/[pcre]/code/branches/pcre16/pcre_study.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_study.c

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

code/trunk/pcre_study.c revision 200 by ph10, Wed Aug 1 09:10:40 2007 UTC code/branches/pcre16/pcre_study.c revision 770 by zherczeg, Mon Nov 28 20:39:30 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 43  supporting functions. */ Line 43  supporting functions. */
43    
44    
45  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
46  #include <config.h>  #include "config.h"
47  #endif  #endif
48    
49  #include "pcre_internal.h"  #include "pcre_internal.h"
50    
51    #define SET_BIT(c) start_bits[c/8] |= (1 << (c&7))
52    
53  /* Returns from set_start_bits() */  /* Returns from set_start_bits() */
54    
55  enum { SSB_FAIL, SSB_DONE, SSB_CONTINUE };  enum { SSB_FAIL, SSB_DONE, SSB_CONTINUE, SSB_UNKNOWN };
56    
57    
58    
59    /*************************************************
60    *   Find the minimum subject length for a group  *
61    *************************************************/
62    
63    /* Scan a parenthesized group and compute the minimum length of subject that
64    is needed to match it. This is a lower bound; it does not mean there is a
65    string of that length that matches. In UTF8 mode, the result is in characters
66    rather than bytes.
67    
68    Arguments:
69      code            pointer to start of group (the bracket)
70      startcode       pointer to start of the whole pattern
71      options         the compiling options
72      int             RECURSE depth
73    
74    Returns:   the minimum length
75               -1 if \C in UTF-8 mode or (*ACCEPT) was encountered
76               -2 internal error (missing capturing bracket)
77               -3 internal error (opcode not listed)
78    */
79    
80    static int
81    find_minlength(const pcre_uchar *code, const pcre_uchar *startcode, int options,
82      int recurse_depth)
83    {
84    int length = -1;
85    BOOL utf8 = (options & PCRE_UTF8) != 0;
86    BOOL had_recurse = FALSE;
87    register int branchlength = 0;
88    register pcre_uchar *cc = (pcre_uchar *)code + 1 + LINK_SIZE;
89    
90    if (*code == OP_CBRA || *code == OP_SCBRA ||
91        *code == OP_CBRAPOS || *code == OP_SCBRAPOS) cc += IMM2_SIZE;
92    
93    /* Scan along the opcodes for this branch. If we get to the end of the
94    branch, check the length against that of the other branches. */
95    
96    for (;;)
97      {
98      int d, min;
99      pcre_uchar *cs, *ce;
100      register int op = *cc;
101    
102      switch (op)
103        {
104        case OP_COND:
105        case OP_SCOND:
106    
107        /* If there is only one branch in a condition, the implied branch has zero
108        length, so we don't add anything. This covers the DEFINE "condition"
109        automatically. */
110    
111        cs = cc + GET(cc, 1);
112        if (*cs != OP_ALT)
113          {
114          cc = cs + 1 + LINK_SIZE;
115          break;
116          }
117    
118        /* Otherwise we can fall through and treat it the same as any other
119        subpattern. */
120    
121        case OP_CBRA:
122        case OP_SCBRA:
123        case OP_BRA:
124        case OP_SBRA:
125        case OP_CBRAPOS:
126        case OP_SCBRAPOS:
127        case OP_BRAPOS:
128        case OP_SBRAPOS:
129        case OP_ONCE:
130        case OP_ONCE_NC:
131        d = find_minlength(cc, startcode, options, recurse_depth);
132        if (d < 0) return d;
133        branchlength += d;
134        do cc += GET(cc, 1); while (*cc == OP_ALT);
135        cc += 1 + LINK_SIZE;
136        break;
137    
138        /* ACCEPT makes things far too complicated; we have to give up. */
139    
140        case OP_ACCEPT:
141        case OP_ASSERT_ACCEPT:
142        return -1;
143    
144        /* Reached end of a branch; if it's a ket it is the end of a nested
145        call. If it's ALT it is an alternation in a nested call. If it is END it's
146        the end of the outer call. All can be handled by the same code. If an
147        ACCEPT was previously encountered, use the length that was in force at that
148        time, and pass back the shortest ACCEPT length. */
149    
150        case OP_ALT:
151        case OP_KET:
152        case OP_KETRMAX:
153        case OP_KETRMIN:
154        case OP_KETRPOS:
155        case OP_END:
156        if (length < 0 || (!had_recurse && branchlength < length))
157          length = branchlength;
158        if (op != OP_ALT) return length;
159        cc += 1 + LINK_SIZE;
160        branchlength = 0;
161        had_recurse = FALSE;
162        break;
163    
164        /* Skip over assertive subpatterns */
165    
166        case OP_ASSERT:
167        case OP_ASSERT_NOT:
168        case OP_ASSERTBACK:
169        case OP_ASSERTBACK_NOT:
170        do cc += GET(cc, 1); while (*cc == OP_ALT);
171        /* Fall through */
172    
173        /* Skip over things that don't match chars */
174    
175        case OP_REVERSE:
176        case OP_CREF:
177        case OP_NCREF:
178        case OP_RREF:
179        case OP_NRREF:
180        case OP_DEF:
181        case OP_CALLOUT:
182        case OP_SOD:
183        case OP_SOM:
184        case OP_EOD:
185        case OP_EODN:
186        case OP_CIRC:
187        case OP_CIRCM:
188        case OP_DOLL:
189        case OP_DOLLM:
190        case OP_NOT_WORD_BOUNDARY:
191        case OP_WORD_BOUNDARY:
192        cc += PRIV(OP_lengths)[*cc];
193        break;
194    
195        /* Skip over a subpattern that has a {0} or {0,x} quantifier */
196    
197        case OP_BRAZERO:
198        case OP_BRAMINZERO:
199        case OP_BRAPOSZERO:
200        case OP_SKIPZERO:
201        cc += PRIV(OP_lengths)[*cc];
202        do cc += GET(cc, 1); while (*cc == OP_ALT);
203        cc += 1 + LINK_SIZE;
204        break;
205    
206        /* Handle literal characters and + repetitions */
207    
208        case OP_CHAR:
209        case OP_CHARI:
210        case OP_NOT:
211        case OP_NOTI:
212        case OP_PLUS:
213        case OP_PLUSI:
214        case OP_MINPLUS:
215        case OP_MINPLUSI:
216        case OP_POSPLUS:
217        case OP_POSPLUSI:
218        case OP_NOTPLUS:
219        case OP_NOTPLUSI:
220        case OP_NOTMINPLUS:
221        case OP_NOTMINPLUSI:
222        case OP_NOTPOSPLUS:
223        case OP_NOTPOSPLUSI:
224        branchlength++;
225        cc += 2;
226    #ifdef SUPPORT_UTF8
227        if (utf8 && cc[-1] >= 0xc0) cc += PRIV(utf8_table4)[cc[-1] & 0x3f];
228    #endif
229        break;
230    
231        case OP_TYPEPLUS:
232        case OP_TYPEMINPLUS:
233        case OP_TYPEPOSPLUS:
234        branchlength++;
235        cc += (cc[1] == OP_PROP || cc[1] == OP_NOTPROP)? 4 : 2;
236        break;
237    
238        /* Handle exact repetitions. The count is already in characters, but we
239        need to skip over a multibyte character in UTF8 mode.  */
240    
241        case OP_EXACT:
242        case OP_EXACTI:
243        case OP_NOTEXACT:
244        case OP_NOTEXACTI:
245        branchlength += GET2(cc,1);
246        cc += 2 + IMM2_SIZE;
247    #ifdef SUPPORT_UTF8
248        if (utf8 && cc[-1] >= 0xc0) cc += PRIV(utf8_table4)[cc[-1] & 0x3f];
249    #endif
250        break;
251    
252        case OP_TYPEEXACT:
253        branchlength += GET2(cc,1);
254        cc += 2 + IMM2_SIZE + ((cc[1 + IMM2_SIZE] == OP_PROP
255          || cc[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
256        break;
257    
258        /* Handle single-char non-literal matchers */
259    
260        case OP_PROP:
261        case OP_NOTPROP:
262        cc += 2;
263        /* Fall through */
264    
265        case OP_NOT_DIGIT:
266        case OP_DIGIT:
267        case OP_NOT_WHITESPACE:
268        case OP_WHITESPACE:
269        case OP_NOT_WORDCHAR:
270        case OP_WORDCHAR:
271        case OP_ANY:
272        case OP_ALLANY:
273        case OP_EXTUNI:
274        case OP_HSPACE:
275        case OP_NOT_HSPACE:
276        case OP_VSPACE:
277        case OP_NOT_VSPACE:
278        branchlength++;
279        cc++;
280        break;
281    
282        /* "Any newline" might match two characters, but it also might match just
283        one. */
284    
285        case OP_ANYNL:
286        branchlength += 1;
287        cc++;
288        break;
289    
290        /* The single-byte matcher means we can't proceed in UTF-8 mode. (In
291        non-UTF-8 mode \C will actually be turned into OP_ALLANY, so won't ever
292        appear, but leave the code, just in case.) */
293    
294        case OP_ANYBYTE:
295    #ifdef SUPPORT_UTF8
296        if (utf8) return -1;
297    #endif
298        branchlength++;
299        cc++;
300        break;
301    
302        /* For repeated character types, we have to test for \p and \P, which have
303        an extra two bytes of parameters. */
304    
305        case OP_TYPESTAR:
306        case OP_TYPEMINSTAR:
307        case OP_TYPEQUERY:
308        case OP_TYPEMINQUERY:
309        case OP_TYPEPOSSTAR:
310        case OP_TYPEPOSQUERY:
311        if (cc[1] == OP_PROP || cc[1] == OP_NOTPROP) cc += 2;
312        cc += PRIV(OP_lengths)[op];
313        break;
314    
315        case OP_TYPEUPTO:
316        case OP_TYPEMINUPTO:
317        case OP_TYPEPOSUPTO:
318        if (cc[1 + IMM2_SIZE] == OP_PROP
319          || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;
320        cc += PRIV(OP_lengths)[op];
321        break;
322    
323        /* Check a class for variable quantification */
324    
325    #if defined SUPPORT_UTF8 || !defined COMPILE_PCRE8
326        case OP_XCLASS:
327        cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
328        /* Fall through */
329    #endif
330    
331        case OP_CLASS:
332        case OP_NCLASS:
333        cc += PRIV(OP_lengths)[OP_CLASS];
334    
335        switch (*cc)
336          {
337          case OP_CRPLUS:
338          case OP_CRMINPLUS:
339          branchlength++;
340          /* Fall through */
341    
342          case OP_CRSTAR:
343          case OP_CRMINSTAR:
344          case OP_CRQUERY:
345          case OP_CRMINQUERY:
346          cc++;
347          break;
348    
349          case OP_CRRANGE:
350          case OP_CRMINRANGE:
351          branchlength += GET2(cc,1);
352          cc += 1 + 2 * IMM2_SIZE;
353          break;
354    
355          default:
356          branchlength++;
357          break;
358          }
359        break;
360    
361        /* Backreferences and subroutine calls are treated in the same way: we find
362        the minimum length for the subpattern. A recursion, however, causes an
363        a flag to be set that causes the length of this branch to be ignored. The
364        logic is that a recursion can only make sense if there is another
365        alternation that stops the recursing. That will provide the minimum length
366        (when no recursion happens). A backreference within the group that it is
367        referencing behaves in the same way.
368    
369        If PCRE_JAVASCRIPT_COMPAT is set, a backreference to an unset bracket
370        matches an empty string (by default it causes a matching failure), so in
371        that case we must set the minimum length to zero. */
372    
373        case OP_REF:
374        case OP_REFI:
375        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
376          {
377          ce = cs = (pcre_uchar *)PRIV(find_bracket)(startcode, utf8, GET2(cc, 1));
378          if (cs == NULL) return -2;
379          do ce += GET(ce, 1); while (*ce == OP_ALT);
380          if (cc > cs && cc < ce)
381            {
382            d = 0;
383            had_recurse = TRUE;
384            }
385          else
386            {
387            d = find_minlength(cs, startcode, options, recurse_depth);
388            }
389          }
390        else d = 0;
391        cc += 1 + IMM2_SIZE;
392    
393        /* Handle repeated back references */
394    
395        switch (*cc)
396          {
397          case OP_CRSTAR:
398          case OP_CRMINSTAR:
399          case OP_CRQUERY:
400          case OP_CRMINQUERY:
401          min = 0;
402          cc++;
403          break;
404    
405          case OP_CRPLUS:
406          case OP_CRMINPLUS:
407          min = 1;
408          cc++;
409          break;
410    
411          case OP_CRRANGE:
412          case OP_CRMINRANGE:
413          min = GET2(cc, 1);
414          cc += 1 + 2 * IMM2_SIZE;
415          break;
416    
417          default:
418          min = 1;
419          break;
420          }
421    
422        branchlength += min * d;
423        break;
424    
425        /* We can easily detect direct recursion, but not mutual recursion. This is
426        caught by a recursion depth count. */
427    
428        case OP_RECURSE:
429        cs = ce = (pcre_uchar *)startcode + GET(cc, 1);
430        do ce += GET(ce, 1); while (*ce == OP_ALT);
431        if ((cc > cs && cc < ce) || recurse_depth > 10)
432          had_recurse = TRUE;
433        else
434          {
435          branchlength += find_minlength(cs, startcode, options, recurse_depth + 1);
436          }
437        cc += 1 + LINK_SIZE;
438        break;
439    
440        /* Anything else does not or need not match a character. We can get the
441        item's length from the table, but for those that can match zero occurrences
442        of a character, we must take special action for UTF-8 characters. As it
443        happens, the "NOT" versions of these opcodes are used at present only for
444        ASCII characters, so they could be omitted from this list. However, in
445        future that may change, so we include them here so as not to leave a
446        gotcha for a future maintainer. */
447    
448        case OP_UPTO:
449        case OP_UPTOI:
450        case OP_NOTUPTO:
451        case OP_NOTUPTOI:
452        case OP_MINUPTO:
453        case OP_MINUPTOI:
454        case OP_NOTMINUPTO:
455        case OP_NOTMINUPTOI:
456        case OP_POSUPTO:
457        case OP_POSUPTOI:
458        case OP_NOTPOSUPTO:
459        case OP_NOTPOSUPTOI:
460    
461        case OP_STAR:
462        case OP_STARI:
463        case OP_NOTSTAR:
464        case OP_NOTSTARI:
465        case OP_MINSTAR:
466        case OP_MINSTARI:
467        case OP_NOTMINSTAR:
468        case OP_NOTMINSTARI:
469        case OP_POSSTAR:
470        case OP_POSSTARI:
471        case OP_NOTPOSSTAR:
472        case OP_NOTPOSSTARI:
473    
474        case OP_QUERY:
475        case OP_QUERYI:
476        case OP_NOTQUERY:
477        case OP_NOTQUERYI:
478        case OP_MINQUERY:
479        case OP_MINQUERYI:
480        case OP_NOTMINQUERY:
481        case OP_NOTMINQUERYI:
482        case OP_POSQUERY:
483        case OP_POSQUERYI:
484        case OP_NOTPOSQUERY:
485        case OP_NOTPOSQUERYI:
486    
487        cc += PRIV(OP_lengths)[op];
488    #ifdef SUPPORT_UTF8
489        if (utf8 && cc[-1] >= 0xc0) cc += PRIV(utf8_table4)[cc[-1] & 0x3f];
490    #endif
491        break;
492    
493        /* Skip these, but we need to add in the name length. */
494    
495        case OP_MARK:
496        case OP_PRUNE_ARG:
497        case OP_SKIP_ARG:
498        case OP_THEN_ARG:
499        cc += PRIV(OP_lengths)[op] + cc[1];
500        break;
501    
502        /* The remaining opcodes are just skipped over. */
503    
504        case OP_CLOSE:
505        case OP_COMMIT:
506        case OP_FAIL:
507        case OP_PRUNE:
508        case OP_SET_SOM:
509        case OP_SKIP:
510        case OP_THEN:
511        cc += PRIV(OP_lengths)[op];
512        break;
513    
514        /* This should not occur: we list all opcodes explicitly so that when
515        new ones get added they are properly considered. */
516    
517        default:
518        return -3;
519        }
520      }
521    /* Control never gets here */
522    }
523    
524    
525    
526  /*************************************************  /*************************************************
527  *      Set a bit and maybe its alternate case    *  *      Set a bit and maybe its alternate case    *
528  *************************************************/  *************************************************/
529    
530  /* 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
531  version of a letter if we are caseless.  corresponding bit for the other version of a letter if we are caseless. In
532    UTF-8 mode, for characters greater than 127, we can only do the caseless thing
533    when Unicode property support is available.
534    
535  Arguments:  Arguments:
536    start_bits    points to the bit map    start_bits    points to the bit map
537    c             is the character    p             points to the character
538    caseless      the caseless flag    caseless      the caseless flag
539    cd            the block with char table pointers    cd            the block with char table pointers
540      utf8          TRUE for UTF-8 mode
541    
542    Returns:        pointer after the character
543    */
544    
545    static const pcre_uchar *
546    set_table_bit(pcre_uint8 *start_bits, const pcre_uchar *p, BOOL caseless,
547      compile_data *cd, BOOL utf8)
548    {
549    unsigned int c = *p;
550    
551    SET_BIT(c);
552    
553    #ifdef SUPPORT_UTF8
554    if (utf8 && c > 127)
555      {
556      GETCHARINC(c, p);
557    #ifdef SUPPORT_UCP
558      if (caseless)
559        {
560        pcre_uint8 buff[8];
561        c = UCD_OTHERCASE(c);
562        (void)PRIV(ord2utf8)(c, buff);
563        SET_BIT(buff[0]);
564        }
565    #endif
566      return p;
567      }
568    #endif
569    
570    /* Not UTF-8 mode, or character is less than 127. */
571    
572    if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
573    return p + 1;
574    }
575    
576    
577    
578    /*************************************************
579    *     Set bits for a positive character type     *
580    *************************************************/
581    
582    /* This function sets starting bits for a character type. In UTF-8 mode, we can
583    only do a direct setting for bytes less than 128, as otherwise there can be
584    confusion with bytes in the middle of UTF-8 characters. In a "traditional"
585    environment, the tables will only recognize ASCII characters anyway, but in at
586    least one Windows environment, some higher bytes bits were set in the tables.
587    So we deal with that case by considering the UTF-8 encoding.
588    
589    Arguments:
590      start_bits     the starting bitmap
591      cbit type      the type of character wanted
592      table_limit    32 for non-UTF-8; 16 for UTF-8
593      cd             the block with char table pointers
594    
595    Returns:         nothing
596    */
597    
598    static void
599    set_type_bits(pcre_uint8 *start_bits, int cbit_type, int table_limit,
600      compile_data *cd)
601    {
602    register int c;
603    for (c = 0; c < table_limit; c++) start_bits[c] |= cd->cbits[c+cbit_type];
604    #ifdef SUPPORT_UTF8
605    if (table_limit == 32) return;
606    for (c = 128; c < 256; c++)
607      {
608      if ((cd->cbits[c/8] & (1 << (c&7))) != 0)
609        {
610        pcre_uint8 buff[8];
611        (void)PRIV(ord2utf8)(c, buff);
612        SET_BIT(buff[0]);
613        }
614      }
615    #endif
616    }
617    
618    
619    /*************************************************
620    *     Set bits for a negative character type     *
621    *************************************************/
622    
623    /* This function sets starting bits for a negative character type such as \D.
624    In UTF-8 mode, we can only do a direct setting for bytes less than 128, as
625    otherwise there can be confusion with bytes in the middle of UTF-8 characters.
626    Unlike in the positive case, where we can set appropriate starting bits for
627    specific high-valued UTF-8 characters, in this case we have to set the bits for
628    all high-valued characters. The lowest is 0xc2, but we overkill by starting at
629    0xc0 (192) for simplicity.
630    
631  Returns:        nothing  Arguments:
632      start_bits     the starting bitmap
633      cbit type      the type of character wanted
634      table_limit    32 for non-UTF-8; 16 for UTF-8
635      cd             the block with char table pointers
636    
637    Returns:         nothing
638  */  */
639    
640  static void  static void
641  set_bit(uschar *start_bits, unsigned int c, BOOL caseless, compile_data *cd)  set_nottype_bits(pcre_uint8 *start_bits, int cbit_type, int table_limit,
642      compile_data *cd)
643  {  {
644  start_bits[c/8] |= (1 << (c&7));  register int c;
645  if (caseless && (cd->ctypes[c] & ctype_letter) != 0)  for (c = 0; c < table_limit; c++) start_bits[c] |= ~cd->cbits[c+cbit_type];
646    start_bits[cd->fcc[c]/8] |= (1 << (cd->fcc[c]&7));  if (table_limit != 32) for (c = 24; c < 32; c++) start_bits[c] = 0xff;
647  }  }
648    
649    
# Line 95  function fails unless the result is SSB_ Line 663  function fails unless the result is SSB_
663  Arguments:  Arguments:
664    code         points to an expression    code         points to an expression
665    start_bits   points to a 32-byte table, initialized to 0    start_bits   points to a 32-byte table, initialized to 0
   caseless     the current state of the caseless flag  
666    utf8         TRUE if in UTF-8 mode    utf8         TRUE if in UTF-8 mode
667    cd           the block with char table pointers    cd           the block with char table pointers
668    
669  Returns:       SSB_FAIL     => Failed to find any starting bytes  Returns:       SSB_FAIL     => Failed to find any starting bytes
670                 SSB_DONE     => Found mandatory starting bytes                 SSB_DONE     => Found mandatory starting bytes
671                 SSB_CONTINUE => Found optional starting bytes                 SSB_CONTINUE => Found optional starting bytes
672                   SSB_UNKNOWN  => Hit an unrecognized opcode
673  */  */
674    
675  static int  static int
676  set_start_bits(const uschar *code, uschar *start_bits, BOOL caseless,  set_start_bits(const pcre_uchar *code, pcre_uint8 *start_bits, BOOL utf8,
677    BOOL utf8, compile_data *cd)    compile_data *cd)
678  {  {
679  register int c;  register int c;
680  int yield = SSB_DONE;  int yield = SSB_DONE;
681    int table_limit = utf8? 16:32;
682    
683  #if 0  #if 0
684  /* ========================================================================= */  /* ========================================================================= */
# Line 130  volatile int dummy; Line 699  volatile int dummy;
699    
700  do  do
701    {    {
   const uschar *tcode = code + (((int)*code == OP_CBRA)? 3:1) + LINK_SIZE;  
702    BOOL try_next = TRUE;    BOOL try_next = TRUE;
703      const pcre_uchar *tcode = code + 1 + LINK_SIZE;
704    
705      if (*code == OP_CBRA || *code == OP_SCBRA ||
706          *code == OP_CBRAPOS || *code == OP_SCBRAPOS) tcode += IMM2_SIZE;
707    
708    while (try_next)    /* Loop for items in this branch */    while (try_next)    /* Loop for items in this branch */
709      {      {
710      int rc;      int rc;
711    
712      switch(*tcode)      switch(*tcode)
713        {        {
714        /* Fail if we reach something we don't understand */        /* If we reach something we don't understand, it means a new opcode has
715          been created that hasn't been added to this code. Hopefully this problem
716          will be discovered during testing. */
717    
718        default:        default:
719          return SSB_UNKNOWN;
720    
721          /* Fail for a valid opcode that implies no starting bits. */
722    
723          case OP_ACCEPT:
724          case OP_ASSERT_ACCEPT:
725          case OP_ALLANY:
726          case OP_ANY:
727          case OP_ANYBYTE:
728          case OP_CIRC:
729          case OP_CIRCM:
730          case OP_CLOSE:
731          case OP_COMMIT:
732          case OP_COND:
733          case OP_CREF:
734          case OP_DEF:
735          case OP_DOLL:
736          case OP_DOLLM:
737          case OP_END:
738          case OP_EOD:
739          case OP_EODN:
740          case OP_EXTUNI:
741          case OP_FAIL:
742          case OP_MARK:
743          case OP_NCREF:
744          case OP_NOT:
745          case OP_NOTEXACT:
746          case OP_NOTEXACTI:
747          case OP_NOTI:
748          case OP_NOTMINPLUS:
749          case OP_NOTMINPLUSI:
750          case OP_NOTMINQUERY:
751          case OP_NOTMINQUERYI:
752          case OP_NOTMINSTAR:
753          case OP_NOTMINSTARI:
754          case OP_NOTMINUPTO:
755          case OP_NOTMINUPTOI:
756          case OP_NOTPLUS:
757          case OP_NOTPLUSI:
758          case OP_NOTPOSPLUS:
759          case OP_NOTPOSPLUSI:
760          case OP_NOTPOSQUERY:
761          case OP_NOTPOSQUERYI:
762          case OP_NOTPOSSTAR:
763          case OP_NOTPOSSTARI:
764          case OP_NOTPOSUPTO:
765          case OP_NOTPOSUPTOI:
766          case OP_NOTPROP:
767          case OP_NOTQUERY:
768          case OP_NOTQUERYI:
769          case OP_NOTSTAR:
770          case OP_NOTSTARI:
771          case OP_NOTUPTO:
772          case OP_NOTUPTOI:
773          case OP_NOT_HSPACE:
774          case OP_NOT_VSPACE:
775          case OP_NRREF:
776          case OP_PROP:
777          case OP_PRUNE:
778          case OP_PRUNE_ARG:
779          case OP_RECURSE:
780          case OP_REF:
781          case OP_REFI:
782          case OP_REVERSE:
783          case OP_RREF:
784          case OP_SCOND:
785          case OP_SET_SOM:
786          case OP_SKIP:
787          case OP_SKIP_ARG:
788          case OP_SOD:
789          case OP_SOM:
790          case OP_THEN:
791          case OP_THEN_ARG:
792    #if defined SUPPORT_UTF8 || !defined COMPILE_PCRE8
793          case OP_XCLASS:
794    #endif
795        return SSB_FAIL;        return SSB_FAIL;
796    
797          /* We can ignore word boundary tests. */
798    
799          case OP_WORD_BOUNDARY:
800          case OP_NOT_WORD_BOUNDARY:
801          tcode++;
802          break;
803    
804        /* 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
805        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
806        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 810  do
810        case OP_SBRA:        case OP_SBRA:
811        case OP_CBRA:        case OP_CBRA:
812        case OP_SCBRA:        case OP_SCBRA:
813          case OP_BRAPOS:
814          case OP_SBRAPOS:
815          case OP_CBRAPOS:
816          case OP_SCBRAPOS:
817        case OP_ONCE:        case OP_ONCE:
818          case OP_ONCE_NC:
819        case OP_ASSERT:        case OP_ASSERT:
820        rc = set_start_bits(tcode, start_bits, caseless, utf8, cd);        rc = set_start_bits(tcode, start_bits, utf8, cd);
821        if (rc == SSB_FAIL) return SSB_FAIL;        if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
822        if (rc == SSB_DONE) try_next = FALSE; else        if (rc == SSB_DONE) try_next = FALSE; else
823          {          {
824          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
# Line 178  do Line 841  do
841        case OP_KET:        case OP_KET:
842        case OP_KETRMAX:        case OP_KETRMAX:
843        case OP_KETRMIN:        case OP_KETRMIN:
844          case OP_KETRPOS:
845        return SSB_CONTINUE;        return SSB_CONTINUE;
846    
847        /* Skip over callout */        /* Skip over callout */
# Line 195  do Line 859  do
859        tcode += 1 + LINK_SIZE;        tcode += 1 + LINK_SIZE;
860        break;        break;
861    
       /* Skip over an option setting, changing the caseless flag */  
   
       case OP_OPT:  
       caseless = (tcode[1] & PCRE_CASELESS) != 0;  
       tcode += 2;  
       break;  
   
862        /* BRAZERO does the bracket, but carries on. */        /* BRAZERO does the bracket, but carries on. */
863    
864        case OP_BRAZERO:        case OP_BRAZERO:
865        case OP_BRAMINZERO:        case OP_BRAMINZERO:
866        if (set_start_bits(++tcode, start_bits, caseless, utf8, cd) == SSB_FAIL)        case OP_BRAPOSZERO:
867          return SSB_FAIL;        rc = set_start_bits(++tcode, start_bits, utf8, cd);
868          if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
869  /* =========================================================================  /* =========================================================================
870        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,
871        which was an old fudge for the benefit of OS/2.        which was an old fudge for the benefit of OS/2.
# Line 217  do Line 875  do
875        tcode += 1 + LINK_SIZE;        tcode += 1 + LINK_SIZE;
876        break;        break;
877    
878          /* SKIPZERO skips the bracket. */
879    
880          case OP_SKIPZERO:
881          tcode++;
882          do tcode += GET(tcode,1); while (*tcode == OP_ALT);
883          tcode += 1 + LINK_SIZE;
884          break;
885    
886        /* Single-char * or ? sets the bit and tries the next item */        /* Single-char * or ? sets the bit and tries the next item */
887    
888        case OP_STAR:        case OP_STAR:
# Line 225  do Line 891  do
891        case OP_QUERY:        case OP_QUERY:
892        case OP_MINQUERY:        case OP_MINQUERY:
893        case OP_POSQUERY:        case OP_POSQUERY:
894        set_bit(start_bits, tcode[1], caseless, cd);        tcode = set_table_bit(start_bits, tcode + 1, FALSE, cd, utf8);
895        tcode += 2;        break;
896  #ifdef SUPPORT_UTF8  
897        if (utf8 && tcode[-1] >= 0xc0)        case OP_STARI:
898          tcode += _pcre_utf8_table4[tcode[-1] & 0x3f];        case OP_MINSTARI:
899  #endif        case OP_POSSTARI:
900          case OP_QUERYI:
901          case OP_MINQUERYI:
902          case OP_POSQUERYI:
903          tcode = set_table_bit(start_bits, tcode + 1, TRUE, cd, utf8);
904        break;        break;
905    
906        /* Single-char upto sets the bit and tries the next */        /* Single-char upto sets the bit and tries the next */
# Line 238  do Line 908  do
908        case OP_UPTO:        case OP_UPTO:
909        case OP_MINUPTO:        case OP_MINUPTO:
910        case OP_POSUPTO:        case OP_POSUPTO:
911        set_bit(start_bits, tcode[3], caseless, cd);        tcode = set_table_bit(start_bits, tcode + 1 + IMM2_SIZE, FALSE, cd, utf8);
       tcode += 4;  
 #ifdef SUPPORT_UTF8  
       if (utf8 && tcode[-1] >= 0xc0)  
         tcode += _pcre_utf8_table4[tcode[-1] & 0x3f];  
 #endif  
912        break;        break;
913    
914        /* At least one single char sets the bit and stops */        case OP_UPTOI:
915          case OP_MINUPTOI:
916          case OP_POSUPTOI:
917          tcode = set_table_bit(start_bits, tcode + 1 + IMM2_SIZE, TRUE, cd, utf8);
918          break;
919    
920        case OP_EXACT:       /* Fall through */        /* At least one single char sets the bit and stops */
       tcode += 2;  
921    
922          case OP_EXACT:
923          tcode += IMM2_SIZE;
924          /* Fall through */
925        case OP_CHAR:        case OP_CHAR:
       case OP_CHARNC:  
926        case OP_PLUS:        case OP_PLUS:
927        case OP_MINPLUS:        case OP_MINPLUS:
928        case OP_POSPLUS:        case OP_POSPLUS:
929        set_bit(start_bits, tcode[1], caseless, cd);        (void)set_table_bit(start_bits, tcode + 1, FALSE, cd, utf8);
930        try_next = FALSE;        try_next = FALSE;
931        break;        break;
932    
933        /* Single character type sets the bits and stops */        case OP_EXACTI:
934          tcode += IMM2_SIZE;
935          /* Fall through */
936          case OP_CHARI:
937          case OP_PLUSI:
938          case OP_MINPLUSI:
939          case OP_POSPLUSI:
940          (void)set_table_bit(start_bits, tcode + 1, TRUE, cd, utf8);
941          try_next = FALSE;
942          break;
943    
944          /* Special spacing and line-terminating items. These recognize specific
945          lists of characters. The difference between VSPACE and ANYNL is that the
946          latter can match the two-character CRLF sequence, but that is not
947          relevant for finding the first character, so their code here is
948          identical. */
949    
950          case OP_HSPACE:
951          SET_BIT(0x09);
952          SET_BIT(0x20);
953          if (utf8)
954            {
955            SET_BIT(0xC2);  /* For U+00A0 */
956            SET_BIT(0xE1);  /* For U+1680, U+180E */
957            SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
958            SET_BIT(0xE3);  /* For U+3000 */
959            }
960          else SET_BIT(0xA0);
961          try_next = FALSE;
962          break;
963    
964          case OP_ANYNL:
965          case OP_VSPACE:
966          SET_BIT(0x0A);
967          SET_BIT(0x0B);
968          SET_BIT(0x0C);
969          SET_BIT(0x0D);
970          if (utf8)
971            {
972            SET_BIT(0xC2);  /* For U+0085 */
973            SET_BIT(0xE2);  /* For U+2028, U+2029 */
974            }
975          else SET_BIT(0x85);
976          try_next = FALSE;
977          break;
978    
979          /* Single character types set the bits and stop. Note that if PCRE_UCP
980          is set, we do not see these op codes because \d etc are converted to
981          properties. Therefore, these apply in the case when only characters less
982          than 256 are recognized to match the types. */
983    
984        case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
985        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_digit];  
986        try_next = FALSE;        try_next = FALSE;
987        break;        break;
988    
989        case OP_DIGIT:        case OP_DIGIT:
990        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_digit];  
991        try_next = FALSE;        try_next = FALSE;
992        break;        break;
993    
994        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to
995        discard it. */        ensure it is set as not whitespace. */
996    
997        case OP_NOT_WHITESPACE:        case OP_NOT_WHITESPACE:
998        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_space, table_limit, cd);
999          {        start_bits[1] |= 0x08;
         int d = cd->cbits[c+cbit_space];  
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= ~d;  
         }  
1000        try_next = FALSE;        try_next = FALSE;
1001        break;        break;
1002    
1003        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to
1004        discard it. */        not set it from the table. */
1005    
1006        case OP_WHITESPACE:        case OP_WHITESPACE:
1007        for (c = 0; c < 32; c++)        c = start_bits[1];    /* Save in case it was already set */
1008          {        set_type_bits(start_bits, cbit_space, table_limit, cd);
1009          int d = cd->cbits[c+cbit_space];        start_bits[1] = (start_bits[1] & ~0x08) | c;
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= d;  
         }  
1010        try_next = FALSE;        try_next = FALSE;
1011        break;        break;
1012    
1013        case OP_NOT_WORDCHAR:        case OP_NOT_WORDCHAR:
1014        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_word];  
1015        try_next = FALSE;        try_next = FALSE;
1016        break;        break;
1017    
1018        case OP_WORDCHAR:        case OP_WORDCHAR:
1019        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_word];  
1020        try_next = FALSE;        try_next = FALSE;
1021        break;        break;
1022    
# Line 317  do Line 1025  do
1025    
1026        case OP_TYPEPLUS:        case OP_TYPEPLUS:
1027        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
1028          case OP_TYPEPOSPLUS:
1029        tcode++;        tcode++;
1030        break;        break;
1031    
1032        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1033        tcode += 3;        tcode += 1 + IMM2_SIZE;
1034        break;        break;
1035    
1036        /* 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 330  do Line 1039  do
1039        case OP_TYPEUPTO:        case OP_TYPEUPTO:
1040        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
1041        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1042        tcode += 2;               /* Fall through */        tcode += IMM2_SIZE;  /* Fall through */
1043    
1044        case OP_TYPESTAR:        case OP_TYPESTAR:
1045        case OP_TYPEMINSTAR:        case OP_TYPEMINSTAR:
# Line 340  do Line 1049  do
1049        case OP_TYPEPOSQUERY:        case OP_TYPEPOSQUERY:
1050        switch(tcode[1])        switch(tcode[1])
1051          {          {
1052            default:
1053          case OP_ANY:          case OP_ANY:
1054            case OP_ALLANY:
1055          return SSB_FAIL;          return SSB_FAIL;
1056    
1057            case OP_HSPACE:
1058            SET_BIT(0x09);
1059            SET_BIT(0x20);
1060            if (utf8)
1061              {
1062              SET_BIT(0xC2);  /* For U+00A0 */
1063              SET_BIT(0xE1);  /* For U+1680, U+180E */
1064              SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
1065              SET_BIT(0xE3);  /* For U+3000 */
1066              }
1067            else SET_BIT(0xA0);
1068            break;
1069    
1070            case OP_ANYNL:
1071            case OP_VSPACE:
1072            SET_BIT(0x0A);
1073            SET_BIT(0x0B);
1074            SET_BIT(0x0C);
1075            SET_BIT(0x0D);
1076            if (utf8)
1077              {
1078              SET_BIT(0xC2);  /* For U+0085 */
1079              SET_BIT(0xE2);  /* For U+2028, U+2029 */
1080              }
1081            else SET_BIT(0x85);
1082            break;
1083    
1084          case OP_NOT_DIGIT:          case OP_NOT_DIGIT:
1085          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_digit];  
1086          break;          break;
1087    
1088          case OP_DIGIT:          case OP_DIGIT:
1089          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_digit];  
1090          break;          break;
1091    
1092          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we have to
1093          discard it. */          ensure it gets set as not whitespace. */
1094    
1095          case OP_NOT_WHITESPACE:          case OP_NOT_WHITESPACE:
1096          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_space, table_limit, cd);
1097            {          start_bits[1] |= 0x08;
           int d = cd->cbits[c+cbit_space];  
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= ~d;  
           }  
1098          break;          break;
1099    
1100          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we have to
1101          discard it. */          avoid setting it. */
1102    
1103          case OP_WHITESPACE:          case OP_WHITESPACE:
1104          for (c = 0; c < 32; c++)          c = start_bits[1];    /* Save in case it was already set */
1105            {          set_type_bits(start_bits, cbit_space, table_limit, cd);
1106            int d = cd->cbits[c+cbit_space];          start_bits[1] = (start_bits[1] & ~0x08) | c;
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= d;  
           }  
1107          break;          break;
1108    
1109          case OP_NOT_WORDCHAR:          case OP_NOT_WORDCHAR:
1110          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_word];  
1111          break;          break;
1112    
1113          case OP_WORDCHAR:          case OP_WORDCHAR:
1114          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_word];  
1115          break;          break;
1116          }          }
1117    
# Line 409  do Line 1136  do
1136    
1137        case OP_CLASS:        case OP_CLASS:
1138          {          {
1139            pcre_uint8 *map;
1140          tcode++;          tcode++;
1141            map = (pcre_uint8 *)tcode;
1142    
1143          /* 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
1144          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 420  do Line 1149  do
1149  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1150          if (utf8)          if (utf8)
1151            {            {
1152            for (c = 0; c < 16; c++) start_bits[c] |= tcode[c];            for (c = 0; c < 16; c++) start_bits[c] |= map[c];
1153            for (c = 128; c < 256; c++)            for (c = 128; c < 256; c++)
1154              {              {
1155              if ((tcode[c/8] && (1 << (c&7))) != 0)              if ((map[c/8] && (1 << (c&7))) != 0)
1156                {                {
1157                int d = (c >> 6) | 0xc0;            /* Set bit for this starter */                int d = (c >> 6) | 0xc0;            /* Set bit for this starter */
1158                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 437  do Line 1166  do
1166          else          else
1167  #endif  #endif
1168            {            {
1169            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];            for (c = 0; c < 32; c++) start_bits[c] |= map[c];
1170            }            }
1171    
1172          /* Advance past the bit map, and act on what follows */          /* Advance past the bit map, and act on what follows. For a zero
1173            minimum repeat, continue; otherwise stop processing. */
1174    
1175          tcode += 32;          tcode += 32 / sizeof(pcre_uchar);
1176          switch (*tcode)          switch (*tcode)
1177            {            {
1178            case OP_CRSTAR:            case OP_CRSTAR:
# Line 454  do Line 1184  do
1184    
1185            case OP_CRRANGE:            case OP_CRRANGE:
1186            case OP_CRMINRANGE:            case OP_CRMINRANGE:
1187            if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;            if (GET2(tcode, 1) == 0) tcode += 1 + 2 * IMM2_SIZE;
1188              else try_next = FALSE;              else try_next = FALSE;
1189            break;            break;
1190    
# Line 476  return yield; Line 1206  return yield;
1206    
1207    
1208    
1209    
1210    
1211  /*************************************************  /*************************************************
1212  *          Study a compiled expression           *  *          Study a compiled expression           *
1213  *************************************************/  *************************************************/
# Line 491  Arguments: Line 1223  Arguments:
1223              set NULL unless error              set NULL unless error
1224    
1225  Returns:    pointer to a pcre_extra block, with study_data filled in and the  Returns:    pointer to a pcre_extra block, with study_data filled in and the
1226                appropriate flag set;                appropriate flags set;
1227              NULL on error or if no optimization possible              NULL on error or if no optimization possible
1228  */  */
1229    
1230  PCRE_EXP_DEFN pcre_extra *  #ifdef COMPILE_PCRE8
1231    PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION
1232  pcre_study(const pcre *external_re, int options, const char **errorptr)  pcre_study(const pcre *external_re, int options, const char **errorptr)
1233    #else
1234    PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION
1235    pcre16_study(const pcre *external_re, int options, const char **errorptr)
1236    #endif
1237  {  {
1238  uschar start_bits[32];  int min;
1239  pcre_extra *extra;  BOOL bits_set = FALSE;
1240    pcre_uint8 start_bits[32];
1241    pcre_extra *extra = NULL;
1242  pcre_study_data *study;  pcre_study_data *study;
1243  const uschar *tables;  const pcre_uint8 *tables;
1244  uschar *code;  pcre_uchar *code;
1245  compile_data compile_block;  compile_data compile_block;
1246  const real_pcre *re = (const real_pcre *)external_re;  const real_pcre *re = (const real_pcre *)external_re;
1247    
# Line 520  if ((options & ~PUBLIC_STUDY_OPTIONS) != Line 1259  if ((options & ~PUBLIC_STUDY_OPTIONS) !=
1259    return NULL;    return NULL;
1260    }    }
1261    
1262  code = (uschar *)re + re->name_table_offset +  code = (pcre_uchar *)re + re->name_table_offset +
1263    (re->name_count * re->name_entry_size);    (re->name_count * re->name_entry_size);
1264    
1265  /* 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
1266  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
1267  at present. */  seeking a list of starting bytes. */
1268    
1269  if ((re->options & (PCRE_ANCHORED|PCRE_FIRSTSET|PCRE_STARTLINE)) != 0)  if ((re->options & PCRE_ANCHORED) == 0 &&
1270    return NULL;      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)
1271      {
1272      int rc;
1273    
1274  /* Set the character tables in the block that is passed around */    /* Set the character tables in the block that is passed around */
1275    
1276  tables = re->tables;    tables = re->tables;
1277  if (tables == NULL)    if (tables == NULL)
1278    (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,      (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1279    (void *)(&tables));      (void *)(&tables));
1280    
1281  compile_block.lcc = tables + lcc_offset;    compile_block.lcc = tables + lcc_offset;
1282  compile_block.fcc = tables + fcc_offset;    compile_block.fcc = tables + fcc_offset;
1283  compile_block.cbits = tables + cbits_offset;    compile_block.cbits = tables + cbits_offset;
1284  compile_block.ctypes = tables + ctypes_offset;    compile_block.ctypes = tables + ctypes_offset;
1285    
1286  /* See if we can find a fixed set of initial characters for the pattern. */    /* See if we can find a fixed set of initial characters for the pattern. */
1287    
1288  memset(start_bits, 0, 32 * sizeof(uschar));    memset(start_bits, 0, 32 * sizeof(pcre_uint8));
1289  if (set_start_bits(code, start_bits, (re->options & PCRE_CASELESS) != 0,    rc = set_start_bits(code, start_bits, (re->options & PCRE_UTF8) != 0,
1290    (re->options & PCRE_UTF8) != 0, &compile_block) != SSB_DONE) return NULL;      &compile_block);
1291      bits_set = rc == SSB_DONE;
1292  /* Get a pcre_extra block and a pcre_study_data block. The study data is put in    if (rc == SSB_UNKNOWN)
1293  the latter, which is pointed to by the former, which may also get additional      {
1294  data set later by the calling program. At the moment, the size of      *errorptr = "internal error: opcode not recognized";
1295  pcre_study_data is fixed. We nevertheless save it in a field for returning via      return NULL;
1296  the pcre_fullinfo() function so that if it becomes variable in the future, we      }
1297  don't have to change that code. */    }
1298    
1299  extra = (pcre_extra *)(pcre_malloc)  /* Find the minimum length of subject string. */
   (sizeof(pcre_extra) + sizeof(pcre_study_data));  
1300    
1301  if (extra == NULL)  switch(min = find_minlength(code, code, re->options, 0))
1302    {    {
1303    *errorptr = "failed to get memory";    case -2: *errorptr = "internal error: missing capturing bracket"; return NULL;
1304    return NULL;    case -3: *errorptr = "internal error: opcode not recognized"; return NULL;
1305      default: break;
1306    }    }
1307    
1308  study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));  /* If a set of starting bytes has been identified, or if the minimum length is
1309  extra->flags = PCRE_EXTRA_STUDY_DATA;  greater than zero, or if JIT optimization has been requested, get a pcre_extra
1310  extra->study_data = study;  block and a pcre_study_data block. The study data is put in the latter, which
1311    is pointed to by the former, which may also get additional data set later by
1312  study->size = sizeof(pcre_study_data);  the calling program. At the moment, the size of pcre_study_data is fixed. We
1313  study->options = PCRE_STUDY_MAPPED;  nevertheless save it in a field for returning via the pcre_fullinfo() function
1314  memcpy(study->start_bits, start_bits, sizeof(start_bits));  so that if it becomes variable in the future, we don't have to change that
1315    code. */
1316    
1317    if (bits_set || min > 0
1318    #ifdef SUPPORT_JIT
1319        || (options & PCRE_STUDY_JIT_COMPILE) != 0
1320    #endif
1321      )
1322      {
1323      extra = (pcre_extra *)(pcre_malloc)
1324        (sizeof(pcre_extra) + sizeof(pcre_study_data));
1325      if (extra == NULL)
1326        {
1327        *errorptr = "failed to get memory";
1328        return NULL;
1329        }
1330    
1331      study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));
1332      extra->flags = PCRE_EXTRA_STUDY_DATA;
1333      extra->study_data = study;
1334    
1335      study->size = sizeof(pcre_study_data);
1336      study->flags = 0;
1337    
1338      if (bits_set)
1339        {
1340        study->flags |= PCRE_STUDY_MAPPED;
1341        memcpy(study->start_bits, start_bits, sizeof(start_bits));
1342        }
1343    
1344      /* Always set the minlength value in the block, because the JIT compiler
1345      makes use of it. However, don't set the bit unless the length is greater than
1346      zero - the interpretive pcre_exec() and pcre_dfa_exec() needn't waste time
1347      checking the zero case. */
1348    
1349      if (min > 0)
1350        {
1351        study->flags |= PCRE_STUDY_MINLEN;
1352        study->minlength = min;
1353        }
1354      else study->minlength = 0;
1355    
1356      /* If JIT support was compiled and requested, attempt the JIT compilation.
1357      If no starting bytes were found, and the minimum length is zero, and JIT
1358      compilation fails, abandon the extra block and return NULL. */
1359    
1360    #ifdef SUPPORT_JIT
1361      extra->executable_jit = NULL;
1362      if ((options & PCRE_STUDY_JIT_COMPILE) != 0) PRIV(jit_compile)(re, extra);
1363      if (study->flags == 0 && (extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) == 0)
1364        {
1365        pcre_free_study(extra);
1366        extra = NULL;
1367        }
1368    #endif
1369      }
1370    
1371  return extra;  return extra;
1372  }  }
1373    
1374    
1375    /*************************************************
1376    *          Free the study data                   *
1377    *************************************************/
1378    
1379    /* This function frees the memory that was obtained by pcre_study().
1380    
1381    Argument:   a pointer to the pcre_extra block
1382    Returns:    nothing
1383    */
1384    
1385    #ifdef COMPILE_PCRE8
1386    PCRE_EXP_DEFN void
1387    pcre_free_study(pcre_extra *extra)
1388    #else
1389    PCRE_EXP_DEFN void
1390    pcre16_free_study(pcre_extra *extra)
1391    #endif
1392    {
1393    #ifdef SUPPORT_JIT
1394    if ((extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 &&
1395         extra->executable_jit != NULL)
1396      PRIV(jit_free)(extra->executable_jit);
1397    #endif
1398    pcre_free(extra);
1399    }
1400    
1401  /* End of pcre_study.c */  /* End of pcre_study.c */
Legend:
Removed from v.200  
changed lines
  Added in v.770
webmaster@exim.org
 ViewVC Help
Powered by ViewVC 1.1.12