/[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 353 by ph10, Mon Jul 7 15:44:24 2008 UTC revision 1486 by ph10, Wed Jun 18 16:48:57 2014 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_DNCREF:
180        case OP_RREF:
181        case OP_DNRREF:
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          case OP_CRPOSPLUS:
346          branchlength++;
347          /* Fall through */
348    
349          case OP_CRSTAR:
350          case OP_CRMINSTAR:
351          case OP_CRQUERY:
352          case OP_CRMINQUERY:
353          case OP_CRPOSSTAR:
354          case OP_CRPOSQUERY:
355          cc++;
356          break;
357    
358          case OP_CRRANGE:
359          case OP_CRMINRANGE:
360          case OP_CRPOSRANGE:
361          branchlength += GET2(cc,1);
362          cc += 1 + 2 * IMM2_SIZE;
363          break;
364    
365          default:
366          branchlength++;
367          break;
368          }
369        break;
370    
371        /* Backreferences and subroutine calls are treated in the same way: we find
372        the minimum length for the subpattern. A recursion, however, causes an
373        a flag to be set that causes the length of this branch to be ignored. The
374        logic is that a recursion can only make sense if there is another
375        alternation that stops the recursing. That will provide the minimum length
376        (when no recursion happens). A backreference within the group that it is
377        referencing behaves in the same way.
378    
379        If PCRE_JAVASCRIPT_COMPAT is set, a backreference to an unset bracket
380        matches an empty string (by default it causes a matching failure), so in
381        that case we must set the minimum length to zero. */
382    
383        case OP_DNREF:     /* Duplicate named pattern back reference */
384        case OP_DNREFI:
385        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
386          {
387          int count = GET2(cc, 1+IMM2_SIZE);
388          pcre_uchar *slot = (pcre_uchar *)re +
389            re->name_table_offset + GET2(cc, 1) * re->name_entry_size;
390          d = INT_MAX;
391          while (count-- > 0)
392            {
393            ce = cs = (pcre_uchar *)PRIV(find_bracket)(startcode, utf, GET2(slot, 0));
394            if (cs == NULL) return -2;
395            do ce += GET(ce, 1); while (*ce == OP_ALT);
396            if (cc > cs && cc < ce)
397              {
398              d = 0;
399              had_recurse = TRUE;
400              break;
401              }
402            else
403              {
404              int dd = find_minlength(re, cs, startcode, options, recurse_depth);
405              if (dd < d) d = dd;
406              }
407            slot += re->name_entry_size;
408            }
409          }
410        else d = 0;
411        cc += 1 + 2*IMM2_SIZE;
412        goto REPEAT_BACK_REFERENCE;
413    
414        case OP_REF:      /* Single back reference */
415        case OP_REFI:
416        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
417          {
418          ce = cs = (pcre_uchar *)PRIV(find_bracket)(startcode, utf, GET2(cc, 1));
419          if (cs == NULL) return -2;
420          do ce += GET(ce, 1); while (*ce == OP_ALT);
421          if (cc > cs && cc < ce)
422            {
423            d = 0;
424            had_recurse = TRUE;
425            }
426          else
427            {
428            d = find_minlength(re, cs, startcode, options, recurse_depth);
429            }
430          }
431        else d = 0;
432        cc += 1 + IMM2_SIZE;
433    
434        /* Handle repeated back references */
435    
436        REPEAT_BACK_REFERENCE:
437        switch (*cc)
438          {
439          case OP_CRSTAR:
440          case OP_CRMINSTAR:
441          case OP_CRQUERY:
442          case OP_CRMINQUERY:
443          case OP_CRPOSSTAR:
444          case OP_CRPOSQUERY:
445          min = 0;
446          cc++;
447          break;
448    
449          case OP_CRPLUS:
450          case OP_CRMINPLUS:
451          case OP_CRPOSPLUS:
452          min = 1;
453          cc++;
454          break;
455    
456          case OP_CRRANGE:
457          case OP_CRMINRANGE:
458          case OP_CRPOSRANGE:
459          min = GET2(cc, 1);
460          cc += 1 + 2 * IMM2_SIZE;
461          break;
462    
463          default:
464          min = 1;
465          break;
466          }
467    
468        branchlength += min * d;
469        break;
470    
471        /* We can easily detect direct recursion, but not mutual recursion. This is
472        caught by a recursion depth count. */
473    
474        case OP_RECURSE:
475        cs = ce = (pcre_uchar *)startcode + GET(cc, 1);
476        do ce += GET(ce, 1); while (*ce == OP_ALT);
477        if ((cc > cs && cc < ce) || recurse_depth > 10)
478          had_recurse = TRUE;
479        else
480          {
481          branchlength += find_minlength(re, cs, startcode, options,
482            recurse_depth + 1);
483          }
484        cc += 1 + LINK_SIZE;
485        break;
486    
487        /* Anything else does not or need not match a character. We can get the
488        item's length from the table, but for those that can match zero occurrences
489        of a character, we must take special action for UTF-8 characters. As it
490        happens, the "NOT" versions of these opcodes are used at present only for
491        ASCII characters, so they could be omitted from this list. However, in
492        future that may change, so we include them here so as not to leave a
493        gotcha for a future maintainer. */
494    
495        case OP_UPTO:
496        case OP_UPTOI:
497        case OP_NOTUPTO:
498        case OP_NOTUPTOI:
499        case OP_MINUPTO:
500        case OP_MINUPTOI:
501        case OP_NOTMINUPTO:
502        case OP_NOTMINUPTOI:
503        case OP_POSUPTO:
504        case OP_POSUPTOI:
505        case OP_NOTPOSUPTO:
506        case OP_NOTPOSUPTOI:
507    
508        case OP_STAR:
509        case OP_STARI:
510        case OP_NOTSTAR:
511        case OP_NOTSTARI:
512        case OP_MINSTAR:
513        case OP_MINSTARI:
514        case OP_NOTMINSTAR:
515        case OP_NOTMINSTARI:
516        case OP_POSSTAR:
517        case OP_POSSTARI:
518        case OP_NOTPOSSTAR:
519        case OP_NOTPOSSTARI:
520    
521        case OP_QUERY:
522        case OP_QUERYI:
523        case OP_NOTQUERY:
524        case OP_NOTQUERYI:
525        case OP_MINQUERY:
526        case OP_MINQUERYI:
527        case OP_NOTMINQUERY:
528        case OP_NOTMINQUERYI:
529        case OP_POSQUERY:
530        case OP_POSQUERYI:
531        case OP_NOTPOSQUERY:
532        case OP_NOTPOSQUERYI:
533    
534        cc += PRIV(OP_lengths)[op];
535    #ifdef SUPPORT_UTF
536        if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
537    #endif
538        break;
539    
540        /* Skip these, but we need to add in the name length. */
541    
542        case OP_MARK:
543        case OP_PRUNE_ARG:
544        case OP_SKIP_ARG:
545        case OP_THEN_ARG:
546        cc += PRIV(OP_lengths)[op] + cc[1];
547        break;
548    
549        /* The remaining opcodes are just skipped over. */
550    
551        case OP_CLOSE:
552        case OP_COMMIT:
553        case OP_FAIL:
554        case OP_PRUNE:
555        case OP_SET_SOM:
556        case OP_SKIP:
557        case OP_THEN:
558        cc += PRIV(OP_lengths)[op];
559        break;
560    
561        /* This should not occur: we list all opcodes explicitly so that when
562        new ones get added they are properly considered. */
563    
564        default:
565        return -3;
566        }
567      }
568    /* Control never gets here */
569    }
570    
571    
572    
573  /*************************************************  /*************************************************
574  *      Set a bit and maybe its alternate case    *  *      Set a bit and maybe its alternate case    *
575  *************************************************/  *************************************************/
576    
577  /* 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
578  version of a letter if we are caseless.  corresponding bit for the other version of a letter if we are caseless. In
579    UTF-8 mode, for characters greater than 127, we can only do the caseless thing
580    when Unicode property support is available.
581    
582  Arguments:  Arguments:
583    start_bits    points to the bit map    start_bits    points to the bit map
584    c             is the character    p             points to the character
585    caseless      the caseless flag    caseless      the caseless flag
586    cd            the block with char table pointers    cd            the block with char table pointers
587      utf           TRUE for UTF-8 / UTF-16 / UTF-32 mode
588    
589  Returns:        nothing  Returns:        pointer after the character
590    */
591    
592    static const pcre_uchar *
593    set_table_bit(pcre_uint8 *start_bits, const pcre_uchar *p, BOOL caseless,
594      compile_data *cd, BOOL utf)
595    {
596    pcre_uint32 c = *p;
597    
598    #ifdef COMPILE_PCRE8
599    SET_BIT(c);
600    
601    #ifdef SUPPORT_UTF
602    if (utf && c > 127)
603      {
604      GETCHARINC(c, p);
605    #ifdef SUPPORT_UCP
606      if (caseless)
607        {
608        pcre_uchar buff[6];
609        c = UCD_OTHERCASE(c);
610        (void)PRIV(ord2utf)(c, buff);
611        SET_BIT(buff[0]);
612        }
613    #endif  /* Not SUPPORT_UCP */
614      return p;
615      }
616    #else   /* Not SUPPORT_UTF */
617    (void)(utf);   /* Stops warning for unused parameter */
618    #endif  /* SUPPORT_UTF */
619    
620    /* Not UTF-8 mode, or character is less than 127. */
621    
622    if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
623    return p + 1;
624    #endif  /* COMPILE_PCRE8 */
625    
626    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
627    if (c > 0xff)
628      {
629      c = 0xff;
630      caseless = FALSE;
631      }
632    SET_BIT(c);
633    
634    #ifdef SUPPORT_UTF
635    if (utf && c > 127)
636      {
637      GETCHARINC(c, p);
638    #ifdef SUPPORT_UCP
639      if (caseless)
640        {
641        c = UCD_OTHERCASE(c);
642        if (c > 0xff)
643          c = 0xff;
644        SET_BIT(c);
645        }
646    #endif  /* SUPPORT_UCP */
647      return p;
648      }
649    #else   /* Not SUPPORT_UTF */
650    (void)(utf);   /* Stops warning for unused parameter */
651    #endif  /* SUPPORT_UTF */
652    
653    if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
654    return p + 1;
655    #endif
656    }
657    
658    
659    
660    /*************************************************
661    *     Set bits for a positive character type     *
662    *************************************************/
663    
664    /* This function sets starting bits for a character type. In UTF-8 mode, we can
665    only do a direct setting for bytes less than 128, as otherwise there can be
666    confusion with bytes in the middle of UTF-8 characters. In a "traditional"
667    environment, the tables will only recognize ASCII characters anyway, but in at
668    least one Windows environment, some higher bytes bits were set in the tables.
669    So we deal with that case by considering the UTF-8 encoding.
670    
671    Arguments:
672      start_bits     the starting bitmap
673      cbit type      the type of character wanted
674      table_limit    32 for non-UTF-8; 16 for UTF-8
675      cd             the block with char table pointers
676    
677    Returns:         nothing
678  */  */
679    
680  static void  static void
681  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,
682      compile_data *cd)
683  {  {
684  start_bits[c/8] |= (1 << (c&7));  register pcre_uint32 c;
685  if (caseless && (cd->ctypes[c] & ctype_letter) != 0)  for (c = 0; c < table_limit; c++) start_bits[c] |= cd->cbits[c+cbit_type];
686    start_bits[cd->fcc[c]/8] |= (1 << (cd->fcc[c]&7));  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
687    if (table_limit == 32) return;
688    for (c = 128; c < 256; c++)
689      {
690      if ((cd->cbits[c/8] & (1 << (c&7))) != 0)
691        {
692        pcre_uchar buff[6];
693        (void)PRIV(ord2utf)(c, buff);
694        SET_BIT(buff[0]);
695        }
696      }
697    #endif
698    }
699    
700    
701    /*************************************************
702    *     Set bits for a negative character type     *
703    *************************************************/
704    
705    /* This function sets starting bits for a negative character type such as \D.
706    In UTF-8 mode, we can only do a direct setting for bytes less than 128, as
707    otherwise there can be confusion with bytes in the middle of UTF-8 characters.
708    Unlike in the positive case, where we can set appropriate starting bits for
709    specific high-valued UTF-8 characters, in this case we have to set the bits for
710    all high-valued characters. The lowest is 0xc2, but we overkill by starting at
711    0xc0 (192) for simplicity.
712    
713    Arguments:
714      start_bits     the starting bitmap
715      cbit type      the type of character wanted
716      table_limit    32 for non-UTF-8; 16 for UTF-8
717      cd             the block with char table pointers
718    
719    Returns:         nothing
720    */
721    
722    static void
723    set_nottype_bits(pcre_uint8 *start_bits, int cbit_type, unsigned int table_limit,
724      compile_data *cd)
725    {
726    register pcre_uint32 c;
727    for (c = 0; c < table_limit; c++) start_bits[c] |= ~cd->cbits[c+cbit_type];
728    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
729    if (table_limit != 32) for (c = 24; c < 32; c++) start_bits[c] = 0xff;
730    #endif
731  }  }
732    
733    
# Line 95  function fails unless the result is SSB_ Line 747  function fails unless the result is SSB_
747  Arguments:  Arguments:
748    code         points to an expression    code         points to an expression
749    start_bits   points to a 32-byte table, initialized to 0    start_bits   points to a 32-byte table, initialized to 0
750    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  
751    cd           the block with char table pointers    cd           the block with char table pointers
752    
753  Returns:       SSB_FAIL     => Failed to find any starting bytes  Returns:       SSB_FAIL     => Failed to find any starting bytes
754                 SSB_DONE     => Found mandatory starting bytes                 SSB_DONE     => Found mandatory starting bytes
755                 SSB_CONTINUE => Found optional starting bytes                 SSB_CONTINUE => Found optional starting bytes
756                   SSB_UNKNOWN  => Hit an unrecognized opcode
757  */  */
758    
759  static int  static int
760  set_start_bits(const uschar *code, uschar *start_bits, BOOL caseless,  set_start_bits(const pcre_uchar *code, pcre_uint8 *start_bits, BOOL utf,
761    BOOL utf8, compile_data *cd)    compile_data *cd)
762  {  {
763  register int c;  register pcre_uint32 c;
764  int yield = SSB_DONE;  int yield = SSB_DONE;
765    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
766    int table_limit = utf? 16:32;
767    #else
768    int table_limit = 32;
769    #endif
770    
771  #if 0  #if 0
772  /* ========================================================================= */  /* ========================================================================= */
# Line 130  volatile int dummy; Line 787  volatile int dummy;
787    
788  do  do
789    {    {
   const uschar *tcode = code + (((int)*code == OP_CBRA)? 3:1) + LINK_SIZE;  
790    BOOL try_next = TRUE;    BOOL try_next = TRUE;
791      const pcre_uchar *tcode = code + 1 + LINK_SIZE;
792    
793      if (*code == OP_CBRA || *code == OP_SCBRA ||
794          *code == OP_CBRAPOS || *code == OP_SCBRAPOS) tcode += IMM2_SIZE;
795    
796    while (try_next)    /* Loop for items in this branch */    while (try_next)    /* Loop for items in this branch */
797      {      {
798      int rc;      int rc;
799    
800      switch(*tcode)      switch(*tcode)
801        {        {
802        /* Fail if we reach something we don't understand */        /* If we reach something we don't understand, it means a new opcode has
803          been created that hasn't been added to this code. Hopefully this problem
804          will be discovered during testing. */
805    
806        default:        default:
807          return SSB_UNKNOWN;
808    
809          /* Fail for a valid opcode that implies no starting bits. */
810    
811          case OP_ACCEPT:
812          case OP_ASSERT_ACCEPT:
813          case OP_ALLANY:
814          case OP_ANY:
815          case OP_ANYBYTE:
816          case OP_CIRC:
817          case OP_CIRCM:
818          case OP_CLOSE:
819          case OP_COMMIT:
820          case OP_COND:
821          case OP_CREF:
822          case OP_DEF:
823          case OP_DNCREF:
824          case OP_DNREF:
825          case OP_DNREFI:
826          case OP_DNRREF:
827          case OP_DOLL:
828          case OP_DOLLM:
829          case OP_END:
830          case OP_EOD:
831          case OP_EODN:
832          case OP_EXTUNI:
833          case OP_FAIL:
834          case OP_MARK:
835          case OP_NOT:
836          case OP_NOTEXACT:
837          case OP_NOTEXACTI:
838          case OP_NOTI:
839          case OP_NOTMINPLUS:
840          case OP_NOTMINPLUSI:
841          case OP_NOTMINQUERY:
842          case OP_NOTMINQUERYI:
843          case OP_NOTMINSTAR:
844          case OP_NOTMINSTARI:
845          case OP_NOTMINUPTO:
846          case OP_NOTMINUPTOI:
847          case OP_NOTPLUS:
848          case OP_NOTPLUSI:
849          case OP_NOTPOSPLUS:
850          case OP_NOTPOSPLUSI:
851          case OP_NOTPOSQUERY:
852          case OP_NOTPOSQUERYI:
853          case OP_NOTPOSSTAR:
854          case OP_NOTPOSSTARI:
855          case OP_NOTPOSUPTO:
856          case OP_NOTPOSUPTOI:
857          case OP_NOTPROP:
858          case OP_NOTQUERY:
859          case OP_NOTQUERYI:
860          case OP_NOTSTAR:
861          case OP_NOTSTARI:
862          case OP_NOTUPTO:
863          case OP_NOTUPTOI:
864          case OP_NOT_HSPACE:
865          case OP_NOT_VSPACE:
866          case OP_PROP:
867          case OP_PRUNE:
868          case OP_PRUNE_ARG:
869          case OP_RECURSE:
870          case OP_REF:
871          case OP_REFI:
872          case OP_REVERSE:
873          case OP_RREF:
874          case OP_SCOND:
875          case OP_SET_SOM:
876          case OP_SKIP:
877          case OP_SKIP_ARG:
878          case OP_SOD:
879          case OP_SOM:
880          case OP_THEN:
881          case OP_THEN_ARG:
882        return SSB_FAIL;        return SSB_FAIL;
883    
884          /* We can ignore word boundary tests. */
885    
886          case OP_WORD_BOUNDARY:
887          case OP_NOT_WORD_BOUNDARY:
888          tcode++;
889          break;
890    
891        /* 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
892        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
893        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 897  do
897        case OP_SBRA:        case OP_SBRA:
898        case OP_CBRA:        case OP_CBRA:
899        case OP_SCBRA:        case OP_SCBRA:
900          case OP_BRAPOS:
901          case OP_SBRAPOS:
902          case OP_CBRAPOS:
903          case OP_SCBRAPOS:
904        case OP_ONCE:        case OP_ONCE:
905          case OP_ONCE_NC:
906        case OP_ASSERT:        case OP_ASSERT:
907        rc = set_start_bits(tcode, start_bits, caseless, utf8, cd);        rc = set_start_bits(tcode, start_bits, utf, cd);
908        if (rc == SSB_FAIL) return SSB_FAIL;        if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
909        if (rc == SSB_DONE) try_next = FALSE; else        if (rc == SSB_DONE) try_next = FALSE; else
910          {          {
911          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
# Line 178  do Line 928  do
928        case OP_KET:        case OP_KET:
929        case OP_KETRMAX:        case OP_KETRMAX:
930        case OP_KETRMIN:        case OP_KETRMIN:
931          case OP_KETRPOS:
932        return SSB_CONTINUE;        return SSB_CONTINUE;
933    
934        /* Skip over callout */        /* Skip over callout */
# Line 195  do Line 946  do
946        tcode += 1 + LINK_SIZE;        tcode += 1 + LINK_SIZE;
947        break;        break;
948    
       /* Skip over an option setting, changing the caseless flag */  
   
       case OP_OPT:  
       caseless = (tcode[1] & PCRE_CASELESS) != 0;  
       tcode += 2;  
       break;  
   
949        /* BRAZERO does the bracket, but carries on. */        /* BRAZERO does the bracket, but carries on. */
950    
951        case OP_BRAZERO:        case OP_BRAZERO:
952        case OP_BRAMINZERO:        case OP_BRAMINZERO:
953        if (set_start_bits(++tcode, start_bits, caseless, utf8, cd) == SSB_FAIL)        case OP_BRAPOSZERO:
954          return SSB_FAIL;        rc = set_start_bits(++tcode, start_bits, utf, cd);
955          if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
956  /* =========================================================================  /* =========================================================================
957        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,
958        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 965  do
965        /* SKIPZERO skips the bracket. */        /* SKIPZERO skips the bracket. */
966    
967        case OP_SKIPZERO:        case OP_SKIPZERO:
968        tcode++;        tcode++;
969        do tcode += GET(tcode,1); while (*tcode == OP_ALT);        do tcode += GET(tcode,1); while (*tcode == OP_ALT);
970        tcode += 1 + LINK_SIZE;        tcode += 1 + LINK_SIZE;
971        break;        break;
# Line 233  do Line 978  do
978        case OP_QUERY:        case OP_QUERY:
979        case OP_MINQUERY:        case OP_MINQUERY:
980        case OP_POSQUERY:        case OP_POSQUERY:
981        set_bit(start_bits, tcode[1], caseless, cd);        tcode = set_table_bit(start_bits, tcode + 1, FALSE, cd, utf);
982        tcode += 2;        break;
983  #ifdef SUPPORT_UTF8  
984        if (utf8 && tcode[-1] >= 0xc0)        case OP_STARI:
985          tcode += _pcre_utf8_table4[tcode[-1] & 0x3f];        case OP_MINSTARI:
986  #endif        case OP_POSSTARI:
987          case OP_QUERYI:
988          case OP_MINQUERYI:
989          case OP_POSQUERYI:
990          tcode = set_table_bit(start_bits, tcode + 1, TRUE, cd, utf);
991        break;        break;
992    
993        /* Single-char upto sets the bit and tries the next */        /* Single-char upto sets the bit and tries the next */
# Line 246  do Line 995  do
995        case OP_UPTO:        case OP_UPTO:
996        case OP_MINUPTO:        case OP_MINUPTO:
997        case OP_POSUPTO:        case OP_POSUPTO:
998        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  
999        break;        break;
1000    
1001        /* At least one single char sets the bit and stops */        case OP_UPTOI:
1002          case OP_MINUPTOI:
1003          case OP_POSUPTOI:
1004          tcode = set_table_bit(start_bits, tcode + 1 + IMM2_SIZE, TRUE, cd, utf);
1005          break;
1006    
1007        case OP_EXACT:       /* Fall through */        /* At least one single char sets the bit and stops */
       tcode += 2;  
1008    
1009          case OP_EXACT:
1010          tcode += IMM2_SIZE;
1011          /* Fall through */
1012        case OP_CHAR:        case OP_CHAR:
       case OP_CHARNC:  
1013        case OP_PLUS:        case OP_PLUS:
1014        case OP_MINPLUS:        case OP_MINPLUS:
1015        case OP_POSPLUS:        case OP_POSPLUS:
1016        set_bit(start_bits, tcode[1], caseless, cd);        (void)set_table_bit(start_bits, tcode + 1, FALSE, cd, utf);
1017          try_next = FALSE;
1018          break;
1019    
1020          case OP_EXACTI:
1021          tcode += IMM2_SIZE;
1022          /* Fall through */
1023          case OP_CHARI:
1024          case OP_PLUSI:
1025          case OP_MINPLUSI:
1026          case OP_POSPLUSI:
1027          (void)set_table_bit(start_bits, tcode + 1, TRUE, cd, utf);
1028          try_next = FALSE;
1029          break;
1030    
1031          /* Special spacing and line-terminating items. These recognize specific
1032          lists of characters. The difference between VSPACE and ANYNL is that the
1033          latter can match the two-character CRLF sequence, but that is not
1034          relevant for finding the first character, so their code here is
1035          identical. */
1036    
1037          case OP_HSPACE:
1038          SET_BIT(CHAR_HT);
1039          SET_BIT(CHAR_SPACE);
1040    #ifdef SUPPORT_UTF
1041          if (utf)
1042            {
1043    #ifdef COMPILE_PCRE8
1044            SET_BIT(0xC2);  /* For U+00A0 */
1045            SET_BIT(0xE1);  /* For U+1680, U+180E */
1046            SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
1047            SET_BIT(0xE3);  /* For U+3000 */
1048    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1049            SET_BIT(0xA0);
1050            SET_BIT(0xFF);  /* For characters > 255 */
1051    #endif  /* COMPILE_PCRE[8|16|32] */
1052            }
1053          else
1054    #endif /* SUPPORT_UTF */
1055            {
1056    #ifndef EBCDIC
1057            SET_BIT(0xA0);
1058    #endif  /* Not EBCDIC */
1059    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1060            SET_BIT(0xFF);  /* For characters > 255 */
1061    #endif  /* COMPILE_PCRE[16|32] */
1062            }
1063          try_next = FALSE;
1064          break;
1065    
1066          case OP_ANYNL:
1067          case OP_VSPACE:
1068          SET_BIT(CHAR_LF);
1069          SET_BIT(CHAR_VT);
1070          SET_BIT(CHAR_FF);
1071          SET_BIT(CHAR_CR);
1072    #ifdef SUPPORT_UTF
1073          if (utf)
1074            {
1075    #ifdef COMPILE_PCRE8
1076            SET_BIT(0xC2);  /* For U+0085 */
1077            SET_BIT(0xE2);  /* For U+2028, U+2029 */
1078    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1079            SET_BIT(CHAR_NEL);
1080            SET_BIT(0xFF);  /* For characters > 255 */
1081    #endif  /* COMPILE_PCRE[8|16|32] */
1082            }
1083          else
1084    #endif /* SUPPORT_UTF */
1085            {
1086            SET_BIT(CHAR_NEL);
1087    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1088            SET_BIT(0xFF);  /* For characters > 255 */
1089    #endif
1090            }
1091        try_next = FALSE;        try_next = FALSE;
1092        break;        break;
1093    
1094        /* Single character type sets the bits and stops */        /* Single character types set the bits and stop. Note that if PCRE_UCP
1095          is set, we do not see these op codes because \d etc are converted to
1096          properties. Therefore, these apply in the case when only characters less
1097          than 256 are recognized to match the types. */
1098    
1099        case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
1100        for (c = 0; c < 32; c++)        set_nottype_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        case OP_DIGIT:        case OP_DIGIT:
1105        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_digit];  
1106        try_next = FALSE;        try_next = FALSE;
1107        break;        break;
1108    
1109        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we no longer
1110        discard it. */        have to play fancy tricks because Perl added VT to its whitespace at
1111          release 5.18. PCRE added it at release 8.34. */
1112    
1113        case OP_NOT_WHITESPACE:        case OP_NOT_WHITESPACE:
1114        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;  
         }  
1115        try_next = FALSE;        try_next = FALSE;
1116        break;        break;
1117    
       /* The cbit_space table has vertical tab as whitespace; we have to  
       discard it. */  
   
1118        case OP_WHITESPACE:        case OP_WHITESPACE:
1119        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;  
         }  
1120        try_next = FALSE;        try_next = FALSE;
1121        break;        break;
1122    
1123        case OP_NOT_WORDCHAR:        case OP_NOT_WORDCHAR:
1124        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_word];  
1125        try_next = FALSE;        try_next = FALSE;
1126        break;        break;
1127    
1128        case OP_WORDCHAR:        case OP_WORDCHAR:
1129        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_word];  
1130        try_next = FALSE;        try_next = FALSE;
1131        break;        break;
1132    
# Line 325  do Line 1135  do
1135    
1136        case OP_TYPEPLUS:        case OP_TYPEPLUS:
1137        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
1138          case OP_TYPEPOSPLUS:
1139        tcode++;        tcode++;
1140        break;        break;
1141    
1142        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1143        tcode += 3;        tcode += 1 + IMM2_SIZE;
1144        break;        break;
1145    
1146        /* 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 338  do Line 1149  do
1149        case OP_TYPEUPTO:        case OP_TYPEUPTO:
1150        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
1151        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1152        tcode += 2;               /* Fall through */        tcode += IMM2_SIZE;  /* Fall through */
1153    
1154        case OP_TYPESTAR:        case OP_TYPESTAR:
1155        case OP_TYPEMINSTAR:        case OP_TYPEMINSTAR:
# Line 348  do Line 1159  do
1159        case OP_TYPEPOSQUERY:        case OP_TYPEPOSQUERY:
1160        switch(tcode[1])        switch(tcode[1])
1161          {          {
1162            default:
1163          case OP_ANY:          case OP_ANY:
1164          case OP_ALLANY:          case OP_ALLANY:
1165          return SSB_FAIL;          return SSB_FAIL;
1166    
1167            case OP_HSPACE:
1168            SET_BIT(CHAR_HT);
1169            SET_BIT(CHAR_SPACE);
1170    #ifdef SUPPORT_UTF
1171            if (utf)
1172              {
1173    #ifdef COMPILE_PCRE8
1174              SET_BIT(0xC2);  /* For U+00A0 */
1175              SET_BIT(0xE1);  /* For U+1680, U+180E */
1176              SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
1177              SET_BIT(0xE3);  /* For U+3000 */
1178    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1179              SET_BIT(0xA0);
1180              SET_BIT(0xFF);  /* For characters > 255 */
1181    #endif  /* COMPILE_PCRE[8|16|32] */
1182              }
1183            else
1184    #endif /* SUPPORT_UTF */
1185    #ifndef EBCDIC
1186              SET_BIT(0xA0);
1187    #endif  /* Not EBCDIC */
1188            break;
1189    
1190            case OP_ANYNL:
1191            case OP_VSPACE:
1192            SET_BIT(CHAR_LF);
1193            SET_BIT(CHAR_VT);
1194            SET_BIT(CHAR_FF);
1195            SET_BIT(CHAR_CR);
1196    #ifdef SUPPORT_UTF
1197            if (utf)
1198              {
1199    #ifdef COMPILE_PCRE8
1200              SET_BIT(0xC2);  /* For U+0085 */
1201              SET_BIT(0xE2);  /* For U+2028, U+2029 */
1202    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1203              SET_BIT(CHAR_NEL);
1204              SET_BIT(0xFF);  /* For characters > 255 */
1205    #endif  /* COMPILE_PCRE16 */
1206              }
1207            else
1208    #endif /* SUPPORT_UTF */
1209              SET_BIT(CHAR_NEL);
1210            break;
1211    
1212          case OP_NOT_DIGIT:          case OP_NOT_DIGIT:
1213          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_digit];  
1214          break;          break;
1215    
1216          case OP_DIGIT:          case OP_DIGIT:
1217          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_digit];  
1218          break;          break;
1219    
1220          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we no longer
1221          discard it. */          have to play fancy tricks because Perl added VT to its whitespace at
1222            release 5.18. PCRE added it at release 8.34. */
1223    
1224          case OP_NOT_WHITESPACE:          case OP_NOT_WHITESPACE:
1225          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;  
           }  
1226          break;          break;
1227    
         /* The cbit_space table has vertical tab as whitespace; we have to  
         discard it. */  
   
1228          case OP_WHITESPACE:          case OP_WHITESPACE:
1229          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;  
           }  
1230          break;          break;
1231    
1232          case OP_NOT_WORDCHAR:          case OP_NOT_WORDCHAR:
1233          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_word];  
1234          break;          break;
1235    
1236          case OP_WORDCHAR:          case OP_WORDCHAR:
1237          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_word];  
1238          break;          break;
1239          }          }
1240    
# Line 406  do Line 1247  do
1247        with a value >= 0xc4 is a potentially valid starter because it starts a        with a value >= 0xc4 is a potentially valid starter because it starts a
1248        character with a value > 255. */        character with a value > 255. */
1249    
1250    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1251          case OP_XCLASS:
1252          if ((tcode[1 + LINK_SIZE] & XCL_HASPROP) != 0)
1253            return SSB_FAIL;
1254          /* All bits are set. */
1255          if ((tcode[1 + LINK_SIZE] & XCL_MAP) == 0 && (tcode[1 + LINK_SIZE] & XCL_NOT) != 0)
1256            return SSB_FAIL;
1257    #endif
1258          /* Fall through */
1259    
1260        case OP_NCLASS:        case OP_NCLASS:
1261  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
1262        if (utf8)        if (utf)
1263          {          {
1264          start_bits[24] |= 0xf0;              /* Bits for 0xc4 - 0xc8 */          start_bits[24] |= 0xf0;              /* Bits for 0xc4 - 0xc8 */
1265          memset(start_bits+25, 0xff, 7);      /* Bits for 0xc9 - 0xff */          memset(start_bits+25, 0xff, 7);      /* Bits for 0xc9 - 0xff */
1266          }          }
1267  #endif  #endif
1268    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1269          SET_BIT(0xFF);                         /* For characters > 255 */
1270    #endif
1271        /* Fall through */        /* Fall through */
1272    
1273        case OP_CLASS:        case OP_CLASS:
1274          {          {
1275          tcode++;          pcre_uint8 *map;
1276    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1277            map = NULL;
1278            if (*tcode == OP_XCLASS)
1279              {
1280              if ((tcode[1 + LINK_SIZE] & XCL_MAP) != 0)
1281                map = (pcre_uint8 *)(tcode + 1 + LINK_SIZE + 1);
1282              tcode += GET(tcode, 1);
1283              }
1284            else
1285    #endif
1286              {
1287              tcode++;
1288              map = (pcre_uint8 *)tcode;
1289              tcode += 32 / sizeof(pcre_uchar);
1290              }
1291    
1292          /* 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
1293          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 426  do Line 1295  do
1295          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
1296          characters in the range 128 - 255. */          characters in the range 128 - 255. */
1297    
1298  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1299          if (utf8)          if (map != NULL)
1300    #endif
1301            {            {
1302            for (c = 0; c < 16; c++) start_bits[c] |= tcode[c];  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
1303            for (c = 128; c < 256; c++)            if (utf)
1304              {              {
1305              if ((tcode[c/8] && (1 << (c&7))) != 0)              for (c = 0; c < 16; c++) start_bits[c] |= map[c];
1306                for (c = 128; c < 256; c++)
1307                {                {
1308                int d = (c >> 6) | 0xc0;            /* Set bit for this starter */                if ((map[c/8] && (1 << (c&7))) != 0)
1309                start_bits[d/8] |= (1 << (d&7));    /* and then skip on to the */                  {
1310                c = (c & 0xc0) + 0x40 - 1;          /* next relevant character. */                  int d = (c >> 6) | 0xc0;            /* Set bit for this starter */
1311                    start_bits[d/8] |= (1 << (d&7));    /* and then skip on to the */
1312                    c = (c & 0xc0) + 0x40 - 1;          /* next relevant character. */
1313                    }
1314                }                }
1315              }              }
1316            }            else
   
         /* In non-UTF-8 mode, the two bit maps are completely compatible. */  
   
         else  
1317  #endif  #endif
1318            {              {
1319            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];              /* In non-UTF-8 mode, the two bit maps are completely compatible. */
1320                for (c = 0; c < 32; c++) start_bits[c] |= map[c];
1321                }
1322            }            }
1323    
1324          /* Advance past the bit map, and act on what follows */          /* Advance past the bit map, and act on what follows. For a zero
1325            minimum repeat, continue; otherwise stop processing. */
1326    
         tcode += 32;  
1327          switch (*tcode)          switch (*tcode)
1328            {            {
1329            case OP_CRSTAR:            case OP_CRSTAR:
1330            case OP_CRMINSTAR:            case OP_CRMINSTAR:
1331            case OP_CRQUERY:            case OP_CRQUERY:
1332            case OP_CRMINQUERY:            case OP_CRMINQUERY:
1333              case OP_CRPOSSTAR:
1334              case OP_CRPOSQUERY:
1335            tcode++;            tcode++;
1336            break;            break;
1337    
1338            case OP_CRRANGE:            case OP_CRRANGE:
1339            case OP_CRMINRANGE:            case OP_CRMINRANGE:
1340            if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;            case OP_CRPOSRANGE:
1341              if (GET2(tcode, 1) == 0) tcode += 1 + 2 * IMM2_SIZE;
1342              else try_next = FALSE;              else try_next = FALSE;
1343            break;            break;
1344    
# Line 485  return yield; Line 1360  return yield;
1360    
1361    
1362    
1363    
1364    
1365  /*************************************************  /*************************************************
1366  *          Study a compiled expression           *  *          Study a compiled expression           *
1367  *************************************************/  *************************************************/
1368    
1369  /* 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
1370  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
1371  which then gets handed back to pcre_exec().  which then gets handed back to pcre_exec().
1372    
1373  Arguments:  Arguments:
# Line 499  Arguments: Line 1376  Arguments:
1376    errorptr  points to where to place error messages;    errorptr  points to where to place error messages;
1377              set NULL unless error              set NULL unless error
1378    
1379  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
1380                appropriate flag set;                the appropriate flags set;
1381              NULL on error or if no optimization possible              NULL on error or if no optimization possible
1382  */  */
1383    
1384  PCRE_EXP_DEFN pcre_extra *  #if defined COMPILE_PCRE8
1385    PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION
1386  pcre_study(const pcre *external_re, int options, const char **errorptr)  pcre_study(const pcre *external_re, int options, const char **errorptr)
1387    #elif defined COMPILE_PCRE16
1388    PCRE_EXP_DEFN pcre16_extra * PCRE_CALL_CONVENTION
1389    pcre16_study(const pcre16 *external_re, int options, const char **errorptr)
1390    #elif defined COMPILE_PCRE32
1391    PCRE_EXP_DEFN pcre32_extra * PCRE_CALL_CONVENTION
1392    pcre32_study(const pcre32 *external_re, int options, const char **errorptr)
1393    #endif
1394  {  {
1395  uschar start_bits[32];  int min;
1396  pcre_extra *extra;  BOOL bits_set = FALSE;
1397    pcre_uint8 start_bits[32];
1398    PUBL(extra) *extra = NULL;
1399  pcre_study_data *study;  pcre_study_data *study;
1400  const uschar *tables;  const pcre_uint8 *tables;
1401  uschar *code;  pcre_uchar *code;
1402  compile_data compile_block;  compile_data compile_block;
1403  const real_pcre *re = (const real_pcre *)external_re;  const REAL_PCRE *re = (const REAL_PCRE *)external_re;
1404    
1405    
1406  *errorptr = NULL;  *errorptr = NULL;
1407    
# Line 523  if (re == NULL || re->magic_number != MA Line 1411  if (re == NULL || re->magic_number != MA
1411    return NULL;    return NULL;
1412    }    }
1413    
1414    if ((re->flags & PCRE_MODE) == 0)
1415      {
1416    #if defined COMPILE_PCRE8
1417      *errorptr = "argument not compiled in 8 bit mode";
1418    #elif defined COMPILE_PCRE16
1419      *errorptr = "argument not compiled in 16 bit mode";
1420    #elif defined COMPILE_PCRE32
1421      *errorptr = "argument not compiled in 32 bit mode";
1422    #endif
1423      return NULL;
1424      }
1425    
1426  if ((options & ~PUBLIC_STUDY_OPTIONS) != 0)  if ((options & ~PUBLIC_STUDY_OPTIONS) != 0)
1427    {    {
1428    *errorptr = "unknown or incorrect option bit(s) set";    *errorptr = "unknown or incorrect option bit(s) set";
1429    return NULL;    return NULL;
1430    }    }
1431    
1432  code = (uschar *)re + re->name_table_offset +  code = (pcre_uchar *)re + re->name_table_offset +
1433    (re->name_count * re->name_entry_size);    (re->name_count * re->name_entry_size);
1434    
1435  /* 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
1436  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
1437  at present. */  seeking a list of starting bytes. */
1438    
1439  if ((re->options & PCRE_ANCHORED) != 0 ||  if ((re->options & PCRE_ANCHORED) == 0 &&
1440      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) != 0)      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)
1441    return NULL;    {
1442      int rc;
1443    
1444  /* Set the character tables in the block that is passed around */    /* Set the character tables in the block that is passed around */
1445    
1446  tables = re->tables;    tables = re->tables;
 if (tables == NULL)  
   (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,  
   (void *)(&tables));  
   
 compile_block.lcc = tables + lcc_offset;  
 compile_block.fcc = tables + fcc_offset;  
 compile_block.cbits = tables + cbits_offset;  
 compile_block.ctypes = tables + ctypes_offset;  
   
 /* See if we can find a fixed set of initial characters for the pattern. */  
   
 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;  
   
 /* 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. */  
1447    
1448  extra = (pcre_extra *)(pcre_malloc)  #if defined COMPILE_PCRE8
1449    (sizeof(pcre_extra) + sizeof(pcre_study_data));    if (tables == NULL)
1450        (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1451        (void *)(&tables));
1452    #elif defined COMPILE_PCRE16
1453      if (tables == NULL)
1454        (void)pcre16_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1455        (void *)(&tables));
1456    #elif defined COMPILE_PCRE32
1457      if (tables == NULL)
1458        (void)pcre32_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1459        (void *)(&tables));
1460    #endif
1461    
1462  if (extra == NULL)    compile_block.lcc = tables + lcc_offset;
1463      compile_block.fcc = tables + fcc_offset;
1464      compile_block.cbits = tables + cbits_offset;
1465      compile_block.ctypes = tables + ctypes_offset;
1466    
1467      /* See if we can find a fixed set of initial characters for the pattern. */
1468    
1469      memset(start_bits, 0, 32 * sizeof(pcre_uint8));
1470      rc = set_start_bits(code, start_bits, (re->options & PCRE_UTF8) != 0,
1471        &compile_block);
1472      bits_set = rc == SSB_DONE;
1473      if (rc == SSB_UNKNOWN)
1474        {
1475        *errorptr = "internal error: opcode not recognized";
1476        return NULL;
1477        }
1478      }
1479    
1480    /* Find the minimum length of subject string. */
1481    
1482    switch(min = find_minlength(re, code, code, re->options, 0))
1483    {    {
1484    *errorptr = "failed to get memory";    case -2: *errorptr = "internal error: missing capturing bracket"; return NULL;
1485    return NULL;    case -3: *errorptr = "internal error: opcode not recognized"; return NULL;
1486      default: break;
1487    }    }
1488    
1489  study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));  /* If a set of starting bytes has been identified, or if the minimum length is
1490  extra->flags = PCRE_EXTRA_STUDY_DATA;  greater than zero, or if JIT optimization has been requested, or if
1491  extra->study_data = study;  PCRE_STUDY_EXTRA_NEEDED is set, get a pcre[16]_extra block and a
1492    pcre_study_data block. The study data is put in the latter, which is pointed to
1493  study->size = sizeof(pcre_study_data);  by the former, which may also get additional data set later by the calling
1494  study->options = PCRE_STUDY_MAPPED;  program. At the moment, the size of pcre_study_data is fixed. We nevertheless
1495  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
1496    becomes variable in the future, we don't have to change that code. */
1497    
1498    if (bits_set || min > 0 || (options & (
1499    #ifdef SUPPORT_JIT
1500        PCRE_STUDY_JIT_COMPILE | PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE |
1501        PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE |
1502    #endif
1503        PCRE_STUDY_EXTRA_NEEDED)) != 0)
1504      {
1505      extra = (PUBL(extra) *)(PUBL(malloc))
1506        (sizeof(PUBL(extra)) + sizeof(pcre_study_data));
1507      if (extra == NULL)
1508        {
1509        *errorptr = "failed to get memory";
1510        return NULL;
1511        }
1512    
1513      study = (pcre_study_data *)((char *)extra + sizeof(PUBL(extra)));
1514      extra->flags = PCRE_EXTRA_STUDY_DATA;
1515      extra->study_data = study;
1516    
1517      study->size = sizeof(pcre_study_data);
1518      study->flags = 0;
1519    
1520      /* Set the start bits always, to avoid unset memory errors if the
1521      study data is written to a file, but set the flag only if any of the bits
1522      are set, to save time looking when none are. */
1523    
1524      if (bits_set)
1525        {
1526        study->flags |= PCRE_STUDY_MAPPED;
1527        memcpy(study->start_bits, start_bits, sizeof(start_bits));
1528        }
1529      else memset(study->start_bits, 0, 32 * sizeof(pcre_uint8));
1530    
1531    #ifdef PCRE_DEBUG
1532      if (bits_set)
1533        {
1534        pcre_uint8 *ptr = start_bits;
1535        int i;
1536    
1537        printf("Start bits:\n");
1538        for (i = 0; i < 32; i++)
1539          printf("%3d: %02x%s", i * 8, *ptr++, ((i + 1) & 0x7) != 0? " " : "\n");
1540        }
1541    #endif
1542    
1543      /* Always set the minlength value in the block, because the JIT compiler
1544      makes use of it. However, don't set the bit unless the length is greater than
1545      zero - the interpretive pcre_exec() and pcre_dfa_exec() needn't waste time
1546      checking the zero case. */
1547    
1548      if (min > 0)
1549        {
1550        study->flags |= PCRE_STUDY_MINLEN;
1551        study->minlength = min;
1552        }
1553      else study->minlength = 0;
1554    
1555      /* If JIT support was compiled and requested, attempt the JIT compilation.
1556      If no starting bytes were found, and the minimum length is zero, and JIT
1557      compilation fails, abandon the extra block and return NULL, unless
1558      PCRE_STUDY_EXTRA_NEEDED is set. */
1559    
1560    #ifdef SUPPORT_JIT
1561      extra->executable_jit = NULL;
1562      if ((options & PCRE_STUDY_JIT_COMPILE) != 0)
1563        PRIV(jit_compile)(re, extra, JIT_COMPILE);
1564      if ((options & PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE) != 0)
1565        PRIV(jit_compile)(re, extra, JIT_PARTIAL_SOFT_COMPILE);
1566      if ((options & PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE) != 0)
1567        PRIV(jit_compile)(re, extra, JIT_PARTIAL_HARD_COMPILE);
1568    
1569      if (study->flags == 0 && (extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) == 0 &&
1570          (options & PCRE_STUDY_EXTRA_NEEDED) == 0)
1571        {
1572    #if defined COMPILE_PCRE8
1573        pcre_free_study(extra);
1574    #elif defined COMPILE_PCRE16
1575        pcre16_free_study(extra);
1576    #elif defined COMPILE_PCRE32
1577        pcre32_free_study(extra);
1578    #endif
1579        extra = NULL;
1580        }
1581    #endif
1582      }
1583    
1584  return extra;  return extra;
1585  }  }
1586    
1587    
1588    /*************************************************
1589    *          Free the study data                   *
1590    *************************************************/
1591    
1592    /* This function frees the memory that was obtained by pcre_study().
1593    
1594    Argument:   a pointer to the pcre[16]_extra block
1595    Returns:    nothing
1596    */
1597    
1598    #if defined COMPILE_PCRE8
1599    PCRE_EXP_DEFN void
1600    pcre_free_study(pcre_extra *extra)
1601    #elif defined COMPILE_PCRE16
1602    PCRE_EXP_DEFN void
1603    pcre16_free_study(pcre16_extra *extra)
1604    #elif defined COMPILE_PCRE32
1605    PCRE_EXP_DEFN void
1606    pcre32_free_study(pcre32_extra *extra)
1607    #endif
1608    {
1609    if (extra == NULL)
1610      return;
1611    #ifdef SUPPORT_JIT
1612    if ((extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 &&
1613         extra->executable_jit != NULL)
1614      PRIV(jit_free)(extra->executable_jit);
1615    #endif
1616    PUBL(free)(extra);
1617    }
1618    
1619  /* End of pcre_study.c */  /* End of pcre_study.c */

Legend:
Removed from v.353  
changed lines
  Added in v.1486

  ViewVC Help
Powered by ViewVC 1.1.5