#define RE_STRING_CHAR_AND_LENGTH STRING_CHAR_AND_LENGTH
#define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \
(c = ((p) == (str2) ? *((end1) - 1) : *((p) - 1)))
+#define MAKE_CHAR(charset, c1, c2) (c1)
#endif /* not emacs */
#ifndef RE_TRANSLATE
/* Fetch the character which ends the range. */
PATFETCH (c1);
- if (SINGLE_BYTE_CHAR_P (c)
- && ! SINGLE_BYTE_CHAR_P (c1))
+ if (SINGLE_BYTE_CHAR_P (c))
{
- /* Handle a range such as \177-\377 in multibyte mode.
- Split that into two ranges,,
- the low one ending at 0237, and the high one
- starting at ...040. */
- /* Unless I'm missing something,
- this line is useless. -sm
- int c1_base = (c1 & ~0177) | 040; */
- SET_RANGE_TABLE_WORK_AREA (range_table_work, c, c1);
- c1 = 0237;
+ if (! SINGLE_BYTE_CHAR_P (c1))
+ {
+ /* Handle a range such as \177-\377 in
+ multibyte mode. Split that into two
+ ranges, the low one ending at 0237, and
+ the high one starting at the smallest
+ character in the charset of C1 and
+ ending at C1. */
+ int charset = CHAR_CHARSET (c1);
+ int c2 = MAKE_CHAR (charset, 0, 0);
+
+ SET_RANGE_TABLE_WORK_AREA (range_table_work,
+ c2, c1);
+ c1 = 0237;
+ }
}
else if (!SAME_CHARSET_P (c, c1))
FREE_STACK_RETURN (REG_ERANGE);
/* After a subexpression? */
(*prev == '(' && (syntax & RE_NO_BK_PARENS || prev_prev_backslash))
/* After an alternative? */
- || (*prev == '|' && (syntax & RE_NO_BK_VBAR || prev_prev_backslash));
+ || (*prev == '|' && (syntax & RE_NO_BK_VBAR || prev_prev_backslash))
+ /* After a shy subexpression? */
+ || ((syntax & RE_SHY_GROUPS) && prev - 2 >= pattern
+ && prev[-1] == '?' && prev[-2] == '('
+ && (syntax & RE_NO_BK_PARENS
+ || (prev - 3 >= pattern && prev[-3] == '\\')));
}
: ((regoff_t) ((ptr) - string2 + size1)))
/* Call before fetching a character with *d. This switches over to
- string2 if necessary. */
+ string2 if necessary.
+ Check re_match_2_internal for a discussion of why end_match_2 might
+ not be within string2 (but be equal to end_match_1 instead). */
#define PREFETCH() \
while (d == dend) \
{ \
dend = end_match_2; \
}
+/* Call before fetching a char with *d if you already checked other limits.
+ This is meant for use in lookahead operations like wordend, etc..
+ where we might need to look at parts of the string that might be
+ outside of the LIMITs (i.e past `stop'). */
+#define PREFETCH_NOLIMIT() \
+ if (d == end1) \
+ { \
+ d = string2; \
+ dend = end_match_2; \
+ } \
/* Test if at very beginning or at very end of the virtual concatenation
of `string1' and `string2'. If only one string, it's `string2'. */
{
int result = re_match_2_internal (bufp, NULL, 0, string, size,
pos, regs, size);
+#if defined (C_ALLOCA) && !defined (REGEX_MALLOC)
alloca (0);
+#endif
return result;
}
#endif /* not emacs */
result = re_match_2_internal (bufp, string1, size1, string2, size2,
pos, regs, stop);
+#if defined (C_ALLOCA) && !defined (REGEX_MALLOC)
alloca (0);
+#endif
return result;
}
for (mcnt = 1; mcnt < num_regs; mcnt++)
regstart[mcnt] = regend[mcnt] = REG_UNSET_VALUE;
- /* Shorten strings to `stop'. */
- if (stop <= size1)
- {
- size1 = stop;
- size2 = 0;
- }
- else if (stop <= size1 + size2)
- size2 = stop - size1;
-
/* We move `string1' into `string2' if the latter's empty -- but not if
`string1' is null. */
if (size2 == 0 && string1 != NULL)
end1 = string1 + size1;
end2 = string2 + size2;
- /* Compute where to stop matching, within the two strings. */
- end_match_1 = end1;
- end_match_2 = end2;
-
/* `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
loop, `d' can be pointing at the end of a string, but it cannot
equal `string2'. */
- if (size1 > 0 && pos <= size1)
+ if (pos >= size1)
{
- d = string1 + pos;
- dend = end_match_1;
+ /* Only match within string2. */
+ d = string2 + pos - size1;
+ dend = end_match_2 = string2 + stop - size1;
+ end_match_1 = end1; /* Just to give it a value. */
}
else
{
- d = string2 + pos - size1;
- dend = end_match_2;
+ if (stop < size1)
+ {
+ /* Only match within string1. */
+ end_match_1 = string1 + stop;
+ /* BEWARE!
+ When we reach end_match_1, PREFETCH normally switches to string2.
+ But in the present case, this means that just doing a PREFETCH
+ makes us jump from `stop' to `gap' within the string.
+ What we really want here is for the search to stop as
+ soon as we hit end_match_1. That's why we set end_match_2
+ to end_match_1 (since PREFETCH fails as soon as we hit
+ end_match_2). */
+ end_match_2 = end_match_1;
+ }
+ else
+ { /* It's important to use this code when stop == size so that
+ moving `d' from end1 to string2 will not prevent the d == dend
+ check from catching the end of string. */
+ end_match_1 = end1;
+ end_match_2 = string2 + stop - size1;
+ }
+ d = string1 + pos;
+ dend = end_match_1;
}
DEBUG_PRINT1 ("The compiled pattern is: ");
{
if (!bufp->not_bol) break;
}
- else if (d[-1] == '\n' && bufp->newline_anchor)
+ else
{
- break;
+ unsigned char c;
+ GET_CHAR_BEFORE_2 (c, d, string1, end1, string2, end2);
+ if (c == '\n' && bufp->newline_anchor)
+ break;
}
/* In all other cases, we fail. */
goto fail;
{
if (!bufp->not_eol) break;
}
-
- /* We have to ``prefetch'' the next character. */
- else if ((d == end1 ? *string2 : *d) == '\n'
- && bufp->newline_anchor)
+ else
{
- break;
+ PREFETCH_NOLIMIT ();
+ if (*d == '\n' && bufp->newline_anchor)
+ break;
}
goto fail;
#ifdef emacs
UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1);
#endif
- PREFETCH ();
+ PREFETCH_NOLIMIT ();
c2 = RE_STRING_CHAR (d, dend - d);
s2 = SYNTAX (c2);
/* Case 3: D is not at the end of string ... */
if (!AT_STRINGS_END (d))
{
- PREFETCH ();
+ PREFETCH_NOLIMIT ();
c2 = RE_STRING_CHAR (d, dend - d);
#ifdef emacs
UPDATE_SYNTAX_TABLE_FORWARD (charpos);
if (!s)
{
if (!re_comp_buf.buffer)
- return gettext ("No previous regular expression");
+ /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */
+ return (char *) gettext ("No previous regular expression");
return 0;
}
{
re_comp_buf.buffer = (unsigned char *) malloc (200);
if (re_comp_buf.buffer == NULL)
- return gettext (re_error_msgid[(int) REG_ESPACE]);
+ /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */
+ return (char *) 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 gettext (re_error_msgid[(int) REG_ESPACE]);
+ /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */
+ return (char *) gettext (re_error_msgid[(int) REG_ESPACE]);
}
/* Since `re_exec' always passes NULL for the `regs' argument, we