PLUS:      case OP_PLUS:
2413      case OP_MINPLUS:      case OP_MINPLUS:
2414      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1934  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 1942  for (code = first_significant_code(code Line 2449  for (code = first_significant_code(code
2449      case OP_KET:      case OP_KET:
2450      case OP_KETRMAX:      case OP_KETRMAX:
2451      case OP_KETRMIN:      case OP_KETRMIN:
2452        case OP_KETRPOS:
2453      case OP_ALT:      case OP_ALT:
2454      return TRUE;      return TRUE;
2455    
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:
2462      case OP_MINSTAR:      case OP_MINSTAR:
2463        case OP_MINSTARI:
2464      case OP_POSSTAR:      case OP_POSSTAR:
2465        case OP_POSSTARI:
2466      case OP_QUERY:      case OP_QUERY:
2467        case OP_QUERYI:
2468      case OP_MINQUERY:      case OP_MINQUERY:
2469        case OP_MINQUERYI:
2470      case OP_POSQUERY:      case OP_POSQUERY:
2471      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      case OP_POSQUERYI:
2472        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2473      break;      break;
2474    
2475      case OP_UPTO:      case OP_UPTO:
2476        case OP_UPTOI:
2477      case OP_MINUPTO:      case OP_MINUPTO:
2478        case OP_MINUPTOI:
2479      case OP_POSUPTO:      case OP_POSUPTO:
2480      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      case OP_POSUPTOI:
2481        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2482      break;      break;
2483  #endif  #endif
2484    
2485        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2486        string. */
2487    
2488        case OP_MARK:
2489        case OP_PRUNE_ARG:
2490        case OP_SKIP_ARG:
2491        code += code[1];
2492        break;
2493    
2494        case OP_THEN_ARG:
2495        code += code[1];
2496        break;
2497    
2498        /* None of the remaining opcodes are required to match a character. */
2499    
2500        default:
2501        break;
2502      }      }
2503    }    }
2504    
# Line 1980  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
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)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2534  {  {
2535  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2536    {    {
2537    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2538        return FALSE;
2539    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2540    }    }
2541  return TRUE;  return TRUE;
# Line 2028  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 2036  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 2073  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 2112  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 2120  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 2173  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, 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 2202  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 = 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);
2766  }  }
2767    
# Line 2254  for (++c; c <= d; c++) Line 2811  for (++c; c <= d; c++)
2811    
2812  return TRUE;  return TRUE;
2813  }  }
2814    
2815    
2816    
2817    /*************************************************
2818    *        Check a character and a property        *
2819    *************************************************/
2820    
2821    /* This function is called by check_auto_possessive() when a property item
2822    is adjacent to a fixed character.
2823    
2824    Arguments:
2825      c            the character
2826      ptype        the property type
2827      pdata        the data for the type
2828      negated      TRUE if it's a negated property (\P or \p{^)
2829    
2830    Returns:       TRUE if auto-possessifying is OK
2831    */
2832    
2833    static BOOL
2834    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2835    {
2836    const ucd_record *prop = GET_UCD(c);
2837    switch(ptype)
2838      {
2839      case PT_LAMP:
2840      return (prop->chartype == ucp_Lu ||
2841              prop->chartype == ucp_Ll ||
2842              prop->chartype == ucp_Lt) == negated;
2843    
2844      case PT_GC:
2845      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2846    
2847      case PT_PC:
2848      return (pdata == prop->chartype) == negated;
2849    
2850      case PT_SC:
2851      return (pdata == prop->script) == negated;
2852    
2853      /* These are specials */
2854    
2855      case PT_ALNUM:
2856      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2857              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2858    
2859      case PT_SPACE:    /* Perl space */
2860      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2861              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2862              == negated;
2863    
2864      case PT_PXSPACE:  /* POSIX space */
2865      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2866              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2867              c == CHAR_FF || c == CHAR_CR)
2868              == negated;
2869    
2870      case PT_WORD:
2871      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2872              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2873              c == CHAR_UNDERSCORE) == negated;
2874      }
2875    return FALSE;
2876    }
2877  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2878    
2879    
# Line 2267  whether the next thing could possibly ma Line 2887  whether the next thing could possibly ma
2887  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2888    
2889  Arguments:  Arguments:
2890    op_code       the repeated op code    previous      pointer to the repeated opcode
2891    this          data for this item, depends on the opcode    utf           TRUE in UTF-8 / UTF-16 mode
   utf8          TRUE in UTF-8 mode  
   utf8_char     used for utf8 character bytes, NULL if not relevant  
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 2279  Returns:        TRUE if possessifying is Line 2897  Returns:        TRUE if possessifying is
2897  */  */
2898    
2899  static BOOL  static BOOL
2900  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const pcre_uchar *previous, BOOL utf,
2901    const uschar *ptr, int options, compile_data *cd)    const pcre_uchar *ptr, int options, compile_data *cd)
2902  {  {
2903  int next;  pcre_int32 c, next;
2904    int op_code = *previous++;
2905    
2906  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2907    
# Line 2293  if ((options & PCRE_EXTENDED) != 0) Line 2912  if ((options & PCRE_EXTENDED) != 0)
2912      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2913      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2914        {        {
2915        while (*(++ptr) != 0)        ptr++;
2916          while (*ptr != 0)
2917            {
2918          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2919            ptr++;
2920    #ifdef SUPPORT_UTF
2921            if (utf) FORWARDCHAR(ptr);
2922    #endif
2923            }
2924        }        }
2925      else break;      else break;
2926      }      }
# Line 2310  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 2330  if ((options & PCRE_EXTENDED) != 0) Line 2954  if ((options & PCRE_EXTENDED) != 0)
2954      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2955      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2956        {        {
2957        while (*(++ptr) != 0)        ptr++;
2958          while (*ptr != 0)
2959            {
2960          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2961            ptr++;
2962    #ifdef SUPPORT_UTF
2963            if (utf) FORWARDCHAR(ptr);
2964    #endif
2965            }
2966        }        }
2967      else break;      else break;
2968      }      }
# Line 2340  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. If the previous is a  /* Now compare the next item with the previous opcode. First, handle cases when
2978  positive single character match, "item" either contains the character or, if  the next item is a character. */
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
   
 /* Handle cases when the next item is a character. */  
2979    
2980  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2981    {    {
2982    case OP_CHAR:    case OP_CHAR:
2983  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2984    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2985  #else  #else
2986    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2987  #endif  #endif
2988    return item != next;    return c != next;
2989    
2990    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2991    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2992    high-valued characters. */    high-valued characters. */
2993    
2994    case OP_CHARNC:    case OP_CHARI:
2995  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2996    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2997    #else
2998      c = *previous;
2999  #endif  #endif
3000    if (item == 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 2380  if (next >= 0) switch(op_code) Line 3008  if (next >= 0) switch(op_code)
3008  #else  #else
3009      othercase = NOTACHAR;      othercase = NOTACHAR;
3010  #endif  #endif
3011      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
3012      }      }
3013    else    else
3014  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
3015    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
3016    
3017    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
3018      opcodes are not used for multi-byte characters, because they are coded using
3019      an XCLASS instead. */
3020    
3021    case OP_NOT:    case OP_NOT:
3022    if (item == next) return TRUE;    return (c = *previous) == next;
3023    if ((options & PCRE_CASELESS) == 0) return FALSE;  
3024      case OP_NOTI:
3025      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 2401  if (next >= 0) switch(op_code) Line 3033  if (next >= 0) switch(op_code)
3033  #else  #else
3034      othercase = NOTACHAR;      othercase = NOTACHAR;
3035  #endif  #endif
3036      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
3037      }      }
3038    else    else
3039  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
3040    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
3041    
3042      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3043      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3044    
3045    case OP_DIGIT:    case OP_DIGIT:
3046    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2448  if (next >= 0) switch(op_code) Line 3083  if (next >= 0) switch(op_code)
3083      case 0x202f:      case 0x202f:
3084      case 0x205f:      case 0x205f:
3085      case 0x3000:      case 0x3000:
3086      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
3087      default:      default:
3088      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
3089      }      }
3090    
3091      case OP_ANYNL:
3092    case OP_VSPACE:    case OP_VSPACE:
3093    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3094    switch(next)    switch(next)
# Line 2464  if (next >= 0) switch(op_code) Line 3100  if (next >= 0) switch(op_code)
3100      case 0x85:      case 0x85:
3101      case 0x2028:      case 0x2028:
3102      case 0x2029:      case 0x2029:
3103      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
3104      default:      default:
3105      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
3106      }      }
3107    
3108    #ifdef SUPPORT_UCP
3109      case OP_PROP:
3110      return check_char_prop(next, previous[0], previous[1], FALSE);
3111    
3112      case OP_NOTPROP:
3113      return check_char_prop(next, previous[0], previous[1], TRUE);
3114    #endif
3115    
3116    default:    default:
3117    return FALSE;    return FALSE;
3118    }    }
3119    
3120    
3121  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3122    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3123    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3124    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3125    replaced by OP_PROP codes when PCRE_UCP is set. */
3126    
3127  switch(op_code)  switch(op_code)
3128    {    {
3129    case OP_CHAR:    case OP_CHAR:
3130    case OP_CHARNC:    case OP_CHARI:
3131  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3132    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3133    #else
3134      c = *previous;
3135  #endif  #endif
3136    switch(-next)    switch(-next)
3137      {      {
3138      case ESC_d:      case ESC_d:
3139      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3140    
3141      case ESC_D:      case ESC_D:
3142      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3143    
3144      case ESC_s:      case ESC_s:
3145      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3146    
3147      case ESC_S:      case ESC_S:
3148      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3149    
3150      case ESC_w:      case ESC_w:
3151      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3152    
3153      case ESC_W:      case ESC_W:
3154      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3155    
3156      case ESC_h:      case ESC_h:
3157      case ESC_H:      case ESC_H:
3158      switch(item)      switch(c)
3159        {        {
3160        case 0x09:        case 0x09:
3161        case 0x20:        case 0x20:
# Line 2533  switch(op_code) Line 3183  switch(op_code)
3183    
3184      case ESC_v:      case ESC_v:
3185      case ESC_V:      case ESC_V:
3186      switch(item)      switch(c)
3187        {        {
3188        case 0x0a:        case 0x0a:
3189        case 0x0b:        case 0x0b:
# Line 2547  switch(op_code) Line 3197  switch(op_code)
3197        return -next == ESC_v;        return -next == ESC_v;
3198        }        }
3199    
3200        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3201        their substitutions and process them. The result will always be either
3202        -ESC_p or -ESC_P. Then fall through to process those values. */
3203    
3204    #ifdef SUPPORT_UCP
3205        case ESC_du:
3206        case ESC_DU:
3207        case ESC_wu:
3208        case ESC_WU:
3209        case ESC_su:
3210        case ESC_SU:
3211          {
3212          int temperrorcode = 0;
3213          ptr = substitutes[-next - ESC_DU];
3214          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3215          if (temperrorcode != 0) return FALSE;
3216          ptr++;    /* For compatibility */
3217          }
3218        /* Fall through */
3219    
3220        case ESC_p:
3221        case ESC_P:
3222          {
3223          int ptype, pdata, errorcodeptr;
3224          BOOL negated;
3225    
3226          ptr--;      /* Make ptr point at the p or P */
3227          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3228          if (ptype < 0) return FALSE;
3229          ptr++;      /* Point past the final curly ket */
3230    
3231          /* If the property item is optional, we have to give up. (When generated
3232          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3233          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 ||
3236            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3237              return FALSE;
3238    
3239          /* Do the property check. */
3240    
3241          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3242          }
3243    #endif
3244    
3245      default:      default:
3246      return FALSE;      return FALSE;
3247      }      }
3248    
3249      /* In principle, support for Unicode properties should be integrated here as
3250      well. It means re-organizing the above code so as to get hold of the property
3251      values before switching on the op-code. However, I wonder how many patterns
3252      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3253      these op-codes are never generated.) */
3254    
3255    case OP_DIGIT:    case OP_DIGIT:
3256    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3257           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3258    
3259    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3260    return next == -ESC_d;    return next == -ESC_d;
3261    
3262    case OP_WHITESPACE:    case OP_WHITESPACE:
3263    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3264    
3265    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3266    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3267    
3268    case OP_HSPACE:    case OP_HSPACE:
3269    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3270             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3271    
3272    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3273    return next == -ESC_h;    return next == -ESC_h;
3274    
3275    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3276      case OP_ANYNL:
3277    case OP_VSPACE:    case OP_VSPACE:
3278    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3279    
3280    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3281    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3282    
3283    case OP_WORDCHAR:    case OP_WORDCHAR:
3284    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3285             next == -ESC_v || next == -ESC_R;
3286    
3287    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3288    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2607  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 2619  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  uschar *previous = NULL;  const pcre_uchar *nestptr = NULL;
3352  uschar *previous_callout = NULL;  pcre_uchar *previous = NULL;
3353  uschar *save_hwm = NULL;  pcre_uchar *previous_callout = NULL;
3354  uschar classbits[32];  pcre_uchar *save_hwm = NULL;
3355    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 DEBUG  #ifdef PCRE_DEBUG
3378  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3379  #endif  #endif
3380    
# Line 2669  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 2702  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    
3433      /* If we are at the end of a nested substitution, revert to the outer level
3434      string. Nesting only happens one level deep. */
3435    
3436      if (c == 0 && nestptr != NULL)
3437        {
3438        ptr = nestptr;
3439        nestptr = NULL;
3440        c = *ptr;
3441        }
3442    
3443    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
3444    previous cycle of this loop. */    previous cycle of this loop. */
3445    
3446    if (lengthptr != NULL)    if (lengthptr != NULL)
3447      {      {
3448  #ifdef DEBUG  #ifdef PCRE_DEBUG
3449      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3450  #endif  #endif
3451      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3452        {        {
3453        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3454        goto FAILED;        goto FAILED;
# Line 2742  for (;; ptr++) Line 3470  for (;; ptr++)
3470        goto FAILED;        goto FAILED;
3471        }        }
3472    
3473      *lengthptr += 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 2753  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 2769  for (;; ptr++) Line 3498  for (;; ptr++)
3498    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3499    reference list. */    reference list. */
3500    
3501    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3502      {      {
3503      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3504      goto FAILED;      goto FAILED;
# Line 2817  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      {      {
3553      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3554      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3555        {        {
3556        while (*(++ptr) != 0)        ptr++;
3557          while (*ptr != 0)
3558          {          {
3559          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3560            ptr++;
3561    #ifdef SUPPORT_UTF
3562            if (utf) FORWARDCHAR(ptr);
3563    #endif
3564          }          }
3565        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3566    
# Line 2849  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 2860  for (;; ptr++) Line 3594  for (;; ptr++)
3594          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3595          goto FAILED;          goto FAILED;
3596          }          }
3597        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3598        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3599        }        }
3600      return TRUE;      return TRUE;
# Line 2871  for (;; ptr++) Line 3605  for (;; ptr++)
3605      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3606    
3607      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3608        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;
3613        }        }
3614      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3615      break;      break;
3616    
3617      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3618      previous = NULL;      previous = NULL;
3619      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
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 2945  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 2965  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 2988  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 3002  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 3048  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 3065  for (;; ptr++) Line 3800  for (;; ptr++)
3800            ptr++;            ptr++;
3801            }            }
3802    
3803          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3804          if (posix_class < 0)          if (posix_class < 0)
3805            {            {
3806            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3079  for (;; ptr++) Line 3814  for (;; ptr++)
3814          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3815            posix_class = 0;            posix_class = 0;
3816    
3817          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3818          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3819          subtract bits that may be in the main map already. At the end we or the  
3820          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3821            if ((options & PCRE_UCP) != 0)
3822              {
3823              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3824              if (posix_substitutes[pc] != NULL)
3825                {
3826                nestptr = tempptr + 1;
3827                ptr = posix_substitutes[pc] - 1;
3828                continue;
3829                }
3830              }
3831    #endif
3832            /* In the non-UCP case, we build the bit map for the POSIX class in a
3833            chunk of local store because we may be adding and subtracting from it,
3834            and we don't want to subtract bits that may be in the main map already.
3835            At the end we or the result into the bit map that is being built. */
3836    
3837          posix_class *= 3;          posix_class *= 3;
3838    
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 3126  for (;; ptr++) Line 3876  for (;; ptr++)
3876    
3877        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3878        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3879        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3880        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so set
3881        to 'or' into the one we are building. We assume they have more than one        class_charcount bigger than one. Unrecognized escapes fall through and
3882        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3883          PCRE_EXTRA is set. */
3884    
3885        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3886          {          {
3887          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3888          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3889    
3890          if (-c == ESC_b) c = CHAR_BS;       /* \b is backspace in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
         else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */  
3891          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3892            {            {
3893            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3152  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            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3908              {              {
3909    #ifdef SUPPORT_UCP
3910                case ESC_du:     /* These are the values given for \d etc */
3911                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3912                case ESC_wu:     /* escape sequence with an appropriate \p */
3913                case ESC_WU:     /* or \P to test Unicode properties instead */
3914                case ESC_su:     /* of the default ASCII testing. */
3915                case ESC_SU:
3916                nestptr = ptr;
3917                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3918                class_charcount -= 2;                /* Undo! */
3919                continue;
3920    #endif
3921              case ESC_d:              case ESC_d:
3922              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3923              continue;              continue;
# Line 3177  for (;; ptr++) Line 3936  for (;; ptr++)
3936              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3937              continue;              continue;
3938    
3939                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3940                if it was previously set by something earlier in the character
3941                class. */
3942    
3943              case ESC_s:              case ESC_s:
3944              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3945              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3946                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3947              continue;              continue;
3948    
3949              case ESC_S:              case ESC_S:
# Line 3188  for (;; ptr++) Line 3952  for (;; ptr++)
3952              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3953              continue;              continue;
3954    
3955              default:    /* Not recognized; fall through */              case ESC_h:
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
           /* We need to deal with \H, \h, \V, and \v in both phases because  
           they use extra memory. */  
   
           if (-c == ESC_h)  
             {  
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;
             }  
3979    
3980            if (-c == ESC_H)              case ESC_H:
             {  
3981              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3982                {                {
3983                int x = 0xff;                int x = 0xff;
# Line 3242  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;
             }  
4022    
4023            if (-c == ESC_v)              case ESC_v:
             {  
4024              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
4025              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
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;
             }  
4039    
4040            if (-c == ESC_V)              case ESC_V:
             {  
4041              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
4042                {                {
4043                int x = 0xff;                int x = 0xff;
# Line 3309  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;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
4070    
4071  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4072            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
4073              {              case ESC_P:
4074              BOOL negated;                {
4075              int pdata;                BOOL negated;
4076              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
4077              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
4078              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
4079              *class_utf8data++ = ((-c == ESC_p) != negated)?                xclass = TRUE;
4080                XCL_PROP : XCL_NOTPROP;                *class_uchardata++ = ((-c == ESC_p) != negated)?
4081              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
4082              *class_utf8data++ = pdata;                *class_uchardata++ = ptype;
4083              class_charcount -= 2;   /* Not a < 256 character */                *class_uchardata++ = pdata;
4084              continue;                class_charcount -= 2;   /* Not a < 256 character */
4085              }                continue;
4086                  }
4087  #endif  #endif
4088            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
4089            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
4090            treated as literals. */              treated as literals. */
4091    
4092            if ((options & PCRE_EXTRA) != 0)              default:
4093              {              if ((options & PCRE_EXTRA) != 0)
4094              *errorcodeptr = ERR7;                {
4095              goto FAILED;                *errorcodeptr = ERR7;
4096                  goto FAILED;
4097                  }
4098                class_charcount -= 2;  /* Undo the default count from above */
4099                c = *ptr;              /* Get the final character and fall through */
4100                break;
4101              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
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 3406  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 3423  for (;; ptr++) Line 4168  for (;; ptr++)
4168            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
4169            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
4170    
4171            /* \b is backspace; \X is literal X; \R is literal R; any other            /* \b is backspace; any other special means the '-' was literal */
           special means the '-' was literal */  
4172    
4173            if (d < 0)            if (d < 0)
4174              {              {
4175              if (d == -ESC_b) d = CHAR_BS;              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = CHAR_X;  
             else if (d == -ESC_R) d = CHAR_R; else  
4176                {                {
4177                ptr = oldptr;                ptr = oldptr;
4178                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3458  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    #elif !(defined COMPILE_PCRE8)
4206            if (d > 255)
4207    #endif
4208    #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
4209            {            {
4210            class_utf8 = TRUE;            xclass = TRUE;
4211    
4212            /* With UCP support, we can find the other case equivalents of            /* With UCP support, we can find the other case equivalents of
4213            the relevant characters. There may be several ranges. Optimize how            the relevant characters. There may be several ranges. Optimize how
# Line 3494  for (;; ptr++) Line 4240  for (;; ptr++)
4240    
4241                if (occ == ocd)                if (occ == ocd)
4242                  {                  {
4243                  *class_utf8data++ = XCL_SINGLE;                  *class_uchardata++ = XCL_SINGLE;
4244                  }                  }
4245                else                else
4246                  {                  {
4247                  *class_utf8data++ = XCL_RANGE;                  *class_uchardata++ = XCL_RANGE;
4248                  class_utf8data += _pcre_ord2utf8(occ, class_utf8data);                  class_uchardata += PRIV(ord2utf)(occ, class_uchardata);
4249                  }                  }
4250                class_utf8data += _pcre_ord2utf8(ocd, class_utf8data);                class_uchardata += PRIV(ord2utf)(ocd, class_uchardata);
4251                }                }
4252              }              }
4253  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
# Line 3509  for (;; ptr++) Line 4255  for (;; ptr++)
4255            /* Now record the original range, possibly modified for UCP caseless            /* Now record the original range, possibly modified for UCP caseless
4256            overlapping ranges. */            overlapping ranges. */
4257    
4258            *class_utf8data++ = XCL_RANGE;            *class_uchardata++ = XCL_RANGE;
4259            class_utf8data += _pcre_ord2utf8(c, class_utf8data);  #ifdef SUPPORT_UTF
4260            class_utf8data += _pcre_ord2utf8(d, class_utf8data);            class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4261              class_uchardata += PRIV(ord2utf)(d, class_uchardata);
4262    #else
4263              *class_uchardata++ = c;
4264              *class_uchardata++ = d;
4265    #endif
4266    
4267            /* With UCP support, we are done. Without UCP support, there is no            /* With UCP support, we are done. Without UCP support, there is no
4268            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
4269            for the smaller ones. */            for the smaller ones. As for 16 bit characters without UTF, we
4270              can still use  */
4271    
4272  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4273            continue;    /* With next character in the class */            continue;    /* With next character in the class */
4274  #else  #else
4275    #ifdef SUPPORT_UTF
4276            if ((options & PCRE_CASELESS) == 0 || c > 127) continue;            if ((options & PCRE_CASELESS) == 0 || c > 127) continue;
   
4277            /* Adjust upper limit and fall through to set up the map */            /* Adjust upper limit and fall through to set up the map */
   
4278            d = 127;            d = 127;
4279    #else
4280              if (c > 255) continue;
4281              /* Adjust upper limit and fall through to set up the map */
4282              d = 255;
4283    #endif  /* SUPPORT_UTF */
4284  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4285            }            }
4286  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 || COMPILE_PCRE16 */
4287    
4288          /* 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
4289          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. */  
4290    
4291          class_charcount += d - c + 1;          class_charcount += d - c + 1;
4292          class_lastchar = d;          class_lastchar = d;
# Line 3560  for (;; ptr++) Line 4314  for (;; ptr++)
4314    
4315        /* Handle a character that cannot go in the bit map */        /* Handle a character that cannot go in the bit map */
4316    
4317  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4318        if (utf8 && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))        if (utf && (c > 255 || ((options & PCRE_CASELESS) != 0 && c > 127)))
4319          {  #elif !(defined COMPILE_PCRE8)
4320          class_utf8 = TRUE;        if (c > 255)
4321          *class_utf8data++ = XCL_SINGLE;  #endif
4322          class_utf8data += _pcre_ord2utf8(c, class_utf8data);  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
4323            {
4324            xclass = TRUE;
4325            *class_uchardata++ = XCL_SINGLE;
4326    #ifdef SUPPORT_UTF
4327            class_uchardata += PRIV(ord2utf)(c, class_uchardata);
4328    #else
4329            *class_uchardata++ = c;
4330    #endif
4331    
4332  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4333          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
# Line 3573  for (;; ptr++) Line 4335  for (;; ptr++)
4335            unsigned int othercase;            unsigned int othercase;
4336            if ((othercase = UCD_OTHERCASE(c)) != c)            if ((othercase = UCD_OTHERCASE(c)) != c)
4337              {              {
4338              *class_utf8data++ = XCL_SINGLE;              *class_uchardata++ = XCL_SINGLE;
4339              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_uchardata += PRIV(ord2utf)(othercase, class_uchardata);
4340              }              }
4341            }            }
4342  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4343    
4344          }          }
4345        else        else
4346  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF || COMPILE_PCRE16 */
   
4347        /* Handle a single-byte character */        /* Handle a single-byte character */
4348          {          {
4349          classbits[c/8] |= (1 << (c&7));          classbits[c/8] |= (1 << (c&7));
# Line 3594  for (;; ptr++) Line 4355  for (;; ptr++)
4355          class_charcount++;          class_charcount++;
4356          class_lastchar = c;          class_lastchar = c;
4357          }          }
4358    
4359        }        }
4360    
4361      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" far above.
4362        If we are at the end of an internal nested string, revert to the outer
4363        string. */
4364    
4365        while (((c = *(++ptr)) != 0 ||
4366               (nestptr != NULL &&
4367                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4368               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4369    
4370      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4371    
4372      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4373        {        {
4374        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4375        goto FAILED;        goto FAILED;
4376        }        }
4377    
   
 /* This code has been disabled because it would mean that \s counts as  
 an explicit \r or \n reference, and that's not really what is wanted. Now  
 we set the flag only if there is a literal "\r" or "\n" in the class. */  
   
 #if 0  
     /* Remember whether \r or \n are in this class */  
   
     if (negate_class)  
       {  
       if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;  
       }  
     else  
       {  
       if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;  
       }  
 #endif  
   
   
4378      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
4379      less than 256. As long as there were no characters >= 128 and there was no      less than 256. As long as there were no characters >= 128 and there was no
4380      use of \p or \P, in other words, no use of any XCLASS features, we can      use of \p or \P, in other words, no use of any XCLASS features, we can
# Line 3632  we set the flag only if there is a liter Line 4382  we set the flag only if there is a liter
4382    
4383      In UTF-8 mode, we can optimize the negative case only if there were no      In UTF-8 mode, we can optimize the negative case only if there were no
4384      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4385      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4386      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4387    
4388      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
4389      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4390      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4391      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstchar to be set. Otherwise, there can be no first
4392      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4393      reqbyte, save the previous value for reinstating. */      of reqchar, save the previous value for reinstating. */
4394    
4395  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4396      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !xclass &&
4397        (!utf8 || !negate_class || class_lastchar < 128))        (!utf || !negate_class || class_lastchar < 128))
4398  #else  #else
4399      if (class_charcount == 1)      if (class_charcount == 1)
4400  #endif  #endif
4401        {        {
4402        zeroreqbyte = reqbyte;        zeroreqchar = reqchar;
4403    
4404        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4405    
4406        if (negate_class)        if (negate_class)
4407          {          {
4408          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4409          zerofirstbyte = firstbyte;          zerofirstchar = firstchar;
4410          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4411          *code++ = class_lastchar;          *code++ = class_lastchar;
4412          break;          break;
4413          }          }
# Line 3666  we set the flag only if there is a liter Line 4416  we set the flag only if there is a liter
4416        then we can handle this with the normal one-character code. */        then we can handle this with the normal one-character code. */
4417    
4418  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4419        if (utf8 && class_lastchar > 127)        if (utf && class_lastchar > 127)
4420          mclength = _pcre_ord2utf8(class_lastchar, mcbuffer);          mclength = PRIV(ord2utf)(class_lastchar, mcbuffer);
4421        else        else
4422  #endif  #endif
4423          {          {
# Line 3679  we set the flag only if there is a liter Line 4429  we set the flag only if there is a liter
4429    
4430      /* 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
4431      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
4432      repeat count. Any reqbyte setting must remain unchanged after any kind of      repeat count. Any reqchar setting must remain unchanged after any kind of
4433      repeat. */      repeat. */
4434    
4435      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstchar == REQ_UNSET) firstchar = REQ_NONE;
4436      zerofirstbyte = firstbyte;      zerofirstchar = firstchar;
4437      zeroreqbyte = reqbyte;      zeroreqchar = reqchar;
4438    
4439      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4440      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4441      such as \S in the class, because in that case all characters > 255 are in      such as \S in the class, and PCRE_UCP is not set, because in that case all
4442      the class, so any that were explicitly given as well can be ignored. If      characters > 255 are in the class, so any that were explicitly given as
4443      (when there are explicit characters > 255 that must be listed) there are no      well can be ignored. If (when there are explicit characters > 255 that must
4444      characters < 256, we can omit the bitmap in the actual compiled code. */      be listed) there are no characters < 256, we can omit the bitmap in the
4445        actual compiled code. */
4446    
4447  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4448      if (class_utf8 && !should_flip_negation)      if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
4449    #endif
4450    #ifndef COMPILE_PCRE8
4451        if (xclass && !should_flip_negation)
4452    #endif
4453    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4454        {        {
4455        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_uchardata++ = XCL_END;    /* Marks the end of extra data */
4456        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
4457        code += LINK_SIZE;        code += LINK_SIZE;
4458        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT:0;
4459    
4460        /* 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;
4461        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 3707  we set the flag only if there is a liter Line 4463  we set the flag only if there is a liter
4463        if (class_charcount > 0)        if (class_charcount > 0)
4464          {          {
4465          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
4466          memmove(code + 32, code, class_utf8data - code);          memmove(code + (32 / sizeof(pcre_uchar)), code,
4467              IN_UCHARS(class_uchardata - code));
4468          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
4469          code = class_utf8data + 32;          code = class_uchardata + (32 / sizeof(pcre_uchar));
4470          }          }
4471        else code = class_utf8data;        else code = class_uchardata;
4472    
4473        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
4474    
# Line 3720  we set the flag only if there is a liter Line 4477  we set the flag only if there is a liter
4477        }        }
4478  #endif  #endif
4479    
4480      /* If there are no characters > 255, set the opcode to OP_CLASS or      /* If there are no characters > 255, or they are all to be included or
4481      OP_NCLASS, depending on whether the whole class was negated and whether      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4482      there were negative specials such as \S in the class. Then copy the 32-byte      whole class was negated and whether there were negative specials such as \S
4483      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4484        negating it if necessary. */
4485    
4486      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4487      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  
4488        {        {
4489          if (negate_class)
4490            for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
4491        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4492        }        }
4493      code += 32;      code += 32 / sizeof(pcre_uchar);
4494      break;      break;
4495    
4496    
# Line 3772  we set the flag only if there is a liter Line 4527  we set the flag only if there is a liter
4527    
4528      if (repeat_min == 0)      if (repeat_min == 0)
4529        {        {
4530        firstbyte = zerofirstbyte;    /* Adjust for zero repeat */        firstchar = zerofirstchar;    /* Adjust for zero repeat */
4531        reqbyte = zeroreqbyte;        /* Ditto */        reqchar = zeroreqchar;        /* Ditto */
4532        }        }
4533    
4534      /* Remember whether this is a variable length repeat */      /* Remember whether this is a variable length repeat */
# Line 3783  we set the flag only if there is a liter Line 4538  we set the flag only if there is a liter
4538      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4539      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4540    
4541      /* 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
4542      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4543    
4544      tempcode = previous;      tempcode = previous;
4545    
# Line 3807  we set the flag only if there is a liter Line 4562  we set the flag only if there is a liter
4562        }        }
4563      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4564    
4565        /* If previous was a recursion call, wrap it in atomic brackets so that
4566        previous becomes the atomic group. All recursions were so wrapped in the
4567        past, but it no longer happens for non-repeated recursions. In fact, the
4568        repeated ones could be re-implemented independently so as not to need this,
4569        but for the moment we rely on the code for repeating groups. */
4570    
4571        if (*previous == OP_RECURSE)
4572          {
4573          memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
4574          *previous = OP_ONCE;
4575          PUT(previous, 1, 2 + 2*LINK_SIZE);
4576          previous[2 + 2*LINK_SIZE] = OP_KET;
4577          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4578          code += 2 + 2 * LINK_SIZE;
4579          length_prevgroup = 3 + 3*LINK_SIZE;
4580    
4581          /* When actually compiling, we need to check whether this was a forward
4582          reference, and if so, adjust the offset. */
4583    
4584          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4585            {
4586            int offset = GET(cd->hwm, -LINK_SIZE);
4587            if (offset == previous + 1 - cd->start_code)
4588              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4589            }
4590          }
4591    
4592        /* Now handle repetition for the different types of item. */
4593    
4594      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4595      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
4596      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
4597      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
4598      instead.  */      instead.  */
4599    
4600      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4601        {        {
4602        /* Deal with UTF-8 characters that take up more than one byte. It's        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4603    
4604          /* Deal with UTF characters that take up more than one character. It's
4605        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
4606        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
4607        length rather than a small character. */        it's a length rather than a small character. */
4608    
4609  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
4610        if (utf8 && (code[-1] & 0x80) != 0)  #ifdef COMPILE_PCRE8
4611          if (utf && (code[-1] & 0x80) != 0)
4612    #endif /* COMPILE_PCRE8 */
4613    #ifdef COMPILE_PCRE16
4614          if (utf && (code[-1] & 0xfc00) == 0xdc00)
4615    #endif /* COMPILE_PCRE8 */
4616          {          {
4617          uschar *lastchar = code - 1;          pcre_uchar *lastchar = code - 1;
4618          while((*lastchar & 0xc0) == 0x80) lastchar--;          BACKCHAR(lastchar);
4619          c = code - lastchar;            /* Length of UTF-8 character */          c = code - lastchar;            /* Length of UTF-8 character */
4620          memcpy(utf8_char, lastchar, c); /* Save the char */          memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
4621          c |= 0x80;                      /* Flag c as a length */          c |= UTF_LENGTH;                /* Flag c as a length */
4622          }          }
4623        else        else
4624  #endif  #endif /* SUPPORT_UTF */
4625    
4626        /* 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
4627        with UTF-8 disabled, or for a UTF-8 character < 128. */        with UTF disabled, or for a single character UTF character. */
4628    
4629          {          {
4630          c = code[-1];          c = code[-1];
4631          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqchar = c | req_caseopt | cd->req_varyopt;
4632          }          }
4633    
4634        /* 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 3847  we set the flag only if there is a liter Line 4638  we set the flag only if there is a liter
4638    
4639        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4640            repeat_max < 0 &&            repeat_max < 0 &&
4641            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf, ptr + 1, options, cd))
             options, cd))  
4642          {          {
4643          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4644          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3860  we set the flag only if there is a liter Line 4650  we set the flag only if there is a liter
4650      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4651      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4652      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4653      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4654      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4655    
4656      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4657        {        {
4658        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4659        c = previous[1];        c = previous[1];
4660   &n