/[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 602 by ph10, Wed May 25 08:29:03 2011 UTC revision 835 by ph10, Wed Dec 28 16:10:09 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 uschar *code, const uschar *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;
# Line 84  BOOL had_recurse = FALSE; Line 87  BOOL had_recurse = FALSE;
87  register int branchlength = 0;  register int branchlength = 0;
88  register uschar *cc = (uschar *)code + 1 + LINK_SIZE;  register uschar *cc = (uschar *)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 += 2;
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 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 178  for (;;) Line 196  for (;;)
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 += _pcre_OP_lengths[*cc];
202      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 191  for (;;) Line 210  for (;;)
210      case OP_NOT:      case OP_NOT:
211      case OP_NOTI:      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
# Line 214  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 += 4;
247  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 251  for (;;) Line 278  for (;;)
278      cc++;      cc++;
279      break;      break;
280    
281      /* "Any newline" might match two characters */      /* "Any newline" might match two characters, but it also might match just
282        one. */
283    
284      case OP_ANYNL:      case OP_ANYNL:
285      branchlength += 2;      branchlength += 1;
286      cc++;      cc++;
287      break;      break;
288    
289      /* 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
290        non-UTF-8 mode \C will actually be turned into OP_ALLANY, so won't ever
291        appear, but leave the code, just in case.) */
292    
293      case OP_ANYBYTE:      case OP_ANYBYTE:
294  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 350  for (;;) Line 380  for (;;)
380          d = 0;          d = 0;
381          had_recurse = TRUE;          had_recurse = TRUE;
382          }          }
383        else d = find_minlength(cs, startcode, options);        else
384            {
385            d = find_minlength(cs, startcode, options, recurse_depth);
386            }
387        }        }
388      else d = 0;      else d = 0;
389      cc += 3;      cc += 3;
# Line 367  for (;;) Line 400  for (;;)
400        cc++;        cc++;
401        break;        break;
402    
403          case OP_CRPLUS:
404          case OP_CRMINPLUS:
405          min = 1;
406          cc++;
407          break;
408    
409        case OP_CRRANGE:        case OP_CRRANGE:
410        case OP_CRMINRANGE:        case OP_CRMINRANGE:
411        min = GET2(cc, 1);        min = GET2(cc, 1);
# Line 381  for (;;) Line 420  for (;;)
420      branchlength += min * d;      branchlength += min * d;
421      break;      break;
422    
423        /* We can easily detect direct recursion, but not mutual recursion. This is
424        caught by a recursion depth count. */
425    
426      case OP_RECURSE:      case OP_RECURSE:
427      cs = ce = (uschar *)startcode + GET(cc, 1);      cs = ce = (uschar *)startcode + GET(cc, 1);
     if (cs == NULL) return -2;  
428      do ce += GET(ce, 1); while (*ce == OP_ALT);      do ce += GET(ce, 1); while (*ce == OP_ALT);
429      if (cc > cs && cc < ce)      if ((cc > cs && cc < ce) || recurse_depth > 10)
430        had_recurse = TRUE;        had_recurse = TRUE;
431      else      else
432        branchlength += find_minlength(cs, startcode, options);        {
433          branchlength += find_minlength(cs, startcode, options, recurse_depth + 1);
434          }
435      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
436      break;      break;
437    
# Line 397  for (;;) Line 440  for (;;)
440      of a character, we must take special action for UTF-8 characters. As it      of a character, we must take special action for UTF-8 characters. As it
441      happens, the "NOT" versions of these opcodes are used at present only for      happens, the "NOT" versions of these opcodes are used at present only for
442      ASCII characters, so they could be omitted from this list. However, in      ASCII characters, so they could be omitted from this list. However, in
443      future that may change, so we leave them in this special case. */      future that may change, so we include them here so as not to leave a
444        gotcha for a future maintainer. */
445    
446      case OP_UPTO:      case OP_UPTO:
447      case OP_UPTOI:      case OP_UPTOI:
# Line 449  for (;;) Line 493  for (;;)
493      case OP_MARK:      case OP_MARK:
494      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
495      case OP_SKIP_ARG:      case OP_SKIP_ARG:
496        case OP_THEN_ARG:
497      cc += _pcre_OP_lengths[op] + cc[1];      cc += _pcre_OP_lengths[op] + cc[1];
498      break;      break;
499    
500      case OP_THEN_ARG:      /* The remaining opcodes are just skipped over. */
501      cc += _pcre_OP_lengths[op] + cc[1+LINK_SIZE];  
502        case OP_CLOSE:
503        case OP_COMMIT:
504        case OP_FAIL:
505        case OP_PRUNE:
506        case OP_SET_SOM:
507        case OP_SKIP:
508        case OP_THEN:
509        cc += _pcre_OP_lengths[op];
510      break;      break;
511    
512      /* For the record, these are the opcodes that are matched by "default":      /* This should not occur: we list all opcodes explicitly so that when
513      OP_ACCEPT, OP_CLOSE, OP_COMMIT, OP_FAIL, OP_PRUNE, OP_SET_SOM, OP_SKIP,      new ones get added they are properly considered. */
     OP_THEN. */  
514    
515      default:      default:
516      cc += _pcre_OP_lengths[op];      return -3;
     break;  
517      }      }
518    }    }
519  /* Control never gets here */  /* Control never gets here */
# Line 608  function fails unless the result is SSB_ Line 659  function fails unless the result is SSB_
659  Arguments:  Arguments:
660    code         points to an expression    code         points to an expression
661    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  
662    utf8         TRUE if in UTF-8 mode    utf8         TRUE if in UTF-8 mode
663    cd           the block with char table pointers    cd           the block with char table pointers
664    
665  Returns:       SSB_FAIL     => Failed to find any starting bytes  Returns:       SSB_FAIL     => Failed to find any starting bytes
666                 SSB_DONE     => Found mandatory starting bytes                 SSB_DONE     => Found mandatory starting bytes
667                 SSB_CONTINUE => Found optional starting bytes                 SSB_CONTINUE => Found optional starting bytes
668                   SSB_UNKNOWN  => Hit an unrecognized opcode
669  */  */
670    
671  static int  static int
672  set_start_bits(const uschar *code, uschar *start_bits, BOOL caseless,  set_start_bits(const uschar *code, uschar *start_bits, BOOL utf8,
673    BOOL utf8, compile_data *cd)    compile_data *cd)
674  {  {
675  register int c;  register int c;
676  int yield = SSB_DONE;  int yield = SSB_DONE;
# Line 644  volatile int dummy; Line 695  volatile int dummy;
695    
696  do  do
697    {    {
   const uschar *tcode = code + (((int)*code == OP_CBRA)? 3:1) + LINK_SIZE;  
698    BOOL try_next = TRUE;    BOOL try_next = TRUE;
699      const uschar *tcode = code + 1 + LINK_SIZE;
700    
701      if (*code == OP_CBRA || *code == OP_SCBRA ||
702          *code == OP_CBRAPOS || *code == OP_SCBRAPOS) tcode += 2;
703    
704    while (try_next)    /* Loop for items in this branch */    while (try_next)    /* Loop for items in this branch */
705      {      {
706      int rc;      int rc;
707    
708      switch(*tcode)      switch(*tcode)
709        {        {
710        /* Fail if we reach something we don't understand */        /* If we reach something we don't understand, it means a new opcode has
711          been created that hasn't been added to this code. Hopefully this problem
712          will be discovered during testing. */
713    
714        default:        default:
715          return SSB_UNKNOWN;
716    
717          /* Fail for a valid opcode that implies no starting bits. */
718    
719          case OP_ACCEPT:
720          case OP_ASSERT_ACCEPT:
721          case OP_ALLANY:
722          case OP_ANY:
723          case OP_ANYBYTE:
724          case OP_CIRC:
725          case OP_CIRCM:
726          case OP_CLOSE:
727          case OP_COMMIT:
728          case OP_COND:
729          case OP_CREF:
730          case OP_DEF:
731          case OP_DOLL:
732          case OP_DOLLM:
733          case OP_END:
734          case OP_EOD:
735          case OP_EODN:
736          case OP_EXTUNI:
737          case OP_FAIL:
738          case OP_MARK:
739          case OP_NCREF:
740          case OP_NOT:
741          case OP_NOTEXACT:
742          case OP_NOTEXACTI:
743          case OP_NOTI:
744          case OP_NOTMINPLUS:
745          case OP_NOTMINPLUSI:
746          case OP_NOTMINQUERY:
747          case OP_NOTMINQUERYI:
748          case OP_NOTMINSTAR:
749          case OP_NOTMINSTARI:
750          case OP_NOTMINUPTO:
751          case OP_NOTMINUPTOI:
752          case OP_NOTPLUS:
753          case OP_NOTPLUSI:
754          case OP_NOTPOSPLUS:
755          case OP_NOTPOSPLUSI:
756          case OP_NOTPOSQUERY:
757          case OP_NOTPOSQUERYI:
758          case OP_NOTPOSSTAR:
759          case OP_NOTPOSSTARI:
760          case OP_NOTPOSUPTO:
761          case OP_NOTPOSUPTOI:
762          case OP_NOTPROP:
763          case OP_NOTQUERY:
764          case OP_NOTQUERYI:
765          case OP_NOTSTAR:
766          case OP_NOTSTARI:
767          case OP_NOTUPTO:
768          case OP_NOTUPTOI:
769          case OP_NOT_HSPACE:
770          case OP_NOT_VSPACE:
771          case OP_NRREF:
772          case OP_PROP:
773          case OP_PRUNE:
774          case OP_PRUNE_ARG:
775          case OP_RECURSE:
776          case OP_REF:
777          case OP_REFI:
778          case OP_REVERSE:
779          case OP_RREF:
780          case OP_SCOND:
781          case OP_SET_SOM:
782          case OP_SKIP:
783          case OP_SKIP_ARG:
784          case OP_SOD:
785          case OP_SOM:
786          case OP_THEN:
787          case OP_THEN_ARG:
788          case OP_XCLASS:
789        return SSB_FAIL;        return SSB_FAIL;
790    
791          /* We can ignore word boundary tests. */
792    
793          case OP_WORD_BOUNDARY:
794          case OP_NOT_WORD_BOUNDARY:
795          tcode++;
796          break;
797    
798        /* 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
799        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
800        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 666  do Line 804  do
804        case OP_SBRA:        case OP_SBRA:
805        case OP_CBRA:        case OP_CBRA:
806        case OP_SCBRA:        case OP_SCBRA:
807          case OP_BRAPOS:
808          case OP_SBRAPOS:
809          case OP_CBRAPOS:
810          case OP_SCBRAPOS:
811        case OP_ONCE:        case OP_ONCE:
812          case OP_ONCE_NC:
813        case OP_ASSERT:        case OP_ASSERT:
814        rc = set_start_bits(tcode, start_bits, caseless, utf8, cd);        rc = set_start_bits(tcode, start_bits, utf8, cd);
815        if (rc == SSB_FAIL) return SSB_FAIL;        if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
816        if (rc == SSB_DONE) try_next = FALSE; else        if (rc == SSB_DONE) try_next = FALSE; else
817          {          {
818          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);          do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
# Line 692  do Line 835  do
835        case OP_KET:        case OP_KET:
836        case OP_KETRMAX:        case OP_KETRMAX:
837        case OP_KETRMIN:        case OP_KETRMIN:
838          case OP_KETRPOS:
839        return SSB_CONTINUE;        return SSB_CONTINUE;
840    
841        /* Skip over callout */        /* Skip over callout */
# Line 713  do Line 857  do
857    
858        case OP_BRAZERO:        case OP_BRAZERO:
859        case OP_BRAMINZERO:        case OP_BRAMINZERO:
860        if (set_start_bits(++tcode, start_bits, caseless, utf8, cd) == SSB_FAIL)        case OP_BRAPOSZERO:
861          return SSB_FAIL;        rc = set_start_bits(++tcode, start_bits, utf8, cd);
862          if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
863  /* =========================================================================  /* =========================================================================
864        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,
865        which was an old fudge for the benefit of OS/2.        which was an old fudge for the benefit of OS/2.
# Line 740  do Line 885  do
885        case OP_QUERY:        case OP_QUERY:
886        case OP_MINQUERY:        case OP_MINQUERY:
887        case OP_POSQUERY:        case OP_POSQUERY:
888        tcode = set_table_bit(start_bits, tcode + 1, caseless, cd, utf8);        tcode = set_table_bit(start_bits, tcode + 1, FALSE, cd, utf8);
889        break;        break;
890    
891        case OP_STARI:        case OP_STARI:
# Line 757  do Line 902  do
902        case OP_UPTO:        case OP_UPTO:
903        case OP_MINUPTO:        case OP_MINUPTO:
904        case OP_POSUPTO:        case OP_POSUPTO:
905        tcode = set_table_bit(start_bits, tcode + 3, caseless, cd, utf8);        tcode = set_table_bit(start_bits, tcode + 3, FALSE, cd, utf8);
906        break;        break;
907    
908        case OP_UPTOI:        case OP_UPTOI:
# Line 775  do Line 920  do
920        case OP_PLUS:        case OP_PLUS:
921        case OP_MINPLUS:        case OP_MINPLUS:
922        case OP_POSPLUS:        case OP_POSPLUS:
923        (void)set_table_bit(start_bits, tcode + 1, caseless, cd, utf8);        (void)set_table_bit(start_bits, tcode + 1, FALSE, cd, utf8);
924        try_next = FALSE;        try_next = FALSE;
925        break;        break;
926    
927          case OP_EXACTI:
928          tcode += 2;
929          /* Fall through */
930        case OP_CHARI:        case OP_CHARI:
931        case OP_PLUSI:        case OP_PLUSI:
932        case OP_MINPLUSI:        case OP_MINPLUSI:
# Line 1013  do Line 1161  do
1161            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];            for (c = 0; c < 32; c++) start_bits[c] |= tcode[c];
1162            }            }
1163    
1164          /* Advance past the bit map, and act on what follows */          /* Advance past the bit map, and act on what follows. For a zero
1165            minimum repeat, continue; otherwise stop processing. */
1166    
1167          tcode += 32;          tcode += 32;
1168          switch (*tcode)          switch (*tcode)
# Line 1049  return yield; Line 1198  return yield;
1198    
1199    
1200    
1201    
1202    
1203  /*************************************************  /*************************************************
1204  *          Study a compiled expression           *  *          Study a compiled expression           *
1205  *************************************************/  *************************************************/
# Line 1074  pcre_study(const pcre *external_re, int Line 1225  pcre_study(const pcre *external_re, int
1225  int min;  int min;
1226  BOOL bits_set = FALSE;  BOOL bits_set = FALSE;
1227  uschar start_bits[32];  uschar start_bits[32];
1228  pcre_extra *extra;  pcre_extra *extra = NULL;
1229  pcre_study_data *study;  pcre_study_data *study;
1230  const uschar *tables;  const uschar *tables;
1231  uschar *code;  uschar *code;
# Line 1105  seeking a list of starting bytes. */ Line 1256  seeking a list of starting bytes. */
1256  if ((re->options & PCRE_ANCHORED) == 0 &&  if ((re->options & PCRE_ANCHORED) == 0 &&
1257      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)      (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)
1258    {    {
1259      int rc;
1260    
1261    /* Set the character tables in the block that is passed around */    /* Set the character tables in the block that is passed around */
1262    
1263    tables = re->tables;    tables = re->tables;
# Line 1120  if ((re->options & PCRE_ANCHORED) == 0 & Line 1273  if ((re->options & PCRE_ANCHORED) == 0 &
1273    /* 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. */
1274    
1275    memset(start_bits, 0, 32 * sizeof(uschar));    memset(start_bits, 0, 32 * sizeof(uschar));
1276    bits_set = set_start_bits(code, start_bits,    rc = set_start_bits(code, start_bits, (re->options & PCRE_UTF8) != 0,
1277      (re->options & PCRE_CASELESS) != 0, (re->options & PCRE_UTF8) != 0,      &compile_block);
1278      &compile_block) == SSB_DONE;    bits_set = rc == SSB_DONE;
1279      if (rc == SSB_UNKNOWN)
1280        {
1281        *errorptr = "internal error: opcode not recognized";
1282        return NULL;
1283        }
1284    }    }
1285    
1286  /* Find the minimum length of subject string. */  /* Find the minimum length of subject string. */
1287    
1288  min = find_minlength(code, code, re->options);  switch(min = find_minlength(code, code, re->options, 0))
1289      {
1290      case -2: *errorptr = "internal error: missing capturing bracket"; return NULL;
1291      case -3: *errorptr = "internal error: opcode not recognized"; return NULL;
1292      default: break;
1293      }
1294    
1295  /* Return NULL if no optimization is possible. */  /* If a set of starting bytes has been identified, or if the minimum length is
1296    greater than zero, or if JIT optimization has been requested, get a pcre_extra
1297    block and a pcre_study_data block. The study data is put in the latter, which
1298    is pointed to by the former, which may also get additional data set later by
1299    the calling program. At the moment, the size of pcre_study_data is fixed. We
1300    nevertheless save it in a field for returning via the pcre_fullinfo() function
1301    so that if it becomes variable in the future, we don't have to change that
1302    code. */
1303    
1304    if (bits_set || min > 0
1305    #ifdef SUPPORT_JIT
1306        || (options & PCRE_STUDY_JIT_COMPILE) != 0
1307    #endif
1308      )
1309      {
1310      extra = (pcre_extra *)(pcre_malloc)
1311        (sizeof(pcre_extra) + sizeof(pcre_study_data));
1312      if (extra == NULL)
1313        {
1314        *errorptr = "failed to get memory";
1315        return NULL;
1316        }
1317    
1318  if (!bits_set && min < 0) return NULL;    study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));
1319      extra->flags = PCRE_EXTRA_STUDY_DATA;
1320      extra->study_data = study;
1321    
1322  /* Get a pcre_extra block and a pcre_study_data block. The study data is put in    study->size = sizeof(pcre_study_data);
1323  the latter, which is pointed to by the former, which may also get additional    study->flags = 0;
 data set later by the calling program. At the moment, the size of  
 pcre_study_data is fixed. We nevertheless save it in a field for returning via  
 the pcre_fullinfo() function so that if it becomes variable in the future, we  
 don't have to change that code. */  
1324    
1325  extra = (pcre_extra *)(pcre_malloc)    if (bits_set)
1326    (sizeof(pcre_extra) + sizeof(pcre_study_data));      {
1327        study->flags |= PCRE_STUDY_MAPPED;
1328        memcpy(study->start_bits, start_bits, sizeof(start_bits));
1329        }
1330    
1331  if (extra == NULL)    /* Always set the minlength value in the block, because the JIT compiler
1332    {    makes use of it. However, don't set the bit unless the length is greater than
1333    *errorptr = "failed to get memory";    zero - the interpretive pcre_exec() and pcre_dfa_exec() needn't waste time
1334    return NULL;    checking the zero case. */
1335    
1336      if (min > 0)
1337        {
1338        study->flags |= PCRE_STUDY_MINLEN;
1339        study->minlength = min;
1340        }
1341      else study->minlength = 0;
1342    
1343      /* If JIT support was compiled and requested, attempt the JIT compilation.
1344      If no starting bytes were found, and the minimum length is zero, and JIT
1345      compilation fails, abandon the extra block and return NULL. */
1346    
1347    #ifdef SUPPORT_JIT
1348      extra->executable_jit = NULL;
1349      if ((options & PCRE_STUDY_JIT_COMPILE) != 0) _pcre_jit_compile(re, extra);
1350      if (study->flags == 0 && (extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) == 0)
1351        {
1352        pcre_free_study(extra);
1353        extra = NULL;
1354        }
1355    #endif
1356    }    }
1357    
1358  study = (pcre_study_data *)((char *)extra + sizeof(pcre_extra));  return extra;
1359  extra->flags = PCRE_EXTRA_STUDY_DATA;  }
 extra->study_data = study;  
1360    
 study->size = sizeof(pcre_study_data);  
 study->flags = 0;  
1361    
1362  if (bits_set)  /*************************************************
1363    {  *          Free the study data                   *
1364    study->flags |= PCRE_STUDY_MAPPED;  *************************************************/
   memcpy(study->start_bits, start_bits, sizeof(start_bits));  
   }  
1365    
1366  if (min >= 0)  /* This function frees the memory that was obtained by pcre_study().
   {  
   study->flags |= PCRE_STUDY_MINLEN;  
   study->minlength = min;  
   }  
1367    
1368  return extra;  Argument:   a pointer to the pcre_extra block
1369    Returns:    nothing
1370    */
1371    
1372    PCRE_EXP_DEFN void
1373    pcre_free_study(pcre_extra *extra)
1374    {
1375    #ifdef SUPPORT_JIT
1376    if ((extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 &&
1377         extra->executable_jit != NULL)
1378      _pcre_jit_free(extra->executable_jit);
1379    #endif
1380    pcre_free(extra);
1381  }  }
1382    
1383  /* End of pcre_study.c */  /* End of pcre_study.c */

Legend:
Removed from v.602  
changed lines
  Added in v.835

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12