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

Diff of /code/trunk/pcre_compile.c

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

revision 457 by ph10, Sat Oct 3 16:24:08 2009 UTC revision 621 by ph10, Mon Jul 18 10:14:09 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    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  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 124  static const short int escapes[] = {
124       -ESC_H,                  0,       -ESC_H,                  0,
125       0,                       -ESC_K,       0,                       -ESC_K,
126       0,                       0,       0,                       0,
127       0,                       0,       -ESC_N,                  0,
128       -ESC_P,                  -ESC_Q,       -ESC_P,                  -ESC_Q,
129       -ESC_R,                  -ESC_S,       -ESC_R,                  -ESC_S,
130       0,                       0,       0,                       0,
# Line 165  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  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,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  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 188  string is built from string macros so th
188  platforms. */  platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199    STRING_ACCEPT0    STRING_ACCEPT0
200    STRING_COMMIT0    STRING_COMMIT0
201    STRING_F0    STRING_F0
# Line 196  static const char verbnames[] = Line 205  static const char verbnames[] =
205    STRING_THEN;    STRING_THEN;
206    
207  static const verbitem verbs[] = {  static const verbitem verbs[] = {
208    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
209    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
210    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
212    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
213    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
214    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 250  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 262  the number of relocations needed when a Line 320  the number of relocations needed when a
320  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
321  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
322  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
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 309  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 335  static const char error_texts[] = Line 397  static const char error_texts[] =
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\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"
399    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
403    "number is too big\0"    "number is too big\0"
# Line 343  static const char error_texts[] = Line 405  static const char error_texts[] =
405    "digit expected after (?+\0"    "digit expected after (?+\0"
406    "] is an invalid data character in JavaScript compatibility mode\0"    "] is an invalid data character in JavaScript compatibility mode\0"
407    /* 65 */    /* 65 */
408    "different names for subpatterns of the same number are not allowed";    "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      ;
413    
414  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
415  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 480  static const unsigned char ebcdic_charta Line 545  static const unsigned char ebcdic_charta
545  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
546    
547  static BOOL  static BOOL
548    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,
549      int *, int *, branch_chain *, compile_data *, int *);      int *, branch_chain *, compile_data *, int *);
550    
551    
552    
# Line 502  static const char * Line 567  static const char *
567  find_error_text(int n)  find_error_text(int n)
568  {  {
569  const char *s = error_texts;  const char *s = error_texts;
570  for (; n > 0; n--) while (*s++ != 0) {};  for (; n > 0; n--)
571      {
572      while (*s++ != 0) {};
573      if (*s == 0) return "Error text not found (please report)";
574      }
575  return s;  return s;
576  }  }
577    
# Line 573  else Line 642  else
642    
643      case CHAR_l:      case CHAR_l:
644      case CHAR_L:      case CHAR_L:
     case CHAR_N:  
645      case CHAR_u:      case CHAR_u:
646      case CHAR_U:      case CHAR_U:
647      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
# Line 774  else Line 842  else
842      break;      break;
843    
844      /* 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.
845      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
846        coding is ASCII-specific, but then the whole concept of \cx is
847      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
848    
849      case CHAR_c:      case CHAR_c:
# Line 784  else Line 853  else
853        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
854        break;        break;
855        }        }
856    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
857  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
858          {
859          *errorcodeptr = ERR68;
860          break;
861          }
862      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
863      c ^= 0x40;      c ^= 0x40;
864  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
865      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
866      c ^= 0xC0;      c ^= 0xC0;
867  #endif  #endif
# Line 811  else Line 884  else
884      }      }
885    }    }
886    
887    /* Perl supports \N{name} for character names, as well as plain \N for "not
888    newline". PCRE does not support \N{name}. */
889    
890    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
891      *errorcodeptr = ERR37;
892    
893    /* If PCRE_UCP is set, we change the values for \d etc. */
894    
895    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
896      c -= (ESC_DU - ESC_D);
897    
898    /* Set the pointer to the final character before returning. */
899    
900  *ptrptr = ptr;  *ptrptr = ptr;
901  return c;  return c;
902  }  }
# Line 1019  top-level call starts at the beginning o Line 1105  top-level call starts at the beginning o
1105  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
1106  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
1107  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
1108  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
1109  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1110  first pass. Recursion is used to keep track of subpatterns that reset the  
1111  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1112    that if (?< or (?' or (?P< is encountered, the name will be correctly
1113    terminated because that is checked in the first pass. There is now one call to
1114    this function in the first pass, to check for a recursive back reference by
1115    name (so that we can make the whole group atomic). In this case, we need check
1116    only up to the current position in the pattern, and that is still OK because
1117    and previous occurrences will have been checked. To make this work, the test
1118    for "end of pattern" is a check against cd->end_pattern in the main loop,
1119    instead of looking for a binary zero. This means that the special first-pass
1120    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1121    processing items within the loop are OK, because afterwards the main loop will
1122    terminate.)
1123    
1124  Arguments:  Arguments:
1125    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1030  Arguments: Line 1127  Arguments:
1127    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1128    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1129    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1130      utf8         TRUE if we are in UTF-8 mode
1131    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1132    
1133  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
# Line 1037  Returns:       the number of the named s Line 1135  Returns:       the number of the named s
1135    
1136  static int  static int
1137  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1138    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1139  {  {
1140  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1141  int start_count = *count;  int start_count = *count;
# Line 1049  dealing with. The very first call may no Line 1147  dealing with. The very first call may no
1147    
1148  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1149    {    {
1150    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1151        ptr[2] == CHAR_VERTICAL_LINE)  
1152      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1153    
1154      /* Handle a normal, unnamed capturing parenthesis. */
1155    
1156      else if (ptr[1] != CHAR_QUESTION_MARK)
1157        {
1158        *count += 1;
1159        if (name == NULL && *count == lorn) return *count;
1160        ptr++;
1161        }
1162    
1163      /* All cases now have (? at the start. Remember when we are in a group
1164      where the parenthesis numbers are duplicated. */
1165    
1166      else if (ptr[2] == CHAR_VERTICAL_LINE)
1167      {      {
1168      ptr += 3;      ptr += 3;
1169      dup_parens = TRUE;      dup_parens = TRUE;
1170      }      }
1171    
1172    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1173    
1174    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1175      {      {
1176      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1177      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1178      }      }
1179    
1180    /* 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
1181    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
1182    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1183    
1184    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1185      {      {
# Line 1079  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1191  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1191        }        }
1192      }      }
1193    
1194    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1195    
1196    else    else
1197      {      {
# Line 1102  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1214  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1214        if (name != NULL && lorn == ptr - thisname &&        if (name != NULL && lorn == ptr - thisname &&
1215            strncmp((const char *)name, (const char *)thisname, lorn) == 0)            strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1216          return *count;          return *count;
1217        term++;        term++;
1218        }        }
1219      }      }
1220    }    }
1221    
1222  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1223  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1224    first-pass call when this value is temporarily adjusted to stop at the current
1225    position. So DO NOT change this to a test for binary zero. */
1226    
1227  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1228    {    {
1229    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1230    
# Line 1148  for (; *ptr != 0; ptr++) Line 1262  for (; *ptr != 0; ptr++)
1262            break;            break;
1263          }          }
1264        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1265          {          {
1266          negate_class = TRUE;          negate_class = TRUE;
1267          ptr++;          ptr++;
1268          }          }
1269        else break;        else break;
1270        }        }
1271    
# Line 1184  for (; *ptr != 0; ptr++) Line 1298  for (; *ptr != 0; ptr++)
1298    
1299    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1300      {      {
1301      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1302        while (*ptr != 0)
1303          {
1304          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1305          ptr++;
1306    #ifdef SUPPORT_UTF8
1307          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1308    #endif
1309          }
1310      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1311      continue;      continue;
1312      }      }
# Line 1193  for (; *ptr != 0; ptr++) Line 1315  for (; *ptr != 0; ptr++)
1315    
1316    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1317      {      {
1318      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1319      if (rc > 0) return rc;      if (rc > 0) return rc;
1320      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1321      }      }
# Line 1201  for (; *ptr != 0; ptr++) Line 1323  for (; *ptr != 0; ptr++)
1323    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1324      {      {
1325      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1326      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1327      }      }
1328    
1329    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1240  Arguments: Line 1361  Arguments:
1361    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1362    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1363    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1364      utf8         TRUE if we are in UTF-8 mode
1365    
1366  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1367  */  */
1368    
1369  static int  static int
1370  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1371      BOOL utf8)
1372  {  {
1373  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1374  int count = 0;  int count = 0;
# Line 1258  matching closing parens. That is why we Line 1381  matching closing parens. That is why we
1381    
1382  for (;;)  for (;;)
1383    {    {
1384    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1385    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1386    }    }
1387    
# Line 1274  return rc; Line 1397  return rc;
1397    
1398  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1399  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
1400  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
1401  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
1402  assertions, and also the \b assertion; for others it does not.  does not.
1403    
1404  Arguments:  Arguments:
1405    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  
1406    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1407    
1408  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1409  */  */
1410    
1411  static const uschar*  static const uschar*
1412  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1413  {  {
1414  for (;;)  for (;;)
1415    {    {
1416    switch ((int)*code)    switch ((int)*code)
1417      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1418      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1419      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1420      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1317  for (;;) Line 1430  for (;;)
1430    
1431      case OP_CALLOUT:      case OP_CALLOUT:
1432      case OP_CREF:      case OP_CREF:
1433        case OP_NCREF:
1434      case OP_RREF:      case OP_RREF:
1435        case OP_NRREF:
1436      case OP_DEF:      case OP_DEF:
1437      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1438      break;      break;
# Line 1338  for (;;) Line 1453  for (;;)
1453    
1454  /* 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,
1455  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.
1456  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
1457  temporarily terminated with OP_END when this function is called.  temporarily terminated with OP_END when this function is called.
1458    
1459  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
1460  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.
1461  However, we cannot do this when the assertion contains subroutine calls,  However, we cannot do this when the assertion contains subroutine calls,
1462  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
1463  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.
1464    
1465  Arguments:  Arguments:
1466    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1467    options  the compiling options    utf8     TRUE in UTF-8 mode
1468    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1469    cd       the "compile data" structure    cd       the "compile data" structure
1470    
1471  Returns:   the fixed length,  Returns:   the fixed length,
1472               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1473               or -2 if \C was encountered               or -2 if \C was encountered
1474               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
1475  */  */
1476    
1477  static int  static int
1478  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1479  {  {
1480  int length = -1;  int length = -1;
1481    
# Line 1377  for (;;) Line 1492  for (;;)
1492    register int op = *cc;    register int op = *cc;
1493    switch (op)    switch (op)
1494      {      {
1495        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1496        OP_BRA (normal non-capturing bracket) because the other variants of these
1497        opcodes are all concerned with unlimited repeated groups, which of course
1498        are not of fixed length. They will cause a -1 response from the default
1499        case of this switch. */
1500    
1501      case OP_CBRA:      case OP_CBRA:
1502      case OP_BRA:      case OP_BRA:
1503      case OP_ONCE:      case OP_ONCE:
1504      case OP_COND:      case OP_COND:
1505      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1506      if (d < 0) return d;      if (d < 0) return d;
1507      branchlength += d;      branchlength += d;
1508      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1390  for (;;) Line 1511  for (;;)
1511    
1512      /* 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
1513      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1514      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1515        Note that we must not include the OP_KETRxxx opcodes here, because they
1516        all imply an unlimited repeat. */
1517    
1518      case OP_ALT:      case OP_ALT:
1519      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1520      case OP_END:      case OP_END:
1521      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1522        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1403  for (;;) Line 1524  for (;;)
1524      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1525      branchlength = 0;      branchlength = 0;
1526      break;      break;
1527    
1528      /* A true recursion implies not fixed length, but a subroutine call may      /* A true recursion implies not fixed length, but a subroutine call may
1529      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
1530      it until the end of the pattern, so return -3. */      it until the end of the pattern, so return -3. */
1531    
1532      case OP_RECURSE:      case OP_RECURSE:
1533      if (!atend) return -3;      if (!atend) return -3;
1534      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1535      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1536      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1537      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1538      if (d < 0) return d;      if (d < 0) return d;
1539      branchlength += d;      branchlength += d;
1540      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1541      break;      break;
1542    
1543      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1544    
# Line 1432  for (;;) Line 1553  for (;;)
1553    
1554      case OP_REVERSE:      case OP_REVERSE:
1555      case OP_CREF:      case OP_CREF:
1556        case OP_NCREF:
1557      case OP_RREF:      case OP_RREF:
1558        case OP_NRREF:
1559      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1560      case OP_CALLOUT:      case OP_CALLOUT:
1561      case OP_SOD:      case OP_SOD:
1562      case OP_SOM:      case OP_SOM:
1563        case OP_SET_SOM:
1564      case OP_EOD:      case OP_EOD:
1565      case OP_EODN:      case OP_EODN:
1566      case OP_CIRC:      case OP_CIRC:
1567        case OP_CIRCM:
1568      case OP_DOLL:      case OP_DOLL:
1569        case OP_DOLLM:
1570      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1571      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1572      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1450  for (;;) Line 1575  for (;;)
1575      /* Handle literal characters */      /* Handle literal characters */
1576    
1577      case OP_CHAR:      case OP_CHAR:
1578      case OP_CHARNC:      case OP_CHARI:
1579      case OP_NOT:      case OP_NOT:
1580        case OP_NOTI:
1581      branchlength++;      branchlength++;
1582      cc += 2;      cc += 2;
1583  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1584      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1585  #endif  #endif
1586      break;      break;
1587    
# Line 1467  for (;;) Line 1592  for (;;)
1592      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1593      cc += 4;      cc += 4;
1594  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1595      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1596  #endif  #endif
1597      break;      break;
1598    
# Line 1552  for (;;) Line 1676  for (;;)
1676    
1677  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1678  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
1679  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
1680  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
1681  length.  length.
1682    
1683  Arguments:  Arguments:
# Line 1570  _pcre_find_bracket(const uschar *code, B Line 1694  _pcre_find_bracket(const uschar *code, B
1694  for (;;)  for (;;)
1695    {    {
1696    register int c = *code;    register int c = *code;
1697    
1698    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1699    
1700    /* 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 1577  for (;;) Line 1702  for (;;)
1702    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. */
1703    
1704    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1705    
1706    /* Handle recursion */    /* Handle recursion */
1707    
1708    else if (c == OP_REVERSE)    else if (c == OP_REVERSE)
1709      {      {
1710      if (number < 0) return (uschar *)code;      if (number < 0) return (uschar *)code;
1711      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1712      }      }
1713    
1714    /* Handle capturing bracket */    /* Handle capturing bracket */
1715    
1716    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1717               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1718      {      {
1719      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1720      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1597  for (;;) Line 1723  for (;;)
1723    
1724    /* 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
1725    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
1726    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1727      must add in its length. */
1728    
1729    else    else
1730      {      {
# Line 1621  for (;;) Line 1748  for (;;)
1748        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1749        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1750        break;        break;
1751    
1752          case OP_MARK:
1753          case OP_PRUNE_ARG:
1754          case OP_SKIP_ARG:
1755          code += code[1];
1756          break;
1757    
1758          case OP_THEN_ARG:
1759          code += code[1+LINK_SIZE];
1760          break;
1761        }        }
1762    
1763      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1635  for (;;) Line 1772  for (;;)
1772      if (utf8) switch(c)      if (utf8) switch(c)
1773        {        {
1774        case OP_CHAR:        case OP_CHAR:
1775        case OP_CHARNC:        case OP_CHARI:
1776        case OP_EXACT:        case OP_EXACT:
1777          case OP_EXACTI:
1778        case OP_UPTO:        case OP_UPTO:
1779          case OP_UPTOI:
1780        case OP_MINUPTO:        case OP_MINUPTO:
1781          case OP_MINUPTOI:
1782        case OP_POSUPTO:        case OP_POSUPTO:
1783          case OP_POSUPTOI:
1784        case OP_STAR:        case OP_STAR:
1785          case OP_STARI:
1786        case OP_MINSTAR:        case OP_MINSTAR:
1787          case OP_MINSTARI:
1788        case OP_POSSTAR:        case OP_POSSTAR:
1789          case OP_POSSTARI:
1790        case OP_PLUS:        case OP_PLUS:
1791          case OP_PLUSI:
1792        case OP_MINPLUS:        case OP_MINPLUS:
1793          case OP_MINPLUSI:
1794        case OP_POSPLUS:        case OP_POSPLUS:
1795          case OP_POSPLUSI:
1796        case OP_QUERY:        case OP_QUERY:
1797          case OP_QUERYI:
1798        case OP_MINQUERY:        case OP_MINQUERY:
1799          case OP_MINQUERYI:
1800        case OP_POSQUERY:        case OP_POSQUERY:
1801          case OP_POSQUERYI:
1802        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1803        break;        break;
1804        }        }
# Line 1692  for (;;) Line 1842  for (;;)
1842    
1843    /* 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
1844    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
1845    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1846      must add in its length. */
1847    
1848    else    else
1849      {      {
# Line 1716  for (;;) Line 1867  for (;;)
1867        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1868        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1869        break;        break;
1870    
1871          case OP_MARK:
1872          case OP_PRUNE_ARG:
1873          case OP_SKIP_ARG:
1874          code += code[1];
1875          break;
1876    
1877          case OP_THEN_ARG:
1878          code += code[1+LINK_SIZE];
1879          break;
1880        }        }
1881    
1882      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1730  for (;;) Line 1891  for (;;)
1891      if (utf8) switch(c)      if (utf8) switch(c)
1892        {        {
1893        case OP_CHAR:        case OP_CHAR:
1894        case OP_CHARNC:        case OP_CHARI:
1895        case OP_EXACT:        case OP_EXACT:
1896          case OP_EXACTI:
1897        case OP_UPTO:        case OP_UPTO:
1898          case OP_UPTOI:
1899        case OP_MINUPTO:        case OP_MINUPTO:
1900          case OP_MINUPTOI:
1901        case OP_POSUPTO:        case OP_POSUPTO:
1902          case OP_POSUPTOI:
1903        case OP_STAR:        case OP_STAR:
1904          case OP_STARI:
1905        case OP_MINSTAR:        case OP_MINSTAR:
1906          case OP_MINSTARI:
1907        case OP_POSSTAR:        case OP_POSSTAR:
1908          case OP_POSSTARI:
1909        case OP_PLUS:        case OP_PLUS:
1910          case OP_PLUSI:
1911        case OP_MINPLUS:        case OP_MINPLUS:
1912          case OP_MINPLUSI:
1913        case OP_POSPLUS:        case OP_POSPLUS:
1914          case OP_POSPLUSI:
1915        case OP_QUERY:        case OP_QUERY:
1916          case OP_QUERYI:
1917        case OP_MINQUERY:        case OP_MINQUERY:
1918          case OP_MINQUERYI:
1919        case OP_POSQUERY:        case OP_POSQUERY:
1920          case OP_POSQUERYI:
1921        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1922        break;        break;
1923        }        }
# Line 1772  Arguments: Line 1946  Arguments:
1946    code        points to start of search    code        points to start of search
1947    endcode     points to where to stop    endcode     points to where to stop
1948    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1949      cd          contains pointers to tables etc.
1950    
1951  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1952  */  */
1953    
1954  static BOOL  static BOOL
1955  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1956      compile_data *cd)
1957  {  {
1958  register int c;  register int c;
1959  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1960       code < endcode;       code < endcode;
1961       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1962    {    {
1963    const uschar *ccode;    const uschar *ccode;
1964    
# Line 1798  for (code = first_significant_code(code Line 1974  for (code = first_significant_code(code
1974      continue;      continue;
1975      }      }
1976    
1977      /* For a recursion/subroutine call, if its end has been reached, which
1978      implies a subroutine call, we can scan it. */
1979    
1980      if (c == OP_RECURSE)
1981        {
1982        BOOL empty_branch = FALSE;
1983        const uschar *scode = cd->start_code + GET(code, 1);
1984        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1985        do
1986          {
1987          if (could_be_empty_branch(scode, endcode, utf8, cd))
1988            {
1989            empty_branch = TRUE;
1990            break;
1991            }
1992          scode += GET(scode, 1);
1993          }
1994        while (*scode == OP_ALT);
1995        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1996        continue;
1997        }
1998    
1999    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2000    
2001    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2002          c == OP_BRAPOSZERO)
2003      {      {
2004      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2005      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1808  for (code = first_significant_code(code Line 2007  for (code = first_significant_code(code
2007      continue;      continue;
2008      }      }
2009    
2010      /* A nested group that is already marked as "could be empty" can just be
2011      skipped. */
2012    
2013      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2014          c == OP_SCBRA || c == OP_SCBRAPOS)
2015        {
2016        do code += GET(code, 1); while (*code == OP_ALT);
2017        c = *code;
2018        continue;
2019        }
2020    
2021    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2022    
2023    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2024          c == OP_CBRA || c == OP_CBRAPOS ||
2025          c == OP_ONCE || c == OP_COND)
2026      {      {
2027      BOOL empty_branch;      BOOL empty_branch;
2028      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1826  for (code = first_significant_code(code Line 2038  for (code = first_significant_code(code
2038        empty_branch = FALSE;        empty_branch = FALSE;
2039        do        do
2040          {          {
2041          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2042            empty_branch = TRUE;            empty_branch = TRUE;
2043          code += GET(code, 1);          code += GET(code, 1);
2044          }          }
# Line 1897  for (code = first_significant_code(code Line 2109  for (code = first_significant_code(code
2109      case OP_ALLANY:      case OP_ALLANY:
2110      case OP_ANYBYTE:      case OP_ANYBYTE:
2111      case OP_CHAR:      case OP_CHAR:
2112      case OP_CHARNC:      case OP_CHARI:
2113      case OP_NOT:      case OP_NOT:
2114        case OP_NOTI:
2115      case OP_PLUS:      case OP_PLUS:
2116      case OP_MINPLUS:      case OP_MINPLUS:
2117      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1938  for (code = first_significant_code(code Line 2151  for (code = first_significant_code(code
2151      case OP_KET:      case OP_KET:
2152      case OP_KETRMAX:      case OP_KETRMAX:
2153      case OP_KETRMIN:      case OP_KETRMIN:
2154        case OP_KETRPOS:
2155      case OP_ALT:      case OP_ALT:
2156      return TRUE;      return TRUE;
2157    
# Line 1946  for (code = first_significant_code(code Line 2160  for (code = first_significant_code(code
2160    
2161  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2162      case OP_STAR:      case OP_STAR:
2163        case OP_STARI:
2164      case OP_MINSTAR:      case OP_MINSTAR:
2165        case OP_MINSTARI:
2166      case OP_POSSTAR:      case OP_POSSTAR:
2167        case OP_POSSTARI:
2168      case OP_QUERY:      case OP_QUERY:
2169        case OP_QUERYI:
2170      case OP_MINQUERY:      case OP_MINQUERY:
2171        case OP_MINQUERYI:
2172      case OP_POSQUERY:      case OP_POSQUERY:
2173        case OP_POSQUERYI:
2174      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2175      break;      break;
2176    
2177      case OP_UPTO:      case OP_UPTO:
2178        case OP_UPTOI:
2179      case OP_MINUPTO:      case OP_MINUPTO:
2180        case OP_MINUPTOI:
2181      case OP_POSUPTO:      case OP_POSUPTO:
2182        case OP_POSUPTOI:
2183      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2184      break;      break;
2185  #endif  #endif
2186    
2187        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2188        string. */
2189    
2190        case OP_MARK:
2191        case OP_PRUNE_ARG:
2192        case OP_SKIP_ARG:
2193        code += code[1];
2194        break;
2195    
2196        case OP_THEN_ARG:
2197        code += code[1+LINK_SIZE];
2198        break;
2199    
2200        /* None of the remaining opcodes are required to match a character. */
2201    
2202        default:
2203        break;
2204      }      }
2205    }    }
2206    
# Line 1982  Arguments: Line 2223  Arguments:
2223    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2224    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2225    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2226      cd          pointers to tables etc
2227    
2228  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2229  */  */
2230    
2231  static BOOL  static BOOL
2232  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2233    BOOL utf8)    BOOL utf8, compile_data *cd)
2234  {  {
2235  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2236    {    {
2237    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2238        return FALSE;
2239    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2240    }    }
2241  return TRUE;  return TRUE;
# Line 2174  auto_callout(uschar *code, const uschar Line 2417  auto_callout(uschar *code, const uschar
2417  {  {
2418  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2419  *code++ = 255;  *code++ = 255;
2420  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2421  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2422  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2423  }  }
2424    
# Line 2200  Returns:             nothing Line 2443  Returns:             nothing
2443  static void  static void
2444  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2445  {  {
2446  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2447  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2448  }  }
2449    
# Line 2250  for (++c; c <= d; c++) Line 2493  for (++c; c <= d; c++)
2493    
2494  return TRUE;  return TRUE;
2495  }  }
2496    
2497    
2498    
2499    /*************************************************
2500    *        Check a character and a property        *
2501    *************************************************/
2502    
2503    /* This function is called by check_auto_possessive() when a property item
2504    is adjacent to a fixed character.
2505    
2506    Arguments:
2507      c            the character
2508      ptype        the property type
2509      pdata        the data for the type
2510      negated      TRUE if it's a negated property (\P or \p{^)
2511    
2512    Returns:       TRUE if auto-possessifying is OK
2513    */
2514    
2515    static BOOL
2516    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2517    {
2518    const ucd_record *prop = GET_UCD(c);
2519    switch(ptype)
2520      {
2521      case PT_LAMP:
2522      return (prop->chartype == ucp_Lu ||
2523              prop->chartype == ucp_Ll ||
2524              prop->chartype == ucp_Lt) == negated;
2525    
2526      case PT_GC:
2527      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2528    
2529      case PT_PC:
2530      return (pdata == prop->chartype) == negated;
2531    
2532      case PT_SC:
2533      return (pdata == prop->script) == negated;
2534    
2535      /* These are specials */
2536    
2537      case PT_ALNUM:
2538      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2539              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2540    
2541      case PT_SPACE:    /* Perl space */
2542      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2543              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2544              == negated;
2545    
2546      case PT_PXSPACE:  /* POSIX space */
2547      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2548              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2549              c == CHAR_FF || c == CHAR_CR)
2550              == negated;
2551    
2552      case PT_WORD:
2553      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2554              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2555              c == CHAR_UNDERSCORE) == negated;
2556      }
2557    return FALSE;
2558    }
2559  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2560    
2561    
# Line 2263  whether the next thing could possibly ma Line 2569  whether the next thing could possibly ma
2569  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2570    
2571  Arguments:  Arguments:
2572    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2573    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2574    ptr           next character in pattern    ptr           next character in pattern
2575    options       options bits    options       options bits
2576    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2275  Returns:        TRUE if possessifying is Line 2579  Returns:        TRUE if possessifying is
2579  */  */
2580    
2581  static BOOL  static BOOL
2582  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2583    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2584  {  {
2585  int next;  int c, next;
2586    int op_code = *previous++;
2587    
2588  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2589    
# Line 2289  if ((options & PCRE_EXTENDED) != 0) Line 2594  if ((options & PCRE_EXTENDED) != 0)
2594      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2595      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2596        {        {
2597        while (*(++ptr) != 0)        ptr++;
2598          while (*ptr != 0)
2599            {
2600          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2601            ptr++;
2602    #ifdef SUPPORT_UTF8
2603            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2604    #endif
2605            }
2606        }        }
2607      else break;      else break;
2608      }      }
# Line 2326  if ((options & PCRE_EXTENDED) != 0) Line 2638  if ((options & PCRE_EXTENDED) != 0)
2638      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2639      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2640        {        {
2641        while (*(++ptr) != 0)        ptr++;
2642          while (*ptr != 0)
2643            {
2644          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2645            ptr++;
2646    #ifdef SUPPORT_UTF8
2647            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2648    #endif
2649            }
2650        }        }
2651      else break;      else break;
2652      }      }
# Line 2339  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2658  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2658    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2659      return FALSE;      return FALSE;
2660    
2661  /* 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
2662  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. */  
2663    
2664  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2665    {    {
2666    case OP_CHAR:    case OP_CHAR:
2667  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2668    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2669  #else  #else
2670    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2671  #endif  #endif
2672    return item != next;    return c != next;
2673    
2674    /* 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
2675    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
2676    high-valued characters. */    high-valued characters. */
2677    
2678    case OP_CHARNC:    case OP_CHARI:
2679  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2680    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2681    #else
2682      c = *previous;
2683  #endif  #endif
2684    if (item == next) return FALSE;    if (c == next) return FALSE;
2685  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2686    if (utf8)    if (utf8)
2687      {      {
# Line 2376  if (next >= 0) switch(op_code) Line 2692  if (next >= 0) switch(op_code)
2692  #else  #else
2693      othercase = NOTACHAR;      othercase = NOTACHAR;
2694  #endif  #endif
2695      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2696      }      }
2697    else    else
2698  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2699    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2700    
2701    /* 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
2702      opcodes are not used for multi-byte characters, because they are coded using
2703      an XCLASS instead. */
2704    
2705    case OP_NOT:    case OP_NOT:
2706    if (item == next) return TRUE;    return (c = *previous) == next;
2707    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2708      case OP_NOTI:
2709      if ((c = *previous) == next) return TRUE;
2710  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2711    if (utf8)    if (utf8)
2712      {      {
# Line 2397  if (next >= 0) switch(op_code) Line 2717  if (next >= 0) switch(op_code)
2717  #else  #else
2718      othercase = NOTACHAR;      othercase = NOTACHAR;
2719  #endif  #endif
2720      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2721      }      }
2722    else    else
2723  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2724    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2725    
2726      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2727      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2728    
2729    case OP_DIGIT:    case OP_DIGIT:
2730    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2444  if (next >= 0) switch(op_code) Line 2767  if (next >= 0) switch(op_code)
2767      case 0x202f:      case 0x202f:
2768      case 0x205f:      case 0x205f:
2769      case 0x3000:      case 0x3000:
2770      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2771      default:      default:
2772      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2773      }      }
2774    
2775      case OP_ANYNL:
2776    case OP_VSPACE:    case OP_VSPACE:
2777    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2778    switch(next)    switch(next)
# Line 2460  if (next >= 0) switch(op_code) Line 2784  if (next >= 0) switch(op_code)
2784      case 0x85:      case 0x85:
2785      case 0x2028:      case 0x2028:
2786      case 0x2029:      case 0x2029:
2787      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2788      default:      default:
2789      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2790      }      }
2791    
2792    #ifdef SUPPORT_UCP
2793      case OP_PROP:
2794      return check_char_prop(next, previous[0], previous[1], FALSE);
2795    
2796      case OP_NOTPROP:
2797      return check_char_prop(next, previous[0], previous[1], TRUE);
2798    #endif
2799    
2800    default:    default:
2801    return FALSE;    return FALSE;
2802    }    }
2803    
2804    
2805  /* 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
2806    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2807    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2808    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2809    replaced by OP_PROP codes when PCRE_UCP is set. */
2810    
2811  switch(op_code)  switch(op_code)
2812    {    {
2813    case OP_CHAR:    case OP_CHAR:
2814    case OP_CHARNC:    case OP_CHARI:
2815  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2816    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2817    #else
2818      c = *previous;
2819  #endif  #endif
2820    switch(-next)    switch(-next)
2821      {      {
2822      case ESC_d:      case ESC_d:
2823      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2824    
2825      case ESC_D:      case ESC_D:
2826      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2827    
2828      case ESC_s:      case ESC_s:
2829      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2830    
2831      case ESC_S:      case ESC_S:
2832      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2833    
2834      case ESC_w:      case ESC_w:
2835      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2836    
2837      case ESC_W:      case ESC_W:
2838      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2839    
2840      case ESC_h:      case ESC_h:
2841      case ESC_H:      case ESC_H:
2842      switch(item)      switch(c)
2843        {        {
2844        case 0x09:        case 0x09:
2845        case 0x20:        case 0x20:
# Line 2529  switch(op_code) Line 2867  switch(op_code)
2867    
2868      case ESC_v:      case ESC_v:
2869      case ESC_V:      case ESC_V:
2870      switch(item)      switch(c)
2871        {        {
2872        case 0x0a:        case 0x0a:
2873        case 0x0b:        case 0x0b:
# Line 2543  switch(op_code) Line 2881  switch(op_code)
2881        return -next == ESC_v;        return -next == ESC_v;
2882        }        }
2883    
2884        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2885        their substitutions and process them. The result will always be either
2886        -ESC_p or -ESC_P. Then fall through to process those values. */
2887    
2888    #ifdef SUPPORT_UCP
2889        case ESC_du:
2890        case ESC_DU:
2891        case ESC_wu:
2892        case ESC_WU:
2893        case ESC_su:
2894        case ESC_SU:
2895          {
2896          int temperrorcode = 0;
2897          ptr = substitutes[-next - ESC_DU];
2898          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2899          if (temperrorcode != 0) return FALSE;
2900          ptr++;    /* For compatibility */
2901          }
2902        /* Fall through */
2903    
2904        case ESC_p:
2905        case ESC_P:
2906          {
2907          int ptype, pdata, errorcodeptr;
2908          BOOL negated;
2909    
2910          ptr--;      /* Make ptr point at the p or P */
2911          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2912          if (ptype < 0) return FALSE;
2913          ptr++;      /* Point past the final curly ket */
2914    
2915          /* If the property item is optional, we have to give up. (When generated
2916          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2917          to the original \d etc. At this point, ptr will point to a zero byte. */
2918    
2919          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2920            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2921              return FALSE;
2922    
2923          /* Do the property check. */
2924    
2925          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2926          }
2927    #endif
2928    
2929      default:      default:
2930      return FALSE;      return FALSE;
2931      }      }
2932    
2933      /* In principle, support for Unicode properties should be integrated here as
2934      well. It means re-organizing the above code so as to get hold of the property
2935      values before switching on the op-code. However, I wonder how many patterns
2936      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2937      these op-codes are never generated.) */
2938    
2939    case OP_DIGIT:    case OP_DIGIT:
2940    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2941           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2942    
2943    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2944    return next == -ESC_d;    return next == -ESC_d;
2945    
2946    case OP_WHITESPACE:    case OP_WHITESPACE:
2947    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2948    
2949    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2950    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2951    
2952    case OP_HSPACE:    case OP_HSPACE:
2953    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2954             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2955    
2956    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2957    return next == -ESC_h;    return next == -ESC_h;
2958    
2959    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2960      case OP_ANYNL:
2961    case OP_VSPACE:    case OP_VSPACE:
2962    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2963    
2964    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2965    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2966    
2967    case OP_WORDCHAR:    case OP_WORDCHAR:
2968    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2969             next == -ESC_v || next == -ESC_R;
2970    
2971    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2972    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2638  BOOL inescq = FALSE; Line 3030  BOOL inescq = FALSE;
3030  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3031  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3032  const uschar *tempptr;  const uschar *tempptr;
3033    const uschar *nestptr = NULL;
3034  uschar *previous = NULL;  uschar *previous = NULL;
3035  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3036  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2654  BOOL utf8 = FALSE; Line 3047  BOOL utf8 = FALSE;
3047  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
3048  #endif  #endif
3049    
3050  #ifdef DEBUG  #ifdef PCRE_DEBUG
3051  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3052  #endif  #endif
3053    
# Line 2708  for (;; ptr++) Line 3101  for (;; ptr++)
3101    
3102    c = *ptr;    c = *ptr;
3103    
3104      /* If we are at the end of a nested substitution, revert to the outer level
3105      string. Nesting only happens one level deep. */
3106    
3107      if (c == 0 && nestptr != NULL)
3108        {
3109        ptr = nestptr;
3110        nestptr = NULL;
3111        c = *ptr;
3112        }
3113    
3114    /* 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
3115    previous cycle of this loop. */    previous cycle of this loop. */
3116    
3117    if (lengthptr != NULL)    if (lengthptr != NULL)
3118      {      {
3119  #ifdef DEBUG  #ifdef PCRE_DEBUG
3120      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3121  #endif  #endif
3122      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3123        {        {
3124        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3125        goto FAILED;        goto FAILED;
# Line 2738  for (;; ptr++) Line 3141  for (;; ptr++)
3141        goto FAILED;        goto FAILED;
3142        }        }
3143    
3144      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3145      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3146    
3147      /* 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
# Line 2765  for (;; ptr++) Line 3168  for (;; ptr++)
3168    /* 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
3169    reference list. */    reference list. */
3170    
3171    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3172      {      {
3173      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3174      goto FAILED;      goto FAILED;
# Line 2820  for (;; ptr++) Line 3223  for (;; ptr++)
3223      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3224      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3225        {        {
3226        while (*(++ptr) != 0)        ptr++;
3227          while (*ptr != 0)
3228          {          {
3229          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3230            ptr++;
3231    #ifdef SUPPORT_UTF8
3232            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3233    #endif
3234          }          }
3235        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3236    
# Line 2856  for (;; ptr++) Line 3264  for (;; ptr++)
3264          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3265          goto FAILED;          goto FAILED;
3266          }          }
3267        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3268        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3269        }        }
3270      return TRUE;      return TRUE;
# Line 2867  for (;; ptr++) Line 3275  for (;; ptr++)
3275      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3276    
3277      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3278        previous = NULL;
3279      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3280        {        {
3281        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3282          *code++ = OP_CIRCM;
3283        }        }
3284      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3285      break;      break;
3286    
3287      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3288      previous = NULL;      previous = NULL;
3289      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3290      break;      break;
3291    
3292      /* 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
# Line 3061  for (;; ptr++) Line 3470  for (;; ptr++)
3470            ptr++;            ptr++;
3471            }            }
3472    
3473          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3474          if (posix_class < 0)          if (posix_class < 0)
3475            {            {
3476            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3075  for (;; ptr++) Line 3484  for (;; ptr++)
3484          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3485            posix_class = 0;            posix_class = 0;
3486    
3487          /* 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
3488          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3489          subtract bits that may be in the main map already. At the end we or the  
3490          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3491            if ((options & PCRE_UCP) != 0)
3492              {
3493              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3494              if (posix_substitutes[pc] != NULL)
3495                {
3496                nestptr = tempptr + 1;
3497                ptr = posix_substitutes[pc] - 1;
3498                continue;
3499                }
3500              }
3501    #endif
3502            /* In the non-UCP case, we build the bit map for the POSIX class in a
3503            chunk of local store because we may be adding and subtracting from it,
3504            and we don't want to subtract bits that may be in the main map already.
3505            At the end we or the result into the bit map that is being built. */
3506    
3507          posix_class *= 3;          posix_class *= 3;
3508    
# Line 3122  for (;; ptr++) Line 3546  for (;; ptr++)
3546    
3547        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3548        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
3549        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
3550        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
3551        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
3552        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3553          PCRE_EXTRA is set. */
3554    
3555        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3556          {          {
3557          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3558          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3559    
3560          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 */  
3561          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3562            {            {
3563            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3151  for (;; ptr++) Line 3574  for (;; ptr++)
3574            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3575            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3576    
3577            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3578              {              {
3579    #ifdef SUPPORT_UCP
3580                case ESC_du:     /* These are the values given for \d etc */
3581                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3582                case ESC_wu:     /* escape sequence with an appropriate \p */
3583                case ESC_WU:     /* or \P to test Unicode properties instead */
3584                case ESC_su:     /* of the default ASCII testing. */
3585                case ESC_SU:
3586                nestptr = ptr;
3587                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3588                class_charcount -= 2;                /* Undo! */
3589                continue;
3590    #endif
3591              case ESC_d:              case ESC_d:
3592              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3593              continue;              continue;
# Line 3173  for (;; ptr++) Line 3606  for (;; ptr++)
3606              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3607              continue;              continue;
3608    
3609                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3610                if it was previously set by something earlier in the character
3611                class. */
3612    
3613              case ESC_s:              case ESC_s:
3614              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3615              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3616                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3617              continue;              continue;
3618    
3619              case ESC_S:              case ESC_S:
# Line 3184  for (;; ptr++) Line 3622  for (;; ptr++)
3622              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3623              continue;              continue;
3624    
3625              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)  
             {  
3626              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3627              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3628              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3221  for (;; ptr++) Line 3646  for (;; ptr++)
3646                }                }
3647  #endif  #endif
3648              continue;              continue;
             }  
3649    
3650            if (-c == ESC_H)              case ESC_H:
             {  
3651              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3652                {                {
3653                int x = 0xff;                int x = 0xff;
# Line 3266  for (;; ptr++) Line 3689  for (;; ptr++)
3689                }                }
3690  #endif  #endif
3691              continue;              continue;
             }  
3692    
3693            if (-c == ESC_v)              case ESC_v:
             {  
3694              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3695              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3696              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3285  for (;; ptr++) Line 3706  for (;; ptr++)
3706                }                }
3707  #endif  #endif
3708              continue;              continue;
             }  
3709    
3710            if (-c == ESC_V)              case ESC_V:
             {  
3711              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3712                {                {
3713                int x = 0xff;                int x = 0xff;
# Line 3318  for (;; ptr++) Line 3737  for (;; ptr++)
3737                }                }
3738  #endif  #endif
3739              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3740    
3741  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3742            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3743              {              case ESC_P:
3744              BOOL negated;                {
3745              int pdata;                BOOL negated;
3746              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3747              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3748              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3749              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3750                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3751              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3752              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3753              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3754              continue;                class_charcount -= 2;   /* Not a < 256 character */
3755              }                continue;
3756                  }
3757  #endif  #endif
3758            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3759            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3760            treated as literals. */              treated as literals. */
3761    
3762            if ((options & PCRE_EXTRA) != 0)              default:
3763              {              if ((options & PCRE_EXTRA) != 0)
3764              *errorcodeptr = ERR7;                {
3765              goto FAILED;                *errorcodeptr = ERR7;
3766                  goto FAILED;
3767                  }
3768                class_charcount -= 2;  /* Undo the default count from above */
3769                c = *ptr;              /* Get the final character and fall through */
3770                break;
3771              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3772            }            }
3773    
3774          /* 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
# Line 3419  for (;; ptr++) Line 3838  for (;; ptr++)
3838            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3839            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3840    
3841            /* \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 */  
3842    
3843            if (d < 0)            if (d < 0)
3844              {              {
3845              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  
3846                {                {
3847                ptr = oldptr;                ptr = oldptr;
3848                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3592  for (;; ptr++) Line 4008  for (;; ptr++)
4008          }          }
4009        }        }
4010    
4011      /* 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.
4012        If we are at the end of an internal nested string, revert to the outer
4013        string. */
4014    
4015        while (((c = *(++ptr)) != 0 ||
4016               (nestptr != NULL &&
4017                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4018               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4019    
4020      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4021    
4022      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4023        {        {
4024        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4025        goto FAILED;        goto FAILED;
4026        }        }
4027    
   
 /* 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  
   
   
4028      /* 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
4029      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
4030      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 3628  we set the flag only if there is a liter Line 4032  we set the flag only if there is a liter
4032    
4033      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
4034      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4035      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4036      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4037    
4038      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
4039      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.
4040      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
4041      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4042      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
4043      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4044    
4045  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4046      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3647  we set the flag only if there is a liter Line 4051  we set the flag only if there is a liter
4051        {        {
4052        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4053    
4054        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4055    
4056        if (negate_class)        if (negate_class)
4057          {          {
4058          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4059          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4060          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4061          *code++ = class_lastchar;          *code++ = class_lastchar;
4062          break;          break;
4063          }          }
# Line 3684  we set the flag only if there is a liter Line 4088  we set the flag only if there is a liter
4088    
4089      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4090      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4091      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
4092      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
4093      (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
4094      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
4095        actual compiled code. */
4096    
4097  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4098      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4099        {        {
4100        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4101        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3716  we set the flag only if there is a liter Line 4121  we set the flag only if there is a liter
4121        }        }
4122  #endif  #endif
4123    
4124      /* 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
4125      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
4126      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
4127      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4128        negating it if necessary. */
4129    
4130      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4131      if (negate_class)      if (negate_class)
# Line 3802  we set the flag only if there is a liter Line 4208  we set the flag only if there is a liter
4208        ptr++;        ptr++;
4209        }        }
4210      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4211    
4212        /* If previous was a recursion call, wrap it in atomic brackets so that
4213        previous becomes the atomic group. All recursions were so wrapped in the
4214        past, but it no longer happens for non-repeated recursions. In fact, the
4215        repeated ones could be re-implemented independently so as not to need this,
4216        but for the moment we rely on the code for repeating groups. */
4217    
4218        if (*previous == OP_RECURSE)
4219          {
4220          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4221          *previous = OP_ONCE;
4222          PUT(previous, 1, 2 + 2*LINK_SIZE);
4223          previous[2 + 2*LINK_SIZE] = OP_KET;
4224          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4225          code += 2 + 2 * LINK_SIZE;
4226          length_prevgroup = 3 + 3*LINK_SIZE;
4227    
4228          /* When actually compiling, we need to check whether this was a forward
4229          reference, and if so, adjust the offset. */
4230    
4231          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4232            {
4233            int offset = GET(cd->hwm, -LINK_SIZE);
4234            if (offset == previous + 1 - cd->start_code)
4235              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4236            }
4237          }
4238    
4239        /* Now handle repetition for the different types of item. */
4240    
4241      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4242      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
# Line 3809  we set the flag only if there is a liter Line 4244  we set the flag only if there is a liter
4244      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 firstbyte
4245      instead.  */      instead.  */
4246    
4247      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4248        {        {
4249          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4250    
4251        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4252        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
4253        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 0x80 to flag that it's a
# Line 3843  we set the flag only if there is a liter Line 4280  we set the flag only if there is a liter
4280    
4281        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4282            repeat_max < 0 &&            repeat_max < 0 &&
4283            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4284          {          {
4285          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4286          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3856  we set the flag only if there is a liter Line 4292  we set the flag only if there is a liter
4292      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4293      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-
4294      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4295      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
4296      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4297    
4298      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4299        {        {
4300        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4301        c = previous[1];        c = previous[1];
4302        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4303            repeat_max < 0 &&            repeat_max < 0 &&
4304            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4305          {          {
4306          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4307          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3889  we set the flag only if there is a liter Line 4325  we set the flag only if there is a liter
4325    
4326        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4327            repeat_max < 0 &&            repeat_max < 0 &&
4328            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4329          {          {
4330          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4331          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3911  we set the flag only if there is a liter Line 4347  we set the flag only if there is a liter
4347    
4348        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4349    
4350        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4351        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4352        removed: */        removed: */
4353    
4354        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4355        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4356    
4357        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4358        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4359    
4360        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4361    
# Line 4058  we set the flag only if there is a liter Line 4494  we set the flag only if there is a liter
4494  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4495               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4496  #endif  #endif
4497               *previous == OP_REF)               *previous == OP_REF ||
4498                 *previous == OP_REFI)
4499        {        {
4500        if (repeat_max == 0)        if (repeat_max == 0)
4501          {          {
# Line 4066  we set the flag only if there is a liter Line 4503  we set the flag only if there is a liter
4503          goto END_REPEAT;          goto END_REPEAT;
4504          }          }
4505    
4506        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4507        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4508        removed: */        removed: */
4509    
# Line 4074  we set the flag only if there is a liter Line 4511  we set the flag only if there is a liter
4511        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4512    
4513        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4514        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4515    
4516        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4517          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 4092  we set the flag only if there is a liter Line 4529  we set the flag only if there is a liter
4529        }        }
4530    
4531      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4532      cases. */      cases. Note that at this point we can encounter only the "basic" BRA and
4533        KET opcodes, as this is the place where they get converted into the more
4534        special varieties. */
4535    
4536      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4537               *previous == OP_ONCE || *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4538        {        {
4539        register int i;        register int i;
4540        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4541        uschar *bralink = NULL;        uschar *bralink = NULL;
4542          uschar *brazeroptr = NULL;
4543    
4544        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
4545    
4546        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
# Line 4110  we set the flag only if there is a liter Line 4549  we set the flag only if there is a liter
4549          goto FAILED;          goto FAILED;
4550          }          }
4551    
       /* If the maximum repeat count is unlimited, find the end of the bracket  
       by scanning through from the start, and compute the offset back to it  
       from the current code pointer. There may be an OP_OPT setting following  
       the final KET, so we can't find the end just by going back from the code  
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
         }  
   
4552        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
4553        OP_BRAZERO in front of it, and because the group appears once in the        OP_BRAZERO in front of it, and because the group appears once in the
4554        data, whereas in other cases it appears the minimum number of times. For        data, whereas in other cases it appears the minimum number of times. For
# Line 4164  we set the flag only if there is a liter Line 4590  we set the flag only if there is a liter
4590              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4591              goto END_REPEAT;              goto END_REPEAT;
4592              }              }
4593              brazeroptr = previous;    /* Save for possessive optimizing */
4594            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4595            }            }
4596    
# Line 4188  we set the flag only if there is a liter Line 4615  we set the flag only if there is a liter
4615            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4616            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4617    
4618            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4619            bralink = previous;            bralink = previous;
4620            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4621            }            }
# Line 4209  we set the flag only if there is a liter Line 4636  we set the flag only if there is a liter
4636            {            {
4637            /* In the pre-compile phase, we don't actually do the replication. We            /* In the pre-compile phase, we don't actually do the replication. We
4638            just adjust the length as if we had. Do some paranoid checks for            just adjust the length as if we had. Do some paranoid checks for
4639            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4640              integer type when available, otherwise double. */
4641    
4642            if (lengthptr != NULL)            if (lengthptr != NULL)
4643              {              {
4644              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4645              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4646                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4647                        (INT64_OR_DOUBLE)INT_MAX ||
4648                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4649                {                {
4650                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4261  we set the flag only if there is a liter Line 4690  we set the flag only if there is a liter
4690          just adjust the length as if we had. For each repetition we must add 1          just adjust the length as if we had. For each repetition we must add 1
4691          to the length for BRAZERO and for all but the last repetition we must          to the length for BRAZERO and for all but the last repetition we must
4692          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4693          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4694            a 64-bit integer type when available, otherwise double. */
4695    
4696          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4697            {            {
4698            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4699                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4700            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4701                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4702                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4703                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4704              {              {
4705              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4294  we set the flag only if there is a liter Line 4724  we set the flag only if there is a liter
4724              {              {
4725              int offset;              int offset;
4726              *code++ = OP_BRA;              *code++ = OP_BRA;
4727              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4728              bralink = code;              bralink = code;
4729              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4730              }              }
# Line 4315  we set the flag only if there is a liter Line 4745  we set the flag only if there is a liter
4745          while (bralink != NULL)          while (bralink != NULL)
4746            {            {
4747            int oldlinkoffset;            int oldlinkoffset;
4748            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4749            uschar *bra = code - offset;            uschar *bra = code - offset;
4750            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4751            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4325  we set the flag only if there is a liter Line 4755  we set the flag only if there is a liter
4755            }            }
4756          }          }
4757    
4758        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
4759        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4760        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
4761        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4762          deal with possessive ONCEs specially.
4763    
4764          Otherwise, if the quantifier was possessive, we convert the BRA code to
4765          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4766          at runtime to detect this kind of subpattern at both the start and at the
4767          end.) The use of special opcodes makes it possible to reduce greatly the
4768          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4769          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4770          the default action below, of wrapping everything inside atomic brackets,
4771          does not happen.
4772    
4773        Then, when we are doing the actual compile phase, check to see whether        Then, when we are doing the actual compile phase, check to see whether
4774        this group is a non-atomic one that could match an empty string. If so,        this group is one that could match an empty string. If so, convert the
4775        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime
4776        that runtime checking can be done. [This check is also applied to        checking can be done. [This check is also applied to ONCE groups at
4777        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4778    
4779        else        else
4780          {          {
4781          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4782          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4783          *ketcode = OP_KETRMAX + repeat_type;  
4784          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4785            if (*bracode == OP_ONCE)
4786              *ketcode = OP_KETRMAX + repeat_type;
4787            else
4788            {            {
4789            uschar *scode = bracode;            if (possessive_quantifier)
4790            do              {
4791                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4792                *ketcode = OP_KETRPOS;
4793                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4794                possessive_quantifier = FALSE;
4795                }
4796              else *ketcode = OP_KETRMAX + repeat_type;
4797    
4798              if (lengthptr == NULL)
4799              {              {
4800              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
4801                do
4802                {                {
4803                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4804                break;                  {
4805                    *bracode += OP_SBRA - OP_BRA;
4806                    break;
4807                    }
4808                  scode += GET(scode, 1);
4809                }                }
4810              scode += GET(scode, 1);              while (*scode == OP_ALT);
4811              }              }
           while (*scode == OP_ALT);  
4812            }            }
4813          }          }
4814        }        }
# Line 4374  we set the flag only if there is a liter Line 4829  we set the flag only if there is a liter
4829        }        }
4830    
4831      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4832      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4833      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4834      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4835      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
4836      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4837      tempcode, not at previous, which might be the first part of a string whose  
4838      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4839        just above, so possessive_quantifier is always FALSE for them at this
4840        stage.
4841    
4842        Note that the repeated item starts at tempcode, not at previous, which
4843        might be the first part of a string whose (former) last char we repeated.
4844    
4845      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4846      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4389  we set the flag only if there is a liter Line 4849  we set the flag only if there is a liter
4849      if (possessive_quantifier)      if (possessive_quantifier)
4850        {        {
4851        int len;        int len;
4852    
4853        if (*tempcode == OP_TYPEEXACT)        if (*tempcode == OP_TYPEEXACT)
4854          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4855            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4856    
4857        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4858          {          {
4859          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
# Line 4401  we set the flag only if there is a liter Line 4861  we set the flag only if there is a liter
4861          if (utf8 && tempcode[-1] >= 0xc0)          if (utf8 && tempcode[-1] >= 0xc0)
4862            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4863  #endif  #endif
4864          }          }
4865    
4866        len = code - tempcode;        len = (int)(code - tempcode);
4867        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4868          {          {
4869          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4411  we set the flag only if there is a liter Line 4871  we set the flag only if there is a liter
4871          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4872          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4873    
4874          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4875          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4876          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4877          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4878    
4879          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4880          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4881          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4882          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4883    
4884            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4885            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4886            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4887            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4888    
4889            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4890            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4891            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4892            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4893    
4894            /* Because we are moving code along, we must ensure that any
4895            pending recursive references are updated. */
4896    
4897          default:          default:
4898            *code = OP_END;
4899            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4900          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4901          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4902          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4457  we set the flag only if there is a liter Line 4932  we set the flag only if there is a liter
4932    
4933      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4934    
4935      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4936             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4937        {        {
4938        int i, namelen;        int i, namelen;
4939          int arglen = 0;
4940        const char *vn = verbnames;        const char *vn = verbnames;
4941        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4942          const uschar *arg = NULL;
4943        previous = NULL;        previous = NULL;
4944        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4945          namelen = (int)(ptr - name);
4946    
4947        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
4948          {          {
4949          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4950          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4951              || *ptr == '_') ptr++;
4952            arglen = (int)(ptr - arg);
4953          }          }
4954    
4955        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4956          {          {
4957          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4958          goto FAILED;          goto FAILED;
4959          }          }
4960        namelen = ptr - name;  
4961          /* Scan the table of verb names */
4962    
4963        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4964          {          {
4965          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4966              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4967            {            {
4968            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
4969              ASSERT_ACCEPT if in an assertion. */
4970    
4971            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
4972              {              {
4973              open_capitem *oc;              open_capitem *oc;
4974              cd->had_accept = TRUE;              if (arglen != 0)
4975                  {
4976                  *errorcodeptr = ERR59;
4977                  goto FAILED;
4978                  }
4979                cd->had_accept = TRUE;
4980              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4981                {                {
4982                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
4983                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
4984                }                }
4985              }              *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
4986            *code++ = verbs[i].op;              }
4987            break;  
4988              /* Handle other cases with/without an argument */
4989    
4990              else if (arglen == 0)
4991                {
4992                if (verbs[i].op < 0)   /* Argument is mandatory */
4993                  {
4994                  *errorcodeptr = ERR66;
4995                  goto FAILED;
4996                  }
4997                *code = verbs[i].op;
4998                if (*code++ == OP_THEN)
4999                  {
5000                  PUT(code, 0, code - bcptr->current_branch - 1);
5001                  code += LINK_SIZE;
5002                  }
5003                }
5004    
5005              else
5006                {
5007                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
5008                  {
5009                  *errorcodeptr = ERR59;
5010                  goto FAILED;
5011                  }
5012                *code = verbs[i].op_arg;
5013                if (*code++ == OP_THEN_ARG)
5014                  {
5015                  PUT(code, 0, code - bcptr->current_branch - 1);
5016                  code += LINK_SIZE;
5017                  }
5018                *code++ = arglen;
5019                memcpy(code, arg, arglen);
5020                code += arglen;
5021                *code++ = 0;
5022                }
5023    
5024              break;  /* Found verb, exit loop */
5025            }            }
5026    
5027          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
5028          }          }
5029        if (i < verbcount) continue;  
5030        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
5031          *errorcodeptr = ERR60;          /* Verb not recognized */
5032        goto FAILED;        goto FAILED;
5033        }        }
5034    
# Line 4617  we set the flag only if there is a liter Line 5147  we set the flag only if there is a liter
5147                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5148            ptr++;            ptr++;
5149            }            }
5150          namelen = ptr - name;          namelen = (int)(ptr - name);
5151    
5152          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5153              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4654  we set the flag only if there is a liter Line 5184  we set the flag only if there is a liter
5184            }            }
5185    
5186          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5187          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
5188            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5189            except they record that the reference was originally to a name. The
5190            information is used to check duplicate names. */
5191    
5192          slot = cd->name_table;          slot = cd->name_table;
5193          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4669  we set the flag only if there is a liter Line 5202  we set the flag only if there is a liter
5202            {            {
5203            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5204            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5205              code[1+LINK_SIZE]++;
5206            }            }
5207    
5208          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5209    
5210          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5211                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5212            {            {
5213            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5214              code[1+LINK_SIZE]++;
5215            }            }
5216    
5217          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4741  we set the flag only if there is a liter Line 5276  we set the flag only if there is a liter
5276          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5277          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5278          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5279            cd->assert_depth += 1;
5280          ptr++;          ptr++;
5281          break;          break;
5282    
# Line 4755  we set the flag only if there is a liter Line 5291  we set the flag only if there is a liter
5291            continue;            continue;
5292            }            }
5293          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5294            cd->assert_depth += 1;
5295          break;          break;
5296    
5297    
# Line 4764  we set the flag only if there is a liter Line 5301  we set the flag only if there is a liter
5301            {            {
5302            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5303            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5304              cd->assert_depth += 1;
5305            ptr += 2;            ptr += 2;
5306            break;            break;
5307    
5308            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5309            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5310              cd->assert_depth += 1;
5311            ptr += 2;            ptr += 2;
5312            break;            break;
5313    
# Line 4808  we set the flag only if there is a liter Line 5347  we set the flag only if there is a liter
5347              goto FAILED;              goto FAILED;
5348              }              }
5349            *code++ = n;            *code++ = n;
5350            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5351            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5352            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5353            }            }
5354          previous = NULL;          previous = NULL;
# Line 4842  we set the flag only if there is a liter Line 5381  we set the flag only if there is a liter
5381            name = ++ptr;            name = ++ptr;
5382    
5383            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5384            namelen = ptr - name;            namelen = (int)(ptr - name);
5385    
5386            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5387    
# Line 4878  we set the flag only if there is a liter Line 5417  we set the flag only if there is a liter
5417            is because the number of names, and hence the table size, is computed            is because the number of names, and hence the table size, is computed
5418            in the pre-compile, and it affects various numbers and pointers which            in the pre-compile, and it affects various numbers and pointers which
5419            would all have to be modified, and the compiled code moved down, if            would all have to be modified, and the compiled code moved down, if
5420            duplicates with the same number were omitted from the table. This            duplicates with the same number were omitted from the table. This
5421            doesn't seem worth the hassle. However, *different* names for the            doesn't seem worth the hassle. However, *different* names for the
5422            same number are not permitted. */            same number are not permitted. */
5423    
# Line 4886  we set the flag only if there is a liter Line 5425  we set the flag only if there is a liter
5425              {              {
5426              BOOL dupname = FALSE;              BOOL dupname = FALSE;
5427              slot = cd->name_table;              slot = cd->name_table;
5428    
5429              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5430                {                {
5431                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4900  we set the flag only if there is a liter Line 5439  we set the flag only if there is a liter
5439                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5440                      goto FAILED;                      goto FAILED;
5441                      }                      }
5442                    else dupname = TRUE;                    else dupname = TRUE;
5443                    }                    }
5444                  else crc = -1;      /* Current name is a substring */                  else crc = -1;      /* Current name is a substring */
5445                  }                  }
5446    
5447                /* Make space in the table and break the loop for an earlier                /* Make space in the table and break the loop for an earlier
5448                name. For a duplicate or later name, carry on. We do this for                name. For a duplicate or later name, carry on. We do this for
5449                duplicates so that in the simple case (when ?(| is not used) they                duplicates so that in the simple case (when ?(| is not used) they
5450                are in order of their numbers. */                are in order of their numbers. */
5451    
5452                if (crc < 0)                if (crc < 0)
5453                  {                  {
5454                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5455                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5456                  break;                  break;
5457                  }                  }
5458    
5459                /* Continue the loop for a later or duplicate name */                /* Continue the loop for a later or duplicate name */
5460    
5461                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5462                }                }
5463    
5464              /* For non-duplicate names, check for a duplicate number before              /* For non-duplicate names, check for a duplicate number before
5465              adding the new name. */              adding the new name. */
5466    
5467              if (!dupname)              if (!dupname)
5468                {                {
5469                uschar *cslot = cd->name_table;                uschar *cslot = cd->name_table;
# Line 4936  we set the flag only if there is a liter Line 5475  we set the flag only if there is a liter
5475                      {                      {
5476                      *errorcodeptr = ERR65;                      *errorcodeptr = ERR65;
5477                      goto FAILED;                      goto FAILED;
5478                      }                      }
5479                    }                    }
5480                  else i--;                  else i--;
5481                  cslot += cd->name_entry_size;                  cslot += cd->name_entry_size;
5482                  }                  }
5483                }                }
5484    
5485              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5486              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
# Line 4972  we set the flag only if there is a liter Line 5511  we set the flag only if there is a liter
5511          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5512          name = ++ptr;          name = ++ptr;
5513          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5514          namelen = ptr - name;          namelen = (int)(ptr - name);
5515    
5516          /* In the pre-compile phase, do a syntax check and set a dummy          /* In the pre-compile phase, do a syntax check. We used to just set
5517          reference number. */          a dummy reference number, because it was not used in the first pass.
5518            However, with the change of recursive back references to be atomic,
5519            we have to look for the number so that this state can be identified, as
5520            otherwise the incorrect length is computed. If it's not a backwards
5521            reference, the dummy number will do. */
5522    
5523          if (lengthptr != NULL)          if (lengthptr != NULL)
5524            {            {
5525              const uschar *temp;
5526    
5527            if (namelen == 0)            if (namelen == 0)
5528              {              {
5529              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 4994  we set the flag only if there is a liter Line 5539  we set the flag only if there is a liter
5539              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5540              goto FAILED;              goto FAILED;
5541              }              }
5542            recno = 0;  
5543              /* The name table does not exist in the first pass, so we cannot
5544              do a simple search as in the code below. Instead, we have to scan the
5545              pattern to find the number. It is important that we scan it only as
5546              far as we have got because the syntax of named subpatterns has not
5547              been checked for the rest of the pattern, and find_parens() assumes
5548              correct syntax. In any case, it's a waste of resources to scan
5549              further. We stop the scan at the current point by temporarily
5550              adjusting the value of cd->endpattern. */
5551    
5552              temp = cd->end_pattern;
5553              cd->end_pattern = ptr;
5554              recno = find_parens(cd, name, namelen,
5555                (options & PCRE_EXTENDED) != 0, utf8);
5556              cd->end_pattern = temp;
5557              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5558            }            }
5559    
5560          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, seek the name in the table. We check the name
# Line 5019  we set the flag only if there is a liter Line 5579  we set the flag only if there is a liter
5579              }              }
5580            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5581                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5582                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5583              {              {
5584              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5585              goto FAILED;              goto FAILED;
# Line 5122  we set the flag only if there is a liter Line 5682  we set the flag only if there is a liter
5682            if (lengthptr == NULL)            if (lengthptr == NULL)
5683              {              {
5684              *code = OP_END;              *code = OP_END;
5685              if (recno != 0)              if (recno != 0)
5686                called = _pcre_find_bracket(cd->start_code, utf8, recno);                called = _pcre_find_bracket(cd->start_code, utf8, recno);
5687    
5688              /* Forward reference */              /* Forward reference */
# Line 5130  we set the flag only if there is a liter Line 5690  we set the flag only if there is a liter
5690              if (called == NULL)              if (called == NULL)
5691                {                {
5692                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5693                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5694                  {                  {
5695                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5696                  goto FAILED;                  goto FAILED;
5697                  }                  }
5698    
5699                  /* Fudge the value of "called" so that when it is inserted as an
5700                  offset below, what it actually inserted is the reference number
5701                  of the group. Then remember the forward reference. */
5702    
5703                called = cd->start_code + recno;                called = cd->start_code + recno;
5704                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5705                }                }
5706    
5707              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
# Line 5144  we set the flag only if there is a liter Line 5709  we set the flag only if there is a liter
5709              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5710    
5711              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5712                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5713                {                {
5714                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5715                goto FAILED;                goto FAILED;
5716                }                }
5717              }              }
5718    
5719            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
5720            "once" brackets. Set up a "previous group" length so that a  
           subsequent quantifier will work. */  
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
   
5721            *code = OP_RECURSE;            *code = OP_RECURSE;
5722            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
5723            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
   
           *code = OP_KET;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
5724            }            }
5725    
5726          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5228  we set the flag only if there is a liter Line 5781  we set the flag only if there is a liter
5781          is necessary to ensure we correctly detect the start of the pattern in          is necessary to ensure we correctly detect the start of the pattern in
5782          both phases.          both phases.
5783    
5784          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, reset the greedy defaults and the
5785          options if this setting actually changes any of them, and reset the          case value for firstbyte and reqbyte. */
         greedy defaults and the case value for firstbyte and reqbyte. */  
5786    
5787          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5788            {            {
# Line 5239  we set the flag only if there is a liter Line 5791  we set the flag only if there is a liter
5791              {              {
5792              cd->external_options = newoptions;              cd->external_options = newoptions;
5793              }              }
5794           else            else
5795              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5796              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5797              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5798              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5799              }              }
5800    
5801            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5802            in subsequent branches. When not at the start of the pattern, this            in subsequent branches. */
           information is also necessary so that a resetting item can be  
           compiled at the end of a group (if we are in a group). */  
5803    
5804            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5805            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5271  we set the flag only if there is a liter Line 5816  we set the flag only if there is a liter
5816          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5817        }       /* End of (? handling */        }       /* End of (? handling */
5818    
5819      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5820      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5821      brackets. */      brackets. */
5822    
5823      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5292  we set the flag only if there is a liter Line 5837  we set the flag only if there is a liter
5837    
5838      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions may not be repeated, but
5839      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a
5840      non-register variable in order to be able to pass its address because some      non-register variable (tempcode) in order to be able to pass its address
5841      compilers complain otherwise. Pass in a new setting for the ims options if      because some compilers complain otherwise. */
     they have changed. */  
5842    
5843      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = (bravalue >= OP_ONCE)? code : NULL;
5844      *code = bravalue;      *code = bravalue;
# Line 5304  we set the flag only if there is a liter Line 5848  we set the flag only if there is a liter
5848    
5849      if (!compile_regex(      if (!compile_regex(
5850           newoptions,                   /* The complete new option state */           newoptions,                   /* The complete new option state */
          options & PCRE_IMS,           /* The previous ims option state */  
5851           &tempcode,                    /* Where to put code (updated) */           &tempcode,                    /* Where to put code (updated) */
5852           &ptr,                         /* Input pointer (updated) */           &ptr,                         /* Input pointer (updated) */
5853           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
# Line 5320  we set the flag only if there is a liter Line 5863  we set the flag only if there is a liter
5863             &length_prevgroup           /* Pre-compile phase */             &length_prevgroup           /* Pre-compile phase */
5864           ))           ))
5865        goto FAILED;        goto FAILED;
5866    
5867        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5868          cd->assert_depth -= 1;
5869    
5870      /* At the end of compiling, code is still pointing to the start of the      /* At the end of compiling, code is still pointing to the start of the
5871      group, while tempcode has been updated to point past the end of the group      group, while tempcode has been updated to point past the end of the group
# Line 5391  we set the flag only if there is a liter Line 5937  we set the flag only if there is a liter
5937          goto FAILED;          goto FAILED;
5938          }          }
5939        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5940        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
5941        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5942        *code++ = OP_KET;        *code++ = OP_KET;
5943        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5464  we set the flag only if there is a liter Line 6010  we set the flag only if there is a liter
6010    
6011      /* ===================================================================*/      /* ===================================================================*/
6012      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
6013      are arranged to be the negation of the corresponding OP_values. For the      are arranged to be the negation of the corresponding OP_values in the
6014      back references, the values are ESC_REF plus the reference number. Only      default case when PCRE_UCP is not set. For the back references, the values
6015      back references and those types that consume a character may be repeated.      are ESC_REF plus the reference number. Only back references and those types
6016      We can test for values between ESC_b and ESC_Z for the latter; this may      that consume a character may be repeated. We can test for values between
6017      have to change if any new ones are ever created. */      ESC_b and ESC_Z for the latter; this may have to change if any new ones are
6018        ever created. */
6019    
6020      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
6021      tempptr = ptr;      tempptr = ptr;
# Line 5576  we set the flag only if there is a liter Line 6123  we set the flag only if there is a liter
6123    
6124        if (-c >= ESC_REF)        if (-c >= ESC_REF)
6125          {          {
6126            open_capitem *oc;
6127          recno = -c - ESC_REF;          recno = -c - ESC_REF;
6128    
6129          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6130          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6131          previous = code;          previous = code;
6132          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6133          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6134          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6135          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
6136    
6137            /* Check to see if this back reference is recursive, that it, it
6138            is inside the group that it references. A flag is set so that the
6139            group can be made atomic. */
6140    
6141            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6142              {
6143              if (oc->number == recno)
6144                {
6145                oc->flag = TRUE;
6146                break;
6147                }
6148              }
6149          }          }
6150    
6151        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5614  we set the flag only if there is a liter Line 6175  we set the flag only if there is a liter
6175  #endif  #endif
6176    
6177        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
6178        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
6179          situation when PCRE_UCP is not set. When it *is* set, we substitute
6180          Unicode property tests. */
6181    
6182        else        else
6183          {          {
6184          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
6185          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
6186              {
6187              nestptr = ptr + 1;                   /* Where to resume */
6188              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6189              }
6190            else
6191    #endif
6192              {
6193              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6194              *code++ = -c;
6195              }
6196          }          }
6197        continue;        continue;
6198        }        }
# Line 5664  we set the flag only if there is a liter Line 6237  we set the flag only if there is a liter
6237    
6238      ONE_CHAR:      ONE_CHAR:
6239      previous = code;      previous = code;
6240      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6241      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6242    
6243      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 5728  return FALSE; Line 6301  return FALSE;
6301  /* On entry, ptr is pointing past the bracket character, but on return it  /* On entry, ptr is pointing past the bracket character, but on return it
6302  points to the closing bracket, or vertical bar, or end of string. The code  points to the closing bracket, or vertical bar, or end of string. The code
6303  variable is pointing at the byte into which the BRA operator has been stored.  variable is pointing at the byte into which the BRA operator has been stored.
 If the ims options are changed at the start (for a (?ims: group) or during any  
 branch, we need to insert an OP_OPT item at the start of every following branch  
 to ensure they get set correctly at run time, and also pass the new options  
 into every subsequent branch compile.  
   
6304  This function is used during the pre-compile phase when we are trying to find  This function is used during the pre-compile phase when we are trying to find
6305  out the amount of memory needed, as well as during the real compile phase. The  out the amount of memory needed, as well as during the real compile phase. The
6306  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6307    
6308  Arguments:  Arguments:
6309    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
   oldims         previous settings of ims option bits  
6310    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6311    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6312    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
# Line 5757  Returns:         TRUE on success Line 6324  Returns:         TRUE on success
6324  */  */
6325    
6326  static BOOL  static BOOL
6327  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6328    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6329    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
6330    int *lengthptr)    int *lengthptr)
# Line 5777  int max_bracount; Line 6344  int max_bracount;
6344  branch_chain bc;  branch_chain bc;
6345    
6346  bc.outer = bcptr;  bc.outer = bcptr;
6347  bc.current = code;  bc.current_branch = code;
6348    
6349  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
6350    
# Line 5796  them global. It tests the value of lengt Line 6363  them global. It tests the value of lengt
6363  pre-compile phase to find out whether anything has yet been compiled or not. */  pre-compile phase to find out whether anything has yet been compiled or not. */
6364    
6365  /* If this is a capturing subpattern, add to the chain of open capturing items  /* If this is a capturing subpattern, add to the chain of open capturing items
6366  so that we can detect them if (*ACCEPT) is encountered. */  so that we can detect them if (*ACCEPT) is encountered. This is also used to
6367    detect groups that contain recursive back references to themselves. Note that
6368    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6369    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6370    
6371  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6372    {    {
6373    capnumber = GET2(code, 1 + LINK_SIZE);    capnumber = GET2(code, 1 + LINK_SIZE);
6374    capitem.number = capnumber;    capitem.number = capnumber;
6375    capitem.next = cd->open_caps;    capitem.next = cd->open_caps;
6376    cd->open_caps = &capitem;    capitem.flag = FALSE;
6377    }    cd->open_caps = &capitem;
6378      }
6379    
6380  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
6381    
# Line 5821  for (;;) Line 6392  for (;;)
6392    
6393    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6394    
   /* Handle a change of ims options at the start of the branch */  
   
   if ((options & PCRE_IMS) != oldims)  
     {  
     *code++ = OP_OPT;  
     *code++ = options & PCRE_IMS;  
     length += 2;  
     }  
   
6395    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6396    
6397    if (lookbehind)    if (lookbehind)
# Line 5900  for (;;) Line 6462  for (;;)
6462    
6463      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6464      put the length into the OP_REVERSE item. Temporarily mark the end of the      put the length into the OP_REVERSE item. Temporarily mark the end of the
6465      branch with OP_END. If the branch contains OP_RECURSE, the result is -3      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
6466      because there may be forward references that we can't check here. Set a      because there may be forward references that we can't check here. Set a
6467      flag to cause another lookbehind check at the end. Why not do it all at the      flag to cause another lookbehind check at the end. Why not do it all at the
6468      end? Because common, erroneous checks are picked up here and the offset of      end? Because common, erroneous checks are picked up here and the offset of
6469      the problem can be shown. */      the problem can be shown. */
6470    
6471      if (lookbehind)      if (lookbehind)
6472        {        {
6473        int fixed_length;        int fixed_length;
6474        *code = OP_END;        *code = OP_END;
6475        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6476            FALSE, cd);
6477        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6478        if (fixed_length == -3)        if (fixed_length == -3)
6479          {          {
6480          cd->check_lookbehind = TRUE;          cd->check_lookbehind = TRUE;
6481          }          }
6482        else if (fixed_length < 0)        else if (fixed_length < 0)
6483          {          {
6484          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
# Line 5931  for (;;) Line 6494  for (;;)
6494    of offsets, with the field in the BRA item now becoming an offset to the    of offsets, with the field in the BRA item now becoming an offset to the
6495    first alternative. If there are no alternatives, it points to the end of the    first alternative. If there are no alternatives, it points to the end of the
6496    group. The length in the terminating ket is always the length of the whole    group. The length in the terminating ket is always the length of the whole
6497    bracketed item. If any of the ims options were changed inside the group,    bracketed item. Return leaving the pointer at the terminating char. */
   compile a resetting op-code following, except at the very end of the pattern.  
   Return leaving the pointer at the terminating char. */  
6498    
6499    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6500      {      {
6501      if (lengthptr == NULL)      if (lengthptr == NULL)
6502        {        {
6503        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6504        do        do
6505          {          {
6506          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 5949  for (;;) Line 6510  for (;;)
6510          }          }
6511        while (branch_length > 0);        while (branch_length > 0);
6512        }        }
   
     /* If it was a capturing subpattern, remove it from the chain. */  
   
     if (capnumber > 0) cd->open_caps = cd->open_caps->next;  
6513    
6514      /* Fill in the ket */      /* Fill in the ket */
6515    
6516      *code = OP_KET;      *code = OP_KET;
6517      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6518      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6519    
6520      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6521        recursive back references. If so, we must wrap it in atomic brackets.
6522        In any event, remove the block from the chain. */
6523    
6524      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)      if (capnumber > 0)
6525        {        {
6526        *code++ = OP_OPT;        if (cd->open_caps->flag)
6527        *code++ = oldims;          {
6528        length += 2;          memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6529              code - start_bracket);
6530            *start_bracket = OP_ONCE;
6531            code += 1 + LINK_SIZE;
6532            PUT(start_bracket, 1, (int)(code - start_bracket));
6533            *code = OP_KET;
6534            PUT(code, 1, (int)(code - start_bracket));
6535            code += 1 + LINK_SIZE;
6536            length += 2 + 2*LINK_SIZE;
6537            }
6538          cd->open_caps = cd->open_caps->next;
6539        }        }
6540    
6541      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
# Line 6008  for (;;) Line 6577  for (;;)
6577    else    else
6578      {      {
6579      *code = OP_ALT;      *code = OP_ALT;
6580      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6581      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6582      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6583      }      }
6584    
# Line 6028  for (;;) Line 6597  for (;;)
6597  /* Try to find out if this is an anchored regular expression. Consider each  /* Try to find out if this is an anchored regular expression. Consider each
6598  alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket  alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket
6599  all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then  all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then
6600  it's anchored. However, if this is a multiline pattern, then only OP_SOD  it's anchored. However, if this is a multiline pattern, then only OP_SOD will
6601  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6602    
6603  We can also consider a regex to be anchored if OP_SOM starts all its branches.  We can also consider a regex to be anchored if OP_SOM starts all its branches.
6604  This is the code for \G, which means "match at start of match position, taking  This is the code for \G, which means "match at start of match position, taking
# Line 6050  of the more common cases more precisely. Line 6619  of the more common cases more precisely.
6619    
6620  Arguments:  Arguments:
6621    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6622    bracket_map    a bitmap of which brackets we are inside while testing; this    bracket_map    a bitmap of which brackets we are inside while testing; this
6623                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6624                    the less precise approach                    the less precise approach
# Line 6060  Returns:     TRUE or FALSE Line 6628  Returns:     TRUE or FALSE
6628  */  */
6629    
6630  static BOOL  static BOOL
6631  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6632    unsigned int backref_map)    unsigned int backref_map)
6633  {  {
6634  do {  do {
6635     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6636       options, PCRE_MULTILINE, FALSE);       FALSE);
6637     register int op = *scode;     register int op = *scode;
6638    
6639     /* Non-capturing brackets */     /* Non-capturing brackets */
6640    
6641     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6642           op == OP_SBRA || op == OP_SBRAPOS)
6643       {       {
6644       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6645       }       }
6646    
6647     /* Capturing brackets */     /* Capturing brackets */
6648    
6649     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6650                op == OP_SCBRA || op == OP_SCBRAPOS)
6651       {       {
6652       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6653       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6654       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6655       }       }
6656    
6657     /* Other brackets */     /* Other brackets */
6658    
6659     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)
6660       {       {
6661       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6662       }       }
6663    
6664     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
# Line 6103  do { Line 6673  do {
6673    
6674     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6675    
6676     else if (op != OP_SOD && op != OP_SOM &&     else if (op != OP_SOD && op != OP_SOM && op != OP_CIRC) return FALSE;
            ((*options & PCRE_MULTILINE) != 0 || op != OP_CIRC))  
      return FALSE;  
6677     code += GET(code, 1);     code += GET(code, 1);
6678     }     }
6679  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6141  is_startline(const uschar *code, unsigne Line 6709  is_startline(const uschar *code, unsigne
6709  {  {
6710  do {  do {
6711     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6712       NULL, 0, FALSE);       FALSE);
6713     register int op = *scode;     register int op = *scode;
6714    
6715     /* If we are at the start of a conditional assertion group, *both* the     /* If we are at the start of a conditional assertion group, *both* the
# Line 6156  do { Line 6724  do {
6724       switch (*scode)       switch (*scode)
6725         {         {
6726         case OP_CREF:         case OP_CREF:
6727           case OP_NCREF:
6728         case OP_RREF:         case OP_RREF:
6729           case OP_NRREF:
6730         case OP_DEF:         case OP_DEF:
6731         return FALSE;         return FALSE;
6732    
# Line 6166  do { Line 6736  do {
6736         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6737         break;         break;
6738         }         }
6739       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6740       op = *scode;       op = *scode;
6741       }       }
6742    
6743     /* Non-capturing brackets */     /* Non-capturing brackets */
6744    
6745     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6746           op == OP_SBRA || op == OP_SBRAPOS)
6747       {       {
6748       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6749       }       }
6750    
6751     /* Capturing brackets */     /* Capturing brackets */
6752    
6753     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6754                op == OP_SCBRA || op == OP_SCBRAPOS)
6755       {       {
6756       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6757       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6203  do { Line 6775  do {
6775    
6776     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6777    
6778     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6779    
6780     /* Move on to the next alternative */     /* Move on to the next alternative */
6781    
# Line 6229  we return that char, otherwise -1. Line 6801  we return that char, otherwise -1.
6801    
6802  Arguments:  Arguments:
6803    code       points to start of expression (the bracket)    code       points to start of expression (the bracket)
   options    pointer to the options (used to check casing changes)  
6804    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6805    
6806  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6807  */  */
6808    
6809  static int  static int
6810  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6811  {  {
6812  register int c = -1;  register int c = -1;
6813  do {  do {
6814     int d;     int d;
6815     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6816       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6817       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6818     register int op = *scode;     register int op = *scode;
6819    
6820     switch(op)     switch(op)
# Line 6251  do { Line 6823  do {
6823       return -1;       return -1;
6824    
6825       case OP_BRA:       case OP_BRA:
6826         case OP_BRAPOS:
6827       case OP_CBRA:       case OP_CBRA:
6828         case OP_SCBRA:
6829         case OP_CBRAPOS:
6830         case OP_SCBRAPOS:
6831       case OP_ASSERT:       case OP_ASSERT:
6832       case OP_ONCE:       case OP_ONCE:
6833       case OP_COND:       case OP_COND:
6834       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6835         return -1;         return -1;
6836       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6837       break;       break;
6838    
6839       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6840       scode += 2;       scode += 2;
6841         /* Fall through */
6842    
6843       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6844       case OP_PLUS:       case OP_PLUS:
6845       case OP_MINPLUS:       case OP_MINPLUS:
6846       case OP_POSPLUS:       case OP_POSPLUS:
6847       if (!inassert) return -1;       if (!inassert) return -1;
6848       if (c < 0)       if (c < 0) c = scode[1];
6849         {         else if (c != scode[1]) return -1;
6850         c = scode[1];       break;
6851         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6852         }       case OP_EXACTI:
6853       else if (c != scode[1]) return -1;       scode += 2;
6854         /* Fall through */
6855    
6856         case OP_CHARI:
6857         case OP_PLUSI:
6858         case OP_MINPLUSI:
6859         case OP_POSPLUSI:
6860         if (!inassert) return -1;
6861         if (c < 0) c = scode[1] | REQ_CASELESS;
6862           else if (c != scode[1]) return -1;
6863       break;       break;
6864       }       }
6865    
# Line 6325  int length = 1;  /* For final END opcode Line 6910  int length = 1;  /* For final END opcode
6910  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6911  int errorcode = 0;  int errorcode = 0;
6912  int skipatstart = 0;  int skipatstart = 0;
6913  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8;
6914  size_t size;  size_t size;
6915  uschar *code;  uschar *code;
6916  const uschar *codestart;  const uschar *codestart;
# Line 6395  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6980  while (ptr[skipatstart] == CHAR_LEFT_PAR
6980    
6981    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6982      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
6983      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
6984        { skipatstart += 6; options |= PCRE_UCP; continue; }
6985      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
6986        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
6987    
6988    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6989      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6419  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7008  while (ptr[skipatstart] == CHAR_LEFT_PAR
7008    else break;    else break;
7009    }    }
7010    
7011  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  utf8 = (options & PCRE_UTF8) != 0;
7012    
7013    /* Can't support UTF8 unless PCRE has been compiled to include the code. The
7014    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7015    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7016    not used here. */
7017    
7018  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7019  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7020       (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7021    {    {
7022    errorcode = ERR44;    errorcode = ERR44;
7023    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6436  if (utf8) Line 7030  if (utf8)
7030    }    }
7031  #endif  #endif
7032    
7033    /* Can't support UCP unless PCRE has been compiled to include the code. */
7034    
7035    #ifndef SUPPORT_UCP
7036    if ((options & PCRE_UCP) != 0)
7037      {
7038      errorcode = ERR67;
7039      goto PCRE_EARLY_ERROR_RETURN;
7040      }
7041    #endif
7042    
7043  /* Check validity of \R options. */  /* Check validity of \R options. */
7044    
7045  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 6529  outside can help speed up starting point Line 7133  outside can help speed up starting point
7133  ptr += skipatstart;  ptr += skipatstart;
7134  code = cworkspace;  code = cworkspace;
7135  *code = OP_BRA;  *code = OP_BRA;
7136  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7137    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7138  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7139    
7140  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 6564  regex compiled on a system with 4-byte p Line 7167  regex compiled on a system with 4-byte p
7167  pointers. */  pointers. */
7168    
7169  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
7170  re->size = size;  re->size = (int)size;
7171  re->options = cd->external_options;  re->options = cd->external_options;
7172  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7173  re->dummy1 = 0;  re->dummy1 = 0;
# Line 6585  field; this time it's used for rememberi Line 7188  field; this time it's used for rememberi
7188  */  */
7189    
7190  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7191    cd->assert_depth = 0;
7192  cd->bracount = 0;  cd->bracount = 0;
7193  cd->names_found = 0;  cd->names_found = 0;
7194  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 6603  of the function here. */ Line 7207  of the function here. */
7207  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7208  code = (uschar *)codestart;  code = (uschar *)codestart;
7209  *code = OP_BRA;  *code = OP_BRA;
7210  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0,
7211    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7212  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7213  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7214  re->flags = cd->external_flags;  re->flags = cd->external_flags;
# Line 6620  if debugging, leave the test till after Line 7224  if debugging, leave the test till after
7224    
7225  *code++ = OP_END;  *code++ = OP_END;
7226    
7227  #ifndef DEBUG  #ifndef PCRE_DEBUG
7228  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
7229  #endif  #endif
7230    
# Line 6635  while (errorcode == 0 && cd->hwm > cwork Line 7239  while (errorcode == 0 && cd->hwm > cwork
7239    recno = GET(codestart, offset);    recno = GET(codestart, offset);
7240    groupptr = _pcre_find_bracket(codestart, utf8, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
7241    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
7242      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7243    }    }
7244    
7245  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
# Line 6643  subpattern. */ Line 7247  subpattern. */
7247    
7248  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
7249    
7250  /* If there were any lookbehind assertions that contained OP_RECURSE  /* If there were any lookbehind assertions that contained OP_RECURSE
7251  (recursions or subroutine calls), a flag is set for them to be checked here,  (recursions or subroutine calls), a flag is set for them to be checked here,
7252  because they may contain forward references. Actual recursions can't be fixed  because they may contain forward references. Actual recursions can't be fixed
7253  length, but subroutine calls can. It is done like this so that those without  length, but subroutine calls can. It is done like this so that those without
# Line 6654  length, and set their lengths. */ Line 7258  length, and set their lengths. */
7258  if (cd->check_lookbehind)  if (cd->check_lookbehind)
7259    {    {
7260    uschar *cc = (uschar *)codestart;    uschar *cc = (uschar *)codestart;
7261    
7262    /* Loop, searching for OP_REVERSE items, and process those that do not have    /* Loop, searching for OP_REVERSE items, and process those that do not have
7263    their length set. (Actually, it will also re-process any that have a length    their length set. (Actually, it will also re-process any that have a length
7264    of zero, but that is a pathological case, and it does no harm.) When we find    of zero, but that is a pathological case, and it does no harm.) When we find
7265    one, we temporarily terminate the branch it is in while we scan it. */    one, we temporarily terminate the branch it is in while we scan it. */
7266    
7267    for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);    for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
7268         cc != NULL;         cc != NULL;
7269         cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))         cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
7270      {      {
7271      if (GET(cc, 1) == 0)      if (GET(cc, 1) == 0)
7272        {        {
7273        int fixed_length;        int fixed_length;
7274        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7275        int end_op = *be;        int end_op = *be;
7276        *be = OP_END;        *be = OP_END;
7277        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7278            cd);
7279        *be = end_op;        *be = end_op;
7280        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7281        if (fixed_length < 0)        if (fixed_length < 0)
7282          {          {
7283          errorcode = (fixed_length == -2)? ERR36 : ERR25;          errorcode = (fixed_length == -2)? ERR36 : ERR25;
7284          break;          break;
7285          }          }
7286        PUT(cc, 1, fixed_length);        PUT(cc, 1, fixed_length);
7287        }        }
7288      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
7289      }      }
7290    }    }
7291    
7292  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
7293    
# Line 6690  if (errorcode != 0) Line 7295  if (errorcode != 0)
7295    {    {
7296    (pcre_free)(re);    (pcre_free)(re);
7297    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
7298    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = (int)(ptr - (const uschar *)pattern);
7299    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
7300    *errorptr = find_error_text(errorcode);    *errorptr = find_error_text(errorcode);
7301    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
# Line 6709  start with ^. and also when all branches Line 7314  start with ^. and also when all branches
7314    
7315  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7316    {    {
7317    int temp_options = re->options;   /* May get changed during these scans */    if (is_anchored(codestart, 0, cd->backref_map))
   if (is_anchored(codestart, &temp_options, 0, cd->backref_map))  
7318      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7319    else    else
7320      {      {
7321      if (firstbyte < 0)      if (firstbyte < 0)
7322        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7323      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7324        {        {
7325        int ch = firstbyte & 255;        int ch = firstbyte & 255;
# Line 6744  if (reqbyte >= 0 && Line 7348  if (reqbyte >= 0 &&
7348  /* Print out the compiled data if debugging is enabled. This is never the  /* Print out the compiled data if debugging is enabled. This is never the
7349  case when building a production library. */  case when building a production library. */
7350    
7351  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
7352  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
7353    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
7354    
# Line 6782  if (code - codestart > length) Line 7385  if (code - codestart > length)
7385    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
7386    return NULL;    return NULL;
7387    }    }
7388  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
7389    
7390  return (pcre *)re;  return (pcre *)re;
7391  }  }

Legend:
Removed from v.457  
changed lines
  Added in v.621

  ViewVC Help
Powered by ViewVC 1.1.5