This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
remove REGMATCH detritus and shrink the size of the backtrack structure
[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.
166f8a29
DM
16 */
17
a687059c
LW
18/* NOTE: this is derived from Henry Spencer's regexp code, and should not
19 * confused with the original package (see point 3 below). Thanks, Henry!
20 */
21
22/* Additional note: this code is very heavily munged from Henry's version
23 * in places. In some spots I've traded clarity for efficiency, so don't
24 * blame Henry for some of the lack of readability.
25 */
26
e50aee73
AD
27/* The names of the functions have been changed from regcomp and
28 * regexec to pregcomp and pregexec in order to avoid conflicts
29 * with the POSIX routines of the same names.
30*/
31
b9d5759e 32#ifdef PERL_EXT_RE_BUILD
54df2634 33#include "re_top.h"
9041c2e3 34#endif
56953603 35
a687059c 36/*
e50aee73 37 * pregcomp and pregexec -- regsub and regerror are not used in perl
a687059c
LW
38 *
39 * Copyright (c) 1986 by University of Toronto.
40 * Written by Henry Spencer. Not derived from licensed software.
41 *
42 * Permission is granted to anyone to use this software for any
43 * purpose on any computer system, and to redistribute it freely,
44 * subject to the following restrictions:
45 *
46 * 1. The author is not responsible for the consequences of use of
47 * this software, no matter how awful, even if they arise
48 * from defects in it.
49 *
50 * 2. The origin of this software must not be misrepresented, either
51 * by explicit claim or by omission.
52 *
53 * 3. Altered versions must be plainly marked as such, and must not
54 * be misrepresented as being the original software.
55 *
56 **** Alterations to Henry's code are...
57 ****
4bb101f2 58 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
b94e2f88 59 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
a687059c 60 ****
9ef589d8
LW
61 **** You may distribute under the terms of either the GNU General Public
62 **** License or the Artistic License, as specified in the README file.
a687059c
LW
63 *
64 * Beware that some of this code is subtly aware of the way operator
65 * precedence is structured in regular expressions. Serious changes in
66 * regular-expression syntax might require a total rethink.
67 */
68#include "EXTERN.h"
864dbfa3 69#define PERL_IN_REGEXEC_C
a687059c 70#include "perl.h"
0f5d15d6 71
54df2634
NC
72#ifdef PERL_IN_XSUB_RE
73# include "re_comp.h"
74#else
75# include "regcomp.h"
76#endif
a687059c 77
c277df42
IZ
78#define RF_tainted 1 /* tainted information used? */
79#define RF_warned 2 /* warned about big count? */
faec1544 80
ab3bbdeb 81#define RF_utf8 8 /* Pattern contains multibyte chars? */
a0ed51b3 82
eb160463 83#define UTF ((PL_reg_flags & RF_utf8) != 0)
ce862d02
IZ
84
85#define RS_init 1 /* eval environment created */
86#define RS_set 2 /* replsv value is set */
c277df42 87
a687059c
LW
88#ifndef STATIC
89#define STATIC static
90#endif
91
32fc9b6a 92#define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
7d3e948e 93
c277df42
IZ
94/*
95 * Forwards.
96 */
97
33b8afdf 98#define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
53c4c00c 99#define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
a0ed51b3 100
3dab1dad
YO
101#define HOPc(pos,off) \
102 (char *)(PL_reg_match_utf8 \
52657f30 103 ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \
3dab1dad
YO
104 : (U8*)(pos + off))
105#define HOPBACKc(pos, off) \
07be1b83
YO
106 (char*)(PL_reg_match_utf8\
107 ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \
108 : (pos - off >= PL_bostr) \
8e11feef 109 ? (U8*)pos - off \
3dab1dad 110 : NULL)
efb30f32 111
e7409c1b 112#define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off))
1aa99e6b 113#define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
1aa99e6b 114
1a4fad37
AL
115#define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
116 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END
117#define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
118#define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
119#define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
120#define LOAD_UTF8_CHARCLASS_MARK() LOAD_UTF8_CHARCLASS(mark, "\xcd\x86")
51371543 121
3dab1dad
YO
122/* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */
123
5f80c4cf 124/* for use after a quantifier and before an EXACT-like node -- japhy */
e2d8ce26
JP
125#define JUMPABLE(rn) ( \
126 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
cca55fe3
JP
127 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
128 OP(rn) == PLUS || OP(rn) == MINMOD || \
3dab1dad 129 (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \
e2d8ce26
JP
130)
131
cca55fe3 132#define HAS_TEXT(rn) ( \
3dab1dad 133 PL_regkind[OP(rn)] == EXACT || PL_regkind[OP(rn)] == REF \
cca55fe3 134)
e2d8ce26 135
a84d97b6
HS
136/*
137 Search for mandatory following text node; for lookahead, the text must
138 follow but for lookbehind (rn->flags != 0) we skip to the next step.
139*/
cca55fe3 140#define FIND_NEXT_IMPT(rn) STMT_START { \
3dab1dad
YO
141 while (JUMPABLE(rn)) { \
142 const OPCODE type = OP(rn); \
143 if (type == SUSPEND || PL_regkind[type] == CURLY) \
e2d8ce26 144 rn = NEXTOPER(NEXTOPER(rn)); \
3dab1dad 145 else if (type == PLUS) \
cca55fe3 146 rn = NEXTOPER(rn); \
3dab1dad 147 else if (type == IFMATCH) \
a84d97b6 148 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
e2d8ce26 149 else rn += NEXT_OFF(rn); \
3dab1dad 150 } \
5f80c4cf 151} STMT_END
74750237 152
c476f425 153
acfe0abc 154static void restore_pos(pTHX_ void *arg);
51371543 155
76e3520e 156STATIC CHECKPOINT
cea2e8a9 157S_regcppush(pTHX_ I32 parenfloor)
a0d0e21e 158{
97aff369 159 dVAR;
a3b680e6 160 const int retval = PL_savestack_ix;
b1ce53c5 161#define REGCP_PAREN_ELEMS 4
a3b680e6 162 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
a0d0e21e 163 int p;
40a82448 164 GET_RE_DEBUG_FLAGS_DECL;
a0d0e21e 165
e49a9654
IH
166 if (paren_elems_to_push < 0)
167 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
168
6bda09f9 169#define REGCP_OTHER_ELEMS 8
4b3c1a47 170 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
3280af22 171 for (p = PL_regsize; p > parenfloor; p--) {
b1ce53c5 172/* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
cf93c79d
IZ
173 SSPUSHINT(PL_regendp[p]);
174 SSPUSHINT(PL_regstartp[p]);
3280af22 175 SSPUSHPTR(PL_reg_start_tmp[p]);
a0d0e21e 176 SSPUSHINT(p);
40a82448
DM
177 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
178 " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n",
179 (UV)p, (IV)PL_regstartp[p],
180 (IV)(PL_reg_start_tmp[p] - PL_bostr),
181 (IV)PL_regendp[p]
182 ));
a0d0e21e 183 }
b1ce53c5 184/* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
6bda09f9
YO
185 SSPUSHPTR(PL_regstartp);
186 SSPUSHPTR(PL_regendp);
3280af22
NIS
187 SSPUSHINT(PL_regsize);
188 SSPUSHINT(*PL_reglastparen);
a01268b5 189 SSPUSHINT(*PL_reglastcloseparen);
3280af22 190 SSPUSHPTR(PL_reginput);
41123dfd
JH
191#define REGCP_FRAME_ELEMS 2
192/* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
193 * are needed for the regexp context stack bookkeeping. */
194 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
b1ce53c5 195 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
41123dfd 196
a0d0e21e
LW
197 return retval;
198}
199
c277df42 200/* These are needed since we do not localize EVAL nodes: */
ab3bbdeb
YO
201#define REGCP_SET(cp) \
202 DEBUG_STATE_r( \
ab3bbdeb 203 PerlIO_printf(Perl_debug_log, \
e4f74956 204 " Setting an EVAL scope, savestack=%"IVdf"\n", \
ab3bbdeb
YO
205 (IV)PL_savestack_ix)); \
206 cp = PL_savestack_ix
c3464db5 207
ab3bbdeb 208#define REGCP_UNWIND(cp) \
e4f74956 209 DEBUG_STATE_r( \
ab3bbdeb 210 if (cp != PL_savestack_ix) \
e4f74956
YO
211 PerlIO_printf(Perl_debug_log, \
212 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
ab3bbdeb
YO
213 (IV)(cp), (IV)PL_savestack_ix)); \
214 regcpblow(cp)
c277df42 215
76e3520e 216STATIC char *
097eb12c 217S_regcppop(pTHX_ const regexp *rex)
a0d0e21e 218{
97aff369 219 dVAR;
b1ce53c5 220 I32 i;
a0d0e21e 221 char *input;
b1ce53c5 222
a3621e74
YO
223 GET_RE_DEBUG_FLAGS_DECL;
224
b1ce53c5 225 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
a0d0e21e 226 i = SSPOPINT;
b1ce53c5
JH
227 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
228 i = SSPOPINT; /* Parentheses elements to pop. */
a0d0e21e 229 input = (char *) SSPOPPTR;
a01268b5 230 *PL_reglastcloseparen = SSPOPINT;
3280af22
NIS
231 *PL_reglastparen = SSPOPINT;
232 PL_regsize = SSPOPINT;
6bda09f9
YO
233 PL_regendp=(I32 *) SSPOPPTR;
234 PL_regstartp=(I32 *) SSPOPPTR;
b1ce53c5 235
6bda09f9 236
b1ce53c5 237 /* Now restore the parentheses context. */
41123dfd
JH
238 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
239 i > 0; i -= REGCP_PAREN_ELEMS) {
1df70142 240 I32 tmps;
097eb12c 241 U32 paren = (U32)SSPOPINT;
3280af22 242 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
cf93c79d
IZ
243 PL_regstartp[paren] = SSPOPINT;
244 tmps = SSPOPINT;
3280af22
NIS
245 if (paren <= *PL_reglastparen)
246 PL_regendp[paren] = tmps;
a3621e74 247 DEBUG_EXECUTE_r(
c3464db5 248 PerlIO_printf(Perl_debug_log,
b900a521 249 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
9041c2e3 250 (UV)paren, (IV)PL_regstartp[paren],
b900a521 251 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
9041c2e3 252 (IV)PL_regendp[paren],
3280af22 253 (paren > *PL_reglastparen ? "(no)" : ""));
c277df42 254 );
a0d0e21e 255 }
a3621e74 256 DEBUG_EXECUTE_r(
bb7a0f54 257 if (*PL_reglastparen + 1 <= rex->nparens) {
c3464db5 258 PerlIO_printf(Perl_debug_log,
faccc32b 259 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
4f639d21 260 (IV)(*PL_reglastparen + 1), (IV)rex->nparens);
c277df42
IZ
261 }
262 );
daf18116 263#if 1
dafc8851
JH
264 /* It would seem that the similar code in regtry()
265 * already takes care of this, and in fact it is in
266 * a better location to since this code can #if 0-ed out
267 * but the code in regtry() is needed or otherwise tests
268 * requiring null fields (pat.t#187 and split.t#{13,14}
daf18116
JH
269 * (as of patchlevel 7877) will fail. Then again,
270 * this code seems to be necessary or otherwise
271 * building DynaLoader will fail:
272 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
273 * --jhi */
bb7a0f54 274 for (i = *PL_reglastparen + 1; (U32)i <= rex->nparens; i++) {
097eb12c
AL
275 if (i > PL_regsize)
276 PL_regstartp[i] = -1;
277 PL_regendp[i] = -1;
a0d0e21e 278 }
dafc8851 279#endif
a0d0e21e
LW
280 return input;
281}
282
02db2b7b 283#define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
a0d0e21e 284
a687059c 285/*
e50aee73 286 * pregexec and friends
a687059c
LW
287 */
288
76234dfb 289#ifndef PERL_IN_XSUB_RE
a687059c 290/*
c277df42 291 - pregexec - match a regexp against a string
a687059c 292 */
c277df42 293I32
864dbfa3 294Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
c3464db5 295 char *strbeg, I32 minend, SV *screamer, U32 nosave)
c277df42
IZ
296/* strend: pointer to null at end of string */
297/* strbeg: real beginning of string */
298/* minend: end of match must be >=minend after stringarg. */
299/* nosave: For optimizations. */
300{
301 return
9041c2e3 302 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
c277df42
IZ
303 nosave ? 0 : REXEC_COPY_STR);
304}
76234dfb 305#endif
22e551b9 306
9041c2e3 307/*
cad2e5aa
JH
308 * Need to implement the following flags for reg_anch:
309 *
310 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
311 * USE_INTUIT_ML
312 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
313 * INTUIT_AUTORITATIVE_ML
314 * INTUIT_ONCE_NOML - Intuit can match in one location only.
315 * INTUIT_ONCE_ML
316 *
317 * Another flag for this function: SECOND_TIME (so that float substrs
318 * with giant delta may be not rechecked).
319 */
320
321/* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
322
3f7c398e 323/* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
cad2e5aa
JH
324 Otherwise, only SvCUR(sv) is used to get strbeg. */
325
326/* XXXX We assume that strpos is strbeg unless sv. */
327
6eb5f6b9
JH
328/* XXXX Some places assume that there is a fixed substring.
329 An update may be needed if optimizer marks as "INTUITable"
330 RExen without fixed substrings. Similarly, it is assumed that
331 lengths of all the strings are no more than minlen, thus they
332 cannot come from lookahead.
333 (Or minlen should take into account lookahead.) */
334
2c2d71f5
JH
335/* A failure to find a constant substring means that there is no need to make
336 an expensive call to REx engine, thus we celebrate a failure. Similarly,
337 finding a substring too deep into the string means that less calls to
30944b6d
IZ
338 regtry() should be needed.
339
340 REx compiler's optimizer found 4 possible hints:
341 a) Anchored substring;
342 b) Fixed substring;
343 c) Whether we are anchored (beginning-of-line or \G);
344 d) First node (of those at offset 0) which may distingush positions;
6eb5f6b9 345 We use a)b)d) and multiline-part of c), and try to find a position in the
30944b6d
IZ
346 string which does not contradict any of them.
347 */
2c2d71f5 348
6eb5f6b9
JH
349/* Most of decisions we do here should have been done at compile time.
350 The nodes of the REx which we used for the search should have been
351 deleted from the finite automaton. */
352
cad2e5aa
JH
353char *
354Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
355 char *strend, U32 flags, re_scream_pos_data *data)
356{
97aff369 357 dVAR;
b7953727 358 register I32 start_shift = 0;
cad2e5aa 359 /* Should be nonnegative! */
b7953727 360 register I32 end_shift = 0;
2c2d71f5
JH
361 register char *s;
362 register SV *check;
a1933d95 363 char *strbeg;
cad2e5aa 364 char *t;
1de06328 365 const bool do_utf8 = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */
cad2e5aa 366 I32 ml_anch;
bd61b366
SS
367 register char *other_last = NULL; /* other substr checked before this */
368 char *check_at = NULL; /* check substr found at this pos */
1df70142 369 const I32 multiline = prog->reganch & PMf_MULTILINE;
30944b6d 370#ifdef DEBUGGING
890ce7af 371 const char * const i_strpos = strpos;
30944b6d 372#endif
a3621e74
YO
373
374 GET_RE_DEBUG_FLAGS_DECL;
375
a30b2f1f 376 RX_MATCH_UTF8_set(prog,do_utf8);
cad2e5aa 377
b8d68ded 378 if (prog->reganch & ROPT_UTF8) {
b8d68ded
JH
379 PL_reg_flags |= RF_utf8;
380 }
ab3bbdeb
YO
381 DEBUG_EXECUTE_r(
382 debug_start_match(prog, do_utf8, strpos, strend,
1de06328
YO
383 sv ? "Guessing start of match in sv for"
384 : "Guessing start of match in string for");
2a782b5b 385 );
cad2e5aa 386
c344f387
JH
387 /* CHR_DIST() would be more correct here but it makes things slow. */
388 if (prog->minlen > strend - strpos) {
a3621e74 389 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
a72c7584 390 "String too short... [re_intuit_start]\n"));
cad2e5aa 391 goto fail;
2c2d71f5 392 }
1de06328 393
a1933d95 394 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
1aa99e6b 395 PL_regeol = strend;
33b8afdf
JH
396 if (do_utf8) {
397 if (!prog->check_utf8 && prog->check_substr)
398 to_utf8_substr(prog);
399 check = prog->check_utf8;
400 } else {
401 if (!prog->check_substr && prog->check_utf8)
402 to_byte_substr(prog);
403 check = prog->check_substr;
404 }
1de06328 405 if (check == &PL_sv_undef) {
a3621e74 406 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1de06328 407 "Non-utf8 string cannot match utf8 check string\n"));
33b8afdf
JH
408 goto fail;
409 }
2c2d71f5 410 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
cad2e5aa
JH
411 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
412 || ( (prog->reganch & ROPT_ANCH_BOL)
7fba1cd6 413 && !multiline ) ); /* Check after \n? */
cad2e5aa 414
7e25d62c
JH
415 if (!ml_anch) {
416 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
417 | ROPT_IMPLICIT)) /* not a real BOL */
3f7c398e 418 /* SvCUR is not set on references: SvRV and SvPVX_const overlap */
7e25d62c
JH
419 && sv && !SvROK(sv)
420 && (strpos != strbeg)) {
a3621e74 421 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
7e25d62c
JH
422 goto fail;
423 }
424 if (prog->check_offset_min == prog->check_offset_max &&
cce850e4 425 !(prog->reganch & ROPT_CANY_SEEN)) {
2c2d71f5 426 /* Substring at constant offset from beg-of-str... */
cad2e5aa
JH
427 I32 slen;
428
1aa99e6b 429 s = HOP3c(strpos, prog->check_offset_min, strend);
1de06328 430
653099ff
GS
431 if (SvTAIL(check)) {
432 slen = SvCUR(check); /* >= 1 */
cad2e5aa 433
9041c2e3 434 if ( strend - s > slen || strend - s < slen - 1
2c2d71f5 435 || (strend - s == slen && strend[-1] != '\n')) {
a3621e74 436 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
2c2d71f5 437 goto fail_finish;
cad2e5aa
JH
438 }
439 /* Now should match s[0..slen-2] */
440 slen--;
3f7c398e 441 if (slen && (*SvPVX_const(check) != *s
cad2e5aa 442 || (slen > 1
3f7c398e 443 && memNE(SvPVX_const(check), s, slen)))) {
2c2d71f5 444 report_neq:
a3621e74 445 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
2c2d71f5
JH
446 goto fail_finish;
447 }
cad2e5aa 448 }
3f7c398e 449 else if (*SvPVX_const(check) != *s
653099ff 450 || ((slen = SvCUR(check)) > 1
3f7c398e 451 && memNE(SvPVX_const(check), s, slen)))
2c2d71f5 452 goto report_neq;
c315bfe8 453 check_at = s;
2c2d71f5 454 goto success_at_start;
7e25d62c 455 }
cad2e5aa 456 }
2c2d71f5 457 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
cad2e5aa 458 s = strpos;
2c2d71f5 459 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
1de06328
YO
460 end_shift = prog->check_end_shift;
461
2c2d71f5 462 if (!ml_anch) {
a3b680e6 463 const I32 end = prog->check_offset_max + CHR_SVLEN(check)
653099ff 464 - (SvTAIL(check) != 0);
a3b680e6 465 const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
2c2d71f5
JH
466
467 if (end_shift < eshift)
468 end_shift = eshift;
469 }
cad2e5aa 470 }
2c2d71f5 471 else { /* Can match at random position */
cad2e5aa
JH
472 ml_anch = 0;
473 s = strpos;
1de06328
YO
474 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
475 end_shift = prog->check_end_shift;
476
477 /* end shift should be non negative here */
cad2e5aa
JH
478 }
479
2c2d71f5 480#ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
0033605d 481 if (end_shift < 0)
1de06328 482 Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ",
b57a0404 483 (IV)end_shift, prog->precomp);
2c2d71f5
JH
484#endif
485
2c2d71f5
JH
486 restart:
487 /* Find a possible match in the region s..strend by looking for
488 the "check" substring in the region corrected by start/end_shift. */
1de06328
YO
489
490 {
491 I32 srch_start_shift = start_shift;
492 I32 srch_end_shift = end_shift;
493 if (srch_start_shift < 0 && strbeg - s > srch_start_shift) {
494 srch_end_shift -= ((strbeg - s) - srch_start_shift);
495 srch_start_shift = strbeg - s;
496 }
6bda09f9 497 DEBUG_OPTIMISE_MORE_r({
1de06328
YO
498 PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n",
499 (IV)prog->check_offset_min,
500 (IV)srch_start_shift,
501 (IV)srch_end_shift,
502 (IV)prog->check_end_shift);
503 });
504
cad2e5aa 505 if (flags & REXEC_SCREAM) {
cad2e5aa 506 I32 p = -1; /* Internal iterator of scream. */
a3b680e6 507 I32 * const pp = data ? data->scream_pos : &p;
cad2e5aa 508
2c2d71f5
JH
509 if (PL_screamfirst[BmRARE(check)] >= 0
510 || ( BmRARE(check) == '\n'
511 && (BmPREVIOUS(check) == SvCUR(check) - 1)
512 && SvTAIL(check) ))
9041c2e3 513 s = screaminstr(sv, check,
1de06328 514 srch_start_shift + (s - strbeg), srch_end_shift, pp, 0);
cad2e5aa 515 else
2c2d71f5 516 goto fail_finish;
4addbd3b
HS
517 /* we may be pointing at the wrong string */
518 if (s && RX_MATCH_COPIED(prog))
3f7c398e 519 s = strbeg + (s - SvPVX_const(sv));
cad2e5aa
JH
520 if (data)
521 *data->scream_olds = s;
522 }
1de06328
YO
523 else {
524 U8* start_point;
525 U8* end_point;
526 if (prog->reganch & ROPT_CANY_SEEN) {
527 start_point= (U8*)(s + srch_start_shift);
528 end_point= (U8*)(strend - srch_end_shift);
529 } else {
530 start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend);
531 end_point= HOP3(strend, -srch_end_shift, strbeg);
532 }
6bda09f9 533 DEBUG_OPTIMISE_MORE_r({
56570a2c 534 PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n",
1de06328 535 (int)(end_point - start_point),
fc8cd66c 536 (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point),
1de06328
YO
537 start_point);
538 });
539
540 s = fbm_instr( start_point, end_point,
7fba1cd6 541 check, multiline ? FBMrf_MULTILINE : 0);
1de06328
YO
542 }
543 }
cad2e5aa
JH
544 /* Update the count-of-usability, remove useless subpatterns,
545 unshift s. */
2c2d71f5 546
ab3bbdeb
YO
547 DEBUG_EXECUTE_r({
548 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
549 SvPVX_const(check), RE_SV_DUMPLEN(check), 30);
550 PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s",
2c2d71f5 551 (s ? "Found" : "Did not find"),
ab3bbdeb
YO
552 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)
553 ? "anchored" : "floating"),
554 quoted,
555 RE_SV_TAIL(check),
556 (s ? " at offset " : "...\n") );
557 });
2c2d71f5
JH
558
559 if (!s)
560 goto fail_finish;
2c2d71f5 561 /* Finish the diagnostic message */
a3621e74 562 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
2c2d71f5 563
1de06328
YO
564 /* XXX dmq: first branch is for positive lookbehind...
565 Our check string is offset from the beginning of the pattern.
566 So we need to do any stclass tests offset forward from that
567 point. I think. :-(
568 */
569
570
571
572 check_at=s;
573
574
2c2d71f5
JH
575 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
576 Start with the other substr.
577 XXXX no SCREAM optimization yet - and a very coarse implementation
a0288114 578 XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
2c2d71f5
JH
579 *always* match. Probably should be marked during compile...
580 Probably it is right to do no SCREAM here...
581 */
582
1de06328
YO
583 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8)
584 : (prog->float_substr && prog->anchored_substr))
585 {
30944b6d 586 /* Take into account the "other" substring. */
2c2d71f5
JH
587 /* XXXX May be hopelessly wrong for UTF... */
588 if (!other_last)
6eb5f6b9 589 other_last = strpos;
33b8afdf 590 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
30944b6d
IZ
591 do_other_anchored:
592 {
890ce7af
AL
593 char * const last = HOP3c(s, -start_shift, strbeg);
594 char *last1, *last2;
be8e71aa 595 char * const saved_s = s;
33b8afdf 596 SV* must;
2c2d71f5 597
2c2d71f5
JH
598 t = s - prog->check_offset_max;
599 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
1d86a7f9 600 && (!do_utf8
0ce71af7 601 || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos))
2c2d71f5 602 && t > strpos)))
6f207bd3 603 NOOP;
2c2d71f5
JH
604 else
605 t = strpos;
1aa99e6b 606 t = HOP3c(t, prog->anchored_offset, strend);
6eb5f6b9
JH
607 if (t < other_last) /* These positions already checked */
608 t = other_last;
1aa99e6b 609 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
2c2d71f5
JH
610 if (last < last1)
611 last1 = last;
1de06328
YO
612 /* XXXX It is not documented what units *_offsets are in.
613 We assume bytes, but this is clearly wrong.
614 Meaning this code needs to be carefully reviewed for errors.
615 dmq.
616 */
617
2c2d71f5 618 /* On end-of-str: see comment below. */
33b8afdf
JH
619 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
620 if (must == &PL_sv_undef) {
621 s = (char*)NULL;
1de06328 622 DEBUG_r(must = prog->anchored_utf8); /* for debug */
33b8afdf
JH
623 }
624 else
625 s = fbm_instr(
626 (unsigned char*)t,
627 HOP3(HOP3(last1, prog->anchored_offset, strend)
628 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
629 must,
7fba1cd6 630 multiline ? FBMrf_MULTILINE : 0
33b8afdf 631 );
ab3bbdeb
YO
632 DEBUG_EXECUTE_r({
633 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
634 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
635 PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s",
2c2d71f5 636 (s ? "Found" : "Contradicts"),
ab3bbdeb
YO
637 quoted, RE_SV_TAIL(must));
638 });
639
640
2c2d71f5
JH
641 if (!s) {
642 if (last1 >= last2) {
a3621e74 643 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2c2d71f5
JH
644 ", giving up...\n"));
645 goto fail_finish;
646 }
a3621e74 647 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2c2d71f5 648 ", trying floating at offset %ld...\n",
be8e71aa 649 (long)(HOP3c(saved_s, 1, strend) - i_strpos)));
1aa99e6b
IH
650 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
651 s = HOP3c(last, 1, strend);
2c2d71f5
JH
652 goto restart;
653 }
654 else {
a3621e74 655 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
30944b6d 656 (long)(s - i_strpos)));
1aa99e6b
IH
657 t = HOP3c(s, -prog->anchored_offset, strbeg);
658 other_last = HOP3c(s, 1, strend);
be8e71aa 659 s = saved_s;
2c2d71f5
JH
660 if (t == strpos)
661 goto try_at_start;
2c2d71f5
JH
662 goto try_at_offset;
663 }
30944b6d 664 }
2c2d71f5
JH
665 }
666 else { /* Take into account the floating substring. */
33b8afdf 667 char *last, *last1;
be8e71aa 668 char * const saved_s = s;
33b8afdf
JH
669 SV* must;
670
671 t = HOP3c(s, -start_shift, strbeg);
672 last1 = last =
673 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
674 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
675 last = HOP3c(t, prog->float_max_offset, strend);
676 s = HOP3c(t, prog->float_min_offset, strend);
677 if (s < other_last)
678 s = other_last;
2c2d71f5 679 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
33b8afdf
JH
680 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
681 /* fbm_instr() takes into account exact value of end-of-str
682 if the check is SvTAIL(ed). Since false positives are OK,
683 and end-of-str is not later than strend we are OK. */
684 if (must == &PL_sv_undef) {
685 s = (char*)NULL;
1de06328 686 DEBUG_r(must = prog->float_utf8); /* for debug message */
33b8afdf
JH
687 }
688 else
2c2d71f5 689 s = fbm_instr((unsigned char*)s,
33b8afdf
JH
690 (unsigned char*)last + SvCUR(must)
691 - (SvTAIL(must)!=0),
7fba1cd6 692 must, multiline ? FBMrf_MULTILINE : 0);
ab3bbdeb
YO
693 DEBUG_EXECUTE_r({
694 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
695 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
696 PerlIO_printf(Perl_debug_log, "%s floating substr %s%s",
33b8afdf 697 (s ? "Found" : "Contradicts"),
ab3bbdeb
YO
698 quoted, RE_SV_TAIL(must));
699 });
33b8afdf
JH
700 if (!s) {
701 if (last1 == last) {
a3621e74 702 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
33b8afdf
JH
703 ", giving up...\n"));
704 goto fail_finish;
2c2d71f5 705 }
a3621e74 706 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
33b8afdf 707 ", trying anchored starting at offset %ld...\n",
be8e71aa 708 (long)(saved_s + 1 - i_strpos)));
33b8afdf
JH
709 other_last = last;
710 s = HOP3c(t, 1, strend);
711 goto restart;
712 }
713 else {
a3621e74 714 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
33b8afdf
JH
715 (long)(s - i_strpos)));
716 other_last = s; /* Fix this later. --Hugo */
be8e71aa 717 s = saved_s;
33b8afdf
JH
718 if (t == strpos)
719 goto try_at_start;
720 goto try_at_offset;
721 }
2c2d71f5 722 }
cad2e5aa 723 }
2c2d71f5 724
1de06328 725
9ef43ace 726 t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos);
1de06328 727
6bda09f9 728 DEBUG_OPTIMISE_MORE_r(
1de06328
YO
729 PerlIO_printf(Perl_debug_log,
730 "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n",
731 (IV)prog->check_offset_min,
732 (IV)prog->check_offset_max,
733 (IV)(s-strpos),
734 (IV)(t-strpos),
735 (IV)(t-s),
736 (IV)(strend-strpos)
737 )
738 );
739
2c2d71f5 740 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
1d86a7f9 741 && (!do_utf8
9ef43ace 742 || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos)))
1de06328
YO
743 && t > strpos)))
744 {
2c2d71f5
JH
745 /* Fixed substring is found far enough so that the match
746 cannot start at strpos. */
747 try_at_offset:
cad2e5aa 748 if (ml_anch && t[-1] != '\n') {
30944b6d
IZ
749 /* Eventually fbm_*() should handle this, but often
750 anchored_offset is not 0, so this check will not be wasted. */
751 /* XXXX In the code below we prefer to look for "^" even in
752 presence of anchored substrings. And we search even
753 beyond the found float position. These pessimizations
754 are historical artefacts only. */
755 find_anchor:
2c2d71f5 756 while (t < strend - prog->minlen) {
cad2e5aa 757 if (*t == '\n') {
4ee3650e 758 if (t < check_at - prog->check_offset_min) {
33b8afdf 759 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
4ee3650e
GS
760 /* Since we moved from the found position,
761 we definitely contradict the found anchored
30944b6d
IZ
762 substr. Due to the above check we do not
763 contradict "check" substr.
764 Thus we can arrive here only if check substr
765 is float. Redo checking for "other"=="fixed".
766 */
9041c2e3 767 strpos = t + 1;
a3621e74 768 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
e4584336 769 PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
30944b6d
IZ
770 goto do_other_anchored;
771 }
4ee3650e
GS
772 /* We don't contradict the found floating substring. */
773 /* XXXX Why not check for STCLASS? */
cad2e5aa 774 s = t + 1;
a3621e74 775 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
e4584336 776 PL_colors[0], PL_colors[1], (long)(s - i_strpos)));
cad2e5aa
JH
777 goto set_useful;
778 }
4ee3650e
GS
779 /* Position contradicts check-string */
780 /* XXXX probably better to look for check-string
781 than for "\n", so one should lower the limit for t? */
a3621e74 782 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
e4584336 783 PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos)));
0e41cd87 784 other_last = strpos = s = t + 1;
cad2e5aa
JH
785 goto restart;
786 }
787 t++;
788 }
a3621e74 789 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
e4584336 790 PL_colors[0], PL_colors[1]));
2c2d71f5 791 goto fail_finish;
cad2e5aa 792 }
f5952150 793 else {
a3621e74 794 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
e4584336 795 PL_colors[0], PL_colors[1]));
f5952150 796 }
cad2e5aa
JH
797 s = t;
798 set_useful:
33b8afdf 799 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
cad2e5aa
JH
800 }
801 else {
f5952150 802 /* The found string does not prohibit matching at strpos,
2c2d71f5 803 - no optimization of calling REx engine can be performed,
f5952150
GS
804 unless it was an MBOL and we are not after MBOL,
805 or a future STCLASS check will fail this. */
2c2d71f5
JH
806 try_at_start:
807 /* Even in this situation we may use MBOL flag if strpos is offset
808 wrt the start of the string. */
05b4157f 809 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
a1933d95 810 && (strpos != strbeg) && strpos[-1] != '\n'
d506a20d
IZ
811 /* May be due to an implicit anchor of m{.*foo} */
812 && !(prog->reganch & ROPT_IMPLICIT))
813 {
cad2e5aa
JH
814 t = strpos;
815 goto find_anchor;
816 }
a3621e74 817 DEBUG_EXECUTE_r( if (ml_anch)
f5952150 818 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
e4584336 819 (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]);
30944b6d 820 );
2c2d71f5 821 success_at_start:
30944b6d 822 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
33b8afdf
JH
823 && (do_utf8 ? (
824 prog->check_utf8 /* Could be deleted already */
825 && --BmUSEFUL(prog->check_utf8) < 0
826 && (prog->check_utf8 == prog->float_utf8)
827 ) : (
828 prog->check_substr /* Could be deleted already */
829 && --BmUSEFUL(prog->check_substr) < 0
830 && (prog->check_substr == prog->float_substr)
831 )))
66e933ab 832 {
cad2e5aa 833 /* If flags & SOMETHING - do not do it many times on the same match */
a3621e74 834 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
33b8afdf
JH
835 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
836 if (do_utf8 ? prog->check_substr : prog->check_utf8)
837 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
a0714e2c
SS
838 prog->check_substr = prog->check_utf8 = NULL; /* disable */
839 prog->float_substr = prog->float_utf8 = NULL; /* clear */
840 check = NULL; /* abort */
cad2e5aa 841 s = strpos;
3cf5c195
IZ
842 /* XXXX This is a remnant of the old implementation. It
843 looks wasteful, since now INTUIT can use many
6eb5f6b9 844 other heuristics. */
cad2e5aa
JH
845 prog->reganch &= ~RE_USE_INTUIT;
846 }
847 else
848 s = strpos;
849 }
850
6eb5f6b9
JH
851 /* Last resort... */
852 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
1de06328
YO
853 /* trie stclasses are too expensive to use here, we are better off to
854 leave it to regmatch itself */
786e8c11 855 if (prog->regstclass && PL_regkind[OP(prog->regstclass)]!=TRIE) {
6eb5f6b9
JH
856 /* minlen == 0 is possible if regstclass is \b or \B,
857 and the fixed substr is ''$.
858 Since minlen is already taken into account, s+1 is before strend;
859 accidentally, minlen >= 1 guaranties no false positives at s + 1
860 even for \b or \B. But (minlen? 1 : 0) below assumes that
861 regstclass does not come from lookahead... */
862 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
863 This leaves EXACTF only, which is dealt with in find_byclass(). */
890ce7af 864 const U8* const str = (U8*)STRING(prog->regstclass);
3dab1dad 865 const int cl_l = (PL_regkind[OP(prog->regstclass)] == EXACT
1aa99e6b 866 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
66e933ab 867 : 1);
1de06328
YO
868 char * endpos;
869 if (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
870 endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend);
871 else if (prog->float_substr || prog->float_utf8)
872 endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend);
873 else
874 endpos= strend;
875
56570a2c 876 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %d s: %d endpos: %d\n",
b57a0404 877 (IV)start_shift, check_at - strbeg, s - strbeg, endpos - strbeg));
1de06328 878
6eb5f6b9 879 t = s;
3b0527fe 880 s = find_byclass(prog, prog->regstclass, s, endpos, NULL);
6eb5f6b9
JH
881 if (!s) {
882#ifdef DEBUGGING
cbbf8932 883 const char *what = NULL;
6eb5f6b9
JH
884#endif
885 if (endpos == strend) {
a3621e74 886 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
6eb5f6b9
JH
887 "Could not match STCLASS...\n") );
888 goto fail;
889 }
a3621e74 890 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
66e933ab 891 "This position contradicts STCLASS...\n") );
653099ff
GS
892 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
893 goto fail;
6eb5f6b9 894 /* Contradict one of substrings */
33b8afdf
JH
895 if (prog->anchored_substr || prog->anchored_utf8) {
896 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
a3621e74 897 DEBUG_EXECUTE_r( what = "anchored" );
6eb5f6b9 898 hop_and_restart:
1aa99e6b 899 s = HOP3c(t, 1, strend);
66e933ab
GS
900 if (s + start_shift + end_shift > strend) {
901 /* XXXX Should be taken into account earlier? */
a3621e74 902 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
66e933ab
GS
903 "Could not match STCLASS...\n") );
904 goto fail;
905 }
5e39e1e5
HS
906 if (!check)
907 goto giveup;
a3621e74 908 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
f5952150 909 "Looking for %s substr starting at offset %ld...\n",
6eb5f6b9
JH
910 what, (long)(s + start_shift - i_strpos)) );
911 goto restart;
912 }
66e933ab 913 /* Have both, check_string is floating */
6eb5f6b9
JH
914 if (t + start_shift >= check_at) /* Contradicts floating=check */
915 goto retry_floating_check;
916 /* Recheck anchored substring, but not floating... */
9041c2e3 917 s = check_at;
5e39e1e5
HS
918 if (!check)
919 goto giveup;
a3621e74 920 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
f5952150 921 "Looking for anchored substr starting at offset %ld...\n",
6eb5f6b9
JH
922 (long)(other_last - i_strpos)) );
923 goto do_other_anchored;
924 }
60e71179
GS
925 /* Another way we could have checked stclass at the
926 current position only: */
927 if (ml_anch) {
928 s = t = t + 1;
5e39e1e5
HS
929 if (!check)
930 goto giveup;
a3621e74 931 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
f5952150 932 "Looking for /%s^%s/m starting at offset %ld...\n",
e4584336 933 PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
60e71179 934 goto try_at_offset;
66e933ab 935 }
33b8afdf 936 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
60e71179 937 goto fail;
6eb5f6b9
JH
938 /* Check is floating subtring. */
939 retry_floating_check:
940 t = check_at - start_shift;
a3621e74 941 DEBUG_EXECUTE_r( what = "floating" );
6eb5f6b9
JH
942 goto hop_and_restart;
943 }
b7953727 944 if (t != s) {
a3621e74 945 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
6eb5f6b9 946 "By STCLASS: moving %ld --> %ld\n",
b7953727
JH
947 (long)(t - i_strpos), (long)(s - i_strpos))
948 );
949 }
950 else {
a3621e74 951 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
b7953727
JH
952 "Does not contradict STCLASS...\n");
953 );
954 }
6eb5f6b9 955 }
5e39e1e5 956 giveup:
a3621e74 957 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
5e39e1e5
HS
958 PL_colors[4], (check ? "Guessed" : "Giving up"),
959 PL_colors[5], (long)(s - i_strpos)) );
cad2e5aa 960 return s;
2c2d71f5
JH
961
962 fail_finish: /* Substring not found */
33b8afdf
JH
963 if (prog->check_substr || prog->check_utf8) /* could be removed already */
964 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
cad2e5aa 965 fail:
a3621e74 966 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
e4584336 967 PL_colors[4], PL_colors[5]));
bd61b366 968 return NULL;
cad2e5aa 969}
9661b544 970
786e8c11 971
3b0527fe 972
4cadc6a9
YO
973#define REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len, uvc, charid, \
974foldlen, foldbuf, uniflags) STMT_START { \
975 switch (trie_type) { \
976 case trie_utf8_fold: \
977 if ( foldlen>0 ) { \
978 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
979 foldlen -= len; \
980 uscan += len; \
981 len=0; \
982 } else { \
983 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
984 uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \
985 foldlen -= UNISKIP( uvc ); \
986 uscan = foldbuf + UNISKIP( uvc ); \
987 } \
988 break; \
989 case trie_utf8: \
990 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
991 break; \
992 case trie_plain: \
993 uvc = (UV)*uc; \
994 len = 1; \
995 } \
996 \
997 if (uvc < 256) { \
998 charid = trie->charmap[ uvc ]; \
999 } \
1000 else { \
1001 charid = 0; \
1002 if (trie->widecharmap) { \
1003 SV** const svpp = hv_fetch(trie->widecharmap, \
1004 (char*)&uvc, sizeof(UV), 0); \
1005 if (svpp) \
1006 charid = (U16)SvIV(*svpp); \
1007 } \
1008 } \
1009} STMT_END
1010
1011#define REXEC_FBC_EXACTISH_CHECK(CoNd) \
1012 if ( (CoNd) \
1013 && (ln == len || \
1014 ibcmp_utf8(s, NULL, 0, do_utf8, \
1015 m, NULL, ln, (bool)UTF)) \
1016 && (!reginfo || regtry(reginfo, s)) ) \
1017 goto got_it; \
1018 else { \
1019 U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \
1020 uvchr_to_utf8(tmpbuf, c); \
1021 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); \
1022 if ( f != c \
1023 && (f == c1 || f == c2) \
1024 && (ln == foldlen || \
1025 !ibcmp_utf8((char *) foldbuf, \
1026 NULL, foldlen, do_utf8, \
1027 m, \
1028 NULL, ln, (bool)UTF)) \
1029 && (!reginfo || regtry(reginfo, s)) ) \
1030 goto got_it; \
1031 } \
1032 s += len
1033
1034#define REXEC_FBC_EXACTISH_SCAN(CoNd) \
1035STMT_START { \
1036 while (s <= e) { \
1037 if ( (CoNd) \
1038 && (ln == 1 || !(OP(c) == EXACTF \
1039 ? ibcmp(s, m, ln) \
1040 : ibcmp_locale(s, m, ln))) \
1041 && (!reginfo || regtry(reginfo, s)) ) \
1042 goto got_it; \
1043 s++; \
1044 } \
1045} STMT_END
1046
1047#define REXEC_FBC_UTF8_SCAN(CoDe) \
1048STMT_START { \
1049 while (s + (uskip = UTF8SKIP(s)) <= strend) { \
1050 CoDe \
1051 s += uskip; \
1052 } \
1053} STMT_END
1054
1055#define REXEC_FBC_SCAN(CoDe) \
1056STMT_START { \
1057 while (s < strend) { \
1058 CoDe \
1059 s++; \
1060 } \
1061} STMT_END
1062
1063#define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \
1064REXEC_FBC_UTF8_SCAN( \
1065 if (CoNd) { \
1066 if (tmp && (!reginfo || regtry(reginfo, s))) \
1067 goto got_it; \
1068 else \
1069 tmp = doevery; \
1070 } \
1071 else \
1072 tmp = 1; \
1073)
1074
1075#define REXEC_FBC_CLASS_SCAN(CoNd) \
1076REXEC_FBC_SCAN( \
1077 if (CoNd) { \
1078 if (tmp && (!reginfo || regtry(reginfo, s))) \
1079 goto got_it; \
1080 else \
1081 tmp = doevery; \
1082 } \
1083 else \
1084 tmp = 1; \
1085)
1086
1087#define REXEC_FBC_TRYIT \
1088if ((!reginfo || regtry(reginfo, s))) \
1089 goto got_it
1090
1091#define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \
1092 if (do_utf8) { \
1093 UtFpReLoAd; \
1094 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1095 } \
1096 else { \
1097 REXEC_FBC_CLASS_SCAN(CoNd); \
1098 } \
1099 break
1100
1101#define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \
1102 PL_reg_flags |= RF_tainted; \
1103 if (do_utf8) { \
1104 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1105 } \
1106 else { \
1107 REXEC_FBC_CLASS_SCAN(CoNd); \
1108 } \
1109 break
1110
786e8c11
YO
1111#define DUMP_EXEC_POS(li,s,doutf8) \
1112 dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8)
1113
1114/* We know what class REx starts with. Try to find this position... */
1115/* if reginfo is NULL, its a dryrun */
1116/* annoyingly all the vars in this routine have different names from their counterparts
1117 in regmatch. /grrr */
1118
3c3eec57 1119STATIC char *
07be1b83
YO
1120S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
1121 const char *strend, const regmatch_info *reginfo)
a687059c 1122{
27da23d5 1123 dVAR;
1df70142 1124 const I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
6eb5f6b9 1125 char *m;
d8093b23 1126 STRLEN ln;
5dab1207 1127 STRLEN lnc;
078c425b 1128 register STRLEN uskip;
d8093b23
G
1129 unsigned int c1;
1130 unsigned int c2;
6eb5f6b9
JH
1131 char *e;
1132 register I32 tmp = 1; /* Scratch variable? */
a3b680e6 1133 register const bool do_utf8 = PL_reg_match_utf8;
cad2e5aa 1134
6eb5f6b9
JH
1135 /* We know what class it must start with. */
1136 switch (OP(c)) {
6eb5f6b9 1137 case ANYOF:
388cc4de 1138 if (do_utf8) {
4cadc6a9 1139 REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
388cc4de 1140 !UTF8_IS_INVARIANT((U8)s[0]) ?
32fc9b6a 1141 reginclass(prog, c, (U8*)s, 0, do_utf8) :
4cadc6a9 1142 REGINCLASS(prog, c, (U8*)s));
388cc4de
HS
1143 }
1144 else {
1145 while (s < strend) {
1146 STRLEN skip = 1;
1147
32fc9b6a 1148 if (REGINCLASS(prog, c, (U8*)s) ||
388cc4de
HS
1149 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
1150 /* The assignment of 2 is intentional:
1151 * for the folded sharp s, the skip is 2. */
1152 (skip = SHARP_S_SKIP))) {
3b0527fe 1153 if (tmp && (!reginfo || regtry(reginfo, s)))
388cc4de
HS
1154 goto got_it;
1155 else
1156 tmp = doevery;
1157 }
1158 else
1159 tmp = 1;
1160 s += skip;
1161 }
a0d0e21e 1162 }
6eb5f6b9 1163 break;
f33976b4 1164 case CANY:
4cadc6a9 1165 REXEC_FBC_SCAN(
3b0527fe 1166 if (tmp && (!reginfo || regtry(reginfo, s)))
f33976b4
DB
1167 goto got_it;
1168 else
1169 tmp = doevery;
4cadc6a9 1170 );
f33976b4 1171 break;
6eb5f6b9 1172 case EXACTF:
5dab1207
NIS
1173 m = STRING(c);
1174 ln = STR_LEN(c); /* length to match in octets/bytes */
1175 lnc = (I32) ln; /* length to match in characters */
1aa99e6b 1176 if (UTF) {
a2a2844f 1177 STRLEN ulen1, ulen2;
5dab1207 1178 U8 *sm = (U8 *) m;
89ebb4a3
JH
1179 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
1180 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
4ad0818d 1181 const U32 uniflags = UTF8_ALLOW_DEFAULT;
a2a2844f
JH
1182
1183 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1184 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1185
89ebb4a3 1186 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
041457d9 1187 0, uniflags);
89ebb4a3 1188 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
041457d9 1189 0, uniflags);
5dab1207
NIS
1190 lnc = 0;
1191 while (sm < ((U8 *) m + ln)) {
1192 lnc++;
1193 sm += UTF8SKIP(sm);
1194 }
1aa99e6b
IH
1195 }
1196 else {
1197 c1 = *(U8*)m;
1198 c2 = PL_fold[c1];
1199 }
6eb5f6b9
JH
1200 goto do_exactf;
1201 case EXACTFL:
5dab1207
NIS
1202 m = STRING(c);
1203 ln = STR_LEN(c);
1204 lnc = (I32) ln;
d8093b23 1205 c1 = *(U8*)m;
6eb5f6b9
JH
1206 c2 = PL_fold_locale[c1];
1207 do_exactf:
db12adc6 1208 e = HOP3c(strend, -((I32)lnc), s);
b3c9acc1 1209
3b0527fe 1210 if (!reginfo && e < s)
6eb5f6b9 1211 e = s; /* Due to minlen logic of intuit() */
1aa99e6b 1212
60a8b682
JH
1213 /* The idea in the EXACTF* cases is to first find the
1214 * first character of the EXACTF* node and then, if
1215 * necessary, case-insensitively compare the full
1216 * text of the node. The c1 and c2 are the first
1217 * characters (though in Unicode it gets a bit
1218 * more complicated because there are more cases
7f16dd3d
JH
1219 * than just upper and lower: one needs to use
1220 * the so-called folding case for case-insensitive
1221 * matching (called "loose matching" in Unicode).
1222 * ibcmp_utf8() will do just that. */
60a8b682 1223
1aa99e6b 1224 if (do_utf8) {
575cac57 1225 UV c, f;
89ebb4a3 1226 U8 tmpbuf [UTF8_MAXBYTES+1];
575cac57 1227 STRLEN len, foldlen;
4ad0818d 1228 const U32 uniflags = UTF8_ALLOW_DEFAULT;
09091399 1229 if (c1 == c2) {
5dab1207
NIS
1230 /* Upper and lower of 1st char are equal -
1231 * probably not a "letter". */
1aa99e6b 1232 while (s <= e) {
89ebb4a3 1233 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
041457d9 1234 uniflags);
4cadc6a9 1235 REXEC_FBC_EXACTISH_CHECK(c == c1);
1aa99e6b 1236 }
09091399
JH
1237 }
1238 else {
1aa99e6b 1239 while (s <= e) {
89ebb4a3 1240 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
041457d9 1241 uniflags);
80aecb99 1242
60a8b682 1243 /* Handle some of the three Greek sigmas cases.
8c01da3c
JH
1244 * Note that not all the possible combinations
1245 * are handled here: some of them are handled
1246 * by the standard folding rules, and some of
1247 * them (the character class or ANYOF cases)
1248 * are handled during compiletime in
1249 * regexec.c:S_regclass(). */
880bd946
JH
1250 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1251 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1252 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
80aecb99 1253
4cadc6a9 1254 REXEC_FBC_EXACTISH_CHECK(c == c1 || c == c2);
1aa99e6b 1255 }
09091399 1256 }
1aa99e6b
IH
1257 }
1258 else {
1259 if (c1 == c2)
4cadc6a9 1260 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1);
1aa99e6b 1261 else
4cadc6a9 1262 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2);
b3c9acc1
IZ
1263 }
1264 break;
bbce6d69 1265 case BOUNDL:
3280af22 1266 PL_reg_flags |= RF_tainted;
bbce6d69 1267 /* FALL THROUGH */
a0d0e21e 1268 case BOUND:
ffc61ed2 1269 if (do_utf8) {
12d33761 1270 if (s == PL_bostr)
ffc61ed2
JH
1271 tmp = '\n';
1272 else {
6136c704 1273 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
4ad0818d 1274 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
ffc61ed2
JH
1275 }
1276 tmp = ((OP(c) == BOUND ?
9041c2e3 1277 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1a4fad37 1278 LOAD_UTF8_CHARCLASS_ALNUM();
4cadc6a9 1279 REXEC_FBC_UTF8_SCAN(
ffc61ed2 1280 if (tmp == !(OP(c) == BOUND ?
bb7a0f54 1281 (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
ffc61ed2
JH
1282 isALNUM_LC_utf8((U8*)s)))
1283 {
1284 tmp = !tmp;
4cadc6a9 1285 REXEC_FBC_TRYIT;
a687059c 1286 }
4cadc6a9 1287 );
a0d0e21e 1288 }
667bb95a 1289 else {
12d33761 1290 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
ffc61ed2 1291 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
4cadc6a9 1292 REXEC_FBC_SCAN(
ffc61ed2
JH
1293 if (tmp ==
1294 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1295 tmp = !tmp;
4cadc6a9 1296 REXEC_FBC_TRYIT;
a0ed51b3 1297 }
4cadc6a9 1298 );
a0ed51b3 1299 }
3b0527fe 1300 if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, s)))
a0ed51b3
LW
1301 goto got_it;
1302 break;
bbce6d69 1303 case NBOUNDL:
3280af22 1304 PL_reg_flags |= RF_tainted;
bbce6d69 1305 /* FALL THROUGH */
a0d0e21e 1306 case NBOUND:
ffc61ed2 1307 if (do_utf8) {
12d33761 1308 if (s == PL_bostr)
ffc61ed2
JH
1309 tmp = '\n';
1310 else {
6136c704 1311 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
4ad0818d 1312 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
ffc61ed2
JH
1313 }
1314 tmp = ((OP(c) == NBOUND ?
9041c2e3 1315 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1a4fad37 1316 LOAD_UTF8_CHARCLASS_ALNUM();
4cadc6a9 1317 REXEC_FBC_UTF8_SCAN(
ffc61ed2 1318 if (tmp == !(OP(c) == NBOUND ?
bb7a0f54 1319 (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
ffc61ed2
JH
1320 isALNUM_LC_utf8((U8*)s)))
1321 tmp = !tmp;
4cadc6a9
YO
1322 else REXEC_FBC_TRYIT;
1323 );
a0d0e21e 1324 }
667bb95a 1325 else {
12d33761 1326 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
ffc61ed2
JH
1327 tmp = ((OP(c) == NBOUND ?
1328 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
4cadc6a9 1329 REXEC_FBC_SCAN(
ffc61ed2
JH
1330 if (tmp ==
1331 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1332 tmp = !tmp;
4cadc6a9
YO
1333 else REXEC_FBC_TRYIT;
1334 );
a0ed51b3 1335 }
3b0527fe 1336 if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, s)))
a0ed51b3
LW
1337 goto got_it;
1338 break;
a0d0e21e 1339 case ALNUM:
4cadc6a9
YO
1340 REXEC_FBC_CSCAN_PRELOAD(
1341 LOAD_UTF8_CHARCLASS_ALNUM(),
1342 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8),
1343 isALNUM(*s)
1344 );
bbce6d69 1345 case ALNUML:
4cadc6a9
YO
1346 REXEC_FBC_CSCAN_TAINT(
1347 isALNUM_LC_utf8((U8*)s),
1348 isALNUM_LC(*s)
1349 );
a0d0e21e 1350 case NALNUM:
4cadc6a9
YO
1351 REXEC_FBC_CSCAN_PRELOAD(
1352 LOAD_UTF8_CHARCLASS_ALNUM(),
1353 !swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8),
1354 !isALNUM(*s)
1355 );
bbce6d69 1356 case NALNUML:
4cadc6a9
YO
1357 REXEC_FBC_CSCAN_TAINT(
1358 !isALNUM_LC_utf8((U8*)s),
1359 !isALNUM_LC(*s)
1360 );
a0d0e21e 1361 case SPACE:
4cadc6a9
YO
1362 REXEC_FBC_CSCAN_PRELOAD(
1363 LOAD_UTF8_CHARCLASS_SPACE(),
1364 *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8),
1365 isSPACE(*s)
1366 );
bbce6d69 1367 case SPACEL:
4cadc6a9
YO
1368 REXEC_FBC_CSCAN_TAINT(
1369 *s == ' ' || isSPACE_LC_utf8((U8*)s),
1370 isSPACE_LC(*s)
1371 );
a0d0e21e 1372 case NSPACE:
4cadc6a9
YO
1373 REXEC_FBC_CSCAN_PRELOAD(
1374 LOAD_UTF8_CHARCLASS_SPACE(),
1375 !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)),
1376 !isSPACE(*s)
1377 );
bbce6d69 1378 case NSPACEL:
4cadc6a9
YO
1379 REXEC_FBC_CSCAN_TAINT(
1380 !(*s == ' ' || isSPACE_LC_utf8((U8*)s)),
1381 !isSPACE_LC(*s)
1382 );
a0d0e21e 1383 case DIGIT:
4cadc6a9
YO
1384 REXEC_FBC_CSCAN_PRELOAD(
1385 LOAD_UTF8_CHARCLASS_DIGIT(),
1386 swash_fetch(PL_utf8_digit,(U8*)s, do_utf8),
1387 isDIGIT(*s)
1388 );
b8c5462f 1389 case DIGITL:
4cadc6a9
YO
1390 REXEC_FBC_CSCAN_TAINT(
1391 isDIGIT_LC_utf8((U8*)s),
1392 isDIGIT_LC(*s)
1393 );
a0d0e21e 1394 case NDIGIT:
4cadc6a9
YO
1395 REXEC_FBC_CSCAN_PRELOAD(
1396 LOAD_UTF8_CHARCLASS_DIGIT(),
1397 !swash_fetch(PL_utf8_digit,(U8*)s, do_utf8),
1398 !isDIGIT(*s)
1399 );
b8c5462f 1400 case NDIGITL:
4cadc6a9
YO
1401 REXEC_FBC_CSCAN_TAINT(
1402 !isDIGIT_LC_utf8((U8*)s),
1403 !isDIGIT_LC(*s)
1404 );
1de06328
YO
1405 case AHOCORASICKC:
1406 case AHOCORASICK:
07be1b83
YO
1407 {
1408 const enum { trie_plain, trie_utf8, trie_utf8_fold }
1409 trie_type = do_utf8 ?
1410 (c->flags == EXACT ? trie_utf8 : trie_utf8_fold)
1411 : trie_plain;
1412 /* what trie are we using right now */
1413 reg_ac_data *aho
1414 = (reg_ac_data*)prog->data->data[ ARG( c ) ];
1415 reg_trie_data *trie=aho->trie;
1416
1417 const char *last_start = strend - trie->minlen;
6148ee25 1418#ifdef DEBUGGING
07be1b83 1419 const char *real_start = s;
6148ee25 1420#endif
07be1b83 1421 STRLEN maxlen = trie->maxlen;
be8e71aa
YO
1422 SV *sv_points;
1423 U8 **points; /* map of where we were in the input string
786e8c11 1424 when reading a given char. For ASCII this
be8e71aa
YO
1425 is unnecessary overhead as the relationship
1426 is always 1:1, but for unicode, especially
1427 case folded unicode this is not true. */
f9e705e8 1428 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
786e8c11
YO
1429 U8 *bitmap=NULL;
1430
07be1b83
YO
1431
1432 GET_RE_DEBUG_FLAGS_DECL;
1433
be8e71aa
YO
1434 /* We can't just allocate points here. We need to wrap it in
1435 * an SV so it gets freed properly if there is a croak while
1436 * running the match */
1437 ENTER;
1438 SAVETMPS;
1439 sv_points=newSV(maxlen * sizeof(U8 *));
1440 SvCUR_set(sv_points,
1441 maxlen * sizeof(U8 *));
1442 SvPOK_on(sv_points);
1443 sv_2mortal(sv_points);
1444 points=(U8**)SvPV_nolen(sv_points );
1de06328
YO
1445 if ( trie_type != trie_utf8_fold
1446 && (trie->bitmap || OP(c)==AHOCORASICKC) )
1447 {
786e8c11
YO
1448 if (trie->bitmap)
1449 bitmap=(U8*)trie->bitmap;
1450 else
1451 bitmap=(U8*)ANYOF_BITMAP(c);
07be1b83 1452 }
786e8c11
YO
1453 /* this is the Aho-Corasick algorithm modified a touch
1454 to include special handling for long "unknown char"
1455 sequences. The basic idea being that we use AC as long
1456 as we are dealing with a possible matching char, when
1457 we encounter an unknown char (and we have not encountered
1458 an accepting state) we scan forward until we find a legal
1459 starting char.
1460 AC matching is basically that of trie matching, except
1461 that when we encounter a failing transition, we fall back
1462 to the current states "fail state", and try the current char
1463 again, a process we repeat until we reach the root state,
1464 state 1, or a legal transition. If we fail on the root state
1465 then we can either terminate if we have reached an accepting
1466 state previously, or restart the entire process from the beginning
1467 if we have not.
1468
1469 */
07be1b83
YO
1470 while (s <= last_start) {
1471 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1472 U8 *uc = (U8*)s;
1473 U16 charid = 0;
1474 U32 base = 1;
1475 U32 state = 1;
1476 UV uvc = 0;
1477 STRLEN len = 0;
1478 STRLEN foldlen = 0;
1479 U8 *uscan = (U8*)NULL;
1480 U8 *leftmost = NULL;
786e8c11
YO
1481#ifdef DEBUGGING
1482 U32 accepted_word= 0;
1483#endif
07be1b83
YO
1484 U32 pointpos = 0;
1485
1486 while ( state && uc <= (U8*)strend ) {
1487 int failed=0;
786e8c11
YO
1488 U32 word = aho->states[ state ].wordnum;
1489
1de06328
YO
1490 if( state==1 ) {
1491 if ( bitmap ) {
1492 DEBUG_TRIE_EXECUTE_r(
1493 if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1494 dump_exec_pos( (char *)uc, c, strend, real_start,
1495 (char *)uc, do_utf8 );
1496 PerlIO_printf( Perl_debug_log,
1497 " Scanning for legal start char...\n");
1498 }
1499 );
1500 while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1501 uc++;
786e8c11 1502 }
1de06328 1503 s= (char *)uc;
786e8c11 1504 }
786e8c11
YO
1505 if (uc >(U8*)last_start) break;
1506 }
1507
1508 if ( word ) {
1509 U8 *lpos= points[ (pointpos - trie->wordlen[word-1] ) % maxlen ];
1510 if (!leftmost || lpos < leftmost) {
1511 DEBUG_r(accepted_word=word);
07be1b83 1512 leftmost= lpos;
786e8c11 1513 }
07be1b83 1514 if (base==0) break;
786e8c11 1515
07be1b83
YO
1516 }
1517 points[pointpos++ % maxlen]= uc;
4cadc6a9
YO
1518 REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len,
1519 uvc, charid, foldlen, foldbuf, uniflags);
786e8c11
YO
1520 DEBUG_TRIE_EXECUTE_r({
1521 dump_exec_pos( (char *)uc, c, strend, real_start,
1522 s, do_utf8 );
07be1b83 1523 PerlIO_printf(Perl_debug_log,
786e8c11
YO
1524 " Charid:%3u CP:%4"UVxf" ",
1525 charid, uvc);
1526 });
07be1b83
YO
1527
1528 do {
6148ee25 1529#ifdef DEBUGGING
786e8c11 1530 word = aho->states[ state ].wordnum;
6148ee25 1531#endif
07be1b83
YO
1532 base = aho->states[ state ].trans.base;
1533
786e8c11
YO
1534 DEBUG_TRIE_EXECUTE_r({
1535 if (failed)
1536 dump_exec_pos( (char *)uc, c, strend, real_start,
1537 s, do_utf8 );
07be1b83 1538 PerlIO_printf( Perl_debug_log,
786e8c11
YO
1539 "%sState: %4"UVxf", word=%"UVxf,
1540 failed ? " Fail transition to " : "",
1541 (UV)state, (UV)word);
1542 });
07be1b83
YO
1543 if ( base ) {
1544 U32 tmp;
1545 if (charid &&
1546 (base + charid > trie->uniquecharcount )
1547 && (base + charid - 1 - trie->uniquecharcount
1548 < trie->lasttrans)
1549 && trie->trans[base + charid - 1 -
1550 trie->uniquecharcount].check == state
1551 && (tmp=trie->trans[base + charid - 1 -
1552 trie->uniquecharcount ].next))
1553 {
786e8c11
YO
1554 DEBUG_TRIE_EXECUTE_r(
1555 PerlIO_printf( Perl_debug_log," - legal\n"));
07be1b83
YO
1556 state = tmp;
1557 break;
1558 }
1559 else {
786e8c11
YO
1560 DEBUG_TRIE_EXECUTE_r(
1561 PerlIO_printf( Perl_debug_log," - fail\n"));
1562 failed = 1;
1563 state = aho->fail[state];
07be1b83
YO
1564 }
1565 }
1566 else {
1567 /* we must be accepting here */
786e8c11
YO
1568 DEBUG_TRIE_EXECUTE_r(
1569 PerlIO_printf( Perl_debug_log," - accepting\n"));
1570 failed = 1;
07be1b83
YO
1571 break;
1572 }
1573 } while(state);
786e8c11 1574 uc += len;
07be1b83
YO
1575 if (failed) {
1576 if (leftmost)
1577 break;
786e8c11 1578 if (!state) state = 1;
07be1b83
YO
1579 }
1580 }
1581 if ( aho->states[ state ].wordnum ) {
1582 U8 *lpos = points[ (pointpos - trie->wordlen[aho->states[ state ].wordnum-1]) % maxlen ];
786e8c11
YO
1583 if (!leftmost || lpos < leftmost) {
1584 DEBUG_r(accepted_word=aho->states[ state ].wordnum);
07be1b83 1585 leftmost = lpos;
786e8c11 1586 }
07be1b83 1587 }
07be1b83
YO
1588 if (leftmost) {
1589 s = (char*)leftmost;
786e8c11
YO
1590 DEBUG_TRIE_EXECUTE_r({
1591 PerlIO_printf(
1592 Perl_debug_log,"Matches word #%"UVxf" at position %d. Trying full pattern...\n",
1593 (UV)accepted_word, s - real_start
1594 );
1595 });
be8e71aa
YO
1596 if (!reginfo || regtry(reginfo, s)) {
1597 FREETMPS;
1598 LEAVE;
07be1b83 1599 goto got_it;
be8e71aa 1600 }
07be1b83 1601 s = HOPc(s,1);
786e8c11
YO
1602 DEBUG_TRIE_EXECUTE_r({
1603 PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n");
1604 });
07be1b83 1605 } else {
786e8c11
YO
1606 DEBUG_TRIE_EXECUTE_r(
1607 PerlIO_printf( Perl_debug_log,"No match.\n"));
07be1b83
YO
1608 break;
1609 }
1610 }
be8e71aa
YO
1611 FREETMPS;
1612 LEAVE;
07be1b83
YO
1613 }
1614 break;
b3c9acc1 1615 default:
3c3eec57
GS
1616 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1617 break;
d6a28714 1618 }
6eb5f6b9
JH
1619 return 0;
1620 got_it:
1621 return s;
1622}
1623
1624/*
1625 - regexec_flags - match a regexp against a string
1626 */
1627I32
1628Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1629 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1630/* strend: pointer to null at end of string */
1631/* strbeg: real beginning of string */
1632/* minend: end of match must be >=minend after stringarg. */
1633/* data: May be used for some additional optimizations. */
1634/* nosave: For optimizations. */
1635{
97aff369 1636 dVAR;
6eb5f6b9
JH
1637 register char *s;
1638 register regnode *c;
1639 register char *startpos = stringarg;
6eb5f6b9
JH
1640 I32 minlen; /* must match at least this many chars */
1641 I32 dontbother = 0; /* how many characters not to try at end */
6eb5f6b9
JH
1642 I32 end_shift = 0; /* Same for the end. */ /* CC */
1643 I32 scream_pos = -1; /* Internal iterator of scream. */
ccac19ea 1644 char *scream_olds = NULL;
3dab1dad 1645 SV* const oreplsv = GvSV(PL_replgv);
f9f4320a 1646 const bool do_utf8 = (bool)DO_UTF8(sv);
2757e526 1647 I32 multiline;
0df25f3d 1648
3b0527fe 1649 regmatch_info reginfo; /* create some info to pass to regtry etc */
a3621e74
YO
1650
1651 GET_RE_DEBUG_FLAGS_DECL;
1652
9d4ba2ae 1653 PERL_UNUSED_ARG(data);
6eb5f6b9
JH
1654
1655 /* Be paranoid... */
1656 if (prog == NULL || startpos == NULL) {
1657 Perl_croak(aTHX_ "NULL regexp parameter");
1658 return 0;
1659 }
1660
2757e526 1661 multiline = prog->reganch & PMf_MULTILINE;
3b0527fe 1662 reginfo.prog = prog;
2757e526 1663
bac06658 1664 RX_MATCH_UTF8_set(prog, do_utf8);
1de06328
YO
1665 DEBUG_EXECUTE_r(
1666 debug_start_match(prog, do_utf8, startpos, strend,
1667 "Matching");
1668 );
bac06658 1669
6eb5f6b9 1670 minlen = prog->minlen;
1de06328
YO
1671
1672 if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) {
a3621e74 1673 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
a72c7584
JH
1674 "String too short [regexec_flags]...\n"));
1675 goto phooey;
1aa99e6b 1676 }
6eb5f6b9 1677
1de06328 1678
6eb5f6b9
JH
1679 /* Check validity of program. */
1680 if (UCHARAT(prog->program) != REG_MAGIC) {
1681 Perl_croak(aTHX_ "corrupted regexp program");
1682 }
1683
1684 PL_reg_flags = 0;
1685 PL_reg_eval_set = 0;
1686 PL_reg_maxiter = 0;
1687
1688 if (prog->reganch & ROPT_UTF8)
1689 PL_reg_flags |= RF_utf8;
1690
1691 /* Mark beginning of line for ^ and lookbehind. */
3b0527fe 1692 reginfo.bol = startpos; /* XXX not used ??? */
6eb5f6b9 1693 PL_bostr = strbeg;
3b0527fe 1694 reginfo.sv = sv;
6eb5f6b9
JH
1695
1696 /* Mark end of line for $ (and such) */
1697 PL_regeol = strend;
1698
1699 /* see how far we have to get to not match where we matched before */
3b0527fe 1700 reginfo.till = startpos+minend;
6eb5f6b9 1701
6eb5f6b9
JH
1702 /* If there is a "must appear" string, look for it. */
1703 s = startpos;
1704
3b0527fe 1705 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to set reginfo->ganch */
6eb5f6b9
JH
1706 MAGIC *mg;
1707
1708 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
3b0527fe 1709 reginfo.ganch = startpos;
6eb5f6b9
JH
1710 else if (sv && SvTYPE(sv) >= SVt_PVMG
1711 && SvMAGIC(sv)
14befaf4
DM
1712 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1713 && mg->mg_len >= 0) {
3b0527fe 1714 reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */
6eb5f6b9 1715 if (prog->reganch & ROPT_ANCH_GPOS) {
3b0527fe 1716 if (s > reginfo.ganch)
6eb5f6b9 1717 goto phooey;
3b0527fe 1718 s = reginfo.ganch;
6eb5f6b9
JH
1719 }
1720 }
1721 else /* pos() not defined */
3b0527fe 1722 reginfo.ganch = strbeg;
6eb5f6b9
JH
1723 }
1724
a0714e2c 1725 if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
6eb5f6b9
JH
1726 re_scream_pos_data d;
1727
1728 d.scream_olds = &scream_olds;
1729 d.scream_pos = &scream_pos;
1730 s = re_intuit_start(prog, sv, s, strend, flags, &d);
3fa9c3d7 1731 if (!s) {
a3621e74 1732 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
6eb5f6b9 1733 goto phooey; /* not present */
3fa9c3d7 1734 }
6eb5f6b9
JH
1735 }
1736
1de06328 1737
6eb5f6b9
JH
1738
1739 /* Simplest case: anchored match need be tried only once. */
1740 /* [unless only anchor is BOL and multiline is set] */
1741 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
3b0527fe 1742 if (s == startpos && regtry(&reginfo, startpos))
6eb5f6b9 1743 goto got_it;
7fba1cd6 1744 else if (multiline || (prog->reganch & ROPT_IMPLICIT)
6eb5f6b9
JH
1745 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1746 {
1747 char *end;
1748
1749 if (minlen)
1750 dontbother = minlen - 1;
1aa99e6b 1751 end = HOP3c(strend, -dontbother, strbeg) - 1;
6eb5f6b9 1752 /* for multiline we only have to try after newlines */
33b8afdf 1753 if (prog->check_substr || prog->check_utf8) {
6eb5f6b9
JH
1754 if (s == startpos)
1755 goto after_try;
1756 while (1) {
3b0527fe 1757 if (regtry(&reginfo, s))
6eb5f6b9
JH
1758 goto got_it;
1759 after_try:
1760 if (s >= end)
1761 goto phooey;
1762 if (prog->reganch & RE_USE_INTUIT) {
1763 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1764 if (!s)
1765 goto phooey;
1766 }
1767 else
1768 s++;
1769 }
1770 } else {
1771 if (s > startpos)
1772 s--;
1773 while (s < end) {
1774 if (*s++ == '\n') { /* don't need PL_utf8skip here */
3b0527fe 1775 if (regtry(&reginfo, s))
6eb5f6b9
JH
1776 goto got_it;
1777 }
1778 }
1779 }
1780 }
1781 goto phooey;
f9f4320a
YO
1782 } else if (ROPT_GPOS_CHECK == (prog->reganch & ROPT_GPOS_CHECK))
1783 {
1784 /* the warning about reginfo.ganch being used without intialization
1785 is bogus -- we set it above, when prog->reganch & ROPT_GPOS_SEEN
1786 and we only enter this block when the same bit is set. */
3b0527fe 1787 if (regtry(&reginfo, reginfo.ganch))
6eb5f6b9
JH
1788 goto got_it;
1789 goto phooey;
1790 }
1791
1792 /* Messy cases: unanchored match. */
33b8afdf 1793 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
6eb5f6b9
JH
1794 /* we have /x+whatever/ */
1795 /* it must be a one character string (XXXX Except UTF?) */
33b8afdf 1796 char ch;
bf93d4cc
GS
1797#ifdef DEBUGGING
1798 int did_match = 0;
1799#endif
33b8afdf
JH
1800 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1801 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
3f7c398e 1802 ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
bf93d4cc 1803
1aa99e6b 1804 if (do_utf8) {
4cadc6a9 1805 REXEC_FBC_SCAN(
6eb5f6b9 1806 if (*s == ch) {
a3621e74 1807 DEBUG_EXECUTE_r( did_match = 1 );
3b0527fe 1808 if (regtry(&reginfo, s)) goto got_it;
6eb5f6b9
JH
1809 s += UTF8SKIP(s);
1810 while (s < strend && *s == ch)
1811 s += UTF8SKIP(s);
1812 }
4cadc6a9 1813 );
6eb5f6b9
JH
1814 }
1815 else {
4cadc6a9 1816 REXEC_FBC_SCAN(
6eb5f6b9 1817 if (*s == ch) {
a3621e74 1818 DEBUG_EXECUTE_r( did_match = 1 );
3b0527fe 1819 if (regtry(&reginfo, s)) goto got_it;
6eb5f6b9
JH
1820 s++;
1821 while (s < strend && *s == ch)
1822 s++;
1823 }
4cadc6a9 1824 );
6eb5f6b9 1825 }
a3621e74 1826 DEBUG_EXECUTE_r(if (!did_match)
bf93d4cc 1827 PerlIO_printf(Perl_debug_log,
b7953727
JH
1828 "Did not find anchored character...\n")
1829 );
6eb5f6b9 1830 }
a0714e2c
SS
1831 else if (prog->anchored_substr != NULL
1832 || prog->anchored_utf8 != NULL
1833 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
33b8afdf
JH
1834 && prog->float_max_offset < strend - s)) {
1835 SV *must;
1836 I32 back_max;
1837 I32 back_min;
1838 char *last;
6eb5f6b9 1839 char *last1; /* Last position checked before */
bf93d4cc
GS
1840#ifdef DEBUGGING
1841 int did_match = 0;
1842#endif
33b8afdf
JH
1843 if (prog->anchored_substr || prog->anchored_utf8) {
1844 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1845 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1846 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1847 back_max = back_min = prog->anchored_offset;
1848 } else {
1849 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1850 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1851 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1852 back_max = prog->float_max_offset;
1853 back_min = prog->float_min_offset;
1854 }
1de06328
YO
1855
1856
33b8afdf
JH
1857 if (must == &PL_sv_undef)
1858 /* could not downgrade utf8 check substring, so must fail */
1859 goto phooey;
1860
1de06328
YO
1861 if (back_min<0) {
1862 last = strend;
1863 } else {
1864 last = HOP3c(strend, /* Cannot start after this */
1865 -(I32)(CHR_SVLEN(must)
1866 - (SvTAIL(must) != 0) + back_min), strbeg);
1867 }
6eb5f6b9
JH
1868 if (s > PL_bostr)
1869 last1 = HOPc(s, -1);
1870 else
1871 last1 = s - 1; /* bogus */
1872
a0288114 1873 /* XXXX check_substr already used to find "s", can optimize if
6eb5f6b9
JH
1874 check_substr==must. */
1875 scream_pos = -1;
1876 dontbother = end_shift;
1877 strend = HOPc(strend, -dontbother);
1878 while ( (s <= last) &&
9041c2e3 1879 ((flags & REXEC_SCREAM)
1de06328 1880 ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg,
6eb5f6b9 1881 end_shift, &scream_pos, 0))
1de06328 1882 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)),
9041c2e3 1883 (unsigned char*)strend, must,
7fba1cd6 1884 multiline ? FBMrf_MULTILINE : 0))) ) {
4addbd3b
HS
1885 /* we may be pointing at the wrong string */
1886 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
3f7c398e 1887 s = strbeg + (s - SvPVX_const(sv));
a3621e74 1888 DEBUG_EXECUTE_r( did_match = 1 );
6eb5f6b9
JH
1889 if (HOPc(s, -back_max) > last1) {
1890 last1 = HOPc(s, -back_min);
1891 s = HOPc(s, -back_max);
1892 }
1893 else {
52657f30 1894 char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
6eb5f6b9
JH
1895
1896 last1 = HOPc(s, -back_min);
52657f30 1897 s = t;
6eb5f6b9 1898 }
1aa99e6b 1899 if (do_utf8) {
6eb5f6b9 1900 while (s <= last1) {
3b0527fe 1901 if (regtry(&reginfo, s))
6eb5f6b9
JH
1902 goto got_it;
1903 s += UTF8SKIP(s);
1904 }
1905 }
1906 else {
1907 while (s <= last1) {
3b0527fe 1908 if (regtry(&reginfo, s))
6eb5f6b9
JH
1909 goto got_it;
1910 s++;
1911 }
1912 }
1913 }
ab3bbdeb
YO
1914 DEBUG_EXECUTE_r(if (!did_match) {
1915 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
1916 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
1917 PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n",
33b8afdf 1918 ((must == prog->anchored_substr || must == prog->anchored_utf8)
bf93d4cc 1919 ? "anchored" : "floating"),
ab3bbdeb
YO
1920 quoted, RE_SV_TAIL(must));
1921 });
6eb5f6b9
JH
1922 goto phooey;
1923 }
786e8c11 1924 else if ( (c = prog->regstclass) ) {
f14c76ed 1925 if (minlen) {
be8e71aa 1926 const OPCODE op = OP(prog->regstclass);
66e933ab 1927 /* don't bother with what can't match */
786e8c11 1928 if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE)
f14c76ed
RGS
1929 strend = HOPc(strend, -(minlen - 1));
1930 }
a3621e74 1931 DEBUG_EXECUTE_r({
be8e71aa 1932 SV * const prop = sv_newmortal();
32fc9b6a 1933 regprop(prog, prop, c);
0df25f3d 1934 {
ab3bbdeb
YO
1935 RE_PV_QUOTED_DECL(quoted,UTF,PERL_DEBUG_PAD_ZERO(1),
1936 s,strend-s,60);
0df25f3d 1937 PerlIO_printf(Perl_debug_log,
ab3bbdeb 1938 "Matching stclass %.*s against %s (%d chars)\n",
e4f74956 1939 (int)SvCUR(prop), SvPVX_const(prop),
ab3bbdeb 1940 quoted, (int)(strend - s));
0df25f3d 1941 }
ffc61ed2 1942 });
3b0527fe 1943 if (find_byclass(prog, c, s, strend, &reginfo))
6eb5f6b9 1944 goto got_it;
07be1b83 1945 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n"));
d6a28714
JH
1946 }
1947 else {
1948 dontbother = 0;
a0714e2c 1949 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
33b8afdf 1950 /* Trim the end. */
d6a28714 1951 char *last;
33b8afdf
JH
1952 SV* float_real;
1953
1954 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1955 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1956 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
d6a28714
JH
1957
1958 if (flags & REXEC_SCREAM) {
33b8afdf 1959 last = screaminstr(sv, float_real, s - strbeg,
d6a28714
JH
1960 end_shift, &scream_pos, 1); /* last one */
1961 if (!last)
ffc61ed2 1962 last = scream_olds; /* Only one occurrence. */
4addbd3b
HS
1963 /* we may be pointing at the wrong string */
1964 else if (RX_MATCH_COPIED(prog))
3f7c398e 1965 s = strbeg + (s - SvPVX_const(sv));
b8c5462f 1966 }
d6a28714
JH
1967 else {
1968 STRLEN len;
cfd0369c 1969 const char * const little = SvPV_const(float_real, len);
d6a28714 1970
33b8afdf 1971 if (SvTAIL(float_real)) {
d6a28714
JH
1972 if (memEQ(strend - len + 1, little, len - 1))
1973 last = strend - len + 1;
7fba1cd6 1974 else if (!multiline)
9041c2e3 1975 last = memEQ(strend - len, little, len)
bd61b366 1976 ? strend - len : NULL;
b8c5462f 1977 else
d6a28714
JH
1978 goto find_last;
1979 } else {
1980 find_last:
9041c2e3 1981 if (len)
d6a28714 1982 last = rninstr(s, strend, little, little + len);
b8c5462f 1983 else
a0288114 1984 last = strend; /* matching "$" */
b8c5462f 1985 }
b8c5462f 1986 }
bf93d4cc 1987 if (last == NULL) {
6bda09f9
YO
1988 DEBUG_EXECUTE_r(
1989 PerlIO_printf(Perl_debug_log,
1990 "%sCan't trim the tail, match fails (should not happen)%s\n",
1991 PL_colors[4], PL_colors[5]));
bf93d4cc
GS
1992 goto phooey; /* Should not happen! */
1993 }
d6a28714
JH
1994 dontbother = strend - last + prog->float_min_offset;
1995 }
1996 if (minlen && (dontbother < minlen))
1997 dontbother = minlen - 1;
1998 strend -= dontbother; /* this one's always in bytes! */
1999 /* We don't know much -- general case. */
1aa99e6b 2000 if (do_utf8) {
d6a28714 2001 for (;;) {
3b0527fe 2002 if (regtry(&reginfo, s))
d6a28714
JH
2003 goto got_it;
2004 if (s >= strend)
2005 break;
b8c5462f 2006 s += UTF8SKIP(s);
d6a28714
JH
2007 };
2008 }
2009 else {
2010 do {
3b0527fe 2011 if (regtry(&reginfo, s))
d6a28714
JH
2012 goto got_it;
2013 } while (s++ < strend);
2014 }
2015 }
2016
2017 /* Failure. */
2018 goto phooey;
2019
2020got_it:
2021 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2022
2023 if (PL_reg_eval_set) {
2024 /* Preserve the current value of $^R */
2025 if (oreplsv != GvSV(PL_replgv))
2026 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2027 restored, the value remains
2028 the same. */
4f639d21 2029 restore_pos(aTHX_ prog);
d6a28714
JH
2030 }
2031
2032 /* make sure $`, $&, $', and $digit will work later */
2033 if ( !(flags & REXEC_NOT_FIRST) ) {
ed252734 2034 RX_MATCH_COPY_FREE(prog);
d6a28714 2035 if (flags & REXEC_COPY_STR) {
be8e71aa 2036 const I32 i = PL_regeol - startpos + (stringarg - strbeg);
f8c7b90f 2037#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
2038 if ((SvIsCOW(sv)
2039 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2040 if (DEBUG_C_TEST) {
2041 PerlIO_printf(Perl_debug_log,
2042 "Copy on write: regexp capture, type %d\n",
2043 (int) SvTYPE(sv));
2044 }
2045 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
d5263905 2046 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
ed252734
NC
2047 assert (SvPOKp(prog->saved_copy));
2048 } else
2049#endif
2050 {
2051 RX_MATCH_COPIED_on(prog);
2052 s = savepvn(strbeg, i);
2053 prog->subbeg = s;
2054 }
d6a28714 2055 prog->sublen = i;
d6a28714
JH
2056 }
2057 else {
2058 prog->subbeg = strbeg;
2059 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2060 }
2061 }
9041c2e3 2062
d6a28714
JH
2063 return 1;
2064
2065phooey:
a3621e74 2066 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
e4584336 2067 PL_colors[4], PL_colors[5]));
d6a28714 2068 if (PL_reg_eval_set)
4f639d21 2069 restore_pos(aTHX_ prog);
d6a28714
JH
2070 return 0;
2071}
2072
6bda09f9 2073
d6a28714
JH
2074/*
2075 - regtry - try match at specific point
2076 */
2077STATIC I32 /* 0 failure, 1 success */
3b0527fe 2078S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos)
d6a28714 2079{
97aff369 2080 dVAR;
d6a28714
JH
2081 register I32 *sp;
2082 register I32 *ep;
2083 CHECKPOINT lastcp;
3b0527fe 2084 regexp *prog = reginfo->prog;
a3621e74 2085 GET_RE_DEBUG_FLAGS_DECL;
d6a28714
JH
2086
2087 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2088 MAGIC *mg;
2089
2090 PL_reg_eval_set = RS_init;
a3621e74 2091 DEBUG_EXECUTE_r(DEBUG_s(
b900a521
JH
2092 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2093 (IV)(PL_stack_sp - PL_stack_base));
d6a28714 2094 ));
e8347627 2095 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
d6a28714
JH
2096 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2097 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2098 SAVETMPS;
2099 /* Apparently this is not needed, judging by wantarray. */
e8347627 2100 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
d6a28714
JH
2101 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2102
3b0527fe 2103 if (reginfo->sv) {
d6a28714 2104 /* Make $_ available to executed code. */
3b0527fe 2105 if (reginfo->sv != DEFSV) {
59f00321 2106 SAVE_DEFSV;
3b0527fe 2107 DEFSV = reginfo->sv;
b8c5462f 2108 }
d6a28714 2109
3b0527fe
DM
2110 if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv)
2111 && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) {
d6a28714 2112 /* prepare for quick setting of pos */
d300d9fa
NC
2113#ifdef PERL_OLD_COPY_ON_WRITE
2114 if (SvIsCOW(sv))
2115 sv_force_normal_flags(sv, 0);
2116#endif
3dab1dad 2117 mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global,
d300d9fa 2118 &PL_vtbl_mglob, NULL, 0);
d6a28714 2119 mg->mg_len = -1;
b8c5462f 2120 }
d6a28714
JH
2121 PL_reg_magic = mg;
2122 PL_reg_oldpos = mg->mg_len;
4f639d21 2123 SAVEDESTRUCTOR_X(restore_pos, prog);
d6a28714 2124 }
09687e5a 2125 if (!PL_reg_curpm) {
a02a5408 2126 Newxz(PL_reg_curpm, 1, PMOP);
09687e5a
AB
2127#ifdef USE_ITHREADS
2128 {
be8e71aa 2129 SV* const repointer = newSViv(0);
577e12cc 2130 /* so we know which PL_regex_padav element is PL_reg_curpm */
35061a7e 2131 SvFLAGS(repointer) |= SVf_BREAK;
09687e5a
AB
2132 av_push(PL_regex_padav,repointer);
2133 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2134 PL_regex_pad = AvARRAY(PL_regex_padav);
2135 }
2136#endif
2137 }
aaa362c4 2138 PM_SETRE(PL_reg_curpm, prog);
d6a28714
JH
2139 PL_reg_oldcurpm = PL_curpm;
2140 PL_curpm = PL_reg_curpm;
2141 if (RX_MATCH_COPIED(prog)) {
2142 /* Here is a serious problem: we cannot rewrite subbeg,
2143 since it may be needed if this match fails. Thus
2144 $` inside (?{}) could fail... */
2145 PL_reg_oldsaved = prog->subbeg;
2146 PL_reg_oldsavedlen = prog->sublen;
f8c7b90f 2147#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
2148 PL_nrs = prog->saved_copy;
2149#endif
d6a28714
JH
2150 RX_MATCH_COPIED_off(prog);
2151 }
2152 else
bd61b366 2153 PL_reg_oldsaved = NULL;
d6a28714
JH
2154 prog->subbeg = PL_bostr;
2155 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2156 }
6bda09f9 2157 DEBUG_EXECUTE_r(PL_reg_starttry = startpos);
973dddac 2158 prog->startp[0] = startpos - PL_bostr;
d6a28714 2159 PL_reginput = startpos;
d6a28714 2160 PL_reglastparen = &prog->lastparen;
a01268b5 2161 PL_reglastcloseparen = &prog->lastcloseparen;
d6a28714 2162 prog->lastparen = 0;
03994de8 2163 prog->lastcloseparen = 0;
d6a28714 2164 PL_regsize = 0;
6bda09f9
YO
2165 PL_regstartp = prog->startp;
2166 PL_regendp = prog->endp;
d6a28714
JH
2167 if (PL_reg_start_tmpl <= prog->nparens) {
2168 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2169 if(PL_reg_start_tmp)
2170 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2171 else
a02a5408 2172 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
d6a28714
JH
2173 }
2174
2175 /* XXXX What this code is doing here?!!! There should be no need
2176 to do this again and again, PL_reglastparen should take care of
3dd2943c 2177 this! --ilya*/
dafc8851
JH
2178
2179 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2180 * Actually, the code in regcppop() (which Ilya may be meaning by
daf18116
JH
2181 * PL_reglastparen), is not needed at all by the test suite
2182 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2183 * enough, for building DynaLoader, or otherwise this
2184 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2185 * will happen. Meanwhile, this code *is* needed for the
2186 * above-mentioned test suite tests to succeed. The common theme
2187 * on those tests seems to be returning null fields from matches.
2188 * --jhi */
dafc8851 2189#if 1
d6a28714
JH
2190 sp = prog->startp;
2191 ep = prog->endp;
2192 if (prog->nparens) {
097eb12c 2193 register I32 i;
eb160463 2194 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
d6a28714
JH
2195 *++sp = -1;
2196 *++ep = -1;
2197 }
2198 }
dafc8851 2199#endif
02db2b7b 2200 REGCP_SET(lastcp);
3b0527fe 2201 if (regmatch(reginfo, prog->program + 1)) {
d6a28714
JH
2202 prog->endp[0] = PL_reginput - PL_bostr;
2203 return 1;
2204 }
02db2b7b 2205 REGCP_UNWIND(lastcp);
d6a28714
JH
2206 return 0;
2207}
2208
02db2b7b 2209
8ba1375e
MJD
2210#define sayYES goto yes
2211#define sayNO goto no
262b90c4 2212#define sayNO_SILENT goto no_silent
8ba1375e 2213
f9f4320a
YO
2214/* we dont use STMT_START/END here because it leads to
2215 "unreachable code" warnings, which are bogus, but distracting. */
2216#define CACHEsayNO \
c476f425
DM
2217 if (ST.cache_mask) \
2218 PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \
f9f4320a 2219 sayNO
3298f257 2220
a3621e74 2221/* this is used to determine how far from the left messages like
265c4333
YO
2222 'failed...' are printed. It should be set such that messages
2223 are inline with the regop output that created them.
a3621e74 2224*/
265c4333 2225#define REPORT_CODE_OFF 32
a3621e74
YO
2226
2227
2228/* Make sure there is a test for this +1 options in re_tests */
2229#define TRIE_INITAL_ACCEPT_BUFFLEN 4;
2230
40a82448
DM
2231#define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */
2232#define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */
9e137952 2233
86545054
DM
2234#define SLAB_FIRST(s) (&(s)->states[0])
2235#define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1])
2236
5d9a96ca
DM
2237/* grab a new slab and return the first slot in it */
2238
2239STATIC regmatch_state *
2240S_push_slab(pTHX)
2241{
54df2634
NC
2242#if PERL_VERSION < 9
2243 dMY_CXT;
2244#endif
5d9a96ca
DM
2245 regmatch_slab *s = PL_regmatch_slab->next;
2246 if (!s) {
2247 Newx(s, 1, regmatch_slab);
2248 s->prev = PL_regmatch_slab;
2249 s->next = NULL;
2250 PL_regmatch_slab->next = s;
2251 }
2252 PL_regmatch_slab = s;
86545054 2253 return SLAB_FIRST(s);
5d9a96ca 2254}
5b47454d 2255
95b24440 2256
40a82448
DM
2257/* push a new state then goto it */
2258
2259#define PUSH_STATE_GOTO(state, node) \
2260 scan = node; \
2261 st->resume_state = state; \
2262 goto push_state;
2263
2264/* push a new state with success backtracking, then goto it */
2265
2266#define PUSH_YES_STATE_GOTO(state, node) \
2267 scan = node; \
2268 st->resume_state = state; \
2269 goto push_yes_state;
2270
aa283a38 2271
aa283a38 2272
d6a28714
JH
2273/*
2274 - regmatch - main matching routine
2275 *
2276 * Conceptually the strategy is simple: check to see whether the current
2277 * node matches, call self recursively to see whether the rest matches,
2278 * and then act accordingly. In practice we make some effort to avoid
2279 * recursion, in particular by going through "ordinary" nodes (that don't
2280 * need to know whether the rest of the match failed) by a loop instead of
2281 * by recursion.
2282 */
2283/* [lwall] I've hoisted the register declarations to the outer block in order to
2284 * maybe save a little bit of pushing and popping on the stack. It also takes
2285 * advantage of machines that use a register save mask on subroutine entry.
95b24440
DM
2286 *
2287 * This function used to be heavily recursive, but since this had the
2288 * effect of blowing the CPU stack on complex regexes, it has been
2289 * restructured to be iterative, and to save state onto the heap rather
2290 * than the stack. Essentially whereever regmatch() used to be called, it
2291 * pushes the current state, notes where to return, then jumps back into
2292 * the main loop.
2293 *
2294 * Originally the structure of this function used to look something like
2295
2296 S_regmatch() {
2297 int a = 1, b = 2;
2298 ...
2299 while (scan != NULL) {
5d9a96ca 2300 a++; // do stuff with a and b
95b24440
DM
2301 ...
2302 switch (OP(scan)) {
2303 case FOO: {
2304 int local = 3;
2305 ...
2306 if (regmatch(...)) // recurse
2307 goto yes;
2308 }
2309 ...
2310 }
2311 }
2312 yes:
2313 return 1;
2314 }
2315
2316 * Now it looks something like this:
2317
5d9a96ca 2318 typedef struct {
95b24440
DM
2319 int a, b, local;
2320 int resume_state;
5d9a96ca 2321 } regmatch_state;
95b24440
DM
2322
2323 S_regmatch() {
5d9a96ca
DM
2324 regmatch_state *st = new();
2325 int depth=0;
2326 st->a++; // do stuff with a and b
95b24440
DM
2327 ...
2328 while (scan != NULL) {
2329 ...
2330 switch (OP(scan)) {
2331 case FOO: {
5d9a96ca 2332 st->local = 3;
95b24440 2333 ...
5d9a96ca
DM
2334 st->scan = scan;
2335 scan = ...;
2336 st->resume_state = resume_FOO;
2337 goto start_recurse; // recurse
95b24440 2338
5d9a96ca
DM
2339 resume_point_FOO:
2340 if (result)
95b24440
DM
2341 goto yes;
2342 }
2343 ...
2344 }
5d9a96ca
DM
2345 start_recurse:
2346 st = new(); push a new state
2347 st->a = 1; st->b = 2;
2348 depth++;
95b24440 2349 }
5d9a96ca 2350 yes:
95b24440 2351 result = 1;
5d9a96ca
DM
2352 if (depth--) {
2353 st = pop();
95b24440
DM
2354 switch (resume_state) {
2355 case resume_FOO:
2356 goto resume_point_FOO;
2357 ...
2358 }
2359 }
2360 return result
2361 }
2362
2363 * WARNING: this means that any line in this function that contains a
2364 * REGMATCH() or TRYPAREN() is actually simulating a recursive call to
2365 * regmatch() using gotos instead. Thus the values of any local variables
2366 * not saved in the regmatch_state structure will have been lost when
2367 * execution resumes on the next line .
5d9a96ca
DM
2368 *
2369 * States (ie the st pointer) are allocated in slabs of about 4K in size.
2370 * PL_regmatch_state always points to the currently active state, and
2371 * PL_regmatch_slab points to the slab currently containing PL_regmatch_state.
2372 * The first time regmatch is called, the first slab is allocated, and is
2373 * never freed until interpreter desctruction. When the slab is full,
2374 * a new one is allocated chained to the end. At exit from regmatch, slabs
2375 * allocated since entry are freed.
d6a28714 2376 */
95b24440 2377
40a82448 2378
5bc10b2c 2379#define DEBUG_STATE_pp(pp) \
265c4333 2380 DEBUG_STATE_r({ \
5bc10b2c
DM
2381 DUMP_EXEC_POS(locinput, scan, do_utf8); \
2382 PerlIO_printf(Perl_debug_log, \
2383 " %*s"pp" %s\n", \
2384 depth*2, "", \
03363afd 2385 reg_name[st->resume_state] ); \
265c4333 2386 });
5bc10b2c 2387
40a82448 2388
3dab1dad 2389#define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1)
95b24440 2390
3df15adc 2391#ifdef DEBUGGING
5bc10b2c 2392
ab3bbdeb
YO
2393STATIC void
2394S_debug_start_match(pTHX_ const regexp *prog, const bool do_utf8,
2395 const char *start, const char *end, const char *blurb)
2396{
2397 const bool utf8_pat= prog->reganch & ROPT_UTF8 ? 1 : 0;
2398 if (!PL_colorset)
2399 reginitcolors();
2400 {
2401 RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0),
2402 prog->precomp, prog->prelen, 60);
2403
2404 RE_PV_QUOTED_DECL(s1, do_utf8, PERL_DEBUG_PAD_ZERO(1),
2405 start, end - start, 60);
2406
2407 PerlIO_printf(Perl_debug_log,
2408 "%s%s REx%s %s against %s\n",
2409 PL_colors[4], blurb, PL_colors[5], s0, s1);
2410
2411 if (do_utf8||utf8_pat)
1de06328
YO
2412 PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n",
2413 utf8_pat ? "pattern" : "",
2414 utf8_pat && do_utf8 ? " and " : "",
2415 do_utf8 ? "string" : ""
ab3bbdeb
YO
2416 );
2417 }
2418}
3df15adc
YO
2419
2420STATIC void
786e8c11
YO
2421S_dump_exec_pos(pTHX_ const char *locinput,
2422 const regnode *scan,
2423 const char *loc_regeol,
2424 const char *loc_bostr,
2425 const char *loc_reg_starttry,
2426 const bool do_utf8)
07be1b83 2427{
786e8c11 2428 const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4];
07be1b83 2429 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
786e8c11 2430 int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput);
07be1b83
YO
2431 /* The part of the string before starttry has one color
2432 (pref0_len chars), between starttry and current
2433 position another one (pref_len - pref0_len chars),
2434 after the current position the third one.
2435 We assume that pref0_len <= pref_len, otherwise we
2436 decrease pref0_len. */
786e8c11
YO
2437 int pref_len = (locinput - loc_bostr) > (5 + taill) - l
2438 ? (5 + taill) - l : locinput - loc_bostr;
07be1b83
YO
2439 int pref0_len;
2440
2441 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2442 pref_len++;
786e8c11
YO
2443 pref0_len = pref_len - (locinput - loc_reg_starttry);
2444 if (l + pref_len < (5 + taill) && l < loc_regeol - locinput)
2445 l = ( loc_regeol - locinput > (5 + taill) - pref_len
2446 ? (5 + taill) - pref_len : loc_regeol - locinput);
07be1b83
YO
2447 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2448 l--;
2449 if (pref0_len < 0)
2450 pref0_len = 0;
2451 if (pref0_len > pref_len)
2452 pref0_len = pref_len;
2453 {
3df15adc 2454 const int is_uni = (do_utf8 && OP(scan) != CANY) ? 1 : 0;
0df25f3d 2455
ab3bbdeb 2456 RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0),
1de06328 2457 (locinput - pref_len),pref0_len, 60, 4, 5);
0df25f3d 2458
ab3bbdeb 2459 RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1),
3df15adc 2460 (locinput - pref_len + pref0_len),
1de06328 2461 pref_len - pref0_len, 60, 2, 3);
0df25f3d 2462
ab3bbdeb 2463 RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2),
1de06328 2464 locinput, loc_regeol - locinput, 10, 0, 1);
0df25f3d 2465
1de06328 2466 const STRLEN tlen=len0+len1+len2;
3df15adc 2467 PerlIO_printf(Perl_debug_log,
ab3bbdeb 2468 "%4"IVdf" <%.*s%.*s%s%.*s>%*s|",
786e8c11 2469 (IV)(locinput - loc_bostr),
07be1b83 2470 len0, s0,
07be1b83 2471 len1, s1,
07be1b83 2472 (docolor ? "" : "> <"),
07be1b83 2473 len2, s2,
f9f4320a 2474 (int)(tlen > 19 ? 0 : 19 - tlen),
07be1b83
YO
2475 "");
2476 }
2477}
3df15adc 2478
07be1b83
YO
2479#endif
2480
d6a28714 2481STATIC I32 /* 0 failure, 1 success */
3b0527fe 2482S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog)
d6a28714 2483{
54df2634
NC
2484#if PERL_VERSION < 9
2485 dMY_CXT;
2486#endif
27da23d5 2487 dVAR;
95b24440 2488 register const bool do_utf8 = PL_reg_match_utf8;
4ad0818d 2489 const U32 uniflags = UTF8_ALLOW_DEFAULT;
95b24440 2490
3b0527fe
DM
2491 regexp *rex = reginfo->prog;
2492
5d9a96ca
DM
2493 regmatch_slab *orig_slab;
2494 regmatch_state *orig_state;
a3621e74 2495
5d9a96ca
DM
2496 /* the current state. This is a cached copy of PL_regmatch_state */
2497 register regmatch_state *st;
95b24440 2498
5d9a96ca
DM
2499 /* cache heavy used fields of st in registers */
2500 register regnode *scan;
2501 register regnode *next;
24d3c4a9
DM
2502 register I32 n = 0; /* general value; init to avoid compiler warning */
2503 register I32 ln = 0; /* len or last; init to avoid compiler warning */
5d9a96ca 2504 register char *locinput = PL_reginput;
5d9a96ca 2505 register I32 nextchr; /* is always set to UCHARAT(locinput) */
24d3c4a9 2506
b69b0499 2507 bool result = 0; /* return value of S_regmatch */
24d3c4a9 2508 int depth = 0; /* depth of backtrack stack */
6bda09f9 2509 int nochange_depth = 0; /* depth of RECURSE recursion with nochange*/
77cb431f
DM
2510 regmatch_state *yes_state = NULL; /* state to pop to on success of
2511 subpattern */
faec1544 2512 regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */
b8591aee 2513 struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */
40a82448 2514 U32 state_num;
10edeb5d 2515
24d3c4a9
DM
2516 /* these three flags are set by various ops to signal information to
2517 * the very next op. They have a useful lifetime of exactly one loop
2518 * iteration, and are not preserved or restored by state pushes/pops
2519 */
2520 bool sw = 0; /* the condition value in (?(cond)a|b) */
2521 bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */
2522 int logical = 0; /* the following EVAL is:
2523 0: (?{...})
2524 1: (?(?{...})X|Y)
2525 2: (??{...})
2526 or the following IFMATCH/UNLESSM is:
2527 false: plain (?=foo)
2528 true: used as a condition: (?(?=foo))
2529 */
2530
95b24440 2531#ifdef DEBUGGING
e68ec53f 2532 GET_RE_DEBUG_FLAGS_DECL;
d6a28714
JH
2533#endif
2534
5d9a96ca
DM
2535 /* on first ever call to regmatch, allocate first slab */
2536 if (!PL_regmatch_slab) {
2537 Newx(PL_regmatch_slab, 1, regmatch_slab);
2538 PL_regmatch_slab->prev = NULL;
2539 PL_regmatch_slab->next = NULL;
86545054 2540 PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab);
5d9a96ca
DM
2541 }
2542
2543 /* remember current high-water mark for exit */
2544 /* XXX this should be done with SAVE* instead */
2545 orig_slab = PL_regmatch_slab;
2546 orig_state = PL_regmatch_state;
2547
2548 /* grab next free state slot */
2549 st = ++PL_regmatch_state;
86545054 2550 if (st > SLAB_LAST(PL_regmatch_slab))
5d9a96ca
DM
2551 st = PL_regmatch_state = S_push_slab(aTHX);
2552
d6a28714
JH
2553 /* Note that nextchr is a byte even in UTF */
2554 nextchr = UCHARAT(locinput);
2555 scan = prog;
2556 while (scan != NULL) {
8ba1375e 2557
a3621e74 2558 DEBUG_EXECUTE_r( {
6136c704 2559 SV * const prop = sv_newmortal();
1de06328 2560 regnode *rnext=regnext(scan);
786e8c11 2561 DUMP_EXEC_POS( locinput, scan, do_utf8 );
32fc9b6a 2562 regprop(rex, prop, scan);
07be1b83
YO
2563
2564 PerlIO_printf(Perl_debug_log,
2565 "%3"IVdf":%*s%s(%"IVdf")\n",
5bc10b2c 2566 (IV)(scan - rex->program), depth*2, "",
07be1b83 2567 SvPVX_const(prop),
1de06328
YO
2568 (PL_regkind[OP(scan)] == END || !rnext) ?
2569 0 : (IV)(rnext - rex->program));
2a782b5b 2570 });
d6a28714
JH
2571
2572 next = scan + NEXT_OFF(scan);
2573 if (next == scan)
2574 next = NULL;
40a82448 2575 state_num = OP(scan);
d6a28714 2576
40a82448
DM
2577 reenter_switch:
2578 switch (state_num) {
d6a28714 2579 case BOL:
7fba1cd6 2580 if (locinput == PL_bostr)
d6a28714 2581 {
3b0527fe 2582 /* reginfo->till = reginfo->bol; */
b8c5462f
JH
2583 break;
2584 }
d6a28714
JH
2585 sayNO;
2586 case MBOL:
12d33761
HS
2587 if (locinput == PL_bostr ||
2588 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
d6a28714 2589 {
b8c5462f
JH
2590 break;
2591 }
d6a28714
JH
2592 sayNO;
2593 case SBOL:
c2a73568 2594 if (locinput == PL_bostr)
b8c5462f 2595 break;
d6a28714
JH
2596 sayNO;
2597 case GPOS:
3b0527fe 2598 if (locinput == reginfo->ganch)
d6a28714
JH
2599 break;
2600 sayNO;
2601 case EOL:
d6a28714
JH
2602 goto seol;
2603 case MEOL:
d6a28714 2604 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
b8c5462f 2605 sayNO;
b8c5462f 2606 break;
d6a28714
JH
2607 case SEOL:
2608 seol:
2609 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
b8c5462f 2610 sayNO;
d6a28714 2611 if (PL_regeol - locinput > 1)
b8c5462f 2612 sayNO;
b8c5462f 2613 break;
d6a28714
JH
2614 case EOS:
2615 if (PL_regeol != locinput)
b8c5462f 2616 sayNO;
d6a28714 2617 break;
ffc61ed2 2618 case SANY:
d6a28714 2619 if (!nextchr && locinput >= PL_regeol)
4633a7c4 2620 sayNO;
f33976b4
DB
2621 if (do_utf8) {
2622 locinput += PL_utf8skip[nextchr];
2623 if (locinput > PL_regeol)
2624 sayNO;
2625 nextchr = UCHARAT(locinput);
2626 }
2627 else
2628 nextchr = UCHARAT(++locinput);
2629 break;
2630 case CANY:
2631 if (!nextchr && locinput >= PL_regeol)
2632 sayNO;
b8c5462f 2633 nextchr = UCHARAT(++locinput);
a0d0e21e 2634 break;
ffc61ed2 2635 case REG_ANY:
1aa99e6b
IH
2636 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2637 sayNO;
2638 if (do_utf8) {
b8c5462f 2639 locinput += PL_utf8skip[nextchr];
d6a28714
JH
2640 if (locinput > PL_regeol)
2641 sayNO;
a0ed51b3 2642 nextchr = UCHARAT(locinput);
a0ed51b3 2643 }
1aa99e6b
IH
2644 else
2645 nextchr = UCHARAT(++locinput);
a0ed51b3 2646 break;
166ba7cd
DM
2647
2648#undef ST
2649#define ST st->u.trie
786e8c11
YO
2650 case TRIEC:
2651 /* In this case the charclass data is available inline so
2652 we can fail fast without a lot of extra overhead.
2653 */
2654 if (scan->flags == EXACT || !do_utf8) {
2655 if(!ANYOF_BITMAP_TEST(scan, *locinput)) {
2656 DEBUG_EXECUTE_r(
2657 PerlIO_printf(Perl_debug_log,
2658 "%*s %sfailed to match trie start class...%s\n",
5bc10b2c 2659 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
786e8c11
YO
2660 );
2661 sayNO_SILENT;
2662 /* NOTREACHED */
2663 }
2664 }
2665 /* FALL THROUGH */
5b47454d 2666 case TRIE:
3dab1dad 2667 {
07be1b83 2668 /* what type of TRIE am I? (utf8 makes this contextual) */
3dab1dad
YO
2669 const enum { trie_plain, trie_utf8, trie_utf8_fold }
2670 trie_type = do_utf8 ?
2671 (scan->flags == EXACT ? trie_utf8 : trie_utf8_fold)
2672 : trie_plain;
2673
2674 /* what trie are we using right now */
be8e71aa 2675 reg_trie_data * const trie
3dab1dad
YO
2676 = (reg_trie_data*)rex->data->data[ ARG( scan ) ];
2677 U32 state = trie->startstate;
166ba7cd 2678
3dab1dad
YO
2679 if (trie->bitmap && trie_type != trie_utf8_fold &&
2680 !TRIE_BITMAP_TEST(trie,*locinput)
2681 ) {
2682 if (trie->states[ state ].wordnum) {
2683 DEBUG_EXECUTE_r(
2684 PerlIO_printf(Perl_debug_log,
2685 "%*s %smatched empty string...%s\n",
5bc10b2c 2686 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3dab1dad
YO
2687 );
2688 break;
2689 } else {
2690 DEBUG_EXECUTE_r(
2691 PerlIO_printf(Perl_debug_log,
786e8c11 2692 "%*s %sfailed to match trie start class...%s\n",
5bc10b2c 2693 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3dab1dad
YO
2694 );
2695 sayNO_SILENT;
2696 }
2697 }
166ba7cd 2698
786e8c11
YO
2699 {
2700 U8 *uc = ( U8* )locinput;
2701
2702 STRLEN len = 0;
2703 STRLEN foldlen = 0;
2704 U8 *uscan = (U8*)NULL;
2705 STRLEN bufflen=0;
2706 SV *sv_accept_buff = NULL;
2707 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
2708
2709 ST.accepted = 0; /* how many accepting states we have seen */
2710 ST.B = next;
2711 ST.jump = trie->jump;
2712
2713#ifdef DEBUGGING
2714 ST.me = scan;
2715#endif
2716
2717
2718
07be1b83
YO
2719 /*
2720 traverse the TRIE keeping track of all accepting states
2721 we transition through until we get to a failing node.
2722 */
2723
a3621e74 2724 while ( state && uc <= (U8*)PL_regeol ) {
786e8c11 2725 U32 base = trie->states[ state ].trans.base;
f9f4320a 2726 UV uvc = 0;
786e8c11
YO
2727 U16 charid;
2728 /* We use charid to hold the wordnum as we don't use it
2729 for charid until after we have done the wordnum logic.
2730 We define an alias just so that the wordnum logic reads
2731 more naturally. */
2732
2733#define got_wordnum charid
2734 got_wordnum = trie->states[ state ].wordnum;
2735
2736 if ( got_wordnum ) {
2737 if ( ! ST.accepted ) {
5b47454d
DM
2738 ENTER;
2739 SAVETMPS;
2740 bufflen = TRIE_INITAL_ACCEPT_BUFFLEN;
2741 sv_accept_buff=newSV(bufflen *
2742 sizeof(reg_trie_accepted) - 1);
786e8c11 2743 SvCUR_set(sv_accept_buff, 0);
5b47454d
DM
2744 SvPOK_on(sv_accept_buff);
2745 sv_2mortal(sv_accept_buff);
166ba7cd
DM
2746 SAVETMPS;
2747 ST.accept_buff =
5b47454d
DM
2748 (reg_trie_accepted*)SvPV_nolen(sv_accept_buff );
2749 }
786e8c11 2750 do {
166ba7cd 2751 if (ST.accepted >= bufflen) {
5b47454d 2752 bufflen *= 2;
166ba7cd 2753 ST.accept_buff =(reg_trie_accepted*)
5b47454d
DM
2754 SvGROW(sv_accept_buff,
2755 bufflen * sizeof(reg_trie_accepted));
2756 }
2757 SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff)
2758 + sizeof(reg_trie_accepted));
a3621e74 2759
786e8c11
YO
2760
2761 ST.accept_buff[ST.accepted].wordnum = got_wordnum;
2762 ST.accept_buff[ST.accepted].endpos = uc;
2763 ++ST.accepted;
2764 } while (trie->nextword && (got_wordnum= trie->nextword[got_wordnum]));
2765 }
2766#undef got_wordnum
a3621e74 2767
07be1b83 2768 DEBUG_TRIE_EXECUTE_r({
786e8c11 2769 DUMP_EXEC_POS( (char *)uc, scan, do_utf8 );
a3621e74 2770 PerlIO_printf( Perl_debug_log,
786e8c11 2771 "%*s %sState: %4"UVxf" Accepted: %4"UVxf" ",
5bc10b2c 2772 2+depth * 2, "", PL_colors[4],
786e8c11 2773 (UV)state, (UV)ST.accepted );
07be1b83 2774 });
a3621e74
YO
2775
2776 if ( base ) {
4cadc6a9
YO
2777 REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len,
2778 uvc, charid, foldlen, foldbuf, uniflags);
a3621e74 2779
5b47454d
DM
2780 if (charid &&
2781 (base + charid > trie->uniquecharcount )
2782 && (base + charid - 1 - trie->uniquecharcount
2783 < trie->lasttrans)
2784 && trie->trans[base + charid - 1 -
2785 trie->uniquecharcount].check == state)
2786 {
2787 state = trie->trans[base + charid - 1 -
2788 trie->uniquecharcount ].next;
2789 }
2790 else {
2791 state = 0;
2792 }
2793 uc += len;
2794
2795 }
2796 else {
a3621e74
YO
2797 state = 0;
2798 }
2799 DEBUG_TRIE_EXECUTE_r(
e4584336 2800 PerlIO_printf( Perl_debug_log,
786e8c11 2801 "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n",
e4584336 2802 charid, uvc, (UV)state, PL_colors[5] );
a3621e74
YO
2803 );
2804 }
166ba7cd 2805 if (!ST.accepted )
a3621e74 2806 sayNO;
a3621e74 2807
166ba7cd
DM
2808 DEBUG_EXECUTE_r(
2809 PerlIO_printf( Perl_debug_log,
2810 "%*s %sgot %"IVdf" possible matches%s\n",
5bc10b2c 2811 REPORT_CODE_OFF + depth * 2, "",
166ba7cd
DM
2812 PL_colors[4], (IV)ST.accepted, PL_colors[5] );
2813 );
786e8c11 2814 }}
166ba7cd
DM
2815
2816 /* FALL THROUGH */
2817
2818 case TRIE_next_fail: /* we failed - try next alterative */
2819
2820 if ( ST.accepted == 1 ) {
2821 /* only one choice left - just continue */
2822 DEBUG_EXECUTE_r({
2823 reg_trie_data * const trie
2824 = (reg_trie_data*)rex->data->data[ ARG(ST.me) ];
2825 SV ** const tmp = RX_DEBUG(reginfo->prog)
2826 ? av_fetch( trie->words, ST.accept_buff[ 0 ].wordnum-1, 0 )
2827 : NULL;
2828 PerlIO_printf( Perl_debug_log,
2829 "%*s %sonly one match left: #%d <%s>%s\n",
5bc10b2c 2830 REPORT_CODE_OFF+depth*2, "", PL_colors[4],
166ba7cd
DM
2831 ST.accept_buff[ 0 ].wordnum,
2832 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",
2833 PL_colors[5] );
2834 });
2835 PL_reginput = (char *)ST.accept_buff[ 0 ].endpos;
2836 /* in this case we free tmps/leave before we call regmatch
2837 as we wont be using accept_buff again. */
2838 FREETMPS;
2839 LEAVE;
2840 locinput = PL_reginput;
2841 nextchr = UCHARAT(locinput);
786e8c11
YO
2842
2843 if ( !ST.jump )
2844 scan = ST.B;
2845 else
2846 scan = ST.B - ST.jump[ST.accept_buff[0].wordnum];
2847
166ba7cd
DM
2848 continue; /* execute rest of RE */
2849 }
2850
2851 if (!ST.accepted-- ) {
2852 FREETMPS;
2853 LEAVE;
2854 sayNO;
2855 }
2856
a3621e74 2857 /*
166ba7cd
DM
2858 There are at least two accepting states left. Presumably
2859 the number of accepting states is going to be low,
2860 typically two. So we simply scan through to find the one
2861 with lowest wordnum. Once we find it, we swap the last
2862 state into its place and decrement the size. We then try to
2863 match the rest of the pattern at the point where the word
2864 ends. If we succeed, control just continues along the
2865 regex; if we fail we return here to try the next accepting
2866 state
2867 */
a3621e74 2868
166ba7cd
DM
2869 {
2870 U32 best = 0;
2871 U32 cur;
2872 for( cur = 1 ; cur <= ST.accepted ; cur++ ) {
2873 DEBUG_TRIE_EXECUTE_r(
f2278c82 2874 PerlIO_printf( Perl_debug_log,
166ba7cd 2875 "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n",
5bc10b2c 2876 REPORT_CODE_OFF + depth * 2, "", PL_colors[4],
166ba7cd
DM
2877 (IV)best, ST.accept_buff[ best ].wordnum, (IV)cur,
2878 ST.accept_buff[ cur ].wordnum, PL_colors[5] );
2879 );
2880
2881 if (ST.accept_buff[cur].wordnum <
2882 ST.accept_buff[best].wordnum)
2883 best = cur;
a3621e74 2884 }
166ba7cd
DM
2885
2886 DEBUG_EXECUTE_r({
2887 reg_trie_data * const trie
2888 = (reg_trie_data*)rex->data->data[ ARG(ST.me) ];
2889 SV ** const tmp = RX_DEBUG(reginfo->prog)
2890 ? av_fetch( trie->words, ST.accept_buff[ best ].wordnum - 1, 0 )
2891 : NULL;
265c4333
YO
2892 regnode *nextop=!ST.jump ?
2893 ST.B :
2894 ST.B - ST.jump[ST.accept_buff[best].wordnum];
2895 PerlIO_printf( Perl_debug_log,
2896 "%*s %strying alternation #%d <%s> at node #%d %s\n",
5bc10b2c 2897 REPORT_CODE_OFF+depth*2, "", PL_colors[4],
166ba7cd 2898 ST.accept_buff[best].wordnum,
265c4333
YO
2899 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",
2900 REG_NODE_NUM(nextop),
166ba7cd
DM
2901 PL_colors[5] );
2902 });
2903
2904 if ( best<ST.accepted ) {
2905 reg_trie_accepted tmp = ST.accept_buff[ best ];
2906 ST.accept_buff[ best ] = ST.accept_buff[ ST.accepted ];
2907 ST.accept_buff[ ST.accepted ] = tmp;
2908 best = ST.accepted;
a3621e74 2909 }
166ba7cd 2910 PL_reginput = (char *)ST.accept_buff[ best ].endpos;
786e8c11
YO
2911 if ( !ST.jump ) {
2912 PUSH_STATE_GOTO(TRIE_next, ST.B);
2913 /* NOTREACHED */
2914 } else {
2915 PUSH_STATE_GOTO(TRIE_next, ST.B - ST.jump[ST.accept_buff[best].wordnum]);
2916 /* NOTREACHED */
2917 }
2918 /* NOTREACHED */
166ba7cd 2919 }
166ba7cd
DM
2920 /* NOTREACHED */
2921
2922#undef ST
2923
95b24440
DM
2924 case EXACT: {
2925 char *s = STRING(scan);
24d3c4a9 2926 ln = STR_LEN(scan);
eb160463 2927 if (do_utf8 != UTF) {
bc517b45 2928 /* The target and the pattern have differing utf8ness. */
1aa99e6b 2929 char *l = locinput;
24d3c4a9 2930 const char * const e = s + ln;
a72c7584 2931
5ff6fc6d
JH
2932 if (do_utf8) {
2933 /* The target is utf8, the pattern is not utf8. */
1aa99e6b 2934 while (s < e) {
a3b680e6 2935 STRLEN ulen;
1aa99e6b 2936 if (l >= PL_regeol)
5ff6fc6d
JH
2937 sayNO;
2938 if (NATIVE_TO_UNI(*(U8*)s) !=
89ebb4a3 2939 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
041457d9 2940 uniflags))
5ff6fc6d 2941 sayNO;
bc517b45 2942 l += ulen;
5ff6fc6d 2943 s ++;
1aa99e6b 2944 }
5ff6fc6d
JH
2945 }
2946 else {
2947 /* The target is not utf8, the pattern is utf8. */
1aa99e6b 2948 while (s < e) {
a3b680e6 2949 STRLEN ulen;
1aa99e6b
IH
2950 if (l >= PL_regeol)
2951 sayNO;
5ff6fc6d 2952 if (NATIVE_TO_UNI(*((U8*)l)) !=
89ebb4a3 2953 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
041457d9 2954 uniflags))
1aa99e6b 2955 sayNO;
bc517b45 2956 s += ulen;
a72c7584 2957 l ++;
1aa99e6b 2958 }
5ff6fc6d 2959 }
1aa99e6b
IH
2960 locinput = l;
2961 nextchr = UCHARAT(locinput);
2962 break;
2963 }
bc517b45 2964 /* The target and the pattern have the same utf8ness. */
d6a28714
JH
2965 /* Inline the first character, for speed. */
2966 if (UCHARAT(s) != nextchr)
2967 sayNO;
24d3c4a9 2968 if (PL_regeol - locinput < ln)
d6a28714 2969 sayNO;
24d3c4a9 2970 if (ln > 1 && memNE(s, locinput, ln))
d6a28714 2971 sayNO;
24d3c4a9 2972 locinput += ln;
d6a28714
JH
2973 nextchr = UCHARAT(locinput);
2974 break;
95b24440 2975 }
d6a28714 2976 case EXACTFL:
b8c5462f
JH
2977 PL_reg_flags |= RF_tainted;
2978 /* FALL THROUGH */
95b24440 2979 case EXACTF: {
be8e71aa 2980 char * const s = STRING(scan);
24d3c4a9 2981 ln = STR_LEN(scan);
d6a28714 2982
d07ddd77
JH
2983 if (do_utf8 || UTF) {
2984 /* Either target or the pattern are utf8. */
be8e71aa 2985 const char * const l = locinput;
d07ddd77 2986 char *e = PL_regeol;
bc517b45 2987
24d3c4a9 2988 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
1feea2c7 2989 l, &e, 0, do_utf8)) {
5486206c
JH
2990 /* One more case for the sharp s:
2991 * pack("U0U*", 0xDF) =~ /ss/i,
2992 * the 0xC3 0x9F are the UTF-8
2993 * byte sequence for the U+00DF. */
2994 if (!(do_utf8 &&
2995 toLOWER(s[0]) == 's' &&
24d3c4a9 2996 ln >= 2 &&
5486206c
JH
2997 toLOWER(s[1]) == 's' &&
2998 (U8)l[0] == 0xC3 &&
2999 e - l >= 2 &&
3000 (U8)l[1] == 0x9F))
3001 sayNO;
3002 }
d07ddd77
JH
3003 locinput = e;
3004 nextchr = UCHARAT(locinput);
3005 break;
a0ed51b3 3006 }
d6a28714 3007
bc517b45
JH
3008 /* Neither the target and the pattern are utf8. */
3009
d6a28714
JH
3010 /* Inline the first character, for speed. */
3011 if (UCHARAT(s) != nextchr &&
3012 UCHARAT(s) != ((OP(scan) == EXACTF)
3013 ? PL_fold : PL_fold_locale)[nextchr])
a0ed51b3 3014 sayNO;
24d3c4a9 3015 if (PL_regeol - locinput < ln)
b8c5462f 3016 sayNO;
24d3c4a9
DM
3017 if (ln > 1 && (OP(scan) == EXACTF
3018 ? ibcmp(s, locinput, ln)
3019 : ibcmp_locale(s, locinput, ln)))
4633a7c4 3020 sayNO;
24d3c4a9 3021 locinput += ln;
d6a28714 3022 nextchr = UCHARAT(locinput);
a0d0e21e 3023 break;
95b24440 3024 }
d6a28714 3025 case ANYOF:
ffc61ed2 3026 if (do_utf8) {
9e55ce06
JH
3027 STRLEN inclasslen = PL_regeol - locinput;
3028
32fc9b6a 3029 if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8))
262b90c4 3030 goto anyof_fail;
ffc61ed2
JH
3031 if (locinput >= PL_regeol)
3032 sayNO;
0f0076b4 3033 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
b8c5462f 3034 nextchr = UCHARAT(locinput);
e0f9d4a8 3035 break;
ffc61ed2
JH
3036 }
3037 else {
3038 if (nextchr < 0)
3039 nextchr = UCHARAT(locinput);
32fc9b6a 3040 if (!REGINCLASS(rex, scan, (U8*)locinput))
262b90c4 3041 goto anyof_fail;
ffc61ed2
JH
3042 if (!nextchr && locinput >= PL_regeol)
3043 sayNO;
3044 nextchr = UCHARAT(++locinput);
e0f9d4a8
JH
3045 break;
3046 }
262b90c4 3047 anyof_fail:
e0f9d4a8
JH
3048 /* If we might have the case of the German sharp s
3049 * in a casefolding Unicode character class. */
3050
ebc501f0
JH
3051 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
3052 locinput += SHARP_S_SKIP;
e0f9d4a8 3053 nextchr = UCHARAT(locinput);
ffc61ed2 3054 }
e0f9d4a8
JH
3055 else
3056 sayNO;
b8c5462f 3057 break;
d6a28714 3058 case ALNUML:
b8c5462f
JH
3059 PL_reg_flags |= RF_tainted;
3060 /* FALL THROUGH */
d6a28714 3061 case ALNUM:
b8c5462f 3062 if (!nextchr)
4633a7c4 3063 sayNO;
ffc61ed2 3064 if (do_utf8) {
1a4fad37 3065 LOAD_UTF8_CHARCLASS_ALNUM();
ffc61ed2 3066 if (!(OP(scan) == ALNUM
bb7a0f54 3067 ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
d6a28714 3068 : isALNUM_LC_utf8((U8*)locinput)))
b8c5462f
JH
3069 {
3070 sayNO;
a0ed51b3 3071 }
b8c5462f 3072 locinput += PL_utf8skip[nextchr];
a0ed51b3
LW
3073 nextchr = UCHARAT(locinput);
3074 break;
3075 }
ffc61ed2 3076 if (!(OP(scan) == ALNUM
d6a28714 3077 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
bbce6d69 3078 sayNO;
b8c5462f 3079 nextchr = UCHARAT(++locinput);
a0d0e21e 3080 break;
d6a28714 3081 case NALNUML:
b8c5462f
JH
3082 PL_reg_flags |= RF_tainted;
3083 /* FALL THROUGH */
d6a28714
JH
3084 case NALNUM:
3085 if (!nextchr && locinput >= PL_regeol)
a0ed51b3 3086 sayNO;
ffc61ed2 3087 if (do_utf8) {
1a4fad37 3088 LOAD_UTF8_CHARCLASS_ALNUM();
ffc61ed2 3089 if (OP(scan) == NALNUM
bb7a0f54 3090 ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
d6a28714
JH
3091 : isALNUM_LC_utf8((U8*)locinput))
3092 {
b8c5462f 3093 sayNO;
d6a28714 3094 }
b8c5462f
JH
3095 locinput += PL_utf8skip[nextchr];
3096 nextchr = UCHARAT(locinput);
3097 break;
3098 }
ffc61ed2 3099 if (OP(scan) == NALNUM
d6a28714 3100 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
b8c5462f 3101 sayNO;
76e3520e 3102 nextchr = UCHARAT(++locinput);
a0d0e21e 3103 break;
d6a28714
JH
3104 case BOUNDL:
3105 case NBOUNDL:
3280af22 3106 PL_reg_flags |= RF_tainted;
bbce6d69 3107 /* FALL THROUGH */
d6a28714
JH
3108 case BOUND:
3109 case NBOUND:
3110 /* was last char in word? */
ffc61ed2 3111 if (do_utf8) {
12d33761 3112 if (locinput == PL_bostr)
24d3c4a9 3113 ln = '\n';
ffc61ed2 3114 else {
a3b680e6 3115 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
9041c2e3 3116
24d3c4a9 3117 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
ffc61ed2
JH
3118 }
3119 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
24d3c4a9 3120 ln = isALNUM_uni(ln);
1a4fad37 3121 LOAD_UTF8_CHARCLASS_ALNUM();
3568d838 3122 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
ffc61ed2
JH
3123 }
3124 else {
24d3c4a9 3125 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
ffc61ed2
JH
3126 n = isALNUM_LC_utf8((U8*)locinput);
3127 }
a0ed51b3 3128 }
d6a28714 3129 else {
24d3c4a9 3130 ln = (locinput != PL_bostr) ?
12d33761 3131 UCHARAT(locinput - 1) : '\n';
ffc61ed2 3132 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
24d3c4a9 3133 ln = isALNUM(ln);
ffc61ed2
JH
3134 n = isALNUM(nextchr);
3135 }
3136 else {
24d3c4a9 3137 ln = isALNUM_LC(ln);
ffc61ed2
JH
3138 n = isALNUM_LC(nextchr);
3139 }
d6a28714 3140 }
24d3c4a9 3141 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
ffc61ed2
JH
3142 OP(scan) == BOUNDL))
3143 sayNO;
a0ed51b3 3144 break;
d6a28714 3145 case SPACEL:
3280af22 3146 PL_reg_flags |= RF_tainted;
bbce6d69 3147 /* FALL THROUGH */
d6a28714 3148 case SPACE:
9442cb0e 3149 if (!nextchr)
4633a7c4 3150 sayNO;
1aa99e6b 3151 if (do_utf8) {
fd400ab9 3152 if (UTF8_IS_CONTINUED(nextchr)) {
1a4fad37 3153 LOAD_UTF8_CHARCLASS_SPACE();
ffc61ed2 3154 if (!(OP(scan) == SPACE
bb7a0f54 3155 ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
ffc61ed2
JH
3156 : isSPACE_LC_utf8((U8*)locinput)))
3157 {
3158 sayNO;
3159 }
3160 locinput += PL_utf8skip[nextchr];
3161 nextchr = UCHARAT(locinput);
3162 break;
d6a28714 3163 }
ffc61ed2
JH
3164 if (!(OP(scan) == SPACE
3165 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3166 sayNO;
3167 nextchr = UCHARAT(++locinput);
3168 }
3169 else {
3170 if (!(OP(scan) == SPACE
3171 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3172 sayNO;
3173 nextchr = UCHARAT(++locinput);
a0ed51b3 3174 }
a0ed51b3 3175 break;
d6a28714 3176 case NSPACEL:
3280af22 3177 PL_reg_flags |= RF_tainted;
bbce6d69 3178 /* FALL THROUGH */
d6a28714 3179 case NSPACE:
9442cb0e 3180 if (!nextchr && locinput >= PL_regeol)
b8c5462f 3181 sayNO;
1aa99e6b 3182 if (do_utf8) {
1a4fad37 3183 LOAD_UTF8_CHARCLASS_SPACE();
ffc61ed2 3184 if (OP(scan) == NSPACE
bb7a0f54 3185 ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
d6a28714 3186 : isSPACE_LC_utf8((U8*)locinput))
b8c5462f
JH
3187 {
3188 sayNO;
3189 }
3190 locinput += PL_utf8skip[nextchr];
3191 nextchr = UCHARAT(locinput);
3192 break;
a0ed51b3 3193 }
ffc61ed2 3194 if (OP(scan) == NSPACE
d6a28714 3195 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
4633a7c4 3196 sayNO;
76e3520e 3197 nextchr = UCHARAT(++locinput);
a0d0e21e 3198 break;
d6a28714 3199 case DIGITL:
a0ed51b3
LW
3200 PL_reg_flags |= RF_tainted;
3201 /* FALL THROUGH */
d6a28714 3202 case DIGIT:
9442cb0e 3203 if (!nextchr)
a0ed51b3 3204 sayNO;
1aa99e6b 3205 if (do_utf8) {
1a4fad37 3206 LOAD_UTF8_CHARCLASS_DIGIT();
ffc61ed2 3207 if (!(OP(scan) == DIGIT
bb7a0f54 3208 ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
9442cb0e 3209 : isDIGIT_LC_utf8((U8*)locinput)))
dfe13c55 3210 {
a0ed51b3 3211 sayNO;
dfe13c55 3212 }
6f06b55f 3213 locinput += PL_utf8skip[nextchr];
a0ed51b3
LW
3214 nextchr = UCHARAT(locinput);
3215 break;
3216 }
ffc61ed2 3217 if (!(OP(scan) == DIGIT
9442cb0e 3218 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
a0ed51b3
LW
3219 sayNO;
3220 nextchr = UCHARAT(++locinput);
3221 break;
d6a28714 3222 case NDIGITL:
b8c5462f
JH
3223 PL_reg_flags |= RF_tainted;
3224 /* FALL THROUGH */
d6a28714 3225 case NDIGIT:
9442cb0e 3226 if (!nextchr && locinput >= PL_regeol)
b8c5462f 3227 sayNO;
1aa99e6b 3228 if (do_utf8) {
1a4fad37 3229 LOAD_UTF8_CHARCLASS_DIGIT();
ffc61ed2 3230 if (OP(scan) == NDIGIT
bb7a0f54 3231 ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
9442cb0e
GS
3232 : isDIGIT_LC_utf8((U8*)locinput))
3233 {
a0ed51b3 3234 sayNO;
9442cb0e 3235 }
6f06b55f 3236 locinput += PL_utf8skip[nextchr];
a0ed51b3
LW
3237 nextchr = UCHARAT(locinput);
3238 break;
3239 }
ffc61ed2 3240 if (OP(scan) == NDIGIT
9442cb0e 3241 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
a0ed51b3
LW
3242 sayNO;
3243 nextchr = UCHARAT(++locinput);
3244 break;
3245 case CLUMP:
b7c83a7e 3246 if (locinput >= PL_regeol)
a0ed51b3 3247 sayNO;
b7c83a7e 3248 if (do_utf8) {
1a4fad37 3249 LOAD_UTF8_CHARCLASS_MARK();
b7c83a7e
JH
3250 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3251 sayNO;
3252 locinput += PL_utf8skip[nextchr];
3253 while (locinput < PL_regeol &&
3254 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3255 locinput += UTF8SKIP(locinput);
3256 if (locinput > PL_regeol)
3257 sayNO;
eb08e2da
JH
3258 }
3259 else
3260 locinput++;
a0ed51b3
LW
3261 nextchr = UCHARAT(locinput);
3262 break;
c8756f30 3263 case REFFL:
3280af22 3264 PL_reg_flags |= RF_tainted;
c8756f30 3265 /* FALL THROUGH */
c277df42 3266 case REF:
95b24440
DM
3267 case REFF: {
3268 char *s;
c277df42 3269 n = ARG(scan); /* which paren pair */
24d3c4a9 3270 ln = PL_regstartp[n];
2c2d71f5 3271 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
24d3c4a9 3272 if ((I32)*PL_reglastparen < n || ln == -1)
af3f8c16 3273 sayNO; /* Do not match unless seen CLOSEn. */
24d3c4a9 3274 if (ln == PL_regendp[n])
a0d0e21e 3275 break;
a0ed51b3 3276
24d3c4a9 3277 s = PL_bostr + ln;
1aa99e6b 3278 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
a0ed51b3 3279 char *l = locinput;
a3b680e6 3280 const char *e = PL_bostr + PL_regendp[n];
a0ed51b3
LW
3281 /*
3282 * Note that we can't do the "other character" lookup trick as
3283 * in the 8-bit case (no pun intended) because in Unicode we
3284 * have to map both upper and title case to lower case.
3285 */
3286 if (OP(scan) == REFF) {
3287 while (s < e) {
a3b680e6
AL
3288 STRLEN ulen1, ulen2;
3289 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3290 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3291
a0ed51b3
LW
3292 if (l >= PL_regeol)
3293 sayNO;
a2a2844f
JH
3294 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3295 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
7114a2d2 3296 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
a0ed51b3 3297 sayNO;
a2a2844f
JH
3298 s += ulen1;
3299 l += ulen2;
a0ed51b3
LW
3300 }
3301 }
3302 locinput = l;
3303 nextchr = UCHARAT(locinput);
3304 break;
3305 }
3306
a0d0e21e 3307 /* Inline the first character, for speed. */
76e3520e 3308 if (UCHARAT(s) != nextchr &&
c8756f30
AK
3309 (OP(scan) == REF ||
3310 (UCHARAT(s) != ((OP(scan) == REFF
22c35a8c 3311 ? PL_fold : PL_fold_locale)[nextchr]))))
4633a7c4 3312 sayNO;
24d3c4a9
DM
3313 ln = PL_regendp[n] - ln;
3314 if (locinput + ln > PL_regeol)
4633a7c4 3315 sayNO;
24d3c4a9
DM
3316 if (ln > 1 && (OP(scan) == REF
3317 ? memNE(s, locinput, ln)
c8756f30 3318 : (OP(scan) == REFF
24d3c4a9
DM
3319 ? ibcmp(s, locinput, ln)
3320 : ibcmp_locale(s, locinput, ln))))
4633a7c4 3321 sayNO;
24d3c4a9 3322 locinput += ln;
76e3520e 3323 nextchr = UCHARAT(locinput);
a0d0e21e 3324 break;
95b24440 3325 }
a0d0e21e
LW
3326
3327 case NOTHING:
c277df42 3328 case TAIL:
a0d0e21e
LW
3329 break;
3330 case BACK:
3331 break;
40a82448
DM
3332
3333#undef ST
3334#define ST st->u.eval
c277df42 3335 {
c277df42 3336 SV *ret;
6bda09f9
YO
3337 regexp *re;
3338 regnode *startpoint;
3339
3340 case SRECURSE:
3341 case RECURSE: /* /(...(?1))/ */
3342 if (cur_eval && cur_eval->locinput==locinput) {
3343 if (cur_eval->u.eval.close_paren == ARG(scan))
3344 Perl_croak(aTHX_ "Infinite recursion in RECURSE in regexp");
3345 if ( ++nochange_depth > MAX_RECURSE_EVAL_NOCHANGE_DEPTH )
3346 Perl_croak(aTHX_ "RECURSE without pos change exceeded limit in regexp");
3347 } else {
3348 nochange_depth = 0;
3349 }
3350 re = rex;
3351 (void)ReREFCNT_inc(rex);
3352 if (OP(scan)==RECURSE) {
3353 startpoint = scan + ARG2L(scan);
3354 ST.close_paren = ARG(scan);
3355 } else {
3356 startpoint = re->program+1;
3357 ST.close_paren = 0;
3358 }
3359 goto eval_recurse_doit;
3360 /* NOTREACHED */
3361 case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */
3362 if (cur_eval && cur_eval->locinput==locinput) {
3363 if ( ++nochange_depth > MAX_RECURSE_EVAL_NOCHANGE_DEPTH )
3364 Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regexp");
3365 } else {
3366 nochange_depth = 0;
3367 }
8e5e9ebe 3368 {
4aabdb9b
DM
3369 /* execute the code in the {...} */
3370 dSP;
6136c704 3371 SV ** const before = SP;
4aabdb9b
DM
3372 OP_4tree * const oop = PL_op;
3373 COP * const ocurcop = PL_curcop;
3374 PAD *old_comppad;
4aabdb9b
DM
3375
3376 n = ARG(scan);
32fc9b6a 3377 PL_op = (OP_4tree*)rex->data->data[n];
4aabdb9b 3378 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
32fc9b6a 3379 PAD_SAVE_LOCAL(old_comppad, (PAD*)rex->data->data[n + 2]);
4aabdb9b
DM
3380 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
3381
8e5e9ebe
RGS
3382 CALLRUNOPS(aTHX); /* Scalar context. */
3383 SPAGAIN;
3384 if (SP == before)
075aa684 3385 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
8e5e9ebe
RGS
3386 else {
3387 ret = POPs;
3388 PUTBACK;
3389 }
4aabdb9b
DM
3390
3391 PL_op = oop;
3392 PAD_RESTORE_LOCAL(old_comppad);
3393 PL_curcop = ocurcop;
24d3c4a9 3394 if (!logical) {
4aabdb9b
DM
3395 /* /(?{...})/ */
3396 sv_setsv(save_scalar(PL_replgv), ret);
4aabdb9b
DM
3397 break;
3398 }
8e5e9ebe 3399 }
24d3c4a9
DM
3400 if (logical == 2) { /* Postponed subexpression: /(??{...})/ */
3401 logical = 0;
4aabdb9b 3402 {
4f639d21
DM
3403 /* extract RE object from returned value; compiling if
3404 * necessary */
3405
6136c704 3406 MAGIC *mg = NULL;
be8e71aa 3407 const SV *sv;
faf82a0b
AE
3408 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
3409 mg = mg_find(sv, PERL_MAGIC_qr);
3410 else if (SvSMAGICAL(ret)) {
3411 if (SvGMAGICAL(ret))
3412 sv_unmagic(ret, PERL_MAGIC_qr);
3413 else
3414 mg = mg_find(ret, PERL_MAGIC_qr);
0f5d15d6 3415 }
faf82a0b 3416
0f5d15d6
IZ
3417 if (mg) {
3418 re = (regexp *)mg->mg_obj;
df0003d4 3419 (void)ReREFCNT_inc(re);
0f5d15d6
IZ
3420 }
3421 else {
3422 STRLEN len;
6136c704 3423 const char * const t = SvPV_const(ret, len);
0f5d15d6 3424 PMOP pm;
a3b680e6 3425 const I32 osize = PL_regsize;
0f5d15d6 3426
5fcd1c1b 3427 Zero(&pm, 1, PMOP);
4aabdb9b 3428 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
f9f4320a 3429 re = CALLREGCOMP((char*)t, (char*)t + len, &pm);
9041c2e3 3430 if (!(SvFLAGS(ret)
faf82a0b
AE
3431 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
3432 | SVs_GMG)))
14befaf4
DM
3433 sv_magic(ret,(SV*)ReREFCNT_inc(re),
3434 PERL_MAGIC_qr,0,0);
0f5d15d6 3435 PL_regsize = osize;
0f5d15d6 3436 }
4aabdb9b 3437 }
6bda09f9
YO
3438 DEBUG_EXECUTE_r(
3439 debug_start_match(re, do_utf8, locinput, PL_regeol,
3440 "Matching embedded");
3441 );
3442 startpoint = re->program + 1;
3443 ST.close_paren = 0; /* only used for RECURSE */
3444 /* borrowed from regtry */
3445 if (PL_reg_start_tmpl <= re->nparens) {
3446 PL_reg_start_tmpl = re->nparens*3/2 + 3;
3447 if(PL_reg_start_tmp)
3448 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
3449 else
3450 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
3451 }
aa283a38 3452
6bda09f9 3453 eval_recurse_doit: /* Share code with RECURSE below this line */
aa283a38 3454 /* run the pattern returned from (??{...}) */
40a82448
DM
3455 ST.cp = regcppush(0); /* Save *all* the positions. */
3456 REGCP_SET(ST.lastcp);
6bda09f9
YO
3457
3458 PL_regstartp = re->startp; /* essentially NOOP on RECURSE */
3459 PL_regendp = re->endp; /* essentially NOOP on RECURSE */
3460
4aabdb9b
DM
3461 *PL_reglastparen = 0;
3462 *PL_reglastcloseparen = 0;
4aabdb9b 3463 PL_reginput = locinput;
4aabdb9b
DM
3464
3465 /* XXXX This is too dramatic a measure... */
3466 PL_reg_maxiter = 0;
3467
faec1544
DM
3468 ST.toggle_reg_flags = PL_reg_flags;
3469 if (re->reganch & ROPT_UTF8)
3470 PL_reg_flags |= RF_utf8;
3471 else
3472 PL_reg_flags &= ~RF_utf8;
3473 ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */
3474
40a82448 3475 ST.prev_rex = rex;
faec1544 3476 ST.prev_curlyx = cur_curlyx;
aa283a38 3477 rex = re;
faec1544 3478 cur_curlyx = NULL;
40a82448 3479 ST.B = next;
faec1544
DM
3480 ST.prev_eval = cur_eval;
3481 cur_eval = st;
faec1544 3482 /* now continue from first node in postoned RE */
6bda09f9 3483 PUSH_YES_STATE_GOTO(EVAL_AB, startpoint);
4aabdb9b 3484 /* NOTREACHED */
a0ed51b3 3485 }
24d3c4a9
DM
3486 /* logical is 1, /(?(?{...})X|Y)/ */
3487 sw = (bool)SvTRUE(ret);
3488 logical = 0;
c277df42
IZ
3489 break;
3490 }
40a82448 3491
faec1544
DM
3492 case EVAL_AB: /* cleanup after a successful (??{A})B */
3493 /* note: this is called twice; first after popping B, then A */
3494 PL_reg_flags ^= ST.toggle_reg_flags;
40a82448
DM
3495 ReREFCNT_dec(rex);
3496 rex = ST.prev_rex;
faec1544
DM
3497 regcpblow(ST.cp);
3498 cur_eval = ST.prev_eval;
3499 cur_curlyx = ST.prev_curlyx;
40a82448
DM
3500 /* XXXX This is too dramatic a measure... */
3501 PL_reg_maxiter = 0;
262b90c4 3502 sayYES;
40a82448 3503
40a82448 3504
faec1544
DM
3505 case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */
3506 /* note: this is called twice; first after popping B, then A */
3507 PL_reg_flags ^= ST.toggle_reg_flags;
3508 ReREFCNT_dec(rex);
3509 rex = ST.prev_rex;
40a82448
DM
3510 PL_reginput = locinput;
3511 REGCP_UNWIND(ST.lastcp);
3512 regcppop(rex);
faec1544
DM
3513 cur_eval = ST.prev_eval;
3514 cur_curlyx = ST.prev_curlyx;
3515 /* XXXX This is too dramatic a measure... */
3516 PL_reg_maxiter = 0;
40a82448 3517 sayNO_SILENT;
40a82448
DM
3518#undef ST
3519
a0d0e21e 3520 case OPEN:
c277df42 3521 n = ARG(scan); /* which paren pair */
3280af22
NIS
3522 PL_reg_start_tmp[n] = locinput;
3523 if (n > PL_regsize)
3524 PL_regsize = n;
a0d0e21e
LW
3525 break;
3526 case CLOSE:
c277df42 3527 n = ARG(scan); /* which paren pair */
cf93c79d
IZ
3528 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
3529 PL_regendp[n] = locinput - PL_bostr;
eb160463 3530 if (n > (I32)*PL_reglastparen)
3280af22 3531 *PL_reglastparen = n;
a01268b5 3532 *PL_reglastcloseparen = n;
6bda09f9
YO
3533 if (cur_eval && cur_eval->u.eval.close_paren == (U32)n) {
3534 goto fake_end;
3535 }
a0d0e21e 3536 break;
c277df42
IZ
3537 case GROUPP:
3538 n = ARG(scan); /* which paren pair */
24d3c4a9 3539 sw = (bool)((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
c277df42
IZ
3540 break;
3541 case IFTHEN:
2c2d71f5 3542 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
24d3c4a9 3543 if (sw)
c277df42
IZ
3544 next = NEXTOPER(NEXTOPER(scan));
3545 else {
3546 next = scan + ARG(scan);
3547 if (OP(next) == IFTHEN) /* Fake one. */
3548 next = NEXTOPER(NEXTOPER(next));
3549 }
3550 break;
3551 case LOGICAL:
24d3c4a9 3552 logical = scan->flags;
c277df42 3553 break;
c476f425 3554
2ab05381 3555/*******************************************************************
2ab05381 3556
c476f425
DM
3557The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/
3558pattern, where A and B are subpatterns. (For simple A, CURLYM or
3559STAR/PLUS/CURLY/CURLYN are used instead.)
2ab05381 3560
c476f425 3561A*B is compiled as <CURLYX><A><WHILEM><B>
2ab05381 3562
c476f425
DM
3563On entry to the subpattern, CURLYX is called. This pushes a CURLYX
3564state, which contains the current count, initialised to -1. It also sets
3565cur_curlyx to point to this state, with any previous value saved in the
3566state block.
2ab05381 3567
c476f425
DM
3568CURLYX then jumps straight to the WHILEM op, rather than executing A,
3569since the pattern may possibly match zero times (i.e. it's a while {} loop
3570rather than a do {} while loop).
2ab05381 3571
c476f425
DM
3572Each entry to WHILEM represents a successful match of A. The count in the
3573CURLYX block is incremented, another WHILEM state is pushed, and execution
3574passes to A or B depending on greediness and the current count.
2ab05381 3575
c476f425
DM
3576For example, if matching against the string a1a2a3b (where the aN are
3577substrings that match /A/), then the match progresses as follows: (the
3578pushed states are interspersed with the bits of strings matched so far):
2ab05381 3579
c476f425
DM
3580 <CURLYX cnt=-1>
3581 <CURLYX cnt=0><WHILEM>
3582 <CURLYX cnt=1><WHILEM> a1 <WHILEM>
3583 <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM>
3584 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM>
3585 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b
2ab05381 3586
c476f425
DM
3587(Contrast this with something like CURLYM, which maintains only a single
3588backtrack state:
2ab05381 3589
c476f425
DM
3590 <CURLYM cnt=0> a1
3591 a1 <CURLYM cnt=1> a2
3592 a1 a2 <CURLYM cnt=2> a3
3593 a1 a2 a3 <CURLYM cnt=3> b
3594)
2ab05381 3595
c476f425
DM
3596Each WHILEM state block marks a point to backtrack to upon partial failure
3597of A or B, and also contains some minor state data related to that
3598iteration. The CURLYX block, pointed to by cur_curlyx, contains the
3599overall state, such as the count, and pointers to the A and B ops.
2ab05381 3600
c476f425
DM
3601This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx
3602must always point to the *current* CURLYX block, the rules are:
2ab05381 3603
c476f425
DM
3604When executing CURLYX, save the old cur_curlyx in the CURLYX state block,
3605and set cur_curlyx to point the new block.
2ab05381 3606
c476f425
DM
3607When popping the CURLYX block after a successful or unsuccessful match,
3608restore the previous cur_curlyx.
2ab05381 3609
c476f425
DM
3610When WHILEM is about to execute B, save the current cur_curlyx, and set it
3611to the outer one saved in the CURLYX block.
2ab05381 3612
c476f425
DM
3613When popping the WHILEM block after a successful or unsuccessful B match,
3614restore the previous cur_curlyx.
2ab05381 3615
c476f425
DM
3616Here's an example for the pattern (AI* BI)*BO
3617I and O refer to inner and outer, C and W refer to CURLYX and WHILEM:
2ab05381 3618
c476f425
DM
3619cur_
3620curlyx backtrack stack
3621------ ---------------
3622NULL
3623CO <CO prev=NULL> <WO>
3624CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
3625CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
3626NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo
2ab05381 3627
c476f425
DM
3628At this point the pattern succeeds, and we work back down the stack to
3629clean up, restoring as we go:
95b24440 3630
c476f425
DM
3631CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
3632CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
3633CO <CO prev=NULL> <WO>
3634NULL
a0374537 3635
c476f425
DM
3636*******************************************************************/
3637
3638#define ST st->u.curlyx
3639
3640 case CURLYX: /* start of /A*B/ (for complex A) */
3641 {
3642 /* No need to save/restore up to this paren */
3643 I32 parenfloor = scan->flags;
3644
3645 assert(next); /* keep Coverity happy */
3646 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3647 next += ARG(next);
3648
3649 /* XXXX Probably it is better to teach regpush to support
3650 parenfloor > PL_regsize... */
3651 if (parenfloor > (I32)*PL_reglastparen)
3652 parenfloor = *PL_reglastparen; /* Pessimization... */
3653
3654 ST.prev_curlyx= cur_curlyx;
3655 cur_curlyx = st;
3656 ST.cp = PL_savestack_ix;
3657
3658 /* these fields contain the state of the current curly.
3659 * they are accessed by subsequent WHILEMs */
3660 ST.parenfloor = parenfloor;
3661 ST.min = ARG1(scan);
3662 ST.max = ARG2(scan);
3663 ST.A = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3664 ST.B = next;
24d3c4a9
DM
3665 ST.minmod = minmod;
3666 minmod = 0;
c476f425
DM
3667 ST.count = -1; /* this will be updated by WHILEM */
3668 ST.lastloc = NULL; /* this will be updated by WHILEM */
3669
3670 PL_reginput = locinput;
3671 PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next));
5f66b61c 3672 /* NOTREACHED */
c476f425 3673 }
a0d0e21e 3674
c476f425
DM
3675 case CURLYX_end: /* just finished matching all of A*B */
3676 regcpblow(ST.cp);
3677 cur_curlyx = ST.prev_curlyx;
3678 sayYES;
3679 /* NOTREACHED */
a0d0e21e 3680
c476f425
DM
3681 case CURLYX_end_fail: /* just failed to match all of A*B */
3682 regcpblow(ST.cp);
3683 cur_curlyx = ST.prev_curlyx;
3684 sayNO;
3685 /* NOTREACHED */
4633a7c4 3686
a0d0e21e 3687
c476f425
DM
3688#undef ST
3689#define ST st->u.whilem
3690
3691 case WHILEM: /* just matched an A in /A*B/ (for complex A) */
3692 {
3693 /* see the discussion above about CURLYX/WHILEM */
3694
3695 I32 n;
3696 assert(cur_curlyx); /* keep Coverity happy */
3697 n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */
3698 ST.save_lastloc = cur_curlyx->u.curlyx.lastloc;
3699 ST.cache_offset = 0;
3700 ST.cache_mask = 0;
3701
3702 PL_reginput = locinput;
3703
3704 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
3705 "%*s whilem: matched %ld out of %ld..%ld\n",
3706 REPORT_CODE_OFF+depth*2, "", (long)n,
3707 (long)cur_curlyx->u.curlyx.min,
3708 (long)cur_curlyx->u.curlyx.max)
3709 );
a0d0e21e 3710
c476f425 3711 /* First just match a string of min A's. */
a0d0e21e 3712
c476f425
DM
3713 if (n < cur_curlyx->u.curlyx.min) {
3714 cur_curlyx->u.curlyx.lastloc = locinput;
3715 PUSH_STATE_GOTO(WHILEM_A_pre, cur_curlyx->u.curlyx.A);
3716 /* NOTREACHED */
3717 }
3718
3719 /* If degenerate A matches "", assume A done. */
3720
3721 if (locinput == cur_curlyx->u.curlyx.lastloc) {
3722 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
3723 "%*s whilem: empty match detected, trying continuation...\n",
3724 REPORT_CODE_OFF+depth*2, "")
3725 );
3726 goto do_whilem_B_max;
3727 }
3728
3729 /* super-linear cache processing */
3730
3731 if (scan->flags) {
a0d0e21e 3732
2c2d71f5 3733 if (!PL_reg_maxiter) {
c476f425
DM
3734 /* start the countdown: Postpone detection until we
3735 * know the match is not *that* much linear. */
2c2d71f5 3736 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
66bf836d
DM
3737 /* possible overflow for long strings and many CURLYX's */
3738 if (PL_reg_maxiter < 0)
3739 PL_reg_maxiter = I32_MAX;
2c2d71f5
JH
3740 PL_reg_leftiter = PL_reg_maxiter;
3741 }
c476f425 3742
2c2d71f5 3743 if (PL_reg_leftiter-- == 0) {
c476f425 3744 /* initialise cache */
3298f257 3745 const I32 size = (PL_reg_maxiter + 7)/8;
2c2d71f5 3746 if (PL_reg_poscache) {
eb160463 3747 if ((I32)PL_reg_poscache_size < size) {
2c2d71f5
JH
3748 Renew(PL_reg_poscache, size, char);
3749 PL_reg_poscache_size = size;
3750 }
3751 Zero(PL_reg_poscache, size, char);
3752 }
3753 else {
3754 PL_reg_poscache_size = size;
a02a5408 3755 Newxz(PL_reg_poscache, size, char);
2c2d71f5 3756 }
c476f425
DM
3757 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
3758 "%swhilem: Detected a super-linear match, switching on caching%s...\n",
3759 PL_colors[4], PL_colors[5])
3760 );
2c2d71f5 3761 }
c476f425 3762
2c2d71f5 3763 if (PL_reg_leftiter < 0) {
c476f425
DM
3764 /* have we already failed at this position? */
3765 I32 offset, mask;
3766 offset = (scan->flags & 0xf) - 1
3767 + (locinput - PL_bostr) * (scan->flags>>4);
3768 mask = 1 << (offset % 8);
3769 offset /= 8;
3770 if (PL_reg_poscache[offset] & mask) {
3771 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
3772 "%*s whilem: (cache) already tried at this position...\n",
3773 REPORT_CODE_OFF+depth*2, "")
2c2d71f5 3774 );
3298f257 3775 sayNO; /* cache records failure */
2c2d71f5 3776 }
c476f425
DM
3777 ST.cache_offset = offset;
3778 ST.cache_mask = mask;
2c2d71f5 3779 }
c476f425 3780 }
2c2d71f5 3781
c476f425 3782 /* Prefer B over A for minimal matching. */
a687059c 3783
c476f425
DM
3784 if (cur_curlyx->u.curlyx.minmod) {
3785 ST.save_curlyx = cur_curlyx;
3786 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
3787 ST.cp = regcppush(ST.save_curlyx->u.curlyx.parenfloor);
3788 REGCP_SET(ST.lastcp);
3789 PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B);
3790 /* NOTREACHED */
3791 }
a0d0e21e 3792
c476f425
DM
3793 /* Prefer A over B for maximal matching. */
3794
3795 if (n < cur_curlyx->u.curlyx.max) { /* More greed allowed? */
3796 ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
3797 cur_curlyx->u.curlyx.lastloc = locinput;
3798 REGCP_SET(ST.lastcp);
3799 PUSH_STATE_GOTO(WHILEM_A_max, cur_curlyx->u.curlyx.A);
3800 /* NOTREACHED */
3801 }
3802 goto do_whilem_B_max;
3803 }
3804 /* NOTREACHED */
3805
3806 case WHILEM_B_min: /* just matched B in a minimal match */
3807 case WHILEM_B_max: /* just matched B in a maximal match */
3808 cur_curlyx = ST.save_curlyx;
3809 sayYES;
3810 /* NOTREACHED */
3811
3812 case WHILEM_B_max_fail: /* just failed to match B in a maximal match */
3813 cur_curlyx = ST.save_curlyx;
3814 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
3815 cur_curlyx->u.curlyx.count--;
3816 CACHEsayNO;
3817 /* NOTREACHED */
3818
3819 case WHILEM_A_min_fail: /* just failed to match A in a minimal match */
3820 REGCP_UNWIND(ST.lastcp);
3821 regcppop(rex);
3822 /* FALL THROUGH */
3823 case WHILEM_A_pre_fail: /* just failed to match even minimal A */
3824 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
3825 cur_curlyx->u.curlyx.count--;
3826 CACHEsayNO;
3827 /* NOTREACHED */
3828
3829 case WHILEM_A_max_fail: /* just failed to match A in a maximal match */
3830 REGCP_UNWIND(ST.lastcp);
3831 regcppop(rex); /* Restore some previous $<digit>s? */
3832 PL_reginput = locinput;
3833 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
3834 "%*s whilem: failed, trying continuation...\n",
3835 REPORT_CODE_OFF+depth*2, "")
3836 );
3837 do_whilem_B_max:
3838 if (cur_curlyx->u.curlyx.count >= REG_INFTY
3839 && ckWARN(WARN_REGEXP)
3840 && !(PL_reg_flags & RF_warned))
3841 {
3842 PL_reg_flags |= RF_warned;
3843 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3844 "Complex regular subexpression recursion",
3845 REG_INFTY - 1);
3846 }
3847
3848 /* now try B */
3849 ST.save_curlyx = cur_curlyx;
3850 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
3851 PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B);
3852 /* NOTREACHED */
3853
3854 case WHILEM_B_min_fail: /* just failed to match B in a minimal match */
3855 cur_curlyx = ST.save_curlyx;
3856 REGCP_UNWIND(ST.lastcp);
3857 regcppop(rex);
3858
3859 if (cur_curlyx->u.curlyx.count >= cur_curlyx->u.curlyx.max) {
3860 /* Maximum greed exceeded */
3861 if (cur_curlyx->u.curlyx.count >= REG_INFTY
3862 && ckWARN(WARN_REGEXP)
3863 && !(PL_reg_flags & RF_warned))
3864 {
3280af22 3865 PL_reg_flags |= RF_warned;
c476f425
DM
3866 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
3867 "%s limit (%d) exceeded",
3868 "Complex regular subexpression recursion",
3869 REG_INFTY - 1);
a0d0e21e 3870 }
c476f425 3871 cur_curlyx->u.curlyx.count--;
3ab3c9b4 3872 CACHEsayNO;
a0d0e21e 3873 }
c476f425
DM
3874
3875 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
3876 "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "")
3877 );
3878 /* Try grabbing another A and see if it helps. */
3879 PL_reginput = locinput;
3880 cur_curlyx->u.curlyx.lastloc = locinput;
3881 ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
3882 REGCP_SET(ST.lastcp);
3883 PUSH_STATE_GOTO(WHILEM_A_min, ST.save_curlyx->u.curlyx.A);
5f66b61c 3884 /* NOTREACHED */
40a82448
DM
3885
3886#undef ST
3887#define ST st->u.branch
3888
3889 case BRANCHJ: /* /(...|A|...)/ with long next pointer */
c277df42
IZ
3890 next = scan + ARG(scan);
3891 if (next == scan)
3892 next = NULL;
40a82448
DM
3893 scan = NEXTOPER(scan);
3894 /* FALL THROUGH */
c277df42 3895
40a82448
DM
3896 case BRANCH: /* /(...|A|...)/ */
3897 scan = NEXTOPER(scan); /* scan now points to inner node */
3898 if (!next || (OP(next) != BRANCH && OP(next) != BRANCHJ))
3899 /* last branch; skip state push and jump direct to node */
3900 continue;
3901 ST.lastparen = *PL_reglastparen;
3902 ST.next_branch = next;
3903 REGCP_SET(ST.cp);
3904 PL_reginput = locinput;
02db2b7b 3905
40a82448
DM
3906 /* Now go into the branch */
3907 PUSH_STATE_GOTO(BRANCH_next, scan);
3908 /* NOTREACHED */
3909
3910 case BRANCH_next_fail: /* that branch failed; try the next, if any */
3911 REGCP_UNWIND(ST.cp);
3912 for (n = *PL_reglastparen; n > ST.lastparen; n--)
3913 PL_regendp[n] = -1;
3914 *PL_reglastparen = n;
3915 scan = ST.next_branch;
3916 /* no more branches? */
3917 if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ))
3918 sayNO;
3919 continue; /* execute next BRANCH[J] op */
3920 /* NOTREACHED */
3921
a0d0e21e 3922 case MINMOD:
24d3c4a9 3923 minmod = 1;
a0d0e21e 3924 break;
40a82448
DM
3925
3926#undef ST
3927#define ST st->u.curlym
3928
3929 case CURLYM: /* /A{m,n}B/ where A is fixed-length */
3930
3931 /* This is an optimisation of CURLYX that enables us to push
3932 * only a single backtracking state, no matter now many matches
3933 * there are in {m,n}. It relies on the pattern being constant
3934 * length, with no parens to influence future backrefs
3935 */
3936
3937 ST.me = scan;
dc45a647 3938 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
40a82448
DM
3939
3940 /* if paren positive, emulate an OPEN/CLOSE around A */
3941 if (ST.me->flags) {
3942 I32 paren = ST.me->flags;
3943 if (paren > PL_regsize)
3944 PL_regsize = paren;
3945 if (paren > (I32)*PL_reglastparen)
3946 *PL_reglastparen = paren;
c277df42 3947 scan += NEXT_OFF(scan); /* Skip former OPEN. */
6407bf3b 3948 }
40a82448
DM
3949 ST.A = scan;
3950 ST.B = next;
3951 ST.alen = 0;
3952 ST.count = 0;
24d3c4a9
DM
3953 ST.minmod = minmod;
3954 minmod = 0;
40a82448
DM
3955 ST.c1 = CHRTEST_UNINIT;
3956 REGCP_SET(ST.cp);
6407bf3b 3957
40a82448
DM
3958 if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */
3959 goto curlym_do_B;
3960
3961 curlym_do_A: /* execute the A in /A{m,n}B/ */
6407bf3b 3962 PL_reginput = locinput;
40a82448
DM
3963 PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */
3964 /* NOTREACHED */
5f80c4cf 3965
40a82448
DM
3966 case CURLYM_A: /* we've just matched an A */
3967 locinput = st->locinput;
3968 nextchr = UCHARAT(locinput);
3969
3970 ST.count++;
3971 /* after first match, determine A's length: u.curlym.alen */
3972 if (ST.count == 1) {
3973 if (PL_reg_match_utf8) {
3974 char *s = locinput;
3975 while (s < PL_reginput) {
3976 ST.alen++;
3977 s += UTF8SKIP(s);
3978 }
3979 }
3980 else {
3981 ST.alen = PL_reginput - locinput;
3982 }
3983 if (ST.alen == 0)
3984 ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me);
3985 }
0cadcf80
DM
3986 DEBUG_EXECUTE_r(
3987 PerlIO_printf(Perl_debug_log,
40a82448 3988 "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n",
5bc10b2c 3989 (int)(REPORT_CODE_OFF+(depth*2)), "",
40a82448 3990 (IV) ST.count, (IV)ST.alen)
0cadcf80
DM
3991 );
3992
40a82448
DM
3993 locinput = PL_reginput;
3994 if (ST.count < (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)))
3995 goto curlym_do_A; /* try to match another A */
3996 goto curlym_do_B; /* try to match B */
5f80c4cf 3997
40a82448
DM
3998 case CURLYM_A_fail: /* just failed to match an A */
3999 REGCP_UNWIND(ST.cp);
4000 if (ST.minmod || ST.count < ARG1(ST.me) /* min*/ )
4001 sayNO;
0cadcf80 4002
40a82448
DM
4003 curlym_do_B: /* execute the B in /A{m,n}B/ */
4004 PL_reginput = locinput;
4005 if (ST.c1 == CHRTEST_UNINIT) {
4006 /* calculate c1 and c2 for possible match of 1st char
4007 * following curly */
4008 ST.c1 = ST.c2 = CHRTEST_VOID;
4009 if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) {
4010 regnode *text_node = ST.B;
4011 if (! HAS_TEXT(text_node))
4012 FIND_NEXT_IMPT(text_node);
4013 if (HAS_TEXT(text_node)
4014 && PL_regkind[OP(text_node)] != REF)
4015 {
4016 ST.c1 = (U8)*STRING(text_node);
4017 ST.c2 =
4018 (OP(text_node) == EXACTF || OP(text_node) == REFF)
4019 ? PL_fold[ST.c1]
4020 : (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
4021 ? PL_fold_locale[ST.c1]
4022 : ST.c1;
c277df42 4023 }
c277df42 4024 }
40a82448
DM
4025 }
4026
4027 DEBUG_EXECUTE_r(
4028 PerlIO_printf(Perl_debug_log,
4029 "%*s CURLYM trying tail with matches=%"IVdf"...\n",
5bc10b2c 4030 (int)(REPORT_CODE_OFF+(depth*2)),
40a82448
DM
4031 "", (IV)ST.count)
4032 );
4033 if (ST.c1 != CHRTEST_VOID
4034 && UCHARAT(PL_reginput) != ST.c1
4035 && UCHARAT(PL_reginput) != ST.c2)
4036 {
4037 /* simulate B failing */
4038 state_num = CURLYM_B_fail;
4039 goto reenter_switch;
4040 }
4041
4042 if (ST.me->flags) {
4043 /* mark current A as captured */
4044 I32 paren = ST.me->flags;
4045 if (ST.count) {
4046 PL_regstartp[paren]
4047 = HOPc(PL_reginput, -ST.alen) - PL_bostr;
4048 PL_regendp[paren] = PL_reginput - PL_bostr;
c277df42 4049 }
40a82448
DM
4050 else
4051 PL_regendp[paren] = -1;
c277df42 4052 }
40a82448 4053 PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */
5f66b61c 4054 /* NOTREACHED */
40a82448
DM
4055
4056 case CURLYM_B_fail: /* just failed to match a B */
4057 REGCP_UNWIND(ST.cp);
4058 if (ST.minmod) {
4059 if (ST.count == ARG2(ST.me) /* max */)
4060 sayNO;
4061 goto curlym_do_A; /* try to match a further A */
4062 }
4063 /* backtrack one A */
4064 if (ST.count == ARG1(ST.me) /* min */)
4065 sayNO;
4066 ST.count--;
4067 locinput = HOPc(locinput, -ST.alen);
4068 goto curlym_do_B; /* try to match B */
4069
c255a977
DM
4070#undef ST
4071#define ST st->u.curly
40a82448 4072
c255a977
DM
4073#define CURLY_SETPAREN(paren, success) \
4074 if (paren) { \
4075 if (success) { \
4076 PL_regstartp[paren] = HOPc(locinput, -1) - PL_bostr; \
4077 PL_regendp[paren] = locinput - PL_bostr; \
4078 } \
4079 else \
4080 PL_regendp[paren] = -1; \
4081 }
4082
4083 case STAR: /* /A*B/ where A is width 1 */
4084 ST.paren = 0;
4085 ST.min = 0;
4086 ST.max = REG_INFTY;
a0d0e21e
LW
4087 scan = NEXTOPER(scan);
4088 goto repeat;
c255a977
DM
4089 case PLUS: /* /A+B/ where A is width 1 */
4090 ST.paren = 0;
4091 ST.min = 1;
4092 ST.max = REG_INFTY;
c277df42 4093 scan = NEXTOPER(scan);
c255a977
DM
4094 goto repeat;
4095 case CURLYN: /* /(A){m,n}B/ where A is width 1 */
4096 ST.paren = scan->flags; /* Which paren to set */
4097 if (ST.paren > PL_regsize)
4098 PL_regsize = ST.paren;
4099 if (ST.paren > (I32)*PL_reglastparen)
4100 *PL_reglastparen = ST.paren;
4101 ST.min = ARG1(scan); /* min to match */
4102 ST.max = ARG2(scan); /* max to match */
4103 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
4104 goto repeat;
4105 case CURLY: /* /A{m,n}B/ where A is width 1 */
4106 ST.paren = 0;
4107 ST.min = ARG1(scan); /* min to match */
4108 ST.max = ARG2(scan); /* max to match */
4109 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
c277df42 4110 repeat:
a0d0e21e
LW
4111 /*
4112 * Lookahead to avoid useless match attempts
4113 * when we know what character comes next.
c255a977 4114 *
5f80c4cf
JP
4115 * Used to only do .*x and .*?x, but now it allows
4116 * for )'s, ('s and (?{ ... })'s to be in the way
4117 * of the quantifier and the EXACT-like node. -- japhy
4118 */
4119
c255a977
DM
4120 if (ST.min > ST.max) /* XXX make this a compile-time check? */
4121 sayNO;
cca55fe3 4122 if (HAS_TEXT(next) || JUMPABLE(next)) {
5f80c4cf
JP
4123 U8 *s;
4124 regnode *text_node = next;
4125
3dab1dad
YO
4126 if (! HAS_TEXT(text_node))
4127 FIND_NEXT_IMPT(text_node);
5f80c4cf 4128
9e137952 4129 if (! HAS_TEXT(text_node))
c255a977 4130 ST.c1 = ST.c2 = CHRTEST_VOID;
5f80c4cf 4131 else {
3dab1dad 4132 if (PL_regkind[OP(text_node)] == REF) {
c255a977 4133 ST.c1 = ST.c2 = CHRTEST_VOID;
44a68960 4134 goto assume_ok_easy;
cca55fe3 4135 }
be8e71aa
YO
4136 else
4137 s = (U8*)STRING(text_node);
5f80c4cf
JP
4138
4139 if (!UTF) {
c255a977 4140 ST.c2 = ST.c1 = *s;
f65d3ee7 4141 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
c255a977 4142 ST.c2 = PL_fold[ST.c1];
f65d3ee7 4143 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
c255a977 4144 ST.c2 = PL_fold_locale[ST.c1];
1aa99e6b 4145 }
5f80c4cf 4146 else { /* UTF */
f65d3ee7 4147 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
a2a2844f 4148 STRLEN ulen1, ulen2;
89ebb4a3
JH
4149 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
4150 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
a2a2844f
JH
4151
4152 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
4153 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
e294cc5d
JH
4154#ifdef EBCDIC
4155 ST.c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN, 0,
4156 ckWARN(WARN_UTF8) ?
4157 0 : UTF8_ALLOW_ANY);
4158 ST.c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN, 0,
4159 ckWARN(WARN_UTF8) ?
4160 0 : UTF8_ALLOW_ANY);
4161#else
c255a977 4162 ST.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
e294cc5d 4163 uniflags);
c255a977 4164 ST.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
e294cc5d
JH
4165 uniflags);
4166#endif
5f80c4cf
JP
4167 }
4168 else {
c255a977 4169 ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
041457d9 4170 uniflags);
5f80c4cf 4171 }
1aa99e6b
IH
4172 }
4173 }
bbce6d69 4174 }
a0d0e21e 4175 else
c255a977 4176 ST.c1 = ST.c2 = CHRTEST_VOID;
cca55fe3 4177 assume_ok_easy:
c255a977
DM
4178
4179 ST.A = scan;
4180 ST.B = next;
3280af22 4181 PL_reginput = locinput;
24d3c4a9
DM
4182 if (minmod) {
4183 minmod = 0;
c255a977 4184 if (ST.min && regrepeat(rex, ST.A, ST.min) < ST.min)
4633a7c4 4185 sayNO;
c255a977 4186 ST.count = ST.min;
a0ed51b3 4187 locinput = PL_reginput;
c255a977
DM
4188 REGCP_SET(ST.cp);
4189 if (ST.c1 == CHRTEST_VOID)
4190 goto curly_try_B_min;
4191
4192 ST.oldloc = locinput;
4193
4194 /* set ST.maxpos to the furthest point along the
4195 * string that could possibly match */
4196 if (ST.max == REG_INFTY) {
4197 ST.maxpos = PL_regeol - 1;
4198 if (do_utf8)
4199 while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos))
4200 ST.maxpos--;
4201 }
4202 else if (do_utf8) {
4203 int m = ST.max - ST.min;
4204 for (ST.maxpos = locinput;
4205 m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--)
4206 ST.maxpos += UTF8SKIP(ST.maxpos);
4207 }
4208 else {
4209 ST.maxpos = locinput + ST.max - ST.min;
4210 if (ST.maxpos >= PL_regeol)
4211 ST.maxpos = PL_regeol - 1;
4212 }
4213 goto curly_try_B_min_known;
4214
4215 }
4216 else {
4217 ST.count = regrepeat(rex, ST.A, ST.max);
4218 locinput = PL_reginput;
4219 if (ST.count < ST.min)
4220 sayNO;
4221 if ((ST.count > ST.min)
4222 && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL))
4223 {
4224 /* A{m,n} must come at the end of the string, there's
4225 * no point in backing off ... */
4226 ST.min = ST.count;
4227 /* ...except that $ and \Z can match before *and* after
4228 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
4229 We may back off by one in this case. */
4230 if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS)
4231 ST.min--;
4232 }
4233 REGCP_SET(ST.cp);
4234 goto curly_try_B_max;
4235 }
4236 /* NOTREACHED */
4237
4238
4239 case CURLY_B_min_known_fail:
4240 /* failed to find B in a non-greedy match where c1,c2 valid */
4241 if (ST.paren && ST.count)
4242 PL_regendp[ST.paren] = -1;
4243
4244 PL_reginput = locinput; /* Could be reset... */
4245 REGCP_UNWIND(ST.cp);
4246 /* Couldn't or didn't -- move forward. */
4247 ST.oldloc = locinput;
4248 if (do_utf8)
4249 locinput += UTF8SKIP(locinput);
4250 else
4251 locinput++;
4252 ST.count++;
4253 curly_try_B_min_known:
4254 /* find the next place where 'B' could work, then call B */
4255 {
4256 int n;
4257 if (do_utf8) {
4258 n = (ST.oldloc == locinput) ? 0 : 1;
4259 if (ST.c1 == ST.c2) {
4260 STRLEN len;
4261 /* set n to utf8_distance(oldloc, locinput) */
4262 while (locinput <= ST.maxpos &&
4263 utf8n_to_uvchr((U8*)locinput,
4264 UTF8_MAXBYTES, &len,
4265 uniflags) != (UV)ST.c1) {
4266 locinput += len;
4267 n++;
4268 }
1aa99e6b
IH
4269 }
4270 else {
c255a977
DM
4271 /* set n to utf8_distance(oldloc, locinput) */
4272 while (locinput <= ST.maxpos) {
4273 STRLEN len;
4274 const UV c = utf8n_to_uvchr((U8*)locinput,
4275 UTF8_MAXBYTES, &len,
4276 uniflags);
4277 if (c == (UV)ST.c1 || c == (UV)ST.c2)
4278 break;
4279 locinput += len;
4280 n++;
1aa99e6b 4281 }
0fe9bf95
IZ
4282 }
4283 }
c255a977
DM
4284 else {
4285 if (ST.c1 == ST.c2) {
4286 while (locinput <= ST.maxpos &&
4287 UCHARAT(locinput) != ST.c1)
4288 locinput++;
bbce6d69 4289 }
c255a977
DM
4290 else {
4291 while (locinput <= ST.maxpos
4292 && UCHARAT(locinput) != ST.c1
4293 && UCHARAT(locinput) != ST.c2)
4294 locinput++;
a0ed51b3 4295 }
c255a977
DM
4296 n = locinput - ST.oldloc;
4297 }
4298 if (locinput > ST.maxpos)
4299 sayNO;
4300 /* PL_reginput == oldloc now */
4301 if (n) {
4302 ST.count += n;
4303 if (regrepeat(rex, ST.A, n) < n)
4633a7c4 4304 sayNO;
a0d0e21e 4305 }
c255a977
DM
4306 PL_reginput = locinput;
4307 CURLY_SETPAREN(ST.paren, ST.count);
4308 PUSH_STATE_GOTO(CURLY_B_min_known, ST.B);
a0d0e21e 4309 }
c255a977
DM
4310 /* NOTREACHED */
4311
4312
4313 case CURLY_B_min_fail:
4314 /* failed to find B in a non-greedy match where c1,c2 invalid */
4315 if (ST.paren && ST.count)
4316 PL_regendp[ST.paren] = -1;
4317
4318 REGCP_UNWIND(ST.cp);
4319 /* failed -- move forward one */
4320 PL_reginput = locinput;
4321 if (regrepeat(rex, ST.A, 1)) {
4322 ST.count++;
a0ed51b3 4323 locinput = PL_reginput;
c255a977
DM
4324 if (ST.count <= ST.max || (ST.max == REG_INFTY &&
4325 ST.count > 0)) /* count overflow ? */
15272685 4326 {
c255a977
DM
4327 curly_try_B_min:
4328 CURLY_SETPAREN(ST.paren, ST.count);
4329 PUSH_STATE_GOTO(CURLY_B_min, ST.B);
a0d0e21e
LW
4330 }
4331 }
4633a7c4 4332 sayNO;
c255a977
DM
4333 /* NOTREACHED */
4334
4335
4336 curly_try_B_max:
4337 /* a successful greedy match: now try to match B */
4338 {
4339 UV c = 0;
4340 if (ST.c1 != CHRTEST_VOID)
4341 c = do_utf8 ? utf8n_to_uvchr((U8*)PL_reginput,
4342 UTF8_MAXBYTES, 0, uniflags)
466787eb 4343 : (UV) UCHARAT(PL_reginput);
c255a977
DM
4344 /* If it could work, try it. */
4345 if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) {
4346 CURLY_SETPAREN(ST.paren, ST.count);
4347 PUSH_STATE_GOTO(CURLY_B_max, ST.B);
4348 /* NOTREACHED */
4349 }
4350 }
4351 /* FALL THROUGH */
4352 case CURLY_B_max_fail:
4353 /* failed to find B in a greedy match */
4354 if (ST.paren && ST.count)
4355 PL_regendp[ST.paren] = -1;
4356
4357 REGCP_UNWIND(ST.cp);
4358 /* back up. */
4359 if (--ST.count < ST.min)
4360 sayNO;
4361 PL_reginput = locinput = HOPc(locinput, -1);
4362 goto curly_try_B_max;
4363
4364#undef ST
4365
4366
a0d0e21e 4367 case END:
6bda09f9 4368 fake_end:
faec1544
DM
4369 if (cur_eval) {
4370 /* we've just finished A in /(??{A})B/; now continue with B */
4371 I32 tmpix;
4372
4373
4374 st->u.eval.toggle_reg_flags
4375 = cur_eval->u.eval.toggle_reg_flags;
4376 PL_reg_flags ^= st->u.eval.toggle_reg_flags;
4377
4378 st->u.eval.prev_rex = rex; /* inner */
4379 rex = cur_eval->u.eval.prev_rex; /* outer */
4380 cur_curlyx = cur_eval->u.eval.prev_curlyx;
4381 ReREFCNT_inc(rex);
4382 st->u.eval.cp = regcppush(0); /* Save *all* the positions. */
4383 REGCP_SET(st->u.eval.lastcp);
4384 PL_reginput = locinput;
4385
4386 /* Restore parens of the outer rex without popping the
4387 * savestack */
4388 tmpix = PL_savestack_ix;
4389 PL_savestack_ix = cur_eval->u.eval.lastcp;
4390 regcppop(rex);
4391 PL_savestack_ix = tmpix;
4392
4393 st->u.eval.prev_eval = cur_eval;
4394 cur_eval = cur_eval->u.eval.prev_eval;
4395 DEBUG_EXECUTE_r(
6bda09f9
YO
4396 PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %x\n",
4397 REPORT_CODE_OFF+depth*2, "",(int)cur_eval););
faec1544
DM
4398 PUSH_YES_STATE_GOTO(EVAL_AB,
4399 st->u.eval.prev_eval->u.eval.B); /* match B */
4400 }
4401
3b0527fe 4402 if (locinput < reginfo->till) {
a3621e74 4403 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
7821416a
IZ
4404 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
4405 PL_colors[4],
4406 (long)(locinput - PL_reg_starttry),
3b0527fe 4407 (long)(reginfo->till - PL_reg_starttry),
7821416a 4408 PL_colors[5]));
262b90c4 4409 sayNO_SILENT; /* Cannot match: too short. */
7821416a
IZ
4410 }
4411 PL_reginput = locinput; /* put where regtry can find it */
262b90c4 4412 sayYES; /* Success! */
dad79028
DM
4413
4414 case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */
4415 DEBUG_EXECUTE_r(
4416 PerlIO_printf(Perl_debug_log,
4417 "%*s %ssubpattern success...%s\n",
5bc10b2c 4418 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]));
3280af22 4419 PL_reginput = locinput; /* put where regtry can find it */
262b90c4 4420 sayYES; /* Success! */
dad79028 4421
40a82448
DM
4422#undef ST
4423#define ST st->u.ifmatch
4424
4425 case SUSPEND: /* (?>A) */
4426 ST.wanted = 1;
9fe1d20c 4427 PL_reginput = locinput;
9041c2e3 4428 goto do_ifmatch;
dad79028 4429
40a82448
DM
4430 case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */
4431 ST.wanted = 0;
dad79028
DM
4432 goto ifmatch_trivial_fail_test;
4433
40a82448
DM
4434 case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */
4435 ST.wanted = 1;
dad79028 4436 ifmatch_trivial_fail_test:
a0ed51b3 4437 if (scan->flags) {
52657f30 4438 char * const s = HOPBACKc(locinput, scan->flags);
dad79028
DM
4439 if (!s) {
4440 /* trivial fail */
24d3c4a9
DM
4441 if (logical) {
4442 logical = 0;
4443 sw = 1 - (bool)ST.wanted;
dad79028 4444 }
40a82448 4445 else if (ST.wanted)
dad79028
DM
4446 sayNO;
4447 next = scan + ARG(scan);
4448 if (next == scan)
4449 next = NULL;
4450 break;
4451 }
efb30f32 4452 PL_reginput = s;
a0ed51b3
LW
4453 }
4454 else
4455 PL_reginput = locinput;
4456
c277df42 4457 do_ifmatch:
40a82448 4458 ST.me = scan;
24d3c4a9 4459 ST.logical = logical;
40a82448
DM
4460 /* execute body of (?...A) */
4461 PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan)));
4462 /* NOTREACHED */
4463
4464 case IFMATCH_A_fail: /* body of (?...A) failed */
4465 ST.wanted = !ST.wanted;
4466 /* FALL THROUGH */
4467
4468 case IFMATCH_A: /* body of (?...A) succeeded */
24d3c4a9
DM
4469 if (ST.logical) {
4470 sw = (bool)ST.wanted;
40a82448
DM
4471 }
4472 else if (!ST.wanted)
4473 sayNO;
4474
4475 if (OP(ST.me) == SUSPEND)
4476 locinput = PL_reginput;
4477 else {
4478 locinput = PL_reginput = st->locinput;
4479 nextchr = UCHARAT(locinput);
4480 }
4481 scan = ST.me + ARG(ST.me);
4482 if (scan == ST.me)
4483 scan = NULL;
4484 continue; /* execute B */
4485
4486#undef ST
dad79028 4487
c277df42 4488 case LONGJMP:
c277df42
IZ
4489 next = scan + ARG(scan);
4490 if (next == scan)
4491 next = NULL;
a0d0e21e
LW
4492 break;
4493 default:
b900a521 4494 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
d7d93a81 4495 PTR2UV(scan), OP(scan));
cea2e8a9 4496 Perl_croak(aTHX_ "regexp memory corruption");
a687059c 4497 }
95b24440 4498
a0d0e21e 4499 scan = next;
95b24440
DM
4500 continue;
4501 /* NOTREACHED */
4502
40a82448
DM
4503 push_yes_state:
4504 /* push a state that backtracks on success */
4505 st->u.yes.prev_yes_state = yes_state;
4506 yes_state = st;
4507 /* FALL THROUGH */
4508 push_state:
4509 /* push a new regex state, then continue at scan */
4510 {
4511 regmatch_state *newst;
4512
5bc10b2c 4513 DEBUG_STATE_pp("push");
40a82448 4514 depth++;
40a82448
DM
4515 st->locinput = locinput;
4516 newst = st+1;
4517 if (newst > SLAB_LAST(PL_regmatch_slab))
4518 newst = S_push_slab(aTHX);
4519 PL_regmatch_state = newst;
786e8c11 4520
40a82448
DM
4521 locinput = PL_reginput;
4522 nextchr = UCHARAT(locinput);
4523 st = newst;
4524 continue;
4525 /* NOTREACHED */
4526 }
a0d0e21e 4527 }
a687059c 4528
a0d0e21e
LW
4529 /*
4530 * We get here only if there's trouble -- normally "case END" is
4531 * the terminating point.
4532 */
cea2e8a9 4533 Perl_croak(aTHX_ "corrupted regexp pointers");
a0d0e21e 4534 /*NOTREACHED*/
4633a7c4
LW
4535 sayNO;
4536
262b90c4 4537yes:
77cb431f
DM
4538 if (yes_state) {
4539 /* we have successfully completed a subexpression, but we must now
4540 * pop to the state marked by yes_state and continue from there */
77cb431f 4541 assert(st != yes_state);
5bc10b2c
DM
4542#ifdef DEBUGGING
4543 while (st != yes_state) {
4544 st--;
4545 if (st < SLAB_FIRST(PL_regmatch_slab)) {
4546 PL_regmatch_slab = PL_regmatch_slab->prev;
4547 st = SLAB_LAST(PL_regmatch_slab);
4548 }
4549 DEBUG_STATE_pp("pop (yes)");
4550 depth--;
4551 }
4552#else
77cb431f
DM
4553 while (yes_state < SLAB_FIRST(PL_regmatch_slab)
4554 || yes_state > SLAB_LAST(PL_regmatch_slab))
4555 {
4556 /* not in this slab, pop slab */
4557 depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1);
4558 PL_regmatch_slab = PL_regmatch_slab->prev;
4559 st = SLAB_LAST(PL_regmatch_slab);
4560 }
4561 depth -= (st - yes_state);
5bc10b2c 4562#endif
77cb431f
DM
4563 st = yes_state;
4564 yes_state = st->u.yes.prev_yes_state;
4565 PL_regmatch_state = st;
4566
24d3c4a9
DM
4567 state_num = st->resume_state;
4568 goto reenter_switch;
77cb431f
DM
4569 }
4570
a3621e74 4571 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
e4584336 4572 PL_colors[4], PL_colors[5]));
02db2b7b 4573
95b24440 4574 result = 1;
aa283a38 4575 goto final_exit;
4633a7c4
LW
4576
4577no:
a3621e74 4578 DEBUG_EXECUTE_r(
7821416a 4579 PerlIO_printf(Perl_debug_log,
786e8c11 4580 "%*s %sfailed...%s\n",
5bc10b2c 4581 REPORT_CODE_OFF+depth*2, "",
786e8c11 4582 PL_colors[4], PL_colors[5])
7821416a 4583 );
aa283a38 4584
262b90c4 4585no_silent:
aa283a38
DM
4586 if (depth) {
4587 /* there's a previous state to backtrack to */
40a82448
DM
4588 st--;
4589 if (st < SLAB_FIRST(PL_regmatch_slab)) {
4590 PL_regmatch_slab = PL_regmatch_slab->prev;
4591 st = SLAB_LAST(PL_regmatch_slab);
4592 }
4593 PL_regmatch_state = st;
40a82448
DM
4594 locinput= st->locinput;
4595 nextchr = UCHARAT(locinput);
4596
5bc10b2c
DM
4597 DEBUG_STATE_pp("pop");
4598 depth--;
262b90c4
DM
4599 if (yes_state == st)
4600 yes_state = st->u.yes.prev_yes_state;
5bc10b2c 4601
24d3c4a9
DM
4602 state_num = st->resume_state + 1; /* failure = success + 1 */
4603 goto reenter_switch;
95b24440 4604 }
24d3c4a9 4605 result = 0;
aa283a38 4606
262b90c4 4607 final_exit:
aa283a38 4608
5d9a96ca
DM
4609 /* restore original high-water mark */
4610 PL_regmatch_slab = orig_slab;
4611 PL_regmatch_state = orig_state;
4612
4613 /* free all slabs above current one */
4614 if (orig_slab->next) {
c4fd8992 4615 regmatch_slab *sl = orig_slab->next;
5d9a96ca
DM
4616 orig_slab->next = NULL;
4617 while (sl) {
c4fd8992 4618 regmatch_slab * const osl = sl;
5d9a96ca 4619 sl = sl->next;
ad65c075 4620 Safefree(osl);
5d9a96ca
DM
4621 }
4622 }
4623
95b24440 4624 return result;
a687059c
LW
4625}
4626
4627/*
4628 - regrepeat - repeatedly match something simple, report how many
4629 */
4630/*
4631 * [This routine now assumes that it will only match on things of length 1.
4632 * That was true before, but now we assume scan - reginput is the count,
a0ed51b3 4633 * rather than incrementing count on every character. [Er, except utf8.]]
a687059c 4634 */
76e3520e 4635STATIC I32
32fc9b6a 4636S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max)
a687059c 4637{
27da23d5 4638 dVAR;
a0d0e21e 4639 register char *scan;
a0d0e21e 4640 register I32 c;
3280af22 4641 register char *loceol = PL_regeol;
a0ed51b3 4642 register I32 hardcount = 0;
53c4c00c 4643 register bool do_utf8 = PL_reg_match_utf8;
a0d0e21e 4644
3280af22 4645 scan = PL_reginput;
faf11cac
HS
4646 if (max == REG_INFTY)
4647 max = I32_MAX;
4648 else if (max < loceol - scan)
7f596f4c 4649 loceol = scan + max;
a0d0e21e 4650 switch (OP(p)) {
22c35a8c 4651 case REG_ANY:
1aa99e6b 4652 if (do_utf8) {
ffc61ed2 4653 loceol = PL_regeol;
1aa99e6b 4654 while (scan < loceol && hardcount < max && *scan != '\n') {
ffc61ed2
JH
4655 scan += UTF8SKIP(scan);
4656 hardcount++;
4657 }
4658 } else {
4659 while (scan < loceol && *scan != '\n')
4660 scan++;
a0ed51b3
LW
4661 }
4662 break;
ffc61ed2 4663 case SANY:
def8e4ea
JH
4664 if (do_utf8) {
4665 loceol = PL_regeol;
a0804c9e 4666 while (scan < loceol && hardcount < max) {
def8e4ea
JH
4667 scan += UTF8SKIP(scan);
4668 hardcount++;
4669 }
4670 }
4671 else
4672 scan = loceol;
a0ed51b3 4673 break;
f33976b4
DB
4674 case CANY:
4675 scan = loceol;
4676 break;
090f7165
JH
4677 case EXACT: /* length of string is 1 */
4678 c = (U8)*STRING(p);
4679 while (scan < loceol && UCHARAT(scan) == c)
4680 scan++;
bbce6d69 4681 break;
4682 case EXACTF: /* length of string is 1 */
cd439c50 4683 c = (U8)*STRING(p);
bbce6d69 4684 while (scan < loceol &&
22c35a8c 4685 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
bbce6d69 4686 scan++;
4687 break;
4688 case EXACTFL: /* length of string is 1 */
3280af22 4689 PL_reg_flags |= RF_tainted;
cd439c50 4690 c = (U8)*STRING(p);
bbce6d69 4691 while (scan < loceol &&
22c35a8c 4692 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
a0d0e21e
LW
4693 scan++;
4694 break;
4695 case ANYOF:
ffc61ed2
JH
4696 if (do_utf8) {
4697 loceol = PL_regeol;
cfc92286 4698 while (hardcount < max && scan < loceol &&
32fc9b6a 4699 reginclass(prog, p, (U8*)scan, 0, do_utf8)) {
ffc61ed2
JH
4700 scan += UTF8SKIP(scan);
4701 hardcount++;
4702 }
4703 } else {
32fc9b6a 4704 while (scan < loceol && REGINCLASS(prog, p, (U8*)scan))
ffc61ed2
JH
4705 scan++;
4706 }
a0d0e21e
LW
4707 break;
4708 case ALNUM:
1aa99e6b 4709 if (do_utf8) {
ffc61ed2 4710 loceol = PL_regeol;
1a4fad37 4711 LOAD_UTF8_CHARCLASS_ALNUM();
1aa99e6b 4712 while (hardcount < max && scan < loceol &&
3568d838 4713 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
ffc61ed2
JH
4714 scan += UTF8SKIP(scan);
4715 hardcount++;
4716 }
4717 } else {
4718 while (scan < loceol && isALNUM(*scan))
4719 scan++;
a0ed51b3
LW
4720 }
4721 break;
bbce6d69 4722 case ALNUML:
3280af22 4723 PL_reg_flags |= RF_tainted;
1aa99e6b 4724 if (do_utf8) {
ffc61ed2 4725 loceol = PL_regeol;
1aa99e6b
IH
4726 while (hardcount < max && scan < loceol &&
4727 isALNUM_LC_utf8((U8*)scan)) {
ffc61ed2
JH
4728 scan += UTF8SKIP(scan);
4729 hardcount++;
4730 }
4731 } else {
4732 while (scan < loceol && isALNUM_LC(*scan))
4733 scan++;
a0ed51b3
LW
4734 }
4735 break;
a0d0e21e 4736 case NALNUM:
1aa99e6b 4737 if (do_utf8) {
ffc61ed2 4738 loceol = PL_regeol;
1a4fad37 4739 LOAD_UTF8_CHARCLASS_ALNUM();
1aa99e6b 4740 while (hardcount < max && scan < loceol &&
3568d838 4741 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
ffc61ed2
JH
4742 scan += UTF8SKIP(scan);
4743 hardcount++;
4744 }
4745 } else {
4746 while (scan < loceol && !isALNUM(*scan))
4747 scan++;
a0ed51b3
LW
4748 }
4749 break;
bbce6d69 4750 case NALNUML:
3280af22 4751 PL_reg_flags |= RF_tainted;
1aa99e6b 4752 if (do_utf8) {
ffc61ed2 4753 loceol = PL_regeol;
1aa99e6b
IH
4754 while (hardcount < max && scan < loceol &&
4755 !isALNUM_LC_utf8((U8*)scan)) {
ffc61ed2
JH
4756 scan += UTF8SKIP(scan);
4757 hardcount++;
4758 }
4759 } else {
4760 while (scan < loceol && !isALNUM_LC(*scan))
4761 scan++;
a0ed51b3
LW
4762 }
4763 break;
a0d0e21e 4764 case SPACE:
1aa99e6b 4765 if (do_utf8) {
ffc61ed2 4766 loceol = PL_regeol;
1a4fad37 4767 LOAD_UTF8_CHARCLASS_SPACE();
1aa99e6b 4768 while (hardcount < max && scan < loceol &&
3568d838
JH
4769 (*scan == ' ' ||
4770 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
ffc61ed2
JH
4771 scan += UTF8SKIP(scan);
4772 hardcount++;
4773 }
4774 } else {
4775 while (scan < loceol && isSPACE(*scan))
4776 scan++;
a0ed51b3
LW
4777 }
4778 break;
bbce6d69 4779 case SPACEL:
3280af22 4780 PL_reg_flags |= RF_tainted;
1aa99e6b 4781 if (do_utf8) {
ffc61ed2 4782 loceol = PL_regeol;
1aa99e6b 4783 while (hardcount < max && scan < loceol &&
ffc61ed2
JH
4784 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4785 scan += UTF8SKIP(scan);
4786 hardcount++;
4787 }
4788 } else {
4789 while (scan < loceol && isSPACE_LC(*scan))
4790 scan++;
a0ed51b3
LW
4791 }
4792 break;
a0d0e21e 4793 case NSPACE:
1aa99e6b 4794 if (do_utf8) {
ffc61ed2 4795 loceol = PL_regeol;
1a4fad37 4796 LOAD_UTF8_CHARCLASS_SPACE();
1aa99e6b 4797 while (hardcount < max && scan < loceol &&
3568d838
JH
4798 !(*scan == ' ' ||
4799 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
ffc61ed2
JH
4800 scan += UTF8SKIP(scan);
4801 hardcount++;
4802 }
4803 } else {
4804 while (scan < loceol && !isSPACE(*scan))
4805 scan++;
4806 break;
a0ed51b3 4807 }
bbce6d69 4808 case NSPACEL:
3280af22 4809 PL_reg_flags |= RF_tainted;
1aa99e6b 4810 if (do_utf8) {
ffc61ed2 4811 loceol = PL_regeol;
1aa99e6b 4812 while (hardcount < max && scan < loceol &&
ffc61ed2
JH
4813 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4814 scan += UTF8SKIP(scan);
4815 hardcount++;
4816 }
4817 } else {
4818 while (scan < loceol && !isSPACE_LC(*scan))
4819 scan++;
a0ed51b3
LW
4820 }
4821 break;
a0d0e21e 4822 case DIGIT:
1aa99e6b 4823 if (do_utf8) {
ffc61ed2 4824 loceol = PL_regeol;
1a4fad37 4825 LOAD_UTF8_CHARCLASS_DIGIT();
1aa99e6b 4826 while (hardcount < max && scan < loceol &&
3568d838 4827 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
ffc61ed2
JH
4828 scan += UTF8SKIP(scan);
4829 hardcount++;
4830 }
4831 } else {
4832 while (scan < loceol && isDIGIT(*scan))
4833 scan++;
a0ed51b3
LW
4834 }
4835 break;
a0d0e21e 4836 case NDIGIT:
1aa99e6b 4837 if (do_utf8) {
ffc61ed2 4838 loceol = PL_regeol;
1a4fad37 4839 LOAD_UTF8_CHARCLASS_DIGIT();
1aa99e6b 4840 while (hardcount < max && scan < loceol &&
3568d838 4841 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
ffc61ed2
JH
4842 scan += UTF8SKIP(scan);
4843 hardcount++;
4844 }
4845 } else {
4846 while (scan < loceol && !isDIGIT(*scan))
4847 scan++;
a0ed51b3
LW
4848 }
4849 break;
a0d0e21e
LW
4850 default: /* Called on something of 0 width. */
4851 break; /* So match right here or not at all. */
4852 }
a687059c 4853
a0ed51b3
LW
4854 if (hardcount)
4855 c = hardcount;
4856 else
4857 c = scan - PL_reginput;
3280af22 4858 PL_reginput = scan;
a687059c 4859
a3621e74 4860 DEBUG_r({
e68ec53f 4861 GET_RE_DEBUG_FLAGS_DECL;
be8e71aa 4862 DEBUG_EXECUTE_r({
e68ec53f
YO
4863 SV * const prop = sv_newmortal();
4864 regprop(prog, prop, p);
4865 PerlIO_printf(Perl_debug_log,
be8e71aa
YO
4866 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4867 REPORT_CODE_OFF+1, "", SvPVX_const(prop),(IV)c,(IV)max);
a3621e74 4868 });
be8e71aa 4869 });
9041c2e3 4870
a0d0e21e 4871 return(c);
a687059c
LW
4872}
4873
c277df42 4874
be8e71aa 4875#if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION)
c277df42 4876/*
ffc61ed2
JH
4877- regclass_swash - prepare the utf8 swash
4878*/
4879
4880SV *
32fc9b6a 4881Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
ffc61ed2 4882{
97aff369 4883 dVAR;
9e55ce06
JH
4884 SV *sw = NULL;
4885 SV *si = NULL;
4886 SV *alt = NULL;
3dab1dad 4887 const struct reg_data * const data = prog ? prog->data : NULL;
ffc61ed2 4888
4f639d21 4889 if (data && data->count) {
a3b680e6 4890 const U32 n = ARG(node);
ffc61ed2 4891
4f639d21
DM
4892 if (data->what[n] == 's') {
4893 SV * const rv = (SV*)data->data[n];
890ce7af 4894 AV * const av = (AV*)SvRV((SV*)rv);
2d03de9c 4895 SV **const ary = AvARRAY(av);
9e55ce06 4896 SV **a, **b;
9041c2e3 4897
711a919c 4898 /* See the end of regcomp.c:S_regclass() for
9e55ce06
JH
4899 * documentation of these array elements. */
4900
b11f357e 4901 si = *ary;
8f7f7219 4902 a = SvROK(ary[1]) ? &ary[1] : 0;
b11f357e
JH
4903 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4904
ffc61ed2
JH
4905 if (a)
4906 sw = *a;
4907 else if (si && doinit) {
4908 sw = swash_init("utf8", "", si, 1, 0);
4909 (void)av_store(av, 1, sw);
4910 }
9e55ce06
JH
4911 if (b)
4912 alt = *b;
ffc61ed2
JH
4913 }
4914 }
4915
9e55ce06
JH
4916 if (listsvp)
4917 *listsvp = si;
4918 if (altsvp)
4919 *altsvp = alt;
ffc61ed2
JH
4920
4921 return sw;
4922}
76234dfb 4923#endif
ffc61ed2
JH
4924
4925/*
ba7b4546 4926 - reginclass - determine if a character falls into a character class
832705d4
JH
4927
4928 The n is the ANYOF regnode, the p is the target string, lenp
4929 is pointer to the maximum length of how far to go in the p
4930 (if the lenp is zero, UTF8SKIP(p) is used),
4931 do_utf8 tells whether the target string is in UTF-8.
4932
bbce6d69 4933 */
4934
76e3520e 4935STATIC bool
32fc9b6a 4936S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8)
bbce6d69 4937{
27da23d5 4938 dVAR;
a3b680e6 4939 const char flags = ANYOF_FLAGS(n);
bbce6d69 4940 bool match = FALSE;
cc07378b 4941 UV c = *p;
ae9ddab8 4942 STRLEN len = 0;
9e55ce06 4943 STRLEN plen;
1aa99e6b 4944
19f67299
TS
4945 if (do_utf8 && !UTF8_IS_INVARIANT(c)) {
4946 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
4ad0818d
DM
4947 (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) | UTF8_CHECK_ONLY);
4948 /* see [perl #37836] for UTF8_ALLOW_ANYUV */
e8a70c6f
SP
4949 if (len == (STRLEN)-1)
4950 Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)");
19f67299 4951 }
bbce6d69 4952
0f0076b4 4953 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
ffc61ed2 4954 if (do_utf8 || (flags & ANYOF_UNICODE)) {
9e55ce06
JH
4955 if (lenp)
4956 *lenp = 0;
ffc61ed2 4957 if (do_utf8 && !ANYOF_RUNTIME(n)) {
ffc61ed2
JH
4958 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4959 match = TRUE;
bbce6d69 4960 }
3568d838 4961 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
1aa99e6b 4962 match = TRUE;
ffc61ed2 4963 if (!match) {
9e55ce06 4964 AV *av;
32fc9b6a 4965 SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av);
ffc61ed2
JH
4966
4967 if (sw) {
3568d838 4968 if (swash_fetch(sw, p, do_utf8))
ffc61ed2
JH
4969 match = TRUE;
4970 else if (flags & ANYOF_FOLD) {
9e55ce06
JH
4971 if (!match && lenp && av) {
4972 I32 i;
9e55ce06 4973 for (i = 0; i <= av_len(av); i++) {
890ce7af 4974 SV* const sv = *av_fetch(av, i, FALSE);
9e55ce06 4975 STRLEN len;
890ce7af 4976 const char * const s = SvPV_const(sv, len);
9e55ce06 4977
061b10df 4978 if (len <= plen && memEQ(s, (char*)p, len)) {
9e55ce06
JH
4979 *lenp = len;
4980 match = TRUE;
4981 break;
4982 }
4983 }
4984 }
4985 if (!match) {
89ebb4a3 4986 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
4a623e43
JH
4987 STRLEN tmplen;
4988
9e55ce06
JH
4989 to_utf8_fold(p, tmpbuf, &tmplen);
4990 if (swash_fetch(sw, tmpbuf, do_utf8))
4991 match = TRUE;
4992 }
ffc61ed2
JH
4993 }
4994 }
bbce6d69 4995 }
9e55ce06 4996 if (match && lenp && *lenp == 0)
0f0076b4 4997 *lenp = UNISKIP(NATIVE_TO_UNI(c));
bbce6d69 4998 }
1aa99e6b 4999 if (!match && c < 256) {
ffc61ed2
JH
5000 if (ANYOF_BITMAP_TEST(n, c))
5001 match = TRUE;
5002 else if (flags & ANYOF_FOLD) {
eb160463 5003 U8 f;
a0ed51b3 5004
ffc61ed2
JH
5005 if (flags & ANYOF_LOCALE) {
5006 PL_reg_flags |= RF_tainted;
5007 f = PL_fold_locale[c];
5008 }
5009 else
5010 f = PL_fold[c];
5011 if (f != c && ANYOF_BITMAP_TEST(n, f))
5012 match = TRUE;
5013 }
5014
5015 if (!match && (flags & ANYOF_CLASS)) {
a0ed51b3 5016 PL_reg_flags |= RF_tainted;
ffc61ed2
JH
5017 if (
5018 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
5019 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
5020 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
5021 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
5022 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
5023 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
5024 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
5025 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
5026 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
5027 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
5028 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
5029 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
5030 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
5031 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
5032 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
5033 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
5034 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
5035 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
5036 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
5037 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
5038 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
5039 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
5040 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
5041 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
5042 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
5043 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
5044 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
5045 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
5046 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
5047 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
5048 ) /* How's that for a conditional? */
5049 {
5050 match = TRUE;
5051 }
a0ed51b3 5052 }
a0ed51b3
LW
5053 }
5054
a0ed51b3
LW
5055 return (flags & ANYOF_INVERT) ? !match : match;
5056}
161b471a 5057
dfe13c55 5058STATIC U8 *
0ce71af7 5059S_reghop3(U8 *s, I32 off, const U8* lim)
9041c2e3 5060{
97aff369 5061 dVAR;
a0ed51b3 5062 if (off >= 0) {
1aa99e6b 5063 while (off-- && s < lim) {
ffc61ed2 5064 /* XXX could check well-formedness here */
a0ed51b3 5065 s += UTF8SKIP(s);
ffc61ed2 5066 }
a0ed51b3
LW
5067 }
5068 else {
1de06328
YO
5069 while (off++ && s > lim) {
5070 s--;
5071 if (UTF8_IS_CONTINUED(*s)) {
5072 while (s > lim && UTF8_IS_CONTINUATION(*s))
5073 s--;
a0ed51b3 5074 }
1de06328 5075 /* XXX could check well-formedness here */
a0ed51b3
LW
5076 }
5077 }
5078 return s;
5079}
161b471a 5080
f9f4320a
YO
5081#ifdef XXX_dmq
5082/* there are a bunch of places where we use two reghop3's that should
5083 be replaced with this routine. but since thats not done yet
5084 we ifdef it out - dmq
5085*/
dfe13c55 5086STATIC U8 *
1de06328
YO
5087S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim)
5088{
5089 dVAR;
5090 if (off >= 0) {
5091 while (off-- && s < rlim) {
5092 /* XXX could check well-formedness here */
5093 s += UTF8SKIP(s);
5094 }
5095 }
5096 else {
5097 while (off++ && s > llim) {
5098 s--;
5099 if (UTF8_IS_CONTINUED(*s)) {
5100 while (s > llim && UTF8_IS_CONTINUATION(*s))
5101 s--;
5102 }
5103 /* XXX could check well-formedness here */
5104 }
5105 }
5106 return s;
5107}
f9f4320a 5108#endif
1de06328
YO
5109
5110STATIC U8 *
0ce71af7 5111S_reghopmaybe3(U8* s, I32 off, const U8* lim)
a0ed51b3 5112{
97aff369 5113 dVAR;
a0ed51b3 5114 if (off >= 0) {
1aa99e6b 5115 while (off-- && s < lim) {
ffc61ed2 5116 /* XXX could check well-formedness here */
a0ed51b3 5117 s += UTF8SKIP(s);
ffc61ed2 5118 }
a0ed51b3 5119 if (off >= 0)
3dab1dad 5120 return NULL;
a0ed51b3
LW
5121 }
5122 else {
1de06328
YO
5123 while (off++ && s > lim) {
5124 s--;
5125 if (UTF8_IS_CONTINUED(*s)) {
5126 while (s > lim && UTF8_IS_CONTINUATION(*s))
5127 s--;
a0ed51b3 5128 }
1de06328 5129 /* XXX could check well-formedness here */
a0ed51b3
LW
5130 }
5131 if (off <= 0)
3dab1dad 5132 return NULL;
a0ed51b3
LW
5133 }
5134 return s;
5135}
51371543 5136
51371543 5137static void
acfe0abc 5138restore_pos(pTHX_ void *arg)
51371543 5139{
97aff369 5140 dVAR;
097eb12c 5141 regexp * const rex = (regexp *)arg;
51371543
GS
5142 if (PL_reg_eval_set) {
5143 if (PL_reg_oldsaved) {
4f639d21
DM
5144 rex->subbeg = PL_reg_oldsaved;
5145 rex->sublen = PL_reg_oldsavedlen;
f8c7b90f 5146#ifdef PERL_OLD_COPY_ON_WRITE
4f639d21 5147 rex->saved_copy = PL_nrs;
ed252734 5148#endif
4f639d21 5149 RX_MATCH_COPIED_on(rex);
51371543
GS
5150 }
5151 PL_reg_magic->mg_len = PL_reg_oldpos;
5152 PL_reg_eval_set = 0;
5153 PL_curpm = PL_reg_oldcurpm;
5154 }
5155}
33b8afdf
JH
5156
5157STATIC void
5158S_to_utf8_substr(pTHX_ register regexp *prog)
5159{
33b8afdf 5160 if (prog->float_substr && !prog->float_utf8) {
097eb12c
AL
5161 SV* const sv = newSVsv(prog->float_substr);
5162 prog->float_utf8 = sv;
33b8afdf
JH
5163 sv_utf8_upgrade(sv);
5164 if (SvTAIL(prog->float_substr))
5165 SvTAIL_on(sv);
5166 if (prog->float_substr == prog->check_substr)
5167 prog->check_utf8 = sv;
5168 }
5169 if (prog->anchored_substr && !prog->anchored_utf8) {
097eb12c
AL
5170 SV* const sv = newSVsv(prog->anchored_substr);
5171 prog->anchored_utf8 = sv;
33b8afdf
JH
5172 sv_utf8_upgrade(sv);
5173 if (SvTAIL(prog->anchored_substr))
5174 SvTAIL_on(sv);
5175 if (prog->anchored_substr == prog->check_substr)
5176 prog->check_utf8 = sv;
5177 }
5178}
5179
5180STATIC void
5181S_to_byte_substr(pTHX_ register regexp *prog)
5182{
97aff369 5183 dVAR;
33b8afdf 5184 if (prog->float_utf8 && !prog->float_substr) {
097eb12c
AL
5185 SV* sv = newSVsv(prog->float_utf8);
5186 prog->float_substr = sv;
33b8afdf
JH
5187 if (sv_utf8_downgrade(sv, TRUE)) {
5188 if (SvTAIL(prog->float_utf8))
5189 SvTAIL_on(sv);
5190 } else {
5191 SvREFCNT_dec(sv);
5192 prog->float_substr = sv = &PL_sv_undef;
5193 }
5194 if (prog->float_utf8 == prog->check_utf8)
5195 prog->check_substr = sv;
5196 }
5197 if (prog->anchored_utf8 && !prog->anchored_substr) {
097eb12c
AL
5198 SV* sv = newSVsv(prog->anchored_utf8);
5199 prog->anchored_substr = sv;
33b8afdf
JH
5200 if (sv_utf8_downgrade(sv, TRUE)) {
5201 if (SvTAIL(prog->anchored_utf8))
5202 SvTAIL_on(sv);
5203 } else {
5204 SvREFCNT_dec(sv);
5205 prog->anchored_substr = sv = &PL_sv_undef;
5206 }
5207 if (prog->anchored_utf8 == prog->check_utf8)
5208 prog->check_substr = sv;
5209 }
5210}
66610fdd
RGS
5211
5212/*
5213 * Local variables:
5214 * c-indentation-style: bsd
5215 * c-basic-offset: 4
5216 * indent-tabs-mode: t
5217 * End:
5218 *
37442d52
RGS
5219 * ex: set ts=8 sts=4 sw=4 noet:
5220 */