代理服务器

开发平台：

Visual C++

gnu_regex.c：源码内容

DEBUG_PRINT2 (" size: %dn", fail_stack.size);
assert (fail_stack.avail >= NUM_NONREG_ITEMS);
DEBUG_POP (&failure_id);
DEBUG_PRINT2 (" Popping failure id: %un", failure_id);
/* If the saved string location is NULL, it came from an
on_failure_keep_string_jump opcode, and we want to throw away the
saved NULL, thus retaining our current position in the string. */
string_temp = POP_FAILURE_ITEM ();
if (string_temp != NULL)
str = (const char *) string_temp;
DEBUG_PRINT2 (" Popping string 0x%x: `", str);
DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2);
DEBUG_PRINT1 ("'n");
pat = (unsigned char *) POP_FAILURE_ITEM ();
DEBUG_PRINT2 (" Popping pattern 0x%x: ", pat);
DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend);
/* Restore register info. */
high_reg = (unsigned) POP_FAILURE_ITEM ();
DEBUG_PRINT2 (" Popping high active reg: %dn", high_reg);
low_reg = (unsigned) POP_FAILURE_ITEM ();
DEBUG_PRINT2 (" Popping low active reg: %dn", low_reg);
for (this_reg = high_reg; this_reg >= low_reg; this_reg--)
{
DEBUG_PRINT2 (" Popping reg: %dn", this_reg);
reg_info[this_reg].word = POP_FAILURE_ITEM ();
DEBUG_PRINT2 (" info: 0x%xn", reg_info[this_reg]);
regend[this_reg] = (const char *) POP_FAILURE_ITEM ();
DEBUG_PRINT2 (" end: 0x%xn", regend[this_reg]);
regstart[this_reg] = (const char *) POP_FAILURE_ITEM ();
DEBUG_PRINT2 (" start: 0x%xn", regstart[this_reg]);
}
DEBUG_STATEMENT (nfailure_points_popped++);
} /* POP_FAILURE_POINT */
/* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in
BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible
characters can start a string that matches the pattern. This fastmap
is used by re_search to skip quickly over impossible starting points.
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.
Returns 0 if we succeed, -2 if an internal error. */
int
re_compile_fastmap (bufp)
struct re_pattern_buffer *bufp;
{
int j, k;
fail_stack_type fail_stack;
#ifndef REGEX_MALLOC
char *destination;
#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;
const unsigned char *p = pattern;
register unsigned char *pend = pattern + size;
/* 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
match the empty string. */
boolean path_can_be_null = true;
/* We aren't doing a `succeed_n' to begin with. */
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 ())
{
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];
}
/* 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
{
/* I guess the idea here is to simply not bother with a fastmap
if a backreference is used, since it's too hard to figure out
the fastmap for the corresponding group. Setting
`can_be_null' stops `re_search_2' from using the fastmap, so
that is all we do. */
case duplicate:
bufp->can_be_null = 1;
return 0;
/* Following are the cases which match a character. These end
with `break'. */
case exactn:
fastmap[p[1]] = 1;
break;
case charset:
for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--)
if (p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH)))
fastmap[j] = 1;
break;
case charset_not:
/* Chars beyond end of map must be allowed. */
for (j = *p * BYTEWIDTH; j < (1 << BYTEWIDTH); j++)
fastmap[j] = 1;
for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--)
if (!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))))
fastmap[j] = 1;
break;
case wordchar:
for (j = 0; j < (1 << BYTEWIDTH); j++)
if (SYNTAX (j) == Sword)
fastmap[j] = 1;
break;
case notwordchar:
for (j = 0; j < (1 << BYTEWIDTH); j++)
if (SYNTAX (j) != Sword)
fastmap[j] = 1;
break;
case anychar:
/* `.' matches anything ... */
for (j = 0; j < (1 << BYTEWIDTH); j++)
fastmap[j] = 1;
/* ... except perhaps newline. */
if (!(bufp->syntax & RE_DOT_NEWLINE))
fastmap['n'] = 0;
/* 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:
k = *p++;
for (j = 0; j < (1 << BYTEWIDTH); j++)
if (SYNTAX (j) == (enum syntaxcode) k)
fastmap[j] = 1;
break;
case notsyntaxspec:
k = *p++;
for (j = 0; j < (1 << BYTEWIDTH); j++)
if (SYNTAX (j) != (enum syntaxcode) k)
fastmap[j] = 1;
break;
/* All cases after this match the empty string. These end with
`continue'. */
case before_dot:
case at_dot:
case after_dot:
continue;
#endif /* not emacs */
case no_op:
case begline:
case endline:
case begbuf:
case endbuf:
case wordbound:
case notwordbound:
case wordbeg:
case wordend:
case push_dummy_failure:
continue;
case jump_n:
case pop_failure_jump:
case maybe_pop_jump:
case jump:
case jump_past_alt:
case dummy_failure_jump:
EXTRACT_NUMBER_AND_INCR (j, p);
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
ordinary jump. For a * loop, it has pushed its failure
point already; if so, discard that as redundant. */
if ((re_opcode_t) *p != on_failure_jump
&& (re_opcode_t) *p != succeed_n)
continue;
p++;
EXTRACT_NUMBER_AND_INCR (j, p);
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)
fail_stack.avail--;
continue;
case on_failure_jump:
case on_failure_keep_string_jump:
handle_on_failure_jump:
EXTRACT_NUMBER_AND_INCR (j, p);
/* For some patterns, e.g., `(a?)?', `p+j' here points to the
end of the pattern. We don't want to push such a point,
since when we restore it above, entering the switch will
increment `p' past the end of the pattern. We don't need
to push such a point since we obviously won't find any more
fastmap entries beyond `pend'. Such a pattern can match
the null string, though. */
if (p + j < pend)
{
if (!PUSH_PATTERN_OP (p + j, fail_stack))
return -2;
}
else
bufp->can_be_null = 1;
if (succeed_n_p)
{
EXTRACT_NUMBER_AND_INCR (k, p); /* Skip the n. */
succeed_n_p = false;
}
continue;
case succeed_n:
/* Get to the number of times to succeed. */
p += 2;
/* Increment p past the n for when k != 0. */
EXTRACT_NUMBER_AND_INCR (k, p);
if (k == 0)
{
p -= 4;
succeed_n_p = true; /* Spaghetti code alert. */
goto handle_on_failure_jump;
}
continue;
case set_number_at:
p += 4;
continue;
case start_memory:
case stop_memory:
p += 2;
continue;
default:
abort (); /* We have listed all the cases. */
} /* switch *p++ */
/* Getting here means we have found the possible starting
characters for one path of the pattern -- and that the empty
string does not match. We need not follow this path further.
Instead, look at the next alternative (remembered on the
stack), or quit if no more. The test at the top of the loop
does these things. */
path_can_be_null = false;
p = pend;
} /* while 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;
return 0;
} /* re_compile_fastmap */
/* Set REGS to hold NUM_REGS registers, storing them in STARTS and
ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use
this memory for recording register information. STARTS and ENDS
must be allocated using the malloc library routine, and must each
be at least NUM_REGS * sizeof (regoff_t) bytes long.
If NUM_REGS == 0, then subsequent matches should allocate their own
register data.
Unless this function is called, the first search or match using
PATTERN_BUFFER will allocate its own register data, without
freeing the old data. */
void
re_set_registers (bufp, regs, num_regs, starts, ends)
struct re_pattern_buffer *bufp;
struct re_registers *regs;
unsigned num_regs;
regoff_t *starts, *ends;
{
if (num_regs)
{
bufp->regs_allocated = REGS_REALLOCATE;
regs->num_regs = num_regs;
regs->start = starts;
regs->end = ends;
}
else
{
bufp->regs_allocated = REGS_UNALLOCATED;
regs->num_regs = 0;
regs->start = regs->end = (regoff_t) 0;
}
}
/* Searching routines. */
/* Like re_search_2, below, but only one string is specified, and
doesn't let you say where to stop matching. */
int
re_search (bufp, string, size, startpos, range, regs)
struct re_pattern_buffer *bufp;
const char *string;
int size, startpos, range;
struct re_registers *regs;
{
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.
We return either the position in the strings at which the match was
found, -1 if no match, or -2 if error (such as failure
stack overflow). */
int
re_search_2 (bufp, string1, size1, string2, size2, startpos, range, regs, stop)
struct re_pattern_buffer *bufp;
const char *string1, *string2;
int size1, size2;
int startpos;
int range;
struct re_registers *regs;
int stop;
{
int val;
register char *fastmap = bufp->fastmap;
register char *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;
else if (endpos > total_size)
range = total_size - startpos;
/* If the search isn't to be a backwards one, don't waste time in a
search for a pattern that must be anchored. */
if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf && range > 0)
{
if (startpos > 0)
return -1;
else
range = 1;
}
/* 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
the first null string. */
if (fastmap && startpos < total_size && !bufp->can_be_null)
{
if (range > 0) /* Searching forwards. */
{
register const char *d;
register int lim = 0;
int irange = range;
if (startpos < size1 && startpos + range >= size1)
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)
while (range > lim
&& !fastmap[(unsigned char)
translate[(unsigned char) *d++]])
range--;
else
while (range > lim && !fastmap[(unsigned char) *d++])
range--;
startpos += irange - range;
}
else /* Searching backwards. */
{
register char c = (size1 == 0 || startpos >= size1
? string2[startpos - size1]
: string1[startpos]);
if (!fastmap[(unsigned char) TRANSLATE (c)])
goto advance;
}
}
/* If can't match the null string, and that's all we have left, fail. */
if (range >= 0 && startpos == total_size && fastmap
&& !bufp->can_be_null)
return -1;
val = re_match_2 (bufp, string1, size1, string2, size2,
startpos, regs, stop);
if (val >= 0)
return startpos;
if (val == -2)
return -2;
advance:
if (!range)
break;
else if (range > 0)
{
range--;
startpos++;
}
else
{
range++;
startpos--;
}
}
return -1;
} /* re_search_2 */
/* Declarations and macros for re_match_2. */
static int bcmp_translate ();
static boolean alt_match_null_string_p (),
common_op_match_null_string_p (),
group_match_null_string_p ();
/* Structure for per-register (a.k.a. per-group) information.
This must not be longer than one word, because we push this value
onto the failure stack. Other register information, such as the
starting and ending positions (which are addresses), and the list of
inner groups (which is a bits list) are maintained in separate
variables.
We are making a (strictly speaking) nonportable assumption here: that
the compiler will pack our bit fields into something that fits into
the type of `word', i.e., is something that fits into one item on the
failure stack. */
typedef union
{
fail_stack_elt_t word;
struct
{
/* This field is one if this group can match the empty string,
zero if not. If not yet determined, `MATCH_NULL_UNSET_VALUE'. */
#define MATCH_NULL_UNSET_VALUE 3
unsigned match_null_string_p : 2;
unsigned is_active : 1;
unsigned matched_something : 1;
unsigned ever_matched_something : 1;
} bits;
} register_info_type;
#define REG_MATCH_NULL_STRING_P(R) ((R).bits.match_null_string_p)
#define IS_ACTIVE(R) ((R).bits.is_active)
#define MATCHED_SOMETHING(R) ((R).bits.matched_something)
#define EVER_MATCHED_SOMETHING(R) ((R).bits.ever_matched_something)
/* Call this when have matched 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)
/* This converts PTR, a pointer into one of the search strings `string1'
and `string2' into an offset from the beginning of that string. */
#define POINTER_TO_OFFSET(ptr)
(FIRST_STRING_P (ptr) ? (ptr) - string1 : (ptr) - string2 + size1)
/* Registers are set to a sentinel when they haven't yet matched. */
#define REG_UNSET_VALUE ((char *) -1)
#define REG_UNSET(e) ((e) == REG_UNSET_VALUE)
/* Macros for dealing with the split strings in re_match_2. */
#define MATCHING_IN_FIRST_STRING (dend == end_match_1)
/* Call before fetching a character with *d. This switches over to
string2 if necessary. */
#define PREFETCH()
while (d == dend)
{
/* End of string2 => fail. */
if (dend == end_match_2)
goto fail;
/* End of string1 => advance to string2. */
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'. */
#define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2)
#define AT_STRINGS_END(d) ((d) == end2)
/* Test if D points to a character which is word-constituent. We have
two special cases to check for: if past the end of string1, look at
the first character in string2; and if before the beginning of
string2, look at the last character in string1. */
#define WORDCHAR_P(d)
(SYNTAX ((d) == end1 ? *string2
: (d) == string2 - 1 ? *(end1 - 1) : *(d))
== Sword)
/* Test if the character before D and the one at D differ with respect
to being word-constituent. */
#define AT_WORD_BOUNDARY(d)
(AT_STRINGS_BEG (d) || AT_STRINGS_END (d)
|| WORDCHAR_P (d - 1) != WORDCHAR_P (d))
/* Free everything we malloc. */
#ifdef REGEX_MALLOC
#define FREE_VAR(var) if (var) free (var); var = NULL
#define FREE_VARIABLES()
do {
FREE_VAR (fail_stack.stack);
FREE_VAR (regstart);
FREE_VAR (regend);
FREE_VAR (old_regstart);
FREE_VAR (old_regend);
FREE_VAR (best_regstart);
FREE_VAR (best_regend);
FREE_VAR (reg_info);
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 */
/* These values must meet several constraints. They must not be valid
register values; since we have a limit of 255 registers (because
we use only one byte in the pattern for the register number), we can
use numbers larger than 255. They must differ by 1, because of
NUM_FAILURE_ITEMS above. And the value for the lowest register must
be larger than the value for the highest register, so we do not try
to actually save any registers when none are active. */
#define NO_HIGHEST_ACTIVE_REG (1 << BYTEWIDTH)
#define NO_LOWEST_ACTIVE_REG (NO_HIGHEST_ACTIVE_REG + 1)
/* Matching routines. */
#ifndef emacs /* Emacs never uses this. */
/* re_match is like re_match_2 except it takes only a single string. */
int
re_match (bufp, string, size, pos, regs)
struct re_pattern_buffer *bufp;
const char *string;
int size, pos;
struct re_registers *regs;
{
return re_match_2 (bufp, NULL, 0, string, size, pos, regs, size);
}
#endif /* not emacs */
/* re_match_2 matches the compiled pattern in BUFP against the
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.
We return -1 if no match, -2 if an internal error (such as the
failure stack overflowing). Otherwise, we return the length of the
matched substring. */
int
re_match_2 (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;
/* Just past the end of the corresponding string. */
const char *end1, *end2;
/* Pointers into string1 and string2, just past the last characters in
each to consider matching. */
const char *end_match_1, *end_match_2;
/* 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;
/* We use this to map every character in the string. */
char *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
restart, regend, and reg_info for all registers corresponding to
the subexpressions we're currently inside, plus the number of such
registers, and, finally, two char *'s. The first char * is where
to resume scanning the pattern; the second one is where to resume
scanning the strings. If the latter is zero, the failure point is
a ``dummy''; if a failure happens and the failure point is a dummy,
it gets discarded and the next next one is tried. */
fail_stack_type fail_stack;
#ifdef DEBUG
static unsigned failure_id = 0;
unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0;
#endif
/* 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;
/* Information on the contents of registers. These are pointers into
the input strings; they record just what was matched (on this
attempt) by a subexpression part of the pattern, that is, the
regnum-th regstart pointer points to where in the pattern we began
matching and the regnum-th regend points to right after where we
stopped matching the regnum-th subexpression. (The zeroth register
keeps track of what the whole pattern matches.) */
const char **regstart, **regend;
/* If a group that's operated upon by a repetition operator fails to
match anything, then the register for its start will need to be
restored because it will have been set to wherever in the string we
are when we last see its open-group operator. Similarly for a
register's end. */
const char **old_regstart, **old_regend;
/* The is_active field of reg_info helps us keep track of which (possibly
nested) subexpressions we are currently in. The matched_something
field of reg_info[reg_num] helps us tell whether or not we have
matched any of the pattern so far this time through the reg_num-th
subexpression. These two fields get reset each time through any
loop their register is in. */
register_info_type *reg_info;
/* The following record the register info as found in the above
variables when we find a match better than any we've seen before.
This happens as we backtrack through the failure points, which in
turn happens only if we have not yet matched the entire string. */
unsigned best_regs_set = false;
const char **best_regstart, **best_regend;
/* 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
any of the other register vectors, and it seems rather a kludge to
treat `best_regend' differently than the rest. So we keep track of
the end of the best match so far in a separate variable. We
initialize this to NULL so that when we backtrack the first time
and need to test it, it's not garbage. */
const char *match_end = NULL;
/* Used when we pop values we don't care about. */
const char **reg_dummy;
register_info_type *reg_info_dummy;
#ifdef DEBUG
/* Counts the total number of registers pushed. */
unsigned num_regs_pushed = 0;
#endif
DEBUG_PRINT1 ("nnEntering re_match_2.n");
INIT_FAIL_STACK ();
/* Do not bother to initialize all the register variables if there are
no groups in the pattern, as it takes a fair amount of time. If
there are groups, we include space for register 0 (the whole
pattern), even though we never use it, since it simplifies the
array indexing. We should fix this. */
if (bufp->re_nsub)
{
regstart = REGEX_TALLOC (num_regs, const char *);
regend = REGEX_TALLOC (num_regs, const char *);
old_regstart = REGEX_TALLOC (num_regs, const char *);
old_regend = REGEX_TALLOC (num_regs, const char *);
best_regstart = REGEX_TALLOC (num_regs, const char *);
best_regend = REGEX_TALLOC (num_regs, const char *);
reg_info = REGEX_TALLOC (num_regs, register_info_type);
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))
{
FREE_VARIABLES ();
return -2;
}
}
#ifdef REGEX_MALLOC
else
{
/* We must initialize all our variables to NULL, so that
`FREE_VARIABLES' doesn't try to free them. */
regstart = regend = old_regstart = old_regend = best_regstart
= best_regend = reg_dummy = NULL;
reg_info = reg_info_dummy = (register_info_type *) NULL;
}
#endif /* REGEX_MALLOC */
/* The starting position is bogus. */
if (pos < 0 || pos > size1 + size2)
{
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]
= 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)
{
string2 = string1;
size2 = size1;
string1 = 0;
size1 = 0;
}
end1 = string1 + size1;
end2 = string2 + size2;
/* Compute where to stop matching, within the two strings. */
if (stop <= size1)
{
end_match_1 = string1 + stop;
end_match_2 = string2;
}
else
{
end_match_1 = end1;
end_match_2 = string2 + stop - size1;
}
/* `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)
{
d = string1 + pos;
dend = end_match_1;
}
else
{
d = string2 + pos - size1;
dend = end_match_2;
}
DEBUG_PRINT1 ("The compiled pattern is: ");
DEBUG_PRINT_COMPILED_PATTERN (bufp, p, pend);
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. */
for (;;)
{
DEBUG_PRINT2 ("n0x%x: ", p);
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)
{
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))
{
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;
}
/* If no failure points, don't restore garbage. */
else if (best_regs_set)
{
restore_best_regs:
/* Restore best match. It may happen that `dend ==
end_match_1' while the restored d is in string2.
For example, the pattern `x.*y.*z' against the
strings `x-' and `y-z-', if the two strings are
not consecutive in memory. */
DEBUG_PRINT1 ("Restoring best registers.n");
d = match_end;
dend = ((d >= string1 && d <= end1)
? end_match_1 : end_match_2);
for (mcnt = 1; mcnt < num_regs; mcnt++)
{
regstart[mcnt] = best_regstart[mcnt];
regend[mcnt] = best_regend[mcnt];
}
}
} /* d != end_match_2 */
DEBUG_PRINT1 ("Accepting match.n");
/* If caller wants register contents data back, do it. */
if (regs && !bufp->no_sub)
{
/* Have the register data arrays been allocated? */
if (bufp->regs_allocated == REGS_UNALLOCATED)
{ /* No. So allocate them with malloc. We need one
extra element beyond `num_regs' for the `-1' marker
GNU code uses. */
regs->num_regs = MAX (RE_NREGS, num_regs + 1);
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;
bufp->regs_allocated = REGS_REALLOCATE;
}
else if (bufp->regs_allocated == REGS_REALLOCATE)
{ /* Yes. If we need more elements than were already
allocated, reallocate them. If we need fewer, just
leave it alone. */
if (regs->num_regs < num_regs + 1)
{
regs->num_regs = num_regs + 1;
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;
}
}
else
assert (bufp->regs_allocated == REGS_FIXED);
/* Convert the pointer data in `regstart' and `regend' to
indices. Register zero has to be set differently,
since we haven't kept track of any info for it. */
if (regs->num_regs > 0)
{
regs->start[0] = pos;
regs->end[0] = (MATCHING_IN_FIRST_STRING ? d - string1
: 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++)
{
if (REG_UNSET (regstart[mcnt]) || REG_UNSET (regend[mcnt]))
regs->start[mcnt] = regs->end[mcnt] = -1;
else
{
regs->start[mcnt] = POINTER_TO_OFFSET (regstart[mcnt]);
regs->end[mcnt] = 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,
because we always allocate enough to have at least one
-1 at the end. */
for (mcnt = num_regs; mcnt < regs->num_regs; mcnt++)
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
: string2 - size1);
DEBUG_PRINT2 ("Returning %d from re_match_2.n", mcnt);
return mcnt;
}
/* Otherwise match next pattern command. */
#ifdef SWITCH_ENUM_BUG
switch ((int) ((re_opcode_t) *p++))
#else
switch ((re_opcode_t) *p++)
#endif
{
/* Ignore these. Used to ignore the n of succeed_n's which
currently have n == 0. */
case no_op:
DEBUG_PRINT1 ("EXECUTING no_op.n");
break;
/* Match the next n pattern characters exactly. The following
byte in the pattern defines n, and the n bytes after that
are the characters to match. */
case exactn:
mcnt = *p++;
DEBUG_PRINT2 ("EXECUTING exactn %d.n", mcnt);
/* This is written out as an if-else so we don't waste time
testing `translate' inside the loop. */
if (translate)
{
do
{
PREFETCH ();
if (translate[(unsigned char) *d++] != (char) *p++)
goto fail;
}
while (--mcnt);
}
else
{
do
{
PREFETCH ();
if (*d++ != (char) *p++) goto fail;
}
while (--mcnt);
}
SET_REGS_MATCHED ();
break;
/* Match any character except possibly a newline or a null. */
case anychar:
DEBUG_PRINT1 ("EXECUTING anychar.n");
PREFETCH ();
if ((!(bufp->syntax & RE_DOT_NEWLINE) && TRANSLATE (*d) == 'n')
|| (bufp->syntax & RE_DOT_NOT_NULL && TRANSLATE (*d) == '00'))
goto fail;
SET_REGS_MATCHED ();
DEBUG_PRINT2 (" Matched `%d'.n", *d);
d++;
break;
case charset:
case charset_not:
{
register unsigned char c;
boolean not = (re_opcode_t) *(p - 1) == charset_not;
DEBUG_PRINT2 ("EXECUTING charset%s.n", not ? "_not" : "");
PREFETCH ();
c = TRANSLATE (*d); /* The character to match. */
/* Cast to `unsigned' instead of `unsigned char' in case the
bit list is a full 32 bytes long. */
if (c < (unsigned) (*p * BYTEWIDTH)
&& p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
not = !not;
p += 1 + *p;
if (!not) goto fail;
SET_REGS_MATCHED ();
d++;
break;
}
/* The beginning of a group is represented by start_memory.
The arguments are the register number in the next byte, and the
number of groups inner to this one in the next. The text
matched within the group is recorded (in the internal
registers data structure) under the register number. */
case start_memory:
DEBUG_PRINT3 ("EXECUTING start_memory %d (%d):n", *p, p[1]);
/* 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])
= group_match_null_string_p (&p1, pend, reg_info);
/* Save the position in the string where we were the last time
we were at this open-group operator in case the group is
operated upon by a repetition operator, e.g., with `(a*)*b'
against `ab'; then we want to ignore where we are now in
the string in case this attempt to match fails. */
old_regstart[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p])
? REG_UNSET (regstart[*p]) ? d : regstart[*p]
: regstart[*p];
DEBUG_PRINT2 (" old_regstart: %dn",
POINTER_TO_OFFSET (old_regstart[*p]));
regstart[*p] = d;
DEBUG_PRINT2 (" regstart: %dn", POINTER_TO_OFFSET (regstart[*p]));
IS_ACTIVE (reg_info[*p]) = 1;
MATCHED_SOMETHING (reg_info[*p]) = 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)
lowest_active_reg = *p;
/* Move past the register number and inner group count. */
p += 2;
break;
/* The stop_memory opcode represents the end of a group. Its
arguments are the same as start_memory's: the register
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*)*)*'
against `aba'; then we want to ignore where we are now in
the string in case this attempt to match fails. */
old_regend[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p])
? REG_UNSET (regend[*p]) ? d : regend[*p]
: regend[*p];
DEBUG_PRINT2 (" old_regend: %dn",
POINTER_TO_OFFSET (old_regend[*p]));
regend[*p] = d;
DEBUG_PRINT2 (" regend: %dn", POINTER_TO_OFFSET (regend[*p]));
/* This register isn't active anymore. */
IS_ACTIVE (reg_info[*p]) = 0;
/* If this was the only register active, nothing is active
anymore. */
if (lowest_active_reg == highest_active_reg)
{
lowest_active_reg = NO_LOWEST_ACTIVE_REG;
highest_active_reg = NO_HIGHEST_ACTIVE_REG;
}
else
{ /* We must scan for the new highest active register, since
it isn't necessarily one less than now: consider
(a(b)c(d(e)f)g). When group 3 ends, after the f), the
new highest active register is 1. */
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
`stop_memory'. For example, in ((.)*) we save
registers 1 and 2 as a result of the *, but when we pop
back to the second ), we are at the stop_memory 1.
Thus, nothing is active. */
if (r == 0)
{
lowest_active_reg = NO_LOWEST_ACTIVE_REG;
highest_active_reg = NO_HIGHEST_ACTIVE_REG;
}
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
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)
&& (p + 2) < pend)
{
boolean is_a_jump_n = false;
p1 = p + 2;
mcnt = 0;
switch ((re_opcode_t) *p1++)
{
case jump_n:
is_a_jump_n = true;
case pop_failure_jump:
case maybe_pop_jump:
case jump:
case dummy_failure_jump:
EXTRACT_NUMBER_AND_INCR (mcnt, 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
by forcing a failure after pushing on the stack the
on_failure_jump's jump in the pattern, and d. */
if (mcnt < 0 && (re_opcode_t) *p1 == on_failure_jump
&& (re_opcode_t) p1[3] == start_memory && p1[4] == *p)
{
/* If this group ever matched anything, then restore
what its registers were before trying this last
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;
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])
regend[r] = old_regend[r];
}
}
p1++;
EXTRACT_NUMBER_AND_INCR (mcnt, p1);
PUSH_FAILURE_POINT (p1 + mcnt, d, -2);
goto fail;
}
}
/* Move past the register number and the inner group count. */
p += 2;
break;
/* <digit> has been turned into a `duplicate' command which is
followed by the numeric value of <digit> as the register number. */
case duplicate:
{
register const char *d2, *dend2;
int regno = *p++; /* Get which register to match against. */
DEBUG_PRINT2 ("EXECUTING duplicate %d.n", regno);
/* 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])
== FIRST_STRING_P (regend[regno]))
? regend[regno] : end_match_1);
for (;;)
{
/* If necessary, advance to next segment in register
contents. */
while (d2 == dend2)
{
if (dend2 == end_match_2) break;
if (dend2 == regend[regno]) break;
/* End of string1 => advance to string2. */
d2 = string2;
dend2 = regend[regno];
}
/* At end of register contents => success */
if (d2 == dend2) break;
/* If necessary, advance to next segment in data. */
PREFETCH ();
/* 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)
: bcmp (d, d2, mcnt))
goto fail;
d += mcnt, d2 += mcnt;
}
}
break;
/* begline matches the empty string at the beginning of the string
(unless `not_bol' is set in `bufp'), and, if
`newline_anchor' is set, after newlines. */
case begline:
DEBUG_PRINT1 ("EXECUTING begline.n");
if (AT_STRINGS_BEG (d))
{
if (!bufp->not_bol) break;
}
else if (d[-1] == 'n' && bufp->newline_anchor)
{
break;
}
/* In all other cases, we fail. */
goto fail;
/* endline is the dual of begline. */
case endline:
DEBUG_PRINT1 ("EXECUTING endline.n");
if (AT_STRINGS_END (d))
{
if (!bufp->not_eol) break;
}
/* We have to ``prefetch'' the next character. */
else if ((d == end1 ? *string2 : *d) == 'n'
&& bufp->newline_anchor)
{
break;
}
goto fail;
/* Match at the very beginning of the data. */
case begbuf:
DEBUG_PRINT1 ("EXECUTING begbuf.n");
if (AT_STRINGS_BEG (d))
break;
goto fail;
/* Match at the very end of the data. */
case endbuf:
DEBUG_PRINT1 ("EXECUTING endbuf.n");
if (AT_STRINGS_END (d))
break;
goto fail;
/* on_failure_keep_string_jump is used to optimize `.*n'. It
pushes NULL as the value for the string on the stack. Then
`pop_failure_point' will keep the current value for the
string, instead of restoring it. To see why, consider
matching `foonbar' against `.*n'. The .* matches the foo;
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
share its code. The only reason to push anything on the
stack at all is that otherwise we would have to change
`anychar's code to do something besides goto fail in this
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);
PUSH_FAILURE_POINT (p + mcnt, NULL, -2);
break;
/* 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
the rest of the alternatives. (They really jump to the
ending jump of the following alternative, because tensioning
these jumps is a hassle.)
Repeats start with an on_failure_jump that points past both
the repetition text and either the following jump or
pop_failure_jump back to this on_failure_jump. */
case on_failure_jump:
on_failure:
DEBUG_PRINT1 ("EXECUTING on_failure_jump");
EXTRACT_NUMBER_AND_INCR (mcnt, p);
DEBUG_PRINT3 (" %d (to 0x%x)", mcnt, p + mcnt);
/* If this on_failure_jump comes right before a group (i.e.,
the original * applied to a group), save the information
for that group and all inner ones, so that if we fail back
to this point, the group's information will be correct.
For example, in (a*)*1, we need the preceding group,
and in ((a*)b*)2, we need the inner group. */
/* We can't use `p' to check ahead because we push
a failure point to `p + mcnt' after we do this. */
p1 = p;
/* We need to skip no_op's before we look for the
start_memory in case this on_failure_jump is happening as
the result of a completed succeed_n, as in (a){1,3}b1
against aba. */
while (p1 < pend && (re_opcode_t) *p1 == no_op)
p1++;
if (p1 < pend && (re_opcode_t) *p1 == start_memory)
{
/* We have a new highest active register now. This will
get reset at the start_memory we are about to get to,
but we will have saved all the registers relevant to
this repetition op, as described above. */
highest_active_reg = *(p1 + 1) + *(p1 + 2);
if (lowest_active_reg == NO_LOWEST_ACTIVE_REG)
lowest_active_reg = *(p1 + 1);
}
DEBUG_PRINT1 (":n");
PUSH_FAILURE_POINT (p + mcnt, d, -2);
break;
/* A smart repeat ends with `maybe_pop_jump'.
We change it to either `pop_failure_jump' or `jump'. */
case maybe_pop_jump:
EXTRACT_NUMBER_AND_INCR (mcnt, p);
DEBUG_PRINT2 ("EXECUTING maybe_pop_jump %d.n", mcnt);
{
register unsigned char *p2 = p;
/* Compare the beginning of the repeat with what in the
pattern follows its end. If we can establish that there
is nothing that they would both match, i.e., that we
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
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. */
/* If we're at the end of the pattern, we can change. */
if (p2 == pend)
{
/* Consider what happens when matching ":(.*)"
against ":/". I don't really understand this code
yet. */
p[-3] = (unsigned char) pop_failure_jump;
DEBUG_PRINT1
(" End of pattern: change to `pop_failure_jump'.n");
}
else if ((re_opcode_t) *p2 == exactn
|| (bufp->newline_anchor && (re_opcode_t) *p2 == endline))
{
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;
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;
/* `not' is equal to 1 if c would match, which means
that we can't change to pop_failure_jump. */
if (!not)
{
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)
{
p[-1] = (unsigned char) jump;
DEBUG_PRINT1 (" Match => jump.n");
goto unconditional_jump;
}
/* Note fall through. */
/* The end of a simple repeat has a pop_failure_jump back to
its matching on_failure_jump, where the latter will push a
failure point. The pop_failure_jump takes off failure
points put on by this pop_failure_jump's matching
on_failure_jump; we got through the pattern to here from the
matching on_failure_jump, so didn't fail. */
case pop_failure_jump:
{
/* We need to pass separate storage for the lowest and
highest registers, even though we don't care about the
actual values. Otherwise, we will restore only one
register from the stack, since lowest will == highest in
`pop_failure_point'. */
unsigned dummy_low_reg, dummy_high_reg;
unsigned char *pdummy;
const char *sdummy;
DEBUG_PRINT1 ("EXECUTING pop_failure_jump.n");
POP_FAILURE_POINT (sdummy, pdummy,
dummy_low_reg, dummy_high_reg,
reg_dummy, reg_dummy, reg_info_dummy);
}
/* Note fall through. */
/* Unconditionally jump (without popping any failure points). */
case jump:
unconditional_jump:
EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */
DEBUG_PRINT2 ("EXECUTING jump %d ", mcnt);
p += mcnt; /* Do the 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:
DEBUG_PRINT1 ("EXECUTING jump_past_alt.n");
goto unconditional_jump;
/* Normally, the on_failure_jump pushes a failure point, which
then gets popped at pop_failure_jump. We will end up at
pop_failure_jump, also, and with a pattern of, say, `a+', we
are skipping over the on_failure_jump, so we have to push
something meaningless for pop_failure_jump to pop. */
case dummy_failure_jump:
DEBUG_PRINT1 ("EXECUTING dummy_failure_jump.n");
/* It doesn't matter what we push for the string here. What
the code at `fail' tests is the value for the pattern. */
PUSH_FAILURE_POINT (0, 0, -2);
goto unconditional_jump;
/* At the end of an alternative, we need to push a dummy failure
point in case we are followed by a `pop_failure_jump', because
we don't want the failure point for the alternative to be
popped. For example, matching `(a|ab)*' against `aab'
requires that we match the `ab' alternative. */
case push_dummy_failure:
DEBUG_PRINT1 ("EXECUTING push_dummy_failure.n");
/* See comments just above at `dummy_failure_jump' about the
two zeroes. */
PUSH_FAILURE_POINT (0, 0, -2);
break;
/* Have to succeed matching what follows at least n times.
After that, handle like `on_failure_jump'. */
case succeed_n:
EXTRACT_NUMBER (mcnt, p + 2);
DEBUG_PRINT2 ("EXECUTING succeed_n %d.n", mcnt);
assert (mcnt >= 0);
/* Originally, this is how many times we HAVE to succeed. */
if (mcnt > 0)
{
mcnt--;
p += 2;
STORE_NUMBER_AND_INCR (p, mcnt);
DEBUG_PRINT3 (" Setting 0x%x to %d.n", p, mcnt);
}
else if (mcnt == 0)
{
DEBUG_PRINT2 (" Setting two bytes from 0x%x to no_op.n", p+2);
p[2] = (unsigned char) no_op;
p[3] = (unsigned char) no_op;
goto on_failure;
}
break;
case jump_n:
EXTRACT_NUMBER (mcnt, p + 2);
DEBUG_PRINT2 ("EXECUTING jump_n %d.n", mcnt);
/* Originally, this is how many times we CAN jump. */
if (mcnt)
{
mcnt--;
STORE_NUMBER (p + 2, mcnt);
goto unconditional_jump;
}
/* If don't have to jump any more, skip over the rest of command. */
else
p += 4;
break;
case set_number_at:
{
DEBUG_PRINT1 ("EXECUTING set_number_at.n");
EXTRACT_NUMBER_AND_INCR (mcnt, p);
p1 = p + mcnt;
EXTRACT_NUMBER_AND_INCR (mcnt, p);
DEBUG_PRINT3 (" Setting 0x%x to %d.n", p1, mcnt);
STORE_NUMBER (p1, mcnt);
break;
}
case wordbound:
DEBUG_PRINT1 ("EXECUTING wordbound.n");
if (AT_WORD_BOUNDARY (d))
break;
goto fail;
case notwordbound:
DEBUG_PRINT1 ("EXECUTING notwordbound.n");
if (AT_WORD_BOUNDARY (d))
goto fail;
break;
case wordbeg:
DEBUG_PRINT1 ("EXECUTING wordbeg.n");
if (WORDCHAR_P (d) && (AT_STRINGS_BEG (d) || !WORDCHAR_P (d - 1)))
break;
goto fail;
case wordend:
DEBUG_PRINT1 ("EXECUTING wordend.n");
if (!AT_STRINGS_BEG (d) && WORDCHAR_P (d - 1)
&& (!WORDCHAR_P (d) || AT_STRINGS_END (d)))
break;
goto fail;
#ifdef emacs
#ifdef emacs19
case before_dot:
DEBUG_PRINT1 ("EXECUTING before_dot.n");
if (PTR_CHAR_POS ((unsigned char *) d) >= point)
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);
mcnt = *p++;
goto matchsyntax;
case wordchar:
DEBUG_PRINT1 ("EXECUTING Emacs wordchar.n");
mcnt = (int) Sword;
matchsyntax:
PREFETCH ();
if (SYNTAX (*d++) != (enum syntaxcode) mcnt)
goto fail;
SET_REGS_MATCHED ();
break;
case notsyntaxspec:
DEBUG_PRINT2 ("EXECUTING notsyntaxspec %d.n", mcnt);
mcnt = *p++;
goto matchnotsyntax;
case notwordchar:
DEBUG_PRINT1 ("EXECUTING Emacs notwordchar.n");
mcnt = (int) Sword;
matchnotsyntax:
PREFETCH ();
if (SYNTAX (*d++) == (enum syntaxcode) mcnt)
goto fail;
SET_REGS_MATCHED ();
break;
#else /* not emacs */
case wordchar:
DEBUG_PRINT1 ("EXECUTING non-Emacs wordchar.n");
PREFETCH ();
if (!WORDCHAR_P (d))
goto fail;
SET_REGS_MATCHED ();
d++;
break;
case notwordchar:
DEBUG_PRINT1 ("EXECUTING non-Emacs notwordchar.n");
PREFETCH ();
if (WORDCHAR_P (d))
goto fail;
SET_REGS_MATCHED ();
d++;
break;
#endif /* not emacs */
default:
abort ();
}
continue; /* Successfully executed one pattern command; keep going. */
/* We goto here if a matching operation fails. */
fail:
if (!FAIL_STACK_EMPTY ())
{ /* A restart point is known. Restore to that state. */
DEBUG_PRINT1 ("nFAIL:n");
POP_FAILURE_POINT (d, p,
lowest_active_reg, highest_active_reg,
regstart, regend, reg_info);
/* If this failure point is a dummy, try the next one. */
if (!p)
goto fail;
/* If we failed to the end of the pattern, don't examine *p. */
assert (p <= pend);
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_n:
is_a_jump_n = true;
case maybe_pop_jump:
case pop_failure_jump:
case jump:
p1 = p + 1;
EXTRACT_NUMBER_AND_INCR (mcnt, p1);
p1 += mcnt;
if ((is_a_jump_n && (re_opcode_t) *p1 == succeed_n)
|| (!is_a_jump_n
&& (re_opcode_t) *p1 == on_failure_jump))
goto fail;
break;
default:
/* do nothing */ ;
}
}
if (d >= string1 && d <= end1)
dend = end_match_1;
}
else
break; /* Matching at this starting point really fails. */
} /* for (;;) */
if (best_regs_set)
goto restore_best_regs;
FREE_VARIABLES ();
return -1; /* Failure to match. */
} /* re_match_2 */
/* Subroutine definitions for re_match_2. */
/* 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.
We don't handle duplicates properly (yet). */
static boolean
group_match_null_string_p (p, end, reg_info)
unsigned char **p, *end;
register_info_type *reg_info;
{
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. */
if (mcnt >= 0)
{
/* Go through the on_failure_jumps of the alternatives,
seeing if any of the alternatives cannot match nothing.
The last alternative starts with only a jump,
whereas the rest start with on_failure_jump and end
with a jump, e.g., here is the pattern for `a|b|c':
/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
So, we have to first go through the first (n-1)
alternatives and then deal with the last one separately. */
/* Deal with the first (n-1) alternatives, which start
with an on_failure_jump (see above) that jumps to right
past a jump_past_alt. */
while ((re_opcode_t) p1[mcnt-3] == jump_past_alt)
{
/* `mcnt' holds how many bytes long the alternative
is, including the ending `jump_past_alt' and
its number. */
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;
/* 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++;
EXTRACT_NUMBER_AND_INCR (mcnt, p1);
if ((re_opcode_t) p1[mcnt-3] != jump_past_alt)
{
/* Get to the beginning of the n-th alternative. */
p1 -= 3;
break;
}
}
/* Deal with the last alternative: go back and get number
of the `jump_past_alt' just before it. `mcnt' contains
the length of the alternative. */
EXTRACT_NUMBER (mcnt, p1 - 2);
if (!alt_match_null_string_p (p1, p1 + mcnt, reg_info))
return false;
p1 += mcnt; /* Get past the n-th alternative. */
} /* if mcnt > 0 */
break;
case stop_memory:
assert (p1[1] == **p);
*p = p1 + 2;
return true;
default:
if (!common_op_match_null_string_p (&p1, end, reg_info))
return false;
}
} /* while p1 < end */
return false;
} /* group_match_null_string_p */
/* 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;
register_info_type *reg_info;
{
int mcnt;
unsigned char *p1 = p;
while (p1 < end)
{
/* 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. */
case on_failure_jump:
p1++;
EXTRACT_NUMBER_AND_INCR (mcnt, p1);
p1 += mcnt;
break;
default:
if (!common_op_match_null_string_p (&p1, end, reg_info))
return false;
}
} /* while p1 < end */
return true;
} /* alt_match_null_string_p */
/* Deals with the ops common to group_match_null_string_p and
alt_match_null_string_p.
Sets P to one after the op and its arguments, if any. */
static boolean
common_op_match_null_string_p (p, end, reg_info)
unsigned char **p, *end;
register_info_type *reg_info;
{
int mcnt;
boolean ret;
int reg_no;
unsigned char *p1 = *p;
switch ((re_opcode_t) *p1++)
{
case no_op:
case begline:
case endline:
case begbuf:
case endbuf:
case wordbeg:
case wordend:
case wordbound:
case notwordbound:
#ifdef emacs
case before_dot:
case at_dot:
case after_dot:
#endif
break;
case start_memory:
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 (REG_MATCH_NULL_STRING_P (reg_info[reg_no]) == MATCH_NULL_UNSET_VALUE)
REG_MATCH_NULL_STRING_P (reg_info[reg_no]) = ret;
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);
if (mcnt >= 0)
p1 += mcnt;
else
return false;
break;
case succeed_n:
/* Get to the number of times to succeed. */
p1 += 2;
EXTRACT_NUMBER_AND_INCR (mcnt, p1);
if (mcnt == 0)
{
p1 -= 4;
EXTRACT_NUMBER_AND_INCR (mcnt, p1);
p1 += mcnt;
}
else
return false;
break;
case duplicate:
if (!REG_MATCH_NULL_STRING_P (reg_info[*p1]))
return false;
break;
case set_number_at:
p1 += 4;
default:
/* All other opcodes mean we cannot match the empty string. */
return false;
}
*p = p1;
return true;
} /* common_op_match_null_string_p */
/* 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;
{
register unsigned char *p1 = s1, *p2 = s2;
while (len)
{
if (translate[*p1++] != translate[*p2++]) return 1;
len--;
}
return 0;
}
/* Entry points for GNU code. */
/* 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 *
re_compile_pattern (pattern, length, bufp)
const char *pattern;
int length;
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];
}
/* Entry points compatible with 4.2 BSD regex library. We don't define
them if this is an Emacs or POSIX compilation. */
#if !defined (emacs) && !defined (_POSIX_SOURCE)
/* BSD has one and only one pattern buffer. */
static struct re_pattern_buffer re_comp_buf;
char *
re_comp (s)
const char *s;
{
reg_errcode_t ret;
if (!s)
{
if (!re_comp_buf.buffer)
return "No previous regular expression";
return 0;
}
if (!re_comp_buf.buffer)
{
re_comp_buf.buffer = (unsigned char *) malloc (200);
if (re_comp_buf.buffer == NULL)
return "Memory exhausted";
re_comp_buf.allocated = 200;
re_comp_buf.fastmap = (char *) malloc (1 << BYTEWIDTH);
if (re_comp_buf.fastmap == NULL)
return "Memory exhausted";
}
/* Since `re_exec' always passes NULL for the `regs' argument, we
don't need to initialize the pattern buffer fields which affect it. */
/* Match anchors at newlines. */
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];
}
int
re_exec (s)
const char *s;
{
const int len = strlen (s);
return
0 <= re_search (&re_comp_buf, s, len, 0, len, (struct re_registers *) 0);
}
#endif /* not emacs and not _POSIX_SOURCE */
/* POSIX.2 functions. Don't define these for Emacs. */
#ifndef emacs
/* regcomp takes a regular expression as a string and compiles it.
PREG is a regex_t *. We do not expect any fields to be initialized,
since POSIX says we shouldn't. Thus, we set
`buffer' to the compiled pattern;
`used' to the length of the compiled pattern;
`syntax' to RE_SYNTAX_POSIX_EXTENDED if the
REG_EXTENDED bit in CFLAGS is set; otherwise, to
RE_SYNTAX_POSIX_BASIC;
`newline_anchor' to REG_NEWLINE being set in CFLAGS;
`fastmap' and `fastmap_accurate' to zero;
`re_nsub' to the number of subexpressions in PATTERN.
PATTERN is the address of the pattern string.
CFLAGS is a series of bits which affect compilation.
If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we
use POSIX basic syntax.
If REG_NEWLINE is set, then . and [^...] don't match newline.
Also, regexec will try a match beginning after every newline.
If REG_ICASE is set, then we considers upper- and lowercase
versions of letters to be equivalent when matching.
If REG_NOSUB is set, then when PREG is passed to regexec, that
routine will report only success or failure, and nothing about the
registers.
It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for
the return codes and their meanings.) */
int
regcomp (preg, pattern, cflags)
regex_t *preg;
const char *pattern;
int cflags;
{
reg_errcode_t ret;
unsigned syntax
= (cflags & REG_EXTENDED) ?
RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC;
/* regex_compile will allocate the space for the compiled pattern. */
preg->buffer = 0;
preg->allocated = 0;
/* 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);
if (preg->translate == NULL)
return (int) REG_ESPACE;
/* Map uppercase characters to corresponding lowercase ones. */
for (i = 0; i < CHAR_SET_SIZE; i++)
preg->translate[i] = ISUPPER (i) ? tolower (i) : i;
}
else
preg->translate = NULL;
/* If REG_NEWLINE is set, newlines are treated differently. */
if (cflags & REG_NEWLINE)
{ /* REG_NEWLINE implies neither . nor [^...] match newline. */
syntax &= ~RE_DOT_NEWLINE;
syntax |= RE_HAT_LISTS_NOT_NEWLINE;
/* It also changes the matching behavior. */
preg->newline_anchor = 1;
}
else
preg->newline_anchor = 0;
preg->no_sub = !!(cflags & REG_NOSUB);
/* 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[];
int eflags;
{
int ret;
struct re_registers regs;
regex_t private_preg;
int len = strlen (string);
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;
regs.start = TALLOC (nmatch, regoff_t);
regs.end = TALLOC (nmatch, regoff_t);
if (regs.start == NULL || regs.end == NULL)
return (int) REG_NOMATCH;
}
/* Perform the searching operation. */
ret = re_search (&private_preg, string, len,
/* start: */ 0, /* range: */ len,
want_reg_info ? &regs : (struct re_registers *) 0);
/* Copy the register information to the POSIX structure. */
if (want_reg_info)
{
if (ret >= 0)
{
unsigned r;
for (r = 0; r < nmatch; r++)
{
pmatch[r].rm_so = regs.start[r];
pmatch[r].rm_eo = regs.end[r];
}
}
/* If we needed the temporary register info, free the space now. */
free (regs.start);
free (regs.end);
}
/* We want zero return to mean success, unlike `re_search'. */
return ret >= 0 ? (int) REG_NOERROR : (int) REG_NOMATCH;
}
/* Returns a message corresponding to an error code, ERRCODE, returned
from either regcomp or regexec. We don't use PREG here. */
size_t
regerror (errcode, preg, errbuf, errbuf_size)
int errcode;
const regex_t *preg;
char *errbuf;
size_t errbuf_size;
{
const char *msg;
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
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_size = strlen (msg) + 1; /* Includes the null. */
if (errbuf_size != 0)
{
if (msg_size > errbuf_size)
{
strncpy (errbuf, msg, errbuf_size - 1);
errbuf[errbuf_size - 1] = 0;
}
else
strcpy (errbuf, msg);
}
return msg_size;
}
/* Free dynamically allocated space used by PREG. */
void
regfree (preg)
regex_t *preg;
{
if (preg->buffer != NULL)
free (preg->buffer);
preg->buffer = NULL;
preg->allocated = 0;
preg->used = 0;
if (preg->fastmap != NULL)
free (preg->fastmap);
preg->fastmap = NULL;
preg->fastmap_accurate = 0;
if (preg->translate != NULL)
free (preg->translate);
preg->translate = NULL;
}
#endif /* not emacs */
/*
Local variables:
make-backup-files: t
version-control: t
trim-versions-without-asking: nil
End:
*/
#endif