/[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 513 by ph10, Mon May 3 11:13:37 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 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 262  the number of relocations needed when a Line 273  the number of relocations needed when a
273  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
274  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
275  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
276  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
277    
278    Each substring ends with \0 to insert a null character. This includes the final
279    substring, so that the whole string ends with \0\0, which can be detected when
280    counting through. */
281    
282  static const char error_texts[] =  static const char error_texts[] =
283    "no error\0"    "no error\0"
# Line 335  static const char error_texts[] = Line 350  static const char error_texts[] =
350    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
351    "\\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"
352    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
353    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
354    /* 60 */    /* 60 */
355    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
356    "number is too big\0"    "number is too big\0"
357    "subpattern name expected\0"    "subpattern name expected\0"
358    "digit expected after (?+\0"    "digit expected after (?+\0"
359    "] is an invalid data character in JavaScript compatibility mode";    "] is an invalid data character in JavaScript compatibility mode\0"
360      /* 65 */
361      "different names for subpatterns of the same number are not allowed\0"
362      "(*MARK) must have an argument\0"
363      ;
364    
365  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
366  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 518  static const char *
518  find_error_text(int n)  find_error_text(int n)
519  {  {
520  const char *s = error_texts;  const char *s = error_texts;
521  for (; n > 0; n--) while (*s++ != 0) {};  for (; n > 0; n--)
522      {
523      while (*s++ != 0) {};
524      if (*s == 0) return "Error text not found (please report)";
525      }
526  return s;  return s;
527  }  }
528    
# Line 1100  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1122  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1122        if (name != NULL && lorn == ptr - thisname &&        if (name != NULL && lorn == ptr - thisname &&
1123            strncmp((const char *)name, (const char *)thisname, lorn) == 0)            strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1124          return *count;          return *count;
1125        term++;        term++;
1126        }        }
1127      }      }
1128    }    }
# Line 1146  for (; *ptr != 0; ptr++) Line 1168  for (; *ptr != 0; ptr++)
1168            break;            break;
1169          }          }
1170        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1171          {          {
1172          negate_class = TRUE;          negate_class = TRUE;
1173          ptr++;          ptr++;
1174          }          }
1175        else break;        else break;
1176        }        }
1177    
# Line 1315  for (;;) Line 1337  for (;;)
1337    
1338      case OP_CALLOUT:      case OP_CALLOUT:
1339      case OP_CREF:      case OP_CREF:
1340        case OP_NCREF:
1341      case OP_RREF:      case OP_RREF:
1342        case OP_NRREF:
1343      case OP_DEF:      case OP_DEF:
1344      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1345      break;      break;
# Line 1331  for (;;) Line 1355  for (;;)
1355    
1356    
1357  /*************************************************  /*************************************************
1358  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1359  *************************************************/  *************************************************/
1360    
1361  /* 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,
1362  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.
1363  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
1364    temporarily terminated with OP_END when this function is called.
1365    
1366    This function is called when a backward assertion is encountered, so that if it
1367    fails, the error message can point to the correct place in the pattern.
1368    However, we cannot do this when the assertion contains subroutine calls,
1369    because they can be forward references. We solve this by remembering this case
1370    and doing the check at the end; a flag specifies which mode we are running in.
1371    
1372  Arguments:  Arguments:
1373    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1374    options  the compiling options    options  the compiling options
1375      atend    TRUE if called when the pattern is complete
1376      cd       the "compile data" structure
1377    
1378  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1379                 or -1 if there is no fixed length,
1380               or -2 if \C was encountered               or -2 if \C was encountered
1381                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1382  */  */
1383    
1384  static int  static int
1385  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1386  {  {
1387  int length = -1;  int length = -1;
1388    
# Line 1360  branch, check the length against that of Line 1395  branch, check the length against that of
1395  for (;;)  for (;;)
1396    {    {
1397    int d;    int d;
1398      uschar *ce, *cs;
1399    register int op = *cc;    register int op = *cc;
1400    switch (op)    switch (op)
1401      {      {
# Line 1367  for (;;) Line 1403  for (;;)
1403      case OP_BRA:      case OP_BRA:
1404      case OP_ONCE:      case OP_ONCE:
1405      case OP_COND:      case OP_COND:
1406      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1407      if (d < 0) return d;      if (d < 0) return d;
1408      branchlength += d;      branchlength += d;
1409      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1390  for (;;) Line 1426  for (;;)
1426      branchlength = 0;      branchlength = 0;
1427      break;      break;
1428    
1429        /* A true recursion implies not fixed length, but a subroutine call may
1430        be OK. If the subroutine is a forward reference, we can't deal with
1431        it until the end of the pattern, so return -3. */
1432    
1433        case OP_RECURSE:
1434        if (!atend) return -3;
1435        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1436        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1437        if (cc > cs && cc < ce) return -1;                /* Recursion */
1438        d = find_fixedlength(cs + 2, options, atend, cd);
1439        if (d < 0) return d;
1440        branchlength += d;
1441        cc += 1 + LINK_SIZE;
1442        break;
1443    
1444      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1445    
1446      case OP_ASSERT:      case OP_ASSERT:
# Line 1403  for (;;) Line 1454  for (;;)
1454    
1455      case OP_REVERSE:      case OP_REVERSE:
1456      case OP_CREF:      case OP_CREF:
1457        case OP_NCREF:
1458      case OP_RREF:      case OP_RREF:
1459        case OP_NRREF:
1460      case OP_DEF:      case OP_DEF:
1461      case OP_OPT:      case OP_OPT:
1462      case OP_CALLOUT:      case OP_CALLOUT:
1463      case OP_SOD:      case OP_SOD:
1464      case OP_SOM:      case OP_SOM:
1465        case OP_SET_SOM:
1466      case OP_EOD:      case OP_EOD:
1467      case OP_EODN:      case OP_EODN:
1468      case OP_CIRC:      case OP_CIRC:
# Line 1426  for (;;) Line 1480  for (;;)
1480      branchlength++;      branchlength++;
1481      cc += 2;      cc += 2;
1482  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1483      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1484        cc += _pcre_utf8_table4[cc[-1] & 0x3f];        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
1485  #endif  #endif
1486      break;      break;
# Line 1438  for (;;) Line 1492  for (;;)
1492      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1493      cc += 4;      cc += 4;
1494  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1495      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1496        cc += _pcre_utf8_table4[cc[-1] & 0x3f];        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
1497  #endif  #endif
1498      break;      break;
# Line 1518  for (;;) Line 1572  for (;;)
1572    
1573    
1574  /*************************************************  /*************************************************
1575  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1576  *************************************************/  *************************************************/
1577    
1578  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1579  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1580    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1581    so that it can be called from pcre_study() when finding the minimum matching
1582    length.
1583    
1584  Arguments:  Arguments:
1585    code        points to start of expression    code        points to start of expression
1586    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1587    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1588    
1589  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
1590  */  */
1591    
1592  static const uschar *  const uschar *
1593  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1594  {  {
1595  for (;;)  for (;;)
1596    {    {
# Line 1546  for (;;) Line 1603  for (;;)
1603    
1604    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1605    
1606      /* Handle recursion */
1607    
1608      else if (c == OP_REVERSE)
1609        {
1610        if (number < 0) return (uschar *)code;
1611        code += _pcre_OP_lengths[c];
1612        }
1613    
1614    /* Handle capturing bracket */    /* Handle capturing bracket */
1615    
1616    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1557  for (;;) Line 1622  for (;;)
1622    
1623    /* 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
1624    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
1625    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1626      must add in its length. */
1627    
1628    else    else
1629      {      {
# Line 1581  for (;;) Line 1647  for (;;)
1647        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1648        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1649        break;        break;
1650    
1651          case OP_MARK:
1652          case OP_PRUNE_ARG:
1653          case OP_SKIP_ARG:
1654          case OP_THEN_ARG:
1655          code += code[1];
1656          break;
1657        }        }
1658    
1659      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1652  for (;;) Line 1725  for (;;)
1725    
1726    /* 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
1727    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
1728    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1729      must add in its length. */
1730    
1731    else    else
1732      {      {
# Line 1676  for (;;) Line 1750  for (;;)
1750        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1751        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1752        break;        break;
1753    
1754          case OP_MARK:
1755          case OP_PRUNE_ARG:
1756          case OP_SKIP_ARG:
1757          case OP_THEN_ARG:
1758          code += code[1];
1759          break;
1760        }        }
1761    
1762      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1732  Arguments: Line 1813  Arguments:
1813    code        points to start of search    code        points to start of search
1814    endcode     points to where to stop    endcode     points to where to stop
1815    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1816      cd          contains pointers to tables etc.
1817    
1818  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1819  */  */
1820    
1821  static BOOL  static BOOL
1822  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1823      compile_data *cd)
1824  {  {
1825  register int c;  register int c;
1826  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 1851  for (code = first_significant_code(code
1851      continue;      continue;
1852      }      }
1853    
1854      /* For a recursion/subroutine call, if its end has been reached, which
1855      implies a subroutine call, we can scan it. */
1856    
1857      if (c == OP_RECURSE)
1858        {
1859        BOOL empty_branch = FALSE;
1860        const uschar *scode = cd->start_code + GET(code, 1);
1861        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1862        do
1863          {
1864          if (could_be_empty_branch(scode, endcode, utf8, cd))
1865            {
1866            empty_branch = TRUE;
1867            break;
1868            }
1869          scode += GET(scode, 1);
1870          }
1871        while (*scode == OP_ALT);
1872        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1873        continue;
1874        }
1875    
1876    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1877    
1878    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 1891  for (code = first_significant_code(code
1891        empty_branch = FALSE;        empty_branch = FALSE;
1892        do        do
1893          {          {
1894          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1895            empty_branch = TRUE;            empty_branch = TRUE;
1896          code += GET(code, 1);          code += GET(code, 1);
1897          }          }
# Line 1913  for (code = first_significant_code(code Line 2018  for (code = first_significant_code(code
2018      case OP_POSQUERY:      case OP_POSQUERY:
2019      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2020      break;      break;
2021    
2022      case OP_UPTO:      case OP_UPTO:
2023      case OP_MINUPTO:      case OP_MINUPTO:
2024      case OP_POSUPTO:      case OP_POSUPTO:
2025      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2026      break;      break;
2027  #endif  #endif
2028    
2029        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2030        string. */
2031    
2032        case OP_MARK:
2033        case OP_PRUNE_ARG:
2034        case OP_SKIP_ARG:
2035        case OP_THEN_ARG:
2036        code += code[1];
2037        break;
2038    
2039        /* None of the remaining opcodes are required to match a character. */
2040    
2041        default:
2042        break;
2043      }      }
2044    }    }
2045    
# Line 1942  Arguments: Line 2062  Arguments:
2062    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2063    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2064    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2065      cd          pointers to tables etc
2066    
2067  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2068  */  */
2069    
2070  static BOOL  static BOOL
2071  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2072    BOOL utf8)    BOOL utf8, compile_data *cd)
2073  {  {
2074  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2075    {    {
2076    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2077        return FALSE;
2078    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2079    }    }
2080  return TRUE;  return TRUE;
# Line 2614  BOOL utf8 = FALSE; Line 2736  BOOL utf8 = FALSE;
2736  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2737  #endif  #endif
2738    
2739  #ifdef DEBUG  #ifdef PCRE_DEBUG
2740  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2741  #endif  #endif
2742    
# Line 2673  for (;; ptr++) Line 2795  for (;; ptr++)
2795    
2796    if (lengthptr != NULL)    if (lengthptr != NULL)
2797      {      {
2798  #ifdef DEBUG  #ifdef PCRE_DEBUG
2799      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2800  #endif  #endif
2801      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
2802        {        {
2803        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
2804        goto FAILED;        goto FAILED;
# Line 2725  for (;; ptr++) Line 2847  for (;; ptr++)
2847    /* 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
2848    reference list. */    reference list. */
2849    
2850    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
2851      {      {
2852      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
2853      goto FAILED;      goto FAILED;
# Line 3082  for (;; ptr++) Line 3204  for (;; ptr++)
3204    
3205        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3206        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
3207        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
3208        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
3209        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
3210        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3211          PCRE_EXTRA is set. */
3212    
3213        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3214          {          {
3215          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3216          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3217    
3218          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 */  
3219          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3220            {            {
3221            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3871  we set the flag only if there is a liter Line 3992  we set the flag only if there is a liter
3992    
3993        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3994    
3995        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
3996        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
3997        removed: */        removed: */
3998    
3999        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4000        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4001    
4002        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4003        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4004    
4005        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4006    
# Line 4026  we set the flag only if there is a liter Line 4147  we set the flag only if there is a liter
4147          goto END_REPEAT;          goto END_REPEAT;
4148          }          }
4149    
4150        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4151        /* This code is obsolete from release 8.00; the restriction was finally        /* This code is obsolete from release 8.00; the restriction was finally
4152        removed: */        removed: */
4153    
# Line 4034  we set the flag only if there is a liter Line 4155  we set the flag only if there is a liter
4155        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4156    
4157        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4158        /*--------------------------------------------------------------------*/        /*--------------------------------------------------------------------*/
4159    
4160        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4161          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 4169  we set the flag only if there is a liter Line 4290  we set the flag only if there is a liter
4290            {            {
4291            /* 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
4292            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
4293            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4294              integer type when available, otherwise double. */
4295    
4296            if (lengthptr != NULL)            if (lengthptr != NULL)
4297              {              {
4298              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4299              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4300                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4301                        (INT64_OR_DOUBLE)INT_MAX ||
4302                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4303                {                {
4304                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4221  we set the flag only if there is a liter Line 4344  we set the flag only if there is a liter
4344          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
4345          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
4346          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
4347          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4348            a 64-bit integer type when available, otherwise double. */
4349    
4350          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4351            {            {
4352            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4353                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4354            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4355                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4356                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4357                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4358              {              {
4359              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4306  we set the flag only if there is a liter Line 4430  we set the flag only if there is a liter
4430            uschar *scode = bracode;            uschar *scode = bracode;
4431            do            do
4432              {              {
4433              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4434                {                {
4435                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4436                break;                break;
# Line 4349  we set the flag only if there is a liter Line 4473  we set the flag only if there is a liter
4473      if (possessive_quantifier)      if (possessive_quantifier)
4474        {        {
4475        int len;        int len;
4476    
4477        if (*tempcode == OP_TYPEEXACT)        if (*tempcode == OP_TYPEEXACT)
4478          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4479            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4480    
4481        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4482          {          {
4483          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
# Line 4361  we set the flag only if there is a liter Line 4485  we set the flag only if there is a liter
4485          if (utf8 && tempcode[-1] >= 0xc0)          if (utf8 && tempcode[-1] >= 0xc0)
4486            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];            tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4487  #endif  #endif
4488          }          }
4489    
4490        len = code - tempcode;        len = code - tempcode;
4491        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4492          {          {
# Line 4381  we set the flag only if there is a liter Line 4505  we set the flag only if there is a liter
4505          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4506          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4507    
4508            /* Because we are moving code along, we must ensure that any
4509            pending recursive references are updated. */
4510    
4511          default:          default:
4512            *code = OP_END;
4513            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4514          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4515          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4516          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4417  we set the flag only if there is a liter Line 4546  we set the flag only if there is a liter
4546    
4547      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4548    
4549      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4550             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4551        {        {
4552        int i, namelen;        int i, namelen;
4553          int arglen = 0;
4554        const char *vn = verbnames;        const char *vn = verbnames;
4555        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4556          const uschar *arg = NULL;
4557        previous = NULL;        previous = NULL;
4558        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4559          namelen = ptr - name;
4560    
4561        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
4562          {          {
4563          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4564          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4565              || *ptr == '_') ptr++;
4566            arglen = ptr - arg;
4567          }          }
4568    
4569        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4570          {          {
4571          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4572          goto FAILED;          goto FAILED;
4573          }          }
4574        namelen = ptr - name;  
4575          /* Scan the table of verb names */
4576    
4577        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4578          {          {
4579          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4580              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4581            {            {
4582            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4583            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4584            break;            if (verbs[i].op == OP_ACCEPT)
4585                {
4586                open_capitem *oc;
4587                cd->had_accept = TRUE;
4588                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4589                  {
4590                  *code++ = OP_CLOSE;
4591                  PUT2INC(code, 0, oc->number);
4592                  }
4593                }
4594    
4595              /* Handle the cases with/without an argument */
4596    
4597              if (arglen == 0)
4598                {
4599                if (verbs[i].op < 0)   /* Argument is mandatory */
4600                  {
4601                  *errorcodeptr = ERR66;
4602                  goto FAILED;
4603                  }
4604                *code++ = verbs[i].op;
4605                }
4606    
4607              else
4608                {
4609                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4610                  {
4611                  *errorcodeptr = ERR59;
4612                  goto FAILED;
4613                  }
4614                *code++ = verbs[i].op_arg;
4615                *code++ = arglen;
4616                memcpy(code, arg, arglen);
4617                code += arglen;
4618                *code++ = 0;
4619                }
4620    
4621              break;  /* Found verb, exit loop */
4622            }            }
4623    
4624          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4625          }          }
4626        if (i < verbcount) continue;  
4627        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4628          *errorcodeptr = ERR60;          /* Verb not recognized */
4629        goto FAILED;        goto FAILED;
4630        }        }
4631    
# Line 4603  we set the flag only if there is a liter Line 4781  we set the flag only if there is a liter
4781            }            }
4782    
4783          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4784          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4785            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4786            except they record that the reference was originally to a name. The
4787            information is used to check duplicate names. */
4788    
4789          slot = cd->name_table;          slot = cd->name_table;
4790          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 4799  we set the flag only if there is a liter
4799            {            {
4800            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4801            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4802              code[1+LINK_SIZE]++;
4803            }            }
4804    
4805          /* 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 4808  we set the flag only if there is a liter
4808                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4809            {            {
4810            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4811              code[1+LINK_SIZE]++;
4812            }            }
4813    
4814          /* 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 5001  we set the flag only if there is a liter
5001                }                }
5002              }              }
5003    
5004            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5005              alphabetical order. Duplicate names for different numbers are
5006              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5007              number are always OK. (An existing number can be re-used if (?|
5008              appears in the pattern.) In either event, a duplicate name results in
5009              a duplicate entry in the table, even if the number is the same. This
5010              is because the number of names, and hence the table size, is computed
5011              in the pre-compile, and it affects various numbers and pointers which
5012              would all have to be modified, and the compiled code moved down, if
5013              duplicates with the same number were omitted from the table. This
5014              doesn't seem worth the hassle. However, *different* names for the
5015              same number are not permitted. */
5016    
5017            else            else
5018              {              {
5019                BOOL dupname = FALSE;
5020              slot = cd->name_table;              slot = cd->name_table;
5021    
5022              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5023                {                {
5024                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 5026  we set the flag only if there is a liter
5026                  {                  {
5027                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5028                    {                    {
5029                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5030                          (options & PCRE_DUPNAMES) == 0)
5031                      {                      {
5032                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5033                      goto FAILED;                      goto FAILED;
5034                      }                      }
5035                      else dupname = TRUE;
5036                    }                    }
5037                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5038                  }                  }
5039    
5040                  /* Make space in the table and break the loop for an earlier
5041                  name. For a duplicate or later name, carry on. We do this for
5042                  duplicates so that in the simple case (when ?(| is not used) they
5043                  are in order of their numbers. */
5044    
5045                if (crc < 0)                if (crc < 0)
5046                  {                  {
5047                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5048                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5049                  break;                  break;
5050                  }                  }
5051    
5052                  /* Continue the loop for a later or duplicate name */
5053    
5054                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5055                }                }
5056    
5057                /* For non-duplicate names, check for a duplicate number before
5058                adding the new name. */
5059    
5060                if (!dupname)
5061                  {
5062                  uschar *cslot = cd->name_table;
5063                  for (i = 0; i < cd->names_found; i++)
5064                    {
5065                    if (cslot != slot)
5066                      {
5067                      if (GET2(cslot, 0) == cd->bracount + 1)
5068                        {
5069                        *errorcodeptr = ERR65;
5070                        goto FAILED;
5071                        }
5072                      }
5073                    else i--;
5074                    cslot += cd->name_entry_size;
5075                    }
5076                  }
5077    
5078              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5079              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5080              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5081              }              }
5082            }            }
5083    
5084          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5085            encountered. */
5086    
         ptr++;                    /* Move past > or ' */  
5087          cd->names_found++;          cd->names_found++;
5088            ptr++;                    /* Move past > or ' */
5089          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5090    
5091    
# Line 5025  we set the flag only if there is a liter Line 5254  we set the flag only if there is a liter
5254            if (lengthptr == NULL)            if (lengthptr == NULL)
5255              {              {
5256              *code = OP_END;              *code = OP_END;
5257              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5258                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5259    
5260              /* Forward reference */              /* Forward reference */
5261    
# Line 5037  we set the flag only if there is a liter Line 5267  we set the flag only if there is a liter
5267                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5268                  goto FAILED;                  goto FAILED;
5269                  }                  }
5270    
5271                  /* Fudge the value of "called" so that when it is inserted as an
5272                  offset below, what it actually inserted is the reference number
5273                  of the group. */
5274    
5275                called = cd->start_code + recno;                called = cd->start_code + recno;
5276                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5277                }                }
# Line 5046  we set the flag only if there is a liter Line 5281  we set the flag only if there is a liter
5281              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5282    
5283              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5284                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5285                {                {
5286                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5287                goto FAILED;                goto FAILED;
# Line 5141  we set the flag only if there is a liter Line 5376  we set the flag only if there is a liter
5376              {              {
5377              cd->external_options = newoptions;              cd->external_options = newoptions;
5378              }              }
5379           else            else
5380              {              {
5381              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5382                {                {
# Line 5173  we set the flag only if there is a liter Line 5408  we set the flag only if there is a liter
5408          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5409        }       /* End of (? handling */        }       /* End of (? handling */
5410    
5411      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5412      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5413      brackets. */      brackets. */
5414    
5415      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5478  we set the flag only if there is a liter Line 5713  we set the flag only if there is a liter
5713    
5714        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5715          {          {
5716            open_capitem *oc;
5717          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5718    
5719          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 5723  we set the flag only if there is a liter
5723          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5724          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5725          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5726    
5727            /* Check to see if this back reference is recursive, that it, it
5728            is inside the group that it references. A flag is set so that the
5729            group can be made atomic. */
5730    
5731            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5732              {
5733              if (oc->number == recno)
5734                {
5735                oc->flag = TRUE;
5736                break;
5737                }
5738              }
5739          }          }
5740    
5741        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5669  uschar *code = *codeptr; Line 5918  uschar *code = *codeptr;
5918  uschar *last_branch = code;  uschar *last_branch = code;
5919  uschar *start_bracket = code;  uschar *start_bracket = code;
5920  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5921    open_capitem capitem;
5922    int capnumber = 0;
5923  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5924  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5925  int length;  int length;
5926  int orig_bracount;  int orig_bracount;
5927  int max_bracount;  int max_bracount;
5928    int old_external_options = cd->external_options;
5929  branch_chain bc;  branch_chain bc;
5930    
5931  bc.outer = bcptr;  bc.outer = bcptr;
5932  bc.current = code;  bc.current_branch = code;
5933    
5934  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
5935    
# Line 5695  the code that abstracts option settings Line 5947  the code that abstracts option settings
5947  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
5948  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. */
5949    
5950    /* If this is a capturing subpattern, add to the chain of open capturing items
5951    so that we can detect them if (*ACCEPT) is encountered. This is also used to
5952    detect groups that contain recursive back references to themselves. */
5953    
5954    if (*code == OP_CBRA)
5955      {
5956      capnumber = GET2(code, 1 + LINK_SIZE);
5957      capitem.number = capnumber;
5958      capitem.next = cd->open_caps;
5959      capitem.flag = FALSE;
5960      cd->open_caps = &capitem;
5961      }
5962    
5963  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5964    
5965  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5739  for (;;) Line 6004  for (;;)
6004      return FALSE;      return FALSE;
6005      }      }
6006    
6007      /* If the external options have changed during this branch, it means that we
6008      are at the top level, and a leading option setting has been encountered. We
6009      need to re-set the original option values to take account of this so that,
6010      during the pre-compile phase, we know to allow for a re-set at the start of
6011      subsequent branches. */
6012    
6013      if (old_external_options != cd->external_options)
6014        oldims = cd->external_options & PCRE_IMS;
6015    
6016    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6017    has fewer than the rest. */    has fewer than the rest. */
6018    
# Line 5789  for (;;) Line 6063  for (;;)
6063    
6064      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6065      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
6066      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
6067        because there may be forward references that we can't check here. Set a
6068        flag to cause another lookbehind check at the end. Why not do it all at the
6069        end? Because common, erroneous checks are picked up here and the offset of
6070        the problem can be shown. */
6071    
6072      if (lookbehind)      if (lookbehind)
6073        {        {
6074        int fixed_length;        int fixed_length;
6075        *code = OP_END;        *code = OP_END;
6076        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
6077        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6078        if (fixed_length < 0)        if (fixed_length == -3)
6079            {
6080            cd->check_lookbehind = TRUE;
6081            }
6082          else if (fixed_length < 0)
6083          {          {
6084          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6085          *ptrptr = ptr;          *ptrptr = ptr;
6086          return FALSE;          return FALSE;
6087          }          }
6088        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6089        }        }
6090      }      }
6091    
# Line 5837  for (;;) Line 6119  for (;;)
6119      PUT(code, 1, code - start_bracket);      PUT(code, 1, code - start_bracket);
6120      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6121    
6122      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6123        recursive back references. If so, we must wrap it in atomic brackets.
6124        In any event, remove the block from the chain. */
6125    
6126        if (capnumber > 0)
6127          {
6128          if (cd->open_caps->flag)
6129            {
6130            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6131              code - start_bracket);
6132            *start_bracket = OP_ONCE;
6133            code += 1 + LINK_SIZE;
6134            PUT(start_bracket, 1, code - start_bracket);
6135            *code = OP_KET;
6136            PUT(code, 1, code - start_bracket);
6137            code += 1 + LINK_SIZE;
6138            length += 2 + 2*LINK_SIZE;
6139            }
6140          cd->open_caps = cd->open_caps->next;
6141          }
6142    
6143        /* Reset options if needed. */
6144    
6145      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6146        {        {
# Line 5886  for (;;) Line 6189  for (;;)
6189      {      {
6190      *code = OP_ALT;      *code = OP_ALT;
6191      PUT(code, 1, code - last_branch);      PUT(code, 1, code - last_branch);
6192      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6193      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6194      }      }
6195    
# Line 6033  do { Line 6336  do {
6336       switch (*scode)       switch (*scode)
6337         {         {
6338         case OP_CREF:         case OP_CREF:
6339           case OP_NCREF:
6340         case OP_RREF:         case OP_RREF:
6341           case OP_NRREF:
6342         case OP_DEF:         case OP_DEF:
6343         return FALSE;         return FALSE;
6344    
# Line 6202  int length = 1;  /* For final END opcode Line 6507  int length = 1;  /* For final END opcode
6507  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6508  int errorcode = 0;  int errorcode = 0;
6509  int skipatstart = 0;  int skipatstart = 0;
6510  #ifdef SUPPORT_UTF8  BOOL utf8 = (options & PCRE_UTF8) != 0;
 BOOL utf8;  
 #endif  
6511  size_t size;  size_t size;
6512  uschar *code;  uschar *code;
6513  const uschar *codestart;  const uschar *codestart;
# Line 6301  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6604  while (ptr[skipatstart] == CHAR_LEFT_PAR
6604  /* 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. */
6605    
6606  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
 utf8 = (options & PCRE_UTF8) != 0;  
6607  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6608       (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6609    {    {
6610    errorcode = ERR44;    errorcode = ERR44;
6611    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
6612    }    }
6613  #else  #else
6614  if ((options & PCRE_UTF8) != 0)  if (utf8)
6615    {    {
6616    errorcode = ERR32;    errorcode = ERR32;
6617    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
# Line 6398  cd->end_pattern = (const uschar *)(patte Line 6700  cd->end_pattern = (const uschar *)(patte
6700  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6701  cd->external_options = options;  cd->external_options = options;
6702  cd->external_flags = 0;  cd->external_flags = 0;
6703    cd->open_caps = NULL;
6704    
6705  /* 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
6706  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 6775  cd->start_code = codestart;
6775  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6776  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6777  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6778    cd->check_lookbehind = FALSE;
6779    cd->open_caps = NULL;
6780    
6781  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6782  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 6802  if debugging, leave the test till after
6802    
6803  *code++ = OP_END;  *code++ = OP_END;
6804    
6805  #ifndef DEBUG  #ifndef PCRE_DEBUG
6806  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
6807  #endif  #endif
6808    
# Line 6510  while (errorcode == 0 && cd->hwm > cwork Line 6815  while (errorcode == 0 && cd->hwm > cwork
6815    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6816    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6817    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6818    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6819    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6820      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6821    }    }
# Line 6520  subpattern. */ Line 6825  subpattern. */
6825    
6826  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6827    
6828    /* If there were any lookbehind assertions that contained OP_RECURSE
6829    (recursions or subroutine calls), a flag is set for them to be checked here,
6830    because they may contain forward references. Actual recursions can't be fixed
6831    length, but subroutine calls can. It is done like this so that those without
6832    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6833    exceptional ones forgo this. We scan the pattern to check that they are fixed
6834    length, and set their lengths. */
6835    
6836    if (cd->check_lookbehind)
6837      {
6838      uschar *cc = (uschar *)codestart;
6839    
6840      /* Loop, searching for OP_REVERSE items, and process those that do not have
6841      their length set. (Actually, it will also re-process any that have a length
6842      of zero, but that is a pathological case, and it does no harm.) When we find
6843      one, we temporarily terminate the branch it is in while we scan it. */
6844    
6845      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6846           cc != NULL;
6847           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6848        {
6849        if (GET(cc, 1) == 0)
6850          {
6851          int fixed_length;
6852          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6853          int end_op = *be;
6854          *be = OP_END;
6855          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6856          *be = end_op;
6857          DPRINTF(("fixed length = %d\n", fixed_length));
6858          if (fixed_length < 0)
6859            {
6860            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6861            break;
6862            }
6863          PUT(cc, 1, fixed_length);
6864          }
6865        cc += 1 + LINK_SIZE;
6866        }
6867      }
6868    
6869  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6870    
6871  if (errorcode != 0)  if (errorcode != 0)
# Line 6580  if (reqbyte >= 0 && Line 6926  if (reqbyte >= 0 &&
6926  /* 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
6927  case when building a production library. */  case when building a production library. */
6928    
6929  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
6930  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6931    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6932    
# Line 6618  if (code - codestart > length) Line 6963  if (code - codestart > length)
6963    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6964    return NULL;    return NULL;
6965    }    }
6966  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
6967    
6968  return (pcre *)re;  return (pcre *)re;
6969  }  }

Legend:
Removed from v.438  
changed lines
  Added in v.513

  ViewVC Help
Powered by ViewVC 1.1.5