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

Diff of /code/trunk/pcre_study.c

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

revision 523 by ph10, Sun May 23 18:48:54 2010 UTC revision 697 by ph10, Mon Sep 19 12:28:24 2011 UTC
# Line 48  supporting functions. */ Line 48  supporting functions. */
48    
49  #include "pcre_internal.h"  #include "pcre_internal.h"
50    
51    #define SET_BIT(c) start_bits[c/8] |= (1 << (c&7))
52    
53  /* Returns from set_start_bits() */  /* Returns from set_start_bits() */
54    
55  enum { SSB_FAIL, SSB_DONE, SSB_CONTINUE };  enum { SSB_FAIL, SSB_DONE, SSB_CONTINUE, SSB_UNKNOWN };
56    
57    
58    
# Line 65  string of that length that matches. In U Line 66  string of that length that matches. In U
66  rather than bytes.  rather than bytes.
67    
68  Arguments:  Arguments:
69    code       pointer to start of group (the bracket)    code            pointer to start of group (the bracket)
70    startcode  pointer to start of the whole pattern    startcode       pointer to start of the whole pattern
71    options    the compiling options    options         the compiling options
72      int             RECURSE depth
73    
74  Returns:   the minimum length  Returns:   the minimum length
75             -1 if \C was encountered             -1 if \C in UTF-8 mode or (*ACCEPT) was encountered
76             -2 internal error (missing capturing bracket)             -2 internal error (missing capturing bracket)
77               -3 internal error (opcode not listed)
78  */  */
79    
80  static int  static int
81  find_minlength(const uschar *code, const uschar *startcode, int options)  find_minlength(const uschar *code, const uschar *startcode, int options,
82      int recurse_depth)
83  {  {
84  int length = -1;  int length = -1;
85  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 83  BOOL had_recurse = FALSE; Line 87  BOOL had_recurse = FALSE;
87  register int branchlength = 0;  register int branchlength = 0;
88  register uschar *cc = (uschar *)code + 1 + LINK_SIZE;  register uschar *cc = (uschar *)code + 1 + LINK_SIZE;
89    
90  if (*code == OP_CBRA || *code == OP_SCBRA) cc += 2;  if (*code == OP_CBRA || *code == OP_SCBRA ||
91        *code == OP_CBRAPOS || *code == OP_SCBRAPOS) cc += 2;
92    
93  /* Scan along the opcodes for this branch. If we get to the end of the  /* Scan along the opcodes for this branch. If we get to the end of the
94  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 117  for (;;) Line 122  for (;;)
122      case OP_SCBRA:      case OP_SCBRA:
123      case OP_BRA:      case OP_BRA:
124      case OP_SBRA:      case OP_SBRA:
125        case OP_CBRAPOS:
126        case OP_SCBRAPOS:
127        case OP_BRAPOS:
128        case OP_SBRAPOS:
129      case OP_ONCE:      case OP_ONCE:
130      d = find_minlength(cc, startcode, options);      d = find_minlength(cc, startcode, options, recurse_depth);
131      if (d < 0) return d;      if (d < 0) return d;
132      branchlength += d;      branchlength += d;
133      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
134      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
135      break;      break;
136    
137      /* Reached end of a branch; if it's a ket it is the end of a nested      /* ACCEPT makes things far too complicated; we have to give up. */
138      call. If it's ALT it is an alternation in a nested call. If it is  
139      END it's the end of the outer call. All can be handled by the same code. */      case OP_ACCEPT:
140        case OP_ASSERT_ACCEPT:
141        return -1;
142    
143        /* Reached end of a branch; if it's a ket it is the end of a nested
144        call. If it's ALT it is an alternation in a nested call. If it is END it's
145        the end of the outer call. All can be handled by the same code. If an
146        ACCEPT was previously encountered, use the length that was in force at that
147        time, and pass back the shortest ACCEPT length. */
148    
149      case OP_ALT:      case OP_ALT:
150      case OP_KET:      case OP_KET:
151      case OP_KETRMAX:      case OP_KETRMAX:
152      case OP_KETRMIN:      case OP_KETRMIN:
153        case OP_KETRPOS:
154      case OP_END:      case OP_END:
155      if (length < 0 || (!had_recurse && branchlength < length))      if (length < 0 || (!had_recurse && branchlength < length))
156        length = branchlength;        length = branchlength;
157      if (*cc != OP_ALT) return length;      if (op != OP_ALT) return length;
158      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
159      branchlength = 0;      branchlength = 0;
160      had_recurse = FALSE;      had_recurse = FALSE;
# Line 159  for (;;) Line 177  for (;;)
177      case OP_RREF:      case OP_RREF:
178      case OP_NRREF:      case OP_NRREF:
179      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
180      case OP_CALLOUT:      case OP_CALLOUT:
181      case OP_SOD:      case OP_SOD:
182      case OP_SOM:      case OP_SOM:
183      case OP_EOD:      case OP_EOD:
184      case OP_EODN:      case OP_EODN:
185      case OP_CIRC:      case OP_CIRC:
186        case OP_CIRCM:
187      case OP_DOLL:      case OP_DOLL:
188        case OP_DOLLM:
189      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
190      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
191      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 176  for (;;) Line 195  for (;;)
195    
196      case OP_BRAZERO:      case OP_BRAZERO:
197      case OP_BRAMINZERO:      case OP_BRAMINZERO:
198        case OP_BRAPOSZERO:
199      case OP_SKIPZERO:      case OP_SKIPZERO:
200      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
201      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 185  for (;;) Line 205  for (;;)
205      /* Handle literal characters and + repetitions */      /* Handle literal characters and + repetitions */
206    
207      case OP_CHAR:      case OP_CHAR:
208      case OP_CHARNC:      case OP_CHARI:
209      case OP_NOT:      case OP_NOT:
210        case OP_NOTI:
211      case OP_PLUS:      case OP_PLUS:
212        case OP_PLUSI:
213      case OP_MINPLUS:      case OP_MINPLUS:
214        case OP_MINPLUSI:
215      case OP_POSPLUS:      case OP_POSPLUS:
216        case OP_POSPLUSI:
217      case OP_NOTPLUS:      case OP_NOTPLUS:
218        case OP_NOTPLUSI:
219      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
220        case OP_NOTMINPLUSI:
221      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
222        case OP_NOTPOSPLUSI:
223      branchlength++;      branchlength++;
224      cc += 2;      cc += 2;
225  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 211  for (;;) Line 238  for (;;)
238      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
239    
240      case OP_EXACT:      case OP_EXACT:
241        case OP_EXACTI:
242      case OP_NOTEXACT:      case OP_NOTEXACT:
243        case OP_NOTEXACTI:
244      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
245      cc += 4;      cc += 4;
246  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 248  for (;;) Line 277  for (;;)
277      cc++;      cc++;
278      break;      break;
279    
280      /* "Any newline" might match two characters */      /* "Any newline" might match two characters, but it also might match just
281        one. */
282    
283      case OP_ANYNL:      case OP_ANYNL:
284      branchlength += 2;      branchlength += 1;
285      cc++;      cc++;
286      break;      break;
287    
# Line 336  for (;;) Line 366  for (;;)
366      that case we must set the minimum length to zero. */      that case we must set the minimum length to zero. */
367    
368      case OP_REF:      case OP_REF:
369        case OP_REFI:
370      if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
371        {        {
372        ce = cs = (uschar *)_pcre_find_bracket(startcode, utf8, GET2(cc, 1));        ce = cs = (uschar *)_pcre_find_bracket(startcode, utf8, GET2(cc, 1));
# Line 346  for (;;) Line 377  for (;;)
377          d = 0;          d = 0;
378          had_recurse = TRUE;          had_recurse = TRUE;
379          }          }
380        else d = find_minlength(cs, startcode, options);        else
381            {
382            d = find_minlength(cs, startcode, options, recurse_depth);
383            }
384        }        }
385      else d = 0;      else d = 0;
386      cc += 3;      cc += 3;
# Line 363  for (;;) Line 397  for (;;)
397        cc++;        cc++;
398        break;        break;
399    
400          case OP_CRPLUS:
401          case OP_CRMINPLUS:
402          min = 1;
403          cc++;
404          break;
405    
406        case OP_CRRANGE:        case OP_CRRANGE:
407        case OP_CRMINRANGE:        case OP_CRMINRANGE:
408        min = GET2(cc, 1);        min = GET2(cc, 1);
# Line 377  for (;;) Line 417  for (;;)
417      branchlength += min * d;      branchlength += min * d;
418      break;      break;
419    
420        /* We can easily detect direct recursion, but not mutual recursion. This is
421        caught by a recursion depth count. */
422    
423      case OP_RECURSE:      case OP_RECURSE:
424      cs = ce = (uschar *)startcode + GET(cc, 1);      cs = ce = (uschar *)startcode + GET(cc, 1);
425      if (cs == NULL) return -2;      if (cs == NULL) return -2;
426      do ce += GET(ce, 1); while (*ce == OP_ALT);      do ce += GET(ce, 1); while (*ce == OP_ALT);
427      if (cc > cs && cc < ce)      if ((cc > cs && cc < ce) || recurse_depth > 10)
428        had_recurse = TRUE;        had_recurse = TRUE;
429      else      else
430        branchlength += find_minlength(cs, startcode, options);        {
431          branchlength += find_minlength(cs, startcode, options, recurse_depth + 1);
432          }
433      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
434      break;      break;
435    
436      /* Anything else does not or need not match a character. We can get the      /* Anything else does not or need not match a character. We can get the
437      item's length from the table, but for those that can match zero occurrences      item's length from the table, but for those that can match zero occurrences
438      of a character, we must take special action for UTF-8 characters. */      of a character, we must take special action for UTF-8 characters. As it
439        happens, the "NOT" versions of these opcodes are used at present only for
440        ASCII characters, so they could be omitted from this list. However, in
441        future that may change, so we include them here so as not to leave a
442        gotcha for a future maintainer. */
443    
444      case OP_UPTO:      case OP_UPTO:
445        case OP_UPTOI:
446      case OP_NOTUPTO:      case OP_NOTUPTO:
447        case OP_NOTUPTOI:
448      case OP_MINUPTO:      case OP_MINUPTO:
449        case OP_MINUPTOI:
450      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
451        case OP_NOTMINUPTOI:
452      case OP_POSUPTO:      case OP_POSUPTO:
453        case OP_POSUPTOI:
454        case OP_NOTPOSUPTO:
455        case OP_NOTPOSUPTOI:
456    
457      case OP_STAR:      case OP_STAR:
458        case OP_STARI:
459        case OP_NOTSTAR:
460        case OP_NOTSTARI:
461      case OP_MINSTAR:      case OP_MINSTAR:
462        case OP_MINSTARI:
463      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
464        case OP_NOTMINSTARI:
465      case OP_POSSTAR:      case OP_POSSTAR:
466        case OP_POSSTARI:
467      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
468        case OP_NOTPOSSTARI:
469    
470      case OP_QUERY:      case OP_QUERY:
471        case OP_QUERYI:
472        case OP_NOTQUERY:
473        case OP_NOTQUERYI:
474      case OP_MINQUERY:      case OP_MINQUERY:
475        case OP_MINQUERYI:
476      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
477        case OP_NOTMINQUERYI:
478      case OP_POSQUERY:      case OP_POSQUERY:
479        case OP_POSQUERYI:
480      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
481        case OP_NOTPOSQUERYI:
482    
483      cc += _pcre_OP_lengths[op];      cc += _pcre_OP_lengths[op];
484  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
485      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
# Line 418  for (;;) Line 491  for (;;)
491      case OP_MARK:      case OP_MARK:
492      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
493      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     case OP_THEN_ARG:  
494      cc += _pcre_OP_lengths[op] + cc[1];      cc += _pcre_OP_lengths[op] + cc[1];
495      break;      break;
496    
497      /* For the record, these are the opcodes that are matched by "default":      case OP_THEN_ARG:
498      OP_ACCEPT, OP_CLOSE, OP_COMMIT, OP_FAIL, OP_PRUNE, OP_SET_SOM, OP_SKIP,      cc += _pcre_OP_lengths[op] + cc[1+LINK_SIZE];
499      OP_THEN. */      break;
500    
501      default:      /* The remaining opcodes are just skipped over. */
502    
503        case OP_CLOSE:
504        case OP_COMMIT:
505        case OP_FAIL:
506        case OP_PRUNE:
507        case OP_SET_SOM:
508        case OP_SKIP:
509        case OP_THEN:
510      cc += _pcre_OP_lengths[op];      cc += _pcre_OP_lengths[op];
511      break;      break;
512    
513        /* This should not occur: we list all opcodes explicitly so that when
514        new ones get added they are properly considered. */
515    
516        default:
517        return -3;
518      }      }
519    }    }
520  /* Control never gets here */  /* Control never gets here */
# Line 440  for (;;) Line 526  for (;;)
526  *      Set a bit and maybe its alternate case    *  *      Set a bit and maybe its alternate case    *
527  *************************************************/  *************************************************/
528    
529  /* Given a character, set its first byte's bit in the table, and also the  /* Given a character, set its first byte's bit in the table, and also the
530  corresponding bit for the other version of a letter if we are caseless. In  corresponding bit for the other version of a letter if we are caseless. In
531  UTF-8 mode, for characters greater than 127, we can only do the caseless thing  UTF-8 mode, for characters greater than 127, we can only do the caseless thing
532  when Unicode property support is available.  when Unicode property support is available.
# Line 450  Arguments: Line 536  Arguments:
536    p             points to the character    p             points to the character
537    caseless      the caseless flag    caseless      the caseless flag
538    cd            the block with char table pointers    cd            the block with char table pointers
539    utf8          TRUE for UTF-8 mode    utf8          TRUE for UTF-8 mode
540    
541  Returns:        pointer after the character  Returns:        pointer after the character
542  */  */
# Line 460  set_table_bit(uschar *start_bits, const Line 546  set_table_bit(uschar *start_bits, const
546    compile_data *cd, BOOL utf8)    compile_data *cd, BOOL utf8)
547  {  {
548  unsigned int c = *p;  unsigned int c = *p;
549  start_bits[c/8] |= (1 << (c&7));  
550    SET_BIT(c);
551    
552  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
553  if (utf8 && c > 127)  if (utf8 && c > 127)
# Line 469  if (utf8 && c > 127) Line 556  if (utf8 && c > 127)
556  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
557    if (caseless)    if (caseless)
558      {      {
559      uschar buff[8];      uschar buff[8];
560      c = UCD_OTHERCASE(c);      c = UCD_OTHERCASE(c);
561      (void)_pcre_ord2utf8(c, buff);      (void)_pcre_ord2utf8(c, buff);
562      c = buff[0];      SET_BIT(buff[0]);
563      start_bits[c/8] |= (1 << (c&7));      }
564      }  #endif
 #endif  
565    return p;    return p;
566    }    }
567  #endif  #endif
568    
569  /* Not UTF-8 mode, or character is less than 127. */  /* Not UTF-8 mode, or character is less than 127. */
570    
571  if (caseless && (cd->ctypes[c] & ctype_letter) != 0)  if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
   start_bits[cd->fcc[c]/8] |= (1 << (cd->fcc[c]&7));  
572  return p + 1;  return p + 1;
573  }  }
574    
575    
576    
577  /*************************************************  /*************************************************
578    *     Set bits for a positive character type     *
579    *************************************************/
580    
581    /* This function sets starting bits for a character type. In UTF-8 mode, we can
582    only do a direct setting for bytes less than 128, as otherwise there can be
583    confusion with bytes in the middle of UTF-8 characters. In a "traditional"
584    environment, the tables will only recognize ASCII characters anyway, but in at
585    least one Windows environment, some higher bytes bits were set in the tables.
586    So we deal with that case by considering the UTF-8 encoding.
587    
588    Arguments:
589      start_bits     the starting bitmap
590      cbit type      the type of character wanted
591      table_limit    32 for non-UTF-8; 16 for UTF-8
592      cd             the block with char table pointers
593    
594    Returns:         nothing
595    */
596    
597    static void
598    set_type_bits(uschar *start_bits, int cbit_type, int table_limit,
599      compile_data *cd)
600    {
601    register int c;
602    for (c = 0; c < table_limit; c++) start_bits[c] |= cd->cbits[c+cbit_type];
603    if (table_limit == 32) return;
604    for (c = 128; c < 256; c++)
605      {
606      if ((cd->cbits[c/8] & (1 << (c&7))) != 0)
607        {
608        uschar buff[8];
609        (void)_pcre_ord2utf8(c, buff);
610        SET_BIT(buff[0]);
611        }
612      }
613    }
614    
615    
616    /*************************************************
617    *     Set bits for a negative character type     *
618    *************************************************/
619    
620    /* This function sets starting bits for a negative character type such as \D.
621    In UTF-8 mode, we can only do a direct setting for bytes less than 128, as
622    otherwise there can be confusion with bytes in the middle of UTF-8 characters.
623    Unlike in the positive case, where we can set appropriate starting bits for
624    specific high-valued UTF-8 characters, in this case we have to set the bits for
625    all high-valued characters. The lowest is 0xc2, but we overkill by starting at
626    0xc0 (192) for simplicity.
627    
628    Arguments:
629      start_bits     the starting bitmap
630      cbit type      the type of character wanted
631      table_limit    32 for non-UTF-8; 16 for UTF-8
632      cd             the block with char table pointers
633    
634    Returns:         nothing
635    */
636    
637    static void
638    set_nottype_bits(uschar *start_bits, int cbit_type, int table_limit,
639      compile_data *cd)
640    {
641    register int c;
642    for (c = 0; c < table_limit; c++) start_bits[c] |= ~cd->cbits[c+cbit_type];
643    if (table_limit != 32) for (c = 24; c < 32; c++) start_bits[c] = 0xff;
644    }
645    
646    
647    
648    /*************************************************
649  *          Create bitmap of starting bytes       *  *          Create bitmap of starting bytes       *
650  *************************************************/  *************************************************/
651    
# Line 504  function fails unless the result is SSB_ Line 660  function fails unless the result is SSB_
660  Arguments:  Arguments:
661    code         points to an expression    code         points to an expression
662    start_bits   points to a 32-byte table, initialized to 0    start_bits   points to a 32-byte table, initialized to 0
   caseless     the current state of the caseless flag  
663    utf8         TRUE if in UTF-8 mode    utf8         TRUE if in UTF-8 mode
664    cd           the block with char table pointers    cd           the block with char table pointers
665    
666  Returns:       SSB_FAIL     => Failed to find any starting bytes  Returns:       SSB_FAIL     => Failed to find any starting bytes
667                 SSB_DONE     => Found mandatory starting bytes                 SSB_DONE     => Found mandatory starting bytes
668                 SSB_CONTINUE => Found optional starting bytes                 SSB_CONTINUE => Found optional starting bytes
669                   SSB_UNKNOWN  => Hit an unrecognized opcode
670  */  */
671    
672  static int  static int
673  set_start_bits(const uschar *code, uschar *start_bits, BOOL caseless,  set_start_bits(const uschar *code, uschar *start_bits, BOOL utf8,
674    BOOL utf8, compile_data *cd)    compile_data *cd)
675  {  {
676  register int c;  register int c;
677  int yield = SSB_DONE;  int yield = SSB_DONE;
678    int table_limit = utf8? 16:32;
679    
680  #if 0  #if 0
681  /* ========================================================================= */  /* ========================================================================= */
# Line 539  volatile int dummy; Line 696  volatile int dummy;
696    
697  do  do
698    {    {
   const uschar *tcode = code + (((int)*code == OP_CBRA)? 3:1) + LINK_SIZE;  
699    BOOL try_next = TRUE;    BOOL try_next = TRUE;
700      const uschar *tcode = code + 1 + LINK_SIZE;
701    
702      if (*code == OP_CBRA || *code == OP_SCBRA ||
703          *code == OP_CBRAPOS || *code == OP_SCBRAPOS) tcode += 2;
704    
705    while (try_next)    /* Loop for items in this branch */    while (try_next)    /* Loop for items in this branch */
706      {      {
707      int rc;      int rc;
708    
709      switch(*tcode)      switch(*tcode)
710        {        {
711        /* Fail if we reach something we don't understand */        /* If we reach something we don't understand, it means a new opcode has
712          been created that hasn't been added to this code. Hopefully this problem
713          will be discovered during testing. */
714    
715        default:        default:
716          return SSB_UNKNOWN;
717    
718          /* Fail for a valid opcode that implies no starting bits. */
719    
720          case OP_ACCEPT:
721          case OP_ASSERT_ACCEPT:
722          case OP_ALLANY:
723          case OP_ANY:
724          case OP_ANYBYTE:
725          case OP_CIRC:
726          case OP_CIRCM:
727          case OP_CLOSE:
728          case OP_COMMIT:
729          case OP_COND:
730          case OP_CREF:
731          case OP_DEF:
732          case OP_DOLL:
733          case OP_DOLLM:
734          case OP_END:
735          case OP_EOD:
736          case OP_EODN:
737          case OP_EXTUNI:
738          case OP_FAIL:
739          case OP_MARK:
740          case OP_NCREF:
741          case OP_NOT:
742          case OP_NOTEXACT:
743          case OP_NOTEXACTI:
744          case OP_NOTI:
745          case OP_NOTMINPLUS:
746          case OP_NOTMINPLUSI:
747          case OP_NOTMINQUERY:
748          case OP_NOTMINQUERYI:
749          case OP_NOTMINSTAR:
750          case OP_NOTMINSTARI:
751          case OP_NOTMINUPTO:
752          case OP_NOTMINUPTOI:
753          case OP_NOTPLUS:
754          case OP_NOTPLUSI:
755          case OP_NOTPOSPLUS:
756          case OP_NOTPOSPLUSI:
757          case OP_NOTPOSQUERY:
758          case OP_NOTPOSQUERYI:
759          case OP_NOTPOSSTAR:
760          case OP_NOTPOSSTARI:
761          case OP_NOTPOSUPTO:
762          case OP_NOTPOSUPTOI:
763          case OP_NOTPROP:
764          case OP_NOTQUERY:
765          case OP_NOTQUERYI:
766          case OP_NOTSTAR:
767          case OP_NOTSTARI:
768          case OP_NOTUPTO:
769          case OP_NOTUPTOI:
770          case OP_NOT_HSPACE:
771          case OP_NOT_VSPACE:
772          case OP_NRREF:
773          case OP_PROP:
774          case OP_PRUNE:
775          case OP_PRUNE_ARG:
776          case OP_RECURSE:
777          case OP_REF:
778          case OP_REFI:
779          case OP_REVERSE:
780          case OP_RREF:
781          case OP_SCOND:
782          case OP_SET_SOM:
783          case OP_SKIP:
784          case OP_SKIP_ARG:
785          case OP_SOD:
786          case OP_SOM:
787          case OP_THEN:
788          case OP_THEN_ARG:
789          case OP_XCLASS:
790        return SSB_FAIL;        return SSB_FAIL;
791    
792          /* We can ignore word boundary tests. */
793    
794          case OP_WORD_BOUNDARY:
795          case OP_NOT_WORD_BOUNDARY:
796          tcode++;
797          break;
798    
799        /* If we hit a bracket or a positive lookahead assertion, recurse to set        /* If we hit a bracket or a positive lookahead assertion, recurse to set
800        bits from within the subpattern. If it can't find anything, we have to        bits from within the subpattern. If it can't find anything, we have to
801        give up. If it finds some mandatory character(s), we are done for this        give up. If it finds some mandatory character(s), we are done for this
# Line 561  do Line 805  do
805        case OP_SBRA:        case OP_SBRA:
806        case OP_CBRA:        case OP_CBRA:
807        case OP_SCBRA:        case OP_SCBRA:
808          case OP_BRAPOS:
809          case OP_SBRAPOS:
810          case OP_CBRAPOS:
811          case OP_SCBRAPOS:
812        case OP_ONCE:        case OP_ONCE:
813        case OP_ASSERT:        case OP_ASSERT:
814        rc = set_start_bits(tcode, start_bits, caseless, utf8, cd);        rc = set_start_bits(tcode, start_bits, utf8, cd);
815        if (rc == SSB_FAIL) return SSB_FAIL;        if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
816        if (rc == SSB_DONE) try_next = FALSE; else        if (rc == SSB_DONE) try_next = FALSE; else
817          {          {
818          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
# Line 587  do Line 835  do
835        case OP_KET:        case OP_KET:
836        case OP_KETRMAX:        case OP_KETRMAX:
837        case OP_KETRMIN:        case OP_KETRMIN:
838          case OP_KETRPOS:
839        return SSB_CONTINUE;        return SSB_CONTINUE;
840    
841        /* Skip over callout */        /* Skip over callout */
# Line 604  do Line 853  do
853        tcode += 1 + LINK_SIZE;        tcode += 1 + LINK_SIZE;
854        break;        break;
855    
       /* Skip over an option setting, changing the caseless flag */  
   
       case OP_OPT:  
       caseless = (tcode[1] & PCRE_CASELESS) != 0;  
       tcode += 2;  
       break;  
   
856        /* BRAZERO does the bracket, but carries on. */        /* BRAZERO does the bracket, but carries on. */
857    
858        case OP_BRAZERO:        case OP_BRAZERO:
859        case OP_BRAMINZERO:        case OP_BRAMINZERO:
860        if (set_start_bits(++tcode, start_bits, caseless, utf8, cd) == SSB_FAIL)        case OP_BRAPOSZERO:
861          return SSB_FAIL;        rc = set_start_bits(++tcode, start_bits, utf8, cd);
862          if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
863  /* =========================================================================  /* =========================================================================
864        See the comment at the head of this function concerning the next line,        See the comment at the head of this function concerning the next line,
865        which was an old fudge for the benefit of OS/2.        which was an old fudge for the benefit of OS/2.
# Line 642  do Line 885  do
885        case OP_QUERY:        case OP_QUERY:
886        case OP_MINQUERY:        case OP_MINQUERY:
887        case OP_POSQUERY:        case OP_POSQUERY:
888        tcode = set_table_bit(start_bits, tcode + 1, caseless, cd, utf8);        tcode = set_table_bit(start_bits, tcode + 1, FALSE, cd, utf8);
889          break;
890    
891          case OP_STARI:
892          case OP_MINSTARI:
893          case OP_POSSTARI:
894          case OP_QUERYI:
895          case OP_MINQUERYI:
896          case OP_POSQUERYI:
897          tcode = set_table_bit(start_bits, tcode + 1, TRUE, cd, utf8);
898        break;        break;
899    
900        /* Single-char upto sets the bit and tries the next */        /* Single-char upto sets the bit and tries the next */
# Line 650  do Line 902  do
902        case OP_UPTO:        case OP_UPTO:
903        case OP_MINUPTO:        case OP_MINUPTO:
904        case OP_POSUPTO:        case OP_POSUPTO:
905        tcode = set_table_bit(start_bits, tcode + 3, caseless, cd, utf8);        tcode = set_table_bit(start_bits, tcode + 3, FALSE, cd, utf8);
906          break;
907    
908          case OP_UPTOI:
909          case OP_MINUPTOI:
910          case OP_POSUPTOI:
911          tcode = set_table_bit(start_bits, tcode + 3, TRUE, cd, utf8);
912        break;        break;
913    
914        /* At least one single char sets the bit and stops */        /* At least one single char sets the bit and stops */
915    
916        case OP_EXACT:       /* Fall through */        case OP_EXACT:
917        tcode += 2;        tcode += 2;
918          /* Fall through */
919        case OP_CHAR:        case OP_CHAR:
       case OP_CHARNC:  
920        case OP_PLUS:        case OP_PLUS:
921        case OP_MINPLUS:        case OP_MINPLUS:
922        case OP_POSPLUS:        case OP_POSPLUS:
923        (void)set_table_bit(start_bits, tcode + 1, caseless, cd, utf8);        (void)set_table_bit(start_bits, tcode + 1, FALSE, cd, utf8);
924        try_next = FALSE;        try_next = FALSE;
925        break;        break;
926    
927        /* Single character types set the bits and stop. Note that if PCRE_UCP        case OP_EXACTI:
928        is set, we do not see these op codes because \d etc are converted to        tcode += 2;
929        properties. Therefore, these apply in the case when only ASCII characters        /* Fall through */
930        are recognized to match the types. */        case OP_CHARI:
931          case OP_PLUSI:
932          case OP_MINPLUSI:
933          case OP_POSPLUSI:
934          (void)set_table_bit(start_bits, tcode + 1, TRUE, cd, utf8);
935          try_next = FALSE;
936          break;
937    
938          /* Special spacing and line-terminating items. These recognize specific
939          lists of characters. The difference between VSPACE and ANYNL is that the
940          latter can match the two-character CRLF sequence, but that is not
941          relevant for finding the first character, so their code here is
942          identical. */
943    
944          case OP_HSPACE:
945          SET_BIT(0x09);
946          SET_BIT(0x20);
947          if (utf8)
948            {
949            SET_BIT(0xC2);  /* For U+00A0 */
950            SET_BIT(0xE1);  /* For U+1680, U+180E */
951            SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
952            SET_BIT(0xE3);  /* For U+3000 */
953            }
954          else SET_BIT(0xA0);
955          try_next = FALSE;
956          break;
957    
958          case OP_ANYNL:
959          case OP_VSPACE:
960          SET_BIT(0x0A);
961          SET_BIT(0x0B);
962          SET_BIT(0x0C);
963          SET_BIT(0x0D);
964          if (utf8)
965            {
966            SET_BIT(0xC2);  /* For U+0085 */
967            SET_BIT(0xE2);  /* For U+2028, U+2029 */
968            }
969          else SET_BIT(0x85);
970          try_next = FALSE;
971          break;
972    
973          /* Single character types set the bits and stop. Note that if PCRE_UCP
974          is set, we do not see these op codes because \d etc are converted to
975          properties. Therefore, these apply in the case when only characters less
976          than 256 are recognized to match the types. */
977    
978        case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
979        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_digit];  
980        try_next = FALSE;        try_next = FALSE;
981        break;        break;
982    
983        case OP_DIGIT:        case OP_DIGIT:
984        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_digit];  
985        try_next = FALSE;        try_next = FALSE;
986        break;        break;
987    
988        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to
989        discard it. */        ensure it is set as not whitespace. */
990    
991        case OP_NOT_WHITESPACE:        case OP_NOT_WHITESPACE:
992        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_space, table_limit, cd);
993          {        start_bits[1] |= 0x08;
         int d = cd->cbits[c+cbit_space];  
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= ~d;  
         }  
994        try_next = FALSE;        try_next = FALSE;
995        break;        break;
996    
997        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to
998        discard it. */        not set it from the table. */
999    
1000        case OP_WHITESPACE:        case OP_WHITESPACE:
1001        for (c = 0; c < 32; c++)        c = start_bits[1];    /* Save in case it was already set */
1002          {        set_type_bits(start_bits, cbit_space, table_limit, cd);
1003          int d = cd->cbits[c+cbit_space];        start_bits[1] = (start_bits[1] & ~0x08) | c;
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= d;  
         }  
1004        try_next = FALSE;        try_next = FALSE;
1005        break;        break;
1006    
1007        case OP_NOT_WORDCHAR:        case OP_NOT_WORDCHAR:
1008        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_word];  
1009        try_next = FALSE;        try_next = FALSE;
1010        break;        break;
1011    
1012        case OP_WORDCHAR:        case OP_WORDCHAR:
1013        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_word];  
1014        try_next = FALSE;        try_next = FALSE;
1015        break;        break;
1016    
# Line 727  do Line 1019  do
1019    
1020        case OP_TYPEPLUS:        case OP_TYPEPLUS:
1021        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
1022          case OP_TYPEPOSPLUS:
1023        tcode++;        tcode++;
1024        break;        break;
1025    
# Line 755  do Line 1048  do
1048          case OP_ALLANY:          case OP_ALLANY:
1049          return SSB_FAIL;          return SSB_FAIL;
1050    
1051            case OP_HSPACE:
1052            SET_BIT(0x09);
1053            SET_BIT(0x20);
1054            if (utf8)
1055              {
1056              SET_BIT(0xC2);  /* For U+00A0 */
1057              SET_BIT(0xE1);  /* For U+1680, U+180E */
1058              SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
1059              SET_BIT(0xE3);  /* For U+3000 */
1060              }
1061            else SET_BIT(0xA0);
1062            break;
1063    
1064            case OP_ANYNL:
1065            case OP_VSPACE:
1066            SET_BIT(0x0A);
1067            SET_BIT(0x0B);
1068            SET_BIT(0x0C);
1069            SET_BIT(0x0D);
1070            if (utf8)
1071              {
1072              SET_BIT(0xC2);  /* For U+0085 */
1073              SET_BIT(0xE2);  /* For U+2028, U+2029 */
1074              }
1075            else SET_BIT(0x85);
1076            break;
1077    
1078          case OP_NOT_DIGIT:          case OP_NOT_DIGIT:
1079          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_digit];  
1080          break;          break;
1081    
1082          case OP_DIGIT:          case OP_DIGIT:
1083          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_digit];  
1084          break;          break;
1085    
1086          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we have to
1087          discard it. */          ensure it gets set as not whitespace. */
1088    
1089          case OP_NOT_WHITESPACE:          case OP_NOT_WHITESPACE:
1090          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_space, table_limit, cd);
1091            {          start_bits[1] |= 0x08;
           int d = cd->cbits[c+cbit_space];  
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= ~d;  
           }  
1092          break;          break;
1093    
1094          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we have to
1095          discard it. */          avoid setting it. */
1096    
1097          case OP_WHITESPACE:          case OP_WHITESPACE:
1098          for (c = 0; c < 32; c++)          c = start_bits[1];    /* Save in case it was already set */
1099            {          set_type_bits(start_bits, cbit_space, table_limit, cd);
1100            int d = cd->cbits[c+cbit_space];          start_bits[1] = (start_bits[1] & ~0x08) | c;
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= d;  
           }  
1101          break;          break;
1102    
1103          case OP_NOT_WORDCHAR:          case OP_NOT_WORDCHAR:
1104          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_word];  
1105          break;          break;
1106    
1107          case OP_WORDCHAR:          case OP_WORDCHAR:
1108          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_word];  
1109          break;          break;
1110          }          }
1111    
# Line 852  do Line 1161  do
1161            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];
1162            }            }
1163    
1164          /* Advance past the bit map, and act on what follows */          /* Advance past the bit map, and act on what follows. For a zero
1165            minimum repeat, continue; otherwise stop processing. */
1166    
1167          tcode += 32;          tcode += 32;
1168          switch (*tcode)          switch (*tcode)
# Line 888  return yield; Line 1198  return yield;
1198    
1199    
1200    
1201    
1202    
1203  /*************************************************  /*************************************************
1204  *          Study a compiled expression           *  *          Study a compiled expression           *
1205  *************************************************/  *************************************************/
# Line 913  pcre_study(const pcre *external_re, int Line 1225  pcre_study(const pcre *external_re, int
1225  int min;  int min;
1226  BOOL bits_set = FALSE;  BOOL bits_set = FALSE;
1227  uschar start_bits[32];  uschar start_bits[32];
1228  pcre_extra *extra;  pcre_extra *extra = NULL;
1229  pcre_study_data *study;  pcre_study_data *study;
1230  const uschar *tables;  const uschar *tables;
1231  uschar *code;  uschar *code;
# Line 944  seeking a list of starting bytes. */ Line 1256  seeking a list of starting bytes. */
1256  if ((re->options & PCRE_ANCHORED) == 0 &&  if ((re->options & PCRE_ANCHORED) == 0 &&
1257      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)
1258    {    {
1259      int rc;
1260    
1261    /* Set the character tables in the block that is passed around */    /* Set the character tables in the block that is passed around */
1262    
1263    tables = re->tables;    tables = re->tables;
# Line 959  if ((re->options & PCRE_ANCHORED) == 0 & Line 1273  if ((re->options & PCRE_ANCHORED) == 0 &
1273    /* See if we can find a fixed set of initial characters for the pattern. */    /* See if we can find a fixed set of initial characters for the pattern. */
1274    
1275    memset(start_bits, 0, 32 * sizeof(uschar));    memset(start_bits, 0, 32 * sizeof(uschar));
1276    bits_set = set_start_bits(code, start_bits,    rc = set_start_bits(code, start_bits, (re->options & PCRE_UTF8) != 0,
1277      (re->options & PCRE_CASELESS) != 0, (re->options & PCRE_UTF8) != 0,      &compile_block);
1278      &compile_block) == SSB_DONE;    bits_set = rc == SSB_DONE;
1279      if (rc == SSB_UNKNOWN)
1280        {
1281        *errorptr = "internal error: opcode not recognized";
1282        return NULL;
1283        }
1284    }    }
1285    
1286  /* Find the minimum length of subject string. */  /* Find the minimum length of subject string. */
1287    
1288  min = find_minlength(code, code, re->options);  switch(min = find_minlength(code, code, re->options, 0))
1289      {
1290      case -2: *errorptr = "internal error: missing capturing bracket"; return NULL;
1291      case -3: *errorptr = "internal error: opcode not recognized"; return NULL;
1292      default: break;
1293      }
1294    
1295    /* If a set of starting bytes has been identified, or if the minimum length is
1296    greater than zero, or if JIT optimization has been requested, get a pcre_extra
1297    block and a pcre_study_data block. The study data is put in the latter, which
1298    is pointed to by the former, which may also get additional data set later by
1299    the calling program. At the moment, the size of pcre_study_data is fixed. We
1300    nevertheless save it in a field for returning via the pcre_fullinfo() function
1301    so that if it becomes variable in the future, we don't have to change that
1302    code. */
1303    
1304    if (bits_set || min > 0
1305    #ifdef SUPPORT_JIT
1306        || (options & PCRE_STUDY_JIT_COMPILE) != 0
1307    #endif
1308      )
1309      {
1310      extra = (pcre_extra *)(pcre_malloc)
1311        (sizeof(pcre_extra) + sizeof(pcre_study_data));
1312      if (extra == NULL)
1313        {
1314        *errorptr = "failed to get memory";
1315        return NULL;
1316        }
1317    
1318  /* Return NULL if no optimization is possible. */    study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));
1319      extra->flags = PCRE_EXTRA_STUDY_DATA;
1320      extra->study_data = study;
1321    
1322  if (!bits_set && min < 0) return NULL;    study->size = sizeof(pcre_study_data);
1323      study->flags = 0;
1324    
1325  /* Get a pcre_extra block and a pcre_study_data block. The study data is put in    if (bits_set)
1326  the latter, which is pointed to by the former, which may also get additional      {
1327  data set later by the calling program. At the moment, the size of      study->flags |= PCRE_STUDY_MAPPED;
1328  pcre_study_data is fixed. We nevertheless save it in a field for returning via      memcpy(study->start_bits, start_bits, sizeof(start_bits));
1329  the pcre_fullinfo() function so that if it becomes variable in the future, we      }
 don't have to change that code. */  
1330    
1331  extra = (pcre_extra *)(pcre_malloc)    /* Always set the minlength value in the block, because the JIT compiler
1332    (sizeof(pcre_extra) + sizeof(pcre_study_data));    makes use of it. However, don't set the bit unless the length is greater than
1333      zero - the interpretive pcre_exec() and pcre_dfa_exec() needn't waste time
1334      checking the zero case. */
1335    
1336  if (extra == NULL)    if (min > 0)
1337    {      {
1338    *errorptr = "failed to get memory";      study->flags |= PCRE_STUDY_MINLEN;
1339    return NULL;      study->minlength = min;
1340        }
1341      else study->minlength = 0;
1342    
1343      /* If JIT support was compiled and requested, attempt the JIT compilation.
1344      If no starting bytes were found, and the minimum length is zero, and JIT
1345      compilation fails, abandon the extra block and return NULL. */
1346    
1347    #ifdef SUPPORT_JIT
1348      extra->executable_jit = NULL;
1349      if ((options & PCRE_STUDY_JIT_COMPILE) != 0) _pcre_jit_compile(re, extra);
1350      if (study->flags == 0 && (extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) == 0)
1351        {
1352        pcre_free_study(extra);
1353        extra = NULL;
1354        }
1355    #endif
1356    }    }
1357    
1358  study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));  return extra;
1359  extra->flags = PCRE_EXTRA_STUDY_DATA;  }
 extra->study_data = study;  
1360    
 study->size = sizeof(pcre_study_data);  
 study->flags = 0;  
1361    
1362  if (bits_set)  /*************************************************
1363    {  *          Free the study data                   *
1364    study->flags |= PCRE_STUDY_MAPPED;  *************************************************/
   memcpy(study->start_bits, start_bits, sizeof(start_bits));  
   }  
1365    
1366  if (min >= 0)  /* This function frees the memory that was obtained by pcre_study().
   {  
   study->flags |= PCRE_STUDY_MINLEN;  
   study->minlength = min;  
   }  
1367    
1368  return extra;  Argument:   a pointer to the pcre_extra block
1369    Returns:    nothing
1370    */
1371    
1372    PCRE_EXP_DEFN void
1373    pcre_free_study(pcre_extra *extra)
1374    {
1375    #ifdef SUPPORT_JIT
1376    if ((extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 &&
1377         extra->executable_jit != NULL)
1378      _pcre_jit_free(extra->executable_jit);
1379    #endif
1380    pcre_free(extra);
1381  }  }
1382    
1383  /* End of pcre_study.c */  /* End of pcre_study.c */

Legend:
Removed from v.523  
changed lines
  Added in v.697

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12