[ViewVC] Diff of: svn/ircd-hybrid-7.2/lib/pcre/pcre

Comparing ircd-hybrid-7.2/lib/pcre/pcre_exec.c (file contents):
Revision 962 by lusky, Sun Oct 2 21:05:51 2005 UTC vs.
Revision 963 by michael, Sat Aug 1 12:23:39 2009 UTC

#	Line 1 \| Line 1
1	–	/* $Id$ */
2	–
1		/*************************************************
2		* Perl-Compatible Regular Expressions *
3		*************************************************/
#	Line 8 \| Line 6
6		and semantics are as close as possible to those of the Perl 5 language.
7
8		Written by Philip Hazel
9	<	Copyright (c) 1997-2005 University of Cambridge
9	>	Copyright (c) 1997-2009 University of Cambridge
10
11		-----------------------------------------------------------------------------
12		Redistribution and use in source and binary forms, with or without
#	Line 45 \| Line 43 \| pattern matching using an NFA algorithm,
43		possible. There are also some static supporting functions. */
44
45
46	+	#define NLBLOCK md /* Block containing newline information */
47	+	#define PSSTART start_subject /* Field containing processed string start */
48	+	#define PSEND end_subject /* Field containing processed string end */
49	+
50		#include "pcre_internal.h"
51
52	+	/* Undefine some potentially clashing cpp symbols */
53
54	<	/* Structure for building a chain of data that actually lives on the
55	<	stack, for holding the values of the subject pointer at the start of each
53	<	subpattern, so as to detect when an empty string has been matched by a
54	<	subpattern - to break infinite loops. When NO_RECURSE is set, these blocks
55	<	are on the heap, not on the stack. */
56	<
57	<	typedef struct eptrblock {
58	<	struct eptrblock *epb_prev;
59	<	const uschar *epb_saved_eptr;
60	<	} eptrblock;
54	>	#undef min
55	>	#undef max
56
57		/* Flag bits for the match() function */
58
59	<	#define match_condassert 0x01 /* Called to check a condition assertion */
60	<	#define match_isgroup 0x02 /* Set if start of bracketed group */
59	>	#define match_condassert 0x01 /* Called to check a condition assertion */
60	>	#define match_cbegroup 0x02 /* Could-be-empty unlimited repeat group */
61
62		/* Non-error returns from the match() function. Error returns are externally
63		defined PCRE_ERROR_xxx codes, which are all negative. */
#	Line 70 \| Line 65 \| defined PCRE_ERROR_xxx codes, which are
65		#define MATCH_MATCH 1
66		#define MATCH_NOMATCH 0
67
68	+	/* Special internal returns from the match() function. Make them sufficiently
69	+	negative to avoid the external error codes. */
70	+
71	+	#define MATCH_COMMIT (-999)
72	+	#define MATCH_PRUNE (-998)
73	+	#define MATCH_SKIP (-997)
74	+	#define MATCH_THEN (-996)
75	+
76		/* Maximum number of ints of offset to save on the stack for recursive calls.
77		If the offset vector is bigger, malloc is used. This should be a multiple of 3,
78		because the offset vector is always a multiple of 3 long. */
#	Line 83 \| Line 86 \| static const char rep_max[] = { 0, 0, 0,
86
87
88
86	–	#ifdef DEBUG
87	–	/*************************************************
88	–	* Debugging function to print chars *
89	–	*************************************************/
90	–
91	–	/* Print a sequence of chars in printable format, stopping at the end of the
92	–	subject if the requested.
93	–
94	–	Arguments:
95	–	p points to characters
96	–	length number to print
97	–	is_subject TRUE if printing from within md->start_subject
98	–	md pointer to matching data block, if is_subject is TRUE
99	–
100	–	Returns: nothing
101	–	*/
102	–
103	–	static void
104	–	pchars(const uschar p, int length, BOOL is_subject, match_data md)
105	–	{
106	–	int c;
107	–	if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
108	–	while (length-- > 0)
109	–	if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);
110	–	}
111	–	#endif
89
90
91
#	Line 130 \| Line 107 \| Returns: TRUE if matched
107		*/
108
109		static BOOL
110	<	match_ref(int offset, register const uschar eptr, int length, match_data md,
110	>	match_ref(int offset, register USPTR eptr, int length, match_data *md,
111		unsigned long int ims)
112		{
113	<	const uschar *p = md->start_subject + md->offset_vector[offset];
113	>	USPTR p = md->start_subject + md->offset_vector[offset];
114
138	–	#ifdef DEBUG
139	–	if (eptr >= md->end_subject)
140	–	printf("matching subject <null>");
141	–	else
142	–	{
143	–	printf("matching subject ");
144	–	pchars(eptr, length, TRUE, md);
145	–	}
146	–	printf(" against backref ");
147	–	pchars(p, length, FALSE, md);
148	–	printf("\n");
149	–	#endif
115
116		/* Always fail if not enough characters left */
117
118		if (length > md->end_subject - eptr) return FALSE;
119
120	<	/* Separate the caselesss case for speed */
120	>	/* Separate the caseless case for speed. In UTF-8 mode we can only do this
121	>	properly if Unicode properties are supported. Otherwise, we can check only
122	>	ASCII characters. */
123
124		if ((ims & PCRE_CASELESS) != 0)
125		{
126	+	#ifdef SUPPORT_UTF8
127	+	#ifdef SUPPORT_UCP
128	+	if (md->utf8)
129	+	{
130	+	USPTR endptr = eptr + length;
131	+	while (eptr < endptr)
132	+	{
133	+	int c, d;
134	+	GETCHARINC(c, eptr);
135	+	GETCHARINC(d, p);
136	+	if (c != d && c != UCD_OTHERCASE(d)) return FALSE;
137	+	}
138	+	}
139	+	else
140	+	#endif
141	+	#endif
142	+
143	+	/* The same code works when not in UTF-8 mode and in UTF-8 mode when there
144	+	is no UCP support. */
145	+
146		while (length-- > 0)
147	<	if (md->lcc[p++] != md->lcc[eptr++]) return FALSE;
147	>	{ if (md->lcc[p++] != md->lcc[eptr++]) return FALSE; }
148		}
149	+
150	+	/* In the caseful case, we can just compare the bytes, whether or not we
151	+	are in UTF-8 mode. */
152	+
153		else
154		{ while (length-- > 0) if (p++ != eptr++) return FALSE; }
155
#	Line 171 \| Line 162 \| return TRUE;
162		****************************************************************************
163		RECURSION IN THE match() FUNCTION
164
165	<	The match() function is highly recursive. Some regular expressions can cause
166	<	it to recurse thousands of times. I was writing for Unix, so I just let it
167	<	call itself recursively. This uses the stack for saving everything that has
168	<	to be saved for a recursive call. On Unix, the stack can be large, and this
169	<	works fine.
170	<
171	<	It turns out that on non-Unix systems there are problems with programs that
172	<	use a lot of stack. (This despite the fact that every last chip has oodles
173	<	of memory these days, and techniques for extending the stack have been known
174	<	for decades.) So....
165	>	The match() function is highly recursive, though not every recursive call
166	>	increases the recursive depth. Nevertheless, some regular expressions can cause
167	>	it to recurse to a great depth. I was writing for Unix, so I just let it call
168	>	itself recursively. This uses the stack for saving everything that has to be
169	>	saved for a recursive call. On Unix, the stack can be large, and this works
170	>	fine.
171	>
172	>	It turns out that on some non-Unix-like systems there are problems with
173	>	programs that use a lot of stack. (This despite the fact that every last chip
174	>	has oodles of memory these days, and techniques for extending the stack have
175	>	been known for decades.) So....
176
177		There is a fudge, triggered by defining NO_RECURSE, which avoids recursive
178		calls by keeping local variables that need to be preserved in blocks of memory
179	<	obtained from malloc instead instead of on the stack. Macros are used to
179	>	obtained from malloc() instead instead of on the stack. Macros are used to
180		achieve this so that the actual code doesn't look very different to what it
181		always used to.
182	+
183	+	The original heap-recursive code used longjmp(). However, it seems that this
184	+	can be very slow on some operating systems. Following a suggestion from Stan
185	+	Switzer, the use of longjmp() has been abolished, at the cost of having to
186	+	provide a unique number for each call to RMATCH. There is no way of generating
187	+	a sequence of numbers at compile time in C. I have given them names, to make
188	+	them stand out more clearly.
189	+
190	+	Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
191	+	FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
192	+	tests. Furthermore, not using longjmp() means that local dynamic variables
193	+	don't have indeterminate values; this has meant that the frame size can be
194	+	reduced because the result can be "passed back" by straight setting of the
195	+	variable instead of being passed in the frame.
196		****************************************************************************
197		***************************************************************************/
198
199	+	/* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
200	+	below must be updated in sync. */
201
202	<	/* These versions of the macros use the stack, as normal */
202	>	enum { RM1=1, RM2, RM3, RM4, RM5, RM6, RM7, RM8, RM9, RM10,
203	>	RM11, RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
204	>	RM21, RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
205	>	RM31, RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
206	>	RM41, RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
207	>	RM51, RM52, RM53, RM54 };
208	>
209	>	/* These versions of the macros use the stack, as normal. There are debugging
210	>	versions and production versions. Note that the "rw" argument of RMATCH isn't
211	>	actuall used in this definition. */
212
213		#ifndef NO_RECURSE
214		#define REGISTER register
215	<	#define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) rx = match(ra,rb,rc,rd,re,rf,rg)
215	>
216	>	#define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \
217	>	rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1)
218		#define RRETURN(ra) return ra
219	+
220		#else
221
222
223	<	/* These versions of the macros manage a private stack on the heap. Note
224	<	that the rd argument of RMATCH isn't actually used. It's the md argument of
225	<	match(), which never changes. */
223	>	/* These versions of the macros manage a private stack on the heap. Note that
224	>	the "rd" argument of RMATCH isn't actually used in this definition. It's the md
225	>	argument of match(), which never changes. */
226
227		#define REGISTER
228
229	<	#define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\
229	>	#define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\
230		{\
231		heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\
232	<	if (setjmp(frame->Xwhere) == 0)\
233	<	{\
234	<	newframe->Xeptr = ra;\
235	<	newframe->Xecode = rb;\
236	<	newframe->Xoffset_top = rc;\
237	<	newframe->Xims = re;\
238	<	newframe->Xeptrb = rf;\
239	<	newframe->Xflags = rg;\
240	<	newframe->Xprevframe = frame;\
241	<	frame = newframe;\
242	<	DPRINTF(("restarting from line %d\n", __LINE__));\
243	<	goto HEAP_RECURSE;\
244	<	}\
245	<	else\
246	<	{\
227	<	DPRINTF(("longjumped back to line %d\n", __LINE__));\
228	<	frame = md->thisframe;\
229	<	rx = frame->Xresult;\
230	<	}\
232	>	frame->Xwhere = rw; \
233	>	newframe->Xeptr = ra;\
234	>	newframe->Xecode = rb;\
235	>	newframe->Xmstart = mstart;\
236	>	newframe->Xoffset_top = rc;\
237	>	newframe->Xims = re;\
238	>	newframe->Xeptrb = rf;\
239	>	newframe->Xflags = rg;\
240	>	newframe->Xrdepth = frame->Xrdepth + 1;\
241	>	newframe->Xprevframe = frame;\
242	>	frame = newframe;\
243	>	DPRINTF(("restarting from line %d\n", __LINE__));\
244	>	goto HEAP_RECURSE;\
245	>	L_##rw:\
246	>	DPRINTF(("jumped back to line %d\n", __LINE__));\
247		}
248
249		#define RRETURN(ra)\
#	Line 237 \| Line 253 \| *match(), which never changes. /**
253		(pcre_stack_free)(newframe);\
254		if (frame != NULL)\
255		{\
256	<	frame->Xresult = ra;\
257	<	md->thisframe = frame;\
242	<	longjmp(frame->Xwhere, 1);\
256	>	rrc = ra;\
257	>	goto HEAP_RETURN;\
258		}\
259		return ra;\
260		}
#	Line 252 \| Line 267 \| typedef struct heapframe {
267
268		/* Function arguments that may change */
269
270	<	const uschar *Xeptr;
270	>	USPTR Xeptr;
271		const uschar *Xecode;
272	+	USPTR Xmstart;
273		int Xoffset_top;
274		long int Xims;
275		eptrblock *Xeptrb;
276		int Xflags;
277	+	unsigned int Xrdepth;
278
279		/* Function local variables */
280
281	<	const uschar *Xcallpat;
282	<	const uschar *Xcharptr;
283	<	const uschar *Xdata;
284	<	const uschar *Xnext;
285	<	const uschar *Xpp;
286	<	const uschar *Xprev;
287	<	const uschar *Xsaved_eptr;
281	>	USPTR Xcallpat;
282	>	#ifdef SUPPORT_UTF8
283	>	USPTR Xcharptr;
284	>	#endif
285	>	USPTR Xdata;
286	>	USPTR Xnext;
287	>	USPTR Xpp;
288	>	USPTR Xprev;
289	>	USPTR Xsaved_eptr;
290
291		recursion_info Xnew_recursive;
292
293		BOOL Xcur_is_word;
294		BOOL Xcondition;
276	–	BOOL Xminimize;
295		BOOL Xprev_is_word;
296
297		unsigned long int Xoriginal_ims;
298
299	+	#ifdef SUPPORT_UCP
300	+	int Xprop_type;
301	+	int Xprop_value;
302	+	int Xprop_fail_result;
303	+	int Xprop_category;
304	+	int Xprop_chartype;
305	+	int Xprop_script;
306	+	int Xoclength;
307	+	uschar Xocchars[8];
308	+	#endif
309	+
310	+	int Xcodelink;
311		int Xctype;
312	<	int Xfc;
312	>	unsigned int Xfc;
313		int Xfi;
314		int Xlength;
315		int Xmax;
#	Line 293 \| Line 323 \| typedef struct heapframe {
323
324		eptrblock Xnewptrb;
325
326	<	/* Place to pass back result, and where to jump back to */
326	>	/* Where to jump back to */
327
328	<	int Xresult;
299	<	jmp_buf Xwhere;
328	>	int Xwhere;
329
330		} heapframe;
331
#	Line 312 \| Line 341 \| typedef struct heapframe {
341		* Match from current position *
342		*************************************************/
343
344	<	/* On entry ecode points to the first opcode, and eptr to the first character
316	<	in the subject string, while eptrb holds the value of eptr at the start of the
317	<	last bracketed group - used for breaking infinite loops matching zero-length
318	<	strings. This function is called recursively in many circumstances. Whenever it
344	>	/* This function is called recursively in many circumstances. Whenever it
345		returns a negative (error) response, the outer incarnation must also return the
346		same response.
347
#	Line 325 \| Line 351 \| performance. Tests using gcc on a SPARC
351		made performance worse.
352
353		Arguments:
354	<	eptr pointer in subject
355	<	ecode position in code
354	>	eptr pointer to current character in subject
355	>	ecode pointer to current position in compiled code
356	>	mstart pointer to the current match start position (can be modified
357	>	by encountering \K)
358		offset_top current top pointer
359		md pointer to "static" info for the match
360		ims current /i, /m, and /s options
#	Line 334 \| Line 362 \| Arguments:
362		brackets - for testing for empty matches
363		flags can contain
364		match_condassert - this is an assertion condition
365	<	match_isgroup - this is the start of a bracketed group
365	>	match_cbegroup - this is the start of an unlimited repeat
366	>	group that can match an empty string
367	>	rdepth the recursion depth
368
369		Returns: MATCH_MATCH if matched ) these values are >= 0
370		MATCH_NOMATCH if failed to match )
371		a negative PCRE_ERROR_xxx value if aborted by an error condition
372	<	(e.g. stopped by recursion limit)
372	>	(e.g. stopped by repeated call or recursion limit)
373		*/
374
375		static int
376	<	match(REGISTER const uschar eptr, REGISTER const uschar ecode,
376	>	match(REGISTER USPTR eptr, REGISTER const uschar *ecode, USPTR mstart,
377		int offset_top, match_data md, unsigned long int ims, eptrblock eptrb,
378	<	int flags)
378	>	int flags, unsigned int rdepth)
379		{
380		/* These variables do not need to be preserved over recursion in this function,
381	<	so they can be ordinary variables in all cases. Mark them with "register"
382	<	because they are used a lot in loops. */
381	>	so they can be ordinary variables in all cases. Mark some of them with
382	>	"register" because they are used a lot in loops. */
383
384	<	register int rrc; /* Returns from recursive calls */
385	<	register int i; /* Used for loops not involving calls to RMATCH() */
386	<	register int c; /* Character values not kept over RMATCH() calls */
387	<	register BOOL utf8; /* Local copy of UTF-8 flag for speed */
384	>	register int rrc; /* Returns from recursive calls */
385	>	register int i; /* Used for loops not involving calls to RMATCH() */
386	>	register unsigned int c; /* Character values not kept over RMATCH() calls */
387	>	register BOOL utf8; /* Local copy of UTF-8 flag for speed */
388	>
389	>	BOOL minimize, possessive; /* Quantifier options */
390	>	int condcode;
391
392		/* When recursion is not being used, all "local" variables that have to be
393		preserved over calls to RMATCH() are part of a "frame" which is obtained from
#	Line 369 \| Line 402 \| frame->Xprevframe = NULL; /*
402
403		frame->Xeptr = eptr;
404		frame->Xecode = ecode;
405	+	frame->Xmstart = mstart;
406		frame->Xoffset_top = offset_top;
407		frame->Xims = ims;
408		frame->Xeptrb = eptrb;
409		frame->Xflags = flags;
410	+	frame->Xrdepth = rdepth;
411
412		/* This is where control jumps back to to effect "recursion" */
413
#	Line 382 \| Line 417 \| HEAP_RECURSE:
417
418		#define eptr frame->Xeptr
419		#define ecode frame->Xecode
420	+	#define mstart frame->Xmstart
421		#define offset_top frame->Xoffset_top
422		#define ims frame->Xims
423		#define eptrb frame->Xeptrb
424		#define flags frame->Xflags
425	+	#define rdepth frame->Xrdepth
426
427		/* Ditto for the local variables */
428
429	+	#ifdef SUPPORT_UTF8
430	+	#define charptr frame->Xcharptr
431	+	#endif
432		#define callpat frame->Xcallpat
433	+	#define codelink frame->Xcodelink
434		#define data frame->Xdata
435		#define next frame->Xnext
436		#define pp frame->Xpp
#	Line 400 \| Line 441 \| HEAP_RECURSE:
441
442		#define cur_is_word frame->Xcur_is_word
443		#define condition frame->Xcondition
403	–	#define minimize frame->Xminimize
444		#define prev_is_word frame->Xprev_is_word
445
446		#define original_ims frame->Xoriginal_ims
447
448	+	#ifdef SUPPORT_UCP
449	+	#define prop_type frame->Xprop_type
450	+	#define prop_value frame->Xprop_value
451	+	#define prop_fail_result frame->Xprop_fail_result
452	+	#define prop_category frame->Xprop_category
453	+	#define prop_chartype frame->Xprop_chartype
454	+	#define prop_script frame->Xprop_script
455	+	#define oclength frame->Xoclength
456	+	#define occhars frame->Xocchars
457	+	#endif
458	+
459		#define ctype frame->Xctype
460		#define fc frame->Xfc
461		#define fi frame->Xfi
#	Line 426 \| Line 477 \| HEAP_RECURSE:
477		get preserved during recursion in the normal way. In this environment, fi and
478		i, and fc and c, can be the same variables. */
479
480	<	#else
480	>	#else /* NO_RECURSE not defined */
481		#define fi i
482		#define fc c
483
484
485	<	const uschar callpat; / them within each of those blocks. */
486	<	const uschar data; / However, in order to accommodate the */
487	<	const uschar next; / version of this code that uses an */
488	<	const uschar pp; / external "stack" implemented on the */
489	<	const uschar prev; / heap, it is easier to declare them */
490	<	const uschar saved_eptr; / all here, so the declarations can */
491	<	/* be cut out in a block. The only */
492	<	recursion_info new_recursive; /* declarations within blocks below are */
493	<	/* for variables that do not have to */
494	<	BOOL cur_is_word; /* be preserved over a recursive call */
495	<	BOOL condition; /* to RMATCH(). */
496	<	BOOL minimize;
485	>	#ifdef SUPPORT_UTF8 /* Many of these variables are used only */
486	>	const uschar charptr; / in small blocks of the code. My normal */
487	>	#endif /* style of coding would have declared */
488	>	const uschar callpat; / them within each of those blocks. */
489	>	const uschar data; / However, in order to accommodate the */
490	>	const uschar next; / version of this code that uses an */
491	>	USPTR pp; /* external "stack" implemented on the */
492	>	const uschar prev; / heap, it is easier to declare them all */
493	>	USPTR saved_eptr; /* here, so the declarations can be cut */
494	>	/* out in a block. The only declarations */
495	>	recursion_info new_recursive; /* within blocks below are for variables */
496	>	/* that do not have to be preserved over */
497	>	BOOL cur_is_word; /* a recursive call to RMATCH(). */
498	>	BOOL condition;
499		BOOL prev_is_word;
500
501		unsigned long int original_ims;
502
503	+	#ifdef SUPPORT_UCP
504	+	int prop_type;
505	+	int prop_value;
506	+	int prop_fail_result;
507	+	int prop_category;
508	+	int prop_chartype;
509	+	int prop_script;
510	+	int oclength;
511	+	uschar occhars[8];
512	+	#endif
513
514	+	int codelink;
515		int ctype;
516		int length;
517		int max;
#	Line 460 \| Line 524 \| int save_offset1, save_offset2, save_off
524		int stacksave[REC_STACK_SAVE_MAX];
525
526		eptrblock newptrb;
527	+	#endif /* NO_RECURSE */
528	+
529	+	/* These statements are here to stop the compiler complaining about unitialized
530	+	variables. */
531	+
532	+	#ifdef SUPPORT_UCP
533	+	prop_value = 0;
534	+	prop_fail_result = 0;
535		#endif
536
537	<	/* OK, now we can get on with the real code of the function. Recursion is
538	<	specified by the macros RMATCH and RRETURN. When NO_RECURSE is not defined,
539	<	these just turn into a recursive call to match() and a "return", respectively.
540	<	However, RMATCH isn't like a function call because it's quite a complicated
541	<	macro. It has to be used in one particular way. This shouldn't, however, impact
542	<	performance when true recursion is being used. */
537	>
538	>	/* This label is used for tail recursion, which is used in a few cases even
539	>	when NO_RECURSE is not defined, in order to reduce the amount of stack that is
540	>	used. Thanks to Ian Taylor for noticing this possibility and sending the
541	>	original patch. */
542	>
543	>	TAIL_RECURSE:
544	>
545	>	/* OK, now we can get on with the real code of the function. Recursive calls
546	>	are specified by the macro RMATCH and RRETURN is used to return. When
547	>	NO_RECURSE is not defined, these just turn into a recursive call to match()
548	>	and a "return", respectively (possibly with some debugging if DEBUG is
549	>	defined). However, RMATCH isn't like a function call because it's quite a
550	>	complicated macro. It has to be used in one particular way. This shouldn't,
551	>	however, impact performance when true recursion is being used. */
552	>
553	>	#ifdef SUPPORT_UTF8
554	>	utf8 = md->utf8; /* Local copy of the flag */
555	>	#else
556	>	utf8 = FALSE;
557	>	#endif
558	>
559	>	/* First check that we haven't called match() too many times, or that we
560	>	haven't exceeded the recursive call limit. */
561
562		if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
563	+	if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
564
565		original_ims = ims; /* Save for resetting on ')' */
475	–	utf8 = md->utf8; /* Local copy of the flag */
566
567	<	/* At the start of a bracketed group, add the current subject pointer to the
568	<	stack of such pointers, to be re-instated at the end of the group when we hit
569	<	the closing ket. When match() is called in other circumstances, we don't add to
570	<	this stack. */
567	>	/* At the start of a group with an unlimited repeat that may match an empty
568	>	string, the match_cbegroup flag is set. When this is the case, add the current
569	>	subject pointer to the chain of such remembered pointers, to be checked when we
570	>	hit the closing ket, in order to break infinite loops that match no characters.
571	>	When match() is called in other circumstances, don't add to the chain. The
572	>	match_cbegroup flag must NOT be used with tail recursion, because the memory
573	>	block that is used is on the stack, so a new one may be required for each
574	>	match(). */
575
576	<	if ((flags & match_isgroup) != 0)
576	>	if ((flags & match_cbegroup) != 0)
577		{
484	–	newptrb.epb_prev = eptrb;
578		newptrb.epb_saved_eptr = eptr;
579	+	newptrb.epb_prev = eptrb;
580		eptrb = &newptrb;
581		}
582
583	<	/* Now start processing the operations. */
583	>	/* Now start processing the opcodes. */
584
585		for (;;)
586		{
587	+	minimize = possessive = FALSE;
588		op = *ecode;
494	–	minimize = FALSE;
589
590		/* For partial matching, remember if we ever hit the end of the subject after
591		matching at least one subject character. */
592
593		if (md->partial &&
594		eptr >= md->end_subject &&
595	<	eptr > md->start_match)
595	>	eptr > mstart)
596		md->hitend = TRUE;
597
598	<	/* Opening capturing bracket. If there is space in the offset vector, save
505	<	the current subject position in the working slot at the top of the vector. We
506	<	mustn't change the current values of the data slot, because they may be set
507	<	from a previous iteration of this group, and be referred to by a reference
508	<	inside the group.
509	<
510	<	If the bracket fails to match, we need to restore this value and also the
511	<	values of the final offsets, in case they were set by a previous iteration of
512	<	the same bracket.
513	<
514	<	If there isn't enough space in the offset vector, treat this as if it were a
515	<	non-capturing bracket. Don't worry about setting the flag for the error case
516	<	here; that is handled in the code for KET. */
517	<
518	<	if (op > OP_BRA)
598	>	switch(op)
599		{
600	<	number = op - OP_BRA;
601	<
522	<	/* For extended extraction brackets (large number), we have to fish out the
523	<	number from a dummy opcode at the start. */
600	>	case OP_FAIL:
601	>	RRETURN(MATCH_NOMATCH);
602
603	<	if (number > EXTRACT_BASIC_MAX)
604	<	number = GET2(ecode, 2+LINK_SIZE);
603	>	case OP_PRUNE:
604	>	RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
605	>	ims, eptrb, flags, RM51);
606	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
607	>	RRETURN(MATCH_PRUNE);
608	>
609	>	case OP_COMMIT:
610	>	RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
611	>	ims, eptrb, flags, RM52);
612	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
613	>	RRETURN(MATCH_COMMIT);
614	>
615	>	case OP_SKIP:
616	>	RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
617	>	ims, eptrb, flags, RM53);
618	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
619	>	md->start_match_ptr = eptr; /* Pass back current position */
620	>	RRETURN(MATCH_SKIP);
621	>
622	>	case OP_THEN:
623	>	RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
624	>	ims, eptrb, flags, RM54);
625	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
626	>	RRETURN(MATCH_THEN);
627	>
628	>	/* Handle a capturing bracket. If there is space in the offset vector, save
629	>	the current subject position in the working slot at the top of the vector.
630	>	We mustn't change the current values of the data slot, because they may be
631	>	set from a previous iteration of this group, and be referred to by a
632	>	reference inside the group.
633	>
634	>	If the bracket fails to match, we need to restore this value and also the
635	>	values of the final offsets, in case they were set by a previous iteration
636	>	of the same bracket.
637	>
638	>	If there isn't enough space in the offset vector, treat this as if it were
639	>	a non-capturing bracket. Don't worry about setting the flag for the error
640	>	case here; that is handled in the code for KET. */
641	>
642	>	case OP_CBRA:
643	>	case OP_SCBRA:
644	>	number = GET2(ecode, 1+LINK_SIZE);
645		offset = number << 1;
646
529	–	#ifdef DEBUG
530	–	printf("start bracket %d subject=", number);
531	–	pchars(eptr, 16, TRUE, md);
532	–	printf("\n");
533	–	#endif
534	–
647		if (offset < md->offset_max)
648		{
649		save_offset1 = md->offset_vector[offset];
#	Line 542 \| Line 654 \| for (;;)
654		DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
655		md->offset_vector[md->offset_end - number] = eptr - md->start_subject;
656
657	+	flags = (op == OP_SCBRA)? match_cbegroup : 0;
658		do
659		{
660	<	RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,
661	<	match_isgroup);
662	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
660	>	RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,
661	>	ims, eptrb, flags, RM1);
662	>	if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
663		md->capture_last = save_capture_last;
664		ecode += GET(ecode, 1);
665		}
#	Line 561 \| Line 674 \| for (;;)
674		RRETURN(MATCH_NOMATCH);
675		}
676
677	<	/* Insufficient room for saving captured contents */
677	>	/* FALL THROUGH ... Insufficient room for saving captured contents. Treat
678	>	as a non-capturing bracket. */
679
680	<	else op = OP_BRA;
681	<	}
680	>	/* VVVVVVVVVVVVVVVVVVVVVVVVV */
681	>	/* VVVVVVVVVVVVVVVVVVVVVVVVV */
682
683	<	/* Other types of node can be handled by a switch */
683	>	DPRINTF(("insufficient capture room: treat as non-capturing\n"));
684
685	<	switch(op)
686	<	{
687	<	case OP_BRA: /* Non-capturing bracket: optimized */
688	<	DPRINTF(("start bracket 0\n"));
689	<	do
685	>	/* VVVVVVVVVVVVVVVVVVVVVVVVV */
686	>	/* VVVVVVVVVVVVVVVVVVVVVVVVV */
687	>
688	>	/* Non-capturing bracket. Loop for all the alternatives. When we get to the
689	>	final alternative within the brackets, we would return the result of a
690	>	recursive call to match() whatever happened. We can reduce stack usage by
691	>	turning this into a tail recursion, except in the case when match_cbegroup
692	>	is set.*/
693	>
694	>	case OP_BRA:
695	>	case OP_SBRA:
696	>	DPRINTF(("start non-capturing bracket\n"));
697	>	flags = (op >= OP_SBRA)? match_cbegroup : 0;
698	>	for (;;)
699		{
700	<	RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,
701	<	match_isgroup);
702	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
700	>	if (ecode[GET(ecode, 1)] != OP_ALT) /* Final alternative */
701	>	{
702	>	if (flags == 0) /* Not a possibly empty group */
703	>	{
704	>	ecode += _pcre_OP_lengths[*ecode];
705	>	DPRINTF(("bracket 0 tail recursion\n"));
706	>	goto TAIL_RECURSE;
707	>	}
708	>
709	>	/* Possibly empty group; can't use tail recursion. */
710	>
711	>	RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,
712	>	eptrb, flags, RM48);
713	>	RRETURN(rrc);
714	>	}
715	>
716	>	/* For non-final alternatives, continue the loop for a NOMATCH result;
717	>	otherwise return. */
718	>
719	>	RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,
720	>	eptrb, flags, RM2);
721	>	if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
722		ecode += GET(ecode, 1);
723		}
724	<	while (*ecode == OP_ALT);
583	<	DPRINTF(("bracket 0 failed\n"));
584	<	RRETURN(MATCH_NOMATCH);
724	>	/* Control never reaches here. */
725
726		/* Conditional group: compilation checked that there are no more than
727		two branches. If the condition is false, skipping the first branch takes us
728		past the end if there is only one branch, but that's OK because that is
729	<	exactly what going to the ket would do. */
729	>	exactly what going to the ket would do. As there is only one branch to be
730	>	obeyed, we can use tail recursion to avoid using another stack frame. */
731
732		case OP_COND:
733	<	if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */
733	>	case OP_SCOND:
734	>	codelink= GET(ecode, 1);
735	>
736	>	/* Because of the way auto-callout works during compile, a callout item is
737	>	inserted between OP_COND and an assertion condition. */
738	>
739	>	if (ecode[LINK_SIZE+1] == OP_CALLOUT)
740	>	{
741	>	if (pcre_callout != NULL)
742	>	{
743	>	pcre_callout_block cb;
744	>	cb.version = 1; /* Version 1 of the callout block */
745	>	cb.callout_number = ecode[LINK_SIZE+2];
746	>	cb.offset_vector = md->offset_vector;
747	>	cb.subject = (PCRE_SPTR)md->start_subject;
748	>	cb.subject_length = md->end_subject - md->start_subject;
749	>	cb.start_match = mstart - md->start_subject;
750	>	cb.current_position = eptr - md->start_subject;
751	>	cb.pattern_position = GET(ecode, LINK_SIZE + 3);
752	>	cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
753	>	cb.capture_top = offset_top/2;
754	>	cb.capture_last = md->capture_last;
755	>	cb.callout_data = md->callout_data;
756	>	if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);
757	>	if (rrc < 0) RRETURN(rrc);
758	>	}
759	>	ecode += _pcre_OP_lengths[OP_CALLOUT];
760	>	}
761	>
762	>	condcode = ecode[LINK_SIZE+1];
763	>
764	>	/* Now see what the actual condition is */
765	>
766	>	if (condcode == OP_RREF) /* Recursion test */
767	>	{
768	>	offset = GET2(ecode, LINK_SIZE + 2); /* Recursion group number*/
769	>	condition = md->recursive != NULL &&
770	>	(offset == RREF_ANY \|\| offset == md->recursive->group_num);
771	>	ecode += condition? 3 : GET(ecode, 1);
772	>	}
773	>
774	>	else if (condcode == OP_CREF) /* Group used test */
775		{
776		offset = GET2(ecode, LINK_SIZE+2) << 1; /* Doubled ref number */
777	<	condition = (offset == CREF_RECURSE * 2)?
778	<	(md->recursive != NULL) :
779	<	(offset < offset_top && md->offset_vector[offset] >= 0);
780	<	RMATCH(rrc, eptr, ecode + (condition?
781	<	(LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1))),
782	<	offset_top, md, ims, eptrb, match_isgroup);
783	<	RRETURN(rrc);
777	>	condition = offset < offset_top && md->offset_vector[offset] >= 0;
778	>	ecode += condition? 3 : GET(ecode, 1);
779	>	}
780	>
781	>	else if (condcode == OP_DEF) /* DEFINE - always false */
782	>	{
783	>	condition = FALSE;
784	>	ecode += GET(ecode, 1);
785		}
786
787		/* The condition is an assertion. Call match() to evaluate it - setting
788	<	the final argument TRUE causes it to stop at the end of an assertion. */
788	>	the final argument match_condassert causes it to stop at the end of an
789	>	assertion. */
790
791		else
792		{
793	<	RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,
794	<	match_condassert \| match_isgroup);
793	>	RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,
794	>	match_condassert, RM3);
795		if (rrc == MATCH_MATCH)
796		{
797	<	ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);
797	>	condition = TRUE;
798	>	ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
799		while (*ecode == OP_ALT) ecode += GET(ecode, 1);
800		}
801	<	else if (rrc != MATCH_NOMATCH)
801	>	else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
802		{
803		RRETURN(rrc); /* Need braces because of following else */
804		}
805	<	else ecode += GET(ecode, 1);
806	<	RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,
807	<	match_isgroup);
808	<	RRETURN(rrc);
805	>	else
806	>	{
807	>	condition = FALSE;
808	>	ecode += codelink;
809	>	}
810		}
625	–	/* Control never reaches here */
811
812	<	/* Skip over conditional reference or large extraction number data if
813	<	encountered. */
812	>	/* We are now at the branch that is to be obeyed. As there is only one,
813	>	we can use tail recursion to avoid using another stack frame, except when
814	>	match_cbegroup is required for an unlimited repeat of a possibly empty
815	>	group. If the second alternative doesn't exist, we can just plough on. */
816
817	<	case OP_CREF:
818	<	case OP_BRANUMBER:
819	<	ecode += 3;
817	>	if (condition \|\| *ecode == OP_ALT)
818	>	{
819	>	ecode += 1 + LINK_SIZE;
820	>	if (op == OP_SCOND) /* Possibly empty group */
821	>	{
822	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);
823	>	RRETURN(rrc);
824	>	}
825	>	else /* Group must match something */
826	>	{
827	>	flags = 0;
828	>	goto TAIL_RECURSE;
829	>	}
830	>	}
831	>	else /* Condition false & no alternative */
832	>	{
833	>	ecode += 1 + LINK_SIZE;
834	>	}
835		break;
836
635	–	/* End of the pattern. If we are in a recursion, we should restore the
636	–	offsets appropriately and continue from after the call. */
837
838	+	/* End of the pattern, either real or forced. If we are in a top-level
839	+	recursion, we should restore the offsets appropriately and continue from
840	+	after the call. */
841	+
842	+	case OP_ACCEPT:
843		case OP_END:
844		if (md->recursive != NULL && md->recursive->group_num == 0)
845		{
846		recursion_info *rec = md->recursive;
847	<	DPRINTF(("Hit the end in a (?0) recursion\n"));
847	>	DPRINTF(("End of pattern in a (?0) recursion\n"));
848		md->recursive = rec->prevrec;
849		memmove(md->offset_vector, rec->offset_save,
850		rec->saved_max * sizeof(int));
851	<	md->start_match = rec->save_start;
851	>	mstart = rec->save_start;
852		ims = original_ims;
853		ecode = rec->after_call;
854		break;
#	Line 652 \| Line 857 \| for (;;)
857		/* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty
858		string - backtracking will then try other alternatives, if any. */
859
860	<	if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);
861	<	md->end_match_ptr = eptr; /* Record where we ended */
862	<	md->end_offset_top = offset_top; /* and how many extracts were taken */
860	>	if (md->notempty && eptr == mstart) RRETURN(MATCH_NOMATCH);
861	>	md->end_match_ptr = eptr; /* Record where we ended */
862	>	md->end_offset_top = offset_top; /* and how many extracts were taken */
863	>	md->start_match_ptr = mstart; /* and the start (\K can modify) */
864		RRETURN(MATCH_MATCH);
865
866		/* Change option settings */
#	Line 675 \| Line 881 \| for (;;)
881		case OP_ASSERTBACK:
882		do
883		{
884	<	RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,
885	<	match_isgroup);
884	>	RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,
885	>	RM4);
886		if (rrc == MATCH_MATCH) break;
887	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
887	>	if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
888		ecode += GET(ecode, 1);
889		}
890		while (*ecode == OP_ALT);
#	Line 702 \| Line 908 \| for (;;)
908		case OP_ASSERTBACK_NOT:
909		do
910		{
911	<	RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,
912	<	match_isgroup);
911	>	RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,
912	>	RM5);
913		if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);
914	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
914	>	if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
915		ecode += GET(ecode,1);
916		}
917		while (*ecode == OP_ALT);
#	Line 721 \| Line 927 \| for (;;)
927		back a number of characters, not bytes. */
928
929		case OP_REVERSE:
930	+	#ifdef SUPPORT_UTF8
931	+	if (utf8)
932	+	{
933	+	i = GET(ecode, 1);
934	+	while (i-- > 0)
935	+	{
936	+	eptr--;
937	+	if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
938	+	BACKCHAR(eptr);
939	+	}
940	+	}
941	+	else
942	+	#endif
943	+
944		/* No UTF-8 support, or not in UTF-8 mode: count is byte count */
945
946		{
947	<	eptr -= GET(ecode,1);
947	>	eptr -= GET(ecode, 1);
948		if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
949		}
950
#	Line 744 \| Line 964 \| for (;;)
964		cb.version = 1; /* Version 1 of the callout block */
965		cb.callout_number = ecode[1];
966		cb.offset_vector = md->offset_vector;
967	<	cb.subject = (const char *)md->start_subject;
967	>	cb.subject = (PCRE_SPTR)md->start_subject;
968		cb.subject_length = md->end_subject - md->start_subject;
969	<	cb.start_match = md->start_match - md->start_subject;
969	>	cb.start_match = mstart - md->start_subject;
970		cb.current_position = eptr - md->start_subject;
971		cb.pattern_position = GET(ecode, 2);
972		cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
#	Line 781 \| Line 1001 \| for (;;)
1001		case OP_RECURSE:
1002		{
1003		callpat = md->start_code + GET(ecode, 1);
1004	<	new_recursive.group_num = *callpat - OP_BRA;
1005	<
786	<	/* For extended extraction brackets (large number), we have to fish out
787	<	the number from a dummy opcode at the start. */
788	<
789	<	if (new_recursive.group_num > EXTRACT_BASIC_MAX)
790	<	new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);
1004	>	new_recursive.group_num = (callpat == md->start_code)? 0 :
1005	>	GET2(callpat, 1 + LINK_SIZE);
1006
1007		/* Add to "recursing stack" */
1008
#	Line 813 \| Line 1028 \| for (;;)
1028
1029		memcpy(new_recursive.offset_save, md->offset_vector,
1030		new_recursive.saved_max * sizeof(int));
1031	<	new_recursive.save_start = md->start_match;
1032	<	md->start_match = eptr;
1031	>	new_recursive.save_start = mstart;
1032	>	mstart = eptr;
1033
1034		/* OK, now we can do the recursion. For each top-level alternative we
1035		restore the offset and recursion data. */
1036
1037		DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1038	+	flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;
1039		do
1040		{
1041	<	RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,
1042	<	eptrb, match_isgroup);
1041	>	RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,
1042	>	md, ims, eptrb, flags, RM6);
1043		if (rrc == MATCH_MATCH)
1044		{
1045	+	DPRINTF(("Recursion matched\n"));
1046		md->recursive = new_recursive.prevrec;
1047		if (new_recursive.offset_save != stacksave)
1048		(pcre_free)(new_recursive.offset_save);
1049		RRETURN(MATCH_MATCH);
1050		}
1051	<	else if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1051	>	else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1052	>	{
1053	>	DPRINTF(("Recursion gave error %d\n", rrc));
1054	>	if (new_recursive.offset_save != stacksave)
1055	>	(pcre_free)(new_recursive.offset_save);
1056	>	RRETURN(rrc);
1057	>	}
1058
1059		md->recursive = &new_recursive;
1060		memcpy(md->offset_vector, new_recursive.offset_save,
#	Line 856 \| Line 1079 \| for (;;)
1079		the end of a normal bracket, leaving the subject pointer. */
1080
1081		case OP_ONCE:
1082	<	{
1083	<	prev = ecode;
861	<	saved_eptr = eptr;
1082	>	prev = ecode;
1083	>	saved_eptr = eptr;
1084
1085	<	do
1086	<	{
1087	<	RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,
1088	<	eptrb, match_isgroup);
1089	<	if (rrc == MATCH_MATCH) break;
1090	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1091	<	ecode += GET(ecode,1);
1092	<	}
871	<	while (*ecode == OP_ALT);
1085	>	do
1086	>	{
1087	>	RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);
1088	>	if (rrc == MATCH_MATCH) break;
1089	>	if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1090	>	ecode += GET(ecode,1);
1091	>	}
1092	>	while (*ecode == OP_ALT);
1093
1094	<	/* If hit the end of the group (which could be repeated), fail */
1094	>	/* If hit the end of the group (which could be repeated), fail */
1095
1096	<	if (ecode != OP_ONCE && ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
1096	>	if (ecode != OP_ONCE && ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
1097
1098	<	/* Continue as from after the assertion, updating the offsets high water
1099	<	mark, since extracts may have been taken. */
1098	>	/* Continue as from after the assertion, updating the offsets high water
1099	>	mark, since extracts may have been taken. */
1100
1101	<	do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1101	>	do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
1102
1103	<	offset_top = md->end_offset_top;
1104	<	eptr = md->end_match_ptr;
1103	>	offset_top = md->end_offset_top;
1104	>	eptr = md->end_match_ptr;
1105
1106	<	/* For a non-repeating ket, just continue at this level. This also
1107	<	happens for a repeating ket if no characters were matched in the group.
1108	<	This is the forcible breaking of infinite loops as implemented in Perl
1109	<	5.005. If there is an options reset, it will get obeyed in the normal
1110	<	course of events. */
1106	>	/* For a non-repeating ket, just continue at this level. This also
1107	>	happens for a repeating ket if no characters were matched in the group.
1108	>	This is the forcible breaking of infinite loops as implemented in Perl
1109	>	5.005. If there is an options reset, it will get obeyed in the normal
1110	>	course of events. */
1111
1112	<	if (*ecode == OP_KET \|\| eptr == saved_eptr)
1113	<	{
1114	<	ecode += 1+LINK_SIZE;
1115	<	break;
1116	<	}
1112	>	if (*ecode == OP_KET \|\| eptr == saved_eptr)
1113	>	{
1114	>	ecode += 1+LINK_SIZE;
1115	>	break;
1116	>	}
1117
1118	<	/* The repeating kets try the rest of the pattern or restart from the
1119	<	preceding bracket, in the appropriate order. We need to reset any options
1120	<	that changed within the bracket before re-running it, so check the next
1121	<	opcode. */
1118	>	/* The repeating kets try the rest of the pattern or restart from the
1119	>	preceding bracket, in the appropriate order. The second "call" of match()
1120	>	uses tail recursion, to avoid using another stack frame. We need to reset
1121	>	any options that changed within the bracket before re-running it, so
1122	>	check the next opcode. */
1123
1124	<	if (ecode[1+LINK_SIZE] == OP_OPT)
1125	<	{
1126	<	ims = (ims & ~PCRE_IMS) \| ecode[4];
1127	<	DPRINTF(("ims set to %02lx at group repeat\n", ims));
1128	<	}
1124	>	if (ecode[1+LINK_SIZE] == OP_OPT)
1125	>	{
1126	>	ims = (ims & ~PCRE_IMS) \| ecode[4];
1127	>	DPRINTF(("ims set to %02lx at group repeat\n", ims));
1128	>	}
1129
1130	<	if (*ecode == OP_KETRMIN)
1131	<	{
1132	<	RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);
1133	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1134	<	RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);
1135	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1136	<	}
915	<	else /* OP_KETRMAX */
916	<	{
917	<	RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);
918	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
919	<	RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);
920	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
921	<	}
1130	>	if (*ecode == OP_KETRMIN)
1131	>	{
1132	>	RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);
1133	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1134	>	ecode = prev;
1135	>	flags = 0;
1136	>	goto TAIL_RECURSE;
1137		}
1138	<	RRETURN(MATCH_NOMATCH);
1138	>	else /* OP_KETRMAX */
1139	>	{
1140	>	RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);
1141	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1142	>	ecode += 1 + LINK_SIZE;
1143	>	flags = 0;
1144	>	goto TAIL_RECURSE;
1145	>	}
1146	>	/* Control never gets here */
1147
1148		/* An alternation is the end of a branch; scan along to find the end of the
1149		bracketed group and go to there. */
#	Line 929 \| Line 1152 \| for (;;)
1152		do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1153		break;
1154
1155	<	/* BRAZERO and BRAMINZERO occur just before a bracket group, indicating
1156	<	that it may occur zero times. It may repeat infinitely, or not at all -
1157	<	i.e. it could be ()* or ()? in the pattern. Brackets with fixed upper
1158	<	repeat limits are compiled as a number of copies, with the optional ones
1159	<	preceded by BRAZERO or BRAMINZERO. */
1155	>	/* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1156	>	indicating that it may occur zero times. It may repeat infinitely, or not
1157	>	at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1158	>	with fixed upper repeat limits are compiled as a number of copies, with the
1159	>	optional ones preceded by BRAZERO or BRAMINZERO. */
1160
1161		case OP_BRAZERO:
1162		{
1163		next = ecode+1;
1164	<	RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup);
1164	>	RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);
1165		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1166		do next += GET(next,1); while (*next == OP_ALT);
1167	<	ecode = next + 1+LINK_SIZE;
1167	>	ecode = next + 1 + LINK_SIZE;
1168		}
1169		break;
1170
1171		case OP_BRAMINZERO:
1172		{
1173		next = ecode+1;
1174	<	do next += GET(next,1); while (*next == OP_ALT);
1175	<	RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,
953	<	match_isgroup);
1174	>	do next += GET(next, 1); while (*next == OP_ALT);
1175	>	RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);
1176		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1177		ecode++;
1178		}
1179		break;
1180
1181	<	/* End of a group, repeated or non-repeating. If we are at the end of
1182	<	an assertion "group", stop matching and return MATCH_MATCH, but record the
1183	<	current high water mark for use by positive assertions. Do this also
1184	<	for the "once" (not-backup up) groups. */
1181	>	case OP_SKIPZERO:
1182	>	{
1183	>	next = ecode+1;
1184	>	do next += GET(next,1); while (*next == OP_ALT);
1185	>	ecode = next + 1 + LINK_SIZE;
1186	>	}
1187	>	break;
1188	>
1189	>	/* End of a group, repeated or non-repeating. */
1190
1191		case OP_KET:
1192		case OP_KETRMIN:
1193		case OP_KETRMAX:
1194	+	prev = ecode - GET(ecode, 1);
1195	+
1196	+	/* If this was a group that remembered the subject start, in order to break
1197	+	infinite repeats of empty string matches, retrieve the subject start from
1198	+	the chain. Otherwise, set it NULL. */
1199	+
1200	+	if (*prev >= OP_SBRA)
1201		{
1202	<	prev = ecode - GET(ecode, 1);
1203	<	saved_eptr = eptrb->epb_saved_eptr;
1202	>	saved_eptr = eptrb->epb_saved_eptr; /* Value at start of group */
1203	>	eptrb = eptrb->epb_prev; /* Backup to previous group */
1204	>	}
1205	>	else saved_eptr = NULL;
1206
1207	<	/* Back up the stack of bracket start pointers. */
1207	>	/* If we are at the end of an assertion group, stop matching and return
1208	>	MATCH_MATCH, but record the current high water mark for use by positive
1209	>	assertions. Do this also for the "once" (atomic) groups. */
1210	>
1211	>	if (prev == OP_ASSERT \|\| prev == OP_ASSERT_NOT \|\|
1212	>	prev == OP_ASSERTBACK \|\| prev == OP_ASSERTBACK_NOT \|\|
1213	>	*prev == OP_ONCE)
1214	>	{
1215	>	md->end_match_ptr = eptr; /* For ONCE */
1216	>	md->end_offset_top = offset_top;
1217	>	RRETURN(MATCH_MATCH);
1218	>	}
1219
1220	<	eptrb = eptrb->epb_prev;
1220	>	/* For capturing groups we have to check the group number back at the start
1221	>	and if necessary complete handling an extraction by setting the offsets and
1222	>	bumping the high water mark. Note that whole-pattern recursion is coded as
1223	>	a recurse into group 0, so it won't be picked up here. Instead, we catch it
1224	>	when the OP_END is reached. Other recursion is handled here. */
1225
1226	<	if (prev == OP_ASSERT \|\| prev == OP_ASSERT_NOT \|\|
1227	<	prev == OP_ASSERTBACK \|\| prev == OP_ASSERTBACK_NOT \|\|
1228	<	*prev == OP_ONCE)
1226	>	if (prev == OP_CBRA \|\| prev == OP_SCBRA)
1227	>	{
1228	>	number = GET2(prev, 1+LINK_SIZE);
1229	>	offset = number << 1;
1230	>
1231	>	md->capture_last = number;
1232	>	if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1233		{
1234	<	md->end_match_ptr = eptr; /* For ONCE */
1235	<	md->end_offset_top = offset_top;
1236	<	RRETURN(MATCH_MATCH);
1234	>	md->offset_vector[offset] =
1235	>	md->offset_vector[md->offset_end - number];
1236	>	md->offset_vector[offset+1] = eptr - md->start_subject;
1237	>	if (offset_top <= offset) offset_top = offset + 2;
1238		}
1239
1240	<	/* In all other cases except a conditional group we have to check the
1241	<	group number back at the start and if necessary complete handling an
986	<	extraction by setting the offsets and bumping the high water mark. */
1240	>	/* Handle a recursively called group. Restore the offsets
1241	>	appropriately and continue from after the call. */
1242
1243	<	if (*prev != OP_COND)
1243	>	if (md->recursive != NULL && md->recursive->group_num == number)
1244		{
1245	<	number = *prev - OP_BRA;
1246	<
1247	<	/* For extended extraction brackets (large number), we have to fish out
1248	<	the number from a dummy opcode at the start. */
1249	<
1250	<	if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);
1251	<	offset = number << 1;
1252	<
1253	<	#ifdef DEBUG
999	<	printf("end bracket %d", number);
1000	<	printf("\n");
1001	<	#endif
1002	<
1003	<	/* Test for a numbered group. This includes groups called as a result
1004	<	of recursion. Note that whole-pattern recursion is coded as a recurse
1005	<	into group 0, so it won't be picked up here. Instead, we catch it when
1006	<	the OP_END is reached. */
1007	<
1008	<	if (number > 0)
1009	<	{
1010	<	md->capture_last = number;
1011	<	if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1012	<	{
1013	<	md->offset_vector[offset] =
1014	<	md->offset_vector[md->offset_end - number];
1015	<	md->offset_vector[offset+1] = eptr - md->start_subject;
1016	<	if (offset_top <= offset) offset_top = offset + 2;
1017	<	}
1018	<
1019	<	/* Handle a recursively called group. Restore the offsets
1020	<	appropriately and continue from after the call. */
1021	<
1022	<	if (md->recursive != NULL && md->recursive->group_num == number)
1023	<	{
1024	<	recursion_info *rec = md->recursive;
1025	<	DPRINTF(("Recursion (%d) succeeded - continuing\n", number));
1026	<	md->recursive = rec->prevrec;
1027	<	md->start_match = rec->save_start;
1028	<	memcpy(md->offset_vector, rec->offset_save,
1029	<	rec->saved_max * sizeof(int));
1030	<	ecode = rec->after_call;
1031	<	ims = original_ims;
1032	<	break;
1033	<	}
1034	<	}
1245	>	recursion_info *rec = md->recursive;
1246	>	DPRINTF(("Recursion (%d) succeeded - continuing\n", number));
1247	>	md->recursive = rec->prevrec;
1248	>	mstart = rec->save_start;
1249	>	memcpy(md->offset_vector, rec->offset_save,
1250	>	rec->saved_max * sizeof(int));
1251	>	ecode = rec->after_call;
1252	>	ims = original_ims;
1253	>	break;
1254		}
1255	+	}
1256
1257	<	/* Reset the value of the ims flags, in case they got changed during
1258	<	the group. */
1257	>	/* For both capturing and non-capturing groups, reset the value of the ims
1258	>	flags, in case they got changed during the group. */
1259
1260	<	ims = original_ims;
1261	<	DPRINTF(("ims reset to %02lx\n", ims));
1260	>	ims = original_ims;
1261	>	DPRINTF(("ims reset to %02lx\n", ims));
1262
1263	<	/* For a non-repeating ket, just continue at this level. This also
1264	<	happens for a repeating ket if no characters were matched in the group.
1265	<	This is the forcible breaking of infinite loops as implemented in Perl
1266	<	5.005. If there is an options reset, it will get obeyed in the normal
1267	<	course of events. */
1263	>	/* For a non-repeating ket, just continue at this level. This also
1264	>	happens for a repeating ket if no characters were matched in the group.
1265	>	This is the forcible breaking of infinite loops as implemented in Perl
1266	>	5.005. If there is an options reset, it will get obeyed in the normal
1267	>	course of events. */
1268
1269	<	if (*ecode == OP_KET \|\| eptr == saved_eptr)
1270	<	{
1271	<	ecode += 1 + LINK_SIZE;
1272	<	break;
1273	<	}
1269	>	if (*ecode == OP_KET \|\| eptr == saved_eptr)
1270	>	{
1271	>	ecode += 1 + LINK_SIZE;
1272	>	break;
1273	>	}
1274
1275	<	/* The repeating kets try the rest of the pattern or restart from the
1276	<	preceding bracket, in the appropriate order. */
1275	>	/* The repeating kets try the rest of the pattern or restart from the
1276	>	preceding bracket, in the appropriate order. In the second case, we can use
1277	>	tail recursion to avoid using another stack frame, unless we have an
1278	>	unlimited repeat of a group that can match an empty string. */
1279
1280	<	if (*ecode == OP_KETRMIN)
1281	<	{
1282	<	RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);
1283	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1284	<	RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);
1285	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1286	<	}
1065	<	else /* OP_KETRMAX */
1280	>	flags = (*prev >= OP_SBRA)? match_cbegroup : 0;
1281	>
1282	>	if (*ecode == OP_KETRMIN)
1283	>	{
1284	>	RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);
1285	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1286	>	if (flags != 0) /* Could match an empty string */
1287		{
1288	<	RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);
1289	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1069	<	RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);
1070	<	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1288	>	RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);
1289	>	RRETURN(rrc);
1290		}
1291	+	ecode = prev;
1292	+	goto TAIL_RECURSE;
1293		}
1294	<
1295	<	RRETURN(MATCH_NOMATCH);
1294	>	else /* OP_KETRMAX */
1295	>	{
1296	>	RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);
1297	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1298	>	ecode += 1 + LINK_SIZE;
1299	>	flags = 0;
1300	>	goto TAIL_RECURSE;
1301	>	}
1302	>	/* Control never gets here */
1303
1304		/* Start of subject unless notbol, or after internal newline if multiline */
1305
#	Line 1079 \| Line 1307 \| for (;;)
1307		if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
1308		if ((ims & PCRE_MULTILINE) != 0)
1309		{
1310	<	if (eptr != md->start_subject && eptr[-1] != NEWLINE)
1310	>	if (eptr != md->start_subject &&
1311	>	(eptr == md->end_subject \|\| !WAS_NEWLINE(eptr)))
1312		RRETURN(MATCH_NOMATCH);
1313		ecode++;
1314		break;
#	Line 1100 \| Line 1329 \| for (;;)
1329		ecode++;
1330		break;
1331
1332	+	/* Reset the start of match point */
1333	+
1334	+	case OP_SET_SOM:
1335	+	mstart = eptr;
1336	+	ecode++;
1337	+	break;
1338	+
1339		/* Assert before internal newline if multiline, or before a terminating
1340		newline unless endonly is set, else end of subject unless noteol is set. */
1341
#	Line 1107 \| Line 1343 \| for (;;)
1343		if ((ims & PCRE_MULTILINE) != 0)
1344		{
1345		if (eptr < md->end_subject)
1346	<	{ if (*eptr != NEWLINE) RRETURN(MATCH_NOMATCH); }
1346	>	{ if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }
1347		else
1348		{ if (md->noteol) RRETURN(MATCH_NOMATCH); }
1349		ecode++;
#	Line 1118 \| Line 1354 \| for (;;)
1354		if (md->noteol) RRETURN(MATCH_NOMATCH);
1355		if (!md->endonly)
1356		{
1357	<	if (eptr < md->end_subject - 1 \|\|
1358	<	(eptr == md->end_subject - 1 && *eptr != NEWLINE))
1357	>	if (eptr != md->end_subject &&
1358	>	(!IS_NEWLINE(eptr) \|\| eptr != md->end_subject - md->nllen))
1359		RRETURN(MATCH_NOMATCH);
1360		ecode++;
1361		break;
1362		}
1363		}
1364	<	/* ... else fall through */
1364	>	/* ... else fall through for endonly */
1365
1366		/* End of subject assertion (\z) */
1367
#	Line 1137 \| Line 1373 \| for (;;)
1373		/* End of subject or ending \n assertion (\Z) */
1374
1375		case OP_EODN:
1376	<	if (eptr < md->end_subject - 1 \|\|
1377	<	(eptr == md->end_subject - 1 && *eptr != NEWLINE)) RRETURN(MATCH_NOMATCH);
1376	>	if (eptr != md->end_subject &&
1377	>	(!IS_NEWLINE(eptr) \|\| eptr != md->end_subject - md->nllen))
1378	>	RRETURN(MATCH_NOMATCH);
1379		ecode++;
1380		break;
1381
#	Line 1148 \| Line 1385 \| for (;;)
1385		case OP_WORD_BOUNDARY:
1386		{
1387
1388	+	/* Find out if the previous and current characters are "word" characters.
1389	+	It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
1390	+	be "non-word" characters. */
1391	+
1392	+	#ifdef SUPPORT_UTF8
1393	+	if (utf8)
1394	+	{
1395	+	if (eptr == md->start_subject) prev_is_word = FALSE; else
1396	+	{
1397	+	USPTR lastptr = eptr - 1;
1398	+	while((*lastptr & 0xc0) == 0x80) lastptr--;
1399	+	GETCHAR(c, lastptr);
1400	+	prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
1401	+	}
1402	+	if (eptr >= md->end_subject) cur_is_word = FALSE; else
1403	+	{
1404	+	GETCHAR(c, eptr);
1405	+	cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
1406	+	}
1407	+	}
1408	+	else
1409	+	#endif
1410	+
1411		/* More streamlined when not in UTF-8 mode */
1412
1413		{
#	Line 1168 \| Line 1428 \| for (;;)
1428		/* Match a single character type; inline for speed */
1429
1430		case OP_ANY:
1431	<	if ((ims & PCRE_DOTALL) == 0 && eptr < md->end_subject && *eptr == NEWLINE)
1432	<	RRETURN(MATCH_NOMATCH);
1431	>	if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
1432	>	/* Fall through */
1433	>
1434	>	case OP_ALLANY:
1435		if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);
1436	+	if (utf8) while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
1437		ecode++;
1438		break;
1439
#	Line 1185 \| Line 1448 \| for (;;)
1448		case OP_NOT_DIGIT:
1449		if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1450		GETCHARINCTEST(c, eptr);
1451	<	if ((md->ctypes[c] & ctype_digit) != 0)
1451	>	if (
1452	>	#ifdef SUPPORT_UTF8
1453	>	c < 256 &&
1454	>	#endif
1455	>	(md->ctypes[c] & ctype_digit) != 0
1456	>	)
1457		RRETURN(MATCH_NOMATCH);
1458		ecode++;
1459		break;
#	Line 1193 \| Line 1461 \| for (;;)
1461		case OP_DIGIT:
1462		if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1463		GETCHARINCTEST(c, eptr);
1464	<	if ((md->ctypes[c] & ctype_digit) == 0)
1464	>	if (
1465	>	#ifdef SUPPORT_UTF8
1466	>	c >= 256 \|\|
1467	>	#endif
1468	>	(md->ctypes[c] & ctype_digit) == 0
1469	>	)
1470		RRETURN(MATCH_NOMATCH);
1471		ecode++;
1472		break;
#	Line 1201 \| Line 1474 \| for (;;)
1474		case OP_NOT_WHITESPACE:
1475		if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1476		GETCHARINCTEST(c, eptr);
1477	<	if ((md->ctypes[c] & ctype_space) != 0)
1477	>	if (
1478	>	#ifdef SUPPORT_UTF8
1479	>	c < 256 &&
1480	>	#endif
1481	>	(md->ctypes[c] & ctype_space) != 0
1482	>	)
1483		RRETURN(MATCH_NOMATCH);
1484		ecode++;
1485		break;
#	Line 1209 \| Line 1487 \| for (;;)
1487		case OP_WHITESPACE:
1488		if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1489		GETCHARINCTEST(c, eptr);
1490	<	if ((md->ctypes[c] & ctype_space) == 0)
1490	>	if (
1491	>	#ifdef SUPPORT_UTF8
1492	>	c >= 256 \|\|
1493	>	#endif
1494	>	(md->ctypes[c] & ctype_space) == 0
1495	>	)
1496		RRETURN(MATCH_NOMATCH);
1497		ecode++;
1498		break;
#	Line 1217 \| Line 1500 \| for (;;)
1500		case OP_NOT_WORDCHAR:
1501		if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1502		GETCHARINCTEST(c, eptr);
1503	<	if ((md->ctypes[c] & ctype_word) != 0)
1503	>	if (
1504	>	#ifdef SUPPORT_UTF8
1505	>	c < 256 &&
1506	>	#endif
1507	>	(md->ctypes[c] & ctype_word) != 0
1508	>	)
1509		RRETURN(MATCH_NOMATCH);
1510		ecode++;
1511		break;
#	Line 1225 \| Line 1513 \| for (;;)
1513		case OP_WORDCHAR:
1514		if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1515		GETCHARINCTEST(c, eptr);
1516	<	if ((md->ctypes[c] & ctype_word) == 0)
1516	>	if (
1517	>	#ifdef SUPPORT_UTF8
1518	>	c >= 256 \|\|
1519	>	#endif
1520	>	(md->ctypes[c] & ctype_word) == 0
1521	>	)
1522		RRETURN(MATCH_NOMATCH);
1523		ecode++;
1524		break;
1525
1526	+	case OP_ANYNL:
1527	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1528	+	GETCHARINCTEST(c, eptr);
1529	+	switch(c)
1530	+	{
1531	+	default: RRETURN(MATCH_NOMATCH);
1532	+	case 0x000d:
1533	+	if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
1534	+	break;
1535	+
1536	+	case 0x000a:
1537	+	break;
1538	+
1539	+	case 0x000b:
1540	+	case 0x000c:
1541	+	case 0x0085:
1542	+	case 0x2028:
1543	+	case 0x2029:
1544	+	if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
1545	+	break;
1546	+	}
1547	+	ecode++;
1548	+	break;
1549	+
1550	+	case OP_NOT_HSPACE:
1551	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1552	+	GETCHARINCTEST(c, eptr);
1553	+	switch(c)
1554	+	{
1555	+	default: break;
1556	+	case 0x09: /* HT */
1557	+	case 0x20: /* SPACE */
1558	+	case 0xa0: /* NBSP */
1559	+	case 0x1680: /* OGHAM SPACE MARK */
1560	+	case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
1561	+	case 0x2000: /* EN QUAD */
1562	+	case 0x2001: /* EM QUAD */
1563	+	case 0x2002: /* EN SPACE */
1564	+	case 0x2003: /* EM SPACE */
1565	+	case 0x2004: /* THREE-PER-EM SPACE */
1566	+	case 0x2005: /* FOUR-PER-EM SPACE */
1567	+	case 0x2006: /* SIX-PER-EM SPACE */
1568	+	case 0x2007: /* FIGURE SPACE */
1569	+	case 0x2008: /* PUNCTUATION SPACE */
1570	+	case 0x2009: /* THIN SPACE */
1571	+	case 0x200A: /* HAIR SPACE */
1572	+	case 0x202f: /* NARROW NO-BREAK SPACE */
1573	+	case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
1574	+	case 0x3000: /* IDEOGRAPHIC SPACE */
1575	+	RRETURN(MATCH_NOMATCH);
1576	+	}
1577	+	ecode++;
1578	+	break;
1579	+
1580	+	case OP_HSPACE:
1581	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1582	+	GETCHARINCTEST(c, eptr);
1583	+	switch(c)
1584	+	{
1585	+	default: RRETURN(MATCH_NOMATCH);
1586	+	case 0x09: /* HT */
1587	+	case 0x20: /* SPACE */
1588	+	case 0xa0: /* NBSP */
1589	+	case 0x1680: /* OGHAM SPACE MARK */
1590	+	case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
1591	+	case 0x2000: /* EN QUAD */
1592	+	case 0x2001: /* EM QUAD */
1593	+	case 0x2002: /* EN SPACE */
1594	+	case 0x2003: /* EM SPACE */
1595	+	case 0x2004: /* THREE-PER-EM SPACE */
1596	+	case 0x2005: /* FOUR-PER-EM SPACE */
1597	+	case 0x2006: /* SIX-PER-EM SPACE */
1598	+	case 0x2007: /* FIGURE SPACE */
1599	+	case 0x2008: /* PUNCTUATION SPACE */
1600	+	case 0x2009: /* THIN SPACE */
1601	+	case 0x200A: /* HAIR SPACE */
1602	+	case 0x202f: /* NARROW NO-BREAK SPACE */
1603	+	case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
1604	+	case 0x3000: /* IDEOGRAPHIC SPACE */
1605	+	break;
1606	+	}
1607	+	ecode++;
1608	+	break;
1609	+
1610	+	case OP_NOT_VSPACE:
1611	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1612	+	GETCHARINCTEST(c, eptr);
1613	+	switch(c)
1614	+	{
1615	+	default: break;
1616	+	case 0x0a: /* LF */
1617	+	case 0x0b: /* VT */
1618	+	case 0x0c: /* FF */
1619	+	case 0x0d: /* CR */
1620	+	case 0x85: /* NEL */
1621	+	case 0x2028: /* LINE SEPARATOR */
1622	+	case 0x2029: /* PARAGRAPH SEPARATOR */
1623	+	RRETURN(MATCH_NOMATCH);
1624	+	}
1625	+	ecode++;
1626	+	break;
1627	+
1628	+	case OP_VSPACE:
1629	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1630	+	GETCHARINCTEST(c, eptr);
1631	+	switch(c)
1632	+	{
1633	+	default: RRETURN(MATCH_NOMATCH);
1634	+	case 0x0a: /* LF */
1635	+	case 0x0b: /* VT */
1636	+	case 0x0c: /* FF */
1637	+	case 0x0d: /* CR */
1638	+	case 0x85: /* NEL */
1639	+	case 0x2028: /* LINE SEPARATOR */
1640	+	case 0x2029: /* PARAGRAPH SEPARATOR */
1641	+	break;
1642	+	}
1643	+	ecode++;
1644	+	break;
1645	+
1646	+	#ifdef SUPPORT_UCP
1647	+	/* Check the next character by Unicode property. We will get here only
1648	+	if the support is in the binary; otherwise a compile-time error occurs. */
1649	+
1650	+	case OP_PROP:
1651	+	case OP_NOTPROP:
1652	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1653	+	GETCHARINCTEST(c, eptr);
1654	+	{
1655	+	const ucd_record *prop = GET_UCD(c);
1656	+
1657	+	switch(ecode[1])
1658	+	{
1659	+	case PT_ANY:
1660	+	if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
1661	+	break;
1662	+
1663	+	case PT_LAMP:
1664	+	if ((prop->chartype == ucp_Lu \|\|
1665	+	prop->chartype == ucp_Ll \|\|
1666	+	prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
1667	+	RRETURN(MATCH_NOMATCH);
1668	+	break;
1669	+
1670	+	case PT_GC:
1671	+	if ((ecode[2] != _pcre_ucp_gentype[prop->chartype]) == (op == OP_PROP))
1672	+	RRETURN(MATCH_NOMATCH);
1673	+	break;
1674	+
1675	+	case PT_PC:
1676	+	if ((ecode[2] != prop->chartype) == (op == OP_PROP))
1677	+	RRETURN(MATCH_NOMATCH);
1678	+	break;
1679	+
1680	+	case PT_SC:
1681	+	if ((ecode[2] != prop->script) == (op == OP_PROP))
1682	+	RRETURN(MATCH_NOMATCH);
1683	+	break;
1684	+
1685	+	default:
1686	+	RRETURN(PCRE_ERROR_INTERNAL);
1687	+	}
1688	+
1689	+	ecode += 3;
1690	+	}
1691	+	break;
1692	+
1693	+	/* Match an extended Unicode sequence. We will get here only if the support
1694	+	is in the binary; otherwise a compile-time error occurs. */
1695	+
1696	+	case OP_EXTUNI:
1697	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1698	+	GETCHARINCTEST(c, eptr);
1699	+	{
1700	+	int category = UCD_CATEGORY(c);
1701	+	if (category == ucp_M) RRETURN(MATCH_NOMATCH);
1702	+	while (eptr < md->end_subject)
1703	+	{
1704	+	int len = 1;
1705	+	if (!utf8) c = *eptr; else
1706	+	{
1707	+	GETCHARLEN(c, eptr, len);
1708	+	}
1709	+	category = UCD_CATEGORY(c);
1710	+	if (category != ucp_M) break;
1711	+	eptr += len;
1712	+	}
1713	+	}
1714	+	ecode++;
1715	+	break;
1716	+	#endif
1717	+
1718
1719		/* Match a back reference, possibly repeatedly. Look past the end of the
1720		item to see if there is repeat information following. The code is similar
#	Line 1242 \| Line 1727 \| for (;;)
1727		case OP_REF:
1728		{
1729		offset = GET2(ecode, 1) << 1; /* Doubled ref number */
1730	<	ecode += 3; /* Advance past item */
1730	>	ecode += 3;
1731
1732	<	/* If the reference is unset, set the length to be longer than the amount
1733	<	of subject left; this ensures that every attempt at a match fails. We
1734	<	can't just fail here, because of the possibility of quantifiers with zero
1735	<	minima. */
1736	<
1737	<	length = (offset >= offset_top \|\| md->offset_vector[offset] < 0)?
1738	<	md->end_subject - eptr + 1 :
1739	<	md->offset_vector[offset+1] - md->offset_vector[offset];
1732	>	/* If the reference is unset, there are two possibilities:
1733	>
1734	>	(a) In the default, Perl-compatible state, set the length to be longer
1735	>	than the amount of subject left; this ensures that every attempt at a
1736	>	match fails. We can't just fail here, because of the possibility of
1737	>	quantifiers with zero minima.
1738	>
1739	>	(b) If the JavaScript compatibility flag is set, set the length to zero
1740	>	so that the back reference matches an empty string.
1741	>
1742	>	Otherwise, set the length to the length of what was matched by the
1743	>	referenced subpattern. */
1744	>
1745	>	if (offset >= offset_top \|\| md->offset_vector[offset] < 0)
1746	>	length = (md->jscript_compat)? 0 : md->end_subject - eptr + 1;
1747	>	else
1748	>	length = md->offset_vector[offset+1] - md->offset_vector[offset];
1749
1750		/* Set up for repetition, or handle the non-repeated case */
1751
#	Line 1311 \| Line 1805 \| for (;;)
1805		{
1806		for (fi = min;; fi++)
1807		{
1808	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
1808	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);
1809		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1810		if (fi >= max \|\| !match_ref(offset, eptr, length, md, ims))
1811		RRETURN(MATCH_NOMATCH);
#	Line 1332 \| Line 1826 \| for (;;)
1826		}
1827		while (eptr >= pp)
1828		{
1829	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
1829	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);
1830		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1831		eptr -= length;
1832		}
#	Line 1391 \| Line 1885 \| for (;;)
1885
1886		/* First, ensure the minimum number of matches are present. */
1887
1888	+	#ifdef SUPPORT_UTF8
1889	+	/* UTF-8 mode */
1890	+	if (utf8)
1891	+	{
1892	+	for (i = 1; i <= min; i++)
1893	+	{
1894	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1895	+	GETCHARINC(c, eptr);
1896	+	if (c > 255)
1897	+	{
1898	+	if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
1899	+	}
1900	+	else
1901	+	{
1902	+	if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
1903	+	}
1904	+	}
1905	+	}
1906	+	else
1907	+	#endif
1908		/* Not UTF-8 mode */
1909		{
1910		for (i = 1; i <= min; i++)
#	Line 1411 \| Line 1925 \| for (;;)
1925
1926		if (minimize)
1927		{
1928	+	#ifdef SUPPORT_UTF8
1929	+	/* UTF-8 mode */
1930	+	if (utf8)
1931	+	{
1932	+	for (fi = min;; fi++)
1933	+	{
1934	+	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);
1935	+	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1936	+	if (fi >= max \|\| eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1937	+	GETCHARINC(c, eptr);
1938	+	if (c > 255)
1939	+	{
1940	+	if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
1941	+	}
1942	+	else
1943	+	{
1944	+	if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
1945	+	}
1946	+	}
1947	+	}
1948	+	else
1949	+	#endif
1950		/* Not UTF-8 mode */
1951		{
1952		for (fi = min;; fi++)
1953		{
1954	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
1954	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);
1955		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1956		if (fi >= max \|\| eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
1957		c = *eptr++;
#	Line 1431 \| Line 1967 \| for (;;)
1967		{
1968		pp = eptr;
1969
1970	+	#ifdef SUPPORT_UTF8
1971	+	/* UTF-8 mode */
1972	+	if (utf8)
1973	+	{
1974	+	for (i = min; i < max; i++)
1975	+	{
1976	+	int len = 1;
1977	+	if (eptr >= md->end_subject) break;
1978	+	GETCHARLEN(c, eptr, len);
1979	+	if (c > 255)
1980	+	{
1981	+	if (op == OP_CLASS) break;
1982	+	}
1983	+	else
1984	+	{
1985	+	if ((data[c/8] & (1 << (c&7))) == 0) break;
1986	+	}
1987	+	eptr += len;
1988	+	}
1989	+	for (;;)
1990	+	{
1991	+	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);
1992	+	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1993	+	if (eptr-- == pp) break; /* Stop if tried at original pos */
1994	+	BACKCHAR(eptr);
1995	+	}
1996	+	}
1997	+	else
1998	+	#endif
1999		/* Not UTF-8 mode */
2000		{
2001		for (i = min; i < max; i++)
#	Line 1442 \| Line 2007 \| for (;;)
2007		}
2008		while (eptr >= pp)
2009		{
2010	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
1446	<	eptr--;
2010	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);
2011		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2012	+	eptr--;
2013		}
2014		}
2015
#	Line 1454 \| Line 2019 \| for (;;)
2019		/* Control never gets here */
2020
2021
2022	+	/* Match an extended character class. This opcode is encountered only
2023	+	when UTF-8 mode mode is supported. Nevertheless, we may not be in UTF-8
2024	+	mode, because Unicode properties are supported in non-UTF-8 mode. */
2025	+
2026	+	#ifdef SUPPORT_UTF8
2027	+	case OP_XCLASS:
2028	+	{
2029	+	data = ecode + 1 + LINK_SIZE; /* Save for matching */
2030	+	ecode += GET(ecode, 1); /* Advance past the item */
2031	+
2032	+	switch (*ecode)
2033	+	{
2034	+	case OP_CRSTAR:
2035	+	case OP_CRMINSTAR:
2036	+	case OP_CRPLUS:
2037	+	case OP_CRMINPLUS:
2038	+	case OP_CRQUERY:
2039	+	case OP_CRMINQUERY:
2040	+	c = *ecode++ - OP_CRSTAR;
2041	+	minimize = (c & 1) != 0;
2042	+	min = rep_min[c]; /* Pick up values from tables; */
2043	+	max = rep_max[c]; /* zero for max => infinity */
2044	+	if (max == 0) max = INT_MAX;
2045	+	break;
2046	+
2047	+	case OP_CRRANGE:
2048	+	case OP_CRMINRANGE:
2049	+	minimize = (*ecode == OP_CRMINRANGE);
2050	+	min = GET2(ecode, 1);
2051	+	max = GET2(ecode, 3);
2052	+	if (max == 0) max = INT_MAX;
2053	+	ecode += 5;
2054	+	break;
2055	+
2056	+	default: /* No repeat follows */
2057	+	min = max = 1;
2058	+	break;
2059	+	}
2060	+
2061	+	/* First, ensure the minimum number of matches are present. */
2062	+
2063	+	for (i = 1; i <= min; i++)
2064	+	{
2065	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
2066	+	GETCHARINCTEST(c, eptr);
2067	+	if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);
2068	+	}
2069	+
2070	+	/* If max == min we can continue with the main loop without the
2071	+	need to recurse. */
2072	+
2073	+	if (min == max) continue;
2074	+
2075	+	/* If minimizing, keep testing the rest of the expression and advancing
2076	+	the pointer while it matches the class. */
2077	+
2078	+	if (minimize)
2079	+	{
2080	+	for (fi = min;; fi++)
2081	+	{
2082	+	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);
2083	+	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2084	+	if (fi >= max \|\| eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
2085	+	GETCHARINCTEST(c, eptr);
2086	+	if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);
2087	+	}
2088	+	/* Control never gets here */
2089	+	}
2090	+
2091	+	/* If maximizing, find the longest possible run, then work backwards. */
2092	+
2093	+	else
2094	+	{
2095	+	pp = eptr;
2096	+	for (i = min; i < max; i++)
2097	+	{
2098	+	int len = 1;
2099	+	if (eptr >= md->end_subject) break;
2100	+	GETCHARLENTEST(c, eptr, len);
2101	+	if (!_pcre_xclass(c, data)) break;
2102	+	eptr += len;
2103	+	}
2104	+	for(;;)
2105	+	{
2106	+	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);
2107	+	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2108	+	if (eptr-- == pp) break; /* Stop if tried at original pos */
2109	+	if (utf8) BACKCHAR(eptr);
2110	+	}
2111	+	RRETURN(MATCH_NOMATCH);
2112	+	}
2113	+
2114	+	/* Control never gets here */
2115	+	}
2116	+	#endif /* End of XCLASS */
2117	+
2118		/* Match a single character, casefully */
2119
2120		case OP_CHAR:
2121	+	#ifdef SUPPORT_UTF8
2122	+	if (utf8)
2123	+	{
2124	+	length = 1;
2125	+	ecode++;
2126	+	GETCHARLEN(fc, ecode, length);
2127	+	if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);
2128	+	while (length-- > 0) if (ecode++ != eptr++) RRETURN(MATCH_NOMATCH);
2129	+	}
2130	+	else
2131	+	#endif
2132	+
2133		/* Non-UTF-8 mode */
2134		{
2135		if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);
#	Line 1468 \| Line 2141 \| for (;;)
2141		/* Match a single character, caselessly */
2142
2143		case OP_CHARNC:
2144	+	#ifdef SUPPORT_UTF8
2145	+	if (utf8)
2146	+	{
2147	+	length = 1;
2148	+	ecode++;
2149	+	GETCHARLEN(fc, ecode, length);
2150	+
2151	+	if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);
2152	+
2153	+	/* If the pattern character's value is < 128, we have only one byte, and
2154	+	can use the fast lookup table. */
2155	+
2156	+	if (fc < 128)
2157	+	{
2158	+	if (md->lcc[ecode++] != md->lcc[eptr++]) RRETURN(MATCH_NOMATCH);
2159	+	}
2160	+
2161	+	/* Otherwise we must pick up the subject character */
2162	+
2163	+	else
2164	+	{
2165	+	unsigned int dc;
2166	+	GETCHARINC(dc, eptr);
2167	+	ecode += length;
2168	+
2169	+	/* If we have Unicode property support, we can use it to test the other
2170	+	case of the character, if there is one. */
2171	+
2172	+	if (fc != dc)
2173	+	{
2174	+	#ifdef SUPPORT_UCP
2175	+	if (dc != UCD_OTHERCASE(fc))
2176	+	#endif
2177	+	RRETURN(MATCH_NOMATCH);
2178	+	}
2179	+	}
2180	+	}
2181	+	else
2182	+	#endif /* SUPPORT_UTF8 */
2183	+
2184		/* Non-UTF-8 mode */
2185		{
2186		if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);
#	Line 1476 \| Line 2189 \| for (;;)
2189		}
2190		break;
2191
2192	<	/* Match a single character repeatedly; different opcodes share code. */
2192	>	/* Match a single character repeatedly. */
2193
2194		case OP_EXACT:
2195		min = max = GET2(ecode, 1);
2196		ecode += 3;
2197		goto REPEATCHAR;
2198
2199	+	case OP_POSUPTO:
2200	+	possessive = TRUE;
2201	+	/* Fall through */
2202	+
2203		case OP_UPTO:
2204		case OP_MINUPTO:
2205		min = 0;
#	Line 1491 \| Line 2208 \| for (;;)
2208		ecode += 3;
2209		goto REPEATCHAR;
2210
2211	+	case OP_POSSTAR:
2212	+	possessive = TRUE;
2213	+	min = 0;
2214	+	max = INT_MAX;
2215	+	ecode++;
2216	+	goto REPEATCHAR;
2217	+
2218	+	case OP_POSPLUS:
2219	+	possessive = TRUE;
2220	+	min = 1;
2221	+	max = INT_MAX;
2222	+	ecode++;
2223	+	goto REPEATCHAR;
2224	+
2225	+	case OP_POSQUERY:
2226	+	possessive = TRUE;
2227	+	min = 0;
2228	+	max = 1;
2229	+	ecode++;
2230	+	goto REPEATCHAR;
2231	+
2232		case OP_STAR:
2233		case OP_MINSTAR:
2234		case OP_PLUS:
#	Line 1508 \| Line 2246 \| for (;;)
2246		the subject. */
2247
2248		REPEATCHAR:
2249	+	#ifdef SUPPORT_UTF8
2250	+	if (utf8)
2251	+	{
2252	+	length = 1;
2253	+	charptr = ecode;
2254	+	GETCHARLEN(fc, ecode, length);
2255	+	if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);
2256	+	ecode += length;
2257	+
2258	+	/* Handle multibyte character matching specially here. There is
2259	+	support for caseless matching if UCP support is present. */
2260	+
2261	+	if (length > 1)
2262	+	{
2263	+	#ifdef SUPPORT_UCP
2264	+	unsigned int othercase;
2265	+	if ((ims & PCRE_CASELESS) != 0 &&
2266	+	(othercase = UCD_OTHERCASE(fc)) != fc)
2267	+	oclength = _pcre_ord2utf8(othercase, occhars);
2268	+	else oclength = 0;
2269	+	#endif /* SUPPORT_UCP */
2270	+
2271	+	for (i = 1; i <= min; i++)
2272	+	{
2273	+	if (memcmp(eptr, charptr, length) == 0) eptr += length;
2274	+	#ifdef SUPPORT_UCP
2275	+	/* Need braces because of following else */
2276	+	else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }
2277	+	else
2278	+	{
2279	+	if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);
2280	+	eptr += oclength;
2281	+	}
2282	+	#else /* without SUPPORT_UCP */
2283	+	else { RRETURN(MATCH_NOMATCH); }
2284	+	#endif /* SUPPORT_UCP */
2285	+	}
2286	+
2287	+	if (min == max) continue;
2288	+
2289	+	if (minimize)
2290	+	{
2291	+	for (fi = min;; fi++)
2292	+	{
2293	+	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);
2294	+	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2295	+	if (fi >= max \|\| eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
2296	+	if (memcmp(eptr, charptr, length) == 0) eptr += length;
2297	+	#ifdef SUPPORT_UCP
2298	+	/* Need braces because of following else */
2299	+	else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }
2300	+	else
2301	+	{
2302	+	if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);
2303	+	eptr += oclength;
2304	+	}
2305	+	#else /* without SUPPORT_UCP */
2306	+	else { RRETURN (MATCH_NOMATCH); }
2307	+	#endif /* SUPPORT_UCP */
2308	+	}
2309	+	/* Control never gets here */
2310	+	}
2311	+
2312	+	else /* Maximize */
2313	+	{
2314	+	pp = eptr;
2315	+	for (i = min; i < max; i++)
2316	+	{
2317	+	if (eptr > md->end_subject - length) break;
2318	+	if (memcmp(eptr, charptr, length) == 0) eptr += length;
2319	+	#ifdef SUPPORT_UCP
2320	+	else if (oclength == 0) break;
2321	+	else
2322	+	{
2323	+	if (memcmp(eptr, occhars, oclength) != 0) break;
2324	+	eptr += oclength;
2325	+	}
2326	+	#else /* without SUPPORT_UCP */
2327	+	else break;
2328	+	#endif /* SUPPORT_UCP */
2329	+	}
2330	+
2331	+	if (possessive) continue;
2332	+	for(;;)
2333	+	{
2334	+	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);
2335	+	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2336	+	if (eptr == pp) RRETURN(MATCH_NOMATCH);
2337	+	#ifdef SUPPORT_UCP
2338	+	eptr--;
2339	+	BACKCHAR(eptr);
2340	+	#else /* without SUPPORT_UCP */
2341	+	eptr -= length;
2342	+	#endif /* SUPPORT_UCP */
2343	+	}
2344	+	}
2345	+	/* Control never gets here */
2346	+	}
2347	+
2348	+	/* If the length of a UTF-8 character is 1, we fall through here, and
2349	+	obey the code as for non-UTF-8 characters below, though in this case the
2350	+	value of fc will always be < 128. */
2351	+	}
2352	+	else
2353	+	#endif /* SUPPORT_UTF8 */
2354	+
2355		/* When not in UTF-8 mode, load a single-byte character. */
2356		{
2357		if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);
#	Line 1536 \| Line 2380 \| for (;;)
2380		{
2381		for (fi = min;; fi++)
2382		{
2383	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
2383	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);
2384		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2385		if (fi >= max \|\| eptr >= md->end_subject \|\|
2386		fc != md->lcc[*eptr++])
#	Line 1544 \| Line 2388 \| for (;;)
2388		}
2389		/* Control never gets here */
2390		}
2391	<	else
2391	>	else /* Maximize */
2392		{
2393		pp = eptr;
2394		for (i = min; i < max; i++)
#	Line 1552 \| Line 2396 \| for (;;)
2396		if (eptr >= md->end_subject \|\| fc != md->lcc[*eptr]) break;
2397		eptr++;
2398		}
2399	+	if (possessive) continue;
2400		while (eptr >= pp)
2401		{
2402	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
2402	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);
2403		eptr--;
2404		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2405		}
#	Line 1573 \| Line 2418 \| for (;;)
2418		{
2419		for (fi = min;; fi++)
2420		{
2421	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
2421	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);
2422		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2423		if (fi >= max \|\| eptr >= md->end_subject \|\| fc != *eptr++)
2424		RRETURN(MATCH_NOMATCH);
2425		}
2426		/* Control never gets here */
2427		}
2428	<	else
2428	>	else /* Maximize */
2429		{
2430		pp = eptr;
2431		for (i = min; i < max; i++)
#	Line 1588 \| Line 2433 \| for (;;)
2433		if (eptr >= md->end_subject \|\| fc != *eptr) break;
2434		eptr++;
2435		}
2436	+	if (possessive) continue;
2437		while (eptr >= pp)
2438		{
2439	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
2439	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);
2440		eptr--;
2441		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2442		}
#	Line 1608 \| Line 2454 \| for (;;)
2454		GETCHARINCTEST(c, eptr);
2455		if ((ims & PCRE_CASELESS) != 0)
2456		{
2457	+	#ifdef SUPPORT_UTF8
2458	+	if (c < 256)
2459	+	#endif
2460		c = md->lcc[c];
2461		if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);
2462		}
#	Line 1637 \| Line 2486 \| for (;;)
2486		ecode += 3;
2487		goto REPEATNOTCHAR;
2488
2489	+	case OP_NOTPOSSTAR:
2490	+	possessive = TRUE;
2491	+	min = 0;
2492	+	max = INT_MAX;
2493	+	ecode++;
2494	+	goto REPEATNOTCHAR;
2495	+
2496	+	case OP_NOTPOSPLUS:
2497	+	possessive = TRUE;
2498	+	min = 1;
2499	+	max = INT_MAX;
2500	+	ecode++;
2501	+	goto REPEATNOTCHAR;
2502	+
2503	+	case OP_NOTPOSQUERY:
2504	+	possessive = TRUE;
2505	+	min = 0;
2506	+	max = 1;
2507	+	ecode++;
2508	+	goto REPEATNOTCHAR;
2509	+
2510	+	case OP_NOTPOSUPTO:
2511	+	possessive = TRUE;
2512	+	min = 0;
2513	+	max = GET2(ecode, 1);
2514	+	ecode += 3;
2515	+	goto REPEATNOTCHAR;
2516	+
2517		case OP_NOTSTAR:
2518		case OP_NOTMINSTAR:
2519		case OP_NOTPLUS:
#	Line 1672 \| Line 2549 \| for (;;)
2549		{
2550		fc = md->lcc[fc];
2551
2552	+	#ifdef SUPPORT_UTF8
2553	+	/* UTF-8 mode */
2554	+	if (utf8)
2555	+	{
2556	+	register unsigned int d;
2557	+	for (i = 1; i <= min; i++)
2558	+	{
2559	+	GETCHARINC(d, eptr);
2560	+	if (d < 256) d = md->lcc[d];
2561	+	if (fc == d) RRETURN(MATCH_NOMATCH);
2562	+	}
2563	+	}
2564	+	else
2565	+	#endif
2566	+
2567		/* Not UTF-8 mode */
2568		{
2569		for (i = 1; i <= min; i++)
#	Line 1682 \| Line 2574 \| for (;;)
2574
2575		if (minimize)
2576		{
2577	+	#ifdef SUPPORT_UTF8
2578	+	/* UTF-8 mode */
2579	+	if (utf8)
2580	+	{
2581	+	register unsigned int d;
2582	+	for (fi = min;; fi++)
2583	+	{
2584	+	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM28);
2585	+	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2586	+	if (fi >= max \|\| eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
2587	+	GETCHARINC(d, eptr);
2588	+	if (d < 256) d = md->lcc[d];
2589	+	if (fc == d) RRETURN(MATCH_NOMATCH);
2590	+
2591	+	}
2592	+	}
2593	+	else
2594	+	#endif
2595		/* Not UTF-8 mode */
2596		{
2597		for (fi = min;; fi++)
2598		{
2599	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
2599	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM29);
2600		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2601		if (fi >= max \|\| eptr >= md->end_subject \|\| fc == md->lcc[*eptr++])
2602		RRETURN(MATCH_NOMATCH);
#	Line 1701 \| Line 2611 \| for (;;)
2611		{
2612		pp = eptr;
2613
2614	+	#ifdef SUPPORT_UTF8
2615	+	/* UTF-8 mode */
2616	+	if (utf8)
2617	+	{
2618	+	register unsigned int d;
2619	+	for (i = min; i < max; i++)
2620	+	{
2621	+	int len = 1;
2622	+	if (eptr >= md->end_subject) break;
2623	+	GETCHARLEN(d, eptr, len);
2624	+	if (d < 256) d = md->lcc[d];
2625	+	if (fc == d) break;
2626	+	eptr += len;
2627	+	}
2628	+	if (possessive) continue;
2629	+	for(;;)
2630	+	{
2631	+	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM30);
2632	+	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2633	+	if (eptr-- == pp) break; /* Stop if tried at original pos */
2634	+	BACKCHAR(eptr);
2635	+	}
2636	+	}
2637	+	else
2638	+	#endif
2639		/* Not UTF-8 mode */
2640		{
2641		for (i = min; i < max; i++)
#	Line 1708 \| Line 2643 \| for (;;)
2643		if (eptr >= md->end_subject \|\| fc == md->lcc[*eptr]) break;
2644		eptr++;
2645		}
2646	+	if (possessive) continue;
2647		while (eptr >= pp)
2648		{
2649	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
2649	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM31);
2650		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2651		eptr--;
2652		}
#	Line 1725 \| Line 2661 \| for (;;)
2661
2662		else
2663		{
2664	+	#ifdef SUPPORT_UTF8
2665	+	/* UTF-8 mode */
2666	+	if (utf8)
2667	+	{
2668	+	register unsigned int d;
2669	+	for (i = 1; i <= min; i++)
2670	+	{
2671	+	GETCHARINC(d, eptr);
2672	+	if (fc == d) RRETURN(MATCH_NOMATCH);
2673	+	}
2674	+	}
2675	+	else
2676	+	#endif
2677		/* Not UTF-8 mode */
2678		{
2679		for (i = 1; i <= min; i++)
#	Line 1735 \| Line 2684 \| for (;;)
2684
2685		if (minimize)
2686		{
2687	+	#ifdef SUPPORT_UTF8
2688	+	/* UTF-8 mode */
2689	+	if (utf8)
2690	+	{
2691	+	register unsigned int d;
2692	+	for (fi = min;; fi++)
2693	+	{
2694	+	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM32);
2695	+	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2696	+	if (fi >= max \|\| eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
2697	+	GETCHARINC(d, eptr);
2698	+	if (fc == d) RRETURN(MATCH_NOMATCH);
2699	+	}
2700	+	}
2701	+	else
2702	+	#endif
2703		/* Not UTF-8 mode */
2704		{
2705		for (fi = min;; fi++)
2706		{
2707	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
2707	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM33);
2708		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2709		if (fi >= max \|\| eptr >= md->end_subject \|\| fc == *eptr++)
2710		RRETURN(MATCH_NOMATCH);
#	Line 1754 \| Line 2719 \| for (;;)
2719		{
2720		pp = eptr;
2721
2722	+	#ifdef SUPPORT_UTF8
2723	+	/* UTF-8 mode */
2724	+	if (utf8)
2725	+	{
2726	+	register unsigned int d;
2727	+	for (i = min; i < max; i++)
2728	+	{
2729	+	int len = 1;
2730	+	if (eptr >= md->end_subject) break;
2731	+	GETCHARLEN(d, eptr, len);
2732	+	if (fc == d) break;
2733	+	eptr += len;
2734	+	}
2735	+	if (possessive) continue;
2736	+	for(;;)
2737	+	{
2738	+	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM34);
2739	+	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2740	+	if (eptr-- == pp) break; /* Stop if tried at original pos */
2741	+	BACKCHAR(eptr);
2742	+	}
2743	+	}
2744	+	else
2745	+	#endif
2746		/* Not UTF-8 mode */
2747		{
2748		for (i = min; i < max; i++)
#	Line 1761 \| Line 2750 \| for (;;)
2750		if (eptr >= md->end_subject \|\| fc == *eptr) break;
2751		eptr++;
2752		}
2753	+	if (possessive) continue;
2754		while (eptr >= pp)
2755		{
2756	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
2756	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM35);
2757		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2758		eptr--;
2759		}
#	Line 1792 \| Line 2782 \| for (;;)
2782		ecode += 3;
2783		goto REPEATTYPE;
2784
2785	+	case OP_TYPEPOSSTAR:
2786	+	possessive = TRUE;
2787	+	min = 0;
2788	+	max = INT_MAX;
2789	+	ecode++;
2790	+	goto REPEATTYPE;
2791	+
2792	+	case OP_TYPEPOSPLUS:
2793	+	possessive = TRUE;
2794	+	min = 1;
2795	+	max = INT_MAX;
2796	+	ecode++;
2797	+	goto REPEATTYPE;
2798	+
2799	+	case OP_TYPEPOSQUERY:
2800	+	possessive = TRUE;
2801	+	min = 0;
2802	+	max = 1;
2803	+	ecode++;
2804	+	goto REPEATTYPE;
2805	+
2806	+	case OP_TYPEPOSUPTO:
2807	+	possessive = TRUE;
2808	+	min = 0;
2809	+	max = GET2(ecode, 1);
2810	+	ecode += 3;
2811	+	goto REPEATTYPE;
2812	+
2813		case OP_TYPESTAR:
2814		case OP_TYPEMINSTAR:
2815		case OP_TYPEPLUS:
#	Line 1811 \| Line 2829 \| for (;;)
2829		REPEATTYPE:
2830		ctype = ecode++; / Code for the character type */
2831
2832	+	#ifdef SUPPORT_UCP
2833	+	if (ctype == OP_PROP \|\| ctype == OP_NOTPROP)
2834	+	{
2835	+	prop_fail_result = ctype == OP_NOTPROP;
2836	+	prop_type = *ecode++;
2837	+	prop_value = *ecode++;
2838	+	}
2839	+	else prop_type = -1;
2840	+	#endif
2841	+
2842		/* First, ensure the minimum number of matches are present. Use inline
2843		code for maximizing the speed, and do the type test once at the start
2844		(i.e. keep it out of the loop). Also we can test that there are at least
#	Line 1822 \| Line 2850 \| for (;;)
2850		if (min > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);
2851		if (min > 0)
2852		{
2853	+	#ifdef SUPPORT_UCP
2854	+	if (prop_type >= 0)
2855	+	{
2856	+	switch(prop_type)
2857	+	{
2858	+	case PT_ANY:
2859	+	if (prop_fail_result) RRETURN(MATCH_NOMATCH);
2860	+	for (i = 1; i <= min; i++)
2861	+	{
2862	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
2863	+	GETCHARINCTEST(c, eptr);
2864	+	}
2865	+	break;
2866	+
2867	+	case PT_LAMP:
2868	+	for (i = 1; i <= min; i++)
2869	+	{
2870	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
2871	+	GETCHARINCTEST(c, eptr);
2872	+	prop_chartype = UCD_CHARTYPE(c);
2873	+	if ((prop_chartype == ucp_Lu \|\|
2874	+	prop_chartype == ucp_Ll \|\|
2875	+	prop_chartype == ucp_Lt) == prop_fail_result)
2876	+	RRETURN(MATCH_NOMATCH);
2877	+	}
2878	+	break;
2879	+
2880	+	case PT_GC:
2881	+	for (i = 1; i <= min; i++)
2882	+	{
2883	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
2884	+	GETCHARINCTEST(c, eptr);
2885	+	prop_category = UCD_CATEGORY(c);
2886	+	if ((prop_category == prop_value) == prop_fail_result)
2887	+	RRETURN(MATCH_NOMATCH);
2888	+	}
2889	+	break;
2890	+
2891	+	case PT_PC:
2892	+	for (i = 1; i <= min; i++)
2893	+	{
2894	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
2895	+	GETCHARINCTEST(c, eptr);
2896	+	prop_chartype = UCD_CHARTYPE(c);
2897	+	if ((prop_chartype == prop_value) == prop_fail_result)
2898	+	RRETURN(MATCH_NOMATCH);
2899	+	}
2900	+	break;
2901	+
2902	+	case PT_SC:
2903	+	for (i = 1; i <= min; i++)
2904	+	{
2905	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
2906	+	GETCHARINCTEST(c, eptr);
2907	+	prop_script = UCD_SCRIPT(c);
2908	+	if ((prop_script == prop_value) == prop_fail_result)
2909	+	RRETURN(MATCH_NOMATCH);
2910	+	}
2911	+	break;
2912	+
2913	+	default:
2914	+	RRETURN(PCRE_ERROR_INTERNAL);
2915	+	}
2916	+	}
2917	+
2918	+	/* Match extended Unicode sequences. We will get here only if the
2919	+	support is in the binary; otherwise a compile-time error occurs. */
2920	+
2921	+	else if (ctype == OP_EXTUNI)
2922	+	{
2923	+	for (i = 1; i <= min; i++)
2924	+	{
2925	+	GETCHARINCTEST(c, eptr);
2926	+	prop_category = UCD_CATEGORY(c);
2927	+	if (prop_category == ucp_M) RRETURN(MATCH_NOMATCH);
2928	+	while (eptr < md->end_subject)
2929	+	{
2930	+	int len = 1;
2931	+	if (!utf8) c = *eptr; else
2932	+	{
2933	+	GETCHARLEN(c, eptr, len);
2934	+	}
2935	+	prop_category = UCD_CATEGORY(c);
2936	+	if (prop_category != ucp_M) break;
2937	+	eptr += len;
2938	+	}
2939	+	}
2940	+	}
2941	+
2942	+	else
2943	+	#endif /* SUPPORT_UCP */
2944	+
2945	+	/* Handle all other cases when the coding is UTF-8 */
2946	+
2947	+	#ifdef SUPPORT_UTF8
2948	+	if (utf8) switch(ctype)
2949	+	{
2950	+	case OP_ANY:
2951	+	for (i = 1; i <= min; i++)
2952	+	{
2953	+	if (eptr >= md->end_subject \|\| IS_NEWLINE(eptr))
2954	+	RRETURN(MATCH_NOMATCH);
2955	+	eptr++;
2956	+	while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
2957	+	}
2958	+	break;
2959	+
2960	+	case OP_ALLANY:
2961	+	for (i = 1; i <= min; i++)
2962	+	{
2963	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
2964	+	eptr++;
2965	+	while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
2966	+	}
2967	+	break;
2968	+
2969	+	case OP_ANYBYTE:
2970	+	eptr += min;
2971	+	break;
2972	+
2973	+	case OP_ANYNL:
2974	+	for (i = 1; i <= min; i++)
2975	+	{
2976	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
2977	+	GETCHARINC(c, eptr);
2978	+	switch(c)
2979	+	{
2980	+	default: RRETURN(MATCH_NOMATCH);
2981	+	case 0x000d:
2982	+	if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
2983	+	break;
2984	+
2985	+	case 0x000a:
2986	+	break;
2987	+
2988	+	case 0x000b:
2989	+	case 0x000c:
2990	+	case 0x0085:
2991	+	case 0x2028:
2992	+	case 0x2029:
2993	+	if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2994	+	break;
2995	+	}
2996	+	}
2997	+	break;
2998	+
2999	+	case OP_NOT_HSPACE:
3000	+	for (i = 1; i <= min; i++)
3001	+	{
3002	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3003	+	GETCHARINC(c, eptr);
3004	+	switch(c)
3005	+	{
3006	+	default: break;
3007	+	case 0x09: /* HT */
3008	+	case 0x20: /* SPACE */
3009	+	case 0xa0: /* NBSP */
3010	+	case 0x1680: /* OGHAM SPACE MARK */
3011	+	case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
3012	+	case 0x2000: /* EN QUAD */
3013	+	case 0x2001: /* EM QUAD */
3014	+	case 0x2002: /* EN SPACE */
3015	+	case 0x2003: /* EM SPACE */
3016	+	case 0x2004: /* THREE-PER-EM SPACE */
3017	+	case 0x2005: /* FOUR-PER-EM SPACE */
3018	+	case 0x2006: /* SIX-PER-EM SPACE */
3019	+	case 0x2007: /* FIGURE SPACE */
3020	+	case 0x2008: /* PUNCTUATION SPACE */
3021	+	case 0x2009: /* THIN SPACE */
3022	+	case 0x200A: /* HAIR SPACE */
3023	+	case 0x202f: /* NARROW NO-BREAK SPACE */
3024	+	case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
3025	+	case 0x3000: /* IDEOGRAPHIC SPACE */
3026	+	RRETURN(MATCH_NOMATCH);
3027	+	}
3028	+	}
3029	+	break;
3030	+
3031	+	case OP_HSPACE:
3032	+	for (i = 1; i <= min; i++)
3033	+	{
3034	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3035	+	GETCHARINC(c, eptr);
3036	+	switch(c)
3037	+	{
3038	+	default: RRETURN(MATCH_NOMATCH);
3039	+	case 0x09: /* HT */
3040	+	case 0x20: /* SPACE */
3041	+	case 0xa0: /* NBSP */
3042	+	case 0x1680: /* OGHAM SPACE MARK */
3043	+	case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
3044	+	case 0x2000: /* EN QUAD */
3045	+	case 0x2001: /* EM QUAD */
3046	+	case 0x2002: /* EN SPACE */
3047	+	case 0x2003: /* EM SPACE */
3048	+	case 0x2004: /* THREE-PER-EM SPACE */
3049	+	case 0x2005: /* FOUR-PER-EM SPACE */
3050	+	case 0x2006: /* SIX-PER-EM SPACE */
3051	+	case 0x2007: /* FIGURE SPACE */
3052	+	case 0x2008: /* PUNCTUATION SPACE */
3053	+	case 0x2009: /* THIN SPACE */
3054	+	case 0x200A: /* HAIR SPACE */
3055	+	case 0x202f: /* NARROW NO-BREAK SPACE */
3056	+	case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
3057	+	case 0x3000: /* IDEOGRAPHIC SPACE */
3058	+	break;
3059	+	}
3060	+	}
3061	+	break;
3062	+
3063	+	case OP_NOT_VSPACE:
3064	+	for (i = 1; i <= min; i++)
3065	+	{
3066	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3067	+	GETCHARINC(c, eptr);
3068	+	switch(c)
3069	+	{
3070	+	default: break;
3071	+	case 0x0a: /* LF */
3072	+	case 0x0b: /* VT */
3073	+	case 0x0c: /* FF */
3074	+	case 0x0d: /* CR */
3075	+	case 0x85: /* NEL */
3076	+	case 0x2028: /* LINE SEPARATOR */
3077	+	case 0x2029: /* PARAGRAPH SEPARATOR */
3078	+	RRETURN(MATCH_NOMATCH);
3079	+	}
3080	+	}
3081	+	break;
3082	+
3083	+	case OP_VSPACE:
3084	+	for (i = 1; i <= min; i++)
3085	+	{
3086	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3087	+	GETCHARINC(c, eptr);
3088	+	switch(c)
3089	+	{
3090	+	default: RRETURN(MATCH_NOMATCH);
3091	+	case 0x0a: /* LF */
3092	+	case 0x0b: /* VT */
3093	+	case 0x0c: /* FF */
3094	+	case 0x0d: /* CR */
3095	+	case 0x85: /* NEL */
3096	+	case 0x2028: /* LINE SEPARATOR */
3097	+	case 0x2029: /* PARAGRAPH SEPARATOR */
3098	+	break;
3099	+	}
3100	+	}
3101	+	break;
3102	+
3103	+	case OP_NOT_DIGIT:
3104	+	for (i = 1; i <= min; i++)
3105	+	{
3106	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3107	+	GETCHARINC(c, eptr);
3108	+	if (c < 128 && (md->ctypes[c] & ctype_digit) != 0)
3109	+	RRETURN(MATCH_NOMATCH);
3110	+	}
3111	+	break;
3112	+
3113	+	case OP_DIGIT:
3114	+	for (i = 1; i <= min; i++)
3115	+	{
3116	+	if (eptr >= md->end_subject \|\|
3117	+	eptr >= 128 \|\| (md->ctypes[eptr++] & ctype_digit) == 0)
3118	+	RRETURN(MATCH_NOMATCH);
3119	+	/* No need to skip more bytes - we know it's a 1-byte character */
3120	+	}
3121	+	break;
3122	+
3123	+	case OP_NOT_WHITESPACE:
3124	+	for (i = 1; i <= min; i++)
3125	+	{
3126	+	if (eptr >= md->end_subject \|\|
3127	+	(eptr < 128 && (md->ctypes[eptr] & ctype_space) != 0))
3128	+	RRETURN(MATCH_NOMATCH);
3129	+	while (++eptr < md->end_subject && (*eptr & 0xc0) == 0x80);
3130	+	}
3131	+	break;
3132	+
3133	+	case OP_WHITESPACE:
3134	+	for (i = 1; i <= min; i++)
3135	+	{
3136	+	if (eptr >= md->end_subject \|\|
3137	+	eptr >= 128 \|\| (md->ctypes[eptr++] & ctype_space) == 0)
3138	+	RRETURN(MATCH_NOMATCH);
3139	+	/* No need to skip more bytes - we know it's a 1-byte character */
3140	+	}
3141	+	break;
3142	+
3143	+	case OP_NOT_WORDCHAR:
3144	+	for (i = 1; i <= min; i++)
3145	+	{
3146	+	if (eptr >= md->end_subject \|\|
3147	+	(eptr < 128 && (md->ctypes[eptr] & ctype_word) != 0))
3148	+	RRETURN(MATCH_NOMATCH);
3149	+	while (++eptr < md->end_subject && (*eptr & 0xc0) == 0x80);
3150	+	}
3151	+	break;
3152	+
3153	+	case OP_WORDCHAR:
3154	+	for (i = 1; i <= min; i++)
3155	+	{
3156	+	if (eptr >= md->end_subject \|\|
3157	+	eptr >= 128 \|\| (md->ctypes[eptr++] & ctype_word) == 0)
3158	+	RRETURN(MATCH_NOMATCH);
3159	+	/* No need to skip more bytes - we know it's a 1-byte character */
3160	+	}
3161	+	break;
3162	+
3163	+	default:
3164	+	RRETURN(PCRE_ERROR_INTERNAL);
3165	+	} /* End switch(ctype) */
3166	+
3167	+	else
3168	+	#endif /* SUPPORT_UTF8 */
3169	+
3170		/* Code for the non-UTF-8 case for minimum matching of operators other
3171	<	than OP_PROP and OP_NOTPROP. */
3171	>	than OP_PROP and OP_NOTPROP. We can assume that there are the minimum
3172	>	number of bytes present, as this was tested above. */
3173
3174		switch(ctype)
3175		{
3176		case OP_ANY:
3177	<	if ((ims & PCRE_DOTALL) == 0)
3177	>	for (i = 1; i <= min; i++)
3178		{
3179	<	for (i = 1; i <= min; i++)
3180	<	if (*eptr++ == NEWLINE) RRETURN(MATCH_NOMATCH);
3179	>	if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
3180	>	eptr++;
3181		}
3182	<	else eptr += min;
3182	>	break;
3183	>
3184	>	case OP_ALLANY:
3185	>	eptr += min;
3186		break;
3187
3188		case OP_ANYBYTE:
3189		eptr += min;
3190		break;
3191
3192	+	/* Because of the CRLF case, we can't assume the minimum number of
3193	+	bytes are present in this case. */
3194	+
3195	+	case OP_ANYNL:
3196	+	for (i = 1; i <= min; i++)
3197	+	{
3198	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3199	+	switch(*eptr++)
3200	+	{
3201	+	default: RRETURN(MATCH_NOMATCH);
3202	+	case 0x000d:
3203	+	if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
3204	+	break;
3205	+	case 0x000a:
3206	+	break;
3207	+
3208	+	case 0x000b:
3209	+	case 0x000c:
3210	+	case 0x0085:
3211	+	if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
3212	+	break;
3213	+	}
3214	+	}
3215	+	break;
3216	+
3217	+	case OP_NOT_HSPACE:
3218	+	for (i = 1; i <= min; i++)
3219	+	{
3220	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3221	+	switch(*eptr++)
3222	+	{
3223	+	default: break;
3224	+	case 0x09: /* HT */
3225	+	case 0x20: /* SPACE */
3226	+	case 0xa0: /* NBSP */
3227	+	RRETURN(MATCH_NOMATCH);
3228	+	}
3229	+	}
3230	+	break;
3231	+
3232	+	case OP_HSPACE:
3233	+	for (i = 1; i <= min; i++)
3234	+	{
3235	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3236	+	switch(*eptr++)
3237	+	{
3238	+	default: RRETURN(MATCH_NOMATCH);
3239	+	case 0x09: /* HT */
3240	+	case 0x20: /* SPACE */
3241	+	case 0xa0: /* NBSP */
3242	+	break;
3243	+	}
3244	+	}
3245	+	break;
3246	+
3247	+	case OP_NOT_VSPACE:
3248	+	for (i = 1; i <= min; i++)
3249	+	{
3250	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3251	+	switch(*eptr++)
3252	+	{
3253	+	default: break;
3254	+	case 0x0a: /* LF */
3255	+	case 0x0b: /* VT */
3256	+	case 0x0c: /* FF */
3257	+	case 0x0d: /* CR */
3258	+	case 0x85: /* NEL */
3259	+	RRETURN(MATCH_NOMATCH);
3260	+	}
3261	+	}
3262	+	break;
3263	+
3264	+	case OP_VSPACE:
3265	+	for (i = 1; i <= min; i++)
3266	+	{
3267	+	if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3268	+	switch(*eptr++)
3269	+	{
3270	+	default: RRETURN(MATCH_NOMATCH);
3271	+	case 0x0a: /* LF */
3272	+	case 0x0b: /* VT */
3273	+	case 0x0c: /* FF */
3274	+	case 0x0d: /* CR */
3275	+	case 0x85: /* NEL */
3276	+	break;
3277	+	}
3278	+	}
3279	+	break;
3280	+
3281		case OP_NOT_DIGIT:
3282		for (i = 1; i <= min; i++)
3283		if ((md->ctypes[*eptr++] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
#	Line 1887 \| Line 3325 \| for (;;)
3325
3326		if (minimize)
3327		{
3328	<	/* Not UTF-8 mode */
3328	>	#ifdef SUPPORT_UCP
3329	>	if (prop_type >= 0)
3330	>	{
3331	>	switch(prop_type)
3332	>	{
3333	>	case PT_ANY:
3334	>	for (fi = min;; fi++)
3335	>	{
3336	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM36);
3337	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3338	>	if (fi >= max \|\| eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3339	>	GETCHARINC(c, eptr);
3340	>	if (prop_fail_result) RRETURN(MATCH_NOMATCH);
3341	>	}
3342	>	/* Control never gets here */
3343	>
3344	>	case PT_LAMP:
3345	>	for (fi = min;; fi++)
3346	>	{
3347	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM37);
3348	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3349	>	if (fi >= max \|\| eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3350	>	GETCHARINC(c, eptr);
3351	>	prop_chartype = UCD_CHARTYPE(c);
3352	>	if ((prop_chartype == ucp_Lu \|\|
3353	>	prop_chartype == ucp_Ll \|\|
3354	>	prop_chartype == ucp_Lt) == prop_fail_result)
3355	>	RRETURN(MATCH_NOMATCH);
3356	>	}
3357	>	/* Control never gets here */
3358	>
3359	>	case PT_GC:
3360	>	for (fi = min;; fi++)
3361	>	{
3362	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM38);
3363	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3364	>	if (fi >= max \|\| eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3365	>	GETCHARINC(c, eptr);
3366	>	prop_category = UCD_CATEGORY(c);
3367	>	if ((prop_category == prop_value) == prop_fail_result)
3368	>	RRETURN(MATCH_NOMATCH);
3369	>	}
3370	>	/* Control never gets here */
3371	>
3372	>	case PT_PC:
3373	>	for (fi = min;; fi++)
3374	>	{
3375	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM39);
3376	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3377	>	if (fi >= max \|\| eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3378	>	GETCHARINC(c, eptr);
3379	>	prop_chartype = UCD_CHARTYPE(c);
3380	>	if ((prop_chartype == prop_value) == prop_fail_result)
3381	>	RRETURN(MATCH_NOMATCH);
3382	>	}
3383	>	/* Control never gets here */
3384	>
3385	>	case PT_SC:
3386	>	for (fi = min;; fi++)
3387	>	{
3388	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM40);
3389	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3390	>	if (fi >= max \|\| eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3391	>	GETCHARINC(c, eptr);
3392	>	prop_script = UCD_SCRIPT(c);
3393	>	if ((prop_script == prop_value) == prop_fail_result)
3394	>	RRETURN(MATCH_NOMATCH);
3395	>	}
3396	>	/* Control never gets here */
3397	>
3398	>	default:
3399	>	RRETURN(PCRE_ERROR_INTERNAL);
3400	>	}
3401	>	}
3402	>
3403	>	/* Match extended Unicode sequences. We will get here only if the
3404	>	support is in the binary; otherwise a compile-time error occurs. */
3405	>
3406	>	else if (ctype == OP_EXTUNI)
3407		{
3408		for (fi = min;; fi++)
3409		{
3410	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
3410	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM41);
3411		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3412		if (fi >= max \|\| eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);
3413	<	c = *eptr++;
3413	>	GETCHARINCTEST(c, eptr);
3414	>	prop_category = UCD_CATEGORY(c);
3415	>	if (prop_category == ucp_M) RRETURN(MATCH_NOMATCH);
3416	>	while (eptr < md->end_subject)
3417	>	{
3418	>	int len = 1;
3419	>	if (!utf8) c = *eptr; else
3420	>	{
3421	>	GETCHARLEN(c, eptr, len);
3422	>	}
3423	>	prop_category = UCD_CATEGORY(c);
3424	>	if (prop_category != ucp_M) break;
3425	>	eptr += len;
3426	>	}
3427	>	}
3428	>	}
3429	>
3430	>	else
3431	>	#endif /* SUPPORT_UCP */
3432	>
3433	>	#ifdef SUPPORT_UTF8
3434	>	/* UTF-8 mode */
3435	>	if (utf8)
3436	>	{
3437	>	for (fi = min;; fi++)
3438	>	{
3439	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM42);
3440	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3441	>	if (fi >= max \|\| eptr >= md->end_subject \|\|
3442	>	(ctype == OP_ANY && IS_NEWLINE(eptr)))
3443	>	RRETURN(MATCH_NOMATCH);
3444	>
3445	>	GETCHARINC(c, eptr);
3446		switch(ctype)
3447		{
3448	<	case OP_ANY:
3449	<	if ((ims & PCRE_DOTALL) == 0 && c == NEWLINE) RRETURN(MATCH_NOMATCH);
3448	>	case OP_ANY: /* This is the non-NL case */
3449	>	case OP_ALLANY:
3450	>	case OP_ANYBYTE:
3451	>	break;
3452	>
3453	>	case OP_ANYNL:
3454	>	switch(c)
3455	>	{
3456	>	default: RRETURN(MATCH_NOMATCH);
3457	>	case 0x000d:
3458	>	if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
3459	>	break;
3460	>	case 0x000a:
3461	>	break;
3462	>
3463	>	case 0x000b:
3464	>	case 0x000c:
3465	>	case 0x0085:
3466	>	case 0x2028:
3467	>	case 0x2029:
3468	>	if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
3469	>	break;
3470	>	}
3471	>	break;
3472	>
3473	>	case OP_NOT_HSPACE:
3474	>	switch(c)
3475	>	{
3476	>	default: break;
3477	>	case 0x09: /* HT */
3478	>	case 0x20: /* SPACE */
3479	>	case 0xa0: /* NBSP */
3480	>	case 0x1680: /* OGHAM SPACE MARK */
3481	>	case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
3482	>	case 0x2000: /* EN QUAD */
3483	>	case 0x2001: /* EM QUAD */
3484	>	case 0x2002: /* EN SPACE */
3485	>	case 0x2003: /* EM SPACE */
3486	>	case 0x2004: /* THREE-PER-EM SPACE */
3487	>	case 0x2005: /* FOUR-PER-EM SPACE */
3488	>	case 0x2006: /* SIX-PER-EM SPACE */
3489	>	case 0x2007: /* FIGURE SPACE */
3490	>	case 0x2008: /* PUNCTUATION SPACE */
3491	>	case 0x2009: /* THIN SPACE */
3492	>	case 0x200A: /* HAIR SPACE */
3493	>	case 0x202f: /* NARROW NO-BREAK SPACE */
3494	>	case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
3495	>	case 0x3000: /* IDEOGRAPHIC SPACE */
3496	>	RRETURN(MATCH_NOMATCH);
3497	>	}
3498	>	break;
3499	>
3500	>	case OP_HSPACE:
3501	>	switch(c)
3502	>	{
3503	>	default: RRETURN(MATCH_NOMATCH);
3504	>	case 0x09: /* HT */
3505	>	case 0x20: /* SPACE */
3506	>	case 0xa0: /* NBSP */
3507	>	case 0x1680: /* OGHAM SPACE MARK */
3508	>	case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
3509	>	case 0x2000: /* EN QUAD */
3510	>	case 0x2001: /* EM QUAD */
3511	>	case 0x2002: /* EN SPACE */
3512	>	case 0x2003: /* EM SPACE */
3513	>	case 0x2004: /* THREE-PER-EM SPACE */
3514	>	case 0x2005: /* FOUR-PER-EM SPACE */
3515	>	case 0x2006: /* SIX-PER-EM SPACE */
3516	>	case 0x2007: /* FIGURE SPACE */
3517	>	case 0x2008: /* PUNCTUATION SPACE */
3518	>	case 0x2009: /* THIN SPACE */
3519	>	case 0x200A: /* HAIR SPACE */
3520	>	case 0x202f: /* NARROW NO-BREAK SPACE */
3521	>	case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
3522	>	case 0x3000: /* IDEOGRAPHIC SPACE */
3523	>	break;
3524	>	}
3525	>	break;
3526	>
3527	>	case OP_NOT_VSPACE:
3528	>	switch(c)
3529	>	{
3530	>	default: break;
3531	>	case 0x0a: /* LF */
3532	>	case 0x0b: /* VT */
3533	>	case 0x0c: /* FF */
3534	>	case 0x0d: /* CR */
3535	>	case 0x85: /* NEL */
3536	>	case 0x2028: /* LINE SEPARATOR */
3537	>	case 0x2029: /* PARAGRAPH SEPARATOR */
3538	>	RRETURN(MATCH_NOMATCH);
3539	>	}
3540	>	break;
3541	>
3542	>	case OP_VSPACE:
3543	>	switch(c)
3544	>	{
3545	>	default: RRETURN(MATCH_NOMATCH);
3546	>	case 0x0a: /* LF */
3547	>	case 0x0b: /* VT */
3548	>	case 0x0c: /* FF */
3549	>	case 0x0d: /* CR */
3550	>	case 0x85: /* NEL */
3551	>	case 0x2028: /* LINE SEPARATOR */
3552	>	case 0x2029: /* PARAGRAPH SEPARATOR */
3553	>	break;
3554	>	}
3555	>	break;
3556	>
3557	>	case OP_NOT_DIGIT:
3558	>	if (c < 256 && (md->ctypes[c] & ctype_digit) != 0)
3559	>	RRETURN(MATCH_NOMATCH);
3560		break;
3561
3562	+	case OP_DIGIT:
3563	+	if (c >= 256 \|\| (md->ctypes[c] & ctype_digit) == 0)
3564	+	RRETURN(MATCH_NOMATCH);
3565	+	break;
3566	+
3567	+	case OP_NOT_WHITESPACE:
3568	+	if (c < 256 && (md->ctypes[c] & ctype_space) != 0)
3569	+	RRETURN(MATCH_NOMATCH);
3570	+	break;
3571	+
3572	+	case OP_WHITESPACE:
3573	+	if (c >= 256 \|\| (md->ctypes[c] & ctype_space) == 0)
3574	+	RRETURN(MATCH_NOMATCH);
3575	+	break;
3576	+
3577	+	case OP_NOT_WORDCHAR:
3578	+	if (c < 256 && (md->ctypes[c] & ctype_word) != 0)
3579	+	RRETURN(MATCH_NOMATCH);
3580	+	break;
3581	+
3582	+	case OP_WORDCHAR:
3583	+	if (c >= 256 \|\| (md->ctypes[c] & ctype_word) == 0)
3584	+	RRETURN(MATCH_NOMATCH);
3585	+	break;
3586	+
3587	+	default:
3588	+	RRETURN(PCRE_ERROR_INTERNAL);
3589	+	}
3590	+	}
3591	+	}
3592	+	else
3593	+	#endif
3594	+	/* Not UTF-8 mode */
3595	+	{
3596	+	for (fi = min;; fi++)
3597	+	{
3598	+	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM43);
3599	+	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3600	+	if (fi >= max \|\| eptr >= md->end_subject \|\|
3601	+	(ctype == OP_ANY && IS_NEWLINE(eptr)))
3602	+	RRETURN(MATCH_NOMATCH);
3603	+
3604	+	c = *eptr++;
3605	+	switch(ctype)
3606	+	{
3607	+	case OP_ANY: /* This is the non-NL case */
3608	+	case OP_ALLANY:
3609		case OP_ANYBYTE:
3610		break;
3611
3612	+	case OP_ANYNL:
3613	+	switch(c)
3614	+	{
3615	+	default: RRETURN(MATCH_NOMATCH);
3616	+	case 0x000d:
3617	+	if (eptr < md->end_subject && *eptr == 0x0a) eptr++;
3618	+	break;
3619	+
3620	+	case 0x000a:
3621	+	break;
3622	+
3623	+	case 0x000b:
3624	+	case 0x000c:
3625	+	case 0x0085:
3626	+	if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
3627	+	break;
3628	+	}
3629	+	break;
3630	+
3631	+	case OP_NOT_HSPACE:
3632	+	switch(c)
3633	+	{
3634	+	default: break;
3635	+	case 0x09: /* HT */
3636	+	case 0x20: /* SPACE */
3637	+	case 0xa0: /* NBSP */
3638	+	RRETURN(MATCH_NOMATCH);
3639	+	}
3640	+	break;
3641	+
3642	+	case OP_HSPACE:
3643	+	switch(c)
3644	+	{
3645	+	default: RRETURN(MATCH_NOMATCH);
3646	+	case 0x09: /* HT */
3647	+	case 0x20: /* SPACE */
3648	+	case 0xa0: /* NBSP */
3649	+	break;
3650	+	}
3651	+	break;
3652	+
3653	+	case OP_NOT_VSPACE:
3654	+	switch(c)
3655	+	{
3656	+	default: break;
3657	+	case 0x0a: /* LF */
3658	+	case 0x0b: /* VT */
3659	+	case 0x0c: /* FF */
3660	+	case 0x0d: /* CR */
3661	+	case 0x85: /* NEL */
3662	+	RRETURN(MATCH_NOMATCH);
3663	+	}
3664	+	break;
3665	+
3666	+	case OP_VSPACE:
3667	+	switch(c)
3668	+	{
3669	+	default: RRETURN(MATCH_NOMATCH);
3670	+	case 0x0a: /* LF */
3671	+	case 0x0b: /* VT */
3672	+	case 0x0c: /* FF */
3673	+	case 0x0d: /* CR */
3674	+	case 0x85: /* NEL */
3675	+	break;
3676	+	}
3677	+	break;
3678	+
3679		case OP_NOT_DIGIT:
3680		if ((md->ctypes[c] & ctype_digit) != 0) RRETURN(MATCH_NOMATCH);
3681		break;
#	Line 1936 \| Line 3708 \| for (;;)
3708		/* Control never gets here */
3709		}
3710
3711	<	/* If maximizing it is worth using inline code for speed, doing the type
3711	>	/* If maximizing, it is worth using inline code for speed, doing the type
3712		test once at the start (i.e. keep it out of the loop). Again, keep the
3713		UTF-8 and UCP stuff separate. */
3714
#	Line 1944 \| Line 3716 \| for (;;)
3716		{
3717		pp = eptr; /* Remember where we started */
3718
3719	<	/* Not UTF-8 mode */
3719	>	#ifdef SUPPORT_UCP
3720	>	if (prop_type >= 0)
3721	>	{
3722	>	switch(prop_type)
3723	>	{
3724	>	case PT_ANY:
3725	>	for (i = min; i < max; i++)
3726	>	{
3727	>	int len = 1;
3728	>	if (eptr >= md->end_subject) break;
3729	>	GETCHARLEN(c, eptr, len);
3730	>	if (prop_fail_result) break;
3731	>	eptr+= len;
3732	>	}
3733	>	break;
3734	>
3735	>	case PT_LAMP:
3736	>	for (i = min; i < max; i++)
3737	>	{
3738	>	int len = 1;
3739	>	if (eptr >= md->end_subject) break;
3740	>	GETCHARLEN(c, eptr, len);
3741	>	prop_chartype = UCD_CHARTYPE(c);
3742	>	if ((prop_chartype == ucp_Lu \|\|
3743	>	prop_chartype == ucp_Ll \|\|
3744	>	prop_chartype == ucp_Lt) == prop_fail_result)
3745	>	break;
3746	>	eptr+= len;
3747	>	}
3748	>	break;
3749	>
3750	>	case PT_GC:
3751	>	for (i = min; i < max; i++)
3752	>	{
3753	>	int len = 1;
3754	>	if (eptr >= md->end_subject) break;
3755	>	GETCHARLEN(c, eptr, len);
3756	>	prop_category = UCD_CATEGORY(c);
3757	>	if ((prop_category == prop_value) == prop_fail_result)
3758	>	break;
3759	>	eptr+= len;
3760	>	}
3761	>	break;
3762	>
3763	>	case PT_PC:
3764	>	for (i = min; i < max; i++)
3765	>	{
3766	>	int len = 1;
3767	>	if (eptr >= md->end_subject) break;
3768	>	GETCHARLEN(c, eptr, len);
3769	>	prop_chartype = UCD_CHARTYPE(c);
3770	>	if ((prop_chartype == prop_value) == prop_fail_result)
3771	>	break;
3772	>	eptr+= len;
3773	>	}
3774	>	break;
3775	>
3776	>	case PT_SC:
3777	>	for (i = min; i < max; i++)
3778	>	{
3779	>	int len = 1;
3780	>	if (eptr >= md->end_subject) break;
3781	>	GETCHARLEN(c, eptr, len);
3782	>	prop_script = UCD_SCRIPT(c);
3783	>	if ((prop_script == prop_value) == prop_fail_result)
3784	>	break;
3785	>	eptr+= len;
3786	>	}
3787	>	break;
3788	>	}
3789	>
3790	>	/* eptr is now past the end of the maximum run */
3791	>
3792	>	if (possessive) continue;
3793	>	for(;;)
3794	>	{
3795	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM44);
3796	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3797	>	if (eptr-- == pp) break; /* Stop if tried at original pos */
3798	>	if (utf8) BACKCHAR(eptr);
3799	>	}
3800	>	}
3801	>
3802	>	/* Match extended Unicode sequences. We will get here only if the
3803	>	support is in the binary; otherwise a compile-time error occurs. */
3804	>
3805	>	else if (ctype == OP_EXTUNI)
3806	>	{
3807	>	for (i = min; i < max; i++)
3808	>	{
3809	>	if (eptr >= md->end_subject) break;
3810	>	GETCHARINCTEST(c, eptr);
3811	>	prop_category = UCD_CATEGORY(c);
3812	>	if (prop_category == ucp_M) break;
3813	>	while (eptr < md->end_subject)
3814	>	{
3815	>	int len = 1;
3816	>	if (!utf8) c = *eptr; else
3817	>	{
3818	>	GETCHARLEN(c, eptr, len);
3819	>	}
3820	>	prop_category = UCD_CATEGORY(c);
3821	>	if (prop_category != ucp_M) break;
3822	>	eptr += len;
3823	>	}
3824	>	}
3825	>
3826	>	/* eptr is now past the end of the maximum run */
3827	>
3828	>	if (possessive) continue;
3829	>	for(;;)
3830	>	{
3831	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM45);
3832	>	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3833	>	if (eptr-- == pp) break; /* Stop if tried at original pos */
3834	>	for (;;) /* Move back over one extended */
3835	>	{
3836	>	int len = 1;
3837	>	if (!utf8) c = *eptr; else
3838	>	{
3839	>	BACKCHAR(eptr);
3840	>	GETCHARLEN(c, eptr, len);
3841	>	}
3842	>	prop_category = UCD_CATEGORY(c);
3843	>	if (prop_category != ucp_M) break;
3844	>	eptr--;
3845	>	}
3846	>	}
3847	>	}
3848	>
3849	>	else
3850	>	#endif /* SUPPORT_UCP */
3851	>
3852	>	#ifdef SUPPORT_UTF8
3853	>	/* UTF-8 mode */
3854	>
3855	>	if (utf8)
3856		{
3857		switch(ctype)
3858		{
3859		case OP_ANY:
3860	<	if ((ims & PCRE_DOTALL) == 0)
3860	>	if (max < INT_MAX)
3861		{
3862		for (i = min; i < max; i++)
3863		{
3864	<	if (eptr >= md->end_subject \|\| *eptr == NEWLINE) break;
3864	>	if (eptr >= md->end_subject \|\| IS_NEWLINE(eptr)) break;
3865		eptr++;
3866	+	while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
3867		}
1959	–	break;
3868		}
1961	–	/* For DOTALL case, fall through and treat as \C */
3869
3870	+	/* Handle unlimited UTF-8 repeat */
3871	+
3872	+	else
3873	+	{
3874	+	for (i = min; i < max; i++)
3875	+	{
3876	+	if (eptr >= md->end_subject \|\| IS_NEWLINE(eptr)) break;
3877	+	eptr++;
3878	+	while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
3879	+	}
3880	+	}
3881	+	break;
3882	+
3883	+	case OP_ALLANY:
3884	+	if (max < INT_MAX)
3885	+	{
3886	+	for (i = min; i < max; i++)
3887	+	{
3888	+	if (eptr >= md->end_subject) break;
3889	+	eptr++;
3890	+	while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;
3891	+	}
3892	+	}
3893	+	else eptr = md->end_subject; /* Unlimited UTF-8 repeat */
3894	+	break;
3895	+
3896	+	/* The byte case is the same as non-UTF8 */
3897	+
3898	+	case OP_ANYBYTE:
3899	+	c = max - min;
3900	+	if (c > (unsigned int)(md->end_subject - eptr))
3901	+	c = md->end_subject - eptr;
3902	+	eptr += c;
3903	+	break;
3904	+
3905	+	case OP_ANYNL:
3906	+	for (i = min; i < max; i++)
3907	+	{
3908	+	int len = 1;
3909	+	if (eptr >= md->end_subject) break;
3910	+	GETCHARLEN(c, eptr, len);
3911	+	if (c == 0x000d)
3912	+	{
3913	+	if (++eptr >= md->end_subject) break;
3914	+	if (*eptr == 0x000a) eptr++;
3915	+	}
3916	+	else
3917	+	{
3918	+	if (c != 0x000a &&
3919	+	(md->bsr_anycrlf \|\|
3920	+	(c != 0x000b && c != 0x000c &&
3921	+	c != 0x0085 && c != 0x2028 && c != 0x2029)))
3922	+	break;
3923	+	eptr += len;
3924	+	}
3925	+	}
3926	+	break;
3927	+
3928	+	case OP_NOT_HSPACE:
3929	+	case OP_HSPACE:
3930	+	for (i = min; i < max; i++)
3931	+	{
3932	+	BOOL gotspace;
3933	+	int len = 1;
3934	+	if (eptr >= md->end_subject) break;
3935	+	GETCHARLEN(c, eptr, len);
3936	+	switch(c)
3937	+	{
3938	+	default: gotspace = FALSE; break;
3939	+	case 0x09: /* HT */
3940	+	case 0x20: /* SPACE */
3941	+	case 0xa0: /* NBSP */
3942	+	case 0x1680: /* OGHAM SPACE MARK */
3943	+	case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
3944	+	case 0x2000: /* EN QUAD */
3945	+	case 0x2001: /* EM QUAD */
3946	+	case 0x2002: /* EN SPACE */
3947	+	case 0x2003: /* EM SPACE */
3948	+	case 0x2004: /* THREE-PER-EM SPACE */
3949	+	case 0x2005: /* FOUR-PER-EM SPACE */
3950	+	case 0x2006: /* SIX-PER-EM SPACE */
3951	+	case 0x2007: /* FIGURE SPACE */
3952	+	case 0x2008: /* PUNCTUATION SPACE */
3953	+	case 0x2009: /* THIN SPACE */
3954	+	case 0x200A: /* HAIR SPACE */
3955	+	case 0x202f: /* NARROW NO-BREAK SPACE */
3956	+	case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
3957	+	case 0x3000: /* IDEOGRAPHIC SPACE */
3958	+	gotspace = TRUE;
3959	+	break;
3960	+	}
3961	+	if (gotspace == (ctype == OP_NOT_HSPACE)) break;
3962	+	eptr += len;
3963	+	}
3964	+	break;
3965	+
3966	+	case OP_NOT_VSPACE:
3967	+	case OP_VSPACE:
3968	+	for (i = min; i < max; i++)
3969	+	{
3970	+	BOOL gotspace;
3971	+	int len = 1;
3972	+	if (eptr >= md->end_subject) break;
3973	+	GETCHARLEN(c, eptr, len);
3974	+	switch(c)
3975	+	{
3976	+	default: gotspace = FALSE; break;
3977	+	case 0x0a: /* LF */
3978	+	case 0x0b: /* VT */
3979	+	case 0x0c: /* FF */
3980	+	case 0x0d: /* CR */
3981	+	case 0x85: /* NEL */
3982	+	case 0x2028: /* LINE SEPARATOR */
3983	+	case 0x2029: /* PARAGRAPH SEPARATOR */
3984	+	gotspace = TRUE;
3985	+	break;
3986	+	}
3987	+	if (gotspace == (ctype == OP_NOT_VSPACE)) break;
3988	+	eptr += len;
3989	+	}
3990	+	break;
3991	+
3992	+	case OP_NOT_DIGIT:
3993	+	for (i = min; i < max; i++)
3994	+	{
3995	+	int len = 1;
3996	+	if (eptr >= md->end_subject) break;
3997	+	GETCHARLEN(c, eptr, len);
3998	+	if (c < 256 && (md->ctypes[c] & ctype_digit) != 0) break;
3999	+	eptr+= len;
4000	+	}
4001	+	break;
4002	+
4003	+	case OP_DIGIT:
4004	+	for (i = min; i < max; i++)
4005	+	{
4006	+	int len = 1;
4007	+	if (eptr >= md->end_subject) break;
4008	+	GETCHARLEN(c, eptr, len);
4009	+	if (c >= 256 \|\|(md->ctypes[c] & ctype_digit) == 0) break;
4010	+	eptr+= len;
4011	+	}
4012	+	break;
4013	+
4014	+	case OP_NOT_WHITESPACE:
4015	+	for (i = min; i < max; i++)
4016	+	{
4017	+	int len = 1;
4018	+	if (eptr >= md->end_subject) break;
4019	+	GETCHARLEN(c, eptr, len);
4020	+	if (c < 256 && (md->ctypes[c] & ctype_space) != 0) break;
4021	+	eptr+= len;
4022	+	}
4023	+	break;
4024	+
4025	+	case OP_WHITESPACE:
4026	+	for (i = min; i < max; i++)
4027	+	{
4028	+	int len = 1;
4029	+	if (eptr >= md->end_subject) break;
4030	+	GETCHARLEN(c, eptr, len);
4031	+	if (c >= 256 \|\|(md->ctypes[c] & ctype_space) == 0) break;
4032	+	eptr+= len;
4033	+	}
4034	+	break;
4035	+
4036	+	case OP_NOT_WORDCHAR:
4037	+	for (i = min; i < max; i++)
4038	+	{
4039	+	int len = 1;
4040	+	if (eptr >= md->end_subject) break;
4041	+	GETCHARLEN(c, eptr, len);
4042	+	if (c < 256 && (md->ctypes[c] & ctype_word) != 0) break;
4043	+	eptr+= len;
4044	+	}
4045	+	break;
4046	+
4047	+	case OP_WORDCHAR:
4048	+	for (i = min; i < max; i++)
4049	+	{
4050	+	int len = 1;
4051	+	if (eptr >= md->end_subject) break;
4052	+	GETCHARLEN(c, eptr, len);
4053	+	if (c >= 256 \|\| (md->ctypes[c] & ctype_word) == 0) break;
4054	+	eptr+= len;
4055	+	}
4056	+	break;
4057	+
4058	+	default:
4059	+	RRETURN(PCRE_ERROR_INTERNAL);
4060	+	}
4061	+
4062	+	/* eptr is now past the end of the maximum run */
4063	+
4064	+	if (possessive) continue;
4065	+	for(;;)
4066	+	{
4067	+	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM46);
4068	+	if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4069	+	if (eptr-- == pp) break; /* Stop if tried at original pos */
4070	+	BACKCHAR(eptr);
4071	+	}
4072	+	}
4073	+	else
4074	+	#endif /* SUPPORT_UTF8 */
4075	+
4076	+	/* Not UTF-8 mode */
4077	+	{
4078	+	switch(ctype)
4079	+	{
4080	+	case OP_ANY:
4081	+	for (i = min; i < max; i++)
4082	+	{
4083	+	if (eptr >= md->end_subject \|\| IS_NEWLINE(eptr)) break;
4084	+	eptr++;
4085	+	}
4086	+	break;
4087	+
4088	+	case OP_ALLANY:
4089		case OP_ANYBYTE:
4090		c = max - min;
4091	<	if (c > md->end_subject - eptr) c = md->end_subject - eptr;
4091	>	if (c > (unsigned int)(md->end_subject - eptr))
4092	>	c = md->end_subject - eptr;
4093		eptr += c;
4094		break;
4095
4096	+	case OP_ANYNL:
4097	+	for (i = min; i < max; i++)
4098	+	{
4099	+	if (eptr >= md->end_subject) break;
4100	+	c = *eptr;
4101	+	if (c == 0x000d)
4102	+	{
4103	+	if (++eptr >= md->end_subject) break;
4104	+	if (*eptr == 0x000a) eptr++;
4105	+	}
4106	+	else
4107	+	{
4108	+	if (c != 0x000a &&
4109	+	(md->bsr_anycrlf \|\|
4110	+	(c != 0x000b && c != 0x000c && c != 0x0085)))
4111	+	break;
4112	+	eptr++;
4113	+	}
4114	+	}
4115	+	break;
4116	+
4117	+	case OP_NOT_HSPACE:
4118	+	for (i = min; i < max; i++)
4119	+	{
4120	+	if (eptr >= md->end_subject) break;
4121	+	c = *eptr;
4122	+	if (c == 0x09 \|\| c == 0x20 \|\| c == 0xa0) break;
4123	+	eptr++;
4124	+	}
4125	+	break;
4126	+
4127	+	case OP_HSPACE:
4128	+	for (i = min; i < max; i++)
4129	+	{
4130	+	if (eptr >= md->end_subject) break;
4131	+	c = *eptr;
4132	+	if (c != 0x09 && c != 0x20 && c != 0xa0) break;
4133	+	eptr++;
4134	+	}
4135	+	break;
4136	+
4137	+	case OP_NOT_VSPACE:
4138	+	for (i = min; i < max; i++)
4139	+	{
4140	+	if (eptr >= md->end_subject) break;
4141	+	c = *eptr;
4142	+	if (c == 0x0a \|\| c == 0x0b \|\| c == 0x0c \|\| c == 0x0d \|\| c == 0x85)
4143	+	break;
4144	+	eptr++;
4145	+	}
4146	+	break;
4147	+
4148	+	case OP_VSPACE:
4149	+	for (i = min; i < max; i++)
4150	+	{
4151	+	if (eptr >= md->end_subject) break;
4152	+	c = *eptr;
4153	+	if (c != 0x0a && c != 0x0b && c != 0x0c && c != 0x0d && c != 0x85)
4154	+	break;
4155	+	eptr++;
4156	+	}
4157	+	break;
4158	+
4159		case OP_NOT_DIGIT:
4160		for (i = min; i < max; i++)
4161		{
#	Line 2026 \| Line 4216 \| for (;;)
4216
4217		/* eptr is now past the end of the maximum run */
4218
4219	+	if (possessive) continue;
4220		while (eptr >= pp)
4221		{
4222	<	RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);
4222	>	RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM47);
4223		eptr--;
4224		if (rrc != MATCH_NOMATCH) RRETURN(rrc);
4225		}
#	Line 2040 \| Line 4231 \| for (;;)
4231		}
4232		/* Control never gets here */
4233
4234	<	/* There's been some horrible disaster. Since all codes > OP_BRA are
4235	<	for capturing brackets, and there shouldn't be any gaps between 0 and
2045	<	OP_BRA, arrival here can only mean there is something seriously wrong
2046	<	in the code above or the OP_xxx definitions. */
4234	>	/* There's been some horrible disaster. Arrival here can only mean there is
4235	>	something seriously wrong in the code above or the OP_xxx definitions. */
4236
4237		default:
4238		DPRINTF(("Unknown opcode %d\n", *ecode));
4239	<	RRETURN(PCRE_ERROR_UNKNOWN_NODE);
4239	>	RRETURN(PCRE_ERROR_UNKNOWN_OPCODE);
4240		}
4241
4242		/* Do not stick any code in here without much thought; it is assumed
#	Line 2056 \| Line 4245 \| for (;;)
4245
4246		} /* End of main loop */
4247		/* Control never reaches here */
4248	+
4249	+
4250	+	/* When compiling to use the heap rather than the stack for recursive calls to
4251	+	match(), the RRETURN() macro jumps here. The number that is saved in
4252	+	frame->Xwhere indicates which label we actually want to return to. */
4253	+
4254	+	#ifdef NO_RECURSE
4255	+	#define LBL(val) case val: goto L_RM##val;
4256	+	HEAP_RETURN:
4257	+	switch (frame->Xwhere)
4258	+	{
4259	+	LBL( 1) LBL( 2) LBL( 3) LBL( 4) LBL( 5) LBL( 6) LBL( 7) LBL( 8)
4260	+	LBL( 9) LBL(10) LBL(11) LBL(12) LBL(13) LBL(14) LBL(15) LBL(17)
4261	+	LBL(19) LBL(24) LBL(25) LBL(26) LBL(27) LBL(29) LBL(31) LBL(33)
4262	+	LBL(35) LBL(43) LBL(47) LBL(48) LBL(49) LBL(50) LBL(51) LBL(52)
4263	+	LBL(53) LBL(54)
4264	+	#ifdef SUPPORT_UTF8
4265	+	LBL(16) LBL(18) LBL(20) LBL(21) LBL(22) LBL(23) LBL(28) LBL(30)
4266	+	LBL(32) LBL(34) LBL(42) LBL(46)
4267	+	#ifdef SUPPORT_UCP
4268	+	LBL(36) LBL(37) LBL(38) LBL(39) LBL(40) LBL(41) LBL(44) LBL(45)
4269	+	#endif /* SUPPORT_UCP */
4270	+	#endif /* SUPPORT_UTF8 */
4271	+	default:
4272	+	DPRINTF(("jump error in pcre match: label %d non-existent\n", frame->Xwhere));
4273	+	return PCRE_ERROR_INTERNAL;
4274	+	}
4275	+	#undef LBL
4276	+	#endif /* NO_RECURSE */
4277		}
4278
4279
#	Line 2068 \| Line 4286 \| Undefine all the macros that were define
4286		#ifdef NO_RECURSE
4287		#undef eptr
4288		#undef ecode
4289	+	#undef mstart
4290		#undef offset_top
4291		#undef ims
4292		#undef eptrb
#	Line 2085 \| Line 4304 \| Undefine all the macros that were define
4304
4305		#undef cur_is_word
4306		#undef condition
2088	–	#undef minimize
4307		#undef prev_is_word
4308
4309		#undef original_ims
#	Line 2141 \| Line 4359 \| Returns: > 0 => success; value
4359		< -1 => some kind of unexpected problem
4360		*/
4361
4362	<	EXPORT int
4362	>	PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
4363		pcre_exec(const pcre argument_re, const pcre_extra extra_data,
4364	<	const char subject, int length, int start_offset, int options, int offsets,
4364	>	PCRE_SPTR subject, int length, int start_offset, int options, int *offsets,
4365		int offsetcount)
4366		{
4367		int rc, resetcount, ocount;
4368		int first_byte = -1;
4369		int req_byte = -1;
4370		int req_byte2 = -1;
4371	<	unsigned long int ims = 0;
4371	>	int newline;
4372	>	unsigned long int ims;
4373		BOOL using_temporary_offsets = FALSE;
4374		BOOL anchored;
4375		BOOL startline;
4376		BOOL firstline;
4377		BOOL first_byte_caseless = FALSE;
4378		BOOL req_byte_caseless = FALSE;
4379	+	BOOL utf8;
4380		match_data match_block;
4381	+	match_data *md = &match_block;
4382		const uschar *tables;
4383		const uschar *start_bits = NULL;
4384	<	const uschar start_match = (const uschar )subject + start_offset;
4385	<	const uschar *end_subject;
4386	<	const uschar *req_byte_ptr = start_match - 1;
4384	>	USPTR start_match = (USPTR)subject + start_offset;
4385	>	USPTR end_subject;
4386	>	USPTR req_byte_ptr = start_match - 1;
4387
4388		pcre_study_data internal_study;
4389		const pcre_study_data *study;
#	Line 2182 \| Line 4403 \| if (offsetcount < 0) return PCRE_ERROR_B
4403		the default values. */
4404
4405		study = NULL;
4406	<	match_block.match_limit = MATCH_LIMIT;
4407	<	match_block.callout_data = NULL;
4406	>	md->match_limit = MATCH_LIMIT;
4407	>	md->match_limit_recursion = MATCH_LIMIT_RECURSION;
4408	>	md->callout_data = NULL;
4409
4410		/* The table pointer is always in native byte order. */
4411
#	Line 2195 \| Line 4417 \| if (extra_data != NULL)
4417		if ((flags & PCRE_EXTRA_STUDY_DATA) != 0)
4418		study = (const pcre_study_data *)extra_data->study_data;
4419		if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0)
4420	<	match_block.match_limit = extra_data->match_limit;
4420	>	md->match_limit = extra_data->match_limit;
4421	>	if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0)
4422	>	md->match_limit_recursion = extra_data->match_limit_recursion;
4423		if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0)
4424	<	match_block.callout_data = extra_data->callout_data;
4424	>	md->callout_data = extra_data->callout_data;
4425		if ((flags & PCRE_EXTRA_TABLES) != 0) tables = extra_data->tables;
4426		}
4427
#	Line 2222 \| Line 4446 \| if (re->magic_number != MAGIC_NUMBER)
4446		/* Set up other data */
4447
4448		anchored = ((re->options \| options) & PCRE_ANCHORED) != 0;
4449	<	startline = (re->options & PCRE_STARTLINE) != 0;
4449	>	startline = (re->flags & PCRE_STARTLINE) != 0;
4450		firstline = (re->options & PCRE_FIRSTLINE) != 0;
4451
4452		/* The code starts after the real_pcre block and the capture name table. */
4453
4454	<	match_block.start_code = (const uschar *)external_re + re->name_table_offset +
4454	>	md->start_code = (const uschar *)external_re + re->name_table_offset +
4455		re->name_count * re->name_entry_size;
4456
4457	<	match_block.start_subject = (const uschar *)subject;
4458	<	match_block.start_offset = start_offset;
4459	<	match_block.end_subject = match_block.start_subject + length;
4460	<	end_subject = match_block.end_subject;
4461	<
4462	<	match_block.endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
4463	<	match_block.utf8 = (re->options & PCRE_UTF8) != 0;
4464	<
4465	<	match_block.notbol = (options & PCRE_NOTBOL) != 0;
4466	<	match_block.noteol = (options & PCRE_NOTEOL) != 0;
4467	<	match_block.notempty = (options & PCRE_NOTEMPTY) != 0;
4468	<	match_block.partial = (options & PCRE_PARTIAL) != 0;
4469	<	match_block.hitend = FALSE;
4457	>	md->start_subject = (USPTR)subject;
4458	>	md->start_offset = start_offset;
4459	>	md->end_subject = md->start_subject + length;
4460	>	end_subject = md->end_subject;
4461	>
4462	>	md->endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
4463	>	utf8 = md->utf8 = (re->options & PCRE_UTF8) != 0;
4464	>	md->jscript_compat = (re->options & PCRE_JAVASCRIPT_COMPAT) != 0;
4465	>
4466	>	md->notbol = (options & PCRE_NOTBOL) != 0;
4467	>	md->noteol = (options & PCRE_NOTEOL) != 0;
4468	>	md->notempty = (options & PCRE_NOTEMPTY) != 0;
4469	>	md->partial = (options & PCRE_PARTIAL) != 0;
4470	>	md->hitend = FALSE;
4471	>
4472	>	md->recursive = NULL; /* No recursion at top level */
4473	>
4474	>	md->lcc = tables + lcc_offset;
4475	>	md->ctypes = tables + ctypes_offset;
4476	>
4477	>	/* Handle different \R options. */
4478	>
4479	>	switch (options & (PCRE_BSR_ANYCRLF\|PCRE_BSR_UNICODE))
4480	>	{
4481	>	case 0:
4482	>	if ((re->options & (PCRE_BSR_ANYCRLF\|PCRE_BSR_UNICODE)) != 0)
4483	>	md->bsr_anycrlf = (re->options & PCRE_BSR_ANYCRLF) != 0;
4484	>	else
4485	>	#ifdef BSR_ANYCRLF
4486	>	md->bsr_anycrlf = TRUE;
4487	>	#else
4488	>	md->bsr_anycrlf = FALSE;
4489	>	#endif
4490	>	break;
4491	>
4492	>	case PCRE_BSR_ANYCRLF:
4493	>	md->bsr_anycrlf = TRUE;
4494	>	break;
4495	>
4496	>	case PCRE_BSR_UNICODE:
4497	>	md->bsr_anycrlf = FALSE;
4498	>	break;
4499
4500	<	match_block.recursive = NULL; /* No recursion at top level */
4500	>	default: return PCRE_ERROR_BADNEWLINE;
4501	>	}
4502
4503	<	match_block.lcc = tables + lcc_offset;
4504	<	match_block.ctypes = tables + ctypes_offset;
4503	>	/* Handle different types of newline. The three bits give eight cases. If
4504	>	nothing is set at run time, whatever was used at compile time applies. */
4505	>
4506	>	switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options :
4507	>	(pcre_uint32)options) & PCRE_NEWLINE_BITS)
4508	>	{
4509	>	case 0: newline = NEWLINE; break; /* Compile-time default */
4510	>	case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
4511	>	case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
4512	>	case PCRE_NEWLINE_CR+
4513	>	PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) \| CHAR_NL; break;
4514	>	case PCRE_NEWLINE_ANY: newline = -1; break;
4515	>	case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
4516	>	default: return PCRE_ERROR_BADNEWLINE;
4517	>	}
4518	>
4519	>	if (newline == -2)
4520	>	{
4521	>	md->nltype = NLTYPE_ANYCRLF;
4522	>	}
4523	>	else if (newline < 0)
4524	>	{
4525	>	md->nltype = NLTYPE_ANY;
4526	>	}
4527	>	else
4528	>	{
4529	>	md->nltype = NLTYPE_FIXED;
4530	>	if (newline > 255)
4531	>	{
4532	>	md->nllen = 2;
4533	>	md->nl[0] = (newline >> 8) & 255;
4534	>	md->nl[1] = newline & 255;
4535	>	}
4536	>	else
4537	>	{
4538	>	md->nllen = 1;
4539	>	md->nl[0] = newline;
4540	>	}
4541	>	}
4542
4543		/* Partial matching is supported only for a restricted set of regexes at the
4544		moment. */
4545
4546	<	if (match_block.partial && (re->options & PCRE_NOPARTIAL) != 0)
4546	>	if (md->partial && (re->flags & PCRE_NOPARTIAL) != 0)
4547		return PCRE_ERROR_BADPARTIAL;
4548
4549	+	/* Check a UTF-8 string if required. Unfortunately there's no way of passing
4550	+	back the character offset. */
4551	+
4552	+	#ifdef SUPPORT_UTF8
4553	+	if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0)
4554	+	{
4555	+	if (_pcre_valid_utf8((USPTR)subject, length) >= 0)
4556	+	return PCRE_ERROR_BADUTF8;
4557	+	if (start_offset > 0 && start_offset < length)
4558	+	{
4559	+	int tb = ((USPTR)subject)[start_offset];
4560	+	if (tb > 127)
4561	+	{
4562	+	tb &= 0xc0;
4563	+	if (tb != 0 && tb != 0xc0) return PCRE_ERROR_BADUTF8_OFFSET;
4564	+	}
4565	+	}
4566	+	}
4567	+	#endif
4568	+
4569		/* The ims options can vary during the matching as a result of the presence
4570		of (?ims) items in the pattern. They are kept in a local variable so that
4571		restoring at the exit of a group is easy. */
#	Line 2271 \| Line 4582 \| ocount = offsetcount - (offsetcount % 3)
4582		if (re->top_backref > 0 && re->top_backref >= ocount/3)
4583		{
4584		ocount = re->top_backref * 3 + 3;
4585	<	match_block.offset_vector = (int )(pcre_malloc)(ocount sizeof(int));
4586	<	if (match_block.offset_vector == NULL) return PCRE_ERROR_NOMEMORY;
4585	>	md->offset_vector = (int )(pcre_malloc)(ocount sizeof(int));
4586	>	if (md->offset_vector == NULL) return PCRE_ERROR_NOMEMORY;
4587		using_temporary_offsets = TRUE;
4588		DPRINTF(("Got memory to hold back references\n"));
4589		}
4590	<	else match_block.offset_vector = offsets;
4590	>	else md->offset_vector = offsets;
4591
4592	<	match_block.offset_end = ocount;
4593	<	match_block.offset_max = (2*ocount)/3;
4594	<	match_block.offset_overflow = FALSE;
4595	<	match_block.capture_last = -1;
4592	>	md->offset_end = ocount;
4593	>	md->offset_max = (2*ocount)/3;
4594	>	md->offset_overflow = FALSE;
4595	>	md->capture_last = -1;
4596
4597		/* Compute the minimum number of offsets that we need to reset each time. Doing
4598		this makes a huge difference to execution time when there aren't many brackets
#	Line 2294 \| Line 4605 \| if (resetcount > offsetcount) resetcount
4605		never be used unless previously set, but they get saved and restored, and so we
4606		initialize them to avoid reading uninitialized locations. */
4607
4608	<	if (match_block.offset_vector != NULL)
4608	>	if (md->offset_vector != NULL)
4609		{
4610	<	register int *iptr = match_block.offset_vector + ocount;
4610	>	register int *iptr = md->offset_vector + ocount;
4611		register int *iend = iptr - resetcount/2 + 1;
4612		while (--iptr >= iend) *iptr = -1;
4613		}
#	Line 2309 \| Line 4620 \| studied, there may be a bitmap of possib
4620
4621		if (!anchored)
4622		{
4623	<	if ((re->options & PCRE_FIRSTSET) != 0)
4623	>	if ((re->flags & PCRE_FIRSTSET) != 0)
4624		{
4625		first_byte = re->first_byte & 255;
4626		if ((first_byte_caseless = ((re->first_byte & REQ_CASELESS) != 0)) == TRUE)
4627	<	first_byte = match_block.lcc[first_byte];
4627	>	first_byte = md->lcc[first_byte];
4628		}
4629		else
4630		if (!startline && study != NULL &&
#	Line 2324 \| Line 4635 \| if (!anchored)
4635		/* For anchored or unanchored matches, there may be a "last known required
4636		character" set. */
4637
4638	<	if ((re->options & PCRE_REQCHSET) != 0)
4638	>	if ((re->flags & PCRE_REQCHSET) != 0)
4639		{
4640		req_byte = re->req_byte & 255;
4641		req_byte_caseless = (re->req_byte & REQ_CASELESS) != 0;
4642		req_byte2 = (tables + fcc_offset)[req_byte]; /* case flipped */
4643		}
4644
4645	+
4646	+	/* ==========================================================================*/
4647	+
4648		/* Loop for handling unanchored repeated matching attempts; for anchored regexs
4649		the loop runs just once. */
4650
4651	<	do
4651	>	for(;;)
4652		{
4653	<	const uschar *save_end_subject = end_subject;
4653	>	USPTR save_end_subject = end_subject;
4654	>	USPTR new_start_match;
4655
4656		/* Reset the maximum number of extractions we might see. */
4657
4658	<	if (match_block.offset_vector != NULL)
4658	>	if (md->offset_vector != NULL)
4659		{
4660	<	register int *iptr = match_block.offset_vector;
4660	>	register int *iptr = md->offset_vector;
4661		register int *iend = iptr + resetcount;
4662		while (iptr < iend) *iptr++ = -1;
4663		}
4664
4665	<	/* Advance to a unique first char if possible. If firstline is TRUE, the
4666	<	start of the match is constrained to the first line of a multiline string.
4667	<	Implement this by temporarily adjusting end_subject so that we stop scanning
4668	<	at a newline. If the match fails at the newline, later code breaks this loop.
4669	<	*/
4665	>	/* If firstline is TRUE, the start of the match is constrained to the first
4666	>	line of a multiline string. That is, the match must be before or at the first
4667	>	newline. Implement this by temporarily adjusting end_subject so that we stop
4668	>	scanning at a newline. If the match fails at the newline, later code breaks
4669	>	this loop. */
4670
4671		if (firstline)
4672		{
4673	<	const uschar *t = start_match;
4674	<	while (t < save_end_subject && *t != '\n') t++;
4673	>	USPTR t = start_match;
4674	>	#ifdef SUPPORT_UTF8
4675	>	if (utf8)
4676	>	{
4677	>	while (t < md->end_subject && !IS_NEWLINE(t))
4678	>	{
4679	>	t++;
4680	>	while (t < end_subject && (*t & 0xc0) == 0x80) t++;
4681	>	}
4682	>	}
4683	>	else
4684	>	#endif
4685	>	while (t < md->end_subject && !IS_NEWLINE(t)) t++;
4686		end_subject = t;
4687		}
4688
4689	<	/* Now test for a unique first byte */
4689	>	/* There are some optimizations that avoid running the match if a known
4690	>	starting point is not found, or if a known later character is not present.
4691	>	However, there is an option that disables these, for testing and for ensuring
4692	>	that all callouts do actually occur. */
4693
4694	<	if (first_byte >= 0)
4694	>	if ((options & PCRE_NO_START_OPTIMIZE) == 0)
4695		{
4696	<	if (first_byte_caseless)
4697	<	while (start_match < end_subject &&
4698	<	match_block.lcc[*start_match] != first_byte)
4699	<	start_match++;
4700	<	else
4701	<	while (start_match < end_subject && *start_match != first_byte)
4702	<	start_match++;
4703	<	}
4696	>	/* Advance to a unique first byte if there is one. */
4697	>
4698	>	if (first_byte >= 0)
4699	>	{
4700	>	if (first_byte_caseless)
4701	>	while (start_match < end_subject && md->lcc[*start_match] != first_byte)
4702	>	start_match++;
4703	>	else
4704	>	while (start_match < end_subject && *start_match != first_byte)
4705	>	start_match++;
4706	>	}
4707
4708	<	/* Or to just after \n for a multiline match if possible */
4708	>	/* Or to just after a linebreak for a multiline match */
4709
4710	<	else if (startline)
2379	<	{
2380	<	if (start_match > match_block.start_subject + start_offset)
4710	>	else if (startline)
4711		{
4712	<	while (start_match < end_subject && start_match[-1] != NEWLINE)
4713	<	start_match++;
4712	>	if (start_match > md->start_subject + start_offset)
4713	>	{
4714	>	#ifdef SUPPORT_UTF8
4715	>	if (utf8)
4716	>	{
4717	>	while (start_match < end_subject && !WAS_NEWLINE(start_match))
4718	>	{
4719	>	start_match++;
4720	>	while(start_match < end_subject && (*start_match & 0xc0) == 0x80)
4721	>	start_match++;
4722	>	}
4723	>	}
4724	>	else
4725	>	#endif
4726	>	while (start_match < end_subject && !WAS_NEWLINE(start_match))
4727	>	start_match++;
4728	>
4729	>	/* If we have just passed a CR and the newline option is ANY or ANYCRLF,
4730	>	and we are now at a LF, advance the match position by one more character.
4731	>	*/
4732	>
4733	>	if (start_match[-1] == CHAR_CR &&
4734	>	(md->nltype == NLTYPE_ANY \|\| md->nltype == NLTYPE_ANYCRLF) &&
4735	>	start_match < end_subject &&
4736	>	*start_match == CHAR_NL)
4737	>	start_match++;
4738	>	}
4739		}
2385	–	}
4740
4741	<	/* Or to a non-unique first char after study */
4741	>	/* Or to a non-unique first byte after study */
4742
4743	<	else if (start_bits != NULL)
2390	<	{
2391	<	while (start_match < end_subject)
4743	>	else if (start_bits != NULL)
4744		{
4745	<	register unsigned int c = *start_match;
4746	<	if ((start_bits[c/8] & (1 << (c&7))) == 0) start_match++; else break;
4745	>	while (start_match < end_subject)
4746	>	{
4747	>	register unsigned int c = *start_match;
4748	>	if ((start_bits[c/8] & (1 << (c&7))) == 0) start_match++;
4749	>	else break;
4750	>	}
4751		}
4752	<	}
4752	>	} /* Starting optimizations */
4753
4754		/* Restore fudged end_subject */
4755
4756		end_subject = save_end_subject;
4757
2402	–	#ifdef DEBUG /* Sigh. Some compilers never learn. */
2403	–	printf(">>>> Match against: ");
2404	–	pchars(start_match, end_subject - start_match, TRUE, &match_block);
2405	–	printf("\n");
2406	–	#endif
2407	–
2408	–	/* If req_byte is set, we know that that character must appear in the subject
2409	–	for the match to succeed. If the first character is set, req_byte must be
2410	–	later in the subject; otherwise the test starts at the match point. This
2411	–	optimization can save a huge amount of backtracking in patterns with nested
2412	–	unlimited repeats that aren't going to match. Writing separate code for
2413	–	cased/caseless versions makes it go faster, as does using an autoincrement
2414	–	and backing off on a match.
2415	–
2416	–	HOWEVER: when the subject string is very, very long, searching to its end can
2417	–	take a long time, and give bad performance on quite ordinary patterns. This
2418	–	showed up when somebody was matching /^C/ on a 32-megabyte string... so we
2419	–	don't do this when the string is sufficiently long.
4758
4759	<	ALSO: this processing is disabled when partial matching is requested.
4760	<	*/
4759	>	/* If req_byte is set, we know that that character must appear in the
4760	>	subject for the match to succeed. If the first character is set, req_byte
4761	>	must be later in the subject; otherwise the test starts at the match point.
4762	>	This optimization can save a huge amount of backtracking in patterns with
4763	>	nested unlimited repeats that aren't going to match. Writing separate code
4764	>	for cased/caseless versions makes it go faster, as does using an
4765	>	autoincrement and backing off on a match.
4766	>
4767	>	HOWEVER: when the subject string is very, very long, searching to its end
4768	>	can take a long time, and give bad performance on quite ordinary patterns.
4769	>	This showed up when somebody was matching something like /^\d+C/ on a
4770	>	32-megabyte string... so we don't do this when the string is sufficiently
4771	>	long.
4772	>
4773	>	ALSO: this processing is disabled when partial matching is requested, or if
4774	>	disabling is explicitly requested. */
4775
4776	<	if (req_byte >= 0 &&
4776	>	if ((options & PCRE_NO_START_OPTIMIZE) == 0 &&
4777	>	req_byte >= 0 &&
4778		end_subject - start_match < REQ_BYTE_MAX &&
4779	<	!match_block.partial)
4779	>	!md->partial)
4780		{
4781	<	register const uschar *p = start_match + ((first_byte >= 0)? 1 : 0);
4781	>	register USPTR p = start_match + ((first_byte >= 0)? 1 : 0);
4782
4783		/* We don't need to repeat the search if we haven't yet reached the
4784		place we found it at last time. */
#	Line 2448 \| Line 4801 \| do
4801		}
4802		}
4803
4804	<	/* If we can't find the required character, break the matching loop */
4804	>	/* If we can't find the required character, break the matching loop,
4805	>	forcing a match failure. */
4806
4807	<	if (p >= end_subject) break;
4807	>	if (p >= end_subject)
4808	>	{
4809	>	rc = MATCH_NOMATCH;
4810	>	break;
4811	>	}
4812
4813		/* If we have found the required character, save the point where we
4814		found it, so that we don't search again next time round the loop if
#	Line 2460 \| Line 4818 \| do
4818		}
4819		}
4820
4821	<	/* When a match occurs, substrings will be set for all internal extractions;
2464	<	we just need to set up the whole thing as substring 0 before returning. If
2465	<	there were too many extractions, set the return code to zero. In the case
2466	<	where we had to get some local store to hold offsets for backreferences, copy
2467	<	those back references that we can. In this case there need not be overflow
2468	<	if certain parts of the pattern were not used. */
2469	<
2470	<	match_block.start_match = start_match;
2471	<	match_block.match_call_count = 0;
2472	<
2473	<	rc = match(start_match, match_block.start_code, 2, &match_block, ims, NULL,
2474	<	match_isgroup);
2475	<
2476	<	/* When the result is no match, if the subject's first character was a
2477	<	newline and the PCRE_FIRSTLINE option is set, break (which will return
2478	<	PCRE_ERROR_NOMATCH). The option requests that a match occur before the first
2479	<	newline in the subject. Otherwise, advance the pointer to the next character
2480	<	and continue - but the continuation will actually happen only when the
2481	<	pattern is not anchored. */
4821	>	/* OK, we can now run the match. */
4822
4823	<	if (rc == MATCH_NOMATCH)
4824	<	{
4825	<	if (firstline && *start_match == NEWLINE) break;
2486	<	start_match++;
2487	<	continue;
2488	<	}
4823	>	md->start_match_ptr = start_match;
4824	>	md->match_call_count = 0;
4825	>	rc = match(start_match, md->start_code, start_match, 2, md, ims, NULL, 0, 0);
4826
4827	<	if (rc != MATCH_MATCH)
4827	>	switch(rc)
4828		{
4829	<	DPRINTF((">>>> error: returning %d\n", rc));
4830	<	return rc;
4829	>	/* NOMATCH and PRUNE advance by one character. THEN at this level acts
4830	>	exactly like PRUNE. */
4831	>
4832	>	case MATCH_NOMATCH:
4833	>	case MATCH_PRUNE:
4834	>	case MATCH_THEN:
4835	>	new_start_match = start_match + 1;
4836	>	#ifdef SUPPORT_UTF8
4837	>	if (utf8)
4838	>	while(new_start_match < end_subject && (*new_start_match & 0xc0) == 0x80)
4839	>	new_start_match++;
4840	>	#endif
4841	>	break;
4842	>
4843	>	/* SKIP passes back the next starting point explicitly. */
4844	>
4845	>	case MATCH_SKIP:
4846	>	new_start_match = md->start_match_ptr;
4847	>	break;
4848	>
4849	>	/* COMMIT disables the bumpalong, but otherwise behaves as NOMATCH. */
4850	>
4851	>	case MATCH_COMMIT:
4852	>	rc = MATCH_NOMATCH;
4853	>	goto ENDLOOP;
4854	>
4855	>	/* Any other return is some kind of error. */
4856	>
4857	>	default:
4858	>	goto ENDLOOP;
4859		}
4860
4861	<	/* We have a match! Copy the offset information from temporary store if
4862	<	necessary */
4861	>	/* Control reaches here for the various types of "no match at this point"
4862	>	result. Reset the code to MATCH_NOMATCH for subsequent checking. */
4863	>
4864	>	rc = MATCH_NOMATCH;
4865	>
4866	>	/* If PCRE_FIRSTLINE is set, the match must happen before or at the first
4867	>	newline in the subject (though it may continue over the newline). Therefore,
4868	>	if we have just failed to match, starting at a newline, do not continue. */
4869	>
4870	>	if (firstline && IS_NEWLINE(start_match)) break;
4871	>
4872	>	/* Advance to new matching position */
4873	>
4874	>	start_match = new_start_match;
4875	>
4876	>	/* Break the loop if the pattern is anchored or if we have passed the end of
4877	>	the subject. */
4878
4879	+	if (anchored \|\| start_match > end_subject) break;
4880	+
4881	+	/* If we have just passed a CR and we are now at a LF, and the pattern does
4882	+	not contain any explicit matches for \r or \n, and the newline option is CRLF
4883	+	or ANY or ANYCRLF, advance the match position by one more character. */
4884	+
4885	+	if (start_match[-1] == CHAR_CR &&
4886	+	start_match < end_subject &&
4887	+	*start_match == CHAR_NL &&
4888	+	(re->flags & PCRE_HASCRORLF) == 0 &&
4889	+	(md->nltype == NLTYPE_ANY \|\|
4890	+	md->nltype == NLTYPE_ANYCRLF \|\|
4891	+	md->nllen == 2))
4892	+	start_match++;
4893	+
4894	+	} /* End of for(;;) "bumpalong" loop */
4895	+
4896	+	/* ==========================================================================*/
4897	+
4898	+	/* We reach here when rc is not MATCH_NOMATCH, or if one of the stopping
4899	+	conditions is true:
4900	+
4901	+	(1) The pattern is anchored or the match was failed by (*COMMIT);
4902	+
4903	+	(2) We are past the end of the subject;
4904	+
4905	+	(3) PCRE_FIRSTLINE is set and we have failed to match at a newline, because
4906	+	this option requests that a match occur at or before the first newline in
4907	+	the subject.
4908	+
4909	+	When we have a match and the offset vector is big enough to deal with any
4910	+	backreferences, captured substring offsets will already be set up. In the case
4911	+	where we had to get some local store to hold offsets for backreference
4912	+	processing, copy those that we can. In this case there need not be overflow if
4913	+	certain parts of the pattern were not used, even though there are more
4914	+	capturing parentheses than vector slots. */
4915	+
4916	+	ENDLOOP:
4917	+
4918	+	if (rc == MATCH_MATCH)
4919	+	{
4920		if (using_temporary_offsets)
4921		{
4922		if (offsetcount >= 4)
4923		{
4924	<	memcpy(offsets + 2, match_block.offset_vector + 2,
4924	>	memcpy(offsets + 2, md->offset_vector + 2,
4925		(offsetcount - 2) * sizeof(int));
4926		DPRINTF(("Copied offsets from temporary memory\n"));
4927		}
4928	<	if (match_block.end_offset_top > offsetcount)
2508	<	match_block.offset_overflow = TRUE;
2509	<
4928	>	if (md->end_offset_top > offsetcount) md->offset_overflow = TRUE;
4929		DPRINTF(("Freeing temporary memory\n"));
4930	<	(pcre_free)(match_block.offset_vector);
4930	>	(pcre_free)(md->offset_vector);
4931		}
4932
4933	<	rc = match_block.offset_overflow? 0 : match_block.end_offset_top/2;
4933	>	/* Set the return code to the number of captured strings, or 0 if there are
4934	>	too many to fit into the vector. */
4935	>
4936	>	rc = md->offset_overflow? 0 : md->end_offset_top/2;
4937	>
4938	>	/* If there is space, set up the whole thing as substring 0. The value of
4939	>	md->start_match_ptr might be modified if \K was encountered on the success
4940	>	matching path. */
4941
4942		if (offsetcount < 2) rc = 0; else
4943		{
4944	<	offsets[0] = start_match - match_block.start_subject;
4945	<	offsets[1] = match_block.end_match_ptr - match_block.start_subject;
4944	>	offsets[0] = md->start_match_ptr - md->start_subject;
4945	>	offsets[1] = md->end_match_ptr - md->start_subject;
4946		}
4947
4948		DPRINTF((">>>> returning %d\n", rc));
4949		return rc;
4950		}
4951
4952	<	/* This "while" is the end of the "do" above */
4953	<
2528	<	while (!anchored && start_match <= end_subject);
4952	>	/* Control gets here if there has been an error, or if the overall match
4953	>	attempt has failed at all permitted starting positions. */
4954
4955		if (using_temporary_offsets)
4956		{
4957		DPRINTF(("Freeing temporary memory\n"));
4958	<	(pcre_free)(match_block.offset_vector);
4958	>	(pcre_free)(md->offset_vector);
4959		}
4960
4961	<	if (match_block.partial && match_block.hitend)
4961	>	if (rc != MATCH_NOMATCH)
4962	>	{
4963	>	DPRINTF((">>>> error: returning %d\n", rc));
4964	>	return rc;
4965	>	}
4966	>	else if (md->partial && md->hitend)
4967		{
4968		DPRINTF((">>>> returning PCRE_ERROR_PARTIAL\n"));
4969		return PCRE_ERROR_PARTIAL;

Diff Legend

-–
+Removed lines
-+
+Added lines
-<
+Changed lines
->
+Changed lines

Comparing ircd-hybrid-7.2/lib/pcre/pcre_exec.c (file contents): Revision 962 by lusky, Sun Oct 2 21:05:51 2005 UTC vs. Revision 963 by michael, Sat Aug 1 12:23:39 2009 UTC

Diff Legend

Comparing ircd-hybrid-7.2/lib/pcre/pcre_exec.c (file contents):
Revision 962 by lusky, Sun Oct 2 21:05:51 2005 UTC vs.
Revision 963 by michael, Sat Aug 1 12:23:39 2009 UTC