/[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 488 by ph10, Mon Jan 11 15:29:42 2010 UTC revision 629 by ph10, Fri Jul 22 09:18:11 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) 1997-2010 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 92  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 119  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 166  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 183  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 197  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 251  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 263  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 310  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 336  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 344  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      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
416  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 481  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,
550      int *, int *, branch_chain *, compile_data *, int *);      int *, branch_chain *, compile_data *, int *);
551    
552    
553    
# Line 503  static const char * Line 568  static const char *
568  find_error_text(int n)  find_error_text(int n)
569  {  {
570  const char *s = error_texts;  const char *s = error_texts;
571  for (; n > 0; n--) while (*s++ != 0) {};  for (; n > 0; n--)
572      {
573      while (*s++ != 0) {};
574      if (*s == 0) return "Error text not found (please report)";
575      }
576  return s;  return s;
577  }  }
578    
# Line 574  else Line 643  else
643    
644      case CHAR_l:      case CHAR_l:
645      case CHAR_L:      case CHAR_L:
     case CHAR_N:  
646      case CHAR_u:      case CHAR_u:
647      case CHAR_U:      case CHAR_U:
648      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
# Line 775  else Line 843  else
843      break;      break;
844    
845      /* 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.
846      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
847        coding is ASCII-specific, but then the whole concept of \cx is
848      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
849    
850      case CHAR_c:      case CHAR_c:
# Line 785  else Line 854  else
854        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
855        break;        break;
856        }        }
857    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
858  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
859          {
860          *errorcodeptr = ERR68;
861          break;
862          }
863      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
864      c ^= 0x40;      c ^= 0x40;
865  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
866      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
867      c ^= 0xC0;      c ^= 0xC0;
868  #endif  #endif
# Line 812  else Line 885  else
885      }      }
886    }    }
887    
888    /* Perl supports \N{name} for character names, as well as plain \N for "not
889    newline". PCRE does not support \N{name}. */
890    
891    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
892      *errorcodeptr = ERR37;
893    
894    /* If PCRE_UCP is set, we change the values for \d etc. */
895    
896    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
897      c -= (ESC_DU - ESC_D);
898    
899    /* Set the pointer to the final character before returning. */
900    
901  *ptrptr = ptr;  *ptrptr = ptr;
902  return c;  return c;
903  }  }
# Line 1020  top-level call starts at the beginning o Line 1106  top-level call starts at the beginning o
1106  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
1107  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
1108  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
1109  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
1110  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1111  first pass. Recursion is used to keep track of subpatterns that reset the  
1112  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1113    that if (?< or (?' or (?P< is encountered, the name will be correctly
1114    terminated because that is checked in the first pass. There is now one call to
1115    this function in the first pass, to check for a recursive back reference by
1116    name (so that we can make the whole group atomic). In this case, we need check
1117    only up to the current position in the pattern, and that is still OK because
1118    and previous occurrences will have been checked. To make this work, the test
1119    for "end of pattern" is a check against cd->end_pattern in the main loop,
1120    instead of looking for a binary zero. This means that the special first-pass
1121    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1122    processing items within the loop are OK, because afterwards the main loop will
1123    terminate.)
1124    
1125  Arguments:  Arguments:
1126    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1031  Arguments: Line 1128  Arguments:
1128    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1129    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1130    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1131      utf8         TRUE if we are in UTF-8 mode
1132    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1133    
1134  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 1038  Returns:       the number of the named s Line 1136  Returns:       the number of the named s
1136    
1137  static int  static int
1138  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,
1139    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1140  {  {
1141  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1142  int start_count = *count;  int start_count = *count;
# Line 1050  dealing with. The very first call may no Line 1148  dealing with. The very first call may no
1148    
1149  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1150    {    {
1151    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1152        ptr[2] == CHAR_VERTICAL_LINE)  
1153      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1154    
1155      /* Handle a normal, unnamed capturing parenthesis. */
1156    
1157      else if (ptr[1] != CHAR_QUESTION_MARK)
1158        {
1159        *count += 1;
1160        if (name == NULL && *count == lorn) return *count;
1161        ptr++;
1162        }
1163    
1164      /* All cases now have (? at the start. Remember when we are in a group
1165      where the parenthesis numbers are duplicated. */
1166    
1167      else if (ptr[2] == CHAR_VERTICAL_LINE)
1168      {      {
1169      ptr += 3;      ptr += 3;
1170      dup_parens = TRUE;      dup_parens = TRUE;
1171      }      }
1172    
1173    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1174    
1175    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1176      {      {
1177      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1178      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1179      }      }
1180    
1181    /* 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
1182    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
1183    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1184    
1185    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1186      {      {
# Line 1080  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1192  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1192        }        }
1193      }      }
1194    
1195    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1196    
1197    else    else
1198      {      {
# Line 1109  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1221  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1221    }    }
1222    
1223  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1224  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1225    first-pass call when this value is temporarily adjusted to stop at the current
1226    position. So DO NOT change this to a test for binary zero. */
1227    
1228  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1229    {    {
1230    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1231    
# Line 1185  for (; *ptr != 0; ptr++) Line 1299  for (; *ptr != 0; ptr++)
1299    
1300    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1301      {      {
1302      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1303        while (*ptr != 0)
1304          {
1305          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1306          ptr++;
1307    #ifdef SUPPORT_UTF8
1308          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1309    #endif
1310          }
1311      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1312      continue;      continue;
1313      }      }
# Line 1194  for (; *ptr != 0; ptr++) Line 1316  for (; *ptr != 0; ptr++)
1316    
1317    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1318      {      {
1319      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1320      if (rc > 0) return rc;      if (rc > 0) return rc;
1321      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1322      }      }
# Line 1202  for (; *ptr != 0; ptr++) Line 1324  for (; *ptr != 0; ptr++)
1324    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1325      {      {
1326      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1327      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1328      }      }
1329    
1330    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1241  Arguments: Line 1362  Arguments:
1362    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1363    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1364    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1365      utf8         TRUE if we are in UTF-8 mode
1366    
1367  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1368  */  */
1369    
1370  static int  static int
1371  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1372      BOOL utf8)
1373  {  {
1374  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1375  int count = 0;  int count = 0;
# Line 1259  matching closing parens. That is why we Line 1382  matching closing parens. That is why we
1382    
1383  for (;;)  for (;;)
1384    {    {
1385    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1386    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1387    }    }
1388    
# Line 1275  return rc; Line 1398  return rc;
1398    
1399  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1400  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
1401  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
1402  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
1403  assertions, and also the \b assertion; for others it does not.  does not.
1404    
1405  Arguments:  Arguments:
1406    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  
1407    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1408    
1409  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1410  */  */
1411    
1412  static const uschar*  static const uschar*
1413  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1414  {  {
1415  for (;;)  for (;;)
1416    {    {
1417    switch ((int)*code)    switch ((int)*code)
1418      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1419      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1420      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1421      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1352  and doing the check at the end; a flag s Line 1465  and doing the check at the end; a flag s
1465    
1466  Arguments:  Arguments:
1467    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1468    options  the compiling options    utf8     TRUE in UTF-8 mode
1469    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1470    cd       the "compile data" structure    cd       the "compile data" structure
1471    
# Line 1363  Returns:   the fixed length, Line 1476  Returns:   the fixed length,
1476  */  */
1477    
1478  static int  static int
1479  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1480  {  {
1481  int length = -1;  int length = -1;
1482    
# Line 1380  for (;;) Line 1493  for (;;)
1493    register int op = *cc;    register int op = *cc;
1494    switch (op)    switch (op)
1495      {      {
1496        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1497        OP_BRA (normal non-capturing bracket) because the other variants of these
1498        opcodes are all concerned with unlimited repeated groups, which of course
1499        are not of fixed length. They will cause a -1 response from the default
1500        case of this switch. */
1501    
1502      case OP_CBRA:      case OP_CBRA:
1503      case OP_BRA:      case OP_BRA:
1504      case OP_ONCE:      case OP_ONCE:
1505      case OP_COND:      case OP_COND:
1506      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1507      if (d < 0) return d;      if (d < 0) return d;
1508      branchlength += d;      branchlength += d;
1509      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1393  for (;;) Line 1512  for (;;)
1512    
1513      /* 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
1514      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
1515      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.
1516        Note that we must not include the OP_KETRxxx opcodes here, because they
1517        all imply an unlimited repeat. */
1518    
1519      case OP_ALT:      case OP_ALT:
1520      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1521      case OP_END:      case OP_END:
1522      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1523        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1416  for (;;) Line 1535  for (;;)
1535      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1536      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1537      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1538      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1539      if (d < 0) return d;      if (d < 0) return d;
1540      branchlength += d;      branchlength += d;
1541      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1439  for (;;) Line 1558  for (;;)
1558      case OP_RREF:      case OP_RREF:
1559      case OP_NRREF:      case OP_NRREF:
1560      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1561      case OP_CALLOUT:      case OP_CALLOUT:
1562      case OP_SOD:      case OP_SOD:
1563      case OP_SOM:      case OP_SOM:
1564        case OP_SET_SOM:
1565      case OP_EOD:      case OP_EOD:
1566      case OP_EODN:      case OP_EODN:
1567      case OP_CIRC:      case OP_CIRC:
1568        case OP_CIRCM:
1569      case OP_DOLL:      case OP_DOLL:
1570        case OP_DOLLM:
1571      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1572      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1573      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1455  for (;;) Line 1576  for (;;)
1576      /* Handle literal characters */      /* Handle literal characters */
1577    
1578      case OP_CHAR:      case OP_CHAR:
1579      case OP_CHARNC:      case OP_CHARI:
1580      case OP_NOT:      case OP_NOT:
1581        case OP_NOTI:
1582      branchlength++;      branchlength++;
1583      cc += 2;      cc += 2;
1584  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1585      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];  
1586  #endif  #endif
1587      break;      break;
1588    
# Line 1472  for (;;) Line 1593  for (;;)
1593      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1594      cc += 4;      cc += 4;
1595  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1596      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];  
1597  #endif  #endif
1598      break;      break;
1599    
# Line 1575  _pcre_find_bracket(const uschar *code, B Line 1695  _pcre_find_bracket(const uschar *code, B
1695  for (;;)  for (;;)
1696    {    {
1697    register int c = *code;    register int c = *code;
1698    
1699    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1700    
1701    /* 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 1593  for (;;) Line 1714  for (;;)
1714    
1715    /* Handle capturing bracket */    /* Handle capturing bracket */
1716    
1717    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1718               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1719      {      {
1720      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1721      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1602  for (;;) Line 1724  for (;;)
1724    
1725    /* 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
1726    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
1727    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1728      must add in its length. */
1729    
1730    else    else
1731      {      {
# Line 1626  for (;;) Line 1749  for (;;)
1749        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1750        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1751        break;        break;
1752    
1753          case OP_MARK:
1754          case OP_PRUNE_ARG:
1755          case OP_SKIP_ARG:
1756          code += code[1];
1757          break;
1758    
1759          case OP_THEN_ARG:
1760          code += code[1+LINK_SIZE];
1761          break;
1762        }        }
1763    
1764      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1640  for (;;) Line 1773  for (;;)
1773      if (utf8) switch(c)      if (utf8) switch(c)
1774        {        {
1775        case OP_CHAR:        case OP_CHAR:
1776        case OP_CHARNC:        case OP_CHARI:
1777        case OP_EXACT:        case OP_EXACT:
1778          case OP_EXACTI:
1779        case OP_UPTO:        case OP_UPTO:
1780          case OP_UPTOI:
1781        case OP_MINUPTO:        case OP_MINUPTO:
1782          case OP_MINUPTOI:
1783        case OP_POSUPTO:        case OP_POSUPTO:
1784          case OP_POSUPTOI:
1785        case OP_STAR:        case OP_STAR:
1786          case OP_STARI:
1787        case OP_MINSTAR:        case OP_MINSTAR:
1788          case OP_MINSTARI:
1789        case OP_POSSTAR:        case OP_POSSTAR:
1790          case OP_POSSTARI:
1791        case OP_PLUS:        case OP_PLUS:
1792          case OP_PLUSI:
1793        case OP_MINPLUS:        case OP_MINPLUS:
1794          case OP_MINPLUSI:
1795        case OP_POSPLUS:        case OP_POSPLUS:
1796          case OP_POSPLUSI:
1797        case OP_QUERY:        case OP_QUERY:
1798          case OP_QUERYI:
1799        case OP_MINQUERY:        case OP_MINQUERY:
1800          case OP_MINQUERYI:
1801        case OP_POSQUERY:        case OP_POSQUERY:
1802          case OP_POSQUERYI:
1803        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1804        break;        break;
1805        }        }
# Line 1697  for (;;) Line 1843  for (;;)
1843    
1844    /* 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
1845    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
1846    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1847      must add in its length. */
1848    
1849    else    else
1850      {      {
# Line 1721  for (;;) Line 1868  for (;;)
1868        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1869        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1870        break;        break;
1871    
1872          case OP_MARK:
1873          case OP_PRUNE_ARG:
1874          case OP_SKIP_ARG:
1875          code += code[1];
1876          break;
1877    
1878          case OP_THEN_ARG:
1879          code += code[1+LINK_SIZE];
1880          break;
1881        }        }
1882    
1883      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1735  for (;;) Line 1892  for (;;)
1892      if (utf8) switch(c)      if (utf8) switch(c)
1893        {        {
1894        case OP_CHAR:        case OP_CHAR:
1895        case OP_CHARNC:        case OP_CHARI:
1896        case OP_EXACT:        case OP_EXACT:
1897          case OP_EXACTI:
1898        case OP_UPTO:        case OP_UPTO:
1899          case OP_UPTOI:
1900        case OP_MINUPTO:        case OP_MINUPTO:
1901          case OP_MINUPTOI:
1902        case OP_POSUPTO:        case OP_POSUPTO:
1903          case OP_POSUPTOI:
1904        case OP_STAR:        case OP_STAR:
1905          case OP_STARI:
1906        case OP_MINSTAR:        case OP_MINSTAR:
1907          case OP_MINSTARI:
1908        case OP_POSSTAR:        case OP_POSSTAR:
1909          case OP_POSSTARI:
1910        case OP_PLUS:        case OP_PLUS:
1911          case OP_PLUSI:
1912        case OP_MINPLUS:        case OP_MINPLUS:
1913          case OP_MINPLUSI:
1914        case OP_POSPLUS:        case OP_POSPLUS:
1915          case OP_POSPLUSI:
1916        case OP_QUERY:        case OP_QUERY:
1917          case OP_QUERYI:
1918        case OP_MINQUERY:        case OP_MINQUERY:
1919          case OP_MINQUERYI:
1920        case OP_POSQUERY:        case OP_POSQUERY:
1921          case OP_POSQUERYI:
1922        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1923        break;        break;
1924        }        }
# Line 1777  Arguments: Line 1947  Arguments:
1947    code        points to start of search    code        points to start of search
1948    endcode     points to where to stop    endcode     points to where to stop
1949    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1950      cd          contains pointers to tables etc.
1951    
1952  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1953  */  */
1954    
1955  static BOOL  static BOOL
1956  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1957      compile_data *cd)
1958  {  {
1959  register int c;  register int c;
1960  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1961       code < endcode;       code < endcode;
1962       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1963    {    {
1964    const uschar *ccode;    const uschar *ccode;
1965    
# Line 1803  for (code = first_significant_code(code Line 1975  for (code = first_significant_code(code
1975      continue;      continue;
1976      }      }
1977    
1978      /* For a recursion/subroutine call, if its end has been reached, which
1979      implies a backward reference subroutine call, we can scan it. If it's a
1980      forward reference subroutine call, we can't. To detect forward reference
1981      we have to scan up the list that is kept in the workspace. This function is
1982      called only when doing the real compile, not during the pre-compile that
1983      measures the size of the compiled pattern. */
1984    
1985      if (c == OP_RECURSE)
1986        {
1987        const uschar *scode;
1988        BOOL empty_branch;
1989    
1990        /* Test for forward reference */
1991    
1992        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1993          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1994    
1995        /* Not a forward reference, test for completed backward reference */
1996    
1997        empty_branch = FALSE;
1998        scode = cd->start_code + GET(code, 1);
1999        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2000    
2001        /* Completed backwards reference */
2002    
2003        do
2004          {
2005          if (could_be_empty_branch(scode, endcode, utf8, cd))
2006            {
2007            empty_branch = TRUE;
2008            break;
2009            }
2010          scode += GET(scode, 1);
2011          }
2012        while (*scode == OP_ALT);
2013    
2014        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2015        continue;
2016        }
2017    
2018    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2019    
2020    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2021          c == OP_BRAPOSZERO)
2022      {      {
2023      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2024      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1813  for (code = first_significant_code(code Line 2026  for (code = first_significant_code(code
2026      continue;      continue;
2027      }      }
2028    
2029      /* A nested group that is already marked as "could be empty" can just be
2030      skipped. */
2031    
2032      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2033          c == OP_SCBRA || c == OP_SCBRAPOS)
2034        {
2035        do code += GET(code, 1); while (*code == OP_ALT);
2036        c = *code;
2037        continue;
2038        }
2039    
2040    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2041    
2042    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2043          c == OP_CBRA || c == OP_CBRAPOS ||
2044          c == OP_ONCE || c == OP_COND)
2045      {      {
2046      BOOL empty_branch;      BOOL empty_branch;
2047      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1831  for (code = first_significant_code(code Line 2057  for (code = first_significant_code(code
2057        empty_branch = FALSE;        empty_branch = FALSE;
2058        do        do
2059          {          {
2060          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2061            empty_branch = TRUE;            empty_branch = TRUE;
2062          code += GET(code, 1);          code += GET(code, 1);
2063          }          }
# Line 1902  for (code = first_significant_code(code Line 2128  for (code = first_significant_code(code
2128      case OP_ALLANY:      case OP_ALLANY:
2129      case OP_ANYBYTE:      case OP_ANYBYTE:
2130      case OP_CHAR:      case OP_CHAR:
2131      case OP_CHARNC:      case OP_CHARI:
2132      case OP_NOT:      case OP_NOT:
2133        case OP_NOTI:
2134      case OP_PLUS:      case OP_PLUS:
2135      case OP_MINPLUS:      case OP_MINPLUS:
2136      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1943  for (code = first_significant_code(code Line 2170  for (code = first_significant_code(code
2170      case OP_KET:      case OP_KET:
2171      case OP_KETRMAX:      case OP_KETRMAX:
2172      case OP_KETRMIN:      case OP_KETRMIN:
2173        case OP_KETRPOS:
2174      case OP_ALT:      case OP_ALT:
2175      return TRUE;      return TRUE;
2176    
# Line 1951  for (code = first_significant_code(code Line 2179  for (code = first_significant_code(code
2179    
2180  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2181      case OP_STAR:      case OP_STAR:
2182        case OP_STARI:
2183      case OP_MINSTAR:      case OP_MINSTAR:
2184        case OP_MINSTARI:
2185      case OP_POSSTAR:      case OP_POSSTAR:
2186        case OP_POSSTARI:
2187      case OP_QUERY:      case OP_QUERY:
2188        case OP_QUERYI:
2189      case OP_MINQUERY:      case OP_MINQUERY:
2190        case OP_MINQUERYI:
2191      case OP_POSQUERY:      case OP_POSQUERY:
2192        case OP_POSQUERYI:
2193      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2194      break;      break;
2195    
2196      case OP_UPTO:      case OP_UPTO:
2197        case OP_UPTOI:
2198      case OP_MINUPTO:      case OP_MINUPTO:
2199        case OP_MINUPTOI:
2200      case OP_POSUPTO:      case OP_POSUPTO:
2201        case OP_POSUPTOI:
2202      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2203      break;      break;
2204  #endif  #endif
2205    
2206        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2207        string. */
2208    
2209        case OP_MARK:
2210        case OP_PRUNE_ARG:
2211        case OP_SKIP_ARG:
2212        code += code[1];
2213        break;
2214    
2215        case OP_THEN_ARG:
2216        code += code[1+LINK_SIZE];
2217        break;
2218    
2219        /* None of the remaining opcodes are required to match a character. */
2220    
2221        default:
2222        break;
2223      }      }
2224    }    }
2225    
# Line 1981  return TRUE; Line 2236  return TRUE;
2236  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2237  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2238  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2239    This function is called only during the real compile, not during the
2240    pre-compile.
2241    
2242  Arguments:  Arguments:
2243    code        points to start of the recursion    code        points to start of the recursion
2244    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2245    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2246    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2247      cd          pointers to tables etc
2248    
2249  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2250  */  */
2251    
2252  static BOOL  static BOOL
2253  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2254    BOOL utf8)    BOOL utf8, compile_data *cd)
2255  {  {
2256  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2257    {    {
2258    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2259      return FALSE;      return FALSE;
2260    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2261    }    }
# Line 2180  auto_callout(uschar *code, const uschar Line 2438  auto_callout(uschar *code, const uschar
2438  {  {
2439  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2440  *code++ = 255;  *code++ = 255;
2441  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2442  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2443  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2444  }  }
2445    
# Line 2206  Returns:             nothing Line 2464  Returns:             nothing
2464  static void  static void
2465  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2466  {  {
2467  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2468  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2469  }  }
2470    
# Line 2256  for (++c; c <= d; c++) Line 2514  for (++c; c <= d; c++)
2514    
2515  return TRUE;  return TRUE;
2516  }  }
2517    
2518    
2519    
2520    /*************************************************
2521    *        Check a character and a property        *
2522    *************************************************/
2523    
2524    /* This function is called by check_auto_possessive() when a property item
2525    is adjacent to a fixed character.
2526    
2527    Arguments:
2528      c            the character
2529      ptype        the property type
2530      pdata        the data for the type
2531      negated      TRUE if it's a negated property (\P or \p{^)
2532    
2533    Returns:       TRUE if auto-possessifying is OK
2534    */
2535    
2536    static BOOL
2537    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2538    {
2539    const ucd_record *prop = GET_UCD(c);
2540    switch(ptype)
2541      {
2542      case PT_LAMP:
2543      return (prop->chartype == ucp_Lu ||
2544              prop->chartype == ucp_Ll ||
2545              prop->chartype == ucp_Lt) == negated;
2546    
2547      case PT_GC:
2548      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2549    
2550      case PT_PC:
2551      return (pdata == prop->chartype) == negated;
2552    
2553      case PT_SC:
2554      return (pdata == prop->script) == negated;
2555    
2556      /* These are specials */
2557    
2558      case PT_ALNUM:
2559      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2560              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2561    
2562      case PT_SPACE:    /* Perl space */
2563      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2564              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2565              == negated;
2566    
2567      case PT_PXSPACE:  /* POSIX space */
2568      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2569              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2570              c == CHAR_FF || c == CHAR_CR)
2571              == negated;
2572    
2573      case PT_WORD:
2574      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2575              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2576              c == CHAR_UNDERSCORE) == negated;
2577      }
2578    return FALSE;
2579    }
2580  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2581    
2582    
# Line 2269  whether the next thing could possibly ma Line 2590  whether the next thing could possibly ma
2590  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2591    
2592  Arguments:  Arguments:
2593    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2594    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2595    ptr           next character in pattern    ptr           next character in pattern
2596    options       options bits    options       options bits
2597    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2281  Returns:        TRUE if possessifying is Line 2600  Returns:        TRUE if possessifying is
2600  */  */
2601    
2602  static BOOL  static BOOL
2603  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2604    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2605  {  {
2606  int next;  int c, next;
2607    int op_code = *previous++;
2608    
2609  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2610    
# Line 2295  if ((options & PCRE_EXTENDED) != 0) Line 2615  if ((options & PCRE_EXTENDED) != 0)
2615      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2616      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2617        {        {
2618        while (*(++ptr) != 0)        ptr++;
2619          while (*ptr != 0)
2620            {
2621          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2622            ptr++;
2623    #ifdef SUPPORT_UTF8
2624            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2625    #endif
2626            }
2627        }        }
2628      else break;      else break;
2629      }      }
# Line 2332  if ((options & PCRE_EXTENDED) != 0) Line 2659  if ((options & PCRE_EXTENDED) != 0)
2659      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2660      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2661        {        {
2662        while (*(++ptr) != 0)        ptr++;
2663          while (*ptr != 0)
2664            {
2665          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2666            ptr++;
2667    #ifdef SUPPORT_UTF8
2668            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2669    #endif
2670            }
2671        }        }
2672      else break;      else break;
2673      }      }
# Line 2345  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2679  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2679    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)
2680      return FALSE;      return FALSE;
2681    
2682  /* 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
2683  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. */  
2684    
2685  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2686    {    {
2687    case OP_CHAR:    case OP_CHAR:
2688  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2689    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2690  #else  #else
2691    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2692  #endif  #endif
2693    return item != next;    return c != next;
2694    
2695    /* 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
2696    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
2697    high-valued characters. */    high-valued characters. */
2698    
2699    case OP_CHARNC:    case OP_CHARI:
2700  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2701    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2702    #else
2703      c = *previous;
2704  #endif  #endif
2705    if (item == next) return FALSE;    if (c == next) return FALSE;
2706  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2707    if (utf8)    if (utf8)
2708      {      {
# Line 2382  if (next >= 0) switch(op_code) Line 2713  if (next >= 0) switch(op_code)
2713  #else  #else
2714      othercase = NOTACHAR;      othercase = NOTACHAR;
2715  #endif  #endif
2716      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2717      }      }
2718    else    else
2719  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2720    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2721    
2722    /* 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
2723      opcodes are not used for multi-byte characters, because they are coded using
2724      an XCLASS instead. */
2725    
2726    case OP_NOT:    case OP_NOT:
2727    if (item == next) return TRUE;    return (c = *previous) == next;
2728    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2729      case OP_NOTI:
2730      if ((c = *previous) == next) return TRUE;
2731  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2732    if (utf8)    if (utf8)
2733      {      {
# Line 2403  if (next >= 0) switch(op_code) Line 2738  if (next >= 0) switch(op_code)
2738  #else  #else
2739      othercase = NOTACHAR;      othercase = NOTACHAR;
2740  #endif  #endif
2741      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2742      }      }
2743    else    else
2744  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2745    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2746    
2747      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2748      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2749    
2750    case OP_DIGIT:    case OP_DIGIT:
2751    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2450  if (next >= 0) switch(op_code) Line 2788  if (next >= 0) switch(op_code)
2788      case 0x202f:      case 0x202f:
2789      case 0x205f:      case 0x205f:
2790      case 0x3000:      case 0x3000:
2791      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2792      default:      default:
2793      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2794      }      }
2795    
2796      case OP_ANYNL:
2797    case OP_VSPACE:    case OP_VSPACE:
2798    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2799    switch(next)    switch(next)
# Line 2466  if (next >= 0) switch(op_code) Line 2805  if (next >= 0) switch(op_code)
2805      case 0x85:      case 0x85:
2806      case 0x2028:      case 0x2028:
2807      case 0x2029:      case 0x2029:
2808      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2809      default:      default:
2810      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2811      }      }
2812    
2813    #ifdef SUPPORT_UCP
2814      case OP_PROP:
2815      return check_char_prop(next, previous[0], previous[1], FALSE);
2816    
2817      case OP_NOTPROP:
2818      return check_char_prop(next, previous[0], previous[1], TRUE);
2819    #endif
2820    
2821    default:    default:
2822    return FALSE;    return FALSE;
2823    }    }
2824    
2825    
2826  /* 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
2827    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2828    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2829    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2830    replaced by OP_PROP codes when PCRE_UCP is set. */
2831    
2832  switch(op_code)  switch(op_code)
2833    {    {
2834    case OP_CHAR:    case OP_CHAR:
2835    case OP_CHARNC:    case OP_CHARI:
2836  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2837    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2838    #else
2839      c = *previous;
2840  #endif  #endif
2841    switch(-next)    switch(-next)
2842      {      {
2843      case ESC_d:      case ESC_d:
2844      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2845    
2846      case ESC_D:      case ESC_D:
2847      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2848    
2849      case ESC_s:      case ESC_s:
2850      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2851    
2852      case ESC_S:      case ESC_S:
2853      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2854    
2855      case ESC_w:      case ESC_w:
2856      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2857    
2858      case ESC_W:      case ESC_W:
2859      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2860    
2861      case ESC_h:      case ESC_h:
2862      case ESC_H:      case ESC_H:
2863      switch(item)      switch(c)
2864        {        {
2865        case 0x09:        case 0x09:
2866        case 0x20:        case 0x20:
# Line 2535  switch(op_code) Line 2888  switch(op_code)
2888    
2889      case ESC_v:      case ESC_v:
2890      case ESC_V:      case ESC_V:
2891      switch(item)      switch(c)
2892        {        {
2893        case 0x0a:        case 0x0a:
2894        case 0x0b:        case 0x0b:
# Line 2549  switch(op_code) Line 2902  switch(op_code)
2902        return -next == ESC_v;        return -next == ESC_v;
2903        }        }
2904    
2905        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2906        their substitutions and process them. The result will always be either
2907        -ESC_p or -ESC_P. Then fall through to process those values. */
2908    
2909    #ifdef SUPPORT_UCP
2910        case ESC_du:
2911        case ESC_DU:
2912        case ESC_wu:
2913        case ESC_WU:
2914        case ESC_su:
2915        case ESC_SU:
2916          {
2917          int temperrorcode = 0;
2918          ptr = substitutes[-next - ESC_DU];
2919          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2920          if (temperrorcode != 0) return FALSE;
2921          ptr++;    /* For compatibility */
2922          }
2923        /* Fall through */
2924    
2925        case ESC_p:
2926        case ESC_P:
2927          {
2928          int ptype, pdata, errorcodeptr;
2929          BOOL negated;
2930    
2931          ptr--;      /* Make ptr point at the p or P */
2932          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2933          if (ptype < 0) return FALSE;
2934          ptr++;      /* Point past the final curly ket */
2935    
2936          /* If the property item is optional, we have to give up. (When generated
2937          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2938          to the original \d etc. At this point, ptr will point to a zero byte. */
2939    
2940          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2941            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2942              return FALSE;
2943    
2944          /* Do the property check. */
2945    
2946          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2947          }
2948    #endif
2949    
2950      default:      default:
2951      return FALSE;      return FALSE;
2952      }      }
2953    
2954      /* In principle, support for Unicode properties should be integrated here as
2955      well. It means re-organizing the above code so as to get hold of the property
2956      values before switching on the op-code. However, I wonder how many patterns
2957      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2958      these op-codes are never generated.) */
2959    
2960    case OP_DIGIT:    case OP_DIGIT:
2961    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2962           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2963    
2964    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2965    return next == -ESC_d;    return next == -ESC_d;
2966    
2967    case OP_WHITESPACE:    case OP_WHITESPACE:
2968    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2969    
2970    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2971    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2972    
2973    case OP_HSPACE:    case OP_HSPACE:
2974    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2975             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2976    
2977    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2978    return next == -ESC_h;    return next == -ESC_h;
2979    
2980    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2981      case OP_ANYNL:
2982    case OP_VSPACE:    case OP_VSPACE:
2983    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2984    
2985    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2986    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2987    
2988    case OP_WORDCHAR:    case OP_WORDCHAR:
2989    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2990             next == -ESC_v || next == -ESC_R;
2991    
2992    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2993    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2644  BOOL inescq = FALSE; Line 3051  BOOL inescq = FALSE;
3051  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3052  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3053  const uschar *tempptr;  const uschar *tempptr;
3054    const uschar *nestptr = NULL;
3055  uschar *previous = NULL;  uschar *previous = NULL;
3056  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3057  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2714  for (;; ptr++) Line 3122  for (;; ptr++)
3122    
3123    c = *ptr;    c = *ptr;
3124    
3125      /* If we are at the end of a nested substitution, revert to the outer level
3126      string. Nesting only happens one level deep. */
3127    
3128      if (c == 0 && nestptr != NULL)
3129        {
3130        ptr = nestptr;
3131        nestptr = NULL;
3132        c = *ptr;
3133        }
3134    
3135    /* 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
3136    previous cycle of this loop. */    previous cycle of this loop. */
3137    
# Line 2722  for (;; ptr++) Line 3140  for (;; ptr++)
3140  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
3141      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3142  #endif  #endif
3143      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3144        {        {
3145        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3146        goto FAILED;        goto FAILED;
# Line 2744  for (;; ptr++) Line 3162  for (;; ptr++)
3162        goto FAILED;        goto FAILED;
3163        }        }
3164    
3165      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3166      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));
3167    
3168      /* 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 2771  for (;; ptr++) Line 3189  for (;; ptr++)
3189    /* 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
3190    reference list. */    reference list. */
3191    
3192    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3193      {      {
3194      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3195      goto FAILED;      goto FAILED;
# Line 2826  for (;; ptr++) Line 3244  for (;; ptr++)
3244      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3245      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3246        {        {
3247        while (*(++ptr) != 0)        ptr++;
3248          while (*ptr != 0)
3249          {          {
3250          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3251            ptr++;
3252    #ifdef SUPPORT_UTF8
3253            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3254    #endif
3255          }          }
3256        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3257    
# Line 2862  for (;; ptr++) Line 3285  for (;; ptr++)
3285          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3286          goto FAILED;          goto FAILED;
3287          }          }
3288        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3289        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3290        }        }
3291      return TRUE;      return TRUE;
# Line 2873  for (;; ptr++) Line 3296  for (;; ptr++)
3296      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3297    
3298      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3299        previous = NULL;
3300      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3301        {        {
3302        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3303          *code++ = OP_CIRCM;
3304        }        }
3305      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3306      break;      break;
3307    
3308      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3309      previous = NULL;      previous = NULL;
3310      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3311      break;      break;
3312    
3313      /* 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 3067  for (;; ptr++) Line 3491  for (;; ptr++)
3491            ptr++;            ptr++;
3492            }            }
3493    
3494          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3495          if (posix_class < 0)          if (posix_class < 0)
3496            {            {
3497            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3081  for (;; ptr++) Line 3505  for (;; ptr++)
3505          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3506            posix_class = 0;            posix_class = 0;
3507    
3508          /* 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
3509          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3510          subtract bits that may be in the main map already. At the end we or the  
3511          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3512            if ((options & PCRE_UCP) != 0)
3513              {
3514              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3515              if (posix_substitutes[pc] != NULL)
3516                {
3517                nestptr = tempptr + 1;
3518                ptr = posix_substitutes[pc] - 1;
3519                continue;
3520                }
3521              }
3522    #endif
3523            /* In the non-UCP case, we build the bit map for the POSIX class in a
3524            chunk of local store because we may be adding and subtracting from it,
3525            and we don't want to subtract bits that may be in the main map already.
3526            At the end we or the result into the bit map that is being built. */
3527    
3528          posix_class *= 3;          posix_class *= 3;
3529    
# Line 3128  for (;; ptr++) Line 3567  for (;; ptr++)
3567    
3568        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3569        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
3570        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
3571        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
3572        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
3573        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3574          PCRE_EXTRA is set. */
3575    
3576        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3577          {          {
3578          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3579          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3580    
3581          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 */  
3582          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3583            {            {
3584            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3157  for (;; ptr++) Line 3595  for (;; ptr++)
3595            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3596            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3597    
3598            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3599              {              {
3600    #ifdef SUPPORT_UCP
3601                case ESC_du:     /* These are the values given for \d etc */
3602                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3603                case ESC_wu:     /* escape sequence with an appropriate \p */
3604                case ESC_WU:     /* or \P to test Unicode properties instead */
3605                case ESC_su:     /* of the default ASCII testing. */
3606                case ESC_SU:
3607                nestptr = ptr;
3608                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3609                class_charcount -= 2;                /* Undo! */
3610                continue;
3611    #endif
3612              case ESC_d:              case ESC_d:
3613              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3614              continue;              continue;
# Line 3179  for (;; ptr++) Line 3627  for (;; ptr++)
3627              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3628              continue;              continue;
3629    
3630                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3631                if it was previously set by something earlier in the character
3632                class. */
3633    
3634              case ESC_s:              case ESC_s:
3635              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3636              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3637                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3638              continue;              continue;
3639    
3640              case ESC_S:              case ESC_S:
# Line 3190  for (;; ptr++) Line 3643  for (;; ptr++)
3643              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3644              continue;              continue;
3645    
3646              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)  
             {  
3647              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3648              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3649              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3227  for (;; ptr++) Line 3667  for (;; ptr++)
3667                }                }
3668  #endif  #endif
3669              continue;              continue;
             }  
3670    
3671            if (-c == ESC_H)              case ESC_H:
             {  
3672              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3673                {                {
3674                int x = 0xff;                int x = 0xff;
# Line 3272  for (;; ptr++) Line 3710  for (;; ptr++)
3710                }                }
3711  #endif  #endif
3712              continue;              continue;
             }  
3713    
3714            if (-c == ESC_v)              case ESC_v:
             {  
3715              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3716              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3717              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3291  for (;; ptr++) Line 3727  for (;; ptr++)
3727                }                }
3728  #endif  #endif
3729              continue;              continue;
             }  
3730    
3731            if (-c == ESC_V)              case ESC_V:
             {  
3732              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3733                {                {
3734                int x = 0xff;                int x = 0xff;
# Line 3324  for (;; ptr++) Line 3758  for (;; ptr++)
3758                }                }
3759  #endif  #endif
3760              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3761    
3762  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3763            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3764              {              case ESC_P:
3765              BOOL negated;                {
3766              int pdata;                BOOL negated;
3767              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3768              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3769              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3770              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3771                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3772              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3773              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3774              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3775              continue;                class_charcount -= 2;   /* Not a < 256 character */
3776              }                continue;
3777                  }
3778  #endif  #endif
3779            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3780            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3781            treated as literals. */              treated as literals. */
3782    
3783            if ((options & PCRE_EXTRA) != 0)              default:
3784              {              if ((options & PCRE_EXTRA) != 0)
3785              *errorcodeptr = ERR7;                {
3786              goto FAILED;                *errorcodeptr = ERR7;
3787                  goto FAILED;
3788                  }
3789                class_charcount -= 2;  /* Undo the default count from above */
3790                c = *ptr;              /* Get the final character and fall through */
3791                break;
3792              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3793            }            }
3794    
3795          /* 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 3425  for (;; ptr++) Line 3859  for (;; ptr++)
3859            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3860            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3861    
3862            /* \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 */  
3863    
3864            if (d < 0)            if (d < 0)
3865              {              {
3866              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  
3867                {                {
3868                ptr = oldptr;                ptr = oldptr;
3869                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3598  for (;; ptr++) Line 4029  for (;; ptr++)
4029          }          }
4030        }        }
4031    
4032      /* 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.
4033        If we are at the end of an internal nested string, revert to the outer
4034        string. */
4035    
4036        while (((c = *(++ptr)) != 0 ||
4037               (nestptr != NULL &&
4038                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4039               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4040    
4041      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4042    
4043      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4044        {        {
4045        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4046        goto FAILED;        goto FAILED;
4047        }        }
4048    
   
 /* 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  
   
   
4049      /* 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
4050      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
4051      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 3634  we set the flag only if there is a liter Line 4053  we set the flag only if there is a liter
4053    
4054      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
4055      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4056      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4057      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4058    
4059      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
4060      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.
4061      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
4062      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
4063      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
4064      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4065    
4066  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4067      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3653  we set the flag only if there is a liter Line 4072  we set the flag only if there is a liter
4072        {        {
4073        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4074    
4075        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4076    
4077        if (negate_class)        if (negate_class)
4078          {          {
4079          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4080          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4081          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4082          *code++ = class_lastchar;          *code++ = class_lastchar;
4083          break;          break;
4084          }          }
# Line 3690  we set the flag only if there is a liter Line 4109  we set the flag only if there is a liter
4109    
4110      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4111      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4112      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
4113      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
4114      (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
4115      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
4116        actual compiled code. */
4117    
4118  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4119      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4120        {        {
4121        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4122        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3722  we set the flag only if there is a liter Line 4142  we set the flag only if there is a liter
4142        }        }
4143  #endif  #endif
4144    
4145      /* 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
4146      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
4147      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
4148      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4149        negating it if necessary. */
4150    
4151      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4152      if (negate_class)      if (negate_class)
# Line 3808  we set the flag only if there is a liter Line 4229  we set the flag only if there is a liter
4229        ptr++;        ptr++;
4230        }        }
4231      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4232    
4233        /* If previous was a recursion call, wrap it in atomic brackets so that
4234        previous becomes the atomic group. All recursions were so wrapped in the
4235        past, but it no longer happens for non-repeated recursions. In fact, the
4236        repeated ones could be re-implemented independently so as not to need this,
4237        but for the moment we rely on the code for repeating groups. */
4238    
4239        if (*previous == OP_RECURSE)
4240          {
4241          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4242          *previous = OP_ONCE;
4243          PUT(previous, 1, 2 + 2*LINK_SIZE);
4244          previous[2 + 2*LINK_SIZE] = OP_KET;
4245          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4246          code += 2 + 2 * LINK_SIZE;
4247          length_prevgroup = 3 + 3*LINK_SIZE;
4248    
4249          /* When actually compiling, we need to check whether this was a forward
4250          reference, and if so, adjust the offset. */
4251    
4252          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4253            {
4254            int offset = GET(cd->hwm, -LINK_SIZE);
4255            if (offset == previous + 1 - cd->start_code)
4256              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4257            }
4258          }
4259    
4260        /* Now handle repetition for the different types of item. */
4261    
4262      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4263      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 3815  we set the flag only if there is a liter Line 4265  we set the flag only if there is a liter
4265      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
4266      instead.  */      instead.  */
4267    
4268      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4269        {        {
4270          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4271    
4272        /* 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
4273        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
4274        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 3849  we set the flag only if there is a liter Line 4301  we set the flag only if there is a liter
4301    
4302        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4303            repeat_max < 0 &&            repeat_max < 0 &&
4304            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4305          {          {
4306          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4307          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3862  we set the flag only if there is a liter Line 4313  we set the flag only if there is a liter
4313      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4314      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-
4315      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4316      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
4317      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4318    
4319      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4320        {        {
4321        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4322        c = previous[1];        c = previous[1];
4323        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4324            repeat_max < 0 &&            repeat_max < 0 &&
4325            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4326          {          {
4327          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4328          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3895  we set the flag only if there is a liter Line 4346  we set the flag only if there is a liter
4346    
4347        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4348            repeat_max < 0 &&            repeat_max < 0 &&
4349            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4350          {          {
4351          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4352          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4064  we set the flag only if there is a liter Line 4515  we set the flag only if there is a liter
4515  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4516               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4517  #endif  #endif
4518               *previous == OP_REF)               *previous == OP_REF ||
4519                 *previous == OP_REFI)
4520        {        {
4521        if (repeat_max == 0)        if (repeat_max == 0)
4522          {          {
# Line 4098  we set the flag only if there is a liter Line 4550  we set the flag only if there is a liter
4550        }        }
4551    
4552      /* 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
4553      cases. */      cases. Note that at this point we can encounter only the "basic" BRA and
4554        KET opcodes, as this is the place where they get converted into the more
4555        special varieties. */
4556    
4557      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4558               *previous == OP_ONCE || *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4559        {        {
4560        register int i;        register int i;
4561        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4562        uschar *bralink = NULL;        uschar *bralink = NULL;
4563          uschar *brazeroptr = NULL;
4564    
4565        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
4566    
4567        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
# Line 4116  we set the flag only if there is a liter Line 4570  we set the flag only if there is a liter
4570          goto FAILED;          goto FAILED;
4571          }          }
4572    
       /* 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;  
         }  
   
4573        /* 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
4574        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
4575        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 4170  we set the flag only if there is a liter Line 4611  we set the flag only if there is a liter
4611              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4612              goto END_REPEAT;              goto END_REPEAT;
4613              }              }
4614              brazeroptr = previous;    /* Save for possessive optimizing */
4615            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4616            }            }
4617    
# Line 4194  we set the flag only if there is a liter Line 4636  we set the flag only if there is a liter
4636            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4637            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4638    
4639            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4640            bralink = previous;            bralink = previous;
4641            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4642            }            }
# Line 4303  we set the flag only if there is a liter Line 4745  we set the flag only if there is a liter
4745              {              {
4746              int offset;              int offset;
4747              *code++ = OP_BRA;              *code++ = OP_BRA;
4748              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4749              bralink = code;              bralink = code;
4750              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4751              }              }
# Line 4324  we set the flag only if there is a liter Line 4766  we set the flag only if there is a liter
4766          while (bralink != NULL)          while (bralink != NULL)
4767            {            {
4768            int oldlinkoffset;            int oldlinkoffset;
4769            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4770            uschar *bra = code - offset;            uschar *bra = code - offset;
4771            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4772            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4334  we set the flag only if there is a liter Line 4776  we set the flag only if there is a liter
4776            }            }
4777          }          }
4778    
4779        /* 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
4780        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4781        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
4782        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4783          deal with possessive ONCEs specially.
4784    
4785          Otherwise, if the quantifier was possessive, we convert the BRA code to
4786          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4787          at runtime to detect this kind of subpattern at both the start and at the
4788          end.) The use of special opcodes makes it possible to reduce greatly the
4789          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4790          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4791          the default action below, of wrapping everything inside atomic brackets,
4792          does not happen.
4793    
4794        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
4795        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
4796        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
4797        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
4798        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4799    
4800        else        else
4801          {          {
4802          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4803          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4804          *ketcode = OP_KETRMAX + repeat_type;  
4805          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4806            if (*bracode == OP_ONCE)
4807              *ketcode = OP_KETRMAX + repeat_type;
4808            else
4809            {            {
4810            uschar *scode = bracode;            if (possessive_quantifier)
           do  
4811              {              {
4812              if (could_be_empty_branch(scode, ketcode, utf8))              *bracode += 1;                   /* Switch to xxxPOS opcodes */
4813                *ketcode = OP_KETRPOS;
4814                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4815                possessive_quantifier = FALSE;
4816                }
4817              else *ketcode = OP_KETRMAX + repeat_type;
4818    
4819              if (lengthptr == NULL)
4820                {
4821                uschar *scode = bracode;
4822                do
4823                {                {
4824                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4825                break;                  {
4826                    *bracode += OP_SBRA - OP_BRA;
4827                    break;
4828                    }
4829                  scode += GET(scode, 1);
4830                }                }
4831              scode += GET(scode, 1);              while (*scode == OP_ALT);
4832              }              }
           while (*scode == OP_ALT);  
4833            }            }
4834          }          }
4835        }        }
# Line 4383  we set the flag only if there is a liter Line 4850  we set the flag only if there is a liter
4850        }        }
4851    
4852      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4853      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4854      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4855      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4856      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
4857      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4858      tempcode, not at previous, which might be the first part of a string whose  
4859      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4860        just above, so possessive_quantifier is always FALSE for them at this
4861        stage.
4862    
4863        Note that the repeated item starts at tempcode, not at previous, which
4864        might be the first part of a string whose (former) last char we repeated.
4865    
4866      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
4867      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 4412  we set the flag only if there is a liter Line 4884  we set the flag only if there is a liter
4884  #endif  #endif
4885          }          }
4886    
4887        len = code - tempcode;        len = (int)(code - tempcode);
4888        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4889          {          {
4890          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4420  we set the flag only if there is a liter Line 4892  we set the flag only if there is a liter
4892          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4893          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4894    
4895          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4896          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4897          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4898          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4899    
4900          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4901          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4902          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4903          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4904    
4905            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4906            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4907            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4908            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4909    
4910            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4911            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4912            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4913            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4914    
4915            /* Because we are moving code along, we must ensure that any
4916            pending recursive references are updated. */
4917    
4918          default:          default:
4919            *code = OP_END;
4920            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4921          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4922          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4923          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4466  we set the flag only if there is a liter Line 4953  we set the flag only if there is a liter
4953    
4954      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4955    
4956      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4957             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4958        {        {
4959        int i, namelen;        int i, namelen;
4960          int arglen = 0;
4961        const char *vn = verbnames;        const char *vn = verbnames;
4962        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4963          const uschar *arg = NULL;
4964        previous = NULL;        previous = NULL;
4965        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4966          namelen = (int)(ptr - name);
4967    
4968        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
4969          {          {
4970          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4971          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4972              || *ptr == '_') ptr++;
4973            arglen = (int)(ptr - arg);
4974          }          }
4975    
4976        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4977          {          {
4978          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4979          goto FAILED;          goto FAILED;
4980          }          }
4981        namelen = ptr - name;  
4982          /* Scan the table of verb names */
4983    
4984        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4985          {          {
4986          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4987              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4988            {            {
4989            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
4990              ASSERT_ACCEPT if in an assertion. */
4991    
4992            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
4993              {              {
4994              open_capitem *oc;              open_capitem *oc;
4995                if (arglen != 0)
4996                  {
4997                  *errorcodeptr = ERR59;
4998                  goto FAILED;
4999                  }
5000              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5001              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5002                {                {
5003                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5004                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5005                }                }
5006                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5007              }              }
5008            *code++ = verbs[i].op;  
5009            break;            /* Handle other cases with/without an argument */
5010    
5011              else if (arglen == 0)
5012                {
5013                if (verbs[i].op < 0)   /* Argument is mandatory */
5014                  {
5015                  *errorcodeptr = ERR66;
5016                  goto FAILED;
5017                  }
5018                *code = verbs[i].op;
5019                if (*code++ == OP_THEN)
5020                  {
5021                  PUT(code, 0, code - bcptr->current_branch - 1);
5022                  code += LINK_SIZE;
5023                  }
5024                }
5025    
5026              else
5027                {
5028                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
5029                  {
5030                  *errorcodeptr = ERR59;
5031                  goto FAILED;
5032                  }
5033                *code = verbs[i].op_arg;
5034                if (*code++ == OP_THEN_ARG)
5035                  {
5036                  PUT(code, 0, code - bcptr->current_branch - 1);
5037                  code += LINK_SIZE;
5038                  }
5039                *code++ = arglen;
5040                memcpy(code, arg, arglen);
5041                code += arglen;
5042                *code++ = 0;
5043                }
5044    
5045              break;  /* Found verb, exit loop */
5046            }            }
5047    
5048          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
5049          }          }
5050        if (i < verbcount) continue;  
5051        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
5052          *errorcodeptr = ERR60;          /* Verb not recognized */
5053        goto FAILED;        goto FAILED;
5054        }        }
5055    
# Line 4626  we set the flag only if there is a liter Line 5168  we set the flag only if there is a liter
5168                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5169            ptr++;            ptr++;
5170            }            }
5171          namelen = ptr - name;          namelen = (int)(ptr - name);
5172    
5173          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5174              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4687  we set the flag only if there is a liter Line 5229  we set the flag only if there is a liter
5229          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5230    
5231          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5232                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5233            {            {
5234            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5235            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 4755  we set the flag only if there is a liter Line 5297  we set the flag only if there is a liter
5297          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5298          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5299          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5300            cd->assert_depth += 1;
5301          ptr++;          ptr++;
5302          break;          break;
5303    
# Line 4769  we set the flag only if there is a liter Line 5312  we set the flag only if there is a liter
5312            continue;            continue;
5313            }            }
5314          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5315            cd->assert_depth += 1;
5316          break;          break;
5317    
5318    
# Line 4778  we set the flag only if there is a liter Line 5322  we set the flag only if there is a liter
5322            {            {
5323            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5324            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5325              cd->assert_depth += 1;
5326            ptr += 2;            ptr += 2;
5327            break;            break;
5328    
5329            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5330            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5331              cd->assert_depth += 1;
5332            ptr += 2;            ptr += 2;
5333            break;            break;
5334    
# Line 4822  we set the flag only if there is a liter Line 5368  we set the flag only if there is a liter
5368              goto FAILED;              goto FAILED;
5369              }              }
5370            *code++ = n;            *code++ = n;
5371            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5372            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5373            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5374            }            }
5375          previous = NULL;          previous = NULL;
# Line 4856  we set the flag only if there is a liter Line 5402  we set the flag only if there is a liter
5402            name = ++ptr;            name = ++ptr;
5403    
5404            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5405            namelen = ptr - name;            namelen = (int)(ptr - name);
5406    
5407            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5408    
# Line 4986  we set the flag only if there is a liter Line 5532  we set the flag only if there is a liter
5532          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5533          name = ++ptr;          name = ++ptr;
5534          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5535          namelen = ptr - name;          namelen = (int)(ptr - name);
5536    
5537          /* 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
5538          reference number. */          a dummy reference number, because it was not used in the first pass.
5539            However, with the change of recursive back references to be atomic,
5540            we have to look for the number so that this state can be identified, as
5541            otherwise the incorrect length is computed. If it's not a backwards
5542            reference, the dummy number will do. */
5543    
5544          if (lengthptr != NULL)          if (lengthptr != NULL)
5545            {            {
5546              const uschar *temp;
5547    
5548            if (namelen == 0)            if (namelen == 0)
5549              {              {
5550              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5008  we set the flag only if there is a liter Line 5560  we set the flag only if there is a liter
5560              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5561              goto FAILED;              goto FAILED;
5562              }              }
5563            recno = 0;  
5564              /* The name table does not exist in the first pass, so we cannot
5565              do a simple search as in the code below. Instead, we have to scan the
5566              pattern to find the number. It is important that we scan it only as
5567              far as we have got because the syntax of named subpatterns has not
5568              been checked for the rest of the pattern, and find_parens() assumes
5569              correct syntax. In any case, it's a waste of resources to scan
5570              further. We stop the scan at the current point by temporarily
5571              adjusting the value of cd->endpattern. */
5572    
5573              temp = cd->end_pattern;
5574              cd->end_pattern = ptr;
5575              recno = find_parens(cd, name, namelen,
5576                (options & PCRE_EXTENDED) != 0, utf8);
5577              cd->end_pattern = temp;
5578              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5579            }            }
5580    
5581          /* 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 5033  we set the flag only if there is a liter Line 5600  we set the flag only if there is a liter
5600              }              }
5601            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5602                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5603                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5604              {              {
5605              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5606              goto FAILED;              goto FAILED;
# Line 5144  we set the flag only if there is a liter Line 5711  we set the flag only if there is a liter
5711              if (called == NULL)              if (called == NULL)
5712                {                {
5713                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5714                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5715                  {                  {
5716                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5717                  goto FAILED;                  goto FAILED;
5718                  }                  }
5719    
5720                  /* Fudge the value of "called" so that when it is inserted as an
5721                  offset below, what it actually inserted is the reference number
5722                  of the group. Then remember the forward reference. */
5723    
5724                called = cd->start_code + recno;                called = cd->start_code + recno;
5725                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5726                }                }
5727    
5728              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
# Line 5158  we set the flag only if there is a liter Line 5730  we set the flag only if there is a liter
5730              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5731    
5732              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5733                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5734                {                {
5735                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5736                goto FAILED;                goto FAILED;
5737                }                }
5738              }              }
5739    
5740            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
5741            "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;  
   
5742            *code = OP_RECURSE;            *code = OP_RECURSE;
5743            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
5744            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;  
5745            }            }
5746    
5747          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5242  we set the flag only if there is a liter Line 5802  we set the flag only if there is a liter
5802          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
5803          both phases.          both phases.
5804    
5805          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
5806          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. */  
5807    
5808          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5809            {            {
# Line 5255  we set the flag only if there is a liter Line 5814  we set the flag only if there is a liter
5814              }              }
5815            else            else
5816              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5817              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5818              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5819              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5820              }              }
5821    
5822            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5823            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). */  
5824    
5825            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5826            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5285  we set the flag only if there is a liter Line 5837  we set the flag only if there is a liter
5837          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5838        }       /* End of (? handling */        }       /* End of (? handling */
5839    
5840      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5841      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5842      brackets. */      brackets. */
5843    
5844      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5306  we set the flag only if there is a liter Line 5858  we set the flag only if there is a liter
5858    
5859      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions may not be repeated, but
5860      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
5861      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
5862      compilers complain otherwise. Pass in a new setting for the ims options if      because some compilers complain otherwise. */
     they have changed. */  
5863    
5864      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = (bravalue >= OP_ONCE)? code : NULL;
5865      *code = bravalue;      *code = bravalue;
# Line 5318  we set the flag only if there is a liter Line 5869  we set the flag only if there is a liter
5869    
5870      if (!compile_regex(      if (!compile_regex(
5871           newoptions,                   /* The complete new option state */           newoptions,                   /* The complete new option state */
          options & PCRE_IMS,           /* The previous ims option state */  
5872           &tempcode,                    /* Where to put code (updated) */           &tempcode,                    /* Where to put code (updated) */
5873           &ptr,                         /* Input pointer (updated) */           &ptr,                         /* Input pointer (updated) */
5874           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
# Line 5334  we set the flag only if there is a liter Line 5884  we set the flag only if there is a liter
5884             &length_prevgroup           /* Pre-compile phase */             &length_prevgroup           /* Pre-compile phase */
5885           ))           ))
5886        goto FAILED;        goto FAILED;
5887    
5888        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5889          cd->assert_depth -= 1;
5890    
5891      /* 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
5892      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 5405  we set the flag only if there is a liter Line 5958  we set the flag only if there is a liter
5958          goto FAILED;          goto FAILED;
5959          }          }
5960        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5961        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
5962        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5963        *code++ = OP_KET;        *code++ = OP_KET;
5964        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5478  we set the flag only if there is a liter Line 6031  we set the flag only if there is a liter
6031    
6032      /* ===================================================================*/      /* ===================================================================*/
6033      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
6034      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
6035      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
6036      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
6037      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
6038      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
6039        ever created. */
6040    
6041      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
6042      tempptr = ptr;      tempptr = ptr;
# Line 5572  we set the flag only if there is a liter Line 6126  we set the flag only if there is a liter
6126          }          }
6127    
6128        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
6129        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6130    
6131        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6132          {          {
6133            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6134              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6135              {
6136              *errorcodeptr = ERR69;
6137              break;
6138              }
6139          is_recurse = FALSE;          is_recurse = FALSE;
6140          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6141            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
6142            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
6143          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
6144          }          }
6145    
6146        /* Back references are handled specially; must disable firstbyte if        /* Back references are handled specially; must disable firstbyte if
6147        not set to cope with cases like (?=(\w+))\1: which would otherwise set        not set to cope with cases like (?=(\w+))\1: which would otherwise set
# Line 5596  we set the flag only if there is a liter Line 6155  we set the flag only if there is a liter
6155          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6156          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6157          previous = code;          previous = code;
6158          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6159          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6160          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6161          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 5642  we set the flag only if there is a liter Line 6201  we set the flag only if there is a liter
6201  #endif  #endif
6202    
6203        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
6204        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
6205          situation when PCRE_UCP is not set. When it *is* set, we substitute
6206          Unicode property tests. */
6207    
6208        else        else
6209          {          {
6210          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
6211          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
6212              {
6213              nestptr = ptr + 1;                   /* Where to resume */
6214              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6215              }
6216            else
6217    #endif
6218              {
6219              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6220              *code++ = -c;
6221              }
6222          }          }
6223        continue;        continue;
6224        }        }
# Line 5692  we set the flag only if there is a liter Line 6263  we set the flag only if there is a liter
6263    
6264      ONE_CHAR:      ONE_CHAR:
6265      previous = code;      previous = code;
6266      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6267      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6268    
6269      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 5756  return FALSE; Line 6327  return FALSE;
6327  /* 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
6328  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
6329  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.  
   
6330  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
6331  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
6332  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6333    
6334  Arguments:  Arguments:
6335    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  
6336    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6337    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6338    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
# Line 5785  Returns:         TRUE on success Line 6350  Returns:         TRUE on success
6350  */  */
6351    
6352  static BOOL  static BOOL
6353  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6354    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6355    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
6356    int *lengthptr)    int *lengthptr)
# Line 5802  int branchfirstbyte, branchreqbyte; Line 6367  int branchfirstbyte, branchreqbyte;
6367  int length;  int length;
6368  int orig_bracount;  int orig_bracount;
6369  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6370  branch_chain bc;  branch_chain bc;
6371    
6372  bc.outer = bcptr;  bc.outer = bcptr;
# Line 5826  pre-compile phase to find out whether an Line 6390  pre-compile phase to find out whether an
6390    
6391  /* 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
6392  so that we can detect them if (*ACCEPT) is encountered. This is also used to  so that we can detect them if (*ACCEPT) is encountered. This is also used to
6393  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6394    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6395    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6396    
6397  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6398    {    {
# Line 5852  for (;;) Line 6418  for (;;)
6418    
6419    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6420    
   /* 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;  
     }  
   
6421    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6422    
6423    if (lookbehind)    if (lookbehind)
# Line 5881  for (;;) Line 6438  for (;;)
6438      return FALSE;      return FALSE;
6439      }      }
6440    
   /* If the external options have changed during this branch, it means that we  
   are at the top level, and a leading option setting has been encountered. We  
   need to re-set the original option values to take account of this so that,  
   during the pre-compile phase, we know to allow for a re-set at the start of  
   subsequent branches. */  
   
   if (old_external_options != cd->external_options)  
     oldims = cd->external_options & PCRE_IMS;  
   
6441    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6442    has fewer than the rest. */    has fewer than the rest. */
6443    
# Line 5950  for (;;) Line 6498  for (;;)
6498        {        {
6499        int fixed_length;        int fixed_length;
6500        *code = OP_END;        *code = OP_END;
6501        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6502            FALSE, cd);
6503        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6504        if (fixed_length == -3)        if (fixed_length == -3)
6505          {          {
# Line 5971  for (;;) Line 6520  for (;;)
6520    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
6521    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
6522    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
6523    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. */  
6524    
6525    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6526      {      {
6527      if (lengthptr == NULL)      if (lengthptr == NULL)
6528        {        {
6529        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6530        do        do
6531          {          {
6532          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 5993  for (;;) Line 6540  for (;;)
6540      /* Fill in the ket */      /* Fill in the ket */
6541    
6542      *code = OP_KET;      *code = OP_KET;
6543      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6544      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6545    
6546      /* If it was a capturing subpattern, check to see if it contained any      /* If it was a capturing subpattern, check to see if it contained any
# Line 6008  for (;;) Line 6555  for (;;)
6555            code - start_bracket);            code - start_bracket);
6556          *start_bracket = OP_ONCE;          *start_bracket = OP_ONCE;
6557          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6558          PUT(start_bracket, 1, code - start_bracket);          PUT(start_bracket, 1, (int)(code - start_bracket));
6559          *code = OP_KET;          *code = OP_KET;
6560          PUT(code, 1, code - start_bracket);          PUT(code, 1, (int)(code - start_bracket));
6561          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6562          length += 2 + 2*LINK_SIZE;          length += 2 + 2*LINK_SIZE;
6563          }          }
6564        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6565        }        }
6566    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6567      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6568    
6569      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6065  for (;;) Line 6603  for (;;)
6603    else    else
6604      {      {
6605      *code = OP_ALT;      *code = OP_ALT;
6606      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6607      bc.current_branch = last_branch = code;      bc.current_branch = last_branch = code;
6608      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6609      }      }
# Line 6085  for (;;) Line 6623  for (;;)
6623  /* 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
6624  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
6625  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
6626  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
6627  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6628    
6629  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.
6630  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 6107  of the more common cases more precisely. Line 6645  of the more common cases more precisely.
6645    
6646  Arguments:  Arguments:
6647    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6648    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
6649                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6650                    the less precise approach                    the less precise approach
# Line 6117  Returns:     TRUE or FALSE Line 6654  Returns:     TRUE or FALSE
6654  */  */
6655    
6656  static BOOL  static BOOL
6657  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6658    unsigned int backref_map)    unsigned int backref_map)
6659  {  {
6660  do {  do {
6661     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6662       options, PCRE_MULTILINE, FALSE);       FALSE);
6663     register int op = *scode;     register int op = *scode;
6664    
6665     /* Non-capturing brackets */     /* Non-capturing brackets */
6666    
6667     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6668           op == OP_SBRA || op == OP_SBRAPOS)
6669       {       {
6670       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6671       }       }
6672    
6673     /* Capturing brackets */     /* Capturing brackets */
6674    
6675     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6676                op == OP_SCBRA || op == OP_SCBRAPOS)
6677       {       {
6678       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6679       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6680       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6681       }       }
6682    
6683     /* Other brackets */     /* Other brackets */
6684    
6685     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)
6686       {       {
6687       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6688       }       }
6689    
6690     /* .* 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 6160  do { Line 6699  do {
6699    
6700     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6701    
6702     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;  
6703     code += GET(code, 1);     code += GET(code, 1);
6704     }     }
6705  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6198  is_startline(const uschar *code, unsigne Line 6735  is_startline(const uschar *code, unsigne
6735  {  {
6736  do {  do {
6737     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6738       NULL, 0, FALSE);       FALSE);
6739     register int op = *scode;     register int op = *scode;
6740    
6741     /* 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 6225  do { Line 6762  do {
6762         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6763         break;         break;
6764         }         }
6765       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6766       op = *scode;       op = *scode;
6767       }       }
6768    
6769     /* Non-capturing brackets */     /* Non-capturing brackets */
6770    
6771     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6772           op == OP_SBRA || op == OP_SBRAPOS)
6773       {       {
6774       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6775       }       }
6776    
6777     /* Capturing brackets */     /* Capturing brackets */
6778    
6779     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6780                op == OP_SCBRA || op == OP_SCBRAPOS)
6781       {       {
6782       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6783       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6262  do { Line 6801  do {
6801    
6802     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6803    
6804     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6805    
6806     /* Move on to the next alternative */     /* Move on to the next alternative */
6807    
# Line 6288  we return that char, otherwise -1. Line 6827  we return that char, otherwise -1.
6827    
6828  Arguments:  Arguments:
6829    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)  
6830    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6831    
6832  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6833  */  */
6834    
6835  static int  static int
6836  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6837  {  {
6838  register int c = -1;  register int c = -1;
6839  do {  do {
6840     int d;     int d;
6841     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6842       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6843       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6844     register int op = *scode;     register int op = *scode;
6845    
6846     switch(op)     switch(op)
# Line 6310  do { Line 6849  do {
6849       return -1;       return -1;
6850    
6851       case OP_BRA:       case OP_BRA:
6852         case OP_BRAPOS:
6853       case OP_CBRA:       case OP_CBRA:
6854         case OP_SCBRA:
6855         case OP_CBRAPOS:
6856         case OP_SCBRAPOS:
6857       case OP_ASSERT:       case OP_ASSERT:
6858       case OP_ONCE:       case OP_ONCE:
6859       case OP_COND:       case OP_COND:
6860       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6861         return -1;         return -1;
6862       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6863       break;       break;
6864    
6865       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6866       scode += 2;       scode += 2;
6867         /* Fall through */
6868    
6869       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6870       case OP_PLUS:       case OP_PLUS:
6871       case OP_MINPLUS:       case OP_MINPLUS:
6872       case OP_POSPLUS:       case OP_POSPLUS:
6873       if (!inassert) return -1;       if (!inassert) return -1;
6874       if (c < 0)       if (c < 0) c = scode[1];
6875         {         else if (c != scode[1]) return -1;
6876         c = scode[1];       break;
6877         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6878         }       case OP_EXACTI:
6879       else if (c != scode[1]) return -1;       scode += 2;
6880         /* Fall through */
6881    
6882         case OP_CHARI:
6883         case OP_PLUSI:
6884         case OP_MINPLUSI:
6885         case OP_POSPLUSI:
6886         if (!inassert) return -1;
6887         if (c < 0) c = scode[1] | REQ_CASELESS;
6888           else if (c != scode[1]) return -1;
6889       break;       break;
6890       }       }
6891    
# Line 6384  int length = 1;  /* For final END opcode Line 6936  int length = 1;  /* For final END opcode
6936  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6937  int errorcode = 0;  int errorcode = 0;
6938  int skipatstart = 0;  int skipatstart = 0;
6939  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8;
6940  size_t size;  size_t size;
6941  uschar *code;  uschar *code;
6942  const uschar *codestart;  const uschar *codestart;
# Line 6454  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7006  while (ptr[skipatstart] == CHAR_LEFT_PAR
7006    
7007    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
7008      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7009      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7010        { skipatstart += 6; options |= PCRE_UCP; continue; }
7011      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7012        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7013    
7014    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
7015      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6478  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7034  while (ptr[skipatstart] == CHAR_LEFT_PAR
7034    else break;    else break;
7035    }    }
7036    
7037  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  utf8 = (options & PCRE_UTF8) != 0;
7038    
7039    /* Can't support UTF8 unless PCRE has been compiled to include the code. The
7040    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7041    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7042    not used here. */
7043    
7044  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7045  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7046       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7047    {    {
7048    errorcode = ERR44;    errorcode = ERR44;
7049    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6495  if (utf8) Line 7056  if (utf8)
7056    }    }
7057  #endif  #endif
7058    
7059    /* Can't support UCP unless PCRE has been compiled to include the code. */
7060    
7061    #ifndef SUPPORT_UCP
7062    if ((options & PCRE_UCP) != 0)
7063      {
7064      errorcode = ERR67;
7065      goto PCRE_EARLY_ERROR_RETURN;
7066      }
7067    #endif
7068    
7069  /* Check validity of \R options. */  /* Check validity of \R options. */
7070    
7071  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 6588  outside can help speed up starting point Line 7159  outside can help speed up starting point
7159  ptr += skipatstart;  ptr += skipatstart;
7160  code = cworkspace;  code = cworkspace;
7161  *code = OP_BRA;  *code = OP_BRA;
7162  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7163    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7164  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7165    
7166  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 6623  regex compiled on a system with 4-byte p Line 7193  regex compiled on a system with 4-byte p
7193  pointers. */  pointers. */
7194    
7195  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
7196  re->size = size;  re->size = (int)size;
7197  re->options = cd->external_options;  re->options = cd->external_options;
7198  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7199  re->dummy1 = 0;  re->dummy1 = 0;
# Line 6644  field; this time it's used for rememberi Line 7214  field; this time it's used for rememberi
7214  */  */
7215    
7216  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7217    cd->assert_depth = 0;
7218  cd->bracount = 0;  cd->bracount = 0;
7219  cd->names_found = 0;  cd->names_found = 0;
7220  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 6662  of the function here. */ Line 7233  of the function here. */
7233  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7234  code = (uschar *)codestart;  code = (uschar *)codestart;
7235  *code = OP_BRA;  *code = OP_BRA;
7236  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0,
7237    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7238  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7239  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7240  re->flags = cd->external_flags;  re->flags = cd->external_flags;
# Line 6694  while (errorcode == 0 && cd->hwm > cwork Line 7265  while (errorcode == 0 && cd->hwm > cwork
7265    recno = GET(codestart, offset);    recno = GET(codestart, offset);
7266    groupptr = _pcre_find_bracket(codestart, utf8, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
7267    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
7268      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7269    }    }
7270    
7271  /* 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 6729  if (cd->check_lookbehind) Line 7300  if (cd->check_lookbehind)
7300        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7301        int end_op = *be;        int end_op = *be;
7302        *be = OP_END;        *be = OP_END;
7303        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7304            cd);
7305        *be = end_op;        *be = end_op;
7306        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7307        if (fixed_length < 0)        if (fixed_length < 0)
# Line 6749  if (errorcode != 0) Line 7321  if (errorcode != 0)
7321    {    {
7322    (pcre_free)(re);    (pcre_free)(re);
7323    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
7324    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = (int)(ptr - (const uschar *)pattern);
7325    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
7326    *errorptr = find_error_text(errorcode);    *errorptr = find_error_text(errorcode);
7327    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
# Line 6768  start with ^. and also when all branches Line 7340  start with ^. and also when all branches
7340    
7341  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7342    {    {
7343    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))  
7344      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7345    else    else
7346      {      {
7347      if (firstbyte < 0)      if (firstbyte < 0)
7348        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7349      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7350        {        {
7351        int ch = firstbyte & 255;        int ch = firstbyte & 255;
# Line 6804  if (reqbyte >= 0 && Line 7375  if (reqbyte >= 0 &&
7375  case when building a production library. */  case when building a production library. */
7376    
7377  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
   
7378  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
7379    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
7380    

Legend:
Removed from v.488  
changed lines
  Added in v.629

  ViewVC Help
Powered by ViewVC 1.1.5