/[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 335 by ph10, Sat Apr 12 14:36:14 2008 UTC revision 620 by ph10, Sun Jul 17 13:53:14 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 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    
59    /*************************************************
60    *   Find the minimum subject length for a group  *
61    *************************************************/
62    
63    /* Scan a parenthesized group and compute the minimum length of subject that
64    is needed to match it. This is a lower bound; it does not mean there is a
65    string of that length that matches. In UTF8 mode, the result is in characters
66    rather than bytes.
67    
68    Arguments:
69      code        pointer to start of group (the bracket)
70      startcode   pointer to start of the whole pattern
71      options     the compiling options
72      had_accept  pointer to flag for (*ACCEPT) encountered
73      int         RECURSE depth
74    
75    Returns:   the minimum length
76               -1 if \C was encountered
77               -2 internal error (missing capturing bracket)
78               -3 internal error (opcode not listed)
79    */
80    
81    static int
82    find_minlength(const uschar *code, const uschar *startcode, int options,
83      BOOL *had_accept_ptr, int recurse_depth)
84    {
85    int length = -1;
86    BOOL utf8 = (options & PCRE_UTF8) != 0;
87    BOOL had_recurse = FALSE;
88    register int branchlength = 0;
89    register uschar *cc = (uschar *)code + 1 + LINK_SIZE;
90    
91    if (*code == OP_CBRA || *code == OP_SCBRA ||
92        *code == OP_CBRAPOS || *code == OP_SCBRAPOS) cc += 2;
93    
94    /* Scan along the opcodes for this branch. If we get to the end of the
95    branch, check the length against that of the other branches. */
96    
97    for (;;)
98      {
99      int d, min;
100      uschar *cs, *ce;
101      register int op = *cc;
102    
103      switch (op)
104        {
105        case OP_COND:
106        case OP_SCOND:
107    
108        /* If there is only one branch in a condition, the implied branch has zero
109        length, so we don't add anything. This covers the DEFINE "condition"
110        automatically. */
111    
112        cs = cc + GET(cc, 1);
113        if (*cs != OP_ALT)
114          {
115          cc = cs + 1 + LINK_SIZE;
116          break;
117          }
118    
119        /* Otherwise we can fall through and treat it the same as any other
120        subpattern. */
121    
122        case OP_CBRA:
123        case OP_SCBRA:
124        case OP_BRA:
125        case OP_SBRA:
126        case OP_CBRAPOS:
127        case OP_SCBRAPOS:
128        case OP_BRAPOS:
129        case OP_SBRAPOS:
130        case OP_ONCE:
131        d = find_minlength(cc, startcode, options, had_accept_ptr, recurse_depth);
132        if (d < 0) return d;
133        branchlength += d;
134        if (*had_accept_ptr) return branchlength;
135        do cc += GET(cc, 1); while (*cc == OP_ALT);
136        cc += 1 + LINK_SIZE;
137        break;
138    
139        /* Reached end of a branch; if it's a ket it is the end of a nested
140        call. If it's ALT it is an alternation in a nested call. If it is END it's
141        the end of the outer call. All can be handled by the same code. If it is
142        ACCEPT, it is essentially the same as END, but we set a flag so that
143        counting stops. */
144    
145        case OP_ACCEPT:
146        case OP_ASSERT_ACCEPT:
147        *had_accept_ptr = TRUE;
148        /* Fall through */
149        case OP_ALT:
150        case OP_KET:
151        case OP_KETRMAX:
152        case OP_KETRMIN:
153        case OP_KETRPOS:
154        case OP_END:
155        if (length < 0 || (!had_recurse && branchlength < length))
156          length = branchlength;
157        if (op != OP_ALT) return length;
158        cc += 1 + LINK_SIZE;
159        branchlength = 0;
160        had_recurse = FALSE;
161        break;
162    
163        /* Skip over assertive subpatterns */
164    
165        case OP_ASSERT:
166        case OP_ASSERT_NOT:
167        case OP_ASSERTBACK:
168        case OP_ASSERTBACK_NOT:
169        do cc += GET(cc, 1); while (*cc == OP_ALT);
170        /* Fall through */
171    
172        /* Skip over things that don't match chars */
173    
174        case OP_REVERSE:
175        case OP_CREF:
176        case OP_NCREF:
177        case OP_RREF:
178        case OP_NRREF:
179        case OP_DEF:
180        case OP_CALLOUT:
181        case OP_SOD:
182        case OP_SOM:
183        case OP_EOD:
184        case OP_EODN:
185        case OP_CIRC:
186        case OP_CIRCM:
187        case OP_DOLL:
188        case OP_DOLLM:
189        case OP_NOT_WORD_BOUNDARY:
190        case OP_WORD_BOUNDARY:
191        cc += _pcre_OP_lengths[*cc];
192        break;
193    
194        /* Skip over a subpattern that has a {0} or {0,x} quantifier */
195    
196        case OP_BRAZERO:
197        case OP_BRAMINZERO:
198        case OP_BRAPOSZERO:
199        case OP_SKIPZERO:
200        cc += _pcre_OP_lengths[*cc];
201        do cc += GET(cc, 1); while (*cc == OP_ALT);
202        cc += 1 + LINK_SIZE;
203        break;
204    
205        /* Handle literal characters and + repetitions */
206    
207        case OP_CHAR:
208        case OP_CHARI:
209        case OP_NOT:
210        case OP_NOTI:
211        case OP_PLUS:
212        case OP_PLUSI:
213        case OP_MINPLUS:
214        case OP_MINPLUSI:
215        case OP_POSPLUS:
216        case OP_POSPLUSI:
217        case OP_NOTPLUS:
218        case OP_NOTPLUSI:
219        case OP_NOTMINPLUS:
220        case OP_NOTMINPLUSI:
221        case OP_NOTPOSPLUS:
222        case OP_NOTPOSPLUSI:
223        branchlength++;
224        cc += 2;
225    #ifdef SUPPORT_UTF8
226        if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
227    #endif
228        break;
229    
230        case OP_TYPEPLUS:
231        case OP_TYPEMINPLUS:
232        case OP_TYPEPOSPLUS:
233        branchlength++;
234        cc += (cc[1] == OP_PROP || cc[1] == OP_NOTPROP)? 4 : 2;
235        break;
236    
237        /* Handle exact repetitions. The count is already in characters, but we
238        need to skip over a multibyte character in UTF8 mode.  */
239    
240        case OP_EXACT:
241        case OP_EXACTI:
242        case OP_NOTEXACT:
243        case OP_NOTEXACTI:
244        branchlength += GET2(cc,1);
245        cc += 4;
246    #ifdef SUPPORT_UTF8
247        if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
248    #endif
249        break;
250    
251        case OP_TYPEEXACT:
252        branchlength += GET2(cc,1);
253        cc += (cc[3] == OP_PROP || cc[3] == OP_NOTPROP)? 6 : 4;
254        break;
255    
256        /* Handle single-char non-literal matchers */
257    
258        case OP_PROP:
259        case OP_NOTPROP:
260        cc += 2;
261        /* Fall through */
262    
263        case OP_NOT_DIGIT:
264        case OP_DIGIT:
265        case OP_NOT_WHITESPACE:
266        case OP_WHITESPACE:
267        case OP_NOT_WORDCHAR:
268        case OP_WORDCHAR:
269        case OP_ANY:
270        case OP_ALLANY:
271        case OP_EXTUNI:
272        case OP_HSPACE:
273        case OP_NOT_HSPACE:
274        case OP_VSPACE:
275        case OP_NOT_VSPACE:
276        branchlength++;
277        cc++;
278        break;
279    
280        /* "Any newline" might match two characters, but it also might match just
281        one. */
282    
283        case OP_ANYNL:
284        branchlength += 1;
285        cc++;
286        break;
287    
288        /* The single-byte matcher means we can't proceed in UTF-8 mode */
289    
290        case OP_ANYBYTE:
291    #ifdef SUPPORT_UTF8
292        if (utf8) return -1;
293    #endif
294        branchlength++;
295        cc++;
296        break;
297    
298        /* For repeated character types, we have to test for \p and \P, which have
299        an extra two bytes of parameters. */
300    
301        case OP_TYPESTAR:
302        case OP_TYPEMINSTAR:
303        case OP_TYPEQUERY:
304        case OP_TYPEMINQUERY:
305        case OP_TYPEPOSSTAR:
306        case OP_TYPEPOSQUERY:
307        if (cc[1] == OP_PROP || cc[1] == OP_NOTPROP) cc += 2;
308        cc += _pcre_OP_lengths[op];
309        break;
310    
311        case OP_TYPEUPTO:
312        case OP_TYPEMINUPTO:
313        case OP_TYPEPOSUPTO:
314        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
315        cc += _pcre_OP_lengths[op];
316        break;
317    
318        /* Check a class for variable quantification */
319    
320    #ifdef SUPPORT_UTF8
321        case OP_XCLASS:
322        cc += GET(cc, 1) - 33;
323        /* Fall through */
324    #endif
325    
326        case OP_CLASS:
327        case OP_NCLASS:
328        cc += 33;
329    
330        switch (*cc)
331          {
332          case OP_CRPLUS:
333          case OP_CRMINPLUS:
334          branchlength++;
335          /* Fall through */
336    
337          case OP_CRSTAR:
338          case OP_CRMINSTAR:
339          case OP_CRQUERY:
340          case OP_CRMINQUERY:
341          cc++;
342          break;
343    
344          case OP_CRRANGE:
345          case OP_CRMINRANGE:
346          branchlength += GET2(cc,1);
347          cc += 5;
348          break;
349    
350          default:
351          branchlength++;
352          break;
353          }
354        break;
355    
356        /* Backreferences and subroutine calls are treated in the same way: we find
357        the minimum length for the subpattern. A recursion, however, causes an
358        a flag to be set that causes the length of this branch to be ignored. The
359        logic is that a recursion can only make sense if there is another
360        alternation that stops the recursing. That will provide the minimum length
361        (when no recursion happens). A backreference within the group that it is
362        referencing behaves in the same way.
363    
364        If PCRE_JAVASCRIPT_COMPAT is set, a backreference to an unset bracket
365        matches an empty string (by default it causes a matching failure), so in
366        that case we must set the minimum length to zero. */
367    
368        case OP_REF:
369        case OP_REFI:
370        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
371          {
372          ce = cs = (uschar *)_pcre_find_bracket(startcode, utf8, GET2(cc, 1));
373          if (cs == NULL) return -2;
374          do ce += GET(ce, 1); while (*ce == OP_ALT);
375          if (cc > cs && cc < ce)
376            {
377            d = 0;
378            had_recurse = TRUE;
379            }
380          else
381            {
382            d = find_minlength(cs, startcode, options, had_accept_ptr,
383              recurse_depth);
384            *had_accept_ptr = FALSE;
385            }
386          }
387        else d = 0;
388        cc += 3;
389    
390        /* Handle repeated back references */
391    
392        switch (*cc)
393          {
394          case OP_CRSTAR:
395          case OP_CRMINSTAR:
396          case OP_CRQUERY:
397          case OP_CRMINQUERY:
398          min = 0;
399          cc++;
400          break;
401    
402          case OP_CRPLUS:
403          case OP_CRMINPLUS:
404          min = 1;
405          cc++;
406          break;
407    
408          case OP_CRRANGE:
409          case OP_CRMINRANGE:
410          min = GET2(cc, 1);
411          cc += 5;
412          break;
413    
414          default:
415          min = 1;
416          break;
417          }
418    
419        branchlength += min * d;
420        break;
421    
422        /* We can easily detect direct recursion, but not mutual recursion. This is
423        caught by a recursion depth count. */
424    
425        case OP_RECURSE:
426        cs = ce = (uschar *)startcode + GET(cc, 1);
427        if (cs == NULL) return -2;
428        do ce += GET(ce, 1); while (*ce == OP_ALT);
429        if ((cc > cs && cc < ce) || recurse_depth > 10)
430          had_recurse = TRUE;
431        else
432          {
433          branchlength += find_minlength(cs, startcode, options, had_accept_ptr,
434            recurse_depth + 1);
435          *had_accept_ptr = FALSE;
436          }
437        cc += 1 + LINK_SIZE;
438        break;
439    
440        /* Anything else does not or need not match a character. We can get the
441        item's length from the table, but for those that can match zero occurrences
442        of a character, we must take special action for UTF-8 characters. As it
443        happens, the "NOT" versions of these opcodes are used at present only for
444        ASCII characters, so they could be omitted from this list. However, in
445        future that may change, so we include them here so as not to leave a
446        gotcha for a future maintainer. */
447    
448        case OP_UPTO:
449        case OP_UPTOI:
450        case OP_NOTUPTO:
451        case OP_NOTUPTOI:
452        case OP_MINUPTO:
453        case OP_MINUPTOI:
454        case OP_NOTMINUPTO:
455        case OP_NOTMINUPTOI:
456        case OP_POSUPTO:
457        case OP_POSUPTOI:
458        case OP_NOTPOSUPTO:
459        case OP_NOTPOSUPTOI:
460    
461        case OP_STAR:
462        case OP_STARI:
463        case OP_NOTSTAR:
464        case OP_NOTSTARI:
465        case OP_MINSTAR:
466        case OP_MINSTARI:
467        case OP_NOTMINSTAR:
468        case OP_NOTMINSTARI:
469        case OP_POSSTAR:
470        case OP_POSSTARI:
471        case OP_NOTPOSSTAR:
472        case OP_NOTPOSSTARI:
473    
474        case OP_QUERY:
475        case OP_QUERYI:
476        case OP_NOTQUERY:
477        case OP_NOTQUERYI:
478        case OP_MINQUERY:
479        case OP_MINQUERYI:
480        case OP_NOTMINQUERY:
481        case OP_NOTMINQUERYI:
482        case OP_POSQUERY:
483        case OP_POSQUERYI:
484        case OP_NOTPOSQUERY:
485        case OP_NOTPOSQUERYI:
486    
487        cc += _pcre_OP_lengths[op];
488    #ifdef SUPPORT_UTF8
489        if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
490    #endif
491        break;
492    
493        /* Skip these, but we need to add in the name length. */
494    
495        case OP_MARK:
496        case OP_PRUNE_ARG:
497        case OP_SKIP_ARG:
498        cc += _pcre_OP_lengths[op] + cc[1];
499        break;
500    
501        case OP_THEN_ARG:
502        cc += _pcre_OP_lengths[op] + cc[1+LINK_SIZE];
503        break;
504    
505        /* The remaining opcodes are just skipped over. */
506    
507        case OP_CLOSE:
508        case OP_COMMIT:
509        case OP_FAIL:
510        case OP_PRUNE:
511        case OP_SET_SOM:
512        case OP_SKIP:
513        case OP_THEN:
514        cc += _pcre_OP_lengths[op];
515        break;
516    
517        /* This should not occur: we list all opcodes explicitly so that when
518        new ones get added they are properly considered. */
519    
520        default:
521        return -3;
522        }
523      }
524    /* Control never gets here */
525    }
526    
527    
528    
529  /*************************************************  /*************************************************
530  *      Set a bit and maybe its alternate case    *  *      Set a bit and maybe its alternate case    *
531  *************************************************/  *************************************************/
532    
533  /* Given a character, set its bit in the table, and also the bit for the other  /* Given a character, set its first byte's bit in the table, and also the
534  version of a letter if we are caseless.  corresponding bit for the other version of a letter if we are caseless. In
535    UTF-8 mode, for characters greater than 127, we can only do the caseless thing
536    when Unicode property support is available.
537    
538  Arguments:  Arguments:
539    start_bits    points to the bit map    start_bits    points to the bit map
540    c             is the character    p             points to the character
541    caseless      the caseless flag    caseless      the caseless flag
542    cd            the block with char table pointers    cd            the block with char table pointers
543      utf8          TRUE for UTF-8 mode
544    
545  Returns:        nothing  Returns:        pointer after the character
546    */
547    
548    static const uschar *
549    set_table_bit(uschar *start_bits, const uschar *p, BOOL caseless,
550      compile_data *cd, BOOL utf8)
551    {
552    unsigned int c = *p;
553    
554    SET_BIT(c);
555    
556    #ifdef SUPPORT_UTF8
557    if (utf8 && c > 127)
558      {
559      GETCHARINC(c, p);
560    #ifdef SUPPORT_UCP
561      if (caseless)
562        {
563        uschar buff[8];
564        c = UCD_OTHERCASE(c);
565        (void)_pcre_ord2utf8(c, buff);
566        SET_BIT(buff[0]);
567        }
568    #endif
569      return p;
570      }
571    #endif
572    
573    /* Not UTF-8 mode, or character is less than 127. */
574    
575    if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
576    return p + 1;
577    }
578    
579    
580    
581    /*************************************************
582    *     Set bits for a positive character type     *
583    *************************************************/
584    
585    /* This function sets starting bits for a character type. In UTF-8 mode, we can
586    only do a direct setting for bytes less than 128, as otherwise there can be
587    confusion with bytes in the middle of UTF-8 characters. In a "traditional"
588    environment, the tables will only recognize ASCII characters anyway, but in at
589    least one Windows environment, some higher bytes bits were set in the tables.
590    So we deal with that case by considering the UTF-8 encoding.
591    
592    Arguments:
593      start_bits     the starting bitmap
594      cbit type      the type of character wanted
595      table_limit    32 for non-UTF-8; 16 for UTF-8
596      cd             the block with char table pointers
597    
598    Returns:         nothing
599    */
600    
601    static void
602    set_type_bits(uschar *start_bits, int cbit_type, int table_limit,
603      compile_data *cd)
604    {
605    register int c;
606    for (c = 0; c < table_limit; c++) start_bits[c] |= cd->cbits[c+cbit_type];
607    if (table_limit == 32) return;
608    for (c = 128; c < 256; c++)
609      {
610      if ((cd->cbits[c/8] & (1 << (c&7))) != 0)
611        {
612        uschar buff[8];
613        (void)_pcre_ord2utf8(c, buff);
614        SET_BIT(buff[0]);
615        }
616      }
617    }
618    
619    
620    /*************************************************
621    *     Set bits for a negative character type     *
622    *************************************************/
623    
624    /* This function sets starting bits for a negative character type such as \D.
625    In UTF-8 mode, we can only do a direct setting for bytes less than 128, as
626    otherwise there can be confusion with bytes in the middle of UTF-8 characters.
627    Unlike in the positive case, where we can set appropriate starting bits for
628    specific high-valued UTF-8 characters, in this case we have to set the bits for
629    all high-valued characters. The lowest is 0xc2, but we overkill by starting at
630    0xc0 (192) for simplicity.
631    
632    Arguments:
633      start_bits     the starting bitmap
634      cbit type      the type of character wanted
635      table_limit    32 for non-UTF-8; 16 for UTF-8
636      cd             the block with char table pointers
637    
638    Returns:         nothing
639  */  */
640    
641  static void  static void
642  set_bit(uschar *start_bits, unsigned int c, BOOL caseless, compile_data *cd)  set_nottype_bits(uschar *start_bits, int cbit_type, int table_limit,
643      compile_data *cd)
644  {  {
645  start_bits[c/8] |= (1 << (c&7));  register int c;
646  if (caseless && (cd->ctypes[c] & ctype_letter) != 0)  for (c = 0; c < table_limit; c++) start_bits[c] |= ~cd->cbits[c+cbit_type];
647    start_bits[cd->fcc[c]/8] |= (1 << (cd->fcc[c]&7));  if (table_limit != 32) for (c = 24; c < 32; c++) start_bits[c] = 0xff;
648  }  }
649    
650    
# Line 95  function fails unless the result is SSB_ Line 664  function fails unless the result is SSB_
664  Arguments:  Arguments:
665    code         points to an expression    code         points to an expression
666    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  
667    utf8         TRUE if in UTF-8 mode    utf8         TRUE if in UTF-8 mode
668    cd           the block with char table pointers    cd           the block with char table pointers
669    
670  Returns:       SSB_FAIL     => Failed to find any starting bytes  Returns:       SSB_FAIL     => Failed to find any starting bytes
671                 SSB_DONE     => Found mandatory starting bytes                 SSB_DONE     => Found mandatory starting bytes
672                 SSB_CONTINUE => Found optional starting bytes                 SSB_CONTINUE => Found optional starting bytes
673                   SSB_UNKNOWN  => Hit an unrecognized opcode
674  */  */
675    
676  static int  static int
677  set_start_bits(const uschar *code, uschar *start_bits, BOOL caseless,  set_start_bits(const uschar *code, uschar *start_bits, BOOL utf8,
678    BOOL utf8, compile_data *cd)    compile_data *cd)
679  {  {
680  register int c;  register int c;
681  int yield = SSB_DONE;  int yield = SSB_DONE;
682    int table_limit = utf8? 16:32;
683    
684  #if 0  #if 0
685  /* ========================================================================= */  /* ========================================================================= */
# Line 130  volatile int dummy; Line 700  volatile int dummy;
700    
701  do  do
702    {    {
   const uschar *tcode = code + (((int)*code == OP_CBRA)? 3:1) + LINK_SIZE;  
703    BOOL try_next = TRUE;    BOOL try_next = TRUE;
704      const uschar *tcode = code + 1 + LINK_SIZE;
705    
706      if (*code == OP_CBRA || *code == OP_SCBRA ||
707          *code == OP_CBRAPOS || *code == OP_SCBRAPOS) tcode += 2;
708    
709    while (try_next)    /* Loop for items in this branch */    while (try_next)    /* Loop for items in this branch */
710      {      {
711      int rc;      int rc;
712    
713      switch(*tcode)      switch(*tcode)
714        {        {
715        /* Fail if we reach something we don't understand */        /* If we reach something we don't understand, it means a new opcode has
716          been created that hasn't been added to this code. Hopefully this problem
717          will be discovered during testing. */
718    
719        default:        default:
720          return SSB_UNKNOWN;
721    
722          /* Fail for a valid opcode that implies no starting bits. */
723    
724          case OP_ACCEPT:
725          case OP_ASSERT_ACCEPT:
726          case OP_ALLANY:
727          case OP_ANY:
728          case OP_ANYBYTE:
729          case OP_CIRC:
730          case OP_CIRCM:
731          case OP_CLOSE:
732          case OP_COMMIT:
733          case OP_COND:
734          case OP_CREF:
735          case OP_DEF:
736          case OP_DOLL:
737          case OP_DOLLM:
738          case OP_END:
739          case OP_EOD:
740          case OP_EODN:
741          case OP_EXTUNI:
742          case OP_FAIL:
743          case OP_MARK:
744          case OP_NCREF:
745          case OP_NOT:
746          case OP_NOTEXACT:
747          case OP_NOTEXACTI:
748          case OP_NOTI:
749          case OP_NOTMINPLUS:
750          case OP_NOTMINPLUSI:
751          case OP_NOTMINQUERY:
752          case OP_NOTMINQUERYI:
753          case OP_NOTMINSTAR:
754          case OP_NOTMINSTARI:
755          case OP_NOTMINUPTO:
756          case OP_NOTMINUPTOI:
757          case OP_NOTPLUS:
758          case OP_NOTPLUSI:
759          case OP_NOTPOSPLUS:
760          case OP_NOTPOSPLUSI:
761          case OP_NOTPOSQUERY:
762          case OP_NOTPOSQUERYI:
763          case OP_NOTPOSSTAR:
764          case OP_NOTPOSSTARI:
765          case OP_NOTPOSUPTO:
766          case OP_NOTPOSUPTOI:
767          case OP_NOTPROP:
768          case OP_NOTQUERY:
769          case OP_NOTQUERYI:
770          case OP_NOTSTAR:
771          case OP_NOTSTARI:
772          case OP_NOTUPTO:
773          case OP_NOTUPTOI:
774          case OP_NOT_HSPACE:
775          case OP_NOT_VSPACE:
776          case OP_NOT_WORD_BOUNDARY:
777          case OP_NRREF:
778          case OP_PROP:
779          case OP_PRUNE:
780          case OP_PRUNE_ARG:
781          case OP_RECURSE:
782          case OP_REF:
783          case OP_REFI:
784          case OP_REVERSE:
785          case OP_RREF:
786          case OP_SCOND:
787          case OP_SET_SOM:
788          case OP_SKIP:
789          case OP_SKIP_ARG:
790          case OP_SOD:
791          case OP_SOM:
792          case OP_THEN:
793          case OP_THEN_ARG:
794          case OP_WORD_BOUNDARY:
795          case OP_XCLASS:
796        return SSB_FAIL;        return SSB_FAIL;
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
# Line 152  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_ASSERT:        case OP_ASSERT:
813        rc = set_start_bits(tcode, start_bits, caseless, utf8, cd);        rc = set_start_bits(tcode, start_bits, utf8, cd);
814        if (rc == SSB_FAIL) return SSB_FAIL;        if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
815        if (rc == SSB_DONE) try_next = FALSE; else        if (rc == SSB_DONE) try_next = FALSE; else
816          {          {
817          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
# Line 178  do Line 834  do
834        case OP_KET:        case OP_KET:
835        case OP_KETRMAX:        case OP_KETRMAX:
836        case OP_KETRMIN:        case OP_KETRMIN:
837          case OP_KETRPOS:
838        return SSB_CONTINUE;        return SSB_CONTINUE;
839    
840        /* Skip over callout */        /* Skip over callout */
# Line 195  do Line 852  do
852        tcode += 1 + LINK_SIZE;        tcode += 1 + LINK_SIZE;
853        break;        break;
854    
       /* Skip over an option setting, changing the caseless flag */  
   
       case OP_OPT:  
       caseless = (tcode[1] & PCRE_CASELESS) != 0;  
       tcode += 2;  
       break;  
   
855        /* BRAZERO does the bracket, but carries on. */        /* BRAZERO does the bracket, but carries on. */
856    
857        case OP_BRAZERO:        case OP_BRAZERO:
858        case OP_BRAMINZERO:        case OP_BRAMINZERO:
859        if (set_start_bits(++tcode, start_bits, caseless, utf8, cd) == SSB_FAIL)        case OP_BRAPOSZERO:
860          return SSB_FAIL;        rc = set_start_bits(++tcode, start_bits, utf8, cd);
861          if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
862  /* =========================================================================  /* =========================================================================
863        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,
864        which was an old fudge for the benefit of OS/2.        which was an old fudge for the benefit of OS/2.
# Line 220  do Line 871  do
871        /* SKIPZERO skips the bracket. */        /* SKIPZERO skips the bracket. */
872    
873        case OP_SKIPZERO:        case OP_SKIPZERO:
874          tcode++;
875        do tcode += GET(tcode,1); while (*tcode == OP_ALT);        do tcode += GET(tcode,1); while (*tcode == OP_ALT);
876        tcode += 1 + LINK_SIZE;        tcode += 1 + LINK_SIZE;
877        break;        break;
# Line 232  do Line 884  do
884        case OP_QUERY:        case OP_QUERY:
885        case OP_MINQUERY:        case OP_MINQUERY:
886        case OP_POSQUERY:        case OP_POSQUERY:
887        set_bit(start_bits, tcode[1], caseless, cd);        tcode = set_table_bit(start_bits, tcode + 1, FALSE, cd, utf8);
888        tcode += 2;        break;
889  #ifdef SUPPORT_UTF8  
890        if (utf8 && tcode[-1] >= 0xc0)        case OP_STARI:
891          tcode += _pcre_utf8_table4[tcode[-1] & 0x3f];        case OP_MINSTARI:
892  #endif        case OP_POSSTARI:
893          case OP_QUERYI:
894          case OP_MINQUERYI:
895          case OP_POSQUERYI:
896          tcode = set_table_bit(start_bits, tcode + 1, TRUE, cd, utf8);
897        break;        break;
898    
899        /* Single-char upto sets the bit and tries the next */        /* Single-char upto sets the bit and tries the next */
# Line 245  do Line 901  do
901        case OP_UPTO:        case OP_UPTO:
902        case OP_MINUPTO:        case OP_MINUPTO:
903        case OP_POSUPTO:        case OP_POSUPTO:
904        set_bit(start_bits, tcode[3], caseless, cd);        tcode = set_table_bit(start_bits, tcode + 3, FALSE, cd, utf8);
905        tcode += 4;        break;
906  #ifdef SUPPORT_UTF8  
907        if (utf8 && tcode[-1] >= 0xc0)        case OP_UPTOI:
908          tcode += _pcre_utf8_table4[tcode[-1] & 0x3f];        case OP_MINUPTOI:
909  #endif        case OP_POSUPTOI:
910          tcode = set_table_bit(start_bits, tcode + 3, TRUE, cd, utf8);
911        break;        break;
912    
913        /* At least one single char sets the bit and stops */        /* At least one single char sets the bit and stops */
914    
915        case OP_EXACT:       /* Fall through */        case OP_EXACT:
916        tcode += 2;        tcode += 2;
917          /* Fall through */
918        case OP_CHAR:        case OP_CHAR:
       case OP_CHARNC:  
919        case OP_PLUS:        case OP_PLUS:
920        case OP_MINPLUS:        case OP_MINPLUS:
921        case OP_POSPLUS:        case OP_POSPLUS:
922        set_bit(start_bits, tcode[1], caseless, cd);        (void)set_table_bit(start_bits, tcode + 1, FALSE, cd, utf8);
923          try_next = FALSE;
924          break;
925    
926          case OP_EXACTI:
927          tcode += 2;
928          /* Fall through */
929          case OP_CHARI:
930          case OP_PLUSI:
931          case OP_MINPLUSI:
932          case OP_POSPLUSI:
933          (void)set_table_bit(start_bits, tcode + 1, TRUE, cd, utf8);
934          try_next = FALSE;
935          break;
936    
937          /* Special spacing and line-terminating items. These recognize specific
938          lists of characters. The difference between VSPACE and ANYNL is that the
939          latter can match the two-character CRLF sequence, but that is not
940          relevant for finding the first character, so their code here is
941          identical. */
942    
943          case OP_HSPACE:
944          SET_BIT(0x09);
945          SET_BIT(0x20);
946          if (utf8)
947            {
948            SET_BIT(0xC2);  /* For U+00A0 */
949            SET_BIT(0xE1);  /* For U+1680, U+180E */
950            SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
951            SET_BIT(0xE3);  /* For U+3000 */
952            }
953          else SET_BIT(0xA0);
954          try_next = FALSE;
955          break;
956    
957          case OP_ANYNL:
958          case OP_VSPACE:
959          SET_BIT(0x0A);
960          SET_BIT(0x0B);
961          SET_BIT(0x0C);
962          SET_BIT(0x0D);
963          if (utf8)
964            {
965            SET_BIT(0xC2);  /* For U+0085 */
966            SET_BIT(0xE2);  /* For U+2028, U+2029 */
967            }
968          else SET_BIT(0x85);
969        try_next = FALSE;        try_next = FALSE;
970        break;        break;
971    
972        /* Single character type sets the bits and stops */        /* Single character types set the bits and stop. Note that if PCRE_UCP
973          is set, we do not see these op codes because \d etc are converted to
974          properties. Therefore, these apply in the case when only characters less
975          than 256 are recognized to match the types. */
976    
977        case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
978        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_digit];  
979        try_next = FALSE;        try_next = FALSE;
980        break;        break;
981    
982        case OP_DIGIT:        case OP_DIGIT:
983        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_digit];  
984        try_next = FALSE;        try_next = FALSE;
985        break;        break;
986    
987        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to
988        discard it. */        ensure it is set as not whitespace. */
989    
990        case OP_NOT_WHITESPACE:        case OP_NOT_WHITESPACE:
991        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_space, table_limit, cd);
992          {        start_bits[1] |= 0x08;
         int d = cd->cbits[c+cbit_space];  
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= ~d;  
         }  
993        try_next = FALSE;        try_next = FALSE;
994        break;        break;
995    
996        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to
997        discard it. */        not set it from the table. */
998    
999        case OP_WHITESPACE:        case OP_WHITESPACE:
1000        for (c = 0; c < 32; c++)        c = start_bits[1];    /* Save in case it was already set */
1001          {        set_type_bits(start_bits, cbit_space, table_limit, cd);
1002          int d = cd->cbits[c+cbit_space];        start_bits[1] = (start_bits[1] & ~0x08) | c;
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= d;  
         }  
1003        try_next = FALSE;        try_next = FALSE;
1004        break;        break;
1005    
1006        case OP_NOT_WORDCHAR:        case OP_NOT_WORDCHAR:
1007        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_word];  
1008        try_next = FALSE;        try_next = FALSE;
1009        break;        break;
1010    
1011        case OP_WORDCHAR:        case OP_WORDCHAR:
1012        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_word];  
1013        try_next = FALSE;        try_next = FALSE;
1014        break;        break;
1015    
# Line 324  do Line 1018  do
1018    
1019        case OP_TYPEPLUS:        case OP_TYPEPLUS:
1020        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
1021          case OP_TYPEPOSPLUS:
1022        tcode++;        tcode++;
1023        break;        break;
1024    
# Line 347  do Line 1042  do
1042        case OP_TYPEPOSQUERY:        case OP_TYPEPOSQUERY:
1043        switch(tcode[1])        switch(tcode[1])
1044          {          {
1045            default:
1046          case OP_ANY:          case OP_ANY:
1047            case OP_ALLANY:
1048          return SSB_FAIL;          return SSB_FAIL;
1049    
1050            case OP_HSPACE:
1051            SET_BIT(0x09);
1052            SET_BIT(0x20);
1053            if (utf8)
1054              {
1055              SET_BIT(0xC2);  /* For U+00A0 */
1056              SET_BIT(0xE1);  /* For U+1680, U+180E */
1057              SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
1058              SET_BIT(0xE3);  /* For U+3000 */
1059              }
1060            else SET_BIT(0xA0);
1061            break;
1062    
1063            case OP_ANYNL:
1064            case OP_VSPACE:
1065            SET_BIT(0x0A);
1066            SET_BIT(0x0B);
1067            SET_BIT(0x0C);
1068            SET_BIT(0x0D);
1069            if (utf8)
1070              {
1071              SET_BIT(0xC2);  /* For U+0085 */
1072              SET_BIT(0xE2);  /* For U+2028, U+2029 */
1073              }
1074            else SET_BIT(0x85);
1075            break;
1076    
1077          case OP_NOT_DIGIT:          case OP_NOT_DIGIT:
1078          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_digit];  
1079          break;          break;
1080    
1081          case OP_DIGIT:          case OP_DIGIT:
1082          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_digit];  
1083          break;          break;
1084    
1085          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we have to
1086          discard it. */          ensure it gets set as not whitespace. */
1087    
1088          case OP_NOT_WHITESPACE:          case OP_NOT_WHITESPACE:
1089          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_space, table_limit, cd);
1090            {          start_bits[1] |= 0x08;
           int d = cd->cbits[c+cbit_space];  
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= ~d;  
           }  
1091          break;          break;
1092    
1093          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we have to
1094          discard it. */          avoid setting it. */
1095    
1096          case OP_WHITESPACE:          case OP_WHITESPACE:
1097          for (c = 0; c < 32; c++)          c = start_bits[1];    /* Save in case it was already set */
1098            {          set_type_bits(start_bits, cbit_space, table_limit, cd);
1099            int d = cd->cbits[c+cbit_space];          start_bits[1] = (start_bits[1] & ~0x08) | c;
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= d;  
           }  
1100          break;          break;
1101    
1102          case OP_NOT_WORDCHAR:          case OP_NOT_WORDCHAR:
1103          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_word];  
1104          break;          break;
1105    
1106          case OP_WORDCHAR:          case OP_WORDCHAR:
1107          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_word];  
1108          break;          break;
1109          }          }
1110    
# Line 447  do Line 1160  do
1160            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];
1161            }            }
1162    
1163          /* Advance past the bit map, and act on what follows */          /* Advance past the bit map, and act on what follows. For a zero
1164            minimum repeat, continue; otherwise stop processing. */
1165    
1166          tcode += 32;          tcode += 32;
1167          switch (*tcode)          switch (*tcode)
# Line 464  do Line 1178  do
1178            if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;            if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;
1179              else try_next = FALSE;              else try_next = FALSE;
1180            break;            break;
1181    
1182            default:            default:
1183            try_next = FALSE;            try_next = FALSE;
1184            break;            break;
# Line 483  return yield; Line 1197  return yield;
1197    
1198    
1199    
1200    
1201    
1202  /*************************************************  /*************************************************
1203  *          Study a compiled expression           *  *          Study a compiled expression           *
1204  *************************************************/  *************************************************/
# Line 498  Arguments: Line 1214  Arguments:
1214              set NULL unless error              set NULL unless error
1215    
1216  Returns:    pointer to a pcre_extra block, with study_data filled in and the  Returns:    pointer to a pcre_extra block, with study_data filled in and the
1217                appropriate flag set;                appropriate flags set;
1218              NULL on error or if no optimization possible              NULL on error or if no optimization possible
1219  */  */
1220    
1221  PCRE_EXP_DEFN pcre_extra *  PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION
1222  pcre_study(const pcre *external_re, int options, const char **errorptr)  pcre_study(const pcre *external_re, int options, const char **errorptr)
1223  {  {
1224    int min;
1225    BOOL bits_set = FALSE;
1226    BOOL had_accept = FALSE;
1227  uschar start_bits[32];  uschar start_bits[32];
1228  pcre_extra *extra;  pcre_extra *extra;
1229  pcre_study_data *study;  pcre_study_data *study;
# Line 531  code = (uschar *)re + re->name_table_off Line 1250  code = (uschar *)re + re->name_table_off
1250    (re->name_count * re->name_entry_size);    (re->name_count * re->name_entry_size);
1251    
1252  /* For an anchored pattern, or an unanchored pattern that has a first char, or  /* For an anchored pattern, or an unanchored pattern that has a first char, or
1253  a multiline pattern that matches only at "line starts", no further processing  a multiline pattern that matches only at "line starts", there is no point in
1254  at present. */  seeking a list of starting bytes. */
1255    
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    return NULL;    {
1259      int rc;
1260    
1261      /* Set the character tables in the block that is passed around */
1262    
1263  /* Set the character tables in the block that is passed around */    tables = re->tables;
1264      if (tables == NULL)
1265        (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1266        (void *)(&tables));
1267    
1268      compile_block.lcc = tables + lcc_offset;
1269      compile_block.fcc = tables + fcc_offset;
1270      compile_block.cbits = tables + cbits_offset;
1271      compile_block.ctypes = tables + ctypes_offset;
1272    
1273      /* See if we can find a fixed set of initial characters for the pattern. */
1274    
1275      memset(start_bits, 0, 32 * sizeof(uschar));
1276      rc = set_start_bits(code, start_bits, (re->options & PCRE_UTF8) != 0,
1277        &compile_block);
1278      bits_set = rc == SSB_DONE;
1279      if (rc == SSB_UNKNOWN) *errorptr = "internal error: opcode not recognized";
1280      }
1281    
1282  tables = re->tables;  /* Find the minimum length of subject string. */
1283  if (tables == NULL)  
1284    (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,  switch(min = find_minlength(code, code, re->options, &had_accept, 0))
1285    (void *)(&tables));    {
1286      case -2: *errorptr = "internal error: missing capturing bracket"; break;
1287  compile_block.lcc = tables + lcc_offset;    case -3: *errorptr = "internal error: opcode not recognized"; break;
1288  compile_block.fcc = tables + fcc_offset;    default: break;
1289  compile_block.cbits = tables + cbits_offset;    }
1290  compile_block.ctypes = tables + ctypes_offset;  
1291    /* Return NULL if there's been an error or if no optimization is possible. */
1292  /* See if we can find a fixed set of initial characters for the pattern. */  
1293    if (*errorptr != NULL || (!bits_set && min < 0)) return NULL;
 memset(start_bits, 0, 32 * sizeof(uschar));  
 if (set_start_bits(code, start_bits, (re->options & PCRE_CASELESS) != 0,  
   (re->options & PCRE_UTF8) != 0, &compile_block) != SSB_DONE) return NULL;  
1294    
1295  /* Get a pcre_extra block and a pcre_study_data block. The study data is put in  /* Get a pcre_extra block and a pcre_study_data block. The study data is put in
1296  the latter, which is pointed to by the former, which may also get additional  the latter, which is pointed to by the former, which may also get additional
# Line 577  extra->flags = PCRE_EXTRA_STUDY_DATA; Line 1313  extra->flags = PCRE_EXTRA_STUDY_DATA;
1313  extra->study_data = study;  extra->study_data = study;
1314    
1315  study->size = sizeof(pcre_study_data);  study->size = sizeof(pcre_study_data);
1316  study->options = PCRE_STUDY_MAPPED;  study->flags = 0;
1317  memcpy(study->start_bits, start_bits, sizeof(start_bits));  
1318    if (bits_set)
1319      {
1320      study->flags |= PCRE_STUDY_MAPPED;
1321      memcpy(study->start_bits, start_bits, sizeof(start_bits));
1322      }
1323    
1324    if (min >= 0)
1325      {
1326      study->flags |= PCRE_STUDY_MINLEN;
1327      study->minlength = min;
1328      }
1329    
1330  return extra;  return extra;
1331  }  }

Legend:
Removed from v.335  
changed lines
  Added in v.620

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12