/*************************************************

*      Perl-Compatible Regular Expressions       *

*************************************************/

/* PCRE is a library of functions to support regular expressions whose syntax

and semantics are as close as possible to those of the Perl 5 language (but see

below for why this module is different).

                       Written by Philip Hazel

           Copyright (c) 1997-2009 University of Cambridge

-----------------------------------------------------------------------------

Redistribution and use in source and binary forms, with or without

modification, are permitted provided that the following conditions are met:

    * Redistributions of source code must retain the above copyright notice,

      this list of conditions and the following disclaimer.

    * Redistributions in binary form must reproduce the above copyright

      notice, this list of conditions and the following disclaimer in the

      documentation and/or other materials provided with the distribution.

    * Neither the name of the University of Cambridge nor the names of its

      contributors may be used to endorse or promote products derived from

      this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"

AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE

IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE

ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE

LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR

CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF

SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS

INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN

CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)

ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE

POSSIBILITY OF SUCH DAMAGE.

-----------------------------------------------------------------------------

*/

/* This module contains the external function pcre_dfa_exec(), which is an

alternative matching function that uses a sort of DFA algorithm (not a true

FSM). This is NOT Perl- compatible, but it has advantages in certain

applications. */

#ifdef HAVE_CONFIG_H

#include "config.h"

#endif

#define NLBLOCK md             /* Block containing newline information */

#define PSSTART start_subject  /* Field containing processed string start */

#define PSEND   end_subject    /* Field containing processed string end */

#include "pcre_internal.h"

/* For use to indent debugging output */

#define SP "                   "

/*************************************************

*      Code parameters and static tables         *

*************************************************/

/* These are offsets that are used to turn the OP_TYPESTAR and friends opcodes

into others, under special conditions. A gap of 20 between the blocks should be

enough. The resulting opcodes don't have to be less than 256 because they are

never stored, so we push them well clear of the normal opcodes. */

#define OP_PROP_EXTRA       300

#define OP_EXTUNI_EXTRA     320

#define OP_ANYNL_EXTRA      340

#define OP_HSPACE_EXTRA     360

#define OP_VSPACE_EXTRA     380

/* This table identifies those opcodes that are followed immediately by a

character that is to be tested in some way. This makes is possible to

centralize the loading of these characters. In the case of Type * etc, the

"character" is the opcode for \D, \d, \S, \s, \W, or \w, which will always be a

small value. ***NOTE*** If the start of this table is modified, the two tables

that follow must also be modified. */

static const uschar coptable[] = {

  0,                             /* End                                    */

  0, 0, 0, 0, 0,                 /* \A, \G, \K, \B, \b                     */

  0, 0, 0, 0, 0, 0,              /* \D, \d, \S, \s, \W, \w                 */

  0, 0, 0,                       /* Any, AllAny, Anybyte                   */

  0, 0, 0,                       /* NOTPROP, PROP, EXTUNI                  */

  0, 0, 0, 0, 0,                 /* \R, \H, \h, \V, \v                     */

  0, 0, 0, 0, 0,                 /* \Z, \z, Opt, ^, $                      */

  1,                             /* Char                                   */

  1,                             /* Charnc                                 */

  1,                             /* not                                    */

  /* Positive single-char repeats                                          */

  1, 1, 1, 1, 1, 1,              /* *, *?, +, +?, ?, ??                    */

  3, 3, 3,                       /* upto, minupto, exact                   */

  1, 1, 1, 3,                    /* *+, ++, ?+, upto+                      */

  /* Negative single-char repeats - only for chars < 256                   */

  1, 1, 1, 1, 1, 1,              /* NOT *, *?, +, +?, ?, ??                */

  3, 3, 3,                       /* NOT upto, minupto, exact               */

  1, 1, 1, 3,                    /* NOT *+, ++, ?+, updo+                  */

  /* Positive type repeats                                                 */

  1, 1, 1, 1, 1, 1,              /* Type *, *?, +, +?, ?, ??               */

  3, 3, 3,                       /* Type upto, minupto, exact              */

  1, 1, 1, 3,                    /* Type *+, ++, ?+, upto+                 */

  /* Character class & ref repeats                                         */

  0, 0, 0, 0, 0, 0,              /* *, *?, +, +?, ?, ??                    */

  0, 0,                          /* CRRANGE, CRMINRANGE                    */

  0,                             /* CLASS                                  */

  0,                             /* NCLASS                                 */

  0,                             /* XCLASS - variable length               */

  0,                             /* REF                                    */

  0,                             /* RECURSE                                */

  0,                             /* CALLOUT                                */

  0,                             /* Alt                                    */

  0,                             /* Ket                                    */

  0,                             /* KetRmax                                */

  0,                             /* KetRmin                                */

  0,                             /* Assert                                 */

  0,                             /* Assert not                             */

  0,                             /* Assert behind                          */

  0,                             /* Assert behind not                      */

  0,                             /* Reverse                                */

  0, 0, 0, 0,                    /* ONCE, BRA, CBRA, COND                  */

  0, 0, 0,                       /* SBRA, SCBRA, SCOND                     */

  0,                             /* CREF                                   */

  0,                             /* RREF                                   */

  0,                             /* DEF                                    */

  0, 0,                          /* BRAZERO, BRAMINZERO                    */

  0, 0, 0, 0,                    /* PRUNE, SKIP, THEN, COMMIT              */

  0, 0, 0                        /* FAIL, ACCEPT, SKIPZERO                 */

};

/* These 2 tables allow for compact code for testing for \D, \d, \S, \s, \W,

and \w */

static const uschar toptable1[] = {

  0, 0, 0, 0, 0, 0,

  ctype_digit, ctype_digit,

  ctype_space, ctype_space,

  ctype_word,  ctype_word,

  0, 0                            /* OP_ANY, OP_ALLANY */

};

static const uschar toptable2[] = {

  0, 0, 0, 0, 0, 0,

  ctype_digit, 0,

  ctype_space, 0,

  ctype_word,  0,

  1, 1                            /* OP_ANY, OP_ALLANY */

};

/* Structure for holding data about a particular state, which is in effect the

current data for an active path through the match tree. It must consist

entirely of ints because the working vector we are passed, and which we put

these structures in, is a vector of ints. */

typedef struct stateblock {

  int offset;                     /* Offset to opcode */

  int count;                      /* Count for repeats */

  int ims;                        /* ims flag bits */

  int data;                       /* Some use extra data */

} stateblock;

#define INTS_PER_STATEBLOCK  (sizeof(stateblock)/sizeof(int))

#ifdef DEBUG

/*************************************************

*             Print character string             *

*************************************************/

/* Character string printing function for debugging.

Arguments:

  p            points to string

  length       number of bytes

  f            where to print

Returns:       nothing

*/

static void

pchars(unsigned char *p, int length, FILE *f)

{

int c;

while (length-- > 0)

  {

  if (isprint(c = *(p++)))

    fprintf(f, "%c", c);

  else

    fprintf(f, "\\x%02x", c);

  }

}

#endif

/*************************************************

*    Execute a Regular Expression - DFA engine   *

*************************************************/

/* This internal function applies a compiled pattern to a subject string,

starting at a given point, using a DFA engine. This function is called from the

external one, possibly multiple times if the pattern is not anchored. The

function calls itself recursively for some kinds of subpattern.

Arguments:

  md                the match_data block with fixed information

  this_start_code   the opening bracket of this subexpression's code

  current_subject   where we currently are in the subject string

  start_offset      start offset in the subject string

  offsets           vector to contain the matching string offsets

  offsetcount       size of same

  workspace         vector of workspace

  wscount           size of same

  ims               the current ims flags

  rlevel            function call recursion level

  recursing         regex recursive call level

Returns:            > 0 => number of match offset pairs placed in offsets

                    = 0 => offsets overflowed; longest matches are present

                     -1 => failed to match

                   < -1 => some kind of unexpected problem

The following macros are used for adding states to the two state vectors (one

for the current character, one for the following character). */

#define ADD_ACTIVE(x,y) \

  if (active_count++ < wscount) \

    { \

    next_active_state->offset = (x); \

    next_active_state->count  = (y); \

    next_active_state->ims    = ims; \

    next_active_state++; \

    DPRINTF(("%.*sADD_ACTIVE(%d,%d)\n", rlevel*2-2, SP, (x), (y))); \

    } \

  else return PCRE_ERROR_DFA_WSSIZE

#define ADD_ACTIVE_DATA(x,y,z) \

  if (active_count++ < wscount) \

    { \

    next_active_state->offset = (x); \

    next_active_state->count  = (y); \

    next_active_state->ims    = ims; \

    next_active_state->data   = (z); \

    next_active_state++; \

    DPRINTF(("%.*sADD_ACTIVE_DATA(%d,%d,%d)\n", rlevel*2-2, SP, (x), (y), (z))); \

    } \

  else return PCRE_ERROR_DFA_WSSIZE

#define ADD_NEW(x,y) \

  if (new_count++ < wscount) \

    { \

    next_new_state->offset = (x); \

    next_new_state->count  = (y); \

    next_new_state->ims    = ims; \

    next_new_state++; \

    DPRINTF(("%.*sADD_NEW(%d,%d)\n", rlevel*2-2, SP, (x), (y))); \

    } \

  else return PCRE_ERROR_DFA_WSSIZE

#define ADD_NEW_DATA(x,y,z) \

  if (new_count++ < wscount) \

    { \

    next_new_state->offset = (x); \

    next_new_state->count  = (y); \

    next_new_state->ims    = ims; \

    next_new_state->data   = (z); \

    next_new_state++; \

    DPRINTF(("%.*sADD_NEW_DATA(%d,%d,%d)\n", rlevel*2-2, SP, (x), (y), (z))); \

    } \

  else return PCRE_ERROR_DFA_WSSIZE

/* And now, here is the code */

static int

internal_dfa_exec(

  dfa_match_data *md,

  const uschar *this_start_code,

  const uschar *current_subject,

  int start_offset,

  int *offsets,

  int offsetcount,

  int *workspace,

  int wscount,

  int ims,

  int  rlevel,

  int  recursing)

{

stateblock *active_states, *new_states, *temp_states;

stateblock *next_active_state, *next_new_state;

const uschar *ctypes, *lcc, *fcc;

const uschar *ptr;

const uschar *end_code, *first_op;

int active_count, new_count, match_count;

/* Some fields in the md block are frequently referenced, so we load them into

independent variables in the hope that this will perform better. */

const uschar *start_subject = md->start_subject;

const uschar *end_subject = md->end_subject;

const uschar *start_code = md->start_code;

#ifdef SUPPORT_UTF8

BOOL utf8 = (md->poptions & PCRE_UTF8) != 0;

#else

BOOL utf8 = FALSE;

#endif

rlevel++;

offsetcount &= (-2);

wscount -= 2;

wscount = (wscount - (wscount % (INTS_PER_STATEBLOCK * 2))) /

          (2 * INTS_PER_STATEBLOCK);

DPRINTF(("\n%.*s---------------------\n"

  "%.*sCall to internal_dfa_exec f=%d r=%d\n",

  rlevel*2-2, SP, rlevel*2-2, SP, rlevel, recursing));

ctypes = md->tables + ctypes_offset;

lcc = md->tables + lcc_offset;

fcc = md->tables + fcc_offset;

match_count = PCRE_ERROR_NOMATCH;   /* A negative number */

active_states = (stateblock *)(workspace + 2);

next_new_state = new_states = active_states + wscount;

new_count = 0;

first_op = this_start_code + 1 + LINK_SIZE +

  ((*this_start_code == OP_CBRA || *this_start_code == OP_SCBRA)? 2:0);

/* The first thing in any (sub) pattern is a bracket of some sort. Push all

the alternative states onto the list, and find out where the end is. This

makes is possible to use this function recursively, when we want to stop at a

matching internal ket rather than at the end.

If the first opcode in the first alternative is OP_REVERSE, we are dealing with

a backward assertion. In that case, we have to find out the maximum amount to

move back, and set up each alternative appropriately. */

if (*first_op == OP_REVERSE)

  {

  int max_back = 0;

  int gone_back;

  end_code = this_start_code;

  do

    {

    int back = GET(end_code, 2+LINK_SIZE);

    if (back > max_back) max_back = back;

    end_code += GET(end_code, 1);

    }

  while (*end_code == OP_ALT);

  /* If we can't go back the amount required for the longest lookbehind

  pattern, go back as far as we can; some alternatives may still be viable. */

#ifdef SUPPORT_UTF8

  /* In character mode we have to step back character by character */

  if (utf8)

    {

    for (gone_back = 0; gone_back < max_back; gone_back++)

      {

      if (current_subject <= start_subject) break;

      current_subject--;

      while (current_subject > start_subject &&

             (*current_subject & 0xc0) == 0x80)

        current_subject--;

      }

    }

  else

#endif

  /* In byte-mode we can do this quickly. */

    {

    gone_back = (current_subject - max_back < start_subject)?

      current_subject - start_subject : max_back;

    current_subject -= gone_back;

    }

  /* Now we can process the individual branches. */

  end_code = this_start_code;

  do

    {

    int back = GET(end_code, 2+LINK_SIZE);

    if (back <= gone_back)

      {

      int bstate = end_code - start_code + 2 + 2*LINK_SIZE;

      ADD_NEW_DATA(-bstate, 0, gone_back - back);

      }

    end_code += GET(end_code, 1);

    }

  while (*end_code == OP_ALT);

 }

/* This is the code for a "normal" subpattern (not a backward assertion). The

start of a whole pattern is always one of these. If we are at the top level,

we may be asked to restart matching from the same point that we reached for a

previous partial match. We still have to scan through the top-level branches to

find the end state. */

else

  {

  end_code = this_start_code;

  /* Restarting */

  if (rlevel == 1 && (md->moptions & PCRE_DFA_RESTART) != 0)

    {

    do { end_code += GET(end_code, 1); } while (*end_code == OP_ALT);

    new_count = workspace[1];

    if (!workspace[0])

      memcpy(new_states, active_states, new_count * sizeof(stateblock));

    }

  /* Not restarting */

  else

    {

    int length = 1 + LINK_SIZE +

      ((*this_start_code == OP_CBRA || *this_start_code == OP_SCBRA)? 2:0);

    do

      {

      ADD_NEW(end_code - start_code + length, 0);

      end_code += GET(end_code, 1);

      length = 1 + LINK_SIZE;

      }

    while (*end_code == OP_ALT);

    }

  }

workspace[0] = 0;    /* Bit indicating which vector is current */

DPRINTF(("%.*sEnd state = %d\n", rlevel*2-2, SP, end_code - start_code));

/* Loop for scanning the subject */

ptr = current_subject;

for (;;)

  {

  int i, j;

  int clen, dlen;

  unsigned int c, d;

  /* Make the new state list into the active state list and empty the

  new state list. */

  temp_states = active_states;

  active_states = new_states;

  new_states = temp_states;

  active_count = new_count;

  new_count = 0;

  workspace[0] ^= 1;              /* Remember for the restarting feature */

  workspace[1] = active_count;

#ifdef DEBUG

  printf("%.*sNext character: rest of subject = \"", rlevel*2-2, SP);

  pchars((uschar *)ptr, strlen((char *)ptr), stdout);

  printf("\"\n");

  printf("%.*sActive states: ", rlevel*2-2, SP);

  for (i = 0; i < active_count; i++)

    printf("%d/%d ", active_states[i].offset, active_states[i].count);

  printf("\n");

#endif

  /* Set the pointers for adding new states */

  next_active_state = active_states + active_count;

  next_new_state = new_states;

  /* Load the current character from the subject outside the loop, as many

  different states may want to look at it, and we assume that at least one

  will. */

  if (ptr < end_subject)

    {

    clen = 1;        /* Number of bytes in the character */

#ifdef SUPPORT_UTF8

    if (utf8) { GETCHARLEN(c, ptr, clen); } else

#endif  /* SUPPORT_UTF8 */

    c = *ptr;

    }

  else

    {

    clen = 0;        /* This indicates the end of the subject */

    c = NOTACHAR;    /* This value should never actually be used */

    }

  /* Scan up the active states and act on each one. The result of an action

  may be to add more states to the currently active list (e.g. on hitting a

  parenthesis) or it may be to put states on the new list, for considering

  when we move the character pointer on. */

  for (i = 0; i < active_count; i++)

    {

    stateblock *current_state = active_states + i;

    const uschar *code;

    int state_offset = current_state->offset;

    int count, codevalue, rrc;

#ifdef DEBUG

    printf ("%.*sProcessing state %d c=", rlevel*2-2, SP, state_offset);

    if (clen == 0) printf("EOL\n");

      else if (c > 32 && c < 127) printf("'%c'\n", c);

        else printf("0x%02x\n", c);

#endif

    /* This variable is referred to implicity in the ADD_xxx macros. */

    ims = current_state->ims;

    /* A negative offset is a special case meaning "hold off going to this

    (negated) state until the number of characters in the data field have

    been skipped". */

    if (state_offset < 0)

      {

      if (current_state->data > 0)

        {

        DPRINTF(("%.*sSkipping this character\n", rlevel*2-2, SP));

        ADD_NEW_DATA(state_offset, current_state->count,

          current_state->data - 1);

        continue;

        }

      else

        {

        current_state->offset = state_offset = -state_offset;

        }

      }

    /* Check for a duplicate state with the same count, and skip if found. */

    for (j = 0; j < i; j++)

      {

      if (active_states[j].offset == state_offset &&

          active_states[j].count == current_state->count)

        {

        DPRINTF(("%.*sDuplicate state: skipped\n", rlevel*2-2, SP));

        goto NEXT_ACTIVE_STATE;

        }

      }

    /* The state offset is the offset to the opcode */

    code = start_code + state_offset;

    codevalue = *code;

    /* If this opcode is followed by an inline character, load it. It is

    tempting to test for the presence of a subject character here, but that

    is wrong, because sometimes zero repetitions of the subject are

    permitted.

    We also use this mechanism for opcodes such as OP_TYPEPLUS that take an

    argument that is not a data character - but is always one byte long. We

    have to take special action to deal with  \P, \p, \H, \h, \V, \v and \X in

    this case. To keep the other cases fast, convert these ones to new opcodes.

    */

    if (coptable[codevalue] > 0)

      {

      dlen = 1;

#ifdef SUPPORT_UTF8

      if (utf8) { GETCHARLEN(d, (code + coptable[codevalue]), dlen); } else

#endif  /* SUPPORT_UTF8 */

      d = code[coptable[codevalue]];

      if (codevalue >= OP_TYPESTAR)

        {

        switch(d)

          {

          case OP_ANYBYTE: return PCRE_ERROR_DFA_UITEM;

          case OP_NOTPROP:

          case OP_PROP: codevalue += OP_PROP_EXTRA; break;

          case OP_ANYNL: codevalue += OP_ANYNL_EXTRA; break;

          case OP_EXTUNI: codevalue += OP_EXTUNI_EXTRA; break;

          case OP_NOT_HSPACE:

          case OP_HSPACE: codevalue += OP_HSPACE_EXTRA; break;

          case OP_NOT_VSPACE:

          case OP_VSPACE: codevalue += OP_VSPACE_EXTRA; break;

          default: break;

          }

        }

      }

    else

      {

      dlen = 0;         /* Not strictly necessary, but compilers moan */

      d = NOTACHAR;     /* if these variables are not set. */

      }

    /* Now process the individual opcodes */

    switch (codevalue)

      {

/* ========================================================================== */

      /* Reached a closing bracket. If not at the end of the pattern, carry

      on with the next opcode. Otherwise, unless we have an empty string and

      PCRE_NOTEMPTY is set, save the match data, shifting up all previous

      matches so we always have the longest first. */

      case OP_KET:

      case OP_KETRMIN:

      case OP_KETRMAX:

      if (code != end_code)

        {

        ADD_ACTIVE(state_offset + 1 + LINK_SIZE, 0);

        if (codevalue != OP_KET)

          {

          ADD_ACTIVE(state_offset - GET(code, 1), 0);

          }

        }

      else if (ptr > current_subject || (md->moptions & PCRE_NOTEMPTY) == 0)

        {

        if (match_count < 0) match_count = (offsetcount >= 2)? 1 : 0;

          else if (match_count > 0 && ++match_count * 2 >= offsetcount)

            match_count = 0;