5 * "One Ring to rule them all, One Ring to find them..."
8 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
9 * confused with the original package (see point 3 below). Thanks, Henry!
12 /* Additional note: this code is very heavily munged from Henry's version
13 * in places. In some spots I've traded clarity for efficiency, so don't
14 * blame Henry for some of the lack of readability.
17 /* The names of the functions have been changed from regcomp and
18 * regexec to pregcomp and pregexec in order to avoid conflicts
19 * with the POSIX routines of the same names.
22 #ifdef PERL_EXT_RE_BUILD
23 /* need to replace pregcomp et al, so enable that */
24 # ifndef PERL_IN_XSUB_RE
25 # define PERL_IN_XSUB_RE
27 /* need access to debugger hooks */
28 # if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
33 #ifdef PERL_IN_XSUB_RE
34 /* We *really* need to overwrite these symbols: */
35 # define Perl_regexec_flags my_regexec
36 # define Perl_regdump my_regdump
37 # define Perl_regprop my_regprop
38 # define Perl_re_intuit_start my_re_intuit_start
39 /* *These* symbols are masked to allow static link. */
40 # define Perl_pregexec my_pregexec
41 # define Perl_reginitcolors my_reginitcolors
42 # define Perl_regclass_swash my_regclass_swash
44 # define PERL_NO_GET_CONTEXT
49 * pregcomp and pregexec -- regsub and regerror are not used in perl
51 * Copyright (c) 1986 by University of Toronto.
52 * Written by Henry Spencer. Not derived from licensed software.
54 * Permission is granted to anyone to use this software for any
55 * purpose on any computer system, and to redistribute it freely,
56 * subject to the following restrictions:
58 * 1. The author is not responsible for the consequences of use of
59 * this software, no matter how awful, even if they arise
62 * 2. The origin of this software must not be misrepresented, either
63 * by explicit claim or by omission.
65 * 3. Altered versions must be plainly marked as such, and must not
66 * be misrepresented as being the original software.
68 **** Alterations to Henry's code are...
70 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
71 **** 2000, 2001, 2002, 2003, 2004, by Larry Wall and others
73 **** You may distribute under the terms of either the GNU General Public
74 **** License or the Artistic License, as specified in the README file.
76 * Beware that some of this code is subtly aware of the way operator
77 * precedence is structured in regular expressions. Serious changes in
78 * regular-expression syntax might require a total rethink.
81 #define PERL_IN_REGEXEC_C
86 #define RF_tainted 1 /* tainted information used? */
87 #define RF_warned 2 /* warned about big count? */
88 #define RF_evaled 4 /* Did an EVAL with setting? */
89 #define RF_utf8 8 /* String contains multibyte chars? */
90 #define RF_false 16 /* odd number of nested negatives */
92 #define UTF ((PL_reg_flags & RF_utf8) != 0)
94 #define RS_init 1 /* eval environment created */
95 #define RS_set 2 /* replsv value is set */
101 #define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
107 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
108 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
110 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
111 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
112 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
113 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
114 #define HOPc(pos,off) ((char*)HOP(pos,off))
115 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
117 #define HOPBACK(pos, off) ( \
118 (PL_reg_match_utf8) \
119 ? reghopmaybe((U8*)pos, -off) \
120 : (pos - off >= PL_bostr) \
124 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
126 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
127 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
128 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
129 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
130 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
131 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
133 #define LOAD_UTF8_CHARCLASS(a,b) STMT_START { if (!CAT2(PL_utf8_,a)) { ENTER; save_re_context(); (void)CAT2(is_utf8_, a)((U8*)b); LEAVE; } } STMT_END
135 /* for use after a quantifier and before an EXACT-like node -- japhy */
136 #define JUMPABLE(rn) ( \
137 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
138 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
139 OP(rn) == PLUS || OP(rn) == MINMOD || \
140 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
143 #define HAS_TEXT(rn) ( \
144 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
148 Search for mandatory following text node; for lookahead, the text must
149 follow but for lookbehind (rn->flags != 0) we skip to the next step.
151 #define FIND_NEXT_IMPT(rn) STMT_START { \
152 while (JUMPABLE(rn)) \
153 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
154 rn = NEXTOPER(NEXTOPER(rn)); \
155 else if (OP(rn) == PLUS) \
157 else if (OP(rn) == IFMATCH) \
158 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
159 else rn += NEXT_OFF(rn); \
162 static void restore_pos(pTHX_ void *arg);
165 S_regcppush(pTHX_ I32 parenfloor)
167 int retval = PL_savestack_ix;
168 #define REGCP_PAREN_ELEMS 4
169 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
172 if (paren_elems_to_push < 0)
173 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
175 #define REGCP_OTHER_ELEMS 6
176 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
177 for (p = PL_regsize; p > parenfloor; p--) {
178 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
179 SSPUSHINT(PL_regendp[p]);
180 SSPUSHINT(PL_regstartp[p]);
181 SSPUSHPTR(PL_reg_start_tmp[p]);
184 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
185 SSPUSHINT(PL_regsize);
186 SSPUSHINT(*PL_reglastparen);
187 SSPUSHINT(*PL_reglastcloseparen);
188 SSPUSHPTR(PL_reginput);
189 #define REGCP_FRAME_ELEMS 2
190 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
191 * are needed for the regexp context stack bookkeeping. */
192 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
193 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
198 /* These are needed since we do not localize EVAL nodes: */
199 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
200 " Setting an EVAL scope, savestack=%"IVdf"\n", \
201 (IV)PL_savestack_ix)); cp = PL_savestack_ix
203 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
204 PerlIO_printf(Perl_debug_log, \
205 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
206 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
216 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
218 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
219 i = SSPOPINT; /* Parentheses elements to pop. */
220 input = (char *) SSPOPPTR;
221 *PL_reglastcloseparen = SSPOPINT;
222 *PL_reglastparen = SSPOPINT;
223 PL_regsize = SSPOPINT;
225 /* Now restore the parentheses context. */
226 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
227 i > 0; i -= REGCP_PAREN_ELEMS) {
228 paren = (U32)SSPOPINT;
229 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
230 PL_regstartp[paren] = SSPOPINT;
232 if (paren <= *PL_reglastparen)
233 PL_regendp[paren] = tmps;
235 PerlIO_printf(Perl_debug_log,
236 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
237 (UV)paren, (IV)PL_regstartp[paren],
238 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
239 (IV)PL_regendp[paren],
240 (paren > *PL_reglastparen ? "(no)" : ""));
244 if ((I32)(*PL_reglastparen + 1) <= PL_regnpar) {
245 PerlIO_printf(Perl_debug_log,
246 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
247 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
251 /* It would seem that the similar code in regtry()
252 * already takes care of this, and in fact it is in
253 * a better location to since this code can #if 0-ed out
254 * but the code in regtry() is needed or otherwise tests
255 * requiring null fields (pat.t#187 and split.t#{13,14}
256 * (as of patchlevel 7877) will fail. Then again,
257 * this code seems to be necessary or otherwise
258 * building DynaLoader will fail:
259 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
261 for (paren = *PL_reglastparen + 1; (I32)paren <= PL_regnpar; paren++) {
262 if ((I32)paren > PL_regsize)
263 PL_regstartp[paren] = -1;
264 PL_regendp[paren] = -1;
271 S_regcp_set_to(pTHX_ I32 ss)
273 I32 tmp = PL_savestack_ix;
275 PL_savestack_ix = ss;
277 PL_savestack_ix = tmp;
281 typedef struct re_cc_state
285 struct re_cc_state *prev;
290 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
292 #define TRYPAREN(paren, n, input) { \
295 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
296 PL_regendp[paren] = input - PL_bostr; \
299 PL_regendp[paren] = -1; \
301 if (regmatch(next)) \
304 PL_regendp[paren] = -1; \
309 * pregexec and friends
313 - pregexec - match a regexp against a string
316 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
317 char *strbeg, I32 minend, SV *screamer, U32 nosave)
318 /* strend: pointer to null at end of string */
319 /* strbeg: real beginning of string */
320 /* minend: end of match must be >=minend after stringarg. */
321 /* nosave: For optimizations. */
324 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
325 nosave ? 0 : REXEC_COPY_STR);
329 S_cache_re(pTHX_ regexp *prog)
331 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
333 PL_regprogram = prog->program;
335 PL_regnpar = prog->nparens;
336 PL_regdata = prog->data;
341 * Need to implement the following flags for reg_anch:
343 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
345 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
346 * INTUIT_AUTORITATIVE_ML
347 * INTUIT_ONCE_NOML - Intuit can match in one location only.
350 * Another flag for this function: SECOND_TIME (so that float substrs
351 * with giant delta may be not rechecked).
354 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
356 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
357 Otherwise, only SvCUR(sv) is used to get strbeg. */
359 /* XXXX We assume that strpos is strbeg unless sv. */
361 /* XXXX Some places assume that there is a fixed substring.
362 An update may be needed if optimizer marks as "INTUITable"
363 RExen without fixed substrings. Similarly, it is assumed that
364 lengths of all the strings are no more than minlen, thus they
365 cannot come from lookahead.
366 (Or minlen should take into account lookahead.) */
368 /* A failure to find a constant substring means that there is no need to make
369 an expensive call to REx engine, thus we celebrate a failure. Similarly,
370 finding a substring too deep into the string means that less calls to
371 regtry() should be needed.
373 REx compiler's optimizer found 4 possible hints:
374 a) Anchored substring;
376 c) Whether we are anchored (beginning-of-line or \G);
377 d) First node (of those at offset 0) which may distingush positions;
378 We use a)b)d) and multiline-part of c), and try to find a position in the
379 string which does not contradict any of them.
382 /* Most of decisions we do here should have been done at compile time.
383 The nodes of the REx which we used for the search should have been
384 deleted from the finite automaton. */
387 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
388 char *strend, U32 flags, re_scream_pos_data *data)
390 register I32 start_shift = 0;
391 /* Should be nonnegative! */
392 register I32 end_shift = 0;
397 int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
399 register char *other_last = Nullch; /* other substr checked before this */
400 char *check_at = Nullch; /* check substr found at this pos */
402 char *i_strpos = strpos;
403 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
405 RX_MATCH_UTF8_set(prog,do_utf8);
407 if (prog->reganch & ROPT_UTF8) {
408 DEBUG_r(PerlIO_printf(Perl_debug_log,
409 "UTF-8 regex...\n"));
410 PL_reg_flags |= RF_utf8;
414 char *s = PL_reg_match_utf8 ?
415 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
417 int len = PL_reg_match_utf8 ?
418 strlen(s) : strend - strpos;
421 if (PL_reg_match_utf8)
422 DEBUG_r(PerlIO_printf(Perl_debug_log,
423 "UTF-8 target...\n"));
424 PerlIO_printf(Perl_debug_log,
425 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
426 PL_colors[4],PL_colors[5],PL_colors[0],
429 (strlen(prog->precomp) > 60 ? "..." : ""),
431 (int)(len > 60 ? 60 : len),
433 (len > 60 ? "..." : "")
437 /* CHR_DIST() would be more correct here but it makes things slow. */
438 if (prog->minlen > strend - strpos) {
439 DEBUG_r(PerlIO_printf(Perl_debug_log,
440 "String too short... [re_intuit_start]\n"));
443 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
446 if (!prog->check_utf8 && prog->check_substr)
447 to_utf8_substr(prog);
448 check = prog->check_utf8;
450 if (!prog->check_substr && prog->check_utf8)
451 to_byte_substr(prog);
452 check = prog->check_substr;
454 if (check == &PL_sv_undef) {
455 DEBUG_r(PerlIO_printf(Perl_debug_log,
456 "Non-utf string cannot match utf check string\n"));
459 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
460 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
461 || ( (prog->reganch & ROPT_ANCH_BOL)
462 && !PL_multiline ) ); /* Check after \n? */
465 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
466 | ROPT_IMPLICIT)) /* not a real BOL */
467 /* SvCUR is not set on references: SvRV and SvPVX overlap */
469 && (strpos != strbeg)) {
470 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
473 if (prog->check_offset_min == prog->check_offset_max &&
474 !(prog->reganch & ROPT_CANY_SEEN)) {
475 /* Substring at constant offset from beg-of-str... */
478 s = HOP3c(strpos, prog->check_offset_min, strend);
480 slen = SvCUR(check); /* >= 1 */
482 if ( strend - s > slen || strend - s < slen - 1
483 || (strend - s == slen && strend[-1] != '\n')) {
484 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
487 /* Now should match s[0..slen-2] */
489 if (slen && (*SvPVX(check) != *s
491 && memNE(SvPVX(check), s, slen)))) {
493 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
497 else if (*SvPVX(check) != *s
498 || ((slen = SvCUR(check)) > 1
499 && memNE(SvPVX(check), s, slen)))
501 goto success_at_start;
504 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
506 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
507 end_shift = prog->minlen - start_shift -
508 CHR_SVLEN(check) + (SvTAIL(check) != 0);
510 I32 end = prog->check_offset_max + CHR_SVLEN(check)
511 - (SvTAIL(check) != 0);
512 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
514 if (end_shift < eshift)
518 else { /* Can match at random position */
521 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
522 /* Should be nonnegative! */
523 end_shift = prog->minlen - start_shift -
524 CHR_SVLEN(check) + (SvTAIL(check) != 0);
527 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
529 Perl_croak(aTHX_ "panic: end_shift");
533 /* Find a possible match in the region s..strend by looking for
534 the "check" substring in the region corrected by start/end_shift. */
535 if (flags & REXEC_SCREAM) {
536 I32 p = -1; /* Internal iterator of scream. */
537 I32 *pp = data ? data->scream_pos : &p;
539 if (PL_screamfirst[BmRARE(check)] >= 0
540 || ( BmRARE(check) == '\n'
541 && (BmPREVIOUS(check) == SvCUR(check) - 1)
543 s = screaminstr(sv, check,
544 start_shift + (s - strbeg), end_shift, pp, 0);
547 /* we may be pointing at the wrong string */
548 if (s && RX_MATCH_COPIED(prog))
549 s = strbeg + (s - SvPVX(sv));
551 *data->scream_olds = s;
553 else if (prog->reganch & ROPT_CANY_SEEN)
554 s = fbm_instr((U8*)(s + start_shift),
555 (U8*)(strend - end_shift),
556 check, PL_multiline ? FBMrf_MULTILINE : 0);
558 s = fbm_instr(HOP3(s, start_shift, strend),
559 HOP3(strend, -end_shift, strbeg),
560 check, PL_multiline ? FBMrf_MULTILINE : 0);
562 /* Update the count-of-usability, remove useless subpatterns,
565 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
566 (s ? "Found" : "Did not find"),
567 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
569 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
571 PL_colors[1], (SvTAIL(check) ? "$" : ""),
572 (s ? " at offset " : "...\n") ) );
579 /* Finish the diagnostic message */
580 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
582 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
583 Start with the other substr.
584 XXXX no SCREAM optimization yet - and a very coarse implementation
585 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
586 *always* match. Probably should be marked during compile...
587 Probably it is right to do no SCREAM here...
590 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
591 /* Take into account the "other" substring. */
592 /* XXXX May be hopelessly wrong for UTF... */
595 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
598 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
602 t = s - prog->check_offset_max;
603 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
605 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
610 t = HOP3c(t, prog->anchored_offset, strend);
611 if (t < other_last) /* These positions already checked */
613 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
616 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
617 /* On end-of-str: see comment below. */
618 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
619 if (must == &PL_sv_undef) {
621 DEBUG_r(must = prog->anchored_utf8); /* for debug */
626 HOP3(HOP3(last1, prog->anchored_offset, strend)
627 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
629 PL_multiline ? FBMrf_MULTILINE : 0
631 DEBUG_r(PerlIO_printf(Perl_debug_log,
632 "%s anchored substr `%s%.*s%s'%s",
633 (s ? "Found" : "Contradicts"),
636 - (SvTAIL(must)!=0)),
638 PL_colors[1], (SvTAIL(must) ? "$" : "")));
640 if (last1 >= last2) {
641 DEBUG_r(PerlIO_printf(Perl_debug_log,
642 ", giving up...\n"));
645 DEBUG_r(PerlIO_printf(Perl_debug_log,
646 ", trying floating at offset %ld...\n",
647 (long)(HOP3c(s1, 1, strend) - i_strpos)));
648 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
649 s = HOP3c(last, 1, strend);
653 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
654 (long)(s - i_strpos)));
655 t = HOP3c(s, -prog->anchored_offset, strbeg);
656 other_last = HOP3c(s, 1, strend);
664 else { /* Take into account the floating substring. */
669 t = HOP3c(s, -start_shift, strbeg);
671 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
672 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
673 last = HOP3c(t, prog->float_max_offset, strend);
674 s = HOP3c(t, prog->float_min_offset, strend);
677 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
678 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
679 /* fbm_instr() takes into account exact value of end-of-str
680 if the check is SvTAIL(ed). Since false positives are OK,
681 and end-of-str is not later than strend we are OK. */
682 if (must == &PL_sv_undef) {
684 DEBUG_r(must = prog->float_utf8); /* for debug message */
687 s = fbm_instr((unsigned char*)s,
688 (unsigned char*)last + SvCUR(must)
690 must, PL_multiline ? FBMrf_MULTILINE : 0);
691 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
692 (s ? "Found" : "Contradicts"),
694 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
696 PL_colors[1], (SvTAIL(must) ? "$" : "")));
699 DEBUG_r(PerlIO_printf(Perl_debug_log,
700 ", giving up...\n"));
703 DEBUG_r(PerlIO_printf(Perl_debug_log,
704 ", trying anchored starting at offset %ld...\n",
705 (long)(s1 + 1 - i_strpos)));
707 s = HOP3c(t, 1, strend);
711 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
712 (long)(s - i_strpos)));
713 other_last = s; /* Fix this later. --Hugo */
722 t = s - prog->check_offset_max;
723 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
725 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
727 /* Fixed substring is found far enough so that the match
728 cannot start at strpos. */
730 if (ml_anch && t[-1] != '\n') {
731 /* Eventually fbm_*() should handle this, but often
732 anchored_offset is not 0, so this check will not be wasted. */
733 /* XXXX In the code below we prefer to look for "^" even in
734 presence of anchored substrings. And we search even
735 beyond the found float position. These pessimizations
736 are historical artefacts only. */
738 while (t < strend - prog->minlen) {
740 if (t < check_at - prog->check_offset_min) {
741 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
742 /* Since we moved from the found position,
743 we definitely contradict the found anchored
744 substr. Due to the above check we do not
745 contradict "check" substr.
746 Thus we can arrive here only if check substr
747 is float. Redo checking for "other"=="fixed".
750 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
751 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
752 goto do_other_anchored;
754 /* We don't contradict the found floating substring. */
755 /* XXXX Why not check for STCLASS? */
757 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
758 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
761 /* Position contradicts check-string */
762 /* XXXX probably better to look for check-string
763 than for "\n", so one should lower the limit for t? */
764 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
765 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
766 other_last = strpos = s = t + 1;
771 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
772 PL_colors[0],PL_colors[1]));
776 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
777 PL_colors[0],PL_colors[1]));
781 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
784 /* The found string does not prohibit matching at strpos,
785 - no optimization of calling REx engine can be performed,
786 unless it was an MBOL and we are not after MBOL,
787 or a future STCLASS check will fail this. */
789 /* Even in this situation we may use MBOL flag if strpos is offset
790 wrt the start of the string. */
791 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
792 && (strpos != strbeg) && strpos[-1] != '\n'
793 /* May be due to an implicit anchor of m{.*foo} */
794 && !(prog->reganch & ROPT_IMPLICIT))
799 DEBUG_r( if (ml_anch)
800 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
801 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
804 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
806 prog->check_utf8 /* Could be deleted already */
807 && --BmUSEFUL(prog->check_utf8) < 0
808 && (prog->check_utf8 == prog->float_utf8)
810 prog->check_substr /* Could be deleted already */
811 && --BmUSEFUL(prog->check_substr) < 0
812 && (prog->check_substr == prog->float_substr)
815 /* If flags & SOMETHING - do not do it many times on the same match */
816 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
817 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
818 if (do_utf8 ? prog->check_substr : prog->check_utf8)
819 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
820 prog->check_substr = prog->check_utf8 = Nullsv; /* disable */
821 prog->float_substr = prog->float_utf8 = Nullsv; /* clear */
822 check = Nullsv; /* abort */
824 /* XXXX This is a remnant of the old implementation. It
825 looks wasteful, since now INTUIT can use many
827 prog->reganch &= ~RE_USE_INTUIT;
834 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
835 if (prog->regstclass) {
836 /* minlen == 0 is possible if regstclass is \b or \B,
837 and the fixed substr is ''$.
838 Since minlen is already taken into account, s+1 is before strend;
839 accidentally, minlen >= 1 guaranties no false positives at s + 1
840 even for \b or \B. But (minlen? 1 : 0) below assumes that
841 regstclass does not come from lookahead... */
842 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
843 This leaves EXACTF only, which is dealt with in find_byclass(). */
844 U8* str = (U8*)STRING(prog->regstclass);
845 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
846 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
848 char *endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
849 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
850 : (prog->float_substr || prog->float_utf8
851 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
854 char *startpos = strbeg;
858 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
863 if (endpos == strend) {
864 DEBUG_r( PerlIO_printf(Perl_debug_log,
865 "Could not match STCLASS...\n") );
868 DEBUG_r( PerlIO_printf(Perl_debug_log,
869 "This position contradicts STCLASS...\n") );
870 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
872 /* Contradict one of substrings */
873 if (prog->anchored_substr || prog->anchored_utf8) {
874 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
875 DEBUG_r( what = "anchored" );
877 s = HOP3c(t, 1, strend);
878 if (s + start_shift + end_shift > strend) {
879 /* XXXX Should be taken into account earlier? */
880 DEBUG_r( PerlIO_printf(Perl_debug_log,
881 "Could not match STCLASS...\n") );
886 DEBUG_r( PerlIO_printf(Perl_debug_log,
887 "Looking for %s substr starting at offset %ld...\n",
888 what, (long)(s + start_shift - i_strpos)) );
891 /* Have both, check_string is floating */
892 if (t + start_shift >= check_at) /* Contradicts floating=check */
893 goto retry_floating_check;
894 /* Recheck anchored substring, but not floating... */
898 DEBUG_r( PerlIO_printf(Perl_debug_log,
899 "Looking for anchored substr starting at offset %ld...\n",
900 (long)(other_last - i_strpos)) );
901 goto do_other_anchored;
903 /* Another way we could have checked stclass at the
904 current position only: */
909 DEBUG_r( PerlIO_printf(Perl_debug_log,
910 "Looking for /%s^%s/m starting at offset %ld...\n",
911 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
914 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
916 /* Check is floating subtring. */
917 retry_floating_check:
918 t = check_at - start_shift;
919 DEBUG_r( what = "floating" );
920 goto hop_and_restart;
923 DEBUG_r(PerlIO_printf(Perl_debug_log,
924 "By STCLASS: moving %ld --> %ld\n",
925 (long)(t - i_strpos), (long)(s - i_strpos))
929 DEBUG_r(PerlIO_printf(Perl_debug_log,
930 "Does not contradict STCLASS...\n");
935 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
936 PL_colors[4], (check ? "Guessed" : "Giving up"),
937 PL_colors[5], (long)(s - i_strpos)) );
940 fail_finish: /* Substring not found */
941 if (prog->check_substr || prog->check_utf8) /* could be removed already */
942 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
944 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
945 PL_colors[4],PL_colors[5]));
949 /* We know what class REx starts with. Try to find this position... */
951 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
953 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
960 register I32 tmp = 1; /* Scratch variable? */
961 register bool do_utf8 = PL_reg_match_utf8;
963 /* We know what class it must start with. */
968 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
969 !UTF8_IS_INVARIANT((U8)s[0]) ?
970 reginclass(c, (U8*)s, 0, do_utf8) :
971 REGINCLASS(c, (U8*)s)) {
972 if (tmp && (norun || regtry(prog, s)))
986 if (REGINCLASS(c, (U8*)s) ||
987 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
988 /* The assignment of 2 is intentional:
989 * for the folded sharp s, the skip is 2. */
990 (skip = SHARP_S_SKIP))) {
991 if (tmp && (norun || regtry(prog, s)))
1003 while (s < strend) {
1004 if (tmp && (norun || regtry(prog, s)))
1013 ln = STR_LEN(c); /* length to match in octets/bytes */
1014 lnc = (I32) ln; /* length to match in characters */
1016 STRLEN ulen1, ulen2;
1018 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
1019 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
1021 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1022 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1024 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN_UCLC,
1025 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1026 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN_UCLC,
1027 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1029 while (sm < ((U8 *) m + ln)) {
1044 c2 = PL_fold_locale[c1];
1046 e = HOP3c(strend, -((I32)lnc), s);
1049 e = s; /* Due to minlen logic of intuit() */
1051 /* The idea in the EXACTF* cases is to first find the
1052 * first character of the EXACTF* node and then, if
1053 * necessary, case-insensitively compare the full
1054 * text of the node. The c1 and c2 are the first
1055 * characters (though in Unicode it gets a bit
1056 * more complicated because there are more cases
1057 * than just upper and lower: one needs to use
1058 * the so-called folding case for case-insensitive
1059 * matching (called "loose matching" in Unicode).
1060 * ibcmp_utf8() will do just that. */
1064 U8 tmpbuf [UTF8_MAXLEN+1];
1065 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
1066 STRLEN len, foldlen;
1069 /* Upper and lower of 1st char are equal -
1070 * probably not a "letter". */
1072 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1074 0 : UTF8_ALLOW_ANY);
1077 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1078 m, (char **)0, ln, (bool)UTF))
1079 && (norun || regtry(prog, s)) )
1082 uvchr_to_utf8(tmpbuf, c);
1083 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1085 && (f == c1 || f == c2)
1086 && (ln == foldlen ||
1087 !ibcmp_utf8((char *) foldbuf,
1088 (char **)0, foldlen, do_utf8,
1090 (char **)0, ln, (bool)UTF))
1091 && (norun || regtry(prog, s)) )
1099 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1101 0 : UTF8_ALLOW_ANY);
1103 /* Handle some of the three Greek sigmas cases.
1104 * Note that not all the possible combinations
1105 * are handled here: some of them are handled
1106 * by the standard folding rules, and some of
1107 * them (the character class or ANYOF cases)
1108 * are handled during compiletime in
1109 * regexec.c:S_regclass(). */
1110 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1111 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1112 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1114 if ( (c == c1 || c == c2)
1116 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1117 m, (char **)0, ln, (bool)UTF))
1118 && (norun || regtry(prog, s)) )
1121 uvchr_to_utf8(tmpbuf, c);
1122 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1124 && (f == c1 || f == c2)
1125 && (ln == foldlen ||
1126 !ibcmp_utf8((char *) foldbuf,
1127 (char **)0, foldlen, do_utf8,
1129 (char **)0, ln, (bool)UTF))
1130 && (norun || regtry(prog, s)) )
1141 && (ln == 1 || !(OP(c) == EXACTF
1143 : ibcmp_locale(s, m, ln)))
1144 && (norun || regtry(prog, s)) )
1150 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1151 && (ln == 1 || !(OP(c) == EXACTF
1153 : ibcmp_locale(s, m, ln)))
1154 && (norun || regtry(prog, s)) )
1161 PL_reg_flags |= RF_tainted;
1168 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1170 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1172 tmp = ((OP(c) == BOUND ?
1173 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1174 LOAD_UTF8_CHARCLASS(alnum,"a");
1175 while (s < strend) {
1176 if (tmp == !(OP(c) == BOUND ?
1177 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1178 isALNUM_LC_utf8((U8*)s)))
1181 if ((norun || regtry(prog, s)))
1188 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1189 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1190 while (s < strend) {
1192 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1194 if ((norun || regtry(prog, s)))
1200 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1204 PL_reg_flags |= RF_tainted;
1211 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1213 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1215 tmp = ((OP(c) == NBOUND ?
1216 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1217 LOAD_UTF8_CHARCLASS(alnum,"a");
1218 while (s < strend) {
1219 if (tmp == !(OP(c) == NBOUND ?
1220 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1221 isALNUM_LC_utf8((U8*)s)))
1223 else if ((norun || regtry(prog, s)))
1229 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1230 tmp = ((OP(c) == NBOUND ?
1231 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1232 while (s < strend) {
1234 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1236 else if ((norun || regtry(prog, s)))
1241 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1246 LOAD_UTF8_CHARCLASS(alnum,"a");
1247 while (s < strend) {
1248 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1249 if (tmp && (norun || regtry(prog, s)))
1260 while (s < strend) {
1262 if (tmp && (norun || regtry(prog, s)))
1274 PL_reg_flags |= RF_tainted;
1276 while (s < strend) {
1277 if (isALNUM_LC_utf8((U8*)s)) {
1278 if (tmp && (norun || regtry(prog, s)))
1289 while (s < strend) {
1290 if (isALNUM_LC(*s)) {
1291 if (tmp && (norun || regtry(prog, s)))
1304 LOAD_UTF8_CHARCLASS(alnum,"a");
1305 while (s < strend) {
1306 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1307 if (tmp && (norun || regtry(prog, s)))
1318 while (s < strend) {
1320 if (tmp && (norun || regtry(prog, s)))
1332 PL_reg_flags |= RF_tainted;
1334 while (s < strend) {
1335 if (!isALNUM_LC_utf8((U8*)s)) {
1336 if (tmp && (norun || regtry(prog, s)))
1347 while (s < strend) {
1348 if (!isALNUM_LC(*s)) {
1349 if (tmp && (norun || regtry(prog, s)))
1362 LOAD_UTF8_CHARCLASS(space," ");
1363 while (s < strend) {
1364 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1365 if (tmp && (norun || regtry(prog, s)))
1376 while (s < strend) {
1378 if (tmp && (norun || regtry(prog, s)))
1390 PL_reg_flags |= RF_tainted;
1392 while (s < strend) {
1393 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1394 if (tmp && (norun || regtry(prog, s)))
1405 while (s < strend) {
1406 if (isSPACE_LC(*s)) {
1407 if (tmp && (norun || regtry(prog, s)))
1420 LOAD_UTF8_CHARCLASS(space," ");
1421 while (s < strend) {
1422 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1423 if (tmp && (norun || regtry(prog, s)))
1434 while (s < strend) {
1436 if (tmp && (norun || regtry(prog, s)))
1448 PL_reg_flags |= RF_tainted;
1450 while (s < strend) {
1451 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1452 if (tmp && (norun || regtry(prog, s)))
1463 while (s < strend) {
1464 if (!isSPACE_LC(*s)) {
1465 if (tmp && (norun || regtry(prog, s)))
1478 LOAD_UTF8_CHARCLASS(digit,"0");
1479 while (s < strend) {
1480 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1481 if (tmp && (norun || regtry(prog, s)))
1492 while (s < strend) {
1494 if (tmp && (norun || regtry(prog, s)))
1506 PL_reg_flags |= RF_tainted;
1508 while (s < strend) {
1509 if (isDIGIT_LC_utf8((U8*)s)) {
1510 if (tmp && (norun || regtry(prog, s)))
1521 while (s < strend) {
1522 if (isDIGIT_LC(*s)) {
1523 if (tmp && (norun || regtry(prog, s)))
1536 LOAD_UTF8_CHARCLASS(digit,"0");
1537 while (s < strend) {
1538 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1539 if (tmp && (norun || regtry(prog, s)))
1550 while (s < strend) {
1552 if (tmp && (norun || regtry(prog, s)))
1564 PL_reg_flags |= RF_tainted;
1566 while (s < strend) {
1567 if (!isDIGIT_LC_utf8((U8*)s)) {
1568 if (tmp && (norun || regtry(prog, s)))
1579 while (s < strend) {
1580 if (!isDIGIT_LC(*s)) {
1581 if (tmp && (norun || regtry(prog, s)))
1593 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1602 - regexec_flags - match a regexp against a string
1605 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1606 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1607 /* strend: pointer to null at end of string */
1608 /* strbeg: real beginning of string */
1609 /* minend: end of match must be >=minend after stringarg. */
1610 /* data: May be used for some additional optimizations. */
1611 /* nosave: For optimizations. */
1614 register regnode *c;
1615 register char *startpos = stringarg;
1616 I32 minlen; /* must match at least this many chars */
1617 I32 dontbother = 0; /* how many characters not to try at end */
1618 /* I32 start_shift = 0; */ /* Offset of the start to find
1619 constant substr. */ /* CC */
1620 I32 end_shift = 0; /* Same for the end. */ /* CC */
1621 I32 scream_pos = -1; /* Internal iterator of scream. */
1623 SV* oreplsv = GvSV(PL_replgv);
1624 bool do_utf8 = DO_UTF8(sv);
1626 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
1627 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
1629 RX_MATCH_UTF8_set(prog,do_utf8);
1635 PL_regnarrate = DEBUG_r_TEST;
1638 /* Be paranoid... */
1639 if (prog == NULL || startpos == NULL) {
1640 Perl_croak(aTHX_ "NULL regexp parameter");
1644 minlen = prog->minlen;
1645 if (strend - startpos < minlen) {
1646 DEBUG_r(PerlIO_printf(Perl_debug_log,
1647 "String too short [regexec_flags]...\n"));
1651 /* Check validity of program. */
1652 if (UCHARAT(prog->program) != REG_MAGIC) {
1653 Perl_croak(aTHX_ "corrupted regexp program");
1657 PL_reg_eval_set = 0;
1660 if (prog->reganch & ROPT_UTF8)
1661 PL_reg_flags |= RF_utf8;
1663 /* Mark beginning of line for ^ and lookbehind. */
1664 PL_regbol = startpos;
1668 /* Mark end of line for $ (and such) */
1671 /* see how far we have to get to not match where we matched before */
1672 PL_regtill = startpos+minend;
1674 /* We start without call_cc context. */
1677 /* If there is a "must appear" string, look for it. */
1680 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1683 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1684 PL_reg_ganch = startpos;
1685 else if (sv && SvTYPE(sv) >= SVt_PVMG
1687 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1688 && mg->mg_len >= 0) {
1689 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1690 if (prog->reganch & ROPT_ANCH_GPOS) {
1691 if (s > PL_reg_ganch)
1696 else /* pos() not defined */
1697 PL_reg_ganch = strbeg;
1700 if (!(flags & REXEC_CHECKED) && (prog->check_substr != Nullsv || prog->check_utf8 != Nullsv)) {
1701 re_scream_pos_data d;
1703 d.scream_olds = &scream_olds;
1704 d.scream_pos = &scream_pos;
1705 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1707 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1708 goto phooey; /* not present */
1714 pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1715 UNI_DISPLAY_REGEX) :
1717 int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1718 char *s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1719 UNI_DISPLAY_REGEX) : startpos;
1720 int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1723 PerlIO_printf(Perl_debug_log,
1724 "%sMatching REx%s `%s%*.*s%s%s' against `%s%.*s%s%s'\n",
1725 PL_colors[4],PL_colors[5],PL_colors[0],
1728 len0 > 60 ? "..." : "",
1730 (int)(len1 > 60 ? 60 : len1),
1732 (len1 > 60 ? "..." : "")
1736 /* Simplest case: anchored match need be tried only once. */
1737 /* [unless only anchor is BOL and multiline is set] */
1738 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1739 if (s == startpos && regtry(prog, startpos))
1741 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1742 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1747 dontbother = minlen - 1;
1748 end = HOP3c(strend, -dontbother, strbeg) - 1;
1749 /* for multiline we only have to try after newlines */
1750 if (prog->check_substr || prog->check_utf8) {
1754 if (regtry(prog, s))
1759 if (prog->reganch & RE_USE_INTUIT) {
1760 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1771 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1772 if (regtry(prog, s))
1779 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1780 if (regtry(prog, PL_reg_ganch))
1785 /* Messy cases: unanchored match. */
1786 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1787 /* we have /x+whatever/ */
1788 /* it must be a one character string (XXXX Except UTF?) */
1793 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1794 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1795 ch = SvPVX(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1798 while (s < strend) {
1800 DEBUG_r( did_match = 1 );
1801 if (regtry(prog, s)) goto got_it;
1803 while (s < strend && *s == ch)
1810 while (s < strend) {
1812 DEBUG_r( did_match = 1 );
1813 if (regtry(prog, s)) goto got_it;
1815 while (s < strend && *s == ch)
1821 DEBUG_r(if (!did_match)
1822 PerlIO_printf(Perl_debug_log,
1823 "Did not find anchored character...\n")
1827 else if (prog->anchored_substr != Nullsv
1828 || prog->anchored_utf8 != Nullsv
1829 || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv)
1830 && prog->float_max_offset < strend - s)) {
1835 char *last1; /* Last position checked before */
1839 if (prog->anchored_substr || prog->anchored_utf8) {
1840 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1841 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1842 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1843 back_max = back_min = prog->anchored_offset;
1845 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1846 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1847 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1848 back_max = prog->float_max_offset;
1849 back_min = prog->float_min_offset;
1851 if (must == &PL_sv_undef)
1852 /* could not downgrade utf8 check substring, so must fail */
1855 last = HOP3c(strend, /* Cannot start after this */
1856 -(I32)(CHR_SVLEN(must)
1857 - (SvTAIL(must) != 0) + back_min), strbeg);
1860 last1 = HOPc(s, -1);
1862 last1 = s - 1; /* bogus */
1864 /* XXXX check_substr already used to find `s', can optimize if
1865 check_substr==must. */
1867 dontbother = end_shift;
1868 strend = HOPc(strend, -dontbother);
1869 while ( (s <= last) &&
1870 ((flags & REXEC_SCREAM)
1871 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1872 end_shift, &scream_pos, 0))
1873 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1874 (unsigned char*)strend, must,
1875 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1876 /* we may be pointing at the wrong string */
1877 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1878 s = strbeg + (s - SvPVX(sv));
1879 DEBUG_r( did_match = 1 );
1880 if (HOPc(s, -back_max) > last1) {
1881 last1 = HOPc(s, -back_min);
1882 s = HOPc(s, -back_max);
1885 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1887 last1 = HOPc(s, -back_min);
1891 while (s <= last1) {
1892 if (regtry(prog, s))
1898 while (s <= last1) {
1899 if (regtry(prog, s))
1905 DEBUG_r(if (!did_match)
1906 PerlIO_printf(Perl_debug_log,
1907 "Did not find %s substr `%s%.*s%s'%s...\n",
1908 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1909 ? "anchored" : "floating"),
1911 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1913 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1917 else if ((c = prog->regstclass)) {
1919 I32 op = (U8)OP(prog->regstclass);
1920 /* don't bother with what can't match */
1921 if (PL_regkind[op] != EXACT && op != CANY)
1922 strend = HOPc(strend, -(minlen - 1));
1925 SV *prop = sv_newmortal();
1933 pv_uni_display(dsv0, (U8*)SvPVX(prop), SvCUR(prop), 60,
1934 UNI_DISPLAY_REGEX) :
1936 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1938 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1939 len1 = UTF ? SvCUR(dsv1) : strend - s;
1940 PerlIO_printf(Perl_debug_log,
1941 "Matching stclass `%*.*s' against `%*.*s'\n",
1945 if (find_byclass(prog, c, s, strend, startpos, 0))
1947 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1951 if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) {
1956 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1957 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1958 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1960 if (flags & REXEC_SCREAM) {
1961 last = screaminstr(sv, float_real, s - strbeg,
1962 end_shift, &scream_pos, 1); /* last one */
1964 last = scream_olds; /* Only one occurrence. */
1965 /* we may be pointing at the wrong string */
1966 else if (RX_MATCH_COPIED(prog))
1967 s = strbeg + (s - SvPVX(sv));
1971 char *little = SvPV(float_real, len);
1973 if (SvTAIL(float_real)) {
1974 if (memEQ(strend - len + 1, little, len - 1))
1975 last = strend - len + 1;
1976 else if (!PL_multiline)
1977 last = memEQ(strend - len, little, len)
1978 ? strend - len : Nullch;
1984 last = rninstr(s, strend, little, little + len);
1986 last = strend; /* matching `$' */
1990 DEBUG_r(PerlIO_printf(Perl_debug_log,
1991 "%sCan't trim the tail, match fails (should not happen)%s\n",
1992 PL_colors[4],PL_colors[5]));
1993 goto phooey; /* Should not happen! */
1995 dontbother = strend - last + prog->float_min_offset;
1997 if (minlen && (dontbother < minlen))
1998 dontbother = minlen - 1;
1999 strend -= dontbother; /* this one's always in bytes! */
2000 /* We don't know much -- general case. */
2003 if (regtry(prog, s))
2012 if (regtry(prog, s))
2014 } while (s++ < strend);
2022 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2024 if (PL_reg_eval_set) {
2025 /* Preserve the current value of $^R */
2026 if (oreplsv != GvSV(PL_replgv))
2027 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2028 restored, the value remains
2030 restore_pos(aTHX_ 0);
2033 /* make sure $`, $&, $', and $digit will work later */
2034 if ( !(flags & REXEC_NOT_FIRST) ) {
2035 RX_MATCH_COPY_FREE(prog);
2036 if (flags & REXEC_COPY_STR) {
2037 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2038 #ifdef PERL_COPY_ON_WRITE
2040 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2042 PerlIO_printf(Perl_debug_log,
2043 "Copy on write: regexp capture, type %d\n",
2046 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2047 prog->subbeg = SvPVX(prog->saved_copy);
2048 assert (SvPOKp(prog->saved_copy));
2052 RX_MATCH_COPIED_on(prog);
2053 s = savepvn(strbeg, i);
2059 prog->subbeg = strbeg;
2060 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2067 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2068 PL_colors[4],PL_colors[5]));
2069 if (PL_reg_eval_set)
2070 restore_pos(aTHX_ 0);
2075 - regtry - try match at specific point
2077 STATIC I32 /* 0 failure, 1 success */
2078 S_regtry(pTHX_ regexp *prog, char *startpos)
2086 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2088 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2091 PL_reg_eval_set = RS_init;
2093 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2094 (IV)(PL_stack_sp - PL_stack_base));
2096 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2097 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2098 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2100 /* Apparently this is not needed, judging by wantarray. */
2101 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2102 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2105 /* Make $_ available to executed code. */
2106 if (PL_reg_sv != DEFSV) {
2111 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2112 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2113 /* prepare for quick setting of pos */
2114 sv_magic(PL_reg_sv, (SV*)0,
2115 PERL_MAGIC_regex_global, Nullch, 0);
2116 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
2120 PL_reg_oldpos = mg->mg_len;
2121 SAVEDESTRUCTOR_X(restore_pos, 0);
2123 if (!PL_reg_curpm) {
2124 Newz(22,PL_reg_curpm, 1, PMOP);
2127 SV* repointer = newSViv(0);
2128 /* so we know which PL_regex_padav element is PL_reg_curpm */
2129 SvFLAGS(repointer) |= SVf_BREAK;
2130 av_push(PL_regex_padav,repointer);
2131 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2132 PL_regex_pad = AvARRAY(PL_regex_padav);
2136 PM_SETRE(PL_reg_curpm, prog);
2137 PL_reg_oldcurpm = PL_curpm;
2138 PL_curpm = PL_reg_curpm;
2139 if (RX_MATCH_COPIED(prog)) {
2140 /* Here is a serious problem: we cannot rewrite subbeg,
2141 since it may be needed if this match fails. Thus
2142 $` inside (?{}) could fail... */
2143 PL_reg_oldsaved = prog->subbeg;
2144 PL_reg_oldsavedlen = prog->sublen;
2145 #ifdef PERL_COPY_ON_WRITE
2146 PL_nrs = prog->saved_copy;
2148 RX_MATCH_COPIED_off(prog);
2151 PL_reg_oldsaved = Nullch;
2152 prog->subbeg = PL_bostr;
2153 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2155 prog->startp[0] = startpos - PL_bostr;
2156 PL_reginput = startpos;
2157 PL_regstartp = prog->startp;
2158 PL_regendp = prog->endp;
2159 PL_reglastparen = &prog->lastparen;
2160 PL_reglastcloseparen = &prog->lastcloseparen;
2161 prog->lastparen = 0;
2162 prog->lastcloseparen = 0;
2164 DEBUG_r(PL_reg_starttry = startpos);
2165 if (PL_reg_start_tmpl <= prog->nparens) {
2166 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2167 if(PL_reg_start_tmp)
2168 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2170 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2173 /* XXXX What this code is doing here?!!! There should be no need
2174 to do this again and again, PL_reglastparen should take care of
2177 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2178 * Actually, the code in regcppop() (which Ilya may be meaning by
2179 * PL_reglastparen), is not needed at all by the test suite
2180 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2181 * enough, for building DynaLoader, or otherwise this
2182 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2183 * will happen. Meanwhile, this code *is* needed for the
2184 * above-mentioned test suite tests to succeed. The common theme
2185 * on those tests seems to be returning null fields from matches.
2190 if (prog->nparens) {
2191 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2198 if (regmatch(prog->program + 1)) {
2199 prog->endp[0] = PL_reginput - PL_bostr;
2202 REGCP_UNWIND(lastcp);
2206 #define RE_UNWIND_BRANCH 1
2207 #define RE_UNWIND_BRANCHJ 2
2211 typedef struct { /* XX: makes sense to enlarge it... */
2215 } re_unwind_generic_t;
2228 } re_unwind_branch_t;
2230 typedef union re_unwind_t {
2232 re_unwind_generic_t generic;
2233 re_unwind_branch_t branch;
2236 #define sayYES goto yes
2237 #define sayNO goto no
2238 #define sayNO_ANYOF goto no_anyof
2239 #define sayYES_FINAL goto yes_final
2240 #define sayYES_LOUD goto yes_loud
2241 #define sayNO_FINAL goto no_final
2242 #define sayNO_SILENT goto do_no
2243 #define saySAME(x) if (x) goto yes; else goto no
2245 #define REPORT_CODE_OFF 24
2248 - regmatch - main matching routine
2250 * Conceptually the strategy is simple: check to see whether the current
2251 * node matches, call self recursively to see whether the rest matches,
2252 * and then act accordingly. In practice we make some effort to avoid
2253 * recursion, in particular by going through "ordinary" nodes (that don't
2254 * need to know whether the rest of the match failed) by a loop instead of
2257 /* [lwall] I've hoisted the register declarations to the outer block in order to
2258 * maybe save a little bit of pushing and popping on the stack. It also takes
2259 * advantage of machines that use a register save mask on subroutine entry.
2261 STATIC I32 /* 0 failure, 1 success */
2262 S_regmatch(pTHX_ regnode *prog)
2264 register regnode *scan; /* Current node. */
2265 regnode *next; /* Next node. */
2266 regnode *inner; /* Next node in internal branch. */
2267 register I32 nextchr; /* renamed nextchr - nextchar colides with
2268 function of same name */
2269 register I32 n; /* no or next */
2270 register I32 ln = 0; /* len or last */
2271 register char *s = Nullch; /* operand or save */
2272 register char *locinput = PL_reginput;
2273 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2274 int minmod = 0, sw = 0, logical = 0;
2277 I32 firstcp = PL_savestack_ix;
2279 register bool do_utf8 = PL_reg_match_utf8;
2281 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2282 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2283 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2290 /* Note that nextchr is a byte even in UTF */
2291 nextchr = UCHARAT(locinput);
2293 while (scan != NULL) {
2296 SV *prop = sv_newmortal();
2297 int docolor = *PL_colors[0];
2298 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2299 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2300 /* The part of the string before starttry has one color
2301 (pref0_len chars), between starttry and current
2302 position another one (pref_len - pref0_len chars),
2303 after the current position the third one.
2304 We assume that pref0_len <= pref_len, otherwise we
2305 decrease pref0_len. */
2306 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2307 ? (5 + taill) - l : locinput - PL_bostr;
2310 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2312 pref0_len = pref_len - (locinput - PL_reg_starttry);
2313 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2314 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2315 ? (5 + taill) - pref_len : PL_regeol - locinput);
2316 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2320 if (pref0_len > pref_len)
2321 pref0_len = pref_len;
2322 regprop(prop, scan);
2325 do_utf8 && OP(scan) != CANY ?
2326 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2327 pref0_len, 60, UNI_DISPLAY_REGEX) :
2328 locinput - pref_len;
2329 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2330 char *s1 = do_utf8 && OP(scan) != CANY ?
2331 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2332 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2333 locinput - pref_len + pref0_len;
2334 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2335 char *s2 = do_utf8 && OP(scan) != CANY ?
2336 pv_uni_display(dsv2, (U8*)locinput,
2337 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2339 int len2 = do_utf8 ? strlen(s2) : l;
2340 PerlIO_printf(Perl_debug_log,
2341 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2342 (IV)(locinput - PL_bostr),
2349 (docolor ? "" : "> <"),
2353 15 - l - pref_len + 1,
2355 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2360 next = scan + NEXT_OFF(scan);
2366 if (locinput == PL_bostr || (PL_multiline &&
2367 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2369 /* regtill = regbol; */
2374 if (locinput == PL_bostr ||
2375 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2381 if (locinput == PL_bostr)
2385 if (locinput == PL_reg_ganch)
2395 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2400 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2402 if (PL_regeol - locinput > 1)
2406 if (PL_regeol != locinput)
2410 if (!nextchr && locinput >= PL_regeol)
2413 locinput += PL_utf8skip[nextchr];
2414 if (locinput > PL_regeol)
2416 nextchr = UCHARAT(locinput);
2419 nextchr = UCHARAT(++locinput);
2422 if (!nextchr && locinput >= PL_regeol)
2424 nextchr = UCHARAT(++locinput);
2427 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2430 locinput += PL_utf8skip[nextchr];
2431 if (locinput > PL_regeol)
2433 nextchr = UCHARAT(locinput);
2436 nextchr = UCHARAT(++locinput);
2441 if (do_utf8 != UTF) {
2442 /* The target and the pattern have differing utf8ness. */
2448 /* The target is utf8, the pattern is not utf8. */
2452 if (NATIVE_TO_UNI(*(U8*)s) !=
2453 utf8n_to_uvuni((U8*)l, UTF8_MAXLEN, &ulen,
2455 0 : UTF8_ALLOW_ANY))
2462 /* The target is not utf8, the pattern is utf8. */
2466 if (NATIVE_TO_UNI(*((U8*)l)) !=
2467 utf8n_to_uvuni((U8*)s, UTF8_MAXLEN, &ulen,
2469 0 : UTF8_ALLOW_ANY))
2476 nextchr = UCHARAT(locinput);
2479 /* The target and the pattern have the same utf8ness. */
2480 /* Inline the first character, for speed. */
2481 if (UCHARAT(s) != nextchr)
2483 if (PL_regeol - locinput < ln)
2485 if (ln > 1 && memNE(s, locinput, ln))
2488 nextchr = UCHARAT(locinput);
2491 PL_reg_flags |= RF_tainted;
2497 if (do_utf8 || UTF) {
2498 /* Either target or the pattern are utf8. */
2500 char *e = PL_regeol;
2502 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2503 l, &e, 0, do_utf8)) {
2504 /* One more case for the sharp s:
2505 * pack("U0U*", 0xDF) =~ /ss/i,
2506 * the 0xC3 0x9F are the UTF-8
2507 * byte sequence for the U+00DF. */
2509 toLOWER(s[0]) == 's' &&
2511 toLOWER(s[1]) == 's' &&
2518 nextchr = UCHARAT(locinput);
2522 /* Neither the target and the pattern are utf8. */
2524 /* Inline the first character, for speed. */
2525 if (UCHARAT(s) != nextchr &&
2526 UCHARAT(s) != ((OP(scan) == EXACTF)
2527 ? PL_fold : PL_fold_locale)[nextchr])
2529 if (PL_regeol - locinput < ln)
2531 if (ln > 1 && (OP(scan) == EXACTF
2532 ? ibcmp(s, locinput, ln)
2533 : ibcmp_locale(s, locinput, ln)))
2536 nextchr = UCHARAT(locinput);
2540 STRLEN inclasslen = PL_regeol - locinput;
2542 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2544 if (locinput >= PL_regeol)
2546 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2547 nextchr = UCHARAT(locinput);
2552 nextchr = UCHARAT(locinput);
2553 if (!REGINCLASS(scan, (U8*)locinput))
2555 if (!nextchr && locinput >= PL_regeol)
2557 nextchr = UCHARAT(++locinput);
2561 /* If we might have the case of the German sharp s
2562 * in a casefolding Unicode character class. */
2564 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2565 locinput += SHARP_S_SKIP;
2566 nextchr = UCHARAT(locinput);
2572 PL_reg_flags |= RF_tainted;
2578 LOAD_UTF8_CHARCLASS(alnum,"a");
2579 if (!(OP(scan) == ALNUM
2580 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2581 : isALNUM_LC_utf8((U8*)locinput)))
2585 locinput += PL_utf8skip[nextchr];
2586 nextchr = UCHARAT(locinput);
2589 if (!(OP(scan) == ALNUM
2590 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2592 nextchr = UCHARAT(++locinput);
2595 PL_reg_flags |= RF_tainted;
2598 if (!nextchr && locinput >= PL_regeol)
2601 LOAD_UTF8_CHARCLASS(alnum,"a");
2602 if (OP(scan) == NALNUM
2603 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2604 : isALNUM_LC_utf8((U8*)locinput))
2608 locinput += PL_utf8skip[nextchr];
2609 nextchr = UCHARAT(locinput);
2612 if (OP(scan) == NALNUM
2613 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2615 nextchr = UCHARAT(++locinput);
2619 PL_reg_flags |= RF_tainted;
2623 /* was last char in word? */
2625 if (locinput == PL_bostr)
2628 U8 *r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2630 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2632 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2633 ln = isALNUM_uni(ln);
2634 LOAD_UTF8_CHARCLASS(alnum,"a");
2635 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2638 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2639 n = isALNUM_LC_utf8((U8*)locinput);
2643 ln = (locinput != PL_bostr) ?
2644 UCHARAT(locinput - 1) : '\n';
2645 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2647 n = isALNUM(nextchr);
2650 ln = isALNUM_LC(ln);
2651 n = isALNUM_LC(nextchr);
2654 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2655 OP(scan) == BOUNDL))
2659 PL_reg_flags |= RF_tainted;
2665 if (UTF8_IS_CONTINUED(nextchr)) {
2666 LOAD_UTF8_CHARCLASS(space," ");
2667 if (!(OP(scan) == SPACE
2668 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2669 : isSPACE_LC_utf8((U8*)locinput)))
2673 locinput += PL_utf8skip[nextchr];
2674 nextchr = UCHARAT(locinput);
2677 if (!(OP(scan) == SPACE
2678 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2680 nextchr = UCHARAT(++locinput);
2683 if (!(OP(scan) == SPACE
2684 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2686 nextchr = UCHARAT(++locinput);
2690 PL_reg_flags |= RF_tainted;
2693 if (!nextchr && locinput >= PL_regeol)
2696 LOAD_UTF8_CHARCLASS(space," ");
2697 if (OP(scan) == NSPACE
2698 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2699 : isSPACE_LC_utf8((U8*)locinput))
2703 locinput += PL_utf8skip[nextchr];
2704 nextchr = UCHARAT(locinput);
2707 if (OP(scan) == NSPACE
2708 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2710 nextchr = UCHARAT(++locinput);
2713 PL_reg_flags |= RF_tainted;
2719 LOAD_UTF8_CHARCLASS(digit,"0");
2720 if (!(OP(scan) == DIGIT
2721 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2722 : isDIGIT_LC_utf8((U8*)locinput)))
2726 locinput += PL_utf8skip[nextchr];
2727 nextchr = UCHARAT(locinput);
2730 if (!(OP(scan) == DIGIT
2731 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2733 nextchr = UCHARAT(++locinput);
2736 PL_reg_flags |= RF_tainted;
2739 if (!nextchr && locinput >= PL_regeol)
2742 LOAD_UTF8_CHARCLASS(digit,"0");
2743 if (OP(scan) == NDIGIT
2744 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2745 : isDIGIT_LC_utf8((U8*)locinput))
2749 locinput += PL_utf8skip[nextchr];
2750 nextchr = UCHARAT(locinput);
2753 if (OP(scan) == NDIGIT
2754 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2756 nextchr = UCHARAT(++locinput);
2759 if (locinput >= PL_regeol)
2762 LOAD_UTF8_CHARCLASS(mark,"~");
2763 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2765 locinput += PL_utf8skip[nextchr];
2766 while (locinput < PL_regeol &&
2767 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2768 locinput += UTF8SKIP(locinput);
2769 if (locinput > PL_regeol)
2774 nextchr = UCHARAT(locinput);
2777 PL_reg_flags |= RF_tainted;
2781 n = ARG(scan); /* which paren pair */
2782 ln = PL_regstartp[n];
2783 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2784 if ((I32)*PL_reglastparen < n || ln == -1)
2785 sayNO; /* Do not match unless seen CLOSEn. */
2786 if (ln == PL_regendp[n])
2790 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2792 char *e = PL_bostr + PL_regendp[n];
2794 * Note that we can't do the "other character" lookup trick as
2795 * in the 8-bit case (no pun intended) because in Unicode we
2796 * have to map both upper and title case to lower case.
2798 if (OP(scan) == REFF) {
2799 STRLEN ulen1, ulen2;
2800 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2801 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2805 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2806 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2807 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2814 nextchr = UCHARAT(locinput);
2818 /* Inline the first character, for speed. */
2819 if (UCHARAT(s) != nextchr &&
2821 (UCHARAT(s) != ((OP(scan) == REFF
2822 ? PL_fold : PL_fold_locale)[nextchr]))))
2824 ln = PL_regendp[n] - ln;
2825 if (locinput + ln > PL_regeol)
2827 if (ln > 1 && (OP(scan) == REF
2828 ? memNE(s, locinput, ln)
2830 ? ibcmp(s, locinput, ln)
2831 : ibcmp_locale(s, locinput, ln))))
2834 nextchr = UCHARAT(locinput);
2845 OP_4tree *oop = PL_op;
2846 COP *ocurcop = PL_curcop;
2849 struct regexp *oreg = PL_reg_re;
2852 PL_op = (OP_4tree*)PL_regdata->data[n];
2853 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2854 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
2855 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2859 CALLRUNOPS(aTHX); /* Scalar context. */
2862 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
2870 PAD_RESTORE_LOCAL(old_comppad);
2871 PL_curcop = ocurcop;
2873 if (logical == 2) { /* Postponed subexpression. */
2875 MAGIC *mg = Null(MAGIC*);
2877 CHECKPOINT cp, lastcp;
2881 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
2882 mg = mg_find(sv, PERL_MAGIC_qr);
2883 else if (SvSMAGICAL(ret)) {
2884 if (SvGMAGICAL(ret))
2885 sv_unmagic(ret, PERL_MAGIC_qr);
2887 mg = mg_find(ret, PERL_MAGIC_qr);
2891 re = (regexp *)mg->mg_obj;
2892 (void)ReREFCNT_inc(re);
2896 char *t = SvPV(ret, len);
2898 char *oprecomp = PL_regprecomp;
2899 I32 osize = PL_regsize;
2900 I32 onpar = PL_regnpar;
2903 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
2904 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2906 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
2908 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2910 PL_regprecomp = oprecomp;
2915 PerlIO_printf(Perl_debug_log,
2916 "Entering embedded `%s%.60s%s%s'\n",
2920 (strlen(re->precomp) > 60 ? "..." : ""))
2923 state.prev = PL_reg_call_cc;
2924 state.cc = PL_regcc;
2925 state.re = PL_reg_re;
2929 cp = regcppush(0); /* Save *all* the positions. */
2932 state.ss = PL_savestack_ix;
2933 *PL_reglastparen = 0;
2934 *PL_reglastcloseparen = 0;
2935 PL_reg_call_cc = &state;
2936 PL_reginput = locinput;
2937 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
2938 ((re->reganch & ROPT_UTF8) != 0);
2939 if (toggleutf) PL_reg_flags ^= RF_utf8;
2941 /* XXXX This is too dramatic a measure... */
2944 if (regmatch(re->program + 1)) {
2945 /* Even though we succeeded, we need to restore
2946 global variables, since we may be wrapped inside
2947 SUSPEND, thus the match may be not finished yet. */
2949 /* XXXX Do this only if SUSPENDed? */
2950 PL_reg_call_cc = state.prev;
2951 PL_regcc = state.cc;
2952 PL_reg_re = state.re;
2953 cache_re(PL_reg_re);
2954 if (toggleutf) PL_reg_flags ^= RF_utf8;
2956 /* XXXX This is too dramatic a measure... */
2959 /* These are needed even if not SUSPEND. */
2965 REGCP_UNWIND(lastcp);
2967 PL_reg_call_cc = state.prev;
2968 PL_regcc = state.cc;
2969 PL_reg_re = state.re;
2970 cache_re(PL_reg_re);
2971 if (toggleutf) PL_reg_flags ^= RF_utf8;
2973 /* XXXX This is too dramatic a measure... */
2983 sv_setsv(save_scalar(PL_replgv), ret);
2989 n = ARG(scan); /* which paren pair */
2990 PL_reg_start_tmp[n] = locinput;
2995 n = ARG(scan); /* which paren pair */
2996 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2997 PL_regendp[n] = locinput - PL_bostr;
2998 if (n > (I32)*PL_reglastparen)
2999 *PL_reglastparen = n;
3000 *PL_reglastcloseparen = n;
3003 n = ARG(scan); /* which paren pair */
3004 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
3007 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3009 next = NEXTOPER(NEXTOPER(scan));
3011 next = scan + ARG(scan);
3012 if (OP(next) == IFTHEN) /* Fake one. */
3013 next = NEXTOPER(NEXTOPER(next));
3017 logical = scan->flags;
3019 /*******************************************************************
3020 PL_regcc contains infoblock about the innermost (...)* loop, and
3021 a pointer to the next outer infoblock.
3023 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3025 1) After matching X, regnode for CURLYX is processed;
3027 2) This regnode creates infoblock on the stack, and calls
3028 regmatch() recursively with the starting point at WHILEM node;
3030 3) Each hit of WHILEM node tries to match A and Z (in the order
3031 depending on the current iteration, min/max of {min,max} and
3032 greediness). The information about where are nodes for "A"
3033 and "Z" is read from the infoblock, as is info on how many times "A"
3034 was already matched, and greediness.
3036 4) After A matches, the same WHILEM node is hit again.
3038 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3039 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3040 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3041 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3042 of the external loop.
3044 Currently present infoblocks form a tree with a stem formed by PL_curcc
3045 and whatever it mentions via ->next, and additional attached trees
3046 corresponding to temporarily unset infoblocks as in "5" above.
3048 In the following picture infoblocks for outer loop of
3049 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3050 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3051 infoblocks are drawn below the "reset" infoblock.
3053 In fact in the picture below we do not show failed matches for Z and T
3054 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3055 more obvious *why* one needs to *temporary* unset infoblocks.]
3057 Matched REx position InfoBlocks Comment
3061 Y A)*?Z)*?T x <- O <- I
3062 YA )*?Z)*?T x <- O <- I
3063 YA A)*?Z)*?T x <- O <- I
3064 YAA )*?Z)*?T x <- O <- I
3065 YAA Z)*?T x <- O # Temporary unset I
3068 YAAZ Y(A)*?Z)*?T x <- O
3071 YAAZY (A)*?Z)*?T x <- O
3074 YAAZY A)*?Z)*?T x <- O <- I
3077 YAAZYA )*?Z)*?T x <- O <- I
3080 YAAZYA Z)*?T x <- O # Temporary unset I
3086 YAAZYAZ T x # Temporary unset O
3093 *******************************************************************/
3096 CHECKPOINT cp = PL_savestack_ix;
3097 /* No need to save/restore up to this paren */
3098 I32 parenfloor = scan->flags;
3100 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3102 cc.oldcc = PL_regcc;
3104 /* XXXX Probably it is better to teach regpush to support
3105 parenfloor > PL_regsize... */
3106 if (parenfloor > (I32)*PL_reglastparen)
3107 parenfloor = *PL_reglastparen; /* Pessimization... */
3108 cc.parenfloor = parenfloor;
3110 cc.min = ARG1(scan);
3111 cc.max = ARG2(scan);
3112 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3116 PL_reginput = locinput;
3117 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3119 PL_regcc = cc.oldcc;
3125 * This is really hard to understand, because after we match
3126 * what we're trying to match, we must make sure the rest of
3127 * the REx is going to match for sure, and to do that we have
3128 * to go back UP the parse tree by recursing ever deeper. And
3129 * if it fails, we have to reset our parent's current state
3130 * that we can try again after backing off.
3133 CHECKPOINT cp, lastcp;
3134 CURCUR* cc = PL_regcc;
3135 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3137 n = cc->cur + 1; /* how many we know we matched */
3138 PL_reginput = locinput;
3141 PerlIO_printf(Perl_debug_log,
3142 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3143 REPORT_CODE_OFF+PL_regindent*2, "",
3144 (long)n, (long)cc->min,
3145 (long)cc->max, PTR2UV(cc))
3148 /* If degenerate scan matches "", assume scan done. */
3150 if (locinput == cc->lastloc && n >= cc->min) {
3151 PL_regcc = cc->oldcc;
3155 PerlIO_printf(Perl_debug_log,
3156 "%*s empty match detected, try continuation...\n",
3157 REPORT_CODE_OFF+PL_regindent*2, "")
3159 if (regmatch(cc->next))
3167 /* First just match a string of min scans. */
3171 cc->lastloc = locinput;
3172 if (regmatch(cc->scan))
3175 cc->lastloc = lastloc;
3180 /* Check whether we already were at this position.
3181 Postpone detection until we know the match is not
3182 *that* much linear. */
3183 if (!PL_reg_maxiter) {
3184 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3185 PL_reg_leftiter = PL_reg_maxiter;
3187 if (PL_reg_leftiter-- == 0) {
3188 I32 size = (PL_reg_maxiter + 7)/8;
3189 if (PL_reg_poscache) {
3190 if ((I32)PL_reg_poscache_size < size) {
3191 Renew(PL_reg_poscache, size, char);
3192 PL_reg_poscache_size = size;
3194 Zero(PL_reg_poscache, size, char);
3197 PL_reg_poscache_size = size;
3198 Newz(29, PL_reg_poscache, size, char);
3201 PerlIO_printf(Perl_debug_log,
3202 "%sDetected a super-linear match, switching on caching%s...\n",
3203 PL_colors[4], PL_colors[5])
3206 if (PL_reg_leftiter < 0) {
3207 I32 o = locinput - PL_bostr, b;
3209 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3212 if (PL_reg_poscache[o] & (1<<b)) {
3214 PerlIO_printf(Perl_debug_log,
3215 "%*s already tried at this position...\n",
3216 REPORT_CODE_OFF+PL_regindent*2, "")
3218 if (PL_reg_flags & RF_false)
3223 PL_reg_poscache[o] |= (1<<b);
3227 /* Prefer next over scan for minimal matching. */
3230 PL_regcc = cc->oldcc;
3233 cp = regcppush(cc->parenfloor);
3235 if (regmatch(cc->next)) {
3237 sayYES; /* All done. */
3239 REGCP_UNWIND(lastcp);
3245 if (n >= cc->max) { /* Maximum greed exceeded? */
3246 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3247 && !(PL_reg_flags & RF_warned)) {
3248 PL_reg_flags |= RF_warned;
3249 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3250 "Complex regular subexpression recursion",
3257 PerlIO_printf(Perl_debug_log,
3258 "%*s trying longer...\n",
3259 REPORT_CODE_OFF+PL_regindent*2, "")
3261 /* Try scanning more and see if it helps. */
3262 PL_reginput = locinput;
3264 cc->lastloc = locinput;
3265 cp = regcppush(cc->parenfloor);
3267 if (regmatch(cc->scan)) {
3271 REGCP_UNWIND(lastcp);
3274 cc->lastloc = lastloc;
3278 /* Prefer scan over next for maximal matching. */
3280 if (n < cc->max) { /* More greed allowed? */
3281 cp = regcppush(cc->parenfloor);
3283 cc->lastloc = locinput;
3285 if (regmatch(cc->scan)) {
3289 REGCP_UNWIND(lastcp);
3290 regcppop(); /* Restore some previous $<digit>s? */
3291 PL_reginput = locinput;
3293 PerlIO_printf(Perl_debug_log,
3294 "%*s failed, try continuation...\n",
3295 REPORT_CODE_OFF+PL_regindent*2, "")
3298 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3299 && !(PL_reg_flags & RF_warned)) {
3300 PL_reg_flags |= RF_warned;
3301 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3302 "Complex regular subexpression recursion",
3306 /* Failed deeper matches of scan, so see if this one works. */
3307 PL_regcc = cc->oldcc;
3310 if (regmatch(cc->next))
3316 cc->lastloc = lastloc;
3321 next = scan + ARG(scan);
3324 inner = NEXTOPER(NEXTOPER(scan));
3327 inner = NEXTOPER(scan);
3331 if (OP(next) != c1) /* No choice. */
3332 next = inner; /* Avoid recursion. */
3334 I32 lastparen = *PL_reglastparen;
3336 re_unwind_branch_t *uw;
3338 /* Put unwinding data on stack */
3339 unwind1 = SSNEWt(1,re_unwind_branch_t);
3340 uw = SSPTRt(unwind1,re_unwind_branch_t);
3343 uw->type = ((c1 == BRANCH)
3345 : RE_UNWIND_BRANCHJ);
3346 uw->lastparen = lastparen;
3348 uw->locinput = locinput;
3349 uw->nextchr = nextchr;
3351 uw->regindent = ++PL_regindent;
3354 REGCP_SET(uw->lastcp);
3356 /* Now go into the first branch */
3369 /* We suppose that the next guy does not need
3370 backtracking: in particular, it is of constant non-zero length,
3371 and has no parenths to influence future backrefs. */
3372 ln = ARG1(scan); /* min to match */
3373 n = ARG2(scan); /* max to match */
3374 paren = scan->flags;
3376 if (paren > PL_regsize)
3378 if (paren > (I32)*PL_reglastparen)
3379 *PL_reglastparen = paren;
3381 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3383 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3384 PL_reginput = locinput;
3387 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3389 locinput = PL_reginput;
3390 if (HAS_TEXT(next) || JUMPABLE(next)) {
3391 regnode *text_node = next;
3393 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3395 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3397 if (PL_regkind[(U8)OP(text_node)] == REF) {
3401 else { c1 = (U8)*STRING(text_node); }
3402 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3404 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3405 c2 = PL_fold_locale[c1];
3414 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3415 /* If it could work, try it. */
3417 UCHARAT(PL_reginput) == c1 ||
3418 UCHARAT(PL_reginput) == c2)
3422 PL_regstartp[paren] =
3423 HOPc(PL_reginput, -l) - PL_bostr;
3424 PL_regendp[paren] = PL_reginput - PL_bostr;
3427 PL_regendp[paren] = -1;
3431 REGCP_UNWIND(lastcp);
3433 /* Couldn't or didn't -- move forward. */
3434 PL_reginput = locinput;
3435 if (regrepeat_hard(scan, 1, &l)) {
3437 locinput = PL_reginput;
3444 n = regrepeat_hard(scan, n, &l);
3445 locinput = PL_reginput;
3447 PerlIO_printf(Perl_debug_log,
3448 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3449 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3453 if (HAS_TEXT(next) || JUMPABLE(next)) {
3454 regnode *text_node = next;
3456 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3458 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3460 if (PL_regkind[(U8)OP(text_node)] == REF) {
3464 else { c1 = (U8)*STRING(text_node); }
3466 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3468 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3469 c2 = PL_fold_locale[c1];
3480 /* If it could work, try it. */
3482 UCHARAT(PL_reginput) == c1 ||
3483 UCHARAT(PL_reginput) == c2)
3486 PerlIO_printf(Perl_debug_log,
3487 "%*s trying tail with n=%"IVdf"...\n",
3488 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3492 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3493 PL_regendp[paren] = PL_reginput - PL_bostr;
3496 PL_regendp[paren] = -1;
3500 REGCP_UNWIND(lastcp);
3502 /* Couldn't or didn't -- back up. */
3504 locinput = HOPc(locinput, -l);
3505 PL_reginput = locinput;
3512 paren = scan->flags; /* Which paren to set */
3513 if (paren > PL_regsize)
3515 if (paren > (I32)*PL_reglastparen)
3516 *PL_reglastparen = paren;
3517 ln = ARG1(scan); /* min to match */
3518 n = ARG2(scan); /* max to match */
3519 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3523 ln = ARG1(scan); /* min to match */
3524 n = ARG2(scan); /* max to match */
3525 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3530 scan = NEXTOPER(scan);
3536 scan = NEXTOPER(scan);
3540 * Lookahead to avoid useless match attempts
3541 * when we know what character comes next.
3545 * Used to only do .*x and .*?x, but now it allows
3546 * for )'s, ('s and (?{ ... })'s to be in the way
3547 * of the quantifier and the EXACT-like node. -- japhy
3550 if (HAS_TEXT(next) || JUMPABLE(next)) {
3552 regnode *text_node = next;
3554 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3556 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3558 if (PL_regkind[(U8)OP(text_node)] == REF) {
3560 goto assume_ok_easy;
3562 else { s = (U8*)STRING(text_node); }
3566 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3568 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3569 c2 = PL_fold_locale[c1];
3572 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3573 STRLEN ulen1, ulen2;
3574 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3575 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3577 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3578 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3580 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXLEN, 0,
3582 0 : UTF8_ALLOW_ANY);
3583 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXLEN, 0,
3585 0 : UTF8_ALLOW_ANY);
3588 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXLEN, 0,
3590 0 : UTF8_ALLOW_ANY);
3598 PL_reginput = locinput;
3602 if (ln && regrepeat(scan, ln) < ln)
3604 locinput = PL_reginput;
3607 char *e; /* Should not check after this */
3608 char *old = locinput;
3611 if (n == REG_INFTY) {
3614 while (UTF8_IS_CONTINUATION(*(U8*)e))
3620 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3624 e = locinput + n - ln;
3629 /* Find place 'next' could work */
3632 while (locinput <= e &&
3633 UCHARAT(locinput) != c1)
3636 while (locinput <= e
3637 && UCHARAT(locinput) != c1
3638 && UCHARAT(locinput) != c2)
3641 count = locinput - old;
3646 /* count initialised to
3647 * utf8_distance(old, locinput) */
3648 while (locinput <= e &&
3649 utf8n_to_uvchr((U8*)locinput,
3652 0 : UTF8_ALLOW_ANY) != (UV)c1) {
3657 /* count initialised to
3658 * utf8_distance(old, locinput) */
3659 while (locinput <= e) {
3660 UV c = utf8n_to_uvchr((U8*)locinput,
3663 0 : UTF8_ALLOW_ANY);
3664 if (c == (UV)c1 || c == (UV)c2)
3673 /* PL_reginput == old now */
3674 if (locinput != old) {
3675 ln = 1; /* Did some */
3676 if (regrepeat(scan, count) < count)
3679 /* PL_reginput == locinput now */
3680 TRYPAREN(paren, ln, locinput);
3681 PL_reginput = locinput; /* Could be reset... */
3682 REGCP_UNWIND(lastcp);
3683 /* Couldn't or didn't -- move forward. */
3686 locinput += UTF8SKIP(locinput);
3693 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3697 c = utf8n_to_uvchr((U8*)PL_reginput,
3700 0 : UTF8_ALLOW_ANY);
3702 c = UCHARAT(PL_reginput);
3703 /* If it could work, try it. */
3704 if (c == (UV)c1 || c == (UV)c2)
3706 TRYPAREN(paren, ln, PL_reginput);
3707 REGCP_UNWIND(lastcp);
3710 /* If it could work, try it. */
3711 else if (c1 == -1000)
3713 TRYPAREN(paren, ln, PL_reginput);
3714 REGCP_UNWIND(lastcp);
3716 /* Couldn't or didn't -- move forward. */
3717 PL_reginput = locinput;
3718 if (regrepeat(scan, 1)) {
3720 locinput = PL_reginput;
3728 n = regrepeat(scan, n);
3729 locinput = PL_reginput;
3730 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3731 ((!PL_multiline && OP(next) != MEOL) ||
3732 OP(next) == SEOL || OP(next) == EOS))
3734 ln = n; /* why back off? */
3735 /* ...because $ and \Z can match before *and* after
3736 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3737 We should back off by one in this case. */
3738 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3747 c = utf8n_to_uvchr((U8*)PL_reginput,
3750 0 : UTF8_ALLOW_ANY);
3752 c = UCHARAT(PL_reginput);
3754 /* If it could work, try it. */
3755 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3757 TRYPAREN(paren, n, PL_reginput);
3758 REGCP_UNWIND(lastcp);
3760 /* Couldn't or didn't -- back up. */
3762 PL_reginput = locinput = HOPc(locinput, -1);
3770 c = utf8n_to_uvchr((U8*)PL_reginput,
3773 0 : UTF8_ALLOW_ANY);
3775 c = UCHARAT(PL_reginput);
3777 /* If it could work, try it. */
3778 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3780 TRYPAREN(paren, n, PL_reginput);
3781 REGCP_UNWIND(lastcp);
3783 /* Couldn't or didn't -- back up. */
3785 PL_reginput = locinput = HOPc(locinput, -1);
3792 if (PL_reg_call_cc) {
3793 re_cc_state *cur_call_cc = PL_reg_call_cc;
3794 CURCUR *cctmp = PL_regcc;
3795 regexp *re = PL_reg_re;
3796 CHECKPOINT cp, lastcp;
3798 cp = regcppush(0); /* Save *all* the positions. */
3800 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3802 PL_reginput = locinput; /* Make position available to
3804 cache_re(PL_reg_call_cc->re);
3805 PL_regcc = PL_reg_call_cc->cc;
3806 PL_reg_call_cc = PL_reg_call_cc->prev;
3807 if (regmatch(cur_call_cc->node)) {
3808 PL_reg_call_cc = cur_call_cc;
3812 REGCP_UNWIND(lastcp);
3814 PL_reg_call_cc = cur_call_cc;
3820 PerlIO_printf(Perl_debug_log,
3821 "%*s continuation failed...\n",
3822 REPORT_CODE_OFF+PL_regindent*2, "")
3826 if (locinput < PL_regtill) {
3827 DEBUG_r(PerlIO_printf(Perl_debug_log,
3828 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3830 (long)(locinput - PL_reg_starttry),
3831 (long)(PL_regtill - PL_reg_starttry),
3833 sayNO_FINAL; /* Cannot match: too short. */
3835 PL_reginput = locinput; /* put where regtry can find it */
3836 sayYES_FINAL; /* Success! */
3838 PL_reginput = locinput; /* put where regtry can find it */
3839 sayYES_LOUD; /* Success! */
3842 PL_reginput = locinput;
3847 s = HOPBACKc(locinput, scan->flags);
3853 PL_reginput = locinput;