/* Extended regular expression matching and search library,
- version 0.11.
+ version 0.12.
(Implements POSIX draft P10003.2/D11.2, except for
internationalization features.)
- Copyright (C) 1985, 89, 90, 91, 92 Free Software Foundation, Inc.
+ Copyright (C) 1993, 1994, 1995 Free Software Foundation, Inc.
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
#define _GNU_SOURCE
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
/* We need this for `regex.h', and perhaps for the Emacs include files. */
#include <sys/types.h>
+/* This is for other GNU distributions with internationalized messages. */
+#if HAVE_LIBINTL_H || defined (_LIBC)
+# include <libintl.h>
+#else
+# define gettext(msgid) (msgid)
+#endif
+
/* The `emacs' switch turns on certain matching commands
that make sense only in Emacs. */
#ifdef emacs
-#include "config.h"
#include "lisp.h"
#include "buffer.h"
#include "syntax.h"
-/* Emacs uses `NULL' as a predicate. */
-#undef NULL
-
#else /* not emacs */
-/* We used to test for `BSTRING' here, but only GCC and Emacs define
- `BSTRING', as far as I know, and neither of them use this code. */
-#if USG || STDC_HEADERS
+/* If we are not linking with Emacs proper,
+ we can't use the relocating allocator
+ even if config.h says that we can. */
+#undef REL_ALLOC
+
+#if defined (STDC_HEADERS) || defined (_LIBC)
+#include <stdlib.h>
+#else
+char *malloc ();
+char *realloc ();
+#endif
+
+/* When used in Emacs's lib-src, we need to get bzero and bcopy somehow.
+ If nothing else has been done, use the method below. */
+#ifdef INHIBIT_STRING_HEADER
+#if !(defined (HAVE_BZERO) && defined (HAVE_BCOPY))
+#if !defined (bzero) && !defined (bcopy)
+#undef INHIBIT_STRING_HEADER
+#endif
+#endif
+#endif
+
+/* This is the normal way of making sure we have a bcopy and a bzero.
+ This is used in most programs--a few other programs avoid this
+ by defining INHIBIT_STRING_HEADER. */
+#ifndef INHIBIT_STRING_HEADER
+#if defined (HAVE_STRING_H) || defined (STDC_HEADERS) || defined (_LIBC)
#include <string.h>
+#ifndef bcmp
#define bcmp(s1, s2, n) memcmp ((s1), (s2), (n))
+#endif
+#ifndef bcopy
#define bcopy(s, d, n) memcpy ((d), (s), (n))
+#endif
+#ifndef bzero
#define bzero(s, n) memset ((s), 0, (n))
+#endif
#else
#include <strings.h>
#endif
-
-#ifdef STDC_HEADERS
-#include <stdlib.h>
-#else
-char *malloc ();
-char *realloc ();
#endif
-
/* Define the syntax stuff for \<, \>, etc. */
/* This must be nonzero for the wordchar and notwordchar pattern
#define Sword 1
#endif
+#ifdef SWITCH_ENUM_BUG
+#define SWITCH_ENUM_CAST(x) ((int)(x))
+#else
+#define SWITCH_ENUM_CAST(x) (x)
+#endif
+
#ifdef SYNTAX_TABLE
extern char *re_syntax_table;
/* Get the interface, including the syntax bits. */
#include "regex.h"
-
/* isalpha etc. are used for the character classes. */
#include <ctype.h>
-#ifndef isgraph
-#define isgraph(c) (isprint (c) && !isspace (c))
+
+/* Jim Meyering writes:
+
+ "... Some ctype macros are valid only for character codes that
+ isascii says are ASCII (SGI's IRIX-4.0.5 is one such system --when
+ using /bin/cc or gcc but without giving an ansi option). So, all
+ ctype uses should be through macros like ISPRINT... If
+ STDC_HEADERS is defined, then autoconf has verified that the ctype
+ macros don't need to be guarded with references to isascii. ...
+ Defining isascii to 1 should let any compiler worth its salt
+ eliminate the && through constant folding." */
+
+#if defined (STDC_HEADERS) || (!defined (isascii) && !defined (HAVE_ISASCII))
+#define ISASCII(c) 1
+#else
+#define ISASCII(c) isascii(c)
+#endif
+
+#ifdef isblank
+#define ISBLANK(c) (ISASCII (c) && isblank (c))
+#else
+#define ISBLANK(c) ((c) == ' ' || (c) == '\t')
#endif
-#ifndef isblank
-#define isblank(c) ((c) == ' ' || (c) == '\t')
+#ifdef isgraph
+#define ISGRAPH(c) (ISASCII (c) && isgraph (c))
+#else
+#define ISGRAPH(c) (ISASCII (c) && isprint (c) && !isspace (c))
#endif
+#define ISPRINT(c) (ISASCII (c) && isprint (c))
+#define ISDIGIT(c) (ISASCII (c) && isdigit (c))
+#define ISALNUM(c) (ISASCII (c) && isalnum (c))
+#define ISALPHA(c) (ISASCII (c) && isalpha (c))
+#define ISCNTRL(c) (ISASCII (c) && iscntrl (c))
+#define ISLOWER(c) (ISASCII (c) && islower (c))
+#define ISPUNCT(c) (ISASCII (c) && ispunct (c))
+#define ISSPACE(c) (ISASCII (c) && isspace (c))
+#define ISUPPER(c) (ISASCII (c) && isupper (c))
+#define ISXDIGIT(c) (ISASCII (c) && isxdigit (c))
+
#ifndef NULL
-#define NULL 0
+#define NULL (void *)0
#endif
/* We remove any previous definition of `SIGN_EXTEND_CHAR',
#undef SIGN_EXTEND_CHAR
#if __STDC__
#define SIGN_EXTEND_CHAR(c) ((signed char) (c))
-#else
+#else /* not __STDC__ */
/* As in Harbison and Steele. */
#define SIGN_EXTEND_CHAR(c) ((((unsigned char) (c)) ^ 128) - 128)
#endif
#define REGEX_ALLOCATE malloc
#define REGEX_REALLOCATE(source, osize, nsize) realloc (source, nsize)
+#define REGEX_FREE free
#else /* not REGEX_MALLOC */
#include <alloca.h>
#else /* not __GNUC__ or HAVE_ALLOCA_H */
#ifndef _AIX /* Already did AIX, up at the top. */
+#if defined (__STDC__) && __STDC__
+void *alloca ();
+#else
char *alloca ();
+#endif
#endif /* not _AIX */
#endif /* not HAVE_ALLOCA_H */
#endif /* not __GNUC__ */
bcopy (source, destination, osize), \
destination)
+/* No need to do anything to free, after alloca. */
+#define REGEX_FREE(arg) ((void)0) /* Do nothing! But inhibit gcc warning. */
+
#endif /* not REGEX_MALLOC */
+/* Define how to allocate the failure stack. */
+
+#if defined (REL_ALLOC) && defined (REGEX_MALLOC)
+
+#define REGEX_ALLOCATE_STACK(size) \
+ r_alloc (&failure_stack_ptr, (size))
+#define REGEX_REALLOCATE_STACK(source, osize, nsize) \
+ r_re_alloc (&failure_stack_ptr, (nsize))
+#define REGEX_FREE_STACK(ptr) \
+ r_alloc_free (&failure_stack_ptr)
+
+#else /* not using relocating allocator */
+
+#ifdef REGEX_MALLOC
+
+#define REGEX_ALLOCATE_STACK malloc
+#define REGEX_REALLOCATE_STACK(source, osize, nsize) realloc (source, nsize)
+#define REGEX_FREE_STACK free
+
+#else /* not REGEX_MALLOC */
+
+#define REGEX_ALLOCATE_STACK alloca
+
+#define REGEX_REALLOCATE_STACK(source, osize, nsize) \
+ REGEX_REALLOCATE (source, osize, nsize)
+/* No need to explicitly free anything. */
+#define REGEX_FREE_STACK(arg)
+
+#endif /* not REGEX_MALLOC */
+#endif /* not using relocating allocator */
+
/* True if `size1' is non-NULL and PTR is pointing anywhere inside
`string1' or just past its end. This works if PTR is NULL, which is
/* (Re)Allocate N items of type T using malloc, or fail. */
#define TALLOC(n, t) ((t *) malloc ((n) * sizeof (t)))
#define RETALLOC(addr, n, t) ((addr) = (t *) realloc (addr, (n) * sizeof (t)))
+#define RETALLOC_IF(addr, n, t) \
+ if (addr) RETALLOC((addr), (n), t); else (addr) = TALLOC ((n), t)
#define REGEX_TALLOC(n, t) ((t *) REGEX_ALLOCATE ((n) * sizeof (t)))
#define BYTEWIDTH 8 /* In bits. */
#define STREQ(s1, s2) ((strcmp (s1, s2) == 0))
+#undef MAX
+#undef MIN
#define MAX(a, b) ((a) > (b) ? (a) : (b))
#define MIN(a, b) ((a) < (b) ? (a) : (b))
typedef char boolean;
#define false 0
#define true 1
+
+static int re_match_2_internal ();
\f
/* These are the command codes that appear in compiled regular
expressions. Some opcodes are followed by argument bytes. A
command code can specify any interpretation whatsoever for its
- arguments. Zero bytes may appear in the compiled regular expression.
-
- The value of `exactn' is needed in search.c (search_buffer) in Emacs.
- So regex.h defines a symbol `RE_EXACTN_VALUE' to be 1; the value of
- `exactn' we use here must also be 1. */
+ arguments. Zero bytes may appear in the compiled regular expression. */
typedef enum
{
no_op = 0,
+ /* Succeed right away--no more backtracking. */
+ succeed,
+
/* Followed by one byte giving n, then by n literal bytes. */
- exactn = 1,
+ exactn,
/* Matches any (more or less) character. */
anychar,
#define DEBUG_PRINT1(x) if (debug) printf (x)
#define DEBUG_PRINT2(x1, x2) if (debug) printf (x1, x2)
#define DEBUG_PRINT3(x1, x2, x3) if (debug) printf (x1, x2, x3)
+#define DEBUG_PRINT4(x1, x2, x3, x4) if (debug) printf (x1, x2, x3, x4)
#define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \
if (debug) print_partial_compiled_pattern (s, e)
#define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \
if (debug) print_double_string (w, s1, sz1, s2, sz2)
-extern void printchar ();
-
/* Print the fastmap in human-readable form. */
void
if (fastmap[i++])
{
was_a_range = 0;
- printchar (i - 1);
+ putchar (i - 1);
while (i < (1 << BYTEWIDTH) && fastmap[i])
{
was_a_range = 1;
if (was_a_range)
{
printf ("-");
- printchar (i - 1);
+ putchar (i - 1);
}
}
}
/* Loop over pattern commands. */
while (p < pend)
{
+ printf ("%d:\t", p - start);
+
switch ((re_opcode_t) *p++)
{
case no_op:
do
{
putchar ('/');
- printchar (*p++);
+ putchar (*p++);
}
while (--mcnt);
break;
case charset:
case charset_not:
{
- register int c;
+ register int c, last = -100;
+ register int in_range = 0;
- printf ("/charset%s",
- (re_opcode_t) *(p - 1) == charset_not ? "_not" : "");
+ printf ("/charset [%s",
+ (re_opcode_t) *(p - 1) == charset_not ? "^" : "");
assert (p + *p < pend);
- for (c = 0; c < *p; c++)
+ for (c = 0; c < 256; c++)
+ if (c / 8 < *p
+ && (p[1 + (c/8)] & (1 << (c % 8))))
+ {
+ /* Are we starting a range? */
+ if (last + 1 == c && ! in_range)
+ {
+ putchar ('-');
+ in_range = 1;
+ }
+ /* Have we broken a range? */
+ else if (last + 1 != c && in_range)
{
- unsigned bit;
- unsigned char map_byte = p[1 + c];
+ putchar (last);
+ in_range = 0;
+ }
- putchar ('/');
+ if (! in_range)
+ putchar (c);
- for (bit = 0; bit < BYTEWIDTH; bit++)
- if (map_byte & (1 << bit))
- printchar (c * BYTEWIDTH + bit);
+ last = c;
}
+
+ if (in_range)
+ putchar (last);
+
+ putchar (']');
+
p += 1 + *p;
- break;
}
+ break;
case begline:
printf ("/begline");
case on_failure_jump:
extract_number_and_incr (&mcnt, &p);
- printf ("/on_failure_jump/0/%d", mcnt);
+ printf ("/on_failure_jump to %d", p + mcnt - start);
break;
case on_failure_keep_string_jump:
extract_number_and_incr (&mcnt, &p);
- printf ("/on_failure_keep_string_jump/0/%d", mcnt);
+ printf ("/on_failure_keep_string_jump to %d", p + mcnt - start);
break;
case dummy_failure_jump:
extract_number_and_incr (&mcnt, &p);
- printf ("/dummy_failure_jump/0/%d", mcnt);
+ printf ("/dummy_failure_jump to %d", p + mcnt - start);
break;
case push_dummy_failure:
case maybe_pop_jump:
extract_number_and_incr (&mcnt, &p);
- printf ("/maybe_pop_jump/0/%d", mcnt);
+ printf ("/maybe_pop_jump to %d", p + mcnt - start);
break;
case pop_failure_jump:
extract_number_and_incr (&mcnt, &p);
- printf ("/pop_failure_jump/0/%d", mcnt);
+ printf ("/pop_failure_jump to %d", p + mcnt - start);
break;
case jump_past_alt:
extract_number_and_incr (&mcnt, &p);
- printf ("/jump_past_alt/0/%d", mcnt);
+ printf ("/jump_past_alt to %d", p + mcnt - start);
break;
case jump:
extract_number_and_incr (&mcnt, &p);
- printf ("/jump/0/%d", mcnt);
+ printf ("/jump to %d", p + mcnt - start);
break;
case succeed_n:
extract_number_and_incr (&mcnt, &p);
extract_number_and_incr (&mcnt2, &p);
- printf ("/succeed_n/0/%d/0/%d", mcnt, mcnt2);
+ printf ("/succeed_n to %d, %d times", p + mcnt - start, mcnt2);
break;
case jump_n:
extract_number_and_incr (&mcnt, &p);
extract_number_and_incr (&mcnt2, &p);
- printf ("/jump_n/0/%d/0/%d", mcnt, mcnt2);
+ printf ("/jump_n to %d, %d times", p + mcnt - start, mcnt2);
break;
case set_number_at:
extract_number_and_incr (&mcnt, &p);
extract_number_and_incr (&mcnt2, &p);
- printf ("/set_number_at/0/%d/0/%d", mcnt, mcnt2);
+ printf ("/set_number_at location %d to %d", p + mcnt - start, mcnt2);
break;
case wordbound:
default:
printf ("?%d", *(p-1));
}
+
+ putchar ('\n');
}
- printf ("/\n");
+
+ printf ("%d:\tend of pattern.\n", p - start);
}
if (FIRST_STRING_P (where))
{
for (this_char = where - string1; this_char < size1; this_char++)
- printchar (string1[this_char]);
+ putchar (string1[this_char]);
where = string2;
}
for (this_char = where - string2; this_char < size2; this_char++)
- printchar (string2[this_char]);
+ putchar (string2[this_char]);
}
}
#define DEBUG_PRINT1(x)
#define DEBUG_PRINT2(x1, x2)
#define DEBUG_PRINT3(x1, x2, x3)
+#define DEBUG_PRINT4(x1, x2, x3, x4)
#define DEBUG_PRINT_COMPILED_PATTERN(p, s, e)
#define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2)
/* Set by `re_set_syntax' to the current regexp syntax to recognize. Can
also be assigned to arbitrarily: each pattern buffer stores its own
syntax, so it can be changed between regex compilations. */
-reg_syntax_t re_syntax_options = RE_SYNTAX_EMACS;
+/* This has no initializer because initialized variables in Emacs
+ become read-only after dumping. */
+reg_syntax_t re_syntax_options;
+
+
+/* Specify the precise syntax of regexps for compilation. This provides
+ for compatibility for various utilities which historically have
+ different, incompatible syntaxes.
+
+ The argument SYNTAX is a bit mask comprised of the various bits
+ defined in regex.h. We return the old syntax. */
+
+reg_syntax_t
+re_set_syntax (syntax)
+ reg_syntax_t syntax;
+{
+ reg_syntax_t ret = re_syntax_options;
+
+ re_syntax_options = syntax;
+ return ret;
+}
+\f
+/* This table gives an error message for each of the error codes listed
+ in regex.h. Obviously the order here has to be same as there.
+ POSIX doesn't require that we do anything for REG_NOERROR,
+ but why not be nice? */
+
+static const char *re_error_msgid[] =
+ { "Success", /* REG_NOERROR */
+ "No match", /* REG_NOMATCH */
+ "Invalid regular expression", /* REG_BADPAT */
+ "Invalid collation character", /* REG_ECOLLATE */
+ "Invalid character class name", /* REG_ECTYPE */
+ "Trailing backslash", /* REG_EESCAPE */
+ "Invalid back reference", /* REG_ESUBREG */
+ "Unmatched [ or [^", /* REG_EBRACK */
+ "Unmatched ( or \\(", /* REG_EPAREN */
+ "Unmatched \\{", /* REG_EBRACE */
+ "Invalid content of \\{\\}", /* REG_BADBR */
+ "Invalid range end", /* REG_ERANGE */
+ "Memory exhausted", /* REG_ESPACE */
+ "Invalid preceding regular expression", /* REG_BADRPT */
+ "Premature end of regular expression", /* REG_EEND */
+ "Regular expression too big", /* REG_ESIZE */
+ "Unmatched ) or \\)", /* REG_ERPAREN */
+ };
+\f
+/* Avoiding alloca during matching, to placate r_alloc. */
+
+/* Define MATCH_MAY_ALLOCATE unless we need to make sure that the
+ searching and matching functions should not call alloca. On some
+ systems, alloca is implemented in terms of malloc, and if we're
+ using the relocating allocator routines, then malloc could cause a
+ relocation, which might (if the strings being searched are in the
+ ralloc heap) shift the data out from underneath the regexp
+ routines.
+
+ Here's another reason to avoid allocation: Emacs
+ processes input from X in a signal handler; processing X input may
+ call malloc; if input arrives while a matching routine is calling
+ malloc, then we're scrod. But Emacs can't just block input while
+ calling matching routines; then we don't notice interrupts when
+ they come in. So, Emacs blocks input around all regexp calls
+ except the matching calls, which it leaves unprotected, in the
+ faith that they will not malloc. */
+
+/* Normally, this is fine. */
+#define MATCH_MAY_ALLOCATE
+
+/* When using GNU C, we are not REALLY using the C alloca, no matter
+ what config.h may say. So don't take precautions for it. */
+#ifdef __GNUC__
+#undef C_ALLOCA
+#endif
+
+/* The match routines may not allocate if (1) they would do it with malloc
+ and (2) it's not safe for them to use malloc.
+ Note that if REL_ALLOC is defined, matching would not use malloc for the
+ failure stack, but we would still use it for the register vectors;
+ so REL_ALLOC should not affect this. */
+#if (defined (C_ALLOCA) || defined (REGEX_MALLOC)) && defined (emacs)
+#undef MATCH_MAY_ALLOCATE
+#endif
+
+\f
+/* Failure stack declarations and macros; both re_compile_fastmap and
+ re_match_2 use a failure stack. These have to be macros because of
+ REGEX_ALLOCATE_STACK. */
+
+
+/* Number of failure points for which to initially allocate space
+ when matching. If this number is exceeded, we allocate more
+ space, so it is not a hard limit. */
+#ifndef INIT_FAILURE_ALLOC
+#define INIT_FAILURE_ALLOC 5
+#endif
+
+/* Roughly the maximum number of failure points on the stack. Would be
+ exactly that if always used MAX_FAILURE_SPACE each time we failed.
+ This is a variable only so users of regex can assign to it; we never
+ change it ourselves. */
+#if defined (MATCH_MAY_ALLOCATE)
+int re_max_failures = 200000;
+#else
+int re_max_failures = 2000;
+#endif
+
+union fail_stack_elt
+{
+ unsigned char *pointer;
+ int integer;
+};
+
+typedef union fail_stack_elt fail_stack_elt_t;
+
+typedef struct
+{
+ fail_stack_elt_t *stack;
+ unsigned size;
+ unsigned avail; /* Offset of next open position. */
+} fail_stack_type;
+
+#define FAIL_STACK_EMPTY() (fail_stack.avail == 0)
+#define FAIL_STACK_PTR_EMPTY() (fail_stack_ptr->avail == 0)
+#define FAIL_STACK_FULL() (fail_stack.avail == fail_stack.size)
+
+
+/* Define macros to initialize and free the failure stack.
+ Do `return -2' if the alloc fails. */
+
+#ifdef MATCH_MAY_ALLOCATE
+#define INIT_FAIL_STACK() \
+ do { \
+ fail_stack.stack = (fail_stack_elt_t *) \
+ REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * sizeof (fail_stack_elt_t)); \
+ \
+ if (fail_stack.stack == NULL) \
+ return -2; \
+ \
+ fail_stack.size = INIT_FAILURE_ALLOC; \
+ fail_stack.avail = 0; \
+ } while (0)
+
+#define RESET_FAIL_STACK() REGEX_FREE_STACK (fail_stack.stack)
+#else
+#define INIT_FAIL_STACK() \
+ do { \
+ fail_stack.avail = 0; \
+ } while (0)
+
+#define RESET_FAIL_STACK()
+#endif
+
+
+/* Double the size of FAIL_STACK, up to approximately `re_max_failures' items.
+
+ Return 1 if succeeds, and 0 if either ran out of memory
+ allocating space for it or it was already too large.
+
+ REGEX_REALLOCATE_STACK requires `destination' be declared. */
+
+#define DOUBLE_FAIL_STACK(fail_stack) \
+ ((fail_stack).size > re_max_failures * MAX_FAILURE_ITEMS \
+ ? 0 \
+ : ((fail_stack).stack = (fail_stack_elt_t *) \
+ REGEX_REALLOCATE_STACK ((fail_stack).stack, \
+ (fail_stack).size * sizeof (fail_stack_elt_t), \
+ ((fail_stack).size << 1) * sizeof (fail_stack_elt_t)), \
+ \
+ (fail_stack).stack == NULL \
+ ? 0 \
+ : ((fail_stack).size <<= 1, \
+ 1)))
+
+
+/* Push pointer POINTER on FAIL_STACK.
+ Return 1 if was able to do so and 0 if ran out of memory allocating
+ space to do so. */
+#define PUSH_PATTERN_OP(POINTER, FAIL_STACK) \
+ ((FAIL_STACK_FULL () \
+ && !DOUBLE_FAIL_STACK (FAIL_STACK)) \
+ ? 0 \
+ : ((FAIL_STACK).stack[(FAIL_STACK).avail++].pointer = POINTER, \
+ 1))
+
+/* Push a pointer value onto the failure stack.
+ Assumes the variable `fail_stack'. Probably should only
+ be called from within `PUSH_FAILURE_POINT'. */
+#define PUSH_FAILURE_POINTER(item) \
+ fail_stack.stack[fail_stack.avail++].pointer = (unsigned char *) (item)
+
+/* This pushes an integer-valued item onto the failure stack.
+ Assumes the variable `fail_stack'. Probably should only
+ be called from within `PUSH_FAILURE_POINT'. */
+#define PUSH_FAILURE_INT(item) \
+ fail_stack.stack[fail_stack.avail++].integer = (item)
+
+/* Push a fail_stack_elt_t value onto the failure stack.
+ Assumes the variable `fail_stack'. Probably should only
+ be called from within `PUSH_FAILURE_POINT'. */
+#define PUSH_FAILURE_ELT(item) \
+ fail_stack.stack[fail_stack.avail++] = (item)
+
+/* These three POP... operations complement the three PUSH... operations.
+ All assume that `fail_stack' is nonempty. */
+#define POP_FAILURE_POINTER() fail_stack.stack[--fail_stack.avail].pointer
+#define POP_FAILURE_INT() fail_stack.stack[--fail_stack.avail].integer
+#define POP_FAILURE_ELT() fail_stack.stack[--fail_stack.avail]
+
+/* Used to omit pushing failure point id's when we're not debugging. */
+#ifdef DEBUG
+#define DEBUG_PUSH PUSH_FAILURE_INT
+#define DEBUG_POP(item_addr) *(item_addr) = POP_FAILURE_INT ()
+#else
+#define DEBUG_PUSH(item)
+#define DEBUG_POP(item_addr)
+#endif
+
+
+/* Push the information about the state we will need
+ if we ever fail back to it.
+
+ Requires variables fail_stack, regstart, regend, reg_info, and
+ num_regs be declared. DOUBLE_FAIL_STACK requires `destination' be
+ declared.
+
+ Does `return FAILURE_CODE' if runs out of memory. */
+
+#define PUSH_FAILURE_POINT(pattern_place, string_place, failure_code) \
+ do { \
+ char *destination; \
+ /* Must be int, so when we don't save any registers, the arithmetic \
+ of 0 + -1 isn't done as unsigned. */ \
+ int this_reg; \
+ \
+ DEBUG_STATEMENT (failure_id++); \
+ DEBUG_STATEMENT (nfailure_points_pushed++); \
+ DEBUG_PRINT2 ("\nPUSH_FAILURE_POINT #%u:\n", failure_id); \
+ DEBUG_PRINT2 (" Before push, next avail: %d\n", (fail_stack).avail);\
+ DEBUG_PRINT2 (" size: %d\n", (fail_stack).size);\
+ \
+ DEBUG_PRINT2 (" slots needed: %d\n", NUM_FAILURE_ITEMS); \
+ DEBUG_PRINT2 (" available: %d\n", REMAINING_AVAIL_SLOTS); \
+ \
+ /* Ensure we have enough space allocated for what we will push. */ \
+ while (REMAINING_AVAIL_SLOTS < NUM_FAILURE_ITEMS) \
+ { \
+ if (!DOUBLE_FAIL_STACK (fail_stack)) \
+ return failure_code; \
+ \
+ DEBUG_PRINT2 ("\n Doubled stack; size now: %d\n", \
+ (fail_stack).size); \
+ DEBUG_PRINT2 (" slots available: %d\n", REMAINING_AVAIL_SLOTS);\
+ } \
+ \
+ /* Push the info, starting with the registers. */ \
+ DEBUG_PRINT1 ("\n"); \
+ \
+ if (!RE_NO_POSIX_BACKTRACKING & bufp->syntax) \
+ for (this_reg = lowest_active_reg; this_reg <= highest_active_reg; \
+ this_reg++) \
+ { \
+ DEBUG_PRINT2 (" Pushing reg: %d\n", this_reg); \
+ DEBUG_STATEMENT (num_regs_pushed++); \
+ \
+ DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \
+ PUSH_FAILURE_POINTER (regstart[this_reg]); \
+ \
+ DEBUG_PRINT2 (" end: 0x%x\n", regend[this_reg]); \
+ PUSH_FAILURE_POINTER (regend[this_reg]); \
+ \
+ DEBUG_PRINT2 (" info: 0x%x\n ", reg_info[this_reg]); \
+ DEBUG_PRINT2 (" match_null=%d", \
+ REG_MATCH_NULL_STRING_P (reg_info[this_reg])); \
+ DEBUG_PRINT2 (" active=%d", IS_ACTIVE (reg_info[this_reg])); \
+ DEBUG_PRINT2 (" matched_something=%d", \
+ MATCHED_SOMETHING (reg_info[this_reg])); \
+ DEBUG_PRINT2 (" ever_matched=%d", \
+ EVER_MATCHED_SOMETHING (reg_info[this_reg])); \
+ DEBUG_PRINT1 ("\n"); \
+ PUSH_FAILURE_ELT (reg_info[this_reg].word); \
+ } \
+ \
+ DEBUG_PRINT2 (" Pushing low active reg: %d\n", lowest_active_reg);\
+ PUSH_FAILURE_INT (lowest_active_reg); \
+ \
+ DEBUG_PRINT2 (" Pushing high active reg: %d\n", highest_active_reg);\
+ PUSH_FAILURE_INT (highest_active_reg); \
+ \
+ DEBUG_PRINT2 (" Pushing pattern 0x%x: ", pattern_place); \
+ DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern_place, pend); \
+ PUSH_FAILURE_POINTER (pattern_place); \
+ \
+ DEBUG_PRINT2 (" Pushing string 0x%x: `", string_place); \
+ DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, \
+ size2); \
+ DEBUG_PRINT1 ("'\n"); \
+ PUSH_FAILURE_POINTER (string_place); \
+ \
+ DEBUG_PRINT2 (" Pushing failure id: %u\n", failure_id); \
+ DEBUG_PUSH (failure_id); \
+ } while (0)
+
+/* This is the number of items that are pushed and popped on the stack
+ for each register. */
+#define NUM_REG_ITEMS 3
+
+/* Individual items aside from the registers. */
+#ifdef DEBUG
+#define NUM_NONREG_ITEMS 5 /* Includes failure point id. */
+#else
+#define NUM_NONREG_ITEMS 4
+#endif
+
+/* We push at most this many items on the stack. */
+#define MAX_FAILURE_ITEMS ((num_regs - 1) * NUM_REG_ITEMS + NUM_NONREG_ITEMS)
+
+/* We actually push this many items. */
+#define NUM_FAILURE_ITEMS \
+ (((RE_NO_POSIX_BACKTRACKING & bufp->syntax \
+ ? 0 : highest_active_reg - lowest_active_reg + 1) \
+ * NUM_REG_ITEMS) \
+ + NUM_NONREG_ITEMS)
+
+/* How many items can still be added to the stack without overflowing it. */
+#define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail)
+
+
+/* Pops what PUSH_FAIL_STACK pushes.
+
+ We restore into the parameters, all of which should be lvalues:
+ STR -- the saved data position.
+ PAT -- the saved pattern position.
+ LOW_REG, HIGH_REG -- the highest and lowest active registers.
+ REGSTART, REGEND -- arrays of string positions.
+ REG_INFO -- array of information about each subexpression.
+
+ Also assumes the variables `fail_stack' and (if debugging), `bufp',
+ `pend', `string1', `size1', `string2', and `size2'. */
+
+#define POP_FAILURE_POINT(str, pat, low_reg, high_reg, regstart, regend, reg_info)\
+{ \
+ DEBUG_STATEMENT (fail_stack_elt_t failure_id;) \
+ int this_reg; \
+ const unsigned char *string_temp; \
+ \
+ assert (!FAIL_STACK_EMPTY ()); \
+ \
+ /* Remove failure points and point to how many regs pushed. */ \
+ DEBUG_PRINT1 ("POP_FAILURE_POINT:\n"); \
+ DEBUG_PRINT2 (" Before pop, next avail: %d\n", fail_stack.avail); \
+ DEBUG_PRINT2 (" size: %d\n", fail_stack.size); \
+ \
+ assert (fail_stack.avail >= NUM_NONREG_ITEMS); \
+ \
+ DEBUG_POP (&failure_id); \
+ DEBUG_PRINT2 (" Popping failure id: %u\n", failure_id); \
+ \
+ /* If the saved string location is NULL, it came from an \
+ on_failure_keep_string_jump opcode, and we want to throw away the \
+ saved NULL, thus retaining our current position in the string. */ \
+ string_temp = POP_FAILURE_POINTER (); \
+ if (string_temp != NULL) \
+ str = (const char *) string_temp; \
+ \
+ DEBUG_PRINT2 (" Popping string 0x%x: `", str); \
+ DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \
+ DEBUG_PRINT1 ("'\n"); \
+ \
+ pat = (unsigned char *) POP_FAILURE_POINTER (); \
+ DEBUG_PRINT2 (" Popping pattern 0x%x: ", pat); \
+ DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \
+ \
+ /* Restore register info. */ \
+ high_reg = (unsigned) POP_FAILURE_INT (); \
+ DEBUG_PRINT2 (" Popping high active reg: %d\n", high_reg); \
+ \
+ low_reg = (unsigned) POP_FAILURE_INT (); \
+ DEBUG_PRINT2 (" Popping low active reg: %d\n", low_reg); \
+ \
+ if (!RE_NO_POSIX_BACKTRACKING & bufp->syntax) \
+ for (this_reg = high_reg; this_reg >= low_reg; this_reg--) \
+ { \
+ DEBUG_PRINT2 (" Popping reg: %d\n", this_reg); \
+ \
+ reg_info[this_reg].word = POP_FAILURE_ELT (); \
+ DEBUG_PRINT2 (" info: 0x%x\n", reg_info[this_reg]); \
+ \
+ regend[this_reg] = (const char *) POP_FAILURE_POINTER (); \
+ DEBUG_PRINT2 (" end: 0x%x\n", regend[this_reg]); \
+ \
+ regstart[this_reg] = (const char *) POP_FAILURE_POINTER (); \
+ DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \
+ } \
+ \
+ set_regs_matched_done = 0; \
+ DEBUG_STATEMENT (nfailure_points_popped++); \
+} /* POP_FAILURE_POINT */
+
+
+\f
+/* Structure for per-register (a.k.a. per-group) information.
+ Other register information, such as the
+ starting and ending positions (which are addresses), and the list of
+ inner groups (which is a bits list) are maintained in separate
+ variables.
+
+ We are making a (strictly speaking) nonportable assumption here: that
+ the compiler will pack our bit fields into something that fits into
+ the type of `word', i.e., is something that fits into one item on the
+ failure stack. */
+typedef union
+{
+ fail_stack_elt_t word;
+ struct
+ {
+ /* This field is one if this group can match the empty string,
+ zero if not. If not yet determined, `MATCH_NULL_UNSET_VALUE'. */
+#define MATCH_NULL_UNSET_VALUE 3
+ unsigned match_null_string_p : 2;
+ unsigned is_active : 1;
+ unsigned matched_something : 1;
+ unsigned ever_matched_something : 1;
+ } bits;
+} register_info_type;
-/* Specify the precise syntax of regexps for compilation. This provides
- for compatibility for various utilities which historically have
- different, incompatible syntaxes.
+#define REG_MATCH_NULL_STRING_P(R) ((R).bits.match_null_string_p)
+#define IS_ACTIVE(R) ((R).bits.is_active)
+#define MATCHED_SOMETHING(R) ((R).bits.matched_something)
+#define EVER_MATCHED_SOMETHING(R) ((R).bits.ever_matched_something)
- The argument SYNTAX is a bit mask comprised of the various bits
- defined in regex.h. We return the old syntax. */
-reg_syntax_t
-re_set_syntax (syntax)
- reg_syntax_t syntax;
-{
- reg_syntax_t ret = re_syntax_options;
-
- re_syntax_options = syntax;
- return ret;
-}
-\f
-/* This table gives an error message for each of the error codes listed
- in regex.h. Obviously the order here has to be same as there. */
+/* Call this when have matched a real character; it sets `matched' flags
+ for the subexpressions which we are currently inside. Also records
+ that those subexprs have matched. */
+#define SET_REGS_MATCHED() \
+ do \
+ { \
+ if (!set_regs_matched_done) \
+ { \
+ unsigned r; \
+ set_regs_matched_done = 1; \
+ for (r = lowest_active_reg; r <= highest_active_reg; r++) \
+ { \
+ MATCHED_SOMETHING (reg_info[r]) \
+ = EVER_MATCHED_SOMETHING (reg_info[r]) \
+ = 1; \
+ } \
+ } \
+ } \
+ while (0)
-static const char *re_error_msg[] =
- { NULL, /* REG_NOERROR */
- "No match", /* REG_NOMATCH */
- "Invalid regular expression", /* REG_BADPAT */
- "Invalid collation character", /* REG_ECOLLATE */
- "Invalid character class name", /* REG_ECTYPE */
- "Trailing backslash", /* REG_EESCAPE */
- "Invalid back reference", /* REG_ESUBREG */
- "Unmatched [ or [^", /* REG_EBRACK */
- "Unmatched ( or \\(", /* REG_EPAREN */
- "Unmatched \\{", /* REG_EBRACE */
- "Invalid content of \\{\\}", /* REG_BADBR */
- "Invalid range end", /* REG_ERANGE */
- "Memory exhausted", /* REG_ESPACE */
- "Invalid preceding regular expression", /* REG_BADRPT */
- "Premature end of regular expression", /* REG_EEND */
- "Regular expression too big", /* REG_ESIZE */
- "Unmatched ) or \\)", /* REG_ERPAREN */
- };
+/* Registers are set to a sentinel when they haven't yet matched. */
+static char reg_unset_dummy;
+#define REG_UNSET_VALUE (®_unset_dummy)
+#define REG_UNSET(e) ((e) == REG_UNSET_VALUE)
\f
/* Subroutine declarations and macros for regex_compile. */
{ if (p != pend) \
{ \
PATFETCH (c); \
- while (isdigit (c)) \
+ while (ISDIGIT (c)) \
{ \
if (num < 0) \
num = 0; \
|| STREQ (string, "punct") || STREQ (string, "graph") \
|| STREQ (string, "cntrl") || STREQ (string, "blank"))
\f
+#ifndef MATCH_MAY_ALLOCATE
+
+/* If we cannot allocate large objects within re_match_2_internal,
+ we make the fail stack and register vectors global.
+ The fail stack, we grow to the maximum size when a regexp
+ is compiled.
+ The register vectors, we adjust in size each time we
+ compile a regexp, according to the number of registers it needs. */
+
+static fail_stack_type fail_stack;
+
+/* Size with which the following vectors are currently allocated.
+ That is so we can make them bigger as needed,
+ but never make them smaller. */
+static int regs_allocated_size;
+
+static const char ** regstart, ** regend;
+static const char ** old_regstart, ** old_regend;
+static const char **best_regstart, **best_regend;
+static register_info_type *reg_info;
+static const char **reg_dummy;
+static register_info_type *reg_info_dummy;
+
+/* Make the register vectors big enough for NUM_REGS registers,
+ but don't make them smaller. */
+
+static
+regex_grow_registers (num_regs)
+ int num_regs;
+{
+ if (num_regs > regs_allocated_size)
+ {
+ RETALLOC_IF (regstart, num_regs, const char *);
+ RETALLOC_IF (regend, num_regs, const char *);
+ RETALLOC_IF (old_regstart, num_regs, const char *);
+ RETALLOC_IF (old_regend, num_regs, const char *);
+ RETALLOC_IF (best_regstart, num_regs, const char *);
+ RETALLOC_IF (best_regend, num_regs, const char *);
+ RETALLOC_IF (reg_info, num_regs, register_info_type);
+ RETALLOC_IF (reg_dummy, num_regs, const char *);
+ RETALLOC_IF (reg_info_dummy, num_regs, register_info_type);
+
+ regs_allocated_size = num_regs;
+ }
+}
+
+#endif /* not MATCH_MAY_ALLOCATE */
+\f
/* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX.
Returns one of error codes defined in `regex.h', or zero for success.
`buffer' is the compiled pattern;
`syntax' is set to SYNTAX;
`used' is set to the length of the compiled pattern;
- `fastmap_accurate' is set to zero;
- `re_nsub' is set to the number of groups in PATTERN;
- `not_bol' and `not_eol' are set to zero.
+ `fastmap_accurate' is zero;
+ `re_nsub' is the number of subexpressions in PATTERN;
+ `not_bol' and `not_eol' are zero;
The `fastmap' and `newline_anchor' fields are neither
examined nor set. */
+/* Return, freeing storage we allocated. */
+#define FREE_STACK_RETURN(value) \
+ return (free (compile_stack.stack), value)
+
static reg_errcode_t
regex_compile (pattern, size, syntax, bufp)
const char *pattern;
they can be reliably used as array indices. */
register unsigned char c, c1;
- /* A random tempory spot in PATTERN. */
+ /* A random temporary spot in PATTERN. */
const char *p1;
/* Points to the end of the buffer, where we should append. */
unsigned debug_count;
for (debug_count = 0; debug_count < size; debug_count++)
- printchar (pattern[debug_count]);
+ putchar (pattern[debug_count]);
putchar ('\n');
}
#endif /* DEBUG */
{ /* Caller did not allocate a buffer. Do it for them. */
bufp->buffer = TALLOC (INIT_BUF_SIZE, unsigned char);
}
- if (!bufp->buffer) return REG_ESPACE;
+ if (!bufp->buffer) FREE_STACK_RETURN (REG_ESPACE);
bufp->allocated = INIT_BUF_SIZE;
}
if (!laststart)
{
if (syntax & RE_CONTEXT_INVALID_OPS)
- return REG_BADRPT;
+ FREE_STACK_RETURN (REG_BADRPT);
else if (!(syntax & RE_CONTEXT_INDEP_OPS))
goto normal_char;
}
else if (syntax & RE_BK_PLUS_QM && c == '\\')
{
- if (p == pend) return REG_EESCAPE;
+ if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
PATFETCH (c1);
if (!(c1 == '+' || c1 == '?'))
the `*'. Do we have to do something analogous here
for null bytes, because of RE_DOT_NOT_NULL? */
if (TRANSLATE (*(p - 2)) == TRANSLATE ('.')
+ && zero_times_ok
&& p < pend && TRANSLATE (*p) == TRANSLATE ('\n')
&& !(syntax & RE_DOT_NEWLINE))
{ /* We have .*\n. */
{
boolean had_char_class = false;
- if (p == pend) return REG_EBRACK;
+ if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
/* Ensure that we have enough space to push a charset: the
opcode, the length count, and the bitset; 34 bytes in all. */
/* Read in characters and ranges, setting map bits. */
for (;;)
{
- if (p == pend) return REG_EBRACK;
+ if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
PATFETCH (c);
/* \ might escape characters inside [...] and [^...]. */
if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\')
{
- if (p == pend) return REG_EESCAPE;
+ if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
PATFETCH (c1);
SET_LIST_BIT (c1);
/* Look ahead to see if it's a range when the last thing
was a character class. */
if (had_char_class && c == '-' && *p != ']')
- return REG_ERANGE;
+ FREE_STACK_RETURN (REG_ERANGE);
/* Look ahead to see if it's a range when the last thing
was a character: if this is a hyphen not at the
{
reg_errcode_t ret
= compile_range (&p, pend, translate, syntax, b);
- if (ret != REG_NOERROR) return ret;
+ if (ret != REG_NOERROR) FREE_STACK_RETURN (ret);
}
else if (p[0] == '-' && p[1] != ']')
PATFETCH (c1);
ret = compile_range (&p, pend, translate, syntax, b);
- if (ret != REG_NOERROR) return ret;
+ if (ret != REG_NOERROR) FREE_STACK_RETURN (ret);
}
/* See if we're at the beginning of a possible character
c1 = 0;
/* If pattern is `[[:'. */
- if (p == pend) return REG_EBRACK;
+ if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
for (;;)
{
boolean is_upper = STREQ (str, "upper");
boolean is_xdigit = STREQ (str, "xdigit");
- if (!IS_CHAR_CLASS (str)) return REG_ECTYPE;
+ if (!IS_CHAR_CLASS (str))
+ FREE_STACK_RETURN (REG_ECTYPE);
/* Throw away the ] at the end of the character
class. */
PATFETCH (c);
- if (p == pend) return REG_EBRACK;
+ if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
for (ch = 0; ch < 1 << BYTEWIDTH; ch++)
{
- if ( (is_alnum && isalnum (ch))
- || (is_alpha && isalpha (ch))
- || (is_blank && isblank (ch))
- || (is_cntrl && iscntrl (ch))
- || (is_digit && isdigit (ch))
- || (is_graph && isgraph (ch))
- || (is_lower && islower (ch))
- || (is_print && isprint (ch))
- || (is_punct && ispunct (ch))
- || (is_space && isspace (ch))
- || (is_upper && isupper (ch))
- || (is_xdigit && isxdigit (ch)))
- SET_LIST_BIT (ch);
+ /* This was split into 3 if's to
+ avoid an arbitrary limit in some compiler. */
+ if ( (is_alnum && ISALNUM (ch))
+ || (is_alpha && ISALPHA (ch))
+ || (is_blank && ISBLANK (ch))
+ || (is_cntrl && ISCNTRL (ch)))
+ SET_LIST_BIT (ch);
+ if ( (is_digit && ISDIGIT (ch))
+ || (is_graph && ISGRAPH (ch))
+ || (is_lower && ISLOWER (ch))
+ || (is_print && ISPRINT (ch)))
+ SET_LIST_BIT (ch);
+ if ( (is_punct && ISPUNCT (ch))
+ || (is_space && ISSPACE (ch))
+ || (is_upper && ISUPPER (ch))
+ || (is_xdigit && ISXDIGIT (ch)))
+ SET_LIST_BIT (ch);
}
had_char_class = true;
}
case '\\':
- if (p == pend) return REG_EESCAPE;
+ if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
/* Do not translate the character after the \, so that we can
distinguish, e.g., \B from \b, even if we normally would
fixup_alt_jump = 0;
laststart = 0;
begalt = b;
+ /* If we've reached MAX_REGNUM groups, then this open
+ won't actually generate any code, so we'll have to
+ clear pending_exact explicitly. */
+ pending_exact = 0;
break;
if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)
goto normal_backslash;
else
- return REG_ERPAREN;
+ FREE_STACK_RETURN (REG_ERPAREN);
handle_close:
if (fixup_alt_jump)
if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)
goto normal_char;
else
- return REG_ERPAREN;
+ FREE_STACK_RETURN (REG_ERPAREN);
/* Since we just checked for an empty stack above, this
``can't happen''. */
: 0;
laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset;
this_group_regnum = COMPILE_STACK_TOP.regnum;
+ /* If we've reached MAX_REGNUM groups, then this open
+ won't actually generate any code, so we'll have to
+ clear pending_exact explicitly. */
+ pending_exact = 0;
/* We're at the end of the group, so now we know how many
groups were inside this one. */
| v | v
a | b | c
- If we are at `b,' then fixup_alt_jump right now points to a
- three-byte space after `a.' We'll put in the jump, set
- fixup_alt_jump to right after `b,' and leave behind three
- bytes which we'll fill in when we get to after `c.' */
+ If we are at `b', then fixup_alt_jump right now points to a
+ three-byte space after `a'. We'll put in the jump, set
+ fixup_alt_jump to right after `b', and leave behind three
+ bytes which we'll fill in when we get to after `c'. */
if (fixup_alt_jump)
STORE_JUMP (jump_past_alt, fixup_alt_jump, b);
if (syntax & RE_NO_BK_BRACES)
goto unfetch_interval;
else
- return REG_EBRACE;
+ FREE_STACK_RETURN (REG_EBRACE);
}
GET_UNSIGNED_NUMBER (lower_bound);
if (syntax & RE_NO_BK_BRACES)
goto unfetch_interval;
else
- return REG_BADBR;
+ FREE_STACK_RETURN (REG_BADBR);
}
if (!(syntax & RE_NO_BK_BRACES))
{
- if (c != '\\') return REG_EBRACE;
+ if (c != '\\') FREE_STACK_RETURN (REG_EBRACE);
PATFETCH (c);
}
if (syntax & RE_NO_BK_BRACES)
goto unfetch_interval;
else
- return REG_BADBR;
+ FREE_STACK_RETURN (REG_BADBR);
}
/* We just parsed a valid interval. */
if (!laststart)
{
if (syntax & RE_CONTEXT_INVALID_OPS)
- return REG_BADRPT;
+ FREE_STACK_RETURN (REG_BADRPT);
else if (syntax & RE_CONTEXT_INDEP_OPS)
laststart = b;
else
we're all done, the pattern will look like:
set_number_at <jump count> <upper bound>
set_number_at <succeed_n count> <lower bound>
- succeed_n <after jump addr> <succed_n count>
+ succeed_n <after jump addr> <succeed_n count>
<body of loop>
jump_n <succeed_n addr> <jump count>
(The upper bound and `jump_n' are omitted if
c1 = c - '0';
if (c1 > regnum)
- return REG_ESUBREG;
+ FREE_STACK_RETURN (REG_ESUBREG);
/* Can't back reference to a subexpression if inside of it. */
if (group_in_compile_stack (compile_stack, c1))
STORE_JUMP (jump_past_alt, fixup_alt_jump, b);
if (!COMPILE_STACK_EMPTY)
- return REG_EPAREN;
+ FREE_STACK_RETURN (REG_EPAREN);
+
+ /* If we don't want backtracking, force success
+ the first time we reach the end of the compiled pattern. */
+ if (syntax & RE_NO_POSIX_BACKTRACKING)
+ BUF_PUSH (succeed);
free (compile_stack.stack);
#ifdef DEBUG
if (debug)
{
- DEBUG_PRINT1 ("\nCompiled pattern: ");
+ DEBUG_PRINT1 ("\nCompiled pattern: \n");
print_compiled_pattern (bufp);
}
#endif /* DEBUG */
+#ifndef MATCH_MAY_ALLOCATE
+ /* Initialize the failure stack to the largest possible stack. This
+ isn't necessary unless we're trying to avoid calling alloca in
+ the search and match routines. */
+ {
+ int num_regs = bufp->re_nsub + 1;
+
+ /* Since DOUBLE_FAIL_STACK refuses to double only if the current size
+ is strictly greater than re_max_failures, the largest possible stack
+ is 2 * re_max_failures failure points. */
+ if (fail_stack.size < (2 * re_max_failures * MAX_FAILURE_ITEMS))
+ {
+ fail_stack.size = (2 * re_max_failures * MAX_FAILURE_ITEMS);
+
+#ifdef emacs
+ if (! fail_stack.stack)
+ fail_stack.stack
+ = (fail_stack_elt_t *) xmalloc (fail_stack.size
+ * sizeof (fail_stack_elt_t));
+ else
+ fail_stack.stack
+ = (fail_stack_elt_t *) xrealloc (fail_stack.stack,
+ (fail_stack.size
+ * sizeof (fail_stack_elt_t)));
+#else /* not emacs */
+ if (! fail_stack.stack)
+ fail_stack.stack
+ = (fail_stack_elt_t *) malloc (fail_stack.size
+ * sizeof (fail_stack_elt_t));
+ else
+ fail_stack.stack
+ = (fail_stack_elt_t *) realloc (fail_stack.stack,
+ (fail_stack.size
+ * sizeof (fail_stack_elt_t)));
+#endif /* not emacs */
+ }
+
+ regex_grow_registers (num_regs);
+ }
+#endif /* not MATCH_MAY_ALLOCATE */
+
return REG_NOERROR;
} /* regex_compile */
\f
{
const char *next = p;
boolean next_backslash = *next == '\\';
- const char *next_next = p + 1 < pend ? p + 1 : NULL;
+ const char *next_next = p + 1 < pend ? p + 1 : 0;
return
/* Before a subexpression? */
{
unsigned this_char;
- const char *p = *p_ptr;
-
- /* Even though the pattern is a signed `char *', we need to fetch into
- `unsigned char's. Reason: if the high bit of the pattern character
- is set, the range endpoints will be negative if we fetch into a
- signed `char *'. */
- unsigned char range_end;
- unsigned char range_start = p[-2];
-
- if (p == pend)
- return REG_ERANGE;
-
- PATFETCH (range_end);
-
- /* Have to increment the pointer into the pattern string, so the
- caller isn't still at the ending character. */
- (*p_ptr)++;
-
- /* If the start is after the end, the range is empty. */
- if (range_start > range_end)
- return syntax & RE_NO_EMPTY_RANGES ? REG_ERANGE : REG_NOERROR;
-
- /* Here we see why `this_char' has to be larger than an `unsigned
- char' -- the range is inclusive, so if `range_end' == 0xff
- (assuming 8-bit characters), we would otherwise go into an infinite
- loop, since all characters <= 0xff. */
- for (this_char = range_start; this_char <= range_end; this_char++)
- {
- SET_LIST_BIT (TRANSLATE (this_char));
- }
-
- return REG_NOERROR;
-}
-\f
-/* Failure stack declarations and macros; both re_compile_fastmap and
- re_match_2 use a failure stack. These have to be macros because of
- REGEX_ALLOCATE. */
-
-
-/* Number of failure points for which to initially allocate space
- when matching. If this number is exceeded, we allocate more
- space, so it is not a hard limit. */
-#ifndef INIT_FAILURE_ALLOC
-#define INIT_FAILURE_ALLOC 5
-#endif
-
-/* Roughly the maximum number of failure points on the stack. Would be
- exactly that if always used MAX_FAILURE_SPACE each time we failed.
- This is a variable only so users of regex can assign to it; we never
- change it ourselves. */
-int re_max_failures = 2000;
-
-typedef const unsigned char *fail_stack_elt_t;
-
-typedef struct
-{
- fail_stack_elt_t *stack;
- unsigned size;
- unsigned avail; /* Offset of next open position. */
-} fail_stack_type;
-
-#define FAIL_STACK_EMPTY() (fail_stack.avail == 0)
-#define FAIL_STACK_PTR_EMPTY() (fail_stack_ptr->avail == 0)
-#define FAIL_STACK_FULL() (fail_stack.avail == fail_stack.size)
-#define FAIL_STACK_TOP() (fail_stack.stack[fail_stack.avail])
-
-
-/* Initialize `fail_stack'. Do `return -2' if the alloc fails. */
-
-#define INIT_FAIL_STACK() \
- do { \
- fail_stack.stack = (fail_stack_elt_t *) \
- REGEX_ALLOCATE (INIT_FAILURE_ALLOC * sizeof (fail_stack_elt_t)); \
- \
- if (fail_stack.stack == NULL) \
- return -2; \
- \
- fail_stack.size = INIT_FAILURE_ALLOC; \
- fail_stack.avail = 0; \
- } while (0)
-
-
-/* Double the size of FAIL_STACK, up to approximately `re_max_failures' items.
-
- Return 1 if succeeds, and 0 if either ran out of memory
- allocating space for it or it was already too large.
-
- REGEX_REALLOCATE requires `destination' be declared. */
-
-#define DOUBLE_FAIL_STACK(fail_stack) \
- ((fail_stack).size > re_max_failures * MAX_FAILURE_ITEMS \
- ? 0 \
- : ((fail_stack).stack = (fail_stack_elt_t *) \
- REGEX_REALLOCATE ((fail_stack).stack, \
- (fail_stack).size * sizeof (fail_stack_elt_t), \
- ((fail_stack).size << 1) * sizeof (fail_stack_elt_t)), \
- \
- (fail_stack).stack == NULL \
- ? 0 \
- : ((fail_stack).size <<= 1, \
- 1)))
-
-
-/* Push PATTERN_OP on FAIL_STACK.
-
- Return 1 if was able to do so and 0 if ran out of memory allocating
- space to do so. */
-#define PUSH_PATTERN_OP(pattern_op, fail_stack) \
- ((FAIL_STACK_FULL () \
- && !DOUBLE_FAIL_STACK (fail_stack)) \
- ? 0 \
- : ((fail_stack).stack[(fail_stack).avail++] = pattern_op, \
- 1))
-
-/* This pushes an item onto the failure stack. Must be a four-byte
- value. Assumes the variable `fail_stack'. Probably should only
- be called from within `PUSH_FAILURE_POINT'. */
-#define PUSH_FAILURE_ITEM(item) \
- fail_stack.stack[fail_stack.avail++] = (fail_stack_elt_t) item
-
-/* The complement operation. Assumes `fail_stack' is nonempty. */
-#define POP_FAILURE_ITEM() fail_stack.stack[--fail_stack.avail]
-
-/* Used to omit pushing failure point id's when we're not debugging. */
-#ifdef DEBUG
-#define DEBUG_PUSH PUSH_FAILURE_ITEM
-#define DEBUG_POP(item_addr) *(item_addr) = POP_FAILURE_ITEM ()
-#else
-#define DEBUG_PUSH(item)
-#define DEBUG_POP(item_addr)
-#endif
-
-
-/* Push the information about the state we will need
- if we ever fail back to it.
-
- Requires variables fail_stack, regstart, regend, reg_info, and
- num_regs be declared. DOUBLE_FAIL_STACK requires `destination' be
- declared.
-
- Does `return FAILURE_CODE' if runs out of memory. */
-
-#define PUSH_FAILURE_POINT(pattern_place, string_place, failure_code) \
- do { \
- char *destination; \
- /* Must be int, so when we don't save any registers, the arithmetic \
- of 0 + -1 isn't done as unsigned. */ \
- int this_reg; \
- \
- DEBUG_STATEMENT (failure_id++); \
- DEBUG_PRINT2 ("\nPUSH_FAILURE_POINT #%u:\n", failure_id); \
- DEBUG_PRINT2 (" Before push, next avail: %d\n", (fail_stack).avail);\
- DEBUG_PRINT2 (" size: %d\n", (fail_stack).size);\
- \
- DEBUG_PRINT2 (" slots needed: %d\n", NUM_FAILURE_ITEMS); \
- DEBUG_PRINT2 (" available: %d\n", REMAINING_AVAIL_SLOTS); \
- \
- /* Ensure we have enough space allocated for what we will push. */ \
- while (REMAINING_AVAIL_SLOTS < NUM_FAILURE_ITEMS) \
- { \
- if (!DOUBLE_FAIL_STACK (fail_stack)) \
- return failure_code; \
- \
- DEBUG_PRINT2 ("\n Doubled stack; size now: %d\n", \
- (fail_stack).size); \
- DEBUG_PRINT2 (" slots available: %d\n", REMAINING_AVAIL_SLOTS);\
- } \
- \
- /* Push the info, starting with the registers. */ \
- DEBUG_PRINT1 ("\n"); \
- \
- for (this_reg = lowest_active_reg; this_reg <= highest_active_reg; \
- this_reg++) \
- { \
- DEBUG_PRINT2 (" Pushing reg: %d\n", this_reg); \
- DEBUG_STATEMENT (num_regs_pushed++); \
- \
- DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \
- PUSH_FAILURE_ITEM (regstart[this_reg]); \
- \
- DEBUG_PRINT2 (" end: 0x%x\n", regend[this_reg]); \
- PUSH_FAILURE_ITEM (regend[this_reg]); \
- \
- DEBUG_PRINT2 (" info: 0x%x\n ", reg_info[this_reg]); \
- DEBUG_PRINT2 (" match_null=%d", \
- REG_MATCH_NULL_STRING_P (reg_info[this_reg])); \
- DEBUG_PRINT2 (" active=%d", IS_ACTIVE (reg_info[this_reg])); \
- DEBUG_PRINT2 (" matched_something=%d", \
- MATCHED_SOMETHING (reg_info[this_reg])); \
- DEBUG_PRINT2 (" ever_matched=%d", \
- EVER_MATCHED_SOMETHING (reg_info[this_reg])); \
- DEBUG_PRINT1 ("\n"); \
- PUSH_FAILURE_ITEM (reg_info[this_reg].word); \
- } \
- \
- DEBUG_PRINT2 (" Pushing low active reg: %d\n", lowest_active_reg);\
- PUSH_FAILURE_ITEM (lowest_active_reg); \
- \
- DEBUG_PRINT2 (" Pushing high active reg: %d\n", highest_active_reg);\
- PUSH_FAILURE_ITEM (highest_active_reg); \
- \
- DEBUG_PRINT2 (" Pushing pattern 0x%x: ", pattern_place); \
- DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern_place, pend); \
- PUSH_FAILURE_ITEM (pattern_place); \
- \
- DEBUG_PRINT2 (" Pushing string 0x%x: `", string_place); \
- DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, \
- size2); \
- DEBUG_PRINT1 ("'\n"); \
- PUSH_FAILURE_ITEM (string_place); \
- \
- DEBUG_PRINT2 (" Pushing failure id: %u\n", failure_id); \
- DEBUG_PUSH (failure_id); \
- } while (0)
-
-/* This is the number of items that are pushed and popped on the stack
- for each register. */
-#define NUM_REG_ITEMS 3
-
-/* Individual items aside from the registers. */
-#ifdef DEBUG
-#define NUM_NONREG_ITEMS 5 /* Includes failure point id. */
-#else
-#define NUM_NONREG_ITEMS 4
-#endif
-
-/* We push at most this many items on the stack. */
-#define MAX_FAILURE_ITEMS ((num_regs - 1) * NUM_REG_ITEMS + NUM_NONREG_ITEMS)
-
-/* We actually push this many items. */
-#define NUM_FAILURE_ITEMS \
- ((highest_active_reg - lowest_active_reg + 1) * NUM_REG_ITEMS \
- + NUM_NONREG_ITEMS)
+ const char *p = *p_ptr;
+ int range_start, range_end;
+
+ if (p == pend)
+ return REG_ERANGE;
-/* How many items can still be added to the stack without overflowing it. */
-#define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail)
+ /* Even though the pattern is a signed `char *', we need to fetch
+ with unsigned char *'s; if the high bit of the pattern character
+ is set, the range endpoints will be negative if we fetch using a
+ signed char *.
+ We also want to fetch the endpoints without translating them; the
+ appropriate translation is done in the bit-setting loop below. */
+ /* The SVR4 compiler on the 3B2 had trouble with unsigned const char *. */
+ range_start = ((const unsigned char *) p)[-2];
+ range_end = ((const unsigned char *) p)[0];
-/* Pops what PUSH_FAIL_STACK pushes.
+ /* Have to increment the pointer into the pattern string, so the
+ caller isn't still at the ending character. */
+ (*p_ptr)++;
- We restore into the parameters, all of which should be lvalues:
- STR -- the saved data position.
- PAT -- the saved pattern position.
- LOW_REG, HIGH_REG -- the highest and lowest active registers.
- REGSTART, REGEND -- arrays of string positions.
- REG_INFO -- array of information about each subexpression.
-
- Also assumes the variables `fail_stack' and (if debugging), `bufp',
- `pend', `string1', `size1', `string2', and `size2'. */
+ /* If the start is after the end, the range is empty. */
+ if (range_start > range_end)
+ return syntax & RE_NO_EMPTY_RANGES ? REG_ERANGE : REG_NOERROR;
-#define POP_FAILURE_POINT(str, pat, low_reg, high_reg, regstart, regend, reg_info)\
-{ \
- DEBUG_STATEMENT (fail_stack_elt_t failure_id;) \
- int this_reg; \
- const unsigned char *string_temp; \
- \
- assert (!FAIL_STACK_EMPTY ()); \
- \
- /* Remove failure points and point to how many regs pushed. */ \
- DEBUG_PRINT1 ("POP_FAILURE_POINT:\n"); \
- DEBUG_PRINT2 (" Before pop, next avail: %d\n", fail_stack.avail); \
- DEBUG_PRINT2 (" size: %d\n", fail_stack.size); \
- \
- assert (fail_stack.avail >= NUM_NONREG_ITEMS); \
- \
- DEBUG_POP (&failure_id); \
- DEBUG_PRINT2 (" Popping failure id: %u\n", failure_id); \
- \
- /* If the saved string location is NULL, it came from an \
- on_failure_keep_string_jump opcode, and we want to throw away the \
- saved NULL, thus retaining our current position in the string. */ \
- string_temp = POP_FAILURE_ITEM (); \
- if (string_temp != NULL) \
- str = (const char *) string_temp; \
- \
- DEBUG_PRINT2 (" Popping string 0x%x: `", str); \
- DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \
- DEBUG_PRINT1 ("'\n"); \
- \
- pat = (unsigned char *) POP_FAILURE_ITEM (); \
- DEBUG_PRINT2 (" Popping pattern 0x%x: ", pat); \
- DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \
- \
- /* Restore register info. */ \
- high_reg = (unsigned) POP_FAILURE_ITEM (); \
- DEBUG_PRINT2 (" Popping high active reg: %d\n", high_reg); \
- \
- low_reg = (unsigned) POP_FAILURE_ITEM (); \
- DEBUG_PRINT2 (" Popping low active reg: %d\n", low_reg); \
- \
- for (this_reg = high_reg; this_reg >= low_reg; this_reg--) \
- { \
- DEBUG_PRINT2 (" Popping reg: %d\n", this_reg); \
- \
- reg_info[this_reg].word = POP_FAILURE_ITEM (); \
- DEBUG_PRINT2 (" info: 0x%x\n", reg_info[this_reg]); \
- \
- regend[this_reg] = (const char *) POP_FAILURE_ITEM (); \
- DEBUG_PRINT2 (" end: 0x%x\n", regend[this_reg]); \
- \
- regstart[this_reg] = (const char *) POP_FAILURE_ITEM (); \
- DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \
- } \
-} /* POP_FAILURE_POINT */
+ /* Here we see why `this_char' has to be larger than an `unsigned
+ char' -- the range is inclusive, so if `range_end' == 0xff
+ (assuming 8-bit characters), we would otherwise go into an infinite
+ loop, since all characters <= 0xff. */
+ for (this_char = range_start; this_char <= range_end; this_char++)
+ {
+ SET_LIST_BIT (TRANSLATE (this_char));
+ }
+
+ return REG_NOERROR;
+}
\f
/* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in
BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible
struct re_pattern_buffer *bufp;
{
int j, k;
+#ifdef MATCH_MAY_ALLOCATE
fail_stack_type fail_stack;
+#endif
#ifndef REGEX_MALLOC
char *destination;
#endif
register char *fastmap = bufp->fastmap;
unsigned char *pattern = bufp->buffer;
unsigned long size = bufp->used;
- const unsigned char *p = pattern;
+ unsigned char *p = pattern;
register unsigned char *pend = pattern + size;
+ /* This holds the pointer to the failure stack, when
+ it is allocated relocatably. */
+#ifdef REL_ALLOC
+ fail_stack_elt_t *failure_stack_ptr;
+#endif
+
/* Assume that each path through the pattern can be null until
proven otherwise. We set this false at the bottom of switch
statement, to which we get only if a particular path doesn't
bufp->fastmap_accurate = 1; /* It will be when we're done. */
bufp->can_be_null = 0;
- while (p != pend || !FAIL_STACK_EMPTY ())
+ while (1)
{
- if (p == pend)
- {
- bufp->can_be_null |= path_can_be_null;
-
- /* Reset for next path. */
- path_can_be_null = true;
-
- p = fail_stack.stack[--fail_stack.avail];
+ if (p == pend || *p == succeed)
+ {
+ /* We have reached the (effective) end of pattern. */
+ if (!FAIL_STACK_EMPTY ())
+ {
+ bufp->can_be_null |= path_can_be_null;
+
+ /* Reset for next path. */
+ path_can_be_null = true;
+
+ p = fail_stack.stack[--fail_stack.avail].pointer;
+
+ continue;
+ }
+ else
+ break;
}
/* We should never be about to go beyond the end of the pattern. */
assert (p < pend);
-#ifdef SWITCH_ENUM_BUG
- switch ((int) ((re_opcode_t) *p++))
-#else
- switch ((re_opcode_t) *p++)
-#endif
+ switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++))
{
/* I guess the idea here is to simply not bother with a fastmap
that is all we do. */
case duplicate:
bufp->can_be_null = 1;
- return 0;
+ goto done;
/* Following are the cases which match a character. These end
case anychar:
- /* `.' matches anything ... */
- for (j = 0; j < (1 << BYTEWIDTH); j++)
- fastmap[j] = 1;
+ {
+ int fastmap_newline = fastmap['\n'];
- /* ... except perhaps newline. */
- if (!(bufp->syntax & RE_DOT_NEWLINE))
- fastmap['\n'] = 0;
+ /* `.' matches anything ... */
+ for (j = 0; j < (1 << BYTEWIDTH); j++)
+ fastmap[j] = 1;
- /* Return if we have already set `can_be_null'; if we have,
- then the fastmap is irrelevant. Something's wrong here. */
- else if (bufp->can_be_null)
- return 0;
+ /* ... except perhaps newline. */
+ if (!(bufp->syntax & RE_DOT_NEWLINE))
+ fastmap['\n'] = fastmap_newline;
- /* Otherwise, have to check alternative paths. */
- break;
+ /* Return if we have already set `can_be_null'; if we have,
+ then the fastmap is irrelevant. Something's wrong here. */
+ else if (bufp->can_be_null)
+ goto done;
+ /* Otherwise, have to check alternative paths. */
+ break;
+ }
#ifdef emacs
case syntaxspec:
case at_dot:
case after_dot:
continue;
-#endif /* not emacs */
+#endif /* emacs */
case no_op:
/* If what's on the stack is where we are now, pop it. */
if (!FAIL_STACK_EMPTY ()
- && fail_stack.stack[fail_stack.avail - 1] == p)
+ && fail_stack.stack[fail_stack.avail - 1].pointer == p)
fail_stack.avail--;
continue;
if (p + j < pend)
{
if (!PUSH_PATTERN_OP (p + j, fail_stack))
- return -2;
+ {
+ RESET_FAIL_STACK ();
+ return -2;
+ }
}
else
bufp->can_be_null = 1;
/* Set `can_be_null' for the last path (also the first path, if the
pattern is empty). */
bufp->can_be_null |= path_can_be_null;
+
+ done:
+ RESET_FAIL_STACK ();
return 0;
} /* re_compile_fastmap */
\f
{
bufp->regs_allocated = REGS_UNALLOCATED;
regs->num_regs = 0;
- regs->start = regs->end = (regoff_t) 0;
+ regs->start = regs->end = (regoff_t *) 0;
}
}
\f
else if (endpos > total_size)
range = total_size - startpos;
- /* Update the fastmap now if not correct already. */
- if (fastmap && !bufp->fastmap_accurate)
- if (re_compile_fastmap (bufp) == -2)
- return -2;
-
/* If the search isn't to be a backwards one, don't waste time in a
- long search for a pattern that says it is anchored. */
- if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf
- && range > 0)
+ search for a pattern that must be anchored. */
+ if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf && range > 0)
{
if (startpos > 0)
return -1;
range = 1;
}
+#ifdef emacs
+ /* In a forward search for something that starts with \=.
+ don't keep searching past point. */
+ if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == at_dot && range > 0)
+ {
+ range = PT - startpos;
+ if (range <= 0)
+ return -1;
+ }
+#endif /* emacs */
+
+ /* Update the fastmap now if not correct already. */
+ if (fastmap && !bufp->fastmap_accurate)
+ if (re_compile_fastmap (bufp) == -2)
+ return -2;
+
+ /* Loop through the string, looking for a place to start matching. */
for (;;)
{
/* If a fastmap is supplied, skip quickly over characters that
inside the loop. */
if (translate)
while (range > lim
- && !fastmap[(unsigned char) translate[*d++]])
+ && !fastmap[(unsigned char)
+ translate[(unsigned char) *d++]])
range--;
else
while (range > lim && !fastmap[(unsigned char) *d++])
? string2[startpos - size1]
: string1[startpos]);
- if (!fastmap[TRANSLATE (c)])
+ if (!fastmap[(unsigned char) TRANSLATE (c)])
goto advance;
}
}
&& !bufp->can_be_null)
return -1;
- val = re_match_2 (bufp, string1, size1, string2, size2,
- startpos, regs, stop);
+ val = re_match_2_internal (bufp, string1, size1, string2, size2,
+ startpos, regs, stop);
+#ifndef REGEX_MALLOC
+#ifdef C_ALLOCA
+ alloca (0);
+#endif
+#endif
+
if (val >= 0)
return startpos;
common_op_match_null_string_p (),
group_match_null_string_p ();
-/* Structure for per-register (a.k.a. per-group) information.
- This must not be longer than one word, because we push this value
- onto the failure stack. Other register information, such as the
- starting and ending positions (which are addresses), and the list of
- inner groups (which is a bits list) are maintained in separate
- variables.
-
- We are making a (strictly speaking) nonportable assumption here: that
- the compiler will pack our bit fields into something that fits into
- the type of `word', i.e., is something that fits into one item on the
- failure stack. */
-typedef union
-{
- fail_stack_elt_t word;
- struct
- {
- /* This field is one if this group can match the empty string,
- zero if not. If not yet determined, `MATCH_NULL_UNSET_VALUE'. */
-#define MATCH_NULL_UNSET_VALUE 3
- unsigned match_null_string_p : 2;
- unsigned is_active : 1;
- unsigned matched_something : 1;
- unsigned ever_matched_something : 1;
- } bits;
-} register_info_type;
-
-#define REG_MATCH_NULL_STRING_P(R) ((R).bits.match_null_string_p)
-#define IS_ACTIVE(R) ((R).bits.is_active)
-#define MATCHED_SOMETHING(R) ((R).bits.matched_something)
-#define EVER_MATCHED_SOMETHING(R) ((R).bits.ever_matched_something)
-
-
-/* Call this when have matched something; it sets `matched' flags for the
- registers corresponding to the group of which we currently are inside.
- Also records whether this group ever matched something. We only care
- about this information at `stop_memory', and then only about the
- previous time through the loop (if the group is starred or whatever).
- So it is ok to clear all the nonactive registers here. */
-#define SET_REGS_MATCHED() \
- do \
- { \
- unsigned r; \
- for (r = lowest_active_reg; r <= highest_active_reg; r++) \
- { \
- MATCHED_SOMETHING (reg_info[r]) \
- = EVER_MATCHED_SOMETHING (reg_info[r]) \
- = 1; \
- } \
- } \
- while (0)
-
-
/* This converts PTR, a pointer into one of the search strings `string1'
and `string2' into an offset from the beginning of that string. */
-#define POINTER_TO_OFFSET(ptr) \
- (FIRST_STRING_P (ptr) ? (ptr) - string1 : (ptr) - string2 + size1)
-
-/* Registers are set to a sentinel when they haven't yet matched. */
-#define REG_UNSET_VALUE ((char *) -1)
-#define REG_UNSET(e) ((e) == REG_UNSET_VALUE)
-
+#define POINTER_TO_OFFSET(ptr) \
+ (FIRST_STRING_P (ptr) \
+ ? ((regoff_t) ((ptr) - string1)) \
+ : ((regoff_t) ((ptr) - string2 + size1)))
/* Macros for dealing with the split strings in re_match_2. */
/* Test if at very beginning or at very end of the virtual concatenation
of `string1' and `string2'. If only one string, it's `string2'. */
-#define AT_STRINGS_BEG() (d == (size1 ? string1 : string2) || !size2)
-#define AT_STRINGS_END() (d == end2)
+#define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2)
+#define AT_STRINGS_END(d) ((d) == end2)
/* Test if D points to a character which is word-constituent. We have
two special cases to check for: if past the end of string1, look at
the first character in string2; and if before the beginning of
- string2, look at the last character in string1.
-
- Assumes `string1' exists, so use in conjunction with AT_STRINGS_BEG (). */
-#define LETTER_P(d) \
+ string2, look at the last character in string1. */
+#define WORDCHAR_P(d) \
(SYNTAX ((d) == end1 ? *string2 \
- : (d) == string2 - 1 ? *(end1 - 1) : *(d)) == Sword)
+ : (d) == string2 - 1 ? *(end1 - 1) : *(d)) \
+ == Sword)
/* Test if the character before D and the one at D differ with respect
to being word-constituent. */
#define AT_WORD_BOUNDARY(d) \
- (AT_STRINGS_BEG () || AT_STRINGS_END () || LETTER_P (d - 1) != LETTER_P (d))
+ (AT_STRINGS_BEG (d) || AT_STRINGS_END (d) \
+ || WORDCHAR_P (d - 1) != WORDCHAR_P (d))
/* Free everything we malloc. */
-#ifdef REGEX_MALLOC
-#define FREE_VAR(var) if (var) free (var); var = NULL
+#ifdef MATCH_MAY_ALLOCATE
+#define FREE_VAR(var) if (var) REGEX_FREE (var); var = NULL
#define FREE_VARIABLES() \
do { \
- FREE_VAR (fail_stack.stack); \
+ REGEX_FREE_STACK (fail_stack.stack); \
FREE_VAR (regstart); \
FREE_VAR (regend); \
FREE_VAR (old_regstart); \
FREE_VAR (reg_dummy); \
FREE_VAR (reg_info_dummy); \
} while (0)
-#else /* not REGEX_MALLOC */
-/* Some MIPS systems (at least) want this to free alloca'd storage. */
-#define FREE_VARIABLES() alloca (0)
-#endif /* not REGEX_MALLOC */
-
+#else
+#define FREE_VARIABLES() ((void)0) /* Do nothing! But inhibit gcc warning. */
+#endif /* not MATCH_MAY_ALLOCATE */
/* These values must meet several constraints. They must not be valid
register values; since we have a limit of 255 registers (because
const char *string;
int size, pos;
struct re_registers *regs;
- {
- return re_match_2 (bufp, NULL, 0, string, size, pos, regs, size);
+{
+ int result = re_match_2_internal (bufp, NULL, 0, string, size,
+ pos, regs, size);
+ alloca (0);
+ return result;
}
#endif /* not emacs */
struct re_registers *regs;
int stop;
{
+ int result = re_match_2_internal (bufp, string1, size1, string2, size2,
+ pos, regs, stop);
+ alloca (0);
+ return result;
+}
+
+/* This is a separate function so that we can force an alloca cleanup
+ afterwards. */
+static int
+re_match_2_internal (bufp, string1, size1, string2, size2, pos, regs, stop)
+ struct re_pattern_buffer *bufp;
+ const char *string1, *string2;
+ int size1, size2;
+ int pos;
+ struct re_registers *regs;
+ int stop;
+{
/* General temporaries. */
int mcnt;
unsigned char *p1;
unsigned char *p = bufp->buffer;
register unsigned char *pend = p + bufp->used;
+ /* Mark the opcode just after a start_memory, so we can test for an
+ empty subpattern when we get to the stop_memory. */
+ unsigned char *just_past_start_mem = 0;
+
/* We use this to map every character in the string. */
char *translate = bufp->translate;
scanning the strings. If the latter is zero, the failure point is
a ``dummy''; if a failure happens and the failure point is a dummy,
it gets discarded and the next next one is tried. */
+#ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */
fail_stack_type fail_stack;
+#endif
#ifdef DEBUG
static unsigned failure_id = 0;
+ unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0;
+#endif
+
+ /* This holds the pointer to the failure stack, when
+ it is allocated relocatably. */
+#ifdef REL_ALLOC
+ fail_stack_elt_t *failure_stack_ptr;
#endif
/* We fill all the registers internally, independent of what we
matching and the regnum-th regend points to right after where we
stopped matching the regnum-th subexpression. (The zeroth register
keeps track of what the whole pattern matches.) */
+#ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
const char **regstart, **regend;
+#endif
/* If a group that's operated upon by a repetition operator fails to
match anything, then the register for its start will need to be
restored because it will have been set to wherever in the string we
are when we last see its open-group operator. Similarly for a
register's end. */
+#ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
const char **old_regstart, **old_regend;
+#endif
/* The is_active field of reg_info helps us keep track of which (possibly
nested) subexpressions we are currently in. The matched_something
matched any of the pattern so far this time through the reg_num-th
subexpression. These two fields get reset each time through any
loop their register is in. */
+#ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */
register_info_type *reg_info;
+#endif
/* The following record the register info as found in the above
variables when we find a match better than any we've seen before.
This happens as we backtrack through the failure points, which in
turn happens only if we have not yet matched the entire string. */
unsigned best_regs_set = false;
+#ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
const char **best_regstart, **best_regend;
+#endif
/* Logically, this is `best_regend[0]'. But we don't want to have to
allocate space for that if we're not allocating space for anything
and need to test it, it's not garbage. */
const char *match_end = NULL;
+ /* This helps SET_REGS_MATCHED avoid doing redundant work. */
+ int set_regs_matched_done = 0;
+
/* Used when we pop values we don't care about. */
+#ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
const char **reg_dummy;
register_info_type *reg_info_dummy;
+#endif
#ifdef DEBUG
/* Counts the total number of registers pushed. */
INIT_FAIL_STACK ();
+#ifdef MATCH_MAY_ALLOCATE
/* Do not bother to initialize all the register variables if there are
no groups in the pattern, as it takes a fair amount of time. If
there are groups, we include space for register 0 (the whole
return -2;
}
}
-#ifdef REGEX_MALLOC
else
{
/* We must initialize all our variables to NULL, so that
- `FREE_VARIABLES' doesn't try to free them. Too bad this isn't
- Lisp, so we could have a list of variables. As it is, */
+ `FREE_VARIABLES' doesn't try to free them. */
regstart = regend = old_regstart = old_regend = best_regstart
= best_regend = reg_dummy = NULL;
reg_info = reg_info_dummy = (register_info_type *) NULL;
}
-#endif /* REGEX_MALLOC */
+#endif /* MATCH_MAY_ALLOCATE */
/* The starting position is bogus. */
if (pos < 0 || pos > size1 + size2)
if (p == pend)
{ /* End of pattern means we might have succeeded. */
- DEBUG_PRINT1 ("End of pattern: ");
- /* If not end of string, try backtracking. Otherwise done. */
+ DEBUG_PRINT1 ("end of pattern ... ");
+
+ /* If we haven't matched the entire string, and we want the
+ longest match, try backtracking. */
if (d != end_match_2)
{
+ /* 1 if this match ends in the same string (string1 or string2)
+ as the best previous match. */
+ boolean same_str_p = (FIRST_STRING_P (match_end)
+ == MATCHING_IN_FIRST_STRING);
+ /* 1 if this match is the best seen so far. */
+ boolean best_match_p;
+
+ /* AIX compiler got confused when this was combined
+ with the previous declaration. */
+ if (same_str_p)
+ best_match_p = d > match_end;
+ else
+ best_match_p = !MATCHING_IN_FIRST_STRING;
+
DEBUG_PRINT1 ("backtracking.\n");
if (!FAIL_STACK_EMPTY ())
{ /* More failure points to try. */
- boolean same_str_p = (FIRST_STRING_P (match_end)
- == MATCHING_IN_FIRST_STRING);
/* If exceeds best match so far, save it. */
- if (!best_regs_set
- || (same_str_p && d > match_end)
- || (!same_str_p && !MATCHING_IN_FIRST_STRING))
+ if (!best_regs_set || best_match_p)
{
best_regs_set = true;
match_end = d;
goto fail;
}
- /* If no failure points, don't restore garbage. */
- else if (best_regs_set)
+ /* If no failure points, don't restore garbage. And if
+ last match is real best match, don't restore second
+ best one. */
+ else if (best_regs_set && !best_match_p)
{
restore_best_regs:
/* Restore best match. It may happen that `dend ==
For example, the pattern `x.*y.*z' against the
strings `x-' and `y-z-', if the two strings are
not consecutive in memory. */
+ DEBUG_PRINT1 ("Restoring best registers.\n");
+
d = match_end;
dend = ((d >= string1 && d <= end1)
? end_match_1 : end_match_2);
}
} /* d != end_match_2 */
- DEBUG_PRINT1 ("\nAccepting match.\n");
+ succeed_label:
+ DEBUG_PRINT1 ("Accepting match.\n");
/* If caller wants register contents data back, do it. */
if (regs && !bufp->no_sub)
regs->start = TALLOC (regs->num_regs, regoff_t);
regs->end = TALLOC (regs->num_regs, regoff_t);
if (regs->start == NULL || regs->end == NULL)
- return -2;
+ {
+ FREE_VARIABLES ();
+ return -2;
+ }
bufp->regs_allocated = REGS_REALLOCATE;
}
else if (bufp->regs_allocated == REGS_REALLOCATE)
RETALLOC (regs->start, regs->num_regs, regoff_t);
RETALLOC (regs->end, regs->num_regs, regoff_t);
if (regs->start == NULL || regs->end == NULL)
- return -2;
+ {
+ FREE_VARIABLES ();
+ return -2;
+ }
}
}
else
- assert (bufp->regs_allocated == REGS_FIXED);
+ {
+ /* These braces fend off a "empty body in an else-statement"
+ warning under GCC when assert expands to nothing. */
+ assert (bufp->regs_allocated == REGS_FIXED);
+ }
/* Convert the pointer data in `regstart' and `regend' to
indices. Register zero has to be set differently,
if (regs->num_regs > 0)
{
regs->start[0] = pos;
- regs->end[0] = (MATCHING_IN_FIRST_STRING ? d - string1
- : d - string2 + size1);
+ regs->end[0] = (MATCHING_IN_FIRST_STRING
+ ? ((regoff_t) (d - string1))
+ : ((regoff_t) (d - string2 + size1)));
}
/* Go through the first `min (num_regs, regs->num_regs)'
regs->start[mcnt] = regs->end[mcnt] = -1;
else
{
- regs->start[mcnt] = POINTER_TO_OFFSET (regstart[mcnt]);
- regs->end[mcnt] = POINTER_TO_OFFSET (regend[mcnt]);
+ regs->start[mcnt]
+ = (regoff_t) POINTER_TO_OFFSET (regstart[mcnt]);
+ regs->end[mcnt]
+ = (regoff_t) POINTER_TO_OFFSET (regend[mcnt]);
}
}
regs->start[mcnt] = regs->end[mcnt] = -1;
} /* regs && !bufp->no_sub */
- FREE_VARIABLES ();
- DEBUG_PRINT2 ("%d registers pushed.\n", num_regs_pushed);
+ DEBUG_PRINT4 ("%u failure points pushed, %u popped (%u remain).\n",
+ nfailure_points_pushed, nfailure_points_popped,
+ nfailure_points_pushed - nfailure_points_popped);
+ DEBUG_PRINT2 ("%u registers pushed.\n", num_regs_pushed);
mcnt = d - pos - (MATCHING_IN_FIRST_STRING
? string1
DEBUG_PRINT2 ("Returning %d from re_match_2.\n", mcnt);
+ FREE_VARIABLES ();
return mcnt;
}
/* Otherwise match next pattern command. */
-#ifdef SWITCH_ENUM_BUG
- switch ((int) ((re_opcode_t) *p++))
-#else
- switch ((re_opcode_t) *p++)
-#endif
+ switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++))
{
/* Ignore these. Used to ignore the n of succeed_n's which
currently have n == 0. */
DEBUG_PRINT1 ("EXECUTING no_op.\n");
break;
+ case succeed:
+ DEBUG_PRINT1 ("EXECUTING succeed.\n");
+ goto succeed_label;
/* Match the next n pattern characters exactly. The following
byte in the pattern defines n, and the n bytes after that
IS_ACTIVE (reg_info[*p]) = 1;
MATCHED_SOMETHING (reg_info[*p]) = 0;
+
+ /* Clear this whenever we change the register activity status. */
+ set_regs_matched_done = 0;
/* This is the new highest active register. */
highest_active_reg = *p;
/* Move past the register number and inner group count. */
p += 2;
+ just_past_start_mem = p;
+
break;
/* This register isn't active anymore. */
IS_ACTIVE (reg_info[*p]) = 0;
-
+
+ /* Clear this whenever we change the register activity status. */
+ set_regs_matched_done = 0;
+
/* If this was the only register active, nothing is active
anymore. */
if (lowest_active_reg == highest_active_reg)
/* If just failed to match something this time around with a
group that's operated on by a repetition operator, try to
- force exit from the ``loop,'' and restore the register
+ force exit from the ``loop'', and restore the register
information for this group that we had before trying this
last match. */
if ((!MATCHED_SOMETHING (reg_info[*p])
- || (re_opcode_t) p[-3] == start_memory)
+ || just_past_start_mem == p - 1)
&& (p + 2) < pend)
{
boolean is_a_jump_n = false;
regstart[r] = old_regstart[r];
/* xx why this test? */
- if ((int) old_regend[r] >= (int) regstart[r])
+ if (old_regend[r] >= regstart[r])
regend[r] = old_regend[r];
}
}
: bcmp (d, d2, mcnt))
goto fail;
d += mcnt, d2 += mcnt;
+
+ /* Do this because we've match some characters. */
+ SET_REGS_MATCHED ();
}
}
break;
case begline:
DEBUG_PRINT1 ("EXECUTING begline.\n");
- if (AT_STRINGS_BEG ())
+ if (AT_STRINGS_BEG (d))
{
if (!bufp->not_bol) break;
}
case endline:
DEBUG_PRINT1 ("EXECUTING endline.\n");
- if (AT_STRINGS_END ())
+ if (AT_STRINGS_END (d))
{
if (!bufp->not_eol) break;
}
/* Match at the very beginning of the data. */
case begbuf:
DEBUG_PRINT1 ("EXECUTING begbuf.\n");
- if (AT_STRINGS_BEG ())
+ if (AT_STRINGS_BEG (d))
break;
goto fail;
/* Match at the very end of the data. */
case endbuf:
DEBUG_PRINT1 ("EXECUTING endbuf.\n");
- if (AT_STRINGS_END ())
+ if (AT_STRINGS_END (d))
break;
goto fail;
the original * applied to a group), save the information
for that group and all inner ones, so that if we fail back
to this point, the group's information will be correct.
- For example, in \(a*\)*\1, we only need the preceding group,
+ For example, in \(a*\)*\1, we need the preceding group,
and in \(\(a*\)b*\)\2, we need the inner group. */
/* We can't use `p' to check ahead because we push
break;
- /* A smart repeat ends with a maybe_pop_jump.
- We change it either to a pop_failure_jump or a jump. */
+ /* A smart repeat ends with `maybe_pop_jump'.
+ We change it to either `pop_failure_jump' or `jump'. */
case maybe_pop_jump:
EXTRACT_NUMBER_AND_INCR (mcnt, p);
DEBUG_PRINT2 ("EXECUTING maybe_pop_jump %d.\n", mcnt);
detect that here, the alternative has put on a dummy
failure point which is what we will end up popping. */
- /* Skip over open/close-group commands. */
- while (p2 + 2 < pend
- && ((re_opcode_t) *p2 == stop_memory
- || (re_opcode_t) *p2 == start_memory))
- p2 += 3; /* Skip over args, too. */
+ /* Skip over open/close-group commands.
+ If what follows this loop is a ...+ construct,
+ look at what begins its body, since we will have to
+ match at least one of that. */
+ while (1)
+ {
+ if (p2 + 2 < pend
+ && ((re_opcode_t) *p2 == stop_memory
+ || (re_opcode_t) *p2 == start_memory))
+ p2 += 3;
+ else if (p2 + 6 < pend
+ && (re_opcode_t) *p2 == dummy_failure_jump)
+ p2 += 6;
+ else
+ break;
+ }
+
+ p1 = p + mcnt;
+ /* p1[0] ... p1[2] are the `on_failure_jump' corresponding
+ to the `maybe_finalize_jump' of this case. Examine what
+ follows. */
/* If we're at the end of the pattern, we can change. */
if (p2 == pend)
- {
+ {
+ /* Consider what happens when matching ":\(.*\)"
+ against ":/". I don't really understand this code
+ yet. */
p[-3] = (unsigned char) pop_failure_jump;
DEBUG_PRINT1
(" End of pattern: change to `pop_failure_jump'.\n");
{
register unsigned char c
= *p2 == (unsigned char) endline ? '\n' : p2[2];
- p1 = p + mcnt;
- /* p1[0] ... p1[2] are the `on_failure_jump' corresponding
- to the `maybe_finalize_jump' of this case. Examine what
- follows. */
if ((re_opcode_t) p1[3] == exactn && p1[5] != c)
- p[-3] = (unsigned char) pop_failure_jump;
+ {
+ p[-3] = (unsigned char) pop_failure_jump;
+ DEBUG_PRINT3 (" %c != %c => pop_failure_jump.\n",
+ c, p1[5]);
+ }
+
else if ((re_opcode_t) p1[3] == charset
|| (re_opcode_t) p1[3] == charset_not)
{
if (!not)
{
p[-3] = (unsigned char) pop_failure_jump;
- DEBUG_PRINT1
- (" No match: change to `pop_failure_jump'.\n");
+ DEBUG_PRINT1 (" No match => pop_failure_jump.\n");
+ }
+ }
+ }
+ else if ((re_opcode_t) *p2 == charset)
+ {
+#ifdef DEBUG
+ register unsigned char c
+ = *p2 == (unsigned char) endline ? '\n' : p2[2];
+#endif
+ if ((re_opcode_t) p1[3] == exactn
+ && ! ((int) p2[1] * BYTEWIDTH > (int) p1[4]
+ && (p2[1 + p1[4] / BYTEWIDTH]
+ & (1 << (p1[4] % BYTEWIDTH)))))
+ {
+ p[-3] = (unsigned char) pop_failure_jump;
+ DEBUG_PRINT3 (" %c != %c => pop_failure_jump.\n",
+ c, p1[5]);
+ }
+
+ else if ((re_opcode_t) p1[3] == charset_not)
+ {
+ int idx;
+ /* We win if the charset_not inside the loop
+ lists every character listed in the charset after. */
+ for (idx = 0; idx < (int) p2[1]; idx++)
+ if (! (p2[2 + idx] == 0
+ || (idx < (int) p1[4]
+ && ((p2[2 + idx] & ~ p1[5 + idx]) == 0))))
+ break;
+
+ if (idx == p2[1])
+ {
+ p[-3] = (unsigned char) pop_failure_jump;
+ DEBUG_PRINT1 (" No match => pop_failure_jump.\n");
+ }
+ }
+ else if ((re_opcode_t) p1[3] == charset)
+ {
+ int idx;
+ /* We win if the charset inside the loop
+ has no overlap with the one after the loop. */
+ for (idx = 0;
+ idx < (int) p2[1] && idx < (int) p1[4];
+ idx++)
+ if ((p2[2 + idx] & p1[5 + idx]) != 0)
+ break;
+
+ if (idx == p2[1] || idx == p1[4])
+ {
+ p[-3] = (unsigned char) pop_failure_jump;
+ DEBUG_PRINT1 (" No match => pop_failure_jump.\n");
}
}
}
if ((re_opcode_t) p[-1] != pop_failure_jump)
{
p[-1] = (unsigned char) jump;
+ DEBUG_PRINT1 (" Match => jump.\n");
goto unconditional_jump;
}
/* Note fall through. */
/* At the end of an alternative, we need to push a dummy failure
- point in case we are followed by a pop_failure_jump', because
+ point in case we are followed by a `pop_failure_jump', because
we don't want the failure point for the alternative to be
popped. For example, matching `(a|ab)*' against `aab'
requires that we match the `ab' alternative. */
case wordbeg:
DEBUG_PRINT1 ("EXECUTING wordbeg.\n");
- if (LETTER_P (d) && (AT_STRINGS_BEG () || !LETTER_P (d - 1)))
+ if (WORDCHAR_P (d) && (AT_STRINGS_BEG (d) || !WORDCHAR_P (d - 1)))
break;
goto fail;
case wordend:
DEBUG_PRINT1 ("EXECUTING wordend.\n");
- if (!AT_STRINGS_BEG () && LETTER_P (d - 1)
- && (!LETTER_P (d) || AT_STRINGS_END ()))
+ if (!AT_STRINGS_BEG (d) && WORDCHAR_P (d - 1)
+ && (!WORDCHAR_P (d) || AT_STRINGS_END (d)))
break;
goto fail;
#ifdef emacs
-#ifdef emacs19
case before_dot:
DEBUG_PRINT1 ("EXECUTING before_dot.\n");
if (PTR_CHAR_POS ((unsigned char *) d) >= point)
if (PTR_CHAR_POS ((unsigned char *) d) <= point)
goto fail;
break;
-#else /* not emacs19 */
- case at_dot:
- DEBUG_PRINT1 ("EXECUTING at_dot.\n");
- if (PTR_CHAR_POS ((unsigned char *) d) + 1 != point)
- goto fail;
- break;
-#endif /* not emacs19 */
case syntaxspec:
DEBUG_PRINT2 ("EXECUTING syntaxspec %d.\n", mcnt);
goto matchsyntax;
case wordchar:
- DEBUG_PRINT1 ("EXECUTING wordchar.\n");
+ DEBUG_PRINT1 ("EXECUTING Emacs wordchar.\n");
mcnt = (int) Sword;
matchsyntax:
PREFETCH ();
- if (SYNTAX (*d++) != (enum syntaxcode) mcnt) goto fail;
+ /* Can't use *d++ here; SYNTAX may be an unsafe macro. */
+ d++;
+ if (SYNTAX (d[-1]) != (enum syntaxcode) mcnt)
+ goto fail;
SET_REGS_MATCHED ();
break;
goto matchnotsyntax;
case notwordchar:
- DEBUG_PRINT1 ("EXECUTING notwordchar.\n");
+ DEBUG_PRINT1 ("EXECUTING Emacs notwordchar.\n");
mcnt = (int) Sword;
- matchnotsyntax: /* We goto here from notsyntaxspec. */
+ matchnotsyntax:
PREFETCH ();
- if (SYNTAX (*d++) == (enum syntaxcode) mcnt) goto fail;
+ /* Can't use *d++ here; SYNTAX may be an unsafe macro. */
+ d++;
+ if (SYNTAX (d[-1]) == (enum syntaxcode) mcnt)
+ goto fail;
SET_REGS_MATCHED ();
break;
case wordchar:
DEBUG_PRINT1 ("EXECUTING non-Emacs wordchar.\n");
PREFETCH ();
- if (!LETTER_P (d))
+ if (!WORDCHAR_P (d))
goto fail;
SET_REGS_MATCHED ();
+ d++;
break;
case notwordchar:
DEBUG_PRINT1 ("EXECUTING non-Emacs notwordchar.\n");
PREFETCH ();
- if (LETTER_P (d))
+ if (WORDCHAR_P (d))
goto fail;
SET_REGS_MATCHED ();
+ d++;
break;
#endif /* not emacs */
ret = regex_compile (pattern, length, re_syntax_options, bufp);
- return re_error_msg[(int) ret];
+ if (!ret)
+ return NULL;
+ return gettext (re_error_msgid[(int) ret]);
}
\f
/* Entry points compatible with 4.2 BSD regex library. We don't define
- them if this is an Emacs or POSIX compilation. */
+ them unless specifically requested. */
-#if !defined (emacs) && !defined (_POSIX_SOURCE)
+#ifdef _REGEX_RE_COMP
/* BSD has one and only one pattern buffer. */
static struct re_pattern_buffer re_comp_buf;
if (!s)
{
if (!re_comp_buf.buffer)
- return "No previous regular expression";
+ return gettext ("No previous regular expression");
return 0;
}
{
re_comp_buf.buffer = (unsigned char *) malloc (200);
if (re_comp_buf.buffer == NULL)
- return "Memory exhausted";
+ return gettext (re_error_msgid[(int) REG_ESPACE]);
re_comp_buf.allocated = 200;
re_comp_buf.fastmap = (char *) malloc (1 << BYTEWIDTH);
if (re_comp_buf.fastmap == NULL)
- return "Memory exhausted";
+ return gettext (re_error_msgid[(int) REG_ESPACE]);
}
/* Since `re_exec' always passes NULL for the `regs' argument, we
ret = regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf);
- /* Yes, we're discarding `const' here. */
- return (char *) re_error_msg[(int) ret];
+ if (!ret)
+ return NULL;
+
+ /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */
+ return (char *) gettext (re_error_msgid[(int) ret]);
}
return
0 <= re_search (&re_comp_buf, s, len, 0, len, (struct re_registers *) 0);
}
-#endif /* not emacs and not _POSIX_SOURCE */
+#endif /* _REGEX_RE_COMP */
\f
/* POSIX.2 functions. Don't define these for Emacs. */
{
reg_errcode_t ret;
unsigned syntax
- = cflags & REG_EXTENDED ? RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC;
+ = (cflags & REG_EXTENDED) ?
+ RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC;
/* regex_compile will allocate the space for the compiled pattern. */
preg->buffer = 0;
+ preg->allocated = 0;
+ preg->used = 0;
/* Don't bother to use a fastmap when searching. This simplifies the
REG_NEWLINE case: if we used a fastmap, we'd have to put all the
/* Map uppercase characters to corresponding lowercase ones. */
for (i = 0; i < CHAR_SET_SIZE; i++)
- preg->translate[i] = isupper (i) ? tolower (i) : i;
+ preg->translate[i] = ISUPPER (i) ? tolower (i) : i;
}
else
preg->translate = NULL;
/* Returns a message corresponding to an error code, ERRCODE, returned
- from either regcomp or regexec. */
+ from either regcomp or regexec. We don't use PREG here. */
size_t
regerror (errcode, preg, errbuf, errbuf_size)
char *errbuf;
size_t errbuf_size;
{
- const char *msg
- = re_error_msg[errcode] == NULL ? "Success" : re_error_msg[errcode];
- size_t msg_size = strlen (msg) + 1; /* Includes the null. */
+ const char *msg;
+ size_t msg_size;
+
+ if (errcode < 0
+ || errcode >= (sizeof (re_error_msgid) / sizeof (re_error_msgid[0])))
+ /* Only error codes returned by the rest of the code should be passed
+ to this routine. If we are given anything else, or if other regex
+ code generates an invalid error code, then the program has a bug.
+ Dump core so we can fix it. */
+ abort ();
+
+ msg = gettext (re_error_msgid[errcode]);
+
+ msg_size = strlen (msg) + 1; /* Includes the null. */
if (errbuf_size != 0)
{
projects / gnulib.git / blobdiff