/[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 538 by ph10, Wed Jun 9 19:30:57 2010 UTC revision 797 by ph10, Sun Dec 11 17:09:37 2011 UTC
# Line 52  supporting functions. */ Line 52  supporting functions. */
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 66  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 84  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 118  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);      case OP_ONCE_NC:
131        d = find_minlength(cc, startcode, options, recurse_depth);
132      if (d < 0) return d;      if (d < 0) return d;
133      branchlength += d;      branchlength += d;
134      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
135      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
136      break;      break;
137    
138        /* ACCEPT makes things far too complicated; we have to give up. */
139    
140        case OP_ACCEPT:
141        case OP_ASSERT_ACCEPT:
142        return -1;
143    
144      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested
145      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is END it's
146      END it's the end of the outer call. All can be handled by the same code. */      the end of the outer call. All can be handled by the same code. If an
147        ACCEPT was previously encountered, use the length that was in force at that
148        time, and pass back the shortest ACCEPT length. */
149    
150      case OP_ALT:      case OP_ALT:
151      case OP_KET:      case OP_KET:
152      case OP_KETRMAX:      case OP_KETRMAX:
153      case OP_KETRMIN:      case OP_KETRMIN:
154        case OP_KETRPOS:
155      case OP_END:      case OP_END:
156      if (length < 0 || (!had_recurse && branchlength < length))      if (length < 0 || (!had_recurse && branchlength < length))
157        length = branchlength;        length = branchlength;
158      if (*cc != OP_ALT) return length;      if (op != OP_ALT) return length;
159      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
160      branchlength = 0;      branchlength = 0;
161      had_recurse = FALSE;      had_recurse = FALSE;
# Line 160  for (;;) Line 178  for (;;)
178      case OP_RREF:      case OP_RREF:
179      case OP_NRREF:      case OP_NRREF:
180      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
181      case OP_CALLOUT:      case OP_CALLOUT:
182      case OP_SOD:      case OP_SOD:
183      case OP_SOM:      case OP_SOM:
184      case OP_EOD:      case OP_EOD:
185      case OP_EODN:      case OP_EODN:
186      case OP_CIRC:      case OP_CIRC:
187        case OP_CIRCM:
188      case OP_DOLL:      case OP_DOLL:
189        case OP_DOLLM:
190      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
191      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
192      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 177  for (;;) Line 196  for (;;)
196    
197      case OP_BRAZERO:      case OP_BRAZERO:
198      case OP_BRAMINZERO:      case OP_BRAMINZERO:
199        case OP_BRAPOSZERO:
200      case OP_SKIPZERO:      case OP_SKIPZERO:
201      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
202      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 186  for (;;) Line 206  for (;;)
206      /* Handle literal characters and + repetitions */      /* Handle literal characters and + repetitions */
207    
208      case OP_CHAR:      case OP_CHAR:
209      case OP_CHARNC:      case OP_CHARI:
210      case OP_NOT:      case OP_NOT:
211        case OP_NOTI:
212      case OP_PLUS:      case OP_PLUS:
213        case OP_PLUSI:
214      case OP_MINPLUS:      case OP_MINPLUS:
215        case OP_MINPLUSI:
216      case OP_POSPLUS:      case OP_POSPLUS:
217        case OP_POSPLUSI:
218      case OP_NOTPLUS:      case OP_NOTPLUS:
219        case OP_NOTPLUSI:
220      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
221        case OP_NOTMINPLUSI:
222      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
223        case OP_NOTPOSPLUSI:
224      branchlength++;      branchlength++;
225      cc += 2;      cc += 2;
226  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 212  for (;;) Line 239  for (;;)
239      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
240    
241      case OP_EXACT:      case OP_EXACT:
242        case OP_EXACTI:
243      case OP_NOTEXACT:      case OP_NOTEXACT:
244        case OP_NOTEXACTI:
245      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
246      cc += 4;      cc += 4;
247  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 249  for (;;) Line 278  for (;;)
278      cc++;      cc++;
279      break;      break;
280    
281      /* "Any newline" might match two characters */      /* "Any newline" might match two characters, but it also might match just
282        one. */
283    
284      case OP_ANYNL:      case OP_ANYNL:
285      branchlength += 2;      branchlength += 1;
286      cc++;      cc++;
287      break;      break;
288    
289      /* The single-byte matcher means we can't proceed in UTF-8 mode */      /* The single-byte matcher means we can't proceed in UTF-8 mode. (In
290        non-UTF-8 mode \C will actually be turned into OP_ALLANY, so won't ever
291        appear, but leave the code, just in case.) */
292    
293      case OP_ANYBYTE:      case OP_ANYBYTE:
294  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 337  for (;;) Line 369  for (;;)
369      that case we must set the minimum length to zero. */      that case we must set the minimum length to zero. */
370    
371      case OP_REF:      case OP_REF:
372        case OP_REFI:
373      if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
374        {        {
375        ce = cs = (uschar *)_pcre_find_bracket(startcode, utf8, GET2(cc, 1));        ce = cs = (uschar *)_pcre_find_bracket(startcode, utf8, GET2(cc, 1));
# Line 347  for (;;) Line 380  for (;;)
380          d = 0;          d = 0;
381          had_recurse = TRUE;          had_recurse = TRUE;
382          }          }
383        else d = find_minlength(cs, startcode, options);        else
384            {
385            d = find_minlength(cs, startcode, options, recurse_depth);
386            }
387        }        }
388      else d = 0;      else d = 0;
389      cc += 3;      cc += 3;
# Line 364  for (;;) Line 400  for (;;)
400        cc++;        cc++;
401        break;        break;
402    
403          case OP_CRPLUS:
404          case OP_CRMINPLUS:
405          min = 1;
406          cc++;
407          break;
408    
409        case OP_CRRANGE:        case OP_CRRANGE:
410        case OP_CRMINRANGE:        case OP_CRMINRANGE:
411        min = GET2(cc, 1);        min = GET2(cc, 1);
# Line 378  for (;;) Line 420  for (;;)
420      branchlength += min * d;      branchlength += min * d;
421      break;      break;
422    
423        /* We can easily detect direct recursion, but not mutual recursion. This is
424        caught by a recursion depth count. */
425    
426      case OP_RECURSE:      case OP_RECURSE:
427      cs = ce = (uschar *)startcode + GET(cc, 1);      cs = ce = (uschar *)startcode + GET(cc, 1);
     if (cs == NULL) return -2;  
428      do ce += GET(ce, 1); while (*ce == OP_ALT);      do ce += GET(ce, 1); while (*ce == OP_ALT);
429      if (cc > cs && cc < ce)      if ((cc > cs && cc < ce) || recurse_depth > 10)
430        had_recurse = TRUE;        had_recurse = TRUE;
431      else      else
432        branchlength += find_minlength(cs, startcode, options);        {
433          branchlength += find_minlength(cs, startcode, options, recurse_depth + 1);
434          }
435      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
436      break;      break;
437    
438      /* 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
439      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
440      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
441        happens, the "NOT" versions of these opcodes are used at present only for
442        ASCII characters, so they could be omitted from this list. However, in
443        future that may change, so we include them here so as not to leave a
444        gotcha for a future maintainer. */
445    
446      case OP_UPTO:      case OP_UPTO:
447        case OP_UPTOI:
448      case OP_NOTUPTO:      case OP_NOTUPTO:
449        case OP_NOTUPTOI:
450      case OP_MINUPTO:      case OP_MINUPTO:
451        case OP_MINUPTOI:
452      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
453        case OP_NOTMINUPTOI:
454      case OP_POSUPTO:      case OP_POSUPTO:
455        case OP_POSUPTOI:
456        case OP_NOTPOSUPTO:
457        case OP_NOTPOSUPTOI:
458    
459      case OP_STAR:      case OP_STAR:
460        case OP_STARI:
461        case OP_NOTSTAR:
462        case OP_NOTSTARI:
463      case OP_MINSTAR:      case OP_MINSTAR:
464        case OP_MINSTARI:
465      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
466        case OP_NOTMINSTARI:
467      case OP_POSSTAR:      case OP_POSSTAR:
468        case OP_POSSTARI:
469      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
470        case OP_NOTPOSSTARI:
471    
472      case OP_QUERY:      case OP_QUERY:
473        case OP_QUERYI:
474        case OP_NOTQUERY:
475        case OP_NOTQUERYI:
476      case OP_MINQUERY:      case OP_MINQUERY:
477        case OP_MINQUERYI:
478      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
479        case OP_NOTMINQUERYI:
480      case OP_POSQUERY:      case OP_POSQUERY:
481        case OP_POSQUERYI:
482      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
483        case OP_NOTPOSQUERYI:
484    
485      cc += _pcre_OP_lengths[op];      cc += _pcre_OP_lengths[op];
486  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
487      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 423  for (;;) Line 497  for (;;)
497      cc += _pcre_OP_lengths[op] + cc[1];      cc += _pcre_OP_lengths[op] + cc[1];
498      break;      break;
499    
500      /* For the record, these are the opcodes that are matched by "default":      /* The remaining opcodes are just skipped over. */
     OP_ACCEPT, OP_CLOSE, OP_COMMIT, OP_FAIL, OP_PRUNE, OP_SET_SOM, OP_SKIP,  
     OP_THEN. */  
501    
502      default:      case OP_CLOSE:
503        case OP_COMMIT:
504        case OP_FAIL:
505        case OP_PRUNE:
506        case OP_SET_SOM:
507        case OP_SKIP:
508        case OP_THEN:
509      cc += _pcre_OP_lengths[op];      cc += _pcre_OP_lengths[op];
510      break;      break;
511    
512        /* This should not occur: we list all opcodes explicitly so that when
513        new ones get added they are properly considered. */
514    
515        default:
516        return -3;
517      }      }
518    }    }
519  /* Control never gets here */  /* Control never gets here */
# Line 490  return p + 1; Line 574  return p + 1;
574    
575    
576  /*************************************************  /*************************************************
577    *     Set bits for a positive character type     *
578    *************************************************/
579    
580    /* This function sets starting bits for a character type. In UTF-8 mode, we can
581    only do a direct setting for bytes less than 128, as otherwise there can be
582    confusion with bytes in the middle of UTF-8 characters. In a "traditional"
583    environment, the tables will only recognize ASCII characters anyway, but in at
584    least one Windows environment, some higher bytes bits were set in the tables.
585    So we deal with that case by considering the UTF-8 encoding.
586    
587    Arguments:
588      start_bits     the starting bitmap
589      cbit type      the type of character wanted
590      table_limit    32 for non-UTF-8; 16 for UTF-8
591      cd             the block with char table pointers
592    
593    Returns:         nothing
594    */
595    
596    static void
597    set_type_bits(uschar *start_bits, int cbit_type, int table_limit,
598      compile_data *cd)
599    {
600    register int c;
601    for (c = 0; c < table_limit; c++) start_bits[c] |= cd->cbits[c+cbit_type];
602    if (table_limit == 32) return;
603    for (c = 128; c < 256; c++)
604      {
605      if ((cd->cbits[c/8] & (1 << (c&7))) != 0)
606        {
607        uschar buff[8];
608        (void)_pcre_ord2utf8(c, buff);
609        SET_BIT(buff[0]);
610        }
611      }
612    }
613    
614    
615    /*************************************************
616    *     Set bits for a negative character type     *
617    *************************************************/
618    
619    /* This function sets starting bits for a negative character type such as \D.
620    In UTF-8 mode, we can only do a direct setting for bytes less than 128, as
621    otherwise there can be confusion with bytes in the middle of UTF-8 characters.
622    Unlike in the positive case, where we can set appropriate starting bits for
623    specific high-valued UTF-8 characters, in this case we have to set the bits for
624    all high-valued characters. The lowest is 0xc2, but we overkill by starting at
625    0xc0 (192) for simplicity.
626    
627    Arguments:
628      start_bits     the starting bitmap
629      cbit type      the type of character wanted
630      table_limit    32 for non-UTF-8; 16 for UTF-8
631      cd             the block with char table pointers
632    
633    Returns:         nothing
634    */
635    
636    static void
637    set_nottype_bits(uschar *start_bits, int cbit_type, int table_limit,
638      compile_data *cd)
639    {
640    register int c;
641    for (c = 0; c < table_limit; c++) start_bits[c] |= ~cd->cbits[c+cbit_type];
642    if (table_limit != 32) for (c = 24; c < 32; c++) start_bits[c] = 0xff;
643    }
644    
645    
646    
647    /*************************************************
648  *          Create bitmap of starting bytes       *  *          Create bitmap of starting bytes       *
649  *************************************************/  *************************************************/
650    
# Line 504  function fails unless the result is SSB_ Line 659  function fails unless the result is SSB_
659  Arguments:  Arguments:
660    code         points to an expression    code         points to an expression
661    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  
662    utf8         TRUE if in UTF-8 mode    utf8         TRUE if in UTF-8 mode
663    cd           the block with char table pointers    cd           the block with char table pointers
664    
665  Returns:       SSB_FAIL     => Failed to find any starting bytes  Returns:       SSB_FAIL     => Failed to find any starting bytes
666                 SSB_DONE     => Found mandatory starting bytes                 SSB_DONE     => Found mandatory starting bytes
667                 SSB_CONTINUE => Found optional starting bytes                 SSB_CONTINUE => Found optional starting bytes
668                   SSB_UNKNOWN  => Hit an unrecognized opcode
669  */  */
670    
671  static int  static int
672  set_start_bits(const uschar *code, uschar *start_bits, BOOL caseless,  set_start_bits(const uschar *code, uschar *start_bits, BOOL utf8,
673    BOOL utf8, compile_data *cd)    compile_data *cd)
674  {  {
675  register int c;  register int c;
676  int yield = SSB_DONE;  int yield = SSB_DONE;
# Line 540  volatile int dummy; Line 695  volatile int dummy;
695    
696  do  do
697    {    {
   const uschar *tcode = code + (((int)*code == OP_CBRA)? 3:1) + LINK_SIZE;  
698    BOOL try_next = TRUE;    BOOL try_next = TRUE;
699      const uschar *tcode = code + 1 + LINK_SIZE;
700    
701      if (*code == OP_CBRA || *code == OP_SCBRA ||
702          *code == OP_CBRAPOS || *code == OP_SCBRAPOS) tcode += 2;
703    
704    while (try_next)    /* Loop for items in this branch */    while (try_next)    /* Loop for items in this branch */
705      {      {
706      int rc;      int rc;
707    
708      switch(*tcode)      switch(*tcode)
709        {        {
710        /* Fail if we reach something we don't understand */        /* If we reach something we don't understand, it means a new opcode has
711          been created that hasn't been added to this code. Hopefully this problem
712          will be discovered during testing. */
713    
714        default:        default:
715          return SSB_UNKNOWN;
716    
717          /* Fail for a valid opcode that implies no starting bits. */
718    
719          case OP_ACCEPT:
720          case OP_ASSERT_ACCEPT:
721          case OP_ALLANY:
722          case OP_ANY:
723          case OP_ANYBYTE:
724          case OP_CIRC:
725          case OP_CIRCM:
726          case OP_CLOSE:
727          case OP_COMMIT:
728          case OP_COND:
729          case OP_CREF:
730          case OP_DEF:
731          case OP_DOLL:
732          case OP_DOLLM:
733          case OP_END:
734          case OP_EOD:
735          case OP_EODN:
736          case OP_EXTUNI:
737          case OP_FAIL:
738          case OP_MARK:
739          case OP_NCREF:
740          case OP_NOT:
741          case OP_NOTEXACT:
742          case OP_NOTEXACTI:
743          case OP_NOTI:
744          case OP_NOTMINPLUS:
745          case OP_NOTMINPLUSI:
746          case OP_NOTMINQUERY:
747          case OP_NOTMINQUERYI:
748          case OP_NOTMINSTAR:
749          case OP_NOTMINSTARI:
750          case OP_NOTMINUPTO:
751          case OP_NOTMINUPTOI:
752          case OP_NOTPLUS:
753          case OP_NOTPLUSI:
754          case OP_NOTPOSPLUS:
755          case OP_NOTPOSPLUSI:
756          case OP_NOTPOSQUERY:
757          case OP_NOTPOSQUERYI:
758          case OP_NOTPOSSTAR:
759          case OP_NOTPOSSTARI:
760          case OP_NOTPOSUPTO:
761          case OP_NOTPOSUPTOI:
762          case OP_NOTPROP:
763          case OP_NOTQUERY:
764          case OP_NOTQUERYI:
765          case OP_NOTSTAR:
766          case OP_NOTSTARI:
767          case OP_NOTUPTO:
768          case OP_NOTUPTOI:
769          case OP_NOT_HSPACE:
770          case OP_NOT_VSPACE:
771          case OP_NRREF:
772          case OP_PROP:
773          case OP_PRUNE:
774          case OP_PRUNE_ARG:
775          case OP_RECURSE:
776          case OP_REF:
777          case OP_REFI:
778          case OP_REVERSE:
779          case OP_RREF:
780          case OP_SCOND:
781          case OP_SET_SOM:
782          case OP_SKIP:
783          case OP_SKIP_ARG:
784          case OP_SOD:
785          case OP_SOM:
786          case OP_THEN:
787          case OP_THEN_ARG:
788          case OP_XCLASS:
789        return SSB_FAIL;        return SSB_FAIL;
790    
791          /* We can ignore word boundary tests. */
792    
793          case OP_WORD_BOUNDARY:
794          case OP_NOT_WORD_BOUNDARY:
795          tcode++;
796          break;
797    
798        /* 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
799        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
800        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 562  do Line 804  do
804        case OP_SBRA:        case OP_SBRA:
805        case OP_CBRA:        case OP_CBRA:
806        case OP_SCBRA:        case OP_SCBRA:
807          case OP_BRAPOS:
808          case OP_SBRAPOS:
809          case OP_CBRAPOS:
810          case OP_SCBRAPOS:
811        case OP_ONCE:        case OP_ONCE:
812          case OP_ONCE_NC:
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 588  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 605  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 643  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 651  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;
925          break;
926    
927          case OP_EXACTI:
928          tcode += 2;
929          /* Fall through */
930          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;        try_next = FALSE;
936        break;        break;
937    
# Line 679  do Line 946  do
946        SET_BIT(0x20);        SET_BIT(0x20);
947        if (utf8)        if (utf8)
948          {          {
949          SET_BIT(0xC2);  /* For U+00A0 */          SET_BIT(0xC2);  /* For U+00A0 */
950          SET_BIT(0xE1);  /* For U+1680, U+180E */          SET_BIT(0xE1);  /* For U+1680, U+180E */
951          SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */          SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
952          SET_BIT(0xE3);  /* For U+3000 */          SET_BIT(0xE3);  /* For U+3000 */
# Line 694  do Line 961  do
961        SET_BIT(0x0B);        SET_BIT(0x0B);
962        SET_BIT(0x0C);        SET_BIT(0x0C);
963        SET_BIT(0x0D);        SET_BIT(0x0D);
964        if (utf8)        if (utf8)
965          {          {
966          SET_BIT(0xC2);  /* For U+0085 */          SET_BIT(0xC2);  /* For U+0085 */
967          SET_BIT(0xE2);  /* For U+2028, U+2029 */          SET_BIT(0xE2);  /* For U+2028, U+2029 */
968          }          }
969        else SET_BIT(0x85);        else SET_BIT(0x85);
970        try_next = FALSE;        try_next = FALSE;
971        break;        break;
972    
973        /* Single character types set the bits and stop. Note that if PCRE_UCP        /* 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        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 ASCII characters        properties. Therefore, these apply in the case when only characters less
976        are recognized to match the types. In UTF-8 mode, we must restrict        than 256 are recognized to match the types. */
       ourselves to bytes less than 128, as otherwise there can be confusion  
       with bytes in the middle of UTF-8 characters. (In a "traditional"  
       environment, the tables will only recognize ASCII characters anyway, but  
       in at least one Windows environment, some higher bytes bits were set in  
       the tables.) */  
977    
978        case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
979        for (c = 0; c < table_limit; 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 < table_limit; 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 < table_limit; 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 < table_limit; 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 < table_limit; 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 < table_limit; 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 802  do Line 1053  do
1053          SET_BIT(0x20);          SET_BIT(0x20);
1054          if (utf8)          if (utf8)
1055            {            {
1056            SET_BIT(0xC2);  /* For U+00A0 */            SET_BIT(0xC2);  /* For U+00A0 */
1057            SET_BIT(0xE1);  /* For U+1680, U+180E */            SET_BIT(0xE1);  /* For U+1680, U+180E */
1058            SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */            SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
1059            SET_BIT(0xE3);  /* For U+3000 */            SET_BIT(0xE3);  /* For U+3000 */
# Line 816  do Line 1067  do
1067          SET_BIT(0x0B);          SET_BIT(0x0B);
1068          SET_BIT(0x0C);          SET_BIT(0x0C);
1069          SET_BIT(0x0D);          SET_BIT(0x0D);
1070          if (utf8)          if (utf8)
1071            {            {
1072            SET_BIT(0xC2);  /* For U+0085 */            SET_BIT(0xC2);  /* For U+0085 */
1073            SET_BIT(0xE2);  /* For U+2028, U+2029 */            SET_BIT(0xE2);  /* For U+2028, U+2029 */
1074            }            }
1075          else SET_BIT(0x85);          else SET_BIT(0x85);
1076          break;          break;
1077    
1078          case OP_NOT_DIGIT:          case OP_NOT_DIGIT:
1079          for (c = 0; c < table_limit; 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 < table_limit; 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 < table_limit; 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 < table_limit; 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 < table_limit; 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 < table_limit; 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 921  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 957  return yield; Line 1198  return yield;
1198    
1199    
1200    
1201    
1202    
1203  /*************************************************  /*************************************************
1204  *          Study a compiled expression           *  *          Study a compiled expression           *
1205  *************************************************/  *************************************************/
# Line 982  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 1013  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 1028  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      study->size = sizeof(pcre_study_data);
1323      study->flags = 0;
1324    
1325      /* Set the start bits always, to avoid unset memory errors if the
1326      study data is written to a file, but set the flag only if any of the bits
1327      are set, to save time looking when none are. */
1328    
1329  if (!bits_set && min < 0) return NULL;    if (bits_set)
1330        {
1331        study->flags |= PCRE_STUDY_MAPPED;
1332        memcpy(study->start_bits, start_bits, sizeof(start_bits));
1333        }
1334      else memset(study->start_bits, 0, 32 * sizeof(uschar));
1335    
1336  /* Get a pcre_extra block and a pcre_study_data block. The study data is put in    /* Always set the minlength value in the block, because the JIT compiler
1337  the latter, which is pointed to by the former, which may also get additional    makes use of it. However, don't set the bit unless the length is greater than
1338  data set later by the calling program. At the moment, the size of    zero - the interpretive pcre_exec() and pcre_dfa_exec() needn't waste time
1339  pcre_study_data is fixed. We nevertheless save it in a field for returning via    checking the zero case. */
 the pcre_fullinfo() function so that if it becomes variable in the future, we  
 don't have to change that code. */  
1340    
1341  extra = (pcre_extra *)(pcre_malloc)    if (min > 0)
1342    (sizeof(pcre_extra) + sizeof(pcre_study_data));      {
1343        study->flags |= PCRE_STUDY_MINLEN;
1344        study->minlength = min;
1345        }
1346      else study->minlength = 0;
1347    
1348  if (extra == NULL)    /* If JIT support was compiled and requested, attempt the JIT compilation.
1349    {    If no starting bytes were found, and the minimum length is zero, and JIT
1350    *errorptr = "failed to get memory";    compilation fails, abandon the extra block and return NULL. */
1351    return NULL;  
1352    #ifdef SUPPORT_JIT
1353      extra->executable_jit = NULL;
1354      if ((options & PCRE_STUDY_JIT_COMPILE) != 0) _pcre_jit_compile(re, extra);
1355      if (study->flags == 0 && (extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) == 0)
1356        {
1357        pcre_free_study(extra);
1358        extra = NULL;
1359        }
1360    #endif
1361    }    }
1362    
1363  study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));  return extra;
1364  extra->flags = PCRE_EXTRA_STUDY_DATA;  }
 extra->study_data = study;  
1365    
 study->size = sizeof(pcre_study_data);  
 study->flags = 0;  
1366    
1367  if (bits_set)  /*************************************************
1368    {  *          Free the study data                   *
1369    study->flags |= PCRE_STUDY_MAPPED;  *************************************************/
   memcpy(study->start_bits, start_bits, sizeof(start_bits));  
   }  
1370    
1371  if (min >= 0)  /* This function frees the memory that was obtained by pcre_study().
   {  
   study->flags |= PCRE_STUDY_MINLEN;  
   study->minlength = min;  
   }  
1372    
1373  return extra;  Argument:   a pointer to the pcre_extra block
1374    Returns:    nothing
1375    */
1376    
1377    PCRE_EXP_DEFN void
1378    pcre_free_study(pcre_extra *extra)
1379    {
1380    #ifdef SUPPORT_JIT
1381    if ((extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 &&
1382         extra->executable_jit != NULL)
1383      _pcre_jit_free(extra->executable_jit);
1384    #endif
1385    pcre_free(extra);
1386  }  }
1387    
1388  /* End of pcre_study.c */  /* End of pcre_study.c */

Legend:
Removed from v.538  
changed lines
  Added in v.797

  ViewVC Help
Powered by ViewVC 1.1.5