/* We need this for `regex.h', and perhaps for the Emacs include files. */
#include <sys/types.h>
+/* This is for other GNU distributions with internationalized messages.
+ 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
that make sense only in Emacs. */
#ifdef emacs
#include "buffer.h"
#include "syntax.h"
-/* Emacs uses `NULL' as a predicate. */
-#undef NULL
-
#else /* not emacs */
#ifdef STDC_HEADERS
#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;
/* 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 (']');
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 */
#define MATCH_MAY_ALLOCATE
/* The match routines may not allocate if (1) they would do it with malloc
- and (2) it's not safe for htem to use 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
DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \
} \
\
+ set_regs_matched_done = 0; \
DEBUG_STATEMENT (nfailure_points_popped++); \
} /* POP_FAILURE_POINT */
/* Call this when have matched a real character; it sets `matched' flags
for the subexpressions which we are currently inside. Also records
that those subexprs have matched. */
-#define SET_REGS_MATCHED() \
- do \
- { \
- unsigned r; \
- for (r = lowest_active_reg; r <= highest_active_reg; r++) \
- { \
- MATCHED_SOMETHING (reg_info[r]) \
- = EVER_MATCHED_SOMETHING (reg_info[r]) \
- = 1; \
- } \
- } \
- while (0)
+#define SET_REGS_MATCHED() \
+ 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; \
+ } \
+ } \
+ else
/* 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)
unsigned debug_count;
for (debug_count = 0; debug_count < size; debug_count++)
- printchar (pattern[debug_count]);
+ putchar (pattern[debug_count]);
putchar ('\n');
}
#endif /* DEBUG */
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);
/* We have succeeded; set the length of the buffer. */
We also want to fetch the endpoints without translating them; the
appropriate translation is done in the bit-setting loop below. */
+ /* The SVR4 compiler on the 3B2 had trouble with unsigned const char *. */
range_start = ((const unsigned char *) p)[-2];
range_end = ((const unsigned char *) p)[0];
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
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;
}
} /* d != end_match_2 */
+ succeed:
DEBUG_PRINT1 ("Accepting match.\n");
/* If caller wants register contents data back, do it. */
}
/* 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)
: bcmp (d, d2, mcnt))
goto fail;
d += mcnt, d2 += mcnt;
+
+ /* Do this because we've match some characters. */
+ 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