(Implements POSIX draft P10003.2/D11.2, except for
internationalization features.)
- Copyright (C) 1993 Free Software Foundation, Inc.
+ Copyright (C) 1993, 1994 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>
-#ifdef HAVE_CONFIG_H
-#include "config.h"
+/* This is for other GNU distributions with internationalized messages.
+ The GNU C Library itself does not yet support such messages. */
+#if HAVE_LIBINTL_H
+# include <libintl.h>
+#else
+# define gettext(msgid) (msgid)
#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
+#include <stdlib.h>
+#else
+char *malloc ();
+char *realloc ();
+#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. */
+#ifndef INHIBIT_STRING_HEADER
#if HAVE_STRING_H || STDC_HEADERS
#include <string.h>
#ifndef bcmp
#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;
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 (isascii) || defined (STDC_HEADERS)
-#undef isascii
-#define isascii(c) 1
+
+#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))
+#define ISBLANK(c) (ISASCII (c) && isblank (c))
#else
#define ISBLANK(c) ((c) == ' ' || (c) == '\t')
#endif
#ifdef isgraph
-#define ISGRAPH(c) (isascii (c) && isgraph (c))
+#define ISGRAPH(c) (ISASCII (c) && isgraph (c))
#else
-#define ISGRAPH(c) (isascii (c) && isprint (c) && !isspace (c))
+#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))
+#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 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,
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);
}
}
}
do
{
putchar ('/');
- printchar (*p++);
+ putchar (*p++);
}
while (--mcnt);
break;
/* 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 (']');
unsigned char *buffer = bufp->buffer;
print_partial_compiled_pattern (buffer, buffer + bufp->used);
- printf ("%d bytes used/%d bytes allocated.\n", bufp->used, bufp->allocated);
+ printf ("%ld bytes used/%ld bytes allocated.\n", bufp->used, bufp->allocated);
if (bufp->fastmap_accurate && bufp->fastmap)
{
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]);
}
}
/* 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
}
\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. */
+ 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_msg[] =
- { NULL, /* REG_NOERROR */
+static const char *re_error_msgid[] =
+ { "Success", /* REG_NOERROR */
"No match", /* REG_NOMATCH */
"Invalid regular expression", /* REG_BADPAT */
"Invalid collation character", /* REG_ECOLLATE */
\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
relocation, which might (if the strings being searched are in the
ralloc heap) shift the data out from underneath the regexp
- routines. */
+ 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
-/* But under some circumstances, it's not. */
-#if defined (REL_ALLOC) && defined (C_ALLOCA)
+/* 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. */
+#if (defined (C_ALLOCA) || defined (REGEX_MALLOC)) && (defined (emacs) || defined (REL_ALLOC))
#undef MATCH_MAY_ALLOCATE
#endif
change it ourselves. */
int re_max_failures = 2000;
-typedef const unsigned char *fail_stack_elt_t;
+typedef unsigned char *fail_stack_elt_t;
typedef struct
{
DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \
} \
\
+ set_regs_matched_done = 0; \
DEBUG_STATEMENT (nfailure_points_popped++); \
} /* POP_FAILURE_POINT */
#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)
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 == '?'))
{
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++)
{
+ /* 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;
}
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
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''. */
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);
/* 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 *) realloc (fail_stack.stack,
- (fail_stack.size
- * sizeof (fail_stack_elt_t)));
- else
- fail_stack.stack =
- (fail_stack_elt_t *) malloc (fail_stack.size
- * sizeof (fail_stack_elt_t));
+ 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 */
+ }
/* Initialize some other variables the matcher uses. */
RETALLOC_IF (regstart, num_regs, const char *);
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. */
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;
/* Assume that each path through the pattern can be null until
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];
+
+ 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
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)
+ return 0;
+ /* Otherwise, have to check alternative paths. */
+ break;
+ }
#ifdef emacs
case syntaxspec:
{
bufp->regs_allocated = REGS_UNALLOCATED;
regs->num_regs = 0;
- regs->start = regs->end = (regoff_t) 0;
+ regs->start = regs->end = (regoff_t *) 0;
}
}
\f
&& !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;
/* 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)
+#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. */
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)
+/* This used to do alloca (0), but now we do that in the caller. */
+#define FREE_VARIABLES() /* Nothing */
#endif /* not REGEX_MALLOC */
#else
#define FREE_VARIABLES() /* Do nothing! */
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;
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;
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 ==
}
} /* d != end_match_2 */
+ succeed:
DEBUG_PRINT1 ("Accepting match.\n");
/* If caller wants register contents data back, do it. */
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]);
}
}
}
/* 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;
/* 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)
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;
: bcmp (d, d2, mcnt))
goto fail;
d += mcnt, d2 += mcnt;
+
+ /* Do this because we've match some characters. */
+ SET_REGS_MATCHED ();
}
}
break;
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)
{
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;
}
}
}
+ 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");
+ }
+ }
+ }
}
p -= 2; /* Point at relative address again. */
if ((re_opcode_t) p[-1] != pop_failure_jump)
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 */
+#if 0 /* not emacs19 */
case at_dot:
DEBUG_PRINT1 ("EXECUTING at_dot.\n");
if (PTR_CHAR_POS ((unsigned char *) d) + 1 != point)
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;
mcnt = (int) Sword;
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;
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. */
size_t msg_size;
if (errcode < 0
- || errcode >= (sizeof (re_error_msg) / sizeof (re_error_msg[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 = 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. */
projects / gnulib.git / blobdiff