This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Merge branch 'blead' of camel:/gitroot/perl into blead
[perl5.git] / regexec.c
CommitLineData
a0d0e21e
LW
1/* regexec.c
2 */
3
4/*
4ac71550
TC
5 * One Ring to rule them all, One Ring to find them
6 &
7 * [p.v of _The Lord of the Rings_, opening poem]
8 * [p.50 of _The Lord of the Rings_, I/iii: "The Shadow of the Past"]
9 * [p.254 of _The Lord of the Rings_, II/ii: "The Council of Elrond"]
a0d0e21e
LW
10 */
11
61296642
DM
12/* This file contains functions for executing a regular expression. See
13 * also regcomp.c which funnily enough, contains functions for compiling
166f8a29 14 * a regular expression.
e4a054ea
DM
15 *
16 * This file is also copied at build time to ext/re/re_exec.c, where
17 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
18 * This causes the main functions to be compiled under new names and with
19 * debugging support added, which makes "use re 'debug'" work.
166f8a29
DM
20 */
21
a687059c
LW
22/* NOTE: this is derived from Henry Spencer's regexp code, and should not
23 * confused with the original package (see point 3 below). Thanks, Henry!
24 */
25
26/* Additional note: this code is very heavily munged from Henry's version
27 * in places. In some spots I've traded clarity for efficiency, so don't
28 * blame Henry for some of the lack of readability.
29 */
30
e50aee73
AD
31/* The names of the functions have been changed from regcomp and
32 * regexec to pregcomp and pregexec in order to avoid conflicts
33 * with the POSIX routines of the same names.
34*/
35
b9d5759e 36#ifdef PERL_EXT_RE_BUILD
54df2634 37#include "re_top.h"
9041c2e3 38#endif
56953603 39
a687059c 40/*
e50aee73 41 * pregcomp and pregexec -- regsub and regerror are not used in perl
a687059c
LW
42 *
43 * Copyright (c) 1986 by University of Toronto.
44 * Written by Henry Spencer. Not derived from licensed software.
45 *
46 * Permission is granted to anyone to use this software for any
47 * purpose on any computer system, and to redistribute it freely,
48 * subject to the following restrictions:
49 *
50 * 1. The author is not responsible for the consequences of use of
51 * this software, no matter how awful, even if they arise
52 * from defects in it.
53 *
54 * 2. The origin of this software must not be misrepresented, either
55 * by explicit claim or by omission.
56 *
57 * 3. Altered versions must be plainly marked as such, and must not
58 * be misrepresented as being the original software.
59 *
60 **** Alterations to Henry's code are...
61 ****
4bb101f2 62 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
1129b882
NC
63 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
64 **** by Larry Wall and others
a687059c 65 ****
9ef589d8
LW
66 **** You may distribute under the terms of either the GNU General Public
67 **** License or the Artistic License, as specified in the README file.
a687059c
LW
68 *
69 * Beware that some of this code is subtly aware of the way operator
70 * precedence is structured in regular expressions. Serious changes in
71 * regular-expression syntax might require a total rethink.
72 */
73#include "EXTERN.h"
864dbfa3 74#define PERL_IN_REGEXEC_C
a687059c 75#include "perl.h"
0f5d15d6 76
54df2634
NC
77#ifdef PERL_IN_XSUB_RE
78# include "re_comp.h"
79#else
80# include "regcomp.h"
81#endif
a687059c 82
c277df42
IZ
83#define RF_tainted 1 /* tainted information used? */
84#define RF_warned 2 /* warned about big count? */
faec1544 85
ab3bbdeb 86#define RF_utf8 8 /* Pattern contains multibyte chars? */
a0ed51b3 87
eb160463 88#define UTF ((PL_reg_flags & RF_utf8) != 0)
ce862d02
IZ
89
90#define RS_init 1 /* eval environment created */
91#define RS_set 2 /* replsv value is set */
c277df42 92
a687059c
LW
93#ifndef STATIC
94#define STATIC static
95#endif
96
32fc9b6a 97#define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
7d3e948e 98
c277df42
IZ
99/*
100 * Forwards.
101 */
102
33b8afdf 103#define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
53c4c00c 104#define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
a0ed51b3 105
3dab1dad
YO
106#define HOPc(pos,off) \
107 (char *)(PL_reg_match_utf8 \
52657f30 108 ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \
3dab1dad
YO
109 : (U8*)(pos + off))
110#define HOPBACKc(pos, off) \
07be1b83
YO
111 (char*)(PL_reg_match_utf8\
112 ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \
113 : (pos - off >= PL_bostr) \
8e11feef 114 ? (U8*)pos - off \
3dab1dad 115 : NULL)
efb30f32 116
e7409c1b 117#define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off))
1aa99e6b 118#define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
1aa99e6b 119
1a4fad37
AL
120#define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
121 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END
122#define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
123#define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
124#define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
125#define LOAD_UTF8_CHARCLASS_MARK() LOAD_UTF8_CHARCLASS(mark, "\xcd\x86")
51371543 126
3dab1dad
YO
127/* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */
128
5f80c4cf 129/* for use after a quantifier and before an EXACT-like node -- japhy */
3e901dc0
YO
130/* it would be nice to rework regcomp.sym to generate this stuff. sigh */
131#define JUMPABLE(rn) ( \
132 OP(rn) == OPEN || \
133 (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \
134 OP(rn) == EVAL || \
cca55fe3
JP
135 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
136 OP(rn) == PLUS || OP(rn) == MINMOD || \
ee9b8eae 137 OP(rn) == KEEPS || (PL_regkind[OP(rn)] == VERB) || \
3dab1dad 138 (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \
e2d8ce26 139)
ee9b8eae 140#define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT)
e2d8ce26 141
ee9b8eae
YO
142#define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF )
143
144#if 0
145/* Currently these are only used when PL_regkind[OP(rn)] == EXACT so
146 we don't need this definition. */
147#define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF )
148#define IS_TEXTF(rn) ( OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF )
149#define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL )
150
151#else
152/* ... so we use this as its faster. */
153#define IS_TEXT(rn) ( OP(rn)==EXACT )
154#define IS_TEXTF(rn) ( OP(rn)==EXACTF )
155#define IS_TEXTFL(rn) ( OP(rn)==EXACTFL )
156
157#endif
e2d8ce26 158
a84d97b6
HS
159/*
160 Search for mandatory following text node; for lookahead, the text must
161 follow but for lookbehind (rn->flags != 0) we skip to the next step.
162*/
cca55fe3 163#define FIND_NEXT_IMPT(rn) STMT_START { \
3dab1dad
YO
164 while (JUMPABLE(rn)) { \
165 const OPCODE type = OP(rn); \
166 if (type == SUSPEND || PL_regkind[type] == CURLY) \
e2d8ce26 167 rn = NEXTOPER(NEXTOPER(rn)); \
3dab1dad 168 else if (type == PLUS) \
cca55fe3 169 rn = NEXTOPER(rn); \
3dab1dad 170 else if (type == IFMATCH) \
a84d97b6 171 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
e2d8ce26 172 else rn += NEXT_OFF(rn); \
3dab1dad 173 } \
5f80c4cf 174} STMT_END
74750237 175
c476f425 176
acfe0abc 177static void restore_pos(pTHX_ void *arg);
51371543 178
76e3520e 179STATIC CHECKPOINT
cea2e8a9 180S_regcppush(pTHX_ I32 parenfloor)
a0d0e21e 181{
97aff369 182 dVAR;
a3b680e6 183 const int retval = PL_savestack_ix;
b1ce53c5 184#define REGCP_PAREN_ELEMS 4
a3b680e6 185 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
a0d0e21e 186 int p;
40a82448 187 GET_RE_DEBUG_FLAGS_DECL;
a0d0e21e 188
e49a9654
IH
189 if (paren_elems_to_push < 0)
190 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
191
f0ab9afb 192#define REGCP_OTHER_ELEMS 7
4b3c1a47 193 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
7f69552c 194
3280af22 195 for (p = PL_regsize; p > parenfloor; p--) {
b1ce53c5 196/* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
f0ab9afb
NC
197 SSPUSHINT(PL_regoffs[p].end);
198 SSPUSHINT(PL_regoffs[p].start);
3280af22 199 SSPUSHPTR(PL_reg_start_tmp[p]);
a0d0e21e 200 SSPUSHINT(p);
e7707071 201 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
40a82448 202 " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n",
f0ab9afb 203 (UV)p, (IV)PL_regoffs[p].start,
40a82448 204 (IV)(PL_reg_start_tmp[p] - PL_bostr),
f0ab9afb 205 (IV)PL_regoffs[p].end
40a82448 206 ));
a0d0e21e 207 }
b1ce53c5 208/* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
f0ab9afb 209 SSPUSHPTR(PL_regoffs);
3280af22
NIS
210 SSPUSHINT(PL_regsize);
211 SSPUSHINT(*PL_reglastparen);
a01268b5 212 SSPUSHINT(*PL_reglastcloseparen);
3280af22 213 SSPUSHPTR(PL_reginput);
41123dfd
JH
214#define REGCP_FRAME_ELEMS 2
215/* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
216 * are needed for the regexp context stack bookkeeping. */
217 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
b1ce53c5 218 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
41123dfd 219
a0d0e21e
LW
220 return retval;
221}
222
c277df42 223/* These are needed since we do not localize EVAL nodes: */
ab3bbdeb
YO
224#define REGCP_SET(cp) \
225 DEBUG_STATE_r( \
ab3bbdeb 226 PerlIO_printf(Perl_debug_log, \
e4f74956 227 " Setting an EVAL scope, savestack=%"IVdf"\n", \
ab3bbdeb
YO
228 (IV)PL_savestack_ix)); \
229 cp = PL_savestack_ix
c3464db5 230
ab3bbdeb 231#define REGCP_UNWIND(cp) \
e4f74956 232 DEBUG_STATE_r( \
ab3bbdeb 233 if (cp != PL_savestack_ix) \
e4f74956
YO
234 PerlIO_printf(Perl_debug_log, \
235 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
ab3bbdeb
YO
236 (IV)(cp), (IV)PL_savestack_ix)); \
237 regcpblow(cp)
c277df42 238
76e3520e 239STATIC char *
097eb12c 240S_regcppop(pTHX_ const regexp *rex)
a0d0e21e 241{
97aff369 242 dVAR;
3b6647e0 243 U32 i;
a0d0e21e 244 char *input;
a3621e74
YO
245 GET_RE_DEBUG_FLAGS_DECL;
246
7918f24d
NC
247 PERL_ARGS_ASSERT_REGCPPOP;
248
b1ce53c5 249 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
a0d0e21e 250 i = SSPOPINT;
b1ce53c5
JH
251 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
252 i = SSPOPINT; /* Parentheses elements to pop. */
a0d0e21e 253 input = (char *) SSPOPPTR;
a01268b5 254 *PL_reglastcloseparen = SSPOPINT;
3280af22
NIS
255 *PL_reglastparen = SSPOPINT;
256 PL_regsize = SSPOPINT;
f0ab9afb 257 PL_regoffs=(regexp_paren_pair *) SSPOPPTR;
b1ce53c5 258
6bda09f9 259
b1ce53c5 260 /* Now restore the parentheses context. */
41123dfd
JH
261 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
262 i > 0; i -= REGCP_PAREN_ELEMS) {
1df70142 263 I32 tmps;
097eb12c 264 U32 paren = (U32)SSPOPINT;
3280af22 265 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
f0ab9afb 266 PL_regoffs[paren].start = SSPOPINT;
cf93c79d 267 tmps = SSPOPINT;
3280af22 268 if (paren <= *PL_reglastparen)
f0ab9afb 269 PL_regoffs[paren].end = tmps;
e7707071 270 DEBUG_BUFFERS_r(
c3464db5 271 PerlIO_printf(Perl_debug_log,
b900a521 272 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
f0ab9afb 273 (UV)paren, (IV)PL_regoffs[paren].start,
b900a521 274 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
f0ab9afb 275 (IV)PL_regoffs[paren].end,
3280af22 276 (paren > *PL_reglastparen ? "(no)" : ""));
c277df42 277 );
a0d0e21e 278 }
e7707071 279 DEBUG_BUFFERS_r(
bb7a0f54 280 if (*PL_reglastparen + 1 <= rex->nparens) {
c3464db5 281 PerlIO_printf(Perl_debug_log,
faccc32b 282 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
4f639d21 283 (IV)(*PL_reglastparen + 1), (IV)rex->nparens);
c277df42
IZ
284 }
285 );
daf18116 286#if 1
dafc8851
JH
287 /* It would seem that the similar code in regtry()
288 * already takes care of this, and in fact it is in
289 * a better location to since this code can #if 0-ed out
290 * but the code in regtry() is needed or otherwise tests
291 * requiring null fields (pat.t#187 and split.t#{13,14}
daf18116
JH
292 * (as of patchlevel 7877) will fail. Then again,
293 * this code seems to be necessary or otherwise
225593e1
DM
294 * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
295 * --jhi updated by dapm */
3b6647e0 296 for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) {
097eb12c 297 if (i > PL_regsize)
f0ab9afb
NC
298 PL_regoffs[i].start = -1;
299 PL_regoffs[i].end = -1;
a0d0e21e 300 }
dafc8851 301#endif
a0d0e21e
LW
302 return input;
303}
304
02db2b7b 305#define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
a0d0e21e 306
a687059c 307/*
e50aee73 308 * pregexec and friends
a687059c
LW
309 */
310
76234dfb 311#ifndef PERL_IN_XSUB_RE
a687059c 312/*
c277df42 313 - pregexec - match a regexp against a string
a687059c 314 */
c277df42 315I32
49d7dfbc 316Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend,
c3464db5 317 char *strbeg, I32 minend, SV *screamer, U32 nosave)
c277df42
IZ
318/* strend: pointer to null at end of string */
319/* strbeg: real beginning of string */
320/* minend: end of match must be >=minend after stringarg. */
321/* nosave: For optimizations. */
322{
7918f24d
NC
323 PERL_ARGS_ASSERT_PREGEXEC;
324
c277df42 325 return
9041c2e3 326 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
c277df42
IZ
327 nosave ? 0 : REXEC_COPY_STR);
328}
76234dfb 329#endif
22e551b9 330
9041c2e3 331/*
cad2e5aa
JH
332 * Need to implement the following flags for reg_anch:
333 *
334 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
335 * USE_INTUIT_ML
336 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
337 * INTUIT_AUTORITATIVE_ML
338 * INTUIT_ONCE_NOML - Intuit can match in one location only.
339 * INTUIT_ONCE_ML
340 *
341 * Another flag for this function: SECOND_TIME (so that float substrs
342 * with giant delta may be not rechecked).
343 */
344
345/* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
346
3f7c398e 347/* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
cad2e5aa
JH
348 Otherwise, only SvCUR(sv) is used to get strbeg. */
349
350/* XXXX We assume that strpos is strbeg unless sv. */
351
6eb5f6b9
JH
352/* XXXX Some places assume that there is a fixed substring.
353 An update may be needed if optimizer marks as "INTUITable"
354 RExen without fixed substrings. Similarly, it is assumed that
355 lengths of all the strings are no more than minlen, thus they
356 cannot come from lookahead.
40d049e4
YO
357 (Or minlen should take into account lookahead.)
358 NOTE: Some of this comment is not correct. minlen does now take account
359 of lookahead/behind. Further research is required. -- demerphq
360
361*/
6eb5f6b9 362
2c2d71f5
JH
363/* A failure to find a constant substring means that there is no need to make
364 an expensive call to REx engine, thus we celebrate a failure. Similarly,
365 finding a substring too deep into the string means that less calls to
30944b6d
IZ
366 regtry() should be needed.
367
368 REx compiler's optimizer found 4 possible hints:
369 a) Anchored substring;
370 b) Fixed substring;
371 c) Whether we are anchored (beginning-of-line or \G);
372 d) First node (of those at offset 0) which may distingush positions;
6eb5f6b9 373 We use a)b)d) and multiline-part of c), and try to find a position in the
30944b6d
IZ
374 string which does not contradict any of them.
375 */
2c2d71f5 376
6eb5f6b9
JH
377/* Most of decisions we do here should have been done at compile time.
378 The nodes of the REx which we used for the search should have been
379 deleted from the finite automaton. */
380
cad2e5aa 381char *
288b8c02 382Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos,
9f61653a 383 char *strend, const U32 flags, re_scream_pos_data *data)
cad2e5aa 384{
97aff369 385 dVAR;
288b8c02 386 struct regexp *const prog = (struct regexp *)SvANY(rx);
b7953727 387 register I32 start_shift = 0;
cad2e5aa 388 /* Should be nonnegative! */
b7953727 389 register I32 end_shift = 0;
2c2d71f5
JH
390 register char *s;
391 register SV *check;
a1933d95 392 char *strbeg;
cad2e5aa 393 char *t;
1de06328 394 const bool do_utf8 = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */
cad2e5aa 395 I32 ml_anch;
bd61b366
SS
396 register char *other_last = NULL; /* other substr checked before this */
397 char *check_at = NULL; /* check substr found at this pos */
bbe252da 398 const I32 multiline = prog->extflags & RXf_PMf_MULTILINE;
f8fc2ecf 399 RXi_GET_DECL(prog,progi);
30944b6d 400#ifdef DEBUGGING
890ce7af 401 const char * const i_strpos = strpos;
30944b6d 402#endif
a3621e74
YO
403 GET_RE_DEBUG_FLAGS_DECL;
404
7918f24d
NC
405 PERL_ARGS_ASSERT_RE_INTUIT_START;
406
288b8c02 407 RX_MATCH_UTF8_set(rx,do_utf8);
cad2e5aa 408
3c8556c3 409 if (RX_UTF8(rx)) {
b8d68ded
JH
410 PL_reg_flags |= RF_utf8;
411 }
ab3bbdeb 412 DEBUG_EXECUTE_r(
efd26800 413 debug_start_match(rx, do_utf8, strpos, strend,
1de06328
YO
414 sv ? "Guessing start of match in sv for"
415 : "Guessing start of match in string for");
2a782b5b 416 );
cad2e5aa 417
c344f387
JH
418 /* CHR_DIST() would be more correct here but it makes things slow. */
419 if (prog->minlen > strend - strpos) {
a3621e74 420 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
a72c7584 421 "String too short... [re_intuit_start]\n"));
cad2e5aa 422 goto fail;
2c2d71f5 423 }
1de06328 424
a1933d95 425 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
1aa99e6b 426 PL_regeol = strend;
33b8afdf
JH
427 if (do_utf8) {
428 if (!prog->check_utf8 && prog->check_substr)
429 to_utf8_substr(prog);
430 check = prog->check_utf8;
431 } else {
432 if (!prog->check_substr && prog->check_utf8)
433 to_byte_substr(prog);
434 check = prog->check_substr;
435 }
1de06328 436 if (check == &PL_sv_undef) {
a3621e74 437 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1de06328 438 "Non-utf8 string cannot match utf8 check string\n"));
33b8afdf
JH
439 goto fail;
440 }
bbe252da
YO
441 if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */
442 ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE)
443 || ( (prog->extflags & RXf_ANCH_BOL)
7fba1cd6 444 && !multiline ) ); /* Check after \n? */
cad2e5aa 445
7e25d62c 446 if (!ml_anch) {
bbe252da
YO
447 if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */
448 && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */
3f7c398e 449 /* SvCUR is not set on references: SvRV and SvPVX_const overlap */
7e25d62c
JH
450 && sv && !SvROK(sv)
451 && (strpos != strbeg)) {
a3621e74 452 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
7e25d62c
JH
453 goto fail;
454 }
455 if (prog->check_offset_min == prog->check_offset_max &&
bbe252da 456 !(prog->extflags & RXf_CANY_SEEN)) {
2c2d71f5 457 /* Substring at constant offset from beg-of-str... */
cad2e5aa
JH
458 I32 slen;
459
1aa99e6b 460 s = HOP3c(strpos, prog->check_offset_min, strend);
1de06328 461
653099ff
GS
462 if (SvTAIL(check)) {
463 slen = SvCUR(check); /* >= 1 */
cad2e5aa 464
9041c2e3 465 if ( strend - s > slen || strend - s < slen - 1
2c2d71f5 466 || (strend - s == slen && strend[-1] != '\n')) {
a3621e74 467 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
2c2d71f5 468 goto fail_finish;
cad2e5aa
JH
469 }
470 /* Now should match s[0..slen-2] */
471 slen--;
3f7c398e 472 if (slen && (*SvPVX_const(check) != *s
cad2e5aa 473 || (slen > 1
3f7c398e 474 && memNE(SvPVX_const(check), s, slen)))) {
2c2d71f5 475 report_neq:
a3621e74 476 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
2c2d71f5
JH
477 goto fail_finish;
478 }
cad2e5aa 479 }
3f7c398e 480 else if (*SvPVX_const(check) != *s
653099ff 481 || ((slen = SvCUR(check)) > 1
3f7c398e 482 && memNE(SvPVX_const(check), s, slen)))
2c2d71f5 483 goto report_neq;
c315bfe8 484 check_at = s;
2c2d71f5 485 goto success_at_start;
7e25d62c 486 }
cad2e5aa 487 }
2c2d71f5 488 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
cad2e5aa 489 s = strpos;
2c2d71f5 490 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
1de06328
YO
491 end_shift = prog->check_end_shift;
492
2c2d71f5 493 if (!ml_anch) {
a3b680e6 494 const I32 end = prog->check_offset_max + CHR_SVLEN(check)
653099ff 495 - (SvTAIL(check) != 0);
a3b680e6 496 const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
2c2d71f5
JH
497
498 if (end_shift < eshift)
499 end_shift = eshift;
500 }
cad2e5aa 501 }
2c2d71f5 502 else { /* Can match at random position */
cad2e5aa
JH
503 ml_anch = 0;
504 s = strpos;
1de06328
YO
505 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
506 end_shift = prog->check_end_shift;
507
508 /* end shift should be non negative here */
cad2e5aa
JH
509 }
510
bcdf7404 511#ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */
0033605d 512 if (end_shift < 0)
1de06328 513 Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ",
220fc49f 514 (IV)end_shift, RX_PRECOMP(prog));
2c2d71f5
JH
515#endif
516
2c2d71f5
JH
517 restart:
518 /* Find a possible match in the region s..strend by looking for
519 the "check" substring in the region corrected by start/end_shift. */
1de06328
YO
520
521 {
522 I32 srch_start_shift = start_shift;
523 I32 srch_end_shift = end_shift;
524 if (srch_start_shift < 0 && strbeg - s > srch_start_shift) {
525 srch_end_shift -= ((strbeg - s) - srch_start_shift);
526 srch_start_shift = strbeg - s;
527 }
6bda09f9 528 DEBUG_OPTIMISE_MORE_r({
1de06328
YO
529 PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n",
530 (IV)prog->check_offset_min,
531 (IV)srch_start_shift,
532 (IV)srch_end_shift,
533 (IV)prog->check_end_shift);
534 });
535
cad2e5aa 536 if (flags & REXEC_SCREAM) {
cad2e5aa 537 I32 p = -1; /* Internal iterator of scream. */
a3b680e6 538 I32 * const pp = data ? data->scream_pos : &p;
cad2e5aa 539
2c2d71f5
JH
540 if (PL_screamfirst[BmRARE(check)] >= 0
541 || ( BmRARE(check) == '\n'
85c508c3 542 && (BmPREVIOUS(check) == SvCUR(check) - 1)
2c2d71f5 543 && SvTAIL(check) ))
9041c2e3 544 s = screaminstr(sv, check,
1de06328 545 srch_start_shift + (s - strbeg), srch_end_shift, pp, 0);
cad2e5aa 546 else
2c2d71f5 547 goto fail_finish;
4addbd3b 548 /* we may be pointing at the wrong string */
07bc277f 549 if (s && RXp_MATCH_COPIED(prog))
3f7c398e 550 s = strbeg + (s - SvPVX_const(sv));
cad2e5aa
JH
551 if (data)
552 *data->scream_olds = s;
553 }
1de06328
YO
554 else {
555 U8* start_point;
556 U8* end_point;
bbe252da 557 if (prog->extflags & RXf_CANY_SEEN) {
1de06328
YO
558 start_point= (U8*)(s + srch_start_shift);
559 end_point= (U8*)(strend - srch_end_shift);
560 } else {
561 start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend);
562 end_point= HOP3(strend, -srch_end_shift, strbeg);
563 }
6bda09f9 564 DEBUG_OPTIMISE_MORE_r({
56570a2c 565 PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n",
1de06328 566 (int)(end_point - start_point),
fc8cd66c 567 (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point),
1de06328
YO
568 start_point);
569 });
570
571 s = fbm_instr( start_point, end_point,
7fba1cd6 572 check, multiline ? FBMrf_MULTILINE : 0);
1de06328
YO
573 }
574 }
cad2e5aa
JH
575 /* Update the count-of-usability, remove useless subpatterns,
576 unshift s. */
2c2d71f5 577
ab3bbdeb
YO
578 DEBUG_EXECUTE_r({
579 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
580 SvPVX_const(check), RE_SV_DUMPLEN(check), 30);
581 PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s",
2c2d71f5 582 (s ? "Found" : "Did not find"),
ab3bbdeb
YO
583 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)
584 ? "anchored" : "floating"),
585 quoted,
586 RE_SV_TAIL(check),
587 (s ? " at offset " : "...\n") );
588 });
2c2d71f5
JH
589
590 if (!s)
591 goto fail_finish;
2c2d71f5 592 /* Finish the diagnostic message */
a3621e74 593 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
2c2d71f5 594
1de06328
YO
595 /* XXX dmq: first branch is for positive lookbehind...
596 Our check string is offset from the beginning of the pattern.
597 So we need to do any stclass tests offset forward from that
598 point. I think. :-(
599 */
600
601
602
603 check_at=s;
604
605
2c2d71f5
JH
606 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
607 Start with the other substr.
608 XXXX no SCREAM optimization yet - and a very coarse implementation
a0288114 609 XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
2c2d71f5
JH
610 *always* match. Probably should be marked during compile...
611 Probably it is right to do no SCREAM here...
612 */
613
1de06328
YO
614 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8)
615 : (prog->float_substr && prog->anchored_substr))
616 {
30944b6d 617 /* Take into account the "other" substring. */
2c2d71f5
JH
618 /* XXXX May be hopelessly wrong for UTF... */
619 if (!other_last)
6eb5f6b9 620 other_last = strpos;
33b8afdf 621 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
30944b6d
IZ
622 do_other_anchored:
623 {
890ce7af
AL
624 char * const last = HOP3c(s, -start_shift, strbeg);
625 char *last1, *last2;
be8e71aa 626 char * const saved_s = s;
33b8afdf 627 SV* must;
2c2d71f5 628
2c2d71f5
JH
629 t = s - prog->check_offset_max;
630 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
1d86a7f9 631 && (!do_utf8
0ce71af7 632 || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos))
2c2d71f5 633 && t > strpos)))
6f207bd3 634 NOOP;
2c2d71f5
JH
635 else
636 t = strpos;
1aa99e6b 637 t = HOP3c(t, prog->anchored_offset, strend);
6eb5f6b9
JH
638 if (t < other_last) /* These positions already checked */
639 t = other_last;
1aa99e6b 640 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
2c2d71f5
JH
641 if (last < last1)
642 last1 = last;
1de06328
YO
643 /* XXXX It is not documented what units *_offsets are in.
644 We assume bytes, but this is clearly wrong.
645 Meaning this code needs to be carefully reviewed for errors.
646 dmq.
647 */
648
2c2d71f5 649 /* On end-of-str: see comment below. */
33b8afdf
JH
650 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
651 if (must == &PL_sv_undef) {
652 s = (char*)NULL;
1de06328 653 DEBUG_r(must = prog->anchored_utf8); /* for debug */
33b8afdf
JH
654 }
655 else
656 s = fbm_instr(
657 (unsigned char*)t,
658 HOP3(HOP3(last1, prog->anchored_offset, strend)
659 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
660 must,
7fba1cd6 661 multiline ? FBMrf_MULTILINE : 0
33b8afdf 662 );
ab3bbdeb
YO
663 DEBUG_EXECUTE_r({
664 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
665 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
666 PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s",
2c2d71f5 667 (s ? "Found" : "Contradicts"),
ab3bbdeb
YO
668 quoted, RE_SV_TAIL(must));
669 });
670
671
2c2d71f5
JH
672 if (!s) {
673 if (last1 >= last2) {
a3621e74 674 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2c2d71f5
JH
675 ", giving up...\n"));
676 goto fail_finish;
677 }
a3621e74 678 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2c2d71f5 679 ", trying floating at offset %ld...\n",
be8e71aa 680 (long)(HOP3c(saved_s, 1, strend) - i_strpos)));
1aa99e6b
IH
681 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
682 s = HOP3c(last, 1, strend);
2c2d71f5
JH
683 goto restart;
684 }
685 else {
a3621e74 686 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
30944b6d 687 (long)(s - i_strpos)));
1aa99e6b
IH
688 t = HOP3c(s, -prog->anchored_offset, strbeg);
689 other_last = HOP3c(s, 1, strend);
be8e71aa 690 s = saved_s;
2c2d71f5
JH
691 if (t == strpos)
692 goto try_at_start;
2c2d71f5
JH
693 goto try_at_offset;
694 }
30944b6d 695 }
2c2d71f5
JH
696 }
697 else { /* Take into account the floating substring. */
33b8afdf 698 char *last, *last1;
be8e71aa 699 char * const saved_s = s;
33b8afdf
JH
700 SV* must;
701
702 t = HOP3c(s, -start_shift, strbeg);
703 last1 = last =
704 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
705 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
706 last = HOP3c(t, prog->float_max_offset, strend);
707 s = HOP3c(t, prog->float_min_offset, strend);
708 if (s < other_last)
709 s = other_last;
2c2d71f5 710 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
33b8afdf
JH
711 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
712 /* fbm_instr() takes into account exact value of end-of-str
713 if the check is SvTAIL(ed). Since false positives are OK,
714 and end-of-str is not later than strend we are OK. */
715 if (must == &PL_sv_undef) {
716 s = (char*)NULL;
1de06328 717 DEBUG_r(must = prog->float_utf8); /* for debug message */
33b8afdf
JH
718 }
719 else
2c2d71f5 720 s = fbm_instr((unsigned char*)s,
33b8afdf
JH
721 (unsigned char*)last + SvCUR(must)
722 - (SvTAIL(must)!=0),
7fba1cd6 723 must, multiline ? FBMrf_MULTILINE : 0);
ab3bbdeb
YO
724 DEBUG_EXECUTE_r({
725 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
726 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
727 PerlIO_printf(Perl_debug_log, "%s floating substr %s%s",
33b8afdf 728 (s ? "Found" : "Contradicts"),
ab3bbdeb
YO
729 quoted, RE_SV_TAIL(must));
730 });
33b8afdf
JH
731 if (!s) {
732 if (last1 == last) {
a3621e74 733 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
33b8afdf
JH
734 ", giving up...\n"));
735 goto fail_finish;
2c2d71f5 736 }
a3621e74 737 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
33b8afdf 738 ", trying anchored starting at offset %ld...\n",
be8e71aa 739 (long)(saved_s + 1 - i_strpos)));
33b8afdf
JH
740 other_last = last;
741 s = HOP3c(t, 1, strend);
742 goto restart;
743 }
744 else {
a3621e74 745 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
33b8afdf
JH
746 (long)(s - i_strpos)));
747 other_last = s; /* Fix this later. --Hugo */
be8e71aa 748 s = saved_s;
33b8afdf
JH
749 if (t == strpos)
750 goto try_at_start;
751 goto try_at_offset;
752 }
2c2d71f5 753 }
cad2e5aa 754 }
2c2d71f5 755
1de06328 756
9ef43ace 757 t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos);
1de06328 758
6bda09f9 759 DEBUG_OPTIMISE_MORE_r(
1de06328
YO
760 PerlIO_printf(Perl_debug_log,
761 "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n",
762 (IV)prog->check_offset_min,
763 (IV)prog->check_offset_max,
764 (IV)(s-strpos),
765 (IV)(t-strpos),
766 (IV)(t-s),
767 (IV)(strend-strpos)
768 )
769 );
770
2c2d71f5 771 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
1d86a7f9 772 && (!do_utf8
9ef43ace 773 || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos)))
1de06328
YO
774 && t > strpos)))
775 {
2c2d71f5
JH
776 /* Fixed substring is found far enough so that the match
777 cannot start at strpos. */
778 try_at_offset:
cad2e5aa 779 if (ml_anch && t[-1] != '\n') {
30944b6d
IZ
780 /* Eventually fbm_*() should handle this, but often
781 anchored_offset is not 0, so this check will not be wasted. */
782 /* XXXX In the code below we prefer to look for "^" even in
783 presence of anchored substrings. And we search even
784 beyond the found float position. These pessimizations
785 are historical artefacts only. */
786 find_anchor:
2c2d71f5 787 while (t < strend - prog->minlen) {
cad2e5aa 788 if (*t == '\n') {
4ee3650e 789 if (t < check_at - prog->check_offset_min) {
33b8afdf 790 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
4ee3650e
GS
791 /* Since we moved from the found position,
792 we definitely contradict the found anchored
30944b6d
IZ
793 substr. Due to the above check we do not
794 contradict "check" substr.
795 Thus we can arrive here only if check substr
796 is float. Redo checking for "other"=="fixed".
797 */
9041c2e3 798 strpos = t + 1;
a3621e74 799 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
e4584336 800 PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
30944b6d
IZ
801 goto do_other_anchored;
802 }
4ee3650e
GS
803 /* We don't contradict the found floating substring. */
804 /* XXXX Why not check for STCLASS? */
cad2e5aa 805 s = t + 1;
a3621e74 806 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
e4584336 807 PL_colors[0], PL_colors[1], (long)(s - i_strpos)));
cad2e5aa
JH
808 goto set_useful;
809 }
4ee3650e
GS
810 /* Position contradicts check-string */
811 /* XXXX probably better to look for check-string
812 than for "\n", so one should lower the limit for t? */
a3621e74 813 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
e4584336 814 PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos)));
0e41cd87 815 other_last = strpos = s = t + 1;
cad2e5aa
JH
816 goto restart;
817 }
818 t++;
819 }
a3621e74 820 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
e4584336 821 PL_colors[0], PL_colors[1]));
2c2d71f5 822 goto fail_finish;
cad2e5aa 823 }
f5952150 824 else {
a3621e74 825 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
e4584336 826 PL_colors[0], PL_colors[1]));
f5952150 827 }
cad2e5aa
JH
828 s = t;
829 set_useful:
33b8afdf 830 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
cad2e5aa
JH
831 }
832 else {
f5952150 833 /* The found string does not prohibit matching at strpos,
2c2d71f5 834 - no optimization of calling REx engine can be performed,
f5952150
GS
835 unless it was an MBOL and we are not after MBOL,
836 or a future STCLASS check will fail this. */
2c2d71f5
JH
837 try_at_start:
838 /* Even in this situation we may use MBOL flag if strpos is offset
839 wrt the start of the string. */
05b4157f 840 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
a1933d95 841 && (strpos != strbeg) && strpos[-1] != '\n'
d506a20d 842 /* May be due to an implicit anchor of m{.*foo} */
bbe252da 843 && !(prog->intflags & PREGf_IMPLICIT))
d506a20d 844 {
cad2e5aa
JH
845 t = strpos;
846 goto find_anchor;
847 }
a3621e74 848 DEBUG_EXECUTE_r( if (ml_anch)
f5952150 849 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
70685ca0 850 (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]);
30944b6d 851 );
2c2d71f5 852 success_at_start:
bbe252da 853 if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */
33b8afdf
JH
854 && (do_utf8 ? (
855 prog->check_utf8 /* Could be deleted already */
856 && --BmUSEFUL(prog->check_utf8) < 0
857 && (prog->check_utf8 == prog->float_utf8)
858 ) : (
859 prog->check_substr /* Could be deleted already */
860 && --BmUSEFUL(prog->check_substr) < 0
861 && (prog->check_substr == prog->float_substr)
862 )))
66e933ab 863 {
cad2e5aa 864 /* If flags & SOMETHING - do not do it many times on the same match */
a3621e74 865 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
33b8afdf
JH
866 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
867 if (do_utf8 ? prog->check_substr : prog->check_utf8)
868 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
a0714e2c
SS
869 prog->check_substr = prog->check_utf8 = NULL; /* disable */
870 prog->float_substr = prog->float_utf8 = NULL; /* clear */
871 check = NULL; /* abort */
cad2e5aa 872 s = strpos;
3cf5c195
IZ
873 /* XXXX This is a remnant of the old implementation. It
874 looks wasteful, since now INTUIT can use many
6eb5f6b9 875 other heuristics. */
bbe252da 876 prog->extflags &= ~RXf_USE_INTUIT;
cad2e5aa
JH
877 }
878 else
879 s = strpos;
880 }
881
6eb5f6b9
JH
882 /* Last resort... */
883 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
1de06328
YO
884 /* trie stclasses are too expensive to use here, we are better off to
885 leave it to regmatch itself */
f8fc2ecf 886 if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) {
6eb5f6b9
JH
887 /* minlen == 0 is possible if regstclass is \b or \B,
888 and the fixed substr is ''$.
889 Since minlen is already taken into account, s+1 is before strend;
890 accidentally, minlen >= 1 guaranties no false positives at s + 1
891 even for \b or \B. But (minlen? 1 : 0) below assumes that
892 regstclass does not come from lookahead... */
893 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
894 This leaves EXACTF only, which is dealt with in find_byclass(). */
f8fc2ecf
YO
895 const U8* const str = (U8*)STRING(progi->regstclass);
896 const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT
897 ? CHR_DIST(str+STR_LEN(progi->regstclass), str)
66e933ab 898 : 1);
1de06328
YO
899 char * endpos;
900 if (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
901 endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend);
902 else if (prog->float_substr || prog->float_utf8)
903 endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend);
904 else
905 endpos= strend;
906
70685ca0
JH
907 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf"\n",
908 (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg)));
1de06328 909
6eb5f6b9 910 t = s;
f8fc2ecf 911 s = find_byclass(prog, progi->regstclass, s, endpos, NULL);
6eb5f6b9
JH
912 if (!s) {
913#ifdef DEBUGGING
cbbf8932 914 const char *what = NULL;
6eb5f6b9
JH
915#endif
916 if (endpos == strend) {
a3621e74 917 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
6eb5f6b9
JH
918 "Could not match STCLASS...\n") );
919 goto fail;
920 }
a3621e74 921 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
66e933ab 922 "This position contradicts STCLASS...\n") );
bbe252da 923 if ((prog->extflags & RXf_ANCH) && !ml_anch)
653099ff 924 goto fail;
6eb5f6b9 925 /* Contradict one of substrings */
33b8afdf
JH
926 if (prog->anchored_substr || prog->anchored_utf8) {
927 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
a3621e74 928 DEBUG_EXECUTE_r( what = "anchored" );
6eb5f6b9 929 hop_and_restart:
1aa99e6b 930 s = HOP3c(t, 1, strend);
66e933ab
GS
931 if (s + start_shift + end_shift > strend) {
932 /* XXXX Should be taken into account earlier? */
a3621e74 933 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
66e933ab
GS
934 "Could not match STCLASS...\n") );
935 goto fail;
936 }
5e39e1e5
HS
937 if (!check)
938 goto giveup;
a3621e74 939 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
f5952150 940 "Looking for %s substr starting at offset %ld...\n",
6eb5f6b9
JH
941 what, (long)(s + start_shift - i_strpos)) );
942 goto restart;
943 }
66e933ab 944 /* Have both, check_string is floating */
6eb5f6b9
JH
945 if (t + start_shift >= check_at) /* Contradicts floating=check */
946 goto retry_floating_check;
947 /* Recheck anchored substring, but not floating... */
9041c2e3 948 s = check_at;
5e39e1e5
HS
949 if (!check)
950 goto giveup;
a3621e74 951 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
f5952150 952 "Looking for anchored substr starting at offset %ld...\n",
6eb5f6b9
JH
953 (long)(other_last - i_strpos)) );
954 goto do_other_anchored;
955 }
60e71179
GS
956 /* Another way we could have checked stclass at the
957 current position only: */
958 if (ml_anch) {
959 s = t = t + 1;
5e39e1e5
HS
960 if (!check)
961 goto giveup;
a3621e74 962 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
f5952150 963 "Looking for /%s^%s/m starting at offset %ld...\n",
e4584336 964 PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
60e71179 965 goto try_at_offset;
66e933ab 966 }
33b8afdf 967 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
60e71179 968 goto fail;
6eb5f6b9
JH
969 /* Check is floating subtring. */
970 retry_floating_check:
971 t = check_at - start_shift;
a3621e74 972 DEBUG_EXECUTE_r( what = "floating" );
6eb5f6b9
JH
973 goto hop_and_restart;
974 }
b7953727 975 if (t != s) {
a3621e74 976 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
6eb5f6b9 977 "By STCLASS: moving %ld --> %ld\n",
b7953727
JH
978 (long)(t - i_strpos), (long)(s - i_strpos))
979 );
980 }
981 else {
a3621e74 982 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
b7953727
JH
983 "Does not contradict STCLASS...\n");
984 );
985 }
6eb5f6b9 986 }
5e39e1e5 987 giveup:
a3621e74 988 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
5e39e1e5
HS
989 PL_colors[4], (check ? "Guessed" : "Giving up"),
990 PL_colors[5], (long)(s - i_strpos)) );
cad2e5aa 991 return s;
2c2d71f5
JH
992
993 fail_finish: /* Substring not found */
33b8afdf
JH
994 if (prog->check_substr || prog->check_utf8) /* could be removed already */
995 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
cad2e5aa 996 fail:
a3621e74 997 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
e4584336 998 PL_colors[4], PL_colors[5]));
bd61b366 999 return NULL;
cad2e5aa 1000}
9661b544 1001
a0a388a1
YO
1002#define DECL_TRIE_TYPE(scan) \
1003 const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \
1004 trie_type = (scan->flags != EXACT) \
1005 ? (do_utf8 ? trie_utf8_fold : (UTF ? trie_latin_utf8_fold : trie_plain)) \
1006 : (do_utf8 ? trie_utf8 : trie_plain)
3b0527fe 1007
55eed653
NC
1008#define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \
1009uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \
c012444f 1010 UV uvc_unfolded = 0; \
4cadc6a9
YO
1011 switch (trie_type) { \
1012 case trie_utf8_fold: \
1013 if ( foldlen>0 ) { \
c012444f 1014 uvc_unfolded = uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
4cadc6a9
YO
1015 foldlen -= len; \
1016 uscan += len; \
1017 len=0; \
1018 } else { \
c012444f 1019 uvc_unfolded = uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
4cadc6a9
YO
1020 uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \
1021 foldlen -= UNISKIP( uvc ); \
1022 uscan = foldbuf + UNISKIP( uvc ); \
1023 } \
1024 break; \
a0a388a1
YO
1025 case trie_latin_utf8_fold: \
1026 if ( foldlen>0 ) { \
1027 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
1028 foldlen -= len; \
1029 uscan += len; \
1030 len=0; \
1031 } else { \
1032 len = 1; \
1033 uvc = to_uni_fold( *(U8*)uc, foldbuf, &foldlen ); \
1034 foldlen -= UNISKIP( uvc ); \
1035 uscan = foldbuf + UNISKIP( uvc ); \
1036 } \
1037 break; \
4cadc6a9
YO
1038 case trie_utf8: \
1039 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
1040 break; \
1041 case trie_plain: \
1042 uvc = (UV)*uc; \
1043 len = 1; \
1044 } \
1045 \
1046 if (uvc < 256) { \
1047 charid = trie->charmap[ uvc ]; \
1048 } \
1049 else { \
1050 charid = 0; \
55eed653
NC
1051 if (widecharmap) { \
1052 SV** const svpp = hv_fetch(widecharmap, \
4cadc6a9
YO
1053 (char*)&uvc, sizeof(UV), 0); \
1054 if (svpp) \
1055 charid = (U16)SvIV(*svpp); \
1056 } \
1057 } \
c012444f
SR
1058 if (!charid && trie_type == trie_utf8_fold && !UTF) { \
1059 charid = trie->charmap[uvc_unfolded]; \
1060 } \
4cadc6a9
YO
1061} STMT_END
1062
a0a388a1
YO
1063#define REXEC_FBC_EXACTISH_CHECK(CoNd) \
1064{ \
1065 char *my_strend= (char *)strend; \
4cadc6a9
YO
1066 if ( (CoNd) \
1067 && (ln == len || \
a0a388a1 1068 !ibcmp_utf8(s, &my_strend, 0, do_utf8, \
4cadc6a9 1069 m, NULL, ln, (bool)UTF)) \
a0a388a1 1070 && (!reginfo || regtry(reginfo, &s)) ) \
4cadc6a9
YO
1071 goto got_it; \
1072 else { \
1073 U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \
1074 uvchr_to_utf8(tmpbuf, c); \
1075 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); \
1076 if ( f != c \
1077 && (f == c1 || f == c2) \
a0a388a1
YO
1078 && (ln == len || \
1079 !ibcmp_utf8(s, &my_strend, 0, do_utf8,\
1080 m, NULL, ln, (bool)UTF)) \
1081 && (!reginfo || regtry(reginfo, &s)) ) \
4cadc6a9
YO
1082 goto got_it; \
1083 } \
a0a388a1
YO
1084} \
1085s += len
4cadc6a9
YO
1086
1087#define REXEC_FBC_EXACTISH_SCAN(CoNd) \
1088STMT_START { \
1089 while (s <= e) { \
1090 if ( (CoNd) \
1091 && (ln == 1 || !(OP(c) == EXACTF \
1092 ? ibcmp(s, m, ln) \
1093 : ibcmp_locale(s, m, ln))) \
24b23f37 1094 && (!reginfo || regtry(reginfo, &s)) ) \
4cadc6a9
YO
1095 goto got_it; \
1096 s++; \
1097 } \
1098} STMT_END
1099
1100#define REXEC_FBC_UTF8_SCAN(CoDe) \
1101STMT_START { \
1102 while (s + (uskip = UTF8SKIP(s)) <= strend) { \
1103 CoDe \
1104 s += uskip; \
1105 } \
1106} STMT_END
1107
1108#define REXEC_FBC_SCAN(CoDe) \
1109STMT_START { \
1110 while (s < strend) { \
1111 CoDe \
1112 s++; \
1113 } \
1114} STMT_END
1115
1116#define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \
1117REXEC_FBC_UTF8_SCAN( \
1118 if (CoNd) { \
24b23f37 1119 if (tmp && (!reginfo || regtry(reginfo, &s))) \
4cadc6a9
YO
1120 goto got_it; \
1121 else \
1122 tmp = doevery; \
1123 } \
1124 else \
1125 tmp = 1; \
1126)
1127
1128#define REXEC_FBC_CLASS_SCAN(CoNd) \
1129REXEC_FBC_SCAN( \
1130 if (CoNd) { \
24b23f37 1131 if (tmp && (!reginfo || regtry(reginfo, &s))) \
4cadc6a9
YO
1132 goto got_it; \
1133 else \
1134 tmp = doevery; \
1135 } \
1136 else \
1137 tmp = 1; \
1138)
1139
1140#define REXEC_FBC_TRYIT \
24b23f37 1141if ((!reginfo || regtry(reginfo, &s))) \
4cadc6a9
YO
1142 goto got_it
1143
e1d1eefb
YO
1144#define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \
1145 if (do_utf8) { \
1146 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1147 } \
1148 else { \
1149 REXEC_FBC_CLASS_SCAN(CoNd); \
1150 } \
1151 break
1152
4cadc6a9
YO
1153#define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \
1154 if (do_utf8) { \
1155 UtFpReLoAd; \
1156 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1157 } \
1158 else { \
1159 REXEC_FBC_CLASS_SCAN(CoNd); \
1160 } \
1161 break
1162
1163#define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \
1164 PL_reg_flags |= RF_tainted; \
1165 if (do_utf8) { \
1166 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1167 } \
1168 else { \
1169 REXEC_FBC_CLASS_SCAN(CoNd); \
1170 } \
1171 break
1172
786e8c11
YO
1173#define DUMP_EXEC_POS(li,s,doutf8) \
1174 dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8)
1175
1176/* We know what class REx starts with. Try to find this position... */
1177/* if reginfo is NULL, its a dryrun */
1178/* annoyingly all the vars in this routine have different names from their counterparts
1179 in regmatch. /grrr */
1180
3c3eec57 1181STATIC char *
07be1b83 1182S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
24b23f37 1183 const char *strend, regmatch_info *reginfo)
a687059c 1184{
27da23d5 1185 dVAR;
bbe252da 1186 const I32 doevery = (prog->intflags & PREGf_SKIP) == 0;
6eb5f6b9 1187 char *m;
d8093b23 1188 STRLEN ln;
5dab1207 1189 STRLEN lnc;
078c425b 1190 register STRLEN uskip;
d8093b23
G
1191 unsigned int c1;
1192 unsigned int c2;
6eb5f6b9
JH
1193 char *e;
1194 register I32 tmp = 1; /* Scratch variable? */
a3b680e6 1195 register const bool do_utf8 = PL_reg_match_utf8;
f8fc2ecf 1196 RXi_GET_DECL(prog,progi);
7918f24d
NC
1197
1198 PERL_ARGS_ASSERT_FIND_BYCLASS;
f8fc2ecf 1199
6eb5f6b9
JH
1200 /* We know what class it must start with. */
1201 switch (OP(c)) {
6eb5f6b9 1202 case ANYOF:
388cc4de 1203 if (do_utf8) {
4cadc6a9 1204 REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
388cc4de 1205 !UTF8_IS_INVARIANT((U8)s[0]) ?
32fc9b6a 1206 reginclass(prog, c, (U8*)s, 0, do_utf8) :
4cadc6a9 1207 REGINCLASS(prog, c, (U8*)s));
388cc4de
HS
1208 }
1209 else {
1210 while (s < strend) {
1211 STRLEN skip = 1;
1212
32fc9b6a 1213 if (REGINCLASS(prog, c, (U8*)s) ||
388cc4de
HS
1214 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
1215 /* The assignment of 2 is intentional:
1216 * for the folded sharp s, the skip is 2. */
1217 (skip = SHARP_S_SKIP))) {
24b23f37 1218 if (tmp && (!reginfo || regtry(reginfo, &s)))
388cc4de
HS
1219 goto got_it;
1220 else
1221 tmp = doevery;
1222 }
1223 else
1224 tmp = 1;
1225 s += skip;
1226 }
a0d0e21e 1227 }
6eb5f6b9 1228 break;
f33976b4 1229 case CANY:
4cadc6a9 1230 REXEC_FBC_SCAN(
24b23f37 1231 if (tmp && (!reginfo || regtry(reginfo, &s)))
f33976b4
DB
1232 goto got_it;
1233 else
1234 tmp = doevery;
4cadc6a9 1235 );
f33976b4 1236 break;
6eb5f6b9 1237 case EXACTF:
5dab1207
NIS
1238 m = STRING(c);
1239 ln = STR_LEN(c); /* length to match in octets/bytes */
1240 lnc = (I32) ln; /* length to match in characters */
1aa99e6b 1241 if (UTF) {
a2a2844f 1242 STRLEN ulen1, ulen2;
5dab1207 1243 U8 *sm = (U8 *) m;
89ebb4a3
JH
1244 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
1245 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
97dc7d3e
RGS
1246 /* used by commented-out code below */
1247 /*const U32 uniflags = UTF8_ALLOW_DEFAULT;*/
a0a388a1
YO
1248
1249 /* XXX: Since the node will be case folded at compile
1250 time this logic is a little odd, although im not
1251 sure that its actually wrong. --dmq */
1252
1253 c1 = to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1254 c2 = to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1255
1256 /* XXX: This is kinda strange. to_utf8_XYZ returns the
1257 codepoint of the first character in the converted
1258 form, yet originally we did the extra step.
1259 No tests fail by commenting this code out however
1260 so Ive left it out. -- dmq.
1261
89ebb4a3 1262 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
041457d9 1263 0, uniflags);
89ebb4a3 1264 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
041457d9 1265 0, uniflags);
a0a388a1
YO
1266 */
1267
5dab1207
NIS
1268 lnc = 0;
1269 while (sm < ((U8 *) m + ln)) {
1270 lnc++;
1271 sm += UTF8SKIP(sm);
1272 }
1aa99e6b
IH
1273 }
1274 else {
1275 c1 = *(U8*)m;
1276 c2 = PL_fold[c1];
1277 }
6eb5f6b9
JH
1278 goto do_exactf;
1279 case EXACTFL:
5dab1207
NIS
1280 m = STRING(c);
1281 ln = STR_LEN(c);
1282 lnc = (I32) ln;
d8093b23 1283 c1 = *(U8*)m;
6eb5f6b9
JH
1284 c2 = PL_fold_locale[c1];
1285 do_exactf:
db12adc6 1286 e = HOP3c(strend, -((I32)lnc), s);
b3c9acc1 1287
3b0527fe 1288 if (!reginfo && e < s)
6eb5f6b9 1289 e = s; /* Due to minlen logic of intuit() */
1aa99e6b 1290
60a8b682
JH
1291 /* The idea in the EXACTF* cases is to first find the
1292 * first character of the EXACTF* node and then, if
1293 * necessary, case-insensitively compare the full
1294 * text of the node. The c1 and c2 are the first
1295 * characters (though in Unicode it gets a bit
1296 * more complicated because there are more cases
7f16dd3d
JH
1297 * than just upper and lower: one needs to use
1298 * the so-called folding case for case-insensitive
1299 * matching (called "loose matching" in Unicode).
1300 * ibcmp_utf8() will do just that. */
60a8b682 1301
a0a388a1 1302 if (do_utf8 || UTF) {
575cac57 1303 UV c, f;
89ebb4a3 1304 U8 tmpbuf [UTF8_MAXBYTES+1];
a0a388a1
YO
1305 STRLEN len = 1;
1306 STRLEN foldlen;
4ad0818d 1307 const U32 uniflags = UTF8_ALLOW_DEFAULT;
09091399 1308 if (c1 == c2) {
5dab1207
NIS
1309 /* Upper and lower of 1st char are equal -
1310 * probably not a "letter". */
1aa99e6b 1311 while (s <= e) {
a0a388a1
YO
1312 if (do_utf8) {
1313 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
041457d9 1314 uniflags);
a0a388a1
YO
1315 } else {
1316 c = *((U8*)s);
1317 }
4cadc6a9 1318 REXEC_FBC_EXACTISH_CHECK(c == c1);
1aa99e6b 1319 }
09091399
JH
1320 }
1321 else {
1aa99e6b 1322 while (s <= e) {
a0a388a1
YO
1323 if (do_utf8) {
1324 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
041457d9 1325 uniflags);
a0a388a1
YO
1326 } else {
1327 c = *((U8*)s);
1328 }
80aecb99 1329
60a8b682 1330 /* Handle some of the three Greek sigmas cases.
8c01da3c
JH
1331 * Note that not all the possible combinations
1332 * are handled here: some of them are handled
1333 * by the standard folding rules, and some of
1334 * them (the character class or ANYOF cases)
1335 * are handled during compiletime in
1336 * regexec.c:S_regclass(). */
880bd946
JH
1337 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1338 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1339 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
80aecb99 1340
4cadc6a9 1341 REXEC_FBC_EXACTISH_CHECK(c == c1 || c == c2);
1aa99e6b 1342 }
09091399 1343 }
1aa99e6b
IH
1344 }
1345 else {
a0a388a1 1346 /* Neither pattern nor string are UTF8 */
1aa99e6b 1347 if (c1 == c2)
4cadc6a9 1348 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1);
1aa99e6b 1349 else
4cadc6a9 1350 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2);
b3c9acc1
IZ
1351 }
1352 break;
bbce6d69 1353 case BOUNDL:
3280af22 1354 PL_reg_flags |= RF_tainted;
bbce6d69 1355 /* FALL THROUGH */
a0d0e21e 1356 case BOUND:
ffc61ed2 1357 if (do_utf8) {
12d33761 1358 if (s == PL_bostr)
ffc61ed2
JH
1359 tmp = '\n';
1360 else {
6136c704 1361 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
4ad0818d 1362 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
ffc61ed2
JH
1363 }
1364 tmp = ((OP(c) == BOUND ?
9041c2e3 1365 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1a4fad37 1366 LOAD_UTF8_CHARCLASS_ALNUM();
4cadc6a9 1367 REXEC_FBC_UTF8_SCAN(
ffc61ed2 1368 if (tmp == !(OP(c) == BOUND ?
bb7a0f54 1369 (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
ffc61ed2
JH
1370 isALNUM_LC_utf8((U8*)s)))
1371 {
1372 tmp = !tmp;
4cadc6a9 1373 REXEC_FBC_TRYIT;
a687059c 1374 }
4cadc6a9 1375 );
a0d0e21e 1376 }
667bb95a 1377 else {
12d33761 1378 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
ffc61ed2 1379 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
4cadc6a9 1380 REXEC_FBC_SCAN(
ffc61ed2
JH
1381 if (tmp ==
1382 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1383 tmp = !tmp;
4cadc6a9 1384 REXEC_FBC_TRYIT;
a0ed51b3 1385 }
4cadc6a9 1386 );
a0ed51b3 1387 }
24b23f37 1388 if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s)))
a0ed51b3
LW
1389 goto got_it;
1390 break;
bbce6d69 1391 case NBOUNDL:
3280af22 1392 PL_reg_flags |= RF_tainted;
bbce6d69 1393 /* FALL THROUGH */
a0d0e21e 1394 case NBOUND:
ffc61ed2 1395 if (do_utf8) {
12d33761 1396 if (s == PL_bostr)
ffc61ed2
JH
1397 tmp = '\n';
1398 else {
6136c704 1399 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
4ad0818d 1400 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
ffc61ed2
JH
1401 }
1402 tmp = ((OP(c) == NBOUND ?
9041c2e3 1403 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1a4fad37 1404 LOAD_UTF8_CHARCLASS_ALNUM();
4cadc6a9 1405 REXEC_FBC_UTF8_SCAN(
ffc61ed2 1406 if (tmp == !(OP(c) == NBOUND ?
bb7a0f54 1407 (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
ffc61ed2
JH
1408 isALNUM_LC_utf8((U8*)s)))
1409 tmp = !tmp;
4cadc6a9
YO
1410 else REXEC_FBC_TRYIT;
1411 );
a0d0e21e 1412 }
667bb95a 1413 else {
12d33761 1414 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
ffc61ed2
JH
1415 tmp = ((OP(c) == NBOUND ?
1416 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
4cadc6a9 1417 REXEC_FBC_SCAN(
ffc61ed2
JH
1418 if (tmp ==
1419 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1420 tmp = !tmp;
4cadc6a9
YO
1421 else REXEC_FBC_TRYIT;
1422 );
a0ed51b3 1423 }
24b23f37 1424 if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, &s)))
a0ed51b3
LW
1425 goto got_it;
1426 break;
a0d0e21e 1427 case ALNUM:
4cadc6a9
YO
1428 REXEC_FBC_CSCAN_PRELOAD(
1429 LOAD_UTF8_CHARCLASS_ALNUM(),
1430 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8),
1431 isALNUM(*s)
1432 );
bbce6d69 1433 case ALNUML:
4cadc6a9
YO
1434 REXEC_FBC_CSCAN_TAINT(
1435 isALNUM_LC_utf8((U8*)s),
1436 isALNUM_LC(*s)
1437 );
a0d0e21e 1438 case NALNUM:
4cadc6a9
YO
1439 REXEC_FBC_CSCAN_PRELOAD(
1440 LOAD_UTF8_CHARCLASS_ALNUM(),
1441 !swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8),
1442 !isALNUM(*s)
1443 );
bbce6d69 1444 case NALNUML:
4cadc6a9
YO
1445 REXEC_FBC_CSCAN_TAINT(
1446 !isALNUM_LC_utf8((U8*)s),
1447 !isALNUM_LC(*s)
1448 );
a0d0e21e 1449 case SPACE:
4cadc6a9
YO
1450 REXEC_FBC_CSCAN_PRELOAD(
1451 LOAD_UTF8_CHARCLASS_SPACE(),
1452 *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8),
1453 isSPACE(*s)
1454 );
bbce6d69 1455 case SPACEL:
4cadc6a9
YO
1456 REXEC_FBC_CSCAN_TAINT(
1457 *s == ' ' || isSPACE_LC_utf8((U8*)s),
1458 isSPACE_LC(*s)
1459 );
a0d0e21e 1460 case NSPACE:
4cadc6a9
YO
1461 REXEC_FBC_CSCAN_PRELOAD(
1462 LOAD_UTF8_CHARCLASS_SPACE(),
1463 !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)),
1464 !isSPACE(*s)
1465 );
bbce6d69 1466 case NSPACEL:
4cadc6a9
YO
1467 REXEC_FBC_CSCAN_TAINT(
1468 !(*s == ' ' || isSPACE_LC_utf8((U8*)s)),
1469 !isSPACE_LC(*s)
1470 );
a0d0e21e 1471 case DIGIT:
4cadc6a9
YO
1472 REXEC_FBC_CSCAN_PRELOAD(
1473 LOAD_UTF8_CHARCLASS_DIGIT(),
1474 swash_fetch(PL_utf8_digit,(U8*)s, do_utf8),
1475 isDIGIT(*s)
1476 );
b8c5462f 1477 case DIGITL:
4cadc6a9
YO
1478 REXEC_FBC_CSCAN_TAINT(
1479 isDIGIT_LC_utf8((U8*)s),
1480 isDIGIT_LC(*s)
1481 );
a0d0e21e 1482 case NDIGIT:
4cadc6a9
YO
1483 REXEC_FBC_CSCAN_PRELOAD(
1484 LOAD_UTF8_CHARCLASS_DIGIT(),
1485 !swash_fetch(PL_utf8_digit,(U8*)s, do_utf8),
1486 !isDIGIT(*s)
1487 );
b8c5462f 1488 case NDIGITL:
4cadc6a9
YO
1489 REXEC_FBC_CSCAN_TAINT(
1490 !isDIGIT_LC_utf8((U8*)s),
1491 !isDIGIT_LC(*s)
1492 );
e1d1eefb
YO
1493 case LNBREAK:
1494 REXEC_FBC_CSCAN(
1495 is_LNBREAK_utf8(s),
1496 is_LNBREAK_latin1(s)
1497 );
1498 case VERTWS:
1499 REXEC_FBC_CSCAN(
1500 is_VERTWS_utf8(s),
1501 is_VERTWS_latin1(s)
1502 );
1503 case NVERTWS:
1504 REXEC_FBC_CSCAN(
1505 !is_VERTWS_utf8(s),
1506 !is_VERTWS_latin1(s)
1507 );
1508 case HORIZWS:
1509 REXEC_FBC_CSCAN(
1510 is_HORIZWS_utf8(s),
1511 is_HORIZWS_latin1(s)
1512 );
1513 case NHORIZWS:
1514 REXEC_FBC_CSCAN(
1515 !is_HORIZWS_utf8(s),
1516 !is_HORIZWS_latin1(s)
1517 );
1de06328
YO
1518 case AHOCORASICKC:
1519 case AHOCORASICK:
07be1b83 1520 {
a0a388a1 1521 DECL_TRIE_TYPE(c);
07be1b83
YO
1522 /* what trie are we using right now */
1523 reg_ac_data *aho
f8fc2ecf 1524 = (reg_ac_data*)progi->data->data[ ARG( c ) ];
3251b653
NC
1525 reg_trie_data *trie
1526 = (reg_trie_data*)progi->data->data[ aho->trie ];
85fbaab2 1527 HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]);
07be1b83
YO
1528
1529 const char *last_start = strend - trie->minlen;
6148ee25 1530#ifdef DEBUGGING
07be1b83 1531 const char *real_start = s;
6148ee25 1532#endif
07be1b83 1533 STRLEN maxlen = trie->maxlen;
be8e71aa
YO
1534 SV *sv_points;
1535 U8 **points; /* map of where we were in the input string
786e8c11 1536 when reading a given char. For ASCII this
be8e71aa 1537 is unnecessary overhead as the relationship
38a44b82
NC
1538 is always 1:1, but for Unicode, especially
1539 case folded Unicode this is not true. */
f9e705e8 1540 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
786e8c11
YO
1541 U8 *bitmap=NULL;
1542
07be1b83
YO
1543
1544 GET_RE_DEBUG_FLAGS_DECL;
1545
be8e71aa
YO
1546 /* We can't just allocate points here. We need to wrap it in
1547 * an SV so it gets freed properly if there is a croak while
1548 * running the match */
1549 ENTER;
1550 SAVETMPS;
1551 sv_points=newSV(maxlen * sizeof(U8 *));
1552 SvCUR_set(sv_points,
1553 maxlen * sizeof(U8 *));
1554 SvPOK_on(sv_points);
1555 sv_2mortal(sv_points);
1556 points=(U8**)SvPV_nolen(sv_points );
1de06328
YO
1557 if ( trie_type != trie_utf8_fold
1558 && (trie->bitmap || OP(c)==AHOCORASICKC) )
1559 {
786e8c11
YO
1560 if (trie->bitmap)
1561 bitmap=(U8*)trie->bitmap;
1562 else
1563 bitmap=(U8*)ANYOF_BITMAP(c);
07be1b83 1564 }
786e8c11
YO
1565 /* this is the Aho-Corasick algorithm modified a touch
1566 to include special handling for long "unknown char"
1567 sequences. The basic idea being that we use AC as long
1568 as we are dealing with a possible matching char, when
1569 we encounter an unknown char (and we have not encountered
1570 an accepting state) we scan forward until we find a legal
1571 starting char.
1572 AC matching is basically that of trie matching, except
1573 that when we encounter a failing transition, we fall back
1574 to the current states "fail state", and try the current char
1575 again, a process we repeat until we reach the root state,
1576 state 1, or a legal transition. If we fail on the root state
1577 then we can either terminate if we have reached an accepting
1578 state previously, or restart the entire process from the beginning
1579 if we have not.
1580
1581 */
07be1b83
YO
1582 while (s <= last_start) {
1583 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1584 U8 *uc = (U8*)s;
1585 U16 charid = 0;
1586 U32 base = 1;
1587 U32 state = 1;
1588 UV uvc = 0;
1589 STRLEN len = 0;
1590 STRLEN foldlen = 0;
1591 U8 *uscan = (U8*)NULL;
1592 U8 *leftmost = NULL;
786e8c11
YO
1593#ifdef DEBUGGING
1594 U32 accepted_word= 0;
1595#endif
07be1b83
YO
1596 U32 pointpos = 0;
1597
1598 while ( state && uc <= (U8*)strend ) {
1599 int failed=0;
786e8c11
YO
1600 U32 word = aho->states[ state ].wordnum;
1601
1de06328
YO
1602 if( state==1 ) {
1603 if ( bitmap ) {
1604 DEBUG_TRIE_EXECUTE_r(
1605 if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1606 dump_exec_pos( (char *)uc, c, strend, real_start,
1607 (char *)uc, do_utf8 );
1608 PerlIO_printf( Perl_debug_log,
1609 " Scanning for legal start char...\n");
1610 }
1611 );
1612 while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1613 uc++;
786e8c11 1614 }
1de06328 1615 s= (char *)uc;
786e8c11 1616 }
786e8c11
YO
1617 if (uc >(U8*)last_start) break;
1618 }
1619
1620 if ( word ) {
1621 U8 *lpos= points[ (pointpos - trie->wordlen[word-1] ) % maxlen ];
1622 if (!leftmost || lpos < leftmost) {
1623 DEBUG_r(accepted_word=word);
07be1b83 1624 leftmost= lpos;
786e8c11 1625 }
07be1b83 1626 if (base==0) break;
786e8c11 1627
07be1b83
YO
1628 }
1629 points[pointpos++ % maxlen]= uc;
55eed653
NC
1630 REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc,
1631 uscan, len, uvc, charid, foldlen,
1632 foldbuf, uniflags);
786e8c11
YO
1633 DEBUG_TRIE_EXECUTE_r({
1634 dump_exec_pos( (char *)uc, c, strend, real_start,
1635 s, do_utf8 );
07be1b83 1636 PerlIO_printf(Perl_debug_log,
786e8c11
YO
1637 " Charid:%3u CP:%4"UVxf" ",
1638 charid, uvc);
1639 });
07be1b83
YO
1640
1641 do {
6148ee25 1642#ifdef DEBUGGING
786e8c11 1643 word = aho->states[ state ].wordnum;
6148ee25 1644#endif
07be1b83
YO
1645 base = aho->states[ state ].trans.base;
1646
786e8c11
YO
1647 DEBUG_TRIE_EXECUTE_r({
1648 if (failed)
1649 dump_exec_pos( (char *)uc, c, strend, real_start,
1650 s, do_utf8 );
07be1b83 1651 PerlIO_printf( Perl_debug_log,
786e8c11
YO
1652 "%sState: %4"UVxf", word=%"UVxf,
1653 failed ? " Fail transition to " : "",
1654 (UV)state, (UV)word);
1655 });
07be1b83
YO
1656 if ( base ) {
1657 U32 tmp;
1658 if (charid &&
1659 (base + charid > trie->uniquecharcount )
1660 && (base + charid - 1 - trie->uniquecharcount
1661 < trie->lasttrans)
1662 && trie->trans[base + charid - 1 -
1663 trie->uniquecharcount].check == state
1664 && (tmp=trie->trans[base + charid - 1 -
1665 trie->uniquecharcount ].next))
1666 {
786e8c11
YO
1667 DEBUG_TRIE_EXECUTE_r(
1668 PerlIO_printf( Perl_debug_log," - legal\n"));
07be1b83
YO
1669 state = tmp;
1670 break;
1671 }
1672 else {
786e8c11
YO
1673 DEBUG_TRIE_EXECUTE_r(
1674 PerlIO_printf( Perl_debug_log," - fail\n"));
1675 failed = 1;
1676 state = aho->fail[state];
07be1b83
YO
1677 }
1678 }
1679 else {
1680 /* we must be accepting here */
786e8c11
YO
1681 DEBUG_TRIE_EXECUTE_r(
1682 PerlIO_printf( Perl_debug_log," - accepting\n"));
1683 failed = 1;
07be1b83
YO
1684 break;
1685 }
1686 } while(state);
786e8c11 1687 uc += len;
07be1b83
YO
1688 if (failed) {
1689 if (leftmost)
1690 break;
786e8c11 1691 if (!state) state = 1;
07be1b83
YO
1692 }
1693 }
1694 if ( aho->states[ state ].wordnum ) {
1695 U8 *lpos = points[ (pointpos - trie->wordlen[aho->states[ state ].wordnum-1]) % maxlen ];
786e8c11
YO
1696 if (!leftmost || lpos < leftmost) {
1697 DEBUG_r(accepted_word=aho->states[ state ].wordnum);
07be1b83 1698 leftmost = lpos;
786e8c11 1699 }
07be1b83 1700 }
07be1b83
YO
1701 if (leftmost) {
1702 s = (char*)leftmost;
786e8c11
YO
1703 DEBUG_TRIE_EXECUTE_r({
1704 PerlIO_printf(
70685ca0
JH
1705 Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n",
1706 (UV)accepted_word, (IV)(s - real_start)
786e8c11
YO
1707 );
1708 });
24b23f37 1709 if (!reginfo || regtry(reginfo, &s)) {
be8e71aa
YO
1710 FREETMPS;
1711 LEAVE;
07be1b83 1712 goto got_it;
be8e71aa 1713 }
07be1b83 1714 s = HOPc(s,1);
786e8c11
YO
1715 DEBUG_TRIE_EXECUTE_r({
1716 PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n");
1717 });
07be1b83 1718 } else {
786e8c11
YO
1719 DEBUG_TRIE_EXECUTE_r(
1720 PerlIO_printf( Perl_debug_log,"No match.\n"));
07be1b83
YO
1721 break;
1722 }
1723 }
be8e71aa
YO
1724 FREETMPS;
1725 LEAVE;
07be1b83
YO
1726 }
1727 break;
b3c9acc1 1728 default:
3c3eec57
GS
1729 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1730 break;
d6a28714 1731 }
6eb5f6b9
JH
1732 return 0;
1733 got_it:
1734 return s;
1735}
1736
e6aad7ab 1737static void
7918f24d
NC
1738S_swap_match_buff (pTHX_ regexp *prog)
1739{
f0ab9afb 1740 regexp_paren_pair *t;
fae667d5 1741
7918f24d
NC
1742 PERL_ARGS_ASSERT_SWAP_MATCH_BUFF;
1743
28d8d7f4 1744 if (!prog->swap) {
fae667d5
YO
1745 /* We have to be careful. If the previous successful match
1746 was from this regex we don't want a subsequent paritally
1747 successful match to clobber the old results.
1748 So when we detect this possibility we add a swap buffer
1749 to the re, and switch the buffer each match. If we fail
1750 we switch it back, otherwise we leave it swapped.
1751 */
f0ab9afb 1752 Newxz(prog->swap, (prog->nparens + 1), regexp_paren_pair);
fae667d5 1753 }
f0ab9afb
NC
1754 t = prog->swap;
1755 prog->swap = prog->offs;
1756 prog->offs = t;
fae667d5
YO
1757}
1758
1759
6eb5f6b9
JH
1760/*
1761 - regexec_flags - match a regexp against a string
1762 */
1763I32
288b8c02 1764Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend,
6eb5f6b9
JH
1765 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1766/* strend: pointer to null at end of string */
1767/* strbeg: real beginning of string */
1768/* minend: end of match must be >=minend after stringarg. */
58e23c8d
YO
1769/* data: May be used for some additional optimizations.
1770 Currently its only used, with a U32 cast, for transmitting
1771 the ganch offset when doing a /g match. This will change */
6eb5f6b9
JH
1772/* nosave: For optimizations. */
1773{
97aff369 1774 dVAR;
288b8c02 1775 struct regexp *const prog = (struct regexp *)SvANY(rx);
24b23f37 1776 /*register*/ char *s;
6eb5f6b9 1777 register regnode *c;
24b23f37 1778 /*register*/ char *startpos = stringarg;
6eb5f6b9
JH
1779 I32 minlen; /* must match at least this many chars */
1780 I32 dontbother = 0; /* how many characters not to try at end */
6eb5f6b9
JH
1781 I32 end_shift = 0; /* Same for the end. */ /* CC */
1782 I32 scream_pos = -1; /* Internal iterator of scream. */
ccac19ea 1783 char *scream_olds = NULL;
f9f4320a 1784 const bool do_utf8 = (bool)DO_UTF8(sv);
2757e526 1785 I32 multiline;
f8fc2ecf 1786 RXi_GET_DECL(prog,progi);
3b0527fe 1787 regmatch_info reginfo; /* create some info to pass to regtry etc */
fae667d5 1788 bool swap_on_fail = 0;
a3621e74
YO
1789 GET_RE_DEBUG_FLAGS_DECL;
1790
7918f24d 1791 PERL_ARGS_ASSERT_REGEXEC_FLAGS;
9d4ba2ae 1792 PERL_UNUSED_ARG(data);
6eb5f6b9
JH
1793
1794 /* Be paranoid... */
1795 if (prog == NULL || startpos == NULL) {
1796 Perl_croak(aTHX_ "NULL regexp parameter");
1797 return 0;
1798 }
1799
bbe252da 1800 multiline = prog->extflags & RXf_PMf_MULTILINE;
288b8c02 1801 reginfo.prog = rx; /* Yes, sorry that this is confusing. */
2757e526 1802
288b8c02 1803 RX_MATCH_UTF8_set(rx, do_utf8);
1de06328 1804 DEBUG_EXECUTE_r(
efd26800 1805 debug_start_match(rx, do_utf8, startpos, strend,
1de06328
YO
1806 "Matching");
1807 );
bac06658 1808
6eb5f6b9 1809 minlen = prog->minlen;
1de06328
YO
1810
1811 if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) {
a3621e74 1812 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
a72c7584
JH
1813 "String too short [regexec_flags]...\n"));
1814 goto phooey;
1aa99e6b 1815 }
6eb5f6b9 1816
1de06328 1817
6eb5f6b9 1818 /* Check validity of program. */
f8fc2ecf 1819 if (UCHARAT(progi->program) != REG_MAGIC) {
6eb5f6b9
JH
1820 Perl_croak(aTHX_ "corrupted regexp program");
1821 }
1822
1823 PL_reg_flags = 0;
1824 PL_reg_eval_set = 0;
1825 PL_reg_maxiter = 0;
1826
3c8556c3 1827 if (RX_UTF8(rx))
6eb5f6b9
JH
1828 PL_reg_flags |= RF_utf8;
1829
1830 /* Mark beginning of line for ^ and lookbehind. */
3b0527fe 1831 reginfo.bol = startpos; /* XXX not used ??? */
6eb5f6b9 1832 PL_bostr = strbeg;
3b0527fe 1833 reginfo.sv = sv;
6eb5f6b9
JH
1834
1835 /* Mark end of line for $ (and such) */
1836 PL_regeol = strend;
1837
1838 /* see how far we have to get to not match where we matched before */
3b0527fe 1839 reginfo.till = startpos+minend;
6eb5f6b9 1840
6eb5f6b9
JH
1841 /* If there is a "must appear" string, look for it. */
1842 s = startpos;
1843
bbe252da 1844 if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */
6eb5f6b9
JH
1845 MAGIC *mg;
1846
1847 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
58e23c8d 1848 reginfo.ganch = startpos + prog->gofs;
6eb5f6b9
JH
1849 else if (sv && SvTYPE(sv) >= SVt_PVMG
1850 && SvMAGIC(sv)
14befaf4
DM
1851 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1852 && mg->mg_len >= 0) {
3b0527fe 1853 reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */
bbe252da 1854 if (prog->extflags & RXf_ANCH_GPOS) {
3b0527fe 1855 if (s > reginfo.ganch)
6eb5f6b9 1856 goto phooey;
58e23c8d 1857 s = reginfo.ganch - prog->gofs;
6eb5f6b9
JH
1858 }
1859 }
58e23c8d 1860 else if (data) {
70685ca0 1861 reginfo.ganch = strbeg + PTR2UV(data);
58e23c8d 1862 } else /* pos() not defined */
3b0527fe 1863 reginfo.ganch = strbeg;
6eb5f6b9 1864 }
288b8c02 1865 if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) {
fae667d5
YO
1866 swap_on_fail = 1;
1867 swap_match_buff(prog); /* do we need a save destructor here for
1868 eval dies? */
c74340f9 1869 }
a0714e2c 1870 if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
6eb5f6b9
JH
1871 re_scream_pos_data d;
1872
1873 d.scream_olds = &scream_olds;
1874 d.scream_pos = &scream_pos;
288b8c02 1875 s = re_intuit_start(rx, sv, s, strend, flags, &d);
3fa9c3d7 1876 if (!s) {
a3621e74 1877 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
6eb5f6b9 1878 goto phooey; /* not present */
3fa9c3d7 1879 }
6eb5f6b9
JH
1880 }
1881
1de06328 1882
6eb5f6b9
JH
1883
1884 /* Simplest case: anchored match need be tried only once. */
1885 /* [unless only anchor is BOL and multiline is set] */
bbe252da 1886 if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) {
24b23f37 1887 if (s == startpos && regtry(&reginfo, &startpos))
6eb5f6b9 1888 goto got_it;
bbe252da
YO
1889 else if (multiline || (prog->intflags & PREGf_IMPLICIT)
1890 || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */
6eb5f6b9
JH
1891 {
1892 char *end;
1893
1894 if (minlen)
1895 dontbother = minlen - 1;
1aa99e6b 1896 end = HOP3c(strend, -dontbother, strbeg) - 1;
6eb5f6b9 1897 /* for multiline we only have to try after newlines */
33b8afdf 1898 if (prog->check_substr || prog->check_utf8) {
6eb5f6b9
JH
1899 if (s == startpos)
1900 goto after_try;
1901 while (1) {
24b23f37 1902 if (regtry(&reginfo, &s))
6eb5f6b9
JH
1903 goto got_it;
1904 after_try:
5339e136 1905 if (s > end)
6eb5f6b9 1906 goto phooey;
bbe252da 1907 if (prog->extflags & RXf_USE_INTUIT) {
288b8c02 1908 s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL);
6eb5f6b9
JH
1909 if (!s)
1910 goto phooey;
1911 }
1912 else
1913 s++;
1914 }
1915 } else {
1916 if (s > startpos)
1917 s--;
1918 while (s < end) {
1919 if (*s++ == '\n') { /* don't need PL_utf8skip here */
24b23f37 1920 if (regtry(&reginfo, &s))
6eb5f6b9
JH
1921 goto got_it;
1922 }
1923 }
1924 }
1925 }
1926 goto phooey;
bbe252da 1927 } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK))
f9f4320a
YO
1928 {
1929 /* the warning about reginfo.ganch being used without intialization
bbe252da 1930 is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN
f9f4320a 1931 and we only enter this block when the same bit is set. */
58e23c8d
YO
1932 char *tmp_s = reginfo.ganch - prog->gofs;
1933 if (regtry(&reginfo, &tmp_s))
6eb5f6b9
JH
1934 goto got_it;
1935 goto phooey;
1936 }
1937
1938 /* Messy cases: unanchored match. */
bbe252da 1939 if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) {
6eb5f6b9
JH
1940 /* we have /x+whatever/ */
1941 /* it must be a one character string (XXXX Except UTF?) */
33b8afdf 1942 char ch;
bf93d4cc
GS
1943#ifdef DEBUGGING
1944 int did_match = 0;
1945#endif
33b8afdf
JH
1946 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1947 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
3f7c398e 1948 ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
bf93d4cc 1949
1aa99e6b 1950 if (do_utf8) {
4cadc6a9 1951 REXEC_FBC_SCAN(
6eb5f6b9 1952 if (*s == ch) {
a3621e74 1953 DEBUG_EXECUTE_r( did_match = 1 );
24b23f37 1954 if (regtry(&reginfo, &s)) goto got_it;
6eb5f6b9
JH
1955 s += UTF8SKIP(s);
1956 while (s < strend && *s == ch)
1957 s += UTF8SKIP(s);
1958 }
4cadc6a9 1959 );
6eb5f6b9
JH
1960 }
1961 else {
4cadc6a9 1962 REXEC_FBC_SCAN(
6eb5f6b9 1963 if (*s == ch) {
a3621e74 1964 DEBUG_EXECUTE_r( did_match = 1 );
24b23f37 1965 if (regtry(&reginfo, &s)) goto got_it;
6eb5f6b9
JH
1966 s++;
1967 while (s < strend && *s == ch)
1968 s++;
1969 }
4cadc6a9 1970 );
6eb5f6b9 1971 }
a3621e74 1972 DEBUG_EXECUTE_r(if (!did_match)
bf93d4cc 1973 PerlIO_printf(Perl_debug_log,
b7953727
JH
1974 "Did not find anchored character...\n")
1975 );
6eb5f6b9 1976 }
a0714e2c
SS
1977 else if (prog->anchored_substr != NULL
1978 || prog->anchored_utf8 != NULL
1979 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
33b8afdf
JH
1980 && prog->float_max_offset < strend - s)) {
1981 SV *must;
1982 I32 back_max;
1983 I32 back_min;
1984 char *last;
6eb5f6b9 1985 char *last1; /* Last position checked before */
bf93d4cc
GS
1986#ifdef DEBUGGING
1987 int did_match = 0;
1988#endif
33b8afdf
JH
1989 if (prog->anchored_substr || prog->anchored_utf8) {
1990 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1991 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1992 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1993 back_max = back_min = prog->anchored_offset;
1994 } else {
1995 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1996 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1997 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1998 back_max = prog->float_max_offset;
1999 back_min = prog->float_min_offset;
2000 }
1de06328
YO
2001
2002
33b8afdf
JH
2003 if (must == &PL_sv_undef)
2004 /* could not downgrade utf8 check substring, so must fail */
2005 goto phooey;
2006
1de06328
YO
2007 if (back_min<0) {
2008 last = strend;
2009 } else {
2010 last = HOP3c(strend, /* Cannot start after this */
2011 -(I32)(CHR_SVLEN(must)
2012 - (SvTAIL(must) != 0) + back_min), strbeg);
2013 }
6eb5f6b9
JH
2014 if (s > PL_bostr)
2015 last1 = HOPc(s, -1);
2016 else
2017 last1 = s - 1; /* bogus */
2018
a0288114 2019 /* XXXX check_substr already used to find "s", can optimize if
6eb5f6b9
JH
2020 check_substr==must. */
2021 scream_pos = -1;
2022 dontbother = end_shift;
2023 strend = HOPc(strend, -dontbother);
2024 while ( (s <= last) &&
9041c2e3 2025 ((flags & REXEC_SCREAM)
1de06328 2026 ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg,
6eb5f6b9 2027 end_shift, &scream_pos, 0))
1de06328 2028 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)),
9041c2e3 2029 (unsigned char*)strend, must,
7fba1cd6 2030 multiline ? FBMrf_MULTILINE : 0))) ) {
4addbd3b 2031 /* we may be pointing at the wrong string */
07bc277f 2032 if ((flags & REXEC_SCREAM) && RXp_MATCH_COPIED(prog))
3f7c398e 2033 s = strbeg + (s - SvPVX_const(sv));
a3621e74 2034 DEBUG_EXECUTE_r( did_match = 1 );
6eb5f6b9
JH
2035 if (HOPc(s, -back_max) > last1) {
2036 last1 = HOPc(s, -back_min);
2037 s = HOPc(s, -back_max);
2038 }
2039 else {
52657f30 2040 char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
6eb5f6b9
JH
2041
2042 last1 = HOPc(s, -back_min);
52657f30 2043 s = t;
6eb5f6b9 2044 }
1aa99e6b 2045 if (do_utf8) {
6eb5f6b9 2046 while (s <= last1) {
24b23f37 2047 if (regtry(&reginfo, &s))
6eb5f6b9
JH
2048 goto got_it;
2049 s += UTF8SKIP(s);
2050 }
2051 }
2052 else {
2053 while (s <= last1) {
24b23f37 2054 if (regtry(&reginfo, &s))
6eb5f6b9
JH
2055 goto got_it;
2056 s++;
2057 }
2058 }
2059 }
ab3bbdeb
YO
2060 DEBUG_EXECUTE_r(if (!did_match) {
2061 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
2062 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
2063 PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n",
33b8afdf 2064 ((must == prog->anchored_substr || must == prog->anchored_utf8)
bf93d4cc 2065 ? "anchored" : "floating"),
ab3bbdeb
YO
2066 quoted, RE_SV_TAIL(must));
2067 });
6eb5f6b9
JH
2068 goto phooey;
2069 }
f8fc2ecf 2070 else if ( (c = progi->regstclass) ) {
f14c76ed 2071 if (minlen) {
f8fc2ecf 2072 const OPCODE op = OP(progi->regstclass);
66e933ab 2073 /* don't bother with what can't match */
786e8c11 2074 if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE)
f14c76ed
RGS
2075 strend = HOPc(strend, -(minlen - 1));
2076 }
a3621e74 2077 DEBUG_EXECUTE_r({
be8e71aa 2078 SV * const prop = sv_newmortal();
32fc9b6a 2079 regprop(prog, prop, c);
0df25f3d 2080 {
02daf0ab 2081 RE_PV_QUOTED_DECL(quoted,do_utf8,PERL_DEBUG_PAD_ZERO(1),
ab3bbdeb 2082 s,strend-s,60);
0df25f3d 2083 PerlIO_printf(Perl_debug_log,
ab3bbdeb 2084 "Matching stclass %.*s against %s (%d chars)\n",
e4f74956 2085 (int)SvCUR(prop), SvPVX_const(prop),
ab3bbdeb 2086 quoted, (int)(strend - s));
0df25f3d 2087 }
ffc61ed2 2088 });
3b0527fe 2089 if (find_byclass(prog, c, s, strend, &reginfo))
6eb5f6b9 2090 goto got_it;
07be1b83 2091 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n"));
d6a28714
JH
2092 }
2093 else {
2094 dontbother = 0;
a0714e2c 2095 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
33b8afdf 2096 /* Trim the end. */
d6a28714 2097 char *last;
33b8afdf
JH
2098 SV* float_real;
2099
2100 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
2101 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
2102 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
d6a28714
JH
2103
2104 if (flags & REXEC_SCREAM) {
33b8afdf 2105 last = screaminstr(sv, float_real, s - strbeg,
d6a28714
JH
2106 end_shift, &scream_pos, 1); /* last one */
2107 if (!last)
ffc61ed2 2108 last = scream_olds; /* Only one occurrence. */
4addbd3b 2109 /* we may be pointing at the wrong string */
07bc277f 2110 else if (RXp_MATCH_COPIED(prog))
3f7c398e 2111 s = strbeg + (s - SvPVX_const(sv));
b8c5462f 2112 }
d6a28714
JH
2113 else {
2114 STRLEN len;
cfd0369c 2115 const char * const little = SvPV_const(float_real, len);
d6a28714 2116
33b8afdf 2117 if (SvTAIL(float_real)) {
d6a28714
JH
2118 if (memEQ(strend - len + 1, little, len - 1))
2119 last = strend - len + 1;
7fba1cd6 2120 else if (!multiline)
9041c2e3 2121 last = memEQ(strend - len, little, len)
bd61b366 2122 ? strend - len : NULL;
b8c5462f 2123 else
d6a28714
JH
2124 goto find_last;
2125 } else {
2126 find_last:
9041c2e3 2127 if (len)
d6a28714 2128 last = rninstr(s, strend, little, little + len);
b8c5462f 2129 else
a0288114 2130 last = strend; /* matching "$" */
b8c5462f 2131 }
b8c5462f 2132 }
bf93d4cc 2133 if (last == NULL) {
6bda09f9
YO
2134 DEBUG_EXECUTE_r(
2135 PerlIO_printf(Perl_debug_log,
2136 "%sCan't trim the tail, match fails (should not happen)%s\n",
2137 PL_colors[4], PL_colors[5]));
bf93d4cc
GS
2138 goto phooey; /* Should not happen! */
2139 }
d6a28714
JH
2140 dontbother = strend - last + prog->float_min_offset;
2141 }
2142 if (minlen && (dontbother < minlen))
2143 dontbother = minlen - 1;
2144 strend -= dontbother; /* this one's always in bytes! */
2145 /* We don't know much -- general case. */
1aa99e6b 2146 if (do_utf8) {
d6a28714 2147 for (;;) {
24b23f37 2148 if (regtry(&reginfo, &s))
d6a28714
JH
2149 goto got_it;
2150 if (s >= strend)
2151 break;
b8c5462f 2152 s += UTF8SKIP(s);
d6a28714
JH
2153 };
2154 }
2155 else {
2156 do {
24b23f37 2157 if (regtry(&reginfo, &s))
d6a28714
JH
2158 goto got_it;
2159 } while (s++ < strend);
2160 }
2161 }
2162
2163 /* Failure. */
2164 goto phooey;
2165
2166got_it:
288b8c02 2167 RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted);
d6a28714 2168
19b95bf0 2169 if (PL_reg_eval_set)
4f639d21 2170 restore_pos(aTHX_ prog);
5daac39c
NC
2171 if (RXp_PAREN_NAMES(prog))
2172 (void)hv_iterinit(RXp_PAREN_NAMES(prog));
d6a28714
JH
2173
2174 /* make sure $`, $&, $', and $digit will work later */
2175 if ( !(flags & REXEC_NOT_FIRST) ) {
288b8c02 2176 RX_MATCH_COPY_FREE(rx);
d6a28714 2177 if (flags & REXEC_COPY_STR) {
be8e71aa 2178 const I32 i = PL_regeol - startpos + (stringarg - strbeg);
f8c7b90f 2179#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
2180 if ((SvIsCOW(sv)
2181 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2182 if (DEBUG_C_TEST) {
2183 PerlIO_printf(Perl_debug_log,
2184 "Copy on write: regexp capture, type %d\n",
2185 (int) SvTYPE(sv));
2186 }
2187 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
d5263905 2188 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
ed252734
NC
2189 assert (SvPOKp(prog->saved_copy));
2190 } else
2191#endif
2192 {
288b8c02 2193 RX_MATCH_COPIED_on(rx);
ed252734
NC
2194 s = savepvn(strbeg, i);
2195 prog->subbeg = s;
2196 }
d6a28714 2197 prog->sublen = i;
d6a28714
JH
2198 }
2199 else {
2200 prog->subbeg = strbeg;
2201 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2202 }
2203 }
9041c2e3 2204
d6a28714
JH
2205 return 1;
2206
2207phooey:
a3621e74 2208 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
e4584336 2209 PL_colors[4], PL_colors[5]));
d6a28714 2210 if (PL_reg_eval_set)
4f639d21 2211 restore_pos(aTHX_ prog);
fae667d5 2212 if (swap_on_fail)
c74340f9 2213 /* we failed :-( roll it back */
fae667d5
YO
2214 swap_match_buff(prog);
2215
d6a28714
JH
2216 return 0;
2217}
2218
6bda09f9 2219
d6a28714
JH
2220/*
2221 - regtry - try match at specific point
2222 */
2223STATIC I32 /* 0 failure, 1 success */
24b23f37 2224S_regtry(pTHX_ regmatch_info *reginfo, char **startpos)
d6a28714 2225{
97aff369 2226 dVAR;
d6a28714 2227 CHECKPOINT lastcp;
288b8c02
NC
2228 REGEXP *const rx = reginfo->prog;
2229 regexp *const prog = (struct regexp *)SvANY(rx);
f8fc2ecf 2230 RXi_GET_DECL(prog,progi);
a3621e74 2231 GET_RE_DEBUG_FLAGS_DECL;
7918f24d
NC
2232
2233 PERL_ARGS_ASSERT_REGTRY;
2234
24b23f37 2235 reginfo->cutpoint=NULL;
d6a28714 2236
bbe252da 2237 if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) {
d6a28714
JH
2238 MAGIC *mg;
2239
2240 PL_reg_eval_set = RS_init;
a3621e74 2241 DEBUG_EXECUTE_r(DEBUG_s(
b900a521
JH
2242 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2243 (IV)(PL_stack_sp - PL_stack_base));
d6a28714 2244 ));
ea8d6ae1 2245 SAVESTACK_CXPOS();
d6a28714
JH
2246 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2247 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2248 SAVETMPS;
2249 /* Apparently this is not needed, judging by wantarray. */
e8347627 2250 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
d6a28714
JH
2251 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2252
3b0527fe 2253 if (reginfo->sv) {
d6a28714 2254 /* Make $_ available to executed code. */
3b0527fe 2255 if (reginfo->sv != DEFSV) {
59f00321 2256 SAVE_DEFSV;
414bf5ae 2257 DEFSV_set(reginfo->sv);
b8c5462f 2258 }
d6a28714 2259
3b0527fe
DM
2260 if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv)
2261 && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) {
d6a28714 2262 /* prepare for quick setting of pos */
d300d9fa 2263#ifdef PERL_OLD_COPY_ON_WRITE
51a9ea20
NC
2264 if (SvIsCOW(reginfo->sv))
2265 sv_force_normal_flags(reginfo->sv, 0);
d300d9fa 2266#endif
3dab1dad 2267 mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global,
d300d9fa 2268 &PL_vtbl_mglob, NULL, 0);
d6a28714 2269 mg->mg_len = -1;
b8c5462f 2270 }
d6a28714
JH
2271 PL_reg_magic = mg;
2272 PL_reg_oldpos = mg->mg_len;
4f639d21 2273 SAVEDESTRUCTOR_X(restore_pos, prog);
d6a28714 2274 }
09687e5a 2275 if (!PL_reg_curpm) {
a02a5408 2276 Newxz(PL_reg_curpm, 1, PMOP);
09687e5a
AB
2277#ifdef USE_ITHREADS
2278 {
14a49a24 2279 SV* const repointer = &PL_sv_undef;
92313705
NC
2280 /* this regexp is also owned by the new PL_reg_curpm, which
2281 will try to free it. */
d2ece331 2282 av_push(PL_regex_padav, repointer);
09687e5a
AB
2283 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2284 PL_regex_pad = AvARRAY(PL_regex_padav);
2285 }
2286#endif
2287 }
86c29d75
NC
2288#ifdef USE_ITHREADS
2289 /* It seems that non-ithreads works both with and without this code.
2290 So for efficiency reasons it seems best not to have the code
2291 compiled when it is not needed. */
92313705
NC
2292 /* This is safe against NULLs: */
2293 ReREFCNT_dec(PM_GETRE(PL_reg_curpm));
2294 /* PM_reg_curpm owns a reference to this regexp. */
2295 ReREFCNT_inc(rx);
86c29d75 2296#endif
288b8c02 2297 PM_SETRE(PL_reg_curpm, rx);
d6a28714
JH
2298 PL_reg_oldcurpm = PL_curpm;
2299 PL_curpm = PL_reg_curpm;
07bc277f 2300 if (RXp_MATCH_COPIED(prog)) {
d6a28714
JH
2301 /* Here is a serious problem: we cannot rewrite subbeg,
2302 since it may be needed if this match fails. Thus
2303 $` inside (?{}) could fail... */
2304 PL_reg_oldsaved = prog->subbeg;
2305 PL_reg_oldsavedlen = prog->sublen;
f8c7b90f 2306#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
2307 PL_nrs = prog->saved_copy;
2308#endif
07bc277f 2309 RXp_MATCH_COPIED_off(prog);
d6a28714
JH
2310 }
2311 else
bd61b366 2312 PL_reg_oldsaved = NULL;
d6a28714
JH
2313 prog->subbeg = PL_bostr;
2314 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2315 }
24b23f37 2316 DEBUG_EXECUTE_r(PL_reg_starttry = *startpos);
f0ab9afb 2317 prog->offs[0].start = *startpos - PL_bostr;
24b23f37 2318 PL_reginput = *startpos;
d6a28714 2319 PL_reglastparen = &prog->lastparen;
a01268b5 2320 PL_reglastcloseparen = &prog->lastcloseparen;
d6a28714 2321 prog->lastparen = 0;
03994de8 2322 prog->lastcloseparen = 0;
d6a28714 2323 PL_regsize = 0;
f0ab9afb 2324 PL_regoffs = prog->offs;
d6a28714
JH
2325 if (PL_reg_start_tmpl <= prog->nparens) {
2326 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2327 if(PL_reg_start_tmp)
2328 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2329 else
a02a5408 2330 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
d6a28714
JH
2331 }
2332
2333 /* XXXX What this code is doing here?!!! There should be no need
2334 to do this again and again, PL_reglastparen should take care of
3dd2943c 2335 this! --ilya*/
dafc8851
JH
2336
2337 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2338 * Actually, the code in regcppop() (which Ilya may be meaning by
daf18116 2339 * PL_reglastparen), is not needed at all by the test suite
225593e1
DM
2340 * (op/regexp, op/pat, op/split), but that code is needed otherwise
2341 * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
2342 * Meanwhile, this code *is* needed for the
daf18116
JH
2343 * above-mentioned test suite tests to succeed. The common theme
2344 * on those tests seems to be returning null fields from matches.
225593e1 2345 * --jhi updated by dapm */
dafc8851 2346#if 1
d6a28714 2347 if (prog->nparens) {
f0ab9afb 2348 regexp_paren_pair *pp = PL_regoffs;
097eb12c 2349 register I32 i;
eb160463 2350 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
f0ab9afb
NC
2351 ++pp;
2352 pp->start = -1;
2353 pp->end = -1;
d6a28714
JH
2354 }
2355 }
dafc8851 2356#endif
02db2b7b 2357 REGCP_SET(lastcp);
f8fc2ecf 2358 if (regmatch(reginfo, progi->program + 1)) {
f0ab9afb 2359 PL_regoffs[0].end = PL_reginput - PL_bostr;
d6a28714
JH
2360 return 1;
2361 }
24b23f37
YO
2362 if (reginfo->cutpoint)
2363 *startpos= reginfo->cutpoint;
02db2b7b 2364 REGCP_UNWIND(lastcp);
d6a28714
JH
2365 return 0;
2366}
2367
02db2b7b 2368
8ba1375e
MJD
2369#define sayYES goto yes
2370#define sayNO goto no
262b90c4 2371#define sayNO_SILENT goto no_silent
8ba1375e 2372
f9f4320a
YO
2373/* we dont use STMT_START/END here because it leads to
2374 "unreachable code" warnings, which are bogus, but distracting. */
2375#define CACHEsayNO \
c476f425
DM
2376 if (ST.cache_mask) \
2377 PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \
f9f4320a 2378 sayNO
3298f257 2379
a3621e74 2380/* this is used to determine how far from the left messages like
265c4333
YO
2381 'failed...' are printed. It should be set such that messages
2382 are inline with the regop output that created them.
a3621e74 2383*/
265c4333 2384#define REPORT_CODE_OFF 32
a3621e74
YO
2385
2386
2387/* Make sure there is a test for this +1 options in re_tests */
2388#define TRIE_INITAL_ACCEPT_BUFFLEN 4;
2389
40a82448
DM
2390#define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */
2391#define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */
9e137952 2392
86545054
DM
2393#define SLAB_FIRST(s) (&(s)->states[0])
2394#define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1])
2395
5d9a96ca
DM
2396/* grab a new slab and return the first slot in it */
2397
2398STATIC regmatch_state *
2399S_push_slab(pTHX)
2400{
a35a87e7 2401#if PERL_VERSION < 9 && !defined(PERL_CORE)
54df2634
NC
2402 dMY_CXT;
2403#endif
5d9a96ca
DM
2404 regmatch_slab *s = PL_regmatch_slab->next;
2405 if (!s) {
2406 Newx(s, 1, regmatch_slab);
2407 s->prev = PL_regmatch_slab;
2408 s->next = NULL;
2409 PL_regmatch_slab->next = s;
2410 }
2411 PL_regmatch_slab = s;
86545054 2412 return SLAB_FIRST(s);
5d9a96ca 2413}
5b47454d 2414
95b24440 2415
40a82448
DM
2416/* push a new state then goto it */
2417
2418#define PUSH_STATE_GOTO(state, node) \
2419 scan = node; \
2420 st->resume_state = state; \
2421 goto push_state;
2422
2423/* push a new state with success backtracking, then goto it */
2424
2425#define PUSH_YES_STATE_GOTO(state, node) \
2426 scan = node; \
2427 st->resume_state = state; \
2428 goto push_yes_state;
2429
aa283a38 2430
aa283a38 2431
d6a28714 2432/*
95b24440 2433
bf1f174e
DM
2434regmatch() - main matching routine
2435
2436This is basically one big switch statement in a loop. We execute an op,
2437set 'next' to point the next op, and continue. If we come to a point which
2438we may need to backtrack to on failure such as (A|B|C), we push a
2439backtrack state onto the backtrack stack. On failure, we pop the top
2440state, and re-enter the loop at the state indicated. If there are no more
2441states to pop, we return failure.
2442
2443Sometimes we also need to backtrack on success; for example /A+/, where
2444after successfully matching one A, we need to go back and try to
2445match another one; similarly for lookahead assertions: if the assertion
2446completes successfully, we backtrack to the state just before the assertion
2447and then carry on. In these cases, the pushed state is marked as
2448'backtrack on success too'. This marking is in fact done by a chain of
2449pointers, each pointing to the previous 'yes' state. On success, we pop to
2450the nearest yes state, discarding any intermediate failure-only states.
2451Sometimes a yes state is pushed just to force some cleanup code to be
2452called at the end of a successful match or submatch; e.g. (??{$re}) uses
2453it to free the inner regex.
2454
2455Note that failure backtracking rewinds the cursor position, while
2456success backtracking leaves it alone.
2457
2458A pattern is complete when the END op is executed, while a subpattern
2459such as (?=foo) is complete when the SUCCESS op is executed. Both of these
2460ops trigger the "pop to last yes state if any, otherwise return true"
2461behaviour.
2462
2463A common convention in this function is to use A and B to refer to the two
2464subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is
2465the subpattern to be matched possibly multiple times, while B is the entire
2466rest of the pattern. Variable and state names reflect this convention.
2467
2468The states in the main switch are the union of ops and failure/success of
2469substates associated with with that op. For example, IFMATCH is the op
2470that does lookahead assertions /(?=A)B/ and so the IFMATCH state means
2471'execute IFMATCH'; while IFMATCH_A is a state saying that we have just
2472successfully matched A and IFMATCH_A_fail is a state saying that we have
2473just failed to match A. Resume states always come in pairs. The backtrack
2474state we push is marked as 'IFMATCH_A', but when that is popped, we resume
2475at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking
2476on success or failure.
2477
2478The struct that holds a backtracking state is actually a big union, with
2479one variant for each major type of op. The variable st points to the
2480top-most backtrack struct. To make the code clearer, within each
2481block of code we #define ST to alias the relevant union.
2482
2483Here's a concrete example of a (vastly oversimplified) IFMATCH
2484implementation:
2485
2486 switch (state) {
2487 ....
2488
2489#define ST st->u.ifmatch
2490
2491 case IFMATCH: // we are executing the IFMATCH op, (?=A)B
2492 ST.foo = ...; // some state we wish to save
95b24440 2493 ...
bf1f174e
DM
2494 // push a yes backtrack state with a resume value of
2495 // IFMATCH_A/IFMATCH_A_fail, then continue execution at the
2496 // first node of A:
2497 PUSH_YES_STATE_GOTO(IFMATCH_A, A);
2498 // NOTREACHED
2499
2500 case IFMATCH_A: // we have successfully executed A; now continue with B
2501 next = B;
2502 bar = ST.foo; // do something with the preserved value
2503 break;
2504
2505 case IFMATCH_A_fail: // A failed, so the assertion failed
2506 ...; // do some housekeeping, then ...
2507 sayNO; // propagate the failure
2508
2509#undef ST
95b24440 2510
bf1f174e
DM
2511 ...
2512 }
95b24440 2513
bf1f174e
DM
2514For any old-timers reading this who are familiar with the old recursive
2515approach, the code above is equivalent to:
95b24440 2516
bf1f174e
DM
2517 case IFMATCH: // we are executing the IFMATCH op, (?=A)B
2518 {
2519 int foo = ...
95b24440 2520 ...
bf1f174e
DM
2521 if (regmatch(A)) {
2522 next = B;
2523 bar = foo;
2524 break;
95b24440 2525 }
bf1f174e
DM
2526 ...; // do some housekeeping, then ...
2527 sayNO; // propagate the failure
95b24440 2528 }
bf1f174e
DM
2529
2530The topmost backtrack state, pointed to by st, is usually free. If you
2531want to claim it, populate any ST.foo fields in it with values you wish to
2532save, then do one of
2533
2534 PUSH_STATE_GOTO(resume_state, node);
2535 PUSH_YES_STATE_GOTO(resume_state, node);
2536
2537which sets that backtrack state's resume value to 'resume_state', pushes a
2538new free entry to the top of the backtrack stack, then goes to 'node'.
2539On backtracking, the free slot is popped, and the saved state becomes the
2540new free state. An ST.foo field in this new top state can be temporarily
2541accessed to retrieve values, but once the main loop is re-entered, it
2542becomes available for reuse.
2543
2544Note that the depth of the backtrack stack constantly increases during the
2545left-to-right execution of the pattern, rather than going up and down with
2546the pattern nesting. For example the stack is at its maximum at Z at the
2547end of the pattern, rather than at X in the following:
2548
2549 /(((X)+)+)+....(Y)+....Z/
2550
2551The only exceptions to this are lookahead/behind assertions and the cut,
2552(?>A), which pop all the backtrack states associated with A before
2553continuing.
2554
2555Bascktrack state structs are allocated in slabs of about 4K in size.
2556PL_regmatch_state and st always point to the currently active state,
2557and PL_regmatch_slab points to the slab currently containing
2558PL_regmatch_state. The first time regmatch() is called, the first slab is
2559allocated, and is never freed until interpreter destruction. When the slab
2560is full, a new one is allocated and chained to the end. At exit from
2561regmatch(), slabs allocated since entry are freed.
2562
2563*/
95b24440 2564
40a82448 2565
5bc10b2c 2566#define DEBUG_STATE_pp(pp) \
265c4333 2567 DEBUG_STATE_r({ \
5bc10b2c
DM
2568 DUMP_EXEC_POS(locinput, scan, do_utf8); \
2569 PerlIO_printf(Perl_debug_log, \
5d458dd8 2570 " %*s"pp" %s%s%s%s%s\n", \
5bc10b2c 2571 depth*2, "", \
13d6edb4 2572 PL_reg_name[st->resume_state], \
5d458dd8
YO
2573 ((st==yes_state||st==mark_state) ? "[" : ""), \
2574 ((st==yes_state) ? "Y" : ""), \
2575 ((st==mark_state) ? "M" : ""), \
2576 ((st==yes_state||st==mark_state) ? "]" : "") \
2577 ); \
265c4333 2578 });
5bc10b2c 2579
40a82448 2580
3dab1dad 2581#define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1)
95b24440 2582
3df15adc 2583#ifdef DEBUGGING
5bc10b2c 2584
ab3bbdeb 2585STATIC void
efd26800 2586S_debug_start_match(pTHX_ const REGEXP *prog, const bool do_utf8,
ab3bbdeb
YO
2587 const char *start, const char *end, const char *blurb)
2588{
efd26800 2589 const bool utf8_pat = RX_UTF8(prog) ? 1 : 0;
7918f24d
NC
2590
2591 PERL_ARGS_ASSERT_DEBUG_START_MATCH;
2592
ab3bbdeb
YO
2593 if (!PL_colorset)
2594 reginitcolors();
2595 {
2596 RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0),
d2c6dc5e 2597 RX_PRECOMP_const(prog), RX_PRELEN(prog), 60);
ab3bbdeb
YO
2598
2599 RE_PV_QUOTED_DECL(s1, do_utf8, PERL_DEBUG_PAD_ZERO(1),
2600 start, end - start, 60);
2601
2602 PerlIO_printf(Perl_debug_log,
2603 "%s%s REx%s %s against %s\n",
2604 PL_colors[4], blurb, PL_colors[5], s0, s1);
2605
2606 if (do_utf8||utf8_pat)
1de06328
YO
2607 PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n",
2608 utf8_pat ? "pattern" : "",
2609 utf8_pat && do_utf8 ? " and " : "",
2610 do_utf8 ? "string" : ""
ab3bbdeb
YO
2611 );
2612 }
2613}
3df15adc
YO
2614
2615STATIC void
786e8c11
YO
2616S_dump_exec_pos(pTHX_ const char *locinput,
2617 const regnode *scan,
2618 const char *loc_regeol,
2619 const char *loc_bostr,
2620 const char *loc_reg_starttry,
2621 const bool do_utf8)
07be1b83 2622{
786e8c11 2623 const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4];
07be1b83 2624 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
786e8c11 2625 int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput);
07be1b83
YO
2626 /* The part of the string before starttry has one color
2627 (pref0_len chars), between starttry and current
2628 position another one (pref_len - pref0_len chars),
2629 after the current position the third one.
2630 We assume that pref0_len <= pref_len, otherwise we
2631 decrease pref0_len. */
786e8c11
YO
2632 int pref_len = (locinput - loc_bostr) > (5 + taill) - l
2633 ? (5 + taill) - l : locinput - loc_bostr;
07be1b83
YO
2634 int pref0_len;
2635
7918f24d
NC
2636 PERL_ARGS_ASSERT_DUMP_EXEC_POS;
2637
07be1b83
YO
2638 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2639 pref_len++;
786e8c11
YO
2640 pref0_len = pref_len - (locinput - loc_reg_starttry);
2641 if (l + pref_len < (5 + taill) && l < loc_regeol - locinput)
2642 l = ( loc_regeol - locinput > (5 + taill) - pref_len
2643 ? (5 + taill) - pref_len : loc_regeol - locinput);
07be1b83
YO
2644 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2645 l--;
2646 if (pref0_len < 0)
2647 pref0_len = 0;
2648 if (pref0_len > pref_len)
2649 pref0_len = pref_len;
2650 {
3df15adc 2651 const int is_uni = (do_utf8 && OP(scan) != CANY) ? 1 : 0;
0df25f3d 2652
ab3bbdeb 2653 RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0),
1de06328 2654 (locinput - pref_len),pref0_len, 60, 4, 5);
0df25f3d 2655
ab3bbdeb 2656 RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1),
3df15adc 2657 (locinput - pref_len + pref0_len),
1de06328 2658 pref_len - pref0_len, 60, 2, 3);
0df25f3d 2659
ab3bbdeb 2660 RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2),
1de06328 2661 locinput, loc_regeol - locinput, 10, 0, 1);
0df25f3d 2662
1de06328 2663 const STRLEN tlen=len0+len1+len2;
3df15adc 2664 PerlIO_printf(Perl_debug_log,
ab3bbdeb 2665 "%4"IVdf" <%.*s%.*s%s%.*s>%*s|",
786e8c11 2666 (IV)(locinput - loc_bostr),
07be1b83 2667 len0, s0,
07be1b83 2668 len1, s1,
07be1b83 2669 (docolor ? "" : "> <"),
07be1b83 2670 len2, s2,
f9f4320a 2671 (int)(tlen > 19 ? 0 : 19 - tlen),
07be1b83
YO
2672 "");
2673 }
2674}
3df15adc 2675
07be1b83
YO
2676#endif
2677
0a4db386
YO
2678/* reg_check_named_buff_matched()
2679 * Checks to see if a named buffer has matched. The data array of
2680 * buffer numbers corresponding to the buffer is expected to reside
2681 * in the regexp->data->data array in the slot stored in the ARG() of
2682 * node involved. Note that this routine doesn't actually care about the
2683 * name, that information is not preserved from compilation to execution.
2684 * Returns the index of the leftmost defined buffer with the given name
2685 * or 0 if non of the buffers matched.
2686 */
2687STATIC I32
7918f24d
NC
2688S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan)
2689{
0a4db386 2690 I32 n;
f8fc2ecf 2691 RXi_GET_DECL(rex,rexi);
ad64d0ec 2692 SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
0a4db386 2693 I32 *nums=(I32*)SvPVX(sv_dat);
7918f24d
NC
2694
2695 PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED;
2696
0a4db386
YO
2697 for ( n=0; n<SvIVX(sv_dat); n++ ) {
2698 if ((I32)*PL_reglastparen >= nums[n] &&
f0ab9afb 2699 PL_regoffs[nums[n]].end != -1)
0a4db386
YO
2700 {
2701 return nums[n];
2702 }
2703 }
2704 return 0;
2705}
2706
2f554ef7
DM
2707
2708/* free all slabs above current one - called during LEAVE_SCOPE */
2709
2710STATIC void
2711S_clear_backtrack_stack(pTHX_ void *p)
2712{
2713 regmatch_slab *s = PL_regmatch_slab->next;
2714 PERL_UNUSED_ARG(p);
2715
2716 if (!s)
2717 return;
2718 PL_regmatch_slab->next = NULL;
2719 while (s) {
2720 regmatch_slab * const osl = s;
2721 s = s->next;
2722 Safefree(osl);
2723 }
2724}
2725
2726
28d8d7f4
YO
2727#define SETREX(Re1,Re2) \
2728 if (PL_reg_eval_set) PM_SETRE((PL_reg_curpm), (Re2)); \
2729 Re1 = (Re2)
2730
d6a28714 2731STATIC I32 /* 0 failure, 1 success */
24b23f37 2732S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog)
d6a28714 2733{
a35a87e7 2734#if PERL_VERSION < 9 && !defined(PERL_CORE)
54df2634
NC
2735 dMY_CXT;
2736#endif
27da23d5 2737 dVAR;
95b24440 2738 register const bool do_utf8 = PL_reg_match_utf8;
4ad0818d 2739 const U32 uniflags = UTF8_ALLOW_DEFAULT;
288b8c02
NC
2740 REGEXP *rex_sv = reginfo->prog;
2741 regexp *rex = (struct regexp *)SvANY(rex_sv);
f8fc2ecf 2742 RXi_GET_DECL(rex,rexi);
2f554ef7 2743 I32 oldsave;
5d9a96ca
DM
2744 /* the current state. This is a cached copy of PL_regmatch_state */
2745 register regmatch_state *st;
5d9a96ca
DM
2746 /* cache heavy used fields of st in registers */
2747 register regnode *scan;
2748 register regnode *next;
438e9bae 2749 register U32 n = 0; /* general value; init to avoid compiler warning */
24d3c4a9 2750 register I32 ln = 0; /* len or last; init to avoid compiler warning */
5d9a96ca 2751 register char *locinput = PL_reginput;
5d9a96ca 2752 register I32 nextchr; /* is always set to UCHARAT(locinput) */
24d3c4a9 2753
b69b0499 2754 bool result = 0; /* return value of S_regmatch */
24d3c4a9 2755 int depth = 0; /* depth of backtrack stack */
4b196cd4
YO
2756 U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */
2757 const U32 max_nochange_depth =
2758 (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ?
2759 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH;
77cb431f
DM
2760 regmatch_state *yes_state = NULL; /* state to pop to on success of
2761 subpattern */
e2e6a0f1
YO
2762 /* mark_state piggy backs on the yes_state logic so that when we unwind
2763 the stack on success we can update the mark_state as we go */
2764 regmatch_state *mark_state = NULL; /* last mark state we have seen */
faec1544 2765 regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */
b8591aee 2766 struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */
40a82448 2767 U32 state_num;
5d458dd8
YO
2768 bool no_final = 0; /* prevent failure from backtracking? */
2769 bool do_cutgroup = 0; /* no_final only until next branch/trie entry */
e2e6a0f1 2770 char *startpoint = PL_reginput;
5d458dd8
YO
2771 SV *popmark = NULL; /* are we looking for a mark? */
2772 SV *sv_commit = NULL; /* last mark name seen in failure */
2773 SV *sv_yes_mark = NULL; /* last mark name we have seen
2774 during a successfull match */
2775 U32 lastopen = 0; /* last open we saw */
2776 bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0;
19b95bf0 2777 SV* const oreplsv = GvSV(PL_replgv);
24d3c4a9
DM
2778 /* these three flags are set by various ops to signal information to
2779 * the very next op. They have a useful lifetime of exactly one loop
2780 * iteration, and are not preserved or restored by state pushes/pops
2781 */
2782 bool sw = 0; /* the condition value in (?(cond)a|b) */
2783 bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */
2784 int logical = 0; /* the following EVAL is:
2785 0: (?{...})
2786 1: (?(?{...})X|Y)
2787 2: (??{...})
2788 or the following IFMATCH/UNLESSM is:
2789 false: plain (?=foo)
2790 true: used as a condition: (?(?=foo))
2791 */
95b24440 2792#ifdef DEBUGGING
e68ec53f 2793 GET_RE_DEBUG_FLAGS_DECL;
d6a28714
JH
2794#endif
2795
7918f24d
NC
2796 PERL_ARGS_ASSERT_REGMATCH;
2797
3b57cd43 2798 DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({
24b23f37 2799 PerlIO_printf(Perl_debug_log,"regmatch start\n");
3b57cd43 2800 }));
5d9a96ca
DM
2801 /* on first ever call to regmatch, allocate first slab */
2802 if (!PL_regmatch_slab) {
2803 Newx(PL_regmatch_slab, 1, regmatch_slab);
2804 PL_regmatch_slab->prev = NULL;
2805 PL_regmatch_slab->next = NULL;
86545054 2806 PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab);
5d9a96ca
DM
2807 }
2808
2f554ef7
DM
2809 oldsave = PL_savestack_ix;
2810 SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL);
2811 SAVEVPTR(PL_regmatch_slab);
2812 SAVEVPTR(PL_regmatch_state);
5d9a96ca
DM
2813
2814 /* grab next free state slot */
2815 st = ++PL_regmatch_state;
86545054 2816 if (st > SLAB_LAST(PL_regmatch_slab))
5d9a96ca
DM
2817 st = PL_regmatch_state = S_push_slab(aTHX);
2818
d6a28714
JH
2819 /* Note that nextchr is a byte even in UTF */
2820 nextchr = UCHARAT(locinput);
2821 scan = prog;
2822 while (scan != NULL) {
8ba1375e 2823
a3621e74 2824 DEBUG_EXECUTE_r( {
6136c704 2825 SV * const prop = sv_newmortal();
1de06328 2826 regnode *rnext=regnext(scan);
786e8c11 2827 DUMP_EXEC_POS( locinput, scan, do_utf8 );
32fc9b6a 2828 regprop(rex, prop, scan);
07be1b83
YO
2829
2830 PerlIO_printf(Perl_debug_log,
2831 "%3"IVdf":%*s%s(%"IVdf")\n",
f8fc2ecf 2832 (IV)(scan - rexi->program), depth*2, "",
07be1b83 2833 SvPVX_const(prop),
1de06328 2834 (PL_regkind[OP(scan)] == END || !rnext) ?
f8fc2ecf 2835 0 : (IV)(rnext - rexi->program));
2a782b5b 2836 });
d6a28714
JH
2837
2838 next = scan + NEXT_OFF(scan);
2839 if (next == scan)
2840 next = NULL;
40a82448 2841 state_num = OP(scan);
d6a28714 2842
40a82448 2843 reenter_switch:
34a81e2b
B
2844
2845 assert(PL_reglastparen == &rex->lastparen);
2846 assert(PL_reglastcloseparen == &rex->lastcloseparen);
2847 assert(PL_regoffs == rex->offs);
2848
40a82448 2849 switch (state_num) {
d6a28714 2850 case BOL:
7fba1cd6 2851 if (locinput == PL_bostr)
d6a28714 2852 {
3b0527fe 2853 /* reginfo->till = reginfo->bol; */
b8c5462f
JH
2854 break;
2855 }
d6a28714
JH
2856 sayNO;
2857 case MBOL:
12d33761
HS
2858 if (locinput == PL_bostr ||
2859 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
d6a28714 2860 {
b8c5462f
JH
2861 break;
2862 }
d6a28714
JH
2863 sayNO;
2864 case SBOL:
c2a73568 2865 if (locinput == PL_bostr)
b8c5462f 2866 break;
d6a28714
JH
2867 sayNO;
2868 case GPOS:
3b0527fe 2869 if (locinput == reginfo->ganch)
d6a28714
JH
2870 break;
2871 sayNO;
ee9b8eae
YO
2872
2873 case KEEPS:
2874 /* update the startpoint */
f0ab9afb 2875 st->u.keeper.val = PL_regoffs[0].start;
ee9b8eae 2876 PL_reginput = locinput;
f0ab9afb 2877 PL_regoffs[0].start = locinput - PL_bostr;
ee9b8eae
YO
2878 PUSH_STATE_GOTO(KEEPS_next, next);
2879 /*NOT-REACHED*/
2880 case KEEPS_next_fail:
2881 /* rollback the start point change */
f0ab9afb 2882 PL_regoffs[0].start = st->u.keeper.val;
ee9b8eae
YO
2883 sayNO_SILENT;
2884 /*NOT-REACHED*/
d6a28714 2885 case EOL:
d6a28714
JH
2886 goto seol;
2887 case MEOL:
d6a28714 2888 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
b8c5462f 2889 sayNO;
b8c5462f 2890 break;
d6a28714
JH
2891 case SEOL:
2892 seol:
2893 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
b8c5462f 2894 sayNO;
d6a28714 2895 if (PL_regeol - locinput > 1)
b8c5462f 2896 sayNO;
b8c5462f 2897 break;
d6a28714
JH
2898 case EOS:
2899 if (PL_regeol != locinput)
b8c5462f 2900 sayNO;
d6a28714 2901 break;
ffc61ed2 2902 case SANY:
d6a28714 2903 if (!nextchr && locinput >= PL_regeol)
4633a7c4 2904 sayNO;
f33976b4
DB
2905 if (do_utf8) {
2906 locinput += PL_utf8skip[nextchr];
2907 if (locinput > PL_regeol)
2908 sayNO;
2909 nextchr = UCHARAT(locinput);
2910 }
2911 else
2912 nextchr = UCHARAT(++locinput);
2913 break;
2914 case CANY:
2915 if (!nextchr && locinput >= PL_regeol)
2916 sayNO;
b8c5462f 2917 nextchr = UCHARAT(++locinput);
a0d0e21e 2918 break;
ffc61ed2 2919 case REG_ANY:
1aa99e6b
IH
2920 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2921 sayNO;
2922 if (do_utf8) {
b8c5462f 2923 locinput += PL_utf8skip[nextchr];
d6a28714
JH
2924 if (locinput > PL_regeol)
2925 sayNO;
a0ed51b3 2926 nextchr = UCHARAT(locinput);
a0ed51b3 2927 }
1aa99e6b
IH
2928 else
2929 nextchr = UCHARAT(++locinput);
a0ed51b3 2930 break;
166ba7cd
DM
2931
2932#undef ST
2933#define ST st->u.trie
786e8c11
YO
2934 case TRIEC:
2935 /* In this case the charclass data is available inline so
2936 we can fail fast without a lot of extra overhead.
2937 */
2938 if (scan->flags == EXACT || !do_utf8) {
2939 if(!ANYOF_BITMAP_TEST(scan, *locinput)) {
2940 DEBUG_EXECUTE_r(
2941 PerlIO_printf(Perl_debug_log,
2942 "%*s %sfailed to match trie start class...%s\n",
5bc10b2c 2943 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
786e8c11
YO
2944 );
2945 sayNO_SILENT;
2946 /* NOTREACHED */
2947 }
2948 }
2949 /* FALL THROUGH */
5b47454d 2950 case TRIE:
3dab1dad 2951 {
07be1b83 2952 /* what type of TRIE am I? (utf8 makes this contextual) */
a0a388a1 2953 DECL_TRIE_TYPE(scan);
3dab1dad
YO
2954
2955 /* what trie are we using right now */
be8e71aa 2956 reg_trie_data * const trie
f8fc2ecf 2957 = (reg_trie_data*)rexi->data->data[ ARG( scan ) ];
85fbaab2 2958 HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]);
3dab1dad 2959 U32 state = trie->startstate;
166ba7cd 2960
3dab1dad
YO
2961 if (trie->bitmap && trie_type != trie_utf8_fold &&
2962 !TRIE_BITMAP_TEST(trie,*locinput)
2963 ) {
2964 if (trie->states[ state ].wordnum) {
2965 DEBUG_EXECUTE_r(
2966 PerlIO_printf(Perl_debug_log,
2967 "%*s %smatched empty string...%s\n",
5bc10b2c 2968 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3dab1dad
YO
2969 );
2970 break;
2971 } else {
2972 DEBUG_EXECUTE_r(
2973 PerlIO_printf(Perl_debug_log,
786e8c11 2974 "%*s %sfailed to match trie start class...%s\n",
5bc10b2c 2975 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3dab1dad
YO
2976 );
2977 sayNO_SILENT;
2978 }
2979 }
166ba7cd 2980
786e8c11
YO
2981 {
2982 U8 *uc = ( U8* )locinput;
2983
2984 STRLEN len = 0;
2985 STRLEN foldlen = 0;
2986 U8 *uscan = (U8*)NULL;
2987 STRLEN bufflen=0;
2988 SV *sv_accept_buff = NULL;
2989 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
2990
2991 ST.accepted = 0; /* how many accepting states we have seen */
2992 ST.B = next;
2993 ST.jump = trie->jump;
786e8c11 2994 ST.me = scan;
07be1b83
YO
2995 /*
2996 traverse the TRIE keeping track of all accepting states
2997 we transition through until we get to a failing node.
2998 */
2999
a3621e74 3000 while ( state && uc <= (U8*)PL_regeol ) {
786e8c11 3001 U32 base = trie->states[ state ].trans.base;
f9f4320a 3002 UV uvc = 0;
786e8c11
YO
3003 U16 charid;
3004 /* We use charid to hold the wordnum as we don't use it
3005 for charid until after we have done the wordnum logic.
3006 We define an alias just so that the wordnum logic reads
3007 more naturally. */
3008
3009#define got_wordnum charid
3010 got_wordnum = trie->states[ state ].wordnum;
3011
3012 if ( got_wordnum ) {
3013 if ( ! ST.accepted ) {
5b47454d 3014 ENTER;
6962fb1a 3015 SAVETMPS; /* XXX is this necessary? dmq */
5b47454d
DM
3016 bufflen = TRIE_INITAL_ACCEPT_BUFFLEN;
3017 sv_accept_buff=newSV(bufflen *
3018 sizeof(reg_trie_accepted) - 1);
786e8c11 3019 SvCUR_set(sv_accept_buff, 0);
5b47454d
DM
3020 SvPOK_on(sv_accept_buff);
3021 sv_2mortal(sv_accept_buff);
166ba7cd
DM
3022 SAVETMPS;
3023 ST.accept_buff =
5b47454d
DM
3024 (reg_trie_accepted*)SvPV_nolen(sv_accept_buff );
3025 }
786e8c11 3026 do {
166ba7cd 3027 if (ST.accepted >= bufflen) {
5b47454d 3028 bufflen *= 2;
166ba7cd 3029 ST.accept_buff =(reg_trie_accepted*)
5b47454d
DM
3030 SvGROW(sv_accept_buff,
3031 bufflen * sizeof(reg_trie_accepted));
3032 }
3033 SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff)
3034 + sizeof(reg_trie_accepted));
a3621e74 3035
786e8c11
YO
3036
3037 ST.accept_buff[ST.accepted].wordnum = got_wordnum;
3038 ST.accept_buff[ST.accepted].endpos = uc;
3039 ++ST.accepted;
3040 } while (trie->nextword && (got_wordnum= trie->nextword[got_wordnum]));
3041 }
3042#undef got_wordnum
a3621e74 3043
07be1b83 3044 DEBUG_TRIE_EXECUTE_r({
786e8c11 3045 DUMP_EXEC_POS( (char *)uc, scan, do_utf8 );
a3621e74 3046 PerlIO_printf( Perl_debug_log,
786e8c11 3047 "%*s %sState: %4"UVxf" Accepted: %4"UVxf" ",
5bc10b2c 3048 2+depth * 2, "", PL_colors[4],
786e8c11 3049 (UV)state, (UV)ST.accepted );
07be1b83 3050 });
a3621e74
YO
3051
3052 if ( base ) {
55eed653
NC
3053 REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc,
3054 uscan, len, uvc, charid, foldlen,
3055 foldbuf, uniflags);
a3621e74 3056
5b47454d
DM
3057 if (charid &&
3058 (base + charid > trie->uniquecharcount )
3059 && (base + charid - 1 - trie->uniquecharcount
3060 < trie->lasttrans)
3061 && trie->trans[base + charid - 1 -
3062 trie->uniquecharcount].check == state)
3063 {
3064 state = trie->trans[base + charid - 1 -
3065 trie->uniquecharcount ].next;
3066 }
3067 else {
3068 state = 0;
3069 }
3070 uc += len;
3071
3072 }
3073 else {
a3621e74
YO
3074 state = 0;
3075 }
3076 DEBUG_TRIE_EXECUTE_r(
e4584336 3077 PerlIO_printf( Perl_debug_log,
786e8c11 3078 "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n",
e4584336 3079 charid, uvc, (UV)state, PL_colors[5] );
a3621e74
YO
3080 );
3081 }
166ba7cd 3082 if (!ST.accepted )
a3621e74 3083 sayNO;
a3621e74 3084
166ba7cd
DM
3085 DEBUG_EXECUTE_r(
3086 PerlIO_printf( Perl_debug_log,
3087 "%*s %sgot %"IVdf" possible matches%s\n",
5bc10b2c 3088 REPORT_CODE_OFF + depth * 2, "",
166ba7cd
DM
3089 PL_colors[4], (IV)ST.accepted, PL_colors[5] );
3090 );
786e8c11 3091 }}
fae667d5
YO
3092 goto trie_first_try; /* jump into the fail handler */
3093 /* NOTREACHED */
166ba7cd 3094 case TRIE_next_fail: /* we failed - try next alterative */
fae667d5
YO
3095 if ( ST.jump) {
3096 REGCP_UNWIND(ST.cp);
3097 for (n = *PL_reglastparen; n > ST.lastparen; n--)
f0ab9afb 3098 PL_regoffs[n].end = -1;
fae667d5
YO
3099 *PL_reglastparen = n;
3100 }
3101 trie_first_try:
5d458dd8
YO
3102 if (do_cutgroup) {
3103 do_cutgroup = 0;
3104 no_final = 0;
3105 }
fae667d5
YO
3106
3107 if ( ST.jump) {
3108 ST.lastparen = *PL_reglastparen;
3109 REGCP_SET(ST.cp);
3110 }
166ba7cd
DM
3111 if ( ST.accepted == 1 ) {
3112 /* only one choice left - just continue */
3113 DEBUG_EXECUTE_r({
2b8b4781 3114 AV *const trie_words
502c6561 3115 = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]);
2b8b4781 3116 SV ** const tmp = av_fetch( trie_words,
5c9f2f80 3117 ST.accept_buff[ 0 ].wordnum-1, 0 );
de734bd5
A
3118 SV *sv= tmp ? sv_newmortal() : NULL;
3119
166ba7cd
DM
3120 PerlIO_printf( Perl_debug_log,
3121 "%*s %sonly one match left: #%d <%s>%s\n",
5bc10b2c 3122 REPORT_CODE_OFF+depth*2, "", PL_colors[4],
166ba7cd 3123 ST.accept_buff[ 0 ].wordnum,
de734bd5
A
3124 tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0,
3125 PL_colors[0], PL_colors[1],
3126 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
3127 )
3128 : "not compiled under -Dr",
166ba7cd
DM
3129 PL_colors[5] );
3130 });
3131 PL_reginput = (char *)ST.accept_buff[ 0 ].endpos;
3132 /* in this case we free tmps/leave before we call regmatch
3133 as we wont be using accept_buff again. */
5d458dd8 3134
166ba7cd
DM
3135 locinput = PL_reginput;
3136 nextchr = UCHARAT(locinput);
5d458dd8
YO
3137 if ( !ST.jump || !ST.jump[ST.accept_buff[0].wordnum])
3138 scan = ST.B;
3139 else
3140 scan = ST.me + ST.jump[ST.accept_buff[0].wordnum];
3141 if (!has_cutgroup) {
3142 FREETMPS;
3143 LEAVE;
3144 } else {
3145 ST.accepted--;
3146 PUSH_YES_STATE_GOTO(TRIE_next, scan);
3147 }
786e8c11 3148
166ba7cd
DM
3149 continue; /* execute rest of RE */
3150 }
fae667d5
YO
3151
3152 if ( !ST.accepted-- ) {
5d458dd8
YO
3153 DEBUG_EXECUTE_r({
3154 PerlIO_printf( Perl_debug_log,
3155 "%*s %sTRIE failed...%s\n",
3156 REPORT_CODE_OFF+depth*2, "",
3157 PL_colors[4],
3158 PL_colors[5] );
3159 });
166ba7cd
DM
3160 FREETMPS;
3161 LEAVE;
5d458dd8 3162 sayNO_SILENT;
fae667d5
YO
3163 /*NOTREACHED*/
3164 }
166ba7cd 3165
a3621e74 3166 /*
166ba7cd
DM
3167 There are at least two accepting states left. Presumably
3168 the number of accepting states is going to be low,
3169 typically two. So we simply scan through to find the one
3170 with lowest wordnum. Once we find it, we swap the last
3171 state into its place and decrement the size. We then try to
3172 match the rest of the pattern at the point where the word
3173 ends. If we succeed, control just continues along the
3174 regex; if we fail we return here to try the next accepting
3175 state
3176 */
a3621e74 3177
166ba7cd
DM
3178 {
3179 U32 best = 0;
3180 U32 cur;
3181 for( cur = 1 ; cur <= ST.accepted ; cur++ ) {
3182 DEBUG_TRIE_EXECUTE_r(
f2278c82 3183 PerlIO_printf( Perl_debug_log,
166ba7cd 3184 "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n",
5bc10b2c 3185 REPORT_CODE_OFF + depth * 2, "", PL_colors[4],
166ba7cd
DM
3186 (IV)best, ST.accept_buff[ best ].wordnum, (IV)cur,
3187 ST.accept_buff[ cur ].wordnum, PL_colors[5] );
3188 );
3189
3190 if (ST.accept_buff[cur].wordnum <
3191 ST.accept_buff[best].wordnum)
3192 best = cur;
a3621e74 3193 }
166ba7cd
DM
3194
3195 DEBUG_EXECUTE_r({
2b8b4781 3196 AV *const trie_words
502c6561 3197 = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]);
2b8b4781 3198 SV ** const tmp = av_fetch( trie_words,
5c9f2f80 3199 ST.accept_buff[ best ].wordnum - 1, 0 );
7f69552c 3200 regnode *nextop=(!ST.jump || !ST.jump[ST.accept_buff[best].wordnum]) ?
265c4333 3201 ST.B :
7f69552c 3202 ST.me + ST.jump[ST.accept_buff[best].wordnum];
de734bd5
A
3203 SV *sv= tmp ? sv_newmortal() : NULL;
3204
265c4333
YO
3205 PerlIO_printf( Perl_debug_log,
3206 "%*s %strying alternation #%d <%s> at node #%d %s\n",
5bc10b2c 3207 REPORT_CODE_OFF+depth*2, "", PL_colors[4],
166ba7cd 3208 ST.accept_buff[best].wordnum,
de734bd5
A
3209 tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0,
3210 PL_colors[0], PL_colors[1],
3211 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
3212 ) : "not compiled under -Dr",
265c4333 3213 REG_NODE_NUM(nextop),
166ba7cd
DM
3214 PL_colors[5] );
3215 });
3216
3217 if ( best<ST.accepted ) {
3218 reg_trie_accepted tmp = ST.accept_buff[ best ];
3219 ST.accept_buff[ best ] = ST.accept_buff[ ST.accepted ];
3220 ST.accept_buff[ ST.accepted ] = tmp;
3221 best = ST.accepted;
a3621e74 3222 }
166ba7cd 3223 PL_reginput = (char *)ST.accept_buff[ best ].endpos;
7f69552c 3224 if ( !ST.jump || !ST.jump[ST.accept_buff[best].wordnum]) {
5d458dd8 3225 scan = ST.B;
786e8c11 3226 } else {
5d458dd8 3227 scan = ST.me + ST.jump[ST.accept_buff[best].wordnum];
5d458dd8 3228 }
6b173516 3229 PUSH_YES_STATE_GOTO(TRIE_next, scan);
786e8c11 3230 /* NOTREACHED */
166ba7cd 3231 }
166ba7cd 3232 /* NOTREACHED */
5d458dd8 3233 case TRIE_next:
6962fb1a
YO
3234 /* we dont want to throw this away, see bug 57042*/
3235 if (oreplsv != GvSV(PL_replgv))
3236 sv_setsv(oreplsv, GvSV(PL_replgv));
5d458dd8
YO
3237 FREETMPS;
3238 LEAVE;
3239 sayYES;
166ba7cd
DM
3240#undef ST
3241
95b24440
DM
3242 case EXACT: {
3243 char *s = STRING(scan);
24d3c4a9 3244 ln = STR_LEN(scan);
eb160463 3245 if (do_utf8 != UTF) {
bc517b45 3246 /* The target and the pattern have differing utf8ness. */
1aa99e6b 3247 char *l = locinput;
24d3c4a9 3248 const char * const e = s + ln;
a72c7584 3249
5ff6fc6d
JH
3250 if (do_utf8) {
3251 /* The target is utf8, the pattern is not utf8. */
1aa99e6b 3252 while (s < e) {
a3b680e6 3253 STRLEN ulen;
1aa99e6b 3254 if (l >= PL_regeol)
5ff6fc6d
JH
3255 sayNO;
3256 if (NATIVE_TO_UNI(*(U8*)s) !=
89ebb4a3 3257 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
041457d9 3258 uniflags))
5ff6fc6d 3259 sayNO;
bc517b45 3260 l += ulen;
5ff6fc6d 3261 s ++;
1aa99e6b 3262 }
5ff6fc6d
JH
3263 }
3264 else {
3265 /* The target is not utf8, the pattern is utf8. */
1aa99e6b 3266 while (s < e) {
a3b680e6 3267 STRLEN ulen;
1aa99e6b
IH
3268 if (l >= PL_regeol)
3269 sayNO;
5ff6fc6d 3270 if (NATIVE_TO_UNI(*((U8*)l)) !=
89ebb4a3 3271 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
041457d9 3272 uniflags))
1aa99e6b 3273 sayNO;
bc517b45 3274 s += ulen;
a72c7584 3275 l ++;
1aa99e6b 3276 }
5ff6fc6d 3277 }
1aa99e6b
IH
3278 locinput = l;
3279 nextchr = UCHARAT(locinput);
3280 break;
3281 }
bc517b45 3282 /* The target and the pattern have the same utf8ness. */
d6a28714
JH
3283 /* Inline the first character, for speed. */
3284 if (UCHARAT(s) != nextchr)
3285 sayNO;
24d3c4a9 3286 if (PL_regeol - locinput < ln)
d6a28714 3287 sayNO;
24d3c4a9 3288 if (ln > 1 && memNE(s, locinput, ln))
d6a28714 3289 sayNO;
24d3c4a9 3290 locinput += ln;
d6a28714
JH
3291 nextchr = UCHARAT(locinput);
3292 break;
95b24440 3293 }
d6a28714 3294 case EXACTFL:
b8c5462f
JH
3295 PL_reg_flags |= RF_tainted;
3296 /* FALL THROUGH */
95b24440 3297 case EXACTF: {
be8e71aa 3298 char * const s = STRING(scan);
24d3c4a9 3299 ln = STR_LEN(scan);
d6a28714 3300
d07ddd77
JH
3301 if (do_utf8 || UTF) {
3302 /* Either target or the pattern are utf8. */
be8e71aa 3303 const char * const l = locinput;
d07ddd77 3304 char *e = PL_regeol;
bc517b45 3305
24d3c4a9 3306 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
1feea2c7 3307 l, &e, 0, do_utf8)) {
5486206c
JH
3308 /* One more case for the sharp s:
3309 * pack("U0U*", 0xDF) =~ /ss/i,
3310 * the 0xC3 0x9F are the UTF-8
3311 * byte sequence for the U+00DF. */
e1d1eefb 3312
5486206c 3313 if (!(do_utf8 &&
e1d1eefb 3314 toLOWER(s[0]) == 's' &&
24d3c4a9 3315 ln >= 2 &&
5486206c
JH
3316 toLOWER(s[1]) == 's' &&
3317 (U8)l[0] == 0xC3 &&
3318 e - l >= 2 &&
3319 (U8)l[1] == 0x9F))
3320 sayNO;
3321 }
d07ddd77
JH
3322 locinput = e;
3323 nextchr = UCHARAT(locinput);
3324 break;
a0ed51b3 3325 }
d6a28714 3326
bc517b45
JH
3327 /* Neither the target and the pattern are utf8. */
3328
d6a28714
JH
3329 /* Inline the first character, for speed. */
3330 if (UCHARAT(s) != nextchr &&
3331 UCHARAT(s) != ((OP(scan) == EXACTF)
3332 ? PL_fold : PL_fold_locale)[nextchr])
a0ed51b3 3333 sayNO;
24d3c4a9 3334 if (PL_regeol - locinput < ln)
b8c5462f 3335 sayNO;
24d3c4a9
DM
3336 if (ln > 1 && (OP(scan) == EXACTF
3337 ? ibcmp(s, locinput, ln)
3338 : ibcmp_locale(s, locinput, ln)))
4633a7c4 3339 sayNO;
24d3c4a9 3340 locinput += ln;
d6a28714 3341 nextchr = UCHARAT(locinput);
a0d0e21e 3342 break;
95b24440 3343 }
d6a28714 3344 case ANYOF:
ffc61ed2 3345 if (do_utf8) {
9e55ce06
JH
3346 STRLEN inclasslen = PL_regeol - locinput;
3347
32fc9b6a 3348 if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8))
262b90c4 3349 goto anyof_fail;
ffc61ed2
JH
3350 if (locinput >= PL_regeol)
3351 sayNO;
0f0076b4 3352 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
b8c5462f 3353 nextchr = UCHARAT(locinput);
e0f9d4a8 3354 break;
ffc61ed2
JH
3355 }
3356 else {
3357 if (nextchr < 0)
3358 nextchr = UCHARAT(locinput);
32fc9b6a 3359 if (!REGINCLASS(rex, scan, (U8*)locinput))
262b90c4 3360 goto anyof_fail;
ffc61ed2
JH
3361 if (!nextchr && locinput >= PL_regeol)
3362 sayNO;
3363 nextchr = UCHARAT(++locinput);
e0f9d4a8
JH
3364 break;
3365 }
262b90c4 3366 anyof_fail:
e0f9d4a8
JH
3367 /* If we might have the case of the German sharp s
3368 * in a casefolding Unicode character class. */
3369
ebc501f0
JH
3370 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
3371 locinput += SHARP_S_SKIP;
e0f9d4a8 3372 nextchr = UCHARAT(locinput);
ffc61ed2 3373 }
e0f9d4a8
JH
3374 else
3375 sayNO;
b8c5462f 3376 break;
d6a28714 3377 case ALNUML:
b8c5462f
JH
3378 PL_reg_flags |= RF_tainted;
3379 /* FALL THROUGH */
d6a28714 3380 case ALNUM:
b8c5462f 3381 if (!nextchr)
4633a7c4 3382 sayNO;
ffc61ed2 3383 if (do_utf8) {
1a4fad37 3384 LOAD_UTF8_CHARCLASS_ALNUM();
ffc61ed2 3385 if (!(OP(scan) == ALNUM
bb7a0f54 3386 ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
d6a28714 3387 : isALNUM_LC_utf8((U8*)locinput)))
b8c5462f
JH
3388 {
3389 sayNO;
a0ed51b3 3390 }
b8c5462f 3391 locinput += PL_utf8skip[nextchr];
a0ed51b3
LW
3392 nextchr = UCHARAT(locinput);
3393 break;
3394 }
ffc61ed2 3395 if (!(OP(scan) == ALNUM
d6a28714 3396 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
bbce6d69 3397 sayNO;
b8c5462f 3398 nextchr = UCHARAT(++locinput);
a0d0e21e 3399 break;
d6a28714 3400 case NALNUML:
b8c5462f
JH
3401 PL_reg_flags |= RF_tainted;
3402 /* FALL THROUGH */
d6a28714
JH
3403 case NALNUM:
3404 if (!nextchr && locinput >= PL_regeol)
a0ed51b3 3405 sayNO;
ffc61ed2 3406 if (do_utf8) {
1a4fad37 3407 LOAD_UTF8_CHARCLASS_ALNUM();
ffc61ed2 3408 if (OP(scan) == NALNUM
bb7a0f54 3409 ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
d6a28714
JH
3410 : isALNUM_LC_utf8((U8*)locinput))
3411 {
b8c5462f 3412 sayNO;
d6a28714 3413 }
b8c5462f
JH
3414 locinput += PL_utf8skip[nextchr];
3415 nextchr = UCHARAT(locinput);
3416 break;
3417 }
ffc61ed2 3418 if (OP(scan) == NALNUM
d6a28714 3419 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
b8c5462f 3420 sayNO;
76e3520e 3421 nextchr = UCHARAT(++locinput);
a0d0e21e 3422 break;
d6a28714
JH
3423 case BOUNDL:
3424 case NBOUNDL:
3280af22 3425 PL_reg_flags |= RF_tainted;
bbce6d69 3426 /* FALL THROUGH */
d6a28714
JH
3427 case BOUND:
3428 case NBOUND:
3429 /* was last char in word? */
ffc61ed2 3430 if (do_utf8) {
12d33761 3431 if (locinput == PL_bostr)
24d3c4a9 3432 ln = '\n';
ffc61ed2 3433 else {
a3b680e6 3434 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
9041c2e3 3435
24d3c4a9 3436 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
ffc61ed2
JH
3437 }
3438 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
24d3c4a9 3439 ln = isALNUM_uni(ln);
1a4fad37 3440 LOAD_UTF8_CHARCLASS_ALNUM();
3568d838 3441 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
ffc61ed2
JH
3442 }
3443 else {
24d3c4a9 3444 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
ffc61ed2
JH
3445 n = isALNUM_LC_utf8((U8*)locinput);
3446 }
a0ed51b3 3447 }
d6a28714 3448 else {
24d3c4a9 3449 ln = (locinput != PL_bostr) ?
12d33761 3450 UCHARAT(locinput - 1) : '\n';
ffc61ed2 3451 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
24d3c4a9 3452 ln = isALNUM(ln);
ffc61ed2
JH
3453 n = isALNUM(nextchr);
3454 }
3455 else {
24d3c4a9 3456 ln = isALNUM_LC(ln);
ffc61ed2
JH
3457 n = isALNUM_LC(nextchr);
3458 }
d6a28714 3459 }
24d3c4a9 3460 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
ffc61ed2
JH
3461 OP(scan) == BOUNDL))
3462 sayNO;
a0ed51b3 3463 break;
d6a28714 3464 case SPACEL:
3280af22 3465 PL_reg_flags |= RF_tainted;
bbce6d69 3466 /* FALL THROUGH */
d6a28714 3467 case SPACE:
9442cb0e 3468 if (!nextchr)
4633a7c4 3469 sayNO;
1aa99e6b 3470 if (do_utf8) {
fd400ab9 3471 if (UTF8_IS_CONTINUED(nextchr)) {
1a4fad37 3472 LOAD_UTF8_CHARCLASS_SPACE();
ffc61ed2 3473 if (!(OP(scan) == SPACE
bb7a0f54 3474 ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
ffc61ed2
JH
3475 : isSPACE_LC_utf8((U8*)locinput)))
3476 {
3477 sayNO;
3478 }
3479 locinput += PL_utf8skip[nextchr];
3480 nextchr = UCHARAT(locinput);
3481 break;
d6a28714 3482 }
ffc61ed2
JH
3483 if (!(OP(scan) == SPACE
3484 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3485 sayNO;
3486 nextchr = UCHARAT(++locinput);
3487 }
3488 else {
3489 if (!(OP(scan) == SPACE
3490 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3491 sayNO;
3492 nextchr = UCHARAT(++locinput);
a0ed51b3 3493 }
a0ed51b3 3494 break;
d6a28714 3495 case NSPACEL:
3280af22 3496 PL_reg_flags |= RF_tainted;
bbce6d69 3497 /* FALL THROUGH */
d6a28714 3498 case NSPACE:
9442cb0e 3499 if (!nextchr && locinput >= PL_regeol)
b8c5462f 3500 sayNO;
1aa99e6b 3501 if (do_utf8) {
1a4fad37 3502 LOAD_UTF8_CHARCLASS_SPACE();
ffc61ed2 3503 if (OP(scan) == NSPACE
bb7a0f54 3504 ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
d6a28714 3505 : isSPACE_LC_utf8((U8*)locinput))
b8c5462f
JH
3506 {
3507 sayNO;
3508 }
3509 locinput += PL_utf8skip[nextchr];
3510 nextchr = UCHARAT(locinput);
3511 break;
a0ed51b3 3512 }
ffc61ed2 3513 if (OP(scan) == NSPACE
d6a28714 3514 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
4633a7c4 3515 sayNO;
76e3520e 3516 nextchr = UCHARAT(++locinput);
a0d0e21e 3517 break;
d6a28714 3518 case DIGITL:
a0ed51b3
LW
3519 PL_reg_flags |= RF_tainted;
3520 /* FALL THROUGH */
d6a28714 3521 case DIGIT:
9442cb0e 3522 if (!nextchr)
a0ed51b3 3523 sayNO;
1aa99e6b 3524 if (do_utf8) {
1a4fad37 3525 LOAD_UTF8_CHARCLASS_DIGIT();
ffc61ed2 3526 if (!(OP(scan) == DIGIT
bb7a0f54 3527 ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
9442cb0e 3528 : isDIGIT_LC_utf8((U8*)locinput)))
dfe13c55 3529 {
a0ed51b3 3530 sayNO;
dfe13c55 3531 }
6f06b55f 3532 locinput += PL_utf8skip[nextchr];
a0ed51b3
LW
3533 nextchr = UCHARAT(locinput);
3534 break;
3535 }
ffc61ed2 3536 if (!(OP(scan) == DIGIT
9442cb0e 3537 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
a0ed51b3
LW
3538 sayNO;
3539 nextchr = UCHARAT(++locinput);
3540 break;
d6a28714 3541 case NDIGITL:
b8c5462f
JH
3542 PL_reg_flags |= RF_tainted;
3543 /* FALL THROUGH */
d6a28714 3544 case NDIGIT:
9442cb0e 3545 if (!nextchr && locinput >= PL_regeol)
b8c5462f 3546 sayNO;
1aa99e6b 3547 if (do_utf8) {
1a4fad37 3548 LOAD_UTF8_CHARCLASS_DIGIT();
ffc61ed2 3549 if (OP(scan) == NDIGIT
bb7a0f54 3550 ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
9442cb0e
GS
3551 : isDIGIT_LC_utf8((U8*)locinput))
3552 {
a0ed51b3 3553 sayNO;
9442cb0e 3554 }
6f06b55f 3555 locinput += PL_utf8skip[nextchr];
a0ed51b3
LW
3556 nextchr = UCHARAT(locinput);
3557 break;
3558 }
ffc61ed2 3559 if (OP(scan) == NDIGIT
9442cb0e 3560 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
a0ed51b3
LW
3561 sayNO;
3562 nextchr = UCHARAT(++locinput);
3563 break;
3564 case CLUMP:
b7c83a7e 3565 if (locinput >= PL_regeol)
a0ed51b3 3566 sayNO;
b7c83a7e 3567 if (do_utf8) {
1a4fad37 3568 LOAD_UTF8_CHARCLASS_MARK();
b7c83a7e
JH
3569 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3570 sayNO;
3571 locinput += PL_utf8skip[nextchr];
3572 while (locinput < PL_regeol &&
3573 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3574 locinput += UTF8SKIP(locinput);
3575 if (locinput > PL_regeol)
3576 sayNO;
eb08e2da
JH
3577 }
3578 else
3579 locinput++;
a0ed51b3
LW
3580 nextchr = UCHARAT(locinput);
3581 break;
81714fb9
YO
3582
3583 case NREFFL:
3584 {
3585 char *s;
ff1157ca 3586 char type;
81714fb9
YO
3587 PL_reg_flags |= RF_tainted;
3588 /* FALL THROUGH */
3589 case NREF:
3590 case NREFF:
ff1157ca 3591 type = OP(scan);
0a4db386
YO
3592 n = reg_check_named_buff_matched(rex,scan);
3593
3594 if ( n ) {
3595 type = REF + ( type - NREF );
3596 goto do_ref;
3597 } else {
81714fb9 3598 sayNO;
0a4db386
YO
3599 }
3600 /* unreached */
c8756f30 3601 case REFFL:
3280af22 3602 PL_reg_flags |= RF_tainted;
c8756f30 3603 /* FALL THROUGH */
c277df42 3604 case REF:
81714fb9 3605 case REFF:
c277df42 3606 n = ARG(scan); /* which paren pair */
81714fb9
YO
3607 type = OP(scan);
3608 do_ref:
f0ab9afb 3609 ln = PL_regoffs[n].start;
2c2d71f5 3610 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3b6647e0 3611 if (*PL_reglastparen < n || ln == -1)
af3f8c16 3612 sayNO; /* Do not match unless seen CLOSEn. */
f0ab9afb 3613 if (ln == PL_regoffs[n].end)
a0d0e21e 3614 break;
a0ed51b3 3615
24d3c4a9 3616 s = PL_bostr + ln;
81714fb9 3617 if (do_utf8 && type != REF) { /* REF can do byte comparison */
a0ed51b3 3618 char *l = locinput;
f0ab9afb 3619 const char *e = PL_bostr + PL_regoffs[n].end;
a0ed51b3
LW
3620 /*
3621 * Note that we can't do the "other character" lookup trick as
3622 * in the 8-bit case (no pun intended) because in Unicode we
3623 * have to map both upper and title case to lower case.
3624 */
81714fb9 3625 if (type == REFF) {
a0ed51b3 3626 while (s < e) {
a3b680e6
AL
3627 STRLEN ulen1, ulen2;
3628 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3629 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3630
a0ed51b3
LW
3631 if (l >= PL_regeol)
3632 sayNO;
a2a2844f
JH
3633 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3634 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
7114a2d2 3635 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
a0ed51b3 3636 sayNO;
a2a2844f
JH
3637 s += ulen1;
3638 l += ulen2;
a0ed51b3
LW
3639 }
3640 }
3641 locinput = l;
3642 nextchr = UCHARAT(locinput);
3643 break;
3644 }
3645
a0d0e21e 3646 /* Inline the first character, for speed. */
76e3520e 3647 if (UCHARAT(s) != nextchr &&
81714fb9
YO
3648 (type == REF ||
3649 (UCHARAT(s) != (type == REFF
3650 ? PL_fold : PL_fold_locale)[nextchr])))
4633a7c4 3651 sayNO;
f0ab9afb 3652 ln = PL_regoffs[n].end - ln;
24d3c4a9 3653 if (locinput + ln > PL_regeol)
4633a7c4 3654 sayNO;
81714fb9 3655 if (ln > 1 && (type == REF
24d3c4a9 3656 ? memNE(s, locinput, ln)
81714fb9 3657 : (type == REFF
24d3c4a9
DM
3658 ? ibcmp(s, locinput, ln)
3659 : ibcmp_locale(s, locinput, ln))))
4633a7c4 3660 sayNO;
24d3c4a9 3661 locinput += ln;
76e3520e 3662 nextchr = UCHARAT(locinput);
a0d0e21e 3663 break;
81714fb9 3664 }
a0d0e21e 3665 case NOTHING:
c277df42 3666 case TAIL:
a0d0e21e
LW
3667 break;
3668 case BACK:
3669 break;
40a82448
DM
3670
3671#undef ST
3672#define ST st->u.eval
c277df42 3673 {
c277df42 3674 SV *ret;
d2f13c59 3675 REGEXP *re_sv;
6bda09f9 3676 regexp *re;
f8fc2ecf 3677 regexp_internal *rei;
1a147d38
YO
3678 regnode *startpoint;
3679
3680 case GOSTART:
e7707071
YO
3681 case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */
3682 if (cur_eval && cur_eval->locinput==locinput) {
24b23f37 3683 if (cur_eval->u.eval.close_paren == (U32)ARG(scan))
1a147d38 3684 Perl_croak(aTHX_ "Infinite recursion in regex");
4b196cd4 3685 if ( ++nochange_depth > max_nochange_depth )
1a147d38
YO
3686 Perl_croak(aTHX_
3687 "Pattern subroutine nesting without pos change"
3688 " exceeded limit in regex");
6bda09f9
YO
3689 } else {
3690 nochange_depth = 0;
1a147d38 3691 }
288b8c02 3692 re_sv = rex_sv;
6bda09f9 3693 re = rex;
f8fc2ecf 3694 rei = rexi;
288b8c02 3695 (void)ReREFCNT_inc(rex_sv);
1a147d38 3696 if (OP(scan)==GOSUB) {
6bda09f9
YO
3697 startpoint = scan + ARG2L(scan);
3698 ST.close_paren = ARG(scan);
3699 } else {
f8fc2ecf 3700 startpoint = rei->program+1;
6bda09f9
YO
3701 ST.close_paren = 0;
3702 }
3703 goto eval_recurse_doit;
3704 /* NOTREACHED */
3705 case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */
3706 if (cur_eval && cur_eval->locinput==locinput) {
4b196cd4 3707 if ( ++nochange_depth > max_nochange_depth )
1a147d38 3708 Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex");
6bda09f9
YO
3709 } else {
3710 nochange_depth = 0;
3711 }
8e5e9ebe 3712 {
4aabdb9b
DM
3713 /* execute the code in the {...} */
3714 dSP;
6136c704 3715 SV ** const before = SP;
4aabdb9b
DM
3716 OP_4tree * const oop = PL_op;
3717 COP * const ocurcop = PL_curcop;
3718 PAD *old_comppad;
4aabdb9b
DM
3719
3720 n = ARG(scan);
f8fc2ecf 3721 PL_op = (OP_4tree*)rexi->data->data[n];
24b23f37
YO
3722 DEBUG_STATE_r( PerlIO_printf(Perl_debug_log,
3723 " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
f8fc2ecf 3724 PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]);
f0ab9afb 3725 PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr;
4aabdb9b 3726
2bf803e2
YO
3727 if (sv_yes_mark) {
3728 SV *sv_mrk = get_sv("REGMARK", 1);
3729 sv_setsv(sv_mrk, sv_yes_mark);
3730 }
3731
8e5e9ebe
RGS
3732 CALLRUNOPS(aTHX); /* Scalar context. */
3733 SPAGAIN;
3734 if (SP == before)
075aa684 3735 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
8e5e9ebe
RGS
3736 else {
3737 ret = POPs;
3738 PUTBACK;
3739 }
4aabdb9b
DM
3740
3741 PL_op = oop;
3742 PAD_RESTORE_LOCAL(old_comppad);
3743 PL_curcop = ocurcop;
24d3c4a9 3744 if (!logical) {
4aabdb9b
DM
3745 /* /(?{...})/ */
3746 sv_setsv(save_scalar(PL_replgv), ret);
4aabdb9b
DM
3747 break;
3748 }
8e5e9ebe 3749 }
24d3c4a9
DM
3750 if (logical == 2) { /* Postponed subexpression: /(??{...})/ */
3751 logical = 0;
4aabdb9b 3752 {
4f639d21
DM
3753 /* extract RE object from returned value; compiling if
3754 * necessary */
6136c704 3755 MAGIC *mg = NULL;
288b8c02 3756 REGEXP *rx = NULL;
5c35adbb
NC
3757
3758 if (SvROK(ret)) {
288b8c02 3759 SV *const sv = SvRV(ret);
5c35adbb
NC
3760
3761 if (SvTYPE(sv) == SVt_REGEXP) {
d2f13c59 3762 rx = (REGEXP*) sv;
5c35adbb
NC
3763 } else if (SvSMAGICAL(sv)) {
3764 mg = mg_find(sv, PERL_MAGIC_qr);
3765 assert(mg);
3766 }
3767 } else if (SvTYPE(ret) == SVt_REGEXP) {
d2f13c59 3768 rx = (REGEXP*) ret;
5c35adbb 3769 } else if (SvSMAGICAL(ret)) {
124ee91a
NC
3770 if (SvGMAGICAL(ret)) {
3771 /* I don't believe that there is ever qr magic
3772 here. */
3773 assert(!mg_find(ret, PERL_MAGIC_qr));
faf82a0b 3774 sv_unmagic(ret, PERL_MAGIC_qr);
124ee91a
NC
3775 }
3776 else {
faf82a0b 3777 mg = mg_find(ret, PERL_MAGIC_qr);
124ee91a
NC
3778 /* testing suggests mg only ends up non-NULL for
3779 scalars who were upgraded and compiled in the
3780 else block below. In turn, this is only
3781 triggered in the "postponed utf8 string" tests
3782 in t/op/pat.t */
3783 }
0f5d15d6 3784 }
faf82a0b 3785
0f5d15d6 3786 if (mg) {
d2f13c59 3787 rx = (REGEXP *) mg->mg_obj; /*XXX:dmq*/
e2560c33 3788 assert(rx);
0f5d15d6 3789 }
288b8c02
NC
3790 if (rx) {
3791 rx = reg_temp_copy(rx);
3792 }
0f5d15d6 3793 else {
c737faaf 3794 U32 pm_flags = 0;
a3b680e6 3795 const I32 osize = PL_regsize;
0f5d15d6 3796
b9ad30b4
NC
3797 if (DO_UTF8(ret)) {
3798 assert (SvUTF8(ret));
3799 } else if (SvUTF8(ret)) {
3800 /* Not doing UTF-8, despite what the SV says. Is
3801 this only if we're trapped in use 'bytes'? */
3802 /* Make a copy of the octet sequence, but without
3803 the flag on, as the compiler now honours the
3804 SvUTF8 flag on ret. */
3805 STRLEN len;
3806 const char *const p = SvPV(ret, len);
3807 ret = newSVpvn_flags(p, len, SVs_TEMP);
3808 }
288b8c02 3809 rx = CALLREGCOMP(ret, pm_flags);
9041c2e3 3810 if (!(SvFLAGS(ret)
faf82a0b 3811 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
3ce3ed55 3812 | SVs_GMG))) {
a2794585
NC
3813 /* This isn't a first class regexp. Instead, it's
3814 caching a regexp onto an existing, Perl visible
3815 scalar. */
ad64d0ec 3816 sv_magic(ret, MUTABLE_SV(rx), PERL_MAGIC_qr, 0, 0);
3ce3ed55 3817 }
0f5d15d6 3818 PL_regsize = osize;
0f5d15d6 3819 }
288b8c02
NC
3820 re_sv = rx;
3821 re = (struct regexp *)SvANY(rx);
4aabdb9b 3822 }
07bc277f 3823 RXp_MATCH_COPIED_off(re);
28d8d7f4
YO
3824 re->subbeg = rex->subbeg;
3825 re->sublen = rex->sublen;
f8fc2ecf 3826 rei = RXi_GET(re);
6bda09f9 3827 DEBUG_EXECUTE_r(
efd26800 3828 debug_start_match(re_sv, do_utf8, locinput, PL_regeol,
6bda09f9
YO
3829 "Matching embedded");
3830 );
f8fc2ecf 3831 startpoint = rei->program + 1;
1a147d38 3832 ST.close_paren = 0; /* only used for GOSUB */
6bda09f9
YO
3833 /* borrowed from regtry */
3834 if (PL_reg_start_tmpl <= re->nparens) {
3835 PL_reg_start_tmpl = re->nparens*3/2 + 3;
3836 if(PL_reg_start_tmp)
3837 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
3838 else
3839 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
dd5def09 3840 }
aa283a38 3841
1a147d38 3842 eval_recurse_doit: /* Share code with GOSUB below this line */
aa283a38 3843 /* run the pattern returned from (??{...}) */
40a82448
DM
3844 ST.cp = regcppush(0); /* Save *all* the positions. */
3845 REGCP_SET(ST.lastcp);
6bda09f9 3846
f0ab9afb 3847 PL_regoffs = re->offs; /* essentially NOOP on GOSUB */
6bda09f9 3848
0357f1fd
ML
3849 /* see regtry, specifically PL_reglast(?:close)?paren is a pointer! (i dont know why) :dmq */
3850 PL_reglastparen = &re->lastparen;
3851 PL_reglastcloseparen = &re->lastcloseparen;
3852 re->lastparen = 0;
3853 re->lastcloseparen = 0;
3854
4aabdb9b 3855 PL_reginput = locinput;
ae0beba1 3856 PL_regsize = 0;
4aabdb9b
DM
3857
3858 /* XXXX This is too dramatic a measure... */
3859 PL_reg_maxiter = 0;
3860
faec1544 3861 ST.toggle_reg_flags = PL_reg_flags;
3c8556c3 3862 if (RX_UTF8(re_sv))
faec1544
DM
3863 PL_reg_flags |= RF_utf8;
3864 else
3865 PL_reg_flags &= ~RF_utf8;
3866 ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */
3867
288b8c02 3868 ST.prev_rex = rex_sv;
faec1544 3869 ST.prev_curlyx = cur_curlyx;
288b8c02
NC
3870 SETREX(rex_sv,re_sv);
3871 rex = re;
f8fc2ecf 3872 rexi = rei;
faec1544 3873 cur_curlyx = NULL;
40a82448 3874 ST.B = next;
faec1544
DM
3875 ST.prev_eval = cur_eval;
3876 cur_eval = st;
faec1544 3877 /* now continue from first node in postoned RE */
6bda09f9 3878 PUSH_YES_STATE_GOTO(EVAL_AB, startpoint);
4aabdb9b 3879 /* NOTREACHED */
a0ed51b3 3880 }
24d3c4a9
DM
3881 /* logical is 1, /(?(?{...})X|Y)/ */
3882 sw = (bool)SvTRUE(ret);
3883 logical = 0;
c277df42
IZ
3884 break;
3885 }
40a82448 3886
faec1544
DM
3887 case EVAL_AB: /* cleanup after a successful (??{A})B */
3888 /* note: this is called twice; first after popping B, then A */
3889 PL_reg_flags ^= ST.toggle_reg_flags;
288b8c02
NC
3890 ReREFCNT_dec(rex_sv);
3891 SETREX(rex_sv,ST.prev_rex);
3892 rex = (struct regexp *)SvANY(rex_sv);
f8fc2ecf 3893 rexi = RXi_GET(rex);
faec1544
DM
3894 regcpblow(ST.cp);
3895 cur_eval = ST.prev_eval;
3896 cur_curlyx = ST.prev_curlyx;
34a81e2b
B
3897
3898 /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
0357f1fd
ML
3899 PL_reglastparen = &rex->lastparen;
3900 PL_reglastcloseparen = &rex->lastcloseparen;
34a81e2b
B
3901 /* also update PL_regoffs */
3902 PL_regoffs = rex->offs;
0357f1fd 3903
40a82448
DM
3904 /* XXXX This is too dramatic a measure... */
3905 PL_reg_maxiter = 0;
e7707071 3906 if ( nochange_depth )
4b196cd4 3907 nochange_depth--;
262b90c4 3908 sayYES;
40a82448 3909
40a82448 3910
faec1544
DM
3911 case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */
3912 /* note: this is called twice; first after popping B, then A */
3913 PL_reg_flags ^= ST.toggle_reg_flags;
288b8c02
NC
3914 ReREFCNT_dec(rex_sv);
3915 SETREX(rex_sv,ST.prev_rex);
3916 rex = (struct regexp *)SvANY(rex_sv);
f8fc2ecf 3917 rexi = RXi_GET(rex);
34a81e2b 3918 /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
0357f1fd
ML
3919 PL_reglastparen = &rex->lastparen;
3920 PL_reglastcloseparen = &rex->lastcloseparen;
3921
40a82448
DM
3922 PL_reginput = locinput;
3923 REGCP_UNWIND(ST.lastcp);
3924 regcppop(rex);
faec1544
DM
3925 cur_eval = ST.prev_eval;
3926 cur_curlyx = ST.prev_curlyx;
3927 /* XXXX This is too dramatic a measure... */
3928 PL_reg_maxiter = 0;
e7707071 3929 if ( nochange_depth )
4b196cd4 3930 nochange_depth--;
40a82448 3931 sayNO_SILENT;
40a82448
DM
3932#undef ST
3933
a0d0e21e 3934 case OPEN:
c277df42 3935 n = ARG(scan); /* which paren pair */
3280af22
NIS
3936 PL_reg_start_tmp[n] = locinput;
3937 if (n > PL_regsize)
3938 PL_regsize = n;
e2e6a0f1 3939 lastopen = n;
a0d0e21e
LW
3940 break;
3941 case CLOSE:
c277df42 3942 n = ARG(scan); /* which paren pair */
f0ab9afb
NC
3943 PL_regoffs[n].start = PL_reg_start_tmp[n] - PL_bostr;
3944 PL_regoffs[n].end = locinput - PL_bostr;
7f69552c
YO
3945 /*if (n > PL_regsize)
3946 PL_regsize = n;*/
3b6647e0 3947 if (n > *PL_reglastparen)
3280af22 3948 *PL_reglastparen = n;
a01268b5 3949 *PL_reglastcloseparen = n;
3b6647e0 3950 if (cur_eval && cur_eval->u.eval.close_paren == n) {
6bda09f9
YO
3951 goto fake_end;
3952 }
a0d0e21e 3953 break;
e2e6a0f1
YO
3954 case ACCEPT:
3955 if (ARG(scan)){
3956 regnode *cursor;
3957 for (cursor=scan;
3958 cursor && OP(cursor)!=END;
3959 cursor=regnext(cursor))
3960 {
3961 if ( OP(cursor)==CLOSE ){
3962 n = ARG(cursor);
3963 if ( n <= lastopen ) {
f0ab9afb
NC
3964 PL_regoffs[n].start
3965 = PL_reg_start_tmp[n] - PL_bostr;
3966 PL_regoffs[n].end = locinput - PL_bostr;
e2e6a0f1
YO
3967 /*if (n > PL_regsize)
3968 PL_regsize = n;*/
3b6647e0 3969 if (n > *PL_reglastparen)
e2e6a0f1
YO
3970 *PL_reglastparen = n;
3971 *PL_reglastcloseparen = n;
3b6647e0
RB
3972 if ( n == ARG(scan) || (cur_eval &&
3973 cur_eval->u.eval.close_paren == n))
e2e6a0f1
YO
3974 break;
3975 }
3976 }
3977 }
3978 }
3979 goto fake_end;
3980 /*NOTREACHED*/
c277df42
IZ
3981 case GROUPP:
3982 n = ARG(scan); /* which paren pair */
f0ab9afb 3983 sw = (bool)(*PL_reglastparen >= n && PL_regoffs[n].end != -1);
c277df42 3984 break;
0a4db386
YO
3985 case NGROUPP:
3986 /* reg_check_named_buff_matched returns 0 for no match */
3987 sw = (bool)(0 < reg_check_named_buff_matched(rex,scan));
3988 break;
1a147d38 3989 case INSUBP:
0a4db386 3990 n = ARG(scan);
3b6647e0 3991 sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n));
0a4db386
YO
3992 break;
3993 case DEFINEP:
3994 sw = 0;
3995 break;
c277df42 3996 case IFTHEN:
2c2d71f5 3997 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
24d3c4a9 3998 if (sw)
c277df42
IZ
3999 next = NEXTOPER(NEXTOPER(scan));
4000 else {
4001 next = scan + ARG(scan);
4002 if (OP(next) == IFTHEN) /* Fake one. */
4003 next = NEXTOPER(NEXTOPER(next));
4004 }
4005 break;
4006 case LOGICAL:
24d3c4a9 4007 logical = scan->flags;
c277df42 4008 break;
c476f425 4009
2ab05381 4010/*******************************************************************
2ab05381 4011
c476f425
DM
4012The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/
4013pattern, where A and B are subpatterns. (For simple A, CURLYM or
4014STAR/PLUS/CURLY/CURLYN are used instead.)
2ab05381 4015
c476f425 4016A*B is compiled as <CURLYX><A><WHILEM><B>
2ab05381 4017
c476f425
DM
4018On entry to the subpattern, CURLYX is called. This pushes a CURLYX
4019state, which contains the current count, initialised to -1. It also sets
4020cur_curlyx to point to this state, with any previous value saved in the
4021state block.
2ab05381 4022
c476f425
DM
4023CURLYX then jumps straight to the WHILEM op, rather than executing A,
4024since the pattern may possibly match zero times (i.e. it's a while {} loop
4025rather than a do {} while loop).
2ab05381 4026
c476f425
DM
4027Each entry to WHILEM represents a successful match of A. The count in the
4028CURLYX block is incremented, another WHILEM state is pushed, and execution
4029passes to A or B depending on greediness and the current count.
2ab05381 4030
c476f425
DM
4031For example, if matching against the string a1a2a3b (where the aN are
4032substrings that match /A/), then the match progresses as follows: (the
4033pushed states are interspersed with the bits of strings matched so far):
2ab05381 4034
c476f425
DM
4035 <CURLYX cnt=-1>
4036 <CURLYX cnt=0><WHILEM>
4037 <CURLYX cnt=1><WHILEM> a1 <WHILEM>
4038 <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM>
4039 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM>
4040 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b
2ab05381 4041
c476f425
DM
4042(Contrast this with something like CURLYM, which maintains only a single
4043backtrack state:
2ab05381 4044
c476f425
DM
4045 <CURLYM cnt=0> a1
4046 a1 <CURLYM cnt=1> a2
4047 a1 a2 <CURLYM cnt=2> a3
4048 a1 a2 a3 <CURLYM cnt=3> b
4049)
2ab05381 4050
c476f425
DM
4051Each WHILEM state block marks a point to backtrack to upon partial failure
4052of A or B, and also contains some minor state data related to that
4053iteration. The CURLYX block, pointed to by cur_curlyx, contains the
4054overall state, such as the count, and pointers to the A and B ops.
2ab05381 4055
c476f425
DM
4056This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx
4057must always point to the *current* CURLYX block, the rules are:
2ab05381 4058
c476f425
DM
4059When executing CURLYX, save the old cur_curlyx in the CURLYX state block,
4060and set cur_curlyx to point the new block.
2ab05381 4061
c476f425
DM
4062When popping the CURLYX block after a successful or unsuccessful match,
4063restore the previous cur_curlyx.
2ab05381 4064
c476f425
DM
4065When WHILEM is about to execute B, save the current cur_curlyx, and set it
4066to the outer one saved in the CURLYX block.
2ab05381 4067
c476f425
DM
4068When popping the WHILEM block after a successful or unsuccessful B match,
4069restore the previous cur_curlyx.
2ab05381 4070
c476f425
DM
4071Here's an example for the pattern (AI* BI)*BO
4072I and O refer to inner and outer, C and W refer to CURLYX and WHILEM:
2ab05381 4073
c476f425
DM
4074cur_
4075curlyx backtrack stack
4076------ ---------------
4077NULL
4078CO <CO prev=NULL> <WO>
4079CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
4080CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
4081NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo
2ab05381 4082
c476f425
DM
4083At this point the pattern succeeds, and we work back down the stack to
4084clean up, restoring as we go:
95b24440 4085
c476f425
DM
4086CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
4087CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
4088CO <CO prev=NULL> <WO>
4089NULL
a0374537 4090
c476f425
DM
4091*******************************************************************/
4092
4093#define ST st->u.curlyx
4094
4095 case CURLYX: /* start of /A*B/ (for complex A) */
4096 {
4097 /* No need to save/restore up to this paren */
4098 I32 parenfloor = scan->flags;
4099
4100 assert(next); /* keep Coverity happy */
4101 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
4102 next += ARG(next);
4103
4104 /* XXXX Probably it is better to teach regpush to support
4105 parenfloor > PL_regsize... */
4106 if (parenfloor > (I32)*PL_reglastparen)
4107 parenfloor = *PL_reglastparen; /* Pessimization... */
4108
4109 ST.prev_curlyx= cur_curlyx;
4110 cur_curlyx = st;
4111 ST.cp = PL_savestack_ix;
4112
4113 /* these fields contain the state of the current curly.
4114 * they are accessed by subsequent WHILEMs */
4115 ST.parenfloor = parenfloor;
4116 ST.min = ARG1(scan);
4117 ST.max = ARG2(scan);
4118 ST.A = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
4119 ST.B = next;
24d3c4a9
DM
4120 ST.minmod = minmod;
4121 minmod = 0;
c476f425
DM
4122 ST.count = -1; /* this will be updated by WHILEM */
4123 ST.lastloc = NULL; /* this will be updated by WHILEM */
4124
4125 PL_reginput = locinput;
4126 PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next));
5f66b61c 4127 /* NOTREACHED */
c476f425 4128 }
a0d0e21e 4129
c476f425 4130 case CURLYX_end: /* just finished matching all of A*B */
c476f425
DM
4131 cur_curlyx = ST.prev_curlyx;
4132 sayYES;
4133 /* NOTREACHED */
a0d0e21e 4134
c476f425
DM
4135 case CURLYX_end_fail: /* just failed to match all of A*B */
4136 regcpblow(ST.cp);
4137 cur_curlyx = ST.prev_curlyx;
4138 sayNO;
4139 /* NOTREACHED */
4633a7c4 4140
a0d0e21e 4141
c476f425
DM
4142#undef ST
4143#define ST st->u.whilem
4144
4145 case WHILEM: /* just matched an A in /A*B/ (for complex A) */
4146 {
4147 /* see the discussion above about CURLYX/WHILEM */
c476f425
DM
4148 I32 n;
4149 assert(cur_curlyx); /* keep Coverity happy */
4150 n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */
4151 ST.save_lastloc = cur_curlyx->u.curlyx.lastloc;
4152 ST.cache_offset = 0;
4153 ST.cache_mask = 0;
4154
4155 PL_reginput = locinput;
4156
4157 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4158 "%*s whilem: matched %ld out of %ld..%ld\n",
4159 REPORT_CODE_OFF+depth*2, "", (long)n,
4160 (long)cur_curlyx->u.curlyx.min,
4161 (long)cur_curlyx->u.curlyx.max)
4162 );
a0d0e21e 4163
c476f425 4164 /* First just match a string of min A's. */
a0d0e21e 4165
c476f425
DM
4166 if (n < cur_curlyx->u.curlyx.min) {
4167 cur_curlyx->u.curlyx.lastloc = locinput;
4168 PUSH_STATE_GOTO(WHILEM_A_pre, cur_curlyx->u.curlyx.A);
4169 /* NOTREACHED */
4170 }
4171
4172 /* If degenerate A matches "", assume A done. */
4173
4174 if (locinput == cur_curlyx->u.curlyx.lastloc) {
4175 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4176 "%*s whilem: empty match detected, trying continuation...\n",
4177 REPORT_CODE_OFF+depth*2, "")
4178 );
4179 goto do_whilem_B_max;
4180 }
4181
4182 /* super-linear cache processing */
4183
4184 if (scan->flags) {
a0d0e21e 4185
2c2d71f5 4186 if (!PL_reg_maxiter) {
c476f425
DM
4187 /* start the countdown: Postpone detection until we
4188 * know the match is not *that* much linear. */
2c2d71f5 4189 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
66bf836d
DM
4190 /* possible overflow for long strings and many CURLYX's */
4191 if (PL_reg_maxiter < 0)
4192 PL_reg_maxiter = I32_MAX;
2c2d71f5
JH
4193 PL_reg_leftiter = PL_reg_maxiter;
4194 }
c476f425 4195
2c2d71f5 4196 if (PL_reg_leftiter-- == 0) {
c476f425 4197 /* initialise cache */
3298f257 4198 const I32 size = (PL_reg_maxiter + 7)/8;
2c2d71f5 4199 if (PL_reg_poscache) {
eb160463 4200 if ((I32)PL_reg_poscache_size < size) {
2c2d71f5
JH
4201 Renew(PL_reg_poscache, size, char);
4202 PL_reg_poscache_size = size;
4203 }
4204 Zero(PL_reg_poscache, size, char);
4205 }
4206 else {
4207 PL_reg_poscache_size = size;
a02a5408 4208 Newxz(PL_reg_poscache, size, char);
2c2d71f5 4209 }
c476f425
DM
4210 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4211 "%swhilem: Detected a super-linear match, switching on caching%s...\n",
4212 PL_colors[4], PL_colors[5])
4213 );
2c2d71f5 4214 }
c476f425 4215
2c2d71f5 4216 if (PL_reg_leftiter < 0) {
c476f425
DM
4217 /* have we already failed at this position? */
4218 I32 offset, mask;
4219 offset = (scan->flags & 0xf) - 1
4220 + (locinput - PL_bostr) * (scan->flags>>4);
4221 mask = 1 << (offset % 8);
4222 offset /= 8;
4223 if (PL_reg_poscache[offset] & mask) {
4224 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4225 "%*s whilem: (cache) already tried at this position...\n",
4226 REPORT_CODE_OFF+depth*2, "")
2c2d71f5 4227 );
3298f257 4228 sayNO; /* cache records failure */
2c2d71f5 4229 }
c476f425
DM
4230 ST.cache_offset = offset;
4231 ST.cache_mask = mask;
2c2d71f5 4232 }
c476f425 4233 }
2c2d71f5 4234
c476f425 4235 /* Prefer B over A for minimal matching. */
a687059c 4236
c476f425
DM
4237 if (cur_curlyx->u.curlyx.minmod) {
4238 ST.save_curlyx = cur_curlyx;
4239 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
4240 ST.cp = regcppush(ST.save_curlyx->u.curlyx.parenfloor);
4241 REGCP_SET(ST.lastcp);
4242 PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B);
4243 /* NOTREACHED */
4244 }
a0d0e21e 4245
c476f425
DM
4246 /* Prefer A over B for maximal matching. */
4247
4248 if (n < cur_curlyx->u.curlyx.max) { /* More greed allowed? */
4249 ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
4250 cur_curlyx->u.curlyx.lastloc = locinput;
4251 REGCP_SET(ST.lastcp);
4252 PUSH_STATE_GOTO(WHILEM_A_max, cur_curlyx->u.curlyx.A);
4253 /* NOTREACHED */
4254 }
4255 goto do_whilem_B_max;
4256 }
4257 /* NOTREACHED */
4258
4259 case WHILEM_B_min: /* just matched B in a minimal match */
4260 case WHILEM_B_max: /* just matched B in a maximal match */
4261 cur_curlyx = ST.save_curlyx;
4262 sayYES;
4263 /* NOTREACHED */
4264
4265 case WHILEM_B_max_fail: /* just failed to match B in a maximal match */
4266 cur_curlyx = ST.save_curlyx;
4267 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
4268 cur_curlyx->u.curlyx.count--;
4269 CACHEsayNO;
4270 /* NOTREACHED */
4271
4272 case WHILEM_A_min_fail: /* just failed to match A in a minimal match */
4273 REGCP_UNWIND(ST.lastcp);
4274 regcppop(rex);
4275 /* FALL THROUGH */
4276 case WHILEM_A_pre_fail: /* just failed to match even minimal A */
4277 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
4278 cur_curlyx->u.curlyx.count--;
4279 CACHEsayNO;
4280 /* NOTREACHED */
4281
4282 case WHILEM_A_max_fail: /* just failed to match A in a maximal match */
4283 REGCP_UNWIND(ST.lastcp);
4284 regcppop(rex); /* Restore some previous $<digit>s? */
4285 PL_reginput = locinput;
4286 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4287 "%*s whilem: failed, trying continuation...\n",
4288 REPORT_CODE_OFF+depth*2, "")
4289 );
4290 do_whilem_B_max:
4291 if (cur_curlyx->u.curlyx.count >= REG_INFTY
4292 && ckWARN(WARN_REGEXP)
4293 && !(PL_reg_flags & RF_warned))
4294 {
4295 PL_reg_flags |= RF_warned;
4296 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
4297 "Complex regular subexpression recursion",
4298 REG_INFTY - 1);
4299 }
4300
4301 /* now try B */
4302 ST.save_curlyx = cur_curlyx;
4303 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
4304 PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B);
4305 /* NOTREACHED */
4306
4307 case WHILEM_B_min_fail: /* just failed to match B in a minimal match */
4308 cur_curlyx = ST.save_curlyx;
4309 REGCP_UNWIND(ST.lastcp);
4310 regcppop(rex);
4311
4312 if (cur_curlyx->u.curlyx.count >= cur_curlyx->u.curlyx.max) {
4313 /* Maximum greed exceeded */
4314 if (cur_curlyx->u.curlyx.count >= REG_INFTY
4315 && ckWARN(WARN_REGEXP)
4316 && !(PL_reg_flags & RF_warned))
4317 {
3280af22 4318 PL_reg_flags |= RF_warned;
c476f425
DM
4319 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
4320 "%s limit (%d) exceeded",
4321 "Complex regular subexpression recursion",
4322 REG_INFTY - 1);
a0d0e21e 4323 }
c476f425 4324 cur_curlyx->u.curlyx.count--;
3ab3c9b4 4325 CACHEsayNO;
a0d0e21e 4326 }
c476f425
DM
4327
4328 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4329 "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "")
4330 );
4331 /* Try grabbing another A and see if it helps. */
4332 PL_reginput = locinput;
4333 cur_curlyx->u.curlyx.lastloc = locinput;
4334 ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
4335 REGCP_SET(ST.lastcp);
4336 PUSH_STATE_GOTO(WHILEM_A_min, ST.save_curlyx->u.curlyx.A);
5f66b61c 4337 /* NOTREACHED */
40a82448
DM
4338
4339#undef ST
4340#define ST st->u.branch
4341
4342 case BRANCHJ: /* /(...|A|...)/ with long next pointer */
c277df42
IZ
4343 next = scan + ARG(scan);
4344 if (next == scan)
4345 next = NULL;
40a82448
DM
4346 scan = NEXTOPER(scan);
4347 /* FALL THROUGH */
c277df42 4348
40a82448
DM
4349 case BRANCH: /* /(...|A|...)/ */
4350 scan = NEXTOPER(scan); /* scan now points to inner node */
40a82448
DM
4351 ST.lastparen = *PL_reglastparen;
4352 ST.next_branch = next;
4353 REGCP_SET(ST.cp);
4354 PL_reginput = locinput;
02db2b7b 4355
40a82448 4356 /* Now go into the branch */
5d458dd8
YO
4357 if (has_cutgroup) {
4358 PUSH_YES_STATE_GOTO(BRANCH_next, scan);
4359 } else {
4360 PUSH_STATE_GOTO(BRANCH_next, scan);
4361 }
40a82448 4362 /* NOTREACHED */
5d458dd8
YO
4363 case CUTGROUP:
4364 PL_reginput = locinput;
4365 sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL :
ad64d0ec 4366 MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
5d458dd8
YO
4367 PUSH_STATE_GOTO(CUTGROUP_next,next);
4368 /* NOTREACHED */
4369 case CUTGROUP_next_fail:
4370 do_cutgroup = 1;
4371 no_final = 1;
4372 if (st->u.mark.mark_name)
4373 sv_commit = st->u.mark.mark_name;
4374 sayNO;
4375 /* NOTREACHED */
4376 case BRANCH_next:
4377 sayYES;
4378 /* NOTREACHED */
40a82448 4379 case BRANCH_next_fail: /* that branch failed; try the next, if any */
5d458dd8
YO
4380 if (do_cutgroup) {
4381 do_cutgroup = 0;
4382 no_final = 0;
4383 }
40a82448
DM
4384 REGCP_UNWIND(ST.cp);
4385 for (n = *PL_reglastparen; n > ST.lastparen; n--)
f0ab9afb 4386 PL_regoffs[n].end = -1;
40a82448 4387 *PL_reglastparen = n;
0a4db386 4388 /*dmq: *PL_reglastcloseparen = n; */
40a82448
DM
4389 scan = ST.next_branch;
4390 /* no more branches? */
5d458dd8
YO
4391 if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) {
4392 DEBUG_EXECUTE_r({
4393 PerlIO_printf( Perl_debug_log,
4394 "%*s %sBRANCH failed...%s\n",
4395 REPORT_CODE_OFF+depth*2, "",
4396 PL_colors[4],
4397 PL_colors[5] );
4398 });
4399 sayNO_SILENT;
4400 }
40a82448
DM
4401 continue; /* execute next BRANCH[J] op */
4402 /* NOTREACHED */
4403
a0d0e21e 4404 case MINMOD:
24d3c4a9 4405 minmod = 1;
a0d0e21e 4406 break;
40a82448
DM
4407
4408#undef ST
4409#define ST st->u.curlym
4410
4411 case CURLYM: /* /A{m,n}B/ where A is fixed-length */
4412
4413 /* This is an optimisation of CURLYX that enables us to push
4414 * only a single backtracking state, no matter now many matches
4415 * there are in {m,n}. It relies on the pattern being constant
4416 * length, with no parens to influence future backrefs
4417 */
4418
4419 ST.me = scan;
dc45a647 4420 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
40a82448
DM
4421
4422 /* if paren positive, emulate an OPEN/CLOSE around A */
4423 if (ST.me->flags) {
3b6647e0 4424 U32 paren = ST.me->flags;
40a82448
DM
4425 if (paren > PL_regsize)
4426 PL_regsize = paren;
3b6647e0 4427 if (paren > *PL_reglastparen)
40a82448 4428 *PL_reglastparen = paren;
c277df42 4429 scan += NEXT_OFF(scan); /* Skip former OPEN. */
6407bf3b 4430 }
40a82448
DM
4431 ST.A = scan;
4432 ST.B = next;
4433 ST.alen = 0;
4434 ST.count = 0;
24d3c4a9
DM
4435 ST.minmod = minmod;
4436 minmod = 0;
40a82448
DM
4437 ST.c1 = CHRTEST_UNINIT;
4438 REGCP_SET(ST.cp);
6407bf3b 4439
40a82448
DM
4440 if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */
4441 goto curlym_do_B;
4442
4443 curlym_do_A: /* execute the A in /A{m,n}B/ */
6407bf3b 4444 PL_reginput = locinput;
40a82448
DM
4445 PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */
4446 /* NOTREACHED */
5f80c4cf 4447
40a82448
DM
4448 case CURLYM_A: /* we've just matched an A */
4449 locinput = st->locinput;
4450 nextchr = UCHARAT(locinput);
4451
4452 ST.count++;
4453 /* after first match, determine A's length: u.curlym.alen */
4454 if (ST.count == 1) {
4455 if (PL_reg_match_utf8) {
4456 char *s = locinput;
4457 while (s < PL_reginput) {
4458 ST.alen++;
4459 s += UTF8SKIP(s);
4460 }
4461 }
4462 else {
4463 ST.alen = PL_reginput - locinput;
4464 }
4465 if (ST.alen == 0)
4466 ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me);
4467 }
0cadcf80
DM
4468 DEBUG_EXECUTE_r(
4469 PerlIO_printf(Perl_debug_log,
40a82448 4470 "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n",
5bc10b2c 4471 (int)(REPORT_CODE_OFF+(depth*2)), "",
40a82448 4472 (IV) ST.count, (IV)ST.alen)
0cadcf80
DM
4473 );
4474
40a82448 4475 locinput = PL_reginput;
0a4db386
YO
4476
4477 if (cur_eval && cur_eval->u.eval.close_paren &&
24b23f37 4478 cur_eval->u.eval.close_paren == (U32)ST.me->flags)
0a4db386
YO
4479 goto fake_end;
4480
c966426a
DM
4481 {
4482 I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me));
4483 if ( max == REG_INFTY || ST.count < max )
4484 goto curlym_do_A; /* try to match another A */
4485 }
40a82448 4486 goto curlym_do_B; /* try to match B */
5f80c4cf 4487
40a82448
DM
4488 case CURLYM_A_fail: /* just failed to match an A */
4489 REGCP_UNWIND(ST.cp);
0a4db386
YO
4490
4491 if (ST.minmod || ST.count < ARG1(ST.me) /* min*/
4492 || (cur_eval && cur_eval->u.eval.close_paren &&
24b23f37 4493 cur_eval->u.eval.close_paren == (U32)ST.me->flags))
40a82448 4494 sayNO;
0cadcf80 4495
40a82448
DM
4496 curlym_do_B: /* execute the B in /A{m,n}B/ */
4497 PL_reginput = locinput;
4498 if (ST.c1 == CHRTEST_UNINIT) {
4499 /* calculate c1 and c2 for possible match of 1st char
4500 * following curly */
4501 ST.c1 = ST.c2 = CHRTEST_VOID;
4502 if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) {
4503 regnode *text_node = ST.B;
4504 if (! HAS_TEXT(text_node))
4505 FIND_NEXT_IMPT(text_node);
ee9b8eae
YO
4506 /* this used to be
4507
4508 (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT)
4509
4510 But the former is redundant in light of the latter.
4511
4512 if this changes back then the macro for
4513 IS_TEXT and friends need to change.
4514 */
4515 if (PL_regkind[OP(text_node)] == EXACT)
40a82448 4516 {
ee9b8eae 4517
40a82448
DM
4518 ST.c1 = (U8)*STRING(text_node);
4519 ST.c2 =
ee9b8eae 4520 (IS_TEXTF(text_node))
40a82448 4521 ? PL_fold[ST.c1]
ee9b8eae 4522 : (IS_TEXTFL(text_node))
40a82448
DM
4523 ? PL_fold_locale[ST.c1]
4524 : ST.c1;
c277df42 4525 }
c277df42 4526 }
40a82448
DM
4527 }
4528
4529 DEBUG_EXECUTE_r(
4530 PerlIO_printf(Perl_debug_log,
4531 "%*s CURLYM trying tail with matches=%"IVdf"...\n",
5bc10b2c 4532 (int)(REPORT_CODE_OFF+(depth*2)),
40a82448
DM
4533 "", (IV)ST.count)
4534 );
4535 if (ST.c1 != CHRTEST_VOID
4536 && UCHARAT(PL_reginput) != ST.c1
4537 && UCHARAT(PL_reginput) != ST.c2)
4538 {
4539 /* simulate B failing */
3e901dc0
YO
4540 DEBUG_OPTIMISE_r(
4541 PerlIO_printf(Perl_debug_log,
4542 "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n",
4543 (int)(REPORT_CODE_OFF+(depth*2)),"",
4544 (IV)ST.c1,(IV)ST.c2
4545 ));
40a82448
DM
4546 state_num = CURLYM_B_fail;
4547 goto reenter_switch;
4548 }
4549
4550 if (ST.me->flags) {
4551 /* mark current A as captured */
4552 I32 paren = ST.me->flags;
4553 if (ST.count) {
f0ab9afb 4554 PL_regoffs[paren].start
40a82448 4555 = HOPc(PL_reginput, -ST.alen) - PL_bostr;
f0ab9afb 4556 PL_regoffs[paren].end = PL_reginput - PL_bostr;
0a4db386 4557 /*dmq: *PL_reglastcloseparen = paren; */
c277df42 4558 }
40a82448 4559 else
f0ab9afb 4560 PL_regoffs[paren].end = -1;
0a4db386 4561 if (cur_eval && cur_eval->u.eval.close_paren &&
24b23f37 4562 cur_eval->u.eval.close_paren == (U32)ST.me->flags)
0a4db386
YO
4563 {
4564 if (ST.count)
4565 goto fake_end;
4566 else
4567 sayNO;
4568 }
c277df42 4569 }
0a4db386 4570
40a82448 4571 PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */
5f66b61c 4572 /* NOTREACHED */
40a82448
DM
4573
4574 case CURLYM_B_fail: /* just failed to match a B */
4575 REGCP_UNWIND(ST.cp);
4576 if (ST.minmod) {
4577 if (ST.count == ARG2(ST.me) /* max */)
4578 sayNO;
4579 goto curlym_do_A; /* try to match a further A */
4580 }
4581 /* backtrack one A */
4582 if (ST.count == ARG1(ST.me) /* min */)
4583 sayNO;
4584 ST.count--;
4585 locinput = HOPc(locinput, -ST.alen);
4586 goto curlym_do_B; /* try to match B */
4587
c255a977
DM
4588#undef ST
4589#define ST st->u.curly
40a82448 4590
c255a977
DM
4591#define CURLY_SETPAREN(paren, success) \
4592 if (paren) { \
4593 if (success) { \
f0ab9afb
NC
4594 PL_regoffs[paren].start = HOPc(locinput, -1) - PL_bostr; \
4595 PL_regoffs[paren].end = locinput - PL_bostr; \
0a4db386 4596 *PL_reglastcloseparen = paren; \
c255a977
DM
4597 } \
4598 else \
f0ab9afb 4599 PL_regoffs[paren].end = -1; \
c255a977
DM
4600 }
4601
4602 case STAR: /* /A*B/ where A is width 1 */
4603 ST.paren = 0;
4604 ST.min = 0;
4605 ST.max = REG_INFTY;
a0d0e21e
LW
4606 scan = NEXTOPER(scan);
4607 goto repeat;
c255a977
DM
4608 case PLUS: /* /A+B/ where A is width 1 */
4609 ST.paren = 0;
4610 ST.min = 1;
4611 ST.max = REG_INFTY;
c277df42 4612 scan = NEXTOPER(scan);
c255a977
DM
4613 goto repeat;
4614 case CURLYN: /* /(A){m,n}B/ where A is width 1 */
4615 ST.paren = scan->flags; /* Which paren to set */
4616 if (ST.paren > PL_regsize)
4617 PL_regsize = ST.paren;
3b6647e0 4618 if (ST.paren > *PL_reglastparen)
c255a977
DM
4619 *PL_reglastparen = ST.paren;
4620 ST.min = ARG1(scan); /* min to match */
4621 ST.max = ARG2(scan); /* max to match */
0a4db386 4622 if (cur_eval && cur_eval->u.eval.close_paren &&
86413ec0 4623 cur_eval->u.eval.close_paren == (U32)ST.paren) {
0a4db386
YO
4624 ST.min=1;
4625 ST.max=1;
4626 }
c255a977
DM
4627 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
4628 goto repeat;
4629 case CURLY: /* /A{m,n}B/ where A is width 1 */
4630 ST.paren = 0;
4631 ST.min = ARG1(scan); /* min to match */
4632 ST.max = ARG2(scan); /* max to match */
4633 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
c277df42 4634 repeat:
a0d0e21e
LW
4635 /*
4636 * Lookahead to avoid useless match attempts
4637 * when we know what character comes next.
c255a977 4638 *
5f80c4cf
JP
4639 * Used to only do .*x and .*?x, but now it allows
4640 * for )'s, ('s and (?{ ... })'s to be in the way
4641 * of the quantifier and the EXACT-like node. -- japhy
4642 */
4643
c255a977
DM
4644 if (ST.min > ST.max) /* XXX make this a compile-time check? */
4645 sayNO;
cca55fe3 4646 if (HAS_TEXT(next) || JUMPABLE(next)) {
5f80c4cf
JP
4647 U8 *s;
4648 regnode *text_node = next;
4649
3dab1dad
YO
4650 if (! HAS_TEXT(text_node))
4651 FIND_NEXT_IMPT(text_node);
5f80c4cf 4652
9e137952 4653 if (! HAS_TEXT(text_node))
c255a977 4654 ST.c1 = ST.c2 = CHRTEST_VOID;
5f80c4cf 4655 else {
ee9b8eae 4656 if ( PL_regkind[OP(text_node)] != EXACT ) {
c255a977 4657 ST.c1 = ST.c2 = CHRTEST_VOID;
44a68960 4658 goto assume_ok_easy;
cca55fe3 4659 }
be8e71aa
YO
4660 else
4661 s = (U8*)STRING(text_node);
ee9b8eae
YO
4662
4663 /* Currently we only get here when
4664
4665 PL_rekind[OP(text_node)] == EXACT
4666
4667 if this changes back then the macro for IS_TEXT and
4668 friends need to change. */
5f80c4cf 4669 if (!UTF) {
c255a977 4670 ST.c2 = ST.c1 = *s;
ee9b8eae 4671 if (IS_TEXTF(text_node))
c255a977 4672 ST.c2 = PL_fold[ST.c1];
ee9b8eae 4673 else if (IS_TEXTFL(text_node))
c255a977 4674 ST.c2 = PL_fold_locale[ST.c1];
1aa99e6b 4675 }
5f80c4cf 4676 else { /* UTF */
ee9b8eae 4677 if (IS_TEXTF(text_node)) {
a2a2844f 4678 STRLEN ulen1, ulen2;
89ebb4a3
JH
4679 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
4680 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
a2a2844f
JH
4681
4682 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
4683 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
e294cc5d
JH
4684#ifdef EBCDIC
4685 ST.c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN, 0,
4686 ckWARN(WARN_UTF8) ?
4687 0 : UTF8_ALLOW_ANY);
4688 ST.c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN, 0,
4689 ckWARN(WARN_UTF8) ?
4690 0 : UTF8_ALLOW_ANY);
4691#else
c255a977 4692 ST.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
e294cc5d 4693 uniflags);
c255a977 4694 ST.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
e294cc5d
JH
4695 uniflags);
4696#endif
5f80c4cf
JP
4697 }
4698 else {
c255a977 4699 ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
041457d9 4700 uniflags);
5f80c4cf 4701 }
1aa99e6b
IH
4702 }
4703 }
bbce6d69 4704 }
a0d0e21e 4705 else
c255a977 4706 ST.c1 = ST.c2 = CHRTEST_VOID;
cca55fe3 4707 assume_ok_easy:
c255a977
DM
4708
4709 ST.A = scan;
4710 ST.B = next;
3280af22 4711 PL_reginput = locinput;
24d3c4a9
DM
4712 if (minmod) {
4713 minmod = 0;
e2e6a0f1 4714 if (ST.min && regrepeat(rex, ST.A, ST.min, depth) < ST.min)
4633a7c4 4715 sayNO;
c255a977 4716 ST.count = ST.min;
a0ed51b3 4717 locinput = PL_reginput;
c255a977
DM
4718 REGCP_SET(ST.cp);
4719 if (ST.c1 == CHRTEST_VOID)
4720 goto curly_try_B_min;
4721
4722 ST.oldloc = locinput;
4723
4724 /* set ST.maxpos to the furthest point along the
4725 * string that could possibly match */
4726 if (ST.max == REG_INFTY) {
4727 ST.maxpos = PL_regeol - 1;
4728 if (do_utf8)
4729 while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos))
4730 ST.maxpos--;
4731 }
4732 else if (do_utf8) {
4733 int m = ST.max - ST.min;
4734 for (ST.maxpos = locinput;
4735 m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--)
4736 ST.maxpos += UTF8SKIP(ST.maxpos);
4737 }
4738 else {
4739 ST.maxpos = locinput + ST.max - ST.min;
4740 if (ST.maxpos >= PL_regeol)
4741 ST.maxpos = PL_regeol - 1;
4742 }
4743 goto curly_try_B_min_known;
4744
4745 }
4746 else {
e2e6a0f1 4747 ST.count = regrepeat(rex, ST.A, ST.max, depth);
c255a977
DM
4748 locinput = PL_reginput;
4749 if (ST.count < ST.min)
4750 sayNO;
4751 if ((ST.count > ST.min)
4752 && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL))
4753 {
4754 /* A{m,n} must come at the end of the string, there's
4755 * no point in backing off ... */
4756 ST.min = ST.count;
4757 /* ...except that $ and \Z can match before *and* after
4758 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
4759 We may back off by one in this case. */
4760 if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS)
4761 ST.min--;
4762 }
4763 REGCP_SET(ST.cp);
4764 goto curly_try_B_max;
4765 }
4766 /* NOTREACHED */
4767
4768
4769 case CURLY_B_min_known_fail:
4770 /* failed to find B in a non-greedy match where c1,c2 valid */
4771 if (ST.paren && ST.count)
f0ab9afb 4772 PL_regoffs[ST.paren].end = -1;
c255a977
DM
4773
4774 PL_reginput = locinput; /* Could be reset... */
4775 REGCP_UNWIND(ST.cp);
4776 /* Couldn't or didn't -- move forward. */
4777 ST.oldloc = locinput;
4778 if (do_utf8)
4779 locinput += UTF8SKIP(locinput);
4780 else
4781 locinput++;
4782 ST.count++;
4783 curly_try_B_min_known:
4784 /* find the next place where 'B' could work, then call B */
4785 {
4786 int n;
4787 if (do_utf8) {
4788 n = (ST.oldloc == locinput) ? 0 : 1;
4789 if (ST.c1 == ST.c2) {
4790 STRLEN len;
4791 /* set n to utf8_distance(oldloc, locinput) */
4792 while (locinput <= ST.maxpos &&
4793 utf8n_to_uvchr((U8*)locinput,
4794 UTF8_MAXBYTES, &len,
4795 uniflags) != (UV)ST.c1) {
4796 locinput += len;
4797 n++;
4798 }
1aa99e6b
IH
4799 }
4800 else {
c255a977
DM
4801 /* set n to utf8_distance(oldloc, locinput) */
4802 while (locinput <= ST.maxpos) {
4803 STRLEN len;
4804 const UV c = utf8n_to_uvchr((U8*)locinput,
4805 UTF8_MAXBYTES, &len,
4806 uniflags);
4807 if (c == (UV)ST.c1 || c == (UV)ST.c2)
4808 break;
4809 locinput += len;
4810 n++;
1aa99e6b 4811 }
0fe9bf95
IZ
4812 }
4813 }
c255a977
DM
4814 else {
4815 if (ST.c1 == ST.c2) {
4816 while (locinput <= ST.maxpos &&
4817 UCHARAT(locinput) != ST.c1)
4818 locinput++;
bbce6d69 4819 }
c255a977
DM
4820 else {
4821 while (locinput <= ST.maxpos
4822 && UCHARAT(locinput) != ST.c1
4823 && UCHARAT(locinput) != ST.c2)
4824 locinput++;
a0ed51b3 4825 }
c255a977
DM
4826 n = locinput - ST.oldloc;
4827 }
4828 if (locinput > ST.maxpos)
4829 sayNO;
4830 /* PL_reginput == oldloc now */
4831 if (n) {
4832 ST.count += n;
e2e6a0f1 4833 if (regrepeat(rex, ST.A, n, depth) < n)
4633a7c4 4834 sayNO;
a0d0e21e 4835 }
c255a977
DM
4836 PL_reginput = locinput;
4837 CURLY_SETPAREN(ST.paren, ST.count);
0a4db386 4838 if (cur_eval && cur_eval->u.eval.close_paren &&
86413ec0 4839 cur_eval->u.eval.close_paren == (U32)ST.paren) {
0a4db386
YO
4840 goto fake_end;
4841 }
c255a977 4842 PUSH_STATE_GOTO(CURLY_B_min_known, ST.B);
a0d0e21e 4843 }
c255a977
DM
4844 /* NOTREACHED */
4845
4846
4847 case CURLY_B_min_fail:
4848 /* failed to find B in a non-greedy match where c1,c2 invalid */
4849 if (ST.paren && ST.count)
f0ab9afb 4850 PL_regoffs[ST.paren].end = -1;
c255a977
DM
4851
4852 REGCP_UNWIND(ST.cp);
4853 /* failed -- move forward one */
4854 PL_reginput = locinput;
e2e6a0f1 4855 if (regrepeat(rex, ST.A, 1, depth)) {
c255a977 4856 ST.count++;
a0ed51b3 4857 locinput = PL_reginput;
c255a977
DM
4858 if (ST.count <= ST.max || (ST.max == REG_INFTY &&
4859 ST.count > 0)) /* count overflow ? */
15272685 4860 {
c255a977
DM
4861 curly_try_B_min:
4862 CURLY_SETPAREN(ST.paren, ST.count);
0a4db386 4863 if (cur_eval && cur_eval->u.eval.close_paren &&
86413ec0 4864 cur_eval->u.eval.close_paren == (U32)ST.paren) {
0a4db386
YO
4865 goto fake_end;
4866 }
c255a977 4867 PUSH_STATE_GOTO(CURLY_B_min, ST.B);
a0d0e21e
LW
4868 }
4869 }
4633a7c4 4870 sayNO;
c255a977
DM
4871 /* NOTREACHED */
4872
4873
4874 curly_try_B_max:
4875 /* a successful greedy match: now try to match B */
40d049e4 4876 if (cur_eval && cur_eval->u.eval.close_paren &&
86413ec0 4877 cur_eval->u.eval.close_paren == (U32)ST.paren) {
40d049e4
YO
4878 goto fake_end;
4879 }
c255a977
DM
4880 {
4881 UV c = 0;
4882 if (ST.c1 != CHRTEST_VOID)
4883 c = do_utf8 ? utf8n_to_uvchr((U8*)PL_reginput,
4884 UTF8_MAXBYTES, 0, uniflags)
466787eb 4885 : (UV) UCHARAT(PL_reginput);
c255a977
DM
4886 /* If it could work, try it. */
4887 if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) {
4888 CURLY_SETPAREN(ST.paren, ST.count);
4889 PUSH_STATE_GOTO(CURLY_B_max, ST.B);
4890 /* NOTREACHED */
4891 }
4892 }
4893 /* FALL THROUGH */
4894 case CURLY_B_max_fail:
4895 /* failed to find B in a greedy match */
4896 if (ST.paren && ST.count)
f0ab9afb 4897 PL_regoffs[ST.paren].end = -1;
c255a977
DM
4898
4899 REGCP_UNWIND(ST.cp);
4900 /* back up. */
4901 if (--ST.count < ST.min)
4902 sayNO;
4903 PL_reginput = locinput = HOPc(locinput, -1);
4904 goto curly_try_B_max;
4905
4906#undef ST
4907
a0d0e21e 4908 case END:
6bda09f9 4909 fake_end:
faec1544
DM
4910 if (cur_eval) {
4911 /* we've just finished A in /(??{A})B/; now continue with B */
4912 I32 tmpix;
faec1544
DM
4913 st->u.eval.toggle_reg_flags
4914 = cur_eval->u.eval.toggle_reg_flags;
4915 PL_reg_flags ^= st->u.eval.toggle_reg_flags;
4916
288b8c02
NC
4917 st->u.eval.prev_rex = rex_sv; /* inner */
4918 SETREX(rex_sv,cur_eval->u.eval.prev_rex);
4919 rex = (struct regexp *)SvANY(rex_sv);
f8fc2ecf 4920 rexi = RXi_GET(rex);
faec1544 4921 cur_curlyx = cur_eval->u.eval.prev_curlyx;
288b8c02 4922 ReREFCNT_inc(rex_sv);
faec1544 4923 st->u.eval.cp = regcppush(0); /* Save *all* the positions. */
34a81e2b
B
4924
4925 /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */
4926 PL_reglastparen = &rex->lastparen;
4927 PL_reglastcloseparen = &rex->lastcloseparen;
4928
faec1544
DM
4929 REGCP_SET(st->u.eval.lastcp);
4930 PL_reginput = locinput;
4931
4932 /* Restore parens of the outer rex without popping the
4933 * savestack */
4934 tmpix = PL_savestack_ix;
4935 PL_savestack_ix = cur_eval->u.eval.lastcp;
4936 regcppop(rex);
4937 PL_savestack_ix = tmpix;
4938
4939 st->u.eval.prev_eval = cur_eval;
4940 cur_eval = cur_eval->u.eval.prev_eval;
4941 DEBUG_EXECUTE_r(
2a49f0f5
JH
4942 PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n",
4943 REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval)););
e7707071
YO
4944 if ( nochange_depth )
4945 nochange_depth--;
4946
4947 PUSH_YES_STATE_GOTO(EVAL_AB,
faec1544
DM
4948 st->u.eval.prev_eval->u.eval.B); /* match B */
4949 }
4950
3b0527fe 4951 if (locinput < reginfo->till) {
a3621e74 4952 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
7821416a
IZ
4953 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
4954 PL_colors[4],
4955 (long)(locinput - PL_reg_starttry),
3b0527fe 4956 (long)(reginfo->till - PL_reg_starttry),
7821416a 4957 PL_colors[5]));
58e23c8d 4958
262b90c4 4959 sayNO_SILENT; /* Cannot match: too short. */
7821416a
IZ
4960 }
4961 PL_reginput = locinput; /* put where regtry can find it */
262b90c4 4962 sayYES; /* Success! */
dad79028
DM
4963
4964 case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */
4965 DEBUG_EXECUTE_r(
4966 PerlIO_printf(Perl_debug_log,
4967 "%*s %ssubpattern success...%s\n",
5bc10b2c 4968 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]));
3280af22 4969 PL_reginput = locinput; /* put where regtry can find it */
262b90c4 4970 sayYES; /* Success! */
dad79028 4971
40a82448
DM
4972#undef ST
4973#define ST st->u.ifmatch
4974
4975 case SUSPEND: /* (?>A) */
4976 ST.wanted = 1;
9fe1d20c 4977 PL_reginput = locinput;
9041c2e3 4978 goto do_ifmatch;
dad79028 4979
40a82448
DM
4980 case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */
4981 ST.wanted = 0;
dad79028
DM
4982 goto ifmatch_trivial_fail_test;
4983
40a82448
DM
4984 case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */
4985 ST.wanted = 1;
dad79028 4986 ifmatch_trivial_fail_test:
a0ed51b3 4987 if (scan->flags) {
52657f30 4988 char * const s = HOPBACKc(locinput, scan->flags);
dad79028
DM
4989 if (!s) {
4990 /* trivial fail */
24d3c4a9
DM
4991 if (logical) {
4992 logical = 0;
4993 sw = 1 - (bool)ST.wanted;
dad79028 4994 }
40a82448 4995 else if (ST.wanted)
dad79028
DM
4996 sayNO;
4997 next = scan + ARG(scan);
4998 if (next == scan)
4999 next = NULL;
5000 break;
5001 }
efb30f32 5002 PL_reginput = s;
a0ed51b3
LW
5003 }
5004 else
5005 PL_reginput = locinput;
5006
c277df42 5007 do_ifmatch:
40a82448 5008 ST.me = scan;
24d3c4a9 5009 ST.logical = logical;
24d786f4
YO
5010 logical = 0; /* XXX: reset state of logical once it has been saved into ST */
5011
40a82448
DM
5012 /* execute body of (?...A) */
5013 PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan)));
5014 /* NOTREACHED */
5015
5016 case IFMATCH_A_fail: /* body of (?...A) failed */
5017 ST.wanted = !ST.wanted;
5018 /* FALL THROUGH */
5019
5020 case IFMATCH_A: /* body of (?...A) succeeded */
24d3c4a9
DM
5021 if (ST.logical) {
5022 sw = (bool)ST.wanted;
40a82448
DM
5023 }
5024 else if (!ST.wanted)
5025 sayNO;
5026
5027 if (OP(ST.me) == SUSPEND)
5028 locinput = PL_reginput;
5029 else {
5030 locinput = PL_reginput = st->locinput;
5031 nextchr = UCHARAT(locinput);
5032 }
5033 scan = ST.me + ARG(ST.me);
5034 if (scan == ST.me)
5035 scan = NULL;
5036 continue; /* execute B */
5037
5038#undef ST
dad79028 5039
c277df42 5040 case LONGJMP:
c277df42
IZ
5041 next = scan + ARG(scan);
5042 if (next == scan)
5043 next = NULL;
a0d0e21e 5044 break;
54612592 5045 case COMMIT:
e2e6a0f1
YO
5046 reginfo->cutpoint = PL_regeol;
5047 /* FALLTHROUGH */
5d458dd8 5048 case PRUNE:
24b23f37 5049 PL_reginput = locinput;
e2e6a0f1 5050 if (!scan->flags)
ad64d0ec 5051 sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
54612592
YO
5052 PUSH_STATE_GOTO(COMMIT_next,next);
5053 /* NOTREACHED */
5054 case COMMIT_next_fail:
5055 no_final = 1;
5056 /* FALLTHROUGH */
7f69552c
YO
5057 case OPFAIL:
5058 sayNO;
e2e6a0f1
YO
5059 /* NOTREACHED */
5060
5061#define ST st->u.mark
5062 case MARKPOINT:
5063 ST.prev_mark = mark_state;
5d458dd8 5064 ST.mark_name = sv_commit = sv_yes_mark
ad64d0ec 5065 = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
e2e6a0f1
YO
5066 mark_state = st;
5067 ST.mark_loc = PL_reginput = locinput;
5068 PUSH_YES_STATE_GOTO(MARKPOINT_next,next);
5069 /* NOTREACHED */
5070 case MARKPOINT_next:
5071 mark_state = ST.prev_mark;
5072 sayYES;
5073 /* NOTREACHED */
5074 case MARKPOINT_next_fail:
5d458dd8 5075 if (popmark && sv_eq(ST.mark_name,popmark))
e2e6a0f1
YO
5076 {
5077 if (ST.mark_loc > startpoint)
5078 reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1);
5079 popmark = NULL; /* we found our mark */
5080 sv_commit = ST.mark_name;
5081
5082 DEBUG_EXECUTE_r({
5d458dd8 5083 PerlIO_printf(Perl_debug_log,
e2e6a0f1
YO
5084 "%*s %ssetting cutpoint to mark:%"SVf"...%s\n",
5085 REPORT_CODE_OFF+depth*2, "",
be2597df 5086 PL_colors[4], SVfARG(sv_commit), PL_colors[5]);
e2e6a0f1
YO
5087 });
5088 }
5089 mark_state = ST.prev_mark;
5d458dd8
YO
5090 sv_yes_mark = mark_state ?
5091 mark_state->u.mark.mark_name : NULL;
e2e6a0f1
YO
5092 sayNO;
5093 /* NOTREACHED */
5d458dd8
YO
5094 case SKIP:
5095 PL_reginput = locinput;
5096 if (scan->flags) {
2bf803e2 5097 /* (*SKIP) : if we fail we cut here*/
5d458dd8 5098 ST.mark_name = NULL;
e2e6a0f1 5099 ST.mark_loc = locinput;
5d458dd8
YO
5100 PUSH_STATE_GOTO(SKIP_next,next);
5101 } else {
2bf803e2 5102 /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was,
5d458dd8
YO
5103 otherwise do nothing. Meaning we need to scan
5104 */
5105 regmatch_state *cur = mark_state;
ad64d0ec 5106 SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]);
5d458dd8
YO
5107
5108 while (cur) {
5109 if ( sv_eq( cur->u.mark.mark_name,
5110 find ) )
5111 {
5112 ST.mark_name = find;
5113 PUSH_STATE_GOTO( SKIP_next, next );
5114 }
5115 cur = cur->u.mark.prev_mark;
5116 }
e2e6a0f1 5117 }
2bf803e2 5118 /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */
5d458dd8
YO
5119 break;
5120 case SKIP_next_fail:
5121 if (ST.mark_name) {
5122 /* (*CUT:NAME) - Set up to search for the name as we
5123 collapse the stack*/
5124 popmark = ST.mark_name;
5125 } else {
5126 /* (*CUT) - No name, we cut here.*/
e2e6a0f1
YO
5127 if (ST.mark_loc > startpoint)
5128 reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1);
5d458dd8
YO
5129 /* but we set sv_commit to latest mark_name if there
5130 is one so they can test to see how things lead to this
5131 cut */
5132 if (mark_state)
5133 sv_commit=mark_state->u.mark.mark_name;
5134 }
e2e6a0f1
YO
5135 no_final = 1;
5136 sayNO;
5137 /* NOTREACHED */
5138#undef ST
32e6a07c
YO
5139 case FOLDCHAR:
5140 n = ARG(scan);
81d4fa0f 5141 if ( n == (U32)what_len_TRICKYFOLD(locinput,do_utf8,ln) ) {
e64b1bd1
YO
5142 locinput += ln;
5143 } else if ( 0xDF == n && !do_utf8 && !UTF ) {
5144 sayNO;
5145 } else {
5146 U8 folded[UTF8_MAXBYTES_CASE+1];
5147 STRLEN foldlen;
5148 const char * const l = locinput;
5149 char *e = PL_regeol;
5150 to_uni_fold(n, folded, &foldlen);
5151
59fe32ea 5152 if (ibcmp_utf8((const char*) folded, 0, foldlen, 1,
e64b1bd1 5153 l, &e, 0, do_utf8)) {
32e6a07c 5154 sayNO;
e64b1bd1
YO
5155 }
5156 locinput = e;
32e6a07c
YO
5157 }
5158 nextchr = UCHARAT(locinput);
5159 break;
e1d1eefb
YO
5160 case LNBREAK:
5161 if ((n=is_LNBREAK(locinput,do_utf8))) {
5162 locinput += n;
5163 nextchr = UCHARAT(locinput);
5164 } else
5165 sayNO;
5166 break;
5167
5168#define CASE_CLASS(nAmE) \
5169 case nAmE: \
5170 if ((n=is_##nAmE(locinput,do_utf8))) { \
5171 locinput += n; \
5172 nextchr = UCHARAT(locinput); \
5173 } else \
5174 sayNO; \
5175 break; \
5176 case N##nAmE: \
5177 if ((n=is_##nAmE(locinput,do_utf8))) { \
5178 sayNO; \
5179 } else { \
5180 locinput += UTF8SKIP(locinput); \
5181 nextchr = UCHARAT(locinput); \
5182 } \
5183 break
5184
5185 CASE_CLASS(VERTWS);
5186 CASE_CLASS(HORIZWS);
5187#undef CASE_CLASS
5188
a0d0e21e 5189 default:
b900a521 5190 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
d7d93a81 5191 PTR2UV(scan), OP(scan));
cea2e8a9 5192 Perl_croak(aTHX_ "regexp memory corruption");
5d458dd8
YO
5193
5194 } /* end switch */
95b24440 5195
5d458dd8
YO
5196 /* switch break jumps here */
5197 scan = next; /* prepare to execute the next op and ... */
5198 continue; /* ... jump back to the top, reusing st */
95b24440
DM
5199 /* NOTREACHED */
5200
40a82448
DM
5201 push_yes_state:
5202 /* push a state that backtracks on success */
5203 st->u.yes.prev_yes_state = yes_state;
5204 yes_state = st;
5205 /* FALL THROUGH */
5206 push_state:
5207 /* push a new regex state, then continue at scan */
5208 {
5209 regmatch_state *newst;
5210
24b23f37
YO
5211 DEBUG_STACK_r({
5212 regmatch_state *cur = st;
5213 regmatch_state *curyes = yes_state;
5214 int curd = depth;
5215 regmatch_slab *slab = PL_regmatch_slab;
5216 for (;curd > -1;cur--,curd--) {
5217 if (cur < SLAB_FIRST(slab)) {
5218 slab = slab->prev;
5219 cur = SLAB_LAST(slab);
5220 }
5221 PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n",
5222 REPORT_CODE_OFF + 2 + depth * 2,"",
13d6edb4 5223 curd, PL_reg_name[cur->resume_state],
24b23f37
YO
5224 (curyes == cur) ? "yes" : ""
5225 );
5226 if (curyes == cur)
5227 curyes = cur->u.yes.prev_yes_state;
5228 }
5229 } else
5230 DEBUG_STATE_pp("push")
5231 );
40a82448 5232 depth++;
40a82448
DM
5233 st->locinput = locinput;
5234 newst = st+1;
5235 if (newst > SLAB_LAST(PL_regmatch_slab))
5236 newst = S_push_slab(aTHX);
5237 PL_regmatch_state = newst;
786e8c11 5238
40a82448
DM
5239 locinput = PL_reginput;
5240 nextchr = UCHARAT(locinput);
5241 st = newst;
5242 continue;
5243 /* NOTREACHED */
5244 }
a0d0e21e 5245 }
a687059c 5246
a0d0e21e
LW
5247 /*
5248 * We get here only if there's trouble -- normally "case END" is
5249 * the terminating point.
5250 */
cea2e8a9 5251 Perl_croak(aTHX_ "corrupted regexp pointers");
a0d0e21e 5252 /*NOTREACHED*/
4633a7c4
LW
5253 sayNO;
5254
262b90c4 5255yes:
77cb431f
DM
5256 if (yes_state) {
5257 /* we have successfully completed a subexpression, but we must now
5258 * pop to the state marked by yes_state and continue from there */
77cb431f 5259 assert(st != yes_state);
5bc10b2c
DM
5260#ifdef DEBUGGING
5261 while (st != yes_state) {
5262 st--;
5263 if (st < SLAB_FIRST(PL_regmatch_slab)) {
5264 PL_regmatch_slab = PL_regmatch_slab->prev;
5265 st = SLAB_LAST(PL_regmatch_slab);
5266 }
e2e6a0f1 5267 DEBUG_STATE_r({
54612592
YO
5268 if (no_final) {
5269 DEBUG_STATE_pp("pop (no final)");
5270 } else {
5271 DEBUG_STATE_pp("pop (yes)");
5272 }
e2e6a0f1 5273 });
5bc10b2c
DM
5274 depth--;
5275 }
5276#else
77cb431f
DM
5277 while (yes_state < SLAB_FIRST(PL_regmatch_slab)
5278 || yes_state > SLAB_LAST(PL_regmatch_slab))
5279 {
5280 /* not in this slab, pop slab */
5281 depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1);
5282 PL_regmatch_slab = PL_regmatch_slab->prev;
5283 st = SLAB_LAST(PL_regmatch_slab);
5284 }
5285 depth -= (st - yes_state);
5bc10b2c 5286#endif
77cb431f
DM
5287 st = yes_state;
5288 yes_state = st->u.yes.prev_yes_state;
5289 PL_regmatch_state = st;
24b23f37 5290
5d458dd8
YO
5291 if (no_final) {
5292 locinput= st->locinput;
5293 nextchr = UCHARAT(locinput);
5294 }
54612592 5295 state_num = st->resume_state + no_final;
24d3c4a9 5296 goto reenter_switch;
77cb431f
DM
5297 }
5298
a3621e74 5299 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
e4584336 5300 PL_colors[4], PL_colors[5]));
02db2b7b 5301
19b95bf0
DM
5302 if (PL_reg_eval_set) {
5303 /* each successfully executed (?{...}) block does the equivalent of
5304 * local $^R = do {...}
5305 * When popping the save stack, all these locals would be undone;
5306 * bypass this by setting the outermost saved $^R to the latest
5307 * value */
5308 if (oreplsv != GvSV(PL_replgv))
5309 sv_setsv(oreplsv, GvSV(PL_replgv));
5310 }
95b24440 5311 result = 1;
aa283a38 5312 goto final_exit;
4633a7c4
LW
5313
5314no:
a3621e74 5315 DEBUG_EXECUTE_r(
7821416a 5316 PerlIO_printf(Perl_debug_log,
786e8c11 5317 "%*s %sfailed...%s\n",
5bc10b2c 5318 REPORT_CODE_OFF+depth*2, "",
786e8c11 5319 PL_colors[4], PL_colors[5])
7821416a 5320 );
aa283a38 5321
262b90c4 5322no_silent:
54612592
YO
5323 if (no_final) {
5324 if (yes_state) {
5325 goto yes;
5326 } else {
5327 goto final_exit;
5328 }
5329 }
aa283a38
DM
5330 if (depth) {
5331 /* there's a previous state to backtrack to */
40a82448
DM
5332 st--;
5333 if (st < SLAB_FIRST(PL_regmatch_slab)) {
5334 PL_regmatch_slab = PL_regmatch_slab->prev;
5335 st = SLAB_LAST(PL_regmatch_slab);
5336 }
5337 PL_regmatch_state = st;
40a82448
DM
5338 locinput= st->locinput;
5339 nextchr = UCHARAT(locinput);
5340
5bc10b2c
DM
5341 DEBUG_STATE_pp("pop");
5342 depth--;
262b90c4
DM
5343 if (yes_state == st)
5344 yes_state = st->u.yes.prev_yes_state;
5bc10b2c 5345
24d3c4a9
DM
5346 state_num = st->resume_state + 1; /* failure = success + 1 */
5347 goto reenter_switch;
95b24440 5348 }
24d3c4a9 5349 result = 0;
aa283a38 5350
262b90c4 5351 final_exit:
bbe252da 5352 if (rex->intflags & PREGf_VERBARG_SEEN) {
5d458dd8
YO
5353 SV *sv_err = get_sv("REGERROR", 1);
5354 SV *sv_mrk = get_sv("REGMARK", 1);
5355 if (result) {
e2e6a0f1 5356 sv_commit = &PL_sv_no;
5d458dd8
YO
5357 if (!sv_yes_mark)
5358 sv_yes_mark = &PL_sv_yes;
5359 } else {
5360 if (!sv_commit)
5361 sv_commit = &PL_sv_yes;
5362 sv_yes_mark = &PL_sv_no;
5363 }
5364 sv_setsv(sv_err, sv_commit);
5365 sv_setsv(sv_mrk, sv_yes_mark);
e2e6a0f1 5366 }
19b95bf0 5367
2f554ef7
DM
5368 /* clean up; in particular, free all slabs above current one */
5369 LEAVE_SCOPE(oldsave);
5d9a96ca 5370
95b24440 5371 return result;
a687059c
LW
5372}
5373
5374/*
5375 - regrepeat - repeatedly match something simple, report how many
5376 */
5377/*
5378 * [This routine now assumes that it will only match on things of length 1.
5379 * That was true before, but now we assume scan - reginput is the count,
a0ed51b3 5380 * rather than incrementing count on every character. [Er, except utf8.]]
a687059c 5381 */
76e3520e 5382STATIC I32
e2e6a0f1 5383S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth)
a687059c 5384{
27da23d5 5385 dVAR;
a0d0e21e 5386 register char *scan;
a0d0e21e 5387 register I32 c;
3280af22 5388 register char *loceol = PL_regeol;
a0ed51b3 5389 register I32 hardcount = 0;
53c4c00c 5390 register bool do_utf8 = PL_reg_match_utf8;
4f55667c
SP
5391#ifndef DEBUGGING
5392 PERL_UNUSED_ARG(depth);
5393#endif
a0d0e21e 5394
7918f24d
NC
5395 PERL_ARGS_ASSERT_REGREPEAT;
5396
3280af22 5397 scan = PL_reginput;
faf11cac
HS
5398 if (max == REG_INFTY)
5399 max = I32_MAX;
5400 else if (max < loceol - scan)
7f596f4c 5401 loceol = scan + max;
a0d0e21e 5402 switch (OP(p)) {
22c35a8c 5403 case REG_ANY:
1aa99e6b 5404 if (do_utf8) {
ffc61ed2 5405 loceol = PL_regeol;
1aa99e6b 5406 while (scan < loceol && hardcount < max && *scan != '\n') {
ffc61ed2
JH
5407 scan += UTF8SKIP(scan);
5408 hardcount++;
5409 }
5410 } else {
5411 while (scan < loceol && *scan != '\n')
5412 scan++;
a0ed51b3
LW
5413 }
5414 break;
ffc61ed2 5415 case SANY:
def8e4ea
JH
5416 if (do_utf8) {
5417 loceol = PL_regeol;
a0804c9e 5418 while (scan < loceol && hardcount < max) {
def8e4ea
JH
5419 scan += UTF8SKIP(scan);
5420 hardcount++;
5421 }
5422 }
5423 else
5424 scan = loceol;
a0ed51b3 5425 break;
f33976b4
DB
5426 case CANY:
5427 scan = loceol;
5428 break;
090f7165
JH
5429 case EXACT: /* length of string is 1 */
5430 c = (U8)*STRING(p);
5431 while (scan < loceol && UCHARAT(scan) == c)
5432 scan++;
bbce6d69 5433 break;
5434 case EXACTF: /* length of string is 1 */
cd439c50 5435 c = (U8)*STRING(p);
bbce6d69 5436 while (scan < loceol &&
22c35a8c 5437 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
bbce6d69 5438 scan++;
5439 break;
5440 case EXACTFL: /* length of string is 1 */
3280af22 5441 PL_reg_flags |= RF_tainted;
cd439c50 5442 c = (U8)*STRING(p);
bbce6d69 5443 while (scan < loceol &&
22c35a8c 5444 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
a0d0e21e
LW
5445 scan++;
5446 break;
5447 case ANYOF:
ffc61ed2
JH
5448 if (do_utf8) {
5449 loceol = PL_regeol;
cfc92286 5450 while (hardcount < max && scan < loceol &&
32fc9b6a 5451 reginclass(prog, p, (U8*)scan, 0, do_utf8)) {
ffc61ed2
JH
5452 scan += UTF8SKIP(scan);
5453 hardcount++;
5454 }
5455 } else {
32fc9b6a 5456 while (scan < loceol && REGINCLASS(prog, p, (U8*)scan))
ffc61ed2
JH
5457 scan++;
5458 }
a0d0e21e
LW
5459 break;
5460 case ALNUM:
1aa99e6b 5461 if (do_utf8) {
ffc61ed2 5462 loceol = PL_regeol;
1a4fad37 5463 LOAD_UTF8_CHARCLASS_ALNUM();
1aa99e6b 5464 while (hardcount < max && scan < loceol &&
3568d838 5465 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
ffc61ed2
JH
5466 scan += UTF8SKIP(scan);
5467 hardcount++;
5468 }
5469 } else {
5470 while (scan < loceol && isALNUM(*scan))
5471 scan++;
a0ed51b3
LW
5472 }
5473 break;
bbce6d69 5474 case ALNUML:
3280af22 5475 PL_reg_flags |= RF_tainted;
1aa99e6b 5476 if (do_utf8) {
ffc61ed2 5477 loceol = PL_regeol;
1aa99e6b
IH
5478 while (hardcount < max && scan < loceol &&
5479 isALNUM_LC_utf8((U8*)scan)) {
ffc61ed2
JH
5480 scan += UTF8SKIP(scan);
5481 hardcount++;
5482 }
5483 } else {
5484 while (scan < loceol && isALNUM_LC(*scan))
5485 scan++;
a0ed51b3
LW
5486 }
5487 break;
a0d0e21e 5488 case NALNUM:
1aa99e6b 5489 if (do_utf8) {
ffc61ed2 5490 loceol = PL_regeol;
1a4fad37 5491 LOAD_UTF8_CHARCLASS_ALNUM();
1aa99e6b 5492 while (hardcount < max && scan < loceol &&
3568d838 5493 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
ffc61ed2
JH
5494 scan += UTF8SKIP(scan);
5495 hardcount++;
5496 }
5497 } else {
5498 while (scan < loceol && !isALNUM(*scan))
5499 scan++;
a0ed51b3
LW
5500 }
5501 break;
bbce6d69 5502 case NALNUML:
3280af22 5503 PL_reg_flags |= RF_tainted;
1aa99e6b 5504 if (do_utf8) {
ffc61ed2 5505 loceol = PL_regeol;
1aa99e6b
IH
5506 while (hardcount < max && scan < loceol &&
5507 !isALNUM_LC_utf8((U8*)scan)) {
ffc61ed2
JH
5508 scan += UTF8SKIP(scan);
5509 hardcount++;
5510 }
5511 } else {
5512 while (scan < loceol && !isALNUM_LC(*scan))
5513 scan++;
a0ed51b3
LW
5514 }
5515 break;
a0d0e21e 5516 case SPACE:
1aa99e6b 5517 if (do_utf8) {
ffc61ed2 5518 loceol = PL_regeol;
1a4fad37 5519 LOAD_UTF8_CHARCLASS_SPACE();
1aa99e6b 5520 while (hardcount < max && scan < loceol &&
3568d838
JH
5521 (*scan == ' ' ||
5522 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
ffc61ed2
JH
5523 scan += UTF8SKIP(scan);
5524 hardcount++;
5525 }
5526 } else {
5527 while (scan < loceol && isSPACE(*scan))
5528 scan++;
a0ed51b3
LW
5529 }
5530 break;
bbce6d69 5531 case SPACEL:
3280af22 5532 PL_reg_flags |= RF_tainted;
1aa99e6b 5533 if (do_utf8) {
ffc61ed2 5534 loceol = PL_regeol;
1aa99e6b 5535 while (hardcount < max && scan < loceol &&
ffc61ed2
JH
5536 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
5537 scan += UTF8SKIP(scan);
5538 hardcount++;
5539 }
5540 } else {
5541 while (scan < loceol && isSPACE_LC(*scan))
5542 scan++;
a0ed51b3
LW
5543 }
5544 break;
a0d0e21e 5545 case NSPACE:
1aa99e6b 5546 if (do_utf8) {
ffc61ed2 5547 loceol = PL_regeol;
1a4fad37 5548 LOAD_UTF8_CHARCLASS_SPACE();
1aa99e6b 5549 while (hardcount < max && scan < loceol &&
3568d838
JH
5550 !(*scan == ' ' ||
5551 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
ffc61ed2
JH
5552 scan += UTF8SKIP(scan);
5553 hardcount++;
5554 }
5555 } else {
5556 while (scan < loceol && !isSPACE(*scan))
5557 scan++;
a0ed51b3 5558 }
0008a298 5559 break;
bbce6d69 5560 case NSPACEL:
3280af22 5561 PL_reg_flags |= RF_tainted;
1aa99e6b 5562 if (do_utf8) {
ffc61ed2 5563 loceol = PL_regeol;
1aa99e6b 5564 while (hardcount < max && scan < loceol &&
ffc61ed2
JH
5565 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
5566 scan += UTF8SKIP(scan);
5567 hardcount++;
5568 }
5569 } else {
5570 while (scan < loceol && !isSPACE_LC(*scan))
5571 scan++;
a0ed51b3
LW
5572 }
5573 break;
a0d0e21e 5574 case DIGIT:
1aa99e6b 5575 if (do_utf8) {
ffc61ed2 5576 loceol = PL_regeol;
1a4fad37 5577 LOAD_UTF8_CHARCLASS_DIGIT();
1aa99e6b 5578 while (hardcount < max && scan < loceol &&
3568d838 5579 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
ffc61ed2
JH
5580 scan += UTF8SKIP(scan);
5581 hardcount++;
5582 }
5583 } else {
5584 while (scan < loceol && isDIGIT(*scan))
5585 scan++;
a0ed51b3
LW
5586 }
5587 break;
a0d0e21e 5588 case NDIGIT:
1aa99e6b 5589 if (do_utf8) {
ffc61ed2 5590 loceol = PL_regeol;
1a4fad37 5591 LOAD_UTF8_CHARCLASS_DIGIT();
1aa99e6b 5592 while (hardcount < max && scan < loceol &&
3568d838 5593 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
ffc61ed2
JH
5594 scan += UTF8SKIP(scan);
5595 hardcount++;
5596 }
5597 } else {
5598 while (scan < loceol && !isDIGIT(*scan))
5599 scan++;
a0ed51b3 5600 }
e1d1eefb
YO
5601 case LNBREAK:
5602 if (do_utf8) {
5603 loceol = PL_regeol;
5604 while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) {
5605 scan += c;
5606 hardcount++;
5607 }
5608 } else {
5609 /*
5610 LNBREAK can match two latin chars, which is ok,
5611 because we have a null terminated string, but we
5612 have to use hardcount in this situation
5613 */
5614 while (scan < loceol && (c=is_LNBREAK_latin1(scan))) {
5615 scan+=c;
5616 hardcount++;
5617 }
5618 }
5619 break;
5620 case HORIZWS:
5621 if (do_utf8) {
5622 loceol = PL_regeol;
5623 while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) {
5624 scan += c;
5625 hardcount++;
5626 }
5627 } else {
5628 while (scan < loceol && is_HORIZWS_latin1(scan))
5629 scan++;
5630 }
a0ed51b3 5631 break;
e1d1eefb
YO
5632 case NHORIZWS:
5633 if (do_utf8) {
5634 loceol = PL_regeol;
5635 while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) {
5636 scan += UTF8SKIP(scan);
5637 hardcount++;
5638 }
5639 } else {
5640 while (scan < loceol && !is_HORIZWS_latin1(scan))
5641 scan++;
5642
5643 }
5644 break;
5645 case VERTWS:
5646 if (do_utf8) {
5647 loceol = PL_regeol;
5648 while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) {
5649 scan += c;
5650 hardcount++;
5651 }
5652 } else {
5653 while (scan < loceol && is_VERTWS_latin1(scan))
5654 scan++;
5655
5656 }
5657 break;
5658 case NVERTWS:
5659 if (do_utf8) {
5660 loceol = PL_regeol;
5661 while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) {
5662 scan += UTF8SKIP(scan);
5663 hardcount++;
5664 }
5665 } else {
5666 while (scan < loceol && !is_VERTWS_latin1(scan))
5667 scan++;
5668
5669 }
5670 break;
5671
a0d0e21e
LW
5672 default: /* Called on something of 0 width. */
5673 break; /* So match right here or not at all. */
5674 }
a687059c 5675
a0ed51b3
LW
5676 if (hardcount)
5677 c = hardcount;
5678 else
5679 c = scan - PL_reginput;
3280af22 5680 PL_reginput = scan;
a687059c 5681
a3621e74 5682 DEBUG_r({
e68ec53f 5683 GET_RE_DEBUG_FLAGS_DECL;
be8e71aa 5684 DEBUG_EXECUTE_r({
e68ec53f
YO
5685 SV * const prop = sv_newmortal();
5686 regprop(prog, prop, p);
5687 PerlIO_printf(Perl_debug_log,
be8e71aa 5688 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
e2e6a0f1 5689 REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max);
a3621e74 5690 });
be8e71aa 5691 });
9041c2e3 5692
a0d0e21e 5693 return(c);
a687059c
LW
5694}
5695
c277df42 5696
be8e71aa 5697#if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION)
c277df42 5698/*
ffc61ed2
JH
5699- regclass_swash - prepare the utf8 swash
5700*/
5701
5702SV *
32fc9b6a 5703Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
ffc61ed2 5704{
97aff369 5705 dVAR;
9e55ce06
JH
5706 SV *sw = NULL;
5707 SV *si = NULL;
5708 SV *alt = NULL;
f8fc2ecf
YO
5709 RXi_GET_DECL(prog,progi);
5710 const struct reg_data * const data = prog ? progi->data : NULL;
ffc61ed2 5711
7918f24d
NC
5712 PERL_ARGS_ASSERT_REGCLASS_SWASH;
5713
4f639d21 5714 if (data && data->count) {
a3b680e6 5715 const U32 n = ARG(node);
ffc61ed2 5716
4f639d21 5717 if (data->what[n] == 's') {
ad64d0ec
NC
5718 SV * const rv = MUTABLE_SV(data->data[n]);
5719 AV * const av = MUTABLE_AV(SvRV(rv));
2d03de9c 5720 SV **const ary = AvARRAY(av);
9e55ce06 5721 SV **a, **b;
9041c2e3 5722
711a919c 5723 /* See the end of regcomp.c:S_regclass() for
9e55ce06
JH
5724 * documentation of these array elements. */
5725
b11f357e 5726 si = *ary;
fe5bfecd
JH
5727 a = SvROK(ary[1]) ? &ary[1] : NULL;
5728 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : NULL;
b11f357e 5729
ffc61ed2
JH
5730 if (a)
5731 sw = *a;
5732 else if (si && doinit) {
5733 sw = swash_init("utf8", "", si, 1, 0);
5734 (void)av_store(av, 1, sw);
5735 }
9e55ce06
JH
5736 if (b)
5737 alt = *b;
ffc61ed2
JH
5738 }
5739 }
5740
9e55ce06
JH
5741 if (listsvp)
5742 *listsvp = si;
5743 if (altsvp)
5744 *altsvp = alt;
ffc61ed2
JH
5745
5746 return sw;
5747}
76234dfb 5748#endif
ffc61ed2
JH
5749
5750/*
ba7b4546 5751 - reginclass - determine if a character falls into a character class
832705d4
JH
5752
5753 The n is the ANYOF regnode, the p is the target string, lenp
5754 is pointer to the maximum length of how far to go in the p
5755 (if the lenp is zero, UTF8SKIP(p) is used),
5756 do_utf8 tells whether the target string is in UTF-8.
5757
bbce6d69 5758 */
5759
76e3520e 5760STATIC bool
32fc9b6a 5761S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8)
bbce6d69 5762{
27da23d5 5763 dVAR;
a3b680e6 5764 const char flags = ANYOF_FLAGS(n);
bbce6d69 5765 bool match = FALSE;
cc07378b 5766 UV c = *p;
ae9ddab8 5767 STRLEN len = 0;
9e55ce06 5768 STRLEN plen;
1aa99e6b 5769
7918f24d
NC
5770 PERL_ARGS_ASSERT_REGINCLASS;
5771
19f67299
TS
5772 if (do_utf8 && !UTF8_IS_INVARIANT(c)) {
5773 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
4ad0818d
DM
5774 (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) | UTF8_CHECK_ONLY);
5775 /* see [perl #37836] for UTF8_ALLOW_ANYUV */
e8a70c6f
SP
5776 if (len == (STRLEN)-1)
5777 Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)");
19f67299 5778 }
bbce6d69 5779
0f0076b4 5780 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
ffc61ed2 5781 if (do_utf8 || (flags & ANYOF_UNICODE)) {
9e55ce06
JH
5782 if (lenp)
5783 *lenp = 0;
ffc61ed2 5784 if (do_utf8 && !ANYOF_RUNTIME(n)) {
ffc61ed2
JH
5785 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
5786 match = TRUE;
bbce6d69 5787 }
3568d838 5788 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
1aa99e6b 5789 match = TRUE;
ffc61ed2 5790 if (!match) {
9e55ce06 5791 AV *av;
32fc9b6a 5792 SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av);
ffc61ed2
JH
5793
5794 if (sw) {
3f0c5693
KW
5795 U8 * utf8_p;
5796 if (do_utf8) {
5797 utf8_p = (U8 *) p;
5798 } else {
5799 STRLEN len = 1;
5800 utf8_p = bytes_to_utf8(p, &len);
5801 }
5802 if (swash_fetch(sw, utf8_p, 1))
ffc61ed2
JH
5803 match = TRUE;
5804 else if (flags & ANYOF_FOLD) {
9e55ce06
JH
5805 if (!match && lenp && av) {
5806 I32 i;
9e55ce06 5807 for (i = 0; i <= av_len(av); i++) {
890ce7af 5808 SV* const sv = *av_fetch(av, i, FALSE);
9e55ce06 5809 STRLEN len;
890ce7af 5810 const char * const s = SvPV_const(sv, len);
3f0c5693 5811 if (len <= plen && memEQ(s, (char*)utf8_p, len)) {
9e55ce06
JH
5812 *lenp = len;
5813 match = TRUE;
5814 break;
5815 }
5816 }
5817 }
5818 if (!match) {
89ebb4a3 5819 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
4a623e43 5820
3f0c5693
KW
5821 STRLEN tmplen;
5822 to_utf8_fold(utf8_p, tmpbuf, &tmplen);
5823 if (swash_fetch(sw, tmpbuf, 1))
9e55ce06
JH
5824 match = TRUE;
5825 }
ffc61ed2 5826 }
b3a04dd3
KW
5827
5828 /* If we allocated a string above, free it */
5829 if (! do_utf8) Safefree(utf8_p);
ffc61ed2 5830 }
bbce6d69 5831 }
9e55ce06 5832 if (match && lenp && *lenp == 0)
0f0076b4 5833 *lenp = UNISKIP(NATIVE_TO_UNI(c));
bbce6d69 5834 }
1aa99e6b 5835 if (!match && c < 256) {
ffc61ed2
JH
5836 if (ANYOF_BITMAP_TEST(n, c))
5837 match = TRUE;
5838 else if (flags & ANYOF_FOLD) {
eb160463 5839 U8 f;
a0ed51b3 5840
ffc61ed2
JH
5841 if (flags & ANYOF_LOCALE) {
5842 PL_reg_flags |= RF_tainted;
5843 f = PL_fold_locale[c];
5844 }
5845 else
5846 f = PL_fold[c];
5847 if (f != c && ANYOF_BITMAP_TEST(n, f))
5848 match = TRUE;
5849 }
5850
5851 if (!match && (flags & ANYOF_CLASS)) {
a0ed51b3 5852 PL_reg_flags |= RF_tainted;
ffc61ed2
JH
5853 if (
5854 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
5855 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
5856 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
5857 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
5858 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
5859 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
5860 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
5861 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
5862 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
5863 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
5864 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
5865 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
5866 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
5867 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
5868 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
5869 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
5870 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
5871 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
5872 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
5873 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
5874 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
5875 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
5876 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
5877 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
5878 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
5879 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
5880 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
5881 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
5882 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
5883 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
5884 ) /* How's that for a conditional? */
5885 {
5886 match = TRUE;
5887 }
a0ed51b3 5888 }
a0ed51b3
LW
5889 }
5890
a0ed51b3
LW
5891 return (flags & ANYOF_INVERT) ? !match : match;
5892}
161b471a 5893
dfe13c55 5894STATIC U8 *
0ce71af7 5895S_reghop3(U8 *s, I32 off, const U8* lim)
9041c2e3 5896{
97aff369 5897 dVAR;
7918f24d
NC
5898
5899 PERL_ARGS_ASSERT_REGHOP3;
5900
a0ed51b3 5901 if (off >= 0) {
1aa99e6b 5902 while (off-- && s < lim) {
ffc61ed2 5903 /* XXX could check well-formedness here */
a0ed51b3 5904 s += UTF8SKIP(s);
ffc61ed2 5905 }
a0ed51b3
LW
5906 }
5907 else {
1de06328
YO
5908 while (off++ && s > lim) {
5909 s--;
5910 if (UTF8_IS_CONTINUED(*s)) {
5911 while (s > lim && UTF8_IS_CONTINUATION(*s))
5912 s--;
a0ed51b3 5913 }
1de06328 5914 /* XXX could check well-formedness here */
a0ed51b3
LW
5915 }
5916 }
5917 return s;
5918}
161b471a 5919
f9f4320a
YO
5920#ifdef XXX_dmq
5921/* there are a bunch of places where we use two reghop3's that should
5922 be replaced with this routine. but since thats not done yet
5923 we ifdef it out - dmq
5924*/
dfe13c55 5925STATIC U8 *
1de06328
YO
5926S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim)
5927{
5928 dVAR;
7918f24d
NC
5929
5930 PERL_ARGS_ASSERT_REGHOP4;
5931
1de06328
YO
5932 if (off >= 0) {
5933 while (off-- && s < rlim) {
5934 /* XXX could check well-formedness here */
5935 s += UTF8SKIP(s);
5936 }
5937 }
5938 else {
5939 while (off++ && s > llim) {
5940 s--;
5941 if (UTF8_IS_CONTINUED(*s)) {
5942 while (s > llim && UTF8_IS_CONTINUATION(*s))
5943 s--;
5944 }
5945 /* XXX could check well-formedness here */
5946 }
5947 }
5948 return s;
5949}
f9f4320a 5950#endif
1de06328
YO
5951
5952STATIC U8 *
0ce71af7 5953S_reghopmaybe3(U8* s, I32 off, const U8* lim)
a0ed51b3 5954{
97aff369 5955 dVAR;
7918f24d
NC
5956
5957 PERL_ARGS_ASSERT_REGHOPMAYBE3;
5958
a0ed51b3 5959 if (off >= 0) {
1aa99e6b 5960 while (off-- && s < lim) {
ffc61ed2 5961 /* XXX could check well-formedness here */
a0ed51b3 5962 s += UTF8SKIP(s);
ffc61ed2 5963 }
a0ed51b3 5964 if (off >= 0)
3dab1dad 5965 return NULL;
a0ed51b3
LW
5966 }
5967 else {
1de06328
YO
5968 while (off++ && s > lim) {
5969 s--;
5970 if (UTF8_IS_CONTINUED(*s)) {
5971 while (s > lim && UTF8_IS_CONTINUATION(*s))
5972 s--;
a0ed51b3 5973 }
1de06328 5974 /* XXX could check well-formedness here */
a0ed51b3
LW
5975 }
5976 if (off <= 0)
3dab1dad 5977 return NULL;
a0ed51b3
LW
5978 }
5979 return s;
5980}
51371543 5981
51371543 5982static void
acfe0abc 5983restore_pos(pTHX_ void *arg)
51371543 5984{
97aff369 5985 dVAR;
097eb12c 5986 regexp * const rex = (regexp *)arg;
51371543
GS
5987 if (PL_reg_eval_set) {
5988 if (PL_reg_oldsaved) {
4f639d21
DM
5989 rex->subbeg = PL_reg_oldsaved;
5990 rex->sublen = PL_reg_oldsavedlen;
f8c7b90f 5991#ifdef PERL_OLD_COPY_ON_WRITE
4f639d21 5992 rex->saved_copy = PL_nrs;
ed252734 5993#endif
07bc277f 5994 RXp_MATCH_COPIED_on(rex);
51371543
GS
5995 }
5996 PL_reg_magic->mg_len = PL_reg_oldpos;
5997 PL_reg_eval_set = 0;
5998 PL_curpm = PL_reg_oldcurpm;
5999 }
6000}
33b8afdf
JH
6001
6002STATIC void
6003S_to_utf8_substr(pTHX_ register regexp *prog)
6004{
a1cac82e 6005 int i = 1;
7918f24d
NC
6006
6007 PERL_ARGS_ASSERT_TO_UTF8_SUBSTR;
6008
a1cac82e
NC
6009 do {
6010 if (prog->substrs->data[i].substr
6011 && !prog->substrs->data[i].utf8_substr) {
6012 SV* const sv = newSVsv(prog->substrs->data[i].substr);
6013 prog->substrs->data[i].utf8_substr = sv;
6014 sv_utf8_upgrade(sv);
610460f9
NC
6015 if (SvVALID(prog->substrs->data[i].substr)) {
6016 const U8 flags = BmFLAGS(prog->substrs->data[i].substr);
6017 if (flags & FBMcf_TAIL) {
6018 /* Trim the trailing \n that fbm_compile added last
6019 time. */
6020 SvCUR_set(sv, SvCUR(sv) - 1);
6021 /* Whilst this makes the SV technically "invalid" (as its
6022 buffer is no longer followed by "\0") when fbm_compile()
6023 adds the "\n" back, a "\0" is restored. */
6024 }
6025 fbm_compile(sv, flags);
6026 }
a1cac82e
NC
6027 if (prog->substrs->data[i].substr == prog->check_substr)
6028 prog->check_utf8 = sv;
6029 }
6030 } while (i--);
33b8afdf
JH
6031}
6032
6033STATIC void
6034S_to_byte_substr(pTHX_ register regexp *prog)
6035{
97aff369 6036 dVAR;
a1cac82e 6037 int i = 1;
7918f24d
NC
6038
6039 PERL_ARGS_ASSERT_TO_BYTE_SUBSTR;
6040
a1cac82e
NC
6041 do {
6042 if (prog->substrs->data[i].utf8_substr
6043 && !prog->substrs->data[i].substr) {
6044 SV* sv = newSVsv(prog->substrs->data[i].utf8_substr);
6045 if (sv_utf8_downgrade(sv, TRUE)) {
610460f9
NC
6046 if (SvVALID(prog->substrs->data[i].utf8_substr)) {
6047 const U8 flags
6048 = BmFLAGS(prog->substrs->data[i].utf8_substr);
6049 if (flags & FBMcf_TAIL) {
6050 /* Trim the trailing \n that fbm_compile added last
6051 time. */
6052 SvCUR_set(sv, SvCUR(sv) - 1);
6053 }
6054 fbm_compile(sv, flags);
6055 }
a1cac82e
NC
6056 } else {
6057 SvREFCNT_dec(sv);
6058 sv = &PL_sv_undef;
6059 }
6060 prog->substrs->data[i].substr = sv;
6061 if (prog->substrs->data[i].utf8_substr == prog->check_utf8)
6062 prog->check_substr = sv;
33b8afdf 6063 }
a1cac82e 6064 } while (i--);
33b8afdf 6065}
66610fdd
RGS
6066
6067/*
6068 * Local variables:
6069 * c-indentation-style: bsd
6070 * c-basic-offset: 4
6071 * indent-tabs-mode: t
6072 * End:
6073 *
37442d52
RGS
6074 * ex: set ts=8 sts=4 sw=4 noet:
6075 */