This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
The correct solution is to reference count the regexp in PL_reg_curpm,
[perl5.git] / regexec.c
CommitLineData
a0d0e21e
LW
1/* regexec.c
2 */
3
4/*
5 * "One Ring to rule them all, One Ring to find them..."
6 */
7
61296642
DM
8/* This file contains functions for executing a regular expression. See
9 * also regcomp.c which funnily enough, contains functions for compiling
166f8a29 10 * a regular expression.
e4a054ea
DM
11 *
12 * This file is also copied at build time to ext/re/re_exec.c, where
13 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
14 * This causes the main functions to be compiled under new names and with
15 * debugging support added, which makes "use re 'debug'" work.
166f8a29
DM
16 */
17
a687059c
LW
18/* NOTE: this is derived from Henry Spencer's regexp code, and should not
19 * confused with the original package (see point 3 below). Thanks, Henry!
20 */
21
22/* Additional note: this code is very heavily munged from Henry's version
23 * in places. In some spots I've traded clarity for efficiency, so don't
24 * blame Henry for some of the lack of readability.
25 */
26
e50aee73
AD
27/* The names of the functions have been changed from regcomp and
28 * regexec to pregcomp and pregexec in order to avoid conflicts
29 * with the POSIX routines of the same names.
30*/
31
b9d5759e 32#ifdef PERL_EXT_RE_BUILD
54df2634 33#include "re_top.h"
9041c2e3 34#endif
56953603 35
a687059c 36/*
e50aee73 37 * pregcomp and pregexec -- regsub and regerror are not used in perl
a687059c
LW
38 *
39 * Copyright (c) 1986 by University of Toronto.
40 * Written by Henry Spencer. Not derived from licensed software.
41 *
42 * Permission is granted to anyone to use this software for any
43 * purpose on any computer system, and to redistribute it freely,
44 * subject to the following restrictions:
45 *
46 * 1. The author is not responsible for the consequences of use of
47 * this software, no matter how awful, even if they arise
48 * from defects in it.
49 *
50 * 2. The origin of this software must not be misrepresented, either
51 * by explicit claim or by omission.
52 *
53 * 3. Altered versions must be plainly marked as such, and must not
54 * be misrepresented as being the original software.
55 *
56 **** Alterations to Henry's code are...
57 ****
4bb101f2 58 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
fdf8c088 59 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007 by Larry Wall and others
a687059c 60 ****
9ef589d8
LW
61 **** You may distribute under the terms of either the GNU General Public
62 **** License or the Artistic License, as specified in the README file.
a687059c
LW
63 *
64 * Beware that some of this code is subtly aware of the way operator
65 * precedence is structured in regular expressions. Serious changes in
66 * regular-expression syntax might require a total rethink.
67 */
68#include "EXTERN.h"
864dbfa3 69#define PERL_IN_REGEXEC_C
a687059c 70#include "perl.h"
0f5d15d6 71
54df2634
NC
72#ifdef PERL_IN_XSUB_RE
73# include "re_comp.h"
74#else
75# include "regcomp.h"
76#endif
a687059c 77
c277df42
IZ
78#define RF_tainted 1 /* tainted information used? */
79#define RF_warned 2 /* warned about big count? */
faec1544 80
ab3bbdeb 81#define RF_utf8 8 /* Pattern contains multibyte chars? */
a0ed51b3 82
eb160463 83#define UTF ((PL_reg_flags & RF_utf8) != 0)
ce862d02
IZ
84
85#define RS_init 1 /* eval environment created */
86#define RS_set 2 /* replsv value is set */
c277df42 87
a687059c
LW
88#ifndef STATIC
89#define STATIC static
90#endif
91
32fc9b6a 92#define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
7d3e948e 93
c277df42
IZ
94/*
95 * Forwards.
96 */
97
33b8afdf 98#define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
53c4c00c 99#define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
a0ed51b3 100
3dab1dad
YO
101#define HOPc(pos,off) \
102 (char *)(PL_reg_match_utf8 \
52657f30 103 ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \
3dab1dad
YO
104 : (U8*)(pos + off))
105#define HOPBACKc(pos, off) \
07be1b83
YO
106 (char*)(PL_reg_match_utf8\
107 ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \
108 : (pos - off >= PL_bostr) \
8e11feef 109 ? (U8*)pos - off \
3dab1dad 110 : NULL)
efb30f32 111
e7409c1b 112#define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off))
1aa99e6b 113#define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
1aa99e6b 114
1a4fad37
AL
115#define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
116 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END
117#define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
118#define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
119#define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
120#define LOAD_UTF8_CHARCLASS_MARK() LOAD_UTF8_CHARCLASS(mark, "\xcd\x86")
51371543 121
3dab1dad
YO
122/* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */
123
5f80c4cf 124/* for use after a quantifier and before an EXACT-like node -- japhy */
3e901dc0
YO
125/* it would be nice to rework regcomp.sym to generate this stuff. sigh */
126#define JUMPABLE(rn) ( \
127 OP(rn) == OPEN || \
128 (OP(rn) == CLOSE && (!cur_eval || cur_eval->u.eval.close_paren != ARG(rn))) || \
129 OP(rn) == EVAL || \
cca55fe3
JP
130 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
131 OP(rn) == PLUS || OP(rn) == MINMOD || \
ee9b8eae 132 OP(rn) == KEEPS || (PL_regkind[OP(rn)] == VERB) || \
3dab1dad 133 (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \
e2d8ce26 134)
ee9b8eae 135#define IS_EXACT(rn) (PL_regkind[OP(rn)] == EXACT)
e2d8ce26 136
ee9b8eae
YO
137#define HAS_TEXT(rn) ( IS_EXACT(rn) || PL_regkind[OP(rn)] == REF )
138
139#if 0
140/* Currently these are only used when PL_regkind[OP(rn)] == EXACT so
141 we don't need this definition. */
142#define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF )
143#define IS_TEXTF(rn) ( OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF )
144#define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL )
145
146#else
147/* ... so we use this as its faster. */
148#define IS_TEXT(rn) ( OP(rn)==EXACT )
149#define IS_TEXTF(rn) ( OP(rn)==EXACTF )
150#define IS_TEXTFL(rn) ( OP(rn)==EXACTFL )
151
152#endif
e2d8ce26 153
a84d97b6
HS
154/*
155 Search for mandatory following text node; for lookahead, the text must
156 follow but for lookbehind (rn->flags != 0) we skip to the next step.
157*/
cca55fe3 158#define FIND_NEXT_IMPT(rn) STMT_START { \
3dab1dad
YO
159 while (JUMPABLE(rn)) { \
160 const OPCODE type = OP(rn); \
161 if (type == SUSPEND || PL_regkind[type] == CURLY) \
e2d8ce26 162 rn = NEXTOPER(NEXTOPER(rn)); \
3dab1dad 163 else if (type == PLUS) \
cca55fe3 164 rn = NEXTOPER(rn); \
3dab1dad 165 else if (type == IFMATCH) \
a84d97b6 166 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
e2d8ce26 167 else rn += NEXT_OFF(rn); \
3dab1dad 168 } \
5f80c4cf 169} STMT_END
74750237 170
c476f425 171
acfe0abc 172static void restore_pos(pTHX_ void *arg);
51371543 173
76e3520e 174STATIC CHECKPOINT
cea2e8a9 175S_regcppush(pTHX_ I32 parenfloor)
a0d0e21e 176{
97aff369 177 dVAR;
a3b680e6 178 const int retval = PL_savestack_ix;
b1ce53c5 179#define REGCP_PAREN_ELEMS 4
a3b680e6 180 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
a0d0e21e 181 int p;
40a82448 182 GET_RE_DEBUG_FLAGS_DECL;
a0d0e21e 183
e49a9654
IH
184 if (paren_elems_to_push < 0)
185 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
186
f0ab9afb 187#define REGCP_OTHER_ELEMS 7
4b3c1a47 188 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
7f69552c 189
3280af22 190 for (p = PL_regsize; p > parenfloor; p--) {
b1ce53c5 191/* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
f0ab9afb
NC
192 SSPUSHINT(PL_regoffs[p].end);
193 SSPUSHINT(PL_regoffs[p].start);
3280af22 194 SSPUSHPTR(PL_reg_start_tmp[p]);
a0d0e21e 195 SSPUSHINT(p);
e7707071 196 DEBUG_BUFFERS_r(PerlIO_printf(Perl_debug_log,
40a82448 197 " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n",
f0ab9afb 198 (UV)p, (IV)PL_regoffs[p].start,
40a82448 199 (IV)(PL_reg_start_tmp[p] - PL_bostr),
f0ab9afb 200 (IV)PL_regoffs[p].end
40a82448 201 ));
a0d0e21e 202 }
b1ce53c5 203/* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
f0ab9afb 204 SSPUSHPTR(PL_regoffs);
3280af22
NIS
205 SSPUSHINT(PL_regsize);
206 SSPUSHINT(*PL_reglastparen);
a01268b5 207 SSPUSHINT(*PL_reglastcloseparen);
3280af22 208 SSPUSHPTR(PL_reginput);
41123dfd
JH
209#define REGCP_FRAME_ELEMS 2
210/* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
211 * are needed for the regexp context stack bookkeeping. */
212 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
b1ce53c5 213 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
41123dfd 214
a0d0e21e
LW
215 return retval;
216}
217
c277df42 218/* These are needed since we do not localize EVAL nodes: */
ab3bbdeb
YO
219#define REGCP_SET(cp) \
220 DEBUG_STATE_r( \
ab3bbdeb 221 PerlIO_printf(Perl_debug_log, \
e4f74956 222 " Setting an EVAL scope, savestack=%"IVdf"\n", \
ab3bbdeb
YO
223 (IV)PL_savestack_ix)); \
224 cp = PL_savestack_ix
c3464db5 225
ab3bbdeb 226#define REGCP_UNWIND(cp) \
e4f74956 227 DEBUG_STATE_r( \
ab3bbdeb 228 if (cp != PL_savestack_ix) \
e4f74956
YO
229 PerlIO_printf(Perl_debug_log, \
230 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
ab3bbdeb
YO
231 (IV)(cp), (IV)PL_savestack_ix)); \
232 regcpblow(cp)
c277df42 233
76e3520e 234STATIC char *
097eb12c 235S_regcppop(pTHX_ const regexp *rex)
a0d0e21e 236{
97aff369 237 dVAR;
3b6647e0 238 U32 i;
a0d0e21e 239 char *input;
b1ce53c5 240
a3621e74
YO
241 GET_RE_DEBUG_FLAGS_DECL;
242
b1ce53c5 243 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
a0d0e21e 244 i = SSPOPINT;
b1ce53c5
JH
245 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
246 i = SSPOPINT; /* Parentheses elements to pop. */
a0d0e21e 247 input = (char *) SSPOPPTR;
a01268b5 248 *PL_reglastcloseparen = SSPOPINT;
3280af22
NIS
249 *PL_reglastparen = SSPOPINT;
250 PL_regsize = SSPOPINT;
f0ab9afb 251 PL_regoffs=(regexp_paren_pair *) SSPOPPTR;
b1ce53c5 252
6bda09f9 253
b1ce53c5 254 /* Now restore the parentheses context. */
41123dfd
JH
255 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
256 i > 0; i -= REGCP_PAREN_ELEMS) {
1df70142 257 I32 tmps;
097eb12c 258 U32 paren = (U32)SSPOPINT;
3280af22 259 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
f0ab9afb 260 PL_regoffs[paren].start = SSPOPINT;
cf93c79d 261 tmps = SSPOPINT;
3280af22 262 if (paren <= *PL_reglastparen)
f0ab9afb 263 PL_regoffs[paren].end = tmps;
e7707071 264 DEBUG_BUFFERS_r(
c3464db5 265 PerlIO_printf(Perl_debug_log,
b900a521 266 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
f0ab9afb 267 (UV)paren, (IV)PL_regoffs[paren].start,
b900a521 268 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
f0ab9afb 269 (IV)PL_regoffs[paren].end,
3280af22 270 (paren > *PL_reglastparen ? "(no)" : ""));
c277df42 271 );
a0d0e21e 272 }
e7707071 273 DEBUG_BUFFERS_r(
bb7a0f54 274 if (*PL_reglastparen + 1 <= rex->nparens) {
c3464db5 275 PerlIO_printf(Perl_debug_log,
faccc32b 276 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
4f639d21 277 (IV)(*PL_reglastparen + 1), (IV)rex->nparens);
c277df42
IZ
278 }
279 );
daf18116 280#if 1
dafc8851
JH
281 /* It would seem that the similar code in regtry()
282 * already takes care of this, and in fact it is in
283 * a better location to since this code can #if 0-ed out
284 * but the code in regtry() is needed or otherwise tests
285 * requiring null fields (pat.t#187 and split.t#{13,14}
daf18116
JH
286 * (as of patchlevel 7877) will fail. Then again,
287 * this code seems to be necessary or otherwise
225593e1
DM
288 * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
289 * --jhi updated by dapm */
3b6647e0 290 for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) {
097eb12c 291 if (i > PL_regsize)
f0ab9afb
NC
292 PL_regoffs[i].start = -1;
293 PL_regoffs[i].end = -1;
a0d0e21e 294 }
dafc8851 295#endif
a0d0e21e
LW
296 return input;
297}
298
02db2b7b 299#define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
a0d0e21e 300
a687059c 301/*
e50aee73 302 * pregexec and friends
a687059c
LW
303 */
304
76234dfb 305#ifndef PERL_IN_XSUB_RE
a687059c 306/*
c277df42 307 - pregexec - match a regexp against a string
a687059c 308 */
c277df42 309I32
49d7dfbc 310Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend,
c3464db5 311 char *strbeg, I32 minend, SV *screamer, U32 nosave)
c277df42
IZ
312/* strend: pointer to null at end of string */
313/* strbeg: real beginning of string */
314/* minend: end of match must be >=minend after stringarg. */
315/* nosave: For optimizations. */
316{
317 return
9041c2e3 318 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
c277df42
IZ
319 nosave ? 0 : REXEC_COPY_STR);
320}
76234dfb 321#endif
22e551b9 322
9041c2e3 323/*
cad2e5aa
JH
324 * Need to implement the following flags for reg_anch:
325 *
326 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
327 * USE_INTUIT_ML
328 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
329 * INTUIT_AUTORITATIVE_ML
330 * INTUIT_ONCE_NOML - Intuit can match in one location only.
331 * INTUIT_ONCE_ML
332 *
333 * Another flag for this function: SECOND_TIME (so that float substrs
334 * with giant delta may be not rechecked).
335 */
336
337/* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
338
3f7c398e 339/* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
cad2e5aa
JH
340 Otherwise, only SvCUR(sv) is used to get strbeg. */
341
342/* XXXX We assume that strpos is strbeg unless sv. */
343
6eb5f6b9
JH
344/* XXXX Some places assume that there is a fixed substring.
345 An update may be needed if optimizer marks as "INTUITable"
346 RExen without fixed substrings. Similarly, it is assumed that
347 lengths of all the strings are no more than minlen, thus they
348 cannot come from lookahead.
40d049e4
YO
349 (Or minlen should take into account lookahead.)
350 NOTE: Some of this comment is not correct. minlen does now take account
351 of lookahead/behind. Further research is required. -- demerphq
352
353*/
6eb5f6b9 354
2c2d71f5
JH
355/* A failure to find a constant substring means that there is no need to make
356 an expensive call to REx engine, thus we celebrate a failure. Similarly,
357 finding a substring too deep into the string means that less calls to
30944b6d
IZ
358 regtry() should be needed.
359
360 REx compiler's optimizer found 4 possible hints:
361 a) Anchored substring;
362 b) Fixed substring;
363 c) Whether we are anchored (beginning-of-line or \G);
364 d) First node (of those at offset 0) which may distingush positions;
6eb5f6b9 365 We use a)b)d) and multiline-part of c), and try to find a position in the
30944b6d
IZ
366 string which does not contradict any of them.
367 */
2c2d71f5 368
6eb5f6b9
JH
369/* Most of decisions we do here should have been done at compile time.
370 The nodes of the REx which we used for the search should have been
371 deleted from the finite automaton. */
372
cad2e5aa 373char *
288b8c02 374Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos,
9f61653a 375 char *strend, const U32 flags, re_scream_pos_data *data)
cad2e5aa 376{
97aff369 377 dVAR;
288b8c02 378 struct regexp *const prog = (struct regexp *)SvANY(rx);
b7953727 379 register I32 start_shift = 0;
cad2e5aa 380 /* Should be nonnegative! */
b7953727 381 register I32 end_shift = 0;
2c2d71f5
JH
382 register char *s;
383 register SV *check;
a1933d95 384 char *strbeg;
cad2e5aa 385 char *t;
1de06328 386 const bool do_utf8 = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */
cad2e5aa 387 I32 ml_anch;
bd61b366
SS
388 register char *other_last = NULL; /* other substr checked before this */
389 char *check_at = NULL; /* check substr found at this pos */
bbe252da 390 const I32 multiline = prog->extflags & RXf_PMf_MULTILINE;
f8fc2ecf 391 RXi_GET_DECL(prog,progi);
30944b6d 392#ifdef DEBUGGING
890ce7af 393 const char * const i_strpos = strpos;
30944b6d 394#endif
a3621e74
YO
395
396 GET_RE_DEBUG_FLAGS_DECL;
397
288b8c02 398 RX_MATCH_UTF8_set(rx,do_utf8);
cad2e5aa 399
3c8556c3 400 if (RX_UTF8(rx)) {
b8d68ded
JH
401 PL_reg_flags |= RF_utf8;
402 }
ab3bbdeb 403 DEBUG_EXECUTE_r(
efd26800 404 debug_start_match(rx, do_utf8, strpos, strend,
1de06328
YO
405 sv ? "Guessing start of match in sv for"
406 : "Guessing start of match in string for");
2a782b5b 407 );
cad2e5aa 408
c344f387
JH
409 /* CHR_DIST() would be more correct here but it makes things slow. */
410 if (prog->minlen > strend - strpos) {
a3621e74 411 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
a72c7584 412 "String too short... [re_intuit_start]\n"));
cad2e5aa 413 goto fail;
2c2d71f5 414 }
1de06328 415
a1933d95 416 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
1aa99e6b 417 PL_regeol = strend;
33b8afdf
JH
418 if (do_utf8) {
419 if (!prog->check_utf8 && prog->check_substr)
420 to_utf8_substr(prog);
421 check = prog->check_utf8;
422 } else {
423 if (!prog->check_substr && prog->check_utf8)
424 to_byte_substr(prog);
425 check = prog->check_substr;
426 }
1de06328 427 if (check == &PL_sv_undef) {
a3621e74 428 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1de06328 429 "Non-utf8 string cannot match utf8 check string\n"));
33b8afdf
JH
430 goto fail;
431 }
bbe252da
YO
432 if (prog->extflags & RXf_ANCH) { /* Match at beg-of-str or after \n */
433 ml_anch = !( (prog->extflags & RXf_ANCH_SINGLE)
434 || ( (prog->extflags & RXf_ANCH_BOL)
7fba1cd6 435 && !multiline ) ); /* Check after \n? */
cad2e5aa 436
7e25d62c 437 if (!ml_anch) {
bbe252da
YO
438 if ( !(prog->extflags & RXf_ANCH_GPOS) /* Checked by the caller */
439 && !(prog->intflags & PREGf_IMPLICIT) /* not a real BOL */
3f7c398e 440 /* SvCUR is not set on references: SvRV and SvPVX_const overlap */
7e25d62c
JH
441 && sv && !SvROK(sv)
442 && (strpos != strbeg)) {
a3621e74 443 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
7e25d62c
JH
444 goto fail;
445 }
446 if (prog->check_offset_min == prog->check_offset_max &&
bbe252da 447 !(prog->extflags & RXf_CANY_SEEN)) {
2c2d71f5 448 /* Substring at constant offset from beg-of-str... */
cad2e5aa
JH
449 I32 slen;
450
1aa99e6b 451 s = HOP3c(strpos, prog->check_offset_min, strend);
1de06328 452
653099ff
GS
453 if (SvTAIL(check)) {
454 slen = SvCUR(check); /* >= 1 */
cad2e5aa 455
9041c2e3 456 if ( strend - s > slen || strend - s < slen - 1
2c2d71f5 457 || (strend - s == slen && strend[-1] != '\n')) {
a3621e74 458 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
2c2d71f5 459 goto fail_finish;
cad2e5aa
JH
460 }
461 /* Now should match s[0..slen-2] */
462 slen--;
3f7c398e 463 if (slen && (*SvPVX_const(check) != *s
cad2e5aa 464 || (slen > 1
3f7c398e 465 && memNE(SvPVX_const(check), s, slen)))) {
2c2d71f5 466 report_neq:
a3621e74 467 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
2c2d71f5
JH
468 goto fail_finish;
469 }
cad2e5aa 470 }
3f7c398e 471 else if (*SvPVX_const(check) != *s
653099ff 472 || ((slen = SvCUR(check)) > 1
3f7c398e 473 && memNE(SvPVX_const(check), s, slen)))
2c2d71f5 474 goto report_neq;
c315bfe8 475 check_at = s;
2c2d71f5 476 goto success_at_start;
7e25d62c 477 }
cad2e5aa 478 }
2c2d71f5 479 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
cad2e5aa 480 s = strpos;
2c2d71f5 481 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
1de06328
YO
482 end_shift = prog->check_end_shift;
483
2c2d71f5 484 if (!ml_anch) {
a3b680e6 485 const I32 end = prog->check_offset_max + CHR_SVLEN(check)
653099ff 486 - (SvTAIL(check) != 0);
a3b680e6 487 const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
2c2d71f5
JH
488
489 if (end_shift < eshift)
490 end_shift = eshift;
491 }
cad2e5aa 492 }
2c2d71f5 493 else { /* Can match at random position */
cad2e5aa
JH
494 ml_anch = 0;
495 s = strpos;
1de06328
YO
496 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
497 end_shift = prog->check_end_shift;
498
499 /* end shift should be non negative here */
cad2e5aa
JH
500 }
501
bcdf7404 502#ifdef QDEBUGGING /* 7/99: reports of failure (with the older version) */
0033605d 503 if (end_shift < 0)
1de06328 504 Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ",
220fc49f 505 (IV)end_shift, RX_PRECOMP(prog));
2c2d71f5
JH
506#endif
507
2c2d71f5
JH
508 restart:
509 /* Find a possible match in the region s..strend by looking for
510 the "check" substring in the region corrected by start/end_shift. */
1de06328
YO
511
512 {
513 I32 srch_start_shift = start_shift;
514 I32 srch_end_shift = end_shift;
515 if (srch_start_shift < 0 && strbeg - s > srch_start_shift) {
516 srch_end_shift -= ((strbeg - s) - srch_start_shift);
517 srch_start_shift = strbeg - s;
518 }
6bda09f9 519 DEBUG_OPTIMISE_MORE_r({
1de06328
YO
520 PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n",
521 (IV)prog->check_offset_min,
522 (IV)srch_start_shift,
523 (IV)srch_end_shift,
524 (IV)prog->check_end_shift);
525 });
526
cad2e5aa 527 if (flags & REXEC_SCREAM) {
cad2e5aa 528 I32 p = -1; /* Internal iterator of scream. */
a3b680e6 529 I32 * const pp = data ? data->scream_pos : &p;
cad2e5aa 530
2c2d71f5
JH
531 if (PL_screamfirst[BmRARE(check)] >= 0
532 || ( BmRARE(check) == '\n'
85c508c3 533 && (BmPREVIOUS(check) == SvCUR(check) - 1)
2c2d71f5 534 && SvTAIL(check) ))
9041c2e3 535 s = screaminstr(sv, check,
1de06328 536 srch_start_shift + (s - strbeg), srch_end_shift, pp, 0);
cad2e5aa 537 else
2c2d71f5 538 goto fail_finish;
4addbd3b 539 /* we may be pointing at the wrong string */
07bc277f 540 if (s && RXp_MATCH_COPIED(prog))
3f7c398e 541 s = strbeg + (s - SvPVX_const(sv));
cad2e5aa
JH
542 if (data)
543 *data->scream_olds = s;
544 }
1de06328
YO
545 else {
546 U8* start_point;
547 U8* end_point;
bbe252da 548 if (prog->extflags & RXf_CANY_SEEN) {
1de06328
YO
549 start_point= (U8*)(s + srch_start_shift);
550 end_point= (U8*)(strend - srch_end_shift);
551 } else {
552 start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend);
553 end_point= HOP3(strend, -srch_end_shift, strbeg);
554 }
6bda09f9 555 DEBUG_OPTIMISE_MORE_r({
56570a2c 556 PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n",
1de06328 557 (int)(end_point - start_point),
fc8cd66c 558 (int)(end_point - start_point) > 20 ? 20 : (int)(end_point - start_point),
1de06328
YO
559 start_point);
560 });
561
562 s = fbm_instr( start_point, end_point,
7fba1cd6 563 check, multiline ? FBMrf_MULTILINE : 0);
1de06328
YO
564 }
565 }
cad2e5aa
JH
566 /* Update the count-of-usability, remove useless subpatterns,
567 unshift s. */
2c2d71f5 568
ab3bbdeb
YO
569 DEBUG_EXECUTE_r({
570 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
571 SvPVX_const(check), RE_SV_DUMPLEN(check), 30);
572 PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s",
2c2d71f5 573 (s ? "Found" : "Did not find"),
ab3bbdeb
YO
574 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)
575 ? "anchored" : "floating"),
576 quoted,
577 RE_SV_TAIL(check),
578 (s ? " at offset " : "...\n") );
579 });
2c2d71f5
JH
580
581 if (!s)
582 goto fail_finish;
2c2d71f5 583 /* Finish the diagnostic message */
a3621e74 584 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
2c2d71f5 585
1de06328
YO
586 /* XXX dmq: first branch is for positive lookbehind...
587 Our check string is offset from the beginning of the pattern.
588 So we need to do any stclass tests offset forward from that
589 point. I think. :-(
590 */
591
592
593
594 check_at=s;
595
596
2c2d71f5
JH
597 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
598 Start with the other substr.
599 XXXX no SCREAM optimization yet - and a very coarse implementation
a0288114 600 XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
2c2d71f5
JH
601 *always* match. Probably should be marked during compile...
602 Probably it is right to do no SCREAM here...
603 */
604
1de06328
YO
605 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8)
606 : (prog->float_substr && prog->anchored_substr))
607 {
30944b6d 608 /* Take into account the "other" substring. */
2c2d71f5
JH
609 /* XXXX May be hopelessly wrong for UTF... */
610 if (!other_last)
6eb5f6b9 611 other_last = strpos;
33b8afdf 612 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
30944b6d
IZ
613 do_other_anchored:
614 {
890ce7af
AL
615 char * const last = HOP3c(s, -start_shift, strbeg);
616 char *last1, *last2;
be8e71aa 617 char * const saved_s = s;
33b8afdf 618 SV* must;
2c2d71f5 619
2c2d71f5
JH
620 t = s - prog->check_offset_max;
621 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
1d86a7f9 622 && (!do_utf8
0ce71af7 623 || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos))
2c2d71f5 624 && t > strpos)))
6f207bd3 625 NOOP;
2c2d71f5
JH
626 else
627 t = strpos;
1aa99e6b 628 t = HOP3c(t, prog->anchored_offset, strend);
6eb5f6b9
JH
629 if (t < other_last) /* These positions already checked */
630 t = other_last;
1aa99e6b 631 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
2c2d71f5
JH
632 if (last < last1)
633 last1 = last;
1de06328
YO
634 /* XXXX It is not documented what units *_offsets are in.
635 We assume bytes, but this is clearly wrong.
636 Meaning this code needs to be carefully reviewed for errors.
637 dmq.
638 */
639
2c2d71f5 640 /* On end-of-str: see comment below. */
33b8afdf
JH
641 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
642 if (must == &PL_sv_undef) {
643 s = (char*)NULL;
1de06328 644 DEBUG_r(must = prog->anchored_utf8); /* for debug */
33b8afdf
JH
645 }
646 else
647 s = fbm_instr(
648 (unsigned char*)t,
649 HOP3(HOP3(last1, prog->anchored_offset, strend)
650 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
651 must,
7fba1cd6 652 multiline ? FBMrf_MULTILINE : 0
33b8afdf 653 );
ab3bbdeb
YO
654 DEBUG_EXECUTE_r({
655 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
656 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
657 PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s",
2c2d71f5 658 (s ? "Found" : "Contradicts"),
ab3bbdeb
YO
659 quoted, RE_SV_TAIL(must));
660 });
661
662
2c2d71f5
JH
663 if (!s) {
664 if (last1 >= last2) {
a3621e74 665 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2c2d71f5
JH
666 ", giving up...\n"));
667 goto fail_finish;
668 }
a3621e74 669 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2c2d71f5 670 ", trying floating at offset %ld...\n",
be8e71aa 671 (long)(HOP3c(saved_s, 1, strend) - i_strpos)));
1aa99e6b
IH
672 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
673 s = HOP3c(last, 1, strend);
2c2d71f5
JH
674 goto restart;
675 }
676 else {
a3621e74 677 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
30944b6d 678 (long)(s - i_strpos)));
1aa99e6b
IH
679 t = HOP3c(s, -prog->anchored_offset, strbeg);
680 other_last = HOP3c(s, 1, strend);
be8e71aa 681 s = saved_s;
2c2d71f5
JH
682 if (t == strpos)
683 goto try_at_start;
2c2d71f5
JH
684 goto try_at_offset;
685 }
30944b6d 686 }
2c2d71f5
JH
687 }
688 else { /* Take into account the floating substring. */
33b8afdf 689 char *last, *last1;
be8e71aa 690 char * const saved_s = s;
33b8afdf
JH
691 SV* must;
692
693 t = HOP3c(s, -start_shift, strbeg);
694 last1 = last =
695 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
696 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
697 last = HOP3c(t, prog->float_max_offset, strend);
698 s = HOP3c(t, prog->float_min_offset, strend);
699 if (s < other_last)
700 s = other_last;
2c2d71f5 701 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
33b8afdf
JH
702 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
703 /* fbm_instr() takes into account exact value of end-of-str
704 if the check is SvTAIL(ed). Since false positives are OK,
705 and end-of-str is not later than strend we are OK. */
706 if (must == &PL_sv_undef) {
707 s = (char*)NULL;
1de06328 708 DEBUG_r(must = prog->float_utf8); /* for debug message */
33b8afdf
JH
709 }
710 else
2c2d71f5 711 s = fbm_instr((unsigned char*)s,
33b8afdf
JH
712 (unsigned char*)last + SvCUR(must)
713 - (SvTAIL(must)!=0),
7fba1cd6 714 must, multiline ? FBMrf_MULTILINE : 0);
ab3bbdeb
YO
715 DEBUG_EXECUTE_r({
716 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
717 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
718 PerlIO_printf(Perl_debug_log, "%s floating substr %s%s",
33b8afdf 719 (s ? "Found" : "Contradicts"),
ab3bbdeb
YO
720 quoted, RE_SV_TAIL(must));
721 });
33b8afdf
JH
722 if (!s) {
723 if (last1 == last) {
a3621e74 724 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
33b8afdf
JH
725 ", giving up...\n"));
726 goto fail_finish;
2c2d71f5 727 }
a3621e74 728 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
33b8afdf 729 ", trying anchored starting at offset %ld...\n",
be8e71aa 730 (long)(saved_s + 1 - i_strpos)));
33b8afdf
JH
731 other_last = last;
732 s = HOP3c(t, 1, strend);
733 goto restart;
734 }
735 else {
a3621e74 736 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
33b8afdf
JH
737 (long)(s - i_strpos)));
738 other_last = s; /* Fix this later. --Hugo */
be8e71aa 739 s = saved_s;
33b8afdf
JH
740 if (t == strpos)
741 goto try_at_start;
742 goto try_at_offset;
743 }
2c2d71f5 744 }
cad2e5aa 745 }
2c2d71f5 746
1de06328 747
9ef43ace 748 t= (char*)HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos);
1de06328 749
6bda09f9 750 DEBUG_OPTIMISE_MORE_r(
1de06328
YO
751 PerlIO_printf(Perl_debug_log,
752 "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n",
753 (IV)prog->check_offset_min,
754 (IV)prog->check_offset_max,
755 (IV)(s-strpos),
756 (IV)(t-strpos),
757 (IV)(t-s),
758 (IV)(strend-strpos)
759 )
760 );
761
2c2d71f5 762 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
1d86a7f9 763 && (!do_utf8
9ef43ace 764 || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (U8*) ((prog->check_offset_max<0) ? strend : strpos)))
1de06328
YO
765 && t > strpos)))
766 {
2c2d71f5
JH
767 /* Fixed substring is found far enough so that the match
768 cannot start at strpos. */
769 try_at_offset:
cad2e5aa 770 if (ml_anch && t[-1] != '\n') {
30944b6d
IZ
771 /* Eventually fbm_*() should handle this, but often
772 anchored_offset is not 0, so this check will not be wasted. */
773 /* XXXX In the code below we prefer to look for "^" even in
774 presence of anchored substrings. And we search even
775 beyond the found float position. These pessimizations
776 are historical artefacts only. */
777 find_anchor:
2c2d71f5 778 while (t < strend - prog->minlen) {
cad2e5aa 779 if (*t == '\n') {
4ee3650e 780 if (t < check_at - prog->check_offset_min) {
33b8afdf 781 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
4ee3650e
GS
782 /* Since we moved from the found position,
783 we definitely contradict the found anchored
30944b6d
IZ
784 substr. Due to the above check we do not
785 contradict "check" substr.
786 Thus we can arrive here only if check substr
787 is float. Redo checking for "other"=="fixed".
788 */
9041c2e3 789 strpos = t + 1;
a3621e74 790 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
e4584336 791 PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
30944b6d
IZ
792 goto do_other_anchored;
793 }
4ee3650e
GS
794 /* We don't contradict the found floating substring. */
795 /* XXXX Why not check for STCLASS? */
cad2e5aa 796 s = t + 1;
a3621e74 797 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
e4584336 798 PL_colors[0], PL_colors[1], (long)(s - i_strpos)));
cad2e5aa
JH
799 goto set_useful;
800 }
4ee3650e
GS
801 /* Position contradicts check-string */
802 /* XXXX probably better to look for check-string
803 than for "\n", so one should lower the limit for t? */
a3621e74 804 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
e4584336 805 PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos)));
0e41cd87 806 other_last = strpos = s = t + 1;
cad2e5aa
JH
807 goto restart;
808 }
809 t++;
810 }
a3621e74 811 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
e4584336 812 PL_colors[0], PL_colors[1]));
2c2d71f5 813 goto fail_finish;
cad2e5aa 814 }
f5952150 815 else {
a3621e74 816 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
e4584336 817 PL_colors[0], PL_colors[1]));
f5952150 818 }
cad2e5aa
JH
819 s = t;
820 set_useful:
33b8afdf 821 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
cad2e5aa
JH
822 }
823 else {
f5952150 824 /* The found string does not prohibit matching at strpos,
2c2d71f5 825 - no optimization of calling REx engine can be performed,
f5952150
GS
826 unless it was an MBOL and we are not after MBOL,
827 or a future STCLASS check will fail this. */
2c2d71f5
JH
828 try_at_start:
829 /* Even in this situation we may use MBOL flag if strpos is offset
830 wrt the start of the string. */
05b4157f 831 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
a1933d95 832 && (strpos != strbeg) && strpos[-1] != '\n'
d506a20d 833 /* May be due to an implicit anchor of m{.*foo} */
bbe252da 834 && !(prog->intflags & PREGf_IMPLICIT))
d506a20d 835 {
cad2e5aa
JH
836 t = strpos;
837 goto find_anchor;
838 }
a3621e74 839 DEBUG_EXECUTE_r( if (ml_anch)
f5952150 840 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
70685ca0 841 (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]);
30944b6d 842 );
2c2d71f5 843 success_at_start:
bbe252da 844 if (!(prog->intflags & PREGf_NAUGHTY) /* XXXX If strpos moved? */
33b8afdf
JH
845 && (do_utf8 ? (
846 prog->check_utf8 /* Could be deleted already */
847 && --BmUSEFUL(prog->check_utf8) < 0
848 && (prog->check_utf8 == prog->float_utf8)
849 ) : (
850 prog->check_substr /* Could be deleted already */
851 && --BmUSEFUL(prog->check_substr) < 0
852 && (prog->check_substr == prog->float_substr)
853 )))
66e933ab 854 {
cad2e5aa 855 /* If flags & SOMETHING - do not do it many times on the same match */
a3621e74 856 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
33b8afdf
JH
857 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
858 if (do_utf8 ? prog->check_substr : prog->check_utf8)
859 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
a0714e2c
SS
860 prog->check_substr = prog->check_utf8 = NULL; /* disable */
861 prog->float_substr = prog->float_utf8 = NULL; /* clear */
862 check = NULL; /* abort */
cad2e5aa 863 s = strpos;
3cf5c195
IZ
864 /* XXXX This is a remnant of the old implementation. It
865 looks wasteful, since now INTUIT can use many
6eb5f6b9 866 other heuristics. */
bbe252da 867 prog->extflags &= ~RXf_USE_INTUIT;
cad2e5aa
JH
868 }
869 else
870 s = strpos;
871 }
872
6eb5f6b9
JH
873 /* Last resort... */
874 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
1de06328
YO
875 /* trie stclasses are too expensive to use here, we are better off to
876 leave it to regmatch itself */
f8fc2ecf 877 if (progi->regstclass && PL_regkind[OP(progi->regstclass)]!=TRIE) {
6eb5f6b9
JH
878 /* minlen == 0 is possible if regstclass is \b or \B,
879 and the fixed substr is ''$.
880 Since minlen is already taken into account, s+1 is before strend;
881 accidentally, minlen >= 1 guaranties no false positives at s + 1
882 even for \b or \B. But (minlen? 1 : 0) below assumes that
883 regstclass does not come from lookahead... */
884 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
885 This leaves EXACTF only, which is dealt with in find_byclass(). */
f8fc2ecf
YO
886 const U8* const str = (U8*)STRING(progi->regstclass);
887 const int cl_l = (PL_regkind[OP(progi->regstclass)] == EXACT
888 ? CHR_DIST(str+STR_LEN(progi->regstclass), str)
66e933ab 889 : 1);
1de06328
YO
890 char * endpos;
891 if (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
892 endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend);
893 else if (prog->float_substr || prog->float_utf8)
894 endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend);
895 else
896 endpos= strend;
897
70685ca0
JH
898 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %"IVdf" s: %"IVdf" endpos: %"IVdf"\n",
899 (IV)start_shift, (IV)(check_at - strbeg), (IV)(s - strbeg), (IV)(endpos - strbeg)));
1de06328 900
6eb5f6b9 901 t = s;
f8fc2ecf 902 s = find_byclass(prog, progi->regstclass, s, endpos, NULL);
6eb5f6b9
JH
903 if (!s) {
904#ifdef DEBUGGING
cbbf8932 905 const char *what = NULL;
6eb5f6b9
JH
906#endif
907 if (endpos == strend) {
a3621e74 908 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
6eb5f6b9
JH
909 "Could not match STCLASS...\n") );
910 goto fail;
911 }
a3621e74 912 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
66e933ab 913 "This position contradicts STCLASS...\n") );
bbe252da 914 if ((prog->extflags & RXf_ANCH) && !ml_anch)
653099ff 915 goto fail;
6eb5f6b9 916 /* Contradict one of substrings */
33b8afdf
JH
917 if (prog->anchored_substr || prog->anchored_utf8) {
918 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
a3621e74 919 DEBUG_EXECUTE_r( what = "anchored" );
6eb5f6b9 920 hop_and_restart:
1aa99e6b 921 s = HOP3c(t, 1, strend);
66e933ab
GS
922 if (s + start_shift + end_shift > strend) {
923 /* XXXX Should be taken into account earlier? */
a3621e74 924 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
66e933ab
GS
925 "Could not match STCLASS...\n") );
926 goto fail;
927 }
5e39e1e5
HS
928 if (!check)
929 goto giveup;
a3621e74 930 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
f5952150 931 "Looking for %s substr starting at offset %ld...\n",
6eb5f6b9
JH
932 what, (long)(s + start_shift - i_strpos)) );
933 goto restart;
934 }
66e933ab 935 /* Have both, check_string is floating */
6eb5f6b9
JH
936 if (t + start_shift >= check_at) /* Contradicts floating=check */
937 goto retry_floating_check;
938 /* Recheck anchored substring, but not floating... */
9041c2e3 939 s = check_at;
5e39e1e5
HS
940 if (!check)
941 goto giveup;
a3621e74 942 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
f5952150 943 "Looking for anchored substr starting at offset %ld...\n",
6eb5f6b9
JH
944 (long)(other_last - i_strpos)) );
945 goto do_other_anchored;
946 }
60e71179
GS
947 /* Another way we could have checked stclass at the
948 current position only: */
949 if (ml_anch) {
950 s = t = t + 1;
5e39e1e5
HS
951 if (!check)
952 goto giveup;
a3621e74 953 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
f5952150 954 "Looking for /%s^%s/m starting at offset %ld...\n",
e4584336 955 PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
60e71179 956 goto try_at_offset;
66e933ab 957 }
33b8afdf 958 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
60e71179 959 goto fail;
6eb5f6b9
JH
960 /* Check is floating subtring. */
961 retry_floating_check:
962 t = check_at - start_shift;
a3621e74 963 DEBUG_EXECUTE_r( what = "floating" );
6eb5f6b9
JH
964 goto hop_and_restart;
965 }
b7953727 966 if (t != s) {
a3621e74 967 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
6eb5f6b9 968 "By STCLASS: moving %ld --> %ld\n",
b7953727
JH
969 (long)(t - i_strpos), (long)(s - i_strpos))
970 );
971 }
972 else {
a3621e74 973 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
b7953727
JH
974 "Does not contradict STCLASS...\n");
975 );
976 }
6eb5f6b9 977 }
5e39e1e5 978 giveup:
a3621e74 979 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
5e39e1e5
HS
980 PL_colors[4], (check ? "Guessed" : "Giving up"),
981 PL_colors[5], (long)(s - i_strpos)) );
cad2e5aa 982 return s;
2c2d71f5
JH
983
984 fail_finish: /* Substring not found */
33b8afdf
JH
985 if (prog->check_substr || prog->check_utf8) /* could be removed already */
986 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
cad2e5aa 987 fail:
a3621e74 988 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
e4584336 989 PL_colors[4], PL_colors[5]));
bd61b366 990 return NULL;
cad2e5aa 991}
9661b544 992
a0a388a1
YO
993#define DECL_TRIE_TYPE(scan) \
994 const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \
995 trie_type = (scan->flags != EXACT) \
996 ? (do_utf8 ? trie_utf8_fold : (UTF ? trie_latin_utf8_fold : trie_plain)) \
997 : (do_utf8 ? trie_utf8 : trie_plain)
3b0527fe 998
55eed653
NC
999#define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \
1000uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \
4cadc6a9
YO
1001 switch (trie_type) { \
1002 case trie_utf8_fold: \
1003 if ( foldlen>0 ) { \
1004 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
1005 foldlen -= len; \
1006 uscan += len; \
1007 len=0; \
1008 } else { \
1009 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
1010 uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \
1011 foldlen -= UNISKIP( uvc ); \
1012 uscan = foldbuf + UNISKIP( uvc ); \
1013 } \
1014 break; \
a0a388a1
YO
1015 case trie_latin_utf8_fold: \
1016 if ( foldlen>0 ) { \
1017 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
1018 foldlen -= len; \
1019 uscan += len; \
1020 len=0; \
1021 } else { \
1022 len = 1; \
1023 uvc = to_uni_fold( *(U8*)uc, foldbuf, &foldlen ); \
1024 foldlen -= UNISKIP( uvc ); \
1025 uscan = foldbuf + UNISKIP( uvc ); \
1026 } \
1027 break; \
4cadc6a9
YO
1028 case trie_utf8: \
1029 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
1030 break; \
1031 case trie_plain: \
1032 uvc = (UV)*uc; \
1033 len = 1; \
1034 } \
1035 \
1036 if (uvc < 256) { \
1037 charid = trie->charmap[ uvc ]; \
1038 } \
1039 else { \
1040 charid = 0; \
55eed653
NC
1041 if (widecharmap) { \
1042 SV** const svpp = hv_fetch(widecharmap, \
4cadc6a9
YO
1043 (char*)&uvc, sizeof(UV), 0); \
1044 if (svpp) \
1045 charid = (U16)SvIV(*svpp); \
1046 } \
1047 } \
1048} STMT_END
1049
a0a388a1
YO
1050#define REXEC_FBC_EXACTISH_CHECK(CoNd) \
1051{ \
1052 char *my_strend= (char *)strend; \
4cadc6a9
YO
1053 if ( (CoNd) \
1054 && (ln == len || \
a0a388a1 1055 !ibcmp_utf8(s, &my_strend, 0, do_utf8, \
4cadc6a9 1056 m, NULL, ln, (bool)UTF)) \
a0a388a1 1057 && (!reginfo || regtry(reginfo, &s)) ) \
4cadc6a9
YO
1058 goto got_it; \
1059 else { \
1060 U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \
1061 uvchr_to_utf8(tmpbuf, c); \
1062 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); \
1063 if ( f != c \
1064 && (f == c1 || f == c2) \
a0a388a1
YO
1065 && (ln == len || \
1066 !ibcmp_utf8(s, &my_strend, 0, do_utf8,\
1067 m, NULL, ln, (bool)UTF)) \
1068 && (!reginfo || regtry(reginfo, &s)) ) \
4cadc6a9
YO
1069 goto got_it; \
1070 } \
a0a388a1
YO
1071} \
1072s += len
4cadc6a9
YO
1073
1074#define REXEC_FBC_EXACTISH_SCAN(CoNd) \
1075STMT_START { \
1076 while (s <= e) { \
1077 if ( (CoNd) \
1078 && (ln == 1 || !(OP(c) == EXACTF \
1079 ? ibcmp(s, m, ln) \
1080 : ibcmp_locale(s, m, ln))) \
24b23f37 1081 && (!reginfo || regtry(reginfo, &s)) ) \
4cadc6a9
YO
1082 goto got_it; \
1083 s++; \
1084 } \
1085} STMT_END
1086
1087#define REXEC_FBC_UTF8_SCAN(CoDe) \
1088STMT_START { \
1089 while (s + (uskip = UTF8SKIP(s)) <= strend) { \
1090 CoDe \
1091 s += uskip; \
1092 } \
1093} STMT_END
1094
1095#define REXEC_FBC_SCAN(CoDe) \
1096STMT_START { \
1097 while (s < strend) { \
1098 CoDe \
1099 s++; \
1100 } \
1101} STMT_END
1102
1103#define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \
1104REXEC_FBC_UTF8_SCAN( \
1105 if (CoNd) { \
24b23f37 1106 if (tmp && (!reginfo || regtry(reginfo, &s))) \
4cadc6a9
YO
1107 goto got_it; \
1108 else \
1109 tmp = doevery; \
1110 } \
1111 else \
1112 tmp = 1; \
1113)
1114
1115#define REXEC_FBC_CLASS_SCAN(CoNd) \
1116REXEC_FBC_SCAN( \
1117 if (CoNd) { \
24b23f37 1118 if (tmp && (!reginfo || regtry(reginfo, &s))) \
4cadc6a9
YO
1119 goto got_it; \
1120 else \
1121 tmp = doevery; \
1122 } \
1123 else \
1124 tmp = 1; \
1125)
1126
1127#define REXEC_FBC_TRYIT \
24b23f37 1128if ((!reginfo || regtry(reginfo, &s))) \
4cadc6a9
YO
1129 goto got_it
1130
e1d1eefb
YO
1131#define REXEC_FBC_CSCAN(CoNdUtF8,CoNd) \
1132 if (do_utf8) { \
1133 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1134 } \
1135 else { \
1136 REXEC_FBC_CLASS_SCAN(CoNd); \
1137 } \
1138 break
1139
4cadc6a9
YO
1140#define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \
1141 if (do_utf8) { \
1142 UtFpReLoAd; \
1143 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1144 } \
1145 else { \
1146 REXEC_FBC_CLASS_SCAN(CoNd); \
1147 } \
1148 break
1149
1150#define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \
1151 PL_reg_flags |= RF_tainted; \
1152 if (do_utf8) { \
1153 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1154 } \
1155 else { \
1156 REXEC_FBC_CLASS_SCAN(CoNd); \
1157 } \
1158 break
1159
786e8c11
YO
1160#define DUMP_EXEC_POS(li,s,doutf8) \
1161 dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8)
1162
1163/* We know what class REx starts with. Try to find this position... */
1164/* if reginfo is NULL, its a dryrun */
1165/* annoyingly all the vars in this routine have different names from their counterparts
1166 in regmatch. /grrr */
1167
3c3eec57 1168STATIC char *
07be1b83 1169S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
24b23f37 1170 const char *strend, regmatch_info *reginfo)
a687059c 1171{
27da23d5 1172 dVAR;
bbe252da 1173 const I32 doevery = (prog->intflags & PREGf_SKIP) == 0;
6eb5f6b9 1174 char *m;
d8093b23 1175 STRLEN ln;
5dab1207 1176 STRLEN lnc;
078c425b 1177 register STRLEN uskip;
d8093b23
G
1178 unsigned int c1;
1179 unsigned int c2;
6eb5f6b9
JH
1180 char *e;
1181 register I32 tmp = 1; /* Scratch variable? */
a3b680e6 1182 register const bool do_utf8 = PL_reg_match_utf8;
f8fc2ecf
YO
1183 RXi_GET_DECL(prog,progi);
1184
6eb5f6b9
JH
1185 /* We know what class it must start with. */
1186 switch (OP(c)) {
6eb5f6b9 1187 case ANYOF:
388cc4de 1188 if (do_utf8) {
4cadc6a9 1189 REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
388cc4de 1190 !UTF8_IS_INVARIANT((U8)s[0]) ?
32fc9b6a 1191 reginclass(prog, c, (U8*)s, 0, do_utf8) :
4cadc6a9 1192 REGINCLASS(prog, c, (U8*)s));
388cc4de
HS
1193 }
1194 else {
1195 while (s < strend) {
1196 STRLEN skip = 1;
1197
32fc9b6a 1198 if (REGINCLASS(prog, c, (U8*)s) ||
388cc4de
HS
1199 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
1200 /* The assignment of 2 is intentional:
1201 * for the folded sharp s, the skip is 2. */
1202 (skip = SHARP_S_SKIP))) {
24b23f37 1203 if (tmp && (!reginfo || regtry(reginfo, &s)))
388cc4de
HS
1204 goto got_it;
1205 else
1206 tmp = doevery;
1207 }
1208 else
1209 tmp = 1;
1210 s += skip;
1211 }
a0d0e21e 1212 }
6eb5f6b9 1213 break;
f33976b4 1214 case CANY:
4cadc6a9 1215 REXEC_FBC_SCAN(
24b23f37 1216 if (tmp && (!reginfo || regtry(reginfo, &s)))
f33976b4
DB
1217 goto got_it;
1218 else
1219 tmp = doevery;
4cadc6a9 1220 );
f33976b4 1221 break;
6eb5f6b9 1222 case EXACTF:
5dab1207
NIS
1223 m = STRING(c);
1224 ln = STR_LEN(c); /* length to match in octets/bytes */
1225 lnc = (I32) ln; /* length to match in characters */
1aa99e6b 1226 if (UTF) {
a2a2844f 1227 STRLEN ulen1, ulen2;
5dab1207 1228 U8 *sm = (U8 *) m;
89ebb4a3
JH
1229 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
1230 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
97dc7d3e
RGS
1231 /* used by commented-out code below */
1232 /*const U32 uniflags = UTF8_ALLOW_DEFAULT;*/
a0a388a1
YO
1233
1234 /* XXX: Since the node will be case folded at compile
1235 time this logic is a little odd, although im not
1236 sure that its actually wrong. --dmq */
1237
1238 c1 = to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1239 c2 = to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1240
1241 /* XXX: This is kinda strange. to_utf8_XYZ returns the
1242 codepoint of the first character in the converted
1243 form, yet originally we did the extra step.
1244 No tests fail by commenting this code out however
1245 so Ive left it out. -- dmq.
1246
89ebb4a3 1247 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
041457d9 1248 0, uniflags);
89ebb4a3 1249 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
041457d9 1250 0, uniflags);
a0a388a1
YO
1251 */
1252
5dab1207
NIS
1253 lnc = 0;
1254 while (sm < ((U8 *) m + ln)) {
1255 lnc++;
1256 sm += UTF8SKIP(sm);
1257 }
1aa99e6b
IH
1258 }
1259 else {
1260 c1 = *(U8*)m;
1261 c2 = PL_fold[c1];
1262 }
6eb5f6b9
JH
1263 goto do_exactf;
1264 case EXACTFL:
5dab1207
NIS
1265 m = STRING(c);
1266 ln = STR_LEN(c);
1267 lnc = (I32) ln;
d8093b23 1268 c1 = *(U8*)m;
6eb5f6b9
JH
1269 c2 = PL_fold_locale[c1];
1270 do_exactf:
db12adc6 1271 e = HOP3c(strend, -((I32)lnc), s);
b3c9acc1 1272
3b0527fe 1273 if (!reginfo && e < s)
6eb5f6b9 1274 e = s; /* Due to minlen logic of intuit() */
1aa99e6b 1275
60a8b682
JH
1276 /* The idea in the EXACTF* cases is to first find the
1277 * first character of the EXACTF* node and then, if
1278 * necessary, case-insensitively compare the full
1279 * text of the node. The c1 and c2 are the first
1280 * characters (though in Unicode it gets a bit
1281 * more complicated because there are more cases
7f16dd3d
JH
1282 * than just upper and lower: one needs to use
1283 * the so-called folding case for case-insensitive
1284 * matching (called "loose matching" in Unicode).
1285 * ibcmp_utf8() will do just that. */
60a8b682 1286
a0a388a1 1287 if (do_utf8 || UTF) {
575cac57 1288 UV c, f;
89ebb4a3 1289 U8 tmpbuf [UTF8_MAXBYTES+1];
a0a388a1
YO
1290 STRLEN len = 1;
1291 STRLEN foldlen;
4ad0818d 1292 const U32 uniflags = UTF8_ALLOW_DEFAULT;
09091399 1293 if (c1 == c2) {
5dab1207
NIS
1294 /* Upper and lower of 1st char are equal -
1295 * probably not a "letter". */
1aa99e6b 1296 while (s <= e) {
a0a388a1
YO
1297 if (do_utf8) {
1298 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
041457d9 1299 uniflags);
a0a388a1
YO
1300 } else {
1301 c = *((U8*)s);
1302 }
4cadc6a9 1303 REXEC_FBC_EXACTISH_CHECK(c == c1);
1aa99e6b 1304 }
09091399
JH
1305 }
1306 else {
1aa99e6b 1307 while (s <= e) {
a0a388a1
YO
1308 if (do_utf8) {
1309 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
041457d9 1310 uniflags);
a0a388a1
YO
1311 } else {
1312 c = *((U8*)s);
1313 }
80aecb99 1314
60a8b682 1315 /* Handle some of the three Greek sigmas cases.
8c01da3c
JH
1316 * Note that not all the possible combinations
1317 * are handled here: some of them are handled
1318 * by the standard folding rules, and some of
1319 * them (the character class or ANYOF cases)
1320 * are handled during compiletime in
1321 * regexec.c:S_regclass(). */
880bd946
JH
1322 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1323 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1324 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
80aecb99 1325
4cadc6a9 1326 REXEC_FBC_EXACTISH_CHECK(c == c1 || c == c2);
1aa99e6b 1327 }
09091399 1328 }
1aa99e6b
IH
1329 }
1330 else {
a0a388a1 1331 /* Neither pattern nor string are UTF8 */
1aa99e6b 1332 if (c1 == c2)
4cadc6a9 1333 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1);
1aa99e6b 1334 else
4cadc6a9 1335 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2);
b3c9acc1
IZ
1336 }
1337 break;
bbce6d69 1338 case BOUNDL:
3280af22 1339 PL_reg_flags |= RF_tainted;
bbce6d69 1340 /* FALL THROUGH */
a0d0e21e 1341 case BOUND:
ffc61ed2 1342 if (do_utf8) {
12d33761 1343 if (s == PL_bostr)
ffc61ed2
JH
1344 tmp = '\n';
1345 else {
6136c704 1346 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
4ad0818d 1347 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
ffc61ed2
JH
1348 }
1349 tmp = ((OP(c) == BOUND ?
9041c2e3 1350 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1a4fad37 1351 LOAD_UTF8_CHARCLASS_ALNUM();
4cadc6a9 1352 REXEC_FBC_UTF8_SCAN(
ffc61ed2 1353 if (tmp == !(OP(c) == BOUND ?
bb7a0f54 1354 (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
ffc61ed2
JH
1355 isALNUM_LC_utf8((U8*)s)))
1356 {
1357 tmp = !tmp;
4cadc6a9 1358 REXEC_FBC_TRYIT;
a687059c 1359 }
4cadc6a9 1360 );
a0d0e21e 1361 }
667bb95a 1362 else {
12d33761 1363 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
ffc61ed2 1364 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
4cadc6a9 1365 REXEC_FBC_SCAN(
ffc61ed2
JH
1366 if (tmp ==
1367 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1368 tmp = !tmp;
4cadc6a9 1369 REXEC_FBC_TRYIT;
a0ed51b3 1370 }
4cadc6a9 1371 );
a0ed51b3 1372 }
24b23f37 1373 if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, &s)))
a0ed51b3
LW
1374 goto got_it;
1375 break;
bbce6d69 1376 case NBOUNDL:
3280af22 1377 PL_reg_flags |= RF_tainted;
bbce6d69 1378 /* FALL THROUGH */
a0d0e21e 1379 case NBOUND:
ffc61ed2 1380 if (do_utf8) {
12d33761 1381 if (s == PL_bostr)
ffc61ed2
JH
1382 tmp = '\n';
1383 else {
6136c704 1384 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
4ad0818d 1385 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
ffc61ed2
JH
1386 }
1387 tmp = ((OP(c) == NBOUND ?
9041c2e3 1388 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1a4fad37 1389 LOAD_UTF8_CHARCLASS_ALNUM();
4cadc6a9 1390 REXEC_FBC_UTF8_SCAN(
ffc61ed2 1391 if (tmp == !(OP(c) == NBOUND ?
bb7a0f54 1392 (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
ffc61ed2
JH
1393 isALNUM_LC_utf8((U8*)s)))
1394 tmp = !tmp;
4cadc6a9
YO
1395 else REXEC_FBC_TRYIT;
1396 );
a0d0e21e 1397 }
667bb95a 1398 else {
12d33761 1399 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
ffc61ed2
JH
1400 tmp = ((OP(c) == NBOUND ?
1401 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
4cadc6a9 1402 REXEC_FBC_SCAN(
ffc61ed2
JH
1403 if (tmp ==
1404 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1405 tmp = !tmp;
4cadc6a9
YO
1406 else REXEC_FBC_TRYIT;
1407 );
a0ed51b3 1408 }
24b23f37 1409 if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, &s)))
a0ed51b3
LW
1410 goto got_it;
1411 break;
a0d0e21e 1412 case ALNUM:
4cadc6a9
YO
1413 REXEC_FBC_CSCAN_PRELOAD(
1414 LOAD_UTF8_CHARCLASS_ALNUM(),
1415 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8),
1416 isALNUM(*s)
1417 );
bbce6d69 1418 case ALNUML:
4cadc6a9
YO
1419 REXEC_FBC_CSCAN_TAINT(
1420 isALNUM_LC_utf8((U8*)s),
1421 isALNUM_LC(*s)
1422 );
a0d0e21e 1423 case NALNUM:
4cadc6a9
YO
1424 REXEC_FBC_CSCAN_PRELOAD(
1425 LOAD_UTF8_CHARCLASS_ALNUM(),
1426 !swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8),
1427 !isALNUM(*s)
1428 );
bbce6d69 1429 case NALNUML:
4cadc6a9
YO
1430 REXEC_FBC_CSCAN_TAINT(
1431 !isALNUM_LC_utf8((U8*)s),
1432 !isALNUM_LC(*s)
1433 );
a0d0e21e 1434 case SPACE:
4cadc6a9
YO
1435 REXEC_FBC_CSCAN_PRELOAD(
1436 LOAD_UTF8_CHARCLASS_SPACE(),
1437 *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8),
1438 isSPACE(*s)
1439 );
bbce6d69 1440 case SPACEL:
4cadc6a9
YO
1441 REXEC_FBC_CSCAN_TAINT(
1442 *s == ' ' || isSPACE_LC_utf8((U8*)s),
1443 isSPACE_LC(*s)
1444 );
a0d0e21e 1445 case NSPACE:
4cadc6a9
YO
1446 REXEC_FBC_CSCAN_PRELOAD(
1447 LOAD_UTF8_CHARCLASS_SPACE(),
1448 !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)),
1449 !isSPACE(*s)
1450 );
bbce6d69 1451 case NSPACEL:
4cadc6a9
YO
1452 REXEC_FBC_CSCAN_TAINT(
1453 !(*s == ' ' || isSPACE_LC_utf8((U8*)s)),
1454 !isSPACE_LC(*s)
1455 );
a0d0e21e 1456 case DIGIT:
4cadc6a9
YO
1457 REXEC_FBC_CSCAN_PRELOAD(
1458 LOAD_UTF8_CHARCLASS_DIGIT(),
1459 swash_fetch(PL_utf8_digit,(U8*)s, do_utf8),
1460 isDIGIT(*s)
1461 );
b8c5462f 1462 case DIGITL:
4cadc6a9
YO
1463 REXEC_FBC_CSCAN_TAINT(
1464 isDIGIT_LC_utf8((U8*)s),
1465 isDIGIT_LC(*s)
1466 );
a0d0e21e 1467 case NDIGIT:
4cadc6a9
YO
1468 REXEC_FBC_CSCAN_PRELOAD(
1469 LOAD_UTF8_CHARCLASS_DIGIT(),
1470 !swash_fetch(PL_utf8_digit,(U8*)s, do_utf8),
1471 !isDIGIT(*s)
1472 );
b8c5462f 1473 case NDIGITL:
4cadc6a9
YO
1474 REXEC_FBC_CSCAN_TAINT(
1475 !isDIGIT_LC_utf8((U8*)s),
1476 !isDIGIT_LC(*s)
1477 );
e1d1eefb
YO
1478 case LNBREAK:
1479 REXEC_FBC_CSCAN(
1480 is_LNBREAK_utf8(s),
1481 is_LNBREAK_latin1(s)
1482 );
1483 case VERTWS:
1484 REXEC_FBC_CSCAN(
1485 is_VERTWS_utf8(s),
1486 is_VERTWS_latin1(s)
1487 );
1488 case NVERTWS:
1489 REXEC_FBC_CSCAN(
1490 !is_VERTWS_utf8(s),
1491 !is_VERTWS_latin1(s)
1492 );
1493 case HORIZWS:
1494 REXEC_FBC_CSCAN(
1495 is_HORIZWS_utf8(s),
1496 is_HORIZWS_latin1(s)
1497 );
1498 case NHORIZWS:
1499 REXEC_FBC_CSCAN(
1500 !is_HORIZWS_utf8(s),
1501 !is_HORIZWS_latin1(s)
1502 );
1de06328
YO
1503 case AHOCORASICKC:
1504 case AHOCORASICK:
07be1b83 1505 {
a0a388a1 1506 DECL_TRIE_TYPE(c);
07be1b83
YO
1507 /* what trie are we using right now */
1508 reg_ac_data *aho
f8fc2ecf 1509 = (reg_ac_data*)progi->data->data[ ARG( c ) ];
3251b653
NC
1510 reg_trie_data *trie
1511 = (reg_trie_data*)progi->data->data[ aho->trie ];
55eed653 1512 HV *widecharmap = (HV*) progi->data->data[ aho->trie + 1 ];
07be1b83
YO
1513
1514 const char *last_start = strend - trie->minlen;
6148ee25 1515#ifdef DEBUGGING
07be1b83 1516 const char *real_start = s;
6148ee25 1517#endif
07be1b83 1518 STRLEN maxlen = trie->maxlen;
be8e71aa
YO
1519 SV *sv_points;
1520 U8 **points; /* map of where we were in the input string
786e8c11 1521 when reading a given char. For ASCII this
be8e71aa 1522 is unnecessary overhead as the relationship
38a44b82
NC
1523 is always 1:1, but for Unicode, especially
1524 case folded Unicode this is not true. */
f9e705e8 1525 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
786e8c11
YO
1526 U8 *bitmap=NULL;
1527
07be1b83
YO
1528
1529 GET_RE_DEBUG_FLAGS_DECL;
1530
be8e71aa
YO
1531 /* We can't just allocate points here. We need to wrap it in
1532 * an SV so it gets freed properly if there is a croak while
1533 * running the match */
1534 ENTER;
1535 SAVETMPS;
1536 sv_points=newSV(maxlen * sizeof(U8 *));
1537 SvCUR_set(sv_points,
1538 maxlen * sizeof(U8 *));
1539 SvPOK_on(sv_points);
1540 sv_2mortal(sv_points);
1541 points=(U8**)SvPV_nolen(sv_points );
1de06328
YO
1542 if ( trie_type != trie_utf8_fold
1543 && (trie->bitmap || OP(c)==AHOCORASICKC) )
1544 {
786e8c11
YO
1545 if (trie->bitmap)
1546 bitmap=(U8*)trie->bitmap;
1547 else
1548 bitmap=(U8*)ANYOF_BITMAP(c);
07be1b83 1549 }
786e8c11
YO
1550 /* this is the Aho-Corasick algorithm modified a touch
1551 to include special handling for long "unknown char"
1552 sequences. The basic idea being that we use AC as long
1553 as we are dealing with a possible matching char, when
1554 we encounter an unknown char (and we have not encountered
1555 an accepting state) we scan forward until we find a legal
1556 starting char.
1557 AC matching is basically that of trie matching, except
1558 that when we encounter a failing transition, we fall back
1559 to the current states "fail state", and try the current char
1560 again, a process we repeat until we reach the root state,
1561 state 1, or a legal transition. If we fail on the root state
1562 then we can either terminate if we have reached an accepting
1563 state previously, or restart the entire process from the beginning
1564 if we have not.
1565
1566 */
07be1b83
YO
1567 while (s <= last_start) {
1568 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1569 U8 *uc = (U8*)s;
1570 U16 charid = 0;
1571 U32 base = 1;
1572 U32 state = 1;
1573 UV uvc = 0;
1574 STRLEN len = 0;
1575 STRLEN foldlen = 0;
1576 U8 *uscan = (U8*)NULL;
1577 U8 *leftmost = NULL;
786e8c11
YO
1578#ifdef DEBUGGING
1579 U32 accepted_word= 0;
1580#endif
07be1b83
YO
1581 U32 pointpos = 0;
1582
1583 while ( state && uc <= (U8*)strend ) {
1584 int failed=0;
786e8c11
YO
1585 U32 word = aho->states[ state ].wordnum;
1586
1de06328
YO
1587 if( state==1 ) {
1588 if ( bitmap ) {
1589 DEBUG_TRIE_EXECUTE_r(
1590 if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1591 dump_exec_pos( (char *)uc, c, strend, real_start,
1592 (char *)uc, do_utf8 );
1593 PerlIO_printf( Perl_debug_log,
1594 " Scanning for legal start char...\n");
1595 }
1596 );
1597 while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1598 uc++;
786e8c11 1599 }
1de06328 1600 s= (char *)uc;
786e8c11 1601 }
786e8c11
YO
1602 if (uc >(U8*)last_start) break;
1603 }
1604
1605 if ( word ) {
1606 U8 *lpos= points[ (pointpos - trie->wordlen[word-1] ) % maxlen ];
1607 if (!leftmost || lpos < leftmost) {
1608 DEBUG_r(accepted_word=word);
07be1b83 1609 leftmost= lpos;
786e8c11 1610 }
07be1b83 1611 if (base==0) break;
786e8c11 1612
07be1b83
YO
1613 }
1614 points[pointpos++ % maxlen]= uc;
55eed653
NC
1615 REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc,
1616 uscan, len, uvc, charid, foldlen,
1617 foldbuf, uniflags);
786e8c11
YO
1618 DEBUG_TRIE_EXECUTE_r({
1619 dump_exec_pos( (char *)uc, c, strend, real_start,
1620 s, do_utf8 );
07be1b83 1621 PerlIO_printf(Perl_debug_log,
786e8c11
YO
1622 " Charid:%3u CP:%4"UVxf" ",
1623 charid, uvc);
1624 });
07be1b83
YO
1625
1626 do {
6148ee25 1627#ifdef DEBUGGING
786e8c11 1628 word = aho->states[ state ].wordnum;
6148ee25 1629#endif
07be1b83
YO
1630 base = aho->states[ state ].trans.base;
1631
786e8c11
YO
1632 DEBUG_TRIE_EXECUTE_r({
1633 if (failed)
1634 dump_exec_pos( (char *)uc, c, strend, real_start,
1635 s, do_utf8 );
07be1b83 1636 PerlIO_printf( Perl_debug_log,
786e8c11
YO
1637 "%sState: %4"UVxf", word=%"UVxf,
1638 failed ? " Fail transition to " : "",
1639 (UV)state, (UV)word);
1640 });
07be1b83
YO
1641 if ( base ) {
1642 U32 tmp;
1643 if (charid &&
1644 (base + charid > trie->uniquecharcount )
1645 && (base + charid - 1 - trie->uniquecharcount
1646 < trie->lasttrans)
1647 && trie->trans[base + charid - 1 -
1648 trie->uniquecharcount].check == state
1649 && (tmp=trie->trans[base + charid - 1 -
1650 trie->uniquecharcount ].next))
1651 {
786e8c11
YO
1652 DEBUG_TRIE_EXECUTE_r(
1653 PerlIO_printf( Perl_debug_log," - legal\n"));
07be1b83
YO
1654 state = tmp;
1655 break;
1656 }
1657 else {
786e8c11
YO
1658 DEBUG_TRIE_EXECUTE_r(
1659 PerlIO_printf( Perl_debug_log," - fail\n"));
1660 failed = 1;
1661 state = aho->fail[state];
07be1b83
YO
1662 }
1663 }
1664 else {
1665 /* we must be accepting here */
786e8c11
YO
1666 DEBUG_TRIE_EXECUTE_r(
1667 PerlIO_printf( Perl_debug_log," - accepting\n"));
1668 failed = 1;
07be1b83
YO
1669 break;
1670 }
1671 } while(state);
786e8c11 1672 uc += len;
07be1b83
YO
1673 if (failed) {
1674 if (leftmost)
1675 break;
786e8c11 1676 if (!state) state = 1;
07be1b83
YO
1677 }
1678 }
1679 if ( aho->states[ state ].wordnum ) {
1680 U8 *lpos = points[ (pointpos - trie->wordlen[aho->states[ state ].wordnum-1]) % maxlen ];
786e8c11
YO
1681 if (!leftmost || lpos < leftmost) {
1682 DEBUG_r(accepted_word=aho->states[ state ].wordnum);
07be1b83 1683 leftmost = lpos;
786e8c11 1684 }
07be1b83 1685 }
07be1b83
YO
1686 if (leftmost) {
1687 s = (char*)leftmost;
786e8c11
YO
1688 DEBUG_TRIE_EXECUTE_r({
1689 PerlIO_printf(
70685ca0
JH
1690 Perl_debug_log,"Matches word #%"UVxf" at position %"IVdf". Trying full pattern...\n",
1691 (UV)accepted_word, (IV)(s - real_start)
786e8c11
YO
1692 );
1693 });
24b23f37 1694 if (!reginfo || regtry(reginfo, &s)) {
be8e71aa
YO
1695 FREETMPS;
1696 LEAVE;
07be1b83 1697 goto got_it;
be8e71aa 1698 }
07be1b83 1699 s = HOPc(s,1);
786e8c11
YO
1700 DEBUG_TRIE_EXECUTE_r({
1701 PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n");
1702 });
07be1b83 1703 } else {
786e8c11
YO
1704 DEBUG_TRIE_EXECUTE_r(
1705 PerlIO_printf( Perl_debug_log,"No match.\n"));
07be1b83
YO
1706 break;
1707 }
1708 }
be8e71aa
YO
1709 FREETMPS;
1710 LEAVE;
07be1b83
YO
1711 }
1712 break;
b3c9acc1 1713 default:
3c3eec57
GS
1714 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1715 break;
d6a28714 1716 }
6eb5f6b9
JH
1717 return 0;
1718 got_it:
1719 return s;
1720}
1721
e6aad7ab 1722static void
fae667d5 1723S_swap_match_buff (pTHX_ regexp *prog) {
f0ab9afb 1724 regexp_paren_pair *t;
fae667d5 1725
28d8d7f4 1726 if (!prog->swap) {
fae667d5
YO
1727 /* We have to be careful. If the previous successful match
1728 was from this regex we don't want a subsequent paritally
1729 successful match to clobber the old results.
1730 So when we detect this possibility we add a swap buffer
1731 to the re, and switch the buffer each match. If we fail
1732 we switch it back, otherwise we leave it swapped.
1733 */
f0ab9afb 1734 Newxz(prog->swap, (prog->nparens + 1), regexp_paren_pair);
fae667d5 1735 }
f0ab9afb
NC
1736 t = prog->swap;
1737 prog->swap = prog->offs;
1738 prog->offs = t;
fae667d5
YO
1739}
1740
1741
6eb5f6b9
JH
1742/*
1743 - regexec_flags - match a regexp against a string
1744 */
1745I32
288b8c02 1746Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *strend,
6eb5f6b9
JH
1747 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1748/* strend: pointer to null at end of string */
1749/* strbeg: real beginning of string */
1750/* minend: end of match must be >=minend after stringarg. */
58e23c8d
YO
1751/* data: May be used for some additional optimizations.
1752 Currently its only used, with a U32 cast, for transmitting
1753 the ganch offset when doing a /g match. This will change */
6eb5f6b9
JH
1754/* nosave: For optimizations. */
1755{
97aff369 1756 dVAR;
288b8c02 1757 struct regexp *const prog = (struct regexp *)SvANY(rx);
24b23f37 1758 /*register*/ char *s;
6eb5f6b9 1759 register regnode *c;
24b23f37 1760 /*register*/ char *startpos = stringarg;
6eb5f6b9
JH
1761 I32 minlen; /* must match at least this many chars */
1762 I32 dontbother = 0; /* how many characters not to try at end */
6eb5f6b9
JH
1763 I32 end_shift = 0; /* Same for the end. */ /* CC */
1764 I32 scream_pos = -1; /* Internal iterator of scream. */
ccac19ea 1765 char *scream_olds = NULL;
f9f4320a 1766 const bool do_utf8 = (bool)DO_UTF8(sv);
2757e526 1767 I32 multiline;
f8fc2ecf 1768 RXi_GET_DECL(prog,progi);
3b0527fe 1769 regmatch_info reginfo; /* create some info to pass to regtry etc */
fae667d5 1770 bool swap_on_fail = 0;
a3621e74
YO
1771
1772 GET_RE_DEBUG_FLAGS_DECL;
1773
9d4ba2ae 1774 PERL_UNUSED_ARG(data);
6eb5f6b9
JH
1775
1776 /* Be paranoid... */
1777 if (prog == NULL || startpos == NULL) {
1778 Perl_croak(aTHX_ "NULL regexp parameter");
1779 return 0;
1780 }
1781
bbe252da 1782 multiline = prog->extflags & RXf_PMf_MULTILINE;
288b8c02 1783 reginfo.prog = rx; /* Yes, sorry that this is confusing. */
2757e526 1784
288b8c02 1785 RX_MATCH_UTF8_set(rx, do_utf8);
1de06328 1786 DEBUG_EXECUTE_r(
efd26800 1787 debug_start_match(rx, do_utf8, startpos, strend,
1de06328
YO
1788 "Matching");
1789 );
bac06658 1790
6eb5f6b9 1791 minlen = prog->minlen;
1de06328
YO
1792
1793 if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) {
a3621e74 1794 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
a72c7584
JH
1795 "String too short [regexec_flags]...\n"));
1796 goto phooey;
1aa99e6b 1797 }
6eb5f6b9 1798
1de06328 1799
6eb5f6b9 1800 /* Check validity of program. */
f8fc2ecf 1801 if (UCHARAT(progi->program) != REG_MAGIC) {
6eb5f6b9
JH
1802 Perl_croak(aTHX_ "corrupted regexp program");
1803 }
1804
1805 PL_reg_flags = 0;
1806 PL_reg_eval_set = 0;
1807 PL_reg_maxiter = 0;
1808
3c8556c3 1809 if (RX_UTF8(rx))
6eb5f6b9
JH
1810 PL_reg_flags |= RF_utf8;
1811
1812 /* Mark beginning of line for ^ and lookbehind. */
3b0527fe 1813 reginfo.bol = startpos; /* XXX not used ??? */
6eb5f6b9 1814 PL_bostr = strbeg;
3b0527fe 1815 reginfo.sv = sv;
6eb5f6b9
JH
1816
1817 /* Mark end of line for $ (and such) */
1818 PL_regeol = strend;
1819
1820 /* see how far we have to get to not match where we matched before */
3b0527fe 1821 reginfo.till = startpos+minend;
6eb5f6b9 1822
6eb5f6b9
JH
1823 /* If there is a "must appear" string, look for it. */
1824 s = startpos;
1825
bbe252da 1826 if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */
6eb5f6b9
JH
1827 MAGIC *mg;
1828
1829 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
58e23c8d 1830 reginfo.ganch = startpos + prog->gofs;
6eb5f6b9
JH
1831 else if (sv && SvTYPE(sv) >= SVt_PVMG
1832 && SvMAGIC(sv)
14befaf4
DM
1833 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1834 && mg->mg_len >= 0) {
3b0527fe 1835 reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */
bbe252da 1836 if (prog->extflags & RXf_ANCH_GPOS) {
3b0527fe 1837 if (s > reginfo.ganch)
6eb5f6b9 1838 goto phooey;
58e23c8d 1839 s = reginfo.ganch - prog->gofs;
6eb5f6b9
JH
1840 }
1841 }
58e23c8d 1842 else if (data) {
70685ca0 1843 reginfo.ganch = strbeg + PTR2UV(data);
58e23c8d 1844 } else /* pos() not defined */
3b0527fe 1845 reginfo.ganch = strbeg;
6eb5f6b9 1846 }
288b8c02 1847 if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) {
fae667d5
YO
1848 swap_on_fail = 1;
1849 swap_match_buff(prog); /* do we need a save destructor here for
1850 eval dies? */
c74340f9 1851 }
a0714e2c 1852 if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
6eb5f6b9
JH
1853 re_scream_pos_data d;
1854
1855 d.scream_olds = &scream_olds;
1856 d.scream_pos = &scream_pos;
288b8c02 1857 s = re_intuit_start(rx, sv, s, strend, flags, &d);
3fa9c3d7 1858 if (!s) {
a3621e74 1859 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
6eb5f6b9 1860 goto phooey; /* not present */
3fa9c3d7 1861 }
6eb5f6b9
JH
1862 }
1863
1de06328 1864
6eb5f6b9
JH
1865
1866 /* Simplest case: anchored match need be tried only once. */
1867 /* [unless only anchor is BOL and multiline is set] */
bbe252da 1868 if (prog->extflags & (RXf_ANCH & ~RXf_ANCH_GPOS)) {
24b23f37 1869 if (s == startpos && regtry(&reginfo, &startpos))
6eb5f6b9 1870 goto got_it;
bbe252da
YO
1871 else if (multiline || (prog->intflags & PREGf_IMPLICIT)
1872 || (prog->extflags & RXf_ANCH_MBOL)) /* XXXX SBOL? */
6eb5f6b9
JH
1873 {
1874 char *end;
1875
1876 if (minlen)
1877 dontbother = minlen - 1;
1aa99e6b 1878 end = HOP3c(strend, -dontbother, strbeg) - 1;
6eb5f6b9 1879 /* for multiline we only have to try after newlines */
33b8afdf 1880 if (prog->check_substr || prog->check_utf8) {
6eb5f6b9
JH
1881 if (s == startpos)
1882 goto after_try;
1883 while (1) {
24b23f37 1884 if (regtry(&reginfo, &s))
6eb5f6b9
JH
1885 goto got_it;
1886 after_try:
5339e136 1887 if (s > end)
6eb5f6b9 1888 goto phooey;
bbe252da 1889 if (prog->extflags & RXf_USE_INTUIT) {
288b8c02 1890 s = re_intuit_start(rx, sv, s + 1, strend, flags, NULL);
6eb5f6b9
JH
1891 if (!s)
1892 goto phooey;
1893 }
1894 else
1895 s++;
1896 }
1897 } else {
1898 if (s > startpos)
1899 s--;
1900 while (s < end) {
1901 if (*s++ == '\n') { /* don't need PL_utf8skip here */
24b23f37 1902 if (regtry(&reginfo, &s))
6eb5f6b9
JH
1903 goto got_it;
1904 }
1905 }
1906 }
1907 }
1908 goto phooey;
bbe252da 1909 } else if (RXf_GPOS_CHECK == (prog->extflags & RXf_GPOS_CHECK))
f9f4320a
YO
1910 {
1911 /* the warning about reginfo.ganch being used without intialization
bbe252da 1912 is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN
f9f4320a 1913 and we only enter this block when the same bit is set. */
58e23c8d
YO
1914 char *tmp_s = reginfo.ganch - prog->gofs;
1915 if (regtry(&reginfo, &tmp_s))
6eb5f6b9
JH
1916 goto got_it;
1917 goto phooey;
1918 }
1919
1920 /* Messy cases: unanchored match. */
bbe252da 1921 if ((prog->anchored_substr || prog->anchored_utf8) && prog->intflags & PREGf_SKIP) {
6eb5f6b9
JH
1922 /* we have /x+whatever/ */
1923 /* it must be a one character string (XXXX Except UTF?) */
33b8afdf 1924 char ch;
bf93d4cc
GS
1925#ifdef DEBUGGING
1926 int did_match = 0;
1927#endif
33b8afdf
JH
1928 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1929 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
3f7c398e 1930 ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
bf93d4cc 1931
1aa99e6b 1932 if (do_utf8) {
4cadc6a9 1933 REXEC_FBC_SCAN(
6eb5f6b9 1934 if (*s == ch) {
a3621e74 1935 DEBUG_EXECUTE_r( did_match = 1 );
24b23f37 1936 if (regtry(&reginfo, &s)) goto got_it;
6eb5f6b9
JH
1937 s += UTF8SKIP(s);
1938 while (s < strend && *s == ch)
1939 s += UTF8SKIP(s);
1940 }
4cadc6a9 1941 );
6eb5f6b9
JH
1942 }
1943 else {
4cadc6a9 1944 REXEC_FBC_SCAN(
6eb5f6b9 1945 if (*s == ch) {
a3621e74 1946 DEBUG_EXECUTE_r( did_match = 1 );
24b23f37 1947 if (regtry(&reginfo, &s)) goto got_it;
6eb5f6b9
JH
1948 s++;
1949 while (s < strend && *s == ch)
1950 s++;
1951 }
4cadc6a9 1952 );
6eb5f6b9 1953 }
a3621e74 1954 DEBUG_EXECUTE_r(if (!did_match)
bf93d4cc 1955 PerlIO_printf(Perl_debug_log,
b7953727
JH
1956 "Did not find anchored character...\n")
1957 );
6eb5f6b9 1958 }
a0714e2c
SS
1959 else if (prog->anchored_substr != NULL
1960 || prog->anchored_utf8 != NULL
1961 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
33b8afdf
JH
1962 && prog->float_max_offset < strend - s)) {
1963 SV *must;
1964 I32 back_max;
1965 I32 back_min;
1966 char *last;
6eb5f6b9 1967 char *last1; /* Last position checked before */
bf93d4cc
GS
1968#ifdef DEBUGGING
1969 int did_match = 0;
1970#endif
33b8afdf
JH
1971 if (prog->anchored_substr || prog->anchored_utf8) {
1972 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1973 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1974 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1975 back_max = back_min = prog->anchored_offset;
1976 } else {
1977 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1978 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1979 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1980 back_max = prog->float_max_offset;
1981 back_min = prog->float_min_offset;
1982 }
1de06328
YO
1983
1984
33b8afdf
JH
1985 if (must == &PL_sv_undef)
1986 /* could not downgrade utf8 check substring, so must fail */
1987 goto phooey;
1988
1de06328
YO
1989 if (back_min<0) {
1990 last = strend;
1991 } else {
1992 last = HOP3c(strend, /* Cannot start after this */
1993 -(I32)(CHR_SVLEN(must)
1994 - (SvTAIL(must) != 0) + back_min), strbeg);
1995 }
6eb5f6b9
JH
1996 if (s > PL_bostr)
1997 last1 = HOPc(s, -1);
1998 else
1999 last1 = s - 1; /* bogus */
2000
a0288114 2001 /* XXXX check_substr already used to find "s", can optimize if
6eb5f6b9
JH
2002 check_substr==must. */
2003 scream_pos = -1;
2004 dontbother = end_shift;
2005 strend = HOPc(strend, -dontbother);
2006 while ( (s <= last) &&
9041c2e3 2007 ((flags & REXEC_SCREAM)
1de06328 2008 ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg,
6eb5f6b9 2009 end_shift, &scream_pos, 0))
1de06328 2010 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)),
9041c2e3 2011 (unsigned char*)strend, must,
7fba1cd6 2012 multiline ? FBMrf_MULTILINE : 0))) ) {
4addbd3b 2013 /* we may be pointing at the wrong string */
07bc277f 2014 if ((flags & REXEC_SCREAM) && RXp_MATCH_COPIED(prog))
3f7c398e 2015 s = strbeg + (s - SvPVX_const(sv));
a3621e74 2016 DEBUG_EXECUTE_r( did_match = 1 );
6eb5f6b9
JH
2017 if (HOPc(s, -back_max) > last1) {
2018 last1 = HOPc(s, -back_min);
2019 s = HOPc(s, -back_max);
2020 }
2021 else {
52657f30 2022 char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
6eb5f6b9
JH
2023
2024 last1 = HOPc(s, -back_min);
52657f30 2025 s = t;
6eb5f6b9 2026 }
1aa99e6b 2027 if (do_utf8) {
6eb5f6b9 2028 while (s <= last1) {
24b23f37 2029 if (regtry(&reginfo, &s))
6eb5f6b9
JH
2030 goto got_it;
2031 s += UTF8SKIP(s);
2032 }
2033 }
2034 else {
2035 while (s <= last1) {
24b23f37 2036 if (regtry(&reginfo, &s))
6eb5f6b9
JH
2037 goto got_it;
2038 s++;
2039 }
2040 }
2041 }
ab3bbdeb
YO
2042 DEBUG_EXECUTE_r(if (!did_match) {
2043 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
2044 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
2045 PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n",
33b8afdf 2046 ((must == prog->anchored_substr || must == prog->anchored_utf8)
bf93d4cc 2047 ? "anchored" : "floating"),
ab3bbdeb
YO
2048 quoted, RE_SV_TAIL(must));
2049 });
6eb5f6b9
JH
2050 goto phooey;
2051 }
f8fc2ecf 2052 else if ( (c = progi->regstclass) ) {
f14c76ed 2053 if (minlen) {
f8fc2ecf 2054 const OPCODE op = OP(progi->regstclass);
66e933ab 2055 /* don't bother with what can't match */
786e8c11 2056 if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE)
f14c76ed
RGS
2057 strend = HOPc(strend, -(minlen - 1));
2058 }
a3621e74 2059 DEBUG_EXECUTE_r({
be8e71aa 2060 SV * const prop = sv_newmortal();
32fc9b6a 2061 regprop(prog, prop, c);
0df25f3d 2062 {
02daf0ab 2063 RE_PV_QUOTED_DECL(quoted,do_utf8,PERL_DEBUG_PAD_ZERO(1),
ab3bbdeb 2064 s,strend-s,60);
0df25f3d 2065 PerlIO_printf(Perl_debug_log,
ab3bbdeb 2066 "Matching stclass %.*s against %s (%d chars)\n",
e4f74956 2067 (int)SvCUR(prop), SvPVX_const(prop),
ab3bbdeb 2068 quoted, (int)(strend - s));
0df25f3d 2069 }
ffc61ed2 2070 });
3b0527fe 2071 if (find_byclass(prog, c, s, strend, &reginfo))
6eb5f6b9 2072 goto got_it;
07be1b83 2073 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n"));
d6a28714
JH
2074 }
2075 else {
2076 dontbother = 0;
a0714e2c 2077 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
33b8afdf 2078 /* Trim the end. */
d6a28714 2079 char *last;
33b8afdf
JH
2080 SV* float_real;
2081
2082 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
2083 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
2084 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
d6a28714
JH
2085
2086 if (flags & REXEC_SCREAM) {
33b8afdf 2087 last = screaminstr(sv, float_real, s - strbeg,
d6a28714
JH
2088 end_shift, &scream_pos, 1); /* last one */
2089 if (!last)
ffc61ed2 2090 last = scream_olds; /* Only one occurrence. */
4addbd3b 2091 /* we may be pointing at the wrong string */
07bc277f 2092 else if (RXp_MATCH_COPIED(prog))
3f7c398e 2093 s = strbeg + (s - SvPVX_const(sv));
b8c5462f 2094 }
d6a28714
JH
2095 else {
2096 STRLEN len;
cfd0369c 2097 const char * const little = SvPV_const(float_real, len);
d6a28714 2098
33b8afdf 2099 if (SvTAIL(float_real)) {
d6a28714
JH
2100 if (memEQ(strend - len + 1, little, len - 1))
2101 last = strend - len + 1;
7fba1cd6 2102 else if (!multiline)
9041c2e3 2103 last = memEQ(strend - len, little, len)
bd61b366 2104 ? strend - len : NULL;
b8c5462f 2105 else
d6a28714
JH
2106 goto find_last;
2107 } else {
2108 find_last:
9041c2e3 2109 if (len)
d6a28714 2110 last = rninstr(s, strend, little, little + len);
b8c5462f 2111 else
a0288114 2112 last = strend; /* matching "$" */
b8c5462f 2113 }
b8c5462f 2114 }
bf93d4cc 2115 if (last == NULL) {
6bda09f9
YO
2116 DEBUG_EXECUTE_r(
2117 PerlIO_printf(Perl_debug_log,
2118 "%sCan't trim the tail, match fails (should not happen)%s\n",
2119 PL_colors[4], PL_colors[5]));
bf93d4cc
GS
2120 goto phooey; /* Should not happen! */
2121 }
d6a28714
JH
2122 dontbother = strend - last + prog->float_min_offset;
2123 }
2124 if (minlen && (dontbother < minlen))
2125 dontbother = minlen - 1;
2126 strend -= dontbother; /* this one's always in bytes! */
2127 /* We don't know much -- general case. */
1aa99e6b 2128 if (do_utf8) {
d6a28714 2129 for (;;) {
24b23f37 2130 if (regtry(&reginfo, &s))
d6a28714
JH
2131 goto got_it;
2132 if (s >= strend)
2133 break;
b8c5462f 2134 s += UTF8SKIP(s);
d6a28714
JH
2135 };
2136 }
2137 else {
2138 do {
24b23f37 2139 if (regtry(&reginfo, &s))
d6a28714
JH
2140 goto got_it;
2141 } while (s++ < strend);
2142 }
2143 }
2144
2145 /* Failure. */
2146 goto phooey;
2147
2148got_it:
288b8c02 2149 RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted);
d6a28714 2150
19b95bf0 2151 if (PL_reg_eval_set)
4f639d21 2152 restore_pos(aTHX_ prog);
5daac39c
NC
2153 if (RXp_PAREN_NAMES(prog))
2154 (void)hv_iterinit(RXp_PAREN_NAMES(prog));
d6a28714
JH
2155
2156 /* make sure $`, $&, $', and $digit will work later */
2157 if ( !(flags & REXEC_NOT_FIRST) ) {
288b8c02 2158 RX_MATCH_COPY_FREE(rx);
d6a28714 2159 if (flags & REXEC_COPY_STR) {
be8e71aa 2160 const I32 i = PL_regeol - startpos + (stringarg - strbeg);
f8c7b90f 2161#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
2162 if ((SvIsCOW(sv)
2163 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2164 if (DEBUG_C_TEST) {
2165 PerlIO_printf(Perl_debug_log,
2166 "Copy on write: regexp capture, type %d\n",
2167 (int) SvTYPE(sv));
2168 }
2169 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
d5263905 2170 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
ed252734
NC
2171 assert (SvPOKp(prog->saved_copy));
2172 } else
2173#endif
2174 {
288b8c02 2175 RX_MATCH_COPIED_on(rx);
ed252734
NC
2176 s = savepvn(strbeg, i);
2177 prog->subbeg = s;
2178 }
d6a28714 2179 prog->sublen = i;
d6a28714
JH
2180 }
2181 else {
2182 prog->subbeg = strbeg;
2183 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2184 }
2185 }
9041c2e3 2186
d6a28714
JH
2187 return 1;
2188
2189phooey:
a3621e74 2190 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
e4584336 2191 PL_colors[4], PL_colors[5]));
d6a28714 2192 if (PL_reg_eval_set)
4f639d21 2193 restore_pos(aTHX_ prog);
fae667d5 2194 if (swap_on_fail)
c74340f9 2195 /* we failed :-( roll it back */
fae667d5
YO
2196 swap_match_buff(prog);
2197
d6a28714
JH
2198 return 0;
2199}
2200
6bda09f9 2201
d6a28714
JH
2202/*
2203 - regtry - try match at specific point
2204 */
2205STATIC I32 /* 0 failure, 1 success */
24b23f37 2206S_regtry(pTHX_ regmatch_info *reginfo, char **startpos)
d6a28714 2207{
97aff369 2208 dVAR;
d6a28714 2209 CHECKPOINT lastcp;
288b8c02
NC
2210 REGEXP *const rx = reginfo->prog;
2211 regexp *const prog = (struct regexp *)SvANY(rx);
f8fc2ecf 2212 RXi_GET_DECL(prog,progi);
a3621e74 2213 GET_RE_DEBUG_FLAGS_DECL;
24b23f37 2214 reginfo->cutpoint=NULL;
d6a28714 2215
bbe252da 2216 if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) {
d6a28714
JH
2217 MAGIC *mg;
2218
2219 PL_reg_eval_set = RS_init;
a3621e74 2220 DEBUG_EXECUTE_r(DEBUG_s(
b900a521
JH
2221 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2222 (IV)(PL_stack_sp - PL_stack_base));
d6a28714 2223 ));
ea8d6ae1 2224 SAVESTACK_CXPOS();
d6a28714
JH
2225 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2226 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2227 SAVETMPS;
2228 /* Apparently this is not needed, judging by wantarray. */
e8347627 2229 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
d6a28714
JH
2230 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2231
3b0527fe 2232 if (reginfo->sv) {
d6a28714 2233 /* Make $_ available to executed code. */
3b0527fe 2234 if (reginfo->sv != DEFSV) {
59f00321 2235 SAVE_DEFSV;
3b0527fe 2236 DEFSV = reginfo->sv;
b8c5462f 2237 }
d6a28714 2238
3b0527fe
DM
2239 if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv)
2240 && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) {
d6a28714 2241 /* prepare for quick setting of pos */
d300d9fa 2242#ifdef PERL_OLD_COPY_ON_WRITE
51a9ea20
NC
2243 if (SvIsCOW(reginfo->sv))
2244 sv_force_normal_flags(reginfo->sv, 0);
d300d9fa 2245#endif
3dab1dad 2246 mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global,
d300d9fa 2247 &PL_vtbl_mglob, NULL, 0);
d6a28714 2248 mg->mg_len = -1;
b8c5462f 2249 }
d6a28714
JH
2250 PL_reg_magic = mg;
2251 PL_reg_oldpos = mg->mg_len;
4f639d21 2252 SAVEDESTRUCTOR_X(restore_pos, prog);
d6a28714 2253 }
09687e5a 2254 if (!PL_reg_curpm) {
a02a5408 2255 Newxz(PL_reg_curpm, 1, PMOP);
09687e5a
AB
2256#ifdef USE_ITHREADS
2257 {
be8e71aa 2258 SV* const repointer = newSViv(0);
92313705
NC
2259 /* this regexp is also owned by the new PL_reg_curpm, which
2260 will try to free it. */
d2ece331 2261 av_push(PL_regex_padav, repointer);
09687e5a
AB
2262 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2263 PL_regex_pad = AvARRAY(PL_regex_padav);
2264 }
2265#endif
2266 }
92313705
NC
2267 /* This is safe against NULLs: */
2268 ReREFCNT_dec(PM_GETRE(PL_reg_curpm));
2269 /* PM_reg_curpm owns a reference to this regexp. */
2270 ReREFCNT_inc(rx);
288b8c02 2271 PM_SETRE(PL_reg_curpm, rx);
d6a28714
JH
2272 PL_reg_oldcurpm = PL_curpm;
2273 PL_curpm = PL_reg_curpm;
07bc277f 2274 if (RXp_MATCH_COPIED(prog)) {
d6a28714
JH
2275 /* Here is a serious problem: we cannot rewrite subbeg,
2276 since it may be needed if this match fails. Thus
2277 $` inside (?{}) could fail... */
2278 PL_reg_oldsaved = prog->subbeg;
2279 PL_reg_oldsavedlen = prog->sublen;
f8c7b90f 2280#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
2281 PL_nrs = prog->saved_copy;
2282#endif
07bc277f 2283 RXp_MATCH_COPIED_off(prog);
d6a28714
JH
2284 }
2285 else
bd61b366 2286 PL_reg_oldsaved = NULL;
d6a28714
JH
2287 prog->subbeg = PL_bostr;
2288 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2289 }
24b23f37 2290 DEBUG_EXECUTE_r(PL_reg_starttry = *startpos);
f0ab9afb 2291 prog->offs[0].start = *startpos - PL_bostr;
24b23f37 2292 PL_reginput = *startpos;
d6a28714 2293 PL_reglastparen = &prog->lastparen;
a01268b5 2294 PL_reglastcloseparen = &prog->lastcloseparen;
d6a28714 2295 prog->lastparen = 0;
03994de8 2296 prog->lastcloseparen = 0;
d6a28714 2297 PL_regsize = 0;
f0ab9afb 2298 PL_regoffs = prog->offs;
d6a28714
JH
2299 if (PL_reg_start_tmpl <= prog->nparens) {
2300 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2301 if(PL_reg_start_tmp)
2302 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2303 else
a02a5408 2304 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
d6a28714
JH
2305 }
2306
2307 /* XXXX What this code is doing here?!!! There should be no need
2308 to do this again and again, PL_reglastparen should take care of
3dd2943c 2309 this! --ilya*/
dafc8851
JH
2310
2311 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2312 * Actually, the code in regcppop() (which Ilya may be meaning by
daf18116 2313 * PL_reglastparen), is not needed at all by the test suite
225593e1
DM
2314 * (op/regexp, op/pat, op/split), but that code is needed otherwise
2315 * this erroneously leaves $1 defined: "1" =~ /^(?:(\d)x)?\d$/
2316 * Meanwhile, this code *is* needed for the
daf18116
JH
2317 * above-mentioned test suite tests to succeed. The common theme
2318 * on those tests seems to be returning null fields from matches.
225593e1 2319 * --jhi updated by dapm */
dafc8851 2320#if 1
d6a28714 2321 if (prog->nparens) {
f0ab9afb 2322 regexp_paren_pair *pp = PL_regoffs;
097eb12c 2323 register I32 i;
eb160463 2324 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
f0ab9afb
NC
2325 ++pp;
2326 pp->start = -1;
2327 pp->end = -1;
d6a28714
JH
2328 }
2329 }
dafc8851 2330#endif
02db2b7b 2331 REGCP_SET(lastcp);
f8fc2ecf 2332 if (regmatch(reginfo, progi->program + 1)) {
f0ab9afb 2333 PL_regoffs[0].end = PL_reginput - PL_bostr;
d6a28714
JH
2334 return 1;
2335 }
24b23f37
YO
2336 if (reginfo->cutpoint)
2337 *startpos= reginfo->cutpoint;
02db2b7b 2338 REGCP_UNWIND(lastcp);
d6a28714
JH
2339 return 0;
2340}
2341
02db2b7b 2342
8ba1375e
MJD
2343#define sayYES goto yes
2344#define sayNO goto no
262b90c4 2345#define sayNO_SILENT goto no_silent
8ba1375e 2346
f9f4320a
YO
2347/* we dont use STMT_START/END here because it leads to
2348 "unreachable code" warnings, which are bogus, but distracting. */
2349#define CACHEsayNO \
c476f425
DM
2350 if (ST.cache_mask) \
2351 PL_reg_poscache[ST.cache_offset] |= ST.cache_mask; \
f9f4320a 2352 sayNO
3298f257 2353
a3621e74 2354/* this is used to determine how far from the left messages like
265c4333
YO
2355 'failed...' are printed. It should be set such that messages
2356 are inline with the regop output that created them.
a3621e74 2357*/
265c4333 2358#define REPORT_CODE_OFF 32
a3621e74
YO
2359
2360
2361/* Make sure there is a test for this +1 options in re_tests */
2362#define TRIE_INITAL_ACCEPT_BUFFLEN 4;
2363
40a82448
DM
2364#define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */
2365#define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */
9e137952 2366
86545054
DM
2367#define SLAB_FIRST(s) (&(s)->states[0])
2368#define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1])
2369
5d9a96ca
DM
2370/* grab a new slab and return the first slot in it */
2371
2372STATIC regmatch_state *
2373S_push_slab(pTHX)
2374{
a35a87e7 2375#if PERL_VERSION < 9 && !defined(PERL_CORE)
54df2634
NC
2376 dMY_CXT;
2377#endif
5d9a96ca
DM
2378 regmatch_slab *s = PL_regmatch_slab->next;
2379 if (!s) {
2380 Newx(s, 1, regmatch_slab);
2381 s->prev = PL_regmatch_slab;
2382 s->next = NULL;
2383 PL_regmatch_slab->next = s;
2384 }
2385 PL_regmatch_slab = s;
86545054 2386 return SLAB_FIRST(s);
5d9a96ca 2387}
5b47454d 2388
95b24440 2389
40a82448
DM
2390/* push a new state then goto it */
2391
2392#define PUSH_STATE_GOTO(state, node) \
2393 scan = node; \
2394 st->resume_state = state; \
2395 goto push_state;
2396
2397/* push a new state with success backtracking, then goto it */
2398
2399#define PUSH_YES_STATE_GOTO(state, node) \
2400 scan = node; \
2401 st->resume_state = state; \
2402 goto push_yes_state;
2403
aa283a38 2404
aa283a38 2405
d6a28714 2406/*
95b24440 2407
bf1f174e
DM
2408regmatch() - main matching routine
2409
2410This is basically one big switch statement in a loop. We execute an op,
2411set 'next' to point the next op, and continue. If we come to a point which
2412we may need to backtrack to on failure such as (A|B|C), we push a
2413backtrack state onto the backtrack stack. On failure, we pop the top
2414state, and re-enter the loop at the state indicated. If there are no more
2415states to pop, we return failure.
2416
2417Sometimes we also need to backtrack on success; for example /A+/, where
2418after successfully matching one A, we need to go back and try to
2419match another one; similarly for lookahead assertions: if the assertion
2420completes successfully, we backtrack to the state just before the assertion
2421and then carry on. In these cases, the pushed state is marked as
2422'backtrack on success too'. This marking is in fact done by a chain of
2423pointers, each pointing to the previous 'yes' state. On success, we pop to
2424the nearest yes state, discarding any intermediate failure-only states.
2425Sometimes a yes state is pushed just to force some cleanup code to be
2426called at the end of a successful match or submatch; e.g. (??{$re}) uses
2427it to free the inner regex.
2428
2429Note that failure backtracking rewinds the cursor position, while
2430success backtracking leaves it alone.
2431
2432A pattern is complete when the END op is executed, while a subpattern
2433such as (?=foo) is complete when the SUCCESS op is executed. Both of these
2434ops trigger the "pop to last yes state if any, otherwise return true"
2435behaviour.
2436
2437A common convention in this function is to use A and B to refer to the two
2438subpatterns (or to the first nodes thereof) in patterns like /A*B/: so A is
2439the subpattern to be matched possibly multiple times, while B is the entire
2440rest of the pattern. Variable and state names reflect this convention.
2441
2442The states in the main switch are the union of ops and failure/success of
2443substates associated with with that op. For example, IFMATCH is the op
2444that does lookahead assertions /(?=A)B/ and so the IFMATCH state means
2445'execute IFMATCH'; while IFMATCH_A is a state saying that we have just
2446successfully matched A and IFMATCH_A_fail is a state saying that we have
2447just failed to match A. Resume states always come in pairs. The backtrack
2448state we push is marked as 'IFMATCH_A', but when that is popped, we resume
2449at IFMATCH_A or IFMATCH_A_fail, depending on whether we are backtracking
2450on success or failure.
2451
2452The struct that holds a backtracking state is actually a big union, with
2453one variant for each major type of op. The variable st points to the
2454top-most backtrack struct. To make the code clearer, within each
2455block of code we #define ST to alias the relevant union.
2456
2457Here's a concrete example of a (vastly oversimplified) IFMATCH
2458implementation:
2459
2460 switch (state) {
2461 ....
2462
2463#define ST st->u.ifmatch
2464
2465 case IFMATCH: // we are executing the IFMATCH op, (?=A)B
2466 ST.foo = ...; // some state we wish to save
95b24440 2467 ...
bf1f174e
DM
2468 // push a yes backtrack state with a resume value of
2469 // IFMATCH_A/IFMATCH_A_fail, then continue execution at the
2470 // first node of A:
2471 PUSH_YES_STATE_GOTO(IFMATCH_A, A);
2472 // NOTREACHED
2473
2474 case IFMATCH_A: // we have successfully executed A; now continue with B
2475 next = B;
2476 bar = ST.foo; // do something with the preserved value
2477 break;
2478
2479 case IFMATCH_A_fail: // A failed, so the assertion failed
2480 ...; // do some housekeeping, then ...
2481 sayNO; // propagate the failure
2482
2483#undef ST
95b24440 2484
bf1f174e
DM
2485 ...
2486 }
95b24440 2487
bf1f174e
DM
2488For any old-timers reading this who are familiar with the old recursive
2489approach, the code above is equivalent to:
95b24440 2490
bf1f174e
DM
2491 case IFMATCH: // we are executing the IFMATCH op, (?=A)B
2492 {
2493 int foo = ...
95b24440 2494 ...
bf1f174e
DM
2495 if (regmatch(A)) {
2496 next = B;
2497 bar = foo;
2498 break;
95b24440 2499 }
bf1f174e
DM
2500 ...; // do some housekeeping, then ...
2501 sayNO; // propagate the failure
95b24440 2502 }
bf1f174e
DM
2503
2504The topmost backtrack state, pointed to by st, is usually free. If you
2505want to claim it, populate any ST.foo fields in it with values you wish to
2506save, then do one of
2507
2508 PUSH_STATE_GOTO(resume_state, node);
2509 PUSH_YES_STATE_GOTO(resume_state, node);
2510
2511which sets that backtrack state's resume value to 'resume_state', pushes a
2512new free entry to the top of the backtrack stack, then goes to 'node'.
2513On backtracking, the free slot is popped, and the saved state becomes the
2514new free state. An ST.foo field in this new top state can be temporarily
2515accessed to retrieve values, but once the main loop is re-entered, it
2516becomes available for reuse.
2517
2518Note that the depth of the backtrack stack constantly increases during the
2519left-to-right execution of the pattern, rather than going up and down with
2520the pattern nesting. For example the stack is at its maximum at Z at the
2521end of the pattern, rather than at X in the following:
2522
2523 /(((X)+)+)+....(Y)+....Z/
2524
2525The only exceptions to this are lookahead/behind assertions and the cut,
2526(?>A), which pop all the backtrack states associated with A before
2527continuing.
2528
2529Bascktrack state structs are allocated in slabs of about 4K in size.
2530PL_regmatch_state and st always point to the currently active state,
2531and PL_regmatch_slab points to the slab currently containing
2532PL_regmatch_state. The first time regmatch() is called, the first slab is
2533allocated, and is never freed until interpreter destruction. When the slab
2534is full, a new one is allocated and chained to the end. At exit from
2535regmatch(), slabs allocated since entry are freed.
2536
2537*/
95b24440 2538
40a82448 2539
5bc10b2c 2540#define DEBUG_STATE_pp(pp) \
265c4333 2541 DEBUG_STATE_r({ \
5bc10b2c
DM
2542 DUMP_EXEC_POS(locinput, scan, do_utf8); \
2543 PerlIO_printf(Perl_debug_log, \
5d458dd8 2544 " %*s"pp" %s%s%s%s%s\n", \
5bc10b2c 2545 depth*2, "", \
13d6edb4 2546 PL_reg_name[st->resume_state], \
5d458dd8
YO
2547 ((st==yes_state||st==mark_state) ? "[" : ""), \
2548 ((st==yes_state) ? "Y" : ""), \
2549 ((st==mark_state) ? "M" : ""), \
2550 ((st==yes_state||st==mark_state) ? "]" : "") \
2551 ); \
265c4333 2552 });
5bc10b2c 2553
40a82448 2554
3dab1dad 2555#define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1)
95b24440 2556
3df15adc 2557#ifdef DEBUGGING
5bc10b2c 2558
ab3bbdeb 2559STATIC void
efd26800 2560S_debug_start_match(pTHX_ const REGEXP *prog, const bool do_utf8,
ab3bbdeb
YO
2561 const char *start, const char *end, const char *blurb)
2562{
efd26800 2563 const bool utf8_pat = RX_UTF8(prog) ? 1 : 0;
ab3bbdeb
YO
2564 if (!PL_colorset)
2565 reginitcolors();
2566 {
2567 RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0),
efd26800 2568 RX_PRECOMP(prog), RX_PRELEN(prog), 60);
ab3bbdeb
YO
2569
2570 RE_PV_QUOTED_DECL(s1, do_utf8, PERL_DEBUG_PAD_ZERO(1),
2571 start, end - start, 60);
2572
2573 PerlIO_printf(Perl_debug_log,
2574 "%s%s REx%s %s against %s\n",
2575 PL_colors[4], blurb, PL_colors[5], s0, s1);
2576
2577 if (do_utf8||utf8_pat)
1de06328
YO
2578 PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n",
2579 utf8_pat ? "pattern" : "",
2580 utf8_pat && do_utf8 ? " and " : "",
2581 do_utf8 ? "string" : ""
ab3bbdeb
YO
2582 );
2583 }
2584}
3df15adc
YO
2585
2586STATIC void
786e8c11
YO
2587S_dump_exec_pos(pTHX_ const char *locinput,
2588 const regnode *scan,
2589 const char *loc_regeol,
2590 const char *loc_bostr,
2591 const char *loc_reg_starttry,
2592 const bool do_utf8)
07be1b83 2593{
786e8c11 2594 const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4];
07be1b83 2595 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
786e8c11 2596 int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput);
07be1b83
YO
2597 /* The part of the string before starttry has one color
2598 (pref0_len chars), between starttry and current
2599 position another one (pref_len - pref0_len chars),
2600 after the current position the third one.
2601 We assume that pref0_len <= pref_len, otherwise we
2602 decrease pref0_len. */
786e8c11
YO
2603 int pref_len = (locinput - loc_bostr) > (5 + taill) - l
2604 ? (5 + taill) - l : locinput - loc_bostr;
07be1b83
YO
2605 int pref0_len;
2606
2607 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2608 pref_len++;
786e8c11
YO
2609 pref0_len = pref_len - (locinput - loc_reg_starttry);
2610 if (l + pref_len < (5 + taill) && l < loc_regeol - locinput)
2611 l = ( loc_regeol - locinput > (5 + taill) - pref_len
2612 ? (5 + taill) - pref_len : loc_regeol - locinput);
07be1b83
YO
2613 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2614 l--;
2615 if (pref0_len < 0)
2616 pref0_len = 0;
2617 if (pref0_len > pref_len)
2618 pref0_len = pref_len;
2619 {
3df15adc 2620 const int is_uni = (do_utf8 && OP(scan) != CANY) ? 1 : 0;
0df25f3d 2621
ab3bbdeb 2622 RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0),
1de06328 2623 (locinput - pref_len),pref0_len, 60, 4, 5);
0df25f3d 2624
ab3bbdeb 2625 RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1),
3df15adc 2626 (locinput - pref_len + pref0_len),
1de06328 2627 pref_len - pref0_len, 60, 2, 3);
0df25f3d 2628
ab3bbdeb 2629 RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2),
1de06328 2630 locinput, loc_regeol - locinput, 10, 0, 1);
0df25f3d 2631
1de06328 2632 const STRLEN tlen=len0+len1+len2;
3df15adc 2633 PerlIO_printf(Perl_debug_log,
ab3bbdeb 2634 "%4"IVdf" <%.*s%.*s%s%.*s>%*s|",
786e8c11 2635 (IV)(locinput - loc_bostr),
07be1b83 2636 len0, s0,
07be1b83 2637 len1, s1,
07be1b83 2638 (docolor ? "" : "> <"),
07be1b83 2639 len2, s2,
f9f4320a 2640 (int)(tlen > 19 ? 0 : 19 - tlen),
07be1b83
YO
2641 "");
2642 }
2643}
3df15adc 2644
07be1b83
YO
2645#endif
2646
0a4db386
YO
2647/* reg_check_named_buff_matched()
2648 * Checks to see if a named buffer has matched. The data array of
2649 * buffer numbers corresponding to the buffer is expected to reside
2650 * in the regexp->data->data array in the slot stored in the ARG() of
2651 * node involved. Note that this routine doesn't actually care about the
2652 * name, that information is not preserved from compilation to execution.
2653 * Returns the index of the leftmost defined buffer with the given name
2654 * or 0 if non of the buffers matched.
2655 */
2656STATIC I32
2657S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) {
2658 I32 n;
f8fc2ecf
YO
2659 RXi_GET_DECL(rex,rexi);
2660 SV *sv_dat=(SV*)rexi->data->data[ ARG( scan ) ];
0a4db386
YO
2661 I32 *nums=(I32*)SvPVX(sv_dat);
2662 for ( n=0; n<SvIVX(sv_dat); n++ ) {
2663 if ((I32)*PL_reglastparen >= nums[n] &&
f0ab9afb 2664 PL_regoffs[nums[n]].end != -1)
0a4db386
YO
2665 {
2666 return nums[n];
2667 }
2668 }
2669 return 0;
2670}
2671
2f554ef7
DM
2672
2673/* free all slabs above current one - called during LEAVE_SCOPE */
2674
2675STATIC void
2676S_clear_backtrack_stack(pTHX_ void *p)
2677{
2678 regmatch_slab *s = PL_regmatch_slab->next;
2679 PERL_UNUSED_ARG(p);
2680
2681 if (!s)
2682 return;
2683 PL_regmatch_slab->next = NULL;
2684 while (s) {
2685 regmatch_slab * const osl = s;
2686 s = s->next;
2687 Safefree(osl);
2688 }
2689}
2690
2691
28d8d7f4
YO
2692#define SETREX(Re1,Re2) \
2693 if (PL_reg_eval_set) PM_SETRE((PL_reg_curpm), (Re2)); \
2694 Re1 = (Re2)
2695
d6a28714 2696STATIC I32 /* 0 failure, 1 success */
24b23f37 2697S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog)
d6a28714 2698{
a35a87e7 2699#if PERL_VERSION < 9 && !defined(PERL_CORE)
54df2634
NC
2700 dMY_CXT;
2701#endif
27da23d5 2702 dVAR;
95b24440 2703 register const bool do_utf8 = PL_reg_match_utf8;
4ad0818d 2704 const U32 uniflags = UTF8_ALLOW_DEFAULT;
95b24440 2705
288b8c02
NC
2706 REGEXP *rex_sv = reginfo->prog;
2707 regexp *rex = (struct regexp *)SvANY(rex_sv);
f8fc2ecf
YO
2708 RXi_GET_DECL(rex,rexi);
2709
2f554ef7 2710 I32 oldsave;
a3621e74 2711
5d9a96ca
DM
2712 /* the current state. This is a cached copy of PL_regmatch_state */
2713 register regmatch_state *st;
95b24440 2714
5d9a96ca
DM
2715 /* cache heavy used fields of st in registers */
2716 register regnode *scan;
2717 register regnode *next;
438e9bae 2718 register U32 n = 0; /* general value; init to avoid compiler warning */
24d3c4a9 2719 register I32 ln = 0; /* len or last; init to avoid compiler warning */
5d9a96ca 2720 register char *locinput = PL_reginput;
5d9a96ca 2721 register I32 nextchr; /* is always set to UCHARAT(locinput) */
24d3c4a9 2722
b69b0499 2723 bool result = 0; /* return value of S_regmatch */
24d3c4a9 2724 int depth = 0; /* depth of backtrack stack */
4b196cd4
YO
2725 U32 nochange_depth = 0; /* depth of GOSUB recursion with nochange */
2726 const U32 max_nochange_depth =
2727 (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ?
2728 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH;
2729
77cb431f
DM
2730 regmatch_state *yes_state = NULL; /* state to pop to on success of
2731 subpattern */
e2e6a0f1
YO
2732 /* mark_state piggy backs on the yes_state logic so that when we unwind
2733 the stack on success we can update the mark_state as we go */
2734 regmatch_state *mark_state = NULL; /* last mark state we have seen */
5d458dd8 2735
faec1544 2736 regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */
b8591aee 2737 struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */
40a82448 2738 U32 state_num;
5d458dd8
YO
2739 bool no_final = 0; /* prevent failure from backtracking? */
2740 bool do_cutgroup = 0; /* no_final only until next branch/trie entry */
e2e6a0f1 2741 char *startpoint = PL_reginput;
5d458dd8
YO
2742 SV *popmark = NULL; /* are we looking for a mark? */
2743 SV *sv_commit = NULL; /* last mark name seen in failure */
2744 SV *sv_yes_mark = NULL; /* last mark name we have seen
2745 during a successfull match */
2746 U32 lastopen = 0; /* last open we saw */
2747 bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0;
19b95bf0
DM
2748
2749 SV* const oreplsv = GvSV(PL_replgv);
58e23c8d 2750
5d458dd8 2751
24d3c4a9
DM
2752 /* these three flags are set by various ops to signal information to
2753 * the very next op. They have a useful lifetime of exactly one loop
2754 * iteration, and are not preserved or restored by state pushes/pops
2755 */
2756 bool sw = 0; /* the condition value in (?(cond)a|b) */
2757 bool minmod = 0; /* the next "{n,m}" is a "{n,m}?" */
2758 int logical = 0; /* the following EVAL is:
2759 0: (?{...})
2760 1: (?(?{...})X|Y)
2761 2: (??{...})
2762 or the following IFMATCH/UNLESSM is:
2763 false: plain (?=foo)
2764 true: used as a condition: (?(?=foo))
2765 */
2766
95b24440 2767#ifdef DEBUGGING
e68ec53f 2768 GET_RE_DEBUG_FLAGS_DECL;
d6a28714
JH
2769#endif
2770
3b57cd43 2771 DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({
24b23f37 2772 PerlIO_printf(Perl_debug_log,"regmatch start\n");
3b57cd43 2773 }));
5d9a96ca
DM
2774 /* on first ever call to regmatch, allocate first slab */
2775 if (!PL_regmatch_slab) {
2776 Newx(PL_regmatch_slab, 1, regmatch_slab);
2777 PL_regmatch_slab->prev = NULL;
2778 PL_regmatch_slab->next = NULL;
86545054 2779 PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab);
5d9a96ca
DM
2780 }
2781
2f554ef7
DM
2782 oldsave = PL_savestack_ix;
2783 SAVEDESTRUCTOR_X(S_clear_backtrack_stack, NULL);
2784 SAVEVPTR(PL_regmatch_slab);
2785 SAVEVPTR(PL_regmatch_state);
5d9a96ca
DM
2786
2787 /* grab next free state slot */
2788 st = ++PL_regmatch_state;
86545054 2789 if (st > SLAB_LAST(PL_regmatch_slab))
5d9a96ca
DM
2790 st = PL_regmatch_state = S_push_slab(aTHX);
2791
d6a28714
JH
2792 /* Note that nextchr is a byte even in UTF */
2793 nextchr = UCHARAT(locinput);
2794 scan = prog;
2795 while (scan != NULL) {
8ba1375e 2796
a3621e74 2797 DEBUG_EXECUTE_r( {
6136c704 2798 SV * const prop = sv_newmortal();
1de06328 2799 regnode *rnext=regnext(scan);
786e8c11 2800 DUMP_EXEC_POS( locinput, scan, do_utf8 );
32fc9b6a 2801 regprop(rex, prop, scan);
07be1b83
YO
2802
2803 PerlIO_printf(Perl_debug_log,
2804 "%3"IVdf":%*s%s(%"IVdf")\n",
f8fc2ecf 2805 (IV)(scan - rexi->program), depth*2, "",
07be1b83 2806 SvPVX_const(prop),
1de06328 2807 (PL_regkind[OP(scan)] == END || !rnext) ?
f8fc2ecf 2808 0 : (IV)(rnext - rexi->program));
2a782b5b 2809 });
d6a28714
JH
2810
2811 next = scan + NEXT_OFF(scan);
2812 if (next == scan)
2813 next = NULL;
40a82448 2814 state_num = OP(scan);
d6a28714 2815
40a82448
DM
2816 reenter_switch:
2817 switch (state_num) {
d6a28714 2818 case BOL:
7fba1cd6 2819 if (locinput == PL_bostr)
d6a28714 2820 {
3b0527fe 2821 /* reginfo->till = reginfo->bol; */
b8c5462f
JH
2822 break;
2823 }
d6a28714
JH
2824 sayNO;
2825 case MBOL:
12d33761
HS
2826 if (locinput == PL_bostr ||
2827 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
d6a28714 2828 {
b8c5462f
JH
2829 break;
2830 }
d6a28714
JH
2831 sayNO;
2832 case SBOL:
c2a73568 2833 if (locinput == PL_bostr)
b8c5462f 2834 break;
d6a28714
JH
2835 sayNO;
2836 case GPOS:
3b0527fe 2837 if (locinput == reginfo->ganch)
d6a28714
JH
2838 break;
2839 sayNO;
ee9b8eae
YO
2840
2841 case KEEPS:
2842 /* update the startpoint */
f0ab9afb 2843 st->u.keeper.val = PL_regoffs[0].start;
ee9b8eae 2844 PL_reginput = locinput;
f0ab9afb 2845 PL_regoffs[0].start = locinput - PL_bostr;
ee9b8eae
YO
2846 PUSH_STATE_GOTO(KEEPS_next, next);
2847 /*NOT-REACHED*/
2848 case KEEPS_next_fail:
2849 /* rollback the start point change */
f0ab9afb 2850 PL_regoffs[0].start = st->u.keeper.val;
ee9b8eae
YO
2851 sayNO_SILENT;
2852 /*NOT-REACHED*/
d6a28714 2853 case EOL:
d6a28714
JH
2854 goto seol;
2855 case MEOL:
d6a28714 2856 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
b8c5462f 2857 sayNO;
b8c5462f 2858 break;
d6a28714
JH
2859 case SEOL:
2860 seol:
2861 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
b8c5462f 2862 sayNO;
d6a28714 2863 if (PL_regeol - locinput > 1)
b8c5462f 2864 sayNO;
b8c5462f 2865 break;
d6a28714
JH
2866 case EOS:
2867 if (PL_regeol != locinput)
b8c5462f 2868 sayNO;
d6a28714 2869 break;
ffc61ed2 2870 case SANY:
d6a28714 2871 if (!nextchr && locinput >= PL_regeol)
4633a7c4 2872 sayNO;
f33976b4
DB
2873 if (do_utf8) {
2874 locinput += PL_utf8skip[nextchr];
2875 if (locinput > PL_regeol)
2876 sayNO;
2877 nextchr = UCHARAT(locinput);
2878 }
2879 else
2880 nextchr = UCHARAT(++locinput);
2881 break;
2882 case CANY:
2883 if (!nextchr && locinput >= PL_regeol)
2884 sayNO;
b8c5462f 2885 nextchr = UCHARAT(++locinput);
a0d0e21e 2886 break;
ffc61ed2 2887 case REG_ANY:
1aa99e6b
IH
2888 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2889 sayNO;
2890 if (do_utf8) {
b8c5462f 2891 locinput += PL_utf8skip[nextchr];
d6a28714
JH
2892 if (locinput > PL_regeol)
2893 sayNO;
a0ed51b3 2894 nextchr = UCHARAT(locinput);
a0ed51b3 2895 }
1aa99e6b
IH
2896 else
2897 nextchr = UCHARAT(++locinput);
a0ed51b3 2898 break;
166ba7cd
DM
2899
2900#undef ST
2901#define ST st->u.trie
786e8c11
YO
2902 case TRIEC:
2903 /* In this case the charclass data is available inline so
2904 we can fail fast without a lot of extra overhead.
2905 */
2906 if (scan->flags == EXACT || !do_utf8) {
2907 if(!ANYOF_BITMAP_TEST(scan, *locinput)) {
2908 DEBUG_EXECUTE_r(
2909 PerlIO_printf(Perl_debug_log,
2910 "%*s %sfailed to match trie start class...%s\n",
5bc10b2c 2911 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
786e8c11
YO
2912 );
2913 sayNO_SILENT;
2914 /* NOTREACHED */
2915 }
2916 }
2917 /* FALL THROUGH */
5b47454d 2918 case TRIE:
3dab1dad 2919 {
07be1b83 2920 /* what type of TRIE am I? (utf8 makes this contextual) */
a0a388a1 2921 DECL_TRIE_TYPE(scan);
3dab1dad
YO
2922
2923 /* what trie are we using right now */
be8e71aa 2924 reg_trie_data * const trie
f8fc2ecf 2925 = (reg_trie_data*)rexi->data->data[ ARG( scan ) ];
55eed653 2926 HV * widecharmap = (HV *)rexi->data->data[ ARG( scan ) + 1 ];
3dab1dad 2927 U32 state = trie->startstate;
166ba7cd 2928
3dab1dad
YO
2929 if (trie->bitmap && trie_type != trie_utf8_fold &&
2930 !TRIE_BITMAP_TEST(trie,*locinput)
2931 ) {
2932 if (trie->states[ state ].wordnum) {
2933 DEBUG_EXECUTE_r(
2934 PerlIO_printf(Perl_debug_log,
2935 "%*s %smatched empty string...%s\n",
5bc10b2c 2936 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3dab1dad
YO
2937 );
2938 break;
2939 } else {
2940 DEBUG_EXECUTE_r(
2941 PerlIO_printf(Perl_debug_log,
786e8c11 2942 "%*s %sfailed to match trie start class...%s\n",
5bc10b2c 2943 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5])
3dab1dad
YO
2944 );
2945 sayNO_SILENT;
2946 }
2947 }
166ba7cd 2948
786e8c11
YO
2949 {
2950 U8 *uc = ( U8* )locinput;
2951
2952 STRLEN len = 0;
2953 STRLEN foldlen = 0;
2954 U8 *uscan = (U8*)NULL;
2955 STRLEN bufflen=0;
2956 SV *sv_accept_buff = NULL;
2957 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
2958
2959 ST.accepted = 0; /* how many accepting states we have seen */
2960 ST.B = next;
2961 ST.jump = trie->jump;
786e8c11 2962 ST.me = scan;
07be1b83
YO
2963 /*
2964 traverse the TRIE keeping track of all accepting states
2965 we transition through until we get to a failing node.
2966 */
2967
a3621e74 2968 while ( state && uc <= (U8*)PL_regeol ) {
786e8c11 2969 U32 base = trie->states[ state ].trans.base;
f9f4320a 2970 UV uvc = 0;
786e8c11
YO
2971 U16 charid;
2972 /* We use charid to hold the wordnum as we don't use it
2973 for charid until after we have done the wordnum logic.
2974 We define an alias just so that the wordnum logic reads
2975 more naturally. */
2976
2977#define got_wordnum charid
2978 got_wordnum = trie->states[ state ].wordnum;
2979
2980 if ( got_wordnum ) {
2981 if ( ! ST.accepted ) {
5b47454d 2982 ENTER;
6b173516 2983 /* SAVETMPS; */ /* XXX is this necessary? dmq */
5b47454d
DM
2984 bufflen = TRIE_INITAL_ACCEPT_BUFFLEN;
2985 sv_accept_buff=newSV(bufflen *
2986 sizeof(reg_trie_accepted) - 1);
786e8c11 2987 SvCUR_set(sv_accept_buff, 0);
5b47454d
DM
2988 SvPOK_on(sv_accept_buff);
2989 sv_2mortal(sv_accept_buff);
166ba7cd
DM
2990 SAVETMPS;
2991 ST.accept_buff =
5b47454d
DM
2992 (reg_trie_accepted*)SvPV_nolen(sv_accept_buff );
2993 }
786e8c11 2994 do {
166ba7cd 2995 if (ST.accepted >= bufflen) {
5b47454d 2996 bufflen *= 2;
166ba7cd 2997 ST.accept_buff =(reg_trie_accepted*)
5b47454d
DM
2998 SvGROW(sv_accept_buff,
2999 bufflen * sizeof(reg_trie_accepted));
3000 }
3001 SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff)
3002 + sizeof(reg_trie_accepted));
a3621e74 3003
786e8c11
YO
3004
3005 ST.accept_buff[ST.accepted].wordnum = got_wordnum;
3006 ST.accept_buff[ST.accepted].endpos = uc;
3007 ++ST.accepted;
3008 } while (trie->nextword && (got_wordnum= trie->nextword[got_wordnum]));
3009 }
3010#undef got_wordnum
a3621e74 3011
07be1b83 3012 DEBUG_TRIE_EXECUTE_r({
786e8c11 3013 DUMP_EXEC_POS( (char *)uc, scan, do_utf8 );
a3621e74 3014 PerlIO_printf( Perl_debug_log,
786e8c11 3015 "%*s %sState: %4"UVxf" Accepted: %4"UVxf" ",
5bc10b2c 3016 2+depth * 2, "", PL_colors[4],
786e8c11 3017 (UV)state, (UV)ST.accepted );
07be1b83 3018 });
a3621e74
YO
3019
3020 if ( base ) {
55eed653
NC
3021 REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc,
3022 uscan, len, uvc, charid, foldlen,
3023 foldbuf, uniflags);
a3621e74 3024
5b47454d
DM
3025 if (charid &&
3026 (base + charid > trie->uniquecharcount )
3027 && (base + charid - 1 - trie->uniquecharcount
3028 < trie->lasttrans)
3029 && trie->trans[base + charid - 1 -
3030 trie->uniquecharcount].check == state)
3031 {
3032 state = trie->trans[base + charid - 1 -
3033 trie->uniquecharcount ].next;
3034 }
3035 else {
3036 state = 0;
3037 }
3038 uc += len;
3039
3040 }
3041 else {
a3621e74
YO
3042 state = 0;
3043 }
3044 DEBUG_TRIE_EXECUTE_r(
e4584336 3045 PerlIO_printf( Perl_debug_log,
786e8c11 3046 "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n",
e4584336 3047 charid, uvc, (UV)state, PL_colors[5] );
a3621e74
YO
3048 );
3049 }
166ba7cd 3050 if (!ST.accepted )
a3621e74 3051 sayNO;
a3621e74 3052
166ba7cd
DM
3053 DEBUG_EXECUTE_r(
3054 PerlIO_printf( Perl_debug_log,
3055 "%*s %sgot %"IVdf" possible matches%s\n",
5bc10b2c 3056 REPORT_CODE_OFF + depth * 2, "",
166ba7cd
DM
3057 PL_colors[4], (IV)ST.accepted, PL_colors[5] );
3058 );
786e8c11 3059 }}
fae667d5
YO
3060 goto trie_first_try; /* jump into the fail handler */
3061 /* NOTREACHED */
166ba7cd 3062 case TRIE_next_fail: /* we failed - try next alterative */
fae667d5
YO
3063 if ( ST.jump) {
3064 REGCP_UNWIND(ST.cp);
3065 for (n = *PL_reglastparen; n > ST.lastparen; n--)
f0ab9afb 3066 PL_regoffs[n].end = -1;
fae667d5
YO
3067 *PL_reglastparen = n;
3068 }
3069 trie_first_try:
5d458dd8
YO
3070 if (do_cutgroup) {
3071 do_cutgroup = 0;
3072 no_final = 0;
3073 }
fae667d5
YO
3074
3075 if ( ST.jump) {
3076 ST.lastparen = *PL_reglastparen;
3077 REGCP_SET(ST.cp);
3078 }
166ba7cd
DM
3079 if ( ST.accepted == 1 ) {
3080 /* only one choice left - just continue */
3081 DEBUG_EXECUTE_r({
2b8b4781
NC
3082 AV *const trie_words
3083 = (AV *) rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET];
3084 SV ** const tmp = av_fetch( trie_words,
5c9f2f80 3085 ST.accept_buff[ 0 ].wordnum-1, 0 );
de734bd5
A
3086 SV *sv= tmp ? sv_newmortal() : NULL;
3087
166ba7cd
DM
3088 PerlIO_printf( Perl_debug_log,
3089 "%*s %sonly one match left: #%d <%s>%s\n",
5bc10b2c 3090 REPORT_CODE_OFF+depth*2, "", PL_colors[4],
166ba7cd 3091 ST.accept_buff[ 0 ].wordnum,
de734bd5
A
3092 tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0,
3093 PL_colors[0], PL_colors[1],
3094 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
3095 )
3096 : "not compiled under -Dr",
166ba7cd
DM
3097 PL_colors[5] );
3098 });
3099 PL_reginput = (char *)ST.accept_buff[ 0 ].endpos;
3100 /* in this case we free tmps/leave before we call regmatch
3101 as we wont be using accept_buff again. */
5d458dd8 3102
166ba7cd
DM
3103 locinput = PL_reginput;
3104 nextchr = UCHARAT(locinput);
5d458dd8
YO
3105 if ( !ST.jump || !ST.jump[ST.accept_buff[0].wordnum])
3106 scan = ST.B;
3107 else
3108 scan = ST.me + ST.jump[ST.accept_buff[0].wordnum];
3109 if (!has_cutgroup) {
3110 FREETMPS;
3111 LEAVE;
3112 } else {
3113 ST.accepted--;
3114 PUSH_YES_STATE_GOTO(TRIE_next, scan);
3115 }
786e8c11 3116
166ba7cd
DM
3117 continue; /* execute rest of RE */
3118 }
fae667d5
YO
3119
3120 if ( !ST.accepted-- ) {
5d458dd8
YO
3121 DEBUG_EXECUTE_r({
3122 PerlIO_printf( Perl_debug_log,
3123 "%*s %sTRIE failed...%s\n",
3124 REPORT_CODE_OFF+depth*2, "",
3125 PL_colors[4],
3126 PL_colors[5] );
3127 });
166ba7cd
DM
3128 FREETMPS;
3129 LEAVE;
5d458dd8 3130 sayNO_SILENT;
fae667d5
YO
3131 /*NOTREACHED*/
3132 }
166ba7cd 3133
a3621e74 3134 /*
166ba7cd
DM
3135 There are at least two accepting states left. Presumably
3136 the number of accepting states is going to be low,
3137 typically two. So we simply scan through to find the one
3138 with lowest wordnum. Once we find it, we swap the last
3139 state into its place and decrement the size. We then try to
3140 match the rest of the pattern at the point where the word
3141 ends. If we succeed, control just continues along the
3142 regex; if we fail we return here to try the next accepting
3143 state
3144 */
a3621e74 3145
166ba7cd
DM
3146 {
3147 U32 best = 0;
3148 U32 cur;
3149 for( cur = 1 ; cur <= ST.accepted ; cur++ ) {
3150 DEBUG_TRIE_EXECUTE_r(
f2278c82 3151 PerlIO_printf( Perl_debug_log,
166ba7cd 3152 "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n",
5bc10b2c 3153 REPORT_CODE_OFF + depth * 2, "", PL_colors[4],
166ba7cd
DM
3154 (IV)best, ST.accept_buff[ best ].wordnum, (IV)cur,
3155 ST.accept_buff[ cur ].wordnum, PL_colors[5] );
3156 );
3157
3158 if (ST.accept_buff[cur].wordnum <
3159 ST.accept_buff[best].wordnum)
3160 best = cur;
a3621e74 3161 }
166ba7cd
DM
3162
3163 DEBUG_EXECUTE_r({
2b8b4781
NC
3164 AV *const trie_words
3165 = (AV *) rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET];
3166 SV ** const tmp = av_fetch( trie_words,
5c9f2f80 3167 ST.accept_buff[ best ].wordnum - 1, 0 );
7f69552c 3168 regnode *nextop=(!ST.jump || !ST.jump[ST.accept_buff[best].wordnum]) ?
265c4333 3169 ST.B :
7f69552c 3170 ST.me + ST.jump[ST.accept_buff[best].wordnum];
de734bd5
A
3171 SV *sv= tmp ? sv_newmortal() : NULL;
3172
265c4333
YO
3173 PerlIO_printf( Perl_debug_log,
3174 "%*s %strying alternation #%d <%s> at node #%d %s\n",
5bc10b2c 3175 REPORT_CODE_OFF+depth*2, "", PL_colors[4],
166ba7cd 3176 ST.accept_buff[best].wordnum,
de734bd5
A
3177 tmp ? pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 0,
3178 PL_colors[0], PL_colors[1],
3179 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
3180 ) : "not compiled under -Dr",
265c4333 3181 REG_NODE_NUM(nextop),
166ba7cd
DM
3182 PL_colors[5] );
3183 });
3184
3185 if ( best<ST.accepted ) {
3186 reg_trie_accepted tmp = ST.accept_buff[ best ];
3187 ST.accept_buff[ best ] = ST.accept_buff[ ST.accepted ];
3188 ST.accept_buff[ ST.accepted ] = tmp;
3189 best = ST.accepted;
a3621e74 3190 }
166ba7cd 3191 PL_reginput = (char *)ST.accept_buff[ best ].endpos;
7f69552c 3192 if ( !ST.jump || !ST.jump[ST.accept_buff[best].wordnum]) {
5d458dd8 3193 scan = ST.B;
786e8c11 3194 } else {
5d458dd8 3195 scan = ST.me + ST.jump[ST.accept_buff[best].wordnum];
5d458dd8 3196 }
6b173516 3197 PUSH_YES_STATE_GOTO(TRIE_next, scan);
786e8c11 3198 /* NOTREACHED */
166ba7cd 3199 }
166ba7cd 3200 /* NOTREACHED */
5d458dd8
YO
3201 case TRIE_next:
3202 FREETMPS;
3203 LEAVE;
3204 sayYES;
166ba7cd
DM
3205#undef ST
3206
95b24440
DM
3207 case EXACT: {
3208 char *s = STRING(scan);
24d3c4a9 3209 ln = STR_LEN(scan);
eb160463 3210 if (do_utf8 != UTF) {
bc517b45 3211 /* The target and the pattern have differing utf8ness. */
1aa99e6b 3212 char *l = locinput;
24d3c4a9 3213 const char * const e = s + ln;
a72c7584 3214
5ff6fc6d
JH
3215 if (do_utf8) {
3216 /* The target is utf8, the pattern is not utf8. */
1aa99e6b 3217 while (s < e) {
a3b680e6 3218 STRLEN ulen;
1aa99e6b 3219 if (l >= PL_regeol)
5ff6fc6d
JH
3220 sayNO;
3221 if (NATIVE_TO_UNI(*(U8*)s) !=
89ebb4a3 3222 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
041457d9 3223 uniflags))
5ff6fc6d 3224 sayNO;
bc517b45 3225 l += ulen;
5ff6fc6d 3226 s ++;
1aa99e6b 3227 }
5ff6fc6d
JH
3228 }
3229 else {
3230 /* The target is not utf8, the pattern is utf8. */
1aa99e6b 3231 while (s < e) {
a3b680e6 3232 STRLEN ulen;
1aa99e6b
IH
3233 if (l >= PL_regeol)
3234 sayNO;
5ff6fc6d 3235 if (NATIVE_TO_UNI(*((U8*)l)) !=
89ebb4a3 3236 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
041457d9 3237 uniflags))
1aa99e6b 3238 sayNO;
bc517b45 3239 s += ulen;
a72c7584 3240 l ++;
1aa99e6b 3241 }
5ff6fc6d 3242 }
1aa99e6b
IH
3243 locinput = l;
3244 nextchr = UCHARAT(locinput);
3245 break;
3246 }
bc517b45 3247 /* The target and the pattern have the same utf8ness. */
d6a28714
JH
3248 /* Inline the first character, for speed. */
3249 if (UCHARAT(s) != nextchr)
3250 sayNO;
24d3c4a9 3251 if (PL_regeol - locinput < ln)
d6a28714 3252 sayNO;
24d3c4a9 3253 if (ln > 1 && memNE(s, locinput, ln))
d6a28714 3254 sayNO;
24d3c4a9 3255 locinput += ln;
d6a28714
JH
3256 nextchr = UCHARAT(locinput);
3257 break;
95b24440 3258 }
d6a28714 3259 case EXACTFL:
b8c5462f
JH
3260 PL_reg_flags |= RF_tainted;
3261 /* FALL THROUGH */
95b24440 3262 case EXACTF: {
be8e71aa 3263 char * const s = STRING(scan);
24d3c4a9 3264 ln = STR_LEN(scan);
d6a28714 3265
d07ddd77
JH
3266 if (do_utf8 || UTF) {
3267 /* Either target or the pattern are utf8. */
be8e71aa 3268 const char * const l = locinput;
d07ddd77 3269 char *e = PL_regeol;
bc517b45 3270
24d3c4a9 3271 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
1feea2c7 3272 l, &e, 0, do_utf8)) {
5486206c
JH
3273 /* One more case for the sharp s:
3274 * pack("U0U*", 0xDF) =~ /ss/i,
3275 * the 0xC3 0x9F are the UTF-8
3276 * byte sequence for the U+00DF. */
e1d1eefb 3277
5486206c 3278 if (!(do_utf8 &&
e1d1eefb 3279 toLOWER(s[0]) == 's' &&
24d3c4a9 3280 ln >= 2 &&
5486206c
JH
3281 toLOWER(s[1]) == 's' &&
3282 (U8)l[0] == 0xC3 &&
3283 e - l >= 2 &&
3284 (U8)l[1] == 0x9F))
3285 sayNO;
3286 }
d07ddd77
JH
3287 locinput = e;
3288 nextchr = UCHARAT(locinput);
3289 break;
a0ed51b3 3290 }
d6a28714 3291
bc517b45
JH
3292 /* Neither the target and the pattern are utf8. */
3293
d6a28714
JH
3294 /* Inline the first character, for speed. */
3295 if (UCHARAT(s) != nextchr &&
3296 UCHARAT(s) != ((OP(scan) == EXACTF)
3297 ? PL_fold : PL_fold_locale)[nextchr])
a0ed51b3 3298 sayNO;
24d3c4a9 3299 if (PL_regeol - locinput < ln)
b8c5462f 3300 sayNO;
24d3c4a9
DM
3301 if (ln > 1 && (OP(scan) == EXACTF
3302 ? ibcmp(s, locinput, ln)
3303 : ibcmp_locale(s, locinput, ln)))
4633a7c4 3304 sayNO;
24d3c4a9 3305 locinput += ln;
d6a28714 3306 nextchr = UCHARAT(locinput);
a0d0e21e 3307 break;
95b24440 3308 }
d6a28714 3309 case ANYOF:
ffc61ed2 3310 if (do_utf8) {
9e55ce06
JH
3311 STRLEN inclasslen = PL_regeol - locinput;
3312
32fc9b6a 3313 if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8))
262b90c4 3314 goto anyof_fail;
ffc61ed2
JH
3315 if (locinput >= PL_regeol)
3316 sayNO;
0f0076b4 3317 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
b8c5462f 3318 nextchr = UCHARAT(locinput);
e0f9d4a8 3319 break;
ffc61ed2
JH
3320 }
3321 else {
3322 if (nextchr < 0)
3323 nextchr = UCHARAT(locinput);
32fc9b6a 3324 if (!REGINCLASS(rex, scan, (U8*)locinput))
262b90c4 3325 goto anyof_fail;
ffc61ed2
JH
3326 if (!nextchr && locinput >= PL_regeol)
3327 sayNO;
3328 nextchr = UCHARAT(++locinput);
e0f9d4a8
JH
3329 break;
3330 }
262b90c4 3331 anyof_fail:
e0f9d4a8
JH
3332 /* If we might have the case of the German sharp s
3333 * in a casefolding Unicode character class. */
3334
ebc501f0
JH
3335 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
3336 locinput += SHARP_S_SKIP;
e0f9d4a8 3337 nextchr = UCHARAT(locinput);
ffc61ed2 3338 }
e0f9d4a8
JH
3339 else
3340 sayNO;
b8c5462f 3341 break;
d6a28714 3342 case ALNUML:
b8c5462f
JH
3343 PL_reg_flags |= RF_tainted;
3344 /* FALL THROUGH */
d6a28714 3345 case ALNUM:
b8c5462f 3346 if (!nextchr)
4633a7c4 3347 sayNO;
ffc61ed2 3348 if (do_utf8) {
1a4fad37 3349 LOAD_UTF8_CHARCLASS_ALNUM();
ffc61ed2 3350 if (!(OP(scan) == ALNUM
bb7a0f54 3351 ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
d6a28714 3352 : isALNUM_LC_utf8((U8*)locinput)))
b8c5462f
JH
3353 {
3354 sayNO;
a0ed51b3 3355 }
b8c5462f 3356 locinput += PL_utf8skip[nextchr];
a0ed51b3
LW
3357 nextchr = UCHARAT(locinput);
3358 break;
3359 }
ffc61ed2 3360 if (!(OP(scan) == ALNUM
d6a28714 3361 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
bbce6d69 3362 sayNO;
b8c5462f 3363 nextchr = UCHARAT(++locinput);
a0d0e21e 3364 break;
d6a28714 3365 case NALNUML:
b8c5462f
JH
3366 PL_reg_flags |= RF_tainted;
3367 /* FALL THROUGH */
d6a28714
JH
3368 case NALNUM:
3369 if (!nextchr && locinput >= PL_regeol)
a0ed51b3 3370 sayNO;
ffc61ed2 3371 if (do_utf8) {
1a4fad37 3372 LOAD_UTF8_CHARCLASS_ALNUM();
ffc61ed2 3373 if (OP(scan) == NALNUM
bb7a0f54 3374 ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
d6a28714
JH
3375 : isALNUM_LC_utf8((U8*)locinput))
3376 {
b8c5462f 3377 sayNO;
d6a28714 3378 }
b8c5462f
JH
3379 locinput += PL_utf8skip[nextchr];
3380 nextchr = UCHARAT(locinput);
3381 break;
3382 }
ffc61ed2 3383 if (OP(scan) == NALNUM
d6a28714 3384 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
b8c5462f 3385 sayNO;
76e3520e 3386 nextchr = UCHARAT(++locinput);
a0d0e21e 3387 break;
d6a28714
JH
3388 case BOUNDL:
3389 case NBOUNDL:
3280af22 3390 PL_reg_flags |= RF_tainted;
bbce6d69 3391 /* FALL THROUGH */
d6a28714
JH
3392 case BOUND:
3393 case NBOUND:
3394 /* was last char in word? */
ffc61ed2 3395 if (do_utf8) {
12d33761 3396 if (locinput == PL_bostr)
24d3c4a9 3397 ln = '\n';
ffc61ed2 3398 else {
a3b680e6 3399 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
9041c2e3 3400
24d3c4a9 3401 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
ffc61ed2
JH
3402 }
3403 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
24d3c4a9 3404 ln = isALNUM_uni(ln);
1a4fad37 3405 LOAD_UTF8_CHARCLASS_ALNUM();
3568d838 3406 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
ffc61ed2
JH
3407 }
3408 else {
24d3c4a9 3409 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
ffc61ed2
JH
3410 n = isALNUM_LC_utf8((U8*)locinput);
3411 }
a0ed51b3 3412 }
d6a28714 3413 else {
24d3c4a9 3414 ln = (locinput != PL_bostr) ?
12d33761 3415 UCHARAT(locinput - 1) : '\n';
ffc61ed2 3416 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
24d3c4a9 3417 ln = isALNUM(ln);
ffc61ed2
JH
3418 n = isALNUM(nextchr);
3419 }
3420 else {
24d3c4a9 3421 ln = isALNUM_LC(ln);
ffc61ed2
JH
3422 n = isALNUM_LC(nextchr);
3423 }
d6a28714 3424 }
24d3c4a9 3425 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
ffc61ed2
JH
3426 OP(scan) == BOUNDL))
3427 sayNO;
a0ed51b3 3428 break;
d6a28714 3429 case SPACEL:
3280af22 3430 PL_reg_flags |= RF_tainted;
bbce6d69 3431 /* FALL THROUGH */
d6a28714 3432 case SPACE:
9442cb0e 3433 if (!nextchr)
4633a7c4 3434 sayNO;
1aa99e6b 3435 if (do_utf8) {
fd400ab9 3436 if (UTF8_IS_CONTINUED(nextchr)) {
1a4fad37 3437 LOAD_UTF8_CHARCLASS_SPACE();
ffc61ed2 3438 if (!(OP(scan) == SPACE
bb7a0f54 3439 ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
ffc61ed2
JH
3440 : isSPACE_LC_utf8((U8*)locinput)))
3441 {
3442 sayNO;
3443 }
3444 locinput += PL_utf8skip[nextchr];
3445 nextchr = UCHARAT(locinput);
3446 break;
d6a28714 3447 }
ffc61ed2
JH
3448 if (!(OP(scan) == SPACE
3449 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3450 sayNO;
3451 nextchr = UCHARAT(++locinput);
3452 }
3453 else {
3454 if (!(OP(scan) == SPACE
3455 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3456 sayNO;
3457 nextchr = UCHARAT(++locinput);
a0ed51b3 3458 }
a0ed51b3 3459 break;
d6a28714 3460 case NSPACEL:
3280af22 3461 PL_reg_flags |= RF_tainted;
bbce6d69 3462 /* FALL THROUGH */
d6a28714 3463 case NSPACE:
9442cb0e 3464 if (!nextchr && locinput >= PL_regeol)
b8c5462f 3465 sayNO;
1aa99e6b 3466 if (do_utf8) {
1a4fad37 3467 LOAD_UTF8_CHARCLASS_SPACE();
ffc61ed2 3468 if (OP(scan) == NSPACE
bb7a0f54 3469 ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
d6a28714 3470 : isSPACE_LC_utf8((U8*)locinput))
b8c5462f
JH
3471 {
3472 sayNO;
3473 }
3474 locinput += PL_utf8skip[nextchr];
3475 nextchr = UCHARAT(locinput);
3476 break;
a0ed51b3 3477 }
ffc61ed2 3478 if (OP(scan) == NSPACE
d6a28714 3479 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
4633a7c4 3480 sayNO;
76e3520e 3481 nextchr = UCHARAT(++locinput);
a0d0e21e 3482 break;
d6a28714 3483 case DIGITL:
a0ed51b3
LW
3484 PL_reg_flags |= RF_tainted;
3485 /* FALL THROUGH */
d6a28714 3486 case DIGIT:
9442cb0e 3487 if (!nextchr)
a0ed51b3 3488 sayNO;
1aa99e6b 3489 if (do_utf8) {
1a4fad37 3490 LOAD_UTF8_CHARCLASS_DIGIT();
ffc61ed2 3491 if (!(OP(scan) == DIGIT
bb7a0f54 3492 ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
9442cb0e 3493 : isDIGIT_LC_utf8((U8*)locinput)))
dfe13c55 3494 {
a0ed51b3 3495 sayNO;
dfe13c55 3496 }
6f06b55f 3497 locinput += PL_utf8skip[nextchr];
a0ed51b3
LW
3498 nextchr = UCHARAT(locinput);
3499 break;
3500 }
ffc61ed2 3501 if (!(OP(scan) == DIGIT
9442cb0e 3502 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
a0ed51b3
LW
3503 sayNO;
3504 nextchr = UCHARAT(++locinput);
3505 break;
d6a28714 3506 case NDIGITL:
b8c5462f
JH
3507 PL_reg_flags |= RF_tainted;
3508 /* FALL THROUGH */
d6a28714 3509 case NDIGIT:
9442cb0e 3510 if (!nextchr && locinput >= PL_regeol)
b8c5462f 3511 sayNO;
1aa99e6b 3512 if (do_utf8) {
1a4fad37 3513 LOAD_UTF8_CHARCLASS_DIGIT();
ffc61ed2 3514 if (OP(scan) == NDIGIT
bb7a0f54 3515 ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
9442cb0e
GS
3516 : isDIGIT_LC_utf8((U8*)locinput))
3517 {
a0ed51b3 3518 sayNO;
9442cb0e 3519 }
6f06b55f 3520 locinput += PL_utf8skip[nextchr];
a0ed51b3
LW
3521 nextchr = UCHARAT(locinput);
3522 break;
3523 }
ffc61ed2 3524 if (OP(scan) == NDIGIT
9442cb0e 3525 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
a0ed51b3
LW
3526 sayNO;
3527 nextchr = UCHARAT(++locinput);
3528 break;
3529 case CLUMP:
b7c83a7e 3530 if (locinput >= PL_regeol)
a0ed51b3 3531 sayNO;
b7c83a7e 3532 if (do_utf8) {
1a4fad37 3533 LOAD_UTF8_CHARCLASS_MARK();
b7c83a7e
JH
3534 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3535 sayNO;
3536 locinput += PL_utf8skip[nextchr];
3537 while (locinput < PL_regeol &&
3538 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3539 locinput += UTF8SKIP(locinput);
3540 if (locinput > PL_regeol)
3541 sayNO;
eb08e2da
JH
3542 }
3543 else
3544 locinput++;
a0ed51b3
LW
3545 nextchr = UCHARAT(locinput);
3546 break;
81714fb9
YO
3547
3548 case NREFFL:
3549 {
3550 char *s;
ff1157ca 3551 char type;
81714fb9
YO
3552 PL_reg_flags |= RF_tainted;
3553 /* FALL THROUGH */
3554 case NREF:
3555 case NREFF:
ff1157ca 3556 type = OP(scan);
0a4db386
YO
3557 n = reg_check_named_buff_matched(rex,scan);
3558
3559 if ( n ) {
3560 type = REF + ( type - NREF );
3561 goto do_ref;
3562 } else {
81714fb9 3563 sayNO;
0a4db386
YO
3564 }
3565 /* unreached */
c8756f30 3566 case REFFL:
3280af22 3567 PL_reg_flags |= RF_tainted;
c8756f30 3568 /* FALL THROUGH */
c277df42 3569 case REF:
81714fb9 3570 case REFF:
c277df42 3571 n = ARG(scan); /* which paren pair */
81714fb9
YO
3572 type = OP(scan);
3573 do_ref:
f0ab9afb 3574 ln = PL_regoffs[n].start;
2c2d71f5 3575 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3b6647e0 3576 if (*PL_reglastparen < n || ln == -1)
af3f8c16 3577 sayNO; /* Do not match unless seen CLOSEn. */
f0ab9afb 3578 if (ln == PL_regoffs[n].end)
a0d0e21e 3579 break;
a0ed51b3 3580
24d3c4a9 3581 s = PL_bostr + ln;
81714fb9 3582 if (do_utf8 && type != REF) { /* REF can do byte comparison */
a0ed51b3 3583 char *l = locinput;
f0ab9afb 3584 const char *e = PL_bostr + PL_regoffs[n].end;
a0ed51b3
LW
3585 /*
3586 * Note that we can't do the "other character" lookup trick as
3587 * in the 8-bit case (no pun intended) because in Unicode we
3588 * have to map both upper and title case to lower case.
3589 */
81714fb9 3590 if (type == REFF) {
a0ed51b3 3591 while (s < e) {
a3b680e6
AL
3592 STRLEN ulen1, ulen2;
3593 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3594 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3595
a0ed51b3
LW
3596 if (l >= PL_regeol)
3597 sayNO;
a2a2844f
JH
3598 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3599 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
7114a2d2 3600 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
a0ed51b3 3601 sayNO;
a2a2844f
JH
3602 s += ulen1;
3603 l += ulen2;
a0ed51b3
LW
3604 }
3605 }
3606 locinput = l;
3607 nextchr = UCHARAT(locinput);
3608 break;
3609 }
3610
a0d0e21e 3611 /* Inline the first character, for speed. */
76e3520e 3612 if (UCHARAT(s) != nextchr &&
81714fb9
YO
3613 (type == REF ||
3614 (UCHARAT(s) != (type == REFF
3615 ? PL_fold : PL_fold_locale)[nextchr])))
4633a7c4 3616 sayNO;
f0ab9afb 3617 ln = PL_regoffs[n].end - ln;
24d3c4a9 3618 if (locinput + ln > PL_regeol)
4633a7c4 3619 sayNO;
81714fb9 3620 if (ln > 1 && (type == REF
24d3c4a9 3621 ? memNE(s, locinput, ln)
81714fb9 3622 : (type == REFF
24d3c4a9
DM
3623 ? ibcmp(s, locinput, ln)
3624 : ibcmp_locale(s, locinput, ln))))
4633a7c4 3625 sayNO;
24d3c4a9 3626 locinput += ln;
76e3520e 3627 nextchr = UCHARAT(locinput);
a0d0e21e 3628 break;
81714fb9 3629 }
a0d0e21e 3630 case NOTHING:
c277df42 3631 case TAIL:
a0d0e21e
LW
3632 break;
3633 case BACK:
3634 break;
40a82448
DM
3635
3636#undef ST
3637#define ST st->u.eval
c277df42 3638 {
c277df42 3639 SV *ret;
d2f13c59 3640 REGEXP *re_sv;
6bda09f9 3641 regexp *re;
f8fc2ecf 3642 regexp_internal *rei;
1a147d38
YO
3643 regnode *startpoint;
3644
3645 case GOSTART:
e7707071
YO
3646 case GOSUB: /* /(...(?1))/ /(...(?&foo))/ */
3647 if (cur_eval && cur_eval->locinput==locinput) {
24b23f37 3648 if (cur_eval->u.eval.close_paren == (U32)ARG(scan))
1a147d38 3649 Perl_croak(aTHX_ "Infinite recursion in regex");
4b196cd4 3650 if ( ++nochange_depth > max_nochange_depth )
1a147d38
YO
3651 Perl_croak(aTHX_
3652 "Pattern subroutine nesting without pos change"
3653 " exceeded limit in regex");
6bda09f9
YO
3654 } else {
3655 nochange_depth = 0;
1a147d38 3656 }
288b8c02 3657 re_sv = rex_sv;
6bda09f9 3658 re = rex;
f8fc2ecf 3659 rei = rexi;
288b8c02 3660 (void)ReREFCNT_inc(rex_sv);
1a147d38 3661 if (OP(scan)==GOSUB) {
6bda09f9
YO
3662 startpoint = scan + ARG2L(scan);
3663 ST.close_paren = ARG(scan);
3664 } else {
f8fc2ecf 3665 startpoint = rei->program+1;
6bda09f9
YO
3666 ST.close_paren = 0;
3667 }
3668 goto eval_recurse_doit;
3669 /* NOTREACHED */
3670 case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */
3671 if (cur_eval && cur_eval->locinput==locinput) {
4b196cd4 3672 if ( ++nochange_depth > max_nochange_depth )
1a147d38 3673 Perl_croak(aTHX_ "EVAL without pos change exceeded limit in regex");
6bda09f9
YO
3674 } else {
3675 nochange_depth = 0;
3676 }
8e5e9ebe 3677 {
4aabdb9b
DM
3678 /* execute the code in the {...} */
3679 dSP;
6136c704 3680 SV ** const before = SP;
4aabdb9b
DM
3681 OP_4tree * const oop = PL_op;
3682 COP * const ocurcop = PL_curcop;
3683 PAD *old_comppad;
4aabdb9b
DM
3684
3685 n = ARG(scan);
f8fc2ecf 3686 PL_op = (OP_4tree*)rexi->data->data[n];
24b23f37
YO
3687 DEBUG_STATE_r( PerlIO_printf(Perl_debug_log,
3688 " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
f8fc2ecf 3689 PAD_SAVE_LOCAL(old_comppad, (PAD*)rexi->data->data[n + 2]);
f0ab9afb 3690 PL_regoffs[0].end = PL_reg_magic->mg_len = locinput - PL_bostr;
4aabdb9b 3691
2bf803e2
YO
3692 if (sv_yes_mark) {
3693 SV *sv_mrk = get_sv("REGMARK", 1);
3694 sv_setsv(sv_mrk, sv_yes_mark);
3695 }
3696
8e5e9ebe
RGS
3697 CALLRUNOPS(aTHX); /* Scalar context. */
3698 SPAGAIN;
3699 if (SP == before)
075aa684 3700 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
8e5e9ebe
RGS
3701 else {
3702 ret = POPs;
3703 PUTBACK;
3704 }
4aabdb9b
DM
3705
3706 PL_op = oop;
3707 PAD_RESTORE_LOCAL(old_comppad);
3708 PL_curcop = ocurcop;
24d3c4a9 3709 if (!logical) {
4aabdb9b
DM
3710 /* /(?{...})/ */
3711 sv_setsv(save_scalar(PL_replgv), ret);
4aabdb9b
DM
3712 break;
3713 }
8e5e9ebe 3714 }
24d3c4a9
DM
3715 if (logical == 2) { /* Postponed subexpression: /(??{...})/ */
3716 logical = 0;
4aabdb9b 3717 {
4f639d21
DM
3718 /* extract RE object from returned value; compiling if
3719 * necessary */
6136c704 3720 MAGIC *mg = NULL;
288b8c02 3721 REGEXP *rx = NULL;
5c35adbb
NC
3722
3723 if (SvROK(ret)) {
288b8c02 3724 SV *const sv = SvRV(ret);
5c35adbb
NC
3725
3726 if (SvTYPE(sv) == SVt_REGEXP) {
d2f13c59 3727 rx = (REGEXP*) sv;
5c35adbb
NC
3728 } else if (SvSMAGICAL(sv)) {
3729 mg = mg_find(sv, PERL_MAGIC_qr);
3730 assert(mg);
3731 }
3732 } else if (SvTYPE(ret) == SVt_REGEXP) {
d2f13c59 3733 rx = (REGEXP*) ret;
5c35adbb 3734 } else if (SvSMAGICAL(ret)) {
124ee91a
NC
3735 if (SvGMAGICAL(ret)) {
3736 /* I don't believe that there is ever qr magic
3737 here. */
3738 assert(!mg_find(ret, PERL_MAGIC_qr));
faf82a0b 3739 sv_unmagic(ret, PERL_MAGIC_qr);
124ee91a
NC
3740 }
3741 else {
faf82a0b 3742 mg = mg_find(ret, PERL_MAGIC_qr);
124ee91a
NC
3743 /* testing suggests mg only ends up non-NULL for
3744 scalars who were upgraded and compiled in the
3745 else block below. In turn, this is only
3746 triggered in the "postponed utf8 string" tests
3747 in t/op/pat.t */
3748 }
0f5d15d6 3749 }
faf82a0b 3750
0f5d15d6 3751 if (mg) {
d2f13c59 3752 rx = (REGEXP *) mg->mg_obj; /*XXX:dmq*/
e2560c33 3753 assert(rx);
0f5d15d6 3754 }
288b8c02
NC
3755 if (rx) {
3756 rx = reg_temp_copy(rx);
3757 }
0f5d15d6 3758 else {
c737faaf 3759 U32 pm_flags = 0;
a3b680e6 3760 const I32 osize = PL_regsize;
0f5d15d6 3761
b9ad30b4
NC
3762 if (DO_UTF8(ret)) {
3763 assert (SvUTF8(ret));
3764 } else if (SvUTF8(ret)) {
3765 /* Not doing UTF-8, despite what the SV says. Is
3766 this only if we're trapped in use 'bytes'? */
3767 /* Make a copy of the octet sequence, but without
3768 the flag on, as the compiler now honours the
3769 SvUTF8 flag on ret. */
3770 STRLEN len;
3771 const char *const p = SvPV(ret, len);
3772 ret = newSVpvn_flags(p, len, SVs_TEMP);
3773 }
288b8c02 3774 rx = CALLREGCOMP(ret, pm_flags);
9041c2e3 3775 if (!(SvFLAGS(ret)
faf82a0b 3776 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
3ce3ed55 3777 | SVs_GMG))) {
a2794585
NC
3778 /* This isn't a first class regexp. Instead, it's
3779 caching a regexp onto an existing, Perl visible
3780 scalar. */
d2f13c59 3781 sv_magic(ret, (SV*) rx, PERL_MAGIC_qr, 0, 0);
3ce3ed55 3782 }
0f5d15d6 3783 PL_regsize = osize;
0f5d15d6 3784 }
288b8c02
NC
3785 re_sv = rx;
3786 re = (struct regexp *)SvANY(rx);
4aabdb9b 3787 }
07bc277f 3788 RXp_MATCH_COPIED_off(re);
28d8d7f4
YO
3789 re->subbeg = rex->subbeg;
3790 re->sublen = rex->sublen;
f8fc2ecf 3791 rei = RXi_GET(re);
6bda09f9 3792 DEBUG_EXECUTE_r(
efd26800 3793 debug_start_match(re_sv, do_utf8, locinput, PL_regeol,
6bda09f9
YO
3794 "Matching embedded");
3795 );
f8fc2ecf 3796 startpoint = rei->program + 1;
1a147d38 3797 ST.close_paren = 0; /* only used for GOSUB */
6bda09f9
YO
3798 /* borrowed from regtry */
3799 if (PL_reg_start_tmpl <= re->nparens) {
3800 PL_reg_start_tmpl = re->nparens*3/2 + 3;
3801 if(PL_reg_start_tmp)
3802 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
3803 else
3804 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
dd5def09 3805 }
aa283a38 3806
1a147d38 3807 eval_recurse_doit: /* Share code with GOSUB below this line */
aa283a38 3808 /* run the pattern returned from (??{...}) */
40a82448
DM
3809 ST.cp = regcppush(0); /* Save *all* the positions. */
3810 REGCP_SET(ST.lastcp);
6bda09f9 3811
f0ab9afb 3812 PL_regoffs = re->offs; /* essentially NOOP on GOSUB */
6bda09f9 3813
0357f1fd
ML
3814 /* see regtry, specifically PL_reglast(?:close)?paren is a pointer! (i dont know why) :dmq */
3815 PL_reglastparen = &re->lastparen;
3816 PL_reglastcloseparen = &re->lastcloseparen;
3817 re->lastparen = 0;
3818 re->lastcloseparen = 0;
3819
4aabdb9b 3820 PL_reginput = locinput;
ae0beba1 3821 PL_regsize = 0;
4aabdb9b
DM
3822
3823 /* XXXX This is too dramatic a measure... */
3824 PL_reg_maxiter = 0;
3825
faec1544 3826 ST.toggle_reg_flags = PL_reg_flags;
3c8556c3 3827 if (RX_UTF8(re_sv))
faec1544
DM
3828 PL_reg_flags |= RF_utf8;
3829 else
3830 PL_reg_flags &= ~RF_utf8;
3831 ST.toggle_reg_flags ^= PL_reg_flags; /* diff of old and new */
3832
288b8c02 3833 ST.prev_rex = rex_sv;
faec1544 3834 ST.prev_curlyx = cur_curlyx;
288b8c02
NC
3835 SETREX(rex_sv,re_sv);
3836 rex = re;
f8fc2ecf 3837 rexi = rei;
faec1544 3838 cur_curlyx = NULL;
40a82448 3839 ST.B = next;
faec1544
DM
3840 ST.prev_eval = cur_eval;
3841 cur_eval = st;
faec1544 3842 /* now continue from first node in postoned RE */
6bda09f9 3843 PUSH_YES_STATE_GOTO(EVAL_AB, startpoint);
4aabdb9b 3844 /* NOTREACHED */
a0ed51b3 3845 }
24d3c4a9
DM
3846 /* logical is 1, /(?(?{...})X|Y)/ */
3847 sw = (bool)SvTRUE(ret);
3848 logical = 0;
c277df42
IZ
3849 break;
3850 }
40a82448 3851
faec1544
DM
3852 case EVAL_AB: /* cleanup after a successful (??{A})B */
3853 /* note: this is called twice; first after popping B, then A */
3854 PL_reg_flags ^= ST.toggle_reg_flags;
288b8c02
NC
3855 ReREFCNT_dec(rex_sv);
3856 SETREX(rex_sv,ST.prev_rex);
3857 rex = (struct regexp *)SvANY(rex_sv);
f8fc2ecf 3858 rexi = RXi_GET(rex);
faec1544
DM
3859 regcpblow(ST.cp);
3860 cur_eval = ST.prev_eval;
3861 cur_curlyx = ST.prev_curlyx;
0357f1fd
ML
3862
3863 PL_reglastparen = &rex->lastparen;
3864 PL_reglastcloseparen = &rex->lastcloseparen;
3865
40a82448
DM
3866 /* XXXX This is too dramatic a measure... */
3867 PL_reg_maxiter = 0;
e7707071 3868 if ( nochange_depth )
4b196cd4 3869 nochange_depth--;
262b90c4 3870 sayYES;
40a82448 3871
40a82448 3872
faec1544
DM
3873 case EVAL_AB_fail: /* unsuccessfully ran A or B in (??{A})B */
3874 /* note: this is called twice; first after popping B, then A */
3875 PL_reg_flags ^= ST.toggle_reg_flags;
288b8c02
NC
3876 ReREFCNT_dec(rex_sv);
3877 SETREX(rex_sv,ST.prev_rex);
3878 rex = (struct regexp *)SvANY(rex_sv);
f8fc2ecf 3879 rexi = RXi_GET(rex);
0357f1fd
ML
3880 PL_reglastparen = &rex->lastparen;
3881 PL_reglastcloseparen = &rex->lastcloseparen;
3882
40a82448
DM
3883 PL_reginput = locinput;
3884 REGCP_UNWIND(ST.lastcp);
3885 regcppop(rex);
faec1544
DM
3886 cur_eval = ST.prev_eval;
3887 cur_curlyx = ST.prev_curlyx;
3888 /* XXXX This is too dramatic a measure... */
3889 PL_reg_maxiter = 0;
e7707071 3890 if ( nochange_depth )
4b196cd4 3891 nochange_depth--;
40a82448 3892 sayNO_SILENT;
40a82448
DM
3893#undef ST
3894
a0d0e21e 3895 case OPEN:
c277df42 3896 n = ARG(scan); /* which paren pair */
3280af22
NIS
3897 PL_reg_start_tmp[n] = locinput;
3898 if (n > PL_regsize)
3899 PL_regsize = n;
e2e6a0f1 3900 lastopen = n;
a0d0e21e
LW
3901 break;
3902 case CLOSE:
c277df42 3903 n = ARG(scan); /* which paren pair */
f0ab9afb
NC
3904 PL_regoffs[n].start = PL_reg_start_tmp[n] - PL_bostr;
3905 PL_regoffs[n].end = locinput - PL_bostr;
7f69552c
YO
3906 /*if (n > PL_regsize)
3907 PL_regsize = n;*/
3b6647e0 3908 if (n > *PL_reglastparen)
3280af22 3909 *PL_reglastparen = n;
a01268b5 3910 *PL_reglastcloseparen = n;
3b6647e0 3911 if (cur_eval && cur_eval->u.eval.close_paren == n) {
6bda09f9
YO
3912 goto fake_end;
3913 }
a0d0e21e 3914 break;
e2e6a0f1
YO
3915 case ACCEPT:
3916 if (ARG(scan)){
3917 regnode *cursor;
3918 for (cursor=scan;
3919 cursor && OP(cursor)!=END;
3920 cursor=regnext(cursor))
3921 {
3922 if ( OP(cursor)==CLOSE ){
3923 n = ARG(cursor);
3924 if ( n <= lastopen ) {
f0ab9afb
NC
3925 PL_regoffs[n].start
3926 = PL_reg_start_tmp[n] - PL_bostr;
3927 PL_regoffs[n].end = locinput - PL_bostr;
e2e6a0f1
YO
3928 /*if (n > PL_regsize)
3929 PL_regsize = n;*/
3b6647e0 3930 if (n > *PL_reglastparen)
e2e6a0f1
YO
3931 *PL_reglastparen = n;
3932 *PL_reglastcloseparen = n;
3b6647e0
RB
3933 if ( n == ARG(scan) || (cur_eval &&
3934 cur_eval->u.eval.close_paren == n))
e2e6a0f1
YO
3935 break;
3936 }
3937 }
3938 }
3939 }
3940 goto fake_end;
3941 /*NOTREACHED*/
c277df42
IZ
3942 case GROUPP:
3943 n = ARG(scan); /* which paren pair */
f0ab9afb 3944 sw = (bool)(*PL_reglastparen >= n && PL_regoffs[n].end != -1);
c277df42 3945 break;
0a4db386
YO
3946 case NGROUPP:
3947 /* reg_check_named_buff_matched returns 0 for no match */
3948 sw = (bool)(0 < reg_check_named_buff_matched(rex,scan));
3949 break;
1a147d38 3950 case INSUBP:
0a4db386 3951 n = ARG(scan);
3b6647e0 3952 sw = (cur_eval && (!n || cur_eval->u.eval.close_paren == n));
0a4db386
YO
3953 break;
3954 case DEFINEP:
3955 sw = 0;
3956 break;
c277df42 3957 case IFTHEN:
2c2d71f5 3958 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
24d3c4a9 3959 if (sw)
c277df42
IZ
3960 next = NEXTOPER(NEXTOPER(scan));
3961 else {
3962 next = scan + ARG(scan);
3963 if (OP(next) == IFTHEN) /* Fake one. */
3964 next = NEXTOPER(NEXTOPER(next));
3965 }
3966 break;
3967 case LOGICAL:
24d3c4a9 3968 logical = scan->flags;
c277df42 3969 break;
c476f425 3970
2ab05381 3971/*******************************************************************
2ab05381 3972
c476f425
DM
3973The CURLYX/WHILEM pair of ops handle the most generic case of the /A*B/
3974pattern, where A and B are subpatterns. (For simple A, CURLYM or
3975STAR/PLUS/CURLY/CURLYN are used instead.)
2ab05381 3976
c476f425 3977A*B is compiled as <CURLYX><A><WHILEM><B>
2ab05381 3978
c476f425
DM
3979On entry to the subpattern, CURLYX is called. This pushes a CURLYX
3980state, which contains the current count, initialised to -1. It also sets
3981cur_curlyx to point to this state, with any previous value saved in the
3982state block.
2ab05381 3983
c476f425
DM
3984CURLYX then jumps straight to the WHILEM op, rather than executing A,
3985since the pattern may possibly match zero times (i.e. it's a while {} loop
3986rather than a do {} while loop).
2ab05381 3987
c476f425
DM
3988Each entry to WHILEM represents a successful match of A. The count in the
3989CURLYX block is incremented, another WHILEM state is pushed, and execution
3990passes to A or B depending on greediness and the current count.
2ab05381 3991
c476f425
DM
3992For example, if matching against the string a1a2a3b (where the aN are
3993substrings that match /A/), then the match progresses as follows: (the
3994pushed states are interspersed with the bits of strings matched so far):
2ab05381 3995
c476f425
DM
3996 <CURLYX cnt=-1>
3997 <CURLYX cnt=0><WHILEM>
3998 <CURLYX cnt=1><WHILEM> a1 <WHILEM>
3999 <CURLYX cnt=2><WHILEM> a1 <WHILEM> a2 <WHILEM>
4000 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM>
4001 <CURLYX cnt=3><WHILEM> a1 <WHILEM> a2 <WHILEM> a3 <WHILEM> b
2ab05381 4002
c476f425
DM
4003(Contrast this with something like CURLYM, which maintains only a single
4004backtrack state:
2ab05381 4005
c476f425
DM
4006 <CURLYM cnt=0> a1
4007 a1 <CURLYM cnt=1> a2
4008 a1 a2 <CURLYM cnt=2> a3
4009 a1 a2 a3 <CURLYM cnt=3> b
4010)
2ab05381 4011
c476f425
DM
4012Each WHILEM state block marks a point to backtrack to upon partial failure
4013of A or B, and also contains some minor state data related to that
4014iteration. The CURLYX block, pointed to by cur_curlyx, contains the
4015overall state, such as the count, and pointers to the A and B ops.
2ab05381 4016
c476f425
DM
4017This is complicated slightly by nested CURLYX/WHILEM's. Since cur_curlyx
4018must always point to the *current* CURLYX block, the rules are:
2ab05381 4019
c476f425
DM
4020When executing CURLYX, save the old cur_curlyx in the CURLYX state block,
4021and set cur_curlyx to point the new block.
2ab05381 4022
c476f425
DM
4023When popping the CURLYX block after a successful or unsuccessful match,
4024restore the previous cur_curlyx.
2ab05381 4025
c476f425
DM
4026When WHILEM is about to execute B, save the current cur_curlyx, and set it
4027to the outer one saved in the CURLYX block.
2ab05381 4028
c476f425
DM
4029When popping the WHILEM block after a successful or unsuccessful B match,
4030restore the previous cur_curlyx.
2ab05381 4031
c476f425
DM
4032Here's an example for the pattern (AI* BI)*BO
4033I and O refer to inner and outer, C and W refer to CURLYX and WHILEM:
2ab05381 4034
c476f425
DM
4035cur_
4036curlyx backtrack stack
4037------ ---------------
4038NULL
4039CO <CO prev=NULL> <WO>
4040CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
4041CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
4042NULL <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi <WO prev=CO> bo
2ab05381 4043
c476f425
DM
4044At this point the pattern succeeds, and we work back down the stack to
4045clean up, restoring as we go:
95b24440 4046
c476f425
DM
4047CO <CO prev=NULL> <WO> <CI prev=CO> <WI> ai <WI prev=CI> bi
4048CI <CO prev=NULL> <WO> <CI prev=CO> <WI> ai
4049CO <CO prev=NULL> <WO>
4050NULL
a0374537 4051
c476f425
DM
4052*******************************************************************/
4053
4054#define ST st->u.curlyx
4055
4056 case CURLYX: /* start of /A*B/ (for complex A) */
4057 {
4058 /* No need to save/restore up to this paren */
4059 I32 parenfloor = scan->flags;
4060
4061 assert(next); /* keep Coverity happy */
4062 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
4063 next += ARG(next);
4064
4065 /* XXXX Probably it is better to teach regpush to support
4066 parenfloor > PL_regsize... */
4067 if (parenfloor > (I32)*PL_reglastparen)
4068 parenfloor = *PL_reglastparen; /* Pessimization... */
4069
4070 ST.prev_curlyx= cur_curlyx;
4071 cur_curlyx = st;
4072 ST.cp = PL_savestack_ix;
4073
4074 /* these fields contain the state of the current curly.
4075 * they are accessed by subsequent WHILEMs */
4076 ST.parenfloor = parenfloor;
4077 ST.min = ARG1(scan);
4078 ST.max = ARG2(scan);
4079 ST.A = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
4080 ST.B = next;
24d3c4a9
DM
4081 ST.minmod = minmod;
4082 minmod = 0;
c476f425
DM
4083 ST.count = -1; /* this will be updated by WHILEM */
4084 ST.lastloc = NULL; /* this will be updated by WHILEM */
4085
4086 PL_reginput = locinput;
4087 PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next));
5f66b61c 4088 /* NOTREACHED */
c476f425 4089 }
a0d0e21e 4090
c476f425 4091 case CURLYX_end: /* just finished matching all of A*B */
c476f425
DM
4092 cur_curlyx = ST.prev_curlyx;
4093 sayYES;
4094 /* NOTREACHED */
a0d0e21e 4095
c476f425
DM
4096 case CURLYX_end_fail: /* just failed to match all of A*B */
4097 regcpblow(ST.cp);
4098 cur_curlyx = ST.prev_curlyx;
4099 sayNO;
4100 /* NOTREACHED */
4633a7c4 4101
a0d0e21e 4102
c476f425
DM
4103#undef ST
4104#define ST st->u.whilem
4105
4106 case WHILEM: /* just matched an A in /A*B/ (for complex A) */
4107 {
4108 /* see the discussion above about CURLYX/WHILEM */
c476f425
DM
4109 I32 n;
4110 assert(cur_curlyx); /* keep Coverity happy */
4111 n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */
4112 ST.save_lastloc = cur_curlyx->u.curlyx.lastloc;
4113 ST.cache_offset = 0;
4114 ST.cache_mask = 0;
4115
4116 PL_reginput = locinput;
4117
4118 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4119 "%*s whilem: matched %ld out of %ld..%ld\n",
4120 REPORT_CODE_OFF+depth*2, "", (long)n,
4121 (long)cur_curlyx->u.curlyx.min,
4122 (long)cur_curlyx->u.curlyx.max)
4123 );
a0d0e21e 4124
c476f425 4125 /* First just match a string of min A's. */
a0d0e21e 4126
c476f425
DM
4127 if (n < cur_curlyx->u.curlyx.min) {
4128 cur_curlyx->u.curlyx.lastloc = locinput;
4129 PUSH_STATE_GOTO(WHILEM_A_pre, cur_curlyx->u.curlyx.A);
4130 /* NOTREACHED */
4131 }
4132
4133 /* If degenerate A matches "", assume A done. */
4134
4135 if (locinput == cur_curlyx->u.curlyx.lastloc) {
4136 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4137 "%*s whilem: empty match detected, trying continuation...\n",
4138 REPORT_CODE_OFF+depth*2, "")
4139 );
4140 goto do_whilem_B_max;
4141 }
4142
4143 /* super-linear cache processing */
4144
4145 if (scan->flags) {
a0d0e21e 4146
2c2d71f5 4147 if (!PL_reg_maxiter) {
c476f425
DM
4148 /* start the countdown: Postpone detection until we
4149 * know the match is not *that* much linear. */
2c2d71f5 4150 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
66bf836d
DM
4151 /* possible overflow for long strings and many CURLYX's */
4152 if (PL_reg_maxiter < 0)
4153 PL_reg_maxiter = I32_MAX;
2c2d71f5
JH
4154 PL_reg_leftiter = PL_reg_maxiter;
4155 }
c476f425 4156
2c2d71f5 4157 if (PL_reg_leftiter-- == 0) {
c476f425 4158 /* initialise cache */
3298f257 4159 const I32 size = (PL_reg_maxiter + 7)/8;
2c2d71f5 4160 if (PL_reg_poscache) {
eb160463 4161 if ((I32)PL_reg_poscache_size < size) {
2c2d71f5
JH
4162 Renew(PL_reg_poscache, size, char);
4163 PL_reg_poscache_size = size;
4164 }
4165 Zero(PL_reg_poscache, size, char);
4166 }
4167 else {
4168 PL_reg_poscache_size = size;
a02a5408 4169 Newxz(PL_reg_poscache, size, char);
2c2d71f5 4170 }
c476f425
DM
4171 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4172 "%swhilem: Detected a super-linear match, switching on caching%s...\n",
4173 PL_colors[4], PL_colors[5])
4174 );
2c2d71f5 4175 }
c476f425 4176
2c2d71f5 4177 if (PL_reg_leftiter < 0) {
c476f425
DM
4178 /* have we already failed at this position? */
4179 I32 offset, mask;
4180 offset = (scan->flags & 0xf) - 1
4181 + (locinput - PL_bostr) * (scan->flags>>4);
4182 mask = 1 << (offset % 8);
4183 offset /= 8;
4184 if (PL_reg_poscache[offset] & mask) {
4185 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
4186 "%*s whilem: (cache) already tried at this position...\n",
4187 REPORT_CODE_OFF+depth*2, "")
2c2d71f5 4188 );
3298f257 4189 sayNO; /* cache records failure */
2c2d71f5 4190 }
c476f425
DM
4191 ST.cache_offset = offset;
4192 ST.cache_mask = mask;
2c2d71f5 4193 }
c476f425 4194 }
2c2d71f5 4195
c476f425 4196 /* Prefer B over A for minimal matching. */
a687059c 4197
c476f425
DM
4198 if (cur_curlyx->u.curlyx.minmod) {
4199 ST.save_curlyx = cur_curlyx;
4200 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
4201 ST.cp = regcppush(ST.save_curlyx->u.curlyx.parenfloor);
4202 REGCP_SET(ST.lastcp);
4203 PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B);
4204 /* NOTREACHED */
4205 }
a0d0e21e 4206
c476f425
DM
4207 /* Prefer A over B for maximal matching. */
4208
4209 if (n < cur_curlyx->u.curlyx.max) { /* More greed allowed? */
4210 ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
4211 cur_curlyx->u.curlyx.lastloc = locinput;
4212 REGCP_SET(ST.lastcp);
4213 PUSH_STATE_GOTO(WHILEM_A_max, cur_curlyx->u.curlyx.A);
4214 /* NOTREACHED */
4215 }
4216 goto do_whilem_B_max;
4217 }
4218 /* NOTREACHED */
4219
4220 case WHILEM_B_min: /* just matched B in a minimal match */
4221 case WHILEM_B_max: /* just matched B in a maximal match */
4222 cur_curlyx = ST.save_curlyx;
4223 sayYES;
4224 /* NOTREACHED */
4225
4226 case WHILEM_B_max_fail: /* just failed to match B in a maximal match */
4227 cur_curlyx = ST.save_curlyx;
4228 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
4229 cur_curlyx->u.curlyx.count--;
4230 CACHEsayNO;
4231 /* NOTREACHED */
4232
4233 case WHILEM_A_min_fail: /* just failed to match A in a minimal match */
4234 REGCP_UNWIND(ST.lastcp);
4235 regcppop(rex);
4236 /* FALL THROUGH */
4237 case WHILEM_A_pre_fail: /* just failed to match even minimal A */
4238 cur_curlyx->u.curlyx.lastloc = ST.save_lastloc;
4239 cur_curlyx->u.curlyx.count--;
4240 CACHEsayNO;
4241 /* NOTREACHED */
4242
4243 case WHILEM_A_max_fail: /* just failed to match A in a maximal match */
4244 REGCP_UNWIND(ST.lastcp);
4245 regcppop(rex); /* Restore some previous $<digit>s? */
4246 PL_reginput = locinput;
4247 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4248 "%*s whilem: failed, trying continuation...\n",
4249 REPORT_CODE_OFF+depth*2, "")
4250 );
4251 do_whilem_B_max:
4252 if (cur_curlyx->u.curlyx.count >= REG_INFTY
4253 && ckWARN(WARN_REGEXP)
4254 && !(PL_reg_flags & RF_warned))
4255 {
4256 PL_reg_flags |= RF_warned;
4257 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
4258 "Complex regular subexpression recursion",
4259 REG_INFTY - 1);
4260 }
4261
4262 /* now try B */
4263 ST.save_curlyx = cur_curlyx;
4264 cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx;
4265 PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B);
4266 /* NOTREACHED */
4267
4268 case WHILEM_B_min_fail: /* just failed to match B in a minimal match */
4269 cur_curlyx = ST.save_curlyx;
4270 REGCP_UNWIND(ST.lastcp);
4271 regcppop(rex);
4272
4273 if (cur_curlyx->u.curlyx.count >= cur_curlyx->u.curlyx.max) {
4274 /* Maximum greed exceeded */
4275 if (cur_curlyx->u.curlyx.count >= REG_INFTY
4276 && ckWARN(WARN_REGEXP)
4277 && !(PL_reg_flags & RF_warned))
4278 {
3280af22 4279 PL_reg_flags |= RF_warned;
c476f425
DM
4280 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
4281 "%s limit (%d) exceeded",
4282 "Complex regular subexpression recursion",
4283 REG_INFTY - 1);
a0d0e21e 4284 }
c476f425 4285 cur_curlyx->u.curlyx.count--;
3ab3c9b4 4286 CACHEsayNO;
a0d0e21e 4287 }
c476f425
DM
4288
4289 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4290 "%*s trying longer...\n", REPORT_CODE_OFF+depth*2, "")
4291 );
4292 /* Try grabbing another A and see if it helps. */
4293 PL_reginput = locinput;
4294 cur_curlyx->u.curlyx.lastloc = locinput;
4295 ST.cp = regcppush(cur_curlyx->u.curlyx.parenfloor);
4296 REGCP_SET(ST.lastcp);
4297 PUSH_STATE_GOTO(WHILEM_A_min, ST.save_curlyx->u.curlyx.A);
5f66b61c 4298 /* NOTREACHED */
40a82448
DM
4299
4300#undef ST
4301#define ST st->u.branch
4302
4303 case BRANCHJ: /* /(...|A|...)/ with long next pointer */
c277df42
IZ
4304 next = scan + ARG(scan);
4305 if (next == scan)
4306 next = NULL;
40a82448
DM
4307 scan = NEXTOPER(scan);
4308 /* FALL THROUGH */
c277df42 4309
40a82448
DM
4310 case BRANCH: /* /(...|A|...)/ */
4311 scan = NEXTOPER(scan); /* scan now points to inner node */
40a82448
DM
4312 ST.lastparen = *PL_reglastparen;
4313 ST.next_branch = next;
4314 REGCP_SET(ST.cp);
4315 PL_reginput = locinput;
02db2b7b 4316
40a82448 4317 /* Now go into the branch */
5d458dd8
YO
4318 if (has_cutgroup) {
4319 PUSH_YES_STATE_GOTO(BRANCH_next, scan);
4320 } else {
4321 PUSH_STATE_GOTO(BRANCH_next, scan);
4322 }
40a82448 4323 /* NOTREACHED */
5d458dd8
YO
4324 case CUTGROUP:
4325 PL_reginput = locinput;
4326 sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL :
f8fc2ecf 4327 (SV*)rexi->data->data[ ARG( scan ) ];
5d458dd8
YO
4328 PUSH_STATE_GOTO(CUTGROUP_next,next);
4329 /* NOTREACHED */
4330 case CUTGROUP_next_fail:
4331 do_cutgroup = 1;
4332 no_final = 1;
4333 if (st->u.mark.mark_name)
4334 sv_commit = st->u.mark.mark_name;
4335 sayNO;
4336 /* NOTREACHED */
4337 case BRANCH_next:
4338 sayYES;
4339 /* NOTREACHED */
40a82448 4340 case BRANCH_next_fail: /* that branch failed; try the next, if any */
5d458dd8
YO
4341 if (do_cutgroup) {
4342 do_cutgroup = 0;
4343 no_final = 0;
4344 }
40a82448
DM
4345 REGCP_UNWIND(ST.cp);
4346 for (n = *PL_reglastparen; n > ST.lastparen; n--)
f0ab9afb 4347 PL_regoffs[n].end = -1;
40a82448 4348 *PL_reglastparen = n;
0a4db386 4349 /*dmq: *PL_reglastcloseparen = n; */
40a82448
DM
4350 scan = ST.next_branch;
4351 /* no more branches? */
5d458dd8
YO
4352 if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ)) {
4353 DEBUG_EXECUTE_r({
4354 PerlIO_printf( Perl_debug_log,
4355 "%*s %sBRANCH failed...%s\n",
4356 REPORT_CODE_OFF+depth*2, "",
4357 PL_colors[4],
4358 PL_colors[5] );
4359 });
4360 sayNO_SILENT;
4361 }
40a82448
DM
4362 continue; /* execute next BRANCH[J] op */
4363 /* NOTREACHED */
4364
a0d0e21e 4365 case MINMOD:
24d3c4a9 4366 minmod = 1;
a0d0e21e 4367 break;
40a82448
DM
4368
4369#undef ST
4370#define ST st->u.curlym
4371
4372 case CURLYM: /* /A{m,n}B/ where A is fixed-length */
4373
4374 /* This is an optimisation of CURLYX that enables us to push
4375 * only a single backtracking state, no matter now many matches
4376 * there are in {m,n}. It relies on the pattern being constant
4377 * length, with no parens to influence future backrefs
4378 */
4379
4380 ST.me = scan;
dc45a647 4381 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
40a82448
DM
4382
4383 /* if paren positive, emulate an OPEN/CLOSE around A */
4384 if (ST.me->flags) {
3b6647e0 4385 U32 paren = ST.me->flags;
40a82448
DM
4386 if (paren > PL_regsize)
4387 PL_regsize = paren;
3b6647e0 4388 if (paren > *PL_reglastparen)
40a82448 4389 *PL_reglastparen = paren;
c277df42 4390 scan += NEXT_OFF(scan); /* Skip former OPEN. */
6407bf3b 4391 }
40a82448
DM
4392 ST.A = scan;
4393 ST.B = next;
4394 ST.alen = 0;
4395 ST.count = 0;
24d3c4a9
DM
4396 ST.minmod = minmod;
4397 minmod = 0;
40a82448
DM
4398 ST.c1 = CHRTEST_UNINIT;
4399 REGCP_SET(ST.cp);
6407bf3b 4400
40a82448
DM
4401 if (!(ST.minmod ? ARG1(ST.me) : ARG2(ST.me))) /* min/max */
4402 goto curlym_do_B;
4403
4404 curlym_do_A: /* execute the A in /A{m,n}B/ */
6407bf3b 4405 PL_reginput = locinput;
40a82448
DM
4406 PUSH_YES_STATE_GOTO(CURLYM_A, ST.A); /* match A */
4407 /* NOTREACHED */
5f80c4cf 4408
40a82448
DM
4409 case CURLYM_A: /* we've just matched an A */
4410 locinput = st->locinput;
4411 nextchr = UCHARAT(locinput);
4412
4413 ST.count++;
4414 /* after first match, determine A's length: u.curlym.alen */
4415 if (ST.count == 1) {
4416 if (PL_reg_match_utf8) {
4417 char *s = locinput;
4418 while (s < PL_reginput) {
4419 ST.alen++;
4420 s += UTF8SKIP(s);
4421 }
4422 }
4423 else {
4424 ST.alen = PL_reginput - locinput;
4425 }
4426 if (ST.alen == 0)
4427 ST.count = ST.minmod ? ARG1(ST.me) : ARG2(ST.me);
4428 }
0cadcf80
DM
4429 DEBUG_EXECUTE_r(
4430 PerlIO_printf(Perl_debug_log,
40a82448 4431 "%*s CURLYM now matched %"IVdf" times, len=%"IVdf"...\n",
5bc10b2c 4432 (int)(REPORT_CODE_OFF+(depth*2)), "",
40a82448 4433 (IV) ST.count, (IV)ST.alen)
0cadcf80
DM
4434 );
4435
40a82448 4436 locinput = PL_reginput;
0a4db386
YO
4437
4438 if (cur_eval && cur_eval->u.eval.close_paren &&
24b23f37 4439 cur_eval->u.eval.close_paren == (U32)ST.me->flags)
0a4db386
YO
4440 goto fake_end;
4441
4442 if ( ST.count < (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)) )
40a82448
DM
4443 goto curlym_do_A; /* try to match another A */
4444 goto curlym_do_B; /* try to match B */
5f80c4cf 4445
40a82448
DM
4446 case CURLYM_A_fail: /* just failed to match an A */
4447 REGCP_UNWIND(ST.cp);
0a4db386
YO
4448
4449 if (ST.minmod || ST.count < ARG1(ST.me) /* min*/
4450 || (cur_eval && cur_eval->u.eval.close_paren &&
24b23f37 4451 cur_eval->u.eval.close_paren == (U32)ST.me->flags))
40a82448 4452 sayNO;
0cadcf80 4453
40a82448
DM
4454 curlym_do_B: /* execute the B in /A{m,n}B/ */
4455 PL_reginput = locinput;
4456 if (ST.c1 == CHRTEST_UNINIT) {
4457 /* calculate c1 and c2 for possible match of 1st char
4458 * following curly */
4459 ST.c1 = ST.c2 = CHRTEST_VOID;
4460 if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) {
4461 regnode *text_node = ST.B;
4462 if (! HAS_TEXT(text_node))
4463 FIND_NEXT_IMPT(text_node);
ee9b8eae
YO
4464 /* this used to be
4465
4466 (HAS_TEXT(text_node) && PL_regkind[OP(text_node)] == EXACT)
4467
4468 But the former is redundant in light of the latter.
4469
4470 if this changes back then the macro for
4471 IS_TEXT and friends need to change.
4472 */
4473 if (PL_regkind[OP(text_node)] == EXACT)
40a82448 4474 {
ee9b8eae 4475
40a82448
DM
4476 ST.c1 = (U8)*STRING(text_node);
4477 ST.c2 =
ee9b8eae 4478 (IS_TEXTF(text_node))
40a82448 4479 ? PL_fold[ST.c1]
ee9b8eae 4480 : (IS_TEXTFL(text_node))
40a82448
DM
4481 ? PL_fold_locale[ST.c1]
4482 : ST.c1;
c277df42 4483 }
c277df42 4484 }
40a82448
DM
4485 }
4486
4487 DEBUG_EXECUTE_r(
4488 PerlIO_printf(Perl_debug_log,
4489 "%*s CURLYM trying tail with matches=%"IVdf"...\n",
5bc10b2c 4490 (int)(REPORT_CODE_OFF+(depth*2)),
40a82448
DM
4491 "", (IV)ST.count)
4492 );
4493 if (ST.c1 != CHRTEST_VOID
4494 && UCHARAT(PL_reginput) != ST.c1
4495 && UCHARAT(PL_reginput) != ST.c2)
4496 {
4497 /* simulate B failing */
3e901dc0
YO
4498 DEBUG_OPTIMISE_r(
4499 PerlIO_printf(Perl_debug_log,
4500 "%*s CURLYM Fast bail c1=%"IVdf" c2=%"IVdf"\n",
4501 (int)(REPORT_CODE_OFF+(depth*2)),"",
4502 (IV)ST.c1,(IV)ST.c2
4503 ));
40a82448
DM
4504 state_num = CURLYM_B_fail;
4505 goto reenter_switch;
4506 }
4507
4508 if (ST.me->flags) {
4509 /* mark current A as captured */
4510 I32 paren = ST.me->flags;
4511 if (ST.count) {
f0ab9afb 4512 PL_regoffs[paren].start
40a82448 4513 = HOPc(PL_reginput, -ST.alen) - PL_bostr;
f0ab9afb 4514 PL_regoffs[paren].end = PL_reginput - PL_bostr;
0a4db386 4515 /*dmq: *PL_reglastcloseparen = paren; */
c277df42 4516 }
40a82448 4517 else
f0ab9afb 4518 PL_regoffs[paren].end = -1;
0a4db386 4519 if (cur_eval && cur_eval->u.eval.close_paren &&
24b23f37 4520 cur_eval->u.eval.close_paren == (U32)ST.me->flags)
0a4db386
YO
4521 {
4522 if (ST.count)
4523 goto fake_end;
4524 else
4525 sayNO;
4526 }
c277df42 4527 }
0a4db386 4528
40a82448 4529 PUSH_STATE_GOTO(CURLYM_B, ST.B); /* match B */
5f66b61c 4530 /* NOTREACHED */
40a82448
DM
4531
4532 case CURLYM_B_fail: /* just failed to match a B */
4533 REGCP_UNWIND(ST.cp);
4534 if (ST.minmod) {
4535 if (ST.count == ARG2(ST.me) /* max */)
4536 sayNO;
4537 goto curlym_do_A; /* try to match a further A */
4538 }
4539 /* backtrack one A */
4540 if (ST.count == ARG1(ST.me) /* min */)
4541 sayNO;
4542 ST.count--;
4543 locinput = HOPc(locinput, -ST.alen);
4544 goto curlym_do_B; /* try to match B */
4545
c255a977
DM
4546#undef ST
4547#define ST st->u.curly
40a82448 4548
c255a977
DM
4549#define CURLY_SETPAREN(paren, success) \
4550 if (paren) { \
4551 if (success) { \
f0ab9afb
NC
4552 PL_regoffs[paren].start = HOPc(locinput, -1) - PL_bostr; \
4553 PL_regoffs[paren].end = locinput - PL_bostr; \
0a4db386 4554 *PL_reglastcloseparen = paren; \
c255a977
DM
4555 } \
4556 else \
f0ab9afb 4557 PL_regoffs[paren].end = -1; \
c255a977
DM
4558 }
4559
4560 case STAR: /* /A*B/ where A is width 1 */
4561 ST.paren = 0;
4562 ST.min = 0;
4563 ST.max = REG_INFTY;
a0d0e21e
LW
4564 scan = NEXTOPER(scan);
4565 goto repeat;
c255a977
DM
4566 case PLUS: /* /A+B/ where A is width 1 */
4567 ST.paren = 0;
4568 ST.min = 1;
4569 ST.max = REG_INFTY;
c277df42 4570 scan = NEXTOPER(scan);
c255a977
DM
4571 goto repeat;
4572 case CURLYN: /* /(A){m,n}B/ where A is width 1 */
4573 ST.paren = scan->flags; /* Which paren to set */
4574 if (ST.paren > PL_regsize)
4575 PL_regsize = ST.paren;
3b6647e0 4576 if (ST.paren > *PL_reglastparen)
c255a977
DM
4577 *PL_reglastparen = ST.paren;
4578 ST.min = ARG1(scan); /* min to match */
4579 ST.max = ARG2(scan); /* max to match */
0a4db386 4580 if (cur_eval && cur_eval->u.eval.close_paren &&
86413ec0 4581 cur_eval->u.eval.close_paren == (U32)ST.paren) {
0a4db386
YO
4582 ST.min=1;
4583 ST.max=1;
4584 }
c255a977
DM
4585 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
4586 goto repeat;
4587 case CURLY: /* /A{m,n}B/ where A is width 1 */
4588 ST.paren = 0;
4589 ST.min = ARG1(scan); /* min to match */
4590 ST.max = ARG2(scan); /* max to match */
4591 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
c277df42 4592 repeat:
a0d0e21e
LW
4593 /*
4594 * Lookahead to avoid useless match attempts
4595 * when we know what character comes next.
c255a977 4596 *
5f80c4cf
JP
4597 * Used to only do .*x and .*?x, but now it allows
4598 * for )'s, ('s and (?{ ... })'s to be in the way
4599 * of the quantifier and the EXACT-like node. -- japhy
4600 */
4601
c255a977
DM
4602 if (ST.min > ST.max) /* XXX make this a compile-time check? */
4603 sayNO;
cca55fe3 4604 if (HAS_TEXT(next) || JUMPABLE(next)) {
5f80c4cf
JP
4605 U8 *s;
4606 regnode *text_node = next;
4607
3dab1dad
YO
4608 if (! HAS_TEXT(text_node))
4609 FIND_NEXT_IMPT(text_node);
5f80c4cf 4610
9e137952 4611 if (! HAS_TEXT(text_node))
c255a977 4612 ST.c1 = ST.c2 = CHRTEST_VOID;
5f80c4cf 4613 else {
ee9b8eae 4614 if ( PL_regkind[OP(text_node)] != EXACT ) {
c255a977 4615 ST.c1 = ST.c2 = CHRTEST_VOID;
44a68960 4616 goto assume_ok_easy;
cca55fe3 4617 }
be8e71aa
YO
4618 else
4619 s = (U8*)STRING(text_node);
ee9b8eae
YO
4620
4621 /* Currently we only get here when
4622
4623 PL_rekind[OP(text_node)] == EXACT
4624
4625 if this changes back then the macro for IS_TEXT and
4626 friends need to change. */
5f80c4cf 4627 if (!UTF) {
c255a977 4628 ST.c2 = ST.c1 = *s;
ee9b8eae 4629 if (IS_TEXTF(text_node))
c255a977 4630 ST.c2 = PL_fold[ST.c1];
ee9b8eae 4631 else if (IS_TEXTFL(text_node))
c255a977 4632 ST.c2 = PL_fold_locale[ST.c1];
1aa99e6b 4633 }
5f80c4cf 4634 else { /* UTF */
ee9b8eae 4635 if (IS_TEXTF(text_node)) {
a2a2844f 4636 STRLEN ulen1, ulen2;
89ebb4a3
JH
4637 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
4638 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
a2a2844f
JH
4639
4640 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
4641 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
e294cc5d
JH
4642#ifdef EBCDIC
4643 ST.c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN, 0,
4644 ckWARN(WARN_UTF8) ?
4645 0 : UTF8_ALLOW_ANY);
4646 ST.c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN, 0,
4647 ckWARN(WARN_UTF8) ?
4648 0 : UTF8_ALLOW_ANY);
4649#else
c255a977 4650 ST.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
e294cc5d 4651 uniflags);
c255a977 4652 ST.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
e294cc5d
JH
4653 uniflags);
4654#endif
5f80c4cf
JP
4655 }
4656 else {
c255a977 4657 ST.c2 = ST.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
041457d9 4658 uniflags);
5f80c4cf 4659 }
1aa99e6b
IH
4660 }
4661 }
bbce6d69 4662 }
a0d0e21e 4663 else
c255a977 4664 ST.c1 = ST.c2 = CHRTEST_VOID;
cca55fe3 4665 assume_ok_easy:
c255a977
DM
4666
4667 ST.A = scan;
4668 ST.B = next;
3280af22 4669 PL_reginput = locinput;
24d3c4a9
DM
4670 if (minmod) {
4671 minmod = 0;
e2e6a0f1 4672 if (ST.min && regrepeat(rex, ST.A, ST.min, depth) < ST.min)
4633a7c4 4673 sayNO;
c255a977 4674 ST.count = ST.min;
a0ed51b3 4675 locinput = PL_reginput;
c255a977
DM
4676 REGCP_SET(ST.cp);
4677 if (ST.c1 == CHRTEST_VOID)
4678 goto curly_try_B_min;
4679
4680 ST.oldloc = locinput;
4681
4682 /* set ST.maxpos to the furthest point along the
4683 * string that could possibly match */
4684 if (ST.max == REG_INFTY) {
4685 ST.maxpos = PL_regeol - 1;
4686 if (do_utf8)
4687 while (UTF8_IS_CONTINUATION(*(U8*)ST.maxpos))
4688 ST.maxpos--;
4689 }
4690 else if (do_utf8) {
4691 int m = ST.max - ST.min;
4692 for (ST.maxpos = locinput;
4693 m >0 && ST.maxpos + UTF8SKIP(ST.maxpos) <= PL_regeol; m--)
4694 ST.maxpos += UTF8SKIP(ST.maxpos);
4695 }
4696 else {
4697 ST.maxpos = locinput + ST.max - ST.min;
4698 if (ST.maxpos >= PL_regeol)
4699 ST.maxpos = PL_regeol - 1;
4700 }
4701 goto curly_try_B_min_known;
4702
4703 }
4704 else {
e2e6a0f1 4705 ST.count = regrepeat(rex, ST.A, ST.max, depth);
c255a977
DM
4706 locinput = PL_reginput;
4707 if (ST.count < ST.min)
4708 sayNO;
4709 if ((ST.count > ST.min)
4710 && (PL_regkind[OP(ST.B)] == EOL) && (OP(ST.B) != MEOL))
4711 {
4712 /* A{m,n} must come at the end of the string, there's
4713 * no point in backing off ... */
4714 ST.min = ST.count;
4715 /* ...except that $ and \Z can match before *and* after
4716 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
4717 We may back off by one in this case. */
4718 if (UCHARAT(PL_reginput - 1) == '\n' && OP(ST.B) != EOS)
4719 ST.min--;
4720 }
4721 REGCP_SET(ST.cp);
4722 goto curly_try_B_max;
4723 }
4724 /* NOTREACHED */
4725
4726
4727 case CURLY_B_min_known_fail:
4728 /* failed to find B in a non-greedy match where c1,c2 valid */
4729 if (ST.paren && ST.count)
f0ab9afb 4730 PL_regoffs[ST.paren].end = -1;
c255a977
DM
4731
4732 PL_reginput = locinput; /* Could be reset... */
4733 REGCP_UNWIND(ST.cp);
4734 /* Couldn't or didn't -- move forward. */
4735 ST.oldloc = locinput;
4736 if (do_utf8)
4737 locinput += UTF8SKIP(locinput);
4738 else
4739 locinput++;
4740 ST.count++;
4741 curly_try_B_min_known:
4742 /* find the next place where 'B' could work, then call B */
4743 {
4744 int n;
4745 if (do_utf8) {
4746 n = (ST.oldloc == locinput) ? 0 : 1;
4747 if (ST.c1 == ST.c2) {
4748 STRLEN len;
4749 /* set n to utf8_distance(oldloc, locinput) */
4750 while (locinput <= ST.maxpos &&
4751 utf8n_to_uvchr((U8*)locinput,
4752 UTF8_MAXBYTES, &len,
4753 uniflags) != (UV)ST.c1) {
4754 locinput += len;
4755 n++;
4756 }
1aa99e6b
IH
4757 }
4758 else {
c255a977
DM
4759 /* set n to utf8_distance(oldloc, locinput) */
4760 while (locinput <= ST.maxpos) {
4761 STRLEN len;
4762 const UV c = utf8n_to_uvchr((U8*)locinput,
4763 UTF8_MAXBYTES, &len,
4764 uniflags);
4765 if (c == (UV)ST.c1 || c == (UV)ST.c2)
4766 break;
4767 locinput += len;
4768 n++;
1aa99e6b 4769 }
0fe9bf95
IZ
4770 }
4771 }
c255a977
DM
4772 else {
4773 if (ST.c1 == ST.c2) {
4774 while (locinput <= ST.maxpos &&
4775 UCHARAT(locinput) != ST.c1)
4776 locinput++;
bbce6d69 4777 }
c255a977
DM
4778 else {
4779 while (locinput <= ST.maxpos
4780 && UCHARAT(locinput) != ST.c1
4781 && UCHARAT(locinput) != ST.c2)
4782 locinput++;
a0ed51b3 4783 }
c255a977
DM
4784 n = locinput - ST.oldloc;
4785 }
4786 if (locinput > ST.maxpos)
4787 sayNO;
4788 /* PL_reginput == oldloc now */
4789 if (n) {
4790 ST.count += n;
e2e6a0f1 4791 if (regrepeat(rex, ST.A, n, depth) < n)
4633a7c4 4792 sayNO;
a0d0e21e 4793 }
c255a977
DM
4794 PL_reginput = locinput;
4795 CURLY_SETPAREN(ST.paren, ST.count);
0a4db386 4796 if (cur_eval && cur_eval->u.eval.close_paren &&
86413ec0 4797 cur_eval->u.eval.close_paren == (U32)ST.paren) {
0a4db386
YO
4798 goto fake_end;
4799 }
c255a977 4800 PUSH_STATE_GOTO(CURLY_B_min_known, ST.B);
a0d0e21e 4801 }
c255a977
DM
4802 /* NOTREACHED */
4803
4804
4805 case CURLY_B_min_fail:
4806 /* failed to find B in a non-greedy match where c1,c2 invalid */
4807 if (ST.paren && ST.count)
f0ab9afb 4808 PL_regoffs[ST.paren].end = -1;
c255a977
DM
4809
4810 REGCP_UNWIND(ST.cp);
4811 /* failed -- move forward one */
4812 PL_reginput = locinput;
e2e6a0f1 4813 if (regrepeat(rex, ST.A, 1, depth)) {
c255a977 4814 ST.count++;
a0ed51b3 4815 locinput = PL_reginput;
c255a977
DM
4816 if (ST.count <= ST.max || (ST.max == REG_INFTY &&
4817 ST.count > 0)) /* count overflow ? */
15272685 4818 {
c255a977
DM
4819 curly_try_B_min:
4820 CURLY_SETPAREN(ST.paren, ST.count);
0a4db386 4821 if (cur_eval && cur_eval->u.eval.close_paren &&
86413ec0 4822 cur_eval->u.eval.close_paren == (U32)ST.paren) {
0a4db386
YO
4823 goto fake_end;
4824 }
c255a977 4825 PUSH_STATE_GOTO(CURLY_B_min, ST.B);
a0d0e21e
LW
4826 }
4827 }
4633a7c4 4828 sayNO;
c255a977
DM
4829 /* NOTREACHED */
4830
4831
4832 curly_try_B_max:
4833 /* a successful greedy match: now try to match B */
40d049e4 4834 if (cur_eval && cur_eval->u.eval.close_paren &&
86413ec0 4835 cur_eval->u.eval.close_paren == (U32)ST.paren) {
40d049e4
YO
4836 goto fake_end;
4837 }
c255a977
DM
4838 {
4839 UV c = 0;
4840 if (ST.c1 != CHRTEST_VOID)
4841 c = do_utf8 ? utf8n_to_uvchr((U8*)PL_reginput,
4842 UTF8_MAXBYTES, 0, uniflags)
466787eb 4843 : (UV) UCHARAT(PL_reginput);
c255a977
DM
4844 /* If it could work, try it. */
4845 if (ST.c1 == CHRTEST_VOID || c == (UV)ST.c1 || c == (UV)ST.c2) {
4846 CURLY_SETPAREN(ST.paren, ST.count);
4847 PUSH_STATE_GOTO(CURLY_B_max, ST.B);
4848 /* NOTREACHED */
4849 }
4850 }
4851 /* FALL THROUGH */
4852 case CURLY_B_max_fail:
4853 /* failed to find B in a greedy match */
4854 if (ST.paren && ST.count)
f0ab9afb 4855 PL_regoffs[ST.paren].end = -1;
c255a977
DM
4856
4857 REGCP_UNWIND(ST.cp);
4858 /* back up. */
4859 if (--ST.count < ST.min)
4860 sayNO;
4861 PL_reginput = locinput = HOPc(locinput, -1);
4862 goto curly_try_B_max;
4863
4864#undef ST
4865
a0d0e21e 4866 case END:
6bda09f9 4867 fake_end:
faec1544
DM
4868 if (cur_eval) {
4869 /* we've just finished A in /(??{A})B/; now continue with B */
4870 I32 tmpix;
faec1544
DM
4871 st->u.eval.toggle_reg_flags
4872 = cur_eval->u.eval.toggle_reg_flags;
4873 PL_reg_flags ^= st->u.eval.toggle_reg_flags;
4874
288b8c02
NC
4875 st->u.eval.prev_rex = rex_sv; /* inner */
4876 SETREX(rex_sv,cur_eval->u.eval.prev_rex);
4877 rex = (struct regexp *)SvANY(rex_sv);
f8fc2ecf 4878 rexi = RXi_GET(rex);
faec1544 4879 cur_curlyx = cur_eval->u.eval.prev_curlyx;
288b8c02 4880 ReREFCNT_inc(rex_sv);
faec1544
DM
4881 st->u.eval.cp = regcppush(0); /* Save *all* the positions. */
4882 REGCP_SET(st->u.eval.lastcp);
4883 PL_reginput = locinput;
4884
4885 /* Restore parens of the outer rex without popping the
4886 * savestack */
4887 tmpix = PL_savestack_ix;
4888 PL_savestack_ix = cur_eval->u.eval.lastcp;
4889 regcppop(rex);
4890 PL_savestack_ix = tmpix;
4891
4892 st->u.eval.prev_eval = cur_eval;
4893 cur_eval = cur_eval->u.eval.prev_eval;
4894 DEBUG_EXECUTE_r(
2a49f0f5
JH
4895 PerlIO_printf(Perl_debug_log, "%*s EVAL trying tail ... %"UVxf"\n",
4896 REPORT_CODE_OFF+depth*2, "",PTR2UV(cur_eval)););
e7707071
YO
4897 if ( nochange_depth )
4898 nochange_depth--;
4899
4900 PUSH_YES_STATE_GOTO(EVAL_AB,
faec1544
DM
4901 st->u.eval.prev_eval->u.eval.B); /* match B */
4902 }
4903
3b0527fe 4904 if (locinput < reginfo->till) {
a3621e74 4905 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
7821416a
IZ
4906 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
4907 PL_colors[4],
4908 (long)(locinput - PL_reg_starttry),
3b0527fe 4909 (long)(reginfo->till - PL_reg_starttry),
7821416a 4910 PL_colors[5]));
58e23c8d 4911
262b90c4 4912 sayNO_SILENT; /* Cannot match: too short. */
7821416a
IZ
4913 }
4914 PL_reginput = locinput; /* put where regtry can find it */
262b90c4 4915 sayYES; /* Success! */
dad79028
DM
4916
4917 case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */
4918 DEBUG_EXECUTE_r(
4919 PerlIO_printf(Perl_debug_log,
4920 "%*s %ssubpattern success...%s\n",
5bc10b2c 4921 REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]));
3280af22 4922 PL_reginput = locinput; /* put where regtry can find it */
262b90c4 4923 sayYES; /* Success! */
dad79028 4924
40a82448
DM
4925#undef ST
4926#define ST st->u.ifmatch
4927
4928 case SUSPEND: /* (?>A) */
4929 ST.wanted = 1;
9fe1d20c 4930 PL_reginput = locinput;
9041c2e3 4931 goto do_ifmatch;
dad79028 4932
40a82448
DM
4933 case UNLESSM: /* -ve lookaround: (?!A), or with flags, (?<!A) */
4934 ST.wanted = 0;
dad79028
DM
4935 goto ifmatch_trivial_fail_test;
4936
40a82448
DM
4937 case IFMATCH: /* +ve lookaround: (?=A), or with flags, (?<=A) */
4938 ST.wanted = 1;
dad79028 4939 ifmatch_trivial_fail_test:
a0ed51b3 4940 if (scan->flags) {
52657f30 4941 char * const s = HOPBACKc(locinput, scan->flags);
dad79028
DM
4942 if (!s) {
4943 /* trivial fail */
24d3c4a9
DM
4944 if (logical) {
4945 logical = 0;
4946 sw = 1 - (bool)ST.wanted;
dad79028 4947 }
40a82448 4948 else if (ST.wanted)
dad79028
DM
4949 sayNO;
4950 next = scan + ARG(scan);
4951 if (next == scan)
4952 next = NULL;
4953 break;
4954 }
efb30f32 4955 PL_reginput = s;
a0ed51b3
LW
4956 }
4957 else
4958 PL_reginput = locinput;
4959
c277df42 4960 do_ifmatch:
40a82448 4961 ST.me = scan;
24d3c4a9 4962 ST.logical = logical;
40a82448
DM
4963 /* execute body of (?...A) */
4964 PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan)));
4965 /* NOTREACHED */
4966
4967 case IFMATCH_A_fail: /* body of (?...A) failed */
4968 ST.wanted = !ST.wanted;
4969 /* FALL THROUGH */
4970
4971 case IFMATCH_A: /* body of (?...A) succeeded */
24d3c4a9
DM
4972 if (ST.logical) {
4973 sw = (bool)ST.wanted;
40a82448
DM
4974 }
4975 else if (!ST.wanted)
4976 sayNO;
4977
4978 if (OP(ST.me) == SUSPEND)
4979 locinput = PL_reginput;
4980 else {
4981 locinput = PL_reginput = st->locinput;
4982 nextchr = UCHARAT(locinput);
4983 }
4984 scan = ST.me + ARG(ST.me);
4985 if (scan == ST.me)
4986 scan = NULL;
4987 continue; /* execute B */
4988
4989#undef ST
dad79028 4990
c277df42 4991 case LONGJMP:
c277df42
IZ
4992 next = scan + ARG(scan);
4993 if (next == scan)
4994 next = NULL;
a0d0e21e 4995 break;
54612592 4996 case COMMIT:
e2e6a0f1
YO
4997 reginfo->cutpoint = PL_regeol;
4998 /* FALLTHROUGH */
5d458dd8 4999 case PRUNE:
24b23f37 5000 PL_reginput = locinput;
e2e6a0f1 5001 if (!scan->flags)
f8fc2ecf 5002 sv_yes_mark = sv_commit = (SV*)rexi->data->data[ ARG( scan ) ];
54612592
YO
5003 PUSH_STATE_GOTO(COMMIT_next,next);
5004 /* NOTREACHED */
5005 case COMMIT_next_fail:
5006 no_final = 1;
5007 /* FALLTHROUGH */
7f69552c
YO
5008 case OPFAIL:
5009 sayNO;
e2e6a0f1
YO
5010 /* NOTREACHED */
5011
5012#define ST st->u.mark
5013 case MARKPOINT:
5014 ST.prev_mark = mark_state;
5d458dd8 5015 ST.mark_name = sv_commit = sv_yes_mark
f8fc2ecf 5016 = (SV*)rexi->data->data[ ARG( scan ) ];
e2e6a0f1
YO
5017 mark_state = st;
5018 ST.mark_loc = PL_reginput = locinput;
5019 PUSH_YES_STATE_GOTO(MARKPOINT_next,next);
5020 /* NOTREACHED */
5021 case MARKPOINT_next:
5022 mark_state = ST.prev_mark;
5023 sayYES;
5024 /* NOTREACHED */
5025 case MARKPOINT_next_fail:
5d458dd8 5026 if (popmark && sv_eq(ST.mark_name,popmark))
e2e6a0f1
YO
5027 {
5028 if (ST.mark_loc > startpoint)
5029 reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1);
5030 popmark = NULL; /* we found our mark */
5031 sv_commit = ST.mark_name;
5032
5033 DEBUG_EXECUTE_r({
5d458dd8 5034 PerlIO_printf(Perl_debug_log,
e2e6a0f1
YO
5035 "%*s %ssetting cutpoint to mark:%"SVf"...%s\n",
5036 REPORT_CODE_OFF+depth*2, "",
be2597df 5037 PL_colors[4], SVfARG(sv_commit), PL_colors[5]);
e2e6a0f1
YO
5038 });
5039 }
5040 mark_state = ST.prev_mark;
5d458dd8
YO
5041 sv_yes_mark = mark_state ?
5042 mark_state->u.mark.mark_name : NULL;
e2e6a0f1
YO
5043 sayNO;
5044 /* NOTREACHED */
5d458dd8
YO
5045 case SKIP:
5046 PL_reginput = locinput;
5047 if (scan->flags) {
2bf803e2 5048 /* (*SKIP) : if we fail we cut here*/
5d458dd8 5049 ST.mark_name = NULL;
e2e6a0f1 5050 ST.mark_loc = locinput;
5d458dd8
YO
5051 PUSH_STATE_GOTO(SKIP_next,next);
5052 } else {
2bf803e2 5053 /* (*SKIP:NAME) : if there is a (*MARK:NAME) fail where it was,
5d458dd8
YO
5054 otherwise do nothing. Meaning we need to scan
5055 */
5056 regmatch_state *cur = mark_state;
f8fc2ecf 5057 SV *find = (SV*)rexi->data->data[ ARG( scan ) ];
5d458dd8
YO
5058
5059 while (cur) {
5060 if ( sv_eq( cur->u.mark.mark_name,
5061 find ) )
5062 {
5063 ST.mark_name = find;
5064 PUSH_STATE_GOTO( SKIP_next, next );
5065 }
5066 cur = cur->u.mark.prev_mark;
5067 }
e2e6a0f1 5068 }
2bf803e2 5069 /* Didn't find our (*MARK:NAME) so ignore this (*SKIP:NAME) */
5d458dd8
YO
5070 break;
5071 case SKIP_next_fail:
5072 if (ST.mark_name) {
5073 /* (*CUT:NAME) - Set up to search for the name as we
5074 collapse the stack*/
5075 popmark = ST.mark_name;
5076 } else {
5077 /* (*CUT) - No name, we cut here.*/
e2e6a0f1
YO
5078 if (ST.mark_loc > startpoint)
5079 reginfo->cutpoint = HOPBACKc(ST.mark_loc, 1);
5d458dd8
YO
5080 /* but we set sv_commit to latest mark_name if there
5081 is one so they can test to see how things lead to this
5082 cut */
5083 if (mark_state)
5084 sv_commit=mark_state->u.mark.mark_name;
5085 }
e2e6a0f1
YO
5086 no_final = 1;
5087 sayNO;
5088 /* NOTREACHED */
5089#undef ST
32e6a07c
YO
5090 case FOLDCHAR:
5091 n = ARG(scan);
81d4fa0f 5092 if ( n == (U32)what_len_TRICKYFOLD(locinput,do_utf8,ln) ) {
e64b1bd1
YO
5093 locinput += ln;
5094 } else if ( 0xDF == n && !do_utf8 && !UTF ) {
5095 sayNO;
5096 } else {
5097 U8 folded[UTF8_MAXBYTES_CASE+1];
5098 STRLEN foldlen;
5099 const char * const l = locinput;
5100 char *e = PL_regeol;
5101 to_uni_fold(n, folded, &foldlen);
5102
59fe32ea 5103 if (ibcmp_utf8((const char*) folded, 0, foldlen, 1,
e64b1bd1 5104 l, &e, 0, do_utf8)) {
32e6a07c 5105 sayNO;
e64b1bd1
YO
5106 }
5107 locinput = e;
32e6a07c
YO
5108 }
5109 nextchr = UCHARAT(locinput);
5110 break;
e1d1eefb
YO
5111 case LNBREAK:
5112 if ((n=is_LNBREAK(locinput,do_utf8))) {
5113 locinput += n;
5114 nextchr = UCHARAT(locinput);
5115 } else
5116 sayNO;
5117 break;
5118
5119#define CASE_CLASS(nAmE) \
5120 case nAmE: \
5121 if ((n=is_##nAmE(locinput,do_utf8))) { \
5122 locinput += n; \
5123 nextchr = UCHARAT(locinput); \
5124 } else \
5125 sayNO; \
5126 break; \
5127 case N##nAmE: \
5128 if ((n=is_##nAmE(locinput,do_utf8))) { \
5129 sayNO; \
5130 } else { \
5131 locinput += UTF8SKIP(locinput); \
5132 nextchr = UCHARAT(locinput); \
5133 } \
5134 break
5135
5136 CASE_CLASS(VERTWS);
5137 CASE_CLASS(HORIZWS);
5138#undef CASE_CLASS
5139
a0d0e21e 5140 default:
b900a521 5141 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
d7d93a81 5142 PTR2UV(scan), OP(scan));
cea2e8a9 5143 Perl_croak(aTHX_ "regexp memory corruption");
5d458dd8
YO
5144
5145 } /* end switch */
95b24440 5146
5d458dd8
YO
5147 /* switch break jumps here */
5148 scan = next; /* prepare to execute the next op and ... */
5149 continue; /* ... jump back to the top, reusing st */
95b24440
DM
5150 /* NOTREACHED */
5151
40a82448
DM
5152 push_yes_state:
5153 /* push a state that backtracks on success */
5154 st->u.yes.prev_yes_state = yes_state;
5155 yes_state = st;
5156 /* FALL THROUGH */
5157 push_state:
5158 /* push a new regex state, then continue at scan */
5159 {
5160 regmatch_state *newst;
5161
24b23f37
YO
5162 DEBUG_STACK_r({
5163 regmatch_state *cur = st;
5164 regmatch_state *curyes = yes_state;
5165 int curd = depth;
5166 regmatch_slab *slab = PL_regmatch_slab;
5167 for (;curd > -1;cur--,curd--) {
5168 if (cur < SLAB_FIRST(slab)) {
5169 slab = slab->prev;
5170 cur = SLAB_LAST(slab);
5171 }
5172 PerlIO_printf(Perl_error_log, "%*s#%-3d %-10s %s\n",
5173 REPORT_CODE_OFF + 2 + depth * 2,"",
13d6edb4 5174 curd, PL_reg_name[cur->resume_state],
24b23f37
YO
5175 (curyes == cur) ? "yes" : ""
5176 );
5177 if (curyes == cur)
5178 curyes = cur->u.yes.prev_yes_state;
5179 }
5180 } else
5181 DEBUG_STATE_pp("push")
5182 );
40a82448 5183 depth++;
40a82448
DM
5184 st->locinput = locinput;
5185 newst = st+1;
5186 if (newst > SLAB_LAST(PL_regmatch_slab))
5187 newst = S_push_slab(aTHX);
5188 PL_regmatch_state = newst;
786e8c11 5189
40a82448
DM
5190 locinput = PL_reginput;
5191 nextchr = UCHARAT(locinput);
5192 st = newst;
5193 continue;
5194 /* NOTREACHED */
5195 }
a0d0e21e 5196 }
a687059c 5197
a0d0e21e
LW
5198 /*
5199 * We get here only if there's trouble -- normally "case END" is
5200 * the terminating point.
5201 */
cea2e8a9 5202 Perl_croak(aTHX_ "corrupted regexp pointers");
a0d0e21e 5203 /*NOTREACHED*/
4633a7c4
LW
5204 sayNO;
5205
262b90c4 5206yes:
77cb431f
DM
5207 if (yes_state) {
5208 /* we have successfully completed a subexpression, but we must now
5209 * pop to the state marked by yes_state and continue from there */
77cb431f 5210 assert(st != yes_state);
5bc10b2c
DM
5211#ifdef DEBUGGING
5212 while (st != yes_state) {
5213 st--;
5214 if (st < SLAB_FIRST(PL_regmatch_slab)) {
5215 PL_regmatch_slab = PL_regmatch_slab->prev;
5216 st = SLAB_LAST(PL_regmatch_slab);
5217 }
e2e6a0f1 5218 DEBUG_STATE_r({
54612592
YO
5219 if (no_final) {
5220 DEBUG_STATE_pp("pop (no final)");
5221 } else {
5222 DEBUG_STATE_pp("pop (yes)");
5223 }
e2e6a0f1 5224 });
5bc10b2c
DM
5225 depth--;
5226 }
5227#else
77cb431f
DM
5228 while (yes_state < SLAB_FIRST(PL_regmatch_slab)
5229 || yes_state > SLAB_LAST(PL_regmatch_slab))
5230 {
5231 /* not in this slab, pop slab */
5232 depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1);
5233 PL_regmatch_slab = PL_regmatch_slab->prev;
5234 st = SLAB_LAST(PL_regmatch_slab);
5235 }
5236 depth -= (st - yes_state);
5bc10b2c 5237#endif
77cb431f
DM
5238 st = yes_state;
5239 yes_state = st->u.yes.prev_yes_state;
5240 PL_regmatch_state = st;
24b23f37 5241
5d458dd8
YO
5242 if (no_final) {
5243 locinput= st->locinput;
5244 nextchr = UCHARAT(locinput);
5245 }
54612592 5246 state_num = st->resume_state + no_final;
24d3c4a9 5247 goto reenter_switch;
77cb431f
DM
5248 }
5249
a3621e74 5250 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
e4584336 5251 PL_colors[4], PL_colors[5]));
02db2b7b 5252
19b95bf0
DM
5253 if (PL_reg_eval_set) {
5254 /* each successfully executed (?{...}) block does the equivalent of
5255 * local $^R = do {...}
5256 * When popping the save stack, all these locals would be undone;
5257 * bypass this by setting the outermost saved $^R to the latest
5258 * value */
5259 if (oreplsv != GvSV(PL_replgv))
5260 sv_setsv(oreplsv, GvSV(PL_replgv));
5261 }
95b24440 5262 result = 1;
aa283a38 5263 goto final_exit;
4633a7c4
LW
5264
5265no:
a3621e74 5266 DEBUG_EXECUTE_r(
7821416a 5267 PerlIO_printf(Perl_debug_log,
786e8c11 5268 "%*s %sfailed...%s\n",
5bc10b2c 5269 REPORT_CODE_OFF+depth*2, "",
786e8c11 5270 PL_colors[4], PL_colors[5])
7821416a 5271 );
aa283a38 5272
262b90c4 5273no_silent:
54612592
YO
5274 if (no_final) {
5275 if (yes_state) {
5276 goto yes;
5277 } else {
5278 goto final_exit;
5279 }
5280 }
aa283a38
DM
5281 if (depth) {
5282 /* there's a previous state to backtrack to */
40a82448
DM
5283 st--;
5284 if (st < SLAB_FIRST(PL_regmatch_slab)) {
5285 PL_regmatch_slab = PL_regmatch_slab->prev;
5286 st = SLAB_LAST(PL_regmatch_slab);
5287 }
5288 PL_regmatch_state = st;
40a82448
DM
5289 locinput= st->locinput;
5290 nextchr = UCHARAT(locinput);
5291
5bc10b2c
DM
5292 DEBUG_STATE_pp("pop");
5293 depth--;
262b90c4
DM
5294 if (yes_state == st)
5295 yes_state = st->u.yes.prev_yes_state;
5bc10b2c 5296
24d3c4a9
DM
5297 state_num = st->resume_state + 1; /* failure = success + 1 */
5298 goto reenter_switch;
95b24440 5299 }
24d3c4a9 5300 result = 0;
aa283a38 5301
262b90c4 5302 final_exit:
bbe252da 5303 if (rex->intflags & PREGf_VERBARG_SEEN) {
5d458dd8
YO
5304 SV *sv_err = get_sv("REGERROR", 1);
5305 SV *sv_mrk = get_sv("REGMARK", 1);
5306 if (result) {
e2e6a0f1 5307 sv_commit = &PL_sv_no;
5d458dd8
YO
5308 if (!sv_yes_mark)
5309 sv_yes_mark = &PL_sv_yes;
5310 } else {
5311 if (!sv_commit)
5312 sv_commit = &PL_sv_yes;
5313 sv_yes_mark = &PL_sv_no;
5314 }
5315 sv_setsv(sv_err, sv_commit);
5316 sv_setsv(sv_mrk, sv_yes_mark);
e2e6a0f1 5317 }
19b95bf0 5318
2f554ef7
DM
5319 /* clean up; in particular, free all slabs above current one */
5320 LEAVE_SCOPE(oldsave);
5d9a96ca 5321
95b24440 5322 return result;
a687059c
LW
5323}
5324
5325/*
5326 - regrepeat - repeatedly match something simple, report how many
5327 */
5328/*
5329 * [This routine now assumes that it will only match on things of length 1.
5330 * That was true before, but now we assume scan - reginput is the count,
a0ed51b3 5331 * rather than incrementing count on every character. [Er, except utf8.]]
a687059c 5332 */
76e3520e 5333STATIC I32
e2e6a0f1 5334S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth)
a687059c 5335{
27da23d5 5336 dVAR;
a0d0e21e 5337 register char *scan;
a0d0e21e 5338 register I32 c;
3280af22 5339 register char *loceol = PL_regeol;
a0ed51b3 5340 register I32 hardcount = 0;
53c4c00c 5341 register bool do_utf8 = PL_reg_match_utf8;
4f55667c
SP
5342#ifndef DEBUGGING
5343 PERL_UNUSED_ARG(depth);
5344#endif
a0d0e21e 5345
3280af22 5346 scan = PL_reginput;
faf11cac
HS
5347 if (max == REG_INFTY)
5348 max = I32_MAX;
5349 else if (max < loceol - scan)
7f596f4c 5350 loceol = scan + max;
a0d0e21e 5351 switch (OP(p)) {
22c35a8c 5352 case REG_ANY:
1aa99e6b 5353 if (do_utf8) {
ffc61ed2 5354 loceol = PL_regeol;
1aa99e6b 5355 while (scan < loceol && hardcount < max && *scan != '\n') {
ffc61ed2
JH
5356 scan += UTF8SKIP(scan);
5357 hardcount++;
5358 }
5359 } else {
5360 while (scan < loceol && *scan != '\n')
5361 scan++;
a0ed51b3
LW
5362 }
5363 break;
ffc61ed2 5364 case SANY:
def8e4ea
JH
5365 if (do_utf8) {
5366 loceol = PL_regeol;
a0804c9e 5367 while (scan < loceol && hardcount < max) {
def8e4ea
JH
5368 scan += UTF8SKIP(scan);
5369 hardcount++;
5370 }
5371 }
5372 else
5373 scan = loceol;
a0ed51b3 5374 break;
f33976b4
DB
5375 case CANY:
5376 scan = loceol;
5377 break;
090f7165
JH
5378 case EXACT: /* length of string is 1 */
5379 c = (U8)*STRING(p);
5380 while (scan < loceol && UCHARAT(scan) == c)
5381 scan++;
bbce6d69 5382 break;
5383 case EXACTF: /* length of string is 1 */
cd439c50 5384 c = (U8)*STRING(p);
bbce6d69 5385 while (scan < loceol &&
22c35a8c 5386 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
bbce6d69 5387 scan++;
5388 break;
5389 case EXACTFL: /* length of string is 1 */
3280af22 5390 PL_reg_flags |= RF_tainted;
cd439c50 5391 c = (U8)*STRING(p);
bbce6d69 5392 while (scan < loceol &&
22c35a8c 5393 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
a0d0e21e
LW
5394 scan++;
5395 break;
5396 case ANYOF:
ffc61ed2
JH
5397 if (do_utf8) {
5398 loceol = PL_regeol;
cfc92286 5399 while (hardcount < max && scan < loceol &&
32fc9b6a 5400 reginclass(prog, p, (U8*)scan, 0, do_utf8)) {
ffc61ed2
JH
5401 scan += UTF8SKIP(scan);
5402 hardcount++;
5403 }
5404 } else {
32fc9b6a 5405 while (scan < loceol && REGINCLASS(prog, p, (U8*)scan))
ffc61ed2
JH
5406 scan++;
5407 }
a0d0e21e
LW
5408 break;
5409 case ALNUM:
1aa99e6b 5410 if (do_utf8) {
ffc61ed2 5411 loceol = PL_regeol;
1a4fad37 5412 LOAD_UTF8_CHARCLASS_ALNUM();
1aa99e6b 5413 while (hardcount < max && scan < loceol &&
3568d838 5414 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
ffc61ed2
JH
5415 scan += UTF8SKIP(scan);
5416 hardcount++;
5417 }
5418 } else {
5419 while (scan < loceol && isALNUM(*scan))
5420 scan++;
a0ed51b3
LW
5421 }
5422 break;
bbce6d69 5423 case ALNUML:
3280af22 5424 PL_reg_flags |= RF_tainted;
1aa99e6b 5425 if (do_utf8) {
ffc61ed2 5426 loceol = PL_regeol;
1aa99e6b
IH
5427 while (hardcount < max && scan < loceol &&
5428 isALNUM_LC_utf8((U8*)scan)) {
ffc61ed2
JH
5429 scan += UTF8SKIP(scan);
5430 hardcount++;
5431 }
5432 } else {
5433 while (scan < loceol && isALNUM_LC(*scan))
5434 scan++;
a0ed51b3
LW
5435 }
5436 break;
a0d0e21e 5437 case NALNUM:
1aa99e6b 5438 if (do_utf8) {
ffc61ed2 5439 loceol = PL_regeol;
1a4fad37 5440 LOAD_UTF8_CHARCLASS_ALNUM();
1aa99e6b 5441 while (hardcount < max && scan < loceol &&
3568d838 5442 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
ffc61ed2
JH
5443 scan += UTF8SKIP(scan);
5444 hardcount++;
5445 }
5446 } else {
5447 while (scan < loceol && !isALNUM(*scan))
5448 scan++;
a0ed51b3
LW
5449 }
5450 break;
bbce6d69 5451 case NALNUML:
3280af22 5452 PL_reg_flags |= RF_tainted;
1aa99e6b 5453 if (do_utf8) {
ffc61ed2 5454 loceol = PL_regeol;
1aa99e6b
IH
5455 while (hardcount < max && scan < loceol &&
5456 !isALNUM_LC_utf8((U8*)scan)) {
ffc61ed2
JH
5457 scan += UTF8SKIP(scan);
5458 hardcount++;
5459 }
5460 } else {
5461 while (scan < loceol && !isALNUM_LC(*scan))
5462 scan++;
a0ed51b3
LW
5463 }
5464 break;
a0d0e21e 5465 case SPACE:
1aa99e6b 5466 if (do_utf8) {
ffc61ed2 5467 loceol = PL_regeol;
1a4fad37 5468 LOAD_UTF8_CHARCLASS_SPACE();
1aa99e6b 5469 while (hardcount < max && scan < loceol &&
3568d838
JH
5470 (*scan == ' ' ||
5471 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
ffc61ed2
JH
5472 scan += UTF8SKIP(scan);
5473 hardcount++;
5474 }
5475 } else {
5476 while (scan < loceol && isSPACE(*scan))
5477 scan++;
a0ed51b3
LW
5478 }
5479 break;
bbce6d69 5480 case SPACEL:
3280af22 5481 PL_reg_flags |= RF_tainted;
1aa99e6b 5482 if (do_utf8) {
ffc61ed2 5483 loceol = PL_regeol;
1aa99e6b 5484 while (hardcount < max && scan < loceol &&
ffc61ed2
JH
5485 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
5486 scan += UTF8SKIP(scan);
5487 hardcount++;
5488 }
5489 } else {
5490 while (scan < loceol && isSPACE_LC(*scan))
5491 scan++;
a0ed51b3
LW
5492 }
5493 break;
a0d0e21e 5494 case NSPACE:
1aa99e6b 5495 if (do_utf8) {
ffc61ed2 5496 loceol = PL_regeol;
1a4fad37 5497 LOAD_UTF8_CHARCLASS_SPACE();
1aa99e6b 5498 while (hardcount < max && scan < loceol &&
3568d838
JH
5499 !(*scan == ' ' ||
5500 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
ffc61ed2
JH
5501 scan += UTF8SKIP(scan);
5502 hardcount++;
5503 }
5504 } else {
5505 while (scan < loceol && !isSPACE(*scan))
5506 scan++;
a0ed51b3 5507 }
0008a298 5508 break;
bbce6d69 5509 case NSPACEL:
3280af22 5510 PL_reg_flags |= RF_tainted;
1aa99e6b 5511 if (do_utf8) {
ffc61ed2 5512 loceol = PL_regeol;
1aa99e6b 5513 while (hardcount < max && scan < loceol &&
ffc61ed2
JH
5514 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
5515 scan += UTF8SKIP(scan);
5516 hardcount++;
5517 }
5518 } else {
5519 while (scan < loceol && !isSPACE_LC(*scan))
5520 scan++;
a0ed51b3
LW
5521 }
5522 break;
a0d0e21e 5523 case DIGIT:
1aa99e6b 5524 if (do_utf8) {
ffc61ed2 5525 loceol = PL_regeol;
1a4fad37 5526 LOAD_UTF8_CHARCLASS_DIGIT();
1aa99e6b 5527 while (hardcount < max && scan < loceol &&
3568d838 5528 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
ffc61ed2
JH
5529 scan += UTF8SKIP(scan);
5530 hardcount++;
5531 }
5532 } else {
5533 while (scan < loceol && isDIGIT(*scan))
5534 scan++;
a0ed51b3
LW
5535 }
5536 break;
a0d0e21e 5537 case NDIGIT:
1aa99e6b 5538 if (do_utf8) {
ffc61ed2 5539 loceol = PL_regeol;
1a4fad37 5540 LOAD_UTF8_CHARCLASS_DIGIT();
1aa99e6b 5541 while (hardcount < max && scan < loceol &&
3568d838 5542 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
ffc61ed2
JH
5543 scan += UTF8SKIP(scan);
5544 hardcount++;
5545 }
5546 } else {
5547 while (scan < loceol && !isDIGIT(*scan))
5548 scan++;
a0ed51b3 5549 }
e1d1eefb
YO
5550 case LNBREAK:
5551 if (do_utf8) {
5552 loceol = PL_regeol;
5553 while (hardcount < max && scan < loceol && (c=is_LNBREAK_utf8(scan))) {
5554 scan += c;
5555 hardcount++;
5556 }
5557 } else {
5558 /*
5559 LNBREAK can match two latin chars, which is ok,
5560 because we have a null terminated string, but we
5561 have to use hardcount in this situation
5562 */
5563 while (scan < loceol && (c=is_LNBREAK_latin1(scan))) {
5564 scan+=c;
5565 hardcount++;
5566 }
5567 }
5568 break;
5569 case HORIZWS:
5570 if (do_utf8) {
5571 loceol = PL_regeol;
5572 while (hardcount < max && scan < loceol && (c=is_HORIZWS_utf8(scan))) {
5573 scan += c;
5574 hardcount++;
5575 }
5576 } else {
5577 while (scan < loceol && is_HORIZWS_latin1(scan))
5578 scan++;
5579 }
a0ed51b3 5580 break;
e1d1eefb
YO
5581 case NHORIZWS:
5582 if (do_utf8) {
5583 loceol = PL_regeol;
5584 while (hardcount < max && scan < loceol && !is_HORIZWS_utf8(scan)) {
5585 scan += UTF8SKIP(scan);
5586 hardcount++;
5587 }
5588 } else {
5589 while (scan < loceol && !is_HORIZWS_latin1(scan))
5590 scan++;
5591
5592 }
5593 break;
5594 case VERTWS:
5595 if (do_utf8) {
5596 loceol = PL_regeol;
5597 while (hardcount < max && scan < loceol && (c=is_VERTWS_utf8(scan))) {
5598 scan += c;
5599 hardcount++;
5600 }
5601 } else {
5602 while (scan < loceol && is_VERTWS_latin1(scan))
5603 scan++;
5604
5605 }
5606 break;
5607 case NVERTWS:
5608 if (do_utf8) {
5609 loceol = PL_regeol;
5610 while (hardcount < max && scan < loceol && !is_VERTWS_utf8(scan)) {
5611 scan += UTF8SKIP(scan);
5612 hardcount++;
5613 }
5614 } else {
5615 while (scan < loceol && !is_VERTWS_latin1(scan))
5616 scan++;
5617
5618 }
5619 break;
5620
a0d0e21e
LW
5621 default: /* Called on something of 0 width. */
5622 break; /* So match right here or not at all. */
5623 }
a687059c 5624
a0ed51b3
LW
5625 if (hardcount)
5626 c = hardcount;
5627 else
5628 c = scan - PL_reginput;
3280af22 5629 PL_reginput = scan;
a687059c 5630
a3621e74 5631 DEBUG_r({
e68ec53f 5632 GET_RE_DEBUG_FLAGS_DECL;
be8e71aa 5633 DEBUG_EXECUTE_r({
e68ec53f
YO
5634 SV * const prop = sv_newmortal();
5635 regprop(prog, prop, p);
5636 PerlIO_printf(Perl_debug_log,
be8e71aa 5637 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
e2e6a0f1 5638 REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max);
a3621e74 5639 });
be8e71aa 5640 });
9041c2e3 5641
a0d0e21e 5642 return(c);
a687059c
LW
5643}
5644
c277df42 5645
be8e71aa 5646#if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION)
c277df42 5647/*
ffc61ed2
JH
5648- regclass_swash - prepare the utf8 swash
5649*/
5650
5651SV *
32fc9b6a 5652Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
ffc61ed2 5653{
97aff369 5654 dVAR;
9e55ce06
JH
5655 SV *sw = NULL;
5656 SV *si = NULL;
5657 SV *alt = NULL;
f8fc2ecf
YO
5658 RXi_GET_DECL(prog,progi);
5659 const struct reg_data * const data = prog ? progi->data : NULL;
ffc61ed2 5660
4f639d21 5661 if (data && data->count) {
a3b680e6 5662 const U32 n = ARG(node);
ffc61ed2 5663
4f639d21
DM
5664 if (data->what[n] == 's') {
5665 SV * const rv = (SV*)data->data[n];
890ce7af 5666 AV * const av = (AV*)SvRV((SV*)rv);
2d03de9c 5667 SV **const ary = AvARRAY(av);
9e55ce06 5668 SV **a, **b;
9041c2e3 5669
711a919c 5670 /* See the end of regcomp.c:S_regclass() for
9e55ce06
JH
5671 * documentation of these array elements. */
5672
b11f357e 5673 si = *ary;
fe5bfecd
JH
5674 a = SvROK(ary[1]) ? &ary[1] : NULL;
5675 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : NULL;
b11f357e 5676
ffc61ed2
JH
5677 if (a)
5678 sw = *a;
5679 else if (si && doinit) {
5680 sw = swash_init("utf8", "", si, 1, 0);
5681 (void)av_store(av, 1, sw);
5682 }
9e55ce06
JH
5683 if (b)
5684 alt = *b;
ffc61ed2
JH
5685 }
5686 }
5687
9e55ce06
JH
5688 if (listsvp)
5689 *listsvp = si;
5690 if (altsvp)
5691 *altsvp = alt;
ffc61ed2
JH
5692
5693 return sw;
5694}
76234dfb 5695#endif
ffc61ed2
JH
5696
5697/*
ba7b4546 5698 - reginclass - determine if a character falls into a character class
832705d4
JH
5699
5700 The n is the ANYOF regnode, the p is the target string, lenp
5701 is pointer to the maximum length of how far to go in the p
5702 (if the lenp is zero, UTF8SKIP(p) is used),
5703 do_utf8 tells whether the target string is in UTF-8.
5704
bbce6d69 5705 */
5706
76e3520e 5707STATIC bool
32fc9b6a 5708S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8)
bbce6d69 5709{
27da23d5 5710 dVAR;
a3b680e6 5711 const char flags = ANYOF_FLAGS(n);
bbce6d69 5712 bool match = FALSE;
cc07378b 5713 UV c = *p;
ae9ddab8 5714 STRLEN len = 0;
9e55ce06 5715 STRLEN plen;
1aa99e6b 5716
19f67299
TS
5717 if (do_utf8 && !UTF8_IS_INVARIANT(c)) {
5718 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
4ad0818d
DM
5719 (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) | UTF8_CHECK_ONLY);
5720 /* see [perl #37836] for UTF8_ALLOW_ANYUV */
e8a70c6f
SP
5721 if (len == (STRLEN)-1)
5722 Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)");
19f67299 5723 }
bbce6d69 5724
0f0076b4 5725 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
ffc61ed2 5726 if (do_utf8 || (flags & ANYOF_UNICODE)) {
9e55ce06
JH
5727 if (lenp)
5728 *lenp = 0;
ffc61ed2 5729 if (do_utf8 && !ANYOF_RUNTIME(n)) {
ffc61ed2
JH
5730 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
5731 match = TRUE;
bbce6d69 5732 }
3568d838 5733 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
1aa99e6b 5734 match = TRUE;
ffc61ed2 5735 if (!match) {
9e55ce06 5736 AV *av;
32fc9b6a 5737 SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av);
ffc61ed2
JH
5738
5739 if (sw) {
3568d838 5740 if (swash_fetch(sw, p, do_utf8))
ffc61ed2
JH
5741 match = TRUE;
5742 else if (flags & ANYOF_FOLD) {
9e55ce06
JH
5743 if (!match && lenp && av) {
5744 I32 i;
9e55ce06 5745 for (i = 0; i <= av_len(av); i++) {
890ce7af 5746 SV* const sv = *av_fetch(av, i, FALSE);
9e55ce06 5747 STRLEN len;
890ce7af 5748 const char * const s = SvPV_const(sv, len);
9e55ce06 5749
061b10df 5750 if (len <= plen && memEQ(s, (char*)p, len)) {
9e55ce06
JH
5751 *lenp = len;
5752 match = TRUE;
5753 break;
5754 }
5755 }
5756 }
5757 if (!match) {
89ebb4a3 5758 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
4a623e43
JH
5759 STRLEN tmplen;
5760
9e55ce06
JH
5761 to_utf8_fold(p, tmpbuf, &tmplen);
5762 if (swash_fetch(sw, tmpbuf, do_utf8))
5763 match = TRUE;
5764 }
ffc61ed2
JH
5765 }
5766 }
bbce6d69 5767 }
9e55ce06 5768 if (match && lenp && *lenp == 0)
0f0076b4 5769 *lenp = UNISKIP(NATIVE_TO_UNI(c));
bbce6d69 5770 }
1aa99e6b 5771 if (!match && c < 256) {
ffc61ed2
JH
5772 if (ANYOF_BITMAP_TEST(n, c))
5773 match = TRUE;
5774 else if (flags & ANYOF_FOLD) {
eb160463 5775 U8 f;
a0ed51b3 5776
ffc61ed2
JH
5777 if (flags & ANYOF_LOCALE) {
5778 PL_reg_flags |= RF_tainted;
5779 f = PL_fold_locale[c];
5780 }
5781 else
5782 f = PL_fold[c];
5783 if (f != c && ANYOF_BITMAP_TEST(n, f))
5784 match = TRUE;
5785 }
5786
5787 if (!match && (flags & ANYOF_CLASS)) {
a0ed51b3 5788 PL_reg_flags |= RF_tainted;
ffc61ed2
JH
5789 if (
5790 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
5791 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
5792 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
5793 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
5794 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
5795 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
5796 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
5797 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
5798 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
5799 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
5800 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
5801 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
5802 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
5803 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
5804 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
5805 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
5806 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
5807 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
5808 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
5809 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
5810 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
5811 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
5812 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
5813 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
5814 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
5815 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
5816 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
5817 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
5818 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
5819 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
5820 ) /* How's that for a conditional? */
5821 {
5822 match = TRUE;
5823 }
a0ed51b3 5824 }
a0ed51b3
LW
5825 }
5826
a0ed51b3
LW
5827 return (flags & ANYOF_INVERT) ? !match : match;
5828}
161b471a 5829
dfe13c55 5830STATIC U8 *
0ce71af7 5831S_reghop3(U8 *s, I32 off, const U8* lim)
9041c2e3 5832{
97aff369 5833 dVAR;
a0ed51b3 5834 if (off >= 0) {
1aa99e6b 5835 while (off-- && s < lim) {
ffc61ed2 5836 /* XXX could check well-formedness here */
a0ed51b3 5837 s += UTF8SKIP(s);
ffc61ed2 5838 }
a0ed51b3
LW
5839 }
5840 else {
1de06328
YO
5841 while (off++ && s > lim) {
5842 s--;
5843 if (UTF8_IS_CONTINUED(*s)) {
5844 while (s > lim && UTF8_IS_CONTINUATION(*s))
5845 s--;
a0ed51b3 5846 }
1de06328 5847 /* XXX could check well-formedness here */
a0ed51b3
LW
5848 }
5849 }
5850 return s;
5851}
161b471a 5852
f9f4320a
YO
5853#ifdef XXX_dmq
5854/* there are a bunch of places where we use two reghop3's that should
5855 be replaced with this routine. but since thats not done yet
5856 we ifdef it out - dmq
5857*/
dfe13c55 5858STATIC U8 *
1de06328
YO
5859S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim)
5860{
5861 dVAR;
5862 if (off >= 0) {
5863 while (off-- && s < rlim) {
5864 /* XXX could check well-formedness here */
5865 s += UTF8SKIP(s);
5866 }
5867 }
5868 else {
5869 while (off++ && s > llim) {
5870 s--;
5871 if (UTF8_IS_CONTINUED(*s)) {
5872 while (s > llim && UTF8_IS_CONTINUATION(*s))
5873 s--;
5874 }
5875 /* XXX could check well-formedness here */
5876 }
5877 }
5878 return s;
5879}
f9f4320a 5880#endif
1de06328
YO
5881
5882STATIC U8 *
0ce71af7 5883S_reghopmaybe3(U8* s, I32 off, const U8* lim)
a0ed51b3 5884{
97aff369 5885 dVAR;
a0ed51b3 5886 if (off >= 0) {
1aa99e6b 5887 while (off-- && s < lim) {
ffc61ed2 5888 /* XXX could check well-formedness here */
a0ed51b3 5889 s += UTF8SKIP(s);
ffc61ed2 5890 }
a0ed51b3 5891 if (off >= 0)
3dab1dad 5892 return NULL;
a0ed51b3
LW
5893 }
5894 else {
1de06328
YO
5895 while (off++ && s > lim) {
5896 s--;
5897 if (UTF8_IS_CONTINUED(*s)) {
5898 while (s > lim && UTF8_IS_CONTINUATION(*s))
5899 s--;
a0ed51b3 5900 }
1de06328 5901 /* XXX could check well-formedness here */
a0ed51b3
LW
5902 }
5903 if (off <= 0)
3dab1dad 5904 return NULL;
a0ed51b3
LW
5905 }
5906 return s;
5907}
51371543 5908
51371543 5909static void
acfe0abc 5910restore_pos(pTHX_ void *arg)
51371543 5911{
97aff369 5912 dVAR;
097eb12c 5913 regexp * const rex = (regexp *)arg;
51371543
GS
5914 if (PL_reg_eval_set) {
5915 if (PL_reg_oldsaved) {
4f639d21
DM
5916 rex->subbeg = PL_reg_oldsaved;
5917 rex->sublen = PL_reg_oldsavedlen;
f8c7b90f 5918#ifdef PERL_OLD_COPY_ON_WRITE
4f639d21 5919 rex->saved_copy = PL_nrs;
ed252734 5920#endif
07bc277f 5921 RXp_MATCH_COPIED_on(rex);
51371543
GS
5922 }
5923 PL_reg_magic->mg_len = PL_reg_oldpos;
5924 PL_reg_eval_set = 0;
5925 PL_curpm = PL_reg_oldcurpm;
5926 }
5927}
33b8afdf
JH
5928
5929STATIC void
5930S_to_utf8_substr(pTHX_ register regexp *prog)
5931{
a1cac82e
NC
5932 int i = 1;
5933 do {
5934 if (prog->substrs->data[i].substr
5935 && !prog->substrs->data[i].utf8_substr) {
5936 SV* const sv = newSVsv(prog->substrs->data[i].substr);
5937 prog->substrs->data[i].utf8_substr = sv;
5938 sv_utf8_upgrade(sv);
610460f9
NC
5939 if (SvVALID(prog->substrs->data[i].substr)) {
5940 const U8 flags = BmFLAGS(prog->substrs->data[i].substr);
5941 if (flags & FBMcf_TAIL) {
5942 /* Trim the trailing \n that fbm_compile added last
5943 time. */
5944 SvCUR_set(sv, SvCUR(sv) - 1);
5945 /* Whilst this makes the SV technically "invalid" (as its
5946 buffer is no longer followed by "\0") when fbm_compile()
5947 adds the "\n" back, a "\0" is restored. */
5948 }
5949 fbm_compile(sv, flags);
5950 }
a1cac82e
NC
5951 if (prog->substrs->data[i].substr == prog->check_substr)
5952 prog->check_utf8 = sv;
5953 }
5954 } while (i--);
33b8afdf
JH
5955}
5956
5957STATIC void
5958S_to_byte_substr(pTHX_ register regexp *prog)
5959{
97aff369 5960 dVAR;
a1cac82e
NC
5961 int i = 1;
5962 do {
5963 if (prog->substrs->data[i].utf8_substr
5964 && !prog->substrs->data[i].substr) {
5965 SV* sv = newSVsv(prog->substrs->data[i].utf8_substr);
5966 if (sv_utf8_downgrade(sv, TRUE)) {
610460f9
NC
5967 if (SvVALID(prog->substrs->data[i].utf8_substr)) {
5968 const U8 flags
5969 = BmFLAGS(prog->substrs->data[i].utf8_substr);
5970 if (flags & FBMcf_TAIL) {
5971 /* Trim the trailing \n that fbm_compile added last
5972 time. */
5973 SvCUR_set(sv, SvCUR(sv) - 1);
5974 }
5975 fbm_compile(sv, flags);
5976 }
a1cac82e
NC
5977 } else {
5978 SvREFCNT_dec(sv);
5979 sv = &PL_sv_undef;
5980 }
5981 prog->substrs->data[i].substr = sv;
5982 if (prog->substrs->data[i].utf8_substr == prog->check_utf8)
5983 prog->check_substr = sv;
33b8afdf 5984 }
a1cac82e 5985 } while (i--);
33b8afdf 5986}
66610fdd
RGS
5987
5988/*
5989 * Local variables:
5990 * c-indentation-style: bsd
5991 * c-basic-offset: 4
5992 * indent-tabs-mode: t
5993 * End:
5994 *
37442d52
RGS
5995 * ex: set ts=8 sts=4 sw=4 noet:
5996 */