/[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 200 by ph10, Wed Aug 1 09:10:40 2007 UTC revision 231 by ph10, Tue Sep 11 11:15:33 2007 UTC
# Line 65  used by pcretest. DEBUG is not defined w Line 65  used by pcretest. DEBUG is not defined w
65    
66  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 133  static const short int escapes[] = { Line 140  static const short int escapes[] = {
140  #endif  #endif
141    
142    
143    /* Table of special "verbs" like (*PRUNE) */
144    
145    typedef struct verbitem {
146      const char *name;
147      int   len;
148      int   op;
149    } verbitem;
150    
151    static verbitem verbs[] = {
152      { "ACCEPT", 6, OP_ACCEPT },
153      { "COMMIT", 6, OP_COMMIT },
154      { "F",      1, OP_FAIL },
155      { "FAIL",   4, OP_FAIL },
156      { "PRUNE",  5, OP_PRUNE },
157      { "SKIP",   4, OP_SKIP  },
158      { "THEN",   4, OP_THEN  }
159    };
160    
161    static int verbcount = sizeof(verbs)/sizeof(verbitem);
162    
163    
164  /* Tables of names of POSIX character classes and their lengths. The list is  /* Tables of names of POSIX character classes and their lengths. The list is
165  terminated by a zero length entry. The first three must be alpha, lower, upper,  terminated by a zero length entry. The first three must be alpha, lower, upper,
166  as this is assumed for handling case independence. */  as this is assumed for handling case independence. */
# Line 206  static const char *error_texts[] = { Line 234  static const char *error_texts[] = {
234    "missing ) after comment",    "missing ) after comment",
235    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply",  /** DEAD **/
236    /* 20 */    /* 20 */
237    "regular expression too large",    "regular expression is too large",
238    "failed to get memory",    "failed to get memory",
239    "unmatched parentheses",    "unmatched parentheses",
240    "internal error: code overflow",    "internal error: code overflow",
# Line 242  static const char *error_texts[] = { Line 270  static const char *error_texts[] = {
270    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",
271    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",
272    /* 50 */    /* 50 */
273    "repeated subpattern is too long",    "repeated subpattern is too long",    /** DEAD **/
274    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)",
275    "internal error: overran compiling workspace",    "internal error: overran compiling workspace",
276    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found",
# Line 251  static const char *error_texts[] = { Line 279  static const char *error_texts[] = {
279    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed",
280    "inconsistent NEWLINE options",    "inconsistent NEWLINE options",
281    "\\g is not followed by a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced name or an optionally braced non-zero number",
282    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number",
283      "(*VERB) with an argument is not supported",
284      /* 60 */
285      "(*VERB) not recognized",
286      "number is too big"
287  };  };
288    
289    
# Line 408  Arguments: Line 440  Arguments:
440    
441  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
442                   negative => a special escape sequence                   negative => a special escape sequence
443                   on error, errorptr is set                   on error, errorcodeptr is set
444  */  */
445    
446  static int  static int
# Line 493  else Line 525  else
525      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
526        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - '0';
527    
528        if (c < 0)
529          {
530          *errorcodeptr = ERR61;
531          break;
532          }
533    
534      if (c == 0 || (braced && *(++ptr) != '}'))      if (c == 0 || (braced && *(++ptr) != '}'))
535        {        {
536        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
537        return 0;        break;
538        }        }
539    
540      if (negated)      if (negated)
# Line 504  else Line 542  else
542        if (c > bracount)        if (c > bracount)
543          {          {
544          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
545          return 0;          break;
546          }          }
547        c = bracount - (c - 1);        c = bracount - (c - 1);
548        }        }
# Line 533  else Line 571  else
571        c -= '0';        c -= '0';
572        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
573          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
574          if (c < 0)
575            {
576            *errorcodeptr = ERR61;
577            break;
578            }
579        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
580          {          {
581          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 628  else Line 671  else
671      if (c == 0)      if (c == 0)
672        {        {
673        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
674        return 0;        break;
675        }        }
676    
677  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
# Line 907  for (; *ptr != 0; ptr++) Line 950  for (; *ptr != 0; ptr++)
950      {      {
951      while (*(++ptr) != ']')      while (*(++ptr) != ']')
952        {        {
953          if (*ptr == 0) return -1;
954        if (*ptr == '\\')        if (*ptr == '\\')
955          {          {
956          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) return -1;
# Line 934  for (; *ptr != 0; ptr++) Line 978  for (; *ptr != 0; ptr++)
978    /* An opening parens must now be a real metacharacter */    /* An opening parens must now be a real metacharacter */
979    
980    if (*ptr != '(') continue;    if (*ptr != '(') continue;
981    if (ptr[1] != '?')    if (ptr[1] != '?' && ptr[1] != '*')
982      {      {
983      count++;      count++;
984      if (name == NULL && count == lorn) return count;      if (name == NULL && count == lorn) return count;
# Line 1062  for (;;) Line 1106  for (;;)
1106    {    {
1107    int d;    int d;
1108    register int op = *cc;    register int op = *cc;
   
1109    switch (op)    switch (op)
1110      {      {
1111      case OP_CBRA:      case OP_CBRA:
# Line 1151  for (;;) Line 1194  for (;;)
1194    
1195      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1196      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1197        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1198      cc += 4;      cc += 4;
1199      break;      break;
1200    
# Line 1259  for (;;) Line 1303  for (;;)
1303      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1304      }      }
1305    
1306    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* Otherwise, we can get the item's length from the table, except that for
1307    a multi-byte character. The length in the table is a minimum, so we have to    repeated character types, we have to test for \p and \P, which have an extra
1308    arrange to skip the extra bytes. */    two bytes of parameters. */
1309    
1310    else    else
1311      {      {
1312        switch(c)
1313          {
1314          case OP_TYPESTAR:
1315          case OP_TYPEMINSTAR:
1316          case OP_TYPEPLUS:
1317          case OP_TYPEMINPLUS:
1318          case OP_TYPEQUERY:
1319          case OP_TYPEMINQUERY:
1320          case OP_TYPEPOSSTAR:
1321          case OP_TYPEPOSPLUS:
1322          case OP_TYPEPOSQUERY:
1323          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1324          break;
1325    
1326          case OP_TYPEUPTO:
1327          case OP_TYPEMINUPTO:
1328          case OP_TYPEEXACT:
1329          case OP_TYPEPOSUPTO:
1330          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1331          break;
1332          }
1333    
1334        /* Add in the fixed length from the table */
1335    
1336      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1337    
1338      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1339      a multi-byte character. The length in the table is a minimum, so we have to
1340      arrange to skip the extra bytes. */
1341    
1342  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1343      if (utf8) switch(c)      if (utf8) switch(c)
1344        {        {
# Line 1323  for (;;) Line 1396  for (;;)
1396    
1397    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1398    
1399    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
1400    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
1401    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1402    
1403    else    else
1404      {      {
1405        switch(c)
1406          {
1407          case OP_TYPESTAR:
1408          case OP_TYPEMINSTAR:
1409          case OP_TYPEPLUS:
1410          case OP_TYPEMINPLUS:
1411          case OP_TYPEQUERY:
1412          case OP_TYPEMINQUERY:
1413          case OP_TYPEPOSSTAR:
1414          case OP_TYPEPOSPLUS:
1415          case OP_TYPEPOSQUERY:
1416          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1417          break;
1418    
1419          case OP_TYPEPOSUPTO:
1420          case OP_TYPEUPTO:
1421          case OP_TYPEMINUPTO:
1422          case OP_TYPEEXACT:
1423          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1424          break;
1425          }
1426    
1427        /* Add in the fixed length from the table */
1428    
1429      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1430    
1431        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1432        by a multi-byte character. The length in the table is a minimum, so we have
1433        to arrange to skip the extra bytes. */
1434    
1435  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1436      if (utf8) switch(c)      if (utf8) switch(c)
1437        {        {
# Line 1402  for (code = first_significant_code(code Line 1503  for (code = first_significant_code(code
1503    
1504    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1505    
1506    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1507      {      {
1508      BOOL empty_branch;      BOOL empty_branch;
1509      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1426  for (code = first_significant_code(code Line 1527  for (code = first_significant_code(code
1527    
1528    switch (c)    switch (c)
1529      {      {
1530      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1531        cannot be represented just by a bit map. This includes negated single
1532        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1533        actual length is stored in the compiled code, so we must update "code"
1534        here. */
1535    
1536  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1537      case OP_XCLASS:      case OP_XCLASS:
1538      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1539      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1540  #endif  #endif
1541    
# Line 1492  for (code = first_significant_code(code Line 1597  for (code = first_significant_code(code
1597      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1598      return FALSE;      return FALSE;
1599    
1600        /* These are going to continue, as they may be empty, but we have to
1601        fudge the length for the \p and \P cases. */
1602    
1603        case OP_TYPESTAR:
1604        case OP_TYPEMINSTAR:
1605        case OP_TYPEPOSSTAR:
1606        case OP_TYPEQUERY:
1607        case OP_TYPEMINQUERY:
1608        case OP_TYPEPOSQUERY:
1609        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1610        break;
1611    
1612        /* Same for these */
1613    
1614        case OP_TYPEUPTO:
1615        case OP_TYPEMINUPTO:
1616        case OP_TYPEPOSUPTO:
1617        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1618        break;
1619    
1620      /* End of branch */      /* End of branch */
1621    
1622      case OP_KET:      case OP_KET:
# Line 1654  adjust_recurse(uschar *group, int adjust Line 1779  adjust_recurse(uschar *group, int adjust
1779    uschar *save_hwm)    uschar *save_hwm)
1780  {  {
1781  uschar *ptr = group;  uschar *ptr = group;
1782    
1783  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1784    {    {
1785    int offset;    int offset;
# Line 2258  for (;; ptr++) Line 2384  for (;; ptr++)
2384      */      */
2385    
2386      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2387    
2388        /* Paranoid check for integer overflow */
2389    
2390        if (OFLOW_MAX - *lengthptr < code - last_code)
2391          {
2392          *errorcodeptr = ERR20;
2393          goto FAILED;
2394          }
2395    
2396      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2397      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2398    
# Line 2370  for (;; ptr++) Line 2505  for (;; ptr++)
2505      *ptrptr = ptr;      *ptrptr = ptr;
2506      if (lengthptr != NULL)      if (lengthptr != NULL)
2507        {        {
2508          if (OFLOW_MAX - *lengthptr < code - last_code)
2509            {
2510            *errorcodeptr = ERR20;
2511            goto FAILED;
2512            }
2513        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2514        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2515        }        }
# Line 2432  for (;; ptr++) Line 2572  for (;; ptr++)
2572        goto FAILED;        goto FAILED;
2573        }        }
2574    
2575      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
2576        if the first few characters (either before or after ^) are \Q\E or \E we
2577        skip them too. This makes for compatibility with Perl. */
2578    
2579      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2580        for (;;)
2581        {        {
       negate_class = TRUE;  
2582        c = *(++ptr);        c = *(++ptr);
2583        }        if (c == '\\')
2584      else          {
2585        {          if (ptr[1] == 'E') ptr++;
2586        negate_class = FALSE;            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2587                else break;
2588            }
2589          else if (!negate_class && c == '^')
2590            negate_class = TRUE;
2591          else break;
2592        }        }
2593    
2594      /* Keep a count of chars with values < 256 so that we can optimize the case      /* Keep a count of chars with values < 256 so that we can optimize the case
# Line 2582  for (;; ptr++) Line 2729  for (;; ptr++)
2729        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
2730        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
2731        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2732        to or into the one we are building. We assume they have more than one        to 'or' into the one we are building. We assume they have more than one
2733        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2734    
2735        if (c == '\\')        if (c == '\\')
# Line 2602  for (;; ptr++) Line 2749  for (;; ptr++)
2749            else inescq = TRUE;            else inescq = TRUE;
2750            continue;            continue;
2751            }            }
2752            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2753    
2754          if (c < 0)          if (c < 0)
2755            {            {
# Line 2828  for (;; ptr++) Line 2976  for (;; ptr++)
2976    
2977        oldptr = ptr;        oldptr = ptr;
2978    
2979          /* Remember \r or \n */
2980    
2981          if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
2982    
2983          /* Check for range */
2984    
2985        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == '-')
2986          {          {
2987          int d;          int d;
# Line 2895  for (;; ptr++) Line 3049  for (;; ptr++)
3049    
3050          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3051    
3052            /* Remember \r or \n */
3053    
3054            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3055    
3056          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
3057          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3058          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3048  for (;; ptr++) Line 3206  for (;; ptr++)
3206        goto FAILED;        goto FAILED;
3207        }        }
3208    
3209    
3210    /* This code has been disabled because it would mean that \s counts as
3211    an explicit \r or \n reference, and that's not really what is wanted. Now
3212    we set the flag only if there is a literal "\r" or "\n" in the class. */
3213    
3214    #if 0
3215        /* Remember whether \r or \n are in this class */
3216    
3217        if (negate_class)
3218          {
3219          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3220          }
3221        else
3222          {
3223          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3224          }
3225    #endif
3226    
3227    
3228      /* 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
3229      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
3230      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
3231      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3232      single-bytes only. This is an historical hangover. Maybe one day we can  
3233      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3234        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3235        operate on single-bytes only. This is an historical hangover. Maybe one day
3236        we can tidy these opcodes to handle multi-byte characters.
3237    
3238      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
3239      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 3063  for (;; ptr++) Line 3243  for (;; ptr++)
3243      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3244    
3245  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3246      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3247            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3248  #else  #else
3249      if (class_charcount == 1)      if (class_charcount == 1)
3250  #endif  #endif
# Line 3338  for (;; ptr++) Line 3516  for (;; ptr++)
3516        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3517        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3518    
3519        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3520    
3521        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3522    
# Line 3488  for (;; ptr++) Line 3666  for (;; ptr++)
3666        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3667        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3668    
3669        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3670    
3671        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3672          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3524  for (;; ptr++) Line 3702  for (;; ptr++)
3702          goto FAILED;          goto FAILED;
3703          }          }
3704    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
3705        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
3706        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
3707        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 3620  for (;; ptr++) Line 3790  for (;; ptr++)
3790          if (repeat_min > 1)          if (repeat_min > 1)
3791            {            {
3792            /* 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
3793            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
3794              potential integer overflow. */
3795    
3796            if (lengthptr != NULL)            if (lengthptr != NULL)
3797              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
3798                int delta = (repeat_min - 1)*length_prevgroup;
3799                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3800                                                                (double)INT_MAX ||
3801                    OFLOW_MAX - *lengthptr < delta)
3802                  {
3803                  *errorcodeptr = ERR20;
3804                  goto FAILED;
3805                  }
3806                *lengthptr += delta;
3807                }
3808    
3809            /* This is compiling for real */            /* This is compiling for real */
3810    
# Line 3661  for (;; ptr++) Line 3842  for (;; ptr++)
3842          /* 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
3843          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
3844          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
3845          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3846            paranoid checks to avoid integer overflow. */
3847    
3848          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
3849            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
3850              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3851                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3852              if ((double)repeat_max *
3853                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3854                      > (double)INT_MAX ||
3855                  OFLOW_MAX - *lengthptr < delta)
3856                {
3857                *errorcodeptr = ERR20;
3858                goto FAILED;
3859                }
3860              *lengthptr += delta;
3861              }
3862    
3863          /* This is compiling for real */          /* This is compiling for real */
3864    
# Line 3817  for (;; ptr++) Line 4010  for (;; ptr++)
4010      /* ===================================================================*/      /* ===================================================================*/
4011      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4012      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4013      parenthesis forms. First deal with the specials; all are introduced by ?,      parenthesis forms.  */
     and the appearance of any of them means that this is not a capturing  
     group. */  
4014    
4015      case '(':      case '(':
4016      newoptions = options;      newoptions = options;
# Line 3828  for (;; ptr++) Line 4019  for (;; ptr++)
4019      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4020      reset_bracount = FALSE;      reset_bracount = FALSE;
4021    
4022      if (*(++ptr) == '?')      /* First deal with various "verbs" that can be introduced by '*'. */
4023    
4024        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4025          {
4026          int i, namelen;
4027          const uschar *name = ++ptr;
4028          previous = NULL;
4029          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4030          if (*ptr == ':')
4031            {
4032            *errorcodeptr = ERR59;   /* Not supported */
4033            goto FAILED;
4034            }
4035          if (*ptr != ')')
4036            {
4037            *errorcodeptr = ERR60;
4038            goto FAILED;
4039            }
4040          namelen = ptr - name;
4041          for (i = 0; i < verbcount; i++)
4042            {
4043            if (namelen == verbs[i].len &&
4044                strncmp((char *)name, verbs[i].name, namelen) == 0)
4045              {
4046              *code = verbs[i].op;
4047              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4048              break;
4049              }
4050            }
4051          if (i < verbcount) continue;
4052          *errorcodeptr = ERR60;
4053          goto FAILED;
4054          }
4055    
4056        /* Deal with the extended parentheses; all are introduced by '?', and the
4057        appearance of any of them means that this is not a capturing group. */
4058    
4059        else if (*ptr == '?')
4060        {        {
4061        int i, set, unset, namelen;        int i, set, unset, namelen;
4062        int *optset;        int *optset;
# Line 4070  for (;; ptr++) Line 4298  for (;; ptr++)
4298    
4299          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4300          case '!':                 /* Negative lookahead */          case '!':                 /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4301          ptr++;          ptr++;
4302            if (*ptr == ')')          /* Optimize (?!) */
4303              {
4304              *code++ = OP_FAIL;
4305              previous = NULL;
4306              continue;
4307              }
4308            bravalue = OP_ASSERT_NOT;
4309          break;          break;
4310    
4311    
# Line 4429  for (;; ptr++) Line 4663  for (;; ptr++)
4663    
4664              case 'J':    /* Record that it changed in the external options */              case 'J':    /* Record that it changed in the external options */
4665              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
4666              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
4667              break;              break;
4668    
4669              case 'i': *optset |= PCRE_CASELESS; break;              case 'i': *optset |= PCRE_CASELESS; break;
# Line 4620  for (;; ptr++) Line 4854  for (;; ptr++)
4854        goto FAILED;        goto FAILED;
4855        }        }
4856    
4857      /* In the pre-compile phase, update the length by the length of the nested      /* In the pre-compile phase, update the length by the length of the group,
4858      group, less the brackets at either end. Then reduce the compiled code to      less the brackets at either end. Then reduce the compiled code to just a
4859      just the brackets so that it doesn't use much memory if it is duplicated by      set of non-capturing brackets so that it doesn't use much memory if it is
4860      a quantifier. */      duplicated by a quantifier.*/
4861    
4862      if (lengthptr != NULL)      if (lengthptr != NULL)
4863        {        {
4864          if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
4865            {
4866            *errorcodeptr = ERR20;
4867            goto FAILED;
4868            }
4869        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
4870        code++;        *code++ = OP_BRA;
4871        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
4872        *code++ = OP_KET;        *code++ = OP_KET;
4873        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
4874          break;    /* No need to waste time with special character handling */
4875        }        }
4876    
4877      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
4878    
4879      else code = tempcode;      code = tempcode;
4880    
4881      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
4882      relevant. */      relevant. */
# Line 4840  for (;; ptr++) Line 5080  for (;; ptr++)
5080      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5081      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5082    
5083        /* Remember if \r or \n were seen */
5084    
5085        if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')
5086          cd->external_flags |= PCRE_HASCRORLF;
5087    
5088      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5089      byte, set it from this character, but revert to none on a zero repeat.      byte, set it from this character, but revert to none on a zero repeat.
5090      Otherwise, leave the firstbyte value alone, and don't change it on a zero      Otherwise, leave the firstbyte value alone, and don't change it on a zero
# Line 5122  for (;;) Line 5367  for (;;)
5367      *ptrptr = ptr;      *ptrptr = ptr;
5368      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
5369      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
5370      if (lengthptr != NULL) *lengthptr += length;      if (lengthptr != NULL)
5371          {
5372          if (OFLOW_MAX - *lengthptr < length)
5373            {
5374            *errorcodeptr = ERR20;
5375            return FALSE;
5376            }
5377          *lengthptr += length;
5378          }
5379      return TRUE;      return TRUE;
5380      }      }
5381    
# Line 5431  real_pcre *re; Line 5684  real_pcre *re;
5684  int length = 1;  /* For final END opcode */  int length = 1;  /* For final END opcode */
5685  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
5686  int errorcode = 0;  int errorcode = 0;
5687    int skipatstart = 0;
5688  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5689  BOOL utf8;  BOOL utf8;
5690  #endif  #endif
# Line 5509  cd->fcc = tables + fcc_offset; Line 5763  cd->fcc = tables + fcc_offset;
5763  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
5764  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
5765    
5766    /* Check for global one-time settings at the start of the pattern, and remember
5767    the offset for later. */
5768    
5769    while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')
5770      {
5771      int newnl = 0;
5772      int newbsr = 0;
5773    
5774      if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)
5775        { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
5776      else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)
5777        { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
5778      else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)
5779        { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
5780      else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)
5781        { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
5782      else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)
5783        { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
5784    
5785      else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)
5786        { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
5787      else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)
5788        { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
5789    
5790      if (newnl != 0)
5791        options = (options & ~PCRE_NEWLINE_BITS) | newnl;
5792      else if (newbsr != 0)
5793        options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
5794      else break;
5795      }
5796    
5797    /* Check validity of \R options. */
5798    
5799    switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
5800      {
5801      case 0:
5802      case PCRE_BSR_ANYCRLF:
5803      case PCRE_BSR_UNICODE:
5804      break;
5805      default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
5806      }
5807    
5808  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
5809  current code allows for fixed one- or two-byte sequences, plus "any" and  current code allows for fixed one- or two-byte sequences, plus "any" and
5810  "anycrlf". */  "anycrlf". */
5811    
5812  switch (options & (PCRE_NEWLINE_CRLF | PCRE_NEWLINE_ANY))  switch (options & PCRE_NEWLINE_BITS)
5813    {    {
5814    case 0: newline = NEWLINE; break;   /* Compile-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
5815    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = '\r'; break;
5816    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = '\n'; break;
5817    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
# Line 5578  cd->hwm = cworkspace; Line 5874  cd->hwm = cworkspace;
5874  cd->start_pattern = (const uschar *)pattern;  cd->start_pattern = (const uschar *)pattern;
5875  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));
5876  cd->req_varyopt = 0;  cd->req_varyopt = 0;
 cd->nopartial = FALSE;  
5877  cd->external_options = options;  cd->external_options = options;
5878    cd->external_flags = 0;
5879    
5880  /* 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
5881  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 5587  been put into the cd block so that they Line 5883  been put into the cd block so that they
5883  found within the regex right at the beginning. Bringing initial option settings  found within the regex right at the beginning. Bringing initial option settings
5884  outside can help speed up starting point checks. */  outside can help speed up starting point checks. */
5885    
5886    ptr += skipatstart;
5887  code = cworkspace;  code = cworkspace;
5888  *code = OP_BRA;  *code = OP_BRA;
5889  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,
# Line 5617  if (re == NULL) Line 5914  if (re == NULL)
5914    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
5915    }    }
5916    
5917  /* Put in the magic number, and save the sizes, initial options, and character  /* Put in the magic number, and save the sizes, initial options, internal
5918  table pointer. NULL is used for the default character tables. The nullpad field  flags, and character table pointer. NULL is used for the default character
5919  is at the end; it's there to help in the case when a regex compiled on a system  tables. The nullpad field is at the end; it's there to help in the case when a
5920  with 4-byte pointers is run on another with 8-byte pointers. */  regex compiled on a system with 4-byte pointers is run on another with 8-byte
5921    pointers. */
5922    
5923  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
5924  re->size = size;  re->size = size;
5925  re->options = cd->external_options;  re->options = cd->external_options;
5926    re->flags = cd->external_flags;
5927  re->dummy1 = 0;  re->dummy1 = 0;
5928  re->first_byte = 0;  re->first_byte = 0;
5929  re->req_byte = 0;  re->req_byte = 0;
# Line 5649  codestart = cd->name_table + re->name_en Line 5948  codestart = cd->name_table + re->name_en
5948  cd->start_code = codestart;  cd->start_code = codestart;
5949  cd->hwm = cworkspace;  cd->hwm = cworkspace;
5950  cd->req_varyopt = 0;  cd->req_varyopt = 0;
5951  cd->nopartial = FALSE;  cd->had_accept = FALSE;
5952    
5953  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
5954  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
5955  of the function here. */  of the function here. */
5956    
5957  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern + skipatstart;
5958  code = (uschar *)codestart;  code = (uschar *)codestart;
5959  *code = OP_BRA;  *code = OP_BRA;
5960  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,
5961    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);
5962  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
5963  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
5964    re->flags = cd->external_flags;
5965    
5966  if (cd->nopartial) re->options |= PCRE_NOPARTIAL;  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */
5967    
5968  /* If not reached end of pattern on success, there's an excess bracket. */  /* If not reached end of pattern on success, there's an excess bracket. */
5969    
# Line 5734  if ((re->options & PCRE_ANCHORED) == 0) Line 6034  if ((re->options & PCRE_ANCHORED) == 0)
6034        int ch = firstbyte & 255;        int ch = firstbyte & 255;
6035        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&
6036           cd->fcc[ch] == ch)? ch : firstbyte;           cd->fcc[ch] == ch)? ch : firstbyte;
6037        re->options |= PCRE_FIRSTSET;        re->flags |= PCRE_FIRSTSET;
6038        }        }
6039      else if (is_startline(codestart, 0, cd->backref_map))      else if (is_startline(codestart, 0, cd->backref_map))
6040        re->options |= PCRE_STARTLINE;        re->flags |= PCRE_STARTLINE;
6041      }      }
6042    }    }
6043    
# Line 5751  if (reqbyte >= 0 && Line 6051  if (reqbyte >= 0 &&
6051    int ch = reqbyte & 255;    int ch = reqbyte & 255;
6052    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&
6053      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;
6054    re->options |= PCRE_REQCHSET;    re->flags |= PCRE_REQCHSET;
6055    }    }
6056    
6057  /* 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
# Line 5762  case when building a production library. Line 6062  case when building a production library.
6062  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6063    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6064    
6065  if (re->options != 0)  printf("Options=%08x\n", re->options);
   {  
   printf("%s%s%s%s%s%s%s%s%s\n",  
     ((re->options & PCRE_NOPARTIAL) != 0)? "nopartial " : "",  
     ((re->options & PCRE_ANCHORED) != 0)? "anchored " : "",  
     ((re->options & PCRE_CASELESS) != 0)? "caseless " : "",  
     ((re->options & PCRE_EXTENDED) != 0)? "extended " : "",  
     ((re->options & PCRE_MULTILINE) != 0)? "multiline " : "",  
     ((re->options & PCRE_DOTALL) != 0)? "dotall " : "",  
     ((re->options & PCRE_DOLLAR_ENDONLY) != 0)? "endonly " : "",  
     ((re->options & PCRE_EXTRA) != 0)? "extra " : "",  
     ((re->options & PCRE_UNGREEDY) != 0)? "ungreedy " : "");  
   }  
6066    
6067  if ((re->options & PCRE_FIRSTSET) != 0)  if ((re->flags & PCRE_FIRSTSET) != 0)
6068    {    {
6069    int ch = re->first_byte & 255;    int ch = re->first_byte & 255;
6070    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?
# Line 5785  if ((re->options & PCRE_FIRSTSET) != 0) Line 6073  if ((re->options & PCRE_FIRSTSET) != 0)
6073      else printf("First char = \\x%02x%s\n", ch, caseless);      else printf("First char = \\x%02x%s\n", ch, caseless);
6074    }    }
6075    
6076  if ((re->options & PCRE_REQCHSET) != 0)  if ((re->flags & PCRE_REQCHSET) != 0)
6077    {    {
6078    int ch = re->req_byte & 255;    int ch = re->req_byte & 255;
6079    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?

Legend:
Removed from v.200  
changed lines
  Added in v.231

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12