/[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 438 by ph10, Sun Sep 6 20:00:47 2009 UTC revision 518 by ph10, Tue May 18 15:47:01 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"
404    "subpattern name expected\0"    "subpattern name expected\0"
405    "digit expected after (?+\0"    "digit expected after (?+\0"
406    "] is an invalid data character in JavaScript compatibility mode";    "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "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 500  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 571  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 809  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 1100  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 1146  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 1315  for (;;) Line 1397  for (;;)
1397    
1398      case OP_CALLOUT:      case OP_CALLOUT:
1399      case OP_CREF:      case OP_CREF:
1400        case OP_NCREF:
1401      case OP_RREF:      case OP_RREF:
1402        case OP_NRREF:
1403      case OP_DEF:      case OP_DEF:
1404      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1405      break;      break;
# Line 1331  for (;;) Line 1415  for (;;)
1415    
1416    
1417  /*************************************************  /*************************************************
1418  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1419  *************************************************/  *************************************************/
1420    
1421  /* Scan a pattern 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.  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.
1425    
1426    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.
1428    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
1430    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
1436      cd       the "compile data" structure
1437    
1438  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1439                 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
1442  */  */
1443    
1444  static int  static int
1445  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1446  {  {
1447  int length = -1;  int length = -1;
1448    
# Line 1360  branch, check the length against that of Line 1455  branch, check the length against that of
1455  for (;;)  for (;;)
1456    {    {
1457    int d;    int d;
1458      uschar *ce, *cs;
1459    register int op = *cc;    register int op = *cc;
1460    switch (op)    switch (op)
1461      {      {
# Line 1367  for (;;) Line 1463  for (;;)
1463      case OP_BRA:      case OP_BRA:
1464      case OP_ONCE:      case OP_ONCE:
1465      case OP_COND:      case OP_COND:
1466      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1467      if (d < 0) return d;      if (d < 0) return d;
1468      branchlength += d;      branchlength += d;
1469      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1390  for (;;) Line 1486  for (;;)
1486      branchlength = 0;      branchlength = 0;
1487      break;      break;
1488    
1489        /* 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
1491        it until the end of the pattern, so return -3. */
1492    
1493        case OP_RECURSE:
1494        if (!atend) return -3;
1495        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1496        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1497        if (cc > cs && cc < ce) return -1;                /* Recursion */
1498        d = find_fixedlength(cs + 2, options, atend, cd);
1499        if (d < 0) return d;
1500        branchlength += d;
1501        cc += 1 + LINK_SIZE;
1502        break;
1503    
1504      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1505    
1506      case OP_ASSERT:      case OP_ASSERT:
# Line 1403  for (;;) Line 1514  for (;;)
1514    
1515      case OP_REVERSE:      case OP_REVERSE:
1516      case OP_CREF:      case OP_CREF:
1517        case OP_NCREF:
1518      case OP_RREF:      case OP_RREF:
1519        case OP_NRREF:
1520      case OP_DEF:      case OP_DEF:
1521      case OP_OPT:      case OP_OPT:
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 1426  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 1438  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 1518  for (;;) Line 1632  for (;;)
1632    
1633    
1634  /*************************************************  /*************************************************
1635  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1636  *************************************************/  *************************************************/
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.  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
1641    so that it can be called from pcre_study() when finding the minimum matching
1642    length.
1643    
1644  Arguments:  Arguments:
1645    code        points to start of expression    code        points to start of expression
1646    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1647    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1648    
1649  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
1650  */  */
1651    
1652  static const uschar *  const uschar *
1653  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1654  {  {
1655  for (;;)  for (;;)
1656    {    {
# Line 1546  for (;;) Line 1663  for (;;)
1663    
1664    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1665    
1666      /* Handle recursion */
1667    
1668      else if (c == OP_REVERSE)
1669        {
1670        if (number < 0) return (uschar *)code;
1671        code += _pcre_OP_lengths[c];
1672        }
1673    
1674    /* Handle capturing bracket */    /* Handle capturing bracket */
1675    
1676    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1557  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 1581  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 1652  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 1676  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 1732  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 1768  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 1786  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 1913  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 1942  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 2598  BOOL inescq = FALSE; Line 2780  BOOL inescq = FALSE;
2780  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2781  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2782  const uschar *tempptr;  const uschar *tempptr;
2783    const uschar *nestptr = NULL;
2784  uschar *previous = NULL;  uschar *previous = NULL;
2785  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2786  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2614  BOOL utf8 = FALSE; Line 2797  BOOL utf8 = FALSE;
2797  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2798  #endif  #endif
2799    
2800  #ifdef DEBUG  #ifdef PCRE_DEBUG
2801  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2802  #endif  #endif
2803    
# Line 2668  for (;; ptr++) Line 2851  for (;; ptr++)
2851    
2852    c = *ptr;    c = *ptr;
2853    
2854      /* If we are at the end of a nested substitution, revert to the outer level
2855      string. Nesting only happens one level deep. */
2856    
2857      if (c == 0 && nestptr != NULL)
2858        {
2859        ptr = nestptr;
2860        nestptr = NULL;
2861        c = *ptr;
2862        }
2863    
2864    /* 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
2865    previous cycle of this loop. */    previous cycle of this loop. */
2866    
2867    if (lengthptr != NULL)    if (lengthptr != NULL)
2868      {      {
2869  #ifdef DEBUG  #ifdef PCRE_DEBUG
2870      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2871  #endif  #endif
2872      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
2873        {        {
2874        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
2875        goto FAILED;        goto FAILED;
# Line 2725  for (;; ptr++) Line 2918  for (;; ptr++)
2918    /* 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
2919    reference list. */    reference list. */
2920    
2921    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
2922      {      {
2923      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
2924      goto FAILED;      goto FAILED;
# Line 2965  for (;; ptr++) Line 3158  for (;; ptr++)
3158          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3159          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3160          }          }
3161    
3162        /* 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
3163        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
3164        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 3034  for (;; ptr++) Line 3227  for (;; ptr++)
3227    
3228          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3229            posix_class = 0;            posix_class = 0;
3230    
3231          /* 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
3232          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3233          subtract bits that may be in the main map already. At the end we or the  
3234          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3235            if ((options & PCRE_UCP) != 0)
3236              {
3237              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3238              if (posix_substitutes[pc] != NULL)
3239                {
3240                nestptr = tempptr + 1;
3241                ptr = posix_substitutes[pc] - 1;
3242                continue;
3243                }
3244              }
3245    #endif
3246            /* In the non-UCP case, we build the bit map for the POSIX class in a
3247            chunk of local store because we may be adding and subtracting from it,
3248            and we don't want to subtract bits that may be in the main map already.
3249            At the end we or the result into the bit map that is being built. */
3250    
3251          posix_class *= 3;          posix_class *= 3;
3252    
# Line 3082  for (;; ptr++) Line 3290  for (;; ptr++)
3290    
3291        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3292        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
3293        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
3294        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
3295        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
3296        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3297          PCRE_EXTRA is set. */
3298    
3299        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3300          {          {
3301          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3302          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3303    
3304          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 */  
3305          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3306            {            {
3307            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3111  for (;; ptr++) Line 3318  for (;; ptr++)
3318            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3319            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3320    
3321            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3322              {              {
3323    #ifdef SUPPORT_UCP
3324                case ESC_du:     /* These are the values given for \d etc */
3325                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3326                case ESC_wu:     /* escape sequence with an appropriate \p */
3327                case ESC_WU:     /* or \P to test Unicode properties instead */
3328                case ESC_su:     /* of the default ASCII testing. */
3329                case ESC_SU:
3330                nestptr = ptr;
3331                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3332                class_charcount -= 2;                /* Undo! */
3333                continue;
3334    #endif
3335              case ESC_d:              case ESC_d:
3336              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3337              continue;              continue;
# Line 3144  for (;; ptr++) Line 3361  for (;; ptr++)
3361              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3362              continue;              continue;
3363    
3364              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)  
             {  
3365              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3366              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3367              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3181  for (;; ptr++) Line 3385  for (;; ptr++)
3385                }                }
3386  #endif  #endif
3387              continue;              continue;
             }  
3388    
3389            if (-c == ESC_H)              case ESC_H:
             {  
3390              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3391                {                {
3392                int x = 0xff;                int x = 0xff;
# Line 3226  for (;; ptr++) Line 3428  for (;; ptr++)
3428                }                }
3429  #endif  #endif
3430              continue;              continue;
             }  
3431    
3432            if (-c == ESC_v)              case ESC_v:
             {  
3433              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3434              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3435              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3245  for (;; ptr++) Line 3445  for (;; ptr++)
3445                }                }
3446  #endif  #endif
3447              continue;              continue;
             }  
3448    
3449            if (-c == ESC_V)              case ESC_V:
             {  
3450              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3451                {                {
3452                int x = 0xff;                int x = 0xff;
# Line 3278  for (;; ptr++) Line 3476  for (;; ptr++)
3476                }                }
3477  #endif  #endif
3478              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3479    
3480  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3481            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3482              {              case ESC_P:
3483              BOOL negated;                {
3484              int pdata;                BOOL negated;
3485              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3486              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3487              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3488              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3489                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3490              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3491              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3492              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3493              continue;                class_charcount -= 2;   /* Not a < 256 character */
3494              }                continue;
3495                  }
3496  #endif  #endif
3497            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3498            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3499            treated as literals. */              treated as literals. */
3500    
3501            if ((options & PCRE_EXTRA) != 0)              default:
3502              {              if ((options & PCRE_EXTRA) != 0)
3503              *errorcodeptr = ERR7;                {
3504              goto FAILED;                *errorcodeptr = ERR7;
3505                  goto FAILED;
3506                  }
3507                class_charcount -= 2;  /* Undo the default count from above */
3508                c = *ptr;              /* Get the final character and fall through */
3509                break;
3510              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3511            }            }
3512    
3513          /* 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 3379  for (;; ptr++) Line 3577  for (;; ptr++)
3577            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3578            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3579    
3580            /* \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 */  
3581    
3582            if (d < 0)            if (d < 0)
3583              {              {
3584              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  
3585                {                {
3586                ptr = oldptr;                ptr = oldptr;
3587                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3552  for (;; ptr++) Line 3747  for (;; ptr++)
3747          }          }
3748        }        }
3749    
3750      /* 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.
3751        If we are at the end of an internal nested string, revert to the outer
3752        string. */
3753    
3754        while (((c = *(++ptr)) != 0 ||
3755               (nestptr != NULL &&
3756                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3757               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3758    
3759      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
3760    
3761      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3762        {        {
3763        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3764        goto FAILED;        goto FAILED;
3765        }        }
3766    
   
 /* 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  
   
   
3767      /* 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
3768      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
3769      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 3597  we set the flag only if there is a liter Line 3780  we set the flag only if there is a liter
3780      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
3781      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
3782      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3783    
3784  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3785      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
3786        (!utf8 || !negate_class || class_lastchar < 128))        (!utf8 || !negate_class || class_lastchar < 128))
# Line 3644  we set the flag only if there is a liter Line 3827  we set the flag only if there is a liter
3827    
3828      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3829      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
3830      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
3831      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
3832      (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
3833      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
3834        actual compiled code. */
3835    
3836  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3837      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
3838        {        {
3839        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3840        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3676  we set the flag only if there is a liter Line 3860  we set the flag only if there is a liter
3860        }        }
3861  #endif  #endif
3862    
3863      /* 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
3864      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
3865      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
3866      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
3867        negating it if necessary. */
3868    
3869      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3870      if (negate_class)      if (negate_class)
# Line 3871  we set the flag only if there is a liter Line 4056  we set the flag only if there is a liter
4056    
4057        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4058    
4059        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4060        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4061        removed: */        removed: */
4062    
4063        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4064        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4065    
4066        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4067        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4068    
4069        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4070    
# Line 4026  we set the flag only if there is a liter Line 4211  we set the flag only if there is a liter
4211          goto END_REPEAT;          goto END_REPEAT;
4212          }          }
4213    
4214        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4215        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4216        removed: */        removed: */
4217    
# Line 4034  we set the flag only if there is a liter Line 4219  we set the flag only if there is a liter
4219        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4220    
4221        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4222        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4223    
4224        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4225          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 4169  we set the flag only if there is a liter Line 4354  we set the flag only if there is a liter
4354            {            {
4355            /* 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
4356            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
4357            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4358              integer type when available, otherwise double. */
4359    
4360            if (lengthptr != NULL)            if (lengthptr != NULL)
4361              {              {
4362              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4363              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4364                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4365                        (INT64_OR_DOUBLE)INT_MAX ||
4366                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4367                {                {
4368                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4221  we set the flag only if there is a liter Line 4408  we set the flag only if there is a liter
4408          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
4409          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
4410          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
4411          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4412            a 64-bit integer type when available, otherwise double. */
4413    
4414          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4415            {            {
4416            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4417                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4418            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4419                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4420                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4421                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4422              {              {
4423              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4306  we set the flag only if there is a liter Line 4494  we set the flag only if there is a liter
4494            uschar *scode = bracode;            uschar *scode = bracode;
4495            do            do
4496              {              {
4497              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4498                {                {
4499                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4500                break;                break;
# Line 4349  we set the flag only if there is a liter Line 4537  we set the flag only if there is a liter
4537      if (possessive_quantifier)      if (possessive_quantifier)
4538        {        {
4539        int len;        int len;
4540    
4541        if (*tempcode == OP_TYPEEXACT)        if (*tempcode == OP_TYPEEXACT)
4542          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4543            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4544    
4545        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4546          {          {
4547          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
# Line 4361  we set the flag only if there is a liter Line 4549  we set the flag only if there is a liter
4549          if (utf8 && tempcode[-1] >= 0xc0)          if (utf8 && tempcode[-1] >= 0xc0)
4550            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4551  #endif  #endif
4552          }          }
4553    
4554        len = code - tempcode;        len = code - tempcode;
4555        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4556          {          {
# Line 4381  we set the flag only if there is a liter Line 4569  we set the flag only if there is a liter
4569          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4570          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4571    
4572            /* Because we are moving code along, we must ensure that any
4573            pending recursive references are updated. */
4574    
4575          default:          default:
4576            *code = OP_END;
4577            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4578          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4579          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4580          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4417  we set the flag only if there is a liter Line 4610  we set the flag only if there is a liter
4610    
4611      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4612    
4613      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4614             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4615        {        {
4616        int i, namelen;        int i, namelen;
4617          int arglen = 0;
4618        const char *vn = verbnames;        const char *vn = verbnames;
4619        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4620          const uschar *arg = NULL;
4621        previous = NULL;        previous = NULL;
4622        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4623          namelen = ptr - name;
4624    
4625        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
4626          {          {
4627          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4628          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4629              || *ptr == '_') ptr++;
4630            arglen = ptr - arg;
4631          }          }
4632    
4633        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4634          {          {
4635          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4636          goto FAILED;          goto FAILED;
4637          }          }
4638        namelen = ptr - name;  
4639          /* Scan the table of verb names */
4640    
4641        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4642          {          {
4643          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4644              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4645            {            {
4646            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4647            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4648            break;            if (verbs[i].op == OP_ACCEPT)
4649                {
4650                open_capitem *oc;
4651                cd->had_accept = TRUE;
4652                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4653                  {
4654                  *code++ = OP_CLOSE;
4655                  PUT2INC(code, 0, oc->number);
4656                  }
4657                }
4658    
4659              /* Handle the cases with/without an argument */
4660    
4661              if (arglen == 0)
4662                {
4663                if (verbs[i].op < 0)   /* Argument is mandatory */
4664                  {
4665                  *errorcodeptr = ERR66;
4666                  goto FAILED;
4667                  }
4668                *code++ = verbs[i].op;
4669                }
4670    
4671              else
4672                {
4673                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4674                  {
4675                  *errorcodeptr = ERR59;
4676                  goto FAILED;
4677                  }
4678                *code++ = verbs[i].op_arg;
4679                *code++ = arglen;
4680                memcpy(code, arg, arglen);
4681                code += arglen;
4682                *code++ = 0;
4683                }
4684    
4685              break;  /* Found verb, exit loop */
4686            }            }
4687    
4688          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4689          }          }
4690        if (i < verbcount) continue;  
4691        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4692          *errorcodeptr = ERR60;          /* Verb not recognized */
4693        goto FAILED;        goto FAILED;
4694        }        }
4695    
# Line 4603  we set the flag only if there is a liter Line 4845  we set the flag only if there is a liter
4845            }            }
4846    
4847          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4848          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4849            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4850            except they record that the reference was originally to a name. The
4851            information is used to check duplicate names. */
4852    
4853          slot = cd->name_table;          slot = cd->name_table;
4854          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4618  we set the flag only if there is a liter Line 4863  we set the flag only if there is a liter
4863            {            {
4864            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4865            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4866              code[1+LINK_SIZE]++;
4867            }            }
4868    
4869          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
# Line 4626  we set the flag only if there is a liter Line 4872  we set the flag only if there is a liter
4872                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4873            {            {
4874            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4875              code[1+LINK_SIZE]++;
4876            }            }
4877    
4878          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4818  we set the flag only if there is a liter Line 5065  we set the flag only if there is a liter
5065                }                }
5066              }              }
5067    
5068            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5069              alphabetical order. Duplicate names for different numbers are
5070              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5071              number are always OK. (An existing number can be re-used if (?|
5072              appears in the pattern.) In either event, a duplicate name results in
5073              a duplicate entry in the table, even if the number is the same. This
5074              is because the number of names, and hence the table size, is computed
5075              in the pre-compile, and it affects various numbers and pointers which
5076              would all have to be modified, and the compiled code moved down, if
5077              duplicates with the same number were omitted from the table. This
5078              doesn't seem worth the hassle. However, *different* names for the
5079              same number are not permitted. */
5080    
5081            else            else
5082              {              {
5083                BOOL dupname = FALSE;
5084              slot = cd->name_table;              slot = cd->name_table;
5085    
5086              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5087                {                {
5088                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4830  we set the flag only if there is a liter Line 5090  we set the flag only if there is a liter
5090                  {                  {
5091                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5092                    {                    {
5093                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5094                          (options & PCRE_DUPNAMES) == 0)
5095                      {                      {
5096                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5097                      goto FAILED;                      goto FAILED;
5098                      }                      }
5099                      else dupname = TRUE;
5100                    }                    }
5101                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5102                  }                  }
5103    
5104                  /* Make space in the table and break the loop for an earlier
5105                  name. For a duplicate or later name, carry on. We do this for
5106                  duplicates so that in the simple case (when ?(| is not used) they
5107                  are in order of their numbers. */
5108    
5109                if (crc < 0)                if (crc < 0)
5110                  {                  {
5111                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5112                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5113                  break;                  break;
5114                  }                  }
5115    
5116                  /* Continue the loop for a later or duplicate name */
5117    
5118                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5119                }                }
5120    
5121                /* For non-duplicate names, check for a duplicate number before
5122                adding the new name. */
5123    
5124                if (!dupname)
5125                  {
5126                  uschar *cslot = cd->name_table;
5127                  for (i = 0; i < cd->names_found; i++)
5128                    {
5129                    if (cslot != slot)
5130                      {
5131                      if (GET2(cslot, 0) == cd->bracount + 1)
5132                        {
5133                        *errorcodeptr = ERR65;
5134                        goto FAILED;
5135                        }
5136                      }
5137                    else i--;
5138                    cslot += cd->name_entry_size;
5139                    }
5140                  }
5141    
5142              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5143              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5144              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5145              }              }
5146            }            }
5147    
5148          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5149            encountered. */
5150    
         ptr++;                    /* Move past > or ' */  
5151          cd->names_found++;          cd->names_found++;
5152            ptr++;                    /* Move past > or ' */
5153          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5154    
5155    
# Line 5025  we set the flag only if there is a liter Line 5318  we set the flag only if there is a liter
5318            if (lengthptr == NULL)            if (lengthptr == NULL)
5319              {              {
5320              *code = OP_END;              *code = OP_END;
5321              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5322                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5323    
5324              /* Forward reference */              /* Forward reference */
5325    
# Line 5037  we set the flag only if there is a liter Line 5331  we set the flag only if there is a liter
5331                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5332                  goto FAILED;                  goto FAILED;
5333                  }                  }
5334    
5335                  /* Fudge the value of "called" so that when it is inserted as an
5336                  offset below, what it actually inserted is the reference number
5337                  of the group. */
5338    
5339                called = cd->start_code + recno;                called = cd->start_code + recno;
5340                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5341                }                }
# Line 5046  we set the flag only if there is a liter Line 5345  we set the flag only if there is a liter
5345              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5346    
5347              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5348                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5349                {                {
5350                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5351                goto FAILED;                goto FAILED;
# Line 5141  we set the flag only if there is a liter Line 5440  we set the flag only if there is a liter
5440              {              {
5441              cd->external_options = newoptions;              cd->external_options = newoptions;
5442              }              }
5443           else            else
5444              {              {
5445              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5446                {                {
# Line 5173  we set the flag only if there is a liter Line 5472  we set the flag only if there is a liter
5472          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5473        }       /* End of (? handling */        }       /* End of (? handling */
5474    
5475      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5476      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5477      brackets. */      brackets. */
5478    
5479      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5366  we set the flag only if there is a liter Line 5665  we set the flag only if there is a liter
5665    
5666      /* ===================================================================*/      /* ===================================================================*/
5667      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
5668      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
5669      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
5670      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
5671      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
5672      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
5673        ever created. */
5674    
5675      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
5676      tempptr = ptr;      tempptr = ptr;
# Line 5478  we set the flag only if there is a liter Line 5778  we set the flag only if there is a liter
5778    
5779        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5780          {          {
5781            open_capitem *oc;
5782          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5783    
5784          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5487  we set the flag only if there is a liter Line 5788  we set the flag only if there is a liter
5788          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5789          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5790          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5791    
5792            /* Check to see if this back reference is recursive, that it, it
5793            is inside the group that it references. A flag is set so that the
5794            group can be made atomic. */
5795    
5796            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5797              {
5798              if (oc->number == recno)
5799                {
5800                oc->flag = TRUE;
5801                break;
5802                }
5803              }
5804          }          }
5805    
5806        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5516  we set the flag only if there is a liter Line 5830  we set the flag only if there is a liter
5830  #endif  #endif
5831    
5832        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
5833        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
5834          situation when PCRE_UCP is not set. When it *is* set, we substitute
5835          Unicode property tests. */
5836    
5837        else        else
5838          {          {
5839          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
5840          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
5841              {
5842              nestptr = ptr + 1;                   /* Where to resume */
5843              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
5844              }
5845            else
5846    #endif
5847              {
5848              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
5849              *code++ = -c;
5850              }
5851          }          }
5852        continue;        continue;
5853        }        }
# Line 5669  uschar *code = *codeptr; Line 5995  uschar *code = *codeptr;
5995  uschar *last_branch = code;  uschar *last_branch = code;
5996  uschar *start_bracket = code;  uschar *start_bracket = code;
5997  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5998    open_capitem capitem;
5999    int capnumber = 0;
6000  int firstbyte, reqbyte;  int firstbyte, reqbyte;
6001  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
6002  int length;  int length;
6003  int orig_bracount;  int orig_bracount;
6004  int max_bracount;  int max_bracount;
6005    int old_external_options = cd->external_options;
6006  branch_chain bc;  branch_chain bc;
6007    
6008  bc.outer = bcptr;  bc.outer = bcptr;
6009  bc.current = code;  bc.current_branch = code;
6010    
6011  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
6012    
# Line 5695  the code that abstracts option settings Line 6024  the code that abstracts option settings
6024  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the
6025  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. */
6026    
6027    /* If this is a capturing subpattern, add to the chain of open capturing items
6028    so that we can detect them if (*ACCEPT) is encountered. This is also used to
6029    detect groups that contain recursive back references to themselves. */
6030    
6031    if (*code == OP_CBRA)
6032      {
6033      capnumber = GET2(code, 1 + LINK_SIZE);
6034      capitem.number = capnumber;
6035      capitem.next = cd->open_caps;
6036      capitem.flag = FALSE;
6037      cd->open_caps = &capitem;
6038      }
6039    
6040  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
6041    
6042  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5739  for (;;) Line 6081  for (;;)
6081      return FALSE;      return FALSE;
6082      }      }
6083    
6084      /* If the external options have changed during this branch, it means that we
6085      are at the top level, and a leading option setting has been encountered. We
6086      need to re-set the original option values to take account of this so that,
6087      during the pre-compile phase, we know to allow for a re-set at the start of
6088      subsequent branches. */
6089    
6090      if (old_external_options != cd->external_options)
6091        oldims = cd->external_options & PCRE_IMS;
6092    
6093    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6094    has fewer than the rest. */    has fewer than the rest. */
6095    
# Line 5789  for (;;) Line 6140  for (;;)
6140    
6141      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6142      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
6143      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
6144        because there may be forward references that we can't check here. Set a
6145        flag to cause another lookbehind check at the end. Why not do it all at the
6146        end? Because common, erroneous checks are picked up here and the offset of
6147        the problem can be shown. */
6148    
6149      if (lookbehind)      if (lookbehind)
6150        {        {
6151        int fixed_length;        int fixed_length;
6152        *code = OP_END;        *code = OP_END;
6153        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
6154        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6155        if (fixed_length < 0)        if (fixed_length == -3)
6156            {
6157            cd->check_lookbehind = TRUE;
6158            }
6159          else if (fixed_length < 0)
6160          {          {
6161          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6162          *ptrptr = ptr;          *ptrptr = ptr;
6163          return FALSE;          return FALSE;
6164          }          }
6165        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6166        }        }
6167      }      }
6168    
# Line 5837  for (;;) Line 6196  for (;;)
6196      PUT(code, 1, code - start_bracket);      PUT(code, 1, code - start_bracket);
6197      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6198    
6199      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6200        recursive back references. If so, we must wrap it in atomic brackets.
6201        In any event, remove the block from the chain. */
6202    
6203        if (capnumber > 0)
6204          {
6205          if (cd->open_caps->flag)
6206            {
6207            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6208              code - start_bracket);
6209            *start_bracket = OP_ONCE;
6210            code += 1 + LINK_SIZE;
6211            PUT(start_bracket, 1, code - start_bracket);
6212            *code = OP_KET;
6213            PUT(code, 1, code - start_bracket);
6214            code += 1 + LINK_SIZE;
6215            length += 2 + 2*LINK_SIZE;
6216            }
6217          cd->open_caps = cd->open_caps->next;
6218          }
6219    
6220        /* Reset options if needed. */
6221    
6222      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6223        {        {
# Line 5886  for (;;) Line 6266  for (;;)
6266      {      {
6267      *code = OP_ALT;      *code = OP_ALT;
6268      PUT(code, 1, code - last_branch);      PUT(code, 1, code - last_branch);
6269      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6270      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6271      }      }
6272    
# Line 6033  do { Line 6413  do {
6413       switch (*scode)       switch (*scode)
6414         {         {
6415         case OP_CREF:         case OP_CREF:
6416           case OP_NCREF:
6417         case OP_RREF:         case OP_RREF:
6418           case OP_NRREF:
6419         case OP_DEF:         case OP_DEF:
6420         return FALSE;         return FALSE;
6421    
# Line 6202  int length = 1;  /* For final END opcode Line 6584  int length = 1;  /* For final END opcode
6584  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6585  int errorcode = 0;  int errorcode = 0;
6586  int skipatstart = 0;  int skipatstart = 0;
 #ifdef SUPPORT_UTF8  
6587  BOOL utf8;  BOOL utf8;
 #endif  
6588  size_t size;  size_t size;
6589  uschar *code;  uschar *code;
6590  const uschar *codestart;  const uschar *codestart;
# Line 6274  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6654  while (ptr[skipatstart] == CHAR_LEFT_PAR
6654    
6655    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6656      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
6657      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
6658        { skipatstart += 6; options |= PCRE_UCP; continue; }
6659    
6660    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6661      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6297  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6679  while (ptr[skipatstart] == CHAR_LEFT_PAR
6679      options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;      options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
6680    else break;    else break;
6681    }    }
6682    
6683    utf8 = (options & PCRE_UTF8) != 0;
6684    
6685  /* 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. */
6686    
6687  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
 utf8 = (options & PCRE_UTF8) != 0;  
6688  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6689       (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6690    {    {
6691    errorcode = ERR44;    errorcode = ERR44;
6692    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
6693    }    }
6694  #else  #else
6695  if ((options & PCRE_UTF8) != 0)  if (utf8)
6696    {    {
6697    errorcode = ERR32;    errorcode = ERR32;
6698    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
6699    }    }
6700  #endif  #endif
6701    
6702    /* Can't support UCP unless PCRE has been compiled to include the code. */
6703    
6704    #ifndef SUPPORT_UCP
6705    if ((options & PCRE_UCP) != 0)
6706      {
6707      errorcode = ERR67;
6708      goto PCRE_EARLY_ERROR_RETURN;
6709      }
6710    #endif
6711    
6712  /* Check validity of \R options. */  /* Check validity of \R options. */
6713    
6714  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 6398  cd->end_pattern = (const uschar *)(patte Line 6791  cd->end_pattern = (const uschar *)(patte
6791  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6792  cd->external_options = options;  cd->external_options = options;
6793  cd->external_flags = 0;  cd->external_flags = 0;
6794    cd->open_caps = NULL;
6795    
6796  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we
6797  don't need to look at the result of the function here. The initial options have  don't need to look at the result of the function here. The initial options have
# Line 6472  cd->start_code = codestart; Line 6866  cd->start_code = codestart;
6866  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6867  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6868  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6869    cd->check_lookbehind = FALSE;
6870    cd->open_caps = NULL;
6871    
6872  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6873  error, errorcode will be set non-zero, so we don't need to look at the result  error, errorcode will be set non-zero, so we don't need to look at the result
# Line 6497  if debugging, leave the test till after Line 6893  if debugging, leave the test till after
6893    
6894  *code++ = OP_END;  *code++ = OP_END;
6895    
6896  #ifndef DEBUG  #ifndef PCRE_DEBUG
6897  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
6898  #endif  #endif
6899    
# Line 6510  while (errorcode == 0 && cd->hwm > cwork Line 6906  while (errorcode == 0 && cd->hwm > cwork
6906    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6907    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6908    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6909    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6910    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6911      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6912    }    }
# Line 6520  subpattern. */ Line 6916  subpattern. */
6916    
6917  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6918    
6919    /* If there were any lookbehind assertions that contained OP_RECURSE
6920    (recursions or subroutine calls), a flag is set for them to be checked here,
6921    because they may contain forward references. Actual recursions can't be fixed
6922    length, but subroutine calls can. It is done like this so that those without
6923    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6924    exceptional ones forgo this. We scan the pattern to check that they are fixed
6925    length, and set their lengths. */
6926    
6927    if (cd->check_lookbehind)
6928      {
6929      uschar *cc = (uschar *)codestart;
6930    
6931      /* Loop, searching for OP_REVERSE items, and process those that do not have
6932      their length set. (Actually, it will also re-process any that have a length
6933      of zero, but that is a pathological case, and it does no harm.) When we find
6934      one, we temporarily terminate the branch it is in while we scan it. */
6935    
6936      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6937           cc != NULL;
6938           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6939        {
6940        if (GET(cc, 1) == 0)
6941          {
6942          int fixed_length;
6943          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6944          int end_op = *be;
6945          *be = OP_END;
6946          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6947          *be = end_op;
6948          DPRINTF(("fixed length = %d\n", fixed_length));
6949          if (fixed_length < 0)
6950            {
6951            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6952            break;
6953            }
6954          PUT(cc, 1, fixed_length);
6955          }
6956        cc += 1 + LINK_SIZE;
6957        }
6958      }
6959    
6960  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6961    
6962  if (errorcode != 0)  if (errorcode != 0)
# Line 6580  if (reqbyte >= 0 && Line 7017  if (reqbyte >= 0 &&
7017  /* 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
7018  case when building a production library. */  case when building a production library. */
7019    
7020  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
7021  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
7022    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
7023    
# Line 6618  if (code - codestart > length) Line 7054  if (code - codestart > length)
7054    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
7055    return NULL;    return NULL;
7056    }    }
7057  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
7058    
7059  return (pcre *)re;  return (pcre *)re;
7060  }  }

Legend:
Removed from v.438  
changed lines
  Added in v.518

  ViewVC Help
Powered by ViewVC 1.1.5