/[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 345 by ph10, Mon Apr 28 15:10:02 2008 UTC revision 1364 by ph10, Sat Oct 5 15:45:11 2013 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) 1997-2012 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      re              compiled pattern block
70      code            pointer to start of group (the bracket)
71      startcode       pointer to start of the whole pattern's code
72      options         the compiling options
73      int             RECURSE depth
74    
75    Returns:   the minimum length
76               -1 if \C in UTF-8 mode or (*ACCEPT) 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 REAL_PCRE *re, const pcre_uchar *code,
83      const pcre_uchar *startcode, int options, int recurse_depth)
84    {
85    int length = -1;
86    /* PCRE_UTF16 has the same value as PCRE_UTF8. */
87    BOOL utf = (options & PCRE_UTF8) != 0;
88    BOOL had_recurse = FALSE;
89    register int branchlength = 0;
90    register pcre_uchar *cc = (pcre_uchar *)code + 1 + LINK_SIZE;
91    
92    if (*code == OP_CBRA || *code == OP_SCBRA ||
93        *code == OP_CBRAPOS || *code == OP_SCBRAPOS) cc += IMM2_SIZE;
94    
95    /* Scan along the opcodes for this branch. If we get to the end of the
96    branch, check the length against that of the other branches. */
97    
98    for (;;)
99      {
100      int d, min;
101      pcre_uchar *cs, *ce;
102      register pcre_uchar op = *cc;
103    
104      switch (op)
105        {
106        case OP_COND:
107        case OP_SCOND:
108    
109        /* If there is only one branch in a condition, the implied branch has zero
110        length, so we don't add anything. This covers the DEFINE "condition"
111        automatically. */
112    
113        cs = cc + GET(cc, 1);
114        if (*cs != OP_ALT)
115          {
116          cc = cs + 1 + LINK_SIZE;
117          break;
118          }
119    
120        /* Otherwise we can fall through and treat it the same as any other
121        subpattern. */
122    
123        case OP_CBRA:
124        case OP_SCBRA:
125        case OP_BRA:
126        case OP_SBRA:
127        case OP_CBRAPOS:
128        case OP_SCBRAPOS:
129        case OP_BRAPOS:
130        case OP_SBRAPOS:
131        case OP_ONCE:
132        case OP_ONCE_NC:
133        d = find_minlength(re, cc, startcode, options, recurse_depth);
134        if (d < 0) return d;
135        branchlength += d;
136        do cc += GET(cc, 1); while (*cc == OP_ALT);
137        cc += 1 + LINK_SIZE;
138        break;
139    
140        /* ACCEPT makes things far too complicated; we have to give up. */
141    
142        case OP_ACCEPT:
143        case OP_ASSERT_ACCEPT:
144        return -1;
145    
146        /* Reached end of a branch; if it's a ket it is the end of a nested
147        call. If it's ALT it is an alternation in a nested call. If it is END it's
148        the end of the outer call. All can be handled by the same code. If an
149        ACCEPT was previously encountered, use the length that was in force at that
150        time, and pass back the shortest ACCEPT length. */
151    
152        case OP_ALT:
153        case OP_KET:
154        case OP_KETRMAX:
155        case OP_KETRMIN:
156        case OP_KETRPOS:
157        case OP_END:
158        if (length < 0 || (!had_recurse && branchlength < length))
159          length = branchlength;
160        if (op != OP_ALT) return length;
161        cc += 1 + LINK_SIZE;
162        branchlength = 0;
163        had_recurse = FALSE;
164        break;
165    
166        /* Skip over assertive subpatterns */
167    
168        case OP_ASSERT:
169        case OP_ASSERT_NOT:
170        case OP_ASSERTBACK:
171        case OP_ASSERTBACK_NOT:
172        do cc += GET(cc, 1); while (*cc == OP_ALT);
173        /* Fall through */
174    
175        /* Skip over things that don't match chars */
176    
177        case OP_REVERSE:
178        case OP_CREF:
179        case OP_NCREF:
180        case OP_RREF:
181        case OP_NRREF:
182        case OP_DEF:
183        case OP_CALLOUT:
184        case OP_SOD:
185        case OP_SOM:
186        case OP_EOD:
187        case OP_EODN:
188        case OP_CIRC:
189        case OP_CIRCM:
190        case OP_DOLL:
191        case OP_DOLLM:
192        case OP_NOT_WORD_BOUNDARY:
193        case OP_WORD_BOUNDARY:
194        cc += PRIV(OP_lengths)[*cc];
195        break;
196    
197        /* Skip over a subpattern that has a {0} or {0,x} quantifier */
198    
199        case OP_BRAZERO:
200        case OP_BRAMINZERO:
201        case OP_BRAPOSZERO:
202        case OP_SKIPZERO:
203        cc += PRIV(OP_lengths)[*cc];
204        do cc += GET(cc, 1); while (*cc == OP_ALT);
205        cc += 1 + LINK_SIZE;
206        break;
207    
208        /* Handle literal characters and + repetitions */
209    
210        case OP_CHAR:
211        case OP_CHARI:
212        case OP_NOT:
213        case OP_NOTI:
214        case OP_PLUS:
215        case OP_PLUSI:
216        case OP_MINPLUS:
217        case OP_MINPLUSI:
218        case OP_POSPLUS:
219        case OP_POSPLUSI:
220        case OP_NOTPLUS:
221        case OP_NOTPLUSI:
222        case OP_NOTMINPLUS:
223        case OP_NOTMINPLUSI:
224        case OP_NOTPOSPLUS:
225        case OP_NOTPOSPLUSI:
226        branchlength++;
227        cc += 2;
228    #ifdef SUPPORT_UTF
229        if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
230    #endif
231        break;
232    
233        case OP_TYPEPLUS:
234        case OP_TYPEMINPLUS:
235        case OP_TYPEPOSPLUS:
236        branchlength++;
237        cc += (cc[1] == OP_PROP || cc[1] == OP_NOTPROP)? 4 : 2;
238        break;
239    
240        /* Handle exact repetitions. The count is already in characters, but we
241        need to skip over a multibyte character in UTF8 mode.  */
242    
243        case OP_EXACT:
244        case OP_EXACTI:
245        case OP_NOTEXACT:
246        case OP_NOTEXACTI:
247        branchlength += GET2(cc,1);
248        cc += 2 + IMM2_SIZE;
249    #ifdef SUPPORT_UTF
250        if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
251    #endif
252        break;
253    
254        case OP_TYPEEXACT:
255        branchlength += GET2(cc,1);
256        cc += 2 + IMM2_SIZE + ((cc[1 + IMM2_SIZE] == OP_PROP
257          || cc[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
258        break;
259    
260        /* Handle single-char non-literal matchers */
261    
262        case OP_PROP:
263        case OP_NOTPROP:
264        cc += 2;
265        /* Fall through */
266    
267        case OP_NOT_DIGIT:
268        case OP_DIGIT:
269        case OP_NOT_WHITESPACE:
270        case OP_WHITESPACE:
271        case OP_NOT_WORDCHAR:
272        case OP_WORDCHAR:
273        case OP_ANY:
274        case OP_ALLANY:
275        case OP_EXTUNI:
276        case OP_HSPACE:
277        case OP_NOT_HSPACE:
278        case OP_VSPACE:
279        case OP_NOT_VSPACE:
280        branchlength++;
281        cc++;
282        break;
283    
284        /* "Any newline" might match two characters, but it also might match just
285        one. */
286    
287        case OP_ANYNL:
288        branchlength += 1;
289        cc++;
290        break;
291    
292        /* The single-byte matcher means we can't proceed in UTF-8 mode. (In
293        non-UTF-8 mode \C will actually be turned into OP_ALLANY, so won't ever
294        appear, but leave the code, just in case.) */
295    
296        case OP_ANYBYTE:
297    #ifdef SUPPORT_UTF
298        if (utf) return -1;
299    #endif
300        branchlength++;
301        cc++;
302        break;
303    
304        /* For repeated character types, we have to test for \p and \P, which have
305        an extra two bytes of parameters. */
306    
307        case OP_TYPESTAR:
308        case OP_TYPEMINSTAR:
309        case OP_TYPEQUERY:
310        case OP_TYPEMINQUERY:
311        case OP_TYPEPOSSTAR:
312        case OP_TYPEPOSQUERY:
313        if (cc[1] == OP_PROP || cc[1] == OP_NOTPROP) cc += 2;
314        cc += PRIV(OP_lengths)[op];
315        break;
316    
317        case OP_TYPEUPTO:
318        case OP_TYPEMINUPTO:
319        case OP_TYPEPOSUPTO:
320        if (cc[1 + IMM2_SIZE] == OP_PROP
321          || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;
322        cc += PRIV(OP_lengths)[op];
323        break;
324    
325        /* Check a class for variable quantification */
326    
327        case OP_CLASS:
328        case OP_NCLASS:
329    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
330        case OP_XCLASS:
331        /* The original code caused an unsigned overflow in 64 bit systems,
332        so now we use a conditional statement. */
333        if (op == OP_XCLASS)
334          cc += GET(cc, 1);
335        else
336          cc += PRIV(OP_lengths)[OP_CLASS];
337    #else
338        cc += PRIV(OP_lengths)[OP_CLASS];
339    #endif
340    
341        switch (*cc)
342          {
343          case OP_CRPLUS:
344          case OP_CRMINPLUS:
345          branchlength++;
346          /* Fall through */
347    
348          case OP_CRSTAR:
349          case OP_CRMINSTAR:
350          case OP_CRQUERY:
351          case OP_CRMINQUERY:
352          cc++;
353          break;
354    
355          case OP_CRRANGE:
356          case OP_CRMINRANGE:
357          branchlength += GET2(cc,1);
358          cc += 1 + 2 * IMM2_SIZE;
359          break;
360    
361          default:
362          branchlength++;
363          break;
364          }
365        break;
366    
367        /* Backreferences and subroutine calls are treated in the same way: we find
368        the minimum length for the subpattern. A recursion, however, causes an
369        a flag to be set that causes the length of this branch to be ignored. The
370        logic is that a recursion can only make sense if there is another
371        alternation that stops the recursing. That will provide the minimum length
372        (when no recursion happens). A backreference within the group that it is
373        referencing behaves in the same way.
374    
375        If PCRE_JAVASCRIPT_COMPAT is set, a backreference to an unset bracket
376        matches an empty string (by default it causes a matching failure), so in
377        that case we must set the minimum length to zero. */
378    
379        case OP_DNREF:     /* Duplicate named pattern back reference */
380        case OP_DNREFI:
381        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
382          {
383          int count = GET2(cc, 1+IMM2_SIZE);
384          pcre_uchar *slot = (pcre_uchar *)re +
385            re->name_table_offset + GET2(cc, 1) * re->name_entry_size;
386          d = INT_MAX;
387          while (count-- > 0)
388            {
389            ce = cs = (pcre_uchar *)PRIV(find_bracket)(startcode, utf, GET2(slot, 0));
390            if (cs == NULL) return -2;
391            do ce += GET(ce, 1); while (*ce == OP_ALT);
392            if (cc > cs && cc < ce)
393              {
394              d = 0;
395              had_recurse = TRUE;
396              break;
397              }
398            else
399              {
400              int dd = find_minlength(re, cs, startcode, options, recurse_depth);
401              if (dd < d) d = dd;
402              }
403            slot += re->name_entry_size;
404            }
405          }
406        else d = 0;
407        cc += 1 + 2*IMM2_SIZE;
408        goto REPEAT_BACK_REFERENCE;
409    
410        case OP_REF:      /* Single back reference */
411        case OP_REFI:
412        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
413          {
414          ce = cs = (pcre_uchar *)PRIV(find_bracket)(startcode, utf, GET2(cc, 1));
415          if (cs == NULL) return -2;
416          do ce += GET(ce, 1); while (*ce == OP_ALT);
417          if (cc > cs && cc < ce)
418            {
419            d = 0;
420            had_recurse = TRUE;
421            }
422          else
423            {
424            d = find_minlength(re, cs, startcode, options, recurse_depth);
425            }
426          }
427        else d = 0;
428        cc += 1 + IMM2_SIZE;
429    
430        /* Handle repeated back references */
431    
432        REPEAT_BACK_REFERENCE:
433        switch (*cc)
434          {
435          case OP_CRSTAR:
436          case OP_CRMINSTAR:
437          case OP_CRQUERY:
438          case OP_CRMINQUERY:
439          min = 0;
440          cc++;
441          break;
442    
443          case OP_CRPLUS:
444          case OP_CRMINPLUS:
445          min = 1;
446          cc++;
447          break;
448    
449          case OP_CRRANGE:
450          case OP_CRMINRANGE:
451          min = GET2(cc, 1);
452          cc += 1 + 2 * IMM2_SIZE;
453          break;
454    
455          default:
456          min = 1;
457          break;
458          }
459    
460        branchlength += min * d;
461        break;
462    
463        /* We can easily detect direct recursion, but not mutual recursion. This is
464        caught by a recursion depth count. */
465    
466        case OP_RECURSE:
467        cs = ce = (pcre_uchar *)startcode + GET(cc, 1);
468        do ce += GET(ce, 1); while (*ce == OP_ALT);
469        if ((cc > cs && cc < ce) || recurse_depth > 10)
470          had_recurse = TRUE;
471        else
472          {
473          branchlength += find_minlength(re, cs, startcode, options,
474            recurse_depth + 1);
475          }
476        cc += 1 + LINK_SIZE;
477        break;
478    
479        /* Anything else does not or need not match a character. We can get the
480        item's length from the table, but for those that can match zero occurrences
481        of a character, we must take special action for UTF-8 characters. As it
482        happens, the "NOT" versions of these opcodes are used at present only for
483        ASCII characters, so they could be omitted from this list. However, in
484        future that may change, so we include them here so as not to leave a
485        gotcha for a future maintainer. */
486    
487        case OP_UPTO:
488        case OP_UPTOI:
489        case OP_NOTUPTO:
490        case OP_NOTUPTOI:
491        case OP_MINUPTO:
492        case OP_MINUPTOI:
493        case OP_NOTMINUPTO:
494        case OP_NOTMINUPTOI:
495        case OP_POSUPTO:
496        case OP_POSUPTOI:
497        case OP_NOTPOSUPTO:
498        case OP_NOTPOSUPTOI:
499    
500        case OP_STAR:
501        case OP_STARI:
502        case OP_NOTSTAR:
503        case OP_NOTSTARI:
504        case OP_MINSTAR:
505        case OP_MINSTARI:
506        case OP_NOTMINSTAR:
507        case OP_NOTMINSTARI:
508        case OP_POSSTAR:
509        case OP_POSSTARI:
510        case OP_NOTPOSSTAR:
511        case OP_NOTPOSSTARI:
512    
513        case OP_QUERY:
514        case OP_QUERYI:
515        case OP_NOTQUERY:
516        case OP_NOTQUERYI:
517        case OP_MINQUERY:
518        case OP_MINQUERYI:
519        case OP_NOTMINQUERY:
520        case OP_NOTMINQUERYI:
521        case OP_POSQUERY:
522        case OP_POSQUERYI:
523        case OP_NOTPOSQUERY:
524        case OP_NOTPOSQUERYI:
525    
526        cc += PRIV(OP_lengths)[op];
527    #ifdef SUPPORT_UTF
528        if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
529    #endif
530        break;
531    
532        /* Skip these, but we need to add in the name length. */
533    
534        case OP_MARK:
535        case OP_PRUNE_ARG:
536        case OP_SKIP_ARG:
537        case OP_THEN_ARG:
538        cc += PRIV(OP_lengths)[op] + cc[1];
539        break;
540    
541        /* The remaining opcodes are just skipped over. */
542    
543        case OP_CLOSE:
544        case OP_COMMIT:
545        case OP_FAIL:
546        case OP_PRUNE:
547        case OP_SET_SOM:
548        case OP_SKIP:
549        case OP_THEN:
550        cc += PRIV(OP_lengths)[op];
551        break;
552    
553        /* This should not occur: we list all opcodes explicitly so that when
554        new ones get added they are properly considered. */
555    
556        default:
557        return -3;
558        }
559      }
560    /* Control never gets here */
561    }
562    
563    
564    
565  /*************************************************  /*************************************************
566  *      Set a bit and maybe its alternate case    *  *      Set a bit and maybe its alternate case    *
567  *************************************************/  *************************************************/
568    
569  /* 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
570  version of a letter if we are caseless.  corresponding bit for the other version of a letter if we are caseless. In
571    UTF-8 mode, for characters greater than 127, we can only do the caseless thing
572    when Unicode property support is available.
573    
574  Arguments:  Arguments:
575    start_bits    points to the bit map    start_bits    points to the bit map
576    c             is the character    p             points to the character
577    caseless      the caseless flag    caseless      the caseless flag
578    cd            the block with char table pointers    cd            the block with char table pointers
579      utf           TRUE for UTF-8 / UTF-16 / UTF-32 mode
580    
581  Returns:        nothing  Returns:        pointer after the character
582    */
583    
584    static const pcre_uchar *
585    set_table_bit(pcre_uint8 *start_bits, const pcre_uchar *p, BOOL caseless,
586      compile_data *cd, BOOL utf)
587    {
588    pcre_uint32 c = *p;
589    
590    #ifdef COMPILE_PCRE8
591    SET_BIT(c);
592    
593    #ifdef SUPPORT_UTF
594    if (utf && c > 127)
595      {
596      GETCHARINC(c, p);
597    #ifdef SUPPORT_UCP
598      if (caseless)
599        {
600        pcre_uchar buff[6];
601        c = UCD_OTHERCASE(c);
602        (void)PRIV(ord2utf)(c, buff);
603        SET_BIT(buff[0]);
604        }
605    #endif  /* Not SUPPORT_UCP */
606      return p;
607      }
608    #else   /* Not SUPPORT_UTF */
609    (void)(utf);   /* Stops warning for unused parameter */
610    #endif  /* SUPPORT_UTF */
611    
612    /* Not UTF-8 mode, or character is less than 127. */
613    
614    if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
615    return p + 1;
616    #endif  /* COMPILE_PCRE8 */
617    
618    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
619    if (c > 0xff)
620      {
621      c = 0xff;
622      caseless = FALSE;
623      }
624    SET_BIT(c);
625    
626    #ifdef SUPPORT_UTF
627    if (utf && c > 127)
628      {
629      GETCHARINC(c, p);
630    #ifdef SUPPORT_UCP
631      if (caseless)
632        {
633        c = UCD_OTHERCASE(c);
634        if (c > 0xff)
635          c = 0xff;
636        SET_BIT(c);
637        }
638    #endif  /* SUPPORT_UCP */
639      return p;
640      }
641    #else   /* Not SUPPORT_UTF */
642    (void)(utf);   /* Stops warning for unused parameter */
643    #endif  /* SUPPORT_UTF */
644    
645    if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
646    return p + 1;
647    #endif
648    }
649    
650    
651    
652    /*************************************************
653    *     Set bits for a positive character type     *
654    *************************************************/
655    
656    /* This function sets starting bits for a character type. In UTF-8 mode, we can
657    only do a direct setting for bytes less than 128, as otherwise there can be
658    confusion with bytes in the middle of UTF-8 characters. In a "traditional"
659    environment, the tables will only recognize ASCII characters anyway, but in at
660    least one Windows environment, some higher bytes bits were set in the tables.
661    So we deal with that case by considering the UTF-8 encoding.
662    
663    Arguments:
664      start_bits     the starting bitmap
665      cbit type      the type of character wanted
666      table_limit    32 for non-UTF-8; 16 for UTF-8
667      cd             the block with char table pointers
668    
669    Returns:         nothing
670  */  */
671    
672  static void  static void
673  set_bit(uschar *start_bits, unsigned int c, BOOL caseless, compile_data *cd)  set_type_bits(pcre_uint8 *start_bits, int cbit_type, unsigned int table_limit,
674      compile_data *cd)
675  {  {
676  start_bits[c/8] |= (1 << (c&7));  register pcre_uint32 c;
677  if (caseless && (cd->ctypes[c] & ctype_letter) != 0)  for (c = 0; c < table_limit; c++) start_bits[c] |= cd->cbits[c+cbit_type];
678    start_bits[cd->fcc[c]/8] |= (1 << (cd->fcc[c]&7));  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
679    if (table_limit == 32) return;
680    for (c = 128; c < 256; c++)
681      {
682      if ((cd->cbits[c/8] & (1 << (c&7))) != 0)
683        {
684        pcre_uchar buff[6];
685        (void)PRIV(ord2utf)(c, buff);
686        SET_BIT(buff[0]);
687        }
688      }
689    #endif
690    }
691    
692    
693    /*************************************************
694    *     Set bits for a negative character type     *
695    *************************************************/
696    
697    /* This function sets starting bits for a negative character type such as \D.
698    In UTF-8 mode, we can only do a direct setting for bytes less than 128, as
699    otherwise there can be confusion with bytes in the middle of UTF-8 characters.
700    Unlike in the positive case, where we can set appropriate starting bits for
701    specific high-valued UTF-8 characters, in this case we have to set the bits for
702    all high-valued characters. The lowest is 0xc2, but we overkill by starting at
703    0xc0 (192) for simplicity.
704    
705    Arguments:
706      start_bits     the starting bitmap
707      cbit type      the type of character wanted
708      table_limit    32 for non-UTF-8; 16 for UTF-8
709      cd             the block with char table pointers
710    
711    Returns:         nothing
712    */
713    
714    static void
715    set_nottype_bits(pcre_uint8 *start_bits, int cbit_type, unsigned int table_limit,
716      compile_data *cd)
717    {
718    register pcre_uint32 c;
719    for (c = 0; c < table_limit; c++) start_bits[c] |= ~cd->cbits[c+cbit_type];
720    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
721    if (table_limit != 32) for (c = 24; c < 32; c++) start_bits[c] = 0xff;
722    #endif
723  }  }
724    
725    
# Line 95  function fails unless the result is SSB_ Line 739  function fails unless the result is SSB_
739  Arguments:  Arguments:
740    code         points to an expression    code         points to an expression
741    start_bits   points to a 32-byte table, initialized to 0    start_bits   points to a 32-byte table, initialized to 0
742    caseless     the current state of the caseless flag    utf          TRUE if in UTF-8 / UTF-16 / UTF-32 mode
   utf8         TRUE if in UTF-8 mode  
743    cd           the block with char table pointers    cd           the block with char table pointers
744    
745  Returns:       SSB_FAIL     => Failed to find any starting bytes  Returns:       SSB_FAIL     => Failed to find any starting bytes
746                 SSB_DONE     => Found mandatory starting bytes                 SSB_DONE     => Found mandatory starting bytes
747                 SSB_CONTINUE => Found optional starting bytes                 SSB_CONTINUE => Found optional starting bytes
748                   SSB_UNKNOWN  => Hit an unrecognized opcode
749  */  */
750    
751  static int  static int
752  set_start_bits(const uschar *code, uschar *start_bits, BOOL caseless,  set_start_bits(const pcre_uchar *code, pcre_uint8 *start_bits, BOOL utf,
753    BOOL utf8, compile_data *cd)    compile_data *cd)
754  {  {
755  register int c;  register pcre_uint32 c;
756  int yield = SSB_DONE;  int yield = SSB_DONE;
757    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
758    int table_limit = utf? 16:32;
759    #else
760    int table_limit = 32;
761    #endif
762    
763  #if 0  #if 0
764  /* ========================================================================= */  /* ========================================================================= */
# Line 130  volatile int dummy; Line 779  volatile int dummy;
779    
780  do  do
781    {    {
   const uschar *tcode = code + (((int)*code == OP_CBRA)? 3:1) + LINK_SIZE;  
782    BOOL try_next = TRUE;    BOOL try_next = TRUE;
783      const pcre_uchar *tcode = code + 1 + LINK_SIZE;
784    
785      if (*code == OP_CBRA || *code == OP_SCBRA ||
786          *code == OP_CBRAPOS || *code == OP_SCBRAPOS) tcode += IMM2_SIZE;
787    
788    while (try_next)    /* Loop for items in this branch */    while (try_next)    /* Loop for items in this branch */
789      {      {
790      int rc;      int rc;
791    
792      switch(*tcode)      switch(*tcode)
793        {        {
794        /* Fail if we reach something we don't understand */        /* If we reach something we don't understand, it means a new opcode has
795          been created that hasn't been added to this code. Hopefully this problem
796          will be discovered during testing. */
797    
798        default:        default:
799          return SSB_UNKNOWN;
800    
801          /* Fail for a valid opcode that implies no starting bits. */
802    
803          case OP_ACCEPT:
804          case OP_ASSERT_ACCEPT:
805          case OP_ALLANY:
806          case OP_ANY:
807          case OP_ANYBYTE:
808          case OP_CIRC:
809          case OP_CIRCM:
810          case OP_CLOSE:
811          case OP_COMMIT:
812          case OP_COND:
813          case OP_CREF:
814          case OP_DEF:
815          case OP_DOLL:
816          case OP_DOLLM:
817          case OP_END:
818          case OP_EOD:
819          case OP_EODN:
820          case OP_EXTUNI:
821          case OP_FAIL:
822          case OP_MARK:
823          case OP_NCREF:
824          case OP_NOT:
825          case OP_NOTEXACT:
826          case OP_NOTEXACTI:
827          case OP_NOTI:
828          case OP_NOTMINPLUS:
829          case OP_NOTMINPLUSI:
830          case OP_NOTMINQUERY:
831          case OP_NOTMINQUERYI:
832          case OP_NOTMINSTAR:
833          case OP_NOTMINSTARI:
834          case OP_NOTMINUPTO:
835          case OP_NOTMINUPTOI:
836          case OP_NOTPLUS:
837          case OP_NOTPLUSI:
838          case OP_NOTPOSPLUS:
839          case OP_NOTPOSPLUSI:
840          case OP_NOTPOSQUERY:
841          case OP_NOTPOSQUERYI:
842          case OP_NOTPOSSTAR:
843          case OP_NOTPOSSTARI:
844          case OP_NOTPOSUPTO:
845          case OP_NOTPOSUPTOI:
846          case OP_NOTPROP:
847          case OP_NOTQUERY:
848          case OP_NOTQUERYI:
849          case OP_NOTSTAR:
850          case OP_NOTSTARI:
851          case OP_NOTUPTO:
852          case OP_NOTUPTOI:
853          case OP_NOT_HSPACE:
854          case OP_NOT_VSPACE:
855          case OP_NRREF:
856          case OP_PROP:
857          case OP_PRUNE:
858          case OP_PRUNE_ARG:
859          case OP_RECURSE:
860          case OP_REF:
861          case OP_REFI:
862          case OP_DNREF:
863          case OP_DNREFI:
864          case OP_REVERSE:
865          case OP_RREF:
866          case OP_SCOND:
867          case OP_SET_SOM:
868          case OP_SKIP:
869          case OP_SKIP_ARG:
870          case OP_SOD:
871          case OP_SOM:
872          case OP_THEN:
873          case OP_THEN_ARG:
874    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
875          case OP_XCLASS:
876    #endif
877        return SSB_FAIL;        return SSB_FAIL;
878    
879          /* We can ignore word boundary tests. */
880    
881          case OP_WORD_BOUNDARY:
882          case OP_NOT_WORD_BOUNDARY:
883          tcode++;
884          break;
885    
886        /* 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
887        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
888        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 152  do Line 892  do
892        case OP_SBRA:        case OP_SBRA:
893        case OP_CBRA:        case OP_CBRA:
894        case OP_SCBRA:        case OP_SCBRA:
895          case OP_BRAPOS:
896          case OP_SBRAPOS:
897          case OP_CBRAPOS:
898          case OP_SCBRAPOS:
899        case OP_ONCE:        case OP_ONCE:
900          case OP_ONCE_NC:
901        case OP_ASSERT:        case OP_ASSERT:
902        rc = set_start_bits(tcode, start_bits, caseless, utf8, cd);        rc = set_start_bits(tcode, start_bits, utf, cd);
903        if (rc == SSB_FAIL) return SSB_FAIL;        if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
904        if (rc == SSB_DONE) try_next = FALSE; else        if (rc == SSB_DONE) try_next = FALSE; else
905          {          {
906          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
# Line 178  do Line 923  do
923        case OP_KET:        case OP_KET:
924        case OP_KETRMAX:        case OP_KETRMAX:
925        case OP_KETRMIN:        case OP_KETRMIN:
926          case OP_KETRPOS:
927        return SSB_CONTINUE;        return SSB_CONTINUE;
928    
929        /* Skip over callout */        /* Skip over callout */
# Line 195  do Line 941  do
941        tcode += 1 + LINK_SIZE;        tcode += 1 + LINK_SIZE;
942        break;        break;
943    
       /* Skip over an option setting, changing the caseless flag */  
   
       case OP_OPT:  
       caseless = (tcode[1] & PCRE_CASELESS) != 0;  
       tcode += 2;  
       break;  
   
944        /* BRAZERO does the bracket, but carries on. */        /* BRAZERO does the bracket, but carries on. */
945    
946        case OP_BRAZERO:        case OP_BRAZERO:
947        case OP_BRAMINZERO:        case OP_BRAMINZERO:
948        if (set_start_bits(++tcode, start_bits, caseless, utf8, cd) == SSB_FAIL)        case OP_BRAPOSZERO:
949          return SSB_FAIL;        rc = set_start_bits(++tcode, start_bits, utf, cd);
950          if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
951  /* =========================================================================  /* =========================================================================
952        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,
953        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 960  do
960        /* SKIPZERO skips the bracket. */        /* SKIPZERO skips the bracket. */
961    
962        case OP_SKIPZERO:        case OP_SKIPZERO:
963          tcode++;
964        do tcode += GET(tcode,1); while (*tcode == OP_ALT);        do tcode += GET(tcode,1); while (*tcode == OP_ALT);
965        tcode += 1 + LINK_SIZE;        tcode += 1 + LINK_SIZE;
966        break;        break;
# Line 232  do Line 973  do
973        case OP_QUERY:        case OP_QUERY:
974        case OP_MINQUERY:        case OP_MINQUERY:
975        case OP_POSQUERY:        case OP_POSQUERY:
976        set_bit(start_bits, tcode[1], caseless, cd);        tcode = set_table_bit(start_bits, tcode + 1, FALSE, cd, utf);
977        tcode += 2;        break;
978  #ifdef SUPPORT_UTF8  
979        if (utf8 && tcode[-1] >= 0xc0)        case OP_STARI:
980          tcode += _pcre_utf8_table4[tcode[-1] & 0x3f];        case OP_MINSTARI:
981  #endif        case OP_POSSTARI:
982          case OP_QUERYI:
983          case OP_MINQUERYI:
984          case OP_POSQUERYI:
985          tcode = set_table_bit(start_bits, tcode + 1, TRUE, cd, utf);
986        break;        break;
987    
988        /* Single-char upto sets the bit and tries the next */        /* Single-char upto sets the bit and tries the next */
# Line 245  do Line 990  do
990        case OP_UPTO:        case OP_UPTO:
991        case OP_MINUPTO:        case OP_MINUPTO:
992        case OP_POSUPTO:        case OP_POSUPTO:
993        set_bit(start_bits, tcode[3], caseless, cd);        tcode = set_table_bit(start_bits, tcode + 1 + IMM2_SIZE, FALSE, cd, utf);
       tcode += 4;  
 #ifdef SUPPORT_UTF8  
       if (utf8 && tcode[-1] >= 0xc0)  
         tcode += _pcre_utf8_table4[tcode[-1] & 0x3f];  
 #endif  
994        break;        break;
995    
996        /* At least one single char sets the bit and stops */        case OP_UPTOI:
997          case OP_MINUPTOI:
998          case OP_POSUPTOI:
999          tcode = set_table_bit(start_bits, tcode + 1 + IMM2_SIZE, TRUE, cd, utf);
1000          break;
1001    
1002        case OP_EXACT:       /* Fall through */        /* At least one single char sets the bit and stops */
       tcode += 2;  
1003    
1004          case OP_EXACT:
1005          tcode += IMM2_SIZE;
1006          /* Fall through */
1007        case OP_CHAR:        case OP_CHAR:
       case OP_CHARNC:  
1008        case OP_PLUS:        case OP_PLUS:
1009        case OP_MINPLUS:        case OP_MINPLUS:
1010        case OP_POSPLUS:        case OP_POSPLUS:
1011        set_bit(start_bits, tcode[1], caseless, cd);        (void)set_table_bit(start_bits, tcode + 1, FALSE, cd, utf);
1012        try_next = FALSE;        try_next = FALSE;
1013        break;        break;
1014    
1015        /* Single character type sets the bits and stops */        case OP_EXACTI:
1016          tcode += IMM2_SIZE;
1017          /* Fall through */
1018          case OP_CHARI:
1019          case OP_PLUSI:
1020          case OP_MINPLUSI:
1021          case OP_POSPLUSI:
1022          (void)set_table_bit(start_bits, tcode + 1, TRUE, cd, utf);
1023          try_next = FALSE;
1024          break;
1025    
1026          /* Special spacing and line-terminating items. These recognize specific
1027          lists of characters. The difference between VSPACE and ANYNL is that the
1028          latter can match the two-character CRLF sequence, but that is not
1029          relevant for finding the first character, so their code here is
1030          identical. */
1031    
1032          case OP_HSPACE:
1033          SET_BIT(CHAR_HT);
1034          SET_BIT(CHAR_SPACE);
1035    #ifdef SUPPORT_UTF
1036          if (utf)
1037            {
1038    #ifdef COMPILE_PCRE8
1039            SET_BIT(0xC2);  /* For U+00A0 */
1040            SET_BIT(0xE1);  /* For U+1680, U+180E */
1041            SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
1042            SET_BIT(0xE3);  /* For U+3000 */
1043    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1044            SET_BIT(0xA0);
1045            SET_BIT(0xFF);  /* For characters > 255 */
1046    #endif  /* COMPILE_PCRE[8|16|32] */
1047            }
1048          else
1049    #endif /* SUPPORT_UTF */
1050            {
1051    #ifndef EBCDIC
1052            SET_BIT(0xA0);
1053    #endif  /* Not EBCDIC */
1054    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1055            SET_BIT(0xFF);  /* For characters > 255 */
1056    #endif  /* COMPILE_PCRE[16|32] */
1057            }
1058          try_next = FALSE;
1059          break;
1060    
1061          case OP_ANYNL:
1062          case OP_VSPACE:
1063          SET_BIT(CHAR_LF);
1064          SET_BIT(CHAR_VT);
1065          SET_BIT(CHAR_FF);
1066          SET_BIT(CHAR_CR);
1067    #ifdef SUPPORT_UTF
1068          if (utf)
1069            {
1070    #ifdef COMPILE_PCRE8
1071            SET_BIT(0xC2);  /* For U+0085 */
1072            SET_BIT(0xE2);  /* For U+2028, U+2029 */
1073    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1074            SET_BIT(CHAR_NEL);
1075            SET_BIT(0xFF);  /* For characters > 255 */
1076    #endif  /* COMPILE_PCRE[8|16|32] */
1077            }
1078          else
1079    #endif /* SUPPORT_UTF */
1080            {
1081            SET_BIT(CHAR_NEL);
1082    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1083            SET_BIT(0xFF);  /* For characters > 255 */
1084    #endif
1085            }
1086          try_next = FALSE;
1087          break;
1088    
1089          /* Single character types set the bits and stop. Note that if PCRE_UCP
1090          is set, we do not see these op codes because \d etc are converted to
1091          properties. Therefore, these apply in the case when only characters less
1092          than 256 are recognized to match the types. */
1093    
1094        case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
1095        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_digit];  
1096        try_next = FALSE;        try_next = FALSE;
1097        break;        break;
1098    
1099        case OP_DIGIT:        case OP_DIGIT:
1100        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_digit];  
1101        try_next = FALSE;        try_next = FALSE;
1102        break;        break;
1103    
1104        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to
1105        discard it. */        ensure it is set as not whitespace. Luckily, the code value is the same
1106          (0x0b) in ASCII and EBCDIC, so we can just adjust the appropriate bit. */
1107    
1108        case OP_NOT_WHITESPACE:        case OP_NOT_WHITESPACE:
1109        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_space, table_limit, cd);
1110          {        start_bits[1] |= 0x08;
         int d = cd->cbits[c+cbit_space];  
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= ~d;  
         }  
1111        try_next = FALSE;        try_next = FALSE;
1112        break;        break;
1113    
1114        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to not
1115        discard it. */        set it from the table. Luckily, the code value is the same (0x0b) in
1116          ASCII and EBCDIC, so we can just adjust the appropriate bit. */
1117    
1118        case OP_WHITESPACE:        case OP_WHITESPACE:
1119        for (c = 0; c < 32; c++)        c = start_bits[1];    /* Save in case it was already set */
1120          {        set_type_bits(start_bits, cbit_space, table_limit, cd);
1121          int d = cd->cbits[c+cbit_space];        start_bits[1] = (start_bits[1] & ~0x08) | c;
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= d;  
         }  
1122        try_next = FALSE;        try_next = FALSE;
1123        break;        break;
1124    
1125        case OP_NOT_WORDCHAR:        case OP_NOT_WORDCHAR:
1126        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_word];  
1127        try_next = FALSE;        try_next = FALSE;
1128        break;        break;
1129    
1130        case OP_WORDCHAR:        case OP_WORDCHAR:
1131        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_word];  
1132        try_next = FALSE;        try_next = FALSE;
1133        break;        break;
1134    
# Line 324  do Line 1137  do
1137    
1138        case OP_TYPEPLUS:        case OP_TYPEPLUS:
1139        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
1140          case OP_TYPEPOSPLUS:
1141        tcode++;        tcode++;
1142        break;        break;
1143    
1144        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1145        tcode += 3;        tcode += 1 + IMM2_SIZE;
1146        break;        break;
1147    
1148        /* Zero or more repeats of character types set the bits and then        /* Zero or more repeats of character types set the bits and then
# Line 337  do Line 1151  do
1151        case OP_TYPEUPTO:        case OP_TYPEUPTO:
1152        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
1153        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1154        tcode += 2;               /* Fall through */        tcode += IMM2_SIZE;  /* Fall through */
1155    
1156        case OP_TYPESTAR:        case OP_TYPESTAR:
1157        case OP_TYPEMINSTAR:        case OP_TYPEMINSTAR:
# Line 347  do Line 1161  do
1161        case OP_TYPEPOSQUERY:        case OP_TYPEPOSQUERY:
1162        switch(tcode[1])        switch(tcode[1])
1163          {          {
1164            default:
1165          case OP_ANY:          case OP_ANY:
1166          case OP_ALLANY:          case OP_ALLANY:
1167          return SSB_FAIL;          return SSB_FAIL;
1168    
1169            case OP_HSPACE:
1170            SET_BIT(CHAR_HT);
1171            SET_BIT(CHAR_SPACE);
1172    #ifdef SUPPORT_UTF
1173            if (utf)
1174              {
1175    #ifdef COMPILE_PCRE8
1176              SET_BIT(0xC2);  /* For U+00A0 */
1177              SET_BIT(0xE1);  /* For U+1680, U+180E */
1178              SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
1179              SET_BIT(0xE3);  /* For U+3000 */
1180    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1181              SET_BIT(0xA0);
1182              SET_BIT(0xFF);  /* For characters > 255 */
1183    #endif  /* COMPILE_PCRE[8|16|32] */
1184              }
1185            else
1186    #endif /* SUPPORT_UTF */
1187    #ifndef EBCDIC
1188              SET_BIT(0xA0);
1189    #endif  /* Not EBCDIC */
1190            break;
1191    
1192            case OP_ANYNL:
1193            case OP_VSPACE:
1194            SET_BIT(CHAR_LF);
1195            SET_BIT(CHAR_VT);
1196            SET_BIT(CHAR_FF);
1197            SET_BIT(CHAR_CR);
1198    #ifdef SUPPORT_UTF
1199            if (utf)
1200              {
1201    #ifdef COMPILE_PCRE8
1202              SET_BIT(0xC2);  /* For U+0085 */
1203              SET_BIT(0xE2);  /* For U+2028, U+2029 */
1204    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1205              SET_BIT(CHAR_NEL);
1206              SET_BIT(0xFF);  /* For characters > 255 */
1207    #endif  /* COMPILE_PCRE16 */
1208              }
1209            else
1210    #endif /* SUPPORT_UTF */
1211              SET_BIT(CHAR_NEL);
1212            break;
1213    
1214          case OP_NOT_DIGIT:          case OP_NOT_DIGIT:
1215          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_digit];  
1216          break;          break;
1217    
1218          case OP_DIGIT:          case OP_DIGIT:
1219          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_digit];  
1220          break;          break;
1221    
1222          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we no longer
1223          discard it. */          have to play fancy tricks because Perl added VT to its whitespace at
1224            release 5.18. PCRE added it at release 8.34. */
1225    
1226          case OP_NOT_WHITESPACE:          case OP_NOT_WHITESPACE:
1227          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_space, table_limit, cd);
           {  
           int d = cd->cbits[c+cbit_space];  
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= ~d;  
           }  
1228          break;          break;
1229    
         /* The cbit_space table has vertical tab as whitespace; we have to  
         discard it. */  
   
1230          case OP_WHITESPACE:          case OP_WHITESPACE:
1231          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_space, table_limit, cd);
           {  
           int d = cd->cbits[c+cbit_space];  
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= d;  
           }  
1232          break;          break;
1233    
1234          case OP_NOT_WORDCHAR:          case OP_NOT_WORDCHAR:
1235          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_word];  
1236          break;          break;
1237    
1238          case OP_WORDCHAR:          case OP_WORDCHAR:
1239          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_word];  
1240          break;          break;
1241          }          }
1242    
# Line 406  do Line 1250  do
1250        character with a value > 255. */        character with a value > 255. */
1251    
1252        case OP_NCLASS:        case OP_NCLASS:
1253  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
1254        if (utf8)        if (utf)
1255          {          {
1256          start_bits[24] |= 0xf0;              /* Bits for 0xc4 - 0xc8 */          start_bits[24] |= 0xf0;              /* Bits for 0xc4 - 0xc8 */
1257          memset(start_bits+25, 0xff, 7);      /* Bits for 0xc9 - 0xff */          memset(start_bits+25, 0xff, 7);      /* Bits for 0xc9 - 0xff */
1258          }          }
1259  #endif  #endif
1260    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1261          SET_BIT(0xFF);                         /* For characters > 255 */
1262    #endif
1263        /* Fall through */        /* Fall through */
1264    
1265        case OP_CLASS:        case OP_CLASS:
1266          {          {
1267            pcre_uint8 *map;
1268          tcode++;          tcode++;
1269            map = (pcre_uint8 *)tcode;
1270    
1271          /* In UTF-8 mode, the bits in a bit map correspond to character          /* In UTF-8 mode, the bits in a bit map correspond to character
1272          values, not to byte values. However, the bit map we are constructing is          values, not to byte values. However, the bit map we are constructing is
# Line 425  do Line 1274  do
1274          value is > 127. In fact, there are only two possible starting bytes for          value is > 127. In fact, there are only two possible starting bytes for
1275          characters in the range 128 - 255. */          characters in the range 128 - 255. */
1276    
1277  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
1278          if (utf8)          if (utf)
1279            {            {
1280            for (c = 0; c < 16; c++) start_bits[c] |= tcode[c];            for (c = 0; c < 16; c++) start_bits[c] |= map[c];
1281            for (c = 128; c < 256; c++)            for (c = 128; c < 256; c++)
1282              {              {
1283              if ((tcode[c/8] && (1 << (c&7))) != 0)              if ((map[c/8] && (1 << (c&7))) != 0)
1284                {                {
1285                int d = (c >> 6) | 0xc0;            /* Set bit for this starter */                int d = (c >> 6) | 0xc0;            /* Set bit for this starter */
1286                start_bits[d/8] |= (1 << (d&7));    /* and then skip on to the */                start_bits[d/8] |= (1 << (d&7));    /* and then skip on to the */
# Line 439  do Line 1288  do
1288                }                }
1289              }              }
1290            }            }
   
         /* In non-UTF-8 mode, the two bit maps are completely compatible. */  
   
1291          else          else
1292  #endif  #endif
1293            {            {
1294            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];            /* In non-UTF-8 mode, the two bit maps are completely compatible. */
1295              for (c = 0; c < 32; c++) start_bits[c] |= map[c];
1296            }            }
1297    
1298          /* Advance past the bit map, and act on what follows */          /* Advance past the bit map, and act on what follows. For a zero
1299            minimum repeat, continue; otherwise stop processing. */
1300    
1301          tcode += 32;          tcode += 32 / sizeof(pcre_uchar);
1302          switch (*tcode)          switch (*tcode)
1303            {            {
1304            case OP_CRSTAR:            case OP_CRSTAR:
# Line 462  do Line 1310  do
1310    
1311            case OP_CRRANGE:            case OP_CRRANGE:
1312            case OP_CRMINRANGE:            case OP_CRMINRANGE:
1313            if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;            if (GET2(tcode, 1) == 0) tcode += 1 + 2 * IMM2_SIZE;
1314              else try_next = FALSE;              else try_next = FALSE;
1315            break;            break;
1316    
# Line 484  return yield; Line 1332  return yield;
1332    
1333    
1334    
1335    
1336    
1337  /*************************************************  /*************************************************
1338  *          Study a compiled expression           *  *          Study a compiled expression           *
1339  *************************************************/  *************************************************/
1340    
1341  /* This function is handed a compiled expression that it must study to produce  /* This function is handed a compiled expression that it must study to produce
1342  information that will speed up the matching. It returns a pcre_extra block  information that will speed up the matching. It returns a pcre[16]_extra block
1343  which then gets handed back to pcre_exec().  which then gets handed back to pcre_exec().
1344    
1345  Arguments:  Arguments:
# Line 498  Arguments: Line 1348  Arguments:
1348    errorptr  points to where to place error messages;    errorptr  points to where to place error messages;
1349              set NULL unless error              set NULL unless error
1350    
1351  Returns:    pointer to a pcre_extra block, with study_data filled in and the  Returns:    pointer to a pcre[16]_extra block, with study_data filled in and
1352                appropriate flag set;                the appropriate flags set;
1353              NULL on error or if no optimization possible              NULL on error or if no optimization possible
1354  */  */
1355    
1356  PCRE_EXP_DEFN pcre_extra *  #if defined COMPILE_PCRE8
1357    PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION
1358  pcre_study(const pcre *external_re, int options, const char **errorptr)  pcre_study(const pcre *external_re, int options, const char **errorptr)
1359    #elif defined COMPILE_PCRE16
1360    PCRE_EXP_DEFN pcre16_extra * PCRE_CALL_CONVENTION
1361    pcre16_study(const pcre16 *external_re, int options, const char **errorptr)
1362    #elif defined COMPILE_PCRE32
1363    PCRE_EXP_DEFN pcre32_extra * PCRE_CALL_CONVENTION
1364    pcre32_study(const pcre32 *external_re, int options, const char **errorptr)
1365    #endif
1366  {  {
1367  uschar start_bits[32];  int min;
1368  pcre_extra *extra;  BOOL bits_set = FALSE;
1369    pcre_uint8 start_bits[32];
1370    PUBL(extra) *extra = NULL;
1371  pcre_study_data *study;  pcre_study_data *study;
1372  const uschar *tables;  const pcre_uint8 *tables;
1373  uschar *code;  pcre_uchar *code;
1374  compile_data compile_block;  compile_data compile_block;
1375  const real_pcre *re = (const real_pcre *)external_re;  const REAL_PCRE *re = (const REAL_PCRE *)external_re;
1376    
1377    
1378  *errorptr = NULL;  *errorptr = NULL;
1379    
# Line 522  if (re == NULL || re->magic_number != MA Line 1383  if (re == NULL || re->magic_number != MA
1383    return NULL;    return NULL;
1384    }    }
1385    
1386    if ((re->flags & PCRE_MODE) == 0)
1387      {
1388    #if defined COMPILE_PCRE8
1389      *errorptr = "argument not compiled in 8 bit mode";
1390    #elif defined COMPILE_PCRE16
1391      *errorptr = "argument not compiled in 16 bit mode";
1392    #elif defined COMPILE_PCRE32
1393      *errorptr = "argument not compiled in 32 bit mode";
1394    #endif
1395      return NULL;
1396      }
1397    
1398  if ((options & ~PUBLIC_STUDY_OPTIONS) != 0)  if ((options & ~PUBLIC_STUDY_OPTIONS) != 0)
1399    {    {
1400    *errorptr = "unknown or incorrect option bit(s) set";    *errorptr = "unknown or incorrect option bit(s) set";
1401    return NULL;    return NULL;
1402    }    }
1403    
1404  code = (uschar *)re + re->name_table_offset +  code = (pcre_uchar *)re + re->name_table_offset +
1405    (re->name_count * re->name_entry_size);    (re->name_count * re->name_entry_size);
1406    
1407  /* 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
1408  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
1409  at present. */  seeking a list of starting bytes. */
1410    
1411  if ((re->options & PCRE_ANCHORED) != 0 ||  if ((re->options & PCRE_ANCHORED) == 0 &&
1412      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) != 0)      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)
1413    return NULL;    {
1414      int rc;
1415    
1416      /* Set the character tables in the block that is passed around */
1417    
1418  /* Set the character tables in the block that is passed around */    tables = re->tables;
1419    
1420  tables = re->tables;  #if defined COMPILE_PCRE8
1421  if (tables == NULL)    if (tables == NULL)
1422    (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,      (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1423    (void *)(&tables));      (void *)(&tables));
1424    #elif defined COMPILE_PCRE16
1425  compile_block.lcc = tables + lcc_offset;    if (tables == NULL)
1426  compile_block.fcc = tables + fcc_offset;      (void)pcre16_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1427  compile_block.cbits = tables + cbits_offset;      (void *)(&tables));
1428  compile_block.ctypes = tables + ctypes_offset;  #elif defined COMPILE_PCRE32
1429      if (tables == NULL)
1430  /* See if we can find a fixed set of initial characters for the pattern. */      (void)pcre32_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1431        (void *)(&tables));
1432  memset(start_bits, 0, 32 * sizeof(uschar));  #endif
 if (set_start_bits(code, start_bits, (re->options & PCRE_CASELESS) != 0,  
   (re->options & PCRE_UTF8) != 0, &compile_block) != SSB_DONE) return NULL;  
   
 /* Get a pcre_extra block and a pcre_study_data block. The study data is put in  
 the latter, which is pointed to by the former, which may also get additional  
 data set later by the calling program. At the moment, the size of  
 pcre_study_data is fixed. We nevertheless save it in a field for returning via  
 the pcre_fullinfo() function so that if it becomes variable in the future, we  
 don't have to change that code. */  
1433    
1434  extra = (pcre_extra *)(pcre_malloc)    compile_block.lcc = tables + lcc_offset;
1435    (sizeof(pcre_extra) + sizeof(pcre_study_data));    compile_block.fcc = tables + fcc_offset;
1436      compile_block.cbits = tables + cbits_offset;
1437      compile_block.ctypes = tables + ctypes_offset;
1438    
1439      /* See if we can find a fixed set of initial characters for the pattern. */
1440    
1441      memset(start_bits, 0, 32 * sizeof(pcre_uint8));
1442      rc = set_start_bits(code, start_bits, (re->options & PCRE_UTF8) != 0,
1443        &compile_block);
1444      bits_set = rc == SSB_DONE;
1445      if (rc == SSB_UNKNOWN)
1446        {
1447        *errorptr = "internal error: opcode not recognized";
1448        return NULL;
1449        }
1450      }
1451    
1452  if (extra == NULL)  /* Find the minimum length of subject string. */
1453    
1454    switch(min = find_minlength(re, code, code, re->options, 0))
1455    {    {
1456    *errorptr = "failed to get memory";    case -2: *errorptr = "internal error: missing capturing bracket"; return NULL;
1457    return NULL;    case -3: *errorptr = "internal error: opcode not recognized"; return NULL;
1458      default: break;
1459    }    }
1460    
1461  study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));  /* If a set of starting bytes has been identified, or if the minimum length is
1462  extra->flags = PCRE_EXTRA_STUDY_DATA;  greater than zero, or if JIT optimization has been requested, or if
1463  extra->study_data = study;  PCRE_STUDY_EXTRA_NEEDED is set, get a pcre[16]_extra block and a
1464    pcre_study_data block. The study data is put in the latter, which is pointed to
1465  study->size = sizeof(pcre_study_data);  by the former, which may also get additional data set later by the calling
1466  study->options = PCRE_STUDY_MAPPED;  program. At the moment, the size of pcre_study_data is fixed. We nevertheless
1467  memcpy(study->start_bits, start_bits, sizeof(start_bits));  save it in a field for returning via the pcre_fullinfo() function so that if it
1468    becomes variable in the future, we don't have to change that code. */
1469    
1470    if (bits_set || min > 0 || (options & (
1471    #ifdef SUPPORT_JIT
1472        PCRE_STUDY_JIT_COMPILE | PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE |
1473        PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE |
1474    #endif
1475        PCRE_STUDY_EXTRA_NEEDED)) != 0)
1476      {
1477      extra = (PUBL(extra) *)(PUBL(malloc))
1478        (sizeof(PUBL(extra)) + sizeof(pcre_study_data));
1479      if (extra == NULL)
1480        {
1481        *errorptr = "failed to get memory";
1482        return NULL;
1483        }
1484    
1485      study = (pcre_study_data *)((char *)extra + sizeof(PUBL(extra)));
1486      extra->flags = PCRE_EXTRA_STUDY_DATA;
1487      extra->study_data = study;
1488    
1489      study->size = sizeof(pcre_study_data);
1490      study->flags = 0;
1491    
1492      /* Set the start bits always, to avoid unset memory errors if the
1493      study data is written to a file, but set the flag only if any of the bits
1494      are set, to save time looking when none are. */
1495    
1496      if (bits_set)
1497        {
1498        study->flags |= PCRE_STUDY_MAPPED;
1499        memcpy(study->start_bits, start_bits, sizeof(start_bits));
1500        }
1501      else memset(study->start_bits, 0, 32 * sizeof(pcre_uint8));
1502    
1503    #ifdef PCRE_DEBUG
1504      if (bits_set)
1505        {
1506        pcre_uint8 *ptr = start_bits;
1507        int i;
1508    
1509        printf("Start bits:\n");
1510        for (i = 0; i < 32; i++)
1511          printf("%3d: %02x%s", i * 8, *ptr++, ((i + 1) & 0x7) != 0? " " : "\n");
1512        }
1513    #endif
1514    
1515      /* Always set the minlength value in the block, because the JIT compiler
1516      makes use of it. However, don't set the bit unless the length is greater than
1517      zero - the interpretive pcre_exec() and pcre_dfa_exec() needn't waste time
1518      checking the zero case. */
1519    
1520      if (min > 0)
1521        {
1522        study->flags |= PCRE_STUDY_MINLEN;
1523        study->minlength = min;
1524        }
1525      else study->minlength = 0;
1526    
1527      /* If JIT support was compiled and requested, attempt the JIT compilation.
1528      If no starting bytes were found, and the minimum length is zero, and JIT
1529      compilation fails, abandon the extra block and return NULL, unless
1530      PCRE_STUDY_EXTRA_NEEDED is set. */
1531    
1532    #ifdef SUPPORT_JIT
1533      extra->executable_jit = NULL;
1534      if ((options & PCRE_STUDY_JIT_COMPILE) != 0)
1535        PRIV(jit_compile)(re, extra, JIT_COMPILE);
1536      if ((options & PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE) != 0)
1537        PRIV(jit_compile)(re, extra, JIT_PARTIAL_SOFT_COMPILE);
1538      if ((options & PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE) != 0)
1539        PRIV(jit_compile)(re, extra, JIT_PARTIAL_HARD_COMPILE);
1540    
1541      if (study->flags == 0 && (extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) == 0 &&
1542          (options & PCRE_STUDY_EXTRA_NEEDED) == 0)
1543        {
1544    #if defined COMPILE_PCRE8
1545        pcre_free_study(extra);
1546    #elif defined COMPILE_PCRE16
1547        pcre16_free_study(extra);
1548    #elif defined COMPILE_PCRE32
1549        pcre32_free_study(extra);
1550    #endif
1551        extra = NULL;
1552        }
1553    #endif
1554      }
1555    
1556  return extra;  return extra;
1557  }  }
1558    
1559    
1560    /*************************************************
1561    *          Free the study data                   *
1562    *************************************************/
1563    
1564    /* This function frees the memory that was obtained by pcre_study().
1565    
1566    Argument:   a pointer to the pcre[16]_extra block
1567    Returns:    nothing
1568    */
1569    
1570    #if defined COMPILE_PCRE8
1571    PCRE_EXP_DEFN void
1572    pcre_free_study(pcre_extra *extra)
1573    #elif defined COMPILE_PCRE16
1574    PCRE_EXP_DEFN void
1575    pcre16_free_study(pcre16_extra *extra)
1576    #elif defined COMPILE_PCRE32
1577    PCRE_EXP_DEFN void
1578    pcre32_free_study(pcre32_extra *extra)
1579    #endif
1580    {
1581    if (extra == NULL)
1582      return;
1583    #ifdef SUPPORT_JIT
1584    if ((extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 &&
1585         extra->executable_jit != NULL)
1586      PRIV(jit_free)(extra->executable_jit);
1587    #endif
1588    PUBL(free)(extra);
1589    }
1590    
1591  /* End of pcre_study.c */  /* End of pcre_study.c */

Legend:
Removed from v.345  
changed lines
  Added in v.1364

  ViewVC Help
Powered by ViewVC 1.1.5