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

Diff of /code/trunk/pcre_compile.c

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

revision 459 by ph10, Sun Oct 4 09:21:39 2009 UTC revision 533 by ph10, Wed Jun 2 19:02:41 2010 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) 1997-2010 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      ;
412    
413  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
414  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 566  static const char *
566  find_error_text(int n)  find_error_text(int n)
567  {  {
568  const char *s = error_texts;  const char *s = error_texts;
569  for (; n > 0; n--) while (*s++ != 0) {};  for (; n > 0; n--)
570      {
571      while (*s++ != 0) {};
572      if (*s == 0) return "Error text not found (please report)";
573      }
574  return s;  return s;
575  }  }
576    
# Line 573  else Line 641  else
641    
642      case CHAR_l:      case CHAR_l:
643      case CHAR_L:      case CHAR_L:
     case CHAR_N:  
644      case CHAR_u:      case CHAR_u:
645      case CHAR_U:      case CHAR_U:
646      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
# Line 811  else Line 878  else
878      }      }
879    }    }
880    
881    /* Perl supports \N{name} for character names, as well as plain \N for "not
882    newline". PCRE does not support \N{name}. */
883    
884    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885      *errorcodeptr = ERR37;
886    
887    /* If PCRE_UCP is set, we change the values for \d etc. */
888    
889    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
890      c -= (ESC_DU - ESC_D);
891    
892    /* Set the pointer to the final character before returning. */
893    
894  *ptrptr = ptr;  *ptrptr = ptr;
895  return c;  return c;
896  }  }
# Line 1102  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1182  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1182        if (name != NULL && lorn == ptr - thisname &&        if (name != NULL && lorn == ptr - thisname &&
1183            strncmp((const char *)name, (const char *)thisname, lorn) == 0)            strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1184          return *count;          return *count;
1185        term++;        term++;
1186        }        }
1187      }      }
1188    }    }
# Line 1148  for (; *ptr != 0; ptr++) Line 1228  for (; *ptr != 0; ptr++)
1228            break;            break;
1229          }          }
1230        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1231          {          {
1232          negate_class = TRUE;          negate_class = TRUE;
1233          ptr++;          ptr++;
1234          }          }
1235        else break;        else break;
1236        }        }
1237    
# Line 1340  for (;;) Line 1420  for (;;)
1420    
1421  /* 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,
1422  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.
1423  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
1424  temporarily terminated with OP_END when this function is called.  temporarily terminated with OP_END when this function is called.
1425    
1426  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
1427  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.
1428  However, we cannot do this when the assertion contains subroutine calls,  However, we cannot do this when the assertion contains subroutine calls,
1429  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
1430  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.
1431    
1432  Arguments:  Arguments:
1433    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1434    options  the compiling options    options  the compiling options
1435    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1436    cd       the "compile data" structure    cd       the "compile data" structure
1437    
1438  Returns:   the fixed length,  Returns:   the fixed length,
1439               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1440               or -2 if \C was encountered               or -2 if \C was encountered
1441               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 1405  for (;;) Line 1485  for (;;)
1485      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1486      branchlength = 0;      branchlength = 0;
1487      break;      break;
1488    
1489      /* A true recursion implies not fixed length, but a subroutine call may      /* A true recursion implies not fixed length, but a subroutine call may
1490      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
1491      it until the end of the pattern, so return -3. */      it until the end of the pattern, so return -3. */
1492    
1493      case OP_RECURSE:      case OP_RECURSE:
1494      if (!atend) return -3;      if (!atend) return -3;
1495      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1496      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1497      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1498      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, options, atend, cd);
1499      if (d < 0) return d;      if (d < 0) return d;
1500      branchlength += d;      branchlength += d;
1501      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1502      break;      break;
1503    
1504      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1505    
# Line 1442  for (;;) Line 1522  for (;;)
1522      case OP_CALLOUT:      case OP_CALLOUT:
1523      case OP_SOD:      case OP_SOD:
1524      case OP_SOM:      case OP_SOM:
1525        case OP_SET_SOM:
1526      case OP_EOD:      case OP_EOD:
1527      case OP_EODN:      case OP_EODN:
1528      case OP_CIRC:      case OP_CIRC:
# Line 1459  for (;;) Line 1540  for (;;)
1540      branchlength++;      branchlength++;
1541      cc += 2;      cc += 2;
1542  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1543      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1544        cc += _pcre_utf8_table4[cc[-1] & 0x3f];        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
1545  #endif  #endif
1546      break;      break;
# Line 1471  for (;;) Line 1552  for (;;)
1552      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1553      cc += 4;      cc += 4;
1554  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1555      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1556        cc += _pcre_utf8_table4[cc[-1] & 0x3f];        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
1557  #endif  #endif
1558      break;      break;
# Line 1556  for (;;) Line 1637  for (;;)
1637    
1638  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1639  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
1640  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
1641  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
1642  length.  length.
1643    
1644  Arguments:  Arguments:
# Line 1581  for (;;) Line 1662  for (;;)
1662    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. */
1663    
1664    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1665    
1666    /* Handle recursion */    /* Handle recursion */
1667    
1668    else if (c == OP_REVERSE)    else if (c == OP_REVERSE)
1669      {      {
1670      if (number < 0) return (uschar *)code;      if (number < 0) return (uschar *)code;
1671      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1672      }      }
1673    
# Line 1601  for (;;) Line 1682  for (;;)
1682    
1683    /* 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
1684    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
1685    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1686      must add in its length. */
1687    
1688    else    else
1689      {      {
# Line 1625  for (;;) Line 1707  for (;;)
1707        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1708        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1709        break;        break;
1710    
1711          case OP_MARK:
1712          case OP_PRUNE_ARG:
1713          case OP_SKIP_ARG:
1714          case OP_THEN_ARG:
1715          code += code[1];
1716          break;
1717        }        }
1718    
1719      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1696  for (;;) Line 1785  for (;;)
1785    
1786    /* 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
1787    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
1788    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1789      must add in its length. */
1790    
1791    else    else
1792      {      {
# Line 1720  for (;;) Line 1810  for (;;)
1810        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1811        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1812        break;        break;
1813    
1814          case OP_MARK:
1815          case OP_PRUNE_ARG:
1816          case OP_SKIP_ARG:
1817          case OP_THEN_ARG:
1818          code += code[1];
1819          break;
1820        }        }
1821    
1822      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1776  Arguments: Line 1873  Arguments:
1873    code        points to start of search    code        points to start of search
1874    endcode     points to where to stop    endcode     points to where to stop
1875    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1876      cd          contains pointers to tables etc.
1877    
1878  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1879  */  */
1880    
1881  static BOOL  static BOOL
1882  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1883      compile_data *cd)
1884  {  {
1885  register int c;  register int c;
1886  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 1812  for (code = first_significant_code(code Line 1911  for (code = first_significant_code(code
1911      continue;      continue;
1912      }      }
1913    
1914      /* For a recursion/subroutine call, if its end has been reached, which
1915      implies a subroutine call, we can scan it. */
1916    
1917      if (c == OP_RECURSE)
1918        {
1919        BOOL empty_branch = FALSE;
1920        const uschar *scode = cd->start_code + GET(code, 1);
1921        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1922        do
1923          {
1924          if (could_be_empty_branch(scode, endcode, utf8, cd))
1925            {
1926            empty_branch = TRUE;
1927            break;
1928            }
1929          scode += GET(scode, 1);
1930          }
1931        while (*scode == OP_ALT);
1932        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1933        continue;
1934        }
1935    
1936    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1937    
1938    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 1830  for (code = first_significant_code(code Line 1951  for (code = first_significant_code(code
1951        empty_branch = FALSE;        empty_branch = FALSE;
1952        do        do
1953          {          {
1954          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1955            empty_branch = TRUE;            empty_branch = TRUE;
1956          code += GET(code, 1);          code += GET(code, 1);
1957          }          }
# Line 1957  for (code = first_significant_code(code Line 2078  for (code = first_significant_code(code
2078      case OP_POSQUERY:      case OP_POSQUERY:
2079      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2080      break;      break;
2081    
2082      case OP_UPTO:      case OP_UPTO:
2083      case OP_MINUPTO:      case OP_MINUPTO:
2084      case OP_POSUPTO:      case OP_POSUPTO:
2085      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2086      break;      break;
2087  #endif  #endif
2088    
2089        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2090        string. */
2091    
2092        case OP_MARK:
2093        case OP_PRUNE_ARG:
2094        case OP_SKIP_ARG:
2095        case OP_THEN_ARG:
2096        code += code[1];
2097        break;
2098    
2099        /* None of the remaining opcodes are required to match a character. */
2100    
2101        default:
2102        break;
2103      }      }
2104    }    }
2105    
# Line 1986  Arguments: Line 2122  Arguments:
2122    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2123    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2124    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2125      cd          pointers to tables etc
2126    
2127  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2128  */  */
2129    
2130  static BOOL  static BOOL
2131  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2132    BOOL utf8)    BOOL utf8, compile_data *cd)
2133  {  {
2134  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2135    {    {
2136    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2137        return FALSE;
2138    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2139    }    }
2140  return TRUE;  return TRUE;
# Line 2178  auto_callout(uschar *code, const uschar Line 2316  auto_callout(uschar *code, const uschar
2316  {  {
2317  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2318  *code++ = 255;  *code++ = 255;
2319  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2320  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2321  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2322  }  }
2323    
# Line 2204  Returns: nothing Line 2342  Returns: nothing
2342  static void  static void
2343  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2344  {  {
2345  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2346  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2347  }  }
2348    
# Line 2254  for (++c; c <= d; c++) Line 2392  for (++c; c <= d; c++)
2392    
2393  return TRUE;  return TRUE;
2394  }  }
2395    
2396    
2397    
2398    /*************************************************
2399    *        Check a character and a property        *
2400    *************************************************/
2401    
2402    /* This function is called by check_auto_possessive() when a property item
2403    is adjacent to a fixed character.
2404    
2405    Arguments:
2406      c            the character
2407      ptype        the property type
2408      pdata        the data for the type
2409      negated      TRUE if it's a negated property (\P or \p{^)
2410    
2411    Returns:       TRUE if auto-possessifying is OK
2412    */
2413    
2414    static BOOL
2415    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2416    {
2417    const ucd_record *prop = GET_UCD(c);
2418    switch(ptype)
2419      {
2420      case PT_LAMP:
2421      return (prop->chartype == ucp_Lu ||
2422              prop->chartype == ucp_Ll ||
2423              prop->chartype == ucp_Lt) == negated;
2424    
2425      case PT_GC:
2426      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2427    
2428      case PT_PC:
2429      return (pdata == prop->chartype) == negated;
2430    
2431      case PT_SC:
2432      return (pdata == prop->script) == negated;
2433    
2434      /* These are specials */
2435    
2436      case PT_ALNUM:
2437      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2438              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2439    
2440      case PT_SPACE:    /* Perl space */
2441      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2442              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2443              == negated;
2444    
2445      case PT_PXSPACE:  /* POSIX space */
2446      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2447              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2448              c == CHAR_FF || c == CHAR_CR)
2449              == negated;
2450    
2451      case PT_WORD:
2452      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2453              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2454              c == CHAR_UNDERSCORE) == negated;
2455      }
2456    return FALSE;
2457    }
2458  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2459    
2460    
# Line 2267  whether the next thing could possibly ma Line 2468  whether the next thing could possibly ma
2468  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2469    
2470  Arguments:  Arguments:
2471    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2472    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2473    ptr           next character in pattern    ptr           next character in pattern
2474    options       options bits    options       options bits
2475    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2279  Returns: TRUE if possessifying is Line 2478  Returns: TRUE if possessifying is
2478  */  */
2479    
2480  static BOOL  static BOOL
2481  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2482    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2483  {  {
2484  int next;  int c, next;
2485    int op_code = *previous++;
2486    
2487  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2488    
# Line 2343  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2543  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2543    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)
2544      return FALSE;      return FALSE;
2545    
2546  /* 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
2547  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. */  
2548    
2549  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2550    {    {
2551    case OP_CHAR:    case OP_CHAR:
2552  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2553    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2554  #else  #else
2555    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2556  #endif  #endif
2557    return item != next;    return c != next;
2558    
2559    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARNC (caseless character) we must check the other case. If we have
2560    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
2561    high-valued characters. */    high-valued characters. */
2562    
2563    case OP_CHARNC:    case OP_CHARNC:
2564  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2565    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2566  #endif  #else
2567    if (item == next) return FALSE;    c = *previous;
2568    #endif
2569      if (c == next) return FALSE;
2570  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2571    if (utf8)    if (utf8)
2572      {      {
# Line 2380  if (next >= 0) switch(op_code) Line 2577  if (next >= 0) switch(op_code)
2577  #else  #else
2578      othercase = NOTACHAR;      othercase = NOTACHAR;
2579  #endif  #endif
2580      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2581      }      }
2582    else    else
2583  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2584    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2585    
2586    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2587    
2588    case OP_NOT:    case OP_NOT:
2589    if (item == next) return TRUE;    if ((c = *previous) == next) return TRUE;
2590    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2591  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2592    if (utf8)    if (utf8)
# Line 2401  if (next >= 0) switch(op_code) Line 2598  if (next >= 0) switch(op_code)
2598  #else  #else
2599      othercase = NOTACHAR;      othercase = NOTACHAR;
2600  #endif  #endif
2601      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2602      }      }
2603    else    else
2604  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2605    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2606    
2607      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2608      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2609    
2610    case OP_DIGIT:    case OP_DIGIT:
2611    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2448  if (next >= 0) switch(op_code) Line 2648  if (next >= 0) switch(op_code)
2648      case 0x202f:      case 0x202f:
2649      case 0x205f:      case 0x205f:
2650      case 0x3000:      case 0x3000:
2651      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2652      default:      default:
2653      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2654      }      }
2655    
2656      case OP_ANYNL:
2657    case OP_VSPACE:    case OP_VSPACE:
2658    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2659    switch(next)    switch(next)
# Line 2464  if (next >= 0) switch(op_code) Line 2665  if (next >= 0) switch(op_code)
2665      case 0x85:      case 0x85:
2666      case 0x2028:      case 0x2028:
2667      case 0x2029:      case 0x2029:
2668      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2669      default:      default:
2670      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2671      }      }
2672    
2673    #ifdef SUPPORT_UCP
2674      case OP_PROP:
2675      return check_char_prop(next, previous[0], previous[1], FALSE);
2676    
2677      case OP_NOTPROP:
2678      return check_char_prop(next, previous[0], previous[1], TRUE);
2679    #endif
2680    
2681    default:    default:
2682    return FALSE;    return FALSE;
2683    }    }
2684    
2685    
2686  /* 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
2687    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2688    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2689    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2690    replaced by OP_PROP codes when PCRE_UCP is set. */
2691    
2692  switch(op_code)  switch(op_code)
2693    {    {
2694    case OP_CHAR:    case OP_CHAR:
2695    case OP_CHARNC:    case OP_CHARNC:
2696  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2697    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2698  #endif  #else
2699      c = *previous;
2700    #endif
2701    switch(-next)    switch(-next)
2702      {      {
2703      case ESC_d:      case ESC_d:
2704      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2705    
2706      case ESC_D:      case ESC_D:
2707      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2708    
2709      case ESC_s:      case ESC_s:
2710      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2711    
2712      case ESC_S:      case ESC_S:
2713      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2714    
2715      case ESC_w:      case ESC_w:
2716      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2717    
2718      case ESC_W:      case ESC_W:
2719      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2720    
2721      case ESC_h:      case ESC_h:
2722      case ESC_H:      case ESC_H:
2723      switch(item)      switch(c)
2724        {        {
2725        case 0x09:        case 0x09:
2726        case 0x20:        case 0x20:
# Line 2533  switch(op_code) Line 2748  switch(op_code)
2748    
2749      case ESC_v:      case ESC_v:
2750      case ESC_V:      case ESC_V:
2751      switch(item)      switch(c)
2752        {        {
2753        case 0x0a:        case 0x0a:
2754        case 0x0b:        case 0x0b:
# Line 2546  switch(op_code) Line 2761  switch(op_code)
2761        default:        default:
2762        return -next == ESC_v;        return -next == ESC_v;
2763        }        }
2764    
2765        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2766        their substitutions and process them. The result will always be either
2767        -ESC_p or -ESC_P. Then fall through to process those values. */
2768    
2769    #ifdef SUPPORT_UCP
2770        case ESC_du:
2771        case ESC_DU:
2772        case ESC_wu:
2773        case ESC_WU:
2774        case ESC_su:
2775        case ESC_SU:
2776          {
2777          int temperrorcode = 0;
2778          ptr = substitutes[-next - ESC_DU];
2779          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2780          if (temperrorcode != 0) return FALSE;
2781          ptr++;    /* For compatibility */
2782          }
2783        /* Fall through */
2784    
2785        case ESC_p:
2786        case ESC_P:
2787          {
2788          int ptype, pdata, errorcodeptr;
2789          BOOL negated;
2790    
2791          ptr--;      /* Make ptr point at the p or P */
2792          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2793          if (ptype < 0) return FALSE;
2794          ptr++;      /* Point past the final curly ket */
2795    
2796          /* If the property item is optional, we have to give up. (When generated
2797          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2798          to the original \d etc. At this point, ptr will point to a zero byte. */
2799    
2800          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2801            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2802              return FALSE;
2803    
2804          /* Do the property check. */
2805    
2806          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2807          }
2808    #endif
2809    
2810      default:      default:
2811      return FALSE;      return FALSE;
2812      }      }
2813    
2814      /* In principle, support for Unicode properties should be integrated here as
2815      well. It means re-organizing the above code so as to get hold of the property
2816      values before switching on the op-code. However, I wonder how many patterns
2817      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2818      these op-codes are never generated.) */
2819    
2820    case OP_DIGIT:    case OP_DIGIT:
2821    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2822           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2823    
2824    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2825    return next == -ESC_d;    return next == -ESC_d;
2826    
2827    case OP_WHITESPACE:    case OP_WHITESPACE:
2828    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2829    
2830    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2831    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2832    
2833    case OP_HSPACE:    case OP_HSPACE:
2834    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2835             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2836    
2837    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2838    return next == -ESC_h;    return next == -ESC_h;
2839    
2840    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2841      case OP_ANYNL:
2842    case OP_VSPACE:    case OP_VSPACE:
2843    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2844    
2845    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2846    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2847    
2848    case OP_WORDCHAR:    case OP_WORDCHAR:
2849    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2850             next == -ESC_v || next == -ESC_R;
2851    
2852    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2853    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2642  BOOL inescq = FALSE; Line 2911  BOOL inescq = FALSE;
2911  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2912  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2913  const uschar *tempptr;  const uschar *tempptr;
2914    const uschar *nestptr = NULL;
2915  uschar *previous = NULL;  uschar *previous = NULL;
2916  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2917  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2658  BOOL utf8 = FALSE; Line 2928  BOOL utf8 = FALSE;
2928  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2929  #endif  #endif
2930    
2931  #ifdef DEBUG  #ifdef PCRE_DEBUG
2932  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2933  #endif  #endif
2934    
# Line 2712  for (;; ptr++) Line 2982  for (;; ptr++)
2982    
2983    c = *ptr;    c = *ptr;
2984    
2985      /* If we are at the end of a nested substitution, revert to the outer level
2986      string. Nesting only happens one level deep. */
2987    
2988      if (c == 0 && nestptr != NULL)
2989        {
2990        ptr = nestptr;
2991        nestptr = NULL;
2992        c = *ptr;
2993        }
2994    
2995    /* 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
2996    previous cycle of this loop. */    previous cycle of this loop. */
2997    
2998    if (lengthptr != NULL)    if (lengthptr != NULL)
2999      {      {
3000  #ifdef DEBUG  #ifdef PCRE_DEBUG
3001      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3002  #endif  #endif
3003      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3004        {        {
3005        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3006        goto FAILED;        goto FAILED;
# Line 2742  for (;; ptr++) Line 3022  for (;; ptr++)
3022        goto FAILED;        goto FAILED;
3023        }        }
3024    
3025      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3026      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));
3027    
3028      /* 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 2769  for (;; ptr++) Line 3049  for (;; ptr++)
3049    /* 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
3050    reference list. */    reference list. */
3051    
3052    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3053      {      {
3054      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3055      goto FAILED;      goto FAILED;
# Line 2860  for (;; ptr++) Line 3140  for (;; ptr++)
3140          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3141          goto FAILED;          goto FAILED;
3142          }          }
3143        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3144        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3145        }        }
3146      return TRUE;      return TRUE;
# Line 3009  for (;; ptr++) Line 3289  for (;; ptr++)
3289          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3290          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3291          }          }
3292    
3293        /* In the pre-compile phase, accumulate the length of any UTF-8 extra        /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3294        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
3295        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion UTF-8 characters no longer overwrite the work space
# Line 3065  for (;; ptr++) Line 3345  for (;; ptr++)
3345            ptr++;            ptr++;
3346            }            }
3347    
3348          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3349          if (posix_class < 0)          if (posix_class < 0)
3350            {            {
3351            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3078  for (;; ptr++) Line 3358  for (;; ptr++)
3358    
3359          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3360            posix_class = 0;            posix_class = 0;
3361    
3362          /* 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
3363          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3364          subtract bits that may be in the main map already. At the end we or the  
3365          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3366            if ((options & PCRE_UCP) != 0)
3367              {
3368              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3369              if (posix_substitutes[pc] != NULL)
3370                {
3371                nestptr = tempptr + 1;
3372                ptr = posix_substitutes[pc] - 1;
3373                continue;
3374                }
3375              }
3376    #endif
3377            /* In the non-UCP case, we build the bit map for the POSIX class in a
3378            chunk of local store because we may be adding and subtracting from it,
3379            and we don't want to subtract bits that may be in the main map already.
3380            At the end we or the result into the bit map that is being built. */
3381    
3382          posix_class *= 3;          posix_class *= 3;
3383    
# Line 3126  for (;; ptr++) Line 3421  for (;; ptr++)
3421    
3422        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3423        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
3424        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
3425        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
3426        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
3427        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3428          PCRE_EXTRA is set. */
3429    
3430        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3431          {          {
3432          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3433          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3434    
3435          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 */  
3436          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3437            {            {
3438            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3155  for (;; ptr++) Line 3449  for (;; ptr++)
3449            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3450            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3451    
3452            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3453              {              {
3454    #ifdef SUPPORT_UCP
3455                case ESC_du:     /* These are the values given for \d etc */
3456                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3457                case ESC_wu:     /* escape sequence with an appropriate \p */
3458                case ESC_WU:     /* or \P to test Unicode properties instead */
3459                case ESC_su:     /* of the default ASCII testing. */
3460                case ESC_SU:
3461                nestptr = ptr;
3462                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3463                class_charcount -= 2;                /* Undo! */
3464                continue;
3465    #endif
3466              case ESC_d:              case ESC_d:
3467              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3468              continue;              continue;
# Line 3188  for (;; ptr++) Line 3492  for (;; ptr++)
3492              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3493              continue;              continue;
3494    
3495              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)  
             {  
3496              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3497              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3498              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3225  for (;; ptr++) Line 3516  for (;; ptr++)
3516                }                }
3517  #endif  #endif
3518              continue;              continue;
             }  
3519    
3520            if (-c == ESC_H)              case ESC_H:
             {  
3521              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3522                {                {
3523                int x = 0xff;                int x = 0xff;
# Line 3270  for (;; ptr++) Line 3559  for (;; ptr++)
3559                }                }
3560  #endif  #endif
3561              continue;              continue;
             }  
3562    
3563            if (-c == ESC_v)              case ESC_v:
             {  
3564              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3565              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3566              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3289  for (;; ptr++) Line 3576  for (;; ptr++)
3576                }                }
3577  #endif  #endif
3578              continue;              continue;
             }  
3579    
3580            if (-c == ESC_V)              case ESC_V:
             {  
3581              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3582                {                {
3583                int x = 0xff;                int x = 0xff;
# Line 3322  for (;; ptr++) Line 3607  for (;; ptr++)
3607                }                }
3608  #endif  #endif
3609              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3610    
3611  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3612            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3613              {              case ESC_P:
3614              BOOL negated;                {
3615              int pdata;                BOOL negated;
3616              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3617              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3618              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3619              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3620                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3621              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3622              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3623              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3624              continue;                class_charcount -= 2;   /* Not a < 256 character */
3625              }                continue;
3626                  }
3627  #endif  #endif
3628            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3629            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3630            treated as literals. */              treated as literals. */
3631    
3632            if ((options & PCRE_EXTRA) != 0)              default:
3633              {              if ((options & PCRE_EXTRA) != 0)
3634              *errorcodeptr = ERR7;                {
3635              goto FAILED;                *errorcodeptr = ERR7;
3636                  goto FAILED;
3637                  }
3638                class_charcount -= 2;  /* Undo the default count from above */
3639                c = *ptr;              /* Get the final character and fall through */
3640                break;
3641              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3642            }            }
3643    
3644          /* 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 3423  for (;; ptr++) Line 3708  for (;; ptr++)
3708            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3709            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3710    
3711            /* \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 */  
3712    
3713            if (d < 0)            if (d < 0)
3714              {              {
3715              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  
3716                {                {
3717                ptr = oldptr;                ptr = oldptr;
3718                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3596  for (;; ptr++) Line 3878  for (;; ptr++)
3878          }          }
3879        }        }
3880    
3881      /* 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.
3882        If we are at the end of an internal nested string, revert to the outer
3883        string. */
3884    
3885        while (((c = *(++ptr)) != 0 ||
3886               (nestptr != NULL &&
3887                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3888               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3889    
3890      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
3891    
3892      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3893        {        {
3894        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3895        goto FAILED;        goto FAILED;
3896        }        }
3897    
   
 /* 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  
   
   
3898      /* 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
3899      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
3900      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 3641  we set the flag only if there is a liter Line 3911  we set the flag only if there is a liter
3911      can cause firstbyte to be set. Otherwise, there can be no first char if      can cause firstbyte to be set. Otherwise, there can be no first char if
3912      this item is first, whatever repeat count may follow. In the case of      this item is first, whatever repeat count may follow. In the case of
3913      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3914    
3915  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3916      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
3917        (!utf8 || !negate_class || class_lastchar < 128))        (!utf8 || !negate_class || class_lastchar < 128))
# Line 3688  we set the flag only if there is a liter Line 3958  we set the flag only if there is a liter
3958    
3959      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3960      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
3961      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
3962      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
3963      (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
3964      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
3965        actual compiled code. */
3966    
3967  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3968      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
3969        {        {
3970        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3971        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3720  we set the flag only if there is a liter Line 3991  we set the flag only if there is a liter
3991        }        }
3992  #endif  #endif
3993    
3994      /* 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
3995      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
3996      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
3997      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
3998        negating it if necessary. */
3999    
4000      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4001      if (negate_class)      if (negate_class)
# Line 3847  we set the flag only if there is a liter Line 4119  we set the flag only if there is a liter
4119    
4120        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4121            repeat_max < 0 &&            repeat_max < 0 &&
4122            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4123          {          {
4124          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4125          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3869  we set the flag only if there is a liter Line 4140  we set the flag only if there is a liter
4140        c = previous[1];        c = previous[1];
4141        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4142            repeat_max < 0 &&            repeat_max < 0 &&
4143            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4144          {          {
4145          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4146          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3893  we set the flag only if there is a liter Line 4164  we set the flag only if there is a liter
4164    
4165        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4166            repeat_max < 0 &&            repeat_max < 0 &&
4167            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4168          {          {
4169          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4170          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3915  we set the flag only if there is a liter Line 4186  we set the flag only if there is a liter
4186    
4187        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4188    
4189        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4190        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4191        removed: */        removed: */
4192    
4193        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4194        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4195    
4196        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4197        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4198    
4199        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4200    
# Line 4070  we set the flag only if there is a liter Line 4341  we set the flag only if there is a liter
4341          goto END_REPEAT;          goto END_REPEAT;
4342          }          }
4343    
4344        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4345        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4346        removed: */        removed: */
4347    
# Line 4078  we set the flag only if there is a liter Line 4349  we set the flag only if there is a liter
4349        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4350    
4351        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4352        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4353    
4354        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4355          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 4103  we set the flag only if there is a liter Line 4374  we set the flag only if there is a liter
4374        {        {
4375        register int i;        register int i;
4376        int ketoffset = 0;        int ketoffset = 0;
4377        int len = code - previous;        int len = (int)(code - previous);
4378        uschar *bralink = NULL;        uschar *bralink = NULL;
4379    
4380        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 4124  we set the flag only if there is a liter Line 4395  we set the flag only if there is a liter
4395          {          {
4396          register uschar *ket = previous;          register uschar *ket = previous;
4397          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4398          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4399          }          }
4400    
4401        /* 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 4192  we set the flag only if there is a liter Line 4463  we set the flag only if there is a liter
4463            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4464            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4465    
4466            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4467            bralink = previous;            bralink = previous;
4468            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4469            }            }
# Line 4213  we set the flag only if there is a liter Line 4484  we set the flag only if there is a liter
4484            {            {
4485            /* 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
4486            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
4487            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4488              integer type when available, otherwise double. */
4489    
4490            if (lengthptr != NULL)            if (lengthptr != NULL)
4491              {              {
4492              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4493              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4494                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4495                        (INT64_OR_DOUBLE)INT_MAX ||
4496                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4497                {                {
4498                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4265  we set the flag only if there is a liter Line 4538  we set the flag only if there is a liter
4538          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
4539          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
4540          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
4541          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4542            a 64-bit integer type when available, otherwise double. */
4543    
4544          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4545            {            {
4546            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4547                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4548            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4549                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4550                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4551                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4552              {              {
4553              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4298  we set the flag only if there is a liter Line 4572  we set the flag only if there is a liter
4572              {              {
4573              int offset;              int offset;
4574              *code++ = OP_BRA;              *code++ = OP_BRA;
4575              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4576              bralink = code;              bralink = code;
4577              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4578              }              }
# Line 4319  we set the flag only if there is a liter Line 4593  we set the flag only if there is a liter
4593          while (bralink != NULL)          while (bralink != NULL)
4594            {            {
4595            int oldlinkoffset;            int oldlinkoffset;
4596            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4597            uschar *bra = code - offset;            uschar *bra = code - offset;
4598            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4599            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4350  we set the flag only if there is a liter Line 4624  we set the flag only if there is a liter
4624            uschar *scode = bracode;            uschar *scode = bracode;
4625            do            do
4626              {              {
4627              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4628                {                {
4629                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4630                break;                break;
# Line 4393  we set the flag only if there is a liter Line 4667  we set the flag only if there is a liter
4667      if (possessive_quantifier)      if (possessive_quantifier)
4668        {        {
4669        int len;        int len;
4670    
4671        if (*tempcode == OP_TYPEEXACT)        if (*tempcode == OP_TYPEEXACT)
4672          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4673            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4674    
4675        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4676          {          {
4677          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
# Line 4405  we set the flag only if there is a liter Line 4679  we set the flag only if there is a liter
4679          if (utf8 && tempcode[-1] >= 0xc0)          if (utf8 && tempcode[-1] >= 0xc0)
4680            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4681  #endif  #endif
4682          }          }
4683    
4684        len = code - tempcode;        len = (int)(code - tempcode);
4685        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4686          {          {
4687          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4425  we set the flag only if there is a liter Line 4699  we set the flag only if there is a liter
4699          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4700          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4701    
4702            /* Because we are moving code along, we must ensure that any
4703            pending recursive references are updated. */
4704    
4705          default:          default:
4706            *code = OP_END;
4707            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4708          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4709          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4710          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4461  we set the flag only if there is a liter Line 4740  we set the flag only if there is a liter
4740    
4741      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4742    
4743      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4744             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4745        {        {
4746        int i, namelen;        int i, namelen;
4747          int arglen = 0;
4748        const char *vn = verbnames;        const char *vn = verbnames;
4749        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4750          const uschar *arg = NULL;
4751        previous = NULL;        previous = NULL;
4752        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4753          namelen = (int)(ptr - name);
4754    
4755        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
4756          {          {
4757          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4758          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4759              || *ptr == '_') ptr++;
4760            arglen = (int)(ptr - arg);
4761          }          }
4762    
4763        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4764          {          {
4765          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4766          goto FAILED;          goto FAILED;
4767          }          }
4768        namelen = ptr - name;  
4769          /* Scan the table of verb names */
4770    
4771        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4772          {          {
4773          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4774              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4775            {            {
4776            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT */
4777    
4778            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
4779              {              {
4780              open_capitem *oc;              open_capitem *oc;
4781              cd->had_accept = TRUE;              cd->had_accept = TRUE;
4782              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4783                {                {
4784                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
4785                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
4786                }                }
4787              }              }
4788            *code++ = verbs[i].op;  
4789            break;            /* Handle the cases with/without an argument */
4790    
4791              if (arglen == 0)
4792                {
4793                if (verbs[i].op < 0)   /* Argument is mandatory */
4794                  {
4795                  *errorcodeptr = ERR66;
4796                  goto FAILED;
4797                  }
4798                *code++ = verbs[i].op;
4799                }
4800    
4801              else
4802                {
4803                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4804                  {
4805                  *errorcodeptr = ERR59;
4806                  goto FAILED;
4807                  }
4808                *code++ = verbs[i].op_arg;
4809                *code++ = arglen;
4810                memcpy(code, arg, arglen);
4811                code += arglen;
4812                *code++ = 0;
4813                }
4814    
4815              break;  /* Found verb, exit loop */
4816            }            }
4817    
4818          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4819          }          }
4820        if (i < verbcount) continue;  
4821        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4822          *errorcodeptr = ERR60;          /* Verb not recognized */
4823        goto FAILED;        goto FAILED;
4824        }        }
4825    
# Line 4621  we set the flag only if there is a liter Line 4938  we set the flag only if there is a liter
4938                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4939            ptr++;            ptr++;
4940            }            }
4941          namelen = ptr - name;          namelen = (int)(ptr - name);
4942    
4943          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
4944              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4658  we set the flag only if there is a liter Line 4975  we set the flag only if there is a liter
4975            }            }
4976    
4977          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4978          name. If we find a name, add one to the opcode to change OP_CREF or          name. If we find a name, add one to the opcode to change OP_CREF or
4979          OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,          OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4980          except they record that the reference was originally to a name. The          except they record that the reference was originally to a name. The
4981          information is used to check duplicate names. */          information is used to check duplicate names. */
4982    
4983          slot = cd->name_table;          slot = cd->name_table;
# Line 4817  we set the flag only if there is a liter Line 5134  we set the flag only if there is a liter
5134              goto FAILED;              goto FAILED;
5135              }              }
5136            *code++ = n;            *code++ = n;
5137            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5138            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5139            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5140            }            }
5141          previous = NULL;          previous = NULL;
# Line 4851  we set the flag only if there is a liter Line 5168  we set the flag only if there is a liter
5168            name = ++ptr;            name = ++ptr;
5169    
5170            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5171            namelen = ptr - name;            namelen = (int)(ptr - name);
5172    
5173            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5174    
# Line 4887  we set the flag only if there is a liter Line 5204  we set the flag only if there is a liter
5204            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
5205            in the pre-compile, and it affects various numbers and pointers which            in the pre-compile, and it affects various numbers and pointers which
5206            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
5207            duplicates with the same number were omitted from the table. This            duplicates with the same number were omitted from the table. This
5208            doesn't seem worth the hassle. However, *different* names for the            doesn't seem worth the hassle. However, *different* names for the
5209            same number are not permitted. */            same number are not permitted. */
5210    
# Line 4895  we set the flag only if there is a liter Line 5212  we set the flag only if there is a liter
5212              {              {
5213              BOOL dupname = FALSE;              BOOL dupname = FALSE;
5214              slot = cd->name_table;              slot = cd->name_table;
5215    
5216              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5217                {                {
5218                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4909  we set the flag only if there is a liter Line 5226  we set the flag only if there is a liter
5226                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5227                      goto FAILED;                      goto FAILED;
5228                      }                      }
5229                    else dupname = TRUE;                    else dupname = TRUE;
5230                    }                    }
5231                  else crc = -1;      /* Current name is a substring */                  else crc = -1;      /* Current name is a substring */
5232                  }                  }
5233    
5234                /* Make space in the table and break the loop for an earlier                /* Make space in the table and break the loop for an earlier
5235                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
5236                duplicates so that in the simple case (when ?(| is not used) they                duplicates so that in the simple case (when ?(| is not used) they
5237                are in order of their numbers. */                are in order of their numbers. */
5238    
5239                if (crc < 0)                if (crc < 0)
5240                  {                  {
5241                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5242                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5243                  break;                  break;
5244                  }                  }
5245    
5246                /* Continue the loop for a later or duplicate name */                /* Continue the loop for a later or duplicate name */
5247    
5248                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5249                }                }
5250    
5251              /* For non-duplicate names, check for a duplicate number before              /* For non-duplicate names, check for a duplicate number before
5252              adding the new name. */              adding the new name. */
5253    
5254              if (!dupname)              if (!dupname)
5255                {                {
5256                uschar *cslot = cd->name_table;                uschar *cslot = cd->name_table;
# Line 4945  we set the flag only if there is a liter Line 5262  we set the flag only if there is a liter
5262                      {                      {
5263                      *errorcodeptr = ERR65;                      *errorcodeptr = ERR65;
5264                      goto FAILED;                      goto FAILED;
5265                      }                      }
5266                    }                    }
5267                  else i--;                  else i--;
5268                  cslot += cd->name_entry_size;                  cslot += cd->name_entry_size;
5269                  }                  }
5270                }                }
5271    
5272              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5273              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
# Line 4981  we set the flag only if there is a liter Line 5298  we set the flag only if there is a liter
5298          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5299          name = ++ptr;          name = ++ptr;
5300          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5301          namelen = ptr - name;          namelen = (int)(ptr - name);
5302    
5303          /* In the pre-compile phase, do a syntax check and set a dummy          /* In the pre-compile phase, do a syntax check and set a dummy
5304          reference number. */          reference number. */
# Line 5131  we set the flag only if there is a liter Line 5448  we set the flag only if there is a liter
5448            if (lengthptr == NULL)            if (lengthptr == NULL)
5449              {              {
5450              *code = OP_END;              *code = OP_END;
5451              if (recno != 0)              if (recno != 0)
5452                called = _pcre_find_bracket(cd->start_code, utf8, recno);                called = _pcre_find_bracket(cd->start_code, utf8, recno);
5453    
5454              /* Forward reference */              /* Forward reference */
# Line 5144  we set the flag only if there is a liter Line 5461  we set the flag only if there is a liter
5461                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5462                  goto FAILED;                  goto FAILED;
5463                  }                  }
5464    
5465                  /* Fudge the value of "called" so that when it is inserted as an
5466                  offset below, what it actually inserted is the reference number
5467                  of the group. */
5468    
5469                called = cd->start_code + recno;                called = cd->start_code + recno;
5470                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));
5471                }                }
5472    
5473              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
# Line 5153  we set the flag only if there is a liter Line 5475  we set the flag only if there is a liter
5475              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5476    
5477              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5478                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5479                {                {
5480                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5481                goto FAILED;                goto FAILED;
# Line 5169  we set the flag only if there is a liter Line 5491  we set the flag only if there is a liter
5491            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5492    
5493            *code = OP_RECURSE;            *code = OP_RECURSE;
5494            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
5495            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5496    
5497            *code = OP_KET;            *code = OP_KET;
# Line 5248  we set the flag only if there is a liter Line 5570  we set the flag only if there is a liter
5570              {              {
5571              cd->external_options = newoptions;              cd->external_options = newoptions;
5572              }              }
5573           else            else
5574              {              {
5575              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5576                {                {
# Line 5280  we set the flag only if there is a liter Line 5602  we set the flag only if there is a liter
5602          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5603        }       /* End of (? handling */        }       /* End of (? handling */
5604    
5605      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5606      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5607      brackets. */      brackets. */
5608    
5609      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5473  we set the flag only if there is a liter Line 5795  we set the flag only if there is a liter
5795    
5796      /* ===================================================================*/      /* ===================================================================*/
5797      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
5798      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
5799      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
5800      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
5801      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
5802      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
5803        ever created. */
5804    
5805      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
5806      tempptr = ptr;      tempptr = ptr;
# Line 5585  we set the flag only if there is a liter Line 5908  we set the flag only if there is a liter
5908    
5909        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5910          {          {
5911            open_capitem *oc;
5912          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5913    
5914          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5594  we set the flag only if there is a liter Line 5918  we set the flag only if there is a liter
5918          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5919          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5920          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5921    
5922            /* Check to see if this back reference is recursive, that it, it
5923            is inside the group that it references. A flag is set so that the
5924            group can be made atomic. */
5925    
5926            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5927              {
5928              if (oc->number == recno)
5929                {
5930                oc->flag = TRUE;
5931                break;
5932                }
5933              }
5934          }          }
5935    
5936        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5623  we set the flag only if there is a liter Line 5960  we set the flag only if there is a liter
5960  #endif  #endif
5961    
5962        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
5963        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
5964          situation when PCRE_UCP is not set. When it *is* set, we substitute
5965          Unicode property tests. */
5966    
5967        else        else
5968          {          {
5969          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
5970          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
5971              {
5972              nestptr = ptr + 1;                   /* Where to resume */
5973              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
5974              }
5975            else
5976    #endif
5977              {
5978              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
5979              *code++ = -c;
5980              }
5981          }          }
5982        continue;        continue;
5983        }        }
# Line 5783  int branchfirstbyte, branchreqbyte; Line 6132  int branchfirstbyte, branchreqbyte;
6132  int length;  int length;
6133  int orig_bracount;  int orig_bracount;
6134  int max_bracount;  int max_bracount;
6135    int old_external_options = cd->external_options;
6136  branch_chain bc;  branch_chain bc;
6137    
6138  bc.outer = bcptr;  bc.outer = bcptr;
6139  bc.current = code;  bc.current_branch = code;
6140    
6141  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
6142    
# Line 5805  them global. It tests the value of lengt Line 6155  them global. It tests the value of lengt
6155  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. */
6156    
6157  /* 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
6158  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
6159    detect groups that contain recursive back references to themselves. */
6160    
6161  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6162    {    {
6163    capnumber = GET2(code, 1 + LINK_SIZE);    capnumber = GET2(code, 1 + LINK_SIZE);
6164    capitem.number = capnumber;    capitem.number = capnumber;
6165    capitem.next = cd->open_caps;    capitem.next = cd->open_caps;
6166    cd->open_caps = &capitem;    capitem.flag = FALSE;
6167    }    cd->open_caps = &capitem;
6168      }
6169    
6170  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
6171    
# Line 5859  for (;;) Line 6211  for (;;)
6211      return FALSE;      return FALSE;
6212      }      }
6213    
6214      /* If the external options have changed during this branch, it means that we
6215      are at the top level, and a leading option setting has been encountered. We
6216      need to re-set the original option values to take account of this so that,
6217      during the pre-compile phase, we know to allow for a re-set at the start of
6218      subsequent branches. */
6219    
6220      if (old_external_options != cd->external_options)
6221        oldims = cd->external_options & PCRE_IMS;
6222    
6223    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6224    has fewer than the rest. */    has fewer than the rest. */
6225    
# Line 5909  for (;;) Line 6270  for (;;)
6270    
6271      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6272      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
6273      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
6274      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
6275      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
6276      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
6277      the problem can be shown. */      the problem can be shown. */
6278    
6279      if (lookbehind)      if (lookbehind)
# Line 5923  for (;;) Line 6284  for (;;)
6284        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6285        if (fixed_length == -3)        if (fixed_length == -3)
6286          {          {
6287          cd->check_lookbehind = TRUE;          cd->check_lookbehind = TRUE;
6288          }          }
6289        else if (fixed_length < 0)        else if (fixed_length < 0)
6290          {          {
6291          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
# Line 5948  for (;;) Line 6309  for (;;)
6309      {      {
6310      if (lengthptr == NULL)      if (lengthptr == NULL)
6311        {        {
6312        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6313        do        do
6314          {          {
6315          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 5958  for (;;) Line 6319  for (;;)
6319          }          }
6320        while (branch_length > 0);        while (branch_length > 0);
6321        }        }
   
     /* If it was a capturing subpattern, remove it from the chain. */  
   
     if (capnumber > 0) cd->open_caps = cd->open_caps->next;  
6322    
6323      /* Fill in the ket */      /* Fill in the ket */
6324    
6325      *code = OP_KET;      *code = OP_KET;
6326      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6327      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6328    
6329      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6330        recursive back references. If so, we must wrap it in atomic brackets.
6331        In any event, remove the block from the chain. */
6332    
6333        if (capnumber > 0)
6334          {
6335          if (cd->open_caps->flag)
6336            {
6337            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6338              code - start_bracket);
6339            *start_bracket = OP_ONCE;
6340            code += 1 + LINK_SIZE;
6341            PUT(start_bracket, 1, (int)(code - start_bracket));
6342            *code = OP_KET;
6343            PUT(code, 1, (int)(code - start_bracket));
6344            code += 1 + LINK_SIZE;
6345            length += 2 + 2*LINK_SIZE;
6346            }
6347          cd->open_caps = cd->open_caps->next;
6348          }
6349    
6350        /* Reset options if needed. */
6351    
6352      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6353        {        {
# Line 6017  for (;;) Line 6395  for (;;)
6395    else    else
6396      {      {
6397      *code = OP_ALT;      *code = OP_ALT;
6398      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6399      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6400      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6401      }      }
6402    
# Line 6336  int length = 1; /* For final END opcode Line 6714  int length = 1; /* For final END opcode
6714  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6715  int errorcode = 0;  int errorcode = 0;
6716  int skipatstart = 0;  int skipatstart = 0;
6717  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8;
6718  size_t size;  size_t size;
6719  uschar *code;  uschar *code;
6720  const uschar *codestart;  const uschar *codestart;
# Line 6406  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6784  while (ptr[skipatstart] == CHAR_LEFT_PAR
6784    
6785    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6786      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
6787      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
6788        { skipatstart += 6; options |= PCRE_UCP; continue; }
6789    
6790    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6791      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6429  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6809  while (ptr[skipatstart] == CHAR_LEFT_PAR
6809      options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;      options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
6810    else break;    else break;
6811    }    }
6812    
6813    utf8 = (options & PCRE_UTF8) != 0;
6814    
6815  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6816    
6817  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
6818  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6819       (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6820    {    {
6821    errorcode = ERR44;    errorcode = ERR44;
6822    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6447  if (utf8) Line 6829  if (utf8)
6829    }    }
6830  #endif  #endif
6831    
6832    /* Can't support UCP unless PCRE has been compiled to include the code. */
6833    
6834    #ifndef SUPPORT_UCP
6835    if ((options & PCRE_UCP) != 0)
6836      {
6837      errorcode = ERR67;
6838      goto PCRE_EARLY_ERROR_RETURN;
6839      }
6840    #endif
6841    
6842  /* Check validity of \R options. */  /* Check validity of \R options. */
6843    
6844  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 6575  regex compiled on a system with 4-byte p Line 6967  regex compiled on a system with 4-byte p
6967  pointers. */  pointers. */
6968    
6969  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
6970  re->size = size;  re->size = (int)size;
6971  re->options = cd->external_options;  re->options = cd->external_options;
6972  re->flags = cd->external_flags;  re->flags = cd->external_flags;
6973  re->dummy1 = 0;  re->dummy1 = 0;
# Line 6631  if debugging, leave the test till after Line 7023  if debugging, leave the test till after
7023    
7024  *code++ = OP_END;  *code++ = OP_END;
7025    
7026  #ifndef DEBUG  #ifndef PCRE_DEBUG
7027  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
7028  #endif  #endif
7029    
# Line 6646  while (errorcode == 0 && cd->hwm > cwork Line 7038  while (errorcode == 0 && cd->hwm > cwork
7038    recno = GET(codestart, offset);    recno = GET(codestart, offset);
7039    groupptr = _pcre_find_bracket(codestart, utf8, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
7040    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
7041      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7042    }    }
7043    
7044  /* 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 6654  subpattern. */ Line 7046  subpattern. */
7046    
7047  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
7048    
7049  /* If there were any lookbehind assertions that contained OP_RECURSE  /* If there were any lookbehind assertions that contained OP_RECURSE
7050  (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,
7051  because they may contain forward references. Actual recursions can't be fixed  because they may contain forward references. Actual recursions can't be fixed
7052  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 6665  length, and set their lengths. */ Line 7057  length, and set their lengths. */
7057  if (cd->check_lookbehind)  if (cd->check_lookbehind)
7058    {    {
7059    uschar *cc = (uschar *)codestart;    uschar *cc = (uschar *)codestart;
7060    
7061    /* 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
7062    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
7063    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
7064    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. */
7065    
7066    for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);    for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
7067         cc != NULL;         cc != NULL;
7068         cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))         cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
7069      {      {
7070      if (GET(cc, 1) == 0)      if (GET(cc, 1) == 0)
7071        {        {
7072        int fixed_length;        int fixed_length;
7073        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7074        int end_op = *be;        int end_op = *be;
7075        *be = OP_END;        *be = OP_END;
7076        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
7077        *be = end_op;        *be = end_op;
# Line 6687  if (cd->check_lookbehind) Line 7079  if (cd->check_lookbehind)
7079        if (fixed_length < 0)        if (fixed_length < 0)
7080          {          {
7081          errorcode = (fixed_length == -2)? ERR36 : ERR25;          errorcode = (fixed_length == -2)? ERR36 : ERR25;
7082          break;          break;
7083          }          }
7084        PUT(cc, 1, fixed_length);        PUT(cc, 1, fixed_length);
7085        }        }
7086      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
7087      }      }
7088    }    }
7089    
7090  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
7091    
# Line 6701  if (errorcode != 0) Line 7093  if (errorcode != 0)
7093    {    {
7094    (pcre_free)(re);    (pcre_free)(re);
7095    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
7096    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = (int)(ptr - (const uschar *)pattern);
7097    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
7098    *errorptr = find_error_text(errorcode);    *errorptr = find_error_text(errorcode);
7099    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
# Line 6755  if (reqbyte >= 0 && Line 7147  if (reqbyte >= 0 &&
7147  /* 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
7148  case when building a production library. */  case when building a production library. */
7149    
7150  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
7151  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
7152    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
7153    
# Line 6793  if (code - codestart > length) Line 7184  if (code - codestart > length)
7184    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
7185    return NULL;    return NULL;
7186    }    }
7187  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
7188    
7189  return (pcre *)re;  return (pcre *)re;
7190  }  }

Legend:
Removed from v.459  
changed lines
  Added in v.533

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12