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

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

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

revision 457 by ph10, Sat Oct 3 16:24:08 2009 UTC revision 602 by ph10, Wed May 25 08:29:03 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2009 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
# Line 118  static const short int escapes[] = { Line 124  static const short int escapes[] = {
124       -ESC_H,                  0,       -ESC_H,                  0,
125       0,                       -ESC_K,       0,                       -ESC_K,
126       0,                       0,       0,                       0,
127       0,                       0,       -ESC_N,                  0,
128       -ESC_P,                  -ESC_Q,       -ESC_P,                  -ESC_Q,
129       -ESC_R,                  -ESC_S,       -ESC_R,                  -ESC_S,
130       0,                       0,       0,                       0,
# Line 165  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 182  string is built from string macros so th Line 188  string is built from string macros so th
188  platforms. */  platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199    STRING_ACCEPT0    STRING_ACCEPT0
200    STRING_COMMIT0    STRING_COMMIT0
201    STRING_F0    STRING_F0
# Line 196  static const char verbnames[] = Line 205  static const char verbnames[] =
205    STRING_THEN;    STRING_THEN;
206    
207  static const verbitem verbs[] = {  static const verbitem verbs[] = {
208    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
209    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
210    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
212    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
213    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
214    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 250  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 262  the number of relocations needed when a Line 320  the number of relocations needed when a
320  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
321  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 309  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 335  static const char error_texts[] = Line 397  static const char error_texts[] =
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
403    "number is too big\0"    "number is too big\0"
# Line 343  static const char error_texts[] = Line 405  static const char error_texts[] =
405    "digit expected after (?+\0"    "digit expected after (?+\0"
406    "] is an invalid data character in JavaScript compatibility mode\0"    "] is an invalid data character in JavaScript compatibility mode\0"
407    /* 65 */    /* 65 */
408    "different names for subpatterns of the same number are not allowed";    "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      ;
413    
414  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
415  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 502  static const char * Line 567  static const char *
567  find_error_text(int n)  find_error_text(int n)
568  {  {
569  const char *s = error_texts;  const char *s = error_texts;
570  for (; n > 0; n--) while (*s++ != 0) {};  for (; n > 0; n--)
571      {
572      while (*s++ != 0) {};
573      if (*s == 0) return "Error text not found (please report)";
574      }
575  return s;  return s;
576  }  }
577    
# Line 573  else Line 642  else
642    
643      case CHAR_l:      case CHAR_l:
644      case CHAR_L:      case CHAR_L:
     case CHAR_N:  
645      case CHAR_u:      case CHAR_u:
646      case CHAR_U:      case CHAR_U:
647      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
# Line 774  else Line 842  else
842      break;      break;
843    
844      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
845      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
846        coding is ASCII-specific, but then the whole concept of \cx is
847      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
848    
849      case CHAR_c:      case CHAR_c:
# Line 784  else Line 853  else
853        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
854        break;        break;
855        }        }
856    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
857  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
858          {
859          *errorcodeptr = ERR68;
860          break;
861          }
862      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
863      c ^= 0x40;      c ^= 0x40;
864  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
865      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
866      c ^= 0xC0;      c ^= 0xC0;
867  #endif  #endif
# Line 811  else Line 884  else
884      }      }
885    }    }
886    
887    /* Perl supports \N{name} for character names, as well as plain \N for "not
888    newline". PCRE does not support \N{name}. */
889    
890    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
891      *errorcodeptr = ERR37;
892    
893    /* If PCRE_UCP is set, we change the values for \d etc. */
894    
895    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
896      c -= (ESC_DU - ESC_D);
897    
898    /* Set the pointer to the final character before returning. */
899    
900  *ptrptr = ptr;  *ptrptr = ptr;
901  return c;  return c;
902  }  }
# Line 1019  top-level call starts at the beginning o Line 1105  top-level call starts at the beginning o
1105  start at a parenthesis. It scans along a pattern's text looking for capturing  start at a parenthesis. It scans along a pattern's text looking for capturing
1106  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1107  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1108  returns when it reaches a given numbered subpattern. We know that if (?P< is  returns when it reaches a given numbered subpattern. Recursion is used to keep
1109  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1110  first pass. Recursion is used to keep track of subpatterns that reset the  
1111  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1112    that if (?< or (?' or (?P< is encountered, the name will be correctly
1113    terminated because that is checked in the first pass. There is now one call to
1114    this function in the first pass, to check for a recursive back reference by
1115    name (so that we can make the whole group atomic). In this case, we need check
1116    only up to the current position in the pattern, and that is still OK because
1117    and previous occurrences will have been checked. To make this work, the test
1118    for "end of pattern" is a check against cd->end_pattern in the main loop,
1119    instead of looking for a binary zero. This means that the special first-pass
1120    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1121    processing items within the loop are OK, because afterwards the main loop will
1122    terminate.)
1123    
1124  Arguments:  Arguments:
1125    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1030  Arguments: Line 1127  Arguments:
1127    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1128    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1129    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1130      utf8         TRUE if we are in UTF-8 mode
1131    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1132    
1133  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
# Line 1037  Returns: the number of the named s Line 1135  Returns: the number of the named s
1135    
1136  static int  static int
1137  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1138    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1139  {  {
1140  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1141  int start_count = *count;  int start_count = *count;
# Line 1049  dealing with. The very first call may no Line 1147  dealing with. The very first call may no
1147    
1148  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1149    {    {
1150    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1151        ptr[2] == CHAR_VERTICAL_LINE)  
1152      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1153    
1154      /* Handle a normal, unnamed capturing parenthesis. */
1155    
1156      else if (ptr[1] != CHAR_QUESTION_MARK)
1157        {
1158        *count += 1;
1159        if (name == NULL && *count == lorn) return *count;
1160        ptr++;
1161        }
1162    
1163      /* All cases now have (? at the start. Remember when we are in a group
1164      where the parenthesis numbers are duplicated. */
1165    
1166      else if (ptr[2] == CHAR_VERTICAL_LINE)
1167      {      {
1168      ptr += 3;      ptr += 3;
1169      dup_parens = TRUE;      dup_parens = TRUE;
1170      }      }
1171    
1172    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1173    
1174    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1175      {      {
1176      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1177      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1178      }      }
1179    
1180    /* Handle a condition. If it is an assertion, just carry on so that it    /* Handle a condition. If it is an assertion, just carry on so that it
1181    is processed as normal. If not, skip to the closing parenthesis of the    is processed as normal. If not, skip to the closing parenthesis of the
1182    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1183    
1184    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1185      {      {
# Line 1079  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1191  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1191        }        }
1192      }      }
1193    
1194    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1195    
1196    else    else
1197      {      {
# Line 1102  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1214  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1214        if (name != NULL && lorn == ptr - thisname &&        if (name != NULL && lorn == ptr - thisname &&
1215            strncmp((const char *)name, (const char *)thisname, lorn) == 0)            strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1216          return *count;          return *count;
1217        term++;        term++;
1218        }        }
1219      }      }
1220    }    }
1221    
1222  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1223  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1224    first-pass call when this value is temporarily adjusted to stop at the current
1225    position. So DO NOT change this to a test for binary zero. */
1226    
1227  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1228    {    {
1229    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1230    
# Line 1148  for (; *ptr != 0; ptr++) Line 1262  for (; *ptr != 0; ptr++)
1262            break;            break;
1263          }          }
1264        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1265          {          {
1266          negate_class = TRUE;          negate_class = TRUE;
1267          ptr++;          ptr++;
1268          }          }
1269        else break;        else break;
1270        }        }
1271    
# Line 1184  for (; *ptr != 0; ptr++) Line 1298  for (; *ptr != 0; ptr++)
1298    
1299    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1300      {      {
1301      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1302        while (*ptr != 0)
1303          {
1304          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1305          ptr++;
1306    #ifdef SUPPORT_UTF8
1307          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1308    #endif
1309          }
1310      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1311      continue;      continue;
1312      }      }
# Line 1193  for (; *ptr != 0; ptr++) Line 1315  for (; *ptr != 0; ptr++)
1315    
1316    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1317      {      {
1318      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1319      if (rc > 0) return rc;      if (rc > 0) return rc;
1320      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1321      }      }
# Line 1201  for (; *ptr != 0; ptr++) Line 1323  for (; *ptr != 0; ptr++)
1323    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1324      {      {
1325      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1326      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1327      }      }
1328    
1329    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1240  Arguments: Line 1361  Arguments:
1361    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1362    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1363    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1364      utf8         TRUE if we are in UTF-8 mode
1365    
1366  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1367  */  */
1368    
1369  static int  static int
1370  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1371      BOOL utf8)
1372  {  {
1373  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1374  int count = 0;  int count = 0;
# Line 1258  matching closing parens. That is why we Line 1381  matching closing parens. That is why we
1381    
1382  for (;;)  for (;;)
1383    {    {
1384    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1385    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1386    }    }
1387    
# Line 1274  return rc; Line 1397  return rc;
1397    
1398  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1399  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1400  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1401  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1402  assertions, and also the \b assertion; for others it does not.  does not.
1403    
1404  Arguments:  Arguments:
1405    code         pointer to the start of the group    code         pointer to the start of the group
# Line 1296  for (;;) Line 1419  for (;;)
1419    {    {
1420    switch ((int)*code)    switch ((int)*code)
1421      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1422      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1423      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1424      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1317  for (;;) Line 1434  for (;;)
1434    
1435      case OP_CALLOUT:      case OP_CALLOUT:
1436      case OP_CREF:      case OP_CREF:
1437        case OP_NCREF:
1438      case OP_RREF:      case OP_RREF:
1439        case OP_NRREF:
1440      case OP_DEF:      case OP_DEF:
1441      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1442      break;      break;
# Line 1338  for (;;) Line 1457  for (;;)
1457    
1458  /* Scan a branch and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1459  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1460  In UTF8 mode, the result is in characters rather than bytes. The branch is  In UTF8 mode, the result is in characters rather than bytes. The branch is
1461  temporarily terminated with OP_END when this function is called.  temporarily terminated with OP_END when this function is called.
1462    
1463  This function is called when a backward assertion is encountered, so that if it  This function is called when a backward assertion is encountered, so that if it
1464  fails, the error message can point to the correct place in the pattern.  fails, the error message can point to the correct place in the pattern.
1465  However, we cannot do this when the assertion contains subroutine calls,  However, we cannot do this when the assertion contains subroutine calls,
1466  because they can be forward references. We solve this by remembering this case  because they can be forward references. We solve this by remembering this case
1467  and doing the check at the end; a flag specifies which mode we are running in.  and doing the check at the end; a flag specifies which mode we are running in.
1468    
1469  Arguments:  Arguments:
1470    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1471    options  the compiling options    options  the compiling options
1472    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1473    cd       the "compile data" structure    cd       the "compile data" structure
1474    
1475  Returns:   the fixed length,  Returns:   the fixed length,
1476               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1477               or -2 if \C was encountered               or -2 if \C was encountered
1478               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
# Line 1403  for (;;) Line 1522  for (;;)
1522      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1523      branchlength = 0;      branchlength = 0;
1524      break;      break;
1525    
1526      /* A true recursion implies not fixed length, but a subroutine call may      /* A true recursion implies not fixed length, but a subroutine call may
1527      be OK. If the subroutine is a forward reference, we can't deal with      be OK. If the subroutine is a forward reference, we can't deal with
1528      it until the end of the pattern, so return -3. */      it until the end of the pattern, so return -3. */
1529    
1530      case OP_RECURSE:      case OP_RECURSE:
1531      if (!atend) return -3;      if (!atend) return -3;
1532      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1533      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1534      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1535      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, options, atend, cd);
1536      if (d < 0) return d;      if (d < 0) return d;
1537      branchlength += d;      branchlength += d;
1538      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1539      break;      break;
1540    
1541      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1542    
# Line 1432  for (;;) Line 1551  for (;;)
1551    
1552      case OP_REVERSE:      case OP_REVERSE:
1553      case OP_CREF:      case OP_CREF:
1554        case OP_NCREF:
1555      case OP_RREF:      case OP_RREF:
1556        case OP_NRREF:
1557      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1558      case OP_CALLOUT:      case OP_CALLOUT:
1559      case OP_SOD:      case OP_SOD:
1560      case OP_SOM:      case OP_SOM:
1561        case OP_SET_SOM:
1562      case OP_EOD:      case OP_EOD:
1563      case OP_EODN:      case OP_EODN:
1564      case OP_CIRC:      case OP_CIRC:
1565        case OP_CIRCM:
1566      case OP_DOLL:      case OP_DOLL:
1567        case OP_DOLLM:
1568      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1569      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1570      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1450  for (;;) Line 1573  for (;;)
1573      /* Handle literal characters */      /* Handle literal characters */
1574    
1575      case OP_CHAR:      case OP_CHAR:
1576      case OP_CHARNC:      case OP_CHARI:
1577      case OP_NOT:      case OP_NOT:
1578        case OP_NOTI:
1579      branchlength++;      branchlength++;
1580      cc += 2;      cc += 2;
1581  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1582      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1583        cc += _pcre_utf8_table4[cc[-1] & 0x3f];        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
1584  #endif  #endif
1585      break;      break;
# Line 1467  for (;;) Line 1591  for (;;)
1591      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1592      cc += 4;      cc += 4;
1593  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1594      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1595        cc += _pcre_utf8_table4[cc[-1] & 0x3f];        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
1596  #endif  #endif
1597      break;      break;
# Line 1552  for (;;) Line 1676  for (;;)
1676    
1677  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1678  capturing bracket with the given number, or, if the number is negative, an  capturing bracket with the given number, or, if the number is negative, an
1679  instance of OP_REVERSE for a lookbehind. The function is global in the C sense  instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1680  so that it can be called from pcre_study() when finding the minimum matching  so that it can be called from pcre_study() when finding the minimum matching
1681  length.  length.
1682    
1683  Arguments:  Arguments:
# Line 1577  for (;;) Line 1701  for (;;)
1701    the table is zero; the actual length is stored in the compiled code. */    the table is zero; the actual length is stored in the compiled code. */
1702    
1703    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1704    
1705    /* Handle recursion */    /* Handle recursion */
1706    
1707    else if (c == OP_REVERSE)    else if (c == OP_REVERSE)
1708      {      {
1709      if (number < 0) return (uschar *)code;      if (number < 0) return (uschar *)code;
1710      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1711      }      }
1712    
# Line 1597  for (;;) Line 1721  for (;;)
1721    
1722    /* 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
1723    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
1724    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1725      must add in its length. */
1726    
1727    else    else
1728      {      {
# Line 1621  for (;;) Line 1746  for (;;)
1746        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1747        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1748        break;        break;
1749    
1750          case OP_MARK:
1751          case OP_PRUNE_ARG:
1752          case OP_SKIP_ARG:
1753          code += code[1];
1754          break;
1755    
1756          case OP_THEN_ARG:
1757          code += code[1+LINK_SIZE];
1758          break;
1759        }        }
1760    
1761      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1635  for (;;) Line 1770  for (;;)
1770      if (utf8) switch(c)      if (utf8) switch(c)
1771        {        {
1772        case OP_CHAR:        case OP_CHAR:
1773        case OP_CHARNC:        case OP_CHARI:
1774        case OP_EXACT:        case OP_EXACT:
1775          case OP_EXACTI:
1776        case OP_UPTO:        case OP_UPTO:
1777          case OP_UPTOI:
1778        case OP_MINUPTO:        case OP_MINUPTO:
1779          case OP_MINUPTOI:
1780        case OP_POSUPTO:        case OP_POSUPTO:
1781          case OP_POSUPTOI:
1782        case OP_STAR:        case OP_STAR:
1783          case OP_STARI:
1784        case OP_MINSTAR:        case OP_MINSTAR:
1785          case OP_MINSTARI:
1786        case OP_POSSTAR:        case OP_POSSTAR:
1787          case OP_POSSTARI:
1788        case OP_PLUS:        case OP_PLUS:
1789          case OP_PLUSI:
1790        case OP_MINPLUS:        case OP_MINPLUS:
1791          case OP_MINPLUSI:
1792        case OP_POSPLUS:        case OP_POSPLUS:
1793          case OP_POSPLUSI:
1794        case OP_QUERY:        case OP_QUERY:
1795          case OP_QUERYI:
1796        case OP_MINQUERY:        case OP_MINQUERY:
1797          case OP_MINQUERYI:
1798        case OP_POSQUERY:        case OP_POSQUERY:
1799          case OP_POSQUERYI:
1800        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1801        break;        break;
1802        }        }
# Line 1692  for (;;) Line 1840  for (;;)
1840    
1841    /* 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
1842    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
1843    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1844      must add in its length. */
1845    
1846    else    else
1847      {      {
# Line 1716  for (;;) Line 1865  for (;;)
1865        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1866        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1867        break;        break;
1868    
1869          case OP_MARK:
1870          case OP_PRUNE_ARG:
1871          case OP_SKIP_ARG:
1872          code += code[1];
1873          break;
1874    
1875          case OP_THEN_ARG:
1876          code += code[1+LINK_SIZE];
1877          break;
1878        }        }
1879    
1880      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1730  for (;;) Line 1889  for (;;)
1889      if (utf8) switch(c)      if (utf8) switch(c)
1890        {        {
1891        case OP_CHAR:        case OP_CHAR:
1892        case OP_CHARNC:        case OP_CHARI:
1893        case OP_EXACT:        case OP_EXACT:
1894          case OP_EXACTI:
1895        case OP_UPTO:        case OP_UPTO:
1896          case OP_UPTOI:
1897        case OP_MINUPTO:        case OP_MINUPTO:
1898          case OP_MINUPTOI:
1899        case OP_POSUPTO:        case OP_POSUPTO:
1900          case OP_POSUPTOI:
1901        case OP_STAR:        case OP_STAR:
1902          case OP_STARI:
1903        case OP_MINSTAR:        case OP_MINSTAR:
1904          case OP_MINSTARI:
1905        case OP_POSSTAR:        case OP_POSSTAR:
1906          case OP_POSSTARI:
1907        case OP_PLUS:        case OP_PLUS:
1908          case OP_PLUSI:
1909        case OP_MINPLUS:        case OP_MINPLUS:
1910          case OP_MINPLUSI:
1911        case OP_POSPLUS:        case OP_POSPLUS:
1912          case OP_POSPLUSI:
1913        case OP_QUERY:        case OP_QUERY:
1914          case OP_QUERYI:
1915        case OP_MINQUERY:        case OP_MINQUERY:
1916          case OP_MINQUERYI:
1917        case OP_POSQUERY:        case OP_POSQUERY:
1918          case OP_POSQUERYI:
1919        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1920        break;        break;
1921        }        }
# Line 1772  Arguments: Line 1944  Arguments:
1944    code        points to start of search    code        points to start of search
1945    endcode     points to where to stop    endcode     points to where to stop
1946    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1947      cd          contains pointers to tables etc.
1948    
1949  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1950  */  */
1951    
1952  static BOOL  static BOOL
1953  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1954      compile_data *cd)
1955  {  {
1956  register int c;  register int c;
1957  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
# Line 1808  for (code = first_significant_code(code Line 1982  for (code = first_significant_code(code
1982      continue;      continue;
1983      }      }
1984    
1985      /* For a recursion/subroutine call, if its end has been reached, which
1986      implies a subroutine call, we can scan it. */
1987    
1988      if (c == OP_RECURSE)
1989        {
1990        BOOL empty_branch = FALSE;
1991        const uschar *scode = cd->start_code + GET(code, 1);
1992        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1993        do
1994          {
1995          if (could_be_empty_branch(scode, endcode, utf8, cd))
1996            {
1997            empty_branch = TRUE;
1998            break;
1999            }
2000          scode += GET(scode, 1);
2001          }
2002        while (*scode == OP_ALT);
2003        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2004        continue;
2005        }
2006    
2007    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2008    
2009    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
# Line 1826  for (code = first_significant_code(code Line 2022  for (code = first_significant_code(code
2022        empty_branch = FALSE;        empty_branch = FALSE;
2023        do        do
2024          {          {
2025          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2026            empty_branch = TRUE;            empty_branch = TRUE;
2027          code += GET(code, 1);          code += GET(code, 1);
2028          }          }
# Line 1897  for (code = first_significant_code(code Line 2093  for (code = first_significant_code(code
2093      case OP_ALLANY:      case OP_ALLANY:
2094      case OP_ANYBYTE:      case OP_ANYBYTE:
2095      case OP_CHAR:      case OP_CHAR:
2096      case OP_CHARNC:      case OP_CHARI:
2097      case OP_NOT:      case OP_NOT:
2098        case OP_NOTI:
2099      case OP_PLUS:      case OP_PLUS:
2100      case OP_MINPLUS:      case OP_MINPLUS:
2101      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1946  for (code = first_significant_code(code Line 2143  for (code = first_significant_code(code
2143    
2144  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2145      case OP_STAR:      case OP_STAR:
2146        case OP_STARI:
2147      case OP_MINSTAR:      case OP_MINSTAR:
2148        case OP_MINSTARI:
2149      case OP_POSSTAR:      case OP_POSSTAR:
2150        case OP_POSSTARI:
2151      case OP_QUERY:      case OP_QUERY:
2152        case OP_QUERYI:
2153      case OP_MINQUERY:      case OP_MINQUERY:
2154        case OP_MINQUERYI:
2155      case OP_POSQUERY:      case OP_POSQUERY:
2156        case OP_POSQUERYI:
2157      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2158      break;      break;
2159    
2160      case OP_UPTO:      case OP_UPTO:
2161        case OP_UPTOI:
2162      case OP_MINUPTO:      case OP_MINUPTO:
2163        case OP_MINUPTOI:
2164      case OP_POSUPTO:      case OP_POSUPTO:
2165        case OP_POSUPTOI:
2166      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2167      break;      break;
2168  #endif  #endif
2169    
2170        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2171        string. */
2172    
2173        case OP_MARK:
2174        case OP_PRUNE_ARG:
2175        case OP_SKIP_ARG:
2176        code += code[1];
2177        break;
2178    
2179        case OP_THEN_ARG:
2180        code += code[1+LINK_SIZE];
2181        break;
2182    
2183        /* None of the remaining opcodes are required to match a character. */
2184    
2185        default:
2186        break;
2187      }      }
2188    }    }
2189    
# Line 1982  Arguments: Line 2206  Arguments:
2206    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2207    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2208    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2209      cd          pointers to tables etc
2210    
2211  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2212  */  */
2213    
2214  static BOOL  static BOOL
2215  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2216    BOOL utf8)    BOOL utf8, compile_data *cd)
2217  {  {
2218  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2219    {    {
2220    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2221        return FALSE;
2222    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2223    }    }
2224  return TRUE;  return TRUE;
# Line 2174  auto_callout(uschar *code, const uschar Line 2400  auto_callout(uschar *code, const uschar
2400  {  {
2401  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2402  *code++ = 255;  *code++ = 255;
2403  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2404  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2405  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2406  }  }
2407    
# Line 2200  Returns: nothing Line 2426  Returns: nothing
2426  static void  static void
2427  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2428  {  {
2429  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2430  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2431  }  }
2432    
# Line 2250  for (++c; c <= d; c++) Line 2476  for (++c; c <= d; c++)
2476    
2477  return TRUE;  return TRUE;
2478  }  }
2479    
2480    
2481    
2482    /*************************************************
2483    *        Check a character and a property        *
2484    *************************************************/
2485    
2486    /* This function is called by check_auto_possessive() when a property item
2487    is adjacent to a fixed character.
2488    
2489    Arguments:
2490      c            the character
2491      ptype        the property type
2492      pdata        the data for the type
2493      negated      TRUE if it's a negated property (\P or \p{^)
2494    
2495    Returns:       TRUE if auto-possessifying is OK
2496    */
2497    
2498    static BOOL
2499    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2500    {
2501    const ucd_record *prop = GET_UCD(c);
2502    switch(ptype)
2503      {
2504      case PT_LAMP:
2505      return (prop->chartype == ucp_Lu ||
2506              prop->chartype == ucp_Ll ||
2507              prop->chartype == ucp_Lt) == negated;
2508    
2509      case PT_GC:
2510      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2511    
2512      case PT_PC:
2513      return (pdata == prop->chartype) == negated;
2514    
2515      case PT_SC:
2516      return (pdata == prop->script) == negated;
2517    
2518      /* These are specials */
2519    
2520      case PT_ALNUM:
2521      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2522              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2523    
2524      case PT_SPACE:    /* Perl space */
2525      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2526              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2527              == negated;
2528    
2529      case PT_PXSPACE:  /* POSIX space */
2530      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2531              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2532              c == CHAR_FF || c == CHAR_CR)
2533              == negated;
2534    
2535      case PT_WORD:
2536      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2537              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2538              c == CHAR_UNDERSCORE) == negated;
2539      }
2540    return FALSE;
2541    }
2542  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2543    
2544    
# Line 2263  whether the next thing could possibly ma Line 2552  whether the next thing could possibly ma
2552  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2553    
2554  Arguments:  Arguments:
2555    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2556    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2557    ptr           next character in pattern    ptr           next character in pattern
2558    options       options bits    options       options bits
2559    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2275  Returns: TRUE if possessifying is Line 2562  Returns: TRUE if possessifying is
2562  */  */
2563    
2564  static BOOL  static BOOL
2565  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2566    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2567  {  {
2568  int next;  int c, next;
2569    int op_code = *previous++;
2570    
2571  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2572    
# Line 2289  if ((options & PCRE_EXTENDED) != 0) Line 2577  if ((options & PCRE_EXTENDED) != 0)
2577      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2578      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2579        {        {
2580        while (*(++ptr) != 0)        ptr++;
2581          while (*ptr != 0)
2582            {
2583          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2584            ptr++;
2585    #ifdef SUPPORT_UTF8
2586            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2587    #endif
2588            }
2589        }        }
2590      else break;      else break;
2591      }      }
# Line 2326  if ((options & PCRE_EXTENDED) != 0) Line 2621  if ((options & PCRE_EXTENDED) != 0)
2621      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2622      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2623        {        {
2624        while (*(++ptr) != 0)        ptr++;
2625          while (*ptr != 0)
2626            {
2627          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2628            ptr++;
2629    #ifdef SUPPORT_UTF8
2630            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2631    #endif
2632            }
2633        }        }
2634      else break;      else break;
2635      }      }
# Line 2339  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2641  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2641    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)
2642      return FALSE;      return FALSE;
2643    
2644  /* 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
2645  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. */  
2646    
2647  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2648    {    {
2649    case OP_CHAR:    case OP_CHAR:
2650  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2651    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2652  #else  #else
2653    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2654  #endif  #endif
2655    return item != next;    return c != next;
2656    
2657    /* 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
2658    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
2659    high-valued characters. */    high-valued characters. */
2660    
2661    case OP_CHARNC:    case OP_CHARI:
2662  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2663    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2664    #else
2665      c = *previous;
2666  #endif  #endif
2667    if (item == next) return FALSE;    if (c == next) return FALSE;
2668  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2669    if (utf8)    if (utf8)
2670      {      {
# Line 2376  if (next >= 0) switch(op_code) Line 2675  if (next >= 0) switch(op_code)
2675  #else  #else
2676      othercase = NOTACHAR;      othercase = NOTACHAR;
2677  #endif  #endif
2678      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2679      }      }
2680    else    else
2681  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2682    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2683    
2684    /* 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
2685      opcodes are not used for multi-byte characters, because they are coded using
2686      an XCLASS instead. */
2687    
2688    case OP_NOT:    case OP_NOT:
2689    if (item == next) return TRUE;    return (c = *previous) == next;
2690    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2691      case OP_NOTI:
2692      if ((c = *previous) == next) return TRUE;
2693  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2694    if (utf8)    if (utf8)
2695      {      {
# Line 2397  if (next >= 0) switch(op_code) Line 2700  if (next >= 0) switch(op_code)
2700  #else  #else
2701      othercase = NOTACHAR;      othercase = NOTACHAR;
2702  #endif  #endif
2703      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2704      }      }
2705    else    else
2706  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2707    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2708    
2709      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2710      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2711    
2712    case OP_DIGIT:    case OP_DIGIT:
2713    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2444  if (next >= 0) switch(op_code) Line 2750  if (next >= 0) switch(op_code)
2750      case 0x202f:      case 0x202f:
2751      case 0x205f:      case 0x205f:
2752      case 0x3000:      case 0x3000:
2753      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2754      default:      default:
2755      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2756      }      }
2757    
2758      case OP_ANYNL:
2759    case OP_VSPACE:    case OP_VSPACE:
2760    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2761    switch(next)    switch(next)
# Line 2460  if (next >= 0) switch(op_code) Line 2767  if (next >= 0) switch(op_code)
2767      case 0x85:      case 0x85:
2768      case 0x2028:      case 0x2028:
2769      case 0x2029:      case 0x2029:
2770      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2771      default:      default:
2772      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2773      }      }
2774    
2775    #ifdef SUPPORT_UCP
2776      case OP_PROP:
2777      return check_char_prop(next, previous[0], previous[1], FALSE);
2778    
2779      case OP_NOTPROP:
2780      return check_char_prop(next, previous[0], previous[1], TRUE);
2781    #endif
2782    
2783    default:    default:
2784    return FALSE;    return FALSE;
2785    }    }
2786    
2787    
2788  /* 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
2789    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2790    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2791    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2792    replaced by OP_PROP codes when PCRE_UCP is set. */
2793    
2794  switch(op_code)  switch(op_code)
2795    {    {
2796    case OP_CHAR:    case OP_CHAR:
2797    case OP_CHARNC:    case OP_CHARI:
2798  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2799    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2800    #else
2801      c = *previous;
2802  #endif  #endif
2803    switch(-next)    switch(-next)
2804      {      {
2805      case ESC_d:      case ESC_d:
2806      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2807    
2808      case ESC_D:      case ESC_D:
2809      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2810    
2811      case ESC_s:      case ESC_s:
2812      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2813    
2814      case ESC_S:      case ESC_S:
2815      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2816    
2817      case ESC_w:      case ESC_w:
2818      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2819    
2820      case ESC_W:      case ESC_W:
2821      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2822    
2823      case ESC_h:      case ESC_h:
2824      case ESC_H:      case ESC_H:
2825      switch(item)      switch(c)
2826        {        {
2827        case 0x09:        case 0x09:
2828        case 0x20:        case 0x20:
# Line 2529  switch(op_code) Line 2850  switch(op_code)
2850    
2851      case ESC_v:      case ESC_v:
2852      case ESC_V:      case ESC_V:
2853      switch(item)      switch(c)
2854        {        {
2855        case 0x0a:        case 0x0a:
2856        case 0x0b:        case 0x0b:
# Line 2543  switch(op_code) Line 2864  switch(op_code)
2864        return -next == ESC_v;        return -next == ESC_v;
2865        }        }
2866    
2867        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2868        their substitutions and process them. The result will always be either
2869        -ESC_p or -ESC_P. Then fall through to process those values. */
2870    
2871    #ifdef SUPPORT_UCP
2872        case ESC_du:
2873        case ESC_DU:
2874        case ESC_wu:
2875        case ESC_WU:
2876        case ESC_su:
2877        case ESC_SU:
2878          {
2879          int temperrorcode = 0;
2880          ptr = substitutes[-next - ESC_DU];
2881          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2882          if (temperrorcode != 0) return FALSE;
2883          ptr++;    /* For compatibility */
2884          }
2885        /* Fall through */
2886    
2887        case ESC_p:
2888        case ESC_P:
2889          {
2890          int ptype, pdata, errorcodeptr;
2891          BOOL negated;
2892    
2893          ptr--;      /* Make ptr point at the p or P */
2894          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2895          if (ptype < 0) return FALSE;
2896          ptr++;      /* Point past the final curly ket */
2897    
2898          /* If the property item is optional, we have to give up. (When generated
2899          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2900          to the original \d etc. At this point, ptr will point to a zero byte. */
2901    
2902          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2903            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2904              return FALSE;
2905    
2906          /* Do the property check. */
2907    
2908          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2909          }
2910    #endif
2911    
2912      default:      default:
2913      return FALSE;      return FALSE;
2914      }      }
2915    
2916      /* In principle, support for Unicode properties should be integrated here as
2917      well. It means re-organizing the above code so as to get hold of the property
2918      values before switching on the op-code. However, I wonder how many patterns
2919      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2920      these op-codes are never generated.) */
2921    
2922    case OP_DIGIT:    case OP_DIGIT:
2923    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2924           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2925    
2926    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2927    return next == -ESC_d;    return next == -ESC_d;
2928    
2929    case OP_WHITESPACE:    case OP_WHITESPACE:
2930    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2931    
2932    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2933    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2934    
2935    case OP_HSPACE:    case OP_HSPACE:
2936    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2937             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2938    
2939    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2940    return next == -ESC_h;    return next == -ESC_h;
2941    
2942    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2943      case OP_ANYNL:
2944    case OP_VSPACE:    case OP_VSPACE:
2945    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2946    
2947    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2948    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2949    
2950    case OP_WORDCHAR:    case OP_WORDCHAR:
2951    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2952             next == -ESC_v || next == -ESC_R;
2953    
2954    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2955    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2638  BOOL inescq = FALSE; Line 3013  BOOL inescq = FALSE;
3013  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3014  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3015  const uschar *tempptr;  const uschar *tempptr;
3016    const uschar *nestptr = NULL;
3017  uschar *previous = NULL;  uschar *previous = NULL;
3018  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3019  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2654  BOOL utf8 = FALSE; Line 3030  BOOL utf8 = FALSE;
3030  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
3031  #endif  #endif
3032    
3033  #ifdef DEBUG  #ifdef PCRE_DEBUG
3034  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3035  #endif  #endif
3036    
# Line 2708  for (;; ptr++) Line 3084  for (;; ptr++)
3084    
3085    c = *ptr;    c = *ptr;
3086    
3087      /* If we are at the end of a nested substitution, revert to the outer level
3088      string. Nesting only happens one level deep. */
3089    
3090      if (c == 0 && nestptr != NULL)
3091        {
3092        ptr = nestptr;
3093        nestptr = NULL;
3094        c = *ptr;
3095        }
3096    
3097    /* 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
3098    previous cycle of this loop. */    previous cycle of this loop. */
3099    
3100    if (lengthptr != NULL)    if (lengthptr != NULL)
3101      {      {
3102  #ifdef DEBUG  #ifdef PCRE_DEBUG
3103      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3104  #endif  #endif
3105      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3106        {        {
3107        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3108        goto FAILED;        goto FAILED;
# Line 2738  for (;; ptr++) Line 3124  for (;; ptr++)
3124        goto FAILED;        goto FAILED;
3125        }        }
3126    
3127      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3128      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));
3129    
3130      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
# Line 2765  for (;; ptr++) Line 3151  for (;; ptr++)
3151    /* 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
3152    reference list. */    reference list. */
3153    
3154    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3155      {      {
3156      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3157      goto FAILED;      goto FAILED;
# Line 2820  for (;; ptr++) Line 3206  for (;; ptr++)
3206      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3207      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3208        {        {
3209        while (*(++ptr) != 0)        ptr++;
3210          while (*ptr != 0)
3211          {          {
3212          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3213            ptr++;
3214    #ifdef SUPPORT_UTF8
3215            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3216    #endif
3217          }          }
3218        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3219    
# Line 2856  for (;; ptr++) Line 3247  for (;; ptr++)
3247          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3248          goto FAILED;          goto FAILED;
3249          }          }
3250        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3251        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3252        }        }
3253      return TRUE;      return TRUE;
# Line 2867  for (;; ptr++) Line 3258  for (;; ptr++)
3258      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3259    
3260      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3261        previous = NULL;
3262      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3263        {        {
3264        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3265          *code++ = OP_CIRCM;
3266        }        }
3267      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3268      break;      break;
3269    
3270      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3271      previous = NULL;      previous = NULL;
3272      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3273      break;      break;
3274    
3275      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
# Line 3061  for (;; ptr++) Line 3453  for (;; ptr++)
3453            ptr++;            ptr++;
3454            }            }
3455    
3456          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3457          if (posix_class < 0)          if (posix_class < 0)
3458            {            {
3459            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3075  for (;; ptr++) Line 3467  for (;; ptr++)
3467          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3468            posix_class = 0;            posix_class = 0;
3469    
3470          /* 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
3471          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3472          subtract bits that may be in the main map already. At the end we or the  
3473          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3474            if ((options & PCRE_UCP) != 0)
3475              {
3476              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3477              if (posix_substitutes[pc] != NULL)
3478                {
3479                nestptr = tempptr + 1;
3480                ptr = posix_substitutes[pc] - 1;
3481                continue;
3482                }
3483              }
3484    #endif
3485            /* In the non-UCP case, we build the bit map for the POSIX class in a
3486            chunk of local store because we may be adding and subtracting from it,
3487            and we don't want to subtract bits that may be in the main map already.
3488            At the end we or the result into the bit map that is being built. */
3489    
3490          posix_class *= 3;          posix_class *= 3;
3491    
# Line 3122  for (;; ptr++) Line 3529  for (;; ptr++)
3529    
3530        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3531        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
3532        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
3533        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
3534        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
3535        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3536          PCRE_EXTRA is set. */
3537    
3538        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3539          {          {
3540          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3541          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3542    
3543          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 */  
3544          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3545            {            {
3546            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3151  for (;; ptr++) Line 3557  for (;; ptr++)
3557            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3558            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3559    
3560            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3561              {              {
3562    #ifdef SUPPORT_UCP
3563                case ESC_du:     /* These are the values given for \d etc */
3564                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3565                case ESC_wu:     /* escape sequence with an appropriate \p */
3566                case ESC_WU:     /* or \P to test Unicode properties instead */
3567                case ESC_su:     /* of the default ASCII testing. */
3568                case ESC_SU:
3569                nestptr = ptr;
3570                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3571                class_charcount -= 2;                /* Undo! */
3572                continue;
3573    #endif
3574              case ESC_d:              case ESC_d:
3575              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3576              continue;              continue;
# Line 3173  for (;; ptr++) Line 3589  for (;; ptr++)
3589              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3590              continue;              continue;
3591    
3592                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3593                if it was previously set by something earlier in the character
3594                class. */
3595    
3596              case ESC_s:              case ESC_s:
3597              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3598              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3599                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3600              continue;              continue;
3601    
3602              case ESC_S:              case ESC_S:
# Line 3184  for (;; ptr++) Line 3605  for (;; ptr++)
3605              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3606              continue;              continue;
3607    
3608              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)  
             {  
3609              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3610              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3611              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3221  for (;; ptr++) Line 3629  for (;; ptr++)
3629                }                }
3630  #endif  #endif
3631              continue;              continue;
             }  
3632    
3633            if (-c == ESC_H)              case ESC_H:
             {  
3634              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3635                {                {
3636                int x = 0xff;                int x = 0xff;
# Line 3266  for (;; ptr++) Line 3672  for (;; ptr++)
3672                }                }
3673  #endif  #endif
3674              continue;              continue;
             }  
3675    
3676            if (-c == ESC_v)              case ESC_v:
             {  
3677              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3678              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3679              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3285  for (;; ptr++) Line 3689  for (;; ptr++)
3689                }                }
3690  #endif  #endif
3691              continue;              continue;
             }  
3692    
3693            if (-c == ESC_V)              case ESC_V:
             {  
3694              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3695                {                {
3696                int x = 0xff;                int x = 0xff;
# Line 3318  for (;; ptr++) Line 3720  for (;; ptr++)
3720                }                }
3721  #endif  #endif
3722              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3723    
3724  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3725            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3726              {              case ESC_P:
3727              BOOL negated;                {
3728              int pdata;                BOOL negated;
3729              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3730              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3731              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3732              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3733                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3734              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3735              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3736              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3737              continue;                class_charcount -= 2;   /* Not a < 256 character */
3738              }                continue;
3739                  }
3740  #endif  #endif
3741            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3742            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3743            treated as literals. */              treated as literals. */
3744    
3745            if ((options & PCRE_EXTRA) != 0)              default:
3746              {              if ((options & PCRE_EXTRA) != 0)
3747              *errorcodeptr = ERR7;                {
3748              goto FAILED;                *errorcodeptr = ERR7;
3749                  goto FAILED;
3750                  }
3751                class_charcount -= 2;  /* Undo the default count from above */
3752                c = *ptr;              /* Get the final character and fall through */
3753                break;
3754              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3755            }            }
3756    
3757          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
# Line 3419  for (;; ptr++) Line 3821  for (;; ptr++)
3821            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3822            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3823    
3824            /* \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 */  
3825    
3826            if (d < 0)            if (d < 0)
3827              {              {
3828              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  
3829                {                {
3830                ptr = oldptr;                ptr = oldptr;
3831                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3592  for (;; ptr++) Line 3991  for (;; ptr++)
3991          }          }
3992        }        }
3993    
3994      /* 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.
3995        If we are at the end of an internal nested string, revert to the outer
3996        string. */
3997    
3998        while (((c = *(++ptr)) != 0 ||
3999               (nestptr != NULL &&
4000                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4001               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4002    
4003      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4004    
4005      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4006        {        {
4007        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4008        goto FAILED;        goto FAILED;
4009        }        }
4010    
   
 /* 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  
   
   
4011      /* 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
4012      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
4013      use of \p or \P, in other words, no use of any XCLASS features, we can      use of \p or \P, in other words, no use of any XCLASS features, we can
# Line 3628  we set the flag only if there is a liter Line 4015  we set the flag only if there is a liter
4015    
4016      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
4017      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4018      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4019      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4020    
4021      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
4022      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.
4023      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
4024      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
4025      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
4026      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4027    
4028  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4029      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3647  we set the flag only if there is a liter Line 4034  we set the flag only if there is a liter
4034        {        {
4035        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4036    
4037        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4038    
4039        if (negate_class)        if (negate_class)
4040          {          {
4041          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4042          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4043          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4044          *code++ = class_lastchar;          *code++ = class_lastchar;
4045          break;          break;
4046          }          }
# Line 3684  we set the flag only if there is a liter Line 4071  we set the flag only if there is a liter
4071    
4072      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4073      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4074      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
4075      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
4076      (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
4077      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
4078        actual compiled code. */
4079    
4080  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4081      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4082        {        {
4083        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4084        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3716  we set the flag only if there is a liter Line 4104  we set the flag only if there is a liter
4104        }        }
4105  #endif  #endif
4106    
4107      /* 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
4108      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
4109      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
4110      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4111        negating it if necessary. */
4112    
4113      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4114      if (negate_class)      if (negate_class)
# Line 3809  we set the flag only if there is a liter Line 4198  we set the flag only if there is a liter
4198      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
4199      instead.  */      instead.  */
4200    
4201      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4202        {        {
4203          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4204    
4205        /* 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
4206        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
4207        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 3843  we set the flag only if there is a liter Line 4234  we set the flag only if there is a liter
4234    
4235        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4236            repeat_max < 0 &&            repeat_max < 0 &&
4237            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4238          {          {
4239          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4240          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3856  we set the flag only if there is a liter Line 4246  we set the flag only if there is a liter
4246      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4247      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-
4248      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4249      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
4250      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4251    
4252      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4253        {        {
4254        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4255        c = previous[1];        c = previous[1];
4256        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4257            repeat_max < 0 &&            repeat_max < 0 &&
4258            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4259          {          {
4260          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4261          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3889  we set the flag only if there is a liter Line 4279  we set the flag only if there is a liter
4279    
4280        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4281            repeat_max < 0 &&            repeat_max < 0 &&
4282            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4283          {          {
4284          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4285          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3911  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 (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4303    
4304        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4305        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4306        removed: */        removed: */
4307    
4308        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4309        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4310    
4311        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4312        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4313    
4314        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4315    
# Line 4058  we set the flag only if there is a liter Line 4448  we set the flag only if there is a liter
4448  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4449               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4450  #endif  #endif
4451               *previous == OP_REF)               *previous == OP_REF ||
4452                 *previous == OP_REFI)
4453        {        {
4454        if (repeat_max == 0)        if (repeat_max == 0)
4455          {          {
# Line 4066  we set the flag only if there is a liter Line 4457  we set the flag only if there is a liter
4457          goto END_REPEAT;          goto END_REPEAT;
4458          }          }
4459    
4460        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4461        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4462        removed: */        removed: */
4463    
# Line 4074  we set the flag only if there is a liter Line 4465  we set the flag only if there is a liter
4465        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4466    
4467        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4468        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4469    
4470        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4471          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 4099  we set the flag only if there is a liter Line 4490  we set the flag only if there is a liter
4490        {        {
4491        register int i;        register int i;
4492        int ketoffset = 0;        int ketoffset = 0;
4493        int len = code - previous;        int len = (int)(code - previous);
4494        uschar *bralink = NULL;        uschar *bralink = NULL;
4495    
4496        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 4112  we set the flag only if there is a liter Line 4503  we set the flag only if there is a liter
4503    
4504        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
4505        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
4506        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. */
       the final KET, so we can't find the end just by going back from the code  
       pointer. */  
4507    
4508        if (repeat_max == -1)        if (repeat_max == -1)
4509          {          {
4510          register uschar *ket = previous;          register uschar *ket = previous;
4511          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4512          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4513          }          }
4514    
4515        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 4188  we set the flag only if there is a liter Line 4577  we set the flag only if there is a liter
4577            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4578            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4579    
4580            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4581            bralink = previous;            bralink = previous;
4582            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4583            }            }
# Line 4209  we set the flag only if there is a liter Line 4598  we set the flag only if there is a liter
4598            {            {
4599            /* In the pre-compile phase, we don't actually do the replication. We            /* In the pre-compile phase, we don't actually do the replication. We
4600            just adjust the length as if we had. Do some paranoid checks for            just adjust the length as if we had. Do some paranoid checks for
4601            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4602              integer type when available, otherwise double. */
4603    
4604            if (lengthptr != NULL)            if (lengthptr != NULL)
4605              {              {
4606              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4607              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4608                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4609                        (INT64_OR_DOUBLE)INT_MAX ||
4610                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4611                {                {
4612                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4261  we set the flag only if there is a liter Line 4652  we set the flag only if there is a liter
4652          just adjust the length as if we had. For each repetition we must add 1          just adjust the length as if we had. For each repetition we must add 1
4653          to the length for BRAZERO and for all but the last repetition we must          to the length for BRAZERO and for all but the last repetition we must
4654          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4655          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4656            a 64-bit integer type when available, otherwise double. */
4657    
4658          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4659            {            {
4660            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4661                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4662            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4663                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4664                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4665                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4666              {              {
4667              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4294  we set the flag only if there is a liter Line 4686  we set the flag only if there is a liter
4686              {              {
4687              int offset;              int offset;
4688              *code++ = OP_BRA;              *code++ = OP_BRA;
4689              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4690              bralink = code;              bralink = code;
4691              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4692              }              }
# Line 4315  we set the flag only if there is a liter Line 4707  we set the flag only if there is a liter
4707          while (bralink != NULL)          while (bralink != NULL)
4708            {            {
4709            int oldlinkoffset;            int oldlinkoffset;
4710            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4711            uschar *bra = code - offset;            uschar *bra = code - offset;
4712            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4713            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4346  we set the flag only if there is a liter Line 4738  we set the flag only if there is a liter
4738            uschar *scode = bracode;            uschar *scode = bracode;
4739            do            do
4740              {              {
4741              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4742                {                {
4743                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4744                break;                break;
# Line 4389  we set the flag only if there is a liter Line 4781  we set the flag only if there is a liter
4781      if (possessive_quantifier)      if (possessive_quantifier)
4782        {        {
4783        int len;        int len;
4784    
4785        if (*tempcode == OP_TYPEEXACT)        if (*tempcode == OP_TYPEEXACT)
4786          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4787            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4788    
4789        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4790          {          {
4791          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
# Line 4401  we set the flag only if there is a liter Line 4793  we set the flag only if there is a liter
4793          if (utf8 && tempcode[-1] >= 0xc0)          if (utf8 && tempcode[-1] >= 0xc0)
4794            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4795  #endif  #endif
4796          }          }
4797    
4798        len = code - tempcode;        len = (int)(code - tempcode);
4799        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4800          {          {
4801          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4411  we set the flag only if there is a liter Line 4803  we set the flag only if there is a liter
4803          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4804          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4805    
4806          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4807          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4808          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4809          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4810    
4811          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4812          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4813          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4814          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4815    
4816            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4817            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4818            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4819            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4820    
4821            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4822            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4823            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4824            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4825    
4826            /* Because we are moving code along, we must ensure that any
4827            pending recursive references are updated. */
4828    
4829          default:          default:
4830            *code = OP_END;
4831            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4832          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4833          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4834          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4457  we set the flag only if there is a liter Line 4864  we set the flag only if there is a liter
4864    
4865      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4866    
4867      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4868             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4869        {        {
4870        int i, namelen;        int i, namelen;
4871          int arglen = 0;
4872        const char *vn = verbnames;        const char *vn = verbnames;
4873        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4874          const uschar *arg = NULL;
4875        previous = NULL;        previous = NULL;
4876        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4877          namelen = (int)(ptr - name);
4878    
4879        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
4880          {          {
4881          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4882          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4883              || *ptr == '_') ptr++;
4884            arglen = (int)(ptr - arg);
4885          }          }
4886    
4887        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4888          {          {
4889          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4890          goto FAILED;          goto FAILED;
4891          }          }
4892        namelen = ptr - name;  
4893          /* Scan the table of verb names */
4894    
4895        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4896          {          {
4897          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4898              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4899            {            {
4900            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT */
4901    
4902            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
4903              {              {
4904              open_capitem *oc;              open_capitem *oc;
4905              cd->had_accept = TRUE;              cd->had_accept = TRUE;
4906              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4907                {                {
4908                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
4909                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
4910                }                }
4911              }              }
4912            *code++ = verbs[i].op;  
4913            break;            /* Handle the cases with/without an argument */
4914    
4915              if (arglen == 0)
4916                {
4917                if (verbs[i].op < 0)   /* Argument is mandatory */
4918                  {
4919                  *errorcodeptr = ERR66;
4920                  goto FAILED;
4921                  }
4922                *code = verbs[i].op;
4923                if (*code++ == OP_THEN)
4924                  {
4925                  PUT(code, 0, code - bcptr->current_branch - 1);
4926                  code += LINK_SIZE;
4927                  }
4928                }
4929    
4930              else
4931                {
4932                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4933                  {
4934                  *errorcodeptr = ERR59;
4935                  goto FAILED;
4936                  }
4937                *code = verbs[i].op_arg;
4938                if (*code++ == OP_THEN_ARG)
4939                  {
4940                  PUT(code, 0, code - bcptr->current_branch - 1);
4941                  code += LINK_SIZE;
4942                  }
4943                *code++ = arglen;
4944                memcpy(code, arg, arglen);
4945                code += arglen;
4946                *code++ = 0;
4947                }
4948    
4949              break;  /* Found verb, exit loop */
4950            }            }
4951    
4952          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4953          }          }
4954        if (i < verbcount) continue;  
4955        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4956          *errorcodeptr = ERR60;          /* Verb not recognized */
4957        goto FAILED;        goto FAILED;
4958        }        }
4959    
# Line 4617  we set the flag only if there is a liter Line 5072  we set the flag only if there is a liter
5072                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5073            ptr++;            ptr++;
5074            }            }
5075          namelen = ptr - name;          namelen = (int)(ptr - name);
5076    
5077          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5078              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4654  we set the flag only if there is a liter Line 5109  we set the flag only if there is a liter
5109            }            }
5110    
5111          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5112          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
5113            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5114            except they record that the reference was originally to a name. The
5115            information is used to check duplicate names. */
5116    
5117          slot = cd->name_table;          slot = cd->name_table;
5118          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4669  we set the flag only if there is a liter Line 5127  we set the flag only if there is a liter
5127            {            {
5128            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5129            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5130              code[1+LINK_SIZE]++;
5131            }            }
5132    
5133          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5134    
5135          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5136                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5137            {            {
5138            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5139              code[1+LINK_SIZE]++;
5140            }            }
5141    
5142          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4808  we set the flag only if there is a liter Line 5268  we set the flag only if there is a liter
5268              goto FAILED;              goto FAILED;
5269              }              }
5270            *code++ = n;            *code++ = n;
5271            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5272            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5273            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5274            }            }
5275          previous = NULL;          previous = NULL;
# Line 4842  we set the flag only if there is a liter Line 5302  we set the flag only if there is a liter
5302            name = ++ptr;            name = ++ptr;
5303    
5304            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5305            namelen = ptr - name;            namelen = (int)(ptr - name);
5306    
5307            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5308    
# Line 4878  we set the flag only if there is a liter Line 5338  we set the flag only if there is a liter
5338            is because the number of names, and hence the table size, is computed            is because the number of names, and hence the table size, is computed
5339            in the pre-compile, and it affects various numbers and pointers which            in the pre-compile, and it affects various numbers and pointers which
5340            would all have to be modified, and the compiled code moved down, if            would all have to be modified, and the compiled code moved down, if
5341            duplicates with the same number were omitted from the table. This            duplicates with the same number were omitted from the table. This
5342            doesn't seem worth the hassle. However, *different* names for the            doesn't seem worth the hassle. However, *different* names for the
5343            same number are not permitted. */            same number are not permitted. */
5344    
# Line 4886  we set the flag only if there is a liter Line 5346  we set the flag only if there is a liter
5346              {              {
5347              BOOL dupname = FALSE;              BOOL dupname = FALSE;
5348              slot = cd->name_table;              slot = cd->name_table;
5349    
5350              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5351                {                {
5352                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4900  we set the flag only if there is a liter Line 5360  we set the flag only if there is a liter
5360                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5361                      goto FAILED;                      goto FAILED;
5362                      }                      }
5363                    else dupname = TRUE;                    else dupname = TRUE;
5364                    }                    }
5365                  else crc = -1;      /* Current name is a substring */                  else crc = -1;      /* Current name is a substring */
5366                  }                  }
5367    
5368                /* Make space in the table and break the loop for an earlier                /* Make space in the table and break the loop for an earlier
5369                name. For a duplicate or later name, carry on. We do this for                name. For a duplicate or later name, carry on. We do this for
5370                duplicates so that in the simple case (when ?(| is not used) they                duplicates so that in the simple case (when ?(| is not used) they
5371                are in order of their numbers. */                are in order of their numbers. */
5372    
5373                if (crc < 0)                if (crc < 0)
5374                  {                  {
5375                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5376                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5377                  break;                  break;
5378                  }                  }
5379    
5380                /* Continue the loop for a later or duplicate name */                /* Continue the loop for a later or duplicate name */
5381    
5382                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5383                }                }
5384    
5385              /* For non-duplicate names, check for a duplicate number before              /* For non-duplicate names, check for a duplicate number before
5386              adding the new name. */              adding the new name. */
5387    
5388              if (!dupname)              if (!dupname)
5389                {                {
5390                uschar *cslot = cd->name_table;                uschar *cslot = cd->name_table;
# Line 4936  we set the flag only if there is a liter Line 5396  we set the flag only if there is a liter
5396                      {                      {
5397                      *errorcodeptr = ERR65;                      *errorcodeptr = ERR65;
5398                      goto FAILED;                      goto FAILED;
5399                      }                      }
5400                    }                    }
5401                  else i--;                  else i--;
5402                  cslot += cd->name_entry_size;                  cslot += cd->name_entry_size;
5403                  }                  }
5404                }                }
5405    
5406              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5407              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
# Line 4972  we set the flag only if there is a liter Line 5432  we set the flag only if there is a liter
5432          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5433          name = ++ptr;          name = ++ptr;
5434          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5435          namelen = ptr - name;          namelen = (int)(ptr - name);
5436    
5437          /* 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
5438          reference number. */          a dummy reference number, because it was not used in the first pass.
5439            However, with the change of recursive back references to be atomic,
5440            we have to look for the number so that this state can be identified, as
5441            otherwise the incorrect length is computed. If it's not a backwards
5442            reference, the dummy number will do. */
5443    
5444          if (lengthptr != NULL)          if (lengthptr != NULL)
5445            {            {
5446              const uschar *temp;
5447    
5448            if (namelen == 0)            if (namelen == 0)
5449              {              {
5450              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 4994  we set the flag only if there is a liter Line 5460  we set the flag only if there is a liter
5460              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5461              goto FAILED;              goto FAILED;
5462              }              }
5463            recno = 0;  
5464              /* The name table does not exist in the first pass, so we cannot
5465              do a simple search as in the code below. Instead, we have to scan the
5466              pattern to find the number. It is important that we scan it only as
5467              far as we have got because the syntax of named subpatterns has not
5468              been checked for the rest of the pattern, and find_parens() assumes
5469              correct syntax. In any case, it's a waste of resources to scan
5470              further. We stop the scan at the current point by temporarily
5471              adjusting the value of cd->endpattern. */
5472    
5473              temp = cd->end_pattern;
5474              cd->end_pattern = ptr;
5475              recno = find_parens(cd, name, namelen,
5476                (options & PCRE_EXTENDED) != 0, utf8);
5477              cd->end_pattern = temp;
5478              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5479            }            }
5480    
5481          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, seek the name in the table. We check the name
# Line 5019  we set the flag only if there is a liter Line 5500  we set the flag only if there is a liter
5500              }              }
5501            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5502                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5503                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5504              {              {
5505              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5506              goto FAILED;              goto FAILED;
# Line 5122  we set the flag only if there is a liter Line 5603  we set the flag only if there is a liter
5603            if (lengthptr == NULL)            if (lengthptr == NULL)
5604              {              {
5605              *code = OP_END;              *code = OP_END;
5606              if (recno != 0)              if (recno != 0)
5607                called = _pcre_find_bracket(cd->start_code, utf8, recno);                called = _pcre_find_bracket(cd->start_code, utf8, recno);
5608    
5609              /* Forward reference */              /* Forward reference */
# Line 5130  we set the flag only if there is a liter Line 5611  we set the flag only if there is a liter
5611              if (called == NULL)              if (called == NULL)
5612                {                {
5613                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5614                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5615                  {                  {
5616                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5617                  goto FAILED;                  goto FAILED;
5618                  }                  }
5619    
5620                  /* Fudge the value of "called" so that when it is inserted as an
5621                  offset below, what it actually inserted is the reference number
5622                  of the group. */
5623    
5624                called = cd->start_code + recno;                called = cd->start_code + recno;
5625                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));
5626                }                }
5627    
5628              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
# Line 5144  we set the flag only if there is a liter Line 5630  we set the flag only if there is a liter
5630              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5631    
5632              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5633                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5634                {                {
5635                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5636                goto FAILED;                goto FAILED;
# Line 5160  we set the flag only if there is a liter Line 5646  we set the flag only if there is a liter
5646            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5647    
5648            *code = OP_RECURSE;            *code = OP_RECURSE;
5649            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
5650            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5651    
5652            *code = OP_KET;            *code = OP_KET;
# Line 5239  we set the flag only if there is a liter Line 5725  we set the flag only if there is a liter
5725              {              {
5726              cd->external_options = newoptions;              cd->external_options = newoptions;
5727              }              }
5728           else            else
5729              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5730              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5731              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5732              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
# Line 5271  we set the flag only if there is a liter Line 5752  we set the flag only if there is a liter
5752          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5753        }       /* End of (? handling */        }       /* End of (? handling */
5754    
5755      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5756      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5757      brackets. */      brackets. */
5758    
5759      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5464  we set the flag only if there is a liter Line 5945  we set the flag only if there is a liter
5945    
5946      /* ===================================================================*/      /* ===================================================================*/
5947      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
5948      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
5949      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
5950      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
5951      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
5952      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
5953        ever created. */
5954    
5955      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
5956      tempptr = ptr;      tempptr = ptr;
# Line 5576  we set the flag only if there is a liter Line 6058  we set the flag only if there is a liter
6058    
6059        if (-c >= ESC_REF)        if (-c >= ESC_REF)
6060          {          {
6061            open_capitem *oc;
6062          recno = -c - ESC_REF;          recno = -c - ESC_REF;
6063    
6064          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6065          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6066          previous = code;          previous = code;
6067          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6068          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6069          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6070          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
6071    
6072            /* Check to see if this back reference is recursive, that it, it
6073            is inside the group that it references. A flag is set so that the
6074            group can be made atomic. */
6075    
6076            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6077              {
6078              if (oc->number == recno)
6079                {
6080                oc->flag = TRUE;
6081                break;
6082                }
6083              }
6084          }          }
6085    
6086        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5614  we set the flag only if there is a liter Line 6110  we set the flag only if there is a liter
6110  #endif  #endif
6111    
6112        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
6113        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
6114          situation when PCRE_UCP is not set. When it *is* set, we substitute
6115          Unicode property tests. */
6116    
6117        else        else
6118          {          {
6119          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
6120          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
6121              {
6122              nestptr = ptr + 1;                   /* Where to resume */
6123              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6124              }
6125            else
6126    #endif
6127              {
6128              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6129              *code++ = -c;
6130              }
6131          }          }
6132        continue;        continue;
6133        }        }
# Line 5664  we set the flag only if there is a liter Line 6172  we set the flag only if there is a liter
6172    
6173      ONE_CHAR:      ONE_CHAR:
6174      previous = code;      previous = code;
6175      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6176      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6177    
6178      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 5728  return FALSE; Line 6236  return FALSE;
6236  /* 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
6237  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
6238  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.  
   
6239  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
6240  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
6241  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
# Line 5774  int branchfirstbyte, branchreqbyte; Line 6277  int branchfirstbyte, branchreqbyte;
6277  int length;  int length;
6278  int orig_bracount;  int orig_bracount;
6279  int max_bracount;  int max_bracount;
6280    int old_external_options = cd->external_options;
6281  branch_chain bc;  branch_chain bc;
6282    
6283  bc.outer = bcptr;  bc.outer = bcptr;
6284  bc.current = code;  bc.current_branch = code;
6285    
6286  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
6287    
# Line 5796  them global. It tests the value of lengt Line 6300  them global. It tests the value of lengt
6300  pre-compile phase to find out whether anything has yet been compiled or not. */  pre-compile phase to find out whether anything has yet been compiled or not. */
6301    
6302  /* 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
6303  so that we can detect them if (*ACCEPT) is encountered. */  so that we can detect them if (*ACCEPT) is encountered. This is also used to
6304    detect groups that contain recursive back references to themselves. */
6305    
6306  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6307    {    {
6308    capnumber = GET2(code, 1 + LINK_SIZE);    capnumber = GET2(code, 1 + LINK_SIZE);
6309    capitem.number = capnumber;    capitem.number = capnumber;
6310    capitem.next = cd->open_caps;    capitem.next = cd->open_caps;
6311    cd->open_caps = &capitem;    capitem.flag = FALSE;
6312    }    cd->open_caps = &capitem;
6313      }
6314    
6315  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
6316    
# Line 5821  for (;;) Line 6327  for (;;)
6327    
6328    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6329    
   /* 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;  
     }  
   
6330    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6331    
6332    if (lookbehind)    if (lookbehind)
# Line 5850  for (;;) Line 6347  for (;;)
6347      return FALSE;      return FALSE;
6348      }      }
6349    
6350      /* If the external options have changed during this branch, it means that we
6351      are at the top level, and a leading option setting has been encountered. We
6352      need to re-set the original option values to take account of this so that,
6353      during the pre-compile phase, we know to allow for a re-set at the start of
6354      subsequent branches. */
6355    
6356      if (old_external_options != cd->external_options)
6357        oldims = cd->external_options & PCRE_IMS;
6358    
6359    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6360    has fewer than the rest. */    has fewer than the rest. */
6361    
# Line 5900  for (;;) Line 6406  for (;;)
6406    
6407      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6408      put the length into the OP_REVERSE item. Temporarily mark the end of the      put the length into the OP_REVERSE item. Temporarily mark the end of the
6409      branch with OP_END. If the branch contains OP_RECURSE, the result is -3      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
6410      because there may be forward references that we can't check here. Set a      because there may be forward references that we can't check here. Set a
6411      flag to cause another lookbehind check at the end. Why not do it all at the      flag to cause another lookbehind check at the end. Why not do it all at the
6412      end? Because common, erroneous checks are picked up here and the offset of      end? Because common, erroneous checks are picked up here and the offset of
6413      the problem can be shown. */      the problem can be shown. */
6414    
6415      if (lookbehind)      if (lookbehind)
# Line 5914  for (;;) Line 6420  for (;;)
6420        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6421        if (fixed_length == -3)        if (fixed_length == -3)
6422          {          {
6423          cd->check_lookbehind = TRUE;          cd->check_lookbehind = TRUE;
6424          }          }
6425        else if (fixed_length < 0)        else if (fixed_length < 0)
6426          {          {
6427          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
# Line 5939  for (;;) Line 6445  for (;;)
6445      {      {
6446      if (lengthptr == NULL)      if (lengthptr == NULL)
6447        {        {
6448        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6449        do        do
6450          {          {
6451          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 5949  for (;;) Line 6455  for (;;)
6455          }          }
6456        while (branch_length > 0);        while (branch_length > 0);
6457        }        }
   
     /* If it was a capturing subpattern, remove it from the chain. */  
   
     if (capnumber > 0) cd->open_caps = cd->open_caps->next;  
6458    
6459      /* Fill in the ket */      /* Fill in the ket */
6460    
6461      *code = OP_KET;      *code = OP_KET;
6462      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6463      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6464    
6465      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6466        recursive back references. If so, we must wrap it in atomic brackets.
6467        In any event, remove the block from the chain. */
6468    
6469      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)      if (capnumber > 0)
6470        {        {
6471        *code++ = OP_OPT;        if (cd->open_caps->flag)
6472        *code++ = oldims;          {
6473        length += 2;          memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6474              code - start_bracket);
6475            *start_bracket = OP_ONCE;
6476            code += 1 + LINK_SIZE;
6477            PUT(start_bracket, 1, (int)(code - start_bracket));
6478            *code = OP_KET;
6479            PUT(code, 1, (int)(code - start_bracket));
6480            code += 1 + LINK_SIZE;
6481            length += 2 + 2*LINK_SIZE;
6482            }
6483          cd->open_caps = cd->open_caps->next;
6484        }        }
6485    
6486      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
# Line 6008  for (;;) Line 6522  for (;;)
6522    else    else
6523      {      {
6524      *code = OP_ALT;      *code = OP_ALT;
6525      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6526      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6527      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6528      }      }
6529    
# Line 6028  for (;;) Line 6542  for (;;)
6542  /* 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
6543  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
6544  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
6545  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
6546  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6547    
6548  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.
6549  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 6103  do { Line 6617  do {
6617    
6618     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6619    
6620     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;  
6621     code += GET(code, 1);     code += GET(code, 1);
6622     }     }
6623  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6156  do { Line 6668  do {
6668       switch (*scode)       switch (*scode)
6669         {         {
6670         case OP_CREF:         case OP_CREF:
6671           case OP_NCREF:
6672         case OP_RREF:         case OP_RREF:
6673           case OP_NRREF:
6674         case OP_DEF:         case OP_DEF:
6675         return FALSE;         return FALSE;
6676    
# Line 6203  do { Line 6717  do {
6717    
6718     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6719    
6720     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6721    
6722     /* Move on to the next alternative */     /* Move on to the next alternative */
6723    
# Line 6264  do { Line 6778  do {
6778       scode += 2;       scode += 2;
6779    
6780       case OP_CHAR:       case OP_CHAR:
6781       case OP_CHARNC:       case OP_CHARI:
6782       case OP_PLUS:       case OP_PLUS:
6783       case OP_MINPLUS:       case OP_MINPLUS:
6784       case OP_POSPLUS:       case OP_POSPLUS:
# Line 6325  int length = 1; /* For final END opcode Line 6839  int length = 1; /* For final END opcode
6839  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6840  int errorcode = 0;  int errorcode = 0;
6841  int skipatstart = 0;  int skipatstart = 0;
6842  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8;
6843  size_t size;  size_t size;
6844  uschar *code;  uschar *code;
6845  const uschar *codestart;  const uschar *codestart;
# Line 6395  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6909  while (ptr[skipatstart] == CHAR_LEFT_PAR
6909    
6910    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6911      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
6912      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
6913        { skipatstart += 6; options |= PCRE_UCP; continue; }
6914      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
6915        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
6916    
6917    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6918      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6419  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6937  while (ptr[skipatstart] == CHAR_LEFT_PAR
6937    else break;    else break;
6938    }    }
6939    
6940  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  utf8 = (options & PCRE_UTF8) != 0;
6941    
6942    /* Can't support UTF8 unless PCRE has been compiled to include the code. The
6943    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
6944    release 8.13. The only use we make of it here is to adjust the offset value to
6945    the end of the string for a short string error, for compatibility with previous
6946    versions. */
6947    
6948  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
6949  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6950       (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1, &errorcode)) >= 0)
6951    {    {
6952    errorcode = ERR44;    errorcode = ERR44;
6953    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6436  if (utf8) Line 6960  if (utf8)
6960    }    }
6961  #endif  #endif
6962    
6963    /* Can't support UCP unless PCRE has been compiled to include the code. */
6964    
6965    #ifndef SUPPORT_UCP
6966    if ((options & PCRE_UCP) != 0)
6967      {
6968      errorcode = ERR67;
6969      goto PCRE_EARLY_ERROR_RETURN;
6970      }
6971    #endif
6972    
6973  /* Check validity of \R options. */  /* Check validity of \R options. */
6974    
6975  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 6564  regex compiled on a system with 4-byte p Line 7098  regex compiled on a system with 4-byte p
7098  pointers. */  pointers. */
7099    
7100  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
7101  re->size = size;  re->size = (int)size;
7102  re->options = cd->external_options;  re->options = cd->external_options;
7103  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7104  re->dummy1 = 0;  re->dummy1 = 0;
# Line 6620  if debugging, leave the test till after Line 7154  if debugging, leave the test till after
7154    
7155  *code++ = OP_END;  *code++ = OP_END;
7156    
7157  #ifndef DEBUG  #ifndef PCRE_DEBUG
7158  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
7159  #endif  #endif
7160    
# Line 6635  while (errorcode == 0 && cd->hwm > cwork Line 7169  while (errorcode == 0 && cd->hwm > cwork
7169    recno = GET(codestart, offset);    recno = GET(codestart, offset);
7170    groupptr = _pcre_find_bracket(codestart, utf8, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
7171    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
7172      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7173    }    }
7174    
7175  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
# Line 6643  subpattern. */ Line 7177  subpattern. */
7177    
7178  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
7179    
7180  /* If there were any lookbehind assertions that contained OP_RECURSE  /* If there were any lookbehind assertions that contained OP_RECURSE
7181  (recursions or subroutine calls), a flag is set for them to be checked here,  (recursions or subroutine calls), a flag is set for them to be checked here,
7182  because they may contain forward references. Actual recursions can't be fixed  because they may contain forward references. Actual recursions can't be fixed
7183  length, but subroutine calls can. It is done like this so that those without  length, but subroutine calls can. It is done like this so that those without
# Line 6654  length, and set their lengths. */ Line 7188  length, and set their lengths. */
7188  if (cd->check_lookbehind)  if (cd->check_lookbehind)
7189    {    {
7190    uschar *cc = (uschar *)codestart;    uschar *cc = (uschar *)codestart;
7191    
7192    /* Loop, searching for OP_REVERSE items, and process those that do not have    /* Loop, searching for OP_REVERSE items, and process those that do not have
7193    their length set. (Actually, it will also re-process any that have a length    their length set. (Actually, it will also re-process any that have a length
7194    of zero, but that is a pathological case, and it does no harm.) When we find    of zero, but that is a pathological case, and it does no harm.) When we find
7195    one, we temporarily terminate the branch it is in while we scan it. */    one, we temporarily terminate the branch it is in while we scan it. */
7196    
7197    for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);    for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
7198         cc != NULL;         cc != NULL;
7199         cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))         cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
7200      {      {
7201      if (GET(cc, 1) == 0)      if (GET(cc, 1) == 0)
7202        {        {
7203        int fixed_length;        int fixed_length;
7204        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7205        int end_op = *be;        int end_op = *be;
7206        *be = OP_END;        *be = OP_END;
7207        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
7208        *be = end_op;        *be = end_op;
# Line 6676  if (cd->check_lookbehind) Line 7210  if (cd->check_lookbehind)
7210        if (fixed_length < 0)        if (fixed_length < 0)
7211          {          {
7212          errorcode = (fixed_length == -2)? ERR36 : ERR25;          errorcode = (fixed_length == -2)? ERR36 : ERR25;
7213          break;          break;
7214          }          }
7215        PUT(cc, 1, fixed_length);        PUT(cc, 1, fixed_length);
7216        }        }
7217      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
7218      }      }
7219    }    }
7220    
7221  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
7222    
# Line 6690  if (errorcode != 0) Line 7224  if (errorcode != 0)
7224    {    {
7225    (pcre_free)(re);    (pcre_free)(re);
7226    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
7227    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = (int)(ptr - (const uschar *)pattern);
7228    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
7229    *errorptr = find_error_text(errorcode);    *errorptr = find_error_text(errorcode);
7230    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
# Line 6744  if (reqbyte >= 0 && Line 7278  if (reqbyte >= 0 &&
7278  /* Print out the compiled data if debugging is enabled. This is never the  /* Print out the compiled data if debugging is enabled. This is never the
7279  case when building a production library. */  case when building a production library. */
7280    
7281  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
7282  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
7283    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
7284    
# Line 6782  if (code - codestart > length) Line 7315  if (code - codestart > length)
7315    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
7316    return NULL;    return NULL;
7317    }    }
7318  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
7319    
7320  return (pcre *)re;  return (pcre *)re;
7321  }  }

Legend:
Removed from v.457  
changed lines
  Added in v.602

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12