Imported new improved free regex implementation. Added (again)
authorPekka Riikonen <priikone@silcnet.org>
Wed, 2 Jan 2008 20:39:42 +0000 (20:39 +0000)
committerPekka Riikonen <priikone@silcnet.org>
Wed, 2 Jan 2008 20:39:42 +0000 (20:39 +0000)
support for SILC_REGEX_NOTBOL and SILC_REGEX_NOTEOLF flags.
Added also support for POSIX bounded repeat expression (a{n,m})
to make it more POSIX compliant.  Added SilcStack support to
the regex too.

lib/silcutil/silcerrno.c
lib/silcutil/silcerrno.h
lib/silcutil/silcregex.c
lib/silcutil/silcregex.h
lib/silcutil/tests/test_silcregex.c

index fae62b500f2a2786137443cba72a70f8a9a0af88..ff809c5f87831e35296623988c77986bd627c72f 100644 (file)
@@ -4,7 +4,7 @@
 
   Author: Pekka Riikonen <priikone@silcnet.org>
 
-  Copyright (C) 2007 Pekka Riikonen
+  Copyright (C) 2007 - 2008 Pekka Riikonen
 
   This program is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -432,6 +432,8 @@ const char *silc_errno_strings[] =
   "Bad match register number",
   "Badly placed special character",
   "Regular expression too complex",
+  "Bad regular expression opcode",
+  "Bad repeat value",
 };
 
 /* Map error to string */
index c13749ac745ca30854c5ee19d53dbb3aa629371a..f8e8f15f5649556ec79d9ac9b9bad3b787481b4d 100644 (file)
@@ -4,7 +4,7 @@
 
   Author: Pekka Riikonen <priikone@silcnet.org>
 
-  Copyright (C) 2007 Pekka Riikonen
+  Copyright (C) 2007 - 2008 Pekka Riikonen
 
   This program is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -132,6 +132,8 @@ typedef enum {
   SILC_ERR_REGEX_REG                   = 102, /* Bad register number */
   SILC_ERR_REGEX_SPECIAL               = 103, /* Unmatched special character */
   SILC_ERR_REGEX_TOO_COMPLEX           = 104, /* Too complex expression */
+  SILC_ERR_REGEX_OPCODE                = 105, /* Bad regex opcode */
+  SILC_ERR_REGEX_REPEAT                = 106, /* Bad repeat value */
 
   SILC_ERR_MAX,
 } SilcResult;
index 1412f20d0dd192394c3af0927b10341abe627931..cdd6ecfaa5431545c7f6a354e5faf2df088eb6af 100644 (file)
@@ -1,41 +1,42 @@
-/*
-
-  regexpr.c
-
-  Author: Tatu Ylonen <ylo@ngs.fi>
-
-  Copyright (c) 1991 Tatu Ylonen, Espoo, Finland
-
-  Permission to use, copy, modify, distribute, and sell this software
-  and its documentation is hereby granted without fee, provided that the
-  above copyright notice appears in all source code copies, the name of
-  Tatu Ylonen is not used to advertise products containing this software
-  or a derivation thereof, and all modified versions are clearly marked
-  as such.
-
-  This software is provided "as is" without express or implied warranty.
-
-  Created: Thu Sep 26 17:14:05 1991 ylo
-  Last modified: Sun Mar 29 16:47:31 1992 ylo
-
-  This code draws many ideas from the regular expression packages by
-  Henry Spencer of the University of Toronto and Richard Stallman of the
-  Free Software Foundation.
-
-  Emacs-specific code and syntax table code is almost directly borrowed
-  from GNU regexp.
+/* regexpr.c
+ *
+ * Author: Tatu Ylonen <ylo@ngs.fi>
+ *
+ * Copyright (c) 1991 Tatu Ylonen, Espoo, Finland
+ *
+ * Permission to use, copy, modify, distribute, and sell this software
+ * and its documentation for any purpose is hereby granted without
+ * fee, provided that the above copyright notice appear in all copies.
+ * This software is provided "as is" without express or implied
+ * warranty.
+ *
+ * Created: Thu Sep 26 17:14:05 1991 ylo
+ * Last modified: Mon Nov  4 17:06:48 1991 ylo
+ * Ported to Think C: 19 Jan 1992 guido@cwi.nl
+ *
+ * This code draws many ideas from the regular expression packages by
+ * Henry Spencer of the University of Toronto and Richard Stallman of
+ * the Free Software Foundation.
+ *
+ * Bugs fixed and lots of reorganization by Jeffrey C. Ollie, April
+ * 1997 Thanks for bug reports and ideas from Andrew Kuchling, Tim
+ * Peters, Guido van Rossum, Ka-Ping Yee, Sjoerd Mullender, and
+ * probably one or two others that I'm forgetting.
+ *
+ */
 
+/*
   The SILC Regex API and modifications by Pekka Riikonen, under the same
-  license as the original code.
+  license as the original code.  We've added following features:
 
+  - RE_NOTBOL            - bol fails to match (conforming POSIX)
+  - RE_NOTEOL            - eol fails to match (conforming POSIX)
+  - RE_REPEAT a{n,m}     - bounded repeat (conforming POSIX)
+  - SilStack support     - compile without real memory allocations
 */
 
 #include "silc.h"
 
-/* Modified for use in SILC Runtime Toolkit.  I think we have disabled many
-   features we could use, for the sake of simple API, which we may want to
-   extend later.  But, we've added RE_NOTBOL and RE_NOTEOL. */
-
 #define RE_NREGS       128     /* number of registers available */
 
 /* bit definitions for syntax */
@@ -49,6 +50,8 @@
 #define RE_NO_GNU_EXTENSIONS   128   /* no gnu extensions */
 #define RE_NOTBOL              256   /* bol fails to match */
 #define RE_NOTEOL              512   /* eol fails to match */
+#define RE_REPEAT             1024   /* bounded repeat, must be quoted without
+                                       RE_NO_BK_VBAR */
 
 /* definitions for some common regexp styles */
 #define RE_SYNTAX_AWK  (RE_NO_BK_PARENS|RE_NO_BK_VBAR|RE_CONTEXT_INDEP_OPS)
 #define RE_SYNTAX_GREP (RE_BK_PLUS_QM|RE_NEWLINE_OR)
 #define RE_SYNTAX_EMACS        0
 
+#define Sword       1
+#define Swhitespace 2
+#define Sdigit      4
+#define Soctaldigit 8
+#define Shexdigit   16
+
 /* Registers */
 typedef struct re_registers {
   int start[RE_NREGS];         /* start offset of region */
   int end[RE_NREGS];           /* end offset of region */
 } *regexp_registers_t;
 
-int re_set_syntax(int syntax);
-/* This sets the syntax to use and returns the previous syntax.  The
-   syntax is specified by a bit mask of the above defined bits. */
-
-SilcResult re_compile_pattern(char *regex, int regex_size, SilcRegex compiled);
-/* This compiles the regexp (given in regex and length in regex_size).
-   This returns NULL if the regexp compiled successfully, and an error
-   message if an error was encountered.  The buffer field must be
-   initialized to a memory area allocated by malloc (or to NULL) before
-   use, and the allocated field must be set to its length (or 0 if buffer is
-   NULL).  Also, the translate field must be set to point to a valid
-   translation table, or NULL if it is not used. */
-
-int re_match(SilcRegex compiled, char *string, int size, int pos,
-            regexp_registers_t regs, unsigned int flags);
-/* This tries to match the regexp against the string.  This returns the
-   length of the matched portion, or -1 if the pattern could not be
-   matched and -2 if an error (such as failure stack overflow) is
-   encountered. */
-
-int re_match_2(SilcRegex compiled, char *string1, int size1,
-              char *string2, int size2, int pos, regexp_registers_t regs,
-              int mstop, unsigned int flags);
-/* This tries to match the regexp to the concatenation of string1 and
-   string2.  This returns the length of the matched portion, or -1 if the
-   pattern could not be matched and -2 if an error (such as failure stack
-   overflow) is encountered. */
-
-int re_search(SilcRegex compiled, char *string, int size, int startpos,
-             int range, regexp_registers_t regs, unsigned int flags);
-/* This rearches for a substring matching the regexp.  This returns the first
-   index at which a match is found.  range specifies at how many positions to
-   try matching; positive values indicate searching forwards, and negative
-   values indicate searching backwards.  mstop specifies the offset beyond
-   which a match must not go.  This returns -1 if no match is found, and
-   -2 if an error (such as failure stack overflow) is encountered. */
-
-int re_search_2(SilcRegex compiled, char *string1, int size1,
-               char *string2, int size2, int startpos, int range,
-               regexp_registers_t regs, int mstop, unsigned int flags);
-/* This is like re_search, but search from the concatenation of string1 and
-   string2.  */
-
-void re_compile_fastmap(SilcRegex compiled);
-/* This computes the fastmap for the regexp.  For this to have any effect,
-   the calling program must have initialized the fastmap field to point
-   to an array of 256 characters. */
-
-#define MACRO_BEGIN do {
-#define MACRO_END } while (0)
+/* The original code blithely assumed that sizeof(short) == 2.  Not
+ * always true.  Original instances of "(short)x" were replaced by
+ * SHORT(x), where SHORT is #defined below.  */
+
+#define SHORT(x) ((x) & 0x8000 ? (x) - 0x10000 : (x))
+
+/* The stack implementation is taken from an idea by Andrew Kuchling.
+ * It's a doubly linked list of arrays. The advantages of this over a
+ * simple linked list are that the number of mallocs required are
+ * reduced. It also makes it possible to statically allocate enough
+ * space so that small patterns don't ever need to call malloc.
+ *
+ * The advantages over a single array is that is periodically
+ * realloced when more space is needed is that we avoid ever copying
+ * the stack. */
+
+/* item_t is the basic stack element.  Defined as a union of
+ * structures so that both registers, failure points, and counters can
+ * be pushed/popped from the stack.  There's nothing built into the
+ * item to keep track of whether a certain stack item is a register, a
+ * failure point, or a counter. */
+
+typedef union item_t
+{
+  struct
+  {
+    int num;
+    int level;
+    unsigned char *start;
+    unsigned char *end;
+  } reg;
+  struct
+  {
+    int count;
+    int level;
+    int phantom;
+    unsigned char *code;
+    unsigned char *text;
+  } fail;
+  struct
+  {
+    int num;
+    int level;
+    int count;
+  } cntr;
+} item_t;
+
+#define STACK_PAGE_SIZE 256
+#define NUM_REGISTERS 256
+
+/* A 'page' of stack items. */
+
+typedef struct item_page_t
+{
+  item_t items[STACK_PAGE_SIZE];
+  struct item_page_t *prev;
+  struct item_page_t *next;
+} item_page_t;
+
+typedef struct match_state
+{
+  /* The number of registers that have been pushed onto the stack
+   * since the last failure point. */
+
+  int count;
+
+  /* Used to control when registers need to be pushed onto the
+   * stack. */
+
+  int level;
+
+  /* The number of failure points on the stack. */
+
+  int point;
+
+  /* Storage for the registers.  Each register consists of two
+   * pointers to characters.  So register N is represented as
+   * start[N] and end[N].  The pointers must be converted to
+   * offsets from the beginning of the string before returning the
+   * registers to the calling program. */
+
+  unsigned char *start[NUM_REGISTERS];
+  unsigned char *end[NUM_REGISTERS];
+
+  /* Keeps track of whether a register has changed recently. */
+
+  int changed[NUM_REGISTERS];
+
+  /* Structure to encapsulate the stack. */
+  struct
+  {
+    /* index into the current page.  If index == 0 and you need
+     * to pop an item, move to the previous page and set index
+     * = STACK_PAGE_SIZE - 1.  Otherwise decrement index to
+     * push a page. If index == STACK_PAGE_SIZE and you need
+     * to push a page move to the next page and set index =
+     * 0. If there is no new next page, allocate a new page
+     * and link it in. Otherwise, increment index to push a
+     * page. */
+
+    int index;
+    item_page_t *current; /* Pointer to the current page. */
+    item_page_t first; /* First page is statically allocated. */
+  } stack;
+} match_state;
+
+/* Initialize a state object */
+
+/* #define NEW_STATE(state) \ */
+/* memset(&state, 0, (void *)(&state.stack) - (void *)(&state)); \ */
+/* state.stack.current = &state.stack.first; \ */
+/* state.stack.first.prev = NULL; \ */
+/* state.stack.first.next = NULL; \ */
+/* state.stack.index = 0; \ */
+/* state.level = 1 */
+
+#define NEW_STATE(state, nregs)                        \
+  {                                            \
+    int i;                                     \
+    for (i = 0; i < nregs; i++)                        \
+      {                                                \
+       state.start[i] = NULL;                  \
+       state.end[i] = NULL;                    \
+       state.changed[i] = 0;                   \
+      }                                                \
+    state.stack.current = &state.stack.first;  \
+    state.stack.first.prev = NULL;             \
+    state.stack.first.next = NULL;             \
+    state.stack.index = 0;                     \
+    state.level = 1;                           \
+    state.count = 0;                           \
+    state.level = 0;                           \
+    state.point = 0;                           \
+  }
+
+/* Free any memory that might have been malloc'd */
+
+#define FREE_STATE(state)                                      \
+  while(state.stack.first.next != NULL)                                \
+    {                                                          \
+      state.stack.current = state.stack.first.next;            \
+      state.stack.first.next = state.stack.current->next;      \
+      silc_sfree(bufp->rstack, state.stack.current);           \
+    }
+
+/* Discard the top 'count' stack items. */
+
+#define STACK_DISCARD(stack, count, on_error)  \
+  stack.index -= count;                                \
+  while (stack.index < 0)                      \
+    {                                          \
+      if (stack.current->prev == NULL)         \
+       on_error;                               \
+      stack.current = stack.current->prev;     \
+      stack.index += STACK_PAGE_SIZE;          \
+    }
+
+/* Store a pointer to the previous item on the stack. Used to pop an
+ * item off of the stack. */
+
+#define STACK_PREV(stack, top, on_error)       \
+  if (stack.index == 0)                                \
+    {                                          \
+      if (stack.current->prev == NULL)         \
+       on_error;                               \
+      stack.current = stack.current->prev;     \
+      stack.index = STACK_PAGE_SIZE - 1;       \
+    }                                          \
+  else                                         \
+    {                                          \
+      stack.index--;                           \
+    }                                          \
+  top = &(stack.current->items[stack.index])
+
+/* Store a pointer to the next item on the stack. Used to push an item
+ * on to the stack. */
+
+#define STACK_NEXT(stack, top, on_error)                               \
+  if (stack.index == STACK_PAGE_SIZE)                                  \
+    {                                                                  \
+      if (stack.current->next == NULL)                                 \
+       {                                                               \
+         stack.current->next =                                         \
+           (item_page_t *)silc_smalloc(bufp->rstack, sizeof(item_page_t)); \
+         if (stack.current->next == NULL)                              \
+           on_error;                                                   \
+         stack.current->next->prev = stack.current;                    \
+         stack.current->next->next = NULL;                             \
+       }                                                               \
+      stack.current = stack.current->next;                             \
+      stack.index = 0;                                                 \
+    }                                                                  \
+  top = &(stack.current->items[stack.index++])
+
+/* Store a pointer to the item that is 'count' items back in the
+ * stack. STACK_BACK(stack, top, 1, on_error) is equivalent to
+ * STACK_TOP(stack, top, on_error).  */
+
+#define STACK_BACK(stack, top, count, on_error) \
+  {                                            \
+    int index;                                 \
+    item_page_t *current;                      \
+    current = stack.current;                   \
+    index = stack.index - (count);             \
+    while (index < 0)                          \
+      {                                                \
+       if (current->prev == NULL)              \
+         on_error;                             \
+       current = current->prev;                \
+       index += STACK_PAGE_SIZE;               \
+      }                                                \
+    top = &(current->items[index]);            \
+  }
+
+/* Store a pointer to the top item on the stack. Execute the
+ * 'on_error' code if there are no items on the stack. */
+
+#define STACK_TOP(stack, top, on_error)                                \
+  if (stack.index == 0)                                                \
+    {                                                          \
+      if (stack.current->prev == NULL)                         \
+       on_error;                                               \
+      top = &(stack.current->prev->items[STACK_PAGE_SIZE - 1]); \
+    }                                                          \
+  else                                                         \
+    {                                                          \
+      top = &(stack.current->items[stack.index - 1]);          \
+    }
+
+/* Test to see if the stack is empty */
+
+#define STACK_EMPTY(stack) ((stack.index == 0) &&              \
+                           (stack.current->prev == NULL))
+
+/* Return the start of register 'reg' */
+
+#define GET_REG_START(state, reg) (state.start[reg])
+
+/* Return the end of register 'reg' */
+
+#define GET_REG_END(state, reg) (state.end[reg])
+
+/* Set the start of register 'reg'. If the state of the register needs
+ * saving, push it on the stack. */
+
+#define SET_REG_START(state, reg, text, on_error)      \
+  if(state.changed[reg] < state.level)                 \
+    {                                                  \
+      item_t *item;                                    \
+      STACK_NEXT(state.stack, item, on_error);         \
+      item->reg.num = reg;                             \
+      item->reg.start = state.start[reg];              \
+      item->reg.end = state.end[reg];                  \
+      item->reg.level = state.changed[reg];            \
+      state.changed[reg] = state.level;                        \
+      state.count++;                                   \
+    }                                                  \
+  state.start[reg] = text
+
+/* Set the end of register 'reg'. If the state of the register needs
+ * saving, push it on the stack. */
+
+#define SET_REG_END(state, reg, text, on_error) \
+  if(state.changed[reg] < state.level)         \
+    {                                          \
+      item_t *item;                            \
+      STACK_NEXT(state.stack, item, on_error); \
+      item->reg.num = reg;                     \
+      item->reg.start = state.start[reg];      \
+      item->reg.end = state.end[reg];          \
+      item->reg.level = state.changed[reg];    \
+      state.changed[reg] = state.level;                \
+      state.count++;                           \
+    }                                          \
+  state.end[reg] = text
+
+#define PUSH_FAILURE(state, xcode, xtext, on_error)    \
+  {                                                    \
+    item_t *item;                                      \
+    STACK_NEXT(state.stack, item, on_error);           \
+    item->fail.code = xcode;                           \
+    item->fail.text = xtext;                           \
+    item->fail.count = state.count;                    \
+    item->fail.level = state.level;                    \
+    item->fail.phantom = 0;                            \
+    state.count = 0;                                   \
+    state.level++;                                     \
+    state.point++;                                     \
+  }
+
+/* Update the last failure point with a new position in the text. */
+
+#define UPDATE_FAILURE(state, xtext, on_error)                 \
+  {                                                            \
+    item_t *item;                                              \
+    STACK_BACK(state.stack, item, state.count + 1, on_error);  \
+    if (!item->fail.phantom)                                   \
+      {                                                                \
+       item_t *item2;                                          \
+       STACK_NEXT(state.stack, item2, on_error);               \
+       item2->fail.code = item->fail.code;                     \
+       item2->fail.text = xtext;                               \
+       item2->fail.count = state.count;                        \
+       item2->fail.level = state.level;                        \
+       item2->fail.phantom = 1;                                \
+       state.count = 0;                                        \
+       state.level++;                                          \
+       state.point++;                                          \
+      }                                                                \
+    else                                                       \
+      {                                                                \
+       STACK_DISCARD(state.stack, state.count, on_error);      \
+       STACK_TOP(state.stack, item, on_error);                 \
+       item->fail.text = xtext;                                \
+       state.count = 0;                                        \
+       state.level++;                                          \
+      }                                                                \
+  }
+
+#define POP_FAILURE(state, xcode, xtext, on_empty, on_error)   \
+  {                                                            \
+    item_t *item;                                              \
+    do                                                         \
+      {                                                                \
+       while(state.count > 0)                                  \
+         {                                                     \
+           STACK_PREV(state.stack, item, on_error);            \
+           state.start[item->reg.num] = item->reg.start;       \
+           state.end[item->reg.num] = item->reg.end;           \
+           state.changed[item->reg.num] = item->reg.level;     \
+           state.count--;                                      \
+         }                                                     \
+       STACK_PREV(state.stack, item, on_empty);                \
+       xcode = item->fail.code;                                \
+       xtext = item->fail.text;                                \
+       state.count = item->fail.count;                         \
+       state.level = item->fail.level;                         \
+       state.point--;                                          \
+      }                                                                \
+    while (item->fail.text == NULL);                           \
+  }
 
 enum regexp_compiled_ops /* opcodes for compiled regexp */
 {
@@ -135,11 +435,9 @@ enum regexp_compiled_ops /* opcodes for compiled regexp */
   Cwordend,            /* match at end of word */
   Cwordbound,          /* match if at word boundary */
   Cnotwordbound,       /* match if not at word boundary */
-#ifdef emacs
-  Cemacs_at_dot,       /* emacs only: matches at dot */
-#endif /* emacs */
   Csyntaxspec,         /* matches syntax code (1 byte follows) */
-  Cnotsyntaxspec       /* matches if syntax code does not match (1 byte foll)*/
+  Cnotsyntaxspec,      /* matches if syntax code does not match (1 byte foll)*/
+  Crepeat1,
 };
 
 enum regexp_syntax_op  /* syntax codes for plain and quoted characters */
@@ -168,12 +466,8 @@ enum regexp_syntax_op      /* syntax codes for plain and quoted characters */
   Rwordend,            /* end of word */
   Rwordbound,          /* word bound */
   Rnotwordbound,       /* not word bound */
-#ifdef emacs
-  Remacs_at_dot,       /* emacs: at dot */
-  Remacs_syntaxspec,   /* syntaxspec */
-  Remacs_notsyntaxspec,        /* notsyntaxspec */
-#endif /* emacs */
-  Rnum_ops
+  Rnum_ops,
+  Ropenrep,            /* opening bounded repeat */
 };
 
 static int re_compile_initialized = 0;
@@ -187,37 +481,14 @@ static int regexp_ansi_sequences;
 #define NUM_LEVELS  5    /* number of precedence levels in use */
 #define MAX_NESTING 100  /* max nesting level of operators */
 
-#ifdef emacs
-
-/* This code is for emacs compatibility only. */
-
-#include "config.h"
-#include "lisp.h"
-#include "buffer.h"
-#include "syntax.h"
-
-/* emacs defines NULL in some strange way? */
-#undef NULL
-#define NULL 0
-
-#else /* emacs */
-
 #define SYNTAX(ch) re_syntax_table[(unsigned char)(ch)]
-#define Sword 1
 
-#ifdef SYNTAX_TABLE
-char *re_syntax_table;
-#else
-static char re_syntax_table[256];
-#endif /* SYNTAX_TABLE */
+unsigned char re_syntax_table[256];
 
-#endif /* emacs */
-
-static void re_compile_initialize()
+void re_compile_initialize(void)
 {
   int a;
 
-#if !defined(emacs) && !defined(SYNTAX_TABLE)
   static int syntax_table_inited = 0;
 
   if (!syntax_table_inited)
@@ -229,9 +500,18 @@ static void re_compile_initialize()
       for (a = 'A'; a <= 'Z'; a++)
        re_syntax_table[a] = Sword;
       for (a = '0'; a <= '9'; a++)
-       re_syntax_table[a] = Sword;
+       re_syntax_table[a] = Sword | Sdigit | Shexdigit;
+      for (a = '0'; a <= '7'; a++)
+       re_syntax_table[a] |= Soctaldigit;
+      for (a = 'A'; a <= 'F'; a++)
+       re_syntax_table[a] |= Shexdigit;
+      for (a = 'a'; a <= 'f'; a++)
+       re_syntax_table[a] |= Shexdigit;
+      re_syntax_table['_'] = Sword;
+      for (a = 9; a <= 13; a++)
+       re_syntax_table[a] = Swhitespace;
+      re_syntax_table[' '] = Swhitespace;
     }
-#endif /* !emacs && !SYNTAX_TABLE */
   re_compile_initialized = 1;
   for (a = 0; a < 256; a++)
     {
@@ -274,11 +554,6 @@ static void re_compile_initialize()
   regexp_plain_ops['.'] = Ranychar;
   if (!(regexp_syntax & RE_NO_GNU_EXTENSIONS))
     {
-#ifdef emacs
-      regexp_quoted_ops['='] = Remacs_at_dot;
-      regexp_quoted_ops['s'] = Remacs_syntaxspec;
-      regexp_quoted_ops['S'] = Remacs_notsyntaxspec;
-#endif /* emacs */
       regexp_quoted_ops['w'] = Rwordchar;
       regexp_quoted_ops['W'] = Rnotwordchar;
       regexp_quoted_ops['<'] = Rwordbeg;
@@ -304,14 +579,24 @@ static void re_compile_initialize()
       regexp_precedences[Rbol] = 3;
       regexp_precedences[Reol] = 3;
     }
+  if (regexp_syntax & RE_REPEAT)
+    {
+      if (regexp_syntax & RE_NO_BK_PARENS)
+       {
+         regexp_plain_ops['{'] = Ropenrep;
+       }
+      else
+       {
+         regexp_quoted_ops['{'] = Ropenrep;
+       }
+    }
   regexp_precedences[Rclosepar] = 1;
   regexp_precedences[Rend] = 0;
   regexp_context_indep_ops = (regexp_syntax & RE_CONTEXT_INDEP_OPS) != 0;
   regexp_ansi_sequences = (regexp_syntax & RE_ANSI_HEX) != 0;
 }
 
-int re_set_syntax(syntax)
-int syntax;
+int re_set_syntax(int syntax)
 {
   int ret;
 
@@ -321,8 +606,7 @@ int syntax;
   return ret;
 }
 
-static int hex_char_to_decimal(ch)
-int ch;
+static int hex_char_to_decimal(int ch)
 {
   if (ch >= '0' && ch <= '9')
     return ch - '0';
@@ -333,186 +617,647 @@ int ch;
   return 16;
 }
 
-SilcResult re_compile_pattern(regex, size, bufp)
-char *regex;
-int size;
-SilcRegex bufp;
+static int re_compile_fastmap_aux(unsigned char *code, int pos,
+                                 unsigned char *visited,
+                                 unsigned char *can_be_null,
+                                 unsigned char *fastmap)
 {
-  int a, pos, op, current_level, level, opcode;
-  int pattern_offset = 0, alloc;
-  int starts[NUM_LEVELS * MAX_NESTING], starts_base;
-  int future_jumps[MAX_NESTING], num_jumps;
-  unsigned char ch = 0;
-  char *pattern, *translate;
-  int next_register, paren_depth, num_open_registers, open_registers[RE_NREGS];
-  int beginning_context;
-
-#define NEXTCHAR(var)                  \
-  MACRO_BEGIN                          \
-    if (pos >= size)                   \
-      goto ends_prematurely;           \
-    (var) = regex[pos];                        \
-    pos++;                             \
-  MACRO_END
-
-#define ALLOC(amount)                          \
-  MACRO_BEGIN                                  \
-    if (pattern_offset+(amount) > alloc)       \
-      {                                                \
-       alloc += 256 + (amount);                \
-       pattern = silc_realloc(pattern, alloc); \
-       if (!pattern)                           \
-         goto out_of_memory;                   \
-      }                                                \
-  MACRO_END
-
-#define STORE(ch) pattern[pattern_offset++] = (ch)
+  int a;
+  int b;
+  int syntaxcode;
 
-#define CURRENT_LEVEL_START (starts[starts_base + current_level])
+  if (visited[pos])
+    return 0;  /* we have already been here */
+  visited[pos] = 1;
+  for (;;)
+    switch (code[pos++]) {
+    case Cend:
+      {
+       *can_be_null = 1;
+       return 0;
+      }
+    case Cbol:
+    case Cbegbuf:
+    case Cendbuf:
+    case Cwordbeg:
+    case Cwordend:
+    case Cwordbound:
+    case Cnotwordbound:
+      {
+       for (a = 0; a < 256; a++)
+         fastmap[a] = 1;
+       break;
+      }
+    case Csyntaxspec:
+      {
+       syntaxcode = code[pos++];
+       for (a = 0; a < 256; a++)
+         if (SYNTAX(a) & syntaxcode)
+           fastmap[a] = 1;
+       return 0;
+      }
+    case Cnotsyntaxspec:
+      {
+       syntaxcode = code[pos++];
+       for (a = 0; a < 256; a++)
+         if (!(SYNTAX(a) & syntaxcode) )
+           fastmap[a] = 1;
+       return 0;
+      }
+    case Ceol:
+      {
+       fastmap['\n'] = 1;
+       if (*can_be_null == 0)
+         *can_be_null = 2; /* can match null, but only at end of buffer*/
+       return 0;
+      }
+    case Cset:
+      {
+       for (a = 0; a < 256/8; a++)
+         if (code[pos + a] != 0)
+           for (b = 0; b < 8; b++)
+             if (code[pos + a] & (1 << b))
+               fastmap[(a << 3) + b] = 1;
+       pos += 256/8;
+       return 0;
+      }
+    case Cexact:
+      {
+       fastmap[(unsigned char)code[pos]] = 1;
+       return 0;
+      }
+    case Canychar:
+      {
+       for (a = 0; a < 256; a++)
+         if (a != '\n')
+           fastmap[a] = 1;
+       return 0;
+      }
+    case Cstart_memory:
+    case Cend_memory:
+      {
+       pos++;
+       break;
+      }
+    case Cmatch_memory:
+      {
+       for (a = 0; a < 256; a++)
+         fastmap[a] = 1;
+       *can_be_null = 1;
+       return 0;
+      }
+    case Cjump:
+    case Cdummy_failure_jump:
+    case Cupdate_failure_jump:
+    case Cstar_jump:
+      {
+       a = (unsigned char)code[pos++];
+       a |= (unsigned char)code[pos++] << 8;
+       pos += (int)SHORT(a);
+       if (visited[pos])
+         {
+           /* argh... the regexp contains empty loops.  This is not
+              good, as this may cause a failure stack overflow when
+              matching.  Oh well. */
+           /* this path leads nowhere; pursue other paths. */
+           return 0;
+         }
+       visited[pos] = 1;
+       break;
+      }
+    case Cfailure_jump:
+      {
+       a = (unsigned char)code[pos++];
+       a |= (unsigned char)code[pos++] << 8;
+       a = pos + (int)SHORT(a);
+       return re_compile_fastmap_aux(code, a, visited, can_be_null, fastmap);
+      }
+    case Crepeat1:
+      {
+       pos += 2;
+       break;
+      }
+    default:
+      {
+       silc_set_errno(SILC_ERR_REGEX_OPCODE);
+       return -1;
+       /*NOTREACHED*/
+      }
+    }
+}
 
-#define SET_LEVEL_START starts[starts_base + current_level] = pattern_offset
+static int re_do_compile_fastmap(unsigned char *buffer, int used, int pos,
+                                 unsigned char *can_be_null,
+                                 unsigned char *fastmap, SilcRegex bufp)
+{
+  unsigned char small_visited[512], *visited;
+  int ret;
 
-#define PUSH_LEVEL_STARTS if (starts_base < (MAX_NESTING-1)*NUM_LEVELS) \
-                           starts_base += NUM_LEVELS;                  \
-                          else                                         \
-                           goto too_complex
+  if (used <= sizeof(small_visited))
+    visited = small_visited;
+  else
+    {
+      silc_stack_push(bufp->rstack, NULL);
+      visited = silc_smalloc(bufp->rstack, used);
+      if (!visited) {
+       silc_stack_pop(bufp->rstack);
+       return 0;
+      }
+    }
+  *can_be_null = 0;
+  memset(fastmap, 0, 256);
+  memset(visited, 0, used);
+  ret = re_compile_fastmap_aux(buffer, pos, visited, can_be_null, fastmap);
+  if (visited != small_visited) {
+    silc_sfree(bufp->rstack, visited);
+    silc_stack_pop(bufp->rstack);
+  }
+  return ret == 0;
+}
 
-#define POP_LEVEL_STARTS starts_base -= NUM_LEVELS
+int re_compile_fastmap(SilcRegex bufp)
+{
+  if (!bufp->fastmap || bufp->fastmap_accurate)
+    return 0;
+  SILC_ASSERT(bufp->used > 0);
+  if (!re_do_compile_fastmap(bufp->buffer,
+                            bufp->used,
+                            0,
+                            &bufp->can_be_null,
+                            bufp->fastmap, bufp))
+    return -1;
+  if (bufp->buffer[0] == Cbol)
+    bufp->anchor = 1;   /* begline */
+  else {
+    if (bufp->buffer[0] == Cbegbuf)
+      bufp->anchor = 2; /* begbuf */
+    else
+      bufp->anchor = 0; /* none */
+  }
+  bufp->fastmap_accurate = 1;
+  return 0;
+}
 
-#define PUT_ADDR(offset,addr)                          \
-  MACRO_BEGIN                                          \
-    int disp = (addr) - (offset) - 2;                  \
-    pattern[(offset)] = disp & 0xff;                   \
-    pattern[(offset)+1] = (disp>>8) & 0xff;            \
-  MACRO_END
-
-#define INSERT_JUMP(pos,type,addr)                     \
-  MACRO_BEGIN                                          \
-    int a, p = (pos), t = (type), ad = (addr);         \
-    for (a = pattern_offset - 1; a >= p; a--)          \
-      pattern[a + 3] = pattern[a];                     \
-    pattern[p] = t;                                    \
-    PUT_ADDR(p+1,ad);                                  \
-    pattern_offset += 3;                               \
-  MACRO_END
+/*
+ * star is coded as:
+ * 1: failure_jump 2
+ *    ... code for operand of star
+ *    star_jump 1
+ * 2: ... code after star
+ *
+ * We change the star_jump to update_failure_jump if we can determine
+ * that it is safe to do so; otherwise we change it to an ordinary
+ * jump.
+ *
+ * plus is coded as
+ *
+ *    jump 2
+ * 1: failure_jump 3
+ * 2: ... code for operand of plus
+ *    star_jump 1
+ * 3: ... code after plus
+ *
+ * For star_jump considerations this is processed identically to star.
+ *
+ */
+
+static int re_optimize_star_jump(SilcRegex bufp, unsigned char *code)
+{
+  unsigned char map[256];
+  unsigned char can_be_null;
+  unsigned char *p1;
+  unsigned char *p2;
+  unsigned char ch;
+  int a;
+  int b;
+  int num_instructions = 0;
 
-#define SETBIT(buf,offset,bit) (buf)[(offset)+(bit)/8] |= (1<<((bit) & 7))
+  a = (unsigned char)*code++;
+  a |= (unsigned char)*code++ << 8;
+  a = (int)SHORT(a);
 
-#define SET_FIELDS                             \
-  MACRO_BEGIN                                  \
-    bufp->allocated = alloc;                   \
-    bufp->buffer = pattern;                    \
-    bufp->used = pattern_offset;               \
-  MACRO_END
-
-#define GETHEX(var)                                            \
-  MACRO_BEGIN                                                  \
-    char gethex_ch, gethex_value;                              \
-    NEXTCHAR(gethex_ch);                                       \
-    gethex_value = hex_char_to_decimal(gethex_ch);             \
-    if (gethex_value == 16)                                    \
-      goto hex_error;                                          \
-    NEXTCHAR(gethex_ch);                                       \
-    gethex_ch = hex_char_to_decimal(gethex_ch);                        \
-    if (gethex_ch == 16)                                       \
-      goto hex_error;                                          \
-    (var) = gethex_value * 16 + gethex_ch;                     \
-  MACRO_END
-
-#define ANSI_TRANSLATE(ch)                             \
-  MACRO_BEGIN                                          \
-    switch (ch)                                                \
-      {                                                        \
-      case 'a':                                                \
-      case 'A':                                                \
-       ch = 7; /* audible bell */                      \
-       break;                                          \
-      case 'b':                                                \
-      case 'B':                                                \
-       ch = 8; /* backspace */                         \
-       break;                                          \
-      case 'f':                                                \
-      case 'F':                                                \
-       ch = 12; /* form feed */                        \
-       break;                                          \
-      case 'n':                                                \
-      case 'N':                                                \
-       ch = 10; /* line feed */                        \
-       break;                                          \
-      case 'r':                                                \
-      case 'R':                                                \
-       ch = 13; /* carriage return */                  \
-       break;                                          \
-      case 't':                                                \
-      case 'T':                                                \
-       ch = 9; /* tab */                               \
-       break;                                          \
-      case 'v':                                                \
-      case 'V':                                                \
-       ch = 11; /* vertical tab */                     \
-       break;                                          \
-      case 'x': /* hex code */                         \
-      case 'X':                                                \
-       GETHEX(ch);                                     \
-       break;                                          \
-      default:                                         \
-       /* other characters passed through */           \
-       if (translate)                                  \
-         ch = translate[(unsigned char)ch];            \
-       break;                                          \
-      }                                                        \
-  MACRO_END
+  p1 = code + a + 3; /* skip the failure_jump */
+  /* Check that the jump is within the pattern */
+  if (p1<bufp->buffer || bufp->buffer+bufp->used<p1)
+    {
+      silc_set_errno(SILC_ERR_OVERFLOW);
+      return 0;
+    }
 
-  if (!re_compile_initialized)
-    re_compile_initialize();
-  bufp->used = 0;
-  bufp->fastmap_accurate = 0;
-  bufp->uses_registers = 0;
-  translate = bufp->translate;
-  pattern = bufp->buffer;
-  alloc = bufp->allocated;
-  if (alloc == 0 || pattern == NULL)
+  SILC_ASSERT(p1[-3] == Cfailure_jump);
+  p2 = code;
+  /* p1 points inside loop, p2 points to after loop */
+  if (!re_do_compile_fastmap(bufp->buffer, bufp->used,
+                            (int)(p2 - bufp->buffer),
+                            &can_be_null, map, bufp))
+    goto make_normal_jump;
+
+  /* If we might introduce a new update point inside the
+   * loop, we can't optimize because then update_jump would
+   * update a wrong failure point.  Thus we have to be
+   * quite careful here.
+   */
+
+  /* loop until we find something that consumes a character */
+ loop_p1:
+  num_instructions++;
+  switch (*p1++)
     {
-      alloc = 256;
-      pattern = silc_malloc(alloc);
-      if (!pattern)
-       goto out_of_memory;
+    case Cbol:
+    case Ceol:
+    case Cbegbuf:
+    case Cendbuf:
+    case Cwordbeg:
+    case Cwordend:
+    case Cwordbound:
+    case Cnotwordbound:
+      {
+       goto loop_p1;
+      }
+    case Cstart_memory:
+    case Cend_memory:
+      {
+       p1++;
+       goto loop_p1;
+      }
+    case Cexact:
+      {
+       ch = (unsigned char)*p1++;
+       if (map[(int)ch])
+         goto make_normal_jump;
+       break;
+      }
+    case Canychar:
+      {
+       for (b = 0; b < 256; b++)
+         if (b != '\n' && map[b])
+           goto make_normal_jump;
+       break;
+      }
+    case Cset:
+      {
+       for (b = 0; b < 256; b++)
+         if ((p1[b >> 3] & (1 << (b & 7))) && map[b])
+           goto make_normal_jump;
+       p1 += 256/8;
+       break;
+      }
+    default:
+      {
+       goto make_normal_jump;
+      }
     }
-  pattern_offset = 0;
-  starts_base = 0;
-  num_jumps = 0;
-  current_level = 0;
-  SET_LEVEL_START;
-  num_open_registers = 0;
-  next_register = 1;
-  paren_depth = 0;
-  beginning_context = 1;
-  op = -1;
-  /* we use Rend dummy to ensure that pending jumps are updated (due to
-     low priority of Rend) before exiting the loop. */
-  pos = 0;
-  while (op != Rend)
+  /* now we know that we can't backtrack. */
+  while (p1 != p2 - 3)
     {
-      if (pos >= size)
-       op = Rend;
-      else
-       {
-         NEXTCHAR(ch);
-         if (translate)
-           ch = translate[(unsigned char)ch];
-         op = regexp_plain_ops[(unsigned char)ch];
-         if (op == Rquote)
-           {
-             NEXTCHAR(ch);
-             op = regexp_quoted_ops[(unsigned char)ch];
-             if (op == Rnormal && regexp_ansi_sequences)
-               ANSI_TRANSLATE(ch);
-           }
-       }
-      level = regexp_precedences[op];
-      /* printf("ch='%c' op=%d level=%d current_level=%d curlevstart=%d\n",
-            ch, op, level, current_level, CURRENT_LEVEL_START); */
-      if (level > current_level)
+      num_instructions++;
+      switch (*p1++)
+       {
+       case Cend:
+         {
+           return 0;
+         }
+       case Cbol:
+       case Ceol:
+       case Canychar:
+       case Cbegbuf:
+       case Cendbuf:
+       case Cwordbeg:
+       case Cwordend:
+       case Cwordbound:
+       case Cnotwordbound:
+         {
+           break;
+         }
+       case Cset:
+         {
+           p1 += 256/8;
+           break;
+         }
+       case Cexact:
+       case Cstart_memory:
+       case Cend_memory:
+       case Cmatch_memory:
+       case Csyntaxspec:
+       case Cnotsyntaxspec:
+         {
+           p1++;
+           break;
+         }
+       case Cjump:
+       case Cstar_jump:
+       case Cfailure_jump:
+       case Cupdate_failure_jump:
+       case Cdummy_failure_jump:
+         {
+           goto make_normal_jump;
+         }
+       default:
+         {
+           return 0;
+         }
+       }
+    }
+
+  /* make_update_jump: */
+  code -= 3;
+  a += 3;  /* jump to after the Cfailure_jump */
+  code[0] = Cupdate_failure_jump;
+  code[1] = a & 0xff;
+  code[2] = a >> 8;
+  if (num_instructions > 1)
+    return 1;
+  SILC_ASSERT(num_instructions == 1);
+  /* if the only instruction matches a single character, we can do
+   * better */
+  p1 = code + 3 + a;   /* start of sole instruction */
+  if (*p1 == Cset || *p1 == Cexact || *p1 == Canychar ||
+      *p1 == Csyntaxspec || *p1 == Cnotsyntaxspec)
+    code[0] = Crepeat1;
+  return 1;
+
+ make_normal_jump:
+  code -= 3;
+  *code = Cjump;
+  return 1;
+}
+
+static int re_optimize(SilcRegex bufp)
+{
+  unsigned char *code;
+
+  code = bufp->buffer;
+
+  while(1)
+    {
+      switch (*code++)
+       {
+       case Cend:
+         {
+           return 1;
+         }
+       case Canychar:
+       case Cbol:
+       case Ceol:
+       case Cbegbuf:
+       case Cendbuf:
+       case Cwordbeg:
+       case Cwordend:
+       case Cwordbound:
+       case Cnotwordbound:
+         {
+           break;
+         }
+       case Cset:
+         {
+           code += 256/8;
+           break;
+         }
+       case Cexact:
+       case Cstart_memory:
+       case Cend_memory:
+       case Cmatch_memory:
+       case Csyntaxspec:
+       case Cnotsyntaxspec:
+         {
+           code++;
+           break;
+         }
+       case Cstar_jump:
+         {
+           if (!re_optimize_star_jump(bufp, code))
+             {
+               return 0;
+             }
+           /* fall through */
+         }
+       case Cupdate_failure_jump:
+       case Cjump:
+       case Cdummy_failure_jump:
+       case Cfailure_jump:
+       case Crepeat1:
+         {
+           code += 2;
+           break;
+         }
+       default:
+         {
+           return 0;
+         }
+       }
+    }
+}
+
+#define NEXTCHAR(var)                          \
+  {                                            \
+    if (pos >= size)                           \
+      goto ends_prematurely;                   \
+    (var) = regex[pos];                                \
+    pos++;                                     \
+  }
+
+#define ALLOC(amount)                          \
+  {                                            \
+    if (pattern_offset+(amount) > alloc)       \
+      {                                                \
+       pattern = silc_srealloc(bufp->rstack, alloc, pattern, \
+                               alloc + 256 + (amount));      \
+       alloc += 256 + (amount);                \
+       if (!pattern)                           \
+         goto out_of_memory;                   \
+      }                                                \
+  }
+
+#define STORE(ch) pattern[pattern_offset++] = (ch)
+
+#define CURRENT_LEVEL_START (starts[starts_base + current_level])
+
+#define SET_LEVEL_START starts[starts_base + current_level] = pattern_offset
+
+#define PUSH_LEVEL_STARTS \
+  if (starts_base < (MAX_NESTING-1)*NUM_LEVELS) \
+    starts_base += NUM_LEVELS; \
+  else \
+    goto too_complex \
+
+#define POP_LEVEL_STARTS starts_base -= NUM_LEVELS
+
+#define PUT_ADDR(offset,addr)                  \
+  {                                            \
+    int disp = (addr) - (offset) - 2;          \
+    pattern[(offset)] = disp & 0xff;           \
+    pattern[(offset)+1] = (disp>>8) & 0xff;    \
+  }
+
+#define INSERT_JUMP(pos,type,addr)             \
+  {                                            \
+    int a, p = (pos), t = (type), ad = (addr); \
+    for (a = pattern_offset - 1; a >= p; a--)  \
+      pattern[a + 3] = pattern[a];             \
+    pattern[p] = t;                            \
+    PUT_ADDR(p+1,ad);                          \
+    pattern_offset += 3;                       \
+  }
+
+#define SETBIT(buf,offset,bit) (buf)[(offset)+(bit)/8] |= (1<<((bit) & 7))
+
+#define SET_FIELDS                             \
+  {                                            \
+    bufp->allocated = alloc;                   \
+    bufp->buffer = pattern;                    \
+    bufp->used = pattern_offset;               \
+  }
+
+#define GETHEX(var)                                    \
+  {                                                    \
+    unsigned char gethex_ch, gethex_value;             \
+    NEXTCHAR(gethex_ch);                               \
+    gethex_value = hex_char_to_decimal(gethex_ch);     \
+    if (gethex_value == 16)                            \
+      goto hex_error;                                  \
+    NEXTCHAR(gethex_ch);                               \
+    gethex_ch = hex_char_to_decimal(gethex_ch);                \
+    if (gethex_ch == 16)                               \
+      goto hex_error;                                  \
+    (var) = gethex_value * 16 + gethex_ch;             \
+  }
+
+#define ANSI_TRANSLATE(ch)                     \
+  {                                            \
+    switch (ch)                                        \
+      {                                                \
+      case 'a':                                        \
+      case 'A':                                        \
+       {                                       \
+         ch = 7; /* audible bell */            \
+         break;                                \
+       }                                       \
+      case 'b':                                        \
+      case 'B':                                        \
+       {                                       \
+         ch = 8; /* backspace */               \
+         break;                                \
+       }                                       \
+      case 'f':                                        \
+      case 'F':                                        \
+       {                                       \
+         ch = 12; /* form feed */              \
+         break;                                \
+       }                                       \
+      case 'n':                                        \
+      case 'N':                                        \
+       {                                       \
+         ch = 10; /* line feed */              \
+         break;                                \
+       }                                       \
+      case 'r':                                        \
+      case 'R':                                        \
+       {                                       \
+         ch = 13; /* carriage return */        \
+         break;                                \
+       }                                       \
+      case 't':                                        \
+      case 'T':                                        \
+       {                                       \
+         ch = 9; /* tab */                     \
+         break;                                \
+       }                                       \
+      case 'v':                                        \
+      case 'V':                                        \
+       {                                       \
+         ch = 11; /* vertical tab */           \
+         break;                                \
+       }                                       \
+      case 'x': /* hex code */                 \
+      case 'X':                                        \
+       {                                       \
+         GETHEX(ch);                           \
+         break;                                \
+       }                                       \
+      default:                                 \
+       {                                       \
+         /* other characters passed through */ \
+         if (translate)                        \
+           ch = translate[(unsigned char)ch];  \
+         break;                                \
+       }                                       \
+      }                                                \
+  }
+
+SilcResult re_compile_pattern(unsigned char *regex, int size, SilcRegex bufp)
+{
+  int a;
+  int pos;
+  int op;
+  int current_level;
+  int level;
+  int opcode;
+  int pattern_offset = 0, alloc;
+  int starts[NUM_LEVELS * MAX_NESTING];
+  int starts_base;
+  int future_jumps[MAX_NESTING];
+  int num_jumps;
+  unsigned char ch = '\0';
+  unsigned char *pattern;
+  unsigned char *translate;
+  int next_register;
+  int paren_depth;
+  int num_open_registers;
+  int open_registers[RE_NREGS];
+  int beginning_context;
+
+  if (!re_compile_initialized)
+    re_compile_initialize();
+  bufp->used = 0;
+  bufp->fastmap_accurate = 0;
+  bufp->uses_registers = 1;
+  bufp->num_registers = 1;
+  translate = bufp->translate;
+  pattern = bufp->buffer;
+  alloc = bufp->allocated;
+  if (alloc == 0 || pattern == NULL)
+    {
+      alloc = 256;
+      pattern = silc_smalloc(bufp->rstack, alloc);
+      if (!pattern)
+       goto out_of_memory;
+    }
+  pattern_offset = 0;
+  starts_base = 0;
+  num_jumps = 0;
+  current_level = 0;
+  SET_LEVEL_START;
+  num_open_registers = 0;
+  next_register = 1;
+  paren_depth = 0;
+  beginning_context = 1;
+  op = -1;
+  /* we use Rend dummy to ensure that pending jumps are updated
+     (due to low priority of Rend) before exiting the loop. */
+  pos = 0;
+  while (op != Rend)
+    {
+      if (pos >= size)
+       op = Rend;
+      else
+       {
+         NEXTCHAR(ch);
+         if (translate)
+           ch = translate[(unsigned char)ch];
+         op = regexp_plain_ops[(unsigned char)ch];
+         if (op == Rquote)
+           {
+             NEXTCHAR(ch);
+             op = regexp_quoted_ops[(unsigned char)ch];
+             if (op == Rnormal && regexp_ansi_sequences)
+               ANSI_TRANSLATE(ch);
+           }
+       }
+      level = regexp_precedences[op];
+      /* printf("ch='%c' op=%d level=%d current_level=%d
+        curlevstart=%d\n", ch, op, level, current_level,
+        CURRENT_LEVEL_START); */
+      if (level > current_level)
        {
          for (current_level++; current_level < level; current_level++)
            SET_LEVEL_START;
@@ -523,158 +1268,191 @@ SilcRegex bufp;
          {
            current_level = level;
            for (;num_jumps > 0 &&
-                future_jumps[num_jumps-1] >= CURRENT_LEVEL_START;
+                  future_jumps[num_jumps-1] >= CURRENT_LEVEL_START;
                 num_jumps--)
              PUT_ADDR(future_jumps[num_jumps-1], pattern_offset);
          }
       switch (op)
        {
        case Rend:
-         break;
+         {
+           break;
+         }
        case Rnormal:
-       normal_char:
-         opcode = Cexact;
-       store_opcode_and_arg: /* opcode & ch must be set */
-         SET_LEVEL_START;
-         ALLOC(2);
-         STORE(opcode);
-         STORE(ch);
-         break;
+         {
+         normal_char:
+           opcode = Cexact;
+         store_opcode_and_arg: /* opcode & ch must be set */
+           SET_LEVEL_START;
+           ALLOC(2);
+           STORE(opcode);
+           STORE(ch);
+           break;
+         }
        case Ranychar:
-         opcode = Canychar;
-       store_opcode:
-         SET_LEVEL_START;
-         ALLOC(1);
-         STORE(opcode);
-         break;
+         {
+           opcode = Canychar;
+         store_opcode:
+           SET_LEVEL_START;
+           ALLOC(1);
+           STORE(opcode);
+           break;
+         }
        case Rquote:
-         abort();
-         /*NOTREACHED*/
+         {
+           SILC_VERIFY(op != Rquote);
+           /*NOTREACHED*/
+         }
        case Rbol:
-         if (!beginning_context) {
-           if (regexp_context_indep_ops)
-             goto op_error;
-           else
-             goto normal_char;
+         {
+           if (!beginning_context) {
+             if (regexp_context_indep_ops)
+               goto op_error;
+             else
+               goto normal_char;
+           }
+           opcode = Cbol;
+           goto store_opcode;
          }
-         opcode = Cbol;
-         goto store_opcode;
        case Reol:
-         if (!((pos >= size) ||
-               ((regexp_syntax & RE_NO_BK_VBAR) ?
-                (regex[pos] == '\174') :
-                (pos+1 < size && regex[pos] == '\134' &&
-                 regex[pos+1] == '\174')) ||
-               ((regexp_syntax & RE_NO_BK_PARENS)?
-                (regex[pos] == ')'):
-                (pos+1 < size && regex[pos] == '\134' &&
-                 regex[pos+1] == ')')))) {
-           if (regexp_context_indep_ops)
-             goto op_error;
-           else
-             goto normal_char;
+         {
+           if (!((pos >= size) ||
+                 ((regexp_syntax & RE_NO_BK_VBAR) ?
+                  (regex[pos] == '\174') :
+                  (pos+1 < size && regex[pos] == '\134' &&
+                   regex[pos+1] == '\174')) ||
+                 ((regexp_syntax & RE_NO_BK_PARENS)?
+                  (regex[pos] == ')'):
+                  (pos+1 < size && regex[pos] == '\134' &&
+                   regex[pos+1] == ')')))) {
+             if (regexp_context_indep_ops)
+               goto op_error;
+             else
+               goto normal_char;
+           }
+           opcode = Ceol;
+           goto store_opcode;
+           /* NOTREACHED */
+           break;
          }
-         opcode = Ceol;
-         goto store_opcode;
-         break;
        case Roptional:
-         if (beginning_context) {
-           if (regexp_context_indep_ops)
-             goto op_error;
-           else
-             goto normal_char;
+         {
+           if (beginning_context) {
+             if (regexp_context_indep_ops)
+               goto op_error;
+             else
+               goto normal_char;
+           }
+           if (CURRENT_LEVEL_START == pattern_offset)
+             break; /* ignore empty patterns for ? */
+           ALLOC(3);
+           INSERT_JUMP(CURRENT_LEVEL_START, Cfailure_jump,
+                       pattern_offset + 3);
+           break;
          }
-         if (CURRENT_LEVEL_START == pattern_offset)
-           break; /* ignore empty patterns for ? */
-         ALLOC(3);
-         INSERT_JUMP(CURRENT_LEVEL_START, Cfailure_jump,
-                     pattern_offset + 3);
-         break;
        case Rstar:
        case Rplus:
-         if (beginning_context) {
-           if (regexp_context_indep_ops)
-             goto op_error;
-           else
-             goto normal_char;
+         {
+         store_jump:
+           if (beginning_context) {
+             if (regexp_context_indep_ops)
+               goto op_error;
+             else
+               goto normal_char;
+           }
+           if (CURRENT_LEVEL_START == pattern_offset)
+             break; /* ignore empty patterns for + and * */
+           ALLOC(9);
+           INSERT_JUMP(CURRENT_LEVEL_START, Cfailure_jump,
+                       pattern_offset + 6);
+           INSERT_JUMP(pattern_offset, Cstar_jump, CURRENT_LEVEL_START);
+           if (op == Rplus)  /* jump over initial failure_jump */
+             INSERT_JUMP(CURRENT_LEVEL_START, Cdummy_failure_jump,
+                         CURRENT_LEVEL_START + 6);
+           break;
          }
-         if (CURRENT_LEVEL_START == pattern_offset)
-           break; /* ignore empty patterns for + and * */
-         ALLOC(9);
-         INSERT_JUMP(CURRENT_LEVEL_START, Cfailure_jump,
-                     pattern_offset + 6);
-         INSERT_JUMP(pattern_offset, Cstar_jump, CURRENT_LEVEL_START);
-         if (op == Rplus)  /* jump over initial failure_jump */
-           INSERT_JUMP(CURRENT_LEVEL_START, Cdummy_failure_jump,
-                       CURRENT_LEVEL_START + 6);
-         break;
        case Ror:
-         ALLOC(6);
-         INSERT_JUMP(CURRENT_LEVEL_START, Cfailure_jump,
-                     pattern_offset + 6);
-         if (num_jumps >= MAX_NESTING)
-           goto too_complex;
-         STORE(Cjump);
-         future_jumps[num_jumps++] = pattern_offset;
-         STORE(0);
-         STORE(0);
-         SET_LEVEL_START;
-         break;
+         {
+           ALLOC(6);
+           INSERT_JUMP(CURRENT_LEVEL_START, Cfailure_jump,
+                       pattern_offset + 6);
+           if (num_jumps >= MAX_NESTING)
+             goto too_complex;
+           STORE(Cjump);
+           future_jumps[num_jumps++] = pattern_offset;
+           STORE(0);
+           STORE(0);
+           SET_LEVEL_START;
+           break;
+         }
        case Ropenpar:
-         SET_LEVEL_START;
-         if (next_register < RE_NREGS)
-           {
-             bufp->uses_registers = 1;
-             ALLOC(2);
-             STORE(Cstart_memory);
-             STORE(next_register);
-             open_registers[num_open_registers++] = next_register;
-             next_register++;
-           }
-         paren_depth++;
-         PUSH_LEVEL_STARTS;
-         current_level = 0;
-         SET_LEVEL_START;
-         break;
+         {
+           SET_LEVEL_START;
+           if (next_register < RE_NREGS)
+             {
+               bufp->uses_registers = 1;
+               ALLOC(2);
+               STORE(Cstart_memory);
+               STORE(next_register);
+               open_registers[num_open_registers++] = next_register;
+               bufp->num_registers++;
+               next_register++;
+             }
+           paren_depth++;
+           PUSH_LEVEL_STARTS;
+           current_level = 0;
+           SET_LEVEL_START;
+           break;
+         }
        case Rclosepar:
-         if (paren_depth <= 0)
-           goto parenthesis_error;
-         POP_LEVEL_STARTS;
-         current_level = regexp_precedences[Ropenpar];
-         paren_depth--;
-         if (paren_depth < num_open_registers)
-           {
-             bufp->uses_registers = 1;
-             ALLOC(2);
-             STORE(Cend_memory);
-             num_open_registers--;
-             STORE(open_registers[num_open_registers]);
-           }
-         break;
+         {
+           if (paren_depth <= 0)
+             goto parenthesis_error;
+           POP_LEVEL_STARTS;
+           current_level = regexp_precedences[Ropenpar];
+           paren_depth--;
+           if (paren_depth < num_open_registers)
+             {
+               bufp->uses_registers = 1;
+               ALLOC(2);
+               STORE(Cend_memory);
+               num_open_registers--;
+               STORE(open_registers[num_open_registers]);
+             }
+           break;
+         }
        case Rmemory:
-         if (ch == '0')
-           goto bad_match_register;
-         assert(ch >= '0' && ch <= '9');
-         bufp->uses_registers = 1;
-         opcode = Cmatch_memory;
-         ch -= '0';
-         goto store_opcode_and_arg;
+         {
+           if (ch == '0')
+             goto bad_match_register;
+           SILC_ASSERT(ch >= '0' && ch <= '9');
+           bufp->uses_registers = 1;
+           opcode = Cmatch_memory;
+           ch -= '0';
+           goto store_opcode_and_arg;
+         }
        case Rextended_memory:
-         NEXTCHAR(ch);
-         if (ch < '0' || ch > '9')
-           goto bad_match_register;
-         NEXTCHAR(a);
-         if (a < '0' || a > '9')
-           goto bad_match_register;
-         ch = 10 * (a - '0') + ch - '0';
-         if (ch <= 0 || ch >= RE_NREGS)
-           goto bad_match_register;
-         bufp->uses_registers = 1;
-         opcode = Cmatch_memory;
-         goto store_opcode_and_arg;
+         {
+           NEXTCHAR(ch);
+           if (ch < '0' || ch > '9')
+             goto bad_match_register;
+           NEXTCHAR(a);
+           if (a < '0' || a > '9')
+             goto bad_match_register;
+           ch = 10 * (a - '0') + ch - '0';
+           if (ch == 0 || ch >= RE_NREGS)
+             goto bad_match_register;
+           bufp->uses_registers = 1;
+           opcode = Cmatch_memory;
+           goto store_opcode_and_arg;
+         }
        case Ropenset:
          {
-           int complement,prev,offset,range,firstchar;
+           int complement;
+           int prev;
+           int offset;
+           int range;
+           int firstchar;
 
            SET_LEVEL_START;
            ALLOC(1+256/8);
@@ -707,7 +1485,7 @@ SilcRegex bufp;
                  }
                if (range)
                  {
-                   for (a = prev; a <= ch; a++)
+                   for (a = prev; a <= (int)ch; a++)
                      SETBIT(pattern, offset, a);
                    prev = -1;
                    range = 0;
@@ -733,62 +1511,195 @@ SilcRegex bufp;
              }
            break;
          }
+       case Ropenrep:
+         {
+           /* The bounded repeat syntax: a{n}, a{n,} and a{n,m}.  The first
+              is compiled as n-1 Rnormals.  The second is compiled as n-1
+              Rnormals and one Rplus.  The third is compiled as n-1 Rnormals
+              and m-n Rnormals with Roptionals.  0 values have special
+              compilation. */
+           int min, max, i;
+
+           if (pos >= size)
+             goto op_error;
+
+           /* Get the preceding atom */
+           if (pos < 2)
+             goto op_error;
+           pos -= 2;
+           NEXTCHAR(a);
+           NEXTCHAR(ch);
+
+           /* Get min value */
+           NEXTCHAR(ch);
+           if (!isdigit(ch))
+             goto repeat_value_error;
+           min = ch - '0';
+           NEXTCHAR(ch);
+           while (isdigit(ch)) {
+             min *= 10;
+             min += ch - '0';
+             NEXTCHAR(ch);
+           }
+
+           if (min > 255)
+             goto repeat_value_error;
+
+           if (ch == '}') {
+             /* The a{n} case */
+
+             if (!min) {
+               /* Will not do any matching with a{0} */
+               pattern_offset -= 2;
+               break;
+             }
+
+             /* Store min - 1 many Cexacts. */
+             for (i = 0; i < min - 1; i++) {
+               SET_LEVEL_START;
+               ALLOC(2);
+               STORE(Cexact);
+               STORE((unsigned char)a);
+             }
+             break;
+           }
+
+           if (ch == ',') {
+             NEXTCHAR(ch);
+
+             if (ch == '}') {
+               /* The a{n,} case */
+
+               if (!min) {
+                 /* Store Rstar with a{0,} */
+                 op = Rstar;
+                 goto store_jump;
+               }
+
+               /* Store min - 1 many Cexacts. */
+               for (i = 0; i < min - 1; i++) {
+                 SET_LEVEL_START;
+                 ALLOC(2);
+                 STORE(Cexact);
+                 STORE((unsigned char)a);
+               }
+
+               /* Store Rplus */
+               op = Rplus;
+               goto store_jump;
+             }
+
+             /* The a{n,m} case */
+
+             /* Get max value */
+             if (!isdigit(ch))
+               goto repeat_value_error;
+             max = ch - '0';
+             NEXTCHAR(ch);
+             while (isdigit(ch)) {
+               max *= 10;
+               max += ch - '0';
+               NEXTCHAR(ch);
+             }
+
+             if (max > 255)
+               goto repeat_value_error;
+             if (min > max)
+               goto repeat_value_error;
+
+             if (!min && !max) {
+               /* Will not do any matching with a{0,0} */
+               pattern_offset -= 2;
+               break;
+             }
+
+             if (ch != '}')
+               goto op_error;
+
+             if (!min)
+               /* Only optional matching with a{0,m}. */
+               pattern_offset -= 2;
+
+             /* Store min - 1 many Cexacts. */
+             for (i = 0; min && i < min - 1; i++) {
+               SET_LEVEL_START;
+               ALLOC(2);
+               STORE(Cexact);
+               STORE((unsigned char)a);
+             }
+
+             /* Store max - min Cexacts and Roptionals. */
+             for (i = 0; i < max - min; i++) {
+               SET_LEVEL_START;
+               ALLOC(2);
+               STORE(Cexact);
+               STORE((unsigned char)a);
+               ALLOC(3);
+               INSERT_JUMP(CURRENT_LEVEL_START, Cfailure_jump,
+                           pattern_offset + 3);
+             }
+             break;
+           }
+
+           goto op_error;
+         }
        case Rbegbuf:
-         opcode = Cbegbuf;
-         goto store_opcode;
+         {
+           opcode = Cbegbuf;
+           goto store_opcode;
+         }
        case Rendbuf:
-         opcode = Cendbuf;
-         goto store_opcode;
+         {
+           opcode = Cendbuf;
+           goto store_opcode;
+         }
        case Rwordchar:
-         opcode = Csyntaxspec;
-         ch = Sword;
-         goto store_opcode_and_arg;
+         {
+           opcode = Csyntaxspec;
+           ch = Sword;
+           goto store_opcode_and_arg;
+         }
        case Rnotwordchar:
-         opcode = Cnotsyntaxspec;
-         ch = Sword;
-         goto store_opcode_and_arg;
+         {
+           opcode = Cnotsyntaxspec;
+           ch = Sword;
+           goto store_opcode_and_arg;
+         }
        case Rwordbeg:
-         opcode = Cwordbeg;
-         goto store_opcode;
+         {
+           opcode = Cwordbeg;
+           goto store_opcode;
+         }
        case Rwordend:
-         opcode = Cwordend;
-         goto store_opcode;
+         {
+           opcode = Cwordend;
+           goto store_opcode;
+         }
        case Rwordbound:
-         opcode = Cwordbound;
-         goto store_opcode;
+         {
+           opcode = Cwordbound;
+           goto store_opcode;
+         }
        case Rnotwordbound:
-         opcode = Cnotwordbound;
-         goto store_opcode;
-#ifdef emacs
-       case Remacs_at_dot:
-         opcode = Cemacs_at_dot;
-         goto store_opcode;
-       case Remacs_syntaxspec:
-         NEXTCHAR(ch);
-         if (translate)
-           ch = translate[(unsigned char)ch];
-         opcode = Csyntaxspec;
-         ch = syntax_spec_code[(unsigned char)ch];
-         goto store_opcode_and_arg;
-       case Remacs_notsyntaxspec:
-         NEXTCHAR(ch);
-         if (translate)
-           ch = translate[(unsigned char)ch];
-         opcode = Cnotsyntaxspec;
-         ch = syntax_spec_code[(unsigned char)ch];
-         goto store_opcode_and_arg;
-#endif /* emacs */
+         {
+           opcode = Cnotwordbound;
+           goto store_opcode;
+         }
        default:
-         abort();
+         {
+           abort();
+         }
        }
       beginning_context = (op == Ropenpar || op == Ror);
     }
   if (starts_base != 0)
     goto parenthesis_error;
-  assert(num_jumps == 0);
+  SILC_ASSERT(num_jumps == 0);
   ALLOC(1);
   STORE(Cend);
   SET_FIELDS;
+  if (!re_optimize(bufp))
+    return SILC_ERR;
   return SILC_OK;
 
  op_error:
@@ -818,667 +1729,484 @@ SilcRegex bufp;
  too_complex:
   SET_FIELDS;
   return SILC_ERR_REGEX_TOO_COMPLEX;
+
+ repeat_value_error:
+  SET_FIELDS;
+  return SILC_ERR_REGEX_REPEAT;
 }
+
 #undef CHARAT
 #undef NEXTCHAR
 #undef GETHEX
-#undef ALLOC
-#undef STORE
-#undef CURRENT_LEVEL_START
-#undef SET_LEVEL_START
-#undef PUSH_LEVEL_STARTS
-#undef POP_LEVEL_STARTS
-#undef PUT_ADDR
-#undef INSERT_JUMP
-#undef SETBIT
-#undef SET_FIELDS
-
-static void re_compile_fastmap_aux(code, pos, visited, can_be_null, fastmap)
-char *code, *visited, *can_be_null, *fastmap;
-int pos;
-{
-  int a, b, syntaxcode;
-
-  if (visited[pos])
-    return;  /* we have already been here */
-  visited[pos] = 1;
-  for (;;)
-    switch (code[pos++])
-      {
-      case Cend:
-       *can_be_null = 1;
-       return;
-      case Cbol:
-      case Cbegbuf:
-      case Cendbuf:
-      case Cwordbeg:
-      case Cwordend:
-      case Cwordbound:
-      case Cnotwordbound:
-#ifdef emacs
-      case Cemacs_at_dot:
-#endif /* emacs */
-       break;
-      case Csyntaxspec:
-       syntaxcode = code[pos++];
-       for (a = 0; a < 256; a++)
-         if (SYNTAX(a) == syntaxcode)
-           fastmap[a] = 1;
-       return;
-      case Cnotsyntaxspec:
-       syntaxcode = code[pos++];
-       for (a = 0; a < 256; a++)
-         if (SYNTAX(a) != syntaxcode)
-           fastmap[a] = 1;
-       return;
-      case Ceol:
-       fastmap['\n'] = 1;
-       if (*can_be_null == 0)
-         *can_be_null = 2;  /* can match null, but only at end of buffer*/
-       return;
-      case Cset:
-       for (a = 0; a < 256/8; a++)
-         if (code[pos + a] != 0)
-           for (b = 0; b < 8; b++)
-             if (code[pos + a] & (1 << b))
-               fastmap[(a << 3) + b] = 1;
-       pos += 256/8;
-       return;
-      case Cexact:
-       fastmap[(unsigned char)code[pos]] = 1;
-       return;
-      case Canychar:
-       for (a = 0; a < 256; a++)
-         if (a != '\n')
-           fastmap[a] = 1;
-       return;
-      case Cstart_memory:
-      case Cend_memory:
-       pos++;
-       break;
-      case Cmatch_memory:
-       for (a = 0; a < 256; a++)
-         fastmap[a] = 1;
-       *can_be_null = 1;
-       return;
-      case Cjump:
-      case Cdummy_failure_jump:
-      case Cupdate_failure_jump:
-      case Cstar_jump:
-       a = (unsigned char)code[pos++];
-       a |= (unsigned char)code[pos++] << 8;
-       pos += (int)(short)a;
-       if (visited[pos])
-         {
-           /* argh... the regexp contains empty loops.  This is not
-              good, as this may cause a failure stack overflow when
-              matching.  Oh well. */
-           /* this path leads nowhere; pursue other paths. */
-           return;
-         }
-       visited[pos] = 1;
-       break;
-      case Cfailure_jump:
-       a = (unsigned char)code[pos++];
-       a |= (unsigned char)code[pos++] << 8;
-       a = pos + (int)(short)a;
-       re_compile_fastmap_aux(code, a, visited, can_be_null, fastmap);
-       break;
-      default:
-       abort();  /* probably some opcode is missing from this switch */
-       /*NOTREACHED*/
-      }
-}
-
-static int re_do_compile_fastmap(buffer, used, pos, can_be_null, fastmap)
-char *buffer, *fastmap, *can_be_null;
-int used, pos;
-{
-  char small_visited[512], *visited;
-
-  if (used <= sizeof(small_visited))
-    visited = small_visited;
-  else
-    {
-      visited = silc_malloc(used);
-      if (!visited)
-       return 0;
-    }
-  *can_be_null = 0;
-  memset(fastmap, 0, 256);
-  memset(visited, 0, used);
-  re_compile_fastmap_aux(buffer, pos, visited, can_be_null, fastmap);
-  if (visited != small_visited)
-    silc_free(visited);
-  return 1;
-}
-
-void re_compile_fastmap(bufp)
-SilcRegex bufp;
-{
-  if (!bufp->fastmap || bufp->fastmap_accurate)
-    return;
-  assert(bufp->used > 0);
-  if (!re_do_compile_fastmap(bufp->buffer, bufp->used, 0, &bufp->can_be_null,
-                            bufp->fastmap))
-    return;
-  if (bufp->buffer[0] == Cbol)
-    bufp->anchor = 1;   /* begline */
-  else
-    if (bufp->buffer[0] == Cbegbuf)
-      bufp->anchor = 2; /* begbuf */
-    else
-      bufp->anchor = 0; /* none */
-  bufp->fastmap_accurate = 1;
-}
-
-#define INITIAL_FAILURES  128  /* initial # failure points to allocate */
-#define MAX_FAILURES     4100  /* max # of failure points before failing */
+#undef ALLOC
+#undef STORE
+#undef CURRENT_LEVEL_START
+#undef SET_LEVEL_START
+#undef PUSH_LEVEL_STARTS
+#undef POP_LEVEL_STARTS
+#undef PUT_ADDR
+#undef INSERT_JUMP
+#undef SETBIT
+#undef SET_FIELDS
 
-int re_match_2(bufp, string1, size1, string2, size2, pos, regs, mstop, flags)
-SilcRegex bufp;
-char *string1, *string2;
-int size1, size2, pos, mstop;
-regexp_registers_t regs;
-unsigned int flags;
-{
-  struct failure_point { char *text, *partend, *code; }
-    *failure_stack_start, *failure_sp, *failure_stack_end,
-    initial_failure_stack[INITIAL_FAILURES];
-  char *code, *translate, *text, *textend, *partend, *part_2_end;
-  char *regstart_text[RE_NREGS], *regstart_partend[RE_NREGS];
-  char *regend_text[RE_NREGS], *regend_partend[RE_NREGS];
-  int a, b, ch, reg, regch, match_end;
-  char *regtext, *regpartend, *regtextend;
-
-#define PREFETCH                                       \
-  MACRO_BEGIN                                          \
-    if (text == partend)                               \
-      {                                                        \
-       if (text == textend)                            \
-         goto fail;                                    \
-       text = string2;                                 \
-       partend = part_2_end;                           \
-      }                                                        \
-  MACRO_END
+#define PREFETCH if (text == textend) goto fail
 
 #define NEXTCHAR(var)                          \
-  MACRO_BEGIN                                  \
-    PREFETCH;                                  \
-    (var) = (unsigned char)*text++;            \
-    if (translate)                             \
-      (var) = (unsigned char)translate[(var)]; \
-  MACRO_END
-
-  assert(pos >= 0 && size1 >= 0 && size2 >= 0 && mstop >= 0);
-  assert(mstop <= size1 + size2);
-  assert(pos <= mstop);
-
-  if (pos <= size1)
-    {
-      text = string1 + pos;
-      if (mstop <= size1)
-       {
-         partend = string1 + mstop;
-         textend = partend;
-       }
-      else
-       {
-         partend = string1 + size1;
-         textend = string2 + mstop - size1;
-       }
-      part_2_end = string2 + mstop - size1;
-    }
-  else
-    {
-      text = string2 + pos - size1;
-      partend = string2 + mstop - size1;
-      textend = partend;
-      part_2_end = partend;
-    }
+  PREFETCH;                                    \
+  var = (unsigned char)*text++;                        \
+  if (translate)                               \
+    var = translate[var]
 
-  if (bufp->uses_registers && regs != NULL)
-    for (a = 0; a < RE_NREGS; a++)
-      regend_text[a] = NULL;
+int re_match(SilcRegex bufp, unsigned char *string, int size, int pos,
+             regexp_registers_t old_regs, unsigned int flags)
+{
+  unsigned char *code;
+  unsigned char *translate;
+  unsigned char *text;
+  unsigned char *textstart;
+  unsigned char *textend;
+  int a;
+  int b;
+  int ch;
+  int reg;
+  int match_end;
+  unsigned char *regstart;
+  unsigned char *regend;
+  int regsize;
+  match_state state;
+
+  SILC_ASSERT(pos >= 0 && size >= 0);
+  SILC_ASSERT(pos <= size);
+
+  text = string + pos;
+  textstart = string;
+  textend = string + size;
 
   code = bufp->buffer;
+
   translate = bufp->translate;
-  failure_stack_start = failure_sp = initial_failure_stack;
-  failure_stack_end = initial_failure_stack + INITIAL_FAILURES;
-
-#if 0
-  /* re_search_2 has already done this, and otherwise we get little benefit
-     from this.  So I'll leave this out. */
-  if (bufp->fastmap_accurate && !bufp->can_be_null &&
-      text != textend &&
-      !bufp->fastmap[translate ?
-                    (unsigned char)translate[(unsigned char)*text] :
-                    (unsigned char)*text])
-    return -1;  /* it can't possibly match */
-#endif
+
+  NEW_STATE(state, bufp->num_registers);
 
  continue_matching:
-  for (;;)
+  switch (*code++)
     {
-      switch (*code++)
-       {
-       case Cend:
-         if (partend != part_2_end)
-           match_end = text - string1;
-         else
-           match_end = text - string2 + size1;
-         if (regs)
+    case Cend:
+      {
+       match_end = text - textstart;
+       if (old_regs)
+         {
+           old_regs->start[0] = pos;
+           old_regs->end[0] = match_end;
+           if (!bufp->uses_registers)
+             {
+               for (a = 1; a < RE_NREGS; a++)
+                 {
+                   old_regs->start[a] = -1;
+                   old_regs->end[a] = -1;
+                 }
+             }
+           else
+             {
+               for (a = 1; a < bufp->num_registers; a++)
+                 {
+                   if ((GET_REG_START(state, a) == NULL) ||
+                       (GET_REG_END(state, a) == NULL))
+                     {
+                       old_regs->start[a] = -1;
+                       old_regs->end[a] = -1;
+                       continue;
+                     }
+                   old_regs->start[a] = GET_REG_START(state, a) - textstart;
+                   old_regs->end[a] = GET_REG_END(state, a) - textstart;
+                 }
+               for (; a < RE_NREGS; a++)
+                 {
+                   old_regs->start[a] = -1;
+                   old_regs->end[a] = -1;
+                 }
+             }
+         }
+       FREE_STATE(state);
+       return match_end - pos;
+      }
+    case Cbol:
+      {
+       if (text == textstart || text[-1] == '\n') {
+         if (flags & RE_NOTBOL)
+           goto fail;
+         goto continue_matching;
+       }
+       goto fail;
+      }
+    case Ceol:
+      {
+       if (text == textend || *text == '\n') {
+         if (flags & RE_NOTEOL)
+           goto fail;
+         goto continue_matching;
+       }
+       goto fail;
+      }
+    case Cset:
+      {
+       NEXTCHAR(ch);
+       if (code[ch/8] & (1<<(ch & 7)))
+         {
+           code += 256/8;
+           goto continue_matching;
+         }
+       goto fail;
+      }
+    case Cexact:
+      {
+       NEXTCHAR(ch);
+       if (ch != (unsigned char)*code++)
+         goto fail;
+       goto continue_matching;
+      }
+    case Canychar:
+      {
+       NEXTCHAR(ch);
+       if (ch == '\n')
+         goto fail;
+       goto continue_matching;
+      }
+    case Cstart_memory:
+      {
+       reg = *code++;
+       SET_REG_START(state, reg, text, goto error);
+       goto continue_matching;
+      }
+    case Cend_memory:
+      {
+       reg = *code++;
+       SET_REG_END(state, reg, text, goto error);
+       goto continue_matching;
+      }
+    case Cmatch_memory:
+      {
+       reg = *code++;
+       regstart = GET_REG_START(state, reg);
+       regend = GET_REG_END(state, reg);
+       if ((regstart == NULL) || (regend == NULL))
+         goto fail;  /* or should we just match nothing? */
+       regsize = regend - regstart;
+
+       if (regsize > (textend - text))
+         goto fail;
+       if(translate)
+         {
+           for (; regstart < regend; regstart++, text++)
+             if (translate[*regstart] != translate[*text])
+               goto fail;
+         }
+       else
+         for (; regstart < regend; regstart++, text++)
+           if (*regstart != *text)
+             goto fail;
+       goto continue_matching;
+      }
+    case Cupdate_failure_jump:
+      {
+       UPDATE_FAILURE(state, text, goto error);
+       /* fall to next case */
+      }
+      /* treat Cstar_jump just like Cjump if it hasn't been optimized */
+    case Cstar_jump:
+    case Cjump:
+      {
+       a = (unsigned char)*code++;
+       a |= (unsigned char)*code++ << 8;
+       code += (int)SHORT(a);
+       if (code<bufp->buffer || bufp->buffer+bufp->used<code) {
+         silc_set_errno(SILC_ERR_OVERFLOW);
+         FREE_STATE(state);
+         return -2;
+       }
+       goto continue_matching;
+      }
+    case Cdummy_failure_jump:
+      {
+       unsigned char *failuredest;
+
+       a = (unsigned char)*code++;
+       a |= (unsigned char)*code++ << 8;
+       a = (int)SHORT(a);
+       SILC_ASSERT(*code == Cfailure_jump);
+       b = (unsigned char)code[1];
+       b |= (unsigned char)code[2] << 8;
+       failuredest = code + (int)SHORT(b) + 3;
+       if (failuredest<bufp->buffer || bufp->buffer+bufp->used < failuredest) {
+         silc_set_errno(SILC_ERR_OVERFLOW);
+         FREE_STATE(state);
+         return -2;
+       }
+       PUSH_FAILURE(state, failuredest, NULL, goto error);
+       code += a;
+       if (code<bufp->buffer || bufp->buffer+bufp->used < code) {
+         silc_set_errno(SILC_ERR_OVERFLOW);
+         FREE_STATE(state);
+         return -2;
+       }
+       goto continue_matching;
+      }
+    case Cfailure_jump:
+      {
+       a = (unsigned char)*code++;
+       a |= (unsigned char)*code++ << 8;
+       a = (int)SHORT(a);
+       if (code+a<bufp->buffer || bufp->buffer+bufp->used < code+a) {
+         silc_set_errno(SILC_ERR_OVERFLOW);
+         FREE_STATE(state);
+         return -2;
+       }
+       PUSH_FAILURE(state, code + a, text, goto error);
+       goto continue_matching;
+      }
+    case Crepeat1:
+      {
+       unsigned char *pinst;
+       a = (unsigned char)*code++;
+       a |= (unsigned char)*code++ << 8;
+       a = (int)SHORT(a);
+       pinst = code + a;
+       if (pinst<bufp->buffer || bufp->buffer+bufp->used<pinst) {
+         silc_set_errno(SILC_ERR_OVERFLOW);
+         FREE_STATE(state);
+         return -2;
+       }
+       /* pinst is sole instruction in loop, and it matches a
+        * single character.  Since Crepeat1 was originally a
+        * Cupdate_failure_jump, we also know that backtracking
+        * is useless: so long as the single-character
+        * expression matches, it must be used.  Also, in the
+        * case of +, we've already matched one character, so +
+        * can't fail: nothing here can cause a failure.  */
+       switch (*pinst++)
+         {
+         case Cset:
            {
-             regs->start[0] = pos;
-             regs->end[0] = match_end;
-             if (!bufp->uses_registers)
+             if (translate)
                {
-                 for (a = 1; a < RE_NREGS; a++)
+                 while (text < textend)
                    {
-                     regs->start[a] = -1;
-                     regs->end[a] = -1;
+                     ch = translate[(unsigned char)*text];
+                     if (pinst[ch/8] & (1<<(ch & 7)))
+                       text++;
+                     else
+                       break;
                    }
                }
              else
                {
-                 for (a = 1; a < RE_NREGS; a++)
+                 while (text < textend)
                    {
-                     if (regend_text[a] == NULL)
-                       {
-                         regs->start[a] = -1;
-                         regs->end[a] = -1;
-                         continue;
-                       }
-                     if (regstart_partend[a] != part_2_end)
-                       regs->start[a] = regstart_text[a] - string1;
+                     ch = (unsigned char)*text;
+                     if (pinst[ch/8] & (1<<(ch & 7)))
+                       text++;
                      else
-                       regs->start[a] = regstart_text[a] - string2 + size1;
-                     if (regend_partend[a] != part_2_end)
-                       regs->end[a] = regend_text[a] - string1;
-                     else
-                       regs->end[a] = regend_text[a] - string2 + size1;
+                       break;
                    }
                }
+             break;
            }
-         if (failure_stack_start != initial_failure_stack)
-           silc_free((char *)failure_stack_start);
-         return match_end - pos;
-       case Cbol:
-         if (text == string1 || text[-1] == '\n') { /* text[-1] always valid */
-           if (flags & RE_NOTBOL)
-             goto fail;
-           break;
-         }
-         goto fail;
-       case Ceol:
-         if (text == string2 + size2 ||
-             (text == string1 + size1 ?
-              (size2 == 0 || *string2 == '\n') :
-              *text == '\n')) {
-           if (flags & RE_NOTEOL)
-             goto fail;
-           break;
-         }
-         goto fail;
-       case Cset:
-         NEXTCHAR(ch);
-         if (code[ch/8] & (1<<(ch & 7)))
+         case Cexact:
            {
-             code += 256/8;
+             ch = (unsigned char)*pinst;
+             if (translate)
+               {
+                 while (text < textend &&
+                        translate[(unsigned char)*text] == ch)
+                   text++;
+               }
+             else
+               {
+                 while (text < textend && (unsigned char)*text == ch)
+                   text++;
+               }
              break;
            }
-         goto fail;
-       case Cexact:
-         NEXTCHAR(ch);
-         if (ch != (unsigned char)*code++)
-           goto fail;
-         break;
-       case Canychar:
-         NEXTCHAR(ch);
-         if (ch == '\n')
-           goto fail;
-         break;
-       case Cstart_memory:
-         reg = *code++;
-         regstart_text[reg] = text;
-         regstart_partend[reg] = partend;
-         break;
-       case Cend_memory:
-         reg = *code++;
-         regend_text[reg] = text;
-         regend_partend[reg] = partend;
-         break;
-       case Cmatch_memory:
-         reg = *code++;
-         if (regend_text[reg] == NULL)
-           goto fail;  /* or should we just match nothing? */
-         regtext = regstart_text[reg];
-         regtextend = regend_text[reg];
-         if (regstart_partend[reg] == regend_partend[reg])
-           regpartend = regtextend;
-         else
-           regpartend = string1 + size1;
-
-         for (;regtext != regtextend;)
+         case Canychar:
            {
-             NEXTCHAR(ch);
-             if (regtext == regpartend)
-               regtext = string2;
-             regch = (unsigned char)*regtext++;
-             if (translate)
-               regch = (unsigned char)translate[regch];
-             if (regch != ch)
-               goto fail;
+             while (text < textend && (unsigned char)*text != '\n')
+               text++;
+             break;
            }
-         break;
-       case Cstar_jump:
-         /* star is coded as:
-              1: failure_jump 2
-                 ... code for operand of star
-                 star_jump 1
-              2: ... code after star
-            We change the star_jump to update_failure_jump if we can determine
-            that it is safe to do so; otherwise we change it to an ordinary
-            jump.
-            plus is coded as
-                 jump 2
-              1: failure_jump 3
-              2: ... code for operand of plus
-                 star_jump 1
-              3: ... code after plus
-            For star_jump considerations this is processed identically
-            to star. */
-         a = (unsigned char)*code++;
-         a |= (unsigned char)*code++ << 8;
-         a = (int)(short)a;
-         {
-           char map[256], can_be_null;
-           char *p1, *p2;
-
-           p1 = code + a + 3; /* skip the failure_jump */
-           assert(p1[-3] == Cfailure_jump);
-           p2 = code;
-           /* p1 points inside loop, p2 points to after loop */
-           if (!re_do_compile_fastmap(bufp->buffer, bufp->used,
-                                      p2 - bufp->buffer, &can_be_null, map))
-             goto make_normal_jump;
-           /* If we might introduce a new update point inside the loop,
-              we can't optimize because then update_jump would update a
-              wrong failure point.  Thus we have to be quite careful here. */
-         loop_p1:
-           /* loop until we find something that consumes a character */
-           switch (*p1++)
-             {
-              case Cbol:
-              case Ceol:
-              case Cbegbuf:
-              case Cendbuf:
-              case Cwordbeg:
-              case Cwordend:
-              case Cwordbound:
-              case Cnotwordbound:
-#ifdef emacs
-              case Cemacs_at_dot:
-#endif /* emacs */
-                goto loop_p1;
-              case Cstart_memory:
-              case Cend_memory:
-                p1++;
-                goto loop_p1;
-             case Cexact:
-               ch = (unsigned char)*p1++;
-               if (map[ch])
-                 goto make_normal_jump;
-               break;
-             case Canychar:
-               for (b = 0; b < 256; b++)
-                 if (b != '\n' && map[b])
-                   goto make_normal_jump;
-               break;
-             case Cset:
-               for (b = 0; b < 256; b++)
-                 if ((p1[b >> 3] & (1 << (b & 7))) && map[b])
-                   goto make_normal_jump;
-               p1 += 256/8;
-               break;
-             default:
-               goto make_normal_jump;
-             }
-           /* now we know that we can't backtrack. */
-           while (p1 != p2 - 3)
-             {
-               switch (*p1++)
-                 {
-                 case Cend:
-                   abort();  /* we certainly shouldn't get this inside loop */
-                   /*NOTREACHED*/
-                 case Cbol:
-                 case Ceol:
-                 case Canychar:
-                 case Cbegbuf:
-                 case Cendbuf:
-                 case Cwordbeg:
-                 case Cwordend:
-                 case Cwordbound:
-                 case Cnotwordbound:
-#ifdef emacs
-                 case Cemacs_at_dot:
-#endif /* emacs */
-                   break;
-                 case Cset:
-                   p1 += 256/8;
-                   break;
-                 case Cexact:
-                 case Cstart_memory:
-                 case Cend_memory:
-                 case Cmatch_memory:
-                 case Csyntaxspec:
-                 case Cnotsyntaxspec:
-                   p1++;
-                   break;
-                 case Cjump:
-                 case Cstar_jump:
-                 case Cfailure_jump:
-                 case Cupdate_failure_jump:
-                 case Cdummy_failure_jump:
-                   goto make_normal_jump;
-                 default:
-                   printf("regexpr.c: processing star_jump: unknown op %d\n", p1[-1]);
-                   break;
-                 }
-             }
-           goto make_update_jump;
-         }
-       make_normal_jump:
-         /* printf("changing to normal jump\n"); */
-         code -= 3;
-         *code = Cjump;
-         break;
-       make_update_jump:
-         /* printf("changing to update jump\n"); */
-         code -= 2;
-         a += 3;  /* jump to after the Cfailure_jump */
-         code[-1] = Cupdate_failure_jump;
-         code[0] = a & 0xff;
-         code[1] = a >> 8;
-         /* fall to next case */
-       case Cupdate_failure_jump:
-         failure_sp[-1].text = text;
-         failure_sp[-1].partend = partend;
-         /* fall to next case */
-       case Cjump:
-         a = (unsigned char)*code++;
-         a |= (unsigned char)*code++ << 8;
-         code += (int)(short)a;
-         break;
-       case Cdummy_failure_jump:
-       case Cfailure_jump:
-         if (failure_sp == failure_stack_end)
+         case Csyntaxspec:
            {
-             if (failure_stack_start != initial_failure_stack)
-               goto error;
-             failure_stack_start = (struct failure_point *)
-               silc_malloc(MAX_FAILURES * sizeof(*failure_stack_start));
-             failure_stack_end = failure_stack_start + MAX_FAILURES;
-             memcpy((char *)failure_stack_start, (char *)initial_failure_stack,
-                    INITIAL_FAILURES * sizeof(*failure_stack_start));
-             failure_sp = failure_stack_start + INITIAL_FAILURES;
+             a = (unsigned char)*pinst;
+             if (translate)
+               {
+                 while (text < textend &&
+                        (SYNTAX(translate[*text]) & a) )
+                   text++;
+               }
+             else
+               {
+                 while (text < textend && (SYNTAX(*text) & a) )
+                   text++;
+               }
+             break;
            }
-         a = (unsigned char)*code++;
-         a |= (unsigned char)*code++ << 8;
-         a = (int)(short)a;
-         if (code[-3] == Cdummy_failure_jump)
-           { /* this is only used in plus */
-             assert(*code == Cfailure_jump);
-             b = (unsigned char)code[1];
-             b |= (unsigned char)code[2] << 8;
-             failure_sp->code = code + (int)(short)b + 3;
-             failure_sp->text = NULL;
-             code += a;
+         case Cnotsyntaxspec:
+           {
+             a = (unsigned char)*pinst;
+             if (translate)
+               {
+                 while (text < textend &&
+                        !(SYNTAX(translate[*text]) & a) )
+                   text++;
+               }
+             else
+               {
+                 while (text < textend && !(SYNTAX(*text) & a) )
+                   text++;
+               }
+             break;
            }
-         else
+         default:
            {
-             failure_sp->code = code + a;
-             failure_sp->text = text;
-             failure_sp->partend = partend;
+             FREE_STATE(state);
+             silc_set_errno(SILC_ERR_REGEX_OPCODE);
+             return -2;
+             /*NOTREACHED*/
            }
-         failure_sp++;
-         break;
-       case Cbegbuf:
-         if (text == string1)
-           break;
+         }
+       /* due to the funky way + and * are compiled, the top
+        * failure- stack entry at this point is actually a
+        * success entry -- update it & pop it */
+       UPDATE_FAILURE(state, text, goto error);
+       goto fail;      /* i.e., succeed <wink/sigh> */
+      }
+    case Cbegbuf:
+      {
+       if (text == textstart)
+         goto continue_matching;
+       goto fail;
+      }
+    case Cendbuf:
+      {
+       if (text == textend)
+         goto continue_matching;
+       goto fail;
+      }
+    case Cwordbeg:
+      {
+       if (text == textend)
          goto fail;
-       case Cendbuf:
-         if (size2 == 0 ? text == string1 + size1 : text == string2 + size2)
-           break;
+       if (!(SYNTAX(*text) & Sword))
          goto fail;
-       case Cwordbeg:
-         if (text == string2 + size2)
-           goto fail;
-         if (size2 == 0 && text == string1 + size1)
-           goto fail;
-         if (SYNTAX(text == string1 + size1 ? *string1 : *text) != Sword)
-           goto fail;
-         if (text == string1)
-           break;
-         if (SYNTAX(text[-1]) != Sword)
-           break;
+       if (text == textstart)
+         goto continue_matching;
+       if (!(SYNTAX(text[-1]) & Sword))
+         goto continue_matching;
+       goto fail;
+      }
+    case Cwordend:
+      {
+       if (text == textstart)
          goto fail;
-       case Cwordend:
-         if (text == string1)
-           goto fail;
-         if (SYNTAX(text[-1]) != Sword)
-           goto fail;
-         if (text == string2 + size2)
-           break;
-         if (size2 == 0 && text == string1 + size1)
-           break;
-         if (SYNTAX(*text) == Sword)
-           goto fail;
-         break;
-       case Cwordbound:
-         /* Note: as in gnu regexp, this also matches at the beginning
-            and end of buffer. */
-         if (text == string1 || text == string2 + size2 ||
-             (size2 == 0 && text == string1 + size1))
-           break;
-         if ((SYNTAX(text[-1]) == Sword) ^
-             (SYNTAX(text == string1 + size1 ? *string2 : *text) == Sword))
-           break;
+       if (!(SYNTAX(text[-1]) & Sword))
          goto fail;
-       case Cnotwordbound:
-         /* Note: as in gnu regexp, this never matches at the beginning
-            and end of buffer. */
-         if (text == string1 || text == string2 + size2 ||
-             (size2 == 0 && text == string1 + size1))
-           goto fail;
-         if (!((SYNTAX(text[-1]) == Sword) ^
-               (SYNTAX(text == string1 + size1 ? *string2 : *text) == Sword)))
-           goto fail;
-         break;
-       case Csyntaxspec:
-         NEXTCHAR(ch);
-         if (SYNTAX(ch) != (unsigned char)*code++)
-           goto fail;
-         break;
-       case Cnotsyntaxspec:
-         NEXTCHAR(ch);
-         if (SYNTAX(ch) != (unsigned char)*code++)
-           break;
+       if (text == textend)
+         goto continue_matching;
+       if (!(SYNTAX(*text) & Sword))
+         goto continue_matching;
+       goto fail;
+      }
+    case Cwordbound:
+      {
+       /* Note: as in gnu regexp, this also matches at the
+        * beginning and end of buffer.  */
+
+       if (text == textstart || text == textend)
+         goto continue_matching;
+       if ((SYNTAX(text[-1]) & Sword) ^ (SYNTAX(*text) & Sword))
+         goto continue_matching;
+       goto fail;
+      }
+    case Cnotwordbound:
+      {
+       /* Note: as in gnu regexp, this never matches at the
+        * beginning and end of buffer.  */
+       if (text == textstart || text == textend)
          goto fail;
-#ifdef emacs
-       case Cemacs_at_dot:
-         if (PTR_CHAR_POS((unsigned char *)text) + 1 != point)
-           goto fail;
-         break;
-#endif /* emacs */
-       default:
-         abort();
-         /*NOTREACHED*/
-       }
+       if (!((SYNTAX(text[-1]) & Sword) ^ (SYNTAX(*text) & Sword)))
+         goto continue_matching;
+       goto fail;
+      }
+    case Csyntaxspec:
+      {
+       NEXTCHAR(ch);
+       if (!(SYNTAX(ch) & (unsigned char)*code++))
+         goto fail;
+       goto continue_matching;
+      }
+    case Cnotsyntaxspec:
+      {
+       NEXTCHAR(ch);
+       if (SYNTAX(ch) & (unsigned char)*code++)
+         goto fail;
+       goto continue_matching;
+      }
+    default:
+      {
+       FREE_STATE(state);
+       silc_set_errno(SILC_ERR_REGEX_OPCODE);
+       return -2;
+       /*NOTREACHED*/
+      }
     }
-  abort();
-  /*NOTREACHED*/
 
+  /* Using "break;" in the above switch statement is equivalent to
+     "goto fail;" */
  fail:
-  if (failure_sp != failure_stack_start)
-    {
-      failure_sp--;
-      text = failure_sp->text;
-      if (text == NULL)
-       goto fail;
-      partend = failure_sp->partend;
-      code = failure_sp->code;
-      goto continue_matching;
-    }
-  if (failure_stack_start != initial_failure_stack)
-    silc_free((char *)failure_stack_start);
+  POP_FAILURE(state, code, text, goto done_matching, goto error);
+  goto continue_matching;
+
+ done_matching:
+  /*   if(translated != NULL) */
+  /*      free(translated); */
+  FREE_STATE(state);
   return -1;
 
  error:
-  if (failure_stack_start != initial_failure_stack)
-    silc_free((char *)failure_stack_start);
+  /*   if (translated != NULL) */
+  /*      free(translated); */
+  FREE_STATE(state);
   return -2;
 }
 
 #undef PREFETCH
 #undef NEXTCHAR
-#undef PUSH_FAILURE
-
-int re_match(bufp, string, size, pos, regs, flags)
-SilcRegex bufp;
-char *string;
-int size, pos;
-regexp_registers_t regs;
-unsigned int flags;
-{
-  return re_match_2(bufp, string, size, (char *)NULL, 0, pos, regs, size,
-                   flags);
-}
 
-int re_search_2(bufp, string1, size1, string2, size2, pos, range, regs,
-               mstop, flags)
-SilcRegex bufp;
-char *string1, *string2;
-int size1, size2, pos, range, mstop;
-regexp_registers_t regs;
-unsigned int flags;
+int re_search(SilcRegex bufp, unsigned char *string, int size, int pos,
+              int range, regexp_registers_t regs, unsigned int flags)
 {
-  char *fastmap, *translate, *text, *partstart, *partend;
-  int dir, ret;
-  char anchor;
+  unsigned char *fastmap;
+  unsigned char *translate;
+  unsigned char *text;
+  unsigned char *partstart;
+  unsigned char *partend;
+  int dir;
+  int ret;
+  unsigned char anchor;
 
-  assert(size1 >= 0 && size2 >= 0 && pos >= 0 && mstop >= 0);
-  assert(pos + range >= 0 && pos + range <= size1 + size2);
-  assert(pos <= mstop);
+  SILC_ASSERT(size >= 0 && pos >= 0);
+  SILC_ASSERT(pos + range >= 0 && pos + range <= size); /* Bugfix by ylo */
 
   fastmap = bufp->fastmap;
   translate = bufp->translate;
-  if (fastmap && !bufp->fastmap_accurate)
-    re_compile_fastmap(bufp);
+  if (fastmap && !bufp->fastmap_accurate) {
+    if (re_compile_fastmap(bufp))
+      return -2;
+  }
+
   anchor = bufp->anchor;
   if (bufp->can_be_null == 1) /* can_be_null == 2: can match null at eob */
     fastmap = NULL;
+
   if (range < 0)
     {
       dir = -1;
@@ -1486,60 +2214,40 @@ unsigned int flags;
     }
   else
     dir = 1;
+
   if (anchor == 2) {
     if (pos != 0)
       return -1;
     else
       range = 0;
   }
+
   for (; range >= 0; range--, pos += dir)
     {
       if (fastmap)
        {
          if (dir == 1)
            { /* searching forwards */
-             if (pos < size1)
-               {
-                 text = string1 + pos;
-                 if (pos + range > size1)
-                   partend = string1 + size1;
-                 else
-                   partend = string1 + pos + range;
-               }
-             else
-               {
-                 text = string2 + pos - size1;
-                 partend = string2 + pos + range - size1;
-               }
+
+             text = string + pos;
+             partend = string + size;
              partstart = text;
              if (translate)
                while (text != partend &&
-                      !fastmap[(unsigned char)
-                               translate[(unsigned char)*text]])
+                      !fastmap[(unsigned char) translate[(unsigned char)*text]])
                  text++;
              else
                while (text != partend && !fastmap[(unsigned char)*text])
                  text++;
              pos += text - partstart;
              range -= text - partstart;
-             if (pos == size1 + size2 && bufp->can_be_null == 0)
+             if (pos == size && bufp->can_be_null == 0)
                return -1;
            }
          else
            { /* searching backwards */
-             if (pos <= size1)
-               {
-                 text = string1 + pos;
-                 partstart = string1 + pos - range;
-               }
-             else
-               {
-                 text = string2 + pos - size1;
-                 if (range < pos - size1)
-                   partstart = string2 + pos - size1 - range;
-                 else
-                   partstart = string2;
-               }
+             text = string + pos;
+             partstart = string + pos - range;
              partend = text;
              if (translate)
                while (text != partstart &&
@@ -1556,14 +2264,11 @@ unsigned int flags;
        }
       if (anchor == 1)
        { /* anchored to begline */
-         if (pos > 0 &&
-             (pos <= size1 ? string1[pos - 1] :
-              string2[pos - size1 - 1]) != '\n')
+         if (pos > 0 && (string[pos - 1] != '\n'))
            continue;
        }
-      assert(pos >= 0 && pos <= size1 + size2);
-      ret = re_match_2(bufp, string1, size1, string2, size2, pos, regs, mstop,
-                      flags);
+      SILC_ASSERT(pos >= 0 && pos <= size);
+      ret = re_match(bufp, string, size, pos, regs, flags);
       if (ret >= 0)
        return pos;
       if (ret == -2)
@@ -1572,17 +2277,6 @@ unsigned int flags;
   return -1;
 }
 
-int re_search(bufp, string, size, startpos, range, regs, flags)
-SilcRegex bufp;
-char *string;
-int size, startpos, range;
-regexp_registers_t regs;
-unsigned int flags;
-{
-  return re_search_2(bufp, string, size, (char *)NULL, 0,
-                    startpos, range, regs, size, flags);
-}
-
 /****************************** SILC Regex API ******************************/
 
 /* Compile regular expression */
@@ -1600,8 +2294,14 @@ SilcBool silc_regex_compile(SilcRegex regexp, const char *regex,
 
   memset(regexp, 0, sizeof(*regexp));
 
+  /* Get global stack, if set, and create child stack. */
+  regexp->rstack = silc_stack_get_global();
+  if (regexp->rstack)
+    regexp->rstack = silc_stack_alloc(512, regexp->rstack);
+
   /* Set syntax */
-  syntax |= (RE_CONTEXT_INDEP_OPS | RE_NO_BK_PARENS | RE_NO_BK_VBAR);
+  syntax |= (RE_CONTEXT_INDEP_OPS | RE_NO_BK_PARENS |
+            RE_NO_BK_VBAR | RE_REPEAT);
   re_set_syntax(syntax);
 
   /* Compile */
@@ -1646,9 +2346,7 @@ SilcBool silc_regex_match(SilcRegex regexp, const char *string,
   ret = re_search(regexp, (char *)string, string_len, 0, string_len,
                  num_match ? &regs : NULL, f);
   if (ret < 0) {
-    if (ret == -2)
-      silc_set_errno(SILC_ERR);
-    else
+    if (ret == -1)
       silc_set_errno(SILC_ERR_NOT_FOUND);
   }
 
@@ -1667,7 +2365,8 @@ SilcBool silc_regex_match(SilcRegex regexp, const char *string,
 
 void silc_regex_free(SilcRegex regexp)
 {
-  silc_free(regexp->buffer);
+  silc_sfree(regexp->rstack, regexp->buffer);
+  silc_stack_free(regexp->rstack);
 }
 
 /* Match string */
index 2adddb37876a4edbbece4d2c5065ad501b7f8f36..b44456f1f726257e84c1453852bdf36ccbfd40a6 100644 (file)
@@ -1,25 +1,19 @@
 /*
 
-  regexpr.h
+  silcregex.h
 
-  Author: Tatu Ylonen <ylo@ngs.fi>
+  Author: Pekka Riikonen <priikone@silcnet.org>
 
-  Copyright (c) 1991 Tatu Ylonen, Espoo, Finland
+  Copyright (C) 2007 Pekka Riikonen
 
-  Permission to use, copy, modify, distribute, and sell this software
-  and its documentation is hereby granted without fee, provided that the
-  above copyright notice appears in all source code copies, the name of
-  Tatu Ylonen is not used to advertise products containing this software
-  or a derivation thereof, and all modified versions are clearly marked
-  as such.
+  This program is free software; you can redistribute it and/or modify
+  it under the terms of the GNU General Public License as published by
+  the Free Software Foundation; version 2 of the License.
 
-  This software is provided "as is" without express or implied warranty.
-
-  Created: Thu Sep 26 17:15:36 1991 ylo
-  Last modified: Fri Jan  3 12:05:45 1992 ylo
-
-  The SILC Regex API by Pekka Riikonen, under the same license as the original
-  code.
+  This program is distributed in the hope that it will be useful,
+  but WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+  GNU General Public License for more details.
 
 */
 
  *
  ***/
 typedef struct SilcRegexObject {
-  char *buffer;                       /* compiled pattern */
-  int allocated;              /* allocated size of compiled pattern */
-  int used;                   /* actual length of compiled pattern */
+  SilcStack rstack;           /* Stack for fast allocations */
+  unsigned char *buffer;       /* compiled pattern */
   char *fastmap;               /* fastmap[ch] is true if ch can start pattern */
   char *translate;            /* translation to apply during comp/match */
+  int allocated;              /* allocated size of compiled pattern */
+  int used;                   /* actual length of compiled pattern */
+  int num_registers;          /* number of registers used */
   char fastmap_accurate;       /* true if fastmap is valid */
   char can_be_null;           /* true if can match empty string */
   char uses_registers;         /* registers used and need to be initialized */
index 87b7a5f40b2628e1ac0783f8efc55608d5c70774..0e7f4527125406c92323ec375721b69a91cec1cf 100644 (file)
@@ -9,6 +9,7 @@ int main(int argc, char **argv)
   SilcRegexMatchStruct match[10];
   int i, num_match = 10;
   char *regex, *string, *sub;
+  SilcBufferStruct bmatch;
 
   if (argc > 1 && !strcmp(argv[1], "-d")) {
     silc_log_debug(TRUE);
@@ -17,6 +18,181 @@ int main(int argc, char **argv)
     silc_log_set_debug_string("*regex*,*errno*");
   }
 
+  regex = "^a{0}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "^a{0,}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "bbbb";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "^a{0,}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "aaaaaaaaa";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "^a{0,0}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "a";
+  SILC_LOG_DEBUG(("DO NOT Match %s", string));
+  if (silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+
+  regex = "^a{3}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "aaa";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "^a{3}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "aaaa";
+  SILC_LOG_DEBUG(("DO NOT Match %s", string));
+  if (silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+
+  regex = "^a{3,5}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "aaa";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "^a{3,5}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "aaaa";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "^a{3,5}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "aaaaaa";
+  SILC_LOG_DEBUG(("DO NOT Match %s", string));
+  if (silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+
+  regex = "^a{3,}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "aaa";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "^a{3,}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "aaaaaaaaaaaaa";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "^a{3,}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "aa";
+  SILC_LOG_DEBUG(("DO NOT Match %s", string));
+  if (silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+
+
+  regex = "a*b";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "b";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "a*b";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "ab";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "a*b";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "aaaab";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+
+  regex = "a+b";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "ab";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "a+b";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "aaaab";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "a+b";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "b";
+  SILC_LOG_DEBUG(("DO NOT Match %s", string));
+  if (silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+
+
+  regex = "ca?b";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "cb";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "ca?b";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "cab";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_buffer_printf(&bmatch, TRUE);
+
+  regex = "ca?b";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "caab";
+  SILC_LOG_DEBUG(("DO NOT Match %s", string));
+  if (silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+
+
+  regex = "^a{2,3}$";
+  SILC_LOG_DEBUG(("Regex %s", regex));
+  string = "aaa";
+  SILC_LOG_DEBUG(("Match %s", string));
+  if (!silc_regex(string, regex, &bmatch, NULL))
+    goto err;
+  silc_file_write(1, silc_buffer_data(&bmatch), silc_buffer_len(&bmatch));
+  fflush(stdout);
+
   regex = "(H..).(o..)";
   SILC_LOG_DEBUG(("Regex %s", regex));
   if (!silc_regex_compile(&reg, regex, 0))