/[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 455 by ph10, Sat Sep 26 19:12:32 2009 UTC revision 914 by zherczeg, Mon Feb 13 06:04:50 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) 1997-2009 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 60  enum { SSB_FAIL, SSB_DONE, SSB_CONTINUE Line 61  enum { SSB_FAIL, SSB_DONE, SSB_CONTINUE
61  *************************************************/  *************************************************/
62    
63  /* Scan a parenthesized group and compute the minimum length of subject that  /* Scan a parenthesized group and compute the minimum length of subject that
64  is needed to match it. This is a lower bound; it does not mean there is a  is needed to match it. This is a lower bound; it does not mean there is a
65  string of that length that matches. In UTF8 mode, the result is in characters  string of that length that matches. In UTF8 mode, the result is in characters
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 int op = *cc;
102    
103    switch (op)    switch (op)
104      {      {
105        case OP_COND:
106        case OP_SCOND:
107    
108        /* If there is only one branch in a condition, the implied branch has zero
109        length, so we don't add anything. This covers the DEFINE "condition"
110        automatically. */
111    
112        cs = cc + GET(cc, 1);
113        if (*cs != OP_ALT)
114          {
115          cc = cs + 1 + LINK_SIZE;
116          break;
117          }
118    
119        /* Otherwise we can fall through and treat it the same as any other
120        subpattern. */
121    
122      case OP_CBRA:      case OP_CBRA:
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      case OP_COND:      case OP_ONCE_NC:
132      case OP_SCOND:      d = find_minlength(cc, startcode, options, recurse_depth);
     d = find_minlength(cc, startcode, options);  
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;
163      break;      break;
164    
165      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
# Line 140  for (;;) Line 175  for (;;)
175    
176      case OP_REVERSE:      case OP_REVERSE:
177      case OP_CREF:      case OP_CREF:
178        case OP_NCREF:
179      case OP_RREF:      case OP_RREF:
180        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 168  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  #ifdef SUPPORT_UTF
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    
232      case OP_TYPEPLUS:      case OP_TYPEPLUS:
233      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
234      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
235      branchlength++;      branchlength++;
236      cc += (cc[1] == OP_PROP || cc[1] == OP_NOTPROP)? 4 : 2;      cc += (cc[1] == OP_PROP || cc[1] == OP_NOTPROP)? 4 : 2;
237      break;      break;
# Line 194  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_NOTEXACT:      case OP_EXACTI:
244        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  #ifdef SUPPORT_UTF
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 223  for (;;) Line 272  for (;;)
272      case OP_ANY:      case OP_ANY:
273      case OP_ALLANY:      case OP_ALLANY:
274      case OP_EXTUNI:      case OP_EXTUNI:
275      case OP_HSPACE:      case OP_HSPACE:
276      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
277      case OP_VSPACE:      case OP_VSPACE:
278      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
279      branchlength++;      branchlength++;
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++;
301      break;      break;
302    
303      /* For repeated character types, we have to test for \p and \P, which have      /* For repeated character types, we have to test for \p and \P, which have
304      an extra two bytes of parameters. */      an extra two bytes of parameters. */
# Line 258  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        {        {
338        case OP_CRPLUS:        case OP_CRPLUS:
339        case OP_CRMINPLUS:        case OP_CRMINPLUS:
340        branchlength++;        branchlength++;
341        /* Fall through */        /* Fall through */
342    
343        case OP_CRSTAR:        case OP_CRSTAR:
344        case OP_CRMINSTAR:        case OP_CRMINSTAR:
345        case OP_CRQUERY:        case OP_CRQUERY:
346        case OP_CRMINQUERY:        case OP_CRMINQUERY:
347        cc++;        cc++;
348        break;        break;
349    
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:
357        branchlength++;        branchlength++;
358        break;        break;
359        }        }
360      break;      break;
361    
362      /* Backreferences and subroutine calls are treated in the same way: we find      /* Backreferences and subroutine calls are treated in the same way: we find
363      the minimum length for the subpattern. A recursion, however, causes an      the minimum length for the subpattern. A recursion, however, causes an
364      a flag to be set that causes the length of this branch to be ignored. The      a flag to be set that causes the length of this branch to be ignored. The
365      logic is that a recursion can only make sense if there is another      logic is that a recursion can only make sense if there is another
366      alternation that stops the recursing. That will provide the minimum length      alternation that stops the recursing. That will provide the minimum length
367      (when no recursion happens). A backreference within the group that it is      (when no recursion happens). A backreference within the group that it is
368      referencing behaves in the same way. */      referencing behaves in the same way.
369    
370        If PCRE_JAVASCRIPT_COMPAT is set, a backreference to an unset bracket
371        matches an empty string (by default it causes a matching failure), so in
372        that case we must set the minimum length to zero. */
373    
374      case OP_REF:      case OP_REF:
375      ce = cs = (uschar *)_pcre_find_bracket(startcode, utf8, GET2(cc, 1));      case OP_REFI:
376      if (cs == NULL) return -2;      if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
     do ce += GET(ce, 1); while (*ce == OP_ALT);  
     if (cc > cs && cc < ce)  
377        {        {
378        d = 0;        ce = cs = (pcre_uchar *)PRIV(find_bracket)(startcode, utf, GET2(cc, 1));
379        had_recurse = TRUE;        if (cs == NULL) return -2;
380        }        do ce += GET(ce, 1); while (*ce == OP_ALT);
381      else d = find_minlength(cs, startcode, options);        if (cc > cs && cc < ce)
382      cc += 3;          {
383            d = 0;
384            had_recurse = TRUE;
385            }
386          else
387            {
388            d = find_minlength(cs, startcode, options, recurse_depth);
389            }
390          }
391        else d = 0;
392        cc += 1 + IMM2_SIZE;
393    
394      /* Handle repeated back references */      /* Handle repeated back references */
395    
396      switch (*cc)      switch (*cc)
397        {        {
398        case OP_CRSTAR:        case OP_CRSTAR:
# Line 337  for (;;) Line 402  for (;;)
402        min = 0;        min = 0;
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:
419        min = 1;        min = 1;
420        break;        break;
421        }        }
422    
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_NOTUPTO:      case OP_UPTOI:
451        case OP_NOTUPTO:
452        case OP_NOTUPTOI:
453      case OP_MINUPTO:      case OP_MINUPTO:
454      case OP_NOTMINUPTO:      case OP_MINUPTOI:
455        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_NOTMINSTAR:      case OP_MINSTARI:
468        case OP_NOTMINSTAR:
469        case OP_NOTMINSTARI:
470      case OP_POSSTAR:      case OP_POSSTAR:
471      case OP_NOTPOSSTAR:      case OP_POSSTARI:
472        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_NOTPOSQUERY:      case OP_POSQUERYI:
485      cc += _pcre_OP_lengths[op];      case OP_NOTPOSQUERY:
486  #ifdef SUPPORT_UTF8      case OP_NOTPOSQUERYI:
487      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
488  #endif      cc += PRIV(OP_lengths)[op];
489    #ifdef SUPPORT_UTF
490        if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
491    #endif
492      break;      break;
493    
494      /* For the record, these are the opcodes that are matched by "default":      /* Skip these, but we need to add in the name length. */
495      OP_ACCEPT, OP_CLOSE, OP_COMMIT, OP_FAIL, OP_PRUNE, OP_SET_SOM, OP_SKIP,  
496      OP_THEN. */      case OP_MARK:
497        case OP_PRUNE_ARG:
498      default:      case OP_SKIP_ARG:
499      cc += _pcre_OP_lengths[op];      case OP_THEN_ARG:
500        cc += PRIV(OP_lengths)[op] + cc[1];
501      break;      break;
502    
503        /* The remaining opcodes are just skipped over. */
504    
505        case OP_CLOSE:
506        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;
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 406  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 mode
542    
543  Returns:        nothing  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
568      return p;
569      }
570    #endif
571    
572    /* Not UTF-8 mode, or character is less than 127. */
573    
574    if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
575    return p + 1;
576    #endif
577    
578    #ifdef COMPILE_PCRE16
579    if (c > 0xff)
580      {
581      c = 0xff;
582      caseless = FALSE;
583      }
584    SET_BIT(c);
585    
586    #ifdef SUPPORT_UTF
587    if (utf && c > 127)
588      {
589      GETCHARINC(c, p);
590    #ifdef SUPPORT_UCP
591      if (caseless)
592        {
593        c = UCD_OTHERCASE(c);
594        if (c > 0xff)
595          c = 0xff;
596        SET_BIT(c);
597        }
598    #endif
599      return p;
600      }
601    #endif
602    
603    if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
604    return p + 1;
605    #endif
606    }
607    
608    
609    
610    /*************************************************
611    *     Set bits for a positive character type     *
612    *************************************************/
613    
614    /* This function sets starting bits for a character type. In UTF-8 mode, we can
615    only do a direct setting for bytes less than 128, as otherwise there can be
616    confusion with bytes in the middle of UTF-8 characters. In a "traditional"
617    environment, the tables will only recognize ASCII characters anyway, but in at
618    least one Windows environment, some higher bytes bits were set in the tables.
619    So we deal with that case by considering the UTF-8 encoding.
620    
621    Arguments:
622      start_bits     the starting bitmap
623      cbit type      the type of character wanted
624      table_limit    32 for non-UTF-8; 16 for UTF-8
625      cd             the block with char table pointers
626    
627    Returns:         nothing
628  */  */
629    
630  static void  static void
631  set_bit(uschar *start_bits, unsigned int c, BOOL caseless, compile_data *cd)  set_type_bits(pcre_uint8 *start_bits, int cbit_type, int table_limit,
632      compile_data *cd)
633  {  {
634  start_bits[c/8] |= (1 << (c&7));  register int c;
635  if (caseless && (cd->ctypes[c] & ctype_letter) != 0)  for (c = 0; c < table_limit; c++) start_bits[c] |= cd->cbits[c+cbit_type];
636    start_bits[cd->fcc[c]/8] |= (1 << (cd->fcc[c]&7));  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
637    if (table_limit == 32) return;
638    for (c = 128; c < 256; c++)
639      {
640      if ((cd->cbits[c/8] & (1 << (c&7))) != 0)
641        {
642        pcre_uchar buff[6];
643        (void)PRIV(ord2utf)(c, buff);
644        SET_BIT(buff[0]);
645        }
646      }
647    #endif
648    }
649    
650    
651    /*************************************************
652    *     Set bits for a negative character type     *
653    *************************************************/
654    
655    /* This function sets starting bits for a negative character type such as \D.
656    In UTF-8 mode, we can only do a direct setting for bytes less than 128, as
657    otherwise there can be confusion with bytes in the middle of UTF-8 characters.
658    Unlike in the positive case, where we can set appropriate starting bits for
659    specific high-valued UTF-8 characters, in this case we have to set the bits for
660    all high-valued characters. The lowest is 0xc2, but we overkill by starting at
661    0xc0 (192) for simplicity.
662    
663    Arguments:
664      start_bits     the starting bitmap
665      cbit type      the type of character wanted
666      table_limit    32 for non-UTF-8; 16 for UTF-8
667      cd             the block with char table pointers
668    
669    Returns:         nothing
670    */
671    
672    static void
673    set_nottype_bits(pcre_uint8 *start_bits, int cbit_type, int table_limit,
674      compile_data *cd)
675    {
676    register int c;
677    for (c = 0; c < table_limit; c++) start_bits[c] |= ~cd->cbits[c+cbit_type];
678    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
679    if (table_limit != 32) for (c = 24; c < 32; c++) start_bits[c] = 0xff;
680    #endif
681  }  }
682    
683    
# Line 443  function fails unless the result is SSB_ Line 697  function fails unless the result is SSB_
697  Arguments:  Arguments:
698    code         points to an expression    code         points to an expression
699    start_bits   points to a 32-byte table, initialized to 0    start_bits   points to a 32-byte table, initialized to 0
700    caseless     the current state of the caseless flag    utf          TRUE if in UTF-8 / UTF-16 mode
   utf8         TRUE if in UTF-8 mode  
701    cd           the block with char table pointers    cd           the block with char table pointers
702    
703  Returns:       SSB_FAIL     => Failed to find any starting bytes  Returns:       SSB_FAIL     => Failed to find any starting bytes
704                 SSB_DONE     => Found mandatory starting bytes                 SSB_DONE     => Found mandatory starting bytes
705                 SSB_CONTINUE => Found optional starting bytes                 SSB_CONTINUE => Found optional starting bytes
706                   SSB_UNKNOWN  => Hit an unrecognized opcode
707  */  */
708    
709  static int  static int
710  set_start_bits(const uschar *code, uschar *start_bits, BOOL caseless,  set_start_bits(const pcre_uchar *code, pcre_uint8 *start_bits, BOOL utf,
711    BOOL utf8, compile_data *cd)    compile_data *cd)
712  {  {
713  register int c;  register int c;
714  int yield = SSB_DONE;  int yield = SSB_DONE;
715    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
716    int table_limit = utf? 16:32;
717    #else
718    int table_limit = 32;
719    #endif
720    
721  #if 0  #if 0
722  /* ========================================================================= */  /* ========================================================================= */
# Line 478  volatile int dummy; Line 737  volatile int dummy;
737    
738  do  do
739    {    {
   const uschar *tcode = code + (((int)*code == OP_CBRA)? 3:1) + LINK_SIZE;  
740    BOOL try_next = TRUE;    BOOL try_next = TRUE;
741      const pcre_uchar *tcode = code + 1 + LINK_SIZE;
742    
743      if (*code == OP_CBRA || *code == OP_SCBRA ||
744          *code == OP_CBRAPOS || *code == OP_SCBRAPOS) tcode += IMM2_SIZE;
745    
746    while (try_next)    /* Loop for items in this branch */    while (try_next)    /* Loop for items in this branch */
747      {      {
748      int rc;      int rc;
749    
750      switch(*tcode)      switch(*tcode)
751        {        {
752        /* Fail if we reach something we don't understand */        /* If we reach something we don't understand, it means a new opcode has
753          been created that hasn't been added to this code. Hopefully this problem
754          will be discovered during testing. */
755    
756        default:        default:
757          return SSB_UNKNOWN;
758    
759          /* Fail for a valid opcode that implies no starting bits. */
760    
761          case OP_ACCEPT:
762          case OP_ASSERT_ACCEPT:
763          case OP_ALLANY:
764          case OP_ANY:
765          case OP_ANYBYTE:
766          case OP_CIRC:
767          case OP_CIRCM:
768          case OP_CLOSE:
769          case OP_COMMIT:
770          case OP_COND:
771          case OP_CREF:
772          case OP_DEF:
773          case OP_DOLL:
774          case OP_DOLLM:
775          case OP_END:
776          case OP_EOD:
777          case OP_EODN:
778          case OP_EXTUNI:
779          case OP_FAIL:
780          case OP_MARK:
781          case OP_NCREF:
782          case OP_NOT:
783          case OP_NOTEXACT:
784          case OP_NOTEXACTI:
785          case OP_NOTI:
786          case OP_NOTMINPLUS:
787          case OP_NOTMINPLUSI:
788          case OP_NOTMINQUERY:
789          case OP_NOTMINQUERYI:
790          case OP_NOTMINSTAR:
791          case OP_NOTMINSTARI:
792          case OP_NOTMINUPTO:
793          case OP_NOTMINUPTOI:
794          case OP_NOTPLUS:
795          case OP_NOTPLUSI:
796          case OP_NOTPOSPLUS:
797          case OP_NOTPOSPLUSI:
798          case OP_NOTPOSQUERY:
799          case OP_NOTPOSQUERYI:
800          case OP_NOTPOSSTAR:
801          case OP_NOTPOSSTARI:
802          case OP_NOTPOSUPTO:
803          case OP_NOTPOSUPTOI:
804          case OP_NOTPROP:
805          case OP_NOTQUERY:
806          case OP_NOTQUERYI:
807          case OP_NOTSTAR:
808          case OP_NOTSTARI:
809          case OP_NOTUPTO:
810          case OP_NOTUPTOI:
811          case OP_NOT_HSPACE:
812          case OP_NOT_VSPACE:
813          case OP_NRREF:
814          case OP_PROP:
815          case OP_PRUNE:
816          case OP_PRUNE_ARG:
817          case OP_RECURSE:
818          case OP_REF:
819          case OP_REFI:
820          case OP_REVERSE:
821          case OP_RREF:
822          case OP_SCOND:
823          case OP_SET_SOM:
824          case OP_SKIP:
825          case OP_SKIP_ARG:
826          case OP_SOD:
827          case OP_SOM:
828          case OP_THEN:
829          case OP_THEN_ARG:
830    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
831          case OP_XCLASS:
832    #endif
833        return SSB_FAIL;        return SSB_FAIL;
834    
835          /* We can ignore word boundary tests. */
836    
837          case OP_WORD_BOUNDARY:
838          case OP_NOT_WORD_BOUNDARY:
839          tcode++;
840          break;
841    
842        /* 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
843        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
844        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 500  do Line 848  do
848        case OP_SBRA:        case OP_SBRA:
849        case OP_CBRA:        case OP_CBRA:
850        case OP_SCBRA:        case OP_SCBRA:
851          case OP_BRAPOS:
852          case OP_SBRAPOS:
853          case OP_CBRAPOS:
854          case OP_SCBRAPOS:
855        case OP_ONCE:        case OP_ONCE:
856          case OP_ONCE_NC:
857        case OP_ASSERT:        case OP_ASSERT:
858        rc = set_start_bits(tcode, start_bits, caseless, utf8, cd);        rc = set_start_bits(tcode, start_bits, utf, cd);
859        if (rc == SSB_FAIL) return SSB_FAIL;        if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
860        if (rc == SSB_DONE) try_next = FALSE; else        if (rc == SSB_DONE) try_next = FALSE; else
861          {          {
862          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
# Line 526  do Line 879  do
879        case OP_KET:        case OP_KET:
880        case OP_KETRMAX:        case OP_KETRMAX:
881        case OP_KETRMIN:        case OP_KETRMIN:
882          case OP_KETRPOS:
883        return SSB_CONTINUE;        return SSB_CONTINUE;
884    
885        /* Skip over callout */        /* Skip over callout */
# Line 543  do Line 897  do
897        tcode += 1 + LINK_SIZE;        tcode += 1 + LINK_SIZE;
898        break;        break;
899    
       /* Skip over an option setting, changing the caseless flag */  
   
       case OP_OPT:  
       caseless = (tcode[1] & PCRE_CASELESS) != 0;  
       tcode += 2;  
       break;  
   
900        /* BRAZERO does the bracket, but carries on. */        /* BRAZERO does the bracket, but carries on. */
901    
902        case OP_BRAZERO:        case OP_BRAZERO:
903        case OP_BRAMINZERO:        case OP_BRAMINZERO:
904        if (set_start_bits(++tcode, start_bits, caseless, utf8, cd) == SSB_FAIL)        case OP_BRAPOSZERO:
905          return SSB_FAIL;        rc = set_start_bits(++tcode, start_bits, utf, cd);
906          if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
907  /* =========================================================================  /* =========================================================================
908        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,
909        which was an old fudge for the benefit of OS/2.        which was an old fudge for the benefit of OS/2.
# Line 581  do Line 929  do
929        case OP_QUERY:        case OP_QUERY:
930        case OP_MINQUERY:        case OP_MINQUERY:
931        case OP_POSQUERY:        case OP_POSQUERY:
932        set_bit(start_bits, tcode[1], caseless, cd);        tcode = set_table_bit(start_bits, tcode + 1, FALSE, cd, utf);
933        tcode += 2;        break;
934  #ifdef SUPPORT_UTF8  
935        if (utf8 && tcode[-1] >= 0xc0)        case OP_STARI:
936          tcode += _pcre_utf8_table4[tcode[-1] & 0x3f];        case OP_MINSTARI:
937  #endif        case OP_POSSTARI:
938          case OP_QUERYI:
939          case OP_MINQUERYI:
940          case OP_POSQUERYI:
941          tcode = set_table_bit(start_bits, tcode + 1, TRUE, cd, utf);
942        break;        break;
943    
944        /* Single-char upto sets the bit and tries the next */        /* Single-char upto sets the bit and tries the next */
# Line 594  do Line 946  do
946        case OP_UPTO:        case OP_UPTO:
947        case OP_MINUPTO:        case OP_MINUPTO:
948        case OP_POSUPTO:        case OP_POSUPTO:
949        set_bit(start_bits, tcode[3], caseless, cd);        tcode = set_table_bit(start_bits, tcode + 1 + IMM2_SIZE, FALSE, cd, utf);
       tcode += 4;  
 #ifdef SUPPORT_UTF8  
       if (utf8 && tcode[-1] >= 0xc0)  
         tcode += _pcre_utf8_table4[tcode[-1] & 0x3f];  
 #endif  
950        break;        break;
951    
952        /* At least one single char sets the bit and stops */        case OP_UPTOI:
953          case OP_MINUPTOI:
954          case OP_POSUPTOI:
955          tcode = set_table_bit(start_bits, tcode + 1 + IMM2_SIZE, TRUE, cd, utf);
956          break;
957    
958        case OP_EXACT:       /* Fall through */        /* At least one single char sets the bit and stops */
       tcode += 2;  
959    
960          case OP_EXACT:
961          tcode += IMM2_SIZE;
962          /* Fall through */
963        case OP_CHAR:        case OP_CHAR:
       case OP_CHARNC:  
964        case OP_PLUS:        case OP_PLUS:
965        case OP_MINPLUS:        case OP_MINPLUS:
966        case OP_POSPLUS:        case OP_POSPLUS:
967        set_bit(start_bits, tcode[1], caseless, cd);        (void)set_table_bit(start_bits, tcode + 1, FALSE, cd, utf);
968          try_next = FALSE;
969          break;
970    
971          case OP_EXACTI:
972          tcode += IMM2_SIZE;
973          /* Fall through */
974          case OP_CHARI:
975          case OP_PLUSI:
976          case OP_MINPLUSI:
977          case OP_POSPLUSI:
978          (void)set_table_bit(start_bits, tcode + 1, TRUE, cd, utf);
979          try_next = FALSE;
980          break;
981    
982          /* Special spacing and line-terminating items. These recognize specific
983          lists of characters. The difference between VSPACE and ANYNL is that the
984          latter can match the two-character CRLF sequence, but that is not
985          relevant for finding the first character, so their code here is
986          identical. */
987    
988          case OP_HSPACE:
989          SET_BIT(0x09);
990          SET_BIT(0x20);
991    #ifdef SUPPORT_UTF
992          if (utf)
993            {
994    #ifdef COMPILE_PCRE8
995            SET_BIT(0xC2);  /* For U+00A0 */
996            SET_BIT(0xE1);  /* For U+1680, U+180E */
997            SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
998            SET_BIT(0xE3);  /* For U+3000 */
999    #endif
1000    #ifdef COMPILE_PCRE16
1001            SET_BIT(0xA0);
1002            SET_BIT(0xFF);  /* For characters > 255 */
1003    #endif
1004            }
1005          else
1006    #endif /* SUPPORT_UTF */
1007            {
1008            SET_BIT(0xA0);
1009    #ifdef COMPILE_PCRE16
1010            SET_BIT(0xFF);  /* For characters > 255 */
1011    #endif
1012            }
1013          try_next = FALSE;
1014          break;
1015    
1016          case OP_ANYNL:
1017          case OP_VSPACE:
1018          SET_BIT(0x0A);
1019          SET_BIT(0x0B);
1020          SET_BIT(0x0C);
1021          SET_BIT(0x0D);
1022    #ifdef SUPPORT_UTF
1023          if (utf)
1024            {
1025    #ifdef COMPILE_PCRE8
1026            SET_BIT(0xC2);  /* For U+0085 */
1027            SET_BIT(0xE2);  /* For U+2028, U+2029 */
1028    #endif
1029    #ifdef COMPILE_PCRE16
1030            SET_BIT(0x85);
1031            SET_BIT(0xFF);  /* For characters > 255 */
1032    #endif
1033            }
1034          else
1035    #endif /* SUPPORT_UTF */
1036            {
1037            SET_BIT(0x85);
1038    #ifdef COMPILE_PCRE16
1039            SET_BIT(0xFF);  /* For characters > 255 */
1040    #endif
1041            }
1042        try_next = FALSE;        try_next = FALSE;
1043        break;        break;
1044    
1045        /* Single character type sets the bits and stops */        /* Single character types set the bits and stop. Note that if PCRE_UCP
1046          is set, we do not see these op codes because \d etc are converted to
1047          properties. Therefore, these apply in the case when only characters less
1048          than 256 are recognized to match the types. */
1049    
1050        case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
1051        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_digit];  
1052        try_next = FALSE;        try_next = FALSE;
1053        break;        break;
1054    
1055        case OP_DIGIT:        case OP_DIGIT:
1056        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_digit, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_digit];  
1057        try_next = FALSE;        try_next = FALSE;
1058        break;        break;
1059    
1060        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to
1061        discard it. */        ensure it is set as not whitespace. */
1062    
1063        case OP_NOT_WHITESPACE:        case OP_NOT_WHITESPACE:
1064        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_space, table_limit, cd);
1065          {        start_bits[1] |= 0x08;
         int d = cd->cbits[c+cbit_space];  
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= ~d;  
         }  
1066        try_next = FALSE;        try_next = FALSE;
1067        break;        break;
1068    
1069        /* The cbit_space table has vertical tab as whitespace; we have to        /* The cbit_space table has vertical tab as whitespace; we have to
1070        discard it. */        not set it from the table. */
1071    
1072        case OP_WHITESPACE:        case OP_WHITESPACE:
1073        for (c = 0; c < 32; c++)        c = start_bits[1];    /* Save in case it was already set */
1074          {        set_type_bits(start_bits, cbit_space, table_limit, cd);
1075          int d = cd->cbits[c+cbit_space];        start_bits[1] = (start_bits[1] & ~0x08) | c;
         if (c == 1) d &= ~0x08;  
         start_bits[c] |= d;  
         }  
1076        try_next = FALSE;        try_next = FALSE;
1077        break;        break;
1078    
1079        case OP_NOT_WORDCHAR:        case OP_NOT_WORDCHAR:
1080        for (c = 0; c < 32; c++)        set_nottype_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= ~cd->cbits[c+cbit_word];  
1081        try_next = FALSE;        try_next = FALSE;
1082        break;        break;
1083    
1084        case OP_WORDCHAR:        case OP_WORDCHAR:
1085        for (c = 0; c < 32; c++)        set_type_bits(start_bits, cbit_word, table_limit, cd);
         start_bits[c] |= cd->cbits[c+cbit_word];  
1086        try_next = FALSE;        try_next = FALSE;
1087        break;        break;
1088    
# Line 673  do Line 1091  do
1091    
1092        case OP_TYPEPLUS:        case OP_TYPEPLUS:
1093        case OP_TYPEMINPLUS:        case OP_TYPEMINPLUS:
1094          case OP_TYPEPOSPLUS:
1095        tcode++;        tcode++;
1096        break;        break;
1097    
1098        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1099        tcode += 3;        tcode += 1 + IMM2_SIZE;
1100        break;        break;
1101    
1102        /* 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 686  do Line 1105  do
1105        case OP_TYPEUPTO:        case OP_TYPEUPTO:
1106        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
1107        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1108        tcode += 2;               /* Fall through */        tcode += IMM2_SIZE;  /* Fall through */
1109    
1110        case OP_TYPESTAR:        case OP_TYPESTAR:
1111        case OP_TYPEMINSTAR:        case OP_TYPEMINSTAR:
# Line 696  do Line 1115  do
1115        case OP_TYPEPOSQUERY:        case OP_TYPEPOSQUERY:
1116        switch(tcode[1])        switch(tcode[1])
1117          {          {
1118            default:
1119          case OP_ANY:          case OP_ANY:
1120          case OP_ALLANY:          case OP_ALLANY:
1121          return SSB_FAIL;          return SSB_FAIL;
1122    
1123            case OP_HSPACE:
1124            SET_BIT(0x09);
1125            SET_BIT(0x20);
1126    #ifdef COMPILE_PCRE8
1127            if (utf)
1128              {
1129    #ifdef COMPILE_PCRE8
1130              SET_BIT(0xC2);  /* For U+00A0 */
1131              SET_BIT(0xE1);  /* For U+1680, U+180E */
1132              SET_BIT(0xE2);  /* For U+2000 - U+200A, U+202F, U+205F */
1133              SET_BIT(0xE3);  /* For U+3000 */
1134    #endif
1135    #ifdef COMPILE_PCRE16
1136              SET_BIT(0xA0);
1137              SET_BIT(0xFF);  /* For characters > 255 */
1138    #endif
1139              }
1140            else
1141    #endif /* SUPPORT_UTF */
1142              SET_BIT(0xA0);
1143            break;
1144    
1145            case OP_ANYNL:
1146            case OP_VSPACE:
1147            SET_BIT(0x0A);
1148            SET_BIT(0x0B);
1149            SET_BIT(0x0C);
1150            SET_BIT(0x0D);
1151    #ifdef COMPILE_PCRE8
1152            if (utf)
1153              {
1154    #ifdef COMPILE_PCRE8
1155              SET_BIT(0xC2);  /* For U+0085 */
1156              SET_BIT(0xE2);  /* For U+2028, U+2029 */
1157    #endif
1158    #ifdef COMPILE_PCRE16
1159              SET_BIT(0x85);
1160              SET_BIT(0xFF);  /* For characters > 255 */
1161    #endif
1162              }
1163            else
1164    #endif /* SUPPORT_UTF */
1165              SET_BIT(0x85);
1166            break;
1167    
1168          case OP_NOT_DIGIT:          case OP_NOT_DIGIT:
1169          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_digit];  
1170          break;          break;
1171    
1172          case OP_DIGIT:          case OP_DIGIT:
1173          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_digit, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_digit];  
1174          break;          break;
1175    
1176          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we have to
1177          discard it. */          ensure it gets set as not whitespace. */
1178    
1179          case OP_NOT_WHITESPACE:          case OP_NOT_WHITESPACE:
1180          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_space, table_limit, cd);
1181            {          start_bits[1] |= 0x08;
           int d = cd->cbits[c+cbit_space];  
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= ~d;  
           }  
1182          break;          break;
1183    
1184          /* The cbit_space table has vertical tab as whitespace; we have to          /* The cbit_space table has vertical tab as whitespace; we have to
1185          discard it. */          avoid setting it. */
1186    
1187          case OP_WHITESPACE:          case OP_WHITESPACE:
1188          for (c = 0; c < 32; c++)          c = start_bits[1];    /* Save in case it was already set */
1189            {          set_type_bits(start_bits, cbit_space, table_limit, cd);
1190            int d = cd->cbits[c+cbit_space];          start_bits[1] = (start_bits[1] & ~0x08) | c;
           if (c == 1) d &= ~0x08;  
           start_bits[c] |= d;  
           }  
1191          break;          break;
1192    
1193          case OP_NOT_WORDCHAR:          case OP_NOT_WORDCHAR:
1194          for (c = 0; c < 32; c++)          set_nottype_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= ~cd->cbits[c+cbit_word];  
1195          break;          break;
1196    
1197          case OP_WORDCHAR:          case OP_WORDCHAR:
1198          for (c = 0; c < 32; c++)          set_type_bits(start_bits, cbit_word, table_limit, cd);
           start_bits[c] |= cd->cbits[c+cbit_word];  
1199          break;          break;
1200          }          }
1201    
# Line 755  do Line 1209  do
1209        character with a value > 255. */        character with a value > 255. */
1210    
1211        case OP_NCLASS:        case OP_NCLASS:
1212  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
1213        if (utf8)        if (utf)
1214          {          {
1215          start_bits[24] |= 0xf0;              /* Bits for 0xc4 - 0xc8 */          start_bits[24] |= 0xf0;              /* Bits for 0xc4 - 0xc8 */
1216          memset(start_bits+25, 0xff, 7);      /* Bits for 0xc9 - 0xff */          memset(start_bits+25, 0xff, 7);      /* Bits for 0xc9 - 0xff */
1217          }          }
1218  #endif  #endif
1219    #ifdef COMPILE_PCRE16
1220          SET_BIT(0xFF);                         /* For characters > 255 */
1221    #endif
1222        /* Fall through */        /* Fall through */
1223    
1224        case OP_CLASS:        case OP_CLASS:
1225          {          {
1226            pcre_uint8 *map;
1227          tcode++;          tcode++;
1228            map = (pcre_uint8 *)tcode;
1229    
1230          /* 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
1231          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 774  do Line 1233  do
1233          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
1234          characters in the range 128 - 255. */          characters in the range 128 - 255. */
1235    
1236  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
1237          if (utf8)          if (utf)
1238            {            {
1239            for (c = 0; c < 16; c++) start_bits[c] |= tcode[c];            for (c = 0; c < 16; c++) start_bits[c] |= map[c];
1240            for (c = 128; c < 256; c++)            for (c = 128; c < 256; c++)
1241              {              {
1242              if ((tcode[c/8] && (1 << (c&7))) != 0)              if ((map[c/8] && (1 << (c&7))) != 0)
1243                {                {
1244                int d = (c >> 6) | 0xc0;            /* Set bit for this starter */                int d = (c >> 6) | 0xc0;            /* Set bit for this starter */
1245                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 788  do Line 1247  do
1247                }                }
1248              }              }
1249            }            }
   
         /* In non-UTF-8 mode, the two bit maps are completely compatible. */  
   
1250          else          else
1251  #endif  #endif
1252            {            {
1253            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];            /* In non-UTF-8 mode, the two bit maps are completely compatible. */
1254              for (c = 0; c < 32; c++) start_bits[c] |= map[c];
1255            }            }
1256    
1257          /* Advance past the bit map, and act on what follows */          /* Advance past the bit map, and act on what follows. For a zero
1258            minimum repeat, continue; otherwise stop processing. */
1259    
1260          tcode += 32;          tcode += 32 / sizeof(pcre_uchar);
1261          switch (*tcode)          switch (*tcode)
1262            {            {
1263            case OP_CRSTAR:            case OP_CRSTAR:
# Line 811  do Line 1269  do
1269    
1270            case OP_CRRANGE:            case OP_CRRANGE:
1271            case OP_CRMINRANGE:            case OP_CRMINRANGE:
1272            if (((tcode[1] << 8) + tcode[2]) == 0) tcode += 5;            if (GET2(tcode, 1) == 0) tcode += 1 + 2 * IMM2_SIZE;
1273              else try_next = FALSE;              else try_next = FALSE;
1274            break;            break;
1275    
# Line 833  return yield; Line 1291  return yield;
1291    
1292    
1293    
1294    
1295    
1296  /*************************************************  /*************************************************
1297  *          Study a compiled expression           *  *          Study a compiled expression           *
1298  *************************************************/  *************************************************/
1299    
1300  /* 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
1301  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
1302  which then gets handed back to pcre_exec().  which then gets handed back to pcre_exec().
1303    
1304  Arguments:  Arguments:
# Line 847  Arguments: Line 1307  Arguments:
1307    errorptr  points to where to place error messages;    errorptr  points to where to place error messages;
1308              set NULL unless error              set NULL unless error
1309    
1310  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
1311                appropriate flags set;                the appropriate flags set;
1312              NULL on error or if no optimization possible              NULL on error or if no optimization possible
1313  */  */
1314    
1315    #ifdef COMPILE_PCRE8
1316  PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION  PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION
1317  pcre_study(const pcre *external_re, int options, const char **errorptr)  pcre_study(const pcre *external_re, int options, const char **errorptr)
1318    #else
1319    PCRE_EXP_DEFN pcre16_extra * PCRE_CALL_CONVENTION
1320    pcre16_study(const pcre16 *external_re, int options, const char **errorptr)
1321    #endif
1322  {  {
1323  int min;  int min;
1324  BOOL bits_set = FALSE;  BOOL bits_set = FALSE;
1325  uschar start_bits[32];  pcre_uint8 start_bits[32];
1326  pcre_extra *extra;  PUBL(extra) *extra = NULL;
1327  pcre_study_data *study;  pcre_study_data *study;
1328  const uschar *tables;  const pcre_uint8 *tables;
1329  uschar *code;  pcre_uchar *code;
1330  compile_data compile_block;  compile_data compile_block;
1331  const real_pcre *re = (const real_pcre *)external_re;  const REAL_PCRE *re = (const REAL_PCRE *)external_re;
1332    
1333  *errorptr = NULL;  *errorptr = NULL;
1334    
# Line 873  if (re == NULL || re->magic_number != MA Line 1338  if (re == NULL || re->magic_number != MA
1338    return NULL;    return NULL;
1339    }    }
1340    
1341    if ((re->flags & PCRE_MODE) == 0)
1342      {
1343    #ifdef COMPILE_PCRE8
1344      *errorptr = "argument is compiled in 16 bit mode";
1345    #else
1346      *errorptr = "argument is compiled in 8 bit mode";
1347    #endif
1348      return NULL;
1349      }
1350    
1351  if ((options & ~PUBLIC_STUDY_OPTIONS) != 0)  if ((options & ~PUBLIC_STUDY_OPTIONS) != 0)
1352    {    {
1353    *errorptr = "unknown or incorrect option bit(s) set";    *errorptr = "unknown or incorrect option bit(s) set";
1354    return NULL;    return NULL;
1355    }    }
1356    
1357  code = (uschar *)re + re->name_table_offset +  code = (pcre_uchar *)re + re->name_table_offset +
1358    (re->name_count * re->name_entry_size);    (re->name_count * re->name_entry_size);
1359    
1360  /* 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
1361  a multiline pattern that matches only at "line starts", there is no point in  a multiline pattern that matches only at "line starts", there is no point in
1362  seeking a list of starting bytes. */  seeking a list of starting bytes. */
1363    
1364  if ((re->options & PCRE_ANCHORED) == 0 &&  if ((re->options & PCRE_ANCHORED) == 0 &&
1365      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)
1366    {    {
1367      int rc;
1368    
1369    /* Set the character tables in the block that is passed around */    /* Set the character tables in the block that is passed around */
1370    
1371    tables = re->tables;    tables = re->tables;
1372    
1373    #ifdef COMPILE_PCRE8
1374    if (tables == NULL)    if (tables == NULL)
1375      (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,      (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1376      (void *)(&tables));      (void *)(&tables));
1377    #else
1378      if (tables == NULL)
1379        (void)pcre16_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1380        (void *)(&tables));
1381    #endif
1382    
1383    compile_block.lcc = tables + lcc_offset;    compile_block.lcc = tables + lcc_offset;
1384    compile_block.fcc = tables + fcc_offset;    compile_block.fcc = tables + fcc_offset;
1385    compile_block.cbits = tables + cbits_offset;    compile_block.cbits = tables + cbits_offset;
1386    compile_block.ctypes = tables + ctypes_offset;    compile_block.ctypes = tables + ctypes_offset;
1387    
1388    /* 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. */
1389    
1390    memset(start_bits, 0, 32 * sizeof(uschar));    memset(start_bits, 0, 32 * sizeof(pcre_uint8));
1391    bits_set = set_start_bits(code, start_bits,    rc = set_start_bits(code, start_bits, (re->options & PCRE_UTF8) != 0,
1392      (re->options & PCRE_CASELESS) != 0, (re->options & PCRE_UTF8) != 0,      &compile_block);
1393      &compile_block) == SSB_DONE;    bits_set = rc == SSB_DONE;
1394      if (rc == SSB_UNKNOWN)
1395        {
1396        *errorptr = "internal error: opcode not recognized";
1397        return NULL;
1398        }
1399    }    }
1400    
1401  /* Find the minimum length of subject string. */  /* Find the minimum length of subject string. */
1402    
1403  min = find_minlength(code, code, re->options);  switch(min = find_minlength(code, code, re->options, 0))
1404      {
1405      case -2: *errorptr = "internal error: missing capturing bracket"; return NULL;
1406      case -3: *errorptr = "internal error: opcode not recognized"; return NULL;
1407      default: break;
1408      }
1409    
1410  /* Return NULL if no optimization is possible. */  /* If a set of starting bytes has been identified, or if the minimum length is
1411    greater than zero, or if JIT optimization has been requested, get a
1412    pcre[16]_extra block and a pcre_study_data block. The study data is put in the
1413    latter, which is pointed to by the former, which may also get additional data
1414    set later by the calling program. At the moment, the size of pcre_study_data
1415    is fixed. We nevertheless save it in a field for returning via the
1416    pcre_fullinfo() function so that if it becomes variable in the future,
1417    we don't have to change that code. */
1418    
1419    if (bits_set || min > 0
1420    #ifdef SUPPORT_JIT
1421        || (options & (PCRE_STUDY_JIT_COMPILE | PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE
1422                     | PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE)) != 0
1423    #endif
1424      )
1425      {
1426      extra = (PUBL(extra) *)(PUBL(malloc))
1427        (sizeof(PUBL(extra)) + sizeof(pcre_study_data));
1428      if (extra == NULL)
1429        {
1430        *errorptr = "failed to get memory";
1431        return NULL;
1432        }
1433    
1434  if (!bits_set && min < 0) return NULL;    study = (pcre_study_data *)((char *)extra + sizeof(PUBL(extra)));
1435      extra->flags = PCRE_EXTRA_STUDY_DATA;
1436      extra->study_data = study;
1437    
1438      study->size = sizeof(pcre_study_data);
1439      study->flags = 0;
1440    
1441      /* Set the start bits always, to avoid unset memory errors if the
1442      study data is written to a file, but set the flag only if any of the bits
1443      are set, to save time looking when none are. */
1444    
1445  /* Get a pcre_extra block and a pcre_study_data block. The study data is put in    if (bits_set)
1446  the latter, which is pointed to by the former, which may also get additional      {
1447  data set later by the calling program. At the moment, the size of      study->flags |= PCRE_STUDY_MAPPED;
1448  pcre_study_data is fixed. We nevertheless save it in a field for returning via      memcpy(study->start_bits, start_bits, sizeof(start_bits));
1449  the pcre_fullinfo() function so that if it becomes variable in the future, we      }
1450  don't have to change that code. */    else memset(study->start_bits, 0, 32 * sizeof(pcre_uint8));
1451    
1452  extra = (pcre_extra *)(pcre_malloc)  #ifdef PCRE_DEBUG
1453    (sizeof(pcre_extra) + sizeof(pcre_study_data));    if (bits_set)
1454        {
1455        pcre_uint8 *ptr = start_bits;
1456        int i;
1457    
1458  if (extra == NULL)      printf("Start bits:\n");
1459    {      for (i = 0; i < 32; i++)
1460    *errorptr = "failed to get memory";        printf("%3d: %02x%s", i * 8, *ptr++, ((i + 1) & 0x7) != 0? " " : "\n");
1461    return NULL;      }
1462    }  #endif
1463    
1464  study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));    /* Always set the minlength value in the block, because the JIT compiler
1465  extra->flags = PCRE_EXTRA_STUDY_DATA;    makes use of it. However, don't set the bit unless the length is greater than
1466  extra->study_data = study;    zero - the interpretive pcre_exec() and pcre_dfa_exec() needn't waste time
1467      checking the zero case. */
1468    
1469  study->size = sizeof(pcre_study_data);    if (min > 0)
1470  study->flags = 0;      {
1471        study->flags |= PCRE_STUDY_MINLEN;
1472        study->minlength = min;
1473        }
1474      else study->minlength = 0;
1475    
1476  if (bits_set)    /* If JIT support was compiled and requested, attempt the JIT compilation.
1477    {    If no starting bytes were found, and the minimum length is zero, and JIT
1478    study->flags |= PCRE_STUDY_MAPPED;    compilation fails, abandon the extra block and return NULL. */
1479    memcpy(study->start_bits, start_bits, sizeof(start_bits));  
1480    #ifdef SUPPORT_JIT
1481      extra->executable_jit = NULL;
1482      if ((options & PCRE_STUDY_JIT_COMPILE) != 0)
1483        PRIV(jit_compile)(re, extra, JIT_COMPILE);
1484      if ((options & PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE) != 0)
1485        PRIV(jit_compile)(re, extra, JIT_PARTIAL_SOFT_COMPILE);
1486      if ((options & PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE) != 0)
1487        PRIV(jit_compile)(re, extra, JIT_PARTIAL_HARD_COMPILE);
1488    
1489      if (study->flags == 0 && (extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) == 0)
1490        {
1491    #ifdef COMPILE_PCRE8
1492        pcre_free_study(extra);
1493    #endif
1494    #ifdef COMPILE_PCRE16
1495        pcre16_free_study(extra);
1496    #endif
1497        extra = NULL;
1498        }
1499    #endif
1500    }    }
   
 if (min >= 0)  
   {  
   study->flags |= PCRE_STUDY_MINLEN;  
   study->minlength = min;  
   }  
1501    
1502  return extra;  return extra;
1503  }  }
1504    
1505    
1506    /*************************************************
1507    *          Free the study data                   *
1508    *************************************************/
1509    
1510    /* This function frees the memory that was obtained by pcre_study().
1511    
1512    Argument:   a pointer to the pcre[16]_extra block
1513    Returns:    nothing
1514    */
1515    
1516    #ifdef COMPILE_PCRE8
1517    PCRE_EXP_DEFN void
1518    pcre_free_study(pcre_extra *extra)
1519    #else
1520    PCRE_EXP_DEFN void
1521    pcre16_free_study(pcre16_extra *extra)
1522    #endif
1523    {
1524    if (extra == NULL)
1525      return;
1526    #ifdef SUPPORT_JIT
1527    if ((extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 &&
1528         extra->executable_jit != NULL)
1529      PRIV(jit_free)(extra->executable_jit);
1530    #endif
1531    PUBL(free)(extra);
1532    }
1533    
1534  /* End of pcre_study.c */  /* End of pcre_study.c */

Legend:
Removed from v.455  
changed lines
  Added in v.914

  ViewVC Help
Powered by ViewVC 1.1.5