/[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 538 by ph10, Wed Jun 9 19:30:57 2010 UTC code/branches/pcre16/pcre_study.c revision 769 by zherczeg, Sun Nov 27 17:40:33 2011 UTC
# Line 52  supporting functions. */ Line 52  supporting functions. */
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 66  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;  BOOL utf8 = (options & PCRE_UTF8) != 0;
86  BOOL had_recurse = FALSE;  BOOL had_recurse = FALSE;
87  register int branchlength = 0;  register int branchlength = 0;
88  register uschar *cc = (uschar *)code + 1 + LINK_SIZE;  register pcre_uchar *cc = (pcre_uchar *)code + 1 + LINK_SIZE;
89    
90  if (*code == OP_CBRA || *code == OP_SCBRA) cc += 2;  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  /* 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. */  branch, check the length against that of the other branches. */
# Line 92  branch, check the length against that of Line 96  branch, check the length against that of
96  for (;;)  for (;;)
97    {    {
98    int d, min;    int d, min;
99    uschar *cs, *ce;    pcre_uchar *cs, *ce;
100    register int op = *cc;    register int op = *cc;
101    
102    switch (op)    switch (op)
# Line 118  for (;;) Line 122  for (;;)
122      case OP_SCBRA:      case OP_SCBRA:
123      case OP_BRA:      case OP_BRA:
124      case OP_SBRA:      case OP_SBRA:
125        case OP_CBRAPOS:
126        case OP_SCBRAPOS:
127        case OP_BRAPOS:
128        case OP_SBRAPOS:
129      case OP_ONCE:      case OP_ONCE:
130      d = find_minlength(cc, startcode, options);      case OP_ONCE_NC:
131        d = find_minlength(cc, startcode, options, recurse_depth);
132      if (d < 0) return d;      if (d < 0) return d;
133      branchlength += d;      branchlength += d;
134      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
135      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
136      break;      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      /* 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      call. If it's ALT it is an alternation in a nested call. If it is END it's
146      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
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:      case OP_ALT:
151      case OP_KET:      case OP_KET:
152      case OP_KETRMAX:      case OP_KETRMAX:
153      case OP_KETRMIN:      case OP_KETRMIN:
154        case OP_KETRPOS:
155      case OP_END:      case OP_END:
156      if (length < 0 || (!had_recurse && branchlength < length))      if (length < 0 || (!had_recurse && branchlength < length))
157        length = branchlength;        length = branchlength;
158      if (*cc != OP_ALT) return length;      if (op != OP_ALT) return length;
159      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
160      branchlength = 0;      branchlength = 0;
161      had_recurse = FALSE;      had_recurse = FALSE;
# Line 160  for (;;) Line 178  for (;;)
178      case OP_RREF:      case OP_RREF:
179      case OP_NRREF:      case OP_NRREF:
180      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
181      case OP_CALLOUT:      case OP_CALLOUT:
182      case OP_SOD:      case OP_SOD:
183      case OP_SOM:      case OP_SOM:
184      case OP_EOD:      case OP_EOD:
185      case OP_EODN:      case OP_EODN:
186      case OP_CIRC:      case OP_CIRC:
187        case OP_CIRCM:
188      case OP_DOLL:      case OP_DOLL:
189        case OP_DOLLM:
190      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
191      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
192      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
193      break;      break;
194    
195      /* Skip over a subpattern that has a {0} or {0,x} quantifier */      /* Skip over a subpattern that has a {0} or {0,x} quantifier */
196    
197      case OP_BRAZERO:      case OP_BRAZERO:
198      case OP_BRAMINZERO:      case OP_BRAMINZERO:
199        case OP_BRAPOSZERO:
200      case OP_SKIPZERO:      case OP_SKIPZERO:
201      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
202      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
203      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
204      break;      break;
# Line 186  for (;;) Line 206  for (;;)
206      /* Handle literal characters and + repetitions */      /* Handle literal characters and + repetitions */
207    
208      case OP_CHAR:      case OP_CHAR:
209      case OP_CHARNC:      case OP_CHARI:
210      case OP_NOT:      case OP_NOT:
211        case OP_NOTI:
212      case OP_PLUS:      case OP_PLUS:
213        case OP_PLUSI:
214      case OP_MINPLUS:      case OP_MINPLUS:
215        case OP_MINPLUSI:
216      case OP_POSPLUS:      case OP_POSPLUS:
217        case OP_POSPLUSI:
218      case OP_NOTPLUS:      case OP_NOTPLUS:
219        case OP_NOTPLUSI:
220      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
221        case OP_NOTMINPLUSI:
222      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
223        case OP_NOTPOSPLUSI:
224      branchlength++;      branchlength++;
225      cc += 2;      cc += 2;
226  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
227      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      if (utf8 && cc[-1] >= 0xc0) cc += PRIV(utf8_table4)[cc[-1] & 0x3f];
228  #endif  #endif
229      break;      break;
230    
# Line 212  for (;;) Line 239  for (;;)
239      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
240    
241      case OP_EXACT:      case OP_EXACT:
242        case OP_EXACTI:
243      case OP_NOTEXACT:      case OP_NOTEXACT:
244        case OP_NOTEXACTI:
245      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
246      cc += 4;      cc += 2 + IMM2_SIZE;
247  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
248      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      if (utf8 && cc[-1] >= 0xc0) cc += PRIV(utf8_table4)[cc[-1] & 0x3f];
249  #endif  #endif
250      break;      break;
251    
252      case OP_TYPEEXACT:      case OP_TYPEEXACT:
253      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
254      cc += (cc[3] == OP_PROP || cc[3] == OP_NOTPROP)? 6 : 4;      cc += 2 + IMM2_SIZE + ((cc[1 + IMM2_SIZE] == OP_PROP
255          || cc[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
256      break;      break;
257    
258      /* Handle single-char non-literal matchers */      /* Handle single-char non-literal matchers */
# Line 249  for (;;) Line 279  for (;;)
279      cc++;      cc++;
280      break;      break;
281    
282      /* "Any newline" might match two characters */      /* "Any newline" might match two characters, but it also might match just
283        one. */
284    
285      case OP_ANYNL:      case OP_ANYNL:
286      branchlength += 2;      branchlength += 1;
287      cc++;      cc++;
288      break;      break;
289    
290      /* 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
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:      case OP_ANYBYTE:
295  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 276  for (;;) Line 309  for (;;)
309      case OP_TYPEPOSSTAR:      case OP_TYPEPOSSTAR:
310      case OP_TYPEPOSQUERY:      case OP_TYPEPOSQUERY:
311      if (cc[1] == OP_PROP || cc[1] == OP_NOTPROP) cc += 2;      if (cc[1] == OP_PROP || cc[1] == OP_NOTPROP) cc += 2;
312      cc += _pcre_OP_lengths[op];      cc += PRIV(OP_lengths)[op];
313      break;      break;
314    
315      case OP_TYPEUPTO:      case OP_TYPEUPTO:
316      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
317      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
318      if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP
319      cc += _pcre_OP_lengths[op];        || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;
320        cc += PRIV(OP_lengths)[op];
321      break;      break;
322    
323      /* Check a class for variable quantification */      /* Check a class for variable quantification */
# Line 315  for (;;) Line 349  for (;;)
349        case OP_CRRANGE:        case OP_CRRANGE:
350        case OP_CRMINRANGE:        case OP_CRMINRANGE:
351        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
352        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
353        break;        break;
354    
355        default:        default:
# Line 337  for (;;) Line 371  for (;;)
371      that case we must set the minimum length to zero. */      that case we must set the minimum length to zero. */
372    
373      case OP_REF:      case OP_REF:
374        case OP_REFI:
375      if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
376        {        {
377        ce = cs = (uschar *)_pcre_find_bracket(startcode, utf8, GET2(cc, 1));        ce = cs = (pcre_uchar *)PRIV(find_bracket)(startcode, utf8, GET2(cc, 1));
378        if (cs == NULL) return -2;        if (cs == NULL) return -2;
379        do ce += GET(ce, 1); while (*ce == OP_ALT);        do ce += GET(ce, 1); while (*ce == OP_ALT);
380        if (cc > cs && cc < ce)        if (cc > cs && cc < ce)
# Line 347  for (;;) Line 382  for (;;)
382          d = 0;          d = 0;
383          had_recurse = TRUE;          had_recurse = TRUE;
384          }          }
385        else d = find_minlength(cs, startcode, options);        else
386            {
387            d = find_minlength(cs, startcode, options, recurse_depth);
388            }
389        }        }
390      else d = 0;      else d = 0;
391      cc += 3;      cc += 1 + IMM2_SIZE;
392    
393      /* Handle repeated back references */      /* Handle repeated back references */
394    
# Line 364  for (;;) Line 402  for (;;)
402        cc++;        cc++;
403        break;        break;
404    
405          case OP_CRPLUS:
406          case OP_CRMINPLUS:
407          min = 1;
408          cc++;
409          break;
410    
411        case OP_CRRANGE:        case OP_CRRANGE:
412        case OP_CRMINRANGE:        case OP_CRMINRANGE:
413        min = GET2(cc, 1);        min = GET2(cc, 1);
414        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
415        break;        break;
416    
417        default:        default:
# Line 378  for (;;) Line 422  for (;;)
422      branchlength += min * d;      branchlength += min * d;
423      break;      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:      case OP_RECURSE:
429      cs = ce = (uschar *)startcode + GET(cc, 1);      cs = ce = (pcre_uchar *)startcode + GET(cc, 1);
     if (cs == NULL) return -2;  
430      do ce += GET(ce, 1); while (*ce == OP_ALT);      do ce += GET(ce, 1); while (*ce == OP_ALT);
431      if (cc > cs && cc < ce)      if ((cc > cs && cc < ce) || recurse_depth > 10)
432        had_recurse = TRUE;        had_recurse = TRUE;
433      else      else
434        branchlength += find_minlength(cs, startcode, options);        {
435          branchlength += find_minlength(cs, startcode, options, recurse_depth + 1);
436          }
437      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
438      break;      break;
439    
440      /* 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
441      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
442      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
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:      case OP_UPTO:
449        case OP_UPTOI:
450      case OP_NOTUPTO:      case OP_NOTUPTO:
451        case OP_NOTUPTOI:
452      case OP_MINUPTO:      case OP_MINUPTO:
453        case OP_MINUPTOI:
454      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
455        case OP_NOTMINUPTOI:
456      case OP_POSUPTO:      case OP_POSUPTO:
457        case OP_POSUPTOI:
458        case OP_NOTPOSUPTO:
459        case OP_NOTPOSUPTOI:
460    
461      case OP_STAR:      case OP_STAR:
462        case OP_STARI:
463        case OP_NOTSTAR:
464        case OP_NOTSTARI:
465      case OP_MINSTAR:      case OP_MINSTAR:
466        case OP_MINSTARI:
467      case OP_NOTMINSTAR:      case OP_NOTMINSTAR:
468        case OP_NOTMINSTARI:
469      case OP_POSSTAR:      case OP_POSSTAR:
470        case OP_POSSTARI:
471      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
472        case OP_NOTPOSSTARI:
473    
474      case OP_QUERY:      case OP_QUERY:
475        case OP_QUERYI:
476        case OP_NOTQUERY:
477        case OP_NOTQUERYI:
478      case OP_MINQUERY:      case OP_MINQUERY:
479        case OP_MINQUERYI:
480      case OP_NOTMINQUERY:      case OP_NOTMINQUERY:
481        case OP_NOTMINQUERYI:
482      case OP_POSQUERY:      case OP_POSQUERY:
483        case OP_POSQUERYI:
484      case OP_NOTPOSQUERY:      case OP_NOTPOSQUERY:
485      cc += _pcre_OP_lengths[op];      case OP_NOTPOSQUERYI:
486    
487        cc += PRIV(OP_lengths)[op];
488  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
489      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];      if (utf8 && cc[-1] >= 0xc0) cc += PRIV(utf8_table4)[cc[-1] & 0x3f];
490  #endif  #endif
491      break;      break;
492    
# Line 420  for (;;) Line 496  for (;;)
496      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
497      case OP_SKIP_ARG:      case OP_SKIP_ARG:
498      case OP_THEN_ARG:      case OP_THEN_ARG:
499      cc += _pcre_OP_lengths[op] + cc[1];      cc += PRIV(OP_lengths)[op] + cc[1];
500      break;      break;
501    
502      /* 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. */  
503    
504      default:      case OP_CLOSE:
505      cc += _pcre_OP_lengths[op];      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;      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 */  /* Control never gets here */
# Line 456  Arguments: Line 542  Arguments:
542  Returns:        pointer after the character  Returns:        pointer after the character
543  */  */
544    
545  static const uschar *  static const pcre_uchar *
546  set_table_bit(uschar *start_bits, const uschar *p, BOOL caseless,  set_table_bit(pcre_uint8 *start_bits, const pcre_uchar *p, BOOL caseless,
547    compile_data *cd, BOOL utf8)    compile_data *cd, BOOL utf8)
548  {  {
549  unsigned int c = *p;  unsigned int c = *p;
# Line 471  if (utf8 && c > 127) Line 557  if (utf8 && c > 127)
557  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
558    if (caseless)    if (caseless)
559      {      {
560      uschar buff[8];      pcre_uint8 buff[8];
561      c = UCD_OTHERCASE(c);      c = UCD_OTHERCASE(c);
562      (void)_pcre_ord2utf8(c, buff);      (void)PRIV(ord2utf8)(c, buff);
563      SET_BIT(buff[0]);      SET_BIT(buff[0]);
564      }      }
565  #endif  #endif
# Line 490  return p + 1; Line 576  return p + 1;
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    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
641    set_nottype_bits(pcre_uint8 *start_bits, int cbit_type, int table_limit,
642      compile_data *cd)
643    {
644    register int c;
645    for (c = 0; c < table_limit; c++) start_bits[c] |= ~cd->cbits[c+cbit_type];
646    if (table_limit != 32) for (c = 24; c < 32; c++) start_bits[c] = 0xff;
647    }
648    
649    
650    
651    /*************************************************
652  *          Create bitmap of starting bytes       *  *          Create bitmap of starting bytes       *
653  *************************************************/  *************************************************/
654    
# Line 504  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;
# Line 540  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          case OP_XCLASS:
793        return SSB_FAIL;        return SSB_FAIL;
794    
795          /* We can ignore word boundary tests. */
796    
797          case OP_WORD_BOUNDARY:
798          case OP_NOT_WORD_BOUNDARY:
799          tcode++;
800          break;
801    
802        /* 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
803        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
804        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 562  do Line 808  do
808        case OP_SBRA:        case OP_SBRA:
809        case OP_CBRA:        case OP_CBRA:
810        case OP_SCBRA:        case OP_SCBRA:
811          case OP_BRAPOS:
812          case OP_SBRAPOS:
813          case OP_CBRAPOS:
814          case OP_SCBRAPOS:
815        case OP_ONCE:        case OP_ONCE:
816          case OP_ONCE_NC:
817        case OP_ASSERT:        case OP_ASSERT:
818        rc = set_start_bits(tcode, start_bits, caseless, utf8, cd);        rc = set_start_bits(tcode, start_bits, utf8, cd);
819        if (rc == SSB_FAIL) return SSB_FAIL;        if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
820        if (rc == SSB_DONE) try_next = FALSE; else        if (rc == SSB_DONE) try_next = FALSE; else
821          {          {
822          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
# Line 588  do Line 839  do
839        case OP_KET:        case OP_KET:
840        case OP_KETRMAX:        case OP_KETRMAX:
841        case OP_KETRMIN:        case OP_KETRMIN:
842          case OP_KETRPOS:
843        return SSB_CONTINUE;        return SSB_CONTINUE;
844    
845        /* Skip over callout */        /* Skip over callout */
# Line 605  do Line 857  do
857        tcode += 1 + LINK_SIZE;        tcode += 1 + LINK_SIZE;
858        break;        break;
859    
       /* Skip over an option setting, changing the caseless flag */  
   
       case OP_OPT:  
       caseless = (tcode[1] & PCRE_CASELESS) != 0;  
       tcode += 2;  
       break;  
   
860        /* BRAZERO does the bracket, but carries on. */        /* BRAZERO does the bracket, but carries on. */
861    
862        case OP_BRAZERO:        case OP_BRAZERO:
863        case OP_BRAMINZERO:        case OP_BRAMINZERO:
864        if (set_start_bits(++tcode, start_bits, caseless, utf8, cd) == SSB_FAIL)        case OP_BRAPOSZERO:
865          return SSB_FAIL;        rc = set_start_bits(++tcode, start_bits, utf8, cd);
866          if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
867  /* =========================================================================  /* =========================================================================
868        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,
869        which was an old fudge for the benefit of OS/2.        which was an old fudge for the benefit of OS/2.
# Line 643  do Line 889  do
889        case OP_QUERY:        case OP_QUERY:
890        case OP_MINQUERY:        case OP_MINQUERY:
891        case OP_POSQUERY:        case OP_POSQUERY:
892        tcode = set_table_bit(start_bits, tcode + 1, caseless, cd, utf8);        tcode = set_table_bit(start_bits, tcode + 1, FALSE, cd, utf8);
893          break;
894    
895          case OP_STARI:
896          case OP_MINSTARI:
897          case OP_POSSTARI:
898          case OP_QUERYI:
899          case OP_MINQUERYI:
900          case OP_POSQUERYI:
901          tcode = set_table_bit(start_bits, tcode + 1, TRUE, cd, utf8);
902        break;        break;
903    
904        /* Single-char upto sets the bit and tries the next */        /* Single-char upto sets the bit and tries the next */
# Line 651  do Line 906  do
906        case OP_UPTO:        case OP_UPTO:
907        case OP_MINUPTO:        case OP_MINUPTO:
908        case OP_POSUPTO:        case OP_POSUPTO:
909        tcode = set_table_bit(start_bits, tcode + 3, caseless, cd, utf8);        tcode = set_table_bit(start_bits, tcode + 1 + IMM2_SIZE, FALSE, cd, utf8);
910        break;        break;
911    
912        /* At least one single char sets the bit and stops */        case OP_UPTOI:
913          case OP_MINUPTOI:
914          case OP_POSUPTOI:
915          tcode = set_table_bit(start_bits, tcode + 1 + IMM2_SIZE, TRUE, cd, utf8);
916          break;
917    
918        case OP_EXACT:       /* Fall through */        /* At least one single char sets the bit and stops */
       tcode += 2;  
919    
920          case OP_EXACT:
921          tcode += IMM2_SIZE;
922          /* Fall through */
923        case OP_CHAR:        case OP_CHAR:
       case OP_CHARNC:  
924        case OP_PLUS:        case OP_PLUS:
925        case OP_MINPLUS:        case OP_MINPLUS:
926        case OP_POSPLUS:        case OP_POSPLUS:
927        (void)set_table_bit(start_bits, tcode + 1, caseless, cd, utf8);        (void)set_table_bit(start_bits, tcode + 1, FALSE, cd, utf8);
928          try_next = FALSE;
929          break;
930    
931          case OP_EXACTI:
932          tcode += IMM2_SIZE;
933          /* Fall through */
934          case OP_CHARI:
935          case OP_PLUSI:
936          case OP_MINPLUSI:
937          case OP_POSPLUSI:
938          (void)set_table_bit(start_bits, tcode + 1, TRUE, cd, utf8);
939        try_next = FALSE;        try_next = FALSE;
940        break;        break;
941    
# Line 679  do Line 950  do
950        SET_BIT(0x20);        SET_BIT(0x20);
951        if (utf8)        if (utf8)
952          {          {
953          SET_BIT(0xC2);  /* For U+00A0 */          SET_BIT(0xC2);  /* For U+00A0 */
954          SET_BIT(0xE1);  /* For U+1680, U+180E */          SET_BIT(0xE1);  /* For U+1680, U+180E */
955          SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */          SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
956          SET_BIT(0xE3);  /* For U+3000 */          SET_BIT(0xE3);  /* For U+3000 */
# Line 694  do Line 965  do
965        SET_BIT(0x0B);        SET_BIT(0x0B);
966        SET_BIT(0x0C);        SET_BIT(0x0C);
967        SET_BIT(0x0D);        SET_BIT(0x0D);
968        if (utf8)        if (utf8)
969          {          {
970          SET_BIT(0xC2);  /* For U+0085 */          SET_BIT(0xC2);  /* For U+0085 */
971          SET_BIT(0xE2);  /* For U+2028, U+2029 */          SET_BIT(0xE2);  /* For U+2028, U+2029 */
972          }          }
973        else SET_BIT(0x85);        else SET_BIT(0x85);
974        try_next = FALSE;        try_next = FALSE;
975        break;        break;
976    
977        /* Single character types set the bits and stop. Note that if PCRE_UCP        /* Single character types set the bits and stop. Note that if PCRE_UCP
978        is set, we do not see these op codes because \d etc are converted to        is set, we do not see these op codes because \d etc are converted to
979        properties. Therefore, these apply in the case when only ASCII characters        properties. Therefore, these apply in the case when only characters less
980        are recognized to match the types. In UTF-8 mode, we must restrict        than 256 are recognized to match the types. */
       ourselves to bytes less than 128, as otherwise there can be confusion  
       with bytes in the middle of UTF-8 characters. (In a "traditional"  
       environment, the tables will only recognize ASCII characters anyway, but  
       in at least one Windows environment, some higher bytes bits were set in  
       the tables.) */  
981    
982        case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
983        for (c = 0; c < table_limit; c++)        set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_digit];  
984        try_next = FALSE;        try_next = FALSE;
985        break;        break;
986    
987        case OP_DIGIT:        case OP_DIGIT:
988        for (c = 0; c < table_limit; c++)        set_type_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_digit];  
989        try_next = FALSE;        try_next = FALSE;
990        break;        break;
991    
992        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to
993        discard it. */        ensure it is set as not whitespace. */
994    
995        case OP_NOT_WHITESPACE:        case OP_NOT_WHITESPACE:
996        for (c = 0; c < table_limit; c++)        set_nottype_bits(start_bits, cbit_space, table_limit, cd);
997          {        start_bits[1] |= 0x08;
         int d = cd->cbits[c+cbit_space];  
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= ~d;  
         }  
998        try_next = FALSE;        try_next = FALSE;
999        break;        break;
1000    
1001        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to
1002        discard it. */        not set it from the table. */
1003    
1004        case OP_WHITESPACE:        case OP_WHITESPACE:
1005        for (c = 0; c < table_limit; c++)        c = start_bits[1];    /* Save in case it was already set */
1006          {        set_type_bits(start_bits, cbit_space, table_limit, cd);
1007          int d = cd->cbits[c+cbit_space];        start_bits[1] = (start_bits[1] & ~0x08) | c;
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= d;  
         }  
1008        try_next = FALSE;        try_next = FALSE;
1009        break;        break;
1010    
1011        case OP_NOT_WORDCHAR:        case OP_NOT_WORDCHAR:
1012        for (c = 0; c < table_limit; c++)        set_nottype_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_word];  
1013        try_next = FALSE;        try_next = FALSE;
1014        break;        break;
1015    
1016        case OP_WORDCHAR:        case OP_WORDCHAR:
1017        for (c = 0; c < table_limit; c++)        set_type_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_word];  
1018        try_next = FALSE;        try_next = FALSE;
1019        break;        break;
1020    
# Line 773  do Line 1028  do
1028        break;        break;
1029    
1030        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1031        tcode += 3;        tcode += 1 + IMM2_SIZE;
1032        break;        break;
1033    
1034        /* 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 782  do Line 1037  do
1037        case OP_TYPEUPTO:        case OP_TYPEUPTO:
1038        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
1039        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1040        tcode += 2;               /* Fall through */        tcode += IMM2_SIZE;  /* Fall through */
1041    
1042        case OP_TYPESTAR:        case OP_TYPESTAR:
1043        case OP_TYPEMINSTAR:        case OP_TYPEMINSTAR:
# Line 802  do Line 1057  do
1057          SET_BIT(0x20);          SET_BIT(0x20);
1058          if (utf8)          if (utf8)
1059            {            {
1060            SET_BIT(0xC2);  /* For U+00A0 */            SET_BIT(0xC2);  /* For U+00A0 */
1061            SET_BIT(0xE1);  /* For U+1680, U+180E */            SET_BIT(0xE1);  /* For U+1680, U+180E */
1062            SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */            SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
1063            SET_BIT(0xE3);  /* For U+3000 */            SET_BIT(0xE3);  /* For U+3000 */
# Line 816  do Line 1071  do
1071          SET_BIT(0x0B);          SET_BIT(0x0B);
1072          SET_BIT(0x0C);          SET_BIT(0x0C);
1073          SET_BIT(0x0D);          SET_BIT(0x0D);
1074          if (utf8)          if (utf8)
1075            {            {
1076            SET_BIT(0xC2);  /* For U+0085 */            SET_BIT(0xC2);  /* For U+0085 */
1077            SET_BIT(0xE2);  /* For U+2028, U+2029 */            SET_BIT(0xE2);  /* For U+2028, U+2029 */
1078            }            }
1079          else SET_BIT(0x85);          else SET_BIT(0x85);
1080          break;          break;
1081    
1082          case OP_NOT_DIGIT:          case OP_NOT_DIGIT:
1083          for (c = 0; c < table_limit; c++)          set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_digit];  
1084          break;          break;
1085    
1086          case OP_DIGIT:          case OP_DIGIT:
1087          for (c = 0; c < table_limit; c++)          set_type_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_digit];  
1088          break;          break;
1089    
1090          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we have to
1091          discard it. */          ensure it gets set as not whitespace. */
1092    
1093          case OP_NOT_WHITESPACE:          case OP_NOT_WHITESPACE:
1094          for (c = 0; c < table_limit; c++)          set_nottype_bits(start_bits, cbit_space, table_limit, cd);
1095            {          start_bits[1] |= 0x08;
           int d = cd->cbits[c+cbit_space];  
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= ~d;  
           }  
1096          break;          break;
1097    
1098          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we have to
1099          discard it. */          avoid setting it. */
1100    
1101          case OP_WHITESPACE:          case OP_WHITESPACE:
1102          for (c = 0; c < table_limit; c++)          c = start_bits[1];    /* Save in case it was already set */
1103            {          set_type_bits(start_bits, cbit_space, table_limit, cd);
1104            int d = cd->cbits[c+cbit_space];          start_bits[1] = (start_bits[1] & ~0x08) | c;
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= d;  
           }  
1105          break;          break;
1106    
1107          case OP_NOT_WORDCHAR:          case OP_NOT_WORDCHAR:
1108          for (c = 0; c < table_limit; c++)          set_nottype_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_word];  
1109          break;          break;
1110    
1111          case OP_WORDCHAR:          case OP_WORDCHAR:
1112          for (c = 0; c < table_limit; c++)          set_type_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_word];  
1113          break;          break;
1114          }          }
1115    
# Line 921  do Line 1165  do
1165            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];
1166            }            }
1167    
1168          /* Advance past the bit map, and act on what follows */          /* Advance past the bit map, and act on what follows. For a zero
1169            minimum repeat, continue; otherwise stop processing. */
1170    
1171          tcode += 32;          tcode += 32;
1172          switch (*tcode)          switch (*tcode)
# Line 935  do Line 1180  do
1180    
1181            case OP_CRRANGE:            case OP_CRRANGE:
1182            case OP_CRMINRANGE:            case OP_CRMINRANGE:
1183            if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;            if (GET2(tcode, 1) == 0) tcode += 1 + 2 * IMM2_SIZE;
1184              else try_next = FALSE;              else try_next = FALSE;
1185            break;            break;
1186    
# Line 957  return yield; Line 1202  return yield;
1202    
1203    
1204    
1205    
1206    
1207  /*************************************************  /*************************************************
1208  *          Study a compiled expression           *  *          Study a compiled expression           *
1209  *************************************************/  *************************************************/
# Line 976  Returns: pointer to a pcre_extra bloc Line 1223  Returns: pointer to a pcre_extra bloc
1223              NULL on error or if no optimization possible              NULL on error or if no optimization possible
1224  */  */
1225    
1226    #ifdef COMPILE_PCRE8
1227  PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION  PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION
1228  pcre_study(const pcre *external_re, int options, const char **errorptr)  pcre_study(const pcre *external_re, int options, const char **errorptr)
1229    #else
1230    PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION
1231    pcre16_study(const pcre *external_re, int options, const char **errorptr)
1232    #endif
1233  {  {
1234  int min;  int min;
1235  BOOL bits_set = FALSE;  BOOL bits_set = FALSE;
1236  uschar start_bits[32];  pcre_uint8 start_bits[32];
1237  pcre_extra *extra;  pcre_extra *extra = NULL;
1238  pcre_study_data *study;  pcre_study_data *study;
1239  const uschar *tables;  const pcre_uint8 *tables;
1240  uschar *code;  pcre_uchar *code;
1241  compile_data compile_block;  compile_data compile_block;
1242  const real_pcre *re = (const real_pcre *)external_re;  const real_pcre *re = (const real_pcre *)external_re;
1243    
# Line 1003  if ((options & ~PUBLIC_STUDY_OPTIONS) != Line 1255  if ((options & ~PUBLIC_STUDY_OPTIONS) !=
1255    return NULL;    return NULL;
1256    }    }
1257    
1258  code = (uschar *)re + re->name_table_offset +  code = (pcre_uchar *)re + re->name_table_offset +
1259    (re->name_count * re->name_entry_size);    (re->name_count * re->name_entry_size);
1260    
1261  /* 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 1013  seeking a list of starting bytes. */ Line 1265  seeking a list of starting bytes. */
1265  if ((re->options & PCRE_ANCHORED) == 0 &&  if ((re->options & PCRE_ANCHORED) == 0 &&
1266      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)
1267    {    {
1268      int rc;
1269    
1270    /* Set the character tables in the block that is passed around */    /* Set the character tables in the block that is passed around */
1271    
1272    tables = re->tables;    tables = re->tables;
# Line 1027  if ((re->options & PCRE_ANCHORED) == 0 & Line 1281  if ((re->options & PCRE_ANCHORED) == 0 &
1281    
1282    /* 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. */
1283    
1284    memset(start_bits, 0, 32 * sizeof(uschar));    memset(start_bits, 0, 32 * sizeof(pcre_uint8));
1285    bits_set = set_start_bits(code, start_bits,    rc = set_start_bits(code, start_bits, (re->options & PCRE_UTF8) != 0,
1286      (re->options & PCRE_CASELESS) != 0, (re->options & PCRE_UTF8) != 0,      &compile_block);
1287      &compile_block) == SSB_DONE;    bits_set = rc == SSB_DONE;
1288      if (rc == SSB_UNKNOWN)
1289        {
1290        *errorptr = "internal error: opcode not recognized";
1291        return NULL;
1292        }
1293    }    }
1294    
1295  /* Find the minimum length of subject string. */  /* Find the minimum length of subject string. */
1296    
1297  min = find_minlength(code, code, re->options);  switch(min = find_minlength(code, code, re->options, 0))
1298      {
1299      case -2: *errorptr = "internal error: missing capturing bracket"; return NULL;
1300      case -3: *errorptr = "internal error: opcode not recognized"; return NULL;
1301      default: break;
1302      }
1303    
1304    /* If a set of starting bytes has been identified, or if the minimum length is
1305    greater than zero, or if JIT optimization has been requested, get a pcre_extra
1306    block and a pcre_study_data block. The study data is put in the latter, which
1307    is pointed to by the former, which may also get additional data set later by
1308    the calling program. At the moment, the size of pcre_study_data is fixed. We
1309    nevertheless save it in a field for returning via the pcre_fullinfo() function
1310    so that if it becomes variable in the future, we don't have to change that
1311    code. */
1312    
1313    if (bits_set || min > 0
1314    #ifdef SUPPORT_JIT
1315        || (options & PCRE_STUDY_JIT_COMPILE) != 0
1316    #endif
1317      )
1318      {
1319      extra = (pcre_extra *)(pcre_malloc)
1320        (sizeof(pcre_extra) + sizeof(pcre_study_data));
1321      if (extra == NULL)
1322        {
1323        *errorptr = "failed to get memory";
1324        return NULL;
1325        }
1326    
1327      study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));
1328      extra->flags = PCRE_EXTRA_STUDY_DATA;
1329      extra->study_data = study;
1330    
1331  /* Return NULL if no optimization is possible. */    study->size = sizeof(pcre_study_data);
1332      study->flags = 0;
1333    
1334  if (!bits_set && min < 0) return NULL;    if (bits_set)
1335        {
1336        study->flags |= PCRE_STUDY_MAPPED;
1337        memcpy(study->start_bits, start_bits, sizeof(start_bits));
1338        }
1339    
1340  /* Get a pcre_extra block and a pcre_study_data block. The study data is put in    /* Always set the minlength value in the block, because the JIT compiler
1341  the latter, which is pointed to by the former, which may also get additional    makes use of it. However, don't set the bit unless the length is greater than
1342  data set later by the calling program. At the moment, the size of    zero - the interpretive pcre_exec() and pcre_dfa_exec() needn't waste time
1343  pcre_study_data is fixed. We nevertheless save it in a field for returning via    checking the zero case. */
 the pcre_fullinfo() function so that if it becomes variable in the future, we  
 don't have to change that code. */  
1344    
1345  extra = (pcre_extra *)(pcre_malloc)    if (min > 0)
1346    (sizeof(pcre_extra) + sizeof(pcre_study_data));      {
1347        study->flags |= PCRE_STUDY_MINLEN;
1348        study->minlength = min;
1349        }
1350      else study->minlength = 0;
1351    
1352  if (extra == NULL)    /* If JIT support was compiled and requested, attempt the JIT compilation.
1353    {    If no starting bytes were found, and the minimum length is zero, and JIT
1354    *errorptr = "failed to get memory";    compilation fails, abandon the extra block and return NULL. */
1355    return NULL;  
1356    #ifdef SUPPORT_JIT
1357      extra->executable_jit = NULL;
1358      if ((options & PCRE_STUDY_JIT_COMPILE) != 0) PRIV(jit_compile)(re, extra);
1359      if (study->flags == 0 && (extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) == 0)
1360        {
1361        pcre_free_study(extra);
1362        extra = NULL;
1363        }
1364    #endif
1365    }    }
1366    
1367  study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));  return extra;
1368  extra->flags = PCRE_EXTRA_STUDY_DATA;  }
 extra->study_data = study;  
1369    
 study->size = sizeof(pcre_study_data);  
 study->flags = 0;  
1370    
1371  if (bits_set)  /*************************************************
1372    {  *          Free the study data                   *
1373    study->flags |= PCRE_STUDY_MAPPED;  *************************************************/
   memcpy(study->start_bits, start_bits, sizeof(start_bits));  
   }  
1374    
1375  if (min >= 0)  /* This function frees the memory that was obtained by pcre_study().
   {  
   study->flags |= PCRE_STUDY_MINLEN;  
   study->minlength = min;  
   }  
1376    
1377  return extra;  Argument:   a pointer to the pcre_extra block
1378    Returns:    nothing
1379    */
1380    
1381    #ifdef COMPILE_PCRE8
1382    PCRE_EXP_DEFN void
1383    pcre_free_study(pcre_extra *extra)
1384    #else
1385    PCRE_EXP_DEFN void
1386    pcre16_free_study(pcre_extra *extra)
1387    #endif
1388    {
1389    #ifdef SUPPORT_JIT
1390    if ((extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 &&
1391         extra->executable_jit != NULL)
1392      PRIV(jit_free)(extra->executable_jit);
1393    #endif
1394    pcre_free(extra);
1395  }  }
1396    
1397  /* End of pcre_study.c */  /* End of pcre_study.c */
Legend:
Removed from v.538  
changed lines
  Added in v.769
webmaster@exim.org
 ViewVC Help
Powered by ViewVC 1.1.12