This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
stop PL_op and PL_curpad referring to different pads at start of doeval
[perl5.git] / regexec.c
CommitLineData
a0d0e21e
LW
1/* regexec.c
2 */
3
4/*
5 * "One Ring to rule them all, One Ring to find them..."
6 */
7
61296642
DM
8/* This file contains functions for executing a regular expression. See
9 * also regcomp.c which funnily enough, contains functions for compiling
166f8a29 10 * a regular expression.
e4a054ea
DM
11 *
12 * This file is also copied at build time to ext/re/re_exec.c, where
13 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
14 * This causes the main functions to be compiled under new names and with
15 * debugging support added, which makes "use re 'debug'" work.
16
166f8a29
DM
17 */
18
a687059c
LW
19/* NOTE: this is derived from Henry Spencer's regexp code, and should not
20 * confused with the original package (see point 3 below). Thanks, Henry!
21 */
22
23/* Additional note: this code is very heavily munged from Henry's version
24 * in places. In some spots I've traded clarity for efficiency, so don't
25 * blame Henry for some of the lack of readability.
26 */
27
e50aee73
AD
28/* The names of the functions have been changed from regcomp and
29 * regexec to pregcomp and pregexec in order to avoid conflicts
30 * with the POSIX routines of the same names.
31*/
32
b9d5759e
AD
33#ifdef PERL_EXT_RE_BUILD
34/* need to replace pregcomp et al, so enable that */
35# ifndef PERL_IN_XSUB_RE
36# define PERL_IN_XSUB_RE
37# endif
38/* need access to debugger hooks */
cad2e5aa 39# if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
b9d5759e
AD
40# define DEBUGGING
41# endif
42#endif
43
44#ifdef PERL_IN_XSUB_RE
d06ea78c 45/* We *really* need to overwrite these symbols: */
56953603
IZ
46# define Perl_regexec_flags my_regexec
47# define Perl_regdump my_regdump
48# define Perl_regprop my_regprop
cad2e5aa 49# define Perl_re_intuit_start my_re_intuit_start
d06ea78c
GS
50/* *These* symbols are masked to allow static link. */
51# define Perl_pregexec my_pregexec
9041c2e3 52# define Perl_reginitcolors my_reginitcolors
490a3f88 53# define Perl_regclass_swash my_regclass_swash
c5be433b
GS
54
55# define PERL_NO_GET_CONTEXT
9041c2e3 56#endif
56953603 57
a687059c 58/*
e50aee73 59 * pregcomp and pregexec -- regsub and regerror are not used in perl
a687059c
LW
60 *
61 * Copyright (c) 1986 by University of Toronto.
62 * Written by Henry Spencer. Not derived from licensed software.
63 *
64 * Permission is granted to anyone to use this software for any
65 * purpose on any computer system, and to redistribute it freely,
66 * subject to the following restrictions:
67 *
68 * 1. The author is not responsible for the consequences of use of
69 * this software, no matter how awful, even if they arise
70 * from defects in it.
71 *
72 * 2. The origin of this software must not be misrepresented, either
73 * by explicit claim or by omission.
74 *
75 * 3. Altered versions must be plainly marked as such, and must not
76 * be misrepresented as being the original software.
77 *
78 **** Alterations to Henry's code are...
79 ****
4bb101f2 80 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
b94e2f88 81 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
a687059c 82 ****
9ef589d8
LW
83 **** You may distribute under the terms of either the GNU General Public
84 **** License or the Artistic License, as specified in the README file.
a687059c
LW
85 *
86 * Beware that some of this code is subtly aware of the way operator
87 * precedence is structured in regular expressions. Serious changes in
88 * regular-expression syntax might require a total rethink.
89 */
90#include "EXTERN.h"
864dbfa3 91#define PERL_IN_REGEXEC_C
a687059c 92#include "perl.h"
0f5d15d6 93
a687059c
LW
94#include "regcomp.h"
95
c277df42
IZ
96#define RF_tainted 1 /* tainted information used? */
97#define RF_warned 2 /* warned about big count? */
ce862d02 98#define RF_evaled 4 /* Did an EVAL with setting? */
a0ed51b3
LW
99#define RF_utf8 8 /* String contains multibyte chars? */
100
eb160463 101#define UTF ((PL_reg_flags & RF_utf8) != 0)
ce862d02
IZ
102
103#define RS_init 1 /* eval environment created */
104#define RS_set 2 /* replsv value is set */
c277df42 105
a687059c
LW
106#ifndef STATIC
107#define STATIC static
108#endif
109
32fc9b6a 110#define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
7d3e948e 111
c277df42
IZ
112/*
113 * Forwards.
114 */
115
33b8afdf 116#define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
53c4c00c 117#define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
a0ed51b3 118
52657f30
AL
119#define HOPc(pos,off) ((char *)(PL_reg_match_utf8 \
120 ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \
121 : (U8*)(pos + off)))
122#define HOPBACKc(pos, off) ((char*) \
123 ((PL_reg_match_utf8) \
124 ? reghopmaybe3((U8*)pos, -off, ((U8*)(off < 0 ? PL_regeol : PL_bostr))) \
efb30f32
HS
125 : (pos - off >= PL_bostr) \
126 ? (U8*)(pos - off) \
52657f30 127 : (U8*)NULL) \
efb30f32 128)
efb30f32 129
1aa99e6b 130#define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
53c4c00c 131#define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
1aa99e6b 132#define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
1aa99e6b 133
1a4fad37
AL
134#define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
135 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END
136#define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
137#define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
138#define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
139#define LOAD_UTF8_CHARCLASS_MARK() LOAD_UTF8_CHARCLASS(mark, "\xcd\x86")
51371543 140
5f80c4cf 141/* for use after a quantifier and before an EXACT-like node -- japhy */
e2d8ce26
JP
142#define JUMPABLE(rn) ( \
143 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
cca55fe3
JP
144 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
145 OP(rn) == PLUS || OP(rn) == MINMOD || \
146 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
e2d8ce26
JP
147)
148
cca55fe3
JP
149#define HAS_TEXT(rn) ( \
150 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
151)
e2d8ce26 152
a84d97b6
HS
153/*
154 Search for mandatory following text node; for lookahead, the text must
155 follow but for lookbehind (rn->flags != 0) we skip to the next step.
156*/
cca55fe3 157#define FIND_NEXT_IMPT(rn) STMT_START { \
e2d8ce26 158 while (JUMPABLE(rn)) \
a84d97b6 159 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
e2d8ce26 160 rn = NEXTOPER(NEXTOPER(rn)); \
cca55fe3
JP
161 else if (OP(rn) == PLUS) \
162 rn = NEXTOPER(rn); \
a84d97b6
HS
163 else if (OP(rn) == IFMATCH) \
164 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
e2d8ce26 165 else rn += NEXT_OFF(rn); \
5f80c4cf 166} STMT_END
74750237 167
acfe0abc 168static void restore_pos(pTHX_ void *arg);
51371543 169
76e3520e 170STATIC CHECKPOINT
cea2e8a9 171S_regcppush(pTHX_ I32 parenfloor)
a0d0e21e 172{
97aff369 173 dVAR;
a3b680e6 174 const int retval = PL_savestack_ix;
b1ce53c5 175#define REGCP_PAREN_ELEMS 4
a3b680e6 176 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
a0d0e21e
LW
177 int p;
178
e49a9654
IH
179 if (paren_elems_to_push < 0)
180 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
181
a01268b5 182#define REGCP_OTHER_ELEMS 6
4b3c1a47 183 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
3280af22 184 for (p = PL_regsize; p > parenfloor; p--) {
b1ce53c5 185/* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
cf93c79d
IZ
186 SSPUSHINT(PL_regendp[p]);
187 SSPUSHINT(PL_regstartp[p]);
3280af22 188 SSPUSHPTR(PL_reg_start_tmp[p]);
a0d0e21e
LW
189 SSPUSHINT(p);
190 }
b1ce53c5 191/* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
3280af22
NIS
192 SSPUSHINT(PL_regsize);
193 SSPUSHINT(*PL_reglastparen);
a01268b5 194 SSPUSHINT(*PL_reglastcloseparen);
3280af22 195 SSPUSHPTR(PL_reginput);
41123dfd
JH
196#define REGCP_FRAME_ELEMS 2
197/* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
198 * are needed for the regexp context stack bookkeeping. */
199 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
b1ce53c5 200 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
41123dfd 201
a0d0e21e
LW
202 return retval;
203}
204
c277df42 205/* These are needed since we do not localize EVAL nodes: */
a3621e74 206# define REGCP_SET(cp) DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, \
faccc32b 207 " Setting an EVAL scope, savestack=%"IVdf"\n", \
02db2b7b 208 (IV)PL_savestack_ix)); cp = PL_savestack_ix
c3464db5 209
a3621e74 210# define REGCP_UNWIND(cp) DEBUG_EXECUTE_r(cp != PL_savestack_ix ? \
c3464db5 211 PerlIO_printf(Perl_debug_log, \
faccc32b 212 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
02db2b7b 213 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
c277df42 214
76e3520e 215STATIC char *
097eb12c 216S_regcppop(pTHX_ const regexp *rex)
a0d0e21e 217{
97aff369 218 dVAR;
b1ce53c5 219 I32 i;
a0d0e21e 220 char *input;
b1ce53c5 221
a3621e74
YO
222 GET_RE_DEBUG_FLAGS_DECL;
223
b1ce53c5 224 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
a0d0e21e 225 i = SSPOPINT;
b1ce53c5
JH
226 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
227 i = SSPOPINT; /* Parentheses elements to pop. */
a0d0e21e 228 input = (char *) SSPOPPTR;
a01268b5 229 *PL_reglastcloseparen = SSPOPINT;
3280af22
NIS
230 *PL_reglastparen = SSPOPINT;
231 PL_regsize = SSPOPINT;
b1ce53c5
JH
232
233 /* Now restore the parentheses context. */
41123dfd
JH
234 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
235 i > 0; i -= REGCP_PAREN_ELEMS) {
1df70142 236 I32 tmps;
097eb12c 237 U32 paren = (U32)SSPOPINT;
3280af22 238 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
cf93c79d
IZ
239 PL_regstartp[paren] = SSPOPINT;
240 tmps = SSPOPINT;
3280af22
NIS
241 if (paren <= *PL_reglastparen)
242 PL_regendp[paren] = tmps;
a3621e74 243 DEBUG_EXECUTE_r(
c3464db5 244 PerlIO_printf(Perl_debug_log,
b900a521 245 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
9041c2e3 246 (UV)paren, (IV)PL_regstartp[paren],
b900a521 247 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
9041c2e3 248 (IV)PL_regendp[paren],
3280af22 249 (paren > *PL_reglastparen ? "(no)" : ""));
c277df42 250 );
a0d0e21e 251 }
a3621e74 252 DEBUG_EXECUTE_r(
4f639d21 253 if ((I32)(*PL_reglastparen + 1) <= rex->nparens) {
c3464db5 254 PerlIO_printf(Perl_debug_log,
faccc32b 255 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
4f639d21 256 (IV)(*PL_reglastparen + 1), (IV)rex->nparens);
c277df42
IZ
257 }
258 );
daf18116 259#if 1
dafc8851
JH
260 /* It would seem that the similar code in regtry()
261 * already takes care of this, and in fact it is in
262 * a better location to since this code can #if 0-ed out
263 * but the code in regtry() is needed or otherwise tests
264 * requiring null fields (pat.t#187 and split.t#{13,14}
daf18116
JH
265 * (as of patchlevel 7877) will fail. Then again,
266 * this code seems to be necessary or otherwise
267 * building DynaLoader will fail:
268 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
269 * --jhi */
097eb12c
AL
270 for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) {
271 if (i > PL_regsize)
272 PL_regstartp[i] = -1;
273 PL_regendp[i] = -1;
a0d0e21e 274 }
dafc8851 275#endif
a0d0e21e
LW
276 return input;
277}
278
02db2b7b 279#define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
a0d0e21e 280
95b24440 281#define TRYPAREN(paren, n, input, where) { \
29d1e993
HS
282 if (paren) { \
283 if (n) { \
284 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
285 PL_regendp[paren] = input - PL_bostr; \
286 } \
287 else \
288 PL_regendp[paren] = -1; \
289 } \
95b24440
DM
290 REGMATCH(next, where); \
291 if (result) \
29d1e993
HS
292 sayYES; \
293 if (paren && n) \
294 PL_regendp[paren] = -1; \
295}
296
297
a687059c 298/*
e50aee73 299 * pregexec and friends
a687059c
LW
300 */
301
302/*
c277df42 303 - pregexec - match a regexp against a string
a687059c 304 */
c277df42 305I32
864dbfa3 306Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
c3464db5 307 char *strbeg, I32 minend, SV *screamer, U32 nosave)
c277df42
IZ
308/* strend: pointer to null at end of string */
309/* strbeg: real beginning of string */
310/* minend: end of match must be >=minend after stringarg. */
311/* nosave: For optimizations. */
312{
313 return
9041c2e3 314 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
c277df42
IZ
315 nosave ? 0 : REXEC_COPY_STR);
316}
0f5d15d6 317
22e551b9 318
9041c2e3 319/*
cad2e5aa
JH
320 * Need to implement the following flags for reg_anch:
321 *
322 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
323 * USE_INTUIT_ML
324 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
325 * INTUIT_AUTORITATIVE_ML
326 * INTUIT_ONCE_NOML - Intuit can match in one location only.
327 * INTUIT_ONCE_ML
328 *
329 * Another flag for this function: SECOND_TIME (so that float substrs
330 * with giant delta may be not rechecked).
331 */
332
333/* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
334
3f7c398e 335/* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
cad2e5aa
JH
336 Otherwise, only SvCUR(sv) is used to get strbeg. */
337
338/* XXXX We assume that strpos is strbeg unless sv. */
339
6eb5f6b9
JH
340/* XXXX Some places assume that there is a fixed substring.
341 An update may be needed if optimizer marks as "INTUITable"
342 RExen without fixed substrings. Similarly, it is assumed that
343 lengths of all the strings are no more than minlen, thus they
344 cannot come from lookahead.
345 (Or minlen should take into account lookahead.) */
346
2c2d71f5
JH
347/* A failure to find a constant substring means that there is no need to make
348 an expensive call to REx engine, thus we celebrate a failure. Similarly,
349 finding a substring too deep into the string means that less calls to
30944b6d
IZ
350 regtry() should be needed.
351
352 REx compiler's optimizer found 4 possible hints:
353 a) Anchored substring;
354 b) Fixed substring;
355 c) Whether we are anchored (beginning-of-line or \G);
356 d) First node (of those at offset 0) which may distingush positions;
6eb5f6b9 357 We use a)b)d) and multiline-part of c), and try to find a position in the
30944b6d
IZ
358 string which does not contradict any of them.
359 */
2c2d71f5 360
6eb5f6b9
JH
361/* Most of decisions we do here should have been done at compile time.
362 The nodes of the REx which we used for the search should have been
363 deleted from the finite automaton. */
364
cad2e5aa
JH
365char *
366Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
367 char *strend, U32 flags, re_scream_pos_data *data)
368{
97aff369 369 dVAR;
b7953727 370 register I32 start_shift = 0;
cad2e5aa 371 /* Should be nonnegative! */
b7953727 372 register I32 end_shift = 0;
2c2d71f5
JH
373 register char *s;
374 register SV *check;
a1933d95 375 char *strbeg;
cad2e5aa 376 char *t;
a3b680e6 377 const int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
cad2e5aa 378 I32 ml_anch;
bd61b366
SS
379 register char *other_last = NULL; /* other substr checked before this */
380 char *check_at = NULL; /* check substr found at this pos */
1df70142 381 const I32 multiline = prog->reganch & PMf_MULTILINE;
30944b6d 382#ifdef DEBUGGING
890ce7af
AL
383 const char * const i_strpos = strpos;
384 SV * const dsv = PERL_DEBUG_PAD_ZERO(0);
30944b6d 385#endif
a3621e74
YO
386
387 GET_RE_DEBUG_FLAGS_DECL;
388
a30b2f1f 389 RX_MATCH_UTF8_set(prog,do_utf8);
cad2e5aa 390
b8d68ded 391 if (prog->reganch & ROPT_UTF8) {
a3621e74 392 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
b8d68ded
JH
393 "UTF-8 regex...\n"));
394 PL_reg_flags |= RF_utf8;
395 }
396
a3621e74 397 DEBUG_EXECUTE_r({
1df70142 398 const char *s = PL_reg_match_utf8 ?
c728cb41
JH
399 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
400 strpos;
1df70142 401 const int len = PL_reg_match_utf8 ?
b8d68ded 402 strlen(s) : strend - strpos;
2a782b5b
JH
403 if (!PL_colorset)
404 reginitcolors();
b8d68ded 405 if (PL_reg_match_utf8)
a3621e74 406 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
b8d68ded 407 "UTF-8 target...\n"));
2a782b5b 408 PerlIO_printf(Perl_debug_log,
a0288114 409 "%sGuessing start of match, REx%s \"%s%.60s%s%s\" against \"%s%.*s%s%s\"...\n",
e4584336 410 PL_colors[4], PL_colors[5], PL_colors[0],
2a782b5b
JH
411 prog->precomp,
412 PL_colors[1],
413 (strlen(prog->precomp) > 60 ? "..." : ""),
414 PL_colors[0],
415 (int)(len > 60 ? 60 : len),
416 s, PL_colors[1],
417 (len > 60 ? "..." : "")
418 );
419 });
cad2e5aa 420
c344f387
JH
421 /* CHR_DIST() would be more correct here but it makes things slow. */
422 if (prog->minlen > strend - strpos) {
a3621e74 423 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
a72c7584 424 "String too short... [re_intuit_start]\n"));
cad2e5aa 425 goto fail;
2c2d71f5 426 }
a1933d95 427 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
1aa99e6b 428 PL_regeol = strend;
33b8afdf
JH
429 if (do_utf8) {
430 if (!prog->check_utf8 && prog->check_substr)
431 to_utf8_substr(prog);
432 check = prog->check_utf8;
433 } else {
434 if (!prog->check_substr && prog->check_utf8)
435 to_byte_substr(prog);
436 check = prog->check_substr;
437 }
438 if (check == &PL_sv_undef) {
a3621e74 439 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
33b8afdf
JH
440 "Non-utf string cannot match utf check string\n"));
441 goto fail;
442 }
2c2d71f5 443 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
cad2e5aa
JH
444 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
445 || ( (prog->reganch & ROPT_ANCH_BOL)
7fba1cd6 446 && !multiline ) ); /* Check after \n? */
cad2e5aa 447
7e25d62c
JH
448 if (!ml_anch) {
449 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
450 | ROPT_IMPLICIT)) /* not a real BOL */
3f7c398e 451 /* SvCUR is not set on references: SvRV and SvPVX_const overlap */
7e25d62c
JH
452 && sv && !SvROK(sv)
453 && (strpos != strbeg)) {
a3621e74 454 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
7e25d62c
JH
455 goto fail;
456 }
457 if (prog->check_offset_min == prog->check_offset_max &&
cce850e4 458 !(prog->reganch & ROPT_CANY_SEEN)) {
2c2d71f5 459 /* Substring at constant offset from beg-of-str... */
cad2e5aa
JH
460 I32 slen;
461
1aa99e6b 462 s = HOP3c(strpos, prog->check_offset_min, strend);
653099ff
GS
463 if (SvTAIL(check)) {
464 slen = SvCUR(check); /* >= 1 */
cad2e5aa 465
9041c2e3 466 if ( strend - s > slen || strend - s < slen - 1
2c2d71f5 467 || (strend - s == slen && strend[-1] != '\n')) {
a3621e74 468 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
2c2d71f5 469 goto fail_finish;
cad2e5aa
JH
470 }
471 /* Now should match s[0..slen-2] */
472 slen--;
3f7c398e 473 if (slen && (*SvPVX_const(check) != *s
cad2e5aa 474 || (slen > 1
3f7c398e 475 && memNE(SvPVX_const(check), s, slen)))) {
2c2d71f5 476 report_neq:
a3621e74 477 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
2c2d71f5
JH
478 goto fail_finish;
479 }
cad2e5aa 480 }
3f7c398e 481 else if (*SvPVX_const(check) != *s
653099ff 482 || ((slen = SvCUR(check)) > 1
3f7c398e 483 && memNE(SvPVX_const(check), s, slen)))
2c2d71f5 484 goto report_neq;
c315bfe8 485 check_at = s;
2c2d71f5 486 goto success_at_start;
7e25d62c 487 }
cad2e5aa 488 }
2c2d71f5 489 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
cad2e5aa 490 s = strpos;
2c2d71f5 491 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
2c2d71f5 492 end_shift = prog->minlen - start_shift -
653099ff 493 CHR_SVLEN(check) + (SvTAIL(check) != 0);
2c2d71f5 494 if (!ml_anch) {
a3b680e6 495 const I32 end = prog->check_offset_max + CHR_SVLEN(check)
653099ff 496 - (SvTAIL(check) != 0);
a3b680e6 497 const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
2c2d71f5
JH
498
499 if (end_shift < eshift)
500 end_shift = eshift;
501 }
cad2e5aa 502 }
2c2d71f5 503 else { /* Can match at random position */
cad2e5aa
JH
504 ml_anch = 0;
505 s = strpos;
2c2d71f5
JH
506 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
507 /* Should be nonnegative! */
508 end_shift = prog->minlen - start_shift -
653099ff 509 CHR_SVLEN(check) + (SvTAIL(check) != 0);
cad2e5aa
JH
510 }
511
2c2d71f5 512#ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
0033605d 513 if (end_shift < 0)
6bbae5e6 514 Perl_croak(aTHX_ "panic: end_shift");
2c2d71f5
JH
515#endif
516
2c2d71f5
JH
517 restart:
518 /* Find a possible match in the region s..strend by looking for
519 the "check" substring in the region corrected by start/end_shift. */
cad2e5aa 520 if (flags & REXEC_SCREAM) {
cad2e5aa 521 I32 p = -1; /* Internal iterator of scream. */
a3b680e6 522 I32 * const pp = data ? data->scream_pos : &p;
cad2e5aa 523
2c2d71f5
JH
524 if (PL_screamfirst[BmRARE(check)] >= 0
525 || ( BmRARE(check) == '\n'
526 && (BmPREVIOUS(check) == SvCUR(check) - 1)
527 && SvTAIL(check) ))
9041c2e3 528 s = screaminstr(sv, check,
2c2d71f5 529 start_shift + (s - strbeg), end_shift, pp, 0);
cad2e5aa 530 else
2c2d71f5 531 goto fail_finish;
4addbd3b
HS
532 /* we may be pointing at the wrong string */
533 if (s && RX_MATCH_COPIED(prog))
3f7c398e 534 s = strbeg + (s - SvPVX_const(sv));
cad2e5aa
JH
535 if (data)
536 *data->scream_olds = s;
537 }
f33976b4 538 else if (prog->reganch & ROPT_CANY_SEEN)
3baa4c62
JH
539 s = fbm_instr((U8*)(s + start_shift),
540 (U8*)(strend - end_shift),
7fba1cd6 541 check, multiline ? FBMrf_MULTILINE : 0);
cad2e5aa 542 else
1aa99e6b
IH
543 s = fbm_instr(HOP3(s, start_shift, strend),
544 HOP3(strend, -end_shift, strbeg),
7fba1cd6 545 check, multiline ? FBMrf_MULTILINE : 0);
cad2e5aa
JH
546
547 /* Update the count-of-usability, remove useless subpatterns,
548 unshift s. */
2c2d71f5 549
a0288114 550 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s %s substr \"%s%.*s%s\"%s%s",
2c2d71f5 551 (s ? "Found" : "Did not find"),
33b8afdf 552 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
2c2d71f5 553 PL_colors[0],
7b0972df 554 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
3f7c398e 555 SvPVX_const(check),
2c2d71f5
JH
556 PL_colors[1], (SvTAIL(check) ? "$" : ""),
557 (s ? " at offset " : "...\n") ) );
558
559 if (!s)
560 goto fail_finish;
561
6eb5f6b9
JH
562 check_at = s;
563
2c2d71f5 564 /* Finish the diagnostic message */
a3621e74 565 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
2c2d71f5
JH
566
567 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
568 Start with the other substr.
569 XXXX no SCREAM optimization yet - and a very coarse implementation
a0288114 570 XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
2c2d71f5
JH
571 *always* match. Probably should be marked during compile...
572 Probably it is right to do no SCREAM here...
573 */
574
33b8afdf 575 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
30944b6d 576 /* Take into account the "other" substring. */
2c2d71f5
JH
577 /* XXXX May be hopelessly wrong for UTF... */
578 if (!other_last)
6eb5f6b9 579 other_last = strpos;
33b8afdf 580 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
30944b6d
IZ
581 do_other_anchored:
582 {
890ce7af
AL
583 char * const last = HOP3c(s, -start_shift, strbeg);
584 char *last1, *last2;
2c2d71f5 585 char *s1 = s;
33b8afdf 586 SV* must;
2c2d71f5 587
2c2d71f5
JH
588 t = s - prog->check_offset_max;
589 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
1d86a7f9 590 && (!do_utf8
1aa99e6b 591 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
2c2d71f5 592 && t > strpos)))
30944b6d 593 /* EMPTY */;
2c2d71f5
JH
594 else
595 t = strpos;
1aa99e6b 596 t = HOP3c(t, prog->anchored_offset, strend);
6eb5f6b9
JH
597 if (t < other_last) /* These positions already checked */
598 t = other_last;
1aa99e6b 599 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
2c2d71f5
JH
600 if (last < last1)
601 last1 = last;
602 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
603 /* On end-of-str: see comment below. */
33b8afdf
JH
604 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
605 if (must == &PL_sv_undef) {
606 s = (char*)NULL;
a3621e74 607 DEBUG_EXECUTE_r(must = prog->anchored_utf8); /* for debug */
33b8afdf
JH
608 }
609 else
610 s = fbm_instr(
611 (unsigned char*)t,
612 HOP3(HOP3(last1, prog->anchored_offset, strend)
613 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
614 must,
7fba1cd6 615 multiline ? FBMrf_MULTILINE : 0
33b8afdf 616 );
a3621e74 617 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
a0288114 618 "%s anchored substr \"%s%.*s%s\"%s",
2c2d71f5
JH
619 (s ? "Found" : "Contradicts"),
620 PL_colors[0],
33b8afdf
JH
621 (int)(SvCUR(must)
622 - (SvTAIL(must)!=0)),
3f7c398e 623 SvPVX_const(must),
33b8afdf 624 PL_colors[1], (SvTAIL(must) ? "$" : "")));
2c2d71f5
JH
625 if (!s) {
626 if (last1 >= last2) {
a3621e74 627 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2c2d71f5
JH
628 ", giving up...\n"));
629 goto fail_finish;
630 }
a3621e74 631 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2c2d71f5 632 ", trying floating at offset %ld...\n",
1aa99e6b
IH
633 (long)(HOP3c(s1, 1, strend) - i_strpos)));
634 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
635 s = HOP3c(last, 1, strend);
2c2d71f5
JH
636 goto restart;
637 }
638 else {
a3621e74 639 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
30944b6d 640 (long)(s - i_strpos)));
1aa99e6b
IH
641 t = HOP3c(s, -prog->anchored_offset, strbeg);
642 other_last = HOP3c(s, 1, strend);
30944b6d 643 s = s1;
2c2d71f5
JH
644 if (t == strpos)
645 goto try_at_start;
2c2d71f5
JH
646 goto try_at_offset;
647 }
30944b6d 648 }
2c2d71f5
JH
649 }
650 else { /* Take into account the floating substring. */
33b8afdf
JH
651 char *last, *last1;
652 char *s1 = s;
653 SV* must;
654
655 t = HOP3c(s, -start_shift, strbeg);
656 last1 = last =
657 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
658 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
659 last = HOP3c(t, prog->float_max_offset, strend);
660 s = HOP3c(t, prog->float_min_offset, strend);
661 if (s < other_last)
662 s = other_last;
2c2d71f5 663 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
33b8afdf
JH
664 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
665 /* fbm_instr() takes into account exact value of end-of-str
666 if the check is SvTAIL(ed). Since false positives are OK,
667 and end-of-str is not later than strend we are OK. */
668 if (must == &PL_sv_undef) {
669 s = (char*)NULL;
a3621e74 670 DEBUG_EXECUTE_r(must = prog->float_utf8); /* for debug message */
33b8afdf
JH
671 }
672 else
2c2d71f5 673 s = fbm_instr((unsigned char*)s,
33b8afdf
JH
674 (unsigned char*)last + SvCUR(must)
675 - (SvTAIL(must)!=0),
7fba1cd6 676 must, multiline ? FBMrf_MULTILINE : 0);
a0288114 677 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s floating substr \"%s%.*s%s\"%s",
33b8afdf
JH
678 (s ? "Found" : "Contradicts"),
679 PL_colors[0],
680 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
3f7c398e 681 SvPVX_const(must),
33b8afdf
JH
682 PL_colors[1], (SvTAIL(must) ? "$" : "")));
683 if (!s) {
684 if (last1 == last) {
a3621e74 685 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
33b8afdf
JH
686 ", giving up...\n"));
687 goto fail_finish;
2c2d71f5 688 }
a3621e74 689 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
33b8afdf
JH
690 ", trying anchored starting at offset %ld...\n",
691 (long)(s1 + 1 - i_strpos)));
692 other_last = last;
693 s = HOP3c(t, 1, strend);
694 goto restart;
695 }
696 else {
a3621e74 697 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
33b8afdf
JH
698 (long)(s - i_strpos)));
699 other_last = s; /* Fix this later. --Hugo */
700 s = s1;
701 if (t == strpos)
702 goto try_at_start;
703 goto try_at_offset;
704 }
2c2d71f5 705 }
cad2e5aa 706 }
2c2d71f5
JH
707
708 t = s - prog->check_offset_max;
2c2d71f5 709 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
1d86a7f9 710 && (!do_utf8
1aa99e6b
IH
711 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
712 && t > strpos))) {
2c2d71f5
JH
713 /* Fixed substring is found far enough so that the match
714 cannot start at strpos. */
715 try_at_offset:
cad2e5aa 716 if (ml_anch && t[-1] != '\n') {
30944b6d
IZ
717 /* Eventually fbm_*() should handle this, but often
718 anchored_offset is not 0, so this check will not be wasted. */
719 /* XXXX In the code below we prefer to look for "^" even in
720 presence of anchored substrings. And we search even
721 beyond the found float position. These pessimizations
722 are historical artefacts only. */
723 find_anchor:
2c2d71f5 724 while (t < strend - prog->minlen) {
cad2e5aa 725 if (*t == '\n') {
4ee3650e 726 if (t < check_at - prog->check_offset_min) {
33b8afdf 727 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
4ee3650e
GS
728 /* Since we moved from the found position,
729 we definitely contradict the found anchored
30944b6d
IZ
730 substr. Due to the above check we do not
731 contradict "check" substr.
732 Thus we can arrive here only if check substr
733 is float. Redo checking for "other"=="fixed".
734 */
9041c2e3 735 strpos = t + 1;
a3621e74 736 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
e4584336 737 PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
30944b6d
IZ
738 goto do_other_anchored;
739 }
4ee3650e
GS
740 /* We don't contradict the found floating substring. */
741 /* XXXX Why not check for STCLASS? */
cad2e5aa 742 s = t + 1;
a3621e74 743 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
e4584336 744 PL_colors[0], PL_colors[1], (long)(s - i_strpos)));
cad2e5aa
JH
745 goto set_useful;
746 }
4ee3650e
GS
747 /* Position contradicts check-string */
748 /* XXXX probably better to look for check-string
749 than for "\n", so one should lower the limit for t? */
a3621e74 750 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
e4584336 751 PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos)));
0e41cd87 752 other_last = strpos = s = t + 1;
cad2e5aa
JH
753 goto restart;
754 }
755 t++;
756 }
a3621e74 757 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
e4584336 758 PL_colors[0], PL_colors[1]));
2c2d71f5 759 goto fail_finish;
cad2e5aa 760 }
f5952150 761 else {
a3621e74 762 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
e4584336 763 PL_colors[0], PL_colors[1]));
f5952150 764 }
cad2e5aa
JH
765 s = t;
766 set_useful:
33b8afdf 767 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
cad2e5aa
JH
768 }
769 else {
f5952150 770 /* The found string does not prohibit matching at strpos,
2c2d71f5 771 - no optimization of calling REx engine can be performed,
f5952150
GS
772 unless it was an MBOL and we are not after MBOL,
773 or a future STCLASS check will fail this. */
2c2d71f5
JH
774 try_at_start:
775 /* Even in this situation we may use MBOL flag if strpos is offset
776 wrt the start of the string. */
05b4157f 777 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
a1933d95 778 && (strpos != strbeg) && strpos[-1] != '\n'
d506a20d
IZ
779 /* May be due to an implicit anchor of m{.*foo} */
780 && !(prog->reganch & ROPT_IMPLICIT))
781 {
cad2e5aa
JH
782 t = strpos;
783 goto find_anchor;
784 }
a3621e74 785 DEBUG_EXECUTE_r( if (ml_anch)
f5952150 786 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
e4584336 787 (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]);
30944b6d 788 );
2c2d71f5 789 success_at_start:
30944b6d 790 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
33b8afdf
JH
791 && (do_utf8 ? (
792 prog->check_utf8 /* Could be deleted already */
793 && --BmUSEFUL(prog->check_utf8) < 0
794 && (prog->check_utf8 == prog->float_utf8)
795 ) : (
796 prog->check_substr /* Could be deleted already */
797 && --BmUSEFUL(prog->check_substr) < 0
798 && (prog->check_substr == prog->float_substr)
799 )))
66e933ab 800 {
cad2e5aa 801 /* If flags & SOMETHING - do not do it many times on the same match */
a3621e74 802 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
33b8afdf
JH
803 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
804 if (do_utf8 ? prog->check_substr : prog->check_utf8)
805 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
a0714e2c
SS
806 prog->check_substr = prog->check_utf8 = NULL; /* disable */
807 prog->float_substr = prog->float_utf8 = NULL; /* clear */
808 check = NULL; /* abort */
cad2e5aa 809 s = strpos;
3cf5c195
IZ
810 /* XXXX This is a remnant of the old implementation. It
811 looks wasteful, since now INTUIT can use many
6eb5f6b9 812 other heuristics. */
cad2e5aa
JH
813 prog->reganch &= ~RE_USE_INTUIT;
814 }
815 else
816 s = strpos;
817 }
818
6eb5f6b9
JH
819 /* Last resort... */
820 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
821 if (prog->regstclass) {
822 /* minlen == 0 is possible if regstclass is \b or \B,
823 and the fixed substr is ''$.
824 Since minlen is already taken into account, s+1 is before strend;
825 accidentally, minlen >= 1 guaranties no false positives at s + 1
826 even for \b or \B. But (minlen? 1 : 0) below assumes that
827 regstclass does not come from lookahead... */
828 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
829 This leaves EXACTF only, which is dealt with in find_byclass(). */
890ce7af 830 const U8* const str = (U8*)STRING(prog->regstclass);
06b5626a 831 const int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
1aa99e6b 832 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
66e933ab 833 : 1);
a3b680e6 834 const char * const endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
1aa99e6b 835 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
33b8afdf 836 : (prog->float_substr || prog->float_utf8
1aa99e6b
IH
837 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
838 cl_l, strend)
839 : strend);
6eb5f6b9
JH
840
841 t = s;
3b0527fe 842 s = find_byclass(prog, prog->regstclass, s, endpos, NULL);
6eb5f6b9
JH
843 if (!s) {
844#ifdef DEBUGGING
cbbf8932 845 const char *what = NULL;
6eb5f6b9
JH
846#endif
847 if (endpos == strend) {
a3621e74 848 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
6eb5f6b9
JH
849 "Could not match STCLASS...\n") );
850 goto fail;
851 }
a3621e74 852 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
66e933ab 853 "This position contradicts STCLASS...\n") );
653099ff
GS
854 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
855 goto fail;
6eb5f6b9 856 /* Contradict one of substrings */
33b8afdf
JH
857 if (prog->anchored_substr || prog->anchored_utf8) {
858 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
a3621e74 859 DEBUG_EXECUTE_r( what = "anchored" );
6eb5f6b9 860 hop_and_restart:
1aa99e6b 861 s = HOP3c(t, 1, strend);
66e933ab
GS
862 if (s + start_shift + end_shift > strend) {
863 /* XXXX Should be taken into account earlier? */
a3621e74 864 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
66e933ab
GS
865 "Could not match STCLASS...\n") );
866 goto fail;
867 }
5e39e1e5
HS
868 if (!check)
869 goto giveup;
a3621e74 870 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
f5952150 871 "Looking for %s substr starting at offset %ld...\n",
6eb5f6b9
JH
872 what, (long)(s + start_shift - i_strpos)) );
873 goto restart;
874 }
66e933ab 875 /* Have both, check_string is floating */
6eb5f6b9
JH
876 if (t + start_shift >= check_at) /* Contradicts floating=check */
877 goto retry_floating_check;
878 /* Recheck anchored substring, but not floating... */
9041c2e3 879 s = check_at;
5e39e1e5
HS
880 if (!check)
881 goto giveup;
a3621e74 882 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
f5952150 883 "Looking for anchored substr starting at offset %ld...\n",
6eb5f6b9
JH
884 (long)(other_last - i_strpos)) );
885 goto do_other_anchored;
886 }
60e71179
GS
887 /* Another way we could have checked stclass at the
888 current position only: */
889 if (ml_anch) {
890 s = t = t + 1;
5e39e1e5
HS
891 if (!check)
892 goto giveup;
a3621e74 893 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
f5952150 894 "Looking for /%s^%s/m starting at offset %ld...\n",
e4584336 895 PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
60e71179 896 goto try_at_offset;
66e933ab 897 }
33b8afdf 898 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
60e71179 899 goto fail;
6eb5f6b9
JH
900 /* Check is floating subtring. */
901 retry_floating_check:
902 t = check_at - start_shift;
a3621e74 903 DEBUG_EXECUTE_r( what = "floating" );
6eb5f6b9
JH
904 goto hop_and_restart;
905 }
b7953727 906 if (t != s) {
a3621e74 907 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
6eb5f6b9 908 "By STCLASS: moving %ld --> %ld\n",
b7953727
JH
909 (long)(t - i_strpos), (long)(s - i_strpos))
910 );
911 }
912 else {
a3621e74 913 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
b7953727
JH
914 "Does not contradict STCLASS...\n");
915 );
916 }
6eb5f6b9 917 }
5e39e1e5 918 giveup:
a3621e74 919 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
5e39e1e5
HS
920 PL_colors[4], (check ? "Guessed" : "Giving up"),
921 PL_colors[5], (long)(s - i_strpos)) );
cad2e5aa 922 return s;
2c2d71f5
JH
923
924 fail_finish: /* Substring not found */
33b8afdf
JH
925 if (prog->check_substr || prog->check_utf8) /* could be removed already */
926 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
cad2e5aa 927 fail:
a3621e74 928 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
e4584336 929 PL_colors[4], PL_colors[5]));
bd61b366 930 return NULL;
cad2e5aa 931}
9661b544 932
6eb5f6b9 933/* We know what class REx starts with. Try to find this position... */
3b0527fe
DM
934/* if reginfo is NULL, its a dryrun */
935
3c3eec57 936STATIC char *
3b0527fe
DM
937S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, const char
938*strend, const regmatch_info *reginfo)
a687059c 939{
27da23d5 940 dVAR;
1df70142 941 const I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
6eb5f6b9 942 char *m;
d8093b23 943 STRLEN ln;
5dab1207 944 STRLEN lnc;
078c425b 945 register STRLEN uskip;
d8093b23
G
946 unsigned int c1;
947 unsigned int c2;
6eb5f6b9
JH
948 char *e;
949 register I32 tmp = 1; /* Scratch variable? */
a3b680e6 950 register const bool do_utf8 = PL_reg_match_utf8;
cad2e5aa 951
6eb5f6b9
JH
952 /* We know what class it must start with. */
953 switch (OP(c)) {
6eb5f6b9 954 case ANYOF:
388cc4de 955 if (do_utf8) {
078c425b 956 while (s + (uskip = UTF8SKIP(s)) <= strend) {
388cc4de
HS
957 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
958 !UTF8_IS_INVARIANT((U8)s[0]) ?
32fc9b6a
DM
959 reginclass(prog, c, (U8*)s, 0, do_utf8) :
960 REGINCLASS(prog, c, (U8*)s)) {
3b0527fe 961 if (tmp && (!reginfo || regtry(reginfo, s)))
388cc4de
HS
962 goto got_it;
963 else
964 tmp = doevery;
965 }
966 else
967 tmp = 1;
078c425b 968 s += uskip;
388cc4de
HS
969 }
970 }
971 else {
972 while (s < strend) {
973 STRLEN skip = 1;
974
32fc9b6a 975 if (REGINCLASS(prog, c, (U8*)s) ||
388cc4de
HS
976 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
977 /* The assignment of 2 is intentional:
978 * for the folded sharp s, the skip is 2. */
979 (skip = SHARP_S_SKIP))) {
3b0527fe 980 if (tmp && (!reginfo || regtry(reginfo, s)))
388cc4de
HS
981 goto got_it;
982 else
983 tmp = doevery;
984 }
985 else
986 tmp = 1;
987 s += skip;
988 }
a0d0e21e 989 }
6eb5f6b9 990 break;
f33976b4
DB
991 case CANY:
992 while (s < strend) {
3b0527fe 993 if (tmp && (!reginfo || regtry(reginfo, s)))
f33976b4
DB
994 goto got_it;
995 else
996 tmp = doevery;
997 s++;
998 }
999 break;
6eb5f6b9 1000 case EXACTF:
5dab1207
NIS
1001 m = STRING(c);
1002 ln = STR_LEN(c); /* length to match in octets/bytes */
1003 lnc = (I32) ln; /* length to match in characters */
1aa99e6b 1004 if (UTF) {
a2a2844f 1005 STRLEN ulen1, ulen2;
5dab1207 1006 U8 *sm = (U8 *) m;
89ebb4a3
JH
1007 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
1008 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
4ad0818d 1009 const U32 uniflags = UTF8_ALLOW_DEFAULT;
a2a2844f
JH
1010
1011 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1012 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1013
89ebb4a3 1014 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
041457d9 1015 0, uniflags);
89ebb4a3 1016 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
041457d9 1017 0, uniflags);
5dab1207
NIS
1018 lnc = 0;
1019 while (sm < ((U8 *) m + ln)) {
1020 lnc++;
1021 sm += UTF8SKIP(sm);
1022 }
1aa99e6b
IH
1023 }
1024 else {
1025 c1 = *(U8*)m;
1026 c2 = PL_fold[c1];
1027 }
6eb5f6b9
JH
1028 goto do_exactf;
1029 case EXACTFL:
5dab1207
NIS
1030 m = STRING(c);
1031 ln = STR_LEN(c);
1032 lnc = (I32) ln;
d8093b23 1033 c1 = *(U8*)m;
6eb5f6b9
JH
1034 c2 = PL_fold_locale[c1];
1035 do_exactf:
db12adc6 1036 e = HOP3c(strend, -((I32)lnc), s);
b3c9acc1 1037
3b0527fe 1038 if (!reginfo && e < s)
6eb5f6b9 1039 e = s; /* Due to minlen logic of intuit() */
1aa99e6b 1040
60a8b682
JH
1041 /* The idea in the EXACTF* cases is to first find the
1042 * first character of the EXACTF* node and then, if
1043 * necessary, case-insensitively compare the full
1044 * text of the node. The c1 and c2 are the first
1045 * characters (though in Unicode it gets a bit
1046 * more complicated because there are more cases
7f16dd3d
JH
1047 * than just upper and lower: one needs to use
1048 * the so-called folding case for case-insensitive
1049 * matching (called "loose matching" in Unicode).
1050 * ibcmp_utf8() will do just that. */
60a8b682 1051
1aa99e6b 1052 if (do_utf8) {
575cac57 1053 UV c, f;
89ebb4a3 1054 U8 tmpbuf [UTF8_MAXBYTES+1];
575cac57 1055 STRLEN len, foldlen;
4ad0818d 1056 const U32 uniflags = UTF8_ALLOW_DEFAULT;
09091399 1057 if (c1 == c2) {
5dab1207
NIS
1058 /* Upper and lower of 1st char are equal -
1059 * probably not a "letter". */
1aa99e6b 1060 while (s <= e) {
89ebb4a3 1061 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
041457d9 1062 uniflags);
80aecb99
JH
1063 if ( c == c1
1064 && (ln == len ||
66423254 1065 ibcmp_utf8(s, (char **)0, 0, do_utf8,
eb160463 1066 m, (char **)0, ln, (bool)UTF))
3b0527fe 1067 && (!reginfo || regtry(reginfo, s)) )
1aa99e6b 1068 goto got_it;
80aecb99 1069 else {
1df70142 1070 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
80aecb99 1071 uvchr_to_utf8(tmpbuf, c);
ac7e0132 1072 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
80aecb99
JH
1073 if ( f != c
1074 && (f == c1 || f == c2)
1075 && (ln == foldlen ||
66423254
JH
1076 !ibcmp_utf8((char *) foldbuf,
1077 (char **)0, foldlen, do_utf8,
d07ddd77 1078 m,
eb160463 1079 (char **)0, ln, (bool)UTF))
3b0527fe 1080 && (!reginfo || regtry(reginfo, s)) )
80aecb99
JH
1081 goto got_it;
1082 }
1aa99e6b
IH
1083 s += len;
1084 }
09091399
JH
1085 }
1086 else {
1aa99e6b 1087 while (s <= e) {
89ebb4a3 1088 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
041457d9 1089 uniflags);
80aecb99 1090
60a8b682 1091 /* Handle some of the three Greek sigmas cases.
8c01da3c
JH
1092 * Note that not all the possible combinations
1093 * are handled here: some of them are handled
1094 * by the standard folding rules, and some of
1095 * them (the character class or ANYOF cases)
1096 * are handled during compiletime in
1097 * regexec.c:S_regclass(). */
880bd946
JH
1098 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1099 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1100 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
80aecb99
JH
1101
1102 if ( (c == c1 || c == c2)
1103 && (ln == len ||
66423254 1104 ibcmp_utf8(s, (char **)0, 0, do_utf8,
eb160463 1105 m, (char **)0, ln, (bool)UTF))
3b0527fe 1106 && (!reginfo || regtry(reginfo, s)) )
1aa99e6b 1107 goto got_it;
80aecb99 1108 else {
1df70142 1109 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
80aecb99 1110 uvchr_to_utf8(tmpbuf, c);
ac7e0132 1111 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
80aecb99
JH
1112 if ( f != c
1113 && (f == c1 || f == c2)
1114 && (ln == foldlen ||
a6872d42 1115 !ibcmp_utf8((char *) foldbuf,
66423254 1116 (char **)0, foldlen, do_utf8,
d07ddd77 1117 m,
eb160463 1118 (char **)0, ln, (bool)UTF))
3b0527fe 1119 && (!reginfo || regtry(reginfo, s)) )
80aecb99
JH
1120 goto got_it;
1121 }
1aa99e6b
IH
1122 s += len;
1123 }
09091399 1124 }
1aa99e6b
IH
1125 }
1126 else {
1127 if (c1 == c2)
1128 while (s <= e) {
1129 if ( *(U8*)s == c1
1130 && (ln == 1 || !(OP(c) == EXACTF
1131 ? ibcmp(s, m, ln)
1132 : ibcmp_locale(s, m, ln)))
3b0527fe 1133 && (!reginfo || regtry(reginfo, s)) )
1aa99e6b
IH
1134 goto got_it;
1135 s++;
1136 }
1137 else
1138 while (s <= e) {
1139 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1140 && (ln == 1 || !(OP(c) == EXACTF
1141 ? ibcmp(s, m, ln)
1142 : ibcmp_locale(s, m, ln)))
3b0527fe 1143 && (!reginfo || regtry(reginfo, s)) )
1aa99e6b
IH
1144 goto got_it;
1145 s++;
1146 }
b3c9acc1
IZ
1147 }
1148 break;
bbce6d69 1149 case BOUNDL:
3280af22 1150 PL_reg_flags |= RF_tainted;
bbce6d69 1151 /* FALL THROUGH */
a0d0e21e 1152 case BOUND:
ffc61ed2 1153 if (do_utf8) {
12d33761 1154 if (s == PL_bostr)
ffc61ed2
JH
1155 tmp = '\n';
1156 else {
6136c704 1157 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
4ad0818d 1158 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
ffc61ed2
JH
1159 }
1160 tmp = ((OP(c) == BOUND ?
9041c2e3 1161 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1a4fad37 1162 LOAD_UTF8_CHARCLASS_ALNUM();
078c425b 1163 while (s + (uskip = UTF8SKIP(s)) <= strend) {
ffc61ed2 1164 if (tmp == !(OP(c) == BOUND ?
3568d838 1165 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
ffc61ed2
JH
1166 isALNUM_LC_utf8((U8*)s)))
1167 {
1168 tmp = !tmp;
3b0527fe 1169 if ((!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1170 goto got_it;
1171 }
078c425b 1172 s += uskip;
a687059c 1173 }
a0d0e21e 1174 }
667bb95a 1175 else {
12d33761 1176 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
ffc61ed2
JH
1177 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1178 while (s < strend) {
1179 if (tmp ==
1180 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1181 tmp = !tmp;
3b0527fe 1182 if ((!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1183 goto got_it;
1184 }
1185 s++;
a0ed51b3 1186 }
a0ed51b3 1187 }
3b0527fe 1188 if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, s)))
a0ed51b3
LW
1189 goto got_it;
1190 break;
bbce6d69 1191 case NBOUNDL:
3280af22 1192 PL_reg_flags |= RF_tainted;
bbce6d69 1193 /* FALL THROUGH */
a0d0e21e 1194 case NBOUND:
ffc61ed2 1195 if (do_utf8) {
12d33761 1196 if (s == PL_bostr)
ffc61ed2
JH
1197 tmp = '\n';
1198 else {
6136c704 1199 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
4ad0818d 1200 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
ffc61ed2
JH
1201 }
1202 tmp = ((OP(c) == NBOUND ?
9041c2e3 1203 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1a4fad37 1204 LOAD_UTF8_CHARCLASS_ALNUM();
078c425b 1205 while (s + (uskip = UTF8SKIP(s)) <= strend) {
ffc61ed2 1206 if (tmp == !(OP(c) == NBOUND ?
3568d838 1207 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
ffc61ed2
JH
1208 isALNUM_LC_utf8((U8*)s)))
1209 tmp = !tmp;
3b0527fe 1210 else if ((!reginfo || regtry(reginfo, s)))
ffc61ed2 1211 goto got_it;
078c425b 1212 s += uskip;
ffc61ed2 1213 }
a0d0e21e 1214 }
667bb95a 1215 else {
12d33761 1216 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
ffc61ed2
JH
1217 tmp = ((OP(c) == NBOUND ?
1218 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1219 while (s < strend) {
1220 if (tmp ==
1221 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1222 tmp = !tmp;
3b0527fe 1223 else if ((!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1224 goto got_it;
1225 s++;
1226 }
a0ed51b3 1227 }
3b0527fe 1228 if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, s)))
a0ed51b3
LW
1229 goto got_it;
1230 break;
a0d0e21e 1231 case ALNUM:
ffc61ed2 1232 if (do_utf8) {
1a4fad37 1233 LOAD_UTF8_CHARCLASS_ALNUM();
078c425b 1234 while (s + (uskip = UTF8SKIP(s)) <= strend) {
3568d838 1235 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
3b0527fe 1236 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1237 goto got_it;
1238 else
1239 tmp = doevery;
1240 }
bbce6d69 1241 else
ffc61ed2 1242 tmp = 1;
078c425b 1243 s += uskip;
bbce6d69 1244 }
bbce6d69 1245 }
ffc61ed2
JH
1246 else {
1247 while (s < strend) {
1248 if (isALNUM(*s)) {
3b0527fe 1249 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1250 goto got_it;
1251 else
1252 tmp = doevery;
1253 }
a0ed51b3 1254 else
ffc61ed2
JH
1255 tmp = 1;
1256 s++;
a0ed51b3 1257 }
a0ed51b3
LW
1258 }
1259 break;
bbce6d69 1260 case ALNUML:
3280af22 1261 PL_reg_flags |= RF_tainted;
ffc61ed2 1262 if (do_utf8) {
078c425b 1263 while (s + (uskip = UTF8SKIP(s)) <= strend) {
ffc61ed2 1264 if (isALNUM_LC_utf8((U8*)s)) {
3b0527fe 1265 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1266 goto got_it;
1267 else
1268 tmp = doevery;
1269 }
a687059c 1270 else
ffc61ed2 1271 tmp = 1;
078c425b 1272 s += uskip;
a0d0e21e 1273 }
a0d0e21e 1274 }
ffc61ed2
JH
1275 else {
1276 while (s < strend) {
1277 if (isALNUM_LC(*s)) {
3b0527fe 1278 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1279 goto got_it;
1280 else
1281 tmp = doevery;
1282 }
a0ed51b3 1283 else
ffc61ed2
JH
1284 tmp = 1;
1285 s++;
a0ed51b3 1286 }
a0ed51b3
LW
1287 }
1288 break;
a0d0e21e 1289 case NALNUM:
ffc61ed2 1290 if (do_utf8) {
1a4fad37 1291 LOAD_UTF8_CHARCLASS_ALNUM();
078c425b 1292 while (s + (uskip = UTF8SKIP(s)) <= strend) {
3568d838 1293 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
3b0527fe 1294 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1295 goto got_it;
1296 else
1297 tmp = doevery;
1298 }
bbce6d69 1299 else
ffc61ed2 1300 tmp = 1;
078c425b 1301 s += uskip;
bbce6d69 1302 }
bbce6d69 1303 }
ffc61ed2
JH
1304 else {
1305 while (s < strend) {
1306 if (!isALNUM(*s)) {
3b0527fe 1307 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1308 goto got_it;
1309 else
1310 tmp = doevery;
1311 }
a0ed51b3 1312 else
ffc61ed2
JH
1313 tmp = 1;
1314 s++;
a0ed51b3 1315 }
a0ed51b3
LW
1316 }
1317 break;
bbce6d69 1318 case NALNUML:
3280af22 1319 PL_reg_flags |= RF_tainted;
ffc61ed2 1320 if (do_utf8) {
078c425b 1321 while (s + (uskip = UTF8SKIP(s)) <= strend) {
ffc61ed2 1322 if (!isALNUM_LC_utf8((U8*)s)) {
3b0527fe 1323 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1324 goto got_it;
1325 else
1326 tmp = doevery;
1327 }
a687059c 1328 else
ffc61ed2 1329 tmp = 1;
078c425b 1330 s += uskip;
a687059c 1331 }
a0d0e21e 1332 }
ffc61ed2
JH
1333 else {
1334 while (s < strend) {
1335 if (!isALNUM_LC(*s)) {
3b0527fe 1336 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1337 goto got_it;
1338 else
1339 tmp = doevery;
1340 }
a0ed51b3 1341 else
ffc61ed2
JH
1342 tmp = 1;
1343 s++;
a0ed51b3 1344 }
a0ed51b3
LW
1345 }
1346 break;
a0d0e21e 1347 case SPACE:
ffc61ed2 1348 if (do_utf8) {
1a4fad37 1349 LOAD_UTF8_CHARCLASS_SPACE();
078c425b 1350 while (s + (uskip = UTF8SKIP(s)) <= strend) {
3568d838 1351 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
3b0527fe 1352 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1353 goto got_it;
1354 else
1355 tmp = doevery;
1356 }
a0d0e21e 1357 else
ffc61ed2 1358 tmp = 1;
078c425b 1359 s += uskip;
2304df62 1360 }
a0d0e21e 1361 }
ffc61ed2
JH
1362 else {
1363 while (s < strend) {
1364 if (isSPACE(*s)) {
3b0527fe 1365 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1366 goto got_it;
1367 else
1368 tmp = doevery;
1369 }
a0ed51b3 1370 else
ffc61ed2
JH
1371 tmp = 1;
1372 s++;
a0ed51b3 1373 }
a0ed51b3
LW
1374 }
1375 break;
bbce6d69 1376 case SPACEL:
3280af22 1377 PL_reg_flags |= RF_tainted;
ffc61ed2 1378 if (do_utf8) {
078c425b 1379 while (s + (uskip = UTF8SKIP(s)) <= strend) {
ffc61ed2 1380 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
3b0527fe 1381 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1382 goto got_it;
1383 else
1384 tmp = doevery;
1385 }
bbce6d69 1386 else
ffc61ed2 1387 tmp = 1;
078c425b 1388 s += uskip;
bbce6d69 1389 }
bbce6d69 1390 }
ffc61ed2
JH
1391 else {
1392 while (s < strend) {
1393 if (isSPACE_LC(*s)) {
3b0527fe 1394 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1395 goto got_it;
1396 else
1397 tmp = doevery;
1398 }
a0ed51b3 1399 else
ffc61ed2
JH
1400 tmp = 1;
1401 s++;
a0ed51b3 1402 }
a0ed51b3
LW
1403 }
1404 break;
a0d0e21e 1405 case NSPACE:
ffc61ed2 1406 if (do_utf8) {
1a4fad37 1407 LOAD_UTF8_CHARCLASS_SPACE();
078c425b 1408 while (s + (uskip = UTF8SKIP(s)) <= strend) {
3568d838 1409 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
3b0527fe 1410 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1411 goto got_it;
1412 else
1413 tmp = doevery;
1414 }
a0d0e21e 1415 else
ffc61ed2 1416 tmp = 1;
078c425b 1417 s += uskip;
a687059c 1418 }
a0d0e21e 1419 }
ffc61ed2
JH
1420 else {
1421 while (s < strend) {
1422 if (!isSPACE(*s)) {
3b0527fe 1423 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1424 goto got_it;
1425 else
1426 tmp = doevery;
1427 }
a0ed51b3 1428 else
ffc61ed2
JH
1429 tmp = 1;
1430 s++;
a0ed51b3 1431 }
a0ed51b3
LW
1432 }
1433 break;
bbce6d69 1434 case NSPACEL:
3280af22 1435 PL_reg_flags |= RF_tainted;
ffc61ed2 1436 if (do_utf8) {
078c425b 1437 while (s + (uskip = UTF8SKIP(s)) <= strend) {
ffc61ed2 1438 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
3b0527fe 1439 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1440 goto got_it;
1441 else
1442 tmp = doevery;
1443 }
bbce6d69 1444 else
ffc61ed2 1445 tmp = 1;
078c425b 1446 s += uskip;
bbce6d69 1447 }
bbce6d69 1448 }
ffc61ed2
JH
1449 else {
1450 while (s < strend) {
1451 if (!isSPACE_LC(*s)) {
3b0527fe 1452 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1453 goto got_it;
1454 else
1455 tmp = doevery;
1456 }
a0ed51b3 1457 else
ffc61ed2
JH
1458 tmp = 1;
1459 s++;
a0ed51b3 1460 }
a0ed51b3
LW
1461 }
1462 break;
a0d0e21e 1463 case DIGIT:
ffc61ed2 1464 if (do_utf8) {
1a4fad37 1465 LOAD_UTF8_CHARCLASS_DIGIT();
078c425b 1466 while (s + (uskip = UTF8SKIP(s)) <= strend) {
3568d838 1467 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
3b0527fe 1468 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1469 goto got_it;
1470 else
1471 tmp = doevery;
1472 }
a0d0e21e 1473 else
ffc61ed2 1474 tmp = 1;
078c425b 1475 s += uskip;
2b69d0c2 1476 }
a0d0e21e 1477 }
ffc61ed2
JH
1478 else {
1479 while (s < strend) {
1480 if (isDIGIT(*s)) {
3b0527fe 1481 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1482 goto got_it;
1483 else
1484 tmp = doevery;
1485 }
a0ed51b3 1486 else
ffc61ed2
JH
1487 tmp = 1;
1488 s++;
a0ed51b3 1489 }
a0ed51b3
LW
1490 }
1491 break;
b8c5462f
JH
1492 case DIGITL:
1493 PL_reg_flags |= RF_tainted;
ffc61ed2 1494 if (do_utf8) {
078c425b 1495 while (s + (uskip = UTF8SKIP(s)) <= strend) {
ffc61ed2 1496 if (isDIGIT_LC_utf8((U8*)s)) {
3b0527fe 1497 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1498 goto got_it;
1499 else
1500 tmp = doevery;
1501 }
b8c5462f 1502 else
ffc61ed2 1503 tmp = 1;
078c425b 1504 s += uskip;
b8c5462f 1505 }
b8c5462f 1506 }
ffc61ed2
JH
1507 else {
1508 while (s < strend) {
1509 if (isDIGIT_LC(*s)) {
3b0527fe 1510 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1511 goto got_it;
1512 else
1513 tmp = doevery;
1514 }
b8c5462f 1515 else
ffc61ed2
JH
1516 tmp = 1;
1517 s++;
b8c5462f 1518 }
b8c5462f
JH
1519 }
1520 break;
a0d0e21e 1521 case NDIGIT:
ffc61ed2 1522 if (do_utf8) {
1a4fad37 1523 LOAD_UTF8_CHARCLASS_DIGIT();
078c425b 1524 while (s + (uskip = UTF8SKIP(s)) <= strend) {
3568d838 1525 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
3b0527fe 1526 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1527 goto got_it;
1528 else
1529 tmp = doevery;
1530 }
a0d0e21e 1531 else
ffc61ed2 1532 tmp = 1;
078c425b 1533 s += uskip;
a687059c 1534 }
a0d0e21e 1535 }
ffc61ed2
JH
1536 else {
1537 while (s < strend) {
1538 if (!isDIGIT(*s)) {
3b0527fe 1539 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1540 goto got_it;
1541 else
1542 tmp = doevery;
1543 }
a0ed51b3 1544 else
ffc61ed2
JH
1545 tmp = 1;
1546 s++;
a0ed51b3 1547 }
a0ed51b3
LW
1548 }
1549 break;
b8c5462f
JH
1550 case NDIGITL:
1551 PL_reg_flags |= RF_tainted;
ffc61ed2 1552 if (do_utf8) {
078c425b 1553 while (s + (uskip = UTF8SKIP(s)) <= strend) {
ffc61ed2 1554 if (!isDIGIT_LC_utf8((U8*)s)) {
3b0527fe 1555 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1556 goto got_it;
1557 else
1558 tmp = doevery;
1559 }
b8c5462f 1560 else
ffc61ed2 1561 tmp = 1;
078c425b 1562 s += uskip;
b8c5462f 1563 }
a0ed51b3 1564 }
ffc61ed2
JH
1565 else {
1566 while (s < strend) {
1567 if (!isDIGIT_LC(*s)) {
3b0527fe 1568 if (tmp && (!reginfo || regtry(reginfo, s)))
ffc61ed2
JH
1569 goto got_it;
1570 else
1571 tmp = doevery;
1572 }
cf93c79d 1573 else
ffc61ed2
JH
1574 tmp = 1;
1575 s++;
b8c5462f 1576 }
b8c5462f
JH
1577 }
1578 break;
b3c9acc1 1579 default:
3c3eec57
GS
1580 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1581 break;
d6a28714 1582 }
6eb5f6b9
JH
1583 return 0;
1584 got_it:
1585 return s;
1586}
1587
1588/*
1589 - regexec_flags - match a regexp against a string
1590 */
1591I32
1592Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1593 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1594/* strend: pointer to null at end of string */
1595/* strbeg: real beginning of string */
1596/* minend: end of match must be >=minend after stringarg. */
1597/* data: May be used for some additional optimizations. */
1598/* nosave: For optimizations. */
1599{
97aff369 1600 dVAR;
6eb5f6b9
JH
1601 register char *s;
1602 register regnode *c;
1603 register char *startpos = stringarg;
6eb5f6b9
JH
1604 I32 minlen; /* must match at least this many chars */
1605 I32 dontbother = 0; /* how many characters not to try at end */
6eb5f6b9
JH
1606 I32 end_shift = 0; /* Same for the end. */ /* CC */
1607 I32 scream_pos = -1; /* Internal iterator of scream. */
ccac19ea 1608 char *scream_olds = NULL;
6eb5f6b9 1609 SV* oreplsv = GvSV(PL_replgv);
1df70142 1610 const bool do_utf8 = DO_UTF8(sv);
2757e526 1611 I32 multiline;
2a782b5b 1612#ifdef DEBUGGING
2757e526
JH
1613 SV* dsv0;
1614 SV* dsv1;
2a782b5b 1615#endif
3b0527fe 1616 regmatch_info reginfo; /* create some info to pass to regtry etc */
a3621e74
YO
1617
1618 GET_RE_DEBUG_FLAGS_DECL;
1619
9d4ba2ae 1620 PERL_UNUSED_ARG(data);
6eb5f6b9
JH
1621
1622 /* Be paranoid... */
1623 if (prog == NULL || startpos == NULL) {
1624 Perl_croak(aTHX_ "NULL regexp parameter");
1625 return 0;
1626 }
1627
2757e526 1628 multiline = prog->reganch & PMf_MULTILINE;
3b0527fe 1629 reginfo.prog = prog;
2757e526
JH
1630
1631#ifdef DEBUGGING
1632 dsv0 = PERL_DEBUG_PAD_ZERO(0);
1633 dsv1 = PERL_DEBUG_PAD_ZERO(1);
1634#endif
1635
bac06658
JH
1636 RX_MATCH_UTF8_set(prog, do_utf8);
1637
6eb5f6b9 1638 minlen = prog->minlen;
61a36c01 1639 if (strend - startpos < minlen) {
a3621e74 1640 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
a72c7584
JH
1641 "String too short [regexec_flags]...\n"));
1642 goto phooey;
1aa99e6b 1643 }
6eb5f6b9 1644
6eb5f6b9
JH
1645 /* Check validity of program. */
1646 if (UCHARAT(prog->program) != REG_MAGIC) {
1647 Perl_croak(aTHX_ "corrupted regexp program");
1648 }
1649
1650 PL_reg_flags = 0;
1651 PL_reg_eval_set = 0;
1652 PL_reg_maxiter = 0;
1653
1654 if (prog->reganch & ROPT_UTF8)
1655 PL_reg_flags |= RF_utf8;
1656
1657 /* Mark beginning of line for ^ and lookbehind. */
3b0527fe 1658 reginfo.bol = startpos; /* XXX not used ??? */
6eb5f6b9 1659 PL_bostr = strbeg;
3b0527fe 1660 reginfo.sv = sv;
6eb5f6b9
JH
1661
1662 /* Mark end of line for $ (and such) */
1663 PL_regeol = strend;
1664
1665 /* see how far we have to get to not match where we matched before */
3b0527fe 1666 reginfo.till = startpos+minend;
6eb5f6b9 1667
6eb5f6b9
JH
1668 /* If there is a "must appear" string, look for it. */
1669 s = startpos;
1670
3b0527fe 1671 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to set reginfo->ganch */
6eb5f6b9
JH
1672 MAGIC *mg;
1673
1674 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
3b0527fe 1675 reginfo.ganch = startpos;
6eb5f6b9
JH
1676 else if (sv && SvTYPE(sv) >= SVt_PVMG
1677 && SvMAGIC(sv)
14befaf4
DM
1678 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1679 && mg->mg_len >= 0) {
3b0527fe 1680 reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */
6eb5f6b9 1681 if (prog->reganch & ROPT_ANCH_GPOS) {
3b0527fe 1682 if (s > reginfo.ganch)
6eb5f6b9 1683 goto phooey;
3b0527fe 1684 s = reginfo.ganch;
6eb5f6b9
JH
1685 }
1686 }
1687 else /* pos() not defined */
3b0527fe 1688 reginfo.ganch = strbeg;
6eb5f6b9
JH
1689 }
1690
a0714e2c 1691 if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
6eb5f6b9
JH
1692 re_scream_pos_data d;
1693
1694 d.scream_olds = &scream_olds;
1695 d.scream_pos = &scream_pos;
1696 s = re_intuit_start(prog, sv, s, strend, flags, &d);
3fa9c3d7 1697 if (!s) {
a3621e74 1698 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
6eb5f6b9 1699 goto phooey; /* not present */
3fa9c3d7 1700 }
6eb5f6b9
JH
1701 }
1702
a3621e74 1703 DEBUG_EXECUTE_r({
1df70142
AL
1704 const char * const s0 = UTF
1705 ? pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1706 UNI_DISPLAY_REGEX)
1707 : prog->precomp;
1708 const int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1709 const char * const s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
c728cb41 1710 UNI_DISPLAY_REGEX) : startpos;
1df70142 1711 const int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
2a782b5b
JH
1712 if (!PL_colorset)
1713 reginitcolors();
1714 PerlIO_printf(Perl_debug_log,
a0288114 1715 "%sMatching REx%s \"%s%*.*s%s%s\" against \"%s%.*s%s%s\"\n",
e4584336 1716 PL_colors[4], PL_colors[5], PL_colors[0],
9e55ce06 1717 len0, len0, s0,
2a782b5b 1718 PL_colors[1],
9e55ce06 1719 len0 > 60 ? "..." : "",
2a782b5b 1720 PL_colors[0],
9e55ce06
JH
1721 (int)(len1 > 60 ? 60 : len1),
1722 s1, PL_colors[1],
1723 (len1 > 60 ? "..." : "")
2a782b5b
JH
1724 );
1725 });
6eb5f6b9
JH
1726
1727 /* Simplest case: anchored match need be tried only once. */
1728 /* [unless only anchor is BOL and multiline is set] */
1729 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
3b0527fe 1730 if (s == startpos && regtry(&reginfo, startpos))
6eb5f6b9 1731 goto got_it;
7fba1cd6 1732 else if (multiline || (prog->reganch & ROPT_IMPLICIT)
6eb5f6b9
JH
1733 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1734 {
1735 char *end;
1736
1737 if (minlen)
1738 dontbother = minlen - 1;
1aa99e6b 1739 end = HOP3c(strend, -dontbother, strbeg) - 1;
6eb5f6b9 1740 /* for multiline we only have to try after newlines */
33b8afdf 1741 if (prog->check_substr || prog->check_utf8) {
6eb5f6b9
JH
1742 if (s == startpos)
1743 goto after_try;
1744 while (1) {
3b0527fe 1745 if (regtry(&reginfo, s))
6eb5f6b9
JH
1746 goto got_it;
1747 after_try:
1748 if (s >= end)
1749 goto phooey;
1750 if (prog->reganch & RE_USE_INTUIT) {
1751 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1752 if (!s)
1753 goto phooey;
1754 }
1755 else
1756 s++;
1757 }
1758 } else {
1759 if (s > startpos)
1760 s--;
1761 while (s < end) {
1762 if (*s++ == '\n') { /* don't need PL_utf8skip here */
3b0527fe 1763 if (regtry(&reginfo, s))
6eb5f6b9
JH
1764 goto got_it;
1765 }
1766 }
1767 }
1768 }
1769 goto phooey;
1770 } else if (prog->reganch & ROPT_ANCH_GPOS) {
3b0527fe 1771 if (regtry(&reginfo, reginfo.ganch))
6eb5f6b9
JH
1772 goto got_it;
1773 goto phooey;
1774 }
1775
1776 /* Messy cases: unanchored match. */
33b8afdf 1777 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
6eb5f6b9
JH
1778 /* we have /x+whatever/ */
1779 /* it must be a one character string (XXXX Except UTF?) */
33b8afdf 1780 char ch;
bf93d4cc
GS
1781#ifdef DEBUGGING
1782 int did_match = 0;
1783#endif
33b8afdf
JH
1784 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1785 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
3f7c398e 1786 ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
bf93d4cc 1787
1aa99e6b 1788 if (do_utf8) {
6eb5f6b9
JH
1789 while (s < strend) {
1790 if (*s == ch) {
a3621e74 1791 DEBUG_EXECUTE_r( did_match = 1 );
3b0527fe 1792 if (regtry(&reginfo, s)) goto got_it;
6eb5f6b9
JH
1793 s += UTF8SKIP(s);
1794 while (s < strend && *s == ch)
1795 s += UTF8SKIP(s);
1796 }
1797 s += UTF8SKIP(s);
1798 }
1799 }
1800 else {
1801 while (s < strend) {
1802 if (*s == ch) {
a3621e74 1803 DEBUG_EXECUTE_r( did_match = 1 );
3b0527fe 1804 if (regtry(&reginfo, s)) goto got_it;
6eb5f6b9
JH
1805 s++;
1806 while (s < strend && *s == ch)
1807 s++;
1808 }
1809 s++;
1810 }
1811 }
a3621e74 1812 DEBUG_EXECUTE_r(if (!did_match)
bf93d4cc 1813 PerlIO_printf(Perl_debug_log,
b7953727
JH
1814 "Did not find anchored character...\n")
1815 );
6eb5f6b9 1816 }
a0714e2c
SS
1817 else if (prog->anchored_substr != NULL
1818 || prog->anchored_utf8 != NULL
1819 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
33b8afdf
JH
1820 && prog->float_max_offset < strend - s)) {
1821 SV *must;
1822 I32 back_max;
1823 I32 back_min;
1824 char *last;
6eb5f6b9 1825 char *last1; /* Last position checked before */
bf93d4cc
GS
1826#ifdef DEBUGGING
1827 int did_match = 0;
1828#endif
33b8afdf
JH
1829 if (prog->anchored_substr || prog->anchored_utf8) {
1830 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1831 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1832 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1833 back_max = back_min = prog->anchored_offset;
1834 } else {
1835 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1836 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1837 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1838 back_max = prog->float_max_offset;
1839 back_min = prog->float_min_offset;
1840 }
1841 if (must == &PL_sv_undef)
1842 /* could not downgrade utf8 check substring, so must fail */
1843 goto phooey;
1844
1845 last = HOP3c(strend, /* Cannot start after this */
1846 -(I32)(CHR_SVLEN(must)
1847 - (SvTAIL(must) != 0) + back_min), strbeg);
6eb5f6b9
JH
1848
1849 if (s > PL_bostr)
1850 last1 = HOPc(s, -1);
1851 else
1852 last1 = s - 1; /* bogus */
1853
a0288114 1854 /* XXXX check_substr already used to find "s", can optimize if
6eb5f6b9
JH
1855 check_substr==must. */
1856 scream_pos = -1;
1857 dontbother = end_shift;
1858 strend = HOPc(strend, -dontbother);
1859 while ( (s <= last) &&
9041c2e3 1860 ((flags & REXEC_SCREAM)
1aa99e6b 1861 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
6eb5f6b9 1862 end_shift, &scream_pos, 0))
1aa99e6b 1863 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
9041c2e3 1864 (unsigned char*)strend, must,
7fba1cd6 1865 multiline ? FBMrf_MULTILINE : 0))) ) {
4addbd3b
HS
1866 /* we may be pointing at the wrong string */
1867 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
3f7c398e 1868 s = strbeg + (s - SvPVX_const(sv));
a3621e74 1869 DEBUG_EXECUTE_r( did_match = 1 );
6eb5f6b9
JH
1870 if (HOPc(s, -back_max) > last1) {
1871 last1 = HOPc(s, -back_min);
1872 s = HOPc(s, -back_max);
1873 }
1874 else {
52657f30 1875 char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
6eb5f6b9
JH
1876
1877 last1 = HOPc(s, -back_min);
52657f30 1878 s = t;
6eb5f6b9 1879 }
1aa99e6b 1880 if (do_utf8) {
6eb5f6b9 1881 while (s <= last1) {
3b0527fe 1882 if (regtry(&reginfo, s))
6eb5f6b9
JH
1883 goto got_it;
1884 s += UTF8SKIP(s);
1885 }
1886 }
1887 else {
1888 while (s <= last1) {
3b0527fe 1889 if (regtry(&reginfo, s))
6eb5f6b9
JH
1890 goto got_it;
1891 s++;
1892 }
1893 }
1894 }
a3621e74 1895 DEBUG_EXECUTE_r(if (!did_match)
b7953727 1896 PerlIO_printf(Perl_debug_log,
a0288114 1897 "Did not find %s substr \"%s%.*s%s\"%s...\n",
33b8afdf 1898 ((must == prog->anchored_substr || must == prog->anchored_utf8)
bf93d4cc
GS
1899 ? "anchored" : "floating"),
1900 PL_colors[0],
1901 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
3f7c398e 1902 SvPVX_const(must),
b7953727
JH
1903 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1904 );
6eb5f6b9
JH
1905 goto phooey;
1906 }
155aba94 1907 else if ((c = prog->regstclass)) {
f14c76ed
RGS
1908 if (minlen) {
1909 I32 op = (U8)OP(prog->regstclass);
66e933ab 1910 /* don't bother with what can't match */
f14c76ed
RGS
1911 if (PL_regkind[op] != EXACT && op != CANY)
1912 strend = HOPc(strend, -(minlen - 1));
1913 }
a3621e74 1914 DEBUG_EXECUTE_r({
ffc61ed2 1915 SV *prop = sv_newmortal();
cfd0369c
NC
1916 const char *s0;
1917 const char *s1;
9e55ce06
JH
1918 int len0;
1919 int len1;
1920
32fc9b6a 1921 regprop(prog, prop, c);
9e55ce06 1922 s0 = UTF ?
3f7c398e 1923 pv_uni_display(dsv0, (U8*)SvPVX_const(prop), SvCUR(prop), 60,
c728cb41 1924 UNI_DISPLAY_REGEX) :
cfd0369c 1925 SvPVX_const(prop);
9e55ce06
JH
1926 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1927 s1 = UTF ?
c728cb41 1928 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
9e55ce06
JH
1929 len1 = UTF ? SvCUR(dsv1) : strend - s;
1930 PerlIO_printf(Perl_debug_log,
a0288114 1931 "Matching stclass \"%*.*s\" against \"%*.*s\"\n",
9e55ce06
JH
1932 len0, len0, s0,
1933 len1, len1, s1);
ffc61ed2 1934 });
3b0527fe 1935 if (find_byclass(prog, c, s, strend, &reginfo))
6eb5f6b9 1936 goto got_it;
a3621e74 1937 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
d6a28714
JH
1938 }
1939 else {
1940 dontbother = 0;
a0714e2c 1941 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
33b8afdf 1942 /* Trim the end. */
d6a28714 1943 char *last;
33b8afdf
JH
1944 SV* float_real;
1945
1946 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1947 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1948 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
d6a28714
JH
1949
1950 if (flags & REXEC_SCREAM) {
33b8afdf 1951 last = screaminstr(sv, float_real, s - strbeg,
d6a28714
JH
1952 end_shift, &scream_pos, 1); /* last one */
1953 if (!last)
ffc61ed2 1954 last = scream_olds; /* Only one occurrence. */
4addbd3b
HS
1955 /* we may be pointing at the wrong string */
1956 else if (RX_MATCH_COPIED(prog))
3f7c398e 1957 s = strbeg + (s - SvPVX_const(sv));
b8c5462f 1958 }
d6a28714
JH
1959 else {
1960 STRLEN len;
cfd0369c 1961 const char * const little = SvPV_const(float_real, len);
d6a28714 1962
33b8afdf 1963 if (SvTAIL(float_real)) {
d6a28714
JH
1964 if (memEQ(strend - len + 1, little, len - 1))
1965 last = strend - len + 1;
7fba1cd6 1966 else if (!multiline)
9041c2e3 1967 last = memEQ(strend - len, little, len)
bd61b366 1968 ? strend - len : NULL;
b8c5462f 1969 else
d6a28714
JH
1970 goto find_last;
1971 } else {
1972 find_last:
9041c2e3 1973 if (len)
d6a28714 1974 last = rninstr(s, strend, little, little + len);
b8c5462f 1975 else
a0288114 1976 last = strend; /* matching "$" */
b8c5462f 1977 }
b8c5462f 1978 }
bf93d4cc 1979 if (last == NULL) {
a3621e74 1980 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
bf93d4cc 1981 "%sCan't trim the tail, match fails (should not happen)%s\n",
e4584336 1982 PL_colors[4], PL_colors[5]));
bf93d4cc
GS
1983 goto phooey; /* Should not happen! */
1984 }
d6a28714
JH
1985 dontbother = strend - last + prog->float_min_offset;
1986 }
1987 if (minlen && (dontbother < minlen))
1988 dontbother = minlen - 1;
1989 strend -= dontbother; /* this one's always in bytes! */
1990 /* We don't know much -- general case. */
1aa99e6b 1991 if (do_utf8) {
d6a28714 1992 for (;;) {
3b0527fe 1993 if (regtry(&reginfo, s))
d6a28714
JH
1994 goto got_it;
1995 if (s >= strend)
1996 break;
b8c5462f 1997 s += UTF8SKIP(s);
d6a28714
JH
1998 };
1999 }
2000 else {
2001 do {
3b0527fe 2002 if (regtry(&reginfo, s))
d6a28714
JH
2003 goto got_it;
2004 } while (s++ < strend);
2005 }
2006 }
2007
2008 /* Failure. */
2009 goto phooey;
2010
2011got_it:
2012 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2013
2014 if (PL_reg_eval_set) {
2015 /* Preserve the current value of $^R */
2016 if (oreplsv != GvSV(PL_replgv))
2017 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2018 restored, the value remains
2019 the same. */
4f639d21 2020 restore_pos(aTHX_ prog);
d6a28714
JH
2021 }
2022
2023 /* make sure $`, $&, $', and $digit will work later */
2024 if ( !(flags & REXEC_NOT_FIRST) ) {
ed252734 2025 RX_MATCH_COPY_FREE(prog);
d6a28714
JH
2026 if (flags & REXEC_COPY_STR) {
2027 I32 i = PL_regeol - startpos + (stringarg - strbeg);
f8c7b90f 2028#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
2029 if ((SvIsCOW(sv)
2030 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2031 if (DEBUG_C_TEST) {
2032 PerlIO_printf(Perl_debug_log,
2033 "Copy on write: regexp capture, type %d\n",
2034 (int) SvTYPE(sv));
2035 }
2036 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
d5263905 2037 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
ed252734
NC
2038 assert (SvPOKp(prog->saved_copy));
2039 } else
2040#endif
2041 {
2042 RX_MATCH_COPIED_on(prog);
2043 s = savepvn(strbeg, i);
2044 prog->subbeg = s;
2045 }
d6a28714 2046 prog->sublen = i;
d6a28714
JH
2047 }
2048 else {
2049 prog->subbeg = strbeg;
2050 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2051 }
2052 }
9041c2e3 2053
d6a28714
JH
2054 return 1;
2055
2056phooey:
a3621e74 2057 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
e4584336 2058 PL_colors[4], PL_colors[5]));
d6a28714 2059 if (PL_reg_eval_set)
4f639d21 2060 restore_pos(aTHX_ prog);
d6a28714
JH
2061 return 0;
2062}
2063
2064/*
2065 - regtry - try match at specific point
2066 */
2067STATIC I32 /* 0 failure, 1 success */
3b0527fe 2068S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos)
d6a28714 2069{
97aff369 2070 dVAR;
d6a28714
JH
2071 register I32 *sp;
2072 register I32 *ep;
2073 CHECKPOINT lastcp;
3b0527fe 2074 regexp *prog = reginfo->prog;
a3621e74 2075 GET_RE_DEBUG_FLAGS_DECL;
d6a28714 2076
02db2b7b
IZ
2077#ifdef DEBUGGING
2078 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2079#endif
d6a28714
JH
2080 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2081 MAGIC *mg;
2082
2083 PL_reg_eval_set = RS_init;
a3621e74 2084 DEBUG_EXECUTE_r(DEBUG_s(
b900a521
JH
2085 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2086 (IV)(PL_stack_sp - PL_stack_base));
d6a28714 2087 ));
e8347627 2088 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
d6a28714
JH
2089 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2090 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2091 SAVETMPS;
2092 /* Apparently this is not needed, judging by wantarray. */
e8347627 2093 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
d6a28714
JH
2094 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2095
3b0527fe 2096 if (reginfo->sv) {
d6a28714 2097 /* Make $_ available to executed code. */
3b0527fe 2098 if (reginfo->sv != DEFSV) {
59f00321 2099 SAVE_DEFSV;
3b0527fe 2100 DEFSV = reginfo->sv;
b8c5462f 2101 }
d6a28714 2102
3b0527fe
DM
2103 if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv)
2104 && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) {
d6a28714 2105 /* prepare for quick setting of pos */
d300d9fa
NC
2106#ifdef PERL_OLD_COPY_ON_WRITE
2107 if (SvIsCOW(sv))
2108 sv_force_normal_flags(sv, 0);
2109#endif
3b0527fe 2110 mg = sv_magicext(reginfo->sv, (SV*)0, PERL_MAGIC_regex_global,
d300d9fa 2111 &PL_vtbl_mglob, NULL, 0);
d6a28714 2112 mg->mg_len = -1;
b8c5462f 2113 }
d6a28714
JH
2114 PL_reg_magic = mg;
2115 PL_reg_oldpos = mg->mg_len;
4f639d21 2116 SAVEDESTRUCTOR_X(restore_pos, prog);
d6a28714 2117 }
09687e5a 2118 if (!PL_reg_curpm) {
a02a5408 2119 Newxz(PL_reg_curpm, 1, PMOP);
09687e5a
AB
2120#ifdef USE_ITHREADS
2121 {
2122 SV* repointer = newSViv(0);
577e12cc 2123 /* so we know which PL_regex_padav element is PL_reg_curpm */
35061a7e 2124 SvFLAGS(repointer) |= SVf_BREAK;
09687e5a
AB
2125 av_push(PL_regex_padav,repointer);
2126 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2127 PL_regex_pad = AvARRAY(PL_regex_padav);
2128 }
2129#endif
2130 }
aaa362c4 2131 PM_SETRE(PL_reg_curpm, prog);
d6a28714
JH
2132 PL_reg_oldcurpm = PL_curpm;
2133 PL_curpm = PL_reg_curpm;
2134 if (RX_MATCH_COPIED(prog)) {
2135 /* Here is a serious problem: we cannot rewrite subbeg,
2136 since it may be needed if this match fails. Thus
2137 $` inside (?{}) could fail... */
2138 PL_reg_oldsaved = prog->subbeg;
2139 PL_reg_oldsavedlen = prog->sublen;
f8c7b90f 2140#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
2141 PL_nrs = prog->saved_copy;
2142#endif
d6a28714
JH
2143 RX_MATCH_COPIED_off(prog);
2144 }
2145 else
bd61b366 2146 PL_reg_oldsaved = NULL;
d6a28714
JH
2147 prog->subbeg = PL_bostr;
2148 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2149 }
973dddac 2150 prog->startp[0] = startpos - PL_bostr;
d6a28714
JH
2151 PL_reginput = startpos;
2152 PL_regstartp = prog->startp;
2153 PL_regendp = prog->endp;
2154 PL_reglastparen = &prog->lastparen;
a01268b5 2155 PL_reglastcloseparen = &prog->lastcloseparen;
d6a28714 2156 prog->lastparen = 0;
03994de8 2157 prog->lastcloseparen = 0;
d6a28714 2158 PL_regsize = 0;
a3621e74 2159 DEBUG_EXECUTE_r(PL_reg_starttry = startpos);
d6a28714
JH
2160 if (PL_reg_start_tmpl <= prog->nparens) {
2161 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2162 if(PL_reg_start_tmp)
2163 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2164 else
a02a5408 2165 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
d6a28714
JH
2166 }
2167
2168 /* XXXX What this code is doing here?!!! There should be no need
2169 to do this again and again, PL_reglastparen should take care of
3dd2943c 2170 this! --ilya*/
dafc8851
JH
2171
2172 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2173 * Actually, the code in regcppop() (which Ilya may be meaning by
daf18116
JH
2174 * PL_reglastparen), is not needed at all by the test suite
2175 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2176 * enough, for building DynaLoader, or otherwise this
2177 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2178 * will happen. Meanwhile, this code *is* needed for the
2179 * above-mentioned test suite tests to succeed. The common theme
2180 * on those tests seems to be returning null fields from matches.
2181 * --jhi */
dafc8851 2182#if 1
d6a28714
JH
2183 sp = prog->startp;
2184 ep = prog->endp;
2185 if (prog->nparens) {
097eb12c 2186 register I32 i;
eb160463 2187 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
d6a28714
JH
2188 *++sp = -1;
2189 *++ep = -1;
2190 }
2191 }
dafc8851 2192#endif
02db2b7b 2193 REGCP_SET(lastcp);
3b0527fe 2194 if (regmatch(reginfo, prog->program + 1)) {
d6a28714
JH
2195 prog->endp[0] = PL_reginput - PL_bostr;
2196 return 1;
2197 }
02db2b7b 2198 REGCP_UNWIND(lastcp);
d6a28714
JH
2199 return 0;
2200}
2201
02db2b7b
IZ
2202#define RE_UNWIND_BRANCH 1
2203#define RE_UNWIND_BRANCHJ 2
2204
2205union re_unwind_t;
2206
2207typedef struct { /* XX: makes sense to enlarge it... */
2208 I32 type;
2209 I32 prev;
2210 CHECKPOINT lastcp;
2211} re_unwind_generic_t;
2212
2213typedef struct {
2214 I32 type;
2215 I32 prev;
2216 CHECKPOINT lastcp;
2217 I32 lastparen;
2218 regnode *next;
2219 char *locinput;
2220 I32 nextchr;
3a2830be 2221 int minmod;
02db2b7b
IZ
2222#ifdef DEBUGGING
2223 int regindent;
2224#endif
2225} re_unwind_branch_t;
2226
2227typedef union re_unwind_t {
2228 I32 type;
2229 re_unwind_generic_t generic;
2230 re_unwind_branch_t branch;
2231} re_unwind_t;
2232
8ba1375e
MJD
2233#define sayYES goto yes
2234#define sayNO goto no
e0f9d4a8 2235#define sayNO_ANYOF goto no_anyof
8ba1375e
MJD
2236#define sayYES_FINAL goto yes_final
2237#define sayYES_LOUD goto yes_loud
2238#define sayNO_FINAL goto no_final
2239#define sayNO_SILENT goto do_no
2240#define saySAME(x) if (x) goto yes; else goto no
2241
3ab3c9b4
HS
2242#define POSCACHE_SUCCESS 0 /* caching success rather than failure */
2243#define POSCACHE_SEEN 1 /* we know what we're caching */
2244#define POSCACHE_START 2 /* the real cache: this bit maps to pos 0 */
2245#define CACHEsayYES STMT_START { \
d8319b27 2246 if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) { \
3ab3c9b4
HS
2247 if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
2248 PL_reg_poscache[0] |= (1<<POSCACHE_SUCCESS) || (1<<POSCACHE_SEEN); \
2249 else if (!(PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
2250 /* cache records failure, but this is success */ \
2251 DEBUG_r( \
2252 PerlIO_printf(Perl_debug_log, \
2253 "%*s (remove success from failure cache)\n", \
2254 REPORT_CODE_OFF+PL_regindent*2, "") \
2255 ); \
d8319b27 2256 PL_reg_poscache[st->u.whilem.cache_offset] &= ~(1<<st->u.whilem.cache_bit); \
3ab3c9b4
HS
2257 } \
2258 } \
2259 sayYES; \
2260} STMT_END
2261#define CACHEsayNO STMT_START { \
d8319b27 2262 if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) { \
3ab3c9b4
HS
2263 if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
2264 PL_reg_poscache[0] |= (1<<POSCACHE_SEEN); \
2265 else if ((PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
2266 /* cache records success, but this is failure */ \
2267 DEBUG_r( \
2268 PerlIO_printf(Perl_debug_log, \
2269 "%*s (remove failure from success cache)\n", \
2270 REPORT_CODE_OFF+PL_regindent*2, "") \
2271 ); \
d8319b27 2272 PL_reg_poscache[st->u.whilem.cache_offset] &= ~(1<<st->u.whilem.cache_bit); \
3ab3c9b4
HS
2273 } \
2274 } \
2275 sayNO; \
2276} STMT_END
2277
a3621e74
YO
2278/* this is used to determine how far from the left messages like
2279 'failed...' are printed. Currently 29 makes these messages line
2280 up with the opcode they refer to. Earlier perls used 25 which
2281 left these messages outdented making reviewing a debug output
2282 quite difficult.
2283*/
2284#define REPORT_CODE_OFF 29
2285
2286
2287/* Make sure there is a test for this +1 options in re_tests */
2288#define TRIE_INITAL_ACCEPT_BUFFLEN 4;
2289
5d9a96ca
DM
2290/* grab a new slab and return the first slot in it */
2291
2292STATIC regmatch_state *
2293S_push_slab(pTHX)
2294{
2295 regmatch_slab *s = PL_regmatch_slab->next;
2296 if (!s) {
2297 Newx(s, 1, regmatch_slab);
2298 s->prev = PL_regmatch_slab;
2299 s->next = NULL;
2300 PL_regmatch_slab->next = s;
2301 }
2302 PL_regmatch_slab = s;
2303 return &s->states[0];
2304}
5b47454d 2305
95b24440
DM
2306/* simulate a recursive call to regmatch */
2307
2308#define REGMATCH(ns, where) \
5d9a96ca
DM
2309 st->scan = scan; \
2310 scan = (ns); \
2311 st->resume_state = resume_##where; \
95b24440
DM
2312 goto start_recurse; \
2313 resume_point_##where:
2314
aa283a38
DM
2315
2316/* push a new regex state. Set newst to point to it */
2317
2318#define PUSH_STATE(newst, resume) \
2319 depth++; \
2320 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "PUSH STATE(%d)\n", depth)); \
2321 st->scan = scan; \
2322 st->next = next; \
2323 st->n = n; \
2324 st->locinput = locinput; \
2325 st->resume_state = resume; \
2326 newst = st+1; \
2327 if (newst > &(PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1])) \
2328 newst = S_push_slab(aTHX); \
2329 PL_regmatch_state = newst; \
2330 newst->cc = 0; \
2331 newst->minmod = 0; \
2332 newst->sw = 0; \
2333 newst->logical = 0; \
2334 newst->unwind = 0; \
2335 locinput = PL_reginput; \
2336 nextchr = UCHARAT(locinput);
2337
2338#define POP_STATE \
2339 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "POP STATE(%d)\n", depth)); \
2340 depth--; \
2341 st--; \
2342 if (st < &PL_regmatch_slab->states[0]) { \
2343 PL_regmatch_slab = PL_regmatch_slab->prev; \
2344 st = &PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]; \
2345 } \
2346 PL_regmatch_state = st; \
2347 scan = st->scan; \
2348 next = st->next; \
2349 n = st->n; \
2350 locinput = st->locinput; \
2351 nextchr = UCHARAT(locinput);
2352
d6a28714
JH
2353/*
2354 - regmatch - main matching routine
2355 *
2356 * Conceptually the strategy is simple: check to see whether the current
2357 * node matches, call self recursively to see whether the rest matches,
2358 * and then act accordingly. In practice we make some effort to avoid
2359 * recursion, in particular by going through "ordinary" nodes (that don't
2360 * need to know whether the rest of the match failed) by a loop instead of
2361 * by recursion.
2362 */
2363/* [lwall] I've hoisted the register declarations to the outer block in order to
2364 * maybe save a little bit of pushing and popping on the stack. It also takes
2365 * advantage of machines that use a register save mask on subroutine entry.
95b24440
DM
2366 *
2367 * This function used to be heavily recursive, but since this had the
2368 * effect of blowing the CPU stack on complex regexes, it has been
2369 * restructured to be iterative, and to save state onto the heap rather
2370 * than the stack. Essentially whereever regmatch() used to be called, it
2371 * pushes the current state, notes where to return, then jumps back into
2372 * the main loop.
2373 *
2374 * Originally the structure of this function used to look something like
2375
2376 S_regmatch() {
2377 int a = 1, b = 2;
2378 ...
2379 while (scan != NULL) {
5d9a96ca 2380 a++; // do stuff with a and b
95b24440
DM
2381 ...
2382 switch (OP(scan)) {
2383 case FOO: {
2384 int local = 3;
2385 ...
2386 if (regmatch(...)) // recurse
2387 goto yes;
2388 }
2389 ...
2390 }
2391 }
2392 yes:
2393 return 1;
2394 }
2395
2396 * Now it looks something like this:
2397
5d9a96ca 2398 typedef struct {
95b24440
DM
2399 int a, b, local;
2400 int resume_state;
5d9a96ca 2401 } regmatch_state;
95b24440
DM
2402
2403 S_regmatch() {
5d9a96ca
DM
2404 regmatch_state *st = new();
2405 int depth=0;
2406 st->a++; // do stuff with a and b
95b24440
DM
2407 ...
2408 while (scan != NULL) {
2409 ...
2410 switch (OP(scan)) {
2411 case FOO: {
5d9a96ca 2412 st->local = 3;
95b24440 2413 ...
5d9a96ca
DM
2414 st->scan = scan;
2415 scan = ...;
2416 st->resume_state = resume_FOO;
2417 goto start_recurse; // recurse
95b24440 2418
5d9a96ca
DM
2419 resume_point_FOO:
2420 if (result)
95b24440
DM
2421 goto yes;
2422 }
2423 ...
2424 }
5d9a96ca
DM
2425 start_recurse:
2426 st = new(); push a new state
2427 st->a = 1; st->b = 2;
2428 depth++;
95b24440 2429 }
5d9a96ca 2430 yes:
95b24440 2431 result = 1;
5d9a96ca
DM
2432 if (depth--) {
2433 st = pop();
95b24440
DM
2434 switch (resume_state) {
2435 case resume_FOO:
2436 goto resume_point_FOO;
2437 ...
2438 }
2439 }
2440 return result
2441 }
2442
2443 * WARNING: this means that any line in this function that contains a
2444 * REGMATCH() or TRYPAREN() is actually simulating a recursive call to
2445 * regmatch() using gotos instead. Thus the values of any local variables
2446 * not saved in the regmatch_state structure will have been lost when
2447 * execution resumes on the next line .
5d9a96ca
DM
2448 *
2449 * States (ie the st pointer) are allocated in slabs of about 4K in size.
2450 * PL_regmatch_state always points to the currently active state, and
2451 * PL_regmatch_slab points to the slab currently containing PL_regmatch_state.
2452 * The first time regmatch is called, the first slab is allocated, and is
2453 * never freed until interpreter desctruction. When the slab is full,
2454 * a new one is allocated chained to the end. At exit from regmatch, slabs
2455 * allocated since entry are freed.
d6a28714 2456 */
95b24440
DM
2457
2458
d6a28714 2459STATIC I32 /* 0 failure, 1 success */
3b0527fe 2460S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog)
d6a28714 2461{
27da23d5 2462 dVAR;
95b24440 2463 register const bool do_utf8 = PL_reg_match_utf8;
4ad0818d 2464 const U32 uniflags = UTF8_ALLOW_DEFAULT;
95b24440 2465
3b0527fe
DM
2466 regexp *rex = reginfo->prog;
2467
5d9a96ca
DM
2468 regmatch_slab *orig_slab;
2469 regmatch_state *orig_state;
a3621e74 2470
5d9a96ca
DM
2471 /* the current state. This is a cached copy of PL_regmatch_state */
2472 register regmatch_state *st;
95b24440 2473
5d9a96ca
DM
2474 /* cache heavy used fields of st in registers */
2475 register regnode *scan;
2476 register regnode *next;
2477 register I32 n = 0; /* initialize to shut up compiler warning */
2478 register char *locinput = PL_reginput;
95b24440 2479
5d9a96ca
DM
2480 /* these variables are NOT saved during a recusive RFEGMATCH: */
2481 register I32 nextchr; /* is always set to UCHARAT(locinput) */
2482 bool result; /* return value of S_regmatch */
2483 regnode *inner; /* Next node in internal branch. */
2484 int depth = 0; /* depth of recursion */
aa283a38
DM
2485 regmatch_state *newst; /* when pushing a state, this is the new one */
2486 regmatch_state *cur_eval = NULL; /* most recent (??{}) state */
95b24440
DM
2487
2488#ifdef DEBUGGING
ab74612d 2489 SV *re_debug_flags = NULL;
a3621e74 2490 GET_RE_DEBUG_FLAGS;
d6a28714
JH
2491 PL_regindent++;
2492#endif
2493
5d9a96ca
DM
2494 /* on first ever call to regmatch, allocate first slab */
2495 if (!PL_regmatch_slab) {
2496 Newx(PL_regmatch_slab, 1, regmatch_slab);
2497 PL_regmatch_slab->prev = NULL;
2498 PL_regmatch_slab->next = NULL;
2499 PL_regmatch_state = &PL_regmatch_slab->states[0] - 1;
2500 }
2501
2502 /* remember current high-water mark for exit */
2503 /* XXX this should be done with SAVE* instead */
2504 orig_slab = PL_regmatch_slab;
2505 orig_state = PL_regmatch_state;
2506
2507 /* grab next free state slot */
2508 st = ++PL_regmatch_state;
2509 if (st > &(PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]))
2510 st = PL_regmatch_state = S_push_slab(aTHX);
2511
2512 st->minmod = 0;
2513 st->sw = 0;
2514 st->logical = 0;
2515 st->unwind = 0;
2516 st->cc = NULL;
d6a28714
JH
2517 /* Note that nextchr is a byte even in UTF */
2518 nextchr = UCHARAT(locinput);
2519 scan = prog;
2520 while (scan != NULL) {
8ba1375e 2521
a3621e74 2522 DEBUG_EXECUTE_r( {
6136c704 2523 SV * const prop = sv_newmortal();
1df70142
AL
2524 const int docolor = *PL_colors[0];
2525 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
1aa99e6b 2526 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
d6a28714
JH
2527 /* The part of the string before starttry has one color
2528 (pref0_len chars), between starttry and current
2529 position another one (pref_len - pref0_len chars),
2530 after the current position the third one.
2531 We assume that pref0_len <= pref_len, otherwise we
2532 decrease pref0_len. */
9041c2e3 2533 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
1aa99e6b
IH
2534 ? (5 + taill) - l : locinput - PL_bostr;
2535 int pref0_len;
d6a28714 2536
df1ffd02 2537 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
1aa99e6b
IH
2538 pref_len++;
2539 pref0_len = pref_len - (locinput - PL_reg_starttry);
d6a28714 2540 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
9041c2e3 2541 l = ( PL_regeol - locinput > (5 + taill) - pref_len
d6a28714 2542 ? (5 + taill) - pref_len : PL_regeol - locinput);
df1ffd02 2543 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
1aa99e6b 2544 l--;
d6a28714
JH
2545 if (pref0_len < 0)
2546 pref0_len = 0;
2547 if (pref0_len > pref_len)
2548 pref0_len = pref_len;
32fc9b6a 2549 regprop(rex, prop, scan);
2a782b5b 2550 {
1df70142 2551 const char * const s0 =
f14c76ed 2552 do_utf8 && OP(scan) != CANY ?
95b24440 2553 pv_uni_display(PERL_DEBUG_PAD(0), (U8*)(locinput - pref_len),
c728cb41 2554 pref0_len, 60, UNI_DISPLAY_REGEX) :
2a782b5b 2555 locinput - pref_len;
1df70142
AL
2556 const int len0 = do_utf8 ? strlen(s0) : pref0_len;
2557 const char * const s1 = do_utf8 && OP(scan) != CANY ?
95b24440
DM
2558 pv_uni_display(PERL_DEBUG_PAD(1),
2559 (U8*)(locinput - pref_len + pref0_len),
c728cb41 2560 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2a782b5b 2561 locinput - pref_len + pref0_len;
1df70142
AL
2562 const int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2563 const char * const s2 = do_utf8 && OP(scan) != CANY ?
95b24440 2564 pv_uni_display(PERL_DEBUG_PAD(2), (U8*)locinput,
c728cb41 2565 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2a782b5b 2566 locinput;
1df70142 2567 const int len2 = do_utf8 ? strlen(s2) : l;
2a782b5b
JH
2568 PerlIO_printf(Perl_debug_log,
2569 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2570 (IV)(locinput - PL_bostr),
2571 PL_colors[4],
2572 len0, s0,
2573 PL_colors[5],
2574 PL_colors[2],
2575 len1, s1,
2576 PL_colors[3],
2577 (docolor ? "" : "> <"),
2578 PL_colors[0],
2579 len2, s2,
2580 PL_colors[1],
2581 15 - l - pref_len + 1,
2582 "",
4f639d21 2583 (IV)(scan - rex->program), PL_regindent*2, "",
3f7c398e 2584 SvPVX_const(prop));
2a782b5b
JH
2585 }
2586 });
d6a28714
JH
2587
2588 next = scan + NEXT_OFF(scan);
2589 if (next == scan)
2590 next = NULL;
2591
2592 switch (OP(scan)) {
2593 case BOL:
7fba1cd6 2594 if (locinput == PL_bostr)
d6a28714 2595 {
3b0527fe 2596 /* reginfo->till = reginfo->bol; */
b8c5462f
JH
2597 break;
2598 }
d6a28714
JH
2599 sayNO;
2600 case MBOL:
12d33761
HS
2601 if (locinput == PL_bostr ||
2602 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
d6a28714 2603 {
b8c5462f
JH
2604 break;
2605 }
d6a28714
JH
2606 sayNO;
2607 case SBOL:
c2a73568 2608 if (locinput == PL_bostr)
b8c5462f 2609 break;
d6a28714
JH
2610 sayNO;
2611 case GPOS:
3b0527fe 2612 if (locinput == reginfo->ganch)
d6a28714
JH
2613 break;
2614 sayNO;
2615 case EOL:
d6a28714
JH
2616 goto seol;
2617 case MEOL:
d6a28714 2618 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
b8c5462f 2619 sayNO;
b8c5462f 2620 break;
d6a28714
JH
2621 case SEOL:
2622 seol:
2623 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
b8c5462f 2624 sayNO;
d6a28714 2625 if (PL_regeol - locinput > 1)
b8c5462f 2626 sayNO;
b8c5462f 2627 break;
d6a28714
JH
2628 case EOS:
2629 if (PL_regeol != locinput)
b8c5462f 2630 sayNO;
d6a28714 2631 break;
ffc61ed2 2632 case SANY:
d6a28714 2633 if (!nextchr && locinput >= PL_regeol)
4633a7c4 2634 sayNO;
f33976b4
DB
2635 if (do_utf8) {
2636 locinput += PL_utf8skip[nextchr];
2637 if (locinput > PL_regeol)
2638 sayNO;
2639 nextchr = UCHARAT(locinput);
2640 }
2641 else
2642 nextchr = UCHARAT(++locinput);
2643 break;
2644 case CANY:
2645 if (!nextchr && locinput >= PL_regeol)
2646 sayNO;
b8c5462f 2647 nextchr = UCHARAT(++locinput);
a0d0e21e 2648 break;
ffc61ed2 2649 case REG_ANY:
1aa99e6b
IH
2650 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2651 sayNO;
2652 if (do_utf8) {
b8c5462f 2653 locinput += PL_utf8skip[nextchr];
d6a28714
JH
2654 if (locinput > PL_regeol)
2655 sayNO;
a0ed51b3 2656 nextchr = UCHARAT(locinput);
a0ed51b3 2657 }
1aa99e6b
IH
2658 else
2659 nextchr = UCHARAT(++locinput);
a0ed51b3 2660 break;
a3621e74
YO
2661
2662
2663
2664 /*
2665 traverse the TRIE keeping track of all accepting states
2666 we transition through until we get to a failing node.
2667
a3621e74
YO
2668
2669 */
5b47454d 2670 case TRIE:
a3621e74
YO
2671 case TRIEF:
2672 case TRIEFL:
2673 {
a3621e74
YO
2674 U8 *uc = ( U8* )locinput;
2675 U32 state = 1;
2676 U16 charid = 0;
2677 U32 base = 0;
2678 UV uvc = 0;
2679 STRLEN len = 0;
2680 STRLEN foldlen = 0;
a3621e74
YO
2681 U8 *uscan = (U8*)NULL;
2682 STRLEN bufflen=0;
95b24440 2683 SV *sv_accept_buff = NULL;
5b47454d
DM
2684 const enum { trie_plain, trie_utf8, trie_uft8_fold }
2685 trie_type = do_utf8 ?
2686 (OP(scan) == TRIE ? trie_utf8 : trie_uft8_fold)
2687 : trie_plain;
2688
7087a21c
NC
2689 /* what trie are we using right now */
2690 reg_trie_data *trie
32fc9b6a 2691 = (reg_trie_data*)rex->data->data[ ARG( scan ) ];
d8319b27 2692 st->u.trie.accepted = 0; /* how many accepting states we have seen */
95b24440 2693 result = 0;
a3621e74
YO
2694
2695 while ( state && uc <= (U8*)PL_regeol ) {
2696
5b47454d 2697 if (trie->states[ state ].wordnum) {
d8319b27 2698 if (!st->u.trie.accepted ) {
5b47454d
DM
2699 ENTER;
2700 SAVETMPS;
2701 bufflen = TRIE_INITAL_ACCEPT_BUFFLEN;
2702 sv_accept_buff=newSV(bufflen *
2703 sizeof(reg_trie_accepted) - 1);
2704 SvCUR_set(sv_accept_buff,
2705 sizeof(reg_trie_accepted));
2706 SvPOK_on(sv_accept_buff);
2707 sv_2mortal(sv_accept_buff);
d8319b27 2708 st->u.trie.accept_buff =
5b47454d
DM
2709 (reg_trie_accepted*)SvPV_nolen(sv_accept_buff );
2710 }
2711 else {
d8319b27 2712 if (st->u.trie.accepted >= bufflen) {
5b47454d 2713 bufflen *= 2;
d8319b27 2714 st->u.trie.accept_buff =(reg_trie_accepted*)
5b47454d
DM
2715 SvGROW(sv_accept_buff,
2716 bufflen * sizeof(reg_trie_accepted));
2717 }
2718 SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff)
2719 + sizeof(reg_trie_accepted));
2720 }
d8319b27
DM
2721 st->u.trie.accept_buff[st->u.trie.accepted].wordnum = trie->states[state].wordnum;
2722 st->u.trie.accept_buff[st->u.trie.accepted].endpos = uc;
2723 ++st->u.trie.accepted;
5b47454d 2724 }
a3621e74
YO
2725
2726 base = trie->states[ state ].trans.base;
2727
2728 DEBUG_TRIE_EXECUTE_r(
2729 PerlIO_printf( Perl_debug_log,
e4584336 2730 "%*s %sState: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ",
a3621e74 2731 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
d8319b27 2732 (UV)state, (UV)base, (UV)st->u.trie.accepted );
a3621e74
YO
2733 );
2734
2735 if ( base ) {
5b47454d
DM
2736 switch (trie_type) {
2737 case trie_uft8_fold:
a3621e74
YO
2738 if ( foldlen>0 ) {
2739 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags );
2740 foldlen -= len;
2741 uscan += len;
2742 len=0;
2743 } else {
1df70142 2744 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
a3621e74
YO
2745 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags );
2746 uvc = to_uni_fold( uvc, foldbuf, &foldlen );
2747 foldlen -= UNISKIP( uvc );
2748 uscan = foldbuf + UNISKIP( uvc );
2749 }
5b47454d
DM
2750 break;
2751 case trie_utf8:
2752 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN,
2753 &len, uniflags );
2754 break;
2755 case trie_plain:
e4584336 2756 uvc = (UV)*uc;
a3621e74
YO
2757 len = 1;
2758 }
2759
5b47454d
DM
2760 if (uvc < 256) {
2761 charid = trie->charmap[ uvc ];
2762 }
2763 else {
2764 charid = 0;
2765 if (trie->widecharmap) {
2766 SV** svpp = (SV**)NULL;
2767 svpp = hv_fetch(trie->widecharmap,
2768 (char*)&uvc, sizeof(UV), 0);
2769 if (svpp)
2770 charid = (U16)SvIV(*svpp);
2771 }
2772 }
a3621e74 2773
5b47454d
DM
2774 if (charid &&
2775 (base + charid > trie->uniquecharcount )
2776 && (base + charid - 1 - trie->uniquecharcount
2777 < trie->lasttrans)
2778 && trie->trans[base + charid - 1 -
2779 trie->uniquecharcount].check == state)
2780 {
2781 state = trie->trans[base + charid - 1 -
2782 trie->uniquecharcount ].next;
2783 }
2784 else {
2785 state = 0;
2786 }
2787 uc += len;
2788
2789 }
2790 else {
a3621e74
YO
2791 state = 0;
2792 }
2793 DEBUG_TRIE_EXECUTE_r(
e4584336
RB
2794 PerlIO_printf( Perl_debug_log,
2795 "Charid:%3x CV:%4"UVxf" After State: %4"UVxf"%s\n",
2796 charid, uvc, (UV)state, PL_colors[5] );
a3621e74
YO
2797 );
2798 }
d8319b27 2799 if (!st->u.trie.accepted )
a3621e74 2800 sayNO;
a3621e74
YO
2801
2802 /*
2803 There was at least one accepting state that we
2804 transitioned through. Presumably the number of accepting
2805 states is going to be low, typically one or two. So we
2806 simply scan through to find the one with lowest wordnum.
2807 Once we find it, we swap the last state into its place
2808 and decrement the size. We then try to match the rest of
2809 the pattern at the point where the word ends, if we
2810 succeed then we end the loop, otherwise the loop
2811 eventually terminates once all of the accepting states
2812 have been tried.
2813 */
a3621e74 2814
d8319b27 2815 if ( st->u.trie.accepted == 1 ) {
a3621e74 2816 DEBUG_EXECUTE_r({
097eb12c 2817 SV ** const tmp = av_fetch( trie->words, st->u.trie.accept_buff[ 0 ].wordnum-1, 0 );
a3621e74
YO
2818 PerlIO_printf( Perl_debug_log,
2819 "%*s %sonly one match : #%d <%s>%s\n",
e4584336 2820 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
d8319b27 2821 st->u.trie.accept_buff[ 0 ].wordnum,
cfd0369c 2822 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",
a3621e74
YO
2823 PL_colors[5] );
2824 });
d8319b27 2825 PL_reginput = (char *)st->u.trie.accept_buff[ 0 ].endpos;
a3621e74
YO
2826 /* in this case we free tmps/leave before we call regmatch
2827 as we wont be using accept_buff again. */
2828 FREETMPS;
2829 LEAVE;
95b24440
DM
2830 REGMATCH(scan + NEXT_OFF(scan), TRIE1);
2831 /*** all unsaved local vars undefined at this point */
a3621e74
YO
2832 } else {
2833 DEBUG_EXECUTE_r(
e4584336 2834 PerlIO_printf( Perl_debug_log,"%*s %sgot %"IVdf" possible matches%s\n",
d8319b27 2835 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], (IV)st->u.trie.accepted,
a3621e74
YO
2836 PL_colors[5] );
2837 );
d8319b27 2838 while ( !result && st->u.trie.accepted-- ) {
a3621e74
YO
2839 U32 best = 0;
2840 U32 cur;
d8319b27 2841 for( cur = 1 ; cur <= st->u.trie.accepted ; cur++ ) {
e4584336
RB
2842 DEBUG_TRIE_EXECUTE_r(
2843 PerlIO_printf( Perl_debug_log,
2844 "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n",
2845 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
d8319b27
DM
2846 (IV)best, st->u.trie.accept_buff[ best ].wordnum, (IV)cur,
2847 st->u.trie.accept_buff[ cur ].wordnum, PL_colors[5] );
e4584336 2848 );
a3621e74 2849
d8319b27
DM
2850 if (st->u.trie.accept_buff[cur].wordnum <
2851 st->u.trie.accept_buff[best].wordnum)
e822a8b4 2852 best = cur;
a3621e74
YO
2853 }
2854 DEBUG_EXECUTE_r({
87830502 2855 reg_trie_data * const trie = (reg_trie_data*)
32fc9b6a 2856 rex->data->data[ARG(scan)];
d8319b27 2857 SV ** const tmp = av_fetch( trie->words, st->u.trie.accept_buff[ best ].wordnum - 1, 0 );
a3621e74 2858 PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at 0x%p%s\n",
e4584336 2859 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
d8319b27 2860 st->u.trie.accept_buff[best].wordnum,
cfd0369c 2861 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",scan,
a3621e74
YO
2862 PL_colors[5] );
2863 });
d8319b27
DM
2864 if ( best<st->u.trie.accepted ) {
2865 reg_trie_accepted tmp = st->u.trie.accept_buff[ best ];
2866 st->u.trie.accept_buff[ best ] = st->u.trie.accept_buff[ st->u.trie.accepted ];
2867 st->u.trie.accept_buff[ st->u.trie.accepted ] = tmp;
2868 best = st->u.trie.accepted;
a3621e74 2869 }
d8319b27 2870 PL_reginput = (char *)st->u.trie.accept_buff[ best ].endpos;
a3621e74
YO
2871
2872 /*
2873 as far as I can tell we only need the SAVETMPS/FREETMPS
2874 for re's with EVAL in them but I'm leaving them in for
2875 all until I can be sure.
2876 */
2877 SAVETMPS;
95b24440
DM
2878 REGMATCH(scan + NEXT_OFF(scan), TRIE2);
2879 /*** all unsaved local vars undefined at this point */
a3621e74
YO
2880 FREETMPS;
2881 }
2882 FREETMPS;
2883 LEAVE;
2884 }
2885
95b24440 2886 if (result) {
a3621e74
YO
2887 sayYES;
2888 } else {
2889 sayNO;
2890 }
2891 }
2892 /* unreached codepoint */
95b24440
DM
2893 case EXACT: {
2894 char *s = STRING(scan);
5d9a96ca 2895 st->ln = STR_LEN(scan);
eb160463 2896 if (do_utf8 != UTF) {
bc517b45 2897 /* The target and the pattern have differing utf8ness. */
1aa99e6b 2898 char *l = locinput;
5d9a96ca 2899 const char *e = s + st->ln;
a72c7584 2900
5ff6fc6d
JH
2901 if (do_utf8) {
2902 /* The target is utf8, the pattern is not utf8. */
1aa99e6b 2903 while (s < e) {
a3b680e6 2904 STRLEN ulen;
1aa99e6b 2905 if (l >= PL_regeol)
5ff6fc6d
JH
2906 sayNO;
2907 if (NATIVE_TO_UNI(*(U8*)s) !=
89ebb4a3 2908 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
041457d9 2909 uniflags))
5ff6fc6d 2910 sayNO;
bc517b45 2911 l += ulen;
5ff6fc6d 2912 s ++;
1aa99e6b 2913 }
5ff6fc6d
JH
2914 }
2915 else {
2916 /* The target is not utf8, the pattern is utf8. */
1aa99e6b 2917 while (s < e) {
a3b680e6 2918 STRLEN ulen;
1aa99e6b
IH
2919 if (l >= PL_regeol)
2920 sayNO;
5ff6fc6d 2921 if (NATIVE_TO_UNI(*((U8*)l)) !=
89ebb4a3 2922 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
041457d9 2923 uniflags))
1aa99e6b 2924 sayNO;
bc517b45 2925 s += ulen;
a72c7584 2926 l ++;
1aa99e6b 2927 }
5ff6fc6d 2928 }
1aa99e6b
IH
2929 locinput = l;
2930 nextchr = UCHARAT(locinput);
2931 break;
2932 }
bc517b45 2933 /* The target and the pattern have the same utf8ness. */
d6a28714
JH
2934 /* Inline the first character, for speed. */
2935 if (UCHARAT(s) != nextchr)
2936 sayNO;
5d9a96ca 2937 if (PL_regeol - locinput < st->ln)
d6a28714 2938 sayNO;
5d9a96ca 2939 if (st->ln > 1 && memNE(s, locinput, st->ln))
d6a28714 2940 sayNO;
5d9a96ca 2941 locinput += st->ln;
d6a28714
JH
2942 nextchr = UCHARAT(locinput);
2943 break;
95b24440 2944 }
d6a28714 2945 case EXACTFL:
b8c5462f
JH
2946 PL_reg_flags |= RF_tainted;
2947 /* FALL THROUGH */
95b24440
DM
2948 case EXACTF: {
2949 char *s = STRING(scan);
5d9a96ca 2950 st->ln = STR_LEN(scan);
d6a28714 2951
d07ddd77
JH
2952 if (do_utf8 || UTF) {
2953 /* Either target or the pattern are utf8. */
d6a28714 2954 char *l = locinput;
d07ddd77 2955 char *e = PL_regeol;
bc517b45 2956
5d9a96ca 2957 if (ibcmp_utf8(s, 0, st->ln, (bool)UTF,
1feea2c7 2958 l, &e, 0, do_utf8)) {
5486206c
JH
2959 /* One more case for the sharp s:
2960 * pack("U0U*", 0xDF) =~ /ss/i,
2961 * the 0xC3 0x9F are the UTF-8
2962 * byte sequence for the U+00DF. */
2963 if (!(do_utf8 &&
2964 toLOWER(s[0]) == 's' &&
5d9a96ca 2965 st->ln >= 2 &&
5486206c
JH
2966 toLOWER(s[1]) == 's' &&
2967 (U8)l[0] == 0xC3 &&
2968 e - l >= 2 &&
2969 (U8)l[1] == 0x9F))
2970 sayNO;
2971 }
d07ddd77
JH
2972 locinput = e;
2973 nextchr = UCHARAT(locinput);
2974 break;
a0ed51b3 2975 }
d6a28714 2976
bc517b45
JH
2977 /* Neither the target and the pattern are utf8. */
2978
d6a28714
JH
2979 /* Inline the first character, for speed. */
2980 if (UCHARAT(s) != nextchr &&
2981 UCHARAT(s) != ((OP(scan) == EXACTF)
2982 ? PL_fold : PL_fold_locale)[nextchr])
a0ed51b3 2983 sayNO;
5d9a96ca 2984 if (PL_regeol - locinput < st->ln)
b8c5462f 2985 sayNO;
5d9a96ca
DM
2986 if (st->ln > 1 && (OP(scan) == EXACTF
2987 ? ibcmp(s, locinput, st->ln)
2988 : ibcmp_locale(s, locinput, st->ln)))
4633a7c4 2989 sayNO;
5d9a96ca 2990 locinput += st->ln;
d6a28714 2991 nextchr = UCHARAT(locinput);
a0d0e21e 2992 break;
95b24440 2993 }
d6a28714 2994 case ANYOF:
ffc61ed2 2995 if (do_utf8) {
9e55ce06
JH
2996 STRLEN inclasslen = PL_regeol - locinput;
2997
32fc9b6a 2998 if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8))
e0f9d4a8 2999 sayNO_ANYOF;
ffc61ed2
JH
3000 if (locinput >= PL_regeol)
3001 sayNO;
0f0076b4 3002 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
b8c5462f 3003 nextchr = UCHARAT(locinput);
e0f9d4a8 3004 break;
ffc61ed2
JH
3005 }
3006 else {
3007 if (nextchr < 0)
3008 nextchr = UCHARAT(locinput);
32fc9b6a 3009 if (!REGINCLASS(rex, scan, (U8*)locinput))
e0f9d4a8 3010 sayNO_ANYOF;
ffc61ed2
JH
3011 if (!nextchr && locinput >= PL_regeol)
3012 sayNO;
3013 nextchr = UCHARAT(++locinput);
e0f9d4a8
JH
3014 break;
3015 }
3016 no_anyof:
3017 /* If we might have the case of the German sharp s
3018 * in a casefolding Unicode character class. */
3019
ebc501f0
JH
3020 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
3021 locinput += SHARP_S_SKIP;
e0f9d4a8 3022 nextchr = UCHARAT(locinput);
ffc61ed2 3023 }
e0f9d4a8
JH
3024 else
3025 sayNO;
b8c5462f 3026 break;
d6a28714 3027 case ALNUML:
b8c5462f
JH
3028 PL_reg_flags |= RF_tainted;
3029 /* FALL THROUGH */
d6a28714 3030 case ALNUM:
b8c5462f 3031 if (!nextchr)
4633a7c4 3032 sayNO;
ffc61ed2 3033 if (do_utf8) {
1a4fad37 3034 LOAD_UTF8_CHARCLASS_ALNUM();
ffc61ed2 3035 if (!(OP(scan) == ALNUM
3568d838 3036 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
d6a28714 3037 : isALNUM_LC_utf8((U8*)locinput)))
b8c5462f
JH
3038 {
3039 sayNO;
a0ed51b3 3040 }
b8c5462f 3041 locinput += PL_utf8skip[nextchr];
a0ed51b3
LW
3042 nextchr = UCHARAT(locinput);
3043 break;
3044 }
ffc61ed2 3045 if (!(OP(scan) == ALNUM
d6a28714 3046 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
bbce6d69 3047 sayNO;
b8c5462f 3048 nextchr = UCHARAT(++locinput);
a0d0e21e 3049 break;
d6a28714 3050 case NALNUML:
b8c5462f
JH
3051 PL_reg_flags |= RF_tainted;
3052 /* FALL THROUGH */
d6a28714
JH
3053 case NALNUM:
3054 if (!nextchr && locinput >= PL_regeol)
a0ed51b3 3055 sayNO;
ffc61ed2 3056 if (do_utf8) {
1a4fad37 3057 LOAD_UTF8_CHARCLASS_ALNUM();
ffc61ed2 3058 if (OP(scan) == NALNUM
3568d838 3059 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
d6a28714
JH
3060 : isALNUM_LC_utf8((U8*)locinput))
3061 {
b8c5462f 3062 sayNO;
d6a28714 3063 }
b8c5462f
JH
3064 locinput += PL_utf8skip[nextchr];
3065 nextchr = UCHARAT(locinput);
3066 break;
3067 }
ffc61ed2 3068 if (OP(scan) == NALNUM
d6a28714 3069 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
b8c5462f 3070 sayNO;
76e3520e 3071 nextchr = UCHARAT(++locinput);
a0d0e21e 3072 break;
d6a28714
JH
3073 case BOUNDL:
3074 case NBOUNDL:
3280af22 3075 PL_reg_flags |= RF_tainted;
bbce6d69 3076 /* FALL THROUGH */
d6a28714
JH
3077 case BOUND:
3078 case NBOUND:
3079 /* was last char in word? */
ffc61ed2 3080 if (do_utf8) {
12d33761 3081 if (locinput == PL_bostr)
5d9a96ca 3082 st->ln = '\n';
ffc61ed2 3083 else {
a3b680e6 3084 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
9041c2e3 3085
4ad0818d 3086 st->ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
ffc61ed2
JH
3087 }
3088 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
5d9a96ca 3089 st->ln = isALNUM_uni(st->ln);
1a4fad37 3090 LOAD_UTF8_CHARCLASS_ALNUM();
3568d838 3091 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
ffc61ed2
JH
3092 }
3093 else {
5d9a96ca 3094 st->ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(st->ln));
ffc61ed2
JH
3095 n = isALNUM_LC_utf8((U8*)locinput);
3096 }
a0ed51b3 3097 }
d6a28714 3098 else {
5d9a96ca 3099 st->ln = (locinput != PL_bostr) ?
12d33761 3100 UCHARAT(locinput - 1) : '\n';
ffc61ed2 3101 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
5d9a96ca 3102 st->ln = isALNUM(st->ln);
ffc61ed2
JH
3103 n = isALNUM(nextchr);
3104 }
3105 else {
5d9a96ca 3106 st->ln = isALNUM_LC(st->ln);
ffc61ed2
JH
3107 n = isALNUM_LC(nextchr);
3108 }
d6a28714 3109 }
5d9a96ca 3110 if (((!st->ln) == (!n)) == (OP(scan) == BOUND ||
ffc61ed2
JH
3111 OP(scan) == BOUNDL))
3112 sayNO;
a0ed51b3 3113 break;
d6a28714 3114 case SPACEL:
3280af22 3115 PL_reg_flags |= RF_tainted;
bbce6d69 3116 /* FALL THROUGH */
d6a28714 3117 case SPACE:
9442cb0e 3118 if (!nextchr)
4633a7c4 3119 sayNO;
1aa99e6b 3120 if (do_utf8) {
fd400ab9 3121 if (UTF8_IS_CONTINUED(nextchr)) {
1a4fad37 3122 LOAD_UTF8_CHARCLASS_SPACE();
ffc61ed2 3123 if (!(OP(scan) == SPACE
3568d838 3124 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
ffc61ed2
JH
3125 : isSPACE_LC_utf8((U8*)locinput)))
3126 {
3127 sayNO;
3128 }
3129 locinput += PL_utf8skip[nextchr];
3130 nextchr = UCHARAT(locinput);
3131 break;
d6a28714 3132 }
ffc61ed2
JH
3133 if (!(OP(scan) == SPACE
3134 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3135 sayNO;
3136 nextchr = UCHARAT(++locinput);
3137 }
3138 else {
3139 if (!(OP(scan) == SPACE
3140 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3141 sayNO;
3142 nextchr = UCHARAT(++locinput);
a0ed51b3 3143 }
a0ed51b3 3144 break;
d6a28714 3145 case NSPACEL:
3280af22 3146 PL_reg_flags |= RF_tainted;
bbce6d69 3147 /* FALL THROUGH */
d6a28714 3148 case NSPACE:
9442cb0e 3149 if (!nextchr && locinput >= PL_regeol)
b8c5462f 3150 sayNO;
1aa99e6b 3151 if (do_utf8) {
1a4fad37 3152 LOAD_UTF8_CHARCLASS_SPACE();
ffc61ed2 3153 if (OP(scan) == NSPACE
3568d838 3154 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
d6a28714 3155 : isSPACE_LC_utf8((U8*)locinput))
b8c5462f
JH
3156 {
3157 sayNO;
3158 }
3159 locinput += PL_utf8skip[nextchr];
3160 nextchr = UCHARAT(locinput);
3161 break;
a0ed51b3 3162 }
ffc61ed2 3163 if (OP(scan) == NSPACE
d6a28714 3164 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
4633a7c4 3165 sayNO;
76e3520e 3166 nextchr = UCHARAT(++locinput);
a0d0e21e 3167 break;
d6a28714 3168 case DIGITL:
a0ed51b3
LW
3169 PL_reg_flags |= RF_tainted;
3170 /* FALL THROUGH */
d6a28714 3171 case DIGIT:
9442cb0e 3172 if (!nextchr)
a0ed51b3 3173 sayNO;
1aa99e6b 3174 if (do_utf8) {
1a4fad37 3175 LOAD_UTF8_CHARCLASS_DIGIT();
ffc61ed2 3176 if (!(OP(scan) == DIGIT
3568d838 3177 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
9442cb0e 3178 : isDIGIT_LC_utf8((U8*)locinput)))
dfe13c55 3179 {
a0ed51b3 3180 sayNO;
dfe13c55 3181 }
6f06b55f 3182 locinput += PL_utf8skip[nextchr];
a0ed51b3
LW
3183 nextchr = UCHARAT(locinput);
3184 break;
3185 }
ffc61ed2 3186 if (!(OP(scan) == DIGIT
9442cb0e 3187 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
a0ed51b3
LW
3188 sayNO;
3189 nextchr = UCHARAT(++locinput);
3190 break;
d6a28714 3191 case NDIGITL:
b8c5462f
JH
3192 PL_reg_flags |= RF_tainted;
3193 /* FALL THROUGH */
d6a28714 3194 case NDIGIT:
9442cb0e 3195 if (!nextchr && locinput >= PL_regeol)
b8c5462f 3196 sayNO;
1aa99e6b 3197 if (do_utf8) {
1a4fad37 3198 LOAD_UTF8_CHARCLASS_DIGIT();
ffc61ed2 3199 if (OP(scan) == NDIGIT
3568d838 3200 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
9442cb0e
GS
3201 : isDIGIT_LC_utf8((U8*)locinput))
3202 {
a0ed51b3 3203 sayNO;
9442cb0e 3204 }
6f06b55f 3205 locinput += PL_utf8skip[nextchr];
a0ed51b3
LW
3206 nextchr = UCHARAT(locinput);
3207 break;
3208 }
ffc61ed2 3209 if (OP(scan) == NDIGIT
9442cb0e 3210 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
a0ed51b3
LW
3211 sayNO;
3212 nextchr = UCHARAT(++locinput);
3213 break;
3214 case CLUMP:
b7c83a7e 3215 if (locinput >= PL_regeol)
a0ed51b3 3216 sayNO;
b7c83a7e 3217 if (do_utf8) {
1a4fad37 3218 LOAD_UTF8_CHARCLASS_MARK();
b7c83a7e
JH
3219 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3220 sayNO;
3221 locinput += PL_utf8skip[nextchr];
3222 while (locinput < PL_regeol &&
3223 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3224 locinput += UTF8SKIP(locinput);
3225 if (locinput > PL_regeol)
3226 sayNO;
eb08e2da
JH
3227 }
3228 else
3229 locinput++;
a0ed51b3
LW
3230 nextchr = UCHARAT(locinput);
3231 break;
c8756f30 3232 case REFFL:
3280af22 3233 PL_reg_flags |= RF_tainted;
c8756f30 3234 /* FALL THROUGH */
c277df42 3235 case REF:
95b24440
DM
3236 case REFF: {
3237 char *s;
c277df42 3238 n = ARG(scan); /* which paren pair */
5d9a96ca 3239 st->ln = PL_regstartp[n];
2c2d71f5 3240 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
5d9a96ca 3241 if ((I32)*PL_reglastparen < n || st->ln == -1)
af3f8c16 3242 sayNO; /* Do not match unless seen CLOSEn. */
5d9a96ca 3243 if (st->ln == PL_regendp[n])
a0d0e21e 3244 break;
a0ed51b3 3245
5d9a96ca 3246 s = PL_bostr + st->ln;
1aa99e6b 3247 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
a0ed51b3 3248 char *l = locinput;
a3b680e6 3249 const char *e = PL_bostr + PL_regendp[n];
a0ed51b3
LW
3250 /*
3251 * Note that we can't do the "other character" lookup trick as
3252 * in the 8-bit case (no pun intended) because in Unicode we
3253 * have to map both upper and title case to lower case.
3254 */
3255 if (OP(scan) == REFF) {
3256 while (s < e) {
a3b680e6
AL
3257 STRLEN ulen1, ulen2;
3258 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3259 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3260
a0ed51b3
LW
3261 if (l >= PL_regeol)
3262 sayNO;
a2a2844f
JH
3263 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3264 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
7114a2d2 3265 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
a0ed51b3 3266 sayNO;
a2a2844f
JH
3267 s += ulen1;
3268 l += ulen2;
a0ed51b3
LW
3269 }
3270 }
3271 locinput = l;
3272 nextchr = UCHARAT(locinput);
3273 break;
3274 }
3275
a0d0e21e 3276 /* Inline the first character, for speed. */
76e3520e 3277 if (UCHARAT(s) != nextchr &&
c8756f30
AK
3278 (OP(scan) == REF ||
3279 (UCHARAT(s) != ((OP(scan) == REFF
22c35a8c 3280 ? PL_fold : PL_fold_locale)[nextchr]))))
4633a7c4 3281 sayNO;
5d9a96ca
DM
3282 st->ln = PL_regendp[n] - st->ln;
3283 if (locinput + st->ln > PL_regeol)
4633a7c4 3284 sayNO;
5d9a96ca
DM
3285 if (st->ln > 1 && (OP(scan) == REF
3286 ? memNE(s, locinput, st->ln)
c8756f30 3287 : (OP(scan) == REFF
5d9a96ca
DM
3288 ? ibcmp(s, locinput, st->ln)
3289 : ibcmp_locale(s, locinput, st->ln))))
4633a7c4 3290 sayNO;
5d9a96ca 3291 locinput += st->ln;
76e3520e 3292 nextchr = UCHARAT(locinput);
a0d0e21e 3293 break;
95b24440 3294 }
a0d0e21e
LW
3295
3296 case NOTHING:
c277df42 3297 case TAIL:
a0d0e21e
LW
3298 break;
3299 case BACK:
3300 break;
c277df42
IZ
3301 case EVAL:
3302 {
c277df42 3303 SV *ret;
8e5e9ebe 3304 {
4aabdb9b
DM
3305 /* execute the code in the {...} */
3306 dSP;
6136c704 3307 SV ** const before = SP;
4aabdb9b
DM
3308 OP_4tree * const oop = PL_op;
3309 COP * const ocurcop = PL_curcop;
3310 PAD *old_comppad;
4aabdb9b
DM
3311
3312 n = ARG(scan);
32fc9b6a 3313 PL_op = (OP_4tree*)rex->data->data[n];
4aabdb9b 3314 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
32fc9b6a 3315 PAD_SAVE_LOCAL(old_comppad, (PAD*)rex->data->data[n + 2]);
4aabdb9b
DM
3316 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
3317
8e5e9ebe
RGS
3318 CALLRUNOPS(aTHX); /* Scalar context. */
3319 SPAGAIN;
3320 if (SP == before)
075aa684 3321 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
8e5e9ebe
RGS
3322 else {
3323 ret = POPs;
3324 PUTBACK;
3325 }
4aabdb9b
DM
3326
3327 PL_op = oop;
3328 PAD_RESTORE_LOCAL(old_comppad);
3329 PL_curcop = ocurcop;
3330 if (!st->logical) {
3331 /* /(?{...})/ */
3332 sv_setsv(save_scalar(PL_replgv), ret);
4aabdb9b
DM
3333 break;
3334 }
8e5e9ebe 3335 }
4aabdb9b
DM
3336 if (st->logical == 2) { /* Postponed subexpression: /(??{...})/ */
3337 regexp *re;
4aabdb9b 3338 {
4f639d21
DM
3339 /* extract RE object from returned value; compiling if
3340 * necessary */
3341
6136c704 3342 MAGIC *mg = NULL;
4aabdb9b 3343 SV *sv;
faf82a0b
AE
3344 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
3345 mg = mg_find(sv, PERL_MAGIC_qr);
3346 else if (SvSMAGICAL(ret)) {
3347 if (SvGMAGICAL(ret))
3348 sv_unmagic(ret, PERL_MAGIC_qr);
3349 else
3350 mg = mg_find(ret, PERL_MAGIC_qr);
0f5d15d6 3351 }
faf82a0b 3352
0f5d15d6
IZ
3353 if (mg) {
3354 re = (regexp *)mg->mg_obj;
df0003d4 3355 (void)ReREFCNT_inc(re);
0f5d15d6
IZ
3356 }
3357 else {
3358 STRLEN len;
6136c704 3359 const char * const t = SvPV_const(ret, len);
0f5d15d6 3360 PMOP pm;
a3b680e6 3361 const I32 osize = PL_regsize;
0f5d15d6 3362
5fcd1c1b 3363 Zero(&pm, 1, PMOP);
4aabdb9b 3364 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
83003860 3365 re = CALLREGCOMP(aTHX_ (char*)t, (char*)t + len, &pm);
9041c2e3 3366 if (!(SvFLAGS(ret)
faf82a0b
AE
3367 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
3368 | SVs_GMG)))
14befaf4
DM
3369 sv_magic(ret,(SV*)ReREFCNT_inc(re),
3370 PERL_MAGIC_qr,0,0);
0f5d15d6 3371 PL_regsize = osize;
0f5d15d6 3372 }
4aabdb9b 3373 }
aa283a38
DM
3374
3375 /* run the pattern returned from (??{...}) */
3376
4aabdb9b
DM
3377 DEBUG_EXECUTE_r(
3378 PerlIO_printf(Perl_debug_log,
3379 "Entering embedded \"%s%.60s%s%s\"\n",
3380 PL_colors[0],
3381 re->precomp,
3382 PL_colors[1],
3383 (strlen(re->precomp) > 60 ? "..." : ""))
3384 );
2c2d71f5 3385
4aabdb9b
DM
3386 st->u.eval.cp = regcppush(0); /* Save *all* the positions. */
3387 REGCP_SET(st->u.eval.lastcp);
4aabdb9b
DM
3388 *PL_reglastparen = 0;
3389 *PL_reglastcloseparen = 0;
4aabdb9b 3390 PL_reginput = locinput;
4aabdb9b
DM
3391
3392 /* XXXX This is too dramatic a measure... */
3393 PL_reg_maxiter = 0;
3394
5d9a96ca 3395 st->logical = 0;
aa283a38
DM
3396 st->u.eval.toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
3397 ((re->reganch & ROPT_UTF8) != 0);
3398 if (st->u.eval.toggleutf) PL_reg_flags ^= RF_utf8;
3399 st->u.eval.prev_rex = rex;
aa283a38 3400 rex = re;
aa283a38
DM
3401
3402 st->u.eval.prev_eval = cur_eval;
3403 st->u.eval.prev_slab = PL_regmatch_slab;
3404 st->u.eval.depth = depth;
3405 cur_eval = st;
3406 PUSH_STATE(newst, resume_EVAL);
3407 st = newst;
3408
3409 /* now continue from first node in postoned RE */
3410 next = re->program + 1;
3411 break;
4aabdb9b 3412 /* NOTREACHED */
a0ed51b3 3413 }
4aabdb9b
DM
3414 /* /(?(?{...})X|Y)/ */
3415 st->sw = SvTRUE(ret);
3416 st->logical = 0;
c277df42
IZ
3417 break;
3418 }
a0d0e21e 3419 case OPEN:
c277df42 3420 n = ARG(scan); /* which paren pair */
3280af22
NIS
3421 PL_reg_start_tmp[n] = locinput;
3422 if (n > PL_regsize)
3423 PL_regsize = n;
a0d0e21e
LW
3424 break;
3425 case CLOSE:
c277df42 3426 n = ARG(scan); /* which paren pair */
cf93c79d
IZ
3427 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
3428 PL_regendp[n] = locinput - PL_bostr;
eb160463 3429 if (n > (I32)*PL_reglastparen)
3280af22 3430 *PL_reglastparen = n;
a01268b5 3431 *PL_reglastcloseparen = n;
a0d0e21e 3432 break;
c277df42
IZ
3433 case GROUPP:
3434 n = ARG(scan); /* which paren pair */
5d9a96ca 3435 st->sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
c277df42
IZ
3436 break;
3437 case IFTHEN:
2c2d71f5 3438 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
5d9a96ca 3439 if (st->sw)
c277df42
IZ
3440 next = NEXTOPER(NEXTOPER(scan));
3441 else {
3442 next = scan + ARG(scan);
3443 if (OP(next) == IFTHEN) /* Fake one. */
3444 next = NEXTOPER(NEXTOPER(next));
3445 }
3446 break;
3447 case LOGICAL:
5d9a96ca 3448 st->logical = scan->flags;
c277df42 3449 break;
2ab05381 3450/*******************************************************************
a0374537
DM
3451 cc points to the regmatch_state associated with the most recent CURLYX.
3452 This struct contains info about the innermost (...)* loop (an
3453 "infoblock"), and a pointer to the next outer cc.
2ab05381
IZ
3454
3455 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3456
95b24440 3457 1) After matching Y, regnode for CURLYX is processed;
2ab05381 3458
a0374537 3459 2) This regnode populates cc, and calls regmatch() recursively
95b24440 3460 with the starting point at WHILEM node;
2ab05381
IZ
3461
3462 3) Each hit of WHILEM node tries to match A and Z (in the order
3463 depending on the current iteration, min/max of {min,max} and
3464 greediness). The information about where are nodes for "A"
a0374537 3465 and "Z" is read from cc, as is info on how many times "A"
2ab05381
IZ
3466 was already matched, and greediness.
3467
3468 4) After A matches, the same WHILEM node is hit again.
3469
95b24440 3470 5) Each time WHILEM is hit, cc is the infoblock created by CURLYX
2ab05381 3471 of the same pair. Thus when WHILEM tries to match Z, it temporarily
95b24440 3472 resets cc, since this Y(A)*Z can be a part of some other loop:
2ab05381
IZ
3473 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3474 of the external loop.
3475
a0374537 3476 Currently present infoblocks form a tree with a stem formed by st->cc
2ab05381
IZ
3477 and whatever it mentions via ->next, and additional attached trees
3478 corresponding to temporarily unset infoblocks as in "5" above.
3479
95b24440 3480 In the following picture, infoblocks for outer loop of
2ab05381
IZ
3481 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3482 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3483 infoblocks are drawn below the "reset" infoblock.
3484
3485 In fact in the picture below we do not show failed matches for Z and T
3486 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3487 more obvious *why* one needs to *temporary* unset infoblocks.]
3488
3489 Matched REx position InfoBlocks Comment
3490 (Y(A)*?Z)*?T x
3491 Y(A)*?Z)*?T x <- O
3492 Y (A)*?Z)*?T x <- O
3493 Y A)*?Z)*?T x <- O <- I
3494 YA )*?Z)*?T x <- O <- I
3495 YA A)*?Z)*?T x <- O <- I
3496 YAA )*?Z)*?T x <- O <- I
3497 YAA Z)*?T x <- O # Temporary unset I
3498 I
3499
3500 YAAZ Y(A)*?Z)*?T x <- O
3501 I
3502
3503 YAAZY (A)*?Z)*?T x <- O
3504 I
3505
3506 YAAZY A)*?Z)*?T x <- O <- I
3507 I
3508
3509 YAAZYA )*?Z)*?T x <- O <- I
3510 I
3511
3512 YAAZYA Z)*?T x <- O # Temporary unset I
3513 I,I
3514
3515 YAAZYAZ )*?T x <- O
3516 I,I
3517
3518 YAAZYAZ T x # Temporary unset O
3519 O
3520 I,I
3521
3522 YAAZYAZT x
3523 O
3524 I,I
3525 *******************************************************************/
95b24440 3526
a0d0e21e 3527 case CURLYX: {
cb434fcc
IZ
3528 /* No need to save/restore up to this paren */
3529 I32 parenfloor = scan->flags;
c277df42 3530
30b2893d 3531 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
c277df42 3532 next += ARG(next);
cb434fcc
IZ
3533 /* XXXX Probably it is better to teach regpush to support
3534 parenfloor > PL_regsize... */
eb160463 3535 if (parenfloor > (I32)*PL_reglastparen)
cb434fcc 3536 parenfloor = *PL_reglastparen; /* Pessimization... */
a0374537 3537
d8319b27
DM
3538 st->u.curlyx.cp = PL_savestack_ix;
3539 st->u.curlyx.outercc = st->cc;
a0374537
DM
3540 st->cc = st;
3541 /* these fields contain the state of the current curly.
3542 * they are accessed by subsequent WHILEMs;
3543 * cur and lastloc are also updated by WHILEM */
d8319b27
DM
3544 st->u.curlyx.parenfloor = parenfloor;
3545 st->u.curlyx.cur = -1; /* this will be updated by WHILEM */
3546 st->u.curlyx.min = ARG1(scan);
3547 st->u.curlyx.max = ARG2(scan);
3548 st->u.curlyx.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3549 st->u.curlyx.lastloc = 0;
a0374537
DM
3550 /* st->next and st->minmod are also read by WHILEM */
3551
3280af22 3552 PL_reginput = locinput;
95b24440
DM
3553 REGMATCH(PREVOPER(next), CURLYX); /* start on the WHILEM */
3554 /*** all unsaved local vars undefined at this point */
d8319b27
DM
3555 regcpblow(st->u.curlyx.cp);
3556 st->cc = st->u.curlyx.outercc;
95b24440 3557 saySAME(result);
a0d0e21e 3558 }
5f66b61c 3559 /* NOTREACHED */
a0d0e21e
LW
3560 case WHILEM: {
3561 /*
3562 * This is really hard to understand, because after we match
3563 * what we're trying to match, we must make sure the rest of
2c2d71f5 3564 * the REx is going to match for sure, and to do that we have
a0d0e21e
LW
3565 * to go back UP the parse tree by recursing ever deeper. And
3566 * if it fails, we have to reset our parent's current state
3567 * that we can try again after backing off.
3568 */
3569
d8319b27
DM
3570 st->u.whilem.lastloc = st->cc->u.curlyx.lastloc; /* Detection of 0-len. */
3571 st->u.whilem.cache_offset = 0;
3572 st->u.whilem.cache_bit = 0;
c277df42 3573
d8319b27 3574 n = st->cc->u.curlyx.cur + 1; /* how many we know we matched */
3280af22 3575 PL_reginput = locinput;
a0d0e21e 3576
a3621e74 3577 DEBUG_EXECUTE_r(
9041c2e3 3578 PerlIO_printf(Perl_debug_log,
91f3b821 3579 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3280af22 3580 REPORT_CODE_OFF+PL_regindent*2, "",
d8319b27
DM
3581 (long)n, (long)st->cc->u.curlyx.min,
3582 (long)st->cc->u.curlyx.max, PTR2UV(st->cc))
c277df42 3583 );
4633a7c4 3584
a0d0e21e
LW
3585 /* If degenerate scan matches "", assume scan done. */
3586
d8319b27
DM
3587 if (locinput == st->cc->u.curlyx.lastloc && n >= st->cc->u.curlyx.min) {
3588 st->u.whilem.savecc = st->cc;
3589 st->cc = st->cc->u.curlyx.outercc;
5d9a96ca 3590 if (st->cc)
d8319b27 3591 st->ln = st->cc->u.curlyx.cur;
a3621e74 3592 DEBUG_EXECUTE_r(
c3464db5
DD
3593 PerlIO_printf(Perl_debug_log,
3594 "%*s empty match detected, try continuation...\n",
3280af22 3595 REPORT_CODE_OFF+PL_regindent*2, "")
c277df42 3596 );
d8319b27 3597 REGMATCH(st->u.whilem.savecc->next, WHILEM1);
95b24440 3598 /*** all unsaved local vars undefined at this point */
d8319b27 3599 st->cc = st->u.whilem.savecc;
95b24440 3600 if (result)
4633a7c4 3601 sayYES;
d8319b27
DM
3602 if (st->cc->u.curlyx.outercc)
3603 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
4633a7c4 3604 sayNO;
a0d0e21e
LW
3605 }
3606
3607 /* First just match a string of min scans. */
3608
d8319b27
DM
3609 if (n < st->cc->u.curlyx.min) {
3610 st->cc->u.curlyx.cur = n;
3611 st->cc->u.curlyx.lastloc = locinput;
3612 REGMATCH(st->cc->u.curlyx.scan, WHILEM2);
95b24440
DM
3613 /*** all unsaved local vars undefined at this point */
3614 if (result)
4633a7c4 3615 sayYES;
d8319b27
DM
3616 st->cc->u.curlyx.cur = n - 1;
3617 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
4633a7c4 3618 sayNO;
a0d0e21e
LW
3619 }
3620
2c2d71f5
JH
3621 if (scan->flags) {
3622 /* Check whether we already were at this position.
3623 Postpone detection until we know the match is not
3624 *that* much linear. */
3625 if (!PL_reg_maxiter) {
3626 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3627 PL_reg_leftiter = PL_reg_maxiter;
3628 }
3629 if (PL_reg_leftiter-- == 0) {
a3b680e6 3630 const I32 size = (PL_reg_maxiter + 7 + POSCACHE_START)/8;
2c2d71f5 3631 if (PL_reg_poscache) {
eb160463 3632 if ((I32)PL_reg_poscache_size < size) {
2c2d71f5
JH
3633 Renew(PL_reg_poscache, size, char);
3634 PL_reg_poscache_size = size;
3635 }
3636 Zero(PL_reg_poscache, size, char);
3637 }
3638 else {
3639 PL_reg_poscache_size = size;
a02a5408 3640 Newxz(PL_reg_poscache, size, char);
2c2d71f5 3641 }
a3621e74 3642 DEBUG_EXECUTE_r(
2c2d71f5
JH
3643 PerlIO_printf(Perl_debug_log,
3644 "%sDetected a super-linear match, switching on caching%s...\n",
3645 PL_colors[4], PL_colors[5])
3646 );
3647 }
3648 if (PL_reg_leftiter < 0) {
d8319b27 3649 st->u.whilem.cache_offset = locinput - PL_bostr;
2c2d71f5 3650
d8319b27
DM
3651 st->u.whilem.cache_offset = (scan->flags & 0xf) - 1 + POSCACHE_START
3652 + st->u.whilem.cache_offset * (scan->flags>>4);
3653 st->u.whilem.cache_bit = st->u.whilem.cache_offset % 8;
3654 st->u.whilem.cache_offset /= 8;
3655 if (PL_reg_poscache[st->u.whilem.cache_offset] & (1<<st->u.whilem.cache_bit)) {
a3621e74 3656 DEBUG_EXECUTE_r(
2c2d71f5
JH
3657 PerlIO_printf(Perl_debug_log,
3658 "%*s already tried at this position...\n",
3659 REPORT_CODE_OFF+PL_regindent*2, "")
3660 );
3ab3c9b4
HS
3661 if (PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))
3662 /* cache records success */
c2b0868c
HS
3663 sayYES;
3664 else
3ab3c9b4 3665 /* cache records failure */
c2b0868c 3666 sayNO_SILENT;
2c2d71f5 3667 }
d8319b27 3668 PL_reg_poscache[st->u.whilem.cache_offset] |= (1<<st->u.whilem.cache_bit);
2c2d71f5
JH
3669 }
3670 }
3671
a0d0e21e
LW
3672 /* Prefer next over scan for minimal matching. */
3673
5d9a96ca 3674 if (st->cc->minmod) {
d8319b27
DM
3675 st->u.whilem.savecc = st->cc;
3676 st->cc = st->cc->u.curlyx.outercc;
5d9a96ca 3677 if (st->cc)
d8319b27
DM
3678 st->ln = st->cc->u.curlyx.cur;
3679 st->u.whilem.cp = regcppush(st->u.whilem.savecc->u.curlyx.parenfloor);
3680 REGCP_SET(st->u.whilem.lastcp);
3681 REGMATCH(st->u.whilem.savecc->next, WHILEM3);
95b24440 3682 /*** all unsaved local vars undefined at this point */
d8319b27 3683 st->cc = st->u.whilem.savecc;
95b24440 3684 if (result) {
d8319b27 3685 regcpblow(st->u.whilem.cp);
3ab3c9b4 3686 CACHEsayYES; /* All done. */
5f05dabc 3687 }
d8319b27 3688 REGCP_UNWIND(st->u.whilem.lastcp);
4f639d21 3689 regcppop(rex);
d8319b27
DM
3690 if (st->cc->u.curlyx.outercc)
3691 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
a0d0e21e 3692
d8319b27 3693 if (n >= st->cc->u.curlyx.max) { /* Maximum greed exceeded? */
9041c2e3 3694 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3280af22
NIS
3695 && !(PL_reg_flags & RF_warned)) {
3696 PL_reg_flags |= RF_warned;
9014280d 3697 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
2f3ca594
GS
3698 "Complex regular subexpression recursion",
3699 REG_INFTY - 1);
c277df42 3700 }
3ab3c9b4 3701 CACHEsayNO;
c277df42 3702 }
a687059c 3703
a3621e74 3704 DEBUG_EXECUTE_r(
c3464db5
DD
3705 PerlIO_printf(Perl_debug_log,
3706 "%*s trying longer...\n",
3280af22 3707 REPORT_CODE_OFF+PL_regindent*2, "")
c277df42 3708 );
a0d0e21e 3709 /* Try scanning more and see if it helps. */
3280af22 3710 PL_reginput = locinput;
d8319b27
DM
3711 st->cc->u.curlyx.cur = n;
3712 st->cc->u.curlyx.lastloc = locinput;
3713 st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor);
3714 REGCP_SET(st->u.whilem.lastcp);
3715 REGMATCH(st->cc->u.curlyx.scan, WHILEM4);
95b24440
DM
3716 /*** all unsaved local vars undefined at this point */
3717 if (result) {
d8319b27 3718 regcpblow(st->u.whilem.cp);
3ab3c9b4 3719 CACHEsayYES;
5f05dabc 3720 }
d8319b27 3721 REGCP_UNWIND(st->u.whilem.lastcp);
4f639d21 3722 regcppop(rex);
d8319b27
DM
3723 st->cc->u.curlyx.cur = n - 1;
3724 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3ab3c9b4 3725 CACHEsayNO;
a0d0e21e
LW
3726 }
3727
3728 /* Prefer scan over next for maximal matching. */
3729
d8319b27
DM
3730 if (n < st->cc->u.curlyx.max) { /* More greed allowed? */
3731 st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor);
3732 st->cc->u.curlyx.cur = n;
3733 st->cc->u.curlyx.lastloc = locinput;
3734 REGCP_SET(st->u.whilem.lastcp);
3735 REGMATCH(st->cc->u.curlyx.scan, WHILEM5);
95b24440
DM
3736 /*** all unsaved local vars undefined at this point */
3737 if (result) {
d8319b27 3738 regcpblow(st->u.whilem.cp);
3ab3c9b4 3739 CACHEsayYES;
5f05dabc 3740 }
d8319b27 3741 REGCP_UNWIND(st->u.whilem.lastcp);
4f639d21 3742 regcppop(rex); /* Restore some previous $<digit>s? */
3280af22 3743 PL_reginput = locinput;
a3621e74 3744 DEBUG_EXECUTE_r(
c3464db5
DD
3745 PerlIO_printf(Perl_debug_log,
3746 "%*s failed, try continuation...\n",
3280af22 3747 REPORT_CODE_OFF+PL_regindent*2, "")
c277df42
IZ
3748 );
3749 }
9041c2e3 3750 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
599cee73 3751 && !(PL_reg_flags & RF_warned)) {
3280af22 3752 PL_reg_flags |= RF_warned;
9014280d 3753 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
cb5d145d
GS
3754 "Complex regular subexpression recursion",
3755 REG_INFTY - 1);
a0d0e21e
LW
3756 }
3757
3758 /* Failed deeper matches of scan, so see if this one works. */
d8319b27
DM
3759 st->u.whilem.savecc = st->cc;
3760 st->cc = st->cc->u.curlyx.outercc;
5d9a96ca 3761 if (st->cc)
d8319b27
DM
3762 st->ln = st->cc->u.curlyx.cur;
3763 REGMATCH(st->u.whilem.savecc->next, WHILEM6);
95b24440 3764 /*** all unsaved local vars undefined at this point */
d8319b27 3765 st->cc = st->u.whilem.savecc;
95b24440 3766 if (result)
3ab3c9b4 3767 CACHEsayYES;
d8319b27
DM
3768 if (st->cc->u.curlyx.outercc)
3769 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3770 st->cc->u.curlyx.cur = n - 1;
3771 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3ab3c9b4 3772 CACHEsayNO;
a0d0e21e 3773 }
5f66b61c 3774 /* NOTREACHED */
9041c2e3 3775 case BRANCHJ:
c277df42
IZ
3776 next = scan + ARG(scan);
3777 if (next == scan)
3778 next = NULL;
3779 inner = NEXTOPER(NEXTOPER(scan));
3780 goto do_branch;
9041c2e3 3781 case BRANCH:
c277df42
IZ
3782 inner = NEXTOPER(scan);
3783 do_branch:
3784 {
e822a8b4
DM
3785 I32 type;
3786 type = OP(scan);
ae5031b3 3787 if (!next || OP(next) != type) /* No choice. */
c277df42 3788 next = inner; /* Avoid recursion. */
a0d0e21e 3789 else {
a3b680e6 3790 const I32 lastparen = *PL_reglastparen;
02db2b7b 3791 /* Put unwinding data on stack */
6136c704
AL
3792 const I32 unwind1 = SSNEWt(1,re_unwind_branch_t);
3793 re_unwind_branch_t * const uw = SSPTRt(unwind1,re_unwind_branch_t);
3794
5d9a96ca
DM
3795 uw->prev = st->unwind;
3796 st->unwind = unwind1;
e822a8b4 3797 uw->type = ((type == BRANCH)
02db2b7b
IZ
3798 ? RE_UNWIND_BRANCH
3799 : RE_UNWIND_BRANCHJ);
3800 uw->lastparen = lastparen;
3801 uw->next = next;
3802 uw->locinput = locinput;
3803 uw->nextchr = nextchr;
3a2830be 3804 uw->minmod = st->minmod;
02db2b7b
IZ
3805#ifdef DEBUGGING
3806 uw->regindent = ++PL_regindent;
3807#endif
c277df42 3808
02db2b7b
IZ
3809 REGCP_SET(uw->lastcp);
3810
3811 /* Now go into the first branch */
3812 next = inner;
a687059c 3813 }
a0d0e21e
LW
3814 }
3815 break;
3816 case MINMOD:
5d9a96ca 3817 st->minmod = 1;
a0d0e21e 3818 break;
c277df42
IZ
3819 case CURLYM:
3820 {
d8319b27 3821 st->u.curlym.l = st->u.curlym.matches = 0;
9041c2e3 3822
c277df42 3823 /* We suppose that the next guy does not need
0e788c72 3824 backtracking: in particular, it is of constant non-zero length,
c277df42 3825 and has no parenths to influence future backrefs. */
5d9a96ca 3826 st->ln = ARG1(scan); /* min to match */
c277df42 3827 n = ARG2(scan); /* max to match */
d8319b27
DM
3828 st->u.curlym.paren = scan->flags;
3829 if (st->u.curlym.paren) {
3830 if (st->u.curlym.paren > PL_regsize)
3831 PL_regsize = st->u.curlym.paren;
3832 if (st->u.curlym.paren > (I32)*PL_reglastparen)
3833 *PL_reglastparen = st->u.curlym.paren;
c277df42 3834 }
dc45a647 3835 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
d8319b27 3836 if (st->u.curlym.paren)
c277df42 3837 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3280af22 3838 PL_reginput = locinput;
d8319b27
DM
3839 st->u.curlym.maxwanted = st->minmod ? st->ln : n;
3840 if (st->u.curlym.maxwanted) {
3841 while (PL_reginput < PL_regeol && st->u.curlym.matches < st->u.curlym.maxwanted) {
95b24440
DM
3842 REGMATCH(scan, CURLYM1);
3843 /*** all unsaved local vars undefined at this point */
3844 if (!result)
6407bf3b 3845 break;
d8319b27
DM
3846 /* on first match, determine length, u.curlym.l */
3847 if (!st->u.curlym.matches++) {
6407bf3b
DM
3848 if (PL_reg_match_utf8) {
3849 char *s = locinput;
3850 while (s < PL_reginput) {
d8319b27 3851 st->u.curlym.l++;
6407bf3b
DM
3852 s += UTF8SKIP(s);
3853 }
3854 }
3855 else {
d8319b27 3856 st->u.curlym.l = PL_reginput - locinput;
6407bf3b 3857 }
d8319b27
DM
3858 if (st->u.curlym.l == 0) {
3859 st->u.curlym.matches = st->u.curlym.maxwanted;
6407bf3b
DM
3860 break;
3861 }
3862 }
3863 locinput = PL_reginput;
3864 }
3865 }
3866
3867 PL_reginput = locinput;
3868
5d9a96ca
DM
3869 if (st->minmod) {
3870 st->minmod = 0;
d8319b27 3871 if (st->ln && st->u.curlym.matches < st->ln)
c277df42 3872 sayNO;
cca55fe3 3873 if (HAS_TEXT(next) || JUMPABLE(next)) {
5f80c4cf
JP
3874 regnode *text_node = next;
3875
cca55fe3 3876 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
5f80c4cf 3877
d8319b27 3878 if (! HAS_TEXT(text_node)) st->u.curlym.c1 = st->u.curlym.c2 = -1000;
5f80c4cf 3879 else {
cca55fe3 3880 if (PL_regkind[(U8)OP(text_node)] == REF) {
d8319b27 3881 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
44a68960 3882 goto assume_ok_MM;
cca55fe3 3883 }
d8319b27 3884 else { st->u.curlym.c1 = (U8)*STRING(text_node); }
af5decee 3885 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
d8319b27 3886 st->u.curlym.c2 = PL_fold[st->u.curlym.c1];
af5decee 3887 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
d8319b27 3888 st->u.curlym.c2 = PL_fold_locale[st->u.curlym.c1];
5f80c4cf 3889 else
d8319b27 3890 st->u.curlym.c2 = st->u.curlym.c1;
5f80c4cf 3891 }
a0ed51b3
LW
3892 }
3893 else
d8319b27 3894 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
cca55fe3 3895 assume_ok_MM:
d8319b27 3896 REGCP_SET(st->u.curlym.lastcp);
5d9a96ca 3897 while (n >= st->ln || (n == REG_INFTY && st->ln > 0)) { /* ln overflow ? */
c277df42 3898 /* If it could work, try it. */
d8319b27
DM
3899 if (st->u.curlym.c1 == -1000 ||
3900 UCHARAT(PL_reginput) == st->u.curlym.c1 ||
3901 UCHARAT(PL_reginput) == st->u.curlym.c2)
c277df42 3902 {
d8319b27 3903 if (st->u.curlym.paren) {
5d9a96ca 3904 if (st->ln) {
d8319b27
DM
3905 PL_regstartp[st->u.curlym.paren] =
3906 HOPc(PL_reginput, -st->u.curlym.l) - PL_bostr;
3907 PL_regendp[st->u.curlym.paren] = PL_reginput - PL_bostr;
a0ed51b3
LW
3908 }
3909 else
d8319b27 3910 PL_regendp[st->u.curlym.paren] = -1;
c277df42 3911 }
95b24440
DM
3912 REGMATCH(next, CURLYM2);
3913 /*** all unsaved local vars undefined at this point */
3914 if (result)
c277df42 3915 sayYES;
d8319b27 3916 REGCP_UNWIND(st->u.curlym.lastcp);
c277df42
IZ
3917 }
3918 /* Couldn't or didn't -- move forward. */
3280af22 3919 PL_reginput = locinput;
95b24440
DM
3920 REGMATCH(scan, CURLYM3);
3921 /*** all unsaved local vars undefined at this point */
3922 if (result) {
5d9a96ca 3923 st->ln++;
3280af22 3924 locinput = PL_reginput;
c277df42
IZ
3925 }
3926 else
3927 sayNO;
3928 }
a0ed51b3
LW
3929 }
3930 else {
a3621e74 3931 DEBUG_EXECUTE_r(
5c0ca799 3932 PerlIO_printf(Perl_debug_log,
95b24440
DM
3933 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3934 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
d8319b27 3935 (IV) st->u.curlym.matches, (IV)st->u.curlym.l)
c277df42 3936 );
d8319b27 3937 if (st->u.curlym.matches >= st->ln) {
cca55fe3 3938 if (HAS_TEXT(next) || JUMPABLE(next)) {
5f80c4cf
JP
3939 regnode *text_node = next;
3940
cca55fe3 3941 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
5f80c4cf 3942
d8319b27 3943 if (! HAS_TEXT(text_node)) st->u.curlym.c1 = st->u.curlym.c2 = -1000;
5f80c4cf 3944 else {
cca55fe3 3945 if (PL_regkind[(U8)OP(text_node)] == REF) {
d8319b27 3946 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
44a68960 3947 goto assume_ok_REG;
cca55fe3 3948 }
d8319b27 3949 else { st->u.curlym.c1 = (U8)*STRING(text_node); }
cca55fe3 3950
af5decee 3951 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
d8319b27 3952 st->u.curlym.c2 = PL_fold[st->u.curlym.c1];
af5decee 3953 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
d8319b27 3954 st->u.curlym.c2 = PL_fold_locale[st->u.curlym.c1];
5f80c4cf 3955 else
d8319b27 3956 st->u.curlym.c2 = st->u.curlym.c1;
5f80c4cf 3957 }
a0ed51b3
LW
3958 }
3959 else
d8319b27 3960 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
c277df42 3961 }
cca55fe3 3962 assume_ok_REG:
d8319b27
DM
3963 REGCP_SET(st->u.curlym.lastcp);
3964 while (st->u.curlym.matches >= st->ln) {
c277df42 3965 /* If it could work, try it. */
d8319b27
DM
3966 if (st->u.curlym.c1 == -1000 ||
3967 UCHARAT(PL_reginput) == st->u.curlym.c1 ||
3968 UCHARAT(PL_reginput) == st->u.curlym.c2)
a0ed51b3 3969 {
a3621e74 3970 DEBUG_EXECUTE_r(
6407bf3b
DM
3971 PerlIO_printf(Perl_debug_log,
3972 "%*s trying tail with matches=%"IVdf"...\n",
3973 (int)(REPORT_CODE_OFF+PL_regindent*2),
d8319b27 3974 "", (IV)st->u.curlym.matches)
a0ed51b3 3975 );
d8319b27
DM
3976 if (st->u.curlym.paren) {
3977 if (st->u.curlym.matches) {
3978 PL_regstartp[st->u.curlym.paren]
3979 = HOPc(PL_reginput, -st->u.curlym.l) - PL_bostr;
3980 PL_regendp[st->u.curlym.paren] = PL_reginput - PL_bostr;
c277df42 3981 }
a0ed51b3 3982 else
d8319b27 3983 PL_regendp[st->u.curlym.paren] = -1;
c277df42 3984 }
95b24440
DM
3985 REGMATCH(next, CURLYM4);
3986 /*** all unsaved local vars undefined at this point */
3987 if (result)
a0ed51b3 3988 sayYES;
d8319b27 3989 REGCP_UNWIND(st->u.curlym.lastcp);
a0ed51b3 3990 }
c277df42 3991 /* Couldn't or didn't -- back up. */
d8319b27
DM
3992 st->u.curlym.matches--;
3993 locinput = HOPc(locinput, -st->u.curlym.l);
3280af22 3994 PL_reginput = locinput;
c277df42
IZ
3995 }
3996 }
3997 sayNO;
5f66b61c 3998 /* NOTREACHED */
c277df42
IZ
3999 break;
4000 }
4001 case CURLYN:
d8319b27
DM
4002 st->u.plus.paren = scan->flags; /* Which paren to set */
4003 if (st->u.plus.paren > PL_regsize)
4004 PL_regsize = st->u.plus.paren;
4005 if (st->u.plus.paren > (I32)*PL_reglastparen)
4006 *PL_reglastparen = st->u.plus.paren;
5d9a96ca 4007 st->ln = ARG1(scan); /* min to match */
c277df42 4008 n = ARG2(scan); /* max to match */
dc45a647 4009 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
c277df42 4010 goto repeat;
a0d0e21e 4011 case CURLY:
d8319b27 4012 st->u.plus.paren = 0;
5d9a96ca 4013 st->ln = ARG1(scan); /* min to match */
a0d0e21e 4014 n = ARG2(scan); /* max to match */
dc45a647 4015 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
a0d0e21e
LW
4016 goto repeat;
4017 case STAR:
5d9a96ca 4018 st->ln = 0;
c277df42 4019 n = REG_INFTY;
a0d0e21e 4020 scan = NEXTOPER(scan);
d8319b27 4021 st->u.plus.paren = 0;
a0d0e21e
LW
4022 goto repeat;
4023 case PLUS:
5d9a96ca 4024 st->ln = 1;
c277df42
IZ
4025 n = REG_INFTY;
4026 scan = NEXTOPER(scan);
d8319b27 4027 st->u.plus.paren = 0;
c277df42 4028 repeat:
a0d0e21e
LW
4029 /*
4030 * Lookahead to avoid useless match attempts
4031 * when we know what character comes next.
4032 */
5f80c4cf
JP
4033
4034 /*
4035 * Used to only do .*x and .*?x, but now it allows
4036 * for )'s, ('s and (?{ ... })'s to be in the way
4037 * of the quantifier and the EXACT-like node. -- japhy
4038 */
4039
cca55fe3 4040 if (HAS_TEXT(next) || JUMPABLE(next)) {
5f80c4cf
JP
4041 U8 *s;
4042 regnode *text_node = next;
4043
cca55fe3 4044 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
5f80c4cf 4045
d8319b27 4046 if (! HAS_TEXT(text_node)) st->u.plus.c1 = st->u.plus.c2 = -1000;
5f80c4cf 4047 else {
cca55fe3 4048 if (PL_regkind[(U8)OP(text_node)] == REF) {
d8319b27 4049 st->u.plus.c1 = st->u.plus.c2 = -1000;
44a68960 4050 goto assume_ok_easy;
cca55fe3
JP
4051 }
4052 else { s = (U8*)STRING(text_node); }
5f80c4cf
JP
4053
4054 if (!UTF) {
d8319b27 4055 st->u.plus.c2 = st->u.plus.c1 = *s;
f65d3ee7 4056 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
d8319b27 4057 st->u.plus.c2 = PL_fold[st->u.plus.c1];
f65d3ee7 4058 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
d8319b27 4059 st->u.plus.c2 = PL_fold_locale[st->u.plus.c1];
1aa99e6b 4060 }
5f80c4cf 4061 else { /* UTF */
f65d3ee7 4062 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
a2a2844f 4063 STRLEN ulen1, ulen2;
89ebb4a3
JH
4064 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
4065 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
a2a2844f
JH
4066
4067 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
4068 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
4069
d8319b27 4070 st->u.plus.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
041457d9 4071 uniflags);
d8319b27 4072 st->u.plus.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
041457d9 4073 uniflags);
5f80c4cf
JP
4074 }
4075 else {
d8319b27 4076 st->u.plus.c2 = st->u.plus.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
041457d9 4077 uniflags);
5f80c4cf 4078 }
1aa99e6b
IH
4079 }
4080 }
bbce6d69 4081 }
a0d0e21e 4082 else
d8319b27 4083 st->u.plus.c1 = st->u.plus.c2 = -1000;
cca55fe3 4084 assume_ok_easy:
3280af22 4085 PL_reginput = locinput;
5d9a96ca
DM
4086 if (st->minmod) {
4087 st->minmod = 0;
32fc9b6a 4088 if (st->ln && regrepeat(rex, scan, st->ln) < st->ln)
4633a7c4 4089 sayNO;
a0ed51b3 4090 locinput = PL_reginput;
d8319b27
DM
4091 REGCP_SET(st->u.plus.lastcp);
4092 if (st->u.plus.c1 != -1000) {
4093 st->u.plus.old = locinput;
4094 st->u.plus.count = 0;
0fe9bf95 4095
1aa99e6b 4096 if (n == REG_INFTY) {
d8319b27 4097 st->u.plus.e = PL_regeol - 1;
1aa99e6b 4098 if (do_utf8)
d8319b27
DM
4099 while (UTF8_IS_CONTINUATION(*(U8*)st->u.plus.e))
4100 st->u.plus.e--;
1aa99e6b
IH
4101 }
4102 else if (do_utf8) {
5d9a96ca 4103 int m = n - st->ln;
d8319b27
DM
4104 for (st->u.plus.e = locinput;
4105 m >0 && st->u.plus.e + UTF8SKIP(st->u.plus.e) <= PL_regeol; m--)
4106 st->u.plus.e += UTF8SKIP(st->u.plus.e);
1aa99e6b
IH
4107 }
4108 else {
d8319b27
DM
4109 st->u.plus.e = locinput + n - st->ln;
4110 if (st->u.plus.e >= PL_regeol)
4111 st->u.plus.e = PL_regeol - 1;
1aa99e6b 4112 }
0fe9bf95
IZ
4113 while (1) {
4114 /* Find place 'next' could work */
1aa99e6b 4115 if (!do_utf8) {
d8319b27
DM
4116 if (st->u.plus.c1 == st->u.plus.c2) {
4117 while (locinput <= st->u.plus.e &&
4118 UCHARAT(locinput) != st->u.plus.c1)
1aa99e6b
IH
4119 locinput++;
4120 } else {
d8319b27
DM
4121 while (locinput <= st->u.plus.e
4122 && UCHARAT(locinput) != st->u.plus.c1
4123 && UCHARAT(locinput) != st->u.plus.c2)
1aa99e6b
IH
4124 locinput++;
4125 }
d8319b27 4126 st->u.plus.count = locinput - st->u.plus.old;
1aa99e6b
IH
4127 }
4128 else {
d8319b27 4129 if (st->u.plus.c1 == st->u.plus.c2) {
a3b680e6 4130 STRLEN len;
872c91ae
JH
4131 /* count initialised to
4132 * utf8_distance(old, locinput) */
d8319b27 4133 while (locinput <= st->u.plus.e &&
872c91ae 4134 utf8n_to_uvchr((U8*)locinput,
89ebb4a3 4135 UTF8_MAXBYTES, &len,
d8319b27 4136 uniflags) != (UV)st->u.plus.c1) {
1aa99e6b 4137 locinput += len;
d8319b27 4138 st->u.plus.count++;
b2f2f093 4139 }
1aa99e6b 4140 } else {
a3b680e6 4141 STRLEN len;
872c91ae
JH
4142 /* count initialised to
4143 * utf8_distance(old, locinput) */
d8319b27 4144 while (locinput <= st->u.plus.e) {
872c91ae 4145 UV c = utf8n_to_uvchr((U8*)locinput,
89ebb4a3 4146 UTF8_MAXBYTES, &len,
041457d9 4147 uniflags);
d8319b27 4148 if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2)
1aa99e6b 4149 break;
b2f2f093 4150 locinput += len;
d8319b27 4151 st->u.plus.count++;
1aa99e6b
IH
4152 }
4153 }
0fe9bf95 4154 }
d8319b27 4155 if (locinput > st->u.plus.e)
0fe9bf95
IZ
4156 sayNO;
4157 /* PL_reginput == old now */
d8319b27 4158 if (locinput != st->u.plus.old) {
5d9a96ca 4159 st->ln = 1; /* Did some */
32fc9b6a 4160 if (regrepeat(rex, scan, st->u.plus.count) < st->u.plus.count)
0fe9bf95
IZ
4161 sayNO;
4162 }
4163 /* PL_reginput == locinput now */
d8319b27 4164 TRYPAREN(st->u.plus.paren, st->ln, locinput, PLUS1);
95b24440 4165 /*** all unsaved local vars undefined at this point */
0fe9bf95 4166 PL_reginput = locinput; /* Could be reset... */
d8319b27 4167 REGCP_UNWIND(st->u.plus.lastcp);
0fe9bf95 4168 /* Couldn't or didn't -- move forward. */
d8319b27 4169 st->u.plus.old = locinput;
1aa99e6b
IH
4170 if (do_utf8)
4171 locinput += UTF8SKIP(locinput);
4172 else
4173 locinput++;
d8319b27 4174 st->u.plus.count = 1;
0fe9bf95
IZ
4175 }
4176 }
4177 else
5d9a96ca 4178 while (n >= st->ln || (n == REG_INFTY && st->ln > 0)) { /* ln overflow ? */
1aa99e6b 4179 UV c;
d8319b27 4180 if (st->u.plus.c1 != -1000) {
1aa99e6b 4181 if (do_utf8)
872c91ae 4182 c = utf8n_to_uvchr((U8*)PL_reginput,
89ebb4a3 4183 UTF8_MAXBYTES, 0,
041457d9 4184 uniflags);
1aa99e6b 4185 else
9041c2e3 4186 c = UCHARAT(PL_reginput);
2390ecbc 4187 /* If it could work, try it. */
d8319b27 4188 if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2)
2390ecbc 4189 {
d8319b27 4190 TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS2);
95b24440 4191 /*** all unsaved local vars undefined at this point */
d8319b27 4192 REGCP_UNWIND(st->u.plus.lastcp);
2390ecbc 4193 }
1aa99e6b 4194 }
a0d0e21e 4195 /* If it could work, try it. */
d8319b27 4196 else if (st->u.plus.c1 == -1000)
bbce6d69 4197 {
d8319b27 4198 TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS3);
95b24440 4199 /*** all unsaved local vars undefined at this point */
d8319b27 4200 REGCP_UNWIND(st->u.plus.lastcp);
bbce6d69 4201 }
c277df42 4202 /* Couldn't or didn't -- move forward. */
a0ed51b3 4203 PL_reginput = locinput;
32fc9b6a 4204 if (regrepeat(rex, scan, 1)) {
5d9a96ca 4205 st->ln++;
a0ed51b3
LW
4206 locinput = PL_reginput;
4207 }
4208 else
4633a7c4 4209 sayNO;
a0d0e21e
LW
4210 }
4211 }
4212 else {
32fc9b6a 4213 n = regrepeat(rex, scan, n);
a0ed51b3 4214 locinput = PL_reginput;
5d9a96ca 4215 if (st->ln < n && PL_regkind[(U8)OP(next)] == EOL &&
7fba1cd6 4216 (OP(next) != MEOL ||
15272685
HS
4217 OP(next) == SEOL || OP(next) == EOS))
4218 {
5d9a96ca 4219 st->ln = n; /* why back off? */
1aeab75a
GS
4220 /* ...because $ and \Z can match before *and* after
4221 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
4222 We should back off by one in this case. */
4223 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
5d9a96ca 4224 st->ln--;
1aeab75a 4225 }
d8319b27 4226 REGCP_SET(st->u.plus.lastcp);
1d5c262f 4227 {
8fa7f367 4228 UV c = 0;
5d9a96ca 4229 while (n >= st->ln) {
d8319b27 4230 if (st->u.plus.c1 != -1000) {
1aa99e6b 4231 if (do_utf8)
872c91ae 4232 c = utf8n_to_uvchr((U8*)PL_reginput,
89ebb4a3 4233 UTF8_MAXBYTES, 0,
041457d9 4234 uniflags);
1aa99e6b 4235 else
9041c2e3 4236 c = UCHARAT(PL_reginput);
1aa99e6b 4237 }
c277df42 4238 /* If it could work, try it. */
d8319b27 4239 if (st->u.plus.c1 == -1000 || c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2)
c277df42 4240 {
d8319b27 4241 TRYPAREN(st->u.plus.paren, n, PL_reginput, PLUS4);
95b24440 4242 /*** all unsaved local vars undefined at this point */
d8319b27 4243 REGCP_UNWIND(st->u.plus.lastcp);
c277df42
IZ
4244 }
4245 /* Couldn't or didn't -- back up. */
4246 n--;
dfe13c55 4247 PL_reginput = locinput = HOPc(locinput, -1);
bbce6d69 4248 }
a0d0e21e
LW
4249 }
4250 }
4633a7c4 4251 sayNO;
c277df42 4252 break;
a0d0e21e 4253 case END:
aa283a38
DM
4254 if (cur_eval) {
4255 /* we have successfully completed the execution of a
4256 * postponed re. Pop all states back to the last EVAL
4257 * then continue with the node following the (??{...})
4258 */
4259
4260 /* this simulates a POP_STATE, except that it pops several
4261 * levels, and doesn't restore locinput */
6136c704 4262
aa283a38
DM
4263 st = cur_eval;
4264 PL_regmatch_slab = st->u.eval.prev_slab;
4265 cur_eval = st->u.eval.prev_eval;
4266 depth = st->u.eval.depth;
6136c704 4267
aa283a38
DM
4268 PL_regmatch_state = st;
4269 scan = st->scan;
4270 next = st->next;
4271 n = st->n;
4272
4273 if (st->u.eval.toggleutf)
4274 PL_reg_flags ^= RF_utf8;
4275 ReREFCNT_dec(rex);
4276 rex = st->u.eval.prev_rex;
aa283a38
DM
4277 /* XXXX This is too dramatic a measure... */
4278 PL_reg_maxiter = 0;
4279
4280 /* Restore parens of the caller without popping the
4281 * savestack */
95b24440
DM
4282 {
4283 I32 tmp = PL_savestack_ix;
aa283a38 4284 PL_savestack_ix = st->u.eval.lastcp;
4f639d21 4285 regcppop(rex);
95b24440
DM
4286 PL_savestack_ix = tmp;
4287 }
6136c704 4288
6136c704 4289
aa283a38
DM
4290 PL_reginput = locinput;
4291 /* resume at node following the (??{...}) */
4292 break;
0f5d15d6 4293
0f5d15d6 4294 }
aa283a38 4295
3b0527fe 4296 if (locinput < reginfo->till) {
a3621e74 4297 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
7821416a
IZ
4298 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
4299 PL_colors[4],
4300 (long)(locinput - PL_reg_starttry),
3b0527fe 4301 (long)(reginfo->till - PL_reg_starttry),
7821416a
IZ
4302 PL_colors[5]));
4303 sayNO_FINAL; /* Cannot match: too short. */
4304 }
4305 PL_reginput = locinput; /* put where regtry can find it */
4306 sayYES_FINAL; /* Success! */
7e5428c5 4307 case SUCCEED:
3280af22 4308 PL_reginput = locinput; /* put where regtry can find it */
7821416a 4309 sayYES_LOUD; /* Success! */
c277df42
IZ
4310 case SUSPEND:
4311 n = 1;
9fe1d20c 4312 PL_reginput = locinput;
9041c2e3 4313 goto do_ifmatch;
a0d0e21e 4314 case UNLESSM:
c277df42 4315 n = 0;
a0ed51b3 4316 if (scan->flags) {
52657f30 4317 char * const s = HOPBACKc(locinput, scan->flags);
efb30f32
HS
4318 if (!s)
4319 goto say_yes;
4320 PL_reginput = s;
a0ed51b3
LW
4321 }
4322 else
4323 PL_reginput = locinput;
c277df42
IZ
4324 goto do_ifmatch;
4325 case IFMATCH:
4326 n = 1;
a0ed51b3 4327 if (scan->flags) {
52657f30 4328 char * const s = HOPBACKc(locinput, scan->flags);
efb30f32
HS
4329 if (!s)
4330 goto say_no;
4331 PL_reginput = s;
a0ed51b3
LW
4332 }
4333 else
4334 PL_reginput = locinput;
4335
c277df42 4336 do_ifmatch:
95b24440
DM
4337 REGMATCH(NEXTOPER(NEXTOPER(scan)), IFMATCH);
4338 /*** all unsaved local vars undefined at this point */
4339 if (result != n) {
c277df42 4340 say_no:
5d9a96ca
DM
4341 if (st->logical) {
4342 st->logical = 0;
4343 st->sw = 0;
c277df42 4344 goto do_longjump;
a0ed51b3
LW
4345 }
4346 else
c277df42
IZ
4347 sayNO;
4348 }
4349 say_yes:
5d9a96ca
DM
4350 if (st->logical) {
4351 st->logical = 0;
4352 st->sw = 1;
c277df42 4353 }
fe44a5e8 4354 if (OP(scan) == SUSPEND) {
3280af22 4355 locinput = PL_reginput;
565764a8 4356 nextchr = UCHARAT(locinput);
fe44a5e8 4357 }
c277df42
IZ
4358 /* FALL THROUGH. */
4359 case LONGJMP:
4360 do_longjump:
4361 next = scan + ARG(scan);
4362 if (next == scan)
4363 next = NULL;
a0d0e21e
LW
4364 break;
4365 default:
b900a521 4366 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
d7d93a81 4367 PTR2UV(scan), OP(scan));
cea2e8a9 4368 Perl_croak(aTHX_ "regexp memory corruption");
a687059c 4369 }
95b24440 4370
02db2b7b 4371 reenter:
a0d0e21e 4372 scan = next;
95b24440
DM
4373 continue;
4374 /* NOTREACHED */
4375
4376 /* simulate recursively calling regmatch(), but without actually
4377 * recursing - ie save the current state on the heap rather than on
4378 * the stack, then re-enter the loop. This avoids complex regexes
4379 * blowing the processor stack */
4380
4381 start_recurse:
4382 {
5d9a96ca
DM
4383 /* push new state */
4384 regmatch_state *oldst = st;
4385
4386 depth++;
4387
4388 /* grab the next free state slot */
4389 st++;
4390 if (st > &(PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]))
4391 st = S_push_slab(aTHX);
4392 PL_regmatch_state = st;
4393
4394 oldst->next = next;
4395 oldst->n = n;
4396 oldst->locinput = locinput;
5d9a96ca
DM
4397
4398 st->cc = oldst->cc;
95b24440
DM
4399 locinput = PL_reginput;
4400 nextchr = UCHARAT(locinput);
5d9a96ca
DM
4401 st->minmod = 0;
4402 st->sw = 0;
4403 st->logical = 0;
4404 st->unwind = 0;
95b24440
DM
4405#ifdef DEBUGGING
4406 PL_regindent++;
4407#endif
4408 }
a0d0e21e 4409 }
a687059c 4410
aa283a38
DM
4411
4412
a0d0e21e
LW
4413 /*
4414 * We get here only if there's trouble -- normally "case END" is
4415 * the terminating point.
4416 */
cea2e8a9 4417 Perl_croak(aTHX_ "corrupted regexp pointers");
a0d0e21e 4418 /*NOTREACHED*/
4633a7c4
LW
4419 sayNO;
4420
7821416a 4421yes_loud:
a3621e74 4422 DEBUG_EXECUTE_r(
7821416a
IZ
4423 PerlIO_printf(Perl_debug_log,
4424 "%*s %scould match...%s\n",
e4584336 4425 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
7821416a
IZ
4426 );
4427 goto yes;
4428yes_final:
a3621e74 4429 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
e4584336 4430 PL_colors[4], PL_colors[5]));
4633a7c4
LW
4431yes:
4432#ifdef DEBUGGING
3280af22 4433 PL_regindent--;
4633a7c4 4434#endif
02db2b7b 4435
95b24440 4436 result = 1;
aa283a38
DM
4437 /* XXX this is duplicate(ish) code to that in the do_no section.
4438 * eventually a yes should just pop the whole stack */
4439 if (depth) {
4440 /* restore previous state and re-enter */
4441 POP_STATE;
4442
4443 switch (st->resume_state) {
4444 case resume_TRIE1:
4445 goto resume_point_TRIE1;
4446 case resume_TRIE2:
4447 goto resume_point_TRIE2;
4448 case resume_EVAL:
4449 break;
4450 case resume_CURLYX:
4451 goto resume_point_CURLYX;
4452 case resume_WHILEM1:
4453 goto resume_point_WHILEM1;
4454 case resume_WHILEM2:
4455 goto resume_point_WHILEM2;
4456 case resume_WHILEM3:
4457 goto resume_point_WHILEM3;
4458 case resume_WHILEM4:
4459 goto resume_point_WHILEM4;
4460 case resume_WHILEM5:
4461 goto resume_point_WHILEM5;
4462 case resume_WHILEM6:
4463 goto resume_point_WHILEM6;
4464 case resume_CURLYM1:
4465 goto resume_point_CURLYM1;
4466 case resume_CURLYM2:
4467 goto resume_point_CURLYM2;
4468 case resume_CURLYM3:
4469 goto resume_point_CURLYM3;
4470 case resume_CURLYM4:
4471 goto resume_point_CURLYM4;
4472 case resume_IFMATCH:
4473 goto resume_point_IFMATCH;
4474 case resume_PLUS1:
4475 goto resume_point_PLUS1;
4476 case resume_PLUS2:
4477 goto resume_point_PLUS2;
4478 case resume_PLUS3:
4479 goto resume_point_PLUS3;
4480 case resume_PLUS4:
4481 goto resume_point_PLUS4;
4482 default:
4483 Perl_croak(aTHX_ "regexp resume memory corruption");
4484 }
4485 }
4486 goto final_exit;
4633a7c4
LW
4487
4488no:
a3621e74 4489 DEBUG_EXECUTE_r(
7821416a
IZ
4490 PerlIO_printf(Perl_debug_log,
4491 "%*s %sfailed...%s\n",
e4584336 4492 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
7821416a
IZ
4493 );
4494 goto do_no;
4495no_final:
4496do_no:
5d9a96ca
DM
4497 if (st->unwind) {
4498 re_unwind_t * const uw = SSPTRt(st->unwind,re_unwind_t);
02db2b7b
IZ
4499
4500 switch (uw->type) {
4501 case RE_UNWIND_BRANCH:
4502 case RE_UNWIND_BRANCHJ:
4503 {
6136c704 4504 re_unwind_branch_t * const uwb = &(uw->branch);
a3b680e6 4505 const I32 lastparen = uwb->lastparen;
9041c2e3 4506
02db2b7b
IZ
4507 REGCP_UNWIND(uwb->lastcp);
4508 for (n = *PL_reglastparen; n > lastparen; n--)
4509 PL_regendp[n] = -1;
4510 *PL_reglastparen = n;
4511 scan = next = uwb->next;
3a2830be 4512 st->minmod = uwb->minmod;
9041c2e3
NIS
4513 if ( !scan ||
4514 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
02db2b7b 4515 ? BRANCH : BRANCHJ) ) { /* Failure */
5d9a96ca 4516 st->unwind = uwb->prev;
02db2b7b
IZ
4517#ifdef DEBUGGING
4518 PL_regindent--;
4519#endif
4520 goto do_no;
4521 }
4522 /* Have more choice yet. Reuse the same uwb. */
02db2b7b
IZ
4523 if ((n = (uwb->type == RE_UNWIND_BRANCH
4524 ? NEXT_OFF(next) : ARG(next))))
4525 next += n;
4526 else
4527 next = NULL; /* XXXX Needn't unwinding in this case... */
4528 uwb->next = next;
4529 next = NEXTOPER(scan);
4530 if (uwb->type == RE_UNWIND_BRANCHJ)
4531 next = NEXTOPER(next);
4532 locinput = uwb->locinput;
4533 nextchr = uwb->nextchr;
4534#ifdef DEBUGGING
4535 PL_regindent = uwb->regindent;
4536#endif
4537
4538 goto reenter;
4539 }
5f66b61c 4540 /* NOTREACHED */
02db2b7b
IZ
4541 default:
4542 Perl_croak(aTHX_ "regexp unwind memory corruption");
4543 }
5f66b61c 4544 /* NOTREACHED */
02db2b7b 4545 }
aa283a38 4546
4633a7c4 4547#ifdef DEBUGGING
3280af22 4548 PL_regindent--;
4633a7c4 4549#endif
95b24440 4550 result = 0;
5d9a96ca 4551
aa283a38
DM
4552 if (depth) {
4553 /* there's a previous state to backtrack to */
4554 POP_STATE;
5d9a96ca 4555 switch (st->resume_state) {
95b24440
DM
4556 case resume_TRIE1:
4557 goto resume_point_TRIE1;
4558 case resume_TRIE2:
4559 goto resume_point_TRIE2;
aa283a38
DM
4560 case resume_EVAL:
4561 /* we have failed an (??{...}). Restore state to the outer re
4562 * then re-throw the failure */
4563 if (st->u.eval.toggleutf)
4564 PL_reg_flags ^= RF_utf8;
4565 ReREFCNT_dec(rex);
4566 rex = st->u.eval.prev_rex;
aa283a38
DM
4567 cur_eval = st->u.eval.prev_eval;
4568
4569 /* XXXX This is too dramatic a measure... */
4570 PL_reg_maxiter = 0;
4571
4572 PL_reginput = locinput;
4573 REGCP_UNWIND(st->u.eval.lastcp);
4574 regcppop(rex);
4575 goto do_no;
4576
95b24440
DM
4577 case resume_CURLYX:
4578 goto resume_point_CURLYX;
4579 case resume_WHILEM1:
4580 goto resume_point_WHILEM1;
4581 case resume_WHILEM2:
4582 goto resume_point_WHILEM2;
4583 case resume_WHILEM3:
4584 goto resume_point_WHILEM3;
4585 case resume_WHILEM4:
4586 goto resume_point_WHILEM4;
4587 case resume_WHILEM5:
4588 goto resume_point_WHILEM5;
4589 case resume_WHILEM6:
4590 goto resume_point_WHILEM6;
4591 case resume_CURLYM1:
4592 goto resume_point_CURLYM1;
4593 case resume_CURLYM2:
4594 goto resume_point_CURLYM2;
4595 case resume_CURLYM3:
4596 goto resume_point_CURLYM3;
4597 case resume_CURLYM4:
4598 goto resume_point_CURLYM4;
4599 case resume_IFMATCH:
4600 goto resume_point_IFMATCH;
4601 case resume_PLUS1:
4602 goto resume_point_PLUS1;
4603 case resume_PLUS2:
4604 goto resume_point_PLUS2;
4605 case resume_PLUS3:
4606 goto resume_point_PLUS3;
4607 case resume_PLUS4:
4608 goto resume_point_PLUS4;
95b24440
DM
4609 default:
4610 Perl_croak(aTHX_ "regexp resume memory corruption");
4611 }
95b24440 4612 }
aa283a38
DM
4613
4614final_exit:
4615
5d9a96ca
DM
4616 /* restore original high-water mark */
4617 PL_regmatch_slab = orig_slab;
4618 PL_regmatch_state = orig_state;
4619
4620 /* free all slabs above current one */
4621 if (orig_slab->next) {
ad65c075 4622 regmatch_slab *osl, *sl = orig_slab->next;
5d9a96ca
DM
4623 orig_slab->next = NULL;
4624 while (sl) {
ad65c075 4625 osl = sl;
5d9a96ca 4626 sl = sl->next;
ad65c075 4627 Safefree(osl);
5d9a96ca
DM
4628 }
4629 }
4630
95b24440
DM
4631 return result;
4632
a687059c
LW
4633}
4634
4635/*
4636 - regrepeat - repeatedly match something simple, report how many
4637 */
4638/*
4639 * [This routine now assumes that it will only match on things of length 1.
4640 * That was true before, but now we assume scan - reginput is the count,
a0ed51b3 4641 * rather than incrementing count on every character. [Er, except utf8.]]
a687059c 4642 */
76e3520e 4643STATIC I32
32fc9b6a 4644S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max)
a687059c 4645{
27da23d5 4646 dVAR;
a0d0e21e 4647 register char *scan;
a0d0e21e 4648 register I32 c;
3280af22 4649 register char *loceol = PL_regeol;
a0ed51b3 4650 register I32 hardcount = 0;
53c4c00c 4651 register bool do_utf8 = PL_reg_match_utf8;
a0d0e21e 4652
3280af22 4653 scan = PL_reginput;
faf11cac
HS
4654 if (max == REG_INFTY)
4655 max = I32_MAX;
4656 else if (max < loceol - scan)
7f596f4c 4657 loceol = scan + max;
a0d0e21e 4658 switch (OP(p)) {
22c35a8c 4659 case REG_ANY:
1aa99e6b 4660 if (do_utf8) {
ffc61ed2 4661 loceol = PL_regeol;
1aa99e6b 4662 while (scan < loceol && hardcount < max && *scan != '\n') {
ffc61ed2
JH
4663 scan += UTF8SKIP(scan);
4664 hardcount++;
4665 }
4666 } else {
4667 while (scan < loceol && *scan != '\n')
4668 scan++;
a0ed51b3
LW
4669 }
4670 break;
ffc61ed2 4671 case SANY:
def8e4ea
JH
4672 if (do_utf8) {
4673 loceol = PL_regeol;
a0804c9e 4674 while (scan < loceol && hardcount < max) {
def8e4ea
JH
4675 scan += UTF8SKIP(scan);
4676 hardcount++;
4677 }
4678 }
4679 else
4680 scan = loceol;
a0ed51b3 4681 break;
f33976b4
DB
4682 case CANY:
4683 scan = loceol;
4684 break;
090f7165
JH
4685 case EXACT: /* length of string is 1 */
4686 c = (U8)*STRING(p);
4687 while (scan < loceol && UCHARAT(scan) == c)
4688 scan++;
bbce6d69 4689 break;
4690 case EXACTF: /* length of string is 1 */
cd439c50 4691 c = (U8)*STRING(p);
bbce6d69 4692 while (scan < loceol &&
22c35a8c 4693 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
bbce6d69 4694 scan++;
4695 break;
4696 case EXACTFL: /* length of string is 1 */
3280af22 4697 PL_reg_flags |= RF_tainted;
cd439c50 4698 c = (U8)*STRING(p);
bbce6d69 4699 while (scan < loceol &&
22c35a8c 4700 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
a0d0e21e
LW
4701 scan++;
4702 break;
4703 case ANYOF:
ffc61ed2
JH
4704 if (do_utf8) {
4705 loceol = PL_regeol;
cfc92286 4706 while (hardcount < max && scan < loceol &&
32fc9b6a 4707 reginclass(prog, p, (U8*)scan, 0, do_utf8)) {
ffc61ed2
JH
4708 scan += UTF8SKIP(scan);
4709 hardcount++;
4710 }
4711 } else {
32fc9b6a 4712 while (scan < loceol && REGINCLASS(prog, p, (U8*)scan))
ffc61ed2
JH
4713 scan++;
4714 }
a0d0e21e
LW
4715 break;
4716 case ALNUM:
1aa99e6b 4717 if (do_utf8) {
ffc61ed2 4718 loceol = PL_regeol;
1a4fad37 4719 LOAD_UTF8_CHARCLASS_ALNUM();
1aa99e6b 4720 while (hardcount < max && scan < loceol &&
3568d838 4721 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
ffc61ed2
JH
4722 scan += UTF8SKIP(scan);
4723 hardcount++;
4724 }
4725 } else {
4726 while (scan < loceol && isALNUM(*scan))
4727 scan++;
a0ed51b3
LW
4728 }
4729 break;
bbce6d69 4730 case ALNUML:
3280af22 4731 PL_reg_flags |= RF_tainted;
1aa99e6b 4732 if (do_utf8) {
ffc61ed2 4733 loceol = PL_regeol;
1aa99e6b
IH
4734 while (hardcount < max && scan < loceol &&
4735 isALNUM_LC_utf8((U8*)scan)) {
ffc61ed2
JH
4736 scan += UTF8SKIP(scan);
4737 hardcount++;
4738 }
4739 } else {
4740 while (scan < loceol && isALNUM_LC(*scan))
4741 scan++;
a0ed51b3
LW
4742 }
4743 break;
a0d0e21e 4744 case NALNUM:
1aa99e6b 4745 if (do_utf8) {
ffc61ed2 4746 loceol = PL_regeol;
1a4fad37 4747 LOAD_UTF8_CHARCLASS_ALNUM();
1aa99e6b 4748 while (hardcount < max && scan < loceol &&
3568d838 4749 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
ffc61ed2
JH
4750 scan += UTF8SKIP(scan);
4751 hardcount++;
4752 }
4753 } else {
4754 while (scan < loceol && !isALNUM(*scan))
4755 scan++;
a0ed51b3
LW
4756 }
4757 break;
bbce6d69 4758 case NALNUML:
3280af22 4759 PL_reg_flags |= RF_tainted;
1aa99e6b 4760 if (do_utf8) {
ffc61ed2 4761 loceol = PL_regeol;
1aa99e6b
IH
4762 while (hardcount < max && scan < loceol &&
4763 !isALNUM_LC_utf8((U8*)scan)) {
ffc61ed2
JH
4764 scan += UTF8SKIP(scan);
4765 hardcount++;
4766 }
4767 } else {
4768 while (scan < loceol && !isALNUM_LC(*scan))
4769 scan++;
a0ed51b3
LW
4770 }
4771 break;
a0d0e21e 4772 case SPACE:
1aa99e6b 4773 if (do_utf8) {
ffc61ed2 4774 loceol = PL_regeol;
1a4fad37 4775 LOAD_UTF8_CHARCLASS_SPACE();
1aa99e6b 4776 while (hardcount < max && scan < loceol &&
3568d838
JH
4777 (*scan == ' ' ||
4778 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
ffc61ed2
JH
4779 scan += UTF8SKIP(scan);
4780 hardcount++;
4781 }
4782 } else {
4783 while (scan < loceol && isSPACE(*scan))
4784 scan++;
a0ed51b3
LW
4785 }
4786 break;
bbce6d69 4787 case SPACEL:
3280af22 4788 PL_reg_flags |= RF_tainted;
1aa99e6b 4789 if (do_utf8) {
ffc61ed2 4790 loceol = PL_regeol;
1aa99e6b 4791 while (hardcount < max && scan < loceol &&
ffc61ed2
JH
4792 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4793 scan += UTF8SKIP(scan);
4794 hardcount++;
4795 }
4796 } else {
4797 while (scan < loceol && isSPACE_LC(*scan))
4798 scan++;
a0ed51b3
LW
4799 }
4800 break;
a0d0e21e 4801 case NSPACE:
1aa99e6b 4802 if (do_utf8) {
ffc61ed2 4803 loceol = PL_regeol;
1a4fad37 4804 LOAD_UTF8_CHARCLASS_SPACE();
1aa99e6b 4805 while (hardcount < max && scan < loceol &&
3568d838
JH
4806 !(*scan == ' ' ||
4807 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
ffc61ed2
JH
4808 scan += UTF8SKIP(scan);
4809 hardcount++;
4810 }
4811 } else {
4812 while (scan < loceol && !isSPACE(*scan))
4813 scan++;
4814 break;
a0ed51b3 4815 }
bbce6d69 4816 case NSPACEL:
3280af22 4817 PL_reg_flags |= RF_tainted;
1aa99e6b 4818 if (do_utf8) {
ffc61ed2 4819 loceol = PL_regeol;
1aa99e6b 4820 while (hardcount < max && scan < loceol &&
ffc61ed2
JH
4821 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4822 scan += UTF8SKIP(scan);
4823 hardcount++;
4824 }
4825 } else {
4826 while (scan < loceol && !isSPACE_LC(*scan))
4827 scan++;
a0ed51b3
LW
4828 }
4829 break;
a0d0e21e 4830 case DIGIT:
1aa99e6b 4831 if (do_utf8) {
ffc61ed2 4832 loceol = PL_regeol;
1a4fad37 4833 LOAD_UTF8_CHARCLASS_DIGIT();
1aa99e6b 4834 while (hardcount < max && scan < loceol &&
3568d838 4835 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
ffc61ed2
JH
4836 scan += UTF8SKIP(scan);
4837 hardcount++;
4838 }
4839 } else {
4840 while (scan < loceol && isDIGIT(*scan))
4841 scan++;
a0ed51b3
LW
4842 }
4843 break;
a0d0e21e 4844 case NDIGIT:
1aa99e6b 4845 if (do_utf8) {
ffc61ed2 4846 loceol = PL_regeol;
1a4fad37 4847 LOAD_UTF8_CHARCLASS_DIGIT();
1aa99e6b 4848 while (hardcount < max && scan < loceol &&
3568d838 4849 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
ffc61ed2
JH
4850 scan += UTF8SKIP(scan);
4851 hardcount++;
4852 }
4853 } else {
4854 while (scan < loceol && !isDIGIT(*scan))
4855 scan++;
a0ed51b3
LW
4856 }
4857 break;
a0d0e21e
LW
4858 default: /* Called on something of 0 width. */
4859 break; /* So match right here or not at all. */
4860 }
a687059c 4861
a0ed51b3
LW
4862 if (hardcount)
4863 c = hardcount;
4864 else
4865 c = scan - PL_reginput;
3280af22 4866 PL_reginput = scan;
a687059c 4867
a3621e74 4868 DEBUG_r({
ab74612d 4869 SV *re_debug_flags = NULL;
6136c704 4870 SV * const prop = sv_newmortal();
a3621e74
YO
4871 GET_RE_DEBUG_FLAGS;
4872 DEBUG_EXECUTE_r({
32fc9b6a 4873 regprop(prog, prop, p);
9041c2e3
NIS
4874 PerlIO_printf(Perl_debug_log,
4875 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3f7c398e 4876 REPORT_CODE_OFF+1, "", SvPVX_const(prop),(IV)c,(IV)max);
c277df42 4877 });
a3621e74 4878 });
9041c2e3 4879
a0d0e21e 4880 return(c);
a687059c
LW
4881}
4882
c277df42
IZ
4883
4884/*
ffc61ed2
JH
4885- regclass_swash - prepare the utf8 swash
4886*/
4887
4888SV *
32fc9b6a 4889Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
ffc61ed2 4890{
97aff369 4891 dVAR;
9e55ce06
JH
4892 SV *sw = NULL;
4893 SV *si = NULL;
4894 SV *alt = NULL;
32fc9b6a 4895 const struct reg_data *data = prog ? prog->data : NULL;
ffc61ed2 4896
4f639d21 4897 if (data && data->count) {
a3b680e6 4898 const U32 n = ARG(node);
ffc61ed2 4899
4f639d21
DM
4900 if (data->what[n] == 's') {
4901 SV * const rv = (SV*)data->data[n];
890ce7af 4902 AV * const av = (AV*)SvRV((SV*)rv);
2d03de9c 4903 SV **const ary = AvARRAY(av);
9e55ce06 4904 SV **a, **b;
9041c2e3 4905
711a919c 4906 /* See the end of regcomp.c:S_regclass() for
9e55ce06
JH
4907 * documentation of these array elements. */
4908
b11f357e 4909 si = *ary;
8f7f7219 4910 a = SvROK(ary[1]) ? &ary[1] : 0;
b11f357e
JH
4911 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4912
ffc61ed2
JH
4913 if (a)
4914 sw = *a;
4915 else if (si && doinit) {
4916 sw = swash_init("utf8", "", si, 1, 0);
4917 (void)av_store(av, 1, sw);
4918 }
9e55ce06
JH
4919 if (b)
4920 alt = *b;
ffc61ed2
JH
4921 }
4922 }
4923
9e55ce06
JH
4924 if (listsvp)
4925 *listsvp = si;
4926 if (altsvp)
4927 *altsvp = alt;
ffc61ed2
JH
4928
4929 return sw;
4930}
4931
4932/*
ba7b4546 4933 - reginclass - determine if a character falls into a character class
832705d4
JH
4934
4935 The n is the ANYOF regnode, the p is the target string, lenp
4936 is pointer to the maximum length of how far to go in the p
4937 (if the lenp is zero, UTF8SKIP(p) is used),
4938 do_utf8 tells whether the target string is in UTF-8.
4939
bbce6d69 4940 */
4941
76e3520e 4942STATIC bool
32fc9b6a 4943S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8)
bbce6d69 4944{
27da23d5 4945 dVAR;
a3b680e6 4946 const char flags = ANYOF_FLAGS(n);
bbce6d69 4947 bool match = FALSE;
cc07378b 4948 UV c = *p;
ae9ddab8 4949 STRLEN len = 0;
9e55ce06 4950 STRLEN plen;
1aa99e6b 4951
19f67299
TS
4952 if (do_utf8 && !UTF8_IS_INVARIANT(c)) {
4953 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
4ad0818d
DM
4954 (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) | UTF8_CHECK_ONLY);
4955 /* see [perl #37836] for UTF8_ALLOW_ANYUV */
19f67299
TS
4956 if (len == (STRLEN)-1)
4957 Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)");
4958 }
bbce6d69 4959
0f0076b4 4960 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
ffc61ed2 4961 if (do_utf8 || (flags & ANYOF_UNICODE)) {
9e55ce06
JH
4962 if (lenp)
4963 *lenp = 0;
ffc61ed2 4964 if (do_utf8 && !ANYOF_RUNTIME(n)) {
ffc61ed2
JH
4965 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4966 match = TRUE;
bbce6d69 4967 }
3568d838 4968 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
1aa99e6b 4969 match = TRUE;
ffc61ed2 4970 if (!match) {
9e55ce06 4971 AV *av;
32fc9b6a 4972 SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av);
ffc61ed2
JH
4973
4974 if (sw) {
3568d838 4975 if (swash_fetch(sw, p, do_utf8))
ffc61ed2
JH
4976 match = TRUE;
4977 else if (flags & ANYOF_FOLD) {
9e55ce06
JH
4978 if (!match && lenp && av) {
4979 I32 i;
9e55ce06 4980 for (i = 0; i <= av_len(av); i++) {
890ce7af 4981 SV* const sv = *av_fetch(av, i, FALSE);
9e55ce06 4982 STRLEN len;
890ce7af 4983 const char * const s = SvPV_const(sv, len);
9e55ce06 4984
061b10df 4985 if (len <= plen && memEQ(s, (char*)p, len)) {
9e55ce06
JH
4986 *lenp = len;
4987 match = TRUE;
4988 break;
4989 }
4990 }
4991 }
4992 if (!match) {
89ebb4a3 4993 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
4a623e43
JH
4994 STRLEN tmplen;
4995
9e55ce06
JH
4996 to_utf8_fold(p, tmpbuf, &tmplen);
4997 if (swash_fetch(sw, tmpbuf, do_utf8))
4998 match = TRUE;
4999 }
ffc61ed2
JH
5000 }
5001 }
bbce6d69 5002 }
9e55ce06 5003 if (match && lenp && *lenp == 0)
0f0076b4 5004 *lenp = UNISKIP(NATIVE_TO_UNI(c));
bbce6d69 5005 }
1aa99e6b 5006 if (!match && c < 256) {
ffc61ed2
JH
5007 if (ANYOF_BITMAP_TEST(n, c))
5008 match = TRUE;
5009 else if (flags & ANYOF_FOLD) {
eb160463 5010 U8 f;
a0ed51b3 5011
ffc61ed2
JH
5012 if (flags & ANYOF_LOCALE) {
5013 PL_reg_flags |= RF_tainted;
5014 f = PL_fold_locale[c];
5015 }
5016 else
5017 f = PL_fold[c];
5018 if (f != c && ANYOF_BITMAP_TEST(n, f))
5019 match = TRUE;
5020 }
5021
5022 if (!match && (flags & ANYOF_CLASS)) {
a0ed51b3 5023 PL_reg_flags |= RF_tainted;
ffc61ed2
JH
5024 if (
5025 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
5026 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
5027 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
5028 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
5029 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
5030 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
5031 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
5032 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
5033 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
5034 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
5035 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
5036 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
5037 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
5038 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
5039 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
5040 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
5041 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
5042 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
5043 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
5044 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
5045 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
5046 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
5047 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
5048 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
5049 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
5050 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
5051 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
5052 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
5053 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
5054 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
5055 ) /* How's that for a conditional? */
5056 {
5057 match = TRUE;
5058 }
a0ed51b3 5059 }
a0ed51b3
LW
5060 }
5061
a0ed51b3
LW
5062 return (flags & ANYOF_INVERT) ? !match : match;
5063}
161b471a 5064
dfe13c55 5065STATIC U8 *
5f66b61c 5066S_reghop3(U8 *s, I32 off, U8* lim)
9041c2e3 5067{
97aff369 5068 dVAR;
a0ed51b3 5069 if (off >= 0) {
1aa99e6b 5070 while (off-- && s < lim) {
ffc61ed2 5071 /* XXX could check well-formedness here */
a0ed51b3 5072 s += UTF8SKIP(s);
ffc61ed2 5073 }
a0ed51b3
LW
5074 }
5075 else {
5076 while (off++) {
1aa99e6b 5077 if (s > lim) {
a0ed51b3 5078 s--;
ffc61ed2 5079 if (UTF8_IS_CONTINUED(*s)) {
1aa99e6b 5080 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
a0ed51b3 5081 s--;
ffc61ed2
JH
5082 }
5083 /* XXX could check well-formedness here */
a0ed51b3
LW
5084 }
5085 }
5086 }
5087 return s;
5088}
161b471a 5089
dfe13c55 5090STATIC U8 *
5f66b61c 5091S_reghopmaybe3(U8* s, I32 off, U8* lim)
a0ed51b3 5092{
97aff369 5093 dVAR;
a0ed51b3 5094 if (off >= 0) {
1aa99e6b 5095 while (off-- && s < lim) {
ffc61ed2 5096 /* XXX could check well-formedness here */
a0ed51b3 5097 s += UTF8SKIP(s);
ffc61ed2 5098 }
a0ed51b3
LW
5099 if (off >= 0)
5100 return 0;
5101 }
5102 else {
5103 while (off++) {
1aa99e6b 5104 if (s > lim) {
a0ed51b3 5105 s--;
ffc61ed2 5106 if (UTF8_IS_CONTINUED(*s)) {
1aa99e6b 5107 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
a0ed51b3 5108 s--;
ffc61ed2
JH
5109 }
5110 /* XXX could check well-formedness here */
a0ed51b3
LW
5111 }
5112 else
5113 break;
5114 }
5115 if (off <= 0)
5116 return 0;
5117 }
5118 return s;
5119}
51371543 5120
51371543 5121static void
acfe0abc 5122restore_pos(pTHX_ void *arg)
51371543 5123{
97aff369 5124 dVAR;
097eb12c 5125 regexp * const rex = (regexp *)arg;
51371543
GS
5126 if (PL_reg_eval_set) {
5127 if (PL_reg_oldsaved) {
4f639d21
DM
5128 rex->subbeg = PL_reg_oldsaved;
5129 rex->sublen = PL_reg_oldsavedlen;
f8c7b90f 5130#ifdef PERL_OLD_COPY_ON_WRITE
4f639d21 5131 rex->saved_copy = PL_nrs;
ed252734 5132#endif
4f639d21 5133 RX_MATCH_COPIED_on(rex);
51371543
GS
5134 }
5135 PL_reg_magic->mg_len = PL_reg_oldpos;
5136 PL_reg_eval_set = 0;
5137 PL_curpm = PL_reg_oldcurpm;
5138 }
5139}
33b8afdf
JH
5140
5141STATIC void
5142S_to_utf8_substr(pTHX_ register regexp *prog)
5143{
33b8afdf 5144 if (prog->float_substr && !prog->float_utf8) {
097eb12c
AL
5145 SV* const sv = newSVsv(prog->float_substr);
5146 prog->float_utf8 = sv;
33b8afdf
JH
5147 sv_utf8_upgrade(sv);
5148 if (SvTAIL(prog->float_substr))
5149 SvTAIL_on(sv);
5150 if (prog->float_substr == prog->check_substr)
5151 prog->check_utf8 = sv;
5152 }
5153 if (prog->anchored_substr && !prog->anchored_utf8) {
097eb12c
AL
5154 SV* const sv = newSVsv(prog->anchored_substr);
5155 prog->anchored_utf8 = sv;
33b8afdf
JH
5156 sv_utf8_upgrade(sv);
5157 if (SvTAIL(prog->anchored_substr))
5158 SvTAIL_on(sv);
5159 if (prog->anchored_substr == prog->check_substr)
5160 prog->check_utf8 = sv;
5161 }
5162}
5163
5164STATIC void
5165S_to_byte_substr(pTHX_ register regexp *prog)
5166{
97aff369 5167 dVAR;
33b8afdf 5168 if (prog->float_utf8 && !prog->float_substr) {
097eb12c
AL
5169 SV* sv = newSVsv(prog->float_utf8);
5170 prog->float_substr = sv;
33b8afdf
JH
5171 if (sv_utf8_downgrade(sv, TRUE)) {
5172 if (SvTAIL(prog->float_utf8))
5173 SvTAIL_on(sv);
5174 } else {
5175 SvREFCNT_dec(sv);
5176 prog->float_substr = sv = &PL_sv_undef;
5177 }
5178 if (prog->float_utf8 == prog->check_utf8)
5179 prog->check_substr = sv;
5180 }
5181 if (prog->anchored_utf8 && !prog->anchored_substr) {
097eb12c
AL
5182 SV* sv = newSVsv(prog->anchored_utf8);
5183 prog->anchored_substr = sv;
33b8afdf
JH
5184 if (sv_utf8_downgrade(sv, TRUE)) {
5185 if (SvTAIL(prog->anchored_utf8))
5186 SvTAIL_on(sv);
5187 } else {
5188 SvREFCNT_dec(sv);
5189 prog->anchored_substr = sv = &PL_sv_undef;
5190 }
5191 if (prog->anchored_utf8 == prog->check_utf8)
5192 prog->check_substr = sv;
5193 }
5194}
66610fdd
RGS
5195
5196/*
5197 * Local variables:
5198 * c-indentation-style: bsd
5199 * c-basic-offset: 4
5200 * indent-tabs-mode: t
5201 * End:
5202 *
37442d52
RGS
5203 * ex: set ts=8 sts=4 sw=4 noet:
5204 */