(Implements POSIX draft P10003.2/D11.2, except for
internationalization features.)
- Copyright (C) 1993 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
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
- Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
+ Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307,
+ USA. */
/* AIX requires this to be the first thing in the file. */
#if defined (_AIX) && !defined (REGEX_MALLOC)
#pragma alloca
#endif
+#undef _GNU_SOURCE
#define _GNU_SOURCE
#ifdef HAVE_CONFIG_H
-#if defined (CONFIG_BROKETS)
-/* We use <config.h> instead of "config.h" so that a compilation
- using -I. -I$srcdir will use ./config.h rather than $srcdir/config.h
- (which it would do because it found this file in $srcdir). */
#include <config.h>
-#else
-#include "config.h"
-#endif
#endif
/* We need this for `regex.h', and perhaps for the Emacs include files. */
#include <sys/types.h>
-#ifdef HAVE_CONFIG_H
-#include "config.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
+
+#ifndef gettext_noop
+/* This define is so xgettext can find the internationalizable
+ strings. */
+#define gettext_noop(String) String
#endif
/* The `emacs' switch turns on certain matching commands
#include "buffer.h"
#include "syntax.h"
-/* Emacs uses `NULL' as a predicate. */
-#undef NULL
-
#else /* not emacs */
-#ifdef 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
-/* 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 HAVE_STRING_H || STDC_HEADERS
+/* 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))
#else
#include <strings.h>
#endif
+#endif
/* Define the syntax stuff for \<, \>, etc. */
/* This must be nonzero for the wordchar and notwordchar pattern
commands in re_match_2. */
-#ifndef Sword
+#ifndef Sword
#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;
#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',
use `alloca' instead of `malloc'. This is because using malloc in
re_search* or re_match* could cause memory leaks when C-g is used in
Emacs; also, malloc is slower and causes storage fragmentation. On
- the other hand, malloc is more portable, and easier to debug.
-
+ the other hand, malloc is more portable, and easier to debug.
+
Because we sometimes use alloca, some routines have to be macros,
not functions -- `alloca'-allocated space disappears at the end of the
function it is called in. */
#define REGEX_ALLOCATE malloc
#define REGEX_REALLOCATE(source, osize, nsize) realloc (source, nsize)
+#define REGEX_FREE free
#else /* not REGEX_MALLOC */
#if HAVE_ALLOCA_H
#include <alloca.h>
#else /* not __GNUC__ or HAVE_ALLOCA_H */
+#if 0 /* It is a bad idea to declare alloca. We always cast the result. */
#ifndef _AIX /* Already did AIX, up at the top. */
char *alloca ();
#endif /* not _AIX */
-#endif /* not HAVE_ALLOCA_H */
+#endif
+#endif /* not HAVE_ALLOCA_H */
#endif /* not __GNUC__ */
#endif /* not alloca */
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
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,
/* Analogously, for end of buffer/string. */
endbuf,
-
+
/* Followed by two byte relative address to which to jump. */
- jump,
+ jump,
/* Same as jump, but marks the end of an alternative. */
jump_past_alt,
/* Followed by two-byte relative address of place to resume at
in case of failure. */
on_failure_jump,
-
+
/* Like on_failure_jump, but pushes a placeholder instead of the
current string position when executed. */
on_failure_keep_string_jump,
-
+
/* Throw away latest failure point and then jump to following
two-byte relative address. */
pop_failure_jump,
int *dest;
unsigned char *source;
{
- int temp = SIGN_EXTEND_CHAR (*(source + 1));
+ int temp = SIGN_EXTEND_CHAR (*(source + 1));
*dest = *source & 0377;
*dest += temp << 8;
}
extract_number_and_incr (destination, source)
int *destination;
unsigned char **source;
-{
+{
extract_number (destination, *source);
*source += 2;
}
if (debug) print_double_string (w, s1, sz1, s2, sz2)
-extern void printchar ();
-
/* Print the fastmap in human-readable form. */
void
char *fastmap;
{
unsigned was_a_range = 0;
- unsigned i = 0;
-
+ unsigned i = 0;
+
while (i < (1 << BYTEWIDTH))
{
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);
}
}
}
- putchar ('\n');
+ putchar ('\n');
}
printf ("(null)\n");
return;
}
-
+
/* Loop over pattern commands. */
while (p < pend)
{
do
{
putchar ('/');
- printchar (*p++);
+ putchar (*p++);
}
while (--mcnt);
break;
printf ("/charset [%s",
(re_opcode_t) *(p - 1) == charset_not ? "^" : "");
-
+
assert (p + *p < pend);
for (c = 0; c < 256; c++)
/* Have we broken a range? */
else if (last + 1 != c && in_range)
{
- printchar (last);
+ putchar (last);
in_range = 0;
}
-
+
if (! in_range)
- printchar (c);
+ putchar (c);
last = c;
}
if (in_range)
- printchar (last);
+ putchar (last);
putchar (']');
case push_dummy_failure:
printf ("/push_dummy_failure");
break;
-
+
case maybe_pop_jump:
extract_number_and_incr (&mcnt, &p);
printf ("/maybe_pop_jump to %d", p + mcnt - start);
case pop_failure_jump:
extract_number_and_incr (&mcnt, &p);
printf ("/pop_failure_jump to %d", p + mcnt - start);
- break;
-
+ break;
+
case jump_past_alt:
extract_number_and_incr (&mcnt, &p);
printf ("/jump_past_alt to %d", p + mcnt - start);
- break;
-
+ break;
+
case jump:
extract_number_and_incr (&mcnt, &p);
printf ("/jump to %d", p + mcnt - start);
break;
- case succeed_n:
+ case succeed_n:
extract_number_and_incr (&mcnt, &p);
extract_number_and_incr (&mcnt2, &p);
printf ("/succeed_n to %d, %d times", p + mcnt - start, mcnt2);
break;
-
- case jump_n:
+
+ case jump_n:
extract_number_and_incr (&mcnt, &p);
extract_number_and_incr (&mcnt2, &p);
printf ("/jump_n to %d, %d times", p + mcnt - start, mcnt2);
break;
-
- case set_number_at:
+
+ case set_number_at:
extract_number_and_incr (&mcnt, &p);
extract_number_and_incr (&mcnt2, &p);
printf ("/set_number_at location %d to %d", p + mcnt - start, mcnt2);
break;
-
+
case wordbound:
printf ("/wordbound");
break;
case wordbeg:
printf ("/wordbeg");
break;
-
+
case wordend:
printf ("/wordend");
-
+
#ifdef emacs
case before_dot:
printf ("/before_dot");
mcnt = *p++;
printf ("/%d", mcnt);
break;
-
+
case notsyntaxspec:
printf ("/notsyntaxspec");
mcnt = *p++;
case wordchar:
printf ("/wordchar");
break;
-
+
case notwordchar:
printf ("/notwordchar");
break;
int size2;
{
unsigned this_char;
-
+
if (where == NULL)
printf ("(null)");
else
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;
+ where = string2;
}
for (this_char = where - string2; this_char < size2; this_char++)
- printchar (string2[this_char]);
+ putchar (string2[this_char]);
}
}
/* 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
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. */
-
-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 */
+ 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[] =
+ {
+ gettext_noop ("Success"), /* REG_NOERROR */
+ gettext_noop ("No match"), /* REG_NOMATCH */
+ gettext_noop ("Invalid regular expression"), /* REG_BADPAT */
+ gettext_noop ("Invalid collation character"), /* REG_ECOLLATE */
+ gettext_noop ("Invalid character class name"), /* REG_ECTYPE */
+ gettext_noop ("Trailing backslash"), /* REG_EESCAPE */
+ gettext_noop ("Invalid back reference"), /* REG_ESUBREG */
+ gettext_noop ("Unmatched [ or [^"), /* REG_EBRACK */
+ gettext_noop ("Unmatched ( or \\("), /* REG_EPAREN */
+ gettext_noop ("Unmatched \\{"), /* REG_EBRACE */
+ gettext_noop ("Invalid content of \\{\\}"), /* REG_BADBR */
+ gettext_noop ("Invalid range end"), /* REG_ERANGE */
+ gettext_noop ("Memory exhausted"), /* REG_ESPACE */
+ gettext_noop ("Invalid preceding regular expression"), /* REG_BADRPT */
+ gettext_noop ("Premature end of regular expression"), /* REG_EEND */
+ gettext_noop ("Regular expression too big"), /* REG_ESIZE */
+ gettext_noop ("Unmatched ) or \\)"), /* REG_ERPAREN */
};
\f
/* Avoiding alloca during matching, to placate r_alloc. */
-/* Define MATCH_MAY_ALLOCATE if we need to make sure that the
+/* 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
ralloc heap) shift the data out from underneath the regexp
routines.
- Here's another reason to avoid allocation: Emacs insists on
- processing input from X in a signal handler; processing X input may
+ 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
/* Normally, this is fine. */
#define MATCH_MAY_ALLOCATE
-/* But under some circumstances, it's not. */
-#if defined (emacs) || (defined (REL_ALLOC) && defined (C_ALLOCA))
+/* 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. */
-
+ REGEX_ALLOCATE_STACK. */
+
/* Number of failure points for which to initially allocate space
when matching. If this number is exceeded, we allocate more
#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.
+ exactly that if always used MAX_FAILURE_ITEMS items 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 = 20000;
+#else
int re_max_failures = 2000;
+#endif
+
+union fail_stack_elt
+{
+ unsigned char *pointer;
+ int integer;
+};
-typedef unsigned char *fail_stack_elt_t;
+typedef union fail_stack_elt fail_stack_elt_t;
typedef struct
{
#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 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 (INIT_FAILURE_ALLOC * sizeof (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 requires `destination' be declared. */
+ 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 ((fail_stack).stack, \
+ REGEX_REALLOCATE_STACK ((fail_stack).stack, \
(fail_stack).size * sizeof (fail_stack_elt_t), \
((fail_stack).size << 1) * sizeof (fail_stack_elt_t)), \
\
1)))
-/* Push PATTERN_OP on FAIL_STACK.
-
+/* 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(pattern_op, fail_stack) \
+#define PUSH_PATTERN_OP(POINTER, FAIL_STACK) \
((FAIL_STACK_FULL () \
- && !DOUBLE_FAIL_STACK (fail_stack)) \
- ? 0 \
- : ((fail_stack).stack[(fail_stack).avail++] = pattern_op, \
- 1))
+ && !DOUBLE_FAIL_STACK (FAIL_STACK)) \
+ ? 0 \
+ : ((FAIL_STACK).stack[(FAIL_STACK).avail++].pointer = POINTER, \
+ 1))
-/* This pushes an item onto the failure stack. Must be a four-byte
- value. Assumes the variable `fail_stack'. Probably should only
+/* 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_ITEM(item) \
- fail_stack.stack[fail_stack.avail++] = (fail_stack_elt_t) item
+#define PUSH_FAILURE_POINTER(item) \
+ fail_stack.stack[fail_stack.avail++].pointer = (unsigned char *) (item)
-/* The complement operation. Assumes `fail_stack' is nonempty. */
-#define POP_FAILURE_ITEM() fail_stack.stack[--fail_stack.avail]
+/* 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_ITEM
-#define DEBUG_POP(item_addr) *(item_addr) = POP_FAILURE_ITEM ()
+#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)
/* Push the information about the state we will need
- if we ever fail back to it.
-
+ 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) \
/* Ensure we have enough space allocated for what we will push. */ \
while (REMAINING_AVAIL_SLOTS < NUM_FAILURE_ITEMS) \
{ \
- if (!DOUBLE_FAIL_STACK (fail_stack)) \
+ if (!DOUBLE_FAIL_STACK (fail_stack)) \
return failure_code; \
\
DEBUG_PRINT2 ("\n Doubled stack; size now: %d\n", \
/* 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++); \
+ if (1) \
+ 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 (" start: 0x%x\n", regstart[this_reg]); \
+ PUSH_FAILURE_POINTER (regstart[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 (" 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_ITEM (lowest_active_reg); \
+ PUSH_FAILURE_INT (lowest_active_reg); \
\
DEBUG_PRINT2 (" Pushing high active reg: %d\n", highest_active_reg);\
- PUSH_FAILURE_ITEM (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_ITEM (pattern_place); \
+ 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_ITEM (string_place); \
+ PUSH_FAILURE_POINTER (string_place); \
\
DEBUG_PRINT2 (" Pushing failure id: %u\n", failure_id); \
DEBUG_PUSH (failure_id); \
#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 used to use (num_regs - 1), which is the number of registers
+ this regexp will save; but that was changed to 5
+ to avoid stack overflow for a regexp with lots of parens. */
+#define MAX_FAILURE_ITEMS (5 * 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)
+#define NUM_FAILURE_ITEMS \
+ (((0 \
+ ? 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)
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 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 (); \
+ string_temp = POP_FAILURE_POINTER (); \
if (string_temp != NULL) \
str = (const char *) string_temp; \
\
DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \
DEBUG_PRINT1 ("'\n"); \
\
- pat = (unsigned char *) POP_FAILURE_ITEM (); \
+ 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_ITEM (); \
+ high_reg = (unsigned) POP_FAILURE_INT (); \
DEBUG_PRINT2 (" Popping high active reg: %d\n", high_reg); \
\
- low_reg = (unsigned) POP_FAILURE_ITEM (); \
+ low_reg = (unsigned) POP_FAILURE_INT (); \
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); \
+ if (1) \
+ 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]); \
+ 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_ITEM (); \
- DEBUG_PRINT2 (" end: 0x%x\n", regend[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_ITEM (); \
- DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \
+ regstart[this_reg] = (const char *) POP_FAILURE_POINTER (); \
+ DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \
+ } \
+ else \
+ { \
+ for (this_reg = highest_active_reg; this_reg > high_reg; this_reg--) \
+ { \
+ reg_info[this_reg].word.integer = 0; \
+ regend[this_reg] = 0; \
+ regstart[this_reg] = 0; \
+ } \
+ highest_active_reg = high_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.
- This must not be longer than one word, because we push this value
- onto the failure stack. Other register information, such as the
+ 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.
-
+ 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;
#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; \
- } \
+ 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)
-
/* Registers are set to a sentinel when they haven't yet matched. */
-#define REG_UNSET_VALUE ((char *) -1)
+static char reg_unset_dummy;
+#define REG_UNSET_VALUE (®_unset_dummy)
#define REG_UNSET(e) ((e) == REG_UNSET_VALUE)
-
-
-\f
-/* How do we implement a missing MATCH_MAY_ALLOCATE?
- We make the fail stack a global thing, and then grow it to
- re_max_failures when we compile. */
-#ifndef MATCH_MAY_ALLOCATE
-static fail_stack_type fail_stack;
-
-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;
-#endif
-
\f
/* Subroutine declarations and macros for regex_compile. */
static boolean group_in_compile_stack ();
static reg_errcode_t compile_range ();
-/* Fetch the next character in the uncompiled pattern---translating it
+/* Fetch the next character in the uncompiled pattern---translating it
if necessary. Also cast from a signed character in the constant
string passed to us by the user to an unsigned char that we can use
as an array index (in, e.g., `translate'). */
+#ifndef PATFETCH
#define PATFETCH(c) \
do {if (p == pend) return REG_EEND; \
c = (unsigned char) *p++; \
- if (translate) c = translate[c]; \
+ if (translate) c = (unsigned char) translate[c]; \
} while (0)
+#endif
/* Fetch the next character in the uncompiled pattern, with no
translation. */
cast the subscript to translate because some data is declared as
`char *', to avoid warnings when a string constant is passed. But
when we use a character as a subscript we must make it unsigned. */
-#define TRANSLATE(d) (translate ? translate[(unsigned char) (d)] : (d))
+#ifndef TRANSLATE
+#define TRANSLATE(d) \
+ (translate ? (char) translate[(unsigned char) (d)] : (d))
+#endif
/* Macros for outputting the compiled pattern into `buffer'. */
pattern_offset_t begalt_offset;
pattern_offset_t fixup_alt_jump;
pattern_offset_t inner_group_offset;
- pattern_offset_t laststart_offset;
+ pattern_offset_t laststart_offset;
regnum_t regnum;
} compile_stack_elt_t;
PATFETCH (c); \
} \
} \
- }
+ }
#define CHAR_CLASS_MAX_LENGTH 6 /* Namely, `xdigit'. */
|| 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.
`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;
`char *' (i.e., signed), we declare these variables as unsigned, so
they can be reliably used as array indices. */
register unsigned char c, c1;
-
+
/* A random temporary spot in PATTERN. */
const char *p1;
/* Points to the end of the buffer, where we should append. */
register unsigned char *b;
-
+
/* Keeps track of unclosed groups. */
compile_stack_type compile_stack;
/* Points to the current (ending) position in the pattern. */
const char *p = pattern;
const char *pend = pattern + size;
-
+
/* How to translate the characters in the pattern. */
- char *translate = bufp->translate;
+ RE_TRANSLATE_TYPE translate = bufp->translate;
/* Address of the count-byte of the most recently inserted `exactn'
command. This makes it possible to tell if a new exact-match
/* Place in the uncompiled pattern (i.e., the {) to
which to go back if the interval is invalid. */
const char *beg_interval;
-
+
/* Address of the place where a forward jump should go to the end of
the containing expression. Each alternative of an `or' -- except the
last -- ends with a forward jump of this sort. */
if (debug)
{
unsigned debug_count;
-
+
for (debug_count = 0; debug_count < size; debug_count++)
- printchar (pattern[debug_count]);
+ putchar (pattern[debug_count]);
putchar ('\n');
}
#endif /* DEBUG */
printer (for debugging) will think there's no pattern. We reset it
at the end. */
bufp->used = 0;
-
+
/* Always count groups, whether or not bufp->no_sub is set. */
- bufp->re_nsub = 0;
+ bufp->re_nsub = 0;
#if !defined (emacs) && !defined (SYNTAX_TABLE)
/* Initialize the syntax table. */
{ /* 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;
}
case '$':
{
if ( /* If at end of pattern, it's an operator. */
- p == pend
+ p == pend
/* If context independent, it's an operator. */
|| syntax & RE_CONTEXT_INDEP_ANCHORS
/* Otherwise, depends on what's next. */
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;
}
{
/* Are we optimizing this jump? */
boolean keep_string_p = false;
-
+
/* 1 means zero (many) matches is allowed. */
char zero_times_ok = 0, many_times_ok = 0;
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 == '?'))
/* Star, etc. applied to an empty pattern is equivalent
to an empty pattern. */
- if (!laststart)
+ if (!laststart)
break;
/* Now we know whether or not zero matches is allowed
{ /* More than one repetition is allowed, so put in at the
end a backward relative jump from `b' to before the next
jump we're going to put in below (which jumps from
- laststart to after this jump).
+ laststart to after this jump).
But if we are at the `*' in the exact sequence `.*\n',
insert an unconditional jump backwards to the .,
{
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. */
/* We test `*p == '^' twice, instead of using an if
statement, so we only need one BUF_PUSH. */
- BUF_PUSH (*p == '^' ? charset_not : charset);
+ BUF_PUSH (*p == '^' ? charset_not : charset);
if (*p == '^')
p++;
/* 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
beginning or the end of a list, then it's the range
operator. */
- if (c == '-'
- && !(p - 2 >= pattern && p[-2] == '[')
+ if (c == '-'
+ && !(p - 2 >= pattern && p[-2] == '[')
&& !(p - 3 >= pattern && p[-3] == '[' && p[-2] == '^')
&& *p != ']')
{
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] != ']')
/* Move past the `-'. */
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 (;;)
{
str[c1] = '\0';
/* If isn't a word bracketed by `[:' and:`]':
- undo the ending character, the letters, and leave
+ undo the ending character, the letters, and leave
the leading `:' and `[' (but set bits for them). */
if (c == ':' && *p == ']')
{
boolean is_space = STREQ (str, "space");
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);
+ PATFETCH (c);
- if (p == pend) return REG_EBRACK;
+ if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
for (ch = 0; ch < 1 << BYTEWIDTH; 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))
- || (is_digit && ISDIGIT (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))
- || (is_punct && ISPUNCT (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);
+ SET_LIST_BIT (ch);
}
had_char_class = true;
}
else
{
c1++;
- while (c1--)
+ while (c1--)
PATUNFETCH;
SET_LIST_BIT ('[');
SET_LIST_BIT (':');
/* Discard any (non)matching list bytes that are all 0 at the
end of the map. Decrease the map-length byte too. */
- while ((int) b[-1] > 0 && b[b[-1] - 1] == 0)
- b[-1]--;
+ while ((int) b[-1] > 0 && b[b[-1] - 1] == 0)
+ b[-1]--;
b += b[-1];
}
break;
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
regnum++;
if (COMPILE_STACK_FULL)
- {
+ {
RETALLOC (compile_stack.stack, compile_stack.size << 1,
compile_stack_elt_t);
if (compile_stack.stack == NULL) return REG_ESPACE;
whole pattern moves because of realloc, they will still
be valid. */
COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer;
- COMPILE_STACK_TOP.fixup_alt_jump
+ COMPILE_STACK_TOP.fixup_alt_jump
= fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0;
COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer;
COMPILE_STACK_TOP.regnum = regnum;
COMPILE_STACK_TOP.inner_group_offset = b - bufp->buffer + 2;
BUF_PUSH_3 (start_memory, regnum, 0);
}
-
+
compile_stack.avail++;
fixup_alt_jump = 0;
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)
`pop_failure_jump' to pop. See comments at
`push_dummy_failure' in `re_match_2'. */
BUF_PUSH (push_dummy_failure);
-
+
/* We allocated space for this jump when we assigned
to `fixup_alt_jump', in the `handle_alt' case below. */
STORE_JUMP (jump_past_alt, fixup_alt_jump, b - 1);
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''. */
as in `(ab)c(de)' -- the second group is #2. */
regnum_t this_group_regnum;
- compile_stack.avail--;
+ compile_stack.avail--;
begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset;
fixup_alt_jump
= COMPILE_STACK_TOP.fixup_alt_jump
- ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1
+ ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1
: 0;
laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset;
this_group_regnum = COMPILE_STACK_TOP.regnum;
{
unsigned char *inner_group_loc
= bufp->buffer + COMPILE_STACK_TOP.inner_group_offset;
-
+
*inner_group_loc = regnum - this_group_regnum;
BUF_PUSH_3 (stop_memory, this_group_regnum,
regnum - this_group_regnum);
jump (put in below, which in turn will jump to the next
(if any) alternative's such jump, etc.). The last such
jump jumps to the correct final destination. A picture:
- _____ _____
- | | | |
- | v | v
- a | b | c
+ _____ _____
+ | | | |
+ | 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
break;
- case '{':
+ case '{':
/* If \{ is a literal. */
if (!(syntax & RE_INTERVALS)
- /* If we're at `\{' and it's not the open-interval
+ /* If we're at `\{' and it's not the open-interval
operator. */
|| ((syntax & RE_INTERVALS) && (syntax & RE_NO_BK_BRACES))
|| (p - 2 == pattern && p == pend))
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;
+ else
+ FREE_STACK_RETURN (REG_BADBR);
}
- if (!(syntax & RE_NO_BK_BRACES))
+ 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;
+ else
+ 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
jump_n <succeed_n addr> <jump count>
(The upper bound and `jump_n' are omitted if
`upper_bound' is 1, though.) */
- else
+ else
{ /* If the upper bound is > 1, we need to insert
more at the end of the loop. */
unsigned nbytes = 10 + (upper_bound > 1) * 10;
lower_bound);
b += 5;
- /* Code to initialize the lower bound. Insert
+ /* Code to initialize the lower bound. Insert
before the `succeed_n'. The `5' is the last two
bytes of this `set_number_at', plus 3 bytes of
the following `succeed_n'. */
{ /* More than one repetition is allowed, so
append a backward jump to the `succeed_n'
that starts this interval.
-
+
When we've reached this during matching,
we'll have matched the interval once, so
jump back only `upper_bound - 1' times. */
so everything is getting moved up by 5.
Conclusion: (b - 2) - (laststart + 3) + 5,
i.e., b - laststart.
-
+
We insert this at the beginning of the loop
so that if we fail during matching, we'll
reinitialize the bounds. */
beg_interval = NULL;
/* normal_char and normal_backslash need `c'. */
- PATFETCH (c);
+ PATFETCH (c);
if (!(syntax & RE_NO_BK_BRACES))
{
BUF_PUSH (at_dot);
break;
- case 's':
+ case 's':
laststart = b;
PATFETCH (c);
BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]);
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))
/* Expects the character in `c'. */
normal_char:
/* If no exactn currently being built. */
- if (!pending_exact
+ if (!pending_exact
/* If last exactn not at current position. */
|| pending_exact + *pending_exact + 1 != b
-
+
/* We have only one byte following the exactn for the count. */
|| *pending_exact == (1 << BYTEWIDTH) - 1
: (p[0] == '\\' && p[1] == '{'))))
{
/* Start building a new exactn. */
-
+
laststart = b;
BUF_PUSH_2 (exactn, 0);
pending_exact = b - 1;
}
-
+
BUF_PUSH (c);
(*pending_exact)++;
break;
} /* switch (c) */
} /* while p != pend */
-
+
/* Through the pattern now. */
-
+
if (fixup_alt_jump)
STORE_JUMP (jump_past_alt, fixup_alt_jump, b);
- if (!COMPILE_STACK_EMPTY)
- return REG_EPAREN;
+ if (!COMPILE_STACK_EMPTY)
+ 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);
/* 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. */
- fail_stack.size = (2 * re_max_failures * MAX_FAILURE_ITEMS);
- if (! fail_stack.stack)
- fail_stack.stack =
- (fail_stack_elt_t *) malloc (fail_stack.size
- * sizeof (fail_stack_elt_t));
-
- /* Initialize some other variables the matcher uses. */
- 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);
+ 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
+#endif /* not MATCH_MAY_ALLOCATE */
return REG_NOERROR;
} /* regex_compile */
re_opcode_t op;
unsigned char *loc;
int arg;
- unsigned char *end;
+ unsigned char *end;
{
register unsigned char *pfrom = end;
register unsigned char *pto = end + 3;
while (pfrom != loc)
*--pto = *--pfrom;
-
+
store_op1 (op, loc, arg);
}
re_opcode_t op;
unsigned char *loc;
int arg1, arg2;
- unsigned char *end;
+ unsigned char *end;
{
register unsigned char *pfrom = end;
register unsigned char *pto = end + 5;
while (pfrom != loc)
*--pto = *--pfrom;
-
+
store_op2 (op, loc, arg1, arg2);
}
{
const char *prev = p - 2;
boolean prev_prev_backslash = prev > pattern && prev[-1] == '\\';
-
+
return
/* After a subexpression? */
(*prev == '(' && (syntax & RE_NO_BK_PARENS || prev_prev_backslash))
{
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? */
(syntax & RE_NO_BK_PARENS ? *next == ')'
}
-/* Returns true if REGNUM is in one of COMPILE_STACK's elements and
+/* Returns true if REGNUM is in one of COMPILE_STACK's elements and
false if it's not. */
static boolean
{
int this_element;
- for (this_element = compile_stack.avail - 1;
- this_element >= 0;
+ for (this_element = compile_stack.avail - 1;
+ this_element >= 0;
this_element--)
if (compile_stack.stack[this_element].regnum == regnum)
return true;
starting character is in `P[-2]'. (`P[-1]' is the character `-'.)
Then we set the translation of all bits between the starting and
ending characters (inclusive) in the compiled pattern B.
-
+
Return an error code.
-
+
We use these short variable names so we can use the same macros as
`regex_compile' itself. */
static reg_errcode_t
compile_range (p_ptr, pend, translate, syntax, b)
const char **p_ptr, *pend;
- char *translate;
+ RE_TRANSLATE_TYPE translate;
reg_syntax_t syntax;
unsigned char *b;
{
const char *p = *p_ptr;
int range_start, range_end;
-
+
if (p == pend)
return REG_ERANGE;
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
+ We also want to fetch the endpoints without translating them; the
appropriate translation is done in the bit-setting loop below. */
- range_start = ((unsigned char *) p)[-2];
- range_end = ((unsigned char *) p)[0];
+ /* 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];
/* Have to increment the pointer into the pattern string, so the
caller isn't still at the ending character. */
{
SET_LIST_BIT (TRANSLATE (this_char));
}
-
+
return REG_NOERROR;
}
\f
The caller must supply the address of a (1 << BYTEWIDTH)-byte data
area as BUFP->fastmap.
-
+
We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in
the pattern buffer.
#endif
/* We don't push any register information onto the failure stack. */
unsigned num_regs = 0;
-
+
register char *fastmap = bufp->fastmap;
unsigned char *pattern = bufp->buffer;
unsigned long size = bufp->used;
unsigned char *p = pattern;
register unsigned char *pend = pattern + size;
+ /* This holds the pointer to the failure stack, when
+ it is allocated relocatably. */
+ fail_stack_elt_t *failure_stack_ptr;
+
/* 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
boolean succeed_n_p = false;
assert (fastmap != NULL && p != NULL);
-
+
INIT_FAIL_STACK ();
bzero (fastmap, 1 << BYTEWIDTH); /* Assume nothing's valid. */
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:
case jump_past_alt:
case dummy_failure_jump:
EXTRACT_NUMBER_AND_INCR (j, p);
- p += j;
+ p += j;
if (j > 0)
continue;
-
+
/* Jump backward implies we just went through the body of a
loop and matched nothing. Opcode jumped to should be
`on_failure_jump' or `succeed_n'. Just treat it like an
p++;
EXTRACT_NUMBER_AND_INCR (j, p);
- p += j;
-
+ p += j;
+
/* 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)
+ if (!FAIL_STACK_EMPTY ()
+ && 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;
case succeed_n:
/* Get to the number of times to succeed. */
- p += 2;
+ p += 2;
/* Increment p past the n for when k != 0. */
EXTRACT_NUMBER_AND_INCR (k, p);
/* 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
int size, startpos, range;
struct re_registers *regs;
{
- return re_search_2 (bufp, NULL, 0, string, size, startpos, range,
+ return re_search_2 (bufp, NULL, 0, string, size, startpos, range,
regs, size);
}
/* Using the compiled pattern in BUFP->buffer, first tries to match the
virtual concatenation of STRING1 and STRING2, starting first at index
STARTPOS, then at STARTPOS + 1, and so on.
-
+
STRING1 and STRING2 have length SIZE1 and SIZE2, respectively.
-
+
RANGE is how far to scan while trying to match. RANGE = 0 means try
only at STARTPOS; in general, the last start tried is STARTPOS +
RANGE.
-
+
In REGS, return the indices of the virtual concatenation of STRING1
and STRING2 that matched the entire BUFP->buffer and its contained
subexpressions.
-
+
Do not consider matching one past the index STOP in the virtual
concatenation of STRING1 and STRING2.
{
int val;
register char *fastmap = bufp->fastmap;
- register char *translate = bufp->translate;
+ register RE_TRANSLATE_TYPE translate = bufp->translate;
int total_size = size1 + size2;
int endpos = startpos + range;
/* Check for out-of-range STARTPOS. */
if (startpos < 0 || startpos > total_size)
return -1;
-
+
/* Fix up RANGE if it might eventually take us outside
- the virtual concatenation of STRING1 and STRING2. */
- if (endpos < -1)
- range = -1 - startpos;
+ the virtual concatenation of STRING1 and STRING2.
+ Make sure we won't move STARTPOS below 0 or above TOTAL_SIZE. */
+ if (endpos < 0)
+ range = 0 - startpos;
else if (endpos > total_size)
range = total_size - startpos;
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
cannot be the start of a match. If the pattern can match the
null string, however, we don't need to skip characters; we want
lim = range - (size1 - startpos);
d = (startpos >= size1 ? string2 - size1 : string1) + startpos;
-
+
/* Written out as an if-else to avoid testing `translate'
inside the loop. */
if (translate)
else /* Searching backwards. */
{
register char c = (size1 == 0 || startpos >= size1
- ? string2[startpos - size1]
+ ? string2[startpos - size1]
: string1[startpos]);
if (!fastmap[(unsigned char) TRANSLATE (c)])
&& !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;
-
+
if (val == -2)
return -2;
advance:
- if (!range)
+ if (!range)
break;
- else if (range > 0)
+ else if (range > 0)
{
- range--;
+ range--;
startpos++;
}
else
{
- range++;
+ range++;
startpos--;
}
}
/* 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) ((d) == (size1 ? string1 : string2) || !size2)
-#define AT_STRINGS_END(d) ((d) == end2)
+#define AT_STRINGS_END(d) ((d) == end2)
/* Test if D points to a character which is word-constituent. We have
: (d) == string2 - 1 ? *(end1 - 1) : *(d)) \
== Sword)
+/* Disabled due to a compiler bug -- see comment at case wordbound */
+#if 0
/* 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 (d) || AT_STRINGS_END (d) \
|| WORDCHAR_P (d - 1) != WORDCHAR_P (d))
-
+#endif
/* Free everything we malloc. */
#ifdef MATCH_MAY_ALLOCATE
-#ifdef REGEX_MALLOC
-#define FREE_VAR(var) if (var) free (var); var = NULL
+#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() /* Do nothing! */
+#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
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 */
the (virtual) concatenation of STRING1 and STRING2 (of length SIZE1
and SIZE2, respectively). We start matching at POS, and stop
matching at STOP.
-
+
If REGS is non-null and the `no_sub' field of BUFP is nonzero, we
store offsets for the substring each group matched in REGS. See the
documentation for exactly how many groups we fill.
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;
/* Where we are in the data, and the end of the current string. */
const char *d, *dend;
-
+
/* Where we are in the pattern, and the end of the pattern. */
unsigned char *p = bufp->buffer;
register unsigned char *pend = p + bufp->used;
unsigned char *just_past_start_mem = 0;
/* We use this to map every character in the string. */
- char *translate = bufp->translate;
+ RE_TRANSLATE_TYPE translate = bufp->translate;
/* Failure point stack. Each place that can handle a failure further
down the line pushes a failure point on this stack. It consists of
unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0;
#endif
+ /* This holds the pointer to the failure stack, when
+ it is allocated relocatably. */
+ fail_stack_elt_t *failure_stack_ptr;
+
/* We fill all the registers internally, independent of what we
return, for use in backreferences. The number here includes
an element for register zero. */
unsigned num_regs = bufp->re_nsub + 1;
-
+
/* The currently active registers. */
unsigned lowest_active_reg = NO_LOWEST_ACTIVE_REG;
unsigned highest_active_reg = NO_HIGHEST_ACTIVE_REG;
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;
+ 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.
+ 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
else (see below). Also, we never need info about register 0 for
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;
#ifdef DEBUG
/* Counts the total number of registers pushed. */
- unsigned num_regs_pushed = 0;
+ unsigned num_regs_pushed = 0;
#endif
DEBUG_PRINT1 ("\n\nEntering re_match_2.\n");
-
+
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
reg_dummy = REGEX_TALLOC (num_regs, const char *);
reg_info_dummy = REGEX_TALLOC (num_regs, register_info_type);
- if (!(regstart && regend && old_regstart && old_regend && reg_info
- && best_regstart && best_regend && reg_dummy && reg_info_dummy))
+ if (!(regstart && regend && old_regstart && old_regend && reg_info
+ && best_regstart && best_regend && reg_dummy && reg_info_dummy))
{
FREE_VARIABLES ();
return -2;
}
}
-#if defined (REGEX_MALLOC)
else
{
/* We must initialize all our variables to NULL, so that
= 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. */
FREE_VARIABLES ();
return -1;
}
-
+
/* Initialize subexpression text positions to -1 to mark ones that no
start_memory/stop_memory has been seen for. Also initialize the
register information struct. */
for (mcnt = 1; mcnt < num_regs; mcnt++)
{
- regstart[mcnt] = regend[mcnt]
+ regstart[mcnt] = regend[mcnt]
= old_regstart[mcnt] = old_regend[mcnt] = REG_UNSET_VALUE;
-
+
REG_MATCH_NULL_STRING_P (reg_info[mcnt]) = MATCH_NULL_UNSET_VALUE;
IS_ACTIVE (reg_info[mcnt]) = 0;
MATCHED_SOMETHING (reg_info[mcnt]) = 0;
EVER_MATCHED_SOMETHING (reg_info[mcnt]) = 0;
}
-
+
/* We move `string1' into `string2' if the latter's empty -- but not if
`string1' is null. */
if (size2 == 0 && string1 != NULL)
end_match_2 = string2 + stop - size1;
}
- /* `p' scans through the pattern as `d' scans through the data.
+ /* `p' scans through the pattern as `d' scans through the data.
`dend' is the end of the input string that `d' points within. `d'
is advanced into the following input string whenever necessary, but
this happens before fetching; therefore, at the beginning of the
DEBUG_PRINT1 ("The string to match is: `");
DEBUG_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2);
DEBUG_PRINT1 ("'\n");
-
+
/* This loops over pattern commands. It exits by returning from the
function if the match is complete, or it drops through if the match
fails at this starting point in the input data. */
if (p == pend)
{ /* End of pattern means we might have succeeded. */
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;
-
+
DEBUG_PRINT1 ("\nSAVING match as best so far.\n");
-
+
for (mcnt = 1; mcnt < num_regs; mcnt++)
{
best_regstart[mcnt] = regstart[mcnt];
best_regend[mcnt] = regend[mcnt];
}
}
- goto fail;
+ 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 ==
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 */
+ succeed_label:
DEBUG_PRINT1 ("Accepting match.\n");
/* If caller wants register contents data back, do it. */
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
? ((regoff_t) (d - string1))
: ((regoff_t) (d - string2 + size1)));
}
-
+
/* Go through the first `min (num_regs, regs->num_regs)'
registers, since that is all we initialized. */
for (mcnt = 1; mcnt < MIN (num_regs, regs->num_regs); mcnt++)
= (regoff_t) POINTER_TO_OFFSET (regend[mcnt]);
}
}
-
+
/* If the regs structure we return has more elements than
were in the pattern, set the extra elements to -1. If
we (re)allocated the registers, this is the case,
regs->start[mcnt] = regs->end[mcnt] = -1;
} /* regs && !bufp->no_sub */
- FREE_VARIABLES ();
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
+ mcnt = d - pos - (MATCHING_IN_FIRST_STRING
+ ? string1
: string2 - size1);
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
do
{
PREFETCH ();
- if (translate[(unsigned char) *d++] != (char) *p++)
+ if ((unsigned char) translate[(unsigned char) *d++]
+ != (unsigned char) *p++)
goto fail;
}
while (--mcnt);
p += 1 + *p;
if (!not) goto fail;
-
+
SET_REGS_MATCHED ();
d++;
break;
/* Find out if this group can match the empty string. */
p1 = p; /* To send to group_match_null_string_p. */
-
+
if (REG_MATCH_NULL_STRING_P (reg_info[*p]) == MATCH_NULL_UNSET_VALUE)
- REG_MATCH_NULL_STRING_P (reg_info[*p])
+ REG_MATCH_NULL_STRING_P (reg_info[*p])
= group_match_null_string_p (&p1, pend, reg_info);
/* Save the position in the string where we were the last time
old_regstart[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p])
? REG_UNSET (regstart[*p]) ? d : regstart[*p]
: regstart[*p];
- DEBUG_PRINT2 (" old_regstart: %d\n",
+ DEBUG_PRINT2 (" old_regstart: %d\n",
POINTER_TO_OFFSET (old_regstart[*p]));
regstart[*p] = d;
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;
-
+
/* If nothing was active before, this is the new lowest active
register. */
if (lowest_active_reg == NO_LOWEST_ACTIVE_REG)
/* Move past the register number and inner group count. */
p += 2;
just_past_start_mem = p;
+
break;
number, and the number of inner groups. */
case stop_memory:
DEBUG_PRINT3 ("EXECUTING stop_memory %d (%d):\n", *p, p[1]);
-
+
/* We need to save the string position the last time we were at
this close-group operator in case the group is operated
upon by a repetition operator, e.g., with `((a*)*(b*)*)*'
old_regend[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p])
? REG_UNSET (regend[*p]) ? d : regend[*p]
: regend[*p];
- DEBUG_PRINT2 (" old_regend: %d\n",
+ DEBUG_PRINT2 (" old_regend: %d\n",
POINTER_TO_OFFSET (old_regend[*p]));
regend[*p] = d;
/* 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)
unsigned char r = *p - 1;
while (r > 0 && !IS_ACTIVE (reg_info[r]))
r--;
-
+
/* If we end up at register zero, that means that we saved
the registers as the result of an `on_failure_jump', not
a `start_memory', and we jumped to past the innermost
else
highest_active_reg = r;
}
-
+
/* 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
last match. */
if ((!MATCHED_SOMETHING (reg_info[*p])
|| just_past_start_mem == p - 1)
- && (p + 2) < pend)
+ && (p + 2) < pend)
{
boolean is_a_jump_n = false;
-
+
p1 = p + 2;
mcnt = 0;
switch ((re_opcode_t) *p1++)
if (is_a_jump_n)
p1 += 2;
break;
-
+
default:
/* do nothing */ ;
}
p1 += mcnt;
-
+
/* If the next operation is a jump backwards in the pattern
to an on_failure_jump right before the start_memory
corresponding to this stop_memory, exit from the loop
failed match, e.g., with `(a*)*b' against `ab' for
regstart[1], and, e.g., with `((a*)*(b*)*)*'
against `aba' for regend[3].
-
+
Also restore the registers for inner groups for,
e.g., `((a*)(b*))*' against `aba' (register 3 would
otherwise get trashed). */
-
+
if (EVER_MATCHED_SOMETHING (reg_info[*p]))
{
- unsigned r;
-
+ unsigned r;
+
EVER_MATCHED_SOMETHING (reg_info[*p]) = 0;
-
+
/* Restore this and inner groups' (if any) registers. */
for (r = *p; r < *p + *(p + 1); r++)
{
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];
- }
+ }
}
p1++;
EXTRACT_NUMBER_AND_INCR (mcnt, p1);
goto fail;
}
}
-
+
/* Move past the register number and the inner group count. */
p += 2;
break;
/* Can't back reference a group which we've never matched. */
if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno]))
goto fail;
-
+
/* Where in input to try to start matching. */
d2 = regstart[regno];
-
+
/* Where to stop matching; if both the place to start and
the place to stop matching are in the same string, then
set to the place to stop, otherwise, for now have to use
the end of the first string. */
- dend2 = ((FIRST_STRING_P (regstart[regno])
+ dend2 = ((FIRST_STRING_P (regstart[regno])
== FIRST_STRING_P (regend[regno]))
? regend[regno] : end_match_1);
for (;;)
/* How many characters left in this segment to match. */
mcnt = dend - d;
-
+
/* Want how many consecutive characters we can match in
one shot, so, if necessary, adjust the count. */
if (mcnt > dend2 - d2)
mcnt = dend2 - d2;
-
+
/* Compare that many; failure if mismatch, else move
past them. */
- if (translate
- ? bcmp_translate (d, d2, mcnt, translate)
+ if (translate
+ ? bcmp_translate (d, d2, mcnt, translate)
: bcmp (d, d2, mcnt))
goto fail;
d += mcnt, d2 += mcnt;
+
+ /* Do this because we've match some characters. */
+ SET_REGS_MATCHED ();
}
}
break;
`newline_anchor' is set, after newlines. */
case begline:
DEBUG_PRINT1 ("EXECUTING begline.\n");
-
+
if (AT_STRINGS_BEG (d))
{
if (!bufp->not_bol) break;
{
if (!bufp->not_eol) break;
}
-
+
/* We have to ``prefetch'' the next character. */
else if ((d == end1 ? *string2 : *d) == '\n'
&& bufp->newline_anchor)
then the . fails against the \n. But the next thing we want
to do is match the \n against the \n; if we restored the
string value, we would be back at the foo.
-
+
Because this is used only in specific cases, we don't need to
check all the things that `on_failure_jump' does, to make
sure the right things get saved on the stack. Hence we don't
case; that seems worse than this. */
case on_failure_keep_string_jump:
DEBUG_PRINT1 ("EXECUTING on_failure_keep_string_jump");
-
+
EXTRACT_NUMBER_AND_INCR (mcnt, p);
DEBUG_PRINT3 (" %d (to 0x%x):\n", mcnt, p + mcnt);
/* Uses of on_failure_jump:
-
+
Each alternative starts with an on_failure_jump that points
to the beginning of the next alternative. Each alternative
except the last ends with a jump that in effect jumps past
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 need the preceding group,
- and in \(\(a*\)b*\)\2, we need the inner group. */
+ and in \(zz\(a*\)b*\)\2, we need the inner group. */
/* We can't use `p' to check ahead because we push
a failure point to `p + mcnt' after we do this. */
would have to backtrack because of (as in, e.g., `a*a')
then we can change to pop_failure_jump, because we'll
never have to backtrack.
-
+
This is not true in the case of alternatives: in
`(a|ab)*' we do need to backtrack to the `ab' alternative
(e.g., if the string was `ab'). But instead of trying to
p1 = p + mcnt;
/* p1[0] ... p1[2] are the `on_failure_jump' corresponding
- to the `maybe_finalize_jump' of this case. Examine what
+ to the `maybe_finalize_jump' of this case. Examine what
follows. */
/* If we're at the end of the pattern, we can change. */
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)
{
int not = (re_opcode_t) p1[3] == charset_not;
-
+
if (c < (unsigned char) (p1[4] * BYTEWIDTH)
&& p1[5 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
not = !not;
}
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
- && ! (p2[1] * BYTEWIDTH > p1[4]
- && (p2[1 + p1[4] / BYTEWIDTH]
- & (1 << (p1[4] % BYTEWIDTH)))))
+ && ! ((int) p2[1] * BYTEWIDTH > (int) p1[5]
+ && (p2[2 + p1[5] / BYTEWIDTH]
+ & (1 << (p1[5] % 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 < p2[1]; idx++)
+ for (idx = 0; idx < (int) p2[1]; idx++)
if (! (p2[2 + idx] == 0
- || (idx < p1[4]
+ || (idx < (int) p1[4]
&& ((p2[2 + idx] & ~ p1[5 + idx]) == 0))))
break;
int idx;
/* We win if the charset inside the loop
has no overlap with the one after the loop. */
- for (idx = 0; idx < p2[1] && idx < p1[4]; idx++)
+ for (idx = 0;
+ idx < (int) p2[1] && idx < (int) p1[4];
+ idx++)
if ((p2[2 + idx] & p1[5 + idx]) != 0)
break;
}
/* Note fall through. */
-
+
/* Unconditionally jump (without popping any failure points). */
case jump:
unconditional_jump:
DEBUG_PRINT2 ("(to 0x%x).\n", p);
break;
-
+
/* We need this opcode so we can detect where alternatives end
in `group_match_null_string_p' et al. */
case jump_past_alt:
/* Have to succeed matching what follows at least n times.
After that, handle like `on_failure_jump'. */
- case succeed_n:
+ case succeed_n:
EXTRACT_NUMBER (mcnt, p + 2);
DEBUG_PRINT2 ("EXECUTING succeed_n %d.\n", mcnt);
goto on_failure;
}
break;
-
- case jump_n:
+
+ case jump_n:
EXTRACT_NUMBER (mcnt, p + 2);
DEBUG_PRINT2 ("EXECUTING jump_n %d.\n", mcnt);
{
mcnt--;
STORE_NUMBER (p + 2, mcnt);
- goto unconditional_jump;
+ goto unconditional_jump;
}
/* If don't have to jump any more, skip over the rest of command. */
- else
- p += 4;
+ else
+ p += 4;
break;
-
+
case set_number_at:
{
DEBUG_PRINT1 ("EXECUTING set_number_at.\n");
break;
}
- case wordbound:
- DEBUG_PRINT1 ("EXECUTING wordbound.\n");
- if (AT_WORD_BOUNDARY (d))
+#if 0
+ /* The DEC Alpha C compiler 3.x generates incorrect code for the
+ test WORDCHAR_P (d - 1) != WORDCHAR_P (d) in the expansion of
+ AT_WORD_BOUNDARY, so this code is disabled. Expanding the
+ macro and introducing temporary variables works around the bug. */
+
+ case wordbound:
+ DEBUG_PRINT1 ("EXECUTING wordbound.\n");
+ if (AT_WORD_BOUNDARY (d))
break;
- goto fail;
+ goto fail;
case notwordbound:
- DEBUG_PRINT1 ("EXECUTING notwordbound.\n");
+ DEBUG_PRINT1 ("EXECUTING notwordbound.\n");
if (AT_WORD_BOUNDARY (d))
goto fail;
- break;
+ break;
+#else
+ case wordbound:
+ {
+ boolean prevchar, thischar;
+
+ DEBUG_PRINT1 ("EXECUTING wordbound.\n");
+ if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d))
+ break;
+
+ prevchar = WORDCHAR_P (d - 1);
+ thischar = WORDCHAR_P (d);
+ if (prevchar != thischar)
+ break;
+ goto fail;
+ }
+
+ case notwordbound:
+ {
+ boolean prevchar, thischar;
+
+ DEBUG_PRINT1 ("EXECUTING notwordbound.\n");
+ if (AT_STRINGS_BEG (d) || AT_STRINGS_END (d))
+ goto fail;
+
+ prevchar = WORDCHAR_P (d - 1);
+ thischar = WORDCHAR_P (d);
+ if (prevchar != thischar)
+ goto fail;
+ break;
+ }
+#endif
case wordbeg:
DEBUG_PRINT1 ("EXECUTING wordbeg.\n");
goto fail;
#ifdef emacs
-#ifdef emacs19
case before_dot:
DEBUG_PRINT1 ("EXECUTING before_dot.\n");
if (PTR_CHAR_POS ((unsigned char *) d) >= point)
goto fail;
break;
-
+
case at_dot:
DEBUG_PRINT1 ("EXECUTING at_dot.\n");
if (PTR_CHAR_POS ((unsigned char *) d) != point)
goto fail;
break;
-
+
case after_dot:
DEBUG_PRINT1 ("EXECUTING after_dot.\n");
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);
SET_REGS_MATCHED ();
d++;
break;
-
+
case notwordchar:
DEBUG_PRINT1 ("EXECUTING non-Emacs notwordchar.\n");
PREFETCH ();
d++;
break;
#endif /* not emacs */
-
+
default:
abort ();
}
if (p < pend)
{
boolean is_a_jump_n = false;
-
+
/* If failed to a backwards jump that's part of a repetition
loop, need to pop this failure point and use the next one. */
switch ((re_opcode_t) *p)
case jump:
p1 = p + 1;
EXTRACT_NUMBER_AND_INCR (mcnt, p1);
- p1 += mcnt;
+ p1 += mcnt;
if ((is_a_jump_n && (re_opcode_t) *p1 == succeed_n)
|| (!is_a_jump_n
/* We are passed P pointing to a register number after a start_memory.
-
+
Return true if the pattern up to the corresponding stop_memory can
match the empty string, and false otherwise.
-
+
If we find the matching stop_memory, sets P to point to one past its number.
Otherwise, sets P to an undefined byte less than or equal to END.
int mcnt;
/* Point to after the args to the start_memory. */
unsigned char *p1 = *p + 2;
-
+
while (p1 < end)
{
/* Skip over opcodes that can match nothing, and return true or
false, as appropriate, when we get to one that can't, or to the
matching stop_memory. */
-
+
switch ((re_opcode_t) *p1)
{
/* Could be either a loop or a series of alternatives. */
case on_failure_jump:
p1++;
EXTRACT_NUMBER_AND_INCR (mcnt, p1);
-
+
/* If the next operation is not a jump backwards in the
pattern. */
/on_failure_jump/0/6/exactn/1/a/jump_past_alt/0/6
/on_failure_jump/0/6/exactn/1/b/jump_past_alt/0/3
- /exactn/1/c
+ /exactn/1/c
So, we have to first go through the first (n-1)
alternatives and then deal with the last one separately. */
is, including the ending `jump_past_alt' and
its number. */
- if (!alt_match_null_string_p (p1, p1 + mcnt - 3,
+ if (!alt_match_null_string_p (p1, p1 + mcnt - 3,
reg_info))
return false;
/* Move to right after this alternative, including the
jump_past_alt. */
- p1 += mcnt;
+ p1 += mcnt;
/* Break if it's the beginning of an n-th alternative
that doesn't begin with an on_failure_jump. */
if ((re_opcode_t) *p1 != on_failure_jump)
break;
-
+
/* Still have to check that it's not an n-th
alternative that starts with an on_failure_jump. */
p1++;
} /* if mcnt > 0 */
break;
-
+
case stop_memory:
assert (p1[1] == **p);
*p = p1 + 2;
return true;
-
- default:
+
+ default:
if (!common_op_match_null_string_p (&p1, end, reg_info))
return false;
}
/* Similar to group_match_null_string_p, but doesn't deal with alternatives:
It expects P to be the first byte of a single alternative and END one
byte past the last. The alternative can contain groups. */
-
+
static boolean
alt_match_null_string_p (p, end, reg_info)
unsigned char *p, *end;
{
int mcnt;
unsigned char *p1 = p;
-
+
while (p1 < end)
{
- /* Skip over opcodes that can match nothing, and break when we get
+ /* Skip over opcodes that can match nothing, and break when we get
to one that can't. */
-
+
switch ((re_opcode_t) *p1)
{
/* It's a loop. */
EXTRACT_NUMBER_AND_INCR (mcnt, p1);
p1 += mcnt;
break;
-
- default:
+
+ default:
if (!common_op_match_null_string_p (&p1, end, reg_info))
return false;
}
/* Deals with the ops common to group_match_null_string_p and
- alt_match_null_string_p.
-
+ alt_match_null_string_p.
+
Sets P to one after the op and its arguments, if any. */
static boolean
reg_no = *p1;
assert (reg_no > 0 && reg_no <= MAX_REGNUM);
ret = group_match_null_string_p (&p1, end, reg_info);
-
+
/* Have to set this here in case we're checking a group which
contains a group and a back reference to it. */
if (!ret)
return false;
break;
-
+
/* If this is an optimized succeed_n for zero times, make the jump. */
case jump:
EXTRACT_NUMBER_AND_INCR (mcnt, p1);
case succeed_n:
/* Get to the number of times to succeed. */
- p1 += 2;
+ p1 += 2;
EXTRACT_NUMBER_AND_INCR (mcnt, p1);
if (mcnt == 0)
return false;
break;
- case duplicate:
+ case duplicate:
if (!REG_MATCH_NULL_STRING_P (reg_info[*p1]))
return false;
break;
/* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN
bytes; nonzero otherwise. */
-
+
static int
bcmp_translate (s1, s2, len, translate)
unsigned char *s1, *s2;
register int len;
- char *translate;
+ RE_TRANSLATE_TYPE translate;
{
register unsigned char *p1 = s1, *p2 = s2;
while (len)
/* re_compile_pattern is the GNU regular expression compiler: it
compiles PATTERN (of length SIZE) and puts the result in BUFP.
Returns 0 if the pattern was valid, otherwise an error string.
-
+
Assumes the `allocated' (and perhaps `buffer') and `translate' fields
are set in BUFP on entry.
-
+
We call regex_compile to do the actual compilation. */
const char *
struct re_pattern_buffer *bufp;
{
reg_errcode_t ret;
-
+
/* GNU code is written to assume at least RE_NREGS registers will be set
(and at least one extra will be -1). */
bufp->regs_allocated = REGS_UNALLOCATED;
-
+
/* And GNU code determines whether or not to get register information
by passing null for the REGS argument to re_match, etc., not by
setting no_sub. */
bufp->no_sub = 0;
-
+
/* Match anchors at newline. */
bufp->newline_anchor = 1;
-
+
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;
const char *s;
{
reg_errcode_t ret;
-
+
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
re_comp_buf.newline_anchor = 1;
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. */
int
regcomp (preg, pattern, cflags)
regex_t *preg;
- const char *pattern;
+ const char *pattern;
int cflags;
{
reg_errcode_t ret;
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
characters after newlines into the fastmap. This way, we just try
every character. */
preg->fastmap = 0;
-
+
if (cflags & REG_ICASE)
{
unsigned i;
-
- preg->translate = (char *) malloc (CHAR_SET_SIZE);
+
+ preg->translate
+ = (RE_TRANSLATE_TYPE) malloc (CHAR_SET_SIZE
+ * sizeof (*(RE_TRANSLATE_TYPE)0));
if (preg->translate == NULL)
return (int) REG_ESPACE;
preg->no_sub = !!(cflags & REG_NOSUB);
- /* POSIX says a null character in the pattern terminates it, so we
+ /* POSIX says a null character in the pattern terminates it, so we
can use strlen here in compiling the pattern. */
ret = regex_compile (pattern, strlen (pattern), syntax, preg);
-
+
/* POSIX doesn't distinguish between an unmatched open-group and an
unmatched close-group: both are REG_EPAREN. */
if (ret == REG_ERPAREN) ret = REG_EPAREN;
-
+
return (int) ret;
}
/* regexec searches for a given pattern, specified by PREG, in the
string STRING.
-
+
If NMATCH is zero or REG_NOSUB was set in the cflags argument to
`regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at
least NMATCH elements, and we set them to the offsets of the
corresponding matched substrings.
-
+
EFLAGS specifies `execution flags' which affect matching: if
REG_NOTBOL is set, then ^ does not match at the beginning of the
string; if REG_NOTEOL is set, then $ does not match at the end.
-
+
We return 0 if we find a match and REG_NOMATCH if not. */
int
regexec (preg, string, nmatch, pmatch, eflags)
const regex_t *preg;
- const char *string;
- size_t nmatch;
- regmatch_t pmatch[];
+ const char *string;
+ size_t nmatch;
+ regmatch_t pmatch[];
int eflags;
{
int ret;
boolean want_reg_info = !preg->no_sub && nmatch > 0;
private_preg = *preg;
-
+
private_preg.not_bol = !!(eflags & REG_NOTBOL);
private_preg.not_eol = !!(eflags & REG_NOTEOL);
-
+
/* The user has told us exactly how many registers to return
information about, via `nmatch'. We have to pass that on to the
matching routines. */
private_preg.regs_allocated = REGS_FIXED;
-
+
if (want_reg_info)
{
regs.num_regs = nmatch;
ret = re_search (&private_preg, string, len,
/* start: */ 0, /* range: */ len,
want_reg_info ? ®s : (struct re_registers *) 0);
-
+
/* Copy the register information to the POSIX structure. */
if (want_reg_info)
{
size_t msg_size;
if (errcode < 0
- || errcode >= (sizeof (re_error_msg) / sizeof (re_error_msg[0])))
- /* Only error codes returned by the rest of the code should be passed
+ || 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 = re_error_msg[errcode];
-
- /* POSIX doesn't require that we do anything in this case, but why
- not be nice. */
- if (! msg)
- msg = "Success";
+ msg = gettext (re_error_msgid[errcode]);
msg_size = strlen (msg) + 1; /* Includes the null. */
-
+
if (errbuf_size != 0)
{
if (msg_size > errbuf_size)
if (preg->buffer != NULL)
free (preg->buffer);
preg->buffer = NULL;
-
+
preg->allocated = 0;
preg->used = 0;
}
#endif /* not emacs */
-\f
-/*
-Local variables:
-make-backup-files: t
-version-control: t
-trim-versions-without-asking: nil
-End:
-*/
projects / gnulib.git / blobdiff