/[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 510 by ph10, Sat Mar 27 17:45:29 2010 UTC revision 1084 by chpe, Tue Oct 16 15:55:28 2012 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) 1997-2010 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    
# Line 65  string of that length that matches. In U Line 66  string of that length that matches. In U
66  rather than bytes.  rather than bytes.
67    
68  Arguments:  Arguments:
69    code       pointer to start of group (the bracket)    code            pointer to start of group (the bracket)
70    startcode  pointer to start of the whole pattern    startcode       pointer to start of the whole pattern
71    options    the compiling options    options         the compiling options
72      int             RECURSE depth
73    
74  Returns:   the minimum length  Returns:   the minimum length
75             -1 if \C was encountered             -1 if \C in UTF-8 mode or (*ACCEPT) was encountered
76             -2 internal error (missing capturing bracket)             -2 internal error (missing capturing bracket)
77               -3 internal error (opcode not listed)
78  */  */
79    
80  static int  static int
81  find_minlength(const uschar *code, const uschar *startcode, int options)  find_minlength(const pcre_uchar *code, const pcre_uchar *startcode, int options,
82      int recurse_depth)
83  {  {
84  int length = -1;  int length = -1;
85  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
86    BOOL utf = (options & PCRE_UTF8) != 0;
87  BOOL had_recurse = FALSE;  BOOL had_recurse = FALSE;
88  register int branchlength = 0;  register int branchlength = 0;
89  register uschar *cc = (uschar *)code + 1 + LINK_SIZE;  register pcre_uchar *cc = (pcre_uchar *)code + 1 + LINK_SIZE;
90    
91  if (*code == OP_CBRA || *code == OP_SCBRA) cc += 2;  if (*code == OP_CBRA || *code == OP_SCBRA ||
92        *code == OP_CBRAPOS || *code == OP_SCBRAPOS) cc += IMM2_SIZE;
93    
94  /* Scan along the opcodes for this branch. If we get to the end of the  /* Scan along the opcodes for this branch. If we get to the end of the
95  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 91  branch, check the length against that of Line 97  branch, check the length against that of
97  for (;;)  for (;;)
98    {    {
99    int d, min;    int d, min;
100    uschar *cs, *ce;    pcre_uchar *cs, *ce;
101    register int op = *cc;    register pcre_uchar op = *cc;
102    
103    switch (op)    switch (op)
104      {      {
# Line 117  for (;;) Line 123  for (;;)
123      case OP_SCBRA:      case OP_SCBRA:
124      case OP_BRA:      case OP_BRA:
125      case OP_SBRA:      case OP_SBRA:
126        case OP_CBRAPOS:
127        case OP_SCBRAPOS:
128        case OP_BRAPOS:
129        case OP_SBRAPOS:
130      case OP_ONCE:      case OP_ONCE:
131      d = find_minlength(cc, startcode, options);      case OP_ONCE_NC:
132        d = find_minlength(cc, startcode, options, recurse_depth);
133      if (d < 0) return d;      if (d < 0) return d;
134      branchlength += d;      branchlength += d;
135      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
136      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
137      break;      break;
138    
139        /* ACCEPT makes things far too complicated; we have to give up. */
140    
141        case OP_ACCEPT:
142        case OP_ASSERT_ACCEPT:
143        return -1;
144    
145      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested
146      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is END it's
147      END it's the end of the outer call. All can be handled by the same code. */      the end of the outer call. All can be handled by the same code. If an
148        ACCEPT was previously encountered, use the length that was in force at that
149        time, and pass back the shortest ACCEPT length. */
150    
151      case OP_ALT:      case OP_ALT:
152      case OP_KET:      case OP_KET:
153      case OP_KETRMAX:      case OP_KETRMAX:
154      case OP_KETRMIN:      case OP_KETRMIN:
155        case OP_KETRPOS:
156      case OP_END:      case OP_END:
157      if (length < 0 || (!had_recurse && branchlength < length))      if (length < 0 || (!had_recurse && branchlength < length))
158        length = branchlength;        length = branchlength;
159      if (*cc != OP_ALT) return length;      if (op != OP_ALT) return length;
160      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
161      branchlength = 0;      branchlength = 0;
162      had_recurse = FALSE;      had_recurse = FALSE;
# Line 159  for (;;) Line 179  for (;;)
179      case OP_RREF:      case OP_RREF:
180      case OP_NRREF:      case OP_NRREF:
181      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
182      case OP_CALLOUT:      case OP_CALLOUT:
183      case OP_SOD:      case OP_SOD:
184      case OP_SOM:      case OP_SOM:
185      case OP_EOD:      case OP_EOD:
186      case OP_EODN:      case OP_EODN:
187      case OP_CIRC:      case OP_CIRC:
188        case OP_CIRCM:
189      case OP_DOLL:      case OP_DOLL:
190        case OP_DOLLM:
191      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
192      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
193      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
194      break;      break;
195    
196      /* Skip over a subpattern that has a {0} or {0,x} quantifier */      /* Skip over a subpattern that has a {0} or {0,x} quantifier */
197    
198      case OP_BRAZERO:      case OP_BRAZERO:
199      case OP_BRAMINZERO:      case OP_BRAMINZERO:
200        case OP_BRAPOSZERO:
201      case OP_SKIPZERO:      case OP_SKIPZERO:
202      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
203      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
204      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
205      break;      break;
# Line 185  for (;;) Line 207  for (;;)
207      /* Handle literal characters and + repetitions */      /* Handle literal characters and + repetitions */
208    
209      case OP_CHAR:      case OP_CHAR:
210      case OP_CHARNC:      case OP_CHARI:
211      case OP_NOT:      case OP_NOT:
212        case OP_NOTI:
213      case OP_PLUS:      case OP_PLUS:
214        case OP_PLUSI:
215      case OP_MINPLUS:      case OP_MINPLUS:
216        case OP_MINPLUSI:
217      case OP_POSPLUS:      case OP_POSPLUS:
218        case OP_POSPLUSI:
219      case OP_NOTPLUS:      case OP_NOTPLUS:
220        case OP_NOTPLUSI:
221      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
222        case OP_NOTMINPLUSI:
223      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
224        case OP_NOTPOSPLUSI:
225      branchlength++;      branchlength++;
226      cc += 2;      cc += 2;
227  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
228      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
229  #endif  #endif
230      break;      break;
231    
# Line 211  for (;;) Line 240  for (;;)
240      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
241    
242      case OP_EXACT:      case OP_EXACT:
243        case OP_EXACTI:
244      case OP_NOTEXACT:      case OP_NOTEXACT:
245        case OP_NOTEXACTI:
246      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
247      cc += 4;      cc += 2 + IMM2_SIZE;
248  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
249      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
250  #endif  #endif
251      break;      break;
252    
253      case OP_TYPEEXACT:      case OP_TYPEEXACT:
254      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
255      cc += (cc[3] == OP_PROP || cc[3] == OP_NOTPROP)? 6 : 4;      cc += 2 + IMM2_SIZE + ((cc[1 + IMM2_SIZE] == OP_PROP
256          || cc[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
257      break;      break;
258    
259      /* Handle single-char non-literal matchers */      /* Handle single-char non-literal matchers */
# Line 248  for (;;) Line 280  for (;;)
280      cc++;      cc++;
281      break;      break;
282    
283      /* "Any newline" might match two characters */      /* "Any newline" might match two characters, but it also might match just
284        one. */
285    
286      case OP_ANYNL:      case OP_ANYNL:
287      branchlength += 2;      branchlength += 1;
288      cc++;      cc++;
289      break;      break;
290    
291      /* The single-byte matcher means we can't proceed in UTF-8 mode */      /* The single-byte matcher means we can't proceed in UTF-8 mode. (In
292        non-UTF-8 mode \C will actually be turned into OP_ALLANY, so won't ever
293        appear, but leave the code, just in case.) */
294    
295      case OP_ANYBYTE:      case OP_ANYBYTE:
296  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
297      if (utf8) return -1;      if (utf) return -1;
298  #endif  #endif
299      branchlength++;      branchlength++;
300      cc++;      cc++;
# Line 275  for (;;) Line 310  for (;;)
310      case OP_TYPEPOSSTAR:      case OP_TYPEPOSSTAR:
311      case OP_TYPEPOSQUERY:      case OP_TYPEPOSQUERY:
312      if (cc[1] == OP_PROP || cc[1] == OP_NOTPROP) cc += 2;      if (cc[1] == OP_PROP || cc[1] == OP_NOTPROP) cc += 2;
313      cc += _pcre_OP_lengths[op];      cc += PRIV(OP_lengths)[op];
314      break;      break;
315    
316      case OP_TYPEUPTO:      case OP_TYPEUPTO:
317      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
318      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
319      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP
320      cc += _pcre_OP_lengths[op];        || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;
321        cc += PRIV(OP_lengths)[op];
322      break;      break;
323    
324      /* Check a class for variable quantification */      /* Check a class for variable quantification */
325    
326  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
327      case OP_XCLASS:      case OP_XCLASS:
328      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
329      /* Fall through */      /* Fall through */
330  #endif  #endif
331    
332      case OP_CLASS:      case OP_CLASS:
333      case OP_NCLASS:      case OP_NCLASS:
334      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
335    
336      switch (*cc)      switch (*cc)
337        {        {
# Line 314  for (;;) Line 350  for (;;)
350        case OP_CRRANGE:        case OP_CRRANGE:
351        case OP_CRMINRANGE:        case OP_CRMINRANGE:
352        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
353        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
354        break;        break;
355    
356        default:        default:
# Line 336  for (;;) Line 372  for (;;)
372      that case we must set the minimum length to zero. */      that case we must set the minimum length to zero. */
373    
374      case OP_REF:      case OP_REF:
375        case OP_REFI:
376      if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
377        {        {
378        ce = cs = (uschar *)_pcre_find_bracket(startcode, utf8, GET2(cc, 1));        ce = cs = (pcre_uchar *)PRIV(find_bracket)(startcode, utf, GET2(cc, 1));
379        if (cs == NULL) return -2;        if (cs == NULL) return -2;
380        do ce += GET(ce, 1); while (*ce == OP_ALT);        do ce += GET(ce, 1); while (*ce == OP_ALT);
381        if (cc > cs && cc < ce)        if (cc > cs && cc < ce)
# Line 346  for (;;) Line 383  for (;;)
383          d = 0;          d = 0;
384          had_recurse = TRUE;          had_recurse = TRUE;
385          }          }
386        else d = find_minlength(cs, startcode, options);        else
387            {
388            d = find_minlength(cs, startcode, options, recurse_depth);
389            }
390        }        }
391      else d = 0;      else d = 0;
392      cc += 3;      cc += 1 + IMM2_SIZE;
393    
394      /* Handle repeated back references */      /* Handle repeated back references */
395    
# Line 363  for (;;) Line 403  for (;;)
403        cc++;        cc++;
404        break;        break;
405    
406          case OP_CRPLUS:
407          case OP_CRMINPLUS:
408          min = 1;
409          cc++;
410          break;
411    
412        case OP_CRRANGE:        case OP_CRRANGE:
413        case OP_CRMINRANGE:        case OP_CRMINRANGE:
414        min = GET2(cc, 1);        min = GET2(cc, 1);
415        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
416        break;        break;
417    
418        default:        default:
# Line 377  for (;;) Line 423  for (;;)
423      branchlength += min * d;      branchlength += min * d;
424      break;      break;
425    
426        /* We can easily detect direct recursion, but not mutual recursion. This is
427        caught by a recursion depth count. */
428    
429      case OP_RECURSE:      case OP_RECURSE:
430      cs = ce = (uschar *)startcode + GET(cc, 1);      cs = ce = (pcre_uchar *)startcode + GET(cc, 1);
     if (cs == NULL) return -2;  
431      do ce += GET(ce, 1); while (*ce == OP_ALT);      do ce += GET(ce, 1); while (*ce == OP_ALT);
432      if (cc > cs && cc < ce)      if ((cc > cs && cc < ce) || recurse_depth > 10)
433        had_recurse = TRUE;        had_recurse = TRUE;
434      else      else
435        branchlength += find_minlength(cs, startcode, options);        {
436          branchlength += find_minlength(cs, startcode, options, recurse_depth + 1);
437          }
438      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
439      break;      break;
440    
441      /* Anything else does not or need not match a character. We can get the      /* Anything else does not or need not match a character. We can get the
442      item's length from the table, but for those that can match zero occurrences      item's length from the table, but for those that can match zero occurrences
443      of a character, we must take special action for UTF-8 characters. */      of a character, we must take special action for UTF-8 characters. As it
444        happens, the "NOT" versions of these opcodes are used at present only for
445        ASCII characters, so they could be omitted from this list. However, in
446        future that may change, so we include them here so as not to leave a
447        gotcha for a future maintainer. */
448    
449      case OP_UPTO:      case OP_UPTO:
450        case OP_UPTOI:
451      case OP_NOTUPTO:      case OP_NOTUPTO:
452        case OP_NOTUPTOI:
453      case OP_MINUPTO:      case OP_MINUPTO:
454        case OP_MINUPTOI:
455      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
456        case OP_NOTMINUPTOI:
457      case OP_POSUPTO:      case OP_POSUPTO:
458        case OP_POSUPTOI:
459        case OP_NOTPOSUPTO:
460        case OP_NOTPOSUPTOI:
461    
462      case OP_STAR:      case OP_STAR:
463        case OP_STARI:
464        case OP_NOTSTAR:
465        case OP_NOTSTARI:
466      case OP_MINSTAR:      case OP_MINSTAR:
467        case OP_MINSTARI:
468      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
469        case OP_NOTMINSTARI:
470      case OP_POSSTAR:      case OP_POSSTAR:
471        case OP_POSSTARI:
472      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
473        case OP_NOTPOSSTARI:
474    
475      case OP_QUERY:      case OP_QUERY:
476        case OP_QUERYI:
477        case OP_NOTQUERY:
478        case OP_NOTQUERYI:
479      case OP_MINQUERY:      case OP_MINQUERY:
480        case OP_MINQUERYI:
481      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
482        case OP_NOTMINQUERYI:
483      case OP_POSQUERY:      case OP_POSQUERY:
484        case OP_POSQUERYI:
485      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
486      cc += _pcre_OP_lengths[op];      case OP_NOTPOSQUERYI:
487  #ifdef SUPPORT_UTF8  
488      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      cc += PRIV(OP_lengths)[op];
489    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
490        if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
491  #endif  #endif
492      break;      break;
493    
494      /* Skip these, but we need to add in the name length. */      /* Skip these, but we need to add in the name length. */
495    
496      case OP_MARK:      case OP_MARK:
497      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
498      case OP_SKIP_ARG:      case OP_SKIP_ARG:
499      case OP_THEN_ARG:      case OP_THEN_ARG:
500      cc += _pcre_OP_lengths[op] + cc[1];      cc += PRIV(OP_lengths)[op] + cc[1];
501      break;      break;
502    
503      /* For the record, these are the opcodes that are matched by "default":      /* The remaining opcodes are just skipped over. */
     OP_ACCEPT, OP_CLOSE, OP_COMMIT, OP_FAIL, OP_PRUNE, OP_SET_SOM, OP_SKIP,  
     OP_THEN. */  
504    
505      default:      case OP_CLOSE:
506      cc += _pcre_OP_lengths[op];      case OP_COMMIT:
507        case OP_FAIL:
508        case OP_PRUNE:
509        case OP_SET_SOM:
510        case OP_SKIP:
511        case OP_THEN:
512        cc += PRIV(OP_lengths)[op];
513      break;      break;
514    
515        /* This should not occur: we list all opcodes explicitly so that when
516        new ones get added they are properly considered. */
517    
518        default:
519        return -3;
520      }      }
521    }    }
522  /* Control never gets here */  /* Control never gets here */
# Line 440  for (;;) Line 528  for (;;)
528  *      Set a bit and maybe its alternate case    *  *      Set a bit and maybe its alternate case    *
529  *************************************************/  *************************************************/
530    
531  /* 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
532  version of a letter if we are caseless.  corresponding bit for the other version of a letter if we are caseless. In
533    UTF-8 mode, for characters greater than 127, we can only do the caseless thing
534    when Unicode property support is available.
535    
536  Arguments:  Arguments:
537    start_bits    points to the bit map    start_bits    points to the bit map
538    c             is the character    p             points to the character
539    caseless      the caseless flag    caseless      the caseless flag
540    cd            the block with char table pointers    cd            the block with char table pointers
541      utf           TRUE for UTF-8 / UTF-16 / UTF-32 mode
542    
543    Returns:        pointer after the character
544    */
545    
546    static const pcre_uchar *
547    set_table_bit(pcre_uint8 *start_bits, const pcre_uchar *p, BOOL caseless,
548      compile_data *cd, BOOL utf)
549    {
550    unsigned int c = *p;
551    
552    #ifdef COMPILE_PCRE8
553    SET_BIT(c);
554    
555    #ifdef SUPPORT_UTF
556    if (utf && c > 127)
557      {
558      GETCHARINC(c, p);
559    #ifdef SUPPORT_UCP
560      if (caseless)
561        {
562        pcre_uchar buff[6];
563        c = UCD_OTHERCASE(c);
564        (void)PRIV(ord2utf)(c, buff);
565        SET_BIT(buff[0]);
566        }
567    #endif  /* Not SUPPORT_UCP */
568      return p;
569      }
570    #else   /* Not SUPPORT_UTF */
571    (void)(utf);   /* Stops warning for unused parameter */
572    #endif  /* SUPPORT_UTF */
573    
574    /* Not UTF-8 mode, or character is less than 127. */
575    
576    if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
577    return p + 1;
578    #endif  /* COMPILE_PCRE8 */
579    
580    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
581    if (c > 0xff)
582      {
583      c = 0xff;
584      caseless = FALSE;
585      }
586    SET_BIT(c);
587    
588    #ifdef SUPPORT_UTF
589    if (utf && c > 127)
590      {
591      GETCHARINC(c, p);
592    #ifdef SUPPORT_UCP
593      if (caseless)
594        {
595        c = UCD_OTHERCASE(c);
596        if (c > 0xff)
597          c = 0xff;
598        SET_BIT(c);
599        }
600    #endif  /* SUPPORT_UCP */
601      return p;
602      }
603    #else   /* Not SUPPORT_UTF */
604    (void)(utf);   /* Stops warning for unused parameter */
605    #endif  /* SUPPORT_UTF */
606    
607    if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
608    return p + 1;
609    #endif
610    }
611    
612    
613    
614    /*************************************************
615    *     Set bits for a positive character type     *
616    *************************************************/
617    
618  Returns:        nothing  /* This function sets starting bits for a character type. In UTF-8 mode, we can
619    only do a direct setting for bytes less than 128, as otherwise there can be
620    confusion with bytes in the middle of UTF-8 characters. In a "traditional"
621    environment, the tables will only recognize ASCII characters anyway, but in at
622    least one Windows environment, some higher bytes bits were set in the tables.
623    So we deal with that case by considering the UTF-8 encoding.
624    
625    Arguments:
626      start_bits     the starting bitmap
627      cbit type      the type of character wanted
628      table_limit    32 for non-UTF-8; 16 for UTF-8
629      cd             the block with char table pointers
630    
631    Returns:         nothing
632  */  */
633    
634  static void  static void
635  set_table_bit(uschar *start_bits, unsigned int c, BOOL caseless,  set_type_bits(pcre_uint8 *start_bits, int cbit_type, int table_limit,
636    compile_data *cd)    compile_data *cd)
637  {  {
638  start_bits[c/8] |= (1 << (c&7));  register int c;
639  if (caseless && (cd->ctypes[c] & ctype_letter) != 0)  for (c = 0; c < table_limit; c++) start_bits[c] |= cd->cbits[c+cbit_type];
640    start_bits[cd->fcc[c]/8] |= (1 << (cd->fcc[c]&7));  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
641    if (table_limit == 32) return;
642    for (c = 128; c < 256; c++)
643      {
644      if ((cd->cbits[c/8] & (1 << (c&7))) != 0)
645        {
646        pcre_uchar buff[6];
647        (void)PRIV(ord2utf)(c, buff);
648        SET_BIT(buff[0]);
649        }
650      }
651    #endif
652    }
653    
654    
655    /*************************************************
656    *     Set bits for a negative character type     *
657    *************************************************/
658    
659    /* This function sets starting bits for a negative character type such as \D.
660    In UTF-8 mode, we can only do a direct setting for bytes less than 128, as
661    otherwise there can be confusion with bytes in the middle of UTF-8 characters.
662    Unlike in the positive case, where we can set appropriate starting bits for
663    specific high-valued UTF-8 characters, in this case we have to set the bits for
664    all high-valued characters. The lowest is 0xc2, but we overkill by starting at
665    0xc0 (192) for simplicity.
666    
667    Arguments:
668      start_bits     the starting bitmap
669      cbit type      the type of character wanted
670      table_limit    32 for non-UTF-8; 16 for UTF-8
671      cd             the block with char table pointers
672    
673    Returns:         nothing
674    */
675    
676    static void
677    set_nottype_bits(pcre_uint8 *start_bits, int cbit_type, int table_limit,
678      compile_data *cd)
679    {
680    register int c;
681    for (c = 0; c < table_limit; c++) start_bits[c] |= ~cd->cbits[c+cbit_type];
682    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
683    if (table_limit != 32) for (c = 24; c < 32; c++) start_bits[c] = 0xff;
684    #endif
685  }  }
686    
687    
# Line 478  function fails unless the result is SSB_ Line 701  function fails unless the result is SSB_
701  Arguments:  Arguments:
702    code         points to an expression    code         points to an expression
703    start_bits   points to a 32-byte table, initialized to 0    start_bits   points to a 32-byte table, initialized to 0
704    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  
705    cd           the block with char table pointers    cd           the block with char table pointers
706    
707  Returns:       SSB_FAIL     => Failed to find any starting bytes  Returns:       SSB_FAIL     => Failed to find any starting bytes
708                 SSB_DONE     => Found mandatory starting bytes                 SSB_DONE     => Found mandatory starting bytes
709                 SSB_CONTINUE => Found optional starting bytes                 SSB_CONTINUE => Found optional starting bytes
710                   SSB_UNKNOWN  => Hit an unrecognized opcode
711  */  */
712    
713  static int  static int
714  set_start_bits(const uschar *code, uschar *start_bits, BOOL caseless,  set_start_bits(const pcre_uchar *code, pcre_uint8 *start_bits, BOOL utf,
715    BOOL utf8, compile_data *cd)    compile_data *cd)
716  {  {
717  register int c;  register int c;
718  int yield = SSB_DONE;  int yield = SSB_DONE;
719    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
720    int table_limit = utf? 16:32;
721    #else
722    int table_limit = 32;
723    #endif
724    
725  #if 0  #if 0
726  /* ========================================================================= */  /* ========================================================================= */
# Line 513  volatile int dummy; Line 741  volatile int dummy;
741    
742  do  do
743    {    {
   const uschar *tcode = code + (((int)*code == OP_CBRA)? 3:1) + LINK_SIZE;  
744    BOOL try_next = TRUE;    BOOL try_next = TRUE;
745      const pcre_uchar *tcode = code + 1 + LINK_SIZE;
746    
747      if (*code == OP_CBRA || *code == OP_SCBRA ||
748          *code == OP_CBRAPOS || *code == OP_SCBRAPOS) tcode += IMM2_SIZE;
749    
750    while (try_next)    /* Loop for items in this branch */    while (try_next)    /* Loop for items in this branch */
751      {      {
752      int rc;      int rc;
753    
754      switch(*tcode)      switch(*tcode)
755        {        {
756        /* Fail if we reach something we don't understand */        /* If we reach something we don't understand, it means a new opcode has
757          been created that hasn't been added to this code. Hopefully this problem
758          will be discovered during testing. */
759    
760        default:        default:
761          return SSB_UNKNOWN;
762    
763          /* Fail for a valid opcode that implies no starting bits. */
764    
765          case OP_ACCEPT:
766          case OP_ASSERT_ACCEPT:
767          case OP_ALLANY:
768          case OP_ANY:
769          case OP_ANYBYTE:
770          case OP_CIRC:
771          case OP_CIRCM:
772          case OP_CLOSE:
773          case OP_COMMIT:
774          case OP_COND:
775          case OP_CREF:
776          case OP_DEF:
777          case OP_DOLL:
778          case OP_DOLLM:
779          case OP_END:
780          case OP_EOD:
781          case OP_EODN:
782          case OP_EXTUNI:
783          case OP_FAIL:
784          case OP_MARK:
785          case OP_NCREF:
786          case OP_NOT:
787          case OP_NOTEXACT:
788          case OP_NOTEXACTI:
789          case OP_NOTI:
790          case OP_NOTMINPLUS:
791          case OP_NOTMINPLUSI:
792          case OP_NOTMINQUERY:
793          case OP_NOTMINQUERYI:
794          case OP_NOTMINSTAR:
795          case OP_NOTMINSTARI:
796          case OP_NOTMINUPTO:
797          case OP_NOTMINUPTOI:
798          case OP_NOTPLUS:
799          case OP_NOTPLUSI:
800          case OP_NOTPOSPLUS:
801          case OP_NOTPOSPLUSI:
802          case OP_NOTPOSQUERY:
803          case OP_NOTPOSQUERYI:
804          case OP_NOTPOSSTAR:
805          case OP_NOTPOSSTARI:
806          case OP_NOTPOSUPTO:
807          case OP_NOTPOSUPTOI:
808          case OP_NOTPROP:
809          case OP_NOTQUERY:
810          case OP_NOTQUERYI:
811          case OP_NOTSTAR:
812          case OP_NOTSTARI:
813          case OP_NOTUPTO:
814          case OP_NOTUPTOI:
815          case OP_NOT_HSPACE:
816          case OP_NOT_VSPACE:
817          case OP_NRREF:
818          case OP_PROP:
819          case OP_PRUNE:
820          case OP_PRUNE_ARG:
821          case OP_RECURSE:
822          case OP_REF:
823          case OP_REFI:
824          case OP_REVERSE:
825          case OP_RREF:
826          case OP_SCOND:
827          case OP_SET_SOM:
828          case OP_SKIP:
829          case OP_SKIP_ARG:
830          case OP_SOD:
831          case OP_SOM:
832          case OP_THEN:
833          case OP_THEN_ARG:
834    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
835          case OP_XCLASS:
836    #endif
837        return SSB_FAIL;        return SSB_FAIL;
838    
839          /* We can ignore word boundary tests. */
840    
841          case OP_WORD_BOUNDARY:
842          case OP_NOT_WORD_BOUNDARY:
843          tcode++;
844          break;
845    
846        /* 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
847        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
848        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 535  do Line 852  do
852        case OP_SBRA:        case OP_SBRA:
853        case OP_CBRA:        case OP_CBRA:
854        case OP_SCBRA:        case OP_SCBRA:
855          case OP_BRAPOS:
856          case OP_SBRAPOS:
857          case OP_CBRAPOS:
858          case OP_SCBRAPOS:
859        case OP_ONCE:        case OP_ONCE:
860          case OP_ONCE_NC:
861        case OP_ASSERT:        case OP_ASSERT:
862        rc = set_start_bits(tcode, start_bits, caseless, utf8, cd);        rc = set_start_bits(tcode, start_bits, utf, cd);
863        if (rc == SSB_FAIL) return SSB_FAIL;        if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
864        if (rc == SSB_DONE) try_next = FALSE; else        if (rc == SSB_DONE) try_next = FALSE; else
865          {          {
866          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
# Line 561  do Line 883  do
883        case OP_KET:        case OP_KET:
884        case OP_KETRMAX:        case OP_KETRMAX:
885        case OP_KETRMIN:        case OP_KETRMIN:
886          case OP_KETRPOS:
887        return SSB_CONTINUE;        return SSB_CONTINUE;
888    
889        /* Skip over callout */        /* Skip over callout */
# Line 578  do Line 901  do
901        tcode += 1 + LINK_SIZE;        tcode += 1 + LINK_SIZE;
902        break;        break;
903    
       /* Skip over an option setting, changing the caseless flag */  
   
       case OP_OPT:  
       caseless = (tcode[1] & PCRE_CASELESS) != 0;  
       tcode += 2;  
       break;  
   
904        /* BRAZERO does the bracket, but carries on. */        /* BRAZERO does the bracket, but carries on. */
905    
906        case OP_BRAZERO:        case OP_BRAZERO:
907        case OP_BRAMINZERO:        case OP_BRAMINZERO:
908        if (set_start_bits(++tcode, start_bits, caseless, utf8, cd) == SSB_FAIL)        case OP_BRAPOSZERO:
909          return SSB_FAIL;        rc = set_start_bits(++tcode, start_bits, utf, cd);
910          if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
911  /* =========================================================================  /* =========================================================================
912        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,
913        which was an old fudge for the benefit of OS/2.        which was an old fudge for the benefit of OS/2.
# Line 616  do Line 933  do
933        case OP_QUERY:        case OP_QUERY:
934        case OP_MINQUERY:        case OP_MINQUERY:
935        case OP_POSQUERY:        case OP_POSQUERY:
936        set_table_bit(start_bits, tcode[1], caseless, cd);        tcode = set_table_bit(start_bits, tcode + 1, FALSE, cd, utf);
937        tcode += 2;        break;
938  #ifdef SUPPORT_UTF8  
939        if (utf8 && tcode[-1] >= 0xc0)        case OP_STARI:
940          tcode += _pcre_utf8_table4[tcode[-1] & 0x3f];        case OP_MINSTARI:
941  #endif        case OP_POSSTARI:
942          case OP_QUERYI:
943          case OP_MINQUERYI:
944          case OP_POSQUERYI:
945          tcode = set_table_bit(start_bits, tcode + 1, TRUE, cd, utf);
946        break;        break;
947    
948        /* Single-char upto sets the bit and tries the next */        /* Single-char upto sets the bit and tries the next */
# Line 629  do Line 950  do
950        case OP_UPTO:        case OP_UPTO:
951        case OP_MINUPTO:        case OP_MINUPTO:
952        case OP_POSUPTO:        case OP_POSUPTO:
953        set_table_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  
954        break;        break;
955    
956        /* At least one single char sets the bit and stops */        case OP_UPTOI:
957          case OP_MINUPTOI:
958          case OP_POSUPTOI:
959          tcode = set_table_bit(start_bits, tcode + 1 + IMM2_SIZE, TRUE, cd, utf);
960          break;
961    
962        case OP_EXACT:       /* Fall through */        /* At least one single char sets the bit and stops */
       tcode += 2;  
963    
964          case OP_EXACT:
965          tcode += IMM2_SIZE;
966          /* Fall through */
967        case OP_CHAR:        case OP_CHAR:
       case OP_CHARNC:  
968        case OP_PLUS:        case OP_PLUS:
969        case OP_MINPLUS:        case OP_MINPLUS:
970        case OP_POSPLUS:        case OP_POSPLUS:
971        set_table_bit(start_bits, tcode[1], caseless, cd);        (void)set_table_bit(start_bits, tcode + 1, FALSE, cd, utf);
972        try_next = FALSE;        try_next = FALSE;
973        break;        break;
974    
975        /* Single character type sets the bits and stops */        case OP_EXACTI:
976          tcode += IMM2_SIZE;
977          /* Fall through */
978          case OP_CHARI:
979          case OP_PLUSI:
980          case OP_MINPLUSI:
981          case OP_POSPLUSI:
982          (void)set_table_bit(start_bits, tcode + 1, TRUE, cd, utf);
983          try_next = FALSE;
984          break;
985    
986          /* Special spacing and line-terminating items. These recognize specific
987          lists of characters. The difference between VSPACE and ANYNL is that the
988          latter can match the two-character CRLF sequence, but that is not
989          relevant for finding the first character, so their code here is
990          identical. */
991    
992          case OP_HSPACE:
993          SET_BIT(CHAR_HT);
994          SET_BIT(CHAR_SPACE);
995    #ifdef SUPPORT_UTF
996          if (utf)
997            {
998    #ifdef COMPILE_PCRE8
999            SET_BIT(0xC2);  /* For U+00A0 */
1000            SET_BIT(0xE1);  /* For U+1680, U+180E */
1001            SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
1002            SET_BIT(0xE3);  /* For U+3000 */
1003    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1004            SET_BIT(0xA0);
1005            SET_BIT(0xFF);  /* For characters > 255 */
1006    #endif  /* COMPILE_PCRE[8|16|32] */
1007            }
1008          else
1009    #endif /* SUPPORT_UTF */
1010            {
1011    #ifndef EBCDIC
1012            SET_BIT(0xA0);
1013    #endif  /* Not EBCDIC */
1014    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1015            SET_BIT(0xFF);  /* For characters > 255 */
1016    #endif  /* COMPILE_PCRE[16|32] */
1017            }
1018          try_next = FALSE;
1019          break;
1020    
1021          case OP_ANYNL:
1022          case OP_VSPACE:
1023          SET_BIT(CHAR_LF);
1024          SET_BIT(CHAR_VT);
1025          SET_BIT(CHAR_FF);
1026          SET_BIT(CHAR_CR);
1027    #ifdef SUPPORT_UTF
1028          if (utf)
1029            {
1030    #ifdef COMPILE_PCRE8
1031            SET_BIT(0xC2);  /* For U+0085 */
1032            SET_BIT(0xE2);  /* For U+2028, U+2029 */
1033    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1034            SET_BIT(CHAR_NEL);
1035            SET_BIT(0xFF);  /* For characters > 255 */
1036    #endif  /* COMPILE_PCRE[8|16|32] */
1037            }
1038          else
1039    #endif /* SUPPORT_UTF */
1040            {
1041            SET_BIT(CHAR_NEL);
1042    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1043            SET_BIT(0xFF);  /* For characters > 255 */
1044    #endif
1045            }
1046          try_next = FALSE;
1047          break;
1048    
1049          /* Single character types set the bits and stop. Note that if PCRE_UCP
1050          is set, we do not see these op codes because \d etc are converted to
1051          properties. Therefore, these apply in the case when only characters less
1052          than 256 are recognized to match the types. */
1053    
1054        case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
1055        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_digit];  
1056        try_next = FALSE;        try_next = FALSE;
1057        break;        break;
1058    
1059        case OP_DIGIT:        case OP_DIGIT:
1060        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_digit];  
1061        try_next = FALSE;        try_next = FALSE;
1062        break;        break;
1063    
1064        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to
1065        discard it. */        ensure it is set as not whitespace. Luckily, the code value is the same
1066          (0x0b) in ASCII and EBCDIC, so we can just adjust the appropriate bit. */
1067    
1068        case OP_NOT_WHITESPACE:        case OP_NOT_WHITESPACE:
1069        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_space, table_limit, cd);
1070          {        start_bits[1] |= 0x08;
         int d = cd->cbits[c+cbit_space];  
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= ~d;  
         }  
1071        try_next = FALSE;        try_next = FALSE;
1072        break;        break;
1073    
1074        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to not
1075        discard it. */        set it from the table. Luckily, the code value is the same (0x0b) in
1076          ASCII and EBCDIC, so we can just adjust the appropriate bit. */
1077    
1078        case OP_WHITESPACE:        case OP_WHITESPACE:
1079        for (c = 0; c < 32; c++)        c = start_bits[1];    /* Save in case it was already set */
1080          {        set_type_bits(start_bits, cbit_space, table_limit, cd);
1081          int d = cd->cbits[c+cbit_space];        start_bits[1] = (start_bits[1] & ~0x08) | c;
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= d;  
         }  
1082        try_next = FALSE;        try_next = FALSE;
1083        break;        break;
1084    
1085        case OP_NOT_WORDCHAR:        case OP_NOT_WORDCHAR:
1086        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_word];  
1087        try_next = FALSE;        try_next = FALSE;
1088        break;        break;
1089    
1090        case OP_WORDCHAR:        case OP_WORDCHAR:
1091        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_word];  
1092        try_next = FALSE;        try_next = FALSE;
1093        break;        break;
1094    
# Line 708  do Line 1097  do
1097    
1098        case OP_TYPEPLUS:        case OP_TYPEPLUS:
1099        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
1100          case OP_TYPEPOSPLUS:
1101        tcode++;        tcode++;
1102        break;        break;
1103    
1104        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1105        tcode += 3;        tcode += 1 + IMM2_SIZE;
1106        break;        break;
1107    
1108        /* 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 721  do Line 1111  do
1111        case OP_TYPEUPTO:        case OP_TYPEUPTO:
1112        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
1113        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1114        tcode += 2;               /* Fall through */        tcode += IMM2_SIZE;  /* Fall through */
1115    
1116        case OP_TYPESTAR:        case OP_TYPESTAR:
1117        case OP_TYPEMINSTAR:        case OP_TYPEMINSTAR:
# Line 731  do Line 1121  do
1121        case OP_TYPEPOSQUERY:        case OP_TYPEPOSQUERY:
1122        switch(tcode[1])        switch(tcode[1])
1123          {          {
1124            default:
1125          case OP_ANY:          case OP_ANY:
1126          case OP_ALLANY:          case OP_ALLANY:
1127          return SSB_FAIL;          return SSB_FAIL;
1128    
1129            case OP_HSPACE:
1130            SET_BIT(CHAR_HT);
1131            SET_BIT(CHAR_SPACE);
1132    #ifdef SUPPORT_UTF
1133            if (utf)
1134              {
1135    #ifdef COMPILE_PCRE8
1136              SET_BIT(0xC2);  /* For U+00A0 */
1137              SET_BIT(0xE1);  /* For U+1680, U+180E */
1138              SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
1139              SET_BIT(0xE3);  /* For U+3000 */
1140    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1141              SET_BIT(0xA0);
1142              SET_BIT(0xFF);  /* For characters > 255 */
1143    #endif  /* COMPILE_PCRE[8|16|32] */
1144              }
1145            else
1146    #endif /* SUPPORT_UTF */
1147    #ifndef EBCDIC
1148              SET_BIT(0xA0);
1149    #endif  /* Not EBCDIC */
1150            break;
1151    
1152            case OP_ANYNL:
1153            case OP_VSPACE:
1154            SET_BIT(CHAR_LF);
1155            SET_BIT(CHAR_VT);
1156            SET_BIT(CHAR_FF);
1157            SET_BIT(CHAR_CR);
1158    #ifdef SUPPORT_UTF
1159            if (utf)
1160              {
1161    #ifdef COMPILE_PCRE8
1162              SET_BIT(0xC2);  /* For U+0085 */
1163              SET_BIT(0xE2);  /* For U+2028, U+2029 */
1164    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1165              SET_BIT(CHAR_NEL);
1166              SET_BIT(0xFF);  /* For characters > 255 */
1167    #endif  /* COMPILE_PCRE16 */
1168              }
1169            else
1170    #endif /* SUPPORT_UTF */
1171              SET_BIT(CHAR_NEL);
1172            break;
1173    
1174          case OP_NOT_DIGIT:          case OP_NOT_DIGIT:
1175          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_digit];  
1176          break;          break;
1177    
1178          case OP_DIGIT:          case OP_DIGIT:
1179          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_digit];  
1180          break;          break;
1181    
1182          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we have to
1183          discard it. */          ensure it gets set as not whitespace. Luckily, the code value is the
1184            same (0x0b) in ASCII and EBCDIC, so we can just adjust the appropriate
1185            bit. */
1186    
1187          case OP_NOT_WHITESPACE:          case OP_NOT_WHITESPACE:
1188          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_space, table_limit, cd);
1189            {          start_bits[1] |= 0x08;
           int d = cd->cbits[c+cbit_space];  
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= ~d;  
           }  
1190          break;          break;
1191    
1192          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we have to
1193          discard it. */          avoid setting it. Luckily, the code value is the same (0x0b) in ASCII
1194            and EBCDIC, so we can just adjust the appropriate bit. */
1195    
1196          case OP_WHITESPACE:          case OP_WHITESPACE:
1197          for (c = 0; c < 32; c++)          c = start_bits[1];    /* Save in case it was already set */
1198            {          set_type_bits(start_bits, cbit_space, table_limit, cd);
1199            int d = cd->cbits[c+cbit_space];          start_bits[1] = (start_bits[1] & ~0x08) | c;
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= d;  
           }  
1200          break;          break;
1201    
1202          case OP_NOT_WORDCHAR:          case OP_NOT_WORDCHAR:
1203          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_word];  
1204          break;          break;
1205    
1206          case OP_WORDCHAR:          case OP_WORDCHAR:
1207          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_word];  
1208          break;          break;
1209          }          }
1210    
# Line 790  do Line 1218  do
1218        character with a value > 255. */        character with a value > 255. */
1219    
1220        case OP_NCLASS:        case OP_NCLASS:
1221  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
1222        if (utf8)        if (utf)
1223          {          {
1224          start_bits[24] |= 0xf0;              /* Bits for 0xc4 - 0xc8 */          start_bits[24] |= 0xf0;              /* Bits for 0xc4 - 0xc8 */
1225          memset(start_bits+25, 0xff, 7);      /* Bits for 0xc9 - 0xff */          memset(start_bits+25, 0xff, 7);      /* Bits for 0xc9 - 0xff */
1226          }          }
1227  #endif  #endif
1228    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1229          SET_BIT(0xFF);                         /* For characters > 255 */
1230    #endif
1231        /* Fall through */        /* Fall through */
1232    
1233        case OP_CLASS:        case OP_CLASS:
1234          {          {
1235            pcre_uint8 *map;
1236          tcode++;          tcode++;
1237            map = (pcre_uint8 *)tcode;
1238    
1239          /* 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
1240          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 809  do Line 1242  do
1242          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
1243          characters in the range 128 - 255. */          characters in the range 128 - 255. */
1244    
1245  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
1246          if (utf8)          if (utf)
1247            {            {
1248            for (c = 0; c < 16; c++) start_bits[c] |= tcode[c];            for (c = 0; c < 16; c++) start_bits[c] |= map[c];
1249            for (c = 128; c < 256; c++)            for (c = 128; c < 256; c++)
1250              {              {
1251              if ((tcode[c/8] && (1 << (c&7))) != 0)              if ((map[c/8] && (1 << (c&7))) != 0)
1252                {                {
1253                int d = (c >> 6) | 0xc0;            /* Set bit for this starter */                int d = (c >> 6) | 0xc0;            /* Set bit for this starter */
1254                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 823  do Line 1256  do
1256                }                }
1257              }              }
1258            }            }
   
         /* In non-UTF-8 mode, the two bit maps are completely compatible. */  
   
1259          else          else
1260  #endif  #endif
1261            {            {
1262            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];            /* In non-UTF-8 mode, the two bit maps are completely compatible. */
1263              for (c = 0; c < 32; c++) start_bits[c] |= map[c];
1264            }            }
1265    
1266          /* Advance past the bit map, and act on what follows */          /* Advance past the bit map, and act on what follows. For a zero
1267            minimum repeat, continue; otherwise stop processing. */
1268    
1269          tcode += 32;          tcode += 32 / sizeof(pcre_uchar);
1270          switch (*tcode)          switch (*tcode)
1271            {            {
1272            case OP_CRSTAR:            case OP_CRSTAR:
# Line 846  do Line 1278  do
1278    
1279            case OP_CRRANGE:            case OP_CRRANGE:
1280            case OP_CRMINRANGE:            case OP_CRMINRANGE:
1281            if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;            if (GET2(tcode, 1) == 0) tcode += 1 + 2 * IMM2_SIZE;
1282              else try_next = FALSE;              else try_next = FALSE;
1283            break;            break;
1284    
# Line 868  return yield; Line 1300  return yield;
1300    
1301    
1302    
1303    
1304    
1305  /*************************************************  /*************************************************
1306  *          Study a compiled expression           *  *          Study a compiled expression           *
1307  *************************************************/  *************************************************/
1308    
1309  /* 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
1310  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
1311  which then gets handed back to pcre_exec().  which then gets handed back to pcre_exec().
1312    
1313  Arguments:  Arguments:
# Line 882  Arguments: Line 1316  Arguments:
1316    errorptr  points to where to place error messages;    errorptr  points to where to place error messages;
1317              set NULL unless error              set NULL unless error
1318    
1319  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
1320                appropriate flags set;                the appropriate flags set;
1321              NULL on error or if no optimization possible              NULL on error or if no optimization possible
1322  */  */
1323    
1324    #if defined COMPILE_PCRE8
1325  PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION  PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION
1326  pcre_study(const pcre *external_re, int options, const char **errorptr)  pcre_study(const pcre *external_re, int options, const char **errorptr)
1327    #elif defined COMPILE_PCRE16
1328    PCRE_EXP_DEFN pcre16_extra * PCRE_CALL_CONVENTION
1329    pcre16_study(const pcre16 *external_re, int options, const char **errorptr)
1330    #elif defined COMPILE_PCRE32
1331    PCRE_EXP_DEFN pcre32_extra * PCRE_CALL_CONVENTION
1332    pcre32_study(const pcre32 *external_re, int options, const char **errorptr)
1333    #endif
1334  {  {
1335  int min;  int min;
1336  BOOL bits_set = FALSE;  BOOL bits_set = FALSE;
1337  uschar start_bits[32];  pcre_uint8 start_bits[32];
1338  pcre_extra *extra;  PUBL(extra) *extra = NULL;
1339  pcre_study_data *study;  pcre_study_data *study;
1340  const uschar *tables;  const pcre_uint8 *tables;
1341  uschar *code;  pcre_uchar *code;
1342  compile_data compile_block;  compile_data compile_block;
1343  const real_pcre *re = (const real_pcre *)external_re;  const REAL_PCRE *re = (const REAL_PCRE *)external_re;
1344    
1345  *errorptr = NULL;  *errorptr = NULL;
1346    
# Line 908  if (re == NULL || re->magic_number != MA Line 1350  if (re == NULL || re->magic_number != MA
1350    return NULL;    return NULL;
1351    }    }
1352    
1353    if ((re->flags & PCRE_MODE) == 0)
1354      {
1355    #if defined COMPILE_PCRE8
1356      *errorptr = "argument not compiled in 8 bit mode";
1357    #elif defined COMPILE_PCRE16
1358      *errorptr = "argument not compiled in 16 bit mode";
1359    #elif defined COMPILE_PCRE32
1360      *errorptr = "argument not compiled in 32 bit mode";
1361    #endif
1362      return NULL;
1363      }
1364    
1365  if ((options & ~PUBLIC_STUDY_OPTIONS) != 0)  if ((options & ~PUBLIC_STUDY_OPTIONS) != 0)
1366    {    {
1367    *errorptr = "unknown or incorrect option bit(s) set";    *errorptr = "unknown or incorrect option bit(s) set";
1368    return NULL;    return NULL;
1369    }    }
1370    
1371  code = (uschar *)re + re->name_table_offset +  code = (pcre_uchar *)re + re->name_table_offset +
1372    (re->name_count * re->name_entry_size);    (re->name_count * re->name_entry_size);
1373    
1374  /* 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
# Line 924  seeking a list of starting bytes. */ Line 1378  seeking a list of starting bytes. */
1378  if ((re->options & PCRE_ANCHORED) == 0 &&  if ((re->options & PCRE_ANCHORED) == 0 &&
1379      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)
1380    {    {
1381      int rc;
1382    
1383    /* Set the character tables in the block that is passed around */    /* Set the character tables in the block that is passed around */
1384    
1385    tables = re->tables;    tables = re->tables;
1386    
1387    #if defined COMPILE_PCRE8
1388    if (tables == NULL)    if (tables == NULL)
1389      (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,      (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1390      (void *)(&tables));      (void *)(&tables));
1391    #elif defined COMPILE_PCRE16
1392      if (tables == NULL)
1393        (void)pcre16_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1394        (void *)(&tables));
1395    #elif defined COMPILE_PCRE32
1396      if (tables == NULL)
1397        (void)pcre32_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1398        (void *)(&tables));
1399    #endif
1400    
1401    compile_block.lcc = tables + lcc_offset;    compile_block.lcc = tables + lcc_offset;
1402    compile_block.fcc = tables + fcc_offset;    compile_block.fcc = tables + fcc_offset;
# Line 938  if ((re->options & PCRE_ANCHORED) == 0 & Line 1405  if ((re->options & PCRE_ANCHORED) == 0 &
1405    
1406    /* 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. */
1407    
1408    memset(start_bits, 0, 32 * sizeof(uschar));    memset(start_bits, 0, 32 * sizeof(pcre_uint8));
1409    bits_set = set_start_bits(code, start_bits,    rc = set_start_bits(code, start_bits, (re->options & PCRE_UTF8) != 0,
1410      (re->options & PCRE_CASELESS) != 0, (re->options & PCRE_UTF8) != 0,      &compile_block);
1411      &compile_block) == SSB_DONE;    bits_set = rc == SSB_DONE;
1412      if (rc == SSB_UNKNOWN)
1413        {
1414        *errorptr = "internal error: opcode not recognized";
1415        return NULL;
1416        }
1417    }    }
1418    
1419  /* Find the minimum length of subject string. */  /* Find the minimum length of subject string. */
1420    
1421  min = find_minlength(code, code, re->options);  switch(min = find_minlength(code, code, re->options, 0))
1422      {
1423      case -2: *errorptr = "internal error: missing capturing bracket"; return NULL;
1424      case -3: *errorptr = "internal error: opcode not recognized"; return NULL;
1425      default: break;
1426      }
1427    
1428  /* Return NULL if no optimization is possible. */  /* If a set of starting bytes has been identified, or if the minimum length is
1429    greater than zero, or if JIT optimization has been requested, or if
1430    PCRE_STUDY_EXTRA_NEEDED is set, get a pcre[16]_extra block and a
1431    pcre_study_data block. The study data is put in the latter, which is pointed to
1432    by the former, which may also get additional data set later by the calling
1433    program. At the moment, the size of pcre_study_data is fixed. We nevertheless
1434    save it in a field for returning via the pcre_fullinfo() function so that if it
1435    becomes variable in the future, we don't have to change that code. */
1436    
1437    if (bits_set || min > 0 || (options & (
1438    #ifdef SUPPORT_JIT
1439        PCRE_STUDY_JIT_COMPILE | PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE |
1440        PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE |
1441    #endif
1442        PCRE_STUDY_EXTRA_NEEDED)) != 0)
1443      {
1444      extra = (PUBL(extra) *)(PUBL(malloc))
1445        (sizeof(PUBL(extra)) + sizeof(pcre_study_data));
1446      if (extra == NULL)
1447        {
1448        *errorptr = "failed to get memory";
1449        return NULL;
1450        }
1451    
1452  if (!bits_set && min < 0) return NULL;    study = (pcre_study_data *)((char *)extra + sizeof(PUBL(extra)));
1453      extra->flags = PCRE_EXTRA_STUDY_DATA;
1454      extra->study_data = study;
1455    
1456      study->size = sizeof(pcre_study_data);
1457      study->flags = 0;
1458    
1459      /* Set the start bits always, to avoid unset memory errors if the
1460      study data is written to a file, but set the flag only if any of the bits
1461      are set, to save time looking when none are. */
1462    
1463  /* Get a pcre_extra block and a pcre_study_data block. The study data is put in    if (bits_set)
1464  the latter, which is pointed to by the former, which may also get additional      {
1465  data set later by the calling program. At the moment, the size of      study->flags |= PCRE_STUDY_MAPPED;
1466  pcre_study_data is fixed. We nevertheless save it in a field for returning via      memcpy(study->start_bits, start_bits, sizeof(start_bits));
1467  the pcre_fullinfo() function so that if it becomes variable in the future, we      }
1468  don't have to change that code. */    else memset(study->start_bits, 0, 32 * sizeof(pcre_uint8));
1469    
1470  extra = (pcre_extra *)(pcre_malloc)  #ifdef PCRE_DEBUG
1471    (sizeof(pcre_extra) + sizeof(pcre_study_data));    if (bits_set)
1472        {
1473        pcre_uint8 *ptr = start_bits;
1474        int i;
1475    
1476  if (extra == NULL)      printf("Start bits:\n");
1477    {      for (i = 0; i < 32; i++)
1478    *errorptr = "failed to get memory";        printf("%3d: %02x%s", i * 8, *ptr++, ((i + 1) & 0x7) != 0? " " : "\n");
1479    return NULL;      }
1480    }  #endif
1481    
1482  study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));    /* Always set the minlength value in the block, because the JIT compiler
1483  extra->flags = PCRE_EXTRA_STUDY_DATA;    makes use of it. However, don't set the bit unless the length is greater than
1484  extra->study_data = study;    zero - the interpretive pcre_exec() and pcre_dfa_exec() needn't waste time
1485      checking the zero case. */
1486    
1487  study->size = sizeof(pcre_study_data);    if (min > 0)
1488  study->flags = 0;      {
1489        study->flags |= PCRE_STUDY_MINLEN;
1490        study->minlength = min;
1491        }
1492      else study->minlength = 0;
1493    
1494  if (bits_set)    /* If JIT support was compiled and requested, attempt the JIT compilation.
1495    {    If no starting bytes were found, and the minimum length is zero, and JIT
1496    study->flags |= PCRE_STUDY_MAPPED;    compilation fails, abandon the extra block and return NULL, unless
1497    memcpy(study->start_bits, start_bits, sizeof(start_bits));    PCRE_STUDY_EXTRA_NEEDED is set. */
1498    }  
1499    #ifdef SUPPORT_JIT
1500      extra->executable_jit = NULL;
1501      if ((options & PCRE_STUDY_JIT_COMPILE) != 0)
1502        PRIV(jit_compile)(re, extra, JIT_COMPILE);
1503      if ((options & PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE) != 0)
1504        PRIV(jit_compile)(re, extra, JIT_PARTIAL_SOFT_COMPILE);
1505      if ((options & PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE) != 0)
1506        PRIV(jit_compile)(re, extra, JIT_PARTIAL_HARD_COMPILE);
1507    
1508  if (min >= 0)    if (study->flags == 0 && (extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) == 0 &&
1509    {        (options & PCRE_STUDY_EXTRA_NEEDED) == 0)
1510    study->flags |= PCRE_STUDY_MINLEN;      {
1511    study->minlength = min;  #if defined COMPILE_PCRE8
1512        pcre_free_study(extra);
1513    #elif defined COMPILE_PCRE16
1514        pcre16_free_study(extra);
1515    #elif defined COMPILE_PCRE32
1516        pcre32_free_study(extra);
1517    #endif
1518        extra = NULL;
1519        }
1520    #endif
1521    }    }
1522    
1523  return extra;  return extra;
1524  }  }
1525    
1526    
1527    /*************************************************
1528    *          Free the study data                   *
1529    *************************************************/
1530    
1531    /* This function frees the memory that was obtained by pcre_study().
1532    
1533    Argument:   a pointer to the pcre[16]_extra block
1534    Returns:    nothing
1535    */
1536    
1537    #if defined COMPILE_PCRE8
1538    PCRE_EXP_DEFN void
1539    pcre_free_study(pcre_extra *extra)
1540    #elif defined COMPILE_PCRE16
1541    PCRE_EXP_DEFN void
1542    pcre16_free_study(pcre16_extra *extra)
1543    #elif defined COMPILE_PCRE32
1544    PCRE_EXP_DEFN void
1545    pcre32_free_study(pcre32_extra *extra)
1546    #endif
1547    {
1548    if (extra == NULL)
1549      return;
1550    #ifdef SUPPORT_JIT
1551    if ((extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 &&
1552         extra->executable_jit != NULL)
1553      PRIV(jit_free)(extra->executable_jit);
1554    #endif
1555    PUBL(free)(extra);
1556    }
1557    
1558  /* End of pcre_study.c */  /* End of pcre_study.c */

Legend:
Removed from v.510  
changed lines
  Added in v.1084

  ViewVC Help
Powered by ViewVC 1.1.5