Embedded/pikapython
1
0
Fork 0
mirror of https://gitee.com/Lyon1998/pikapython.git synced 2025-01-22 17:12:55 +08:00
Code Issues Packages Projects Releases Wiki Activity
pikapython/port/linux/package/pikascript/pikascript-lib/re/pcre_exec.c
Lyon 0adbb62369 format and push pkgs, fix SOFT_IIC
2023-07-15 14:59:56 +08:00

5221 lines
214 KiB
C
Raw Blame History

/* This module contains pcre_exec(), the externally visible function that does
pattern matching using an NFA algorithm, trying to mimic Perl as closely as
possible. There are also some static supporting functions. */
#include "re_config.h"
#define NLBLOCK md /* Block containing newline information */
#define PSSTART start_subject /* Field containing processed string start */
#define PSEND end_subject /* Field containing processed string end */
#include "pcre_internal.h"
/* Undefine some potentially clashing cpp symbols */
#undef min
#undef max
/* Flag bits for the match() function */
#define match_condassert 0x01 /* Called to check a condition assertion */
#define match_cbegroup 0x02 /* Could-be-empty unlimited repeat group */
/* Non-error returns from the match() function. Error returns are externally
defined PCRE_ERROR_xxx codes, which are all negative. */
#define MATCH_MATCH 1
#define MATCH_NOMATCH 0
/* Special internal returns from the match() function. Make them sufficiently
negative to avoid the external error codes. */
#define MATCH_COMMIT (-999)
#define MATCH_PRUNE (-998)
#define MATCH_SKIP (-997)
#define MATCH_THEN (-996)
/* Maximum number of ints of offset to save on the stack for recursive calls.
If the offset vector is bigger, malloc is used. This should be a multiple of 3,
because the offset vector is always a multiple of 3 long. */
#define REC_STACK_SAVE_MAX 30
/* Min and max values for the common repeats; for the maxima, 0 => infinity */
static const char rep_min[] = {0, 0, 1, 1, 0, 0};
static const char rep_max[] = {0, 0, 0, 0, 1, 1};
#ifdef DEBUG
/*************************************************
* Debugging function to print chars *
*************************************************/
/* Print a sequence of chars in printable format, stopping at the end of the
subject if the requested.
Arguments:
p points to characters
length number to print
is_subject TRUE if printing from within md->start_subject
md pointer to matching data block, if is_subject is TRUE
Returns: nothing
*/
static void pchars(const uschar* p,
int length,
BOOL is_subject,
match_data* md) {
unsigned int c;
if (is_subject && length > md->end_subject - p)
length = md->end_subject - p;
while (length-- > 0)
if (isprint(c = *(p++)))
printf("%c", c);
else
printf("\\x%02x", c);
}
#endif
/*************************************************
* Match a back-reference *
*************************************************/
/* If a back reference hasn't been set, the length that is passed is greater
than the number of characters left in the string, so the match fails.
Arguments:
offset index into the offset vector
eptr points into the subject
length length to be matched
md points to match data block
ims the ims flags
Returns: TRUE if matched
*/
static BOOL match_ref(int offset,
register USPTR eptr,
int length,
match_data* md,
unsigned long int ims) {
USPTR p = md->start_subject + md->offset_vector[offset];
#ifdef DEBUG
if (eptr >= md->end_subject)
printf("matching subject <null>");
else {
printf("matching subject ");
pchars(eptr, length, TRUE, md);
}
printf(" against backref ");
pchars(p, length, FALSE, md);
printf("\n");
#endif
/* Always fail if not enough characters left */
if (length > md->end_subject - eptr)
return FALSE;
/* Separate the caselesss case for speed */
if ((ims & PCRE_CASELESS) != 0) {
while (length-- > 0)
if (md->lcc[*p++] != md->lcc[*eptr++])
return FALSE;
} else {
while (length-- > 0)
if (*p++ != *eptr++)
return FALSE;
}
return TRUE;
}
/***************************************************************************
****************************************************************************
RECURSION IN THE match() FUNCTION
The match() function is highly recursive, though not every recursive call
increases the recursive depth. Nevertheless, some regular expressions can cause
it to recurse to a great depth. I was writing for Unix, so I just let it call
itself recursively. This uses the stack for saving everything that has to be
saved for a recursive call. On Unix, the stack can be large, and this works
fine.
It turns out that on some non-Unix-like systems there are problems with
programs that use a lot of stack. (This despite the fact that every last chip
has oodles of memory these days, and techniques for extending the stack have
been known for decades.) So....
There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
calls by keeping local variables that need to be preserved in blocks of memory
obtained from malloc() instead instead of on the stack. Macros are used to
achieve this so that the actual code doesn't look very different to what it
always used to.
The original heap-recursive code used longjmp(). However, it seems that this
can be very slow on some operating systems. Following a suggestion from Stan
Switzer, the use of longjmp() has been abolished, at the cost of having to
provide a unique number for each call to RMATCH. There is no way of generating
a sequence of numbers at compile time in C. I have given them names, to make
them stand out more clearly.
Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
tests. Furthermore, not using longjmp() means that local dynamic variables
don't have indeterminate values; this has meant that the frame size can be
reduced because the result can be "passed back" by straight setting of the
variable instead of being passed in the frame.
****************************************************************************
***************************************************************************/
/* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
below must be updated in sync. */
enum {
RM1 = 1,
RM2,
RM3,
RM4,
RM5,
RM6,
RM7,
RM8,
RM9,
RM10,
RM11,
RM12,
RM13,
RM14,
RM15,
RM16,
RM17,
RM18,
RM19,
RM20,
RM21,
RM22,
RM23,
RM24,
RM25,
RM26,
RM27,
RM28,
RM29,
RM30,
RM31,
RM32,
RM33,
RM34,
RM35,
RM36,
RM37,
RM38,
RM39,
RM40,
RM41,
RM42,
RM43,
RM44,
RM45,
RM46,
RM47,
RM48,
RM49,
RM50,
RM51,
RM52,
RM53,
RM54
};
/* These versions of the macros use the stack, as normal. There are debugging
versions and production versions. Note that the "rw" argument of RMATCH isn't
actuall used in this definition. */
#ifndef NO_RECURSE
#define REGISTER register
#ifdef DEBUG
#define RMATCH(ra, rb, rc, rd, re, rf, rg, rw) \
{ \
printf("match() called in line %d\n", __LINE__); \
rrc = match(ra, rb, mstart, rc, rd, re, rf, rg, rdepth + 1); \
printf("to line %d\n", __LINE__); \
}
#define RRETURN(ra) \
{ \
printf("match() returned %d from line %d ", ra, __LINE__); \
return ra; \
}
#else
#define RMATCH(ra, rb, rc, rd, re, rf, rg, rw) \
rrc = match(ra, rb, mstart, rc, rd, re, rf, rg, rdepth + 1)
#define RRETURN(ra) return ra
#endif
#else
/* These versions of the macros manage a private stack on the heap. Note that
the "rd" argument of RMATCH isn't actually used in this definition. It's the md
argument of match(), which never changes. */
#define REGISTER
#define RMATCH(ra, rb, rc, rd, re, rf, rg, rw) \
{ \
heapframe* newframe = (pcre_stack_malloc)(sizeof(heapframe)); \
frame->Xwhere = rw; \
newframe->Xeptr = ra; \
newframe->Xecode = rb; \
newframe->Xmstart = mstart; \
newframe->Xoffset_top = rc; \
newframe->Xims = re; \
newframe->Xeptrb = rf; \
newframe->Xflags = rg; \
newframe->Xrdepth = frame->Xrdepth + 1; \
newframe->Xprevframe = frame; \
frame = newframe; \
DPRINTF(("restarting from line %d\n", __LINE__)); \
goto HEAP_RECURSE; \
L_##rw : DPRINTF(("jumped back to line %d\n", __LINE__)); \
}
#define RRETURN(ra) \
{ \
heapframe* newframe = frame; \
frame = newframe->Xprevframe; \
(pcre_stack_free)(newframe); \
if (frame != NULL) { \
rrc = ra; \
goto HEAP_RETURN; \
} \
return ra; \
}
/* Structure for remembering the local variables in a private frame */
typedef struct heapframe {
struct heapframe* Xprevframe;
/* Function arguments that may change */
const uschar* Xeptr;
const uschar* Xecode;
const uschar* Xmstart;
int Xoffset_top;
long int Xims;
eptrblock* Xeptrb;
int Xflags;
unsigned int Xrdepth;
/* Function local variables */
const uschar* Xcallpat;
const uschar* Xcharptr;
const uschar* Xdata;
const uschar* Xnext;
const uschar* Xpp;
const uschar* Xprev;
const uschar* Xsaved_eptr;
recursion_info Xnew_recursive;
BOOL Xcur_is_word;
BOOL Xcondition;
BOOL Xprev_is_word;
unsigned long int Xoriginal_ims;
#ifdef SUPPORT_UCP
int Xprop_type;
int Xprop_value;
int Xprop_fail_result;
int Xprop_category;
int Xprop_chartype;
int Xprop_script;
int Xoclength;
uschar Xocchars[8];
#endif
int Xctype;
unsigned int Xfc;
int Xfi;
int Xlength;
int Xmax;
int Xmin;
int Xnumber;
int Xoffset;
int Xop;
int Xsave_capture_last;
int Xsave_offset1, Xsave_offset2, Xsave_offset3;
int Xstacksave[REC_STACK_SAVE_MAX];
eptrblock Xnewptrb;
/* Where to jump back to */
int Xwhere;
} heapframe;
#endif
/***************************************************************************
***************************************************************************/
/*************************************************
* Match from current position *
*************************************************/
/* This function is called recursively in many circumstances. Whenever it
returns a negative (error) response, the outer incarnation must also return the
same response.
Performance note: It might be tempting to extract commonly used fields from the
md structure (e.g. utf8, end_subject) into individual variables to improve
performance. Tests using gcc on a SPARC disproved this; in the first case, it
made performance worse.
Arguments:
eptr pointer to current character in subject
ecode pointer to current position in compiled code
mstart pointer to the current match start position (can be modified
by encountering \K)
offset_top current top pointer
md pointer to "static" info for the match
ims current /i, /m, and /s options
eptrb pointer to chain of blocks containing eptr at start of
brackets - for testing for empty matches
flags can contain
match_condassert - this is an assertion condition
match_cbegroup - this is the start of an unlimited repeat
group that can match an empty string
rdepth the recursion depth
Returns: MATCH_MATCH if matched ) these values are >= 0
MATCH_NOMATCH if failed to match )
a negative PCRE_ERROR_xxx value if aborted by an error condition
(e.g. stopped by repeated call or recursion limit)
*/
static int match(REGISTER USPTR eptr,
REGISTER const uschar* ecode,
const uschar* mstart,
int offset_top,
match_data* md,
unsigned long int ims,
eptrblock* eptrb,
int flags,
unsigned int rdepth) {
/* These variables do not need to be preserved over recursion in this
function, so they can be ordinary variables in all cases. Mark some of them
with "register" because they are used a lot in loops. */
register int rrc; /* Returns from recursive calls */
register int i; /* Used for loops not involving calls to RMATCH() */
register unsigned int c; /* Character values not kept over RMATCH() calls */
register BOOL utf8; /* Local copy of UTF-8 flag for speed */
BOOL minimize, possessive; /* Quantifier options */
/* When recursion is not being used, all "local" variables that have to be
preserved over calls to RMATCH() are part of a "frame" which is obtained
from heap storage. Set up the top-level frame here; others are obtained from
the heap whenever RMATCH() does a "recursion". See the macro definitions
above. */
#ifdef NO_RECURSE
heapframe* frame = (pcre_stack_malloc)(sizeof(heapframe));
frame->Xprevframe = NULL; /* Marks the top level */
/* Copy in the original argument variables */
frame->Xeptr = eptr;
frame->Xecode = ecode;
frame->Xmstart = mstart;
frame->Xoffset_top = offset_top;
frame->Xims = ims;
frame->Xeptrb = eptrb;
frame->Xflags = flags;
frame->Xrdepth = rdepth;
/* This is where control jumps back to to effect "recursion" */
HEAP_RECURSE:
/* Macros make the argument variables come from the current frame */
#define eptr frame->Xeptr
#define ecode frame->Xecode
#define mstart frame->Xmstart
#define offset_top frame->Xoffset_top
#define ims frame->Xims
#define eptrb frame->Xeptrb
#define flags frame->Xflags
#define rdepth frame->Xrdepth
/* Ditto for the local variables */
#ifdef SUPPORT_UTF8
#define charptr frame->Xcharptr
#endif
#define callpat frame->Xcallpat
#define data frame->Xdata
#define next frame->Xnext
#define pp frame->Xpp
#define prev frame->Xprev
#define saved_eptr frame->Xsaved_eptr
#define new_recursive frame->Xnew_recursive
#define cur_is_word frame->Xcur_is_word
#define condition frame->Xcondition
#define prev_is_word frame->Xprev_is_word
#define original_ims frame->Xoriginal_ims
#ifdef SUPPORT_UCP
#define prop_type frame->Xprop_type
#define prop_value frame->Xprop_value
#define prop_fail_result frame->Xprop_fail_result
#define prop_category frame->Xprop_category
#define prop_chartype frame->Xprop_chartype
#define prop_script frame->Xprop_script
#define oclength frame->Xoclength
#define occhars frame->Xocchars
#endif
#define ctype frame->Xctype
#define fc frame->Xfc
#define fi frame->Xfi
#define length frame->Xlength
#define max frame->Xmax
#define min frame->Xmin
#define number frame->Xnumber
#define offset frame->Xoffset
#define op frame->Xop
#define save_capture_last frame->Xsave_capture_last
#define save_offset1 frame->Xsave_offset1
#define save_offset2 frame->Xsave_offset2
#define save_offset3 frame->Xsave_offset3
#define stacksave frame->Xstacksave
#define newptrb frame->Xnewptrb
/* When recursion is being used, local variables are allocated on the stack
and get preserved during recursion in the normal way. In this environment,
fi and i, and fc and c, can be the same variables. */
#else /* NO_RECURSE not defined */
#define fi i
#define fc c
#ifdef SUPPORT_UTF8 /* Many of these variables are used only */
const uschar* charptr; /* in small blocks of the code. My normal */
#endif /* style of coding would have declared */
const uschar* callpat; /* them within each of those blocks. */
const uschar* data; /* However, in order to accommodate the */
const uschar* next; /* version of this code that uses an */
USPTR pp; /* external "stack" implemented on the */
const uschar* prev; /* heap, it is easier to declare them all */
USPTR saved_eptr; /* here, so the declarations can be cut */
/* out in a block. The only declarations */
recursion_info new_recursive; /* within blocks below are for variables */
/* that do not have to be preserved over */
BOOL cur_is_word; /* a recursive call to RMATCH(). */
BOOL condition;
BOOL prev_is_word;
unsigned long int original_ims;
#ifdef SUPPORT_UCP
int prop_type;
int prop_value;
int prop_fail_result;
int prop_category;
int prop_chartype;
int prop_script;
int oclength;
uschar occhars[8];
#endif
int ctype;
int length;
int max;
int min;
int number;
int offset;
int op;
int save_capture_last;
int save_offset1, save_offset2, save_offset3;
int stacksave[REC_STACK_SAVE_MAX];
eptrblock newptrb;
#endif /* NO_RECURSE */
/* These statements are here to stop the compiler complaining about
unitialized variables. */
#ifdef SUPPORT_UCP
prop_value = 0;
prop_fail_result = 0;
#endif
/* This label is used for tail recursion, which is used in a few cases even
when NO_RECURSE is not defined, in order to reduce the amount of stack that
is used. Thanks to Ian Taylor for noticing this possibility and sending the
original patch. */
TAIL_RECURSE:
/* OK, now we can get on with the real code of the function. Recursive calls
are specified by the macro RMATCH and RRETURN is used to return. When
NO_RECURSE is *not* defined, these just turn into a recursive call to
match() and a "return", respectively (possibly with some debugging if DEBUG
is defined). However, RMATCH isn't like a function call because it's quite a
complicated macro. It has to be used in one particular way. This shouldn't,
however, impact performance when true recursion is being used. */
#ifdef SUPPORT_UTF8
utf8 = md->utf8; /* Local copy of the flag */
#else
utf8 = FALSE;
#endif
/* First check that we haven't called match() too many times, or that we
haven't exceeded the recursive call limit. */
if (md->match_call_count++ >= md->match_limit)
RRETURN(PCRE_ERROR_MATCHLIMIT);
if (rdepth >= md->match_limit_recursion)
RRETURN(PCRE_ERROR_RECURSIONLIMIT);
original_ims = ims; /* Save for resetting on ')' */
/* At the start of a group with an unlimited repeat that may match an empty
string, the match_cbegroup flag is set. When this is the case, add the
current subject pointer to the chain of such remembered pointers, to be
checked when we hit the closing ket, in order to break infinite loops that
match no characters. When match() is called in other circumstances, don't
add to the chain. The match_cbegroup flag must NOT be used with tail
recursion, because the memory block that is used is on the stack, so a new
one may be required for each match(). */
if ((flags & match_cbegroup) != 0) {
newptrb.epb_saved_eptr = eptr;
newptrb.epb_prev = eptrb;
eptrb = &newptrb;
}
/* Now start processing the opcodes. */
for (;;) {
minimize = possessive = FALSE;
op = *ecode;
/* For partial matching, remember if we ever hit the end of the subject
after matching at least one subject character. */
if (md->partial && eptr >= md->end_subject && eptr > mstart)
md->hitend = TRUE;
switch (op) {
case OP_FAIL:
RRETURN(MATCH_NOMATCH);
case OP_PRUNE:
RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
ims, eptrb, flags, RM51);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
RRETURN(MATCH_PRUNE);
case OP_COMMIT:
RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
ims, eptrb, flags, RM52);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
RRETURN(MATCH_COMMIT);
case OP_SKIP:
RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
ims, eptrb, flags, RM53);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
md->start_match_ptr = eptr; /* Pass back current position */
RRETURN(MATCH_SKIP);
case OP_THEN:
RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
ims, eptrb, flags, RM54);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
RRETURN(MATCH_THEN);
/* Handle a capturing bracket. If there is space in the offset
vector, save the current subject position in the working slot at
the top of the vector. We mustn't change the current values of
the data slot, because they may be set from a previous iteration
of this group, and be referred to by a reference inside the
group.
If the bracket fails to match, we need to restore this value and
also the values of the final offsets, in case they were set by a
previous iteration of the same bracket.
If there isn't enough space in the offset vector, treat this as
if it were a non-capturing bracket. Don't worry about setting
the flag for the error case here; that is handled in the code
for KET. */
case OP_CBRA:
case OP_SCBRA:
number = GET2(ecode, 1 + LINK_SIZE);
offset = number << 1;
#ifdef DEBUG
printf("start bracket %d\n", number);
printf("subject=");
pchars(eptr, 16, TRUE, md);
printf("\n");
#endif
if (offset < md->offset_max) {
save_offset1 = md->offset_vector[offset];
save_offset2 = md->offset_vector[offset + 1];
save_offset3 = md->offset_vector[md->offset_end - number];
save_capture_last = md->capture_last;
DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2,
save_offset3));
md->offset_vector[md->offset_end - number] =
eptr - md->start_subject;
flags = (op == OP_SCBRA) ? match_cbegroup : 0;
do {
RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode],
offset_top, md, ims, eptrb, flags, RM1);
if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
RRETURN(rrc);
md->capture_last = save_capture_last;
ecode += GET(ecode, 1);
} while (*ecode == OP_ALT);
DPRINTF(("bracket %d failed\n", number));
md->offset_vector[offset] = save_offset1;
md->offset_vector[offset + 1] = save_offset2;
md->offset_vector[md->offset_end - number] = save_offset3;
RRETURN(MATCH_NOMATCH);
}
/* FALL THROUGH ... Insufficient room for saving captured
contents. Treat as a non-capturing bracket. */
/* VVVVVVVVVVVVVVVVVVVVVVVVV */
/* VVVVVVVVVVVVVVVVVVVVVVVVV */
DPRINTF(
("insufficient capture room: treat as non-capturing\n"));
/* VVVVVVVVVVVVVVVVVVVVVVVVV */
/* VVVVVVVVVVVVVVVVVVVVVVVVV */
/* Non-capturing bracket. Loop for all the alternatives. When we
get to the final alternative within the brackets, we would
return the result of a recursive call to match() whatever
happened. We can reduce stack usage by turning this into a tail
recursion, except in the case when match_cbegroup is set.*/
case OP_BRA:
case OP_SBRA:
DPRINTF(("start non-capturing bracket\n"));
flags = (op >= OP_SBRA) ? match_cbegroup : 0;
for (;;) {
if (ecode[GET(ecode, 1)] != OP_ALT) /* Final alternative */
{
if (flags == 0) /* Not a possibly empty group */
{
ecode += _pcre_OP_lengths[*ecode];
DPRINTF(("bracket 0 tail recursion\n"));
goto TAIL_RECURSE;
}
/* Possibly empty group; can't use tail recursion. */
RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode],
offset_top, md, ims, eptrb, flags, RM48);
RRETURN(rrc);
}
/* For non-final alternatives, continue the loop for a
NOMATCH result; otherwise return. */
RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top,
md, ims, eptrb, flags, RM2);
if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
RRETURN(rrc);
ecode += GET(ecode, 1);
}
/* Control never reaches here. */
/* Conditional group: compilation checked that there are no more
than two branches. If the condition is false, skipping the first
branch takes us past the end if there is only one branch, but
that's OK because that is exactly what going to the ket would
do. As there is only one branch to be obeyed, we can use tail
recursion to avoid using another stack frame. */
case OP_COND:
case OP_SCOND:
if (ecode[LINK_SIZE + 1] == OP_RREF) /* Recursion test */
{
offset =
GET2(ecode, LINK_SIZE + 2); /* Recursion group number*/
condition = md->recursive != NULL &&
(offset == RREF_ANY ||
offset == md->recursive->group_num);
ecode += condition ? 3 : GET(ecode, 1);
}
else if (ecode[LINK_SIZE + 1] == OP_CREF) /* Group used test */
{
offset = GET2(ecode, LINK_SIZE + 2)
<< 1; /* Doubled ref number */
condition =
offset < offset_top && md->offset_vector[offset] >= 0;
ecode += condition ? 3 : GET(ecode, 1);
}
else if (ecode[LINK_SIZE + 1] ==
OP_DEF) /* DEFINE - always false */
{
condition = FALSE;
ecode += GET(ecode, 1);
}
/* The condition is an assertion. Call match() to evaluate it -
setting the final argument match_condassert causes it to stop at
the end of an assertion. */
else {
RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,
NULL, match_condassert, RM3);
if (rrc == MATCH_MATCH) {
condition = TRUE;
ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
while (*ecode == OP_ALT)
ecode += GET(ecode, 1);
} else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) {
RRETURN(
rrc); /* Need braces because of following else */
} else {
condition = FALSE;
ecode += GET(ecode, 1);
}
}
/* We are now at the branch that is to be obeyed. As there is
only one, we can use tail recursion to avoid using another stack
frame, except when match_cbegroup is required for an unlimited
repeat of a possibly empty group. If the second alternative
doesn't exist, we can just plough on. */
if (condition || *ecode == OP_ALT) {
ecode += 1 + LINK_SIZE;
if (op == OP_SCOND) /* Possibly empty group */
{
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
match_cbegroup, RM49);
RRETURN(rrc);
} else /* Group must match something */
{
flags = 0;
goto TAIL_RECURSE;
}
} else /* Condition false & no 2nd alternative */
{
ecode += 1 + LINK_SIZE;
}
break;
/* End of the pattern, either real or forced. If we are in a
top-level recursion, we should restore the offsets appropriately
and continue from after the call. */
case OP_ACCEPT:
case OP_END:
if (md->recursive != NULL && md->recursive->group_num == 0) {
recursion_info* rec = md->recursive;
DPRINTF(("End of pattern in a (?0) recursion\n"));
md->recursive = rec->prevrec;
memmove(md->offset_vector, rec->offset_save,
rec->saved_max * sizeof(int));
mstart = rec->save_start;
ims = original_ims;
ecode = rec->after_call;
break;
}
/* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched
an empty string - backtracking will then try other alternatives,
if any. */
if (md->notempty && eptr == mstart)
RRETURN(MATCH_NOMATCH);
md->end_match_ptr = eptr; /* Record where we ended */
md->end_offset_top =
offset_top; /* and how many extracts were taken */
md->start_match_ptr =
mstart; /* and the start (\K can modify) */
RRETURN(MATCH_MATCH);
/* Change option settings */
case OP_OPT:
ims = ecode[1];
ecode += 2;
DPRINTF(("ims set to %02lx\n", ims));
break;
/* Assertion brackets. Check the alternative branches in turn -
the matching won't pass the KET for an assertion. If any one
branch matches, the assertion is true. Lookbehind assertions
have an OP_REVERSE item at the start of each branch to move the
current point backwards, so the code at this level is identical
to the lookahead case. */
case OP_ASSERT:
case OP_ASSERTBACK:
do {
RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,
NULL, 0, RM4);
if (rrc == MATCH_MATCH)
break;
if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
RRETURN(rrc);
ecode += GET(ecode, 1);
} while (*ecode == OP_ALT);
if (*ecode == OP_KET)
RRETURN(MATCH_NOMATCH);
/* If checking an assertion for a condition, return MATCH_MATCH.
*/
if ((flags & match_condassert) != 0)
RRETURN(MATCH_MATCH);
/* Continue from after the assertion, updating the offsets high
water mark, since extracts may have been taken during the
assertion. */
do
ecode += GET(ecode, 1);
while (*ecode == OP_ALT);
ecode += 1 + LINK_SIZE;
offset_top = md->end_offset_top;
continue;
/* Negative assertion: all branches must fail to match */
case OP_ASSERT_NOT:
case OP_ASSERTBACK_NOT:
do {
RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,
NULL, 0, RM5);
if (rrc == MATCH_MATCH)
RRETURN(MATCH_NOMATCH);
if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
RRETURN(rrc);
ecode += GET(ecode, 1);
} while (*ecode == OP_ALT);
if ((flags & match_condassert) != 0)
RRETURN(MATCH_MATCH);
ecode += 1 + LINK_SIZE;
continue;
/* Move the subject pointer back. This occurs only at the start
of each branch of a lookbehind assertion. If we are too close to
the start to move back, this match function fails. When working
with UTF-8 we move back a number of characters, not bytes. */
case OP_REVERSE:
#ifdef SUPPORT_UTF8
if (utf8) {
i = GET(ecode, 1);
while (i-- > 0) {
eptr--;
if (eptr < md->start_subject)
RRETURN(MATCH_NOMATCH);
BACKCHAR(eptr);
}
} else
#endif
/* No UTF-8 support, or not in UTF-8 mode: count is byte count
*/
{
eptr -= GET(ecode, 1);
if (eptr < md->start_subject)
RRETURN(MATCH_NOMATCH);
}
/* Skip to next op code */
ecode += 1 + LINK_SIZE;
break;
/* The callout item calls an external function, if one is
provided, passing details of the match so far. This is mainly
for debugging, though the function is able to force a failure.
*/
case OP_CALLOUT:
if (pcre_callout != NULL) {
pcre_callout_block cb;
cb.version = 1; /* Version 1 of the callout block */
cb.callout_number = ecode[1];
cb.offset_vector = md->offset_vector;
cb.subject = (PCRE_SPTR)md->start_subject;
cb.subject_length = md->end_subject - md->start_subject;
cb.start_match = mstart - md->start_subject;
cb.current_position = eptr - md->start_subject;
cb.pattern_position = GET(ecode, 2);
cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
cb.capture_top = offset_top / 2;
cb.capture_last = md->capture_last;
cb.callout_data = md->callout_data;
if ((rrc = (*pcre_callout)(&cb)) > 0)
RRETURN(MATCH_NOMATCH);
if (rrc < 0)
RRETURN(rrc);
}
ecode += 2 + 2 * LINK_SIZE;
break;
/* Recursion either matches the current regex, or some
subexpression. The offset data is the offset to the starting
bracket from the start of the whole pattern. (This is so that it
works from duplicated subpatterns.)
If there are any capturing brackets started but not finished, we
have to save their starting points and reinstate them after the
recursion. However, we don't know how many such there are
(offset_top records the completed total) so we just have to save
all the potential data. There may be up to 65535 such values,
which is too large to put on the stack, but using malloc for
small numbers seems expensive. As a compromise, the stack is
used when there are no more than REC_STACK_SAVE_MAX values to
store; otherwise malloc is used. A problem is what to do if the
malloc fails ... there is no way of returning to the top level
with an error. Save the top REC_STACK_SAVE_MAX values on the
stack, and accept that the rest may be wrong.
There are also other values that have to be saved. We use a
chained sequence of blocks that actually live on the stack.
Thanks to Robin Houston for the original version of this logic.
*/
case OP_RECURSE: {
callpat = md->start_code + GET(ecode, 1);
new_recursive.group_num = (callpat == md->start_code)
? 0
: GET2(callpat, 1 + LINK_SIZE);
/* Add to "recursing stack" */
new_recursive.prevrec = md->recursive;
md->recursive = &new_recursive;
/* Find where to continue from afterwards */
ecode += 1 + LINK_SIZE;
new_recursive.after_call = ecode;
/* Now save the offset data. */
new_recursive.saved_max = md->offset_end;
if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
new_recursive.offset_save = stacksave;
else {
new_recursive.offset_save =
(int*)(pcre_malloc)(new_recursive.saved_max *
sizeof(int));
if (new_recursive.offset_save == NULL)
RRETURN(PCRE_ERROR_NOMEMORY);
}
memcpy(new_recursive.offset_save, md->offset_vector,
new_recursive.saved_max * sizeof(int));
new_recursive.save_start = mstart;
mstart = eptr;
/* OK, now we can do the recursion. For each top-level
alternative we restore the offset and recursion data. */
DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
flags = (*callpat >= OP_SBRA) ? match_cbegroup : 0;
do {
RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat],
offset_top, md, ims, eptrb, flags, RM6);
if (rrc == MATCH_MATCH) {
DPRINTF(("Recursion matched\n"));
md->recursive = new_recursive.prevrec;
if (new_recursive.offset_save != stacksave)
(pcre_free)(new_recursive.offset_save);
RRETURN(MATCH_MATCH);
} else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) {
DPRINTF(("Recursion gave error %d\n", rrc));
RRETURN(rrc);
}
md->recursive = &new_recursive;
memcpy(md->offset_vector, new_recursive.offset_save,
new_recursive.saved_max * sizeof(int));
callpat += GET(callpat, 1);
} while (*callpat == OP_ALT);
DPRINTF(("Recursion didn't match\n"));
md->recursive = new_recursive.prevrec;
if (new_recursive.offset_save != stacksave)
(pcre_free)(new_recursive.offset_save);
RRETURN(MATCH_NOMATCH);
}
/* Control never reaches here */
/* "Once" brackets are like assertion brackets except that after
a match, the point in the subject string is not moved back. Thus
there can never be a move back into the brackets. Friedl calls
these "atomic" subpatterns. Check the alternative branches in
turn - the matching won't pass the KET for this kind of
subpattern. If any one branch matches, we carry on as at the end
of a normal bracket, leaving the subject pointer. */
case OP_ONCE:
prev = ecode;
saved_eptr = eptr;
do {
RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,
eptrb, 0, RM7);
if (rrc == MATCH_MATCH)
break;
if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
RRETURN(rrc);
ecode += GET(ecode, 1);
} while (*ecode == OP_ALT);
/* If hit the end of the group (which could be repeated), fail
*/
if (*ecode != OP_ONCE && *ecode != OP_ALT)
RRETURN(MATCH_NOMATCH);
/* Continue as from after the assertion, updating the offsets
high water mark, since extracts may have been taken. */
do
ecode += GET(ecode, 1);
while (*ecode == OP_ALT);
offset_top = md->end_offset_top;
eptr = md->end_match_ptr;
/* For a non-repeating ket, just continue at this level. This
also happens for a repeating ket if no characters were matched
in the group. This is the forcible breaking of infinite loops as
implemented in Perl 5.005. If there is an options reset, it will
get obeyed in the normal course of events. */
if (*ecode == OP_KET || eptr == saved_eptr) {
ecode += 1 + LINK_SIZE;
break;
}
/* The repeating kets try the rest of the pattern or restart
from the preceding bracket, in the appropriate order. The second
"call" of match() uses tail recursion, to avoid using another
stack frame. We need to reset any options that changed within
the bracket before re-running it, so check the next opcode. */
if (ecode[1 + LINK_SIZE] == OP_OPT) {
ims = (ims & ~PCRE_IMS) | ecode[4];
DPRINTF(("ims set to %02lx at group repeat\n", ims));
}
if (*ecode == OP_KETRMIN) {
RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,
eptrb, 0, RM8);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
ecode = prev;
flags = 0;
goto TAIL_RECURSE;
} else /* OP_KETRMAX */
{
RMATCH(eptr, prev, offset_top, md, ims, eptrb,
match_cbegroup, RM9);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
ecode += 1 + LINK_SIZE;
flags = 0;
goto TAIL_RECURSE;
}
/* Control never gets here */
/* An alternation is the end of a branch; scan along to find the
end of the bracketed group and go to there. */
case OP_ALT:
do
ecode += GET(ecode, 1);
while (*ecode == OP_ALT);
break;
/* BRAZERO and BRAMINZERO occur just before a bracket group,
indicating that it may occur zero times. It may repeat
infinitely, or not at all - i.e. it could be ()* or ()? in the
pattern. Brackets with fixed upper repeat limits are compiled as
a number of copies, with the optional ones preceded by BRAZERO
or BRAMINZERO. */
case OP_BRAZERO: {
next = ecode + 1;
RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
do
next += GET(next, 1);
while (*next == OP_ALT);
ecode = next + 1 + LINK_SIZE;
} break;
case OP_BRAMINZERO: {
next = ecode + 1;
do
next += GET(next, 1);
while (*next == OP_ALT);
RMATCH(eptr, next + 1 + LINK_SIZE, offset_top, md, ims, eptrb,
0, RM11);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
ecode++;
} break;
/* End of a group, repeated or non-repeating. */
case OP_KET:
case OP_KETRMIN:
case OP_KETRMAX:
prev = ecode - GET(ecode, 1);
/* If this was a group that remembered the subject start, in
order to break infinite repeats of empty string matches,
retrieve the subject start from the chain. Otherwise, set it
NULL. */
if (*prev >= OP_SBRA) {
saved_eptr =
eptrb->epb_saved_eptr; /* Value at start of group */
eptrb = eptrb->epb_prev; /* Backup to previous group */
} else
saved_eptr = NULL;
/* If we are at the end of an assertion group, stop matching and
return MATCH_MATCH, but record the current high water mark for
use by positive assertions. Do this also for the "once" (atomic)
groups. */
if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
*prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||
*prev == OP_ONCE) {
md->end_match_ptr = eptr; /* For ONCE */
md->end_offset_top = offset_top;
RRETURN(MATCH_MATCH);
}
/* For capturing groups we have to check the group number back
at the start and if necessary complete handling an extraction by
setting the offsets and bumping the high water mark. Note that
whole-pattern recursion is coded as a recurse into group 0, so
it won't be picked up here. Instead, we catch it when the OP_END
is reached. Other recursion is handled here. */
if (*prev == OP_CBRA || *prev == OP_SCBRA) {
number = GET2(prev, 1 + LINK_SIZE);
offset = number << 1;
#ifdef DEBUG
printf("end bracket %d", number);
printf("\n");
#endif
md->capture_last = number;
if (offset >= md->offset_max)
md->offset_overflow = TRUE;
else {
md->offset_vector[offset] =
md->offset_vector[md->offset_end - number];
md->offset_vector[offset + 1] =
eptr - md->start_subject;
if (offset_top <= offset)
offset_top = offset + 2;
}
/* Handle a recursively called group. Restore the offsets
appropriately and continue from after the call. */
if (md->recursive != NULL &&
md->recursive->group_num == number) {
recursion_info* rec = md->recursive;
DPRINTF(("Recursion (%d) succeeded - continuing\n",
number));
md->recursive = rec->prevrec;
mstart = rec->save_start;
memcpy(md->offset_vector, rec->offset_save,
rec->saved_max * sizeof(int));
ecode = rec->after_call;
ims = original_ims;
break;
}
}
/* For both capturing and non-capturing groups, reset the value
of the ims flags, in case they got changed during the group. */
ims = original_ims;
DPRINTF(("ims reset to %02lx\n", ims));
/* For a non-repeating ket, just continue at this level. This
also happens for a repeating ket if no characters were matched
in the group. This is the forcible breaking of infinite loops as
implemented in Perl 5.005. If there is an options reset, it will
get obeyed in the normal course of events. */
if (*ecode == OP_KET || eptr == saved_eptr) {
ecode += 1 + LINK_SIZE;
break;
}
/* The repeating kets try the rest of the pattern or restart
from the preceding bracket, in the appropriate order. In the
second case, we can use tail recursion to avoid using another
stack frame, unless we have an unlimited repeat of a group that
can match an empty string. */
flags = (*prev >= OP_SBRA) ? match_cbegroup : 0;
if (*ecode == OP_KETRMIN) {
RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,
eptrb, 0, RM12);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (flags != 0) /* Could match an empty string */
{
RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags,
RM50);
RRETURN(rrc);
}
ecode = prev;
goto TAIL_RECURSE;
} else /* OP_KETRMAX */
{
RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
ecode += 1 + LINK_SIZE;
flags = 0;
goto TAIL_RECURSE;
}
/* Control never gets here */
/* Start of subject unless notbol, or after internal newline if
* multiline */
case OP_CIRC:
if (md->notbol && eptr == md->start_subject)
RRETURN(MATCH_NOMATCH);
if ((ims & PCRE_MULTILINE) != 0) {
if (eptr != md->start_subject &&
(eptr == md->end_subject || !WAS_NEWLINE(eptr)))
RRETURN(MATCH_NOMATCH);
ecode++;
break;
}
/* ... else fall through */
/* Start of subject assertion */
case OP_SOD:
if (eptr != md->start_subject)
RRETURN(MATCH_NOMATCH);
ecode++;
break;
/* Start of match assertion */
case OP_SOM:
if (eptr != md->start_subject + md->start_offset)
RRETURN(MATCH_NOMATCH);
ecode++;
break;
/* Reset the start of match point */
case OP_SET_SOM:
mstart = eptr;
ecode++;
break;
/* Assert before internal newline if multiline, or before a
terminating newline unless endonly is set, else end of subject
unless noteol is set. */
case OP_DOLL:
if ((ims & PCRE_MULTILINE) != 0) {
if (eptr < md->end_subject) {
if (!IS_NEWLINE(eptr))
RRETURN(MATCH_NOMATCH);
} else {
if (md->noteol)
RRETURN(MATCH_NOMATCH);
}
ecode++;
break;
} else {
if (md->noteol)
RRETURN(MATCH_NOMATCH);
if (!md->endonly) {
if (eptr != md->end_subject &&
(!IS_NEWLINE(eptr) ||
eptr != md->end_subject - md->nllen))
RRETURN(MATCH_NOMATCH);
ecode++;
break;
}
}
/* ... else fall through for endonly */
/* End of subject assertion (\z) */
case OP_EOD:
if (eptr < md->end_subject)
RRETURN(MATCH_NOMATCH);
ecode++;
break;
/* End of subject or ending \n assertion (\Z) */
case OP_EODN:
if (eptr != md->end_subject &&
(!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
RRETURN(MATCH_NOMATCH);
ecode++;
break;
/* Word boundary assertions */
case OP_NOT_WORD_BOUNDARY:
case OP_WORD_BOUNDARY: {
/* Find out if the previous and current characters are "word"
characters. It takes a bit more work in UTF-8 mode. Characters >
255 are assumed to be "non-word" characters. */
#ifdef SUPPORT_UTF8
if (utf8) {
if (eptr == md->start_subject)
prev_is_word = FALSE;
else {
const uschar* lastptr = eptr - 1;
while ((*lastptr & 0xc0) == 0x80)
lastptr--;
GETCHAR(c, lastptr);
prev_is_word =
c < 256 && (md->ctypes[c] & ctype_word) != 0;
}
if (eptr >= md->end_subject)
cur_is_word = FALSE;
else {
GETCHAR(c, eptr);
cur_is_word =
c < 256 && (md->ctypes[c] & ctype_word) != 0;
}
} else
#endif
/* More streamlined when not in UTF-8 mode */
{
prev_is_word = (eptr != md->start_subject) &&
((md->ctypes[eptr[-1]] & ctype_word) != 0);
cur_is_word = (eptr < md->end_subject) &&
((md->ctypes[*eptr] & ctype_word) != 0);
}
/* Now see if the situation is what we want */
if ((*ecode++ == OP_WORD_BOUNDARY)
? cur_is_word == prev_is_word
: cur_is_word != prev_is_word)
RRETURN(MATCH_NOMATCH);
} break;
/* Match a single character type; inline for speed */
case OP_ANY:
if ((ims & PCRE_DOTALL) == 0) {
if (IS_NEWLINE(eptr))
RRETURN(MATCH_NOMATCH);
}
if (eptr++ >= md->end_subject)
RRETURN(MATCH_NOMATCH);
if (utf8)
while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80)
eptr++;
ecode++;
break;
/* Match a single byte, even in UTF-8 mode. This opcode really
does match any byte, even newline, independent of the setting of
PCRE_DOTALL. */
case OP_ANYBYTE:
if (eptr++ >= md->end_subject)
RRETURN(MATCH_NOMATCH);
ecode++;
break;
case OP_NOT_DIGIT:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
if (
#ifdef SUPPORT_UTF8
c < 256 &&
#endif
(md->ctypes[c] & ctype_digit) != 0)
RRETURN(MATCH_NOMATCH);
ecode++;
break;
case OP_DIGIT:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
if (
#ifdef SUPPORT_UTF8
c >= 256 ||
#endif
(md->ctypes[c] & ctype_digit) == 0)
RRETURN(MATCH_NOMATCH);
ecode++;
break;
case OP_NOT_WHITESPACE:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
if (
#ifdef SUPPORT_UTF8
c < 256 &&
#endif
(md->ctypes[c] & ctype_space) != 0)
RRETURN(MATCH_NOMATCH);
ecode++;
break;
case OP_WHITESPACE:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
if (
#ifdef SUPPORT_UTF8
c >= 256 ||
#endif
(md->ctypes[c] & ctype_space) == 0)
RRETURN(MATCH_NOMATCH);
ecode++;
break;
case OP_NOT_WORDCHAR:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
if (
#ifdef SUPPORT_UTF8
c < 256 &&
#endif
(md->ctypes[c] & ctype_word) != 0)
RRETURN(MATCH_NOMATCH);
ecode++;
break;
case OP_WORDCHAR:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
if (
#ifdef SUPPORT_UTF8
c >= 256 ||
#endif
(md->ctypes[c] & ctype_word) == 0)
RRETURN(MATCH_NOMATCH);
ecode++;
break;
case OP_ANYNL:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
switch (c) {
default:
RRETURN(MATCH_NOMATCH);
case 0x000d:
if (eptr < md->end_subject && *eptr == 0x0a)
eptr++;
break;
case 0x000a:
break;
case 0x000b:
case 0x000c:
case 0x0085:
case 0x2028:
case 0x2029:
if (md->bsr_anycrlf)
RRETURN(MATCH_NOMATCH);
break;
}
ecode++;
break;
case OP_NOT_HSPACE:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
switch (c) {
default:
break;
case 0x09: /* HT */
case 0x20: /* SPACE */
case 0xa0: /* NBSP */
case 0x1680: /* OGHAM SPACE MARK */
case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
case 0x2000: /* EN QUAD */
case 0x2001: /* EM QUAD */
case 0x2002: /* EN SPACE */
case 0x2003: /* EM SPACE */
case 0x2004: /* THREE-PER-EM SPACE */
case 0x2005: /* FOUR-PER-EM SPACE */
case 0x2006: /* SIX-PER-EM SPACE */
case 0x2007: /* FIGURE SPACE */
case 0x2008: /* PUNCTUATION SPACE */
case 0x2009: /* THIN SPACE */
case 0x200A: /* HAIR SPACE */
case 0x202f: /* NARROW NO-BREAK SPACE */
case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
case 0x3000: /* IDEOGRAPHIC SPACE */
RRETURN(MATCH_NOMATCH);
}
ecode++;
break;
case OP_HSPACE:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
switch (c) {
default:
RRETURN(MATCH_NOMATCH);
case 0x09: /* HT */
case 0x20: /* SPACE */
case 0xa0: /* NBSP */
case 0x1680: /* OGHAM SPACE MARK */
case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
case 0x2000: /* EN QUAD */
case 0x2001: /* EM QUAD */
case 0x2002: /* EN SPACE */
case 0x2003: /* EM SPACE */
case 0x2004: /* THREE-PER-EM SPACE */
case 0x2005: /* FOUR-PER-EM SPACE */
case 0x2006: /* SIX-PER-EM SPACE */
case 0x2007: /* FIGURE SPACE */
case 0x2008: /* PUNCTUATION SPACE */
case 0x2009: /* THIN SPACE */
case 0x200A: /* HAIR SPACE */
case 0x202f: /* NARROW NO-BREAK SPACE */
case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
case 0x3000: /* IDEOGRAPHIC SPACE */
break;
}
ecode++;
break;
case OP_NOT_VSPACE:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
switch (c) {
default:
break;
case 0x0a: /* LF */
case 0x0b: /* VT */
case 0x0c: /* FF */
case 0x0d: /* CR */
case 0x85: /* NEL */
case 0x2028: /* LINE SEPARATOR */
case 0x2029: /* PARAGRAPH SEPARATOR */
RRETURN(MATCH_NOMATCH);
}
ecode++;
break;
case OP_VSPACE:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
switch (c) {
default:
RRETURN(MATCH_NOMATCH);
case 0x0a: /* LF */
case 0x0b: /* VT */
case 0x0c: /* FF */
case 0x0d: /* CR */
case 0x85: /* NEL */
case 0x2028: /* LINE SEPARATOR */
case 0x2029: /* PARAGRAPH SEPARATOR */
break;
}
ecode++;
break;
#ifdef SUPPORT_UCP
/* Check the next character by Unicode property. We will get
here only if the support is in the binary; otherwise a
compile-time error occurs. */
case OP_PROP:
case OP_NOTPROP:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
{
int chartype, script;
int category = _pcre_ucp_findprop(c, &chartype, &script);
switch (ecode[1]) {
case PT_ANY:
if (op == OP_NOTPROP)
RRETURN(MATCH_NOMATCH);
break;
case PT_LAMP:
if ((chartype == ucp_Lu || chartype == ucp_Ll ||
chartype == ucp_Lt) == (op == OP_NOTPROP))
RRETURN(MATCH_NOMATCH);
break;
case PT_GC:
if ((ecode[2] != category) == (op == OP_PROP))
RRETURN(MATCH_NOMATCH);
break;
case PT_PC:
if ((ecode[2] != chartype) == (op == OP_PROP))
RRETURN(MATCH_NOMATCH);
break;
case PT_SC:
if ((ecode[2] != script) == (op == OP_PROP))
RRETURN(MATCH_NOMATCH);
break;
default:
RRETURN(PCRE_ERROR_INTERNAL);
}
ecode += 3;
}
break;
/* Match an extended Unicode sequence. We will get here only if
the support is in the binary; otherwise a compile-time error
occurs. */
case OP_EXTUNI:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
{
int chartype, script;
int category = _pcre_ucp_findprop(c, &chartype, &script);
if (category == ucp_M)
RRETURN(MATCH_NOMATCH);
while (eptr < md->end_subject) {
int len = 1;
if (!utf8)
c = *eptr;
else {
GETCHARLEN(c, eptr, len);
}
category = _pcre_ucp_findprop(c, &chartype, &script);
if (category != ucp_M)
break;
eptr += len;
}
}
ecode++;
break;
#endif
/* Match a back reference, possibly repeatedly. Look past the
end of the item to see if there is repeat information following.
The code is similar to that for character classes, but repeated
for efficiency. Then obey similar code to character type repeats
- written out again for speed. However, if the referenced string
is the empty string, always treat it as matched, any number of
times (otherwise there could be infinite loops). */
case OP_REF: {
offset = GET2(ecode, 1) << 1; /* Doubled ref number */
ecode += 3; /* Advance past item */
/* If the reference is unset, set the length to be longer than
the amount of subject left; this ensures that every attempt at a
match fails. We can't just fail here, because of the possibility
of quantifiers with zero minima. */
length = (offset >= offset_top || md->offset_vector[offset] < 0)
? md->end_subject - eptr + 1
: md->offset_vector[offset + 1] -
md->offset_vector[offset];
/* Set up for repetition, or handle the non-repeated case */
switch (*ecode) {
case OP_CRSTAR:
case OP_CRMINSTAR:
case OP_CRPLUS:
case OP_CRMINPLUS:
case OP_CRQUERY:
case OP_CRMINQUERY:
c = *ecode++ - OP_CRSTAR;
minimize = (c & 1) != 0;
min = rep_min[c]; /* Pick up values from tables; */
max = rep_max[c]; /* zero for max => infinity */
if (max == 0)
max = INT_MAX;
break;
case OP_CRRANGE:
case OP_CRMINRANGE:
minimize = (*ecode == OP_CRMINRANGE);
min = GET2(ecode, 1);
max = GET2(ecode, 3);
if (max == 0)
max = INT_MAX;
ecode += 5;
break;
default: /* No repeat follows */
if (!match_ref(offset, eptr, length, md, ims))
RRETURN(MATCH_NOMATCH);
eptr += length;
continue; /* With the main loop */
}
/* If the length of the reference is zero, just continue with
the main loop. */
if (length == 0)
continue;
/* First, ensure the minimum number of matches are present. We
get back the length of the reference string explicitly rather
than passing the address of eptr, so that eptr can be a register
variable. */
for (i = 1; i <= min; i++) {
if (!match_ref(offset, eptr, length, md, ims))
RRETURN(MATCH_NOMATCH);
eptr += length;
}
/* If min = max, continue at the same level without recursion.
They are not both allowed to be zero. */
if (min == max)
continue;
/* If minimizing, keep trying and advancing the pointer */
if (minimize) {
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM14);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max ||
!match_ref(offset, eptr, length, md, ims))
RRETURN(MATCH_NOMATCH);
eptr += length;
}
/* Control never gets here */
}
/* If maximizing, find the longest string and work backwards */
else {
pp = eptr;
for (i = min; i < max; i++) {
if (!match_ref(offset, eptr, length, md, ims))
break;
eptr += length;
}
while (eptr >= pp) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM15);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
eptr -= length;
}
RRETURN(MATCH_NOMATCH);
}
}
/* Control never gets here */
/* Match a bit-mapped character class, possibly repeatedly. This
op code is used when all the characters in the class have values
in the range 0-255, and either the matching is caseful, or the
characters are in the range 0-127 when UTF-8 processing is
enabled. The only difference between OP_CLASS and OP_NCLASS
occurs when a data character outside the range is encountered.
First, look past the end of the item to see if there is repeat
information following. Then obey similar code to character type
repeats - written out again for speed. */
case OP_NCLASS:
case OP_CLASS: {
data = ecode + 1; /* Save for matching */
ecode += 33; /* Advance past the item */
switch (*ecode) {
case OP_CRSTAR:
case OP_CRMINSTAR:
case OP_CRPLUS:
case OP_CRMINPLUS:
case OP_CRQUERY:
case OP_CRMINQUERY:
c = *ecode++ - OP_CRSTAR;
minimize = (c & 1) != 0;
min = rep_min[c]; /* Pick up values from tables; */
max = rep_max[c]; /* zero for max => infinity */
if (max == 0)
max = INT_MAX;
break;
case OP_CRRANGE:
case OP_CRMINRANGE:
minimize = (*ecode == OP_CRMINRANGE);
min = GET2(ecode, 1);
max = GET2(ecode, 3);
if (max == 0)
max = INT_MAX;
ecode += 5;
break;
default: /* No repeat follows */
min = max = 1;
break;
}
/* First, ensure the minimum number of matches are present.
*/
#ifdef SUPPORT_UTF8
/* UTF-8 mode */
if (utf8) {
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
if (c > 255) {
if (op == OP_CLASS)
RRETURN(MATCH_NOMATCH);
} else {
if ((data[c / 8] & (1 << (c & 7))) == 0)
RRETURN(MATCH_NOMATCH);
}
}
} else
#endif
/* Not UTF-8 mode */
{
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
c = *eptr++;
if ((data[c / 8] & (1 << (c & 7))) == 0)
RRETURN(MATCH_NOMATCH);
}
}
/* If max == min we can continue with the main loop without the
need to recurse. */
if (min == max)
continue;
/* If minimizing, keep testing the rest of the expression and
advancing the pointer while it matches the class. */
if (minimize) {
#ifdef SUPPORT_UTF8
/* UTF-8 mode */
if (utf8) {
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM16);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
if (c > 255) {
if (op == OP_CLASS)
RRETURN(MATCH_NOMATCH);
} else {
if ((data[c / 8] & (1 << (c & 7))) == 0)
RRETURN(MATCH_NOMATCH);
}
}
} else
#endif
/* Not UTF-8 mode */
{
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM17);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
c = *eptr++;
if ((data[c / 8] & (1 << (c & 7))) == 0)
RRETURN(MATCH_NOMATCH);
}
}
/* Control never gets here */
}
/* If maximizing, find the longest possible run, then work
backwards. */
else {
pp = eptr;
#ifdef SUPPORT_UTF8
/* UTF-8 mode */
if (utf8) {
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
if (c > 255) {
if (op == OP_CLASS)
break;
} else {
if ((data[c / 8] & (1 << (c & 7))) == 0)
break;
}
eptr += len;
}
for (;;) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM18);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (eptr-- == pp)
break; /* Stop if tried at original pos */
BACKCHAR(eptr);
}
} else
#endif
/* Not UTF-8 mode */
{
for (i = min; i < max; i++) {
if (eptr >= md->end_subject)
break;
c = *eptr;
if ((data[c / 8] & (1 << (c & 7))) == 0)
break;
eptr++;
}
while (eptr >= pp) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM19);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
eptr--;
}
}
RRETURN(MATCH_NOMATCH);
}
}
/* Control never gets here */
/* Match an extended character class. This opcode is encountered
only in UTF-8 mode, because that's the only time it is compiled. */
#ifdef SUPPORT_UTF8
case OP_XCLASS: {
data = ecode + 1 + LINK_SIZE; /* Save for matching */
ecode += GET(ecode, 1); /* Advance past the item */
switch (*ecode) {
case OP_CRSTAR:
case OP_CRMINSTAR:
case OP_CRPLUS:
case OP_CRMINPLUS:
case OP_CRQUERY:
case OP_CRMINQUERY:
c = *ecode++ - OP_CRSTAR;
minimize = (c & 1) != 0;
min = rep_min[c]; /* Pick up values from tables; */
max = rep_max[c]; /* zero for max => infinity */
if (max == 0)
max = INT_MAX;
break;
case OP_CRRANGE:
case OP_CRMINRANGE:
minimize = (*ecode == OP_CRMINRANGE);
min = GET2(ecode, 1);
max = GET2(ecode, 3);
if (max == 0)
max = INT_MAX;
ecode += 5;
break;
default: /* No repeat follows */
min = max = 1;
break;
}
/* First, ensure the minimum number of matches are present. */
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
if (!_pcre_xclass(c, data))
RRETURN(MATCH_NOMATCH);
}
/* If max == min we can continue with the main loop without the
need to recurse. */
if (min == max)
continue;
/* If minimizing, keep testing the rest of the expression and
advancing the pointer while it matches the class. */
if (minimize) {
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM20);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
if (!_pcre_xclass(c, data))
RRETURN(MATCH_NOMATCH);
}
/* Control never gets here */
}
/* If maximizing, find the longest possible run, then work
backwards. */
else {
pp = eptr;
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
if (!_pcre_xclass(c, data))
break;
eptr += len;
}
for (;;) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM21);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (eptr-- == pp)
break; /* Stop if tried at original pos */
if (utf8)
BACKCHAR(eptr);
}
RRETURN(MATCH_NOMATCH);
}
/* Control never gets here */
}
#endif /* End of XCLASS */
/* Match a single character, casefully */
case OP_CHAR:
#ifdef SUPPORT_UTF8
if (utf8) {
length = 1;
ecode++;
GETCHARLEN(fc, ecode, length);
if (length > md->end_subject - eptr)
RRETURN(MATCH_NOMATCH);
while (length-- > 0)
if (*ecode++ != *eptr++)
RRETURN(MATCH_NOMATCH);
} else
#endif
/* Non-UTF-8 mode */
{
if (md->end_subject - eptr < 1)
RRETURN(MATCH_NOMATCH);
if (ecode[1] != *eptr++)
RRETURN(MATCH_NOMATCH);
ecode += 2;
}
break;
/* Match a single character, caselessly */
case OP_CHARNC:
#ifdef SUPPORT_UTF8
if (utf8) {
length = 1;
ecode++;
GETCHARLEN(fc, ecode, length);
if (length > md->end_subject - eptr)
RRETURN(MATCH_NOMATCH);
/* If the pattern character's value is < 128, we have only
one byte, and can use the fast lookup table. */
if (fc < 128) {
if (md->lcc[*ecode++] != md->lcc[*eptr++])
RRETURN(MATCH_NOMATCH);
}
/* Otherwise we must pick up the subject character */
else {
unsigned int dc;
GETCHARINC(dc, eptr);
ecode += length;
/* If we have Unicode property support, we can use it to
test the other case of the character, if there is one.
*/
if (fc != dc) {
#ifdef SUPPORT_UCP
if (dc != _pcre_ucp_othercase(fc))
#endif
RRETURN(MATCH_NOMATCH);
}
}
} else
#endif /* SUPPORT_UTF8 */
/* Non-UTF-8 mode */
{
if (md->end_subject - eptr < 1)
RRETURN(MATCH_NOMATCH);
if (md->lcc[ecode[1]] != md->lcc[*eptr++])
RRETURN(MATCH_NOMATCH);
ecode += 2;
}
break;
/* Match a single character repeatedly. */
case OP_EXACT:
min = max = GET2(ecode, 1);
ecode += 3;
goto REPEATCHAR;
case OP_POSUPTO:
possessive = TRUE;
/* Fall through */
case OP_UPTO:
case OP_MINUPTO:
min = 0;
max = GET2(ecode, 1);
minimize = *ecode == OP_MINUPTO;
ecode += 3;
goto REPEATCHAR;
case OP_POSSTAR:
possessive = TRUE;
min = 0;
max = INT_MAX;
ecode++;
goto REPEATCHAR;
case OP_POSPLUS:
possessive = TRUE;
min = 1;
max = INT_MAX;
ecode++;
goto REPEATCHAR;
case OP_POSQUERY:
possessive = TRUE;
min = 0;
max = 1;
ecode++;
goto REPEATCHAR;
case OP_STAR:
case OP_MINSTAR:
case OP_PLUS:
case OP_MINPLUS:
case OP_QUERY:
case OP_MINQUERY:
c = *ecode++ - OP_STAR;
minimize = (c & 1) != 0;
min = rep_min[c]; /* Pick up values from tables; */
max = rep_max[c]; /* zero for max => infinity */
if (max == 0)
max = INT_MAX;
/* Common code for all repeated single-character matches. We can
give up quickly if there are fewer than the minimum number of
characters left in the subject. */
REPEATCHAR:
#ifdef SUPPORT_UTF8
if (utf8) {
length = 1;
charptr = ecode;
GETCHARLEN(fc, ecode, length);
if (min * length > md->end_subject - eptr)
RRETURN(MATCH_NOMATCH);
ecode += length;
/* Handle multibyte character matching specially here. There
is support for caseless matching if UCP support is present.
*/
if (length > 1) {
#ifdef SUPPORT_UCP
unsigned int othercase;
if ((ims & PCRE_CASELESS) != 0 &&
(othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)
oclength = _pcre_ord2utf8(othercase, occhars);
else
oclength = 0;
#endif /* SUPPORT_UCP */
for (i = 1; i <= min; i++) {
if (memcmp(eptr, charptr, length) == 0)
eptr += length;
#ifdef SUPPORT_UCP
/* Need braces because of following else */
else if (oclength == 0) {
RRETURN(MATCH_NOMATCH);
} else {
if (memcmp(eptr, occhars, oclength) != 0)
RRETURN(MATCH_NOMATCH);
eptr += oclength;
}
#else /* without SUPPORT_UCP */
else {
RRETURN(MATCH_NOMATCH);
}
#endif /* SUPPORT_UCP */
}
if (min == max)
continue;
if (minimize) {
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM22);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
if (memcmp(eptr, charptr, length) == 0)
eptr += length;
#ifdef SUPPORT_UCP
/* Need braces because of following else */
else if (oclength == 0) {
RRETURN(MATCH_NOMATCH);
} else {
if (memcmp(eptr, occhars, oclength) != 0)
RRETURN(MATCH_NOMATCH);
eptr += oclength;
}
#else /* without SUPPORT_UCP */
else {
RRETURN(MATCH_NOMATCH);
}
#endif /* SUPPORT_UCP */
}
/* Control never gets here */
}
else /* Maximize */
{
pp = eptr;
for (i = min; i < max; i++) {
if (eptr > md->end_subject - length)
break;
if (memcmp(eptr, charptr, length) == 0)
eptr += length;
#ifdef SUPPORT_UCP
else if (oclength == 0)
break;
else {
if (memcmp(eptr, occhars, oclength) != 0)
break;
eptr += oclength;
}
#else /* without SUPPORT_UCP */
else
break;
#endif /* SUPPORT_UCP */
}
if (possessive)
continue;
for (;;) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM23);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (eptr == pp)
RRETURN(MATCH_NOMATCH);
#ifdef SUPPORT_UCP
eptr--;
BACKCHAR(eptr);
#else /* without SUPPORT_UCP */
eptr -= length;
#endif /* SUPPORT_UCP */
}
}
/* Control never gets here */
}
/* If the length of a UTF-8 character is 1, we fall through
here, and obey the code as for non-UTF-8 characters below,
though in this case the value of fc will always be < 128. */
} else
#endif /* SUPPORT_UTF8 */
/* When not in UTF-8 mode, load a single-byte character. */
{
if (min > md->end_subject - eptr)
RRETURN(MATCH_NOMATCH);
fc = *ecode++;
}
/* The value of fc at this point is always less than 256, though
we may or may not be in UTF-8 mode. The code is duplicated for
the caseless and caseful cases, for speed, since matching
characters is likely to be quite common. First, ensure the
minimum number of matches are present. If min = max, continue at
the same level without recursing. Otherwise, if minimizing, keep
trying the rest of the expression and advancing one matching
character if failing, up to the maximum. Alternatively, if
maximizing, find the maximum number of characters and work
backwards. */
DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min,
max, max, eptr));
if ((ims & PCRE_CASELESS) != 0) {
fc = md->lcc[fc];
for (i = 1; i <= min; i++)
if (fc != md->lcc[*eptr++])
RRETURN(MATCH_NOMATCH);
if (min == max)
continue;
if (minimize) {
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM24);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject ||
fc != md->lcc[*eptr++])
RRETURN(MATCH_NOMATCH);
}
/* Control never gets here */
} else /* Maximize */
{
pp = eptr;
for (i = min; i < max; i++) {
if (eptr >= md->end_subject || fc != md->lcc[*eptr])
break;
eptr++;
}
if (possessive)
continue;
while (eptr >= pp) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM25);
eptr--;
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
}
RRETURN(MATCH_NOMATCH);
}
/* Control never gets here */
}
/* Caseful comparisons (includes all multi-byte characters) */
else {
for (i = 1; i <= min; i++)
if (fc != *eptr++)
RRETURN(MATCH_NOMATCH);
if (min == max)
continue;
if (minimize) {
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM26);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject ||
fc != *eptr++)
RRETURN(MATCH_NOMATCH);
}
/* Control never gets here */
} else /* Maximize */
{
pp = eptr;
for (i = min; i < max; i++) {
if (eptr >= md->end_subject || fc != *eptr)
break;
eptr++;
}
if (possessive)
continue;
while (eptr >= pp) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM27);
eptr--;
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
}
RRETURN(MATCH_NOMATCH);
}
}
/* Control never gets here */
/* Match a negated single one-byte character. The character we
are checking can be multibyte. */
case OP_NOT:
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
ecode++;
GETCHARINCTEST(c, eptr);
if ((ims & PCRE_CASELESS) != 0) {
#ifdef SUPPORT_UTF8
if (c < 256)
#endif
c = md->lcc[c];
if (md->lcc[*ecode++] == c)
RRETURN(MATCH_NOMATCH);
} else {
if (*ecode++ == c)
RRETURN(MATCH_NOMATCH);
}
break;
/* Match a negated single one-byte character repeatedly. This is
almost a repeat of the code for a repeated single character, but
I haven't found a nice way of commoning these up that doesn't
require a test of the positive/negative option for each
character match. Maybe that wouldn't add very much to the time
taken, but character matching *is* what this is all about... */
case OP_NOTEXACT:
min = max = GET2(ecode, 1);
ecode += 3;
goto REPEATNOTCHAR;
case OP_NOTUPTO:
case OP_NOTMINUPTO:
min = 0;
max = GET2(ecode, 1);
minimize = *ecode == OP_NOTMINUPTO;
ecode += 3;
goto REPEATNOTCHAR;
case OP_NOTPOSSTAR:
possessive = TRUE;
min = 0;
max = INT_MAX;
ecode++;
goto REPEATNOTCHAR;
case OP_NOTPOSPLUS:
possessive = TRUE;
min = 1;
max = INT_MAX;
ecode++;
goto REPEATNOTCHAR;
case OP_NOTPOSQUERY:
possessive = TRUE;
min = 0;
max = 1;
ecode++;
goto REPEATNOTCHAR;
case OP_NOTPOSUPTO:
possessive = TRUE;
min = 0;
max = GET2(ecode, 1);
ecode += 3;
goto REPEATNOTCHAR;
case OP_NOTSTAR:
case OP_NOTMINSTAR:
case OP_NOTPLUS:
case OP_NOTMINPLUS:
case OP_NOTQUERY:
case OP_NOTMINQUERY:
c = *ecode++ - OP_NOTSTAR;
minimize = (c & 1) != 0;
min = rep_min[c]; /* Pick up values from tables; */
max = rep_max[c]; /* zero for max => infinity */
if (max == 0)
max = INT_MAX;
/* Common code for all repeated single-byte matches. We can give
up quickly if there are fewer than the minimum number of bytes
left in the subject. */
REPEATNOTCHAR:
if (min > md->end_subject - eptr)
RRETURN(MATCH_NOMATCH);
fc = *ecode++;
/* The code is duplicated for the caseless and caseful cases,
for speed, since matching characters is likely to be quite
common. First, ensure the minimum number of matches are present.
If min = max, continue at the same level without recursing.
Otherwise, if minimizing, keep trying the rest of the expression
and advancing one matching character if failing, up to the
maximum. Alternatively, if maximizing, find the maximum number
of characters and work backwards. */
DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n",
fc, min, max, max, eptr));
if ((ims & PCRE_CASELESS) != 0) {
fc = md->lcc[fc];
#ifdef SUPPORT_UTF8
/* UTF-8 mode */
if (utf8) {
register unsigned int d;
for (i = 1; i <= min; i++) {
GETCHARINC(d, eptr);
if (d < 256)
d = md->lcc[d];
if (fc == d)
RRETURN(MATCH_NOMATCH);
}
} else
#endif
/* Not UTF-8 mode */
{
for (i = 1; i <= min; i++)
if (fc == md->lcc[*eptr++])
RRETURN(MATCH_NOMATCH);
}
if (min == max)
continue;
if (minimize) {
#ifdef SUPPORT_UTF8
/* UTF-8 mode */
if (utf8) {
register unsigned int d;
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM28);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
GETCHARINC(d, eptr);
if (d < 256)
d = md->lcc[d];
if (fi >= max || eptr >= md->end_subject ||
fc == d)
RRETURN(MATCH_NOMATCH);
}
} else
#endif
/* Not UTF-8 mode */
{
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM29);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject ||
fc == md->lcc[*eptr++])
RRETURN(MATCH_NOMATCH);
}
}
/* Control never gets here */
}
/* Maximize case */
else {
pp = eptr;
#ifdef SUPPORT_UTF8
/* UTF-8 mode */
if (utf8) {
register unsigned int d;
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(d, eptr, len);
if (d < 256)
d = md->lcc[d];
if (fc == d)
break;
eptr += len;
}
if (possessive)
continue;
for (;;) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM30);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (eptr-- == pp)
break; /* Stop if tried at original pos */
BACKCHAR(eptr);
}
} else
#endif
/* Not UTF-8 mode */
{
for (i = min; i < max; i++) {
if (eptr >= md->end_subject ||
fc == md->lcc[*eptr])
break;
eptr++;
}
if (possessive)
continue;
while (eptr >= pp) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM31);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
eptr--;
}
}
RRETURN(MATCH_NOMATCH);
}
/* Control never gets here */
}
/* Caseful comparisons */
else {
#ifdef SUPPORT_UTF8
/* UTF-8 mode */
if (utf8) {
register unsigned int d;
for (i = 1; i <= min; i++) {
GETCHARINC(d, eptr);
if (fc == d)
RRETURN(MATCH_NOMATCH);
}
} else
#endif
/* Not UTF-8 mode */
{
for (i = 1; i <= min; i++)
if (fc == *eptr++)
RRETURN(MATCH_NOMATCH);
}
if (min == max)
continue;
if (minimize) {
#ifdef SUPPORT_UTF8
/* UTF-8 mode */
if (utf8) {
register unsigned int d;
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM32);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
GETCHARINC(d, eptr);
if (fi >= max || eptr >= md->end_subject ||
fc == d)
RRETURN(MATCH_NOMATCH);
}
} else
#endif
/* Not UTF-8 mode */
{
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM33);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject ||
fc == *eptr++)
RRETURN(MATCH_NOMATCH);
}
}
/* Control never gets here */
}
/* Maximize case */
else {
pp = eptr;
#ifdef SUPPORT_UTF8
/* UTF-8 mode */
if (utf8) {
register unsigned int d;
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(d, eptr, len);
if (fc == d)
break;
eptr += len;
}
if (possessive)
continue;
for (;;) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM34);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (eptr-- == pp)
break; /* Stop if tried at original pos */
BACKCHAR(eptr);
}
} else
#endif
/* Not UTF-8 mode */
{
for (i = min; i < max; i++) {
if (eptr >= md->end_subject || fc == *eptr)
break;
eptr++;
}
if (possessive)
continue;
while (eptr >= pp) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM35);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
eptr--;
}
}
RRETURN(MATCH_NOMATCH);
}
}
/* Control never gets here */
/* Match a single character type repeatedly; several different
opcodes share code. This is very similar to the code for single
characters, but we repeat it in the interests of efficiency. */
case OP_TYPEEXACT:
min = max = GET2(ecode, 1);
minimize = TRUE;
ecode += 3;
goto REPEATTYPE;
case OP_TYPEUPTO:
case OP_TYPEMINUPTO:
min = 0;
max = GET2(ecode, 1);
minimize = *ecode == OP_TYPEMINUPTO;
ecode += 3;
goto REPEATTYPE;
case OP_TYPEPOSSTAR:
possessive = TRUE;
min = 0;
max = INT_MAX;
ecode++;
goto REPEATTYPE;
case OP_TYPEPOSPLUS:
possessive = TRUE;
min = 1;
max = INT_MAX;
ecode++;
goto REPEATTYPE;
case OP_TYPEPOSQUERY:
possessive = TRUE;
min = 0;
max = 1;
ecode++;
goto REPEATTYPE;
case OP_TYPEPOSUPTO:
possessive = TRUE;
min = 0;
max = GET2(ecode, 1);
ecode += 3;
goto REPEATTYPE;
case OP_TYPESTAR:
case OP_TYPEMINSTAR:
case OP_TYPEPLUS:
case OP_TYPEMINPLUS:
case OP_TYPEQUERY:
case OP_TYPEMINQUERY:
c = *ecode++ - OP_TYPESTAR;
minimize = (c & 1) != 0;
min = rep_min[c]; /* Pick up values from tables; */
max = rep_max[c]; /* zero for max => infinity */
if (max == 0)
max = INT_MAX;
/* Common code for all repeated single character type matches.
Note that in UTF-8 mode, '.' matches a character of any length,
but for the other character types, the valid characters are all
one-byte long. */
REPEATTYPE:
ctype = *ecode++; /* Code for the character type */
#ifdef SUPPORT_UCP
if (ctype == OP_PROP || ctype == OP_NOTPROP) {
prop_fail_result = ctype == OP_NOTPROP;
prop_type = *ecode++;
prop_value = *ecode++;
} else
prop_type = -1;
#endif
/* First, ensure the minimum number of matches are present. Use
inline code for maximizing the speed, and do the type test once
at the start (i.e. keep it out of the loop). Also we can test
that there are at least the minimum number of bytes before we
start. This isn't as effective in UTF-8 mode, but it does no
harm. Separate the UTF-8 code completely as that is tidier. Also
separate the UCP code, which can be the same for both UTF-8 and
single-bytes. */
if (min > md->end_subject - eptr)
RRETURN(MATCH_NOMATCH);
if (min > 0) {
#ifdef SUPPORT_UCP
if (prop_type >= 0) {
switch (prop_type) {
case PT_ANY:
if (prop_fail_result)
RRETURN(MATCH_NOMATCH);
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
}
break;
case PT_LAMP:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if ((prop_chartype == ucp_Lu ||
prop_chartype == ucp_Ll ||
prop_chartype == ucp_Lt) ==
prop_fail_result)
RRETURN(MATCH_NOMATCH);
}
break;
case PT_GC:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if ((prop_category == prop_value) ==
prop_fail_result)
RRETURN(MATCH_NOMATCH);
}
break;
case PT_PC:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if ((prop_chartype == prop_value) ==
prop_fail_result)
RRETURN(MATCH_NOMATCH);
}
break;
case PT_SC:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if ((prop_script == prop_value) ==
prop_fail_result)
RRETURN(MATCH_NOMATCH);
}
break;
default:
RRETURN(PCRE_ERROR_INTERNAL);
}
}
/* Match extended Unicode sequences. We will get here only
if the support is in the binary; otherwise a compile-time
error occurs. */
else if (ctype == OP_EXTUNI) {
for (i = 1; i <= min; i++) {
GETCHARINCTEST(c, eptr);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if (prop_category == ucp_M)
RRETURN(MATCH_NOMATCH);
while (eptr < md->end_subject) {
int len = 1;
if (!utf8)
c = *eptr;
else {
GETCHARLEN(c, eptr, len);
}
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if (prop_category != ucp_M)
break;
eptr += len;
}
}
}
else
#endif /* SUPPORT_UCP */
/* Handle all other cases when the coding is UTF-8 */
#ifdef SUPPORT_UTF8
if (utf8)
switch (ctype) {
case OP_ANY:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject ||
((ims & PCRE_DOTALL) == 0 &&
IS_NEWLINE(eptr)))
RRETURN(MATCH_NOMATCH);
eptr++;
while (eptr < md->end_subject &&
(*eptr & 0xc0) == 0x80)
eptr++;
}
break;
case OP_ANYBYTE:
eptr += min;
break;
case OP_ANYNL:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
switch (c) {
default:
RRETURN(MATCH_NOMATCH);
case 0x000d:
if (eptr < md->end_subject &&
*eptr == 0x0a)
eptr++;
break;
case 0x000a:
break;
case 0x000b:
case 0x000c:
case 0x0085:
case 0x2028:
case 0x2029:
if (md->bsr_anycrlf)
RRETURN(MATCH_NOMATCH);
break;
}
}
break;
case OP_NOT_HSPACE:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
switch (c) {
default:
break;
case 0x09: /* HT */
case 0x20: /* SPACE */
case 0xa0: /* NBSP */
case 0x1680: /* OGHAM SPACE MARK */
case 0x180e: /* MONGOLIAN VOWEL
SEPARATOR */
case 0x2000: /* EN QUAD */
case 0x2001: /* EM QUAD */
case 0x2002: /* EN SPACE */
case 0x2003: /* EM SPACE */
case 0x2004: /* THREE-PER-EM SPACE */
case 0x2005: /* FOUR-PER-EM SPACE */
case 0x2006: /* SIX-PER-EM SPACE */
case 0x2007: /* FIGURE SPACE */
case 0x2008: /* PUNCTUATION SPACE */
case 0x2009: /* THIN SPACE */
case 0x200A: /* HAIR SPACE */
case 0x202f: /* NARROW NO-BREAK SPACE */
case 0x205f: /* MEDIUM MATHEMATICAL
SPACE */
case 0x3000: /* IDEOGRAPHIC SPACE */
RRETURN(MATCH_NOMATCH);
}
}
break;
case OP_HSPACE:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
switch (c) {
default:
RRETURN(MATCH_NOMATCH);
case 0x09: /* HT */
case 0x20: /* SPACE */
case 0xa0: /* NBSP */
case 0x1680: /* OGHAM SPACE MARK */
case 0x180e: /* MONGOLIAN VOWEL
SEPARATOR */
case 0x2000: /* EN QUAD */
case 0x2001: /* EM QUAD */
case 0x2002: /* EN SPACE */
case 0x2003: /* EM SPACE */
case 0x2004: /* THREE-PER-EM SPACE */
case 0x2005: /* FOUR-PER-EM SPACE */
case 0x2006: /* SIX-PER-EM SPACE */
case 0x2007: /* FIGURE SPACE */
case 0x2008: /* PUNCTUATION SPACE */
case 0x2009: /* THIN SPACE */
case 0x200A: /* HAIR SPACE */
case 0x202f: /* NARROW NO-BREAK SPACE */
case 0x205f: /* MEDIUM MATHEMATICAL
SPACE */
case 0x3000: /* IDEOGRAPHIC SPACE */
break;
}
}
break;
case OP_NOT_VSPACE:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
switch (c) {
default:
break;
case 0x0a: /* LF */
case 0x0b: /* VT */
case 0x0c: /* FF */
case 0x0d: /* CR */
case 0x85: /* NEL */
case 0x2028: /* LINE SEPARATOR */
case 0x2029: /* PARAGRAPH SEPARATOR */
RRETURN(MATCH_NOMATCH);
}
}
break;
case OP_VSPACE:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
switch (c) {
default:
RRETURN(MATCH_NOMATCH);
case 0x0a: /* LF */
case 0x0b: /* VT */
case 0x0c: /* FF */
case 0x0d: /* CR */
case 0x85: /* NEL */
case 0x2028: /* LINE SEPARATOR */
case 0x2029: /* PARAGRAPH SEPARATOR */
break;
}
}
break;
case OP_NOT_DIGIT:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
if (c < 128 &&
(md->ctypes[c] & ctype_digit) != 0)
RRETURN(MATCH_NOMATCH);
}
break;
case OP_DIGIT:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject ||
*eptr >= 128 ||
(md->ctypes[*eptr++] & ctype_digit) ==
0)
RRETURN(MATCH_NOMATCH);
/* No need to skip more bytes - we know it's
* a 1-byte character */
}
break;
case OP_NOT_WHITESPACE:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject ||
(*eptr < 128 && (md->ctypes[*eptr] &
ctype_space) != 0))
RRETURN(MATCH_NOMATCH);
while (++eptr < md->end_subject &&
(*eptr & 0xc0) == 0x80)
;
}
break;
case OP_WHITESPACE:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject ||
*eptr >= 128 ||
(md->ctypes[*eptr++] & ctype_space) ==
0)
RRETURN(MATCH_NOMATCH);
/* No need to skip more bytes - we know it's
* a 1-byte character */
}
break;
case OP_NOT_WORDCHAR:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject ||
(*eptr < 128 &&
(md->ctypes[*eptr] & ctype_word) != 0))
RRETURN(MATCH_NOMATCH);
while (++eptr < md->end_subject &&
(*eptr & 0xc0) == 0x80)
;
}
break;
case OP_WORDCHAR:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject ||
*eptr >= 128 ||
(md->ctypes[*eptr++] & ctype_word) == 0)
RRETURN(MATCH_NOMATCH);
/* No need to skip more bytes - we know it's
* a 1-byte character */
}
break;
default:
RRETURN(PCRE_ERROR_INTERNAL);
} /* End switch(ctype) */
else
#endif /* SUPPORT_UTF8 */
/* Code for the non-UTF-8 case for minimum matching of
operators other than OP_PROP and OP_NOTPROP. We can
assume that there are the minimum number of bytes
present, as this was tested above. */
switch (ctype) {
case OP_ANY:
if ((ims & PCRE_DOTALL) == 0) {
for (i = 1; i <= min; i++) {
if (IS_NEWLINE(eptr))
RRETURN(MATCH_NOMATCH);
eptr++;
}
} else
eptr += min;
break;
case OP_ANYBYTE:
eptr += min;
break;
/* Because of the CRLF case, we can't assume the
minimum number of bytes are present in this
case. */
case OP_ANYNL:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
switch (*eptr++) {
default:
RRETURN(MATCH_NOMATCH);
case 0x000d:
if (eptr < md->end_subject &&
*eptr == 0x0a)
eptr++;
break;
case 0x000a:
break;
case 0x000b:
case 0x000c:
case 0x0085:
if (md->bsr_anycrlf)
RRETURN(MATCH_NOMATCH);
break;
}
}
break;
case OP_NOT_HSPACE:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
switch (*eptr++) {
default:
break;
case 0x09: /* HT */
case 0x20: /* SPACE */
case 0xa0: /* NBSP */
RRETURN(MATCH_NOMATCH);
}
}
break;
case OP_HSPACE:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
switch (*eptr++) {
default:
RRETURN(MATCH_NOMATCH);
case 0x09: /* HT */
case 0x20: /* SPACE */
case 0xa0: /* NBSP */
break;
}
}
break;
case OP_NOT_VSPACE:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
switch (*eptr++) {
default:
break;
case 0x0a: /* LF */
case 0x0b: /* VT */
case 0x0c: /* FF */
case 0x0d: /* CR */
case 0x85: /* NEL */
RRETURN(MATCH_NOMATCH);
}
}
break;
case OP_VSPACE:
for (i = 1; i <= min; i++) {
if (eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
switch (*eptr++) {
default:
RRETURN(MATCH_NOMATCH);
case 0x0a: /* LF */
case 0x0b: /* VT */
case 0x0c: /* FF */
case 0x0d: /* CR */
case 0x85: /* NEL */
break;
}
}
break;
case OP_NOT_DIGIT:
for (i = 1; i <= min; i++)
if ((md->ctypes[*eptr++] & ctype_digit) !=
0)
RRETURN(MATCH_NOMATCH);
break;
case OP_DIGIT:
for (i = 1; i <= min; i++)
if ((md->ctypes[*eptr++] & ctype_digit) ==
0)
RRETURN(MATCH_NOMATCH);
break;
case OP_NOT_WHITESPACE:
for (i = 1; i <= min; i++)
if ((md->ctypes[*eptr++] & ctype_space) !=
0)
RRETURN(MATCH_NOMATCH);
break;
case OP_WHITESPACE:
for (i = 1; i <= min; i++)
if ((md->ctypes[*eptr++] & ctype_space) ==
0)
RRETURN(MATCH_NOMATCH);
break;
case OP_NOT_WORDCHAR:
for (i = 1; i <= min; i++)
if ((md->ctypes[*eptr++] & ctype_word) != 0)
RRETURN(MATCH_NOMATCH);
break;
case OP_WORDCHAR:
for (i = 1; i <= min; i++)
if ((md->ctypes[*eptr++] & ctype_word) == 0)
RRETURN(MATCH_NOMATCH);
break;
default:
RRETURN(PCRE_ERROR_INTERNAL);
}
}
/* If min = max, continue at the same level without recursing */
if (min == max)
continue;
/* If minimizing, we have to test the rest of the pattern before
each subsequent match. Again, separate the UTF-8 case for speed,
and also separate the UCP cases. */
if (minimize) {
#ifdef SUPPORT_UCP
if (prop_type >= 0) {
switch (prop_type) {
case PT_ANY:
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims,
eptrb, 0, RM36);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
if (prop_fail_result)
RRETURN(MATCH_NOMATCH);
}
/* Control never gets here */
case PT_LAMP:
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims,
eptrb, 0, RM37);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if ((prop_chartype == ucp_Lu ||
prop_chartype == ucp_Ll ||
prop_chartype == ucp_Lt) ==
prop_fail_result)
RRETURN(MATCH_NOMATCH);
}
/* Control never gets here */
case PT_GC:
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims,
eptrb, 0, RM38);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if ((prop_category == prop_value) ==
prop_fail_result)
RRETURN(MATCH_NOMATCH);
}
/* Control never gets here */
case PT_PC:
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims,
eptrb, 0, RM39);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if ((prop_chartype == prop_value) ==
prop_fail_result)
RRETURN(MATCH_NOMATCH);
}
/* Control never gets here */
case PT_SC:
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims,
eptrb, 0, RM40);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if ((prop_script == prop_value) ==
prop_fail_result)
RRETURN(MATCH_NOMATCH);
}
/* Control never gets here */
default:
RRETURN(PCRE_ERROR_INTERNAL);
}
}
/* Match extended Unicode sequences. We will get here only
if the support is in the binary; otherwise a compile-time
error occurs. */
else if (ctype == OP_EXTUNI) {
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM41);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject)
RRETURN(MATCH_NOMATCH);
GETCHARINCTEST(c, eptr);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if (prop_category == ucp_M)
RRETURN(MATCH_NOMATCH);
while (eptr < md->end_subject) {
int len = 1;
if (!utf8)
c = *eptr;
else {
GETCHARLEN(c, eptr, len);
}
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if (prop_category != ucp_M)
break;
eptr += len;
}
}
}
else
#endif /* SUPPORT_UCP */
#ifdef SUPPORT_UTF8
/* UTF-8 mode */
if (utf8) {
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM42);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject ||
(ctype == OP_ANY &&
(ims & PCRE_DOTALL) == 0 &&
IS_NEWLINE(eptr)))
RRETURN(MATCH_NOMATCH);
GETCHARINC(c, eptr);
switch (ctype) {
case OP_ANY: /* This is the DOTALL case */
break;
case OP_ANYBYTE:
break;
case OP_ANYNL:
switch (c) {
default:
RRETURN(MATCH_NOMATCH);
case 0x000d:
if (eptr < md->end_subject &&
*eptr == 0x0a)
eptr++;
break;
case 0x000a:
break;
case 0x000b:
case 0x000c:
case 0x0085:
case 0x2028:
case 0x2029:
if (md->bsr_anycrlf)
RRETURN(MATCH_NOMATCH);
break;
}
break;
case OP_NOT_HSPACE:
switch (c) {
default:
break;
case 0x09: /* HT */
case 0x20: /* SPACE */
case 0xa0: /* NBSP */
case 0x1680: /* OGHAM SPACE MARK */
case 0x180e: /* MONGOLIAN VOWEL
SEPARATOR */
case 0x2000: /* EN QUAD */
case 0x2001: /* EM QUAD */
case 0x2002: /* EN SPACE */
case 0x2003: /* EM SPACE */
case 0x2004: /* THREE-PER-EM SPACE
*/
case 0x2005: /* FOUR-PER-EM SPACE */
case 0x2006: /* SIX-PER-EM SPACE */
case 0x2007: /* FIGURE SPACE */
case 0x2008: /* PUNCTUATION SPACE */
case 0x2009: /* THIN SPACE */
case 0x200A: /* HAIR SPACE */
case 0x202f: /* NARROW NO-BREAK
SPACE */
case 0x205f: /* MEDIUM MATHEMATICAL
SPACE */
case 0x3000: /* IDEOGRAPHIC SPACE */
RRETURN(MATCH_NOMATCH);
}
break;
case OP_HSPACE:
switch (c) {
default:
RRETURN(MATCH_NOMATCH);
case 0x09: /* HT */
case 0x20: /* SPACE */
case 0xa0: /* NBSP */
case 0x1680: /* OGHAM SPACE MARK */
case 0x180e: /* MONGOLIAN VOWEL
SEPARATOR */
case 0x2000: /* EN QUAD */
case 0x2001: /* EM QUAD */
case 0x2002: /* EN SPACE */
case 0x2003: /* EM SPACE */
case 0x2004: /* THREE-PER-EM SPACE
*/
case 0x2005: /* FOUR-PER-EM SPACE */
case 0x2006: /* SIX-PER-EM SPACE */
case 0x2007: /* FIGURE SPACE */
case 0x2008: /* PUNCTUATION SPACE */
case 0x2009: /* THIN SPACE */
case 0x200A: /* HAIR SPACE */
case 0x202f: /* NARROW NO-BREAK
SPACE */
case 0x205f: /* MEDIUM MATHEMATICAL
SPACE */
case 0x3000: /* IDEOGRAPHIC SPACE */
break;
}
break;
case OP_NOT_VSPACE:
switch (c) {
default:
break;
case 0x0a: /* LF */
case 0x0b: /* VT */
case 0x0c: /* FF */
case 0x0d: /* CR */
case 0x85: /* NEL */
case 0x2028: /* LINE SEPARATOR */
case 0x2029: /* PARAGRAPH SEPARATOR
*/
RRETURN(MATCH_NOMATCH);
}
break;
case OP_VSPACE:
switch (c) {
default:
RRETURN(MATCH_NOMATCH);
case 0x0a: /* LF */
case 0x0b: /* VT */
case 0x0c: /* FF */
case 0x0d: /* CR */
case 0x85: /* NEL */
case 0x2028: /* LINE SEPARATOR */
case 0x2029: /* PARAGRAPH SEPARATOR
*/
break;
}
break;
case OP_NOT_DIGIT:
if (c < 256 &&
(md->ctypes[c] & ctype_digit) != 0)
RRETURN(MATCH_NOMATCH);
break;
case OP_DIGIT:
if (c >= 256 ||
(md->ctypes[c] & ctype_digit) == 0)
RRETURN(MATCH_NOMATCH);
break;
case OP_NOT_WHITESPACE:
if (c < 256 &&
(md->ctypes[c] & ctype_space) != 0)
RRETURN(MATCH_NOMATCH);
break;
case OP_WHITESPACE:
if (c >= 256 ||
(md->ctypes[c] & ctype_space) == 0)
RRETURN(MATCH_NOMATCH);
break;
case OP_NOT_WORDCHAR:
if (c < 256 &&
(md->ctypes[c] & ctype_word) != 0)
RRETURN(MATCH_NOMATCH);
break;
case OP_WORDCHAR:
if (c >= 256 ||
(md->ctypes[c] & ctype_word) == 0)
RRETURN(MATCH_NOMATCH);
break;
default:
RRETURN(PCRE_ERROR_INTERNAL);
}
}
} else
#endif
/* Not UTF-8 mode */
{
for (fi = min;; fi++) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM43);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (fi >= max || eptr >= md->end_subject ||
((ims & PCRE_DOTALL) == 0 &&
IS_NEWLINE(eptr)))
RRETURN(MATCH_NOMATCH);
c = *eptr++;
switch (ctype) {
case OP_ANY: /* This is the DOTALL case */
break;
case OP_ANYBYTE:
break;
case OP_ANYNL:
switch (c) {
default:
RRETURN(MATCH_NOMATCH);
case 0x000d:
if (eptr < md->end_subject &&
*eptr == 0x0a)
eptr++;
break;
case 0x000a:
break;
case 0x000b:
case 0x000c:
case 0x0085:
if (md->bsr_anycrlf)
RRETURN(MATCH_NOMATCH);
break;
}
break;
case OP_NOT_HSPACE:
switch (c) {
default:
break;
case 0x09: /* HT */
case 0x20: /* SPACE */
case 0xa0: /* NBSP */
RRETURN(MATCH_NOMATCH);
}
break;
case OP_HSPACE:
switch (c) {
default:
RRETURN(MATCH_NOMATCH);
case 0x09: /* HT */
case 0x20: /* SPACE */
case 0xa0: /* NBSP */
break;
}
break;
case OP_NOT_VSPACE:
switch (c) {
default:
break;
case 0x0a: /* LF */
case 0x0b: /* VT */
case 0x0c: /* FF */
case 0x0d: /* CR */
case 0x85: /* NEL */
RRETURN(MATCH_NOMATCH);
}
break;
case OP_VSPACE:
switch (c) {
default:
RRETURN(MATCH_NOMATCH);
case 0x0a: /* LF */
case 0x0b: /* VT */
case 0x0c: /* FF */
case 0x0d: /* CR */
case 0x85: /* NEL */
break;
}
break;
case OP_NOT_DIGIT:
if ((md->ctypes[c] & ctype_digit) != 0)
RRETURN(MATCH_NOMATCH);
break;
case OP_DIGIT:
if ((md->ctypes[c] & ctype_digit) == 0)
RRETURN(MATCH_NOMATCH);
break;
case OP_NOT_WHITESPACE:
if ((md->ctypes[c] & ctype_space) != 0)
RRETURN(MATCH_NOMATCH);
break;
case OP_WHITESPACE:
if ((md->ctypes[c] & ctype_space) == 0)
RRETURN(MATCH_NOMATCH);
break;
case OP_NOT_WORDCHAR:
if ((md->ctypes[c] & ctype_word) != 0)
RRETURN(MATCH_NOMATCH);
break;
case OP_WORDCHAR:
if ((md->ctypes[c] & ctype_word) == 0)
RRETURN(MATCH_NOMATCH);
break;
default:
RRETURN(PCRE_ERROR_INTERNAL);
}
}
}
/* Control never gets here */
}
/* If maximizing, it is worth using inline code for speed, doing
the type test once at the start (i.e. keep it out of the loop).
Again, keep the UTF-8 and UCP stuff separate. */
else {
pp = eptr; /* Remember where we started */
#ifdef SUPPORT_UCP
if (prop_type >= 0) {
switch (prop_type) {
case PT_ANY:
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
if (prop_fail_result)
break;
eptr += len;
}
break;
case PT_LAMP:
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if ((prop_chartype == ucp_Lu ||
prop_chartype == ucp_Ll ||
prop_chartype == ucp_Lt) ==
prop_fail_result)
break;
eptr += len;
}
break;
case PT_GC:
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if ((prop_category == prop_value) ==
prop_fail_result)
break;
eptr += len;
}
break;
case PT_PC:
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if ((prop_chartype == prop_value) ==
prop_fail_result)
break;
eptr += len;
}
break;
case PT_SC:
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if ((prop_script == prop_value) ==
prop_fail_result)
break;
eptr += len;
}
break;
}
/* eptr is now past the end of the maximum run */
if (possessive)
continue;
for (;;) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM44);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (eptr-- == pp)
break; /* Stop if tried at original pos */
if (utf8)
BACKCHAR(eptr);
}
}
/* Match extended Unicode sequences. We will get here only
if the support is in the binary; otherwise a compile-time
error occurs. */
else if (ctype == OP_EXTUNI) {
for (i = min; i < max; i++) {
if (eptr >= md->end_subject)
break;
GETCHARINCTEST(c, eptr);
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if (prop_category == ucp_M)
break;
while (eptr < md->end_subject) {
int len = 1;
if (!utf8)
c = *eptr;
else {
GETCHARLEN(c, eptr, len);
}
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if (prop_category != ucp_M)
break;
eptr += len;
}
}
/* eptr is now past the end of the maximum run */
if (possessive)
continue;
for (;;) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0,
RM45);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (eptr-- == pp)
break; /* Stop if tried at original pos */
for (;;) /* Move back over one extended */
{
int len = 1;
if (!utf8)
c = *eptr;
else {
BACKCHAR(eptr);
GETCHARLEN(c, eptr, len);
}
prop_category = _pcre_ucp_findprop(
c, &prop_chartype, &prop_script);
if (prop_category != ucp_M)
break;
eptr--;
}
}
}
else
#endif /* SUPPORT_UCP */
#ifdef SUPPORT_UTF8
/* UTF-8 mode */
if (utf8) {
switch (ctype) {
case OP_ANY:
if (max < INT_MAX) {
if ((ims & PCRE_DOTALL) == 0) {
for (i = min; i < max; i++) {
if (eptr >= md->end_subject ||
IS_NEWLINE(eptr))
break;
eptr++;
while (eptr < md->end_subject &&
(*eptr & 0xc0) == 0x80)
eptr++;
}
} else {
for (i = min; i < max; i++) {
if (eptr >= md->end_subject)
break;
eptr++;
while (eptr < md->end_subject &&
(*eptr & 0xc0) == 0x80)
eptr++;
}
}
}
/* Handle unlimited UTF-8 repeat */
else {
if ((ims & PCRE_DOTALL) == 0) {
for (i = min; i < max; i++) {
if (eptr >= md->end_subject ||
IS_NEWLINE(eptr))
break;
eptr++;
while (eptr < md->end_subject &&
(*eptr & 0xc0) == 0x80)
eptr++;
}
} else {
eptr = md->end_subject;
}
}
break;
/* The byte case is the same as non-UTF8 */
case OP_ANYBYTE:
c = max - min;
if (c >
(unsigned int)(md->end_subject - eptr))
c = md->end_subject - eptr;
eptr += c;
break;
case OP_ANYNL:
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
if (c == 0x000d) {
if (++eptr >= md->end_subject)
break;
if (*eptr == 0x000a)
eptr++;
} else {
if (c != 0x000a &&
(md->bsr_anycrlf ||
(c != 0x000b && c != 0x000c &&
c != 0x0085 && c != 0x2028 &&
c != 0x2029)))
break;
eptr += len;
}
}
break;
case OP_NOT_HSPACE:
case OP_HSPACE:
for (i = min; i < max; i++) {
BOOL gotspace;
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
switch (c) {
default:
gotspace = FALSE;
break;
case 0x09: /* HT */
case 0x20: /* SPACE */
case 0xa0: /* NBSP */
case 0x1680: /* OGHAM SPACE MARK */
case 0x180e: /* MONGOLIAN VOWEL
SEPARATOR */
case 0x2000: /* EN QUAD */
case 0x2001: /* EM QUAD */
case 0x2002: /* EN SPACE */
case 0x2003: /* EM SPACE */
case 0x2004: /* THREE-PER-EM SPACE
*/
case 0x2005: /* FOUR-PER-EM SPACE */
case 0x2006: /* SIX-PER-EM SPACE */
case 0x2007: /* FIGURE SPACE */
case 0x2008: /* PUNCTUATION SPACE */
case 0x2009: /* THIN SPACE */
case 0x200A: /* HAIR SPACE */
case 0x202f: /* NARROW NO-BREAK
SPACE */
case 0x205f: /* MEDIUM MATHEMATICAL
SPACE */
case 0x3000: /* IDEOGRAPHIC SPACE */
gotspace = TRUE;
break;
}
if (gotspace ==
(ctype == OP_NOT_HSPACE))
break;
eptr += len;
}
break;
case OP_NOT_VSPACE:
case OP_VSPACE:
for (i = min; i < max; i++) {
BOOL gotspace;
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
switch (c) {
default:
gotspace = FALSE;
break;
case 0x0a: /* LF */
case 0x0b: /* VT */
case 0x0c: /* FF */
case 0x0d: /* CR */
case 0x85: /* NEL */
case 0x2028: /* LINE SEPARATOR */
case 0x2029: /* PARAGRAPH SEPARATOR
*/
gotspace = TRUE;
break;
}
if (gotspace ==
(ctype == OP_NOT_VSPACE))
break;
eptr += len;
}
break;
case OP_NOT_DIGIT:
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
if (c < 256 &&
(md->ctypes[c] & ctype_digit) != 0)
break;
eptr += len;
}
break;
case OP_DIGIT:
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
if (c >= 256 ||
(md->ctypes[c] & ctype_digit) == 0)
break;
eptr += len;
}
break;
case OP_NOT_WHITESPACE:
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
if (c < 256 &&
(md->ctypes[c] & ctype_space) != 0)
break;
eptr += len;
}
break;
case OP_WHITESPACE:
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
if (c >= 256 ||
(md->ctypes[c] & ctype_space) == 0)
break;
eptr += len;
}
break;
case OP_NOT_WORDCHAR:
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
if (c < 256 &&
(md->ctypes[c] & ctype_word) != 0)
break;
eptr += len;
}
break;
case OP_WORDCHAR:
for (i = min; i < max; i++) {
int len = 1;
if (eptr >= md->end_subject)
break;
GETCHARLEN(c, eptr, len);
if (c >= 256 ||
(md->ctypes[c] & ctype_word) == 0)
break;
eptr += len;
}
break;
default:
RRETURN(PCRE_ERROR_INTERNAL);
}
/* eptr is now past the end of the maximum run */
if (possessive)
continue;
for (;;) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM46);
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
if (eptr-- == pp)
break; /* Stop if tried at original pos */
BACKCHAR(eptr);
}
} else
#endif /* SUPPORT_UTF8 */
/* Not UTF-8 mode */
{
switch (ctype) {
case OP_ANY:
if ((ims & PCRE_DOTALL) == 0) {
for (i = min; i < max; i++) {
if (eptr >= md->end_subject ||
IS_NEWLINE(eptr))
break;
eptr++;
}
break;
}
/* For DOTALL case, fall through and treat
* as \C */
case OP_ANYBYTE:
c = max - min;
if (c >
(unsigned int)(md->end_subject - eptr))
c = md->end_subject - eptr;
eptr += c;
break;
case OP_ANYNL:
for (i = min; i < max; i++) {
if (eptr >= md->end_subject)
break;
c = *eptr;
if (c == 0x000d) {
if (++eptr >= md->end_subject)
break;
if (*eptr == 0x000a)
eptr++;
} else {
if (c != 0x000a &&
(md->bsr_anycrlf ||
(c != 0x000b && c != 0x000c &&
c != 0x0085)))
break;
eptr++;
}
}
break;
case OP_NOT_HSPACE:
for (i = min; i < max; i++) {
if (eptr >= md->end_subject)
break;
c = *eptr;
if (c == 0x09 || c == 0x20 || c == 0xa0)
break;
eptr++;
}
break;
case OP_HSPACE:
for (i = min; i < max; i++) {
if (eptr >= md->end_subject)
break;
c = *eptr;
if (c != 0x09 && c != 0x20 && c != 0xa0)
break;
eptr++;
}
break;
case OP_NOT_VSPACE:
for (i = min; i < max; i++) {
if (eptr >= md->end_subject)
break;
c = *eptr;
if (c == 0x0a || c == 0x0b ||
c == 0x0c || c == 0x0d || c == 0x85)
break;
eptr++;
}
break;
case OP_VSPACE:
for (i = min; i < max; i++) {
if (eptr >= md->end_subject)
break;
c = *eptr;
if (c != 0x0a && c != 0x0b &&
c != 0x0c && c != 0x0d && c != 0x85)
break;
eptr++;
}
break;
case OP_NOT_DIGIT:
for (i = min; i < max; i++) {
if (eptr >= md->end_subject ||
(md->ctypes[*eptr] & ctype_digit) !=
0)
break;
eptr++;
}
break;
case OP_DIGIT:
for (i = min; i < max; i++) {
if (eptr >= md->end_subject ||
(md->ctypes[*eptr] & ctype_digit) ==
0)
break;
eptr++;
}
break;
case OP_NOT_WHITESPACE:
for (i = min; i < max; i++) {
if (eptr >= md->end_subject ||
(md->ctypes[*eptr] & ctype_space) !=
0)
break;
eptr++;
}
break;
case OP_WHITESPACE:
for (i = min; i < max; i++) {
if (eptr >= md->end_subject ||
(md->ctypes[*eptr] & ctype_space) ==
0)
break;
eptr++;
}
break;
case OP_NOT_WORDCHAR:
for (i = min; i < max; i++) {
if (eptr >= md->end_subject ||
(md->ctypes[*eptr] & ctype_word) !=
0)
break;
eptr++;
}
break;
case OP_WORDCHAR:
for (i = min; i < max; i++) {
if (eptr >= md->end_subject ||
(md->ctypes[*eptr] & ctype_word) ==
0)
break;
eptr++;
}
break;
default:
RRETURN(PCRE_ERROR_INTERNAL);
}
/* eptr is now past the end of the maximum run */
if (possessive)
continue;
while (eptr >= pp) {
RMATCH(eptr, ecode, offset_top, md, ims, eptrb,
0, RM47);
eptr--;
if (rrc != MATCH_NOMATCH)
RRETURN(rrc);
}
}
/* Get here if we can't make it match with any permitted
* repetitions */
RRETURN(MATCH_NOMATCH);
}
/* Control never gets here */
/* There's been some horrible disaster. Arrival here can only
mean there is something seriously wrong in the code above or the
OP_xxx definitions. */
default:
DPRINTF(("Unknown opcode %d\n", *ecode));
RRETURN(PCRE_ERROR_UNKNOWN_OPCODE);
}
/* Do not stick any code in here without much thought; it is assumed
that "continue" in the code above comes out to here to repeat the main
loop. */
} /* End of main loop */
/* Control never reaches here */
/* When compiling to use the heap rather than the stack for recursive calls
to match(), the RRETURN() macro jumps here. The number that is saved in
frame->Xwhere indicates which label we actually want to return to. */
#ifdef NO_RECURSE
#define LBL(val) \
case val: \
goto L_RM##val;
HEAP_RETURN:
switch (frame->Xwhere) {
LBL(1)
LBL(2)
LBL(3)
LBL(4)
LBL(5)
LBL(6)
LBL(7)
LBL(8)
LBL(9)
LBL(10)
LBL(11)
LBL(12) LBL(13) LBL(14) LBL(15) LBL(17) LBL(19) LBL(24) LBL(25) LBL(26)
LBL(27) LBL(29) LBL(31) LBL(33) LBL(35) LBL(43) LBL(47) LBL(48)
LBL(49) LBL(50) LBL(51) LBL(52) LBL(53) LBL(54)
#ifdef SUPPORT_UTF8
LBL(16) LBL(18) LBL(20) LBL(21) LBL(22) LBL(23) LBL(28)
LBL(30) LBL(32) LBL(34) LBL(42) LBL(46)
#ifdef SUPPORT_UCP
LBL(36) LBL(37) LBL(38) LBL(39) LBL(40) LBL(41)
LBL(44) LBL(45)
#endif /* SUPPORT_UCP */
#endif /* SUPPORT_UTF8 */
default
: DPRINTF(("jump error in pcre match: label %d non-existent\n",
frame->Xwhere));
return PCRE_ERROR_INTERNAL;
}
#undef LBL
#endif /* NO_RECURSE */
}
/***************************************************************************
****************************************************************************
RECURSION IN THE match() FUNCTION
Undefine all the macros that were defined above to handle this. */
#ifdef NO_RECURSE
#undef eptr
#undef ecode
#undef mstart
#undef offset_top
#undef ims
#undef eptrb
#undef flags
#undef callpat
#undef charptr
#undef data
#undef next
#undef pp
#undef prev
#undef saved_eptr
#undef new_recursive
#undef cur_is_word
#undef condition
#undef prev_is_word
#undef original_ims
#undef ctype
#undef length
#undef max
#undef min
#undef number
#undef offset
#undef op
#undef save_capture_last
#undef save_offset1
#undef save_offset2
#undef save_offset3
#undef stacksave
#undef newptrb
#endif
/* These two are defined as macros in both cases */
#undef fc
#undef fi
/***************************************************************************
***************************************************************************/
/*************************************************
* Execute a Regular Expression *
*************************************************/
/* This function applies a compiled re to a subject string and picks out
portions of the string if it matches. Two elements in the vector are set for
each substring: the offsets to the start and end of the substring.
Arguments:
argument_re points to the compiled expression
extra_data points to extra data or is NULL
subject points to the subject string
length length of subject string (may contain binary zeros)
start_offset where to start in the subject string
options option bits
offsets points to a vector of ints to be filled in with offsets
offsetcount the number of elements in the vector
Returns: > 0 => success; value is the number of elements filled in
= 0 => success, but offsets is not big enough
-1 => failed to match
< -1 => some kind of unexpected problem
*/
int pcre_exec(const pcre* argument_re,
const pcre_extra* extra_data,
PCRE_SPTR subject,
int length,
int start_offset,
int options,
int* offsets,
int offsetcount) {
int rc, resetcount, ocount;
int first_byte = -1;
int req_byte = -1;
int req_byte2 = -1;
int newline;
unsigned long int ims;
BOOL using_temporary_offsets = FALSE;
BOOL anchored;
BOOL startline;
BOOL firstline;
BOOL first_byte_caseless = FALSE;
BOOL req_byte_caseless = FALSE;
BOOL utf8;
match_data match_block;
match_data* md = &match_block;
const uschar* tables;
const uschar* start_bits = NULL;
USPTR start_match = (USPTR)subject + start_offset;
USPTR end_subject;
USPTR req_byte_ptr = start_match - 1;
pcre_study_data internal_study;
const pcre_study_data* study;
real_pcre internal_re;
const real_pcre* external_re = (const real_pcre*)argument_re;
const real_pcre* re = external_re;
/* Plausibility checks */
if ((options & ~PUBLIC_EXEC_OPTIONS) != 0)
return PCRE_ERROR_BADOPTION;
if (re == NULL || subject == NULL || (offsets == NULL && offsetcount > 0))
return PCRE_ERROR_NULL;
if (offsetcount < 0)
return PCRE_ERROR_BADCOUNT;
/* Fish out the optional data from the extra_data structure, first setting
the default values. */
study = NULL;
md->match_limit = MATCH_LIMIT;
md->match_limit_recursion = MATCH_LIMIT_RECURSION;
md->callout_data = NULL;
/* The table pointer is always in native byte order. */
tables = external_re->tables;
if (extra_data != NULL) {
register unsigned int flags = extra_data->flags;
if ((flags & PCRE_EXTRA_STUDY_DATA) != 0)
study = (const pcre_study_data*)extra_data->study_data;
if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0)
md->match_limit = extra_data->match_limit;
if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0)
md->match_limit_recursion = extra_data->match_limit_recursion;
if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0)
md->callout_data = extra_data->callout_data;
if ((flags & PCRE_EXTRA_TABLES) != 0)
tables = extra_data->tables;
}
/* If the exec call supplied NULL for tables, use the inbuilt ones. This
is a feature that makes it possible to save compiled regex and re-use them
in other programs later. */
if (tables == NULL)
tables = _pcre_default_tables;
/* Check that the first field in the block is the magic number. If it is
not, test for a regex that was compiled on a host of opposite endianness. If
this is the case, flipped values are put in internal_re and internal_study
if there was study data too. */
if (re->magic_number != MAGIC_NUMBER) {
re = _pcre_try_flipped(re, &internal_re, study, &internal_study);
if (re == NULL)
return PCRE_ERROR_BADMAGIC;
if (study != NULL)
study = &internal_study;
}
/* Set up other data */
anchored = ((re->options | options) & PCRE_ANCHORED) != 0;
startline = (re->flags & PCRE_STARTLINE) != 0;
firstline = (re->options & PCRE_FIRSTLINE) != 0;
/* The code starts after the real_pcre block and the capture name table. */
md->start_code = (const uschar*)external_re + re->name_table_offset +
re->name_count * re->name_entry_size;
md->start_subject = (USPTR)subject;
md->start_offset = start_offset;
md->end_subject = md->start_subject + length;
end_subject = md->end_subject;
md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;
md->notbol = (options & PCRE_NOTBOL) != 0;
md->noteol = (options & PCRE_NOTEOL) != 0;
md->notempty = (options & PCRE_NOTEMPTY) != 0;
md->partial = (options & PCRE_PARTIAL) != 0;
md->hitend = FALSE;
md->recursive = NULL; /* No recursion at top level */
md->lcc = tables + lcc_offset;
md->ctypes = tables + ctypes_offset;
/* Handle different \R options. */
switch (options & (PCRE_BSR_ANYCRLF | PCRE_BSR_UNICODE)) {
case 0:
if ((re->options & (PCRE_BSR_ANYCRLF | PCRE_BSR_UNICODE)) != 0)
md->bsr_anycrlf = (re->options & PCRE_BSR_ANYCRLF) != 0;
else
#ifdef BSR_ANYCRLF
md->bsr_anycrlf = TRUE;
#else
md->bsr_anycrlf = FALSE;
#endif
break;
case PCRE_BSR_ANYCRLF:
md->bsr_anycrlf = TRUE;
break;
case PCRE_BSR_UNICODE:
md->bsr_anycrlf = FALSE;
break;
default:
return PCRE_ERROR_BADNEWLINE;
}
/* Handle different types of newline. The three bits give eight cases. If
nothing is set at run time, whatever was used at compile time applies. */
switch ((((options & PCRE_NEWLINE_BITS) == 0) ? re->options
: (pcre_uint32)options) &
PCRE_NEWLINE_BITS) {
case 0:
newline = NEWLINE;
break; /* Compile-time default */
case PCRE_NEWLINE_CR:
newline = '\r';
break;
case PCRE_NEWLINE_LF:
newline = '\n';
break;
case PCRE_NEWLINE_CR + PCRE_NEWLINE_LF:
newline = ('\r' << 8) | '\n';
break;
case PCRE_NEWLINE_ANY:
newline = -1;
break;
case PCRE_NEWLINE_ANYCRLF:
newline = -2;
break;
default:
return PCRE_ERROR_BADNEWLINE;
}
if (newline == -2) {
md->nltype = NLTYPE_ANYCRLF;
} else if (newline < 0) {
md->nltype = NLTYPE_ANY;
} else {
md->nltype = NLTYPE_FIXED;
if (newline > 255) {
md->nllen = 2;
md->nl[0] = (newline >> 8) & 255;
md->nl[1] = newline & 255;
} else {
md->nllen = 1;
md->nl[0] = newline;
}
}
/* Partial matching is supported only for a restricted set of regexes at the
moment. */
if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
return PCRE_ERROR_BADPARTIAL;
/* Check a UTF-8 string if required. Unfortunately there's no way of
passing back the character offset. */
#ifdef SUPPORT_UTF8
if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0) {
if (_pcre_valid_utf8((uschar*)subject, length) >= 0)
return PCRE_ERROR_BADUTF8;
if (start_offset > 0 && start_offset < length) {
int tb = ((uschar*)subject)[start_offset];
if (tb > 127) {
tb &= 0xc0;
if (tb != 0 && tb != 0xc0)
return PCRE_ERROR_BADUTF8_OFFSET;
}
}
}
#endif
/* The ims options can vary during the matching as a result of the presence
of (?ims) items in the pattern. They are kept in a local variable so that
restoring at the exit of a group is easy. */
ims = re->options & (PCRE_CASELESS | PCRE_MULTILINE | PCRE_DOTALL);
/* If the expression has got more back references than the offsets supplied
can hold, we get a temporary chunk of working store to use during the
matching. Otherwise, we can use the vector supplied, rounding down its size
to a multiple of 3. */
ocount = offsetcount - (offsetcount % 3);
if (re->top_backref > 0 && re->top_backref >= ocount / 3) {
ocount = re->top_backref * 3 + 3;
md->offset_vector = (int*)(pcre_malloc)(ocount * sizeof(int));
if (md->offset_vector == NULL)
return PCRE_ERROR_NOMEMORY;
using_temporary_offsets = TRUE;
DPRINTF(("Got memory to hold back references\n"));
} else
md->offset_vector = offsets;
md->offset_end = ocount;
md->offset_max = (2 * ocount) / 3;
md->offset_overflow = FALSE;
md->capture_last = -1;
/* Compute the minimum number of offsets that we need to reset each time.
Doing this makes a huge difference to execution time when there aren't many
brackets in the pattern. */
resetcount = 2 + re->top_bracket * 2;
if (resetcount > offsetcount)
resetcount = ocount;
/* Reset the working variable associated with each extraction. These should
never be used unless previously set, but they get saved and restored, and so
we initialize them to avoid reading uninitialized locations. */
if (md->offset_vector != NULL) {
register int* iptr = md->offset_vector + ocount;
register int* iend = iptr - resetcount / 2 + 1;
while (--iptr >= iend)
*iptr = -1;
}
/* Set up the first character to match, if available. The first_byte value
is never set for an anchored regular expression, but the anchoring may be
forced at run time, so we have to test for anchoring. The first char may be
unset for an unanchored pattern, of course. If there's no first char and the
pattern was studied, there may be a bitmap of possible first characters. */
if (!anchored) {
if ((re->flags & PCRE_FIRSTSET) != 0) {
first_byte = re->first_byte & 255;
if ((first_byte_caseless =
((re->first_byte & REQ_CASELESS) != 0)) == TRUE)
first_byte = md->lcc[first_byte];
} else if (!startline && study != NULL &&
(study->options & PCRE_STUDY_MAPPED) != 0)
start_bits = study->start_bits;
}
/* For anchored or unanchored matches, there may be a "last known required
character" set. */
if ((re->flags & PCRE_REQCHSET) != 0) {
req_byte = re->req_byte & 255;
req_byte_caseless = (re->req_byte & REQ_CASELESS) != 0;
req_byte2 = (tables + fcc_offset)[req_byte]; /* case flipped */
}
/* ==========================================================================*/
/* Loop for handling unanchored repeated matching attempts; for anchored
regexs the loop runs just once. */
for (;;) {
USPTR save_end_subject = end_subject;
USPTR new_start_match;
/* Reset the maximum number of extractions we might see. */
if (md->offset_vector != NULL) {
register int* iptr = md->offset_vector;
register int* iend = iptr + resetcount;
while (iptr < iend)
*iptr++ = -1;
}
/* Advance to a unique first char if possible. If firstline is TRUE, the
start of the match is constrained to the first line of a multiline
string. That is, the match must be before or at the first newline.
Implement this by temporarily adjusting end_subject so that we stop
scanning at a newline. If the match fails at the newline, later code
breaks this loop. */
if (firstline) {
USPTR t = start_match;
while (t < md->end_subject && !IS_NEWLINE(t))
t++;
end_subject = t;
}
/* Now test for a unique first byte */
if (first_byte >= 0) {
if (first_byte_caseless)
while (start_match < end_subject &&
md->lcc[*start_match] != first_byte) {
NEXTCHAR(start_match);
}
else
while (start_match < end_subject &&
*start_match != first_byte) {
NEXTCHAR(start_match);
}
}
/* Or to just after a linebreak for a multiline match if possible */
else if (startline) {
if (start_match > md->start_subject + start_offset) {
while (start_match <= end_subject &&
!WAS_NEWLINE(start_match)) {
NEXTCHAR(start_match);
}
/* If we have just passed a CR and the newline option is ANY or
ANYCRLF, and we are now at a LF, advance the match position by
one more character.
*/
if (start_match[-1] == '\r' &&
(md->nltype == NLTYPE_ANY ||
md->nltype == NLTYPE_ANYCRLF) &&
start_match < end_subject && *start_match == '\n')
start_match++;
}
}
/* Or to a non-unique first char after study */
else if (start_bits != NULL) {
while (start_match < end_subject) {
register unsigned int c = *start_match;
if ((start_bits[c / 8] & (1 << (c & 7))) == 0) {
NEXTCHAR(start_match);
} else
break;
}
}
/* Restore fudged end_subject */
end_subject = save_end_subject;
#ifdef DEBUG /* Sigh. Some compilers never learn. */
printf(">>>> Match against: ");
pchars(start_match, end_subject - start_match, TRUE, md);
printf("\n");
#endif
/* If req_byte is set, we know that that character must appear in the
subject for the match to succeed. If the first character is set,
req_byte must be later in the subject; otherwise the test starts at the
match point. This optimization can save a huge amount of backtracking in
patterns with nested unlimited repeats that aren't going to match.
Writing separate code for cased/caseless versions makes it go faster, as
does using an autoincrement and backing off on a match.
HOWEVER: when the subject string is very, very long, searching to its
end can take a long time, and give bad performance on quite ordinary
patterns. This showed up when somebody was matching something like
/^\d+C/ on a 32-megabyte string... so we don't do this when the string
is sufficiently long.
ALSO: this processing is disabled when partial matching is requested.
*/
if (req_byte >= 0 && end_subject - start_match < REQ_BYTE_MAX &&
!md->partial) {
register USPTR p = start_match + ((first_byte >= 0) ? 1 : 0);
/* We don't need to repeat the search if we haven't yet reached the
place we found it at last time. */
if (p > req_byte_ptr) {
if (req_byte_caseless) {
while (p < end_subject) {
register int pp = *p++;
if (pp == req_byte || pp == req_byte2) {
p--;
break;
}
}
} else {
while (p < end_subject) {
if (*p++ == req_byte) {
p--;
break;
}
}
}
/* If we can't find the required character, break the matching
loop, forcing a match failure. */
if (p >= end_subject) {
rc = MATCH_NOMATCH;
break;
}
/* If we have found the required character, save the point where
we found it, so that we don't search again next time round the
loop if the start hasn't passed this character yet. */
req_byte_ptr = p;
}
}
/* OK, we can now run the match. */
md->start_match_ptr = start_match;
md->match_call_count = 0;
rc = match(start_match, md->start_code, start_match, 2, md, ims, NULL,
0, 0);
switch (rc) {
/* NOMATCH and PRUNE advance by one character. THEN at this
level acts exactly like PRUNE. */
case MATCH_NOMATCH:
case MATCH_PRUNE:
case MATCH_THEN:
new_start_match = start_match + 1;
#ifdef SUPPORT_UTF8
if (utf8)
while (new_start_match < end_subject &&
(*new_start_match & 0xc0) == 0x80)
new_start_match++;
#endif
break;
/* SKIP passes back the next starting point explicitly. */
case MATCH_SKIP:
new_start_match = md->start_match_ptr;
break;
/* COMMIT disables the bumpalong, but otherwise behaves as
* NOMATCH. */
case MATCH_COMMIT:
rc = MATCH_NOMATCH;
goto ENDLOOP;
/* Any other return is some kind of error. */
default:
goto ENDLOOP;
}
/* Control reaches here for the various types of "no match at this
point" result. Reset the code to MATCH_NOMATCH for subsequent checking.
*/
rc = MATCH_NOMATCH;
/* If PCRE_FIRSTLINE is set, the match must happen before or at the
first newline in the subject (though it may continue over the newline).
Therefore, if we have just failed to match, starting at a newline, do
not continue. */
if (firstline && IS_NEWLINE(start_match))
break;
/* Advance to new matching position */
start_match = new_start_match;
/* Break the loop if the pattern is anchored or if we have passed the
end of the subject. */
if (anchored || start_match > end_subject)
break;
/* If we have just passed a CR and we are now at a LF, and the pattern
does not contain any explicit matches for \r or \n, and the newline
option is CRLF or ANY or ANYCRLF, advance the match position by one more
character. */
if (start_match[-1] == '\r' && start_match < end_subject &&
*start_match == '\n' && (re->flags & PCRE_HASCRORLF) == 0 &&
(md->nltype == NLTYPE_ANY || md->nltype == NLTYPE_ANYCRLF ||
md->nllen == 2))
start_match++;
} /* End of for(;;) "bumpalong" loop */
/* ==========================================================================*/
/* We reach here when rc is not MATCH_NOMATCH, or if one of the stopping
conditions is true:
(1) The pattern is anchored or the match was failed by (*COMMIT);
(2) We are past the end of the subject;
(3) PCRE_FIRSTLINE is set and we have failed to match at a newline, because
this option requests that a match occur at or before the first newline
in the subject.
When we have a match and the offset vector is big enough to deal with any
backreferences, captured substring offsets will already be set up. In the
case where we had to get some local store to hold offsets for backreference
processing, copy those that we can. In this case there need not be overflow
if certain parts of the pattern were not used, even though there are more
capturing parentheses than vector slots. */
ENDLOOP:
if (rc == MATCH_MATCH) {
if (using_temporary_offsets) {
if (offsetcount >= 4) {
memcpy(offsets + 2, md->offset_vector + 2,
(offsetcount - 2) * sizeof(int));
DPRINTF(("Copied offsets from temporary memory\n"));
}
if (md->end_offset_top > offsetcount)
md->offset_overflow = TRUE;
DPRINTF(("Freeing temporary memory\n"));
(pcre_free)(md->offset_vector);
}
/* Set the return code to the number of captured strings, or 0 if there
are too many to fit into the vector. */
rc = md->offset_overflow ? 0 : md->end_offset_top / 2;
/* If there is space, set up the whole thing as substring 0. The value
of md->start_match_ptr might be modified if \K was encountered on the
success matching path. */
if (offsetcount < 2)
rc = 0;
else {
offsets[0] = md->start_match_ptr - md->start_subject;
offsets[1] = md->end_match_ptr - md->start_subject;
}
DPRINTF((">>>> returning %d\n", rc));
return rc;
}
/* Control gets here if there has been an error, or if the overall match
attempt has failed at all permitted starting positions. */
if (using_temporary_offsets) {
DPRINTF(("Freeing temporary memory\n"));
(pcre_free)(md->offset_vector);
}
if (rc != MATCH_NOMATCH) {
DPRINTF((">>>> error: returning %d\n", rc));
return rc;
} else if (md->partial && md->hitend) {
DPRINTF((">>>> returning PCRE_ERROR_PARTIAL\n"));
return PCRE_ERROR_PARTIAL;
} else {
DPRINTF((">>>> returning PCRE_ERROR_NOMATCH\n"));
return PCRE_ERROR_NOMATCH;
}
}
/* End of pcre_exec.c */
Reference in New Issue View Git Blame Copy Permalink