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, 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) {
2107 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
2112 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2113 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2114 /* prepare for quick setting of pos */
2115 sv_magic(PL_reg_sv, (SV*)0,
2116 PERL_MAGIC_regex_global, Nullch, 0);
2117 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
2121 PL_reg_oldpos = mg->mg_len;
2122 SAVEDESTRUCTOR_X(restore_pos, 0);
2124 if (!PL_reg_curpm) {
2125 Newz(22,PL_reg_curpm, 1, PMOP);
2128 SV* repointer = newSViv(0);
2129 /* so we know which PL_regex_padav element is PL_reg_curpm */
2130 SvFLAGS(repointer) |= SVf_BREAK;
2131 av_push(PL_regex_padav,repointer);
2132 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2133 PL_regex_pad = AvARRAY(PL_regex_padav);
2137 PM_SETRE(PL_reg_curpm, prog);
2138 PL_reg_oldcurpm = PL_curpm;
2139 PL_curpm = PL_reg_curpm;
2140 if (RX_MATCH_COPIED(prog)) {
2141 /* Here is a serious problem: we cannot rewrite subbeg,
2142 since it may be needed if this match fails. Thus
2143 $` inside (?{}) could fail... */
2144 PL_reg_oldsaved = prog->subbeg;
2145 PL_reg_oldsavedlen = prog->sublen;
2146 #ifdef PERL_COPY_ON_WRITE
2147 PL_nrs = prog->saved_copy;
2149 RX_MATCH_COPIED_off(prog);
2152 PL_reg_oldsaved = Nullch;
2153 prog->subbeg = PL_bostr;
2154 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2156 prog->startp[0] = startpos - PL_bostr;
2157 PL_reginput = startpos;
2158 PL_regstartp = prog->startp;
2159 PL_regendp = prog->endp;
2160 PL_reglastparen = &prog->lastparen;
2161 PL_reglastcloseparen = &prog->lastcloseparen;
2162 prog->lastparen = 0;
2163 prog->lastcloseparen = 0;
2165 DEBUG_r(PL_reg_starttry = startpos);
2166 if (PL_reg_start_tmpl <= prog->nparens) {
2167 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2168 if(PL_reg_start_tmp)
2169 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2171 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2174 /* XXXX What this code is doing here?!!! There should be no need
2175 to do this again and again, PL_reglastparen should take care of
2178 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2179 * Actually, the code in regcppop() (which Ilya may be meaning by
2180 * PL_reglastparen), is not needed at all by the test suite
2181 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2182 * enough, for building DynaLoader, or otherwise this
2183 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2184 * will happen. Meanwhile, this code *is* needed for the
2185 * above-mentioned test suite tests to succeed. The common theme
2186 * on those tests seems to be returning null fields from matches.
2191 if (prog->nparens) {
2192 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2199 if (regmatch(prog->program + 1)) {
2200 prog->endp[0] = PL_reginput - PL_bostr;
2203 REGCP_UNWIND(lastcp);
2207 #define RE_UNWIND_BRANCH 1
2208 #define RE_UNWIND_BRANCHJ 2
2212 typedef struct { /* XX: makes sense to enlarge it... */
2216 } re_unwind_generic_t;
2229 } re_unwind_branch_t;
2231 typedef union re_unwind_t {
2233 re_unwind_generic_t generic;
2234 re_unwind_branch_t branch;
2237 #define sayYES goto yes
2238 #define sayNO goto no
2239 #define sayNO_ANYOF goto no_anyof
2240 #define sayYES_FINAL goto yes_final
2241 #define sayYES_LOUD goto yes_loud
2242 #define sayNO_FINAL goto no_final
2243 #define sayNO_SILENT goto do_no
2244 #define saySAME(x) if (x) goto yes; else goto no
2246 #define REPORT_CODE_OFF 24
2249 - regmatch - main matching routine
2251 * Conceptually the strategy is simple: check to see whether the current
2252 * node matches, call self recursively to see whether the rest matches,
2253 * and then act accordingly. In practice we make some effort to avoid
2254 * recursion, in particular by going through "ordinary" nodes (that don't
2255 * need to know whether the rest of the match failed) by a loop instead of
2258 /* [lwall] I've hoisted the register declarations to the outer block in order to
2259 * maybe save a little bit of pushing and popping on the stack. It also takes
2260 * advantage of machines that use a register save mask on subroutine entry.
2262 STATIC I32 /* 0 failure, 1 success */
2263 S_regmatch(pTHX_ regnode *prog)
2265 register regnode *scan; /* Current node. */
2266 regnode *next; /* Next node. */
2267 regnode *inner; /* Next node in internal branch. */
2268 register I32 nextchr; /* renamed nextchr - nextchar colides with
2269 function of same name */
2270 register I32 n; /* no or next */
2271 register I32 ln = 0; /* len or last */
2272 register char *s = Nullch; /* operand or save */
2273 register char *locinput = PL_reginput;
2274 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2275 int minmod = 0, sw = 0, logical = 0;
2278 I32 firstcp = PL_savestack_ix;
2280 register bool do_utf8 = PL_reg_match_utf8;
2282 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2283 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2284 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2291 /* Note that nextchr is a byte even in UTF */
2292 nextchr = UCHARAT(locinput);
2294 while (scan != NULL) {
2297 SV *prop = sv_newmortal();
2298 int docolor = *PL_colors[0];
2299 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2300 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2301 /* The part of the string before starttry has one color
2302 (pref0_len chars), between starttry and current
2303 position another one (pref_len - pref0_len chars),
2304 after the current position the third one.
2305 We assume that pref0_len <= pref_len, otherwise we
2306 decrease pref0_len. */
2307 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2308 ? (5 + taill) - l : locinput - PL_bostr;
2311 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2313 pref0_len = pref_len - (locinput - PL_reg_starttry);
2314 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2315 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2316 ? (5 + taill) - pref_len : PL_regeol - locinput);
2317 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2321 if (pref0_len > pref_len)
2322 pref0_len = pref_len;
2323 regprop(prop, scan);
2326 do_utf8 && OP(scan) != CANY ?
2327 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2328 pref0_len, 60, UNI_DISPLAY_REGEX) :
2329 locinput - pref_len;
2330 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2331 char *s1 = do_utf8 && OP(scan) != CANY ?
2332 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2333 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2334 locinput - pref_len + pref0_len;
2335 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2336 char *s2 = do_utf8 && OP(scan) != CANY ?
2337 pv_uni_display(dsv2, (U8*)locinput,
2338 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2340 int len2 = do_utf8 ? strlen(s2) : l;
2341 PerlIO_printf(Perl_debug_log,
2342 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2343 (IV)(locinput - PL_bostr),
2350 (docolor ? "" : "> <"),
2354 15 - l - pref_len + 1,
2356 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2361 next = scan + NEXT_OFF(scan);
2367 if (locinput == PL_bostr || (PL_multiline &&
2368 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2370 /* regtill = regbol; */
2375 if (locinput == PL_bostr ||
2376 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2382 if (locinput == PL_bostr)
2386 if (locinput == PL_reg_ganch)
2396 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2401 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2403 if (PL_regeol - locinput > 1)
2407 if (PL_regeol != locinput)
2411 if (!nextchr && locinput >= PL_regeol)
2414 locinput += PL_utf8skip[nextchr];
2415 if (locinput > PL_regeol)
2417 nextchr = UCHARAT(locinput);
2420 nextchr = UCHARAT(++locinput);
2423 if (!nextchr && locinput >= PL_regeol)
2425 nextchr = UCHARAT(++locinput);
2428 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2431 locinput += PL_utf8skip[nextchr];
2432 if (locinput > PL_regeol)
2434 nextchr = UCHARAT(locinput);
2437 nextchr = UCHARAT(++locinput);
2442 if (do_utf8 != UTF) {
2443 /* The target and the pattern have differing utf8ness. */
2449 /* The target is utf8, the pattern is not utf8. */
2453 if (NATIVE_TO_UNI(*(U8*)s) !=
2454 utf8n_to_uvuni((U8*)l, UTF8_MAXLEN, &ulen,
2456 0 : UTF8_ALLOW_ANY))
2463 /* The target is not utf8, the pattern is utf8. */
2467 if (NATIVE_TO_UNI(*((U8*)l)) !=
2468 utf8n_to_uvuni((U8*)s, UTF8_MAXLEN, &ulen,
2470 0 : UTF8_ALLOW_ANY))
2477 nextchr = UCHARAT(locinput);
2480 /* The target and the pattern have the same utf8ness. */
2481 /* Inline the first character, for speed. */
2482 if (UCHARAT(s) != nextchr)
2484 if (PL_regeol - locinput < ln)
2486 if (ln > 1 && memNE(s, locinput, ln))
2489 nextchr = UCHARAT(locinput);
2492 PL_reg_flags |= RF_tainted;
2498 if (do_utf8 || UTF) {
2499 /* Either target or the pattern are utf8. */
2501 char *e = PL_regeol;
2503 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2504 l, &e, 0, do_utf8)) {
2505 /* One more case for the sharp s:
2506 * pack("U0U*", 0xDF) =~ /ss/i,
2507 * the 0xC3 0x9F are the UTF-8
2508 * byte sequence for the U+00DF. */
2510 toLOWER(s[0]) == 's' &&
2512 toLOWER(s[1]) == 's' &&
2519 nextchr = UCHARAT(locinput);
2523 /* Neither the target and the pattern are utf8. */
2525 /* Inline the first character, for speed. */
2526 if (UCHARAT(s) != nextchr &&
2527 UCHARAT(s) != ((OP(scan) == EXACTF)
2528 ? PL_fold : PL_fold_locale)[nextchr])
2530 if (PL_regeol - locinput < ln)
2532 if (ln > 1 && (OP(scan) == EXACTF
2533 ? ibcmp(s, locinput, ln)
2534 : ibcmp_locale(s, locinput, ln)))
2537 nextchr = UCHARAT(locinput);
2541 STRLEN inclasslen = PL_regeol - locinput;
2543 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2545 if (locinput >= PL_regeol)
2547 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2548 nextchr = UCHARAT(locinput);
2553 nextchr = UCHARAT(locinput);
2554 if (!REGINCLASS(scan, (U8*)locinput))
2556 if (!nextchr && locinput >= PL_regeol)
2558 nextchr = UCHARAT(++locinput);
2562 /* If we might have the case of the German sharp s
2563 * in a casefolding Unicode character class. */
2565 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2566 locinput += SHARP_S_SKIP;
2567 nextchr = UCHARAT(locinput);
2573 PL_reg_flags |= RF_tainted;
2579 LOAD_UTF8_CHARCLASS(alnum,"a");
2580 if (!(OP(scan) == ALNUM
2581 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2582 : isALNUM_LC_utf8((U8*)locinput)))
2586 locinput += PL_utf8skip[nextchr];
2587 nextchr = UCHARAT(locinput);
2590 if (!(OP(scan) == ALNUM
2591 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2593 nextchr = UCHARAT(++locinput);
2596 PL_reg_flags |= RF_tainted;
2599 if (!nextchr && locinput >= PL_regeol)
2602 LOAD_UTF8_CHARCLASS(alnum,"a");
2603 if (OP(scan) == NALNUM
2604 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2605 : isALNUM_LC_utf8((U8*)locinput))
2609 locinput += PL_utf8skip[nextchr];
2610 nextchr = UCHARAT(locinput);
2613 if (OP(scan) == NALNUM
2614 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2616 nextchr = UCHARAT(++locinput);
2620 PL_reg_flags |= RF_tainted;
2624 /* was last char in word? */
2626 if (locinput == PL_bostr)
2629 U8 *r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2631 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2633 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2634 ln = isALNUM_uni(ln);
2635 LOAD_UTF8_CHARCLASS(alnum,"a");
2636 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2639 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2640 n = isALNUM_LC_utf8((U8*)locinput);
2644 ln = (locinput != PL_bostr) ?
2645 UCHARAT(locinput - 1) : '\n';
2646 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2648 n = isALNUM(nextchr);
2651 ln = isALNUM_LC(ln);
2652 n = isALNUM_LC(nextchr);
2655 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2656 OP(scan) == BOUNDL))
2660 PL_reg_flags |= RF_tainted;
2666 if (UTF8_IS_CONTINUED(nextchr)) {
2667 LOAD_UTF8_CHARCLASS(space," ");
2668 if (!(OP(scan) == SPACE
2669 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2670 : isSPACE_LC_utf8((U8*)locinput)))
2674 locinput += PL_utf8skip[nextchr];
2675 nextchr = UCHARAT(locinput);
2678 if (!(OP(scan) == SPACE
2679 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2681 nextchr = UCHARAT(++locinput);
2684 if (!(OP(scan) == SPACE
2685 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2687 nextchr = UCHARAT(++locinput);
2691 PL_reg_flags |= RF_tainted;
2694 if (!nextchr && locinput >= PL_regeol)
2697 LOAD_UTF8_CHARCLASS(space," ");
2698 if (OP(scan) == NSPACE
2699 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2700 : isSPACE_LC_utf8((U8*)locinput))
2704 locinput += PL_utf8skip[nextchr];
2705 nextchr = UCHARAT(locinput);
2708 if (OP(scan) == NSPACE
2709 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2711 nextchr = UCHARAT(++locinput);
2714 PL_reg_flags |= RF_tainted;
2720 LOAD_UTF8_CHARCLASS(digit,"0");
2721 if (!(OP(scan) == DIGIT
2722 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2723 : isDIGIT_LC_utf8((U8*)locinput)))
2727 locinput += PL_utf8skip[nextchr];
2728 nextchr = UCHARAT(locinput);
2731 if (!(OP(scan) == DIGIT
2732 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2734 nextchr = UCHARAT(++locinput);
2737 PL_reg_flags |= RF_tainted;
2740 if (!nextchr && locinput >= PL_regeol)
2743 LOAD_UTF8_CHARCLASS(digit,"0");
2744 if (OP(scan) == NDIGIT
2745 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2746 : isDIGIT_LC_utf8((U8*)locinput))
2750 locinput += PL_utf8skip[nextchr];
2751 nextchr = UCHARAT(locinput);
2754 if (OP(scan) == NDIGIT
2755 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2757 nextchr = UCHARAT(++locinput);
2760 if (locinput >= PL_regeol)
2763 LOAD_UTF8_CHARCLASS(mark,"~");
2764 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2766 locinput += PL_utf8skip[nextchr];
2767 while (locinput < PL_regeol &&
2768 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2769 locinput += UTF8SKIP(locinput);
2770 if (locinput > PL_regeol)
2775 nextchr = UCHARAT(locinput);
2778 PL_reg_flags |= RF_tainted;
2782 n = ARG(scan); /* which paren pair */
2783 ln = PL_regstartp[n];
2784 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2785 if ((I32)*PL_reglastparen < n || ln == -1)
2786 sayNO; /* Do not match unless seen CLOSEn. */
2787 if (ln == PL_regendp[n])
2791 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2793 char *e = PL_bostr + PL_regendp[n];
2795 * Note that we can't do the "other character" lookup trick as
2796 * in the 8-bit case (no pun intended) because in Unicode we
2797 * have to map both upper and title case to lower case.
2799 if (OP(scan) == REFF) {
2800 STRLEN ulen1, ulen2;
2801 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2802 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2806 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2807 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2808 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2815 nextchr = UCHARAT(locinput);
2819 /* Inline the first character, for speed. */
2820 if (UCHARAT(s) != nextchr &&
2822 (UCHARAT(s) != ((OP(scan) == REFF
2823 ? PL_fold : PL_fold_locale)[nextchr]))))
2825 ln = PL_regendp[n] - ln;
2826 if (locinput + ln > PL_regeol)
2828 if (ln > 1 && (OP(scan) == REF
2829 ? memNE(s, locinput, ln)
2831 ? ibcmp(s, locinput, ln)
2832 : ibcmp_locale(s, locinput, ln))))
2835 nextchr = UCHARAT(locinput);
2846 OP_4tree *oop = PL_op;
2847 COP *ocurcop = PL_curcop;
2850 struct regexp *oreg = PL_reg_re;
2853 PL_op = (OP_4tree*)PL_regdata->data[n];
2854 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2855 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
2856 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2860 CALLRUNOPS(aTHX); /* Scalar context. */
2863 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
2871 PAD_RESTORE_LOCAL(old_comppad);
2872 PL_curcop = ocurcop;
2874 if (logical == 2) { /* Postponed subexpression. */
2876 MAGIC *mg = Null(MAGIC*);
2878 CHECKPOINT cp, lastcp;
2882 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
2883 mg = mg_find(sv, PERL_MAGIC_qr);
2884 else if (SvSMAGICAL(ret)) {
2885 if (SvGMAGICAL(ret))
2886 sv_unmagic(ret, PERL_MAGIC_qr);
2888 mg = mg_find(ret, PERL_MAGIC_qr);
2892 re = (regexp *)mg->mg_obj;
2893 (void)ReREFCNT_inc(re);
2897 char *t = SvPV(ret, len);
2899 char *oprecomp = PL_regprecomp;
2900 I32 osize = PL_regsize;
2901 I32 onpar = PL_regnpar;
2904 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
2905 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2907 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
2909 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2911 PL_regprecomp = oprecomp;
2916 PerlIO_printf(Perl_debug_log,
2917 "Entering embedded `%s%.60s%s%s'\n",
2921 (strlen(re->precomp) > 60 ? "..." : ""))
2924 state.prev = PL_reg_call_cc;
2925 state.cc = PL_regcc;
2926 state.re = PL_reg_re;
2930 cp = regcppush(0); /* Save *all* the positions. */
2933 state.ss = PL_savestack_ix;
2934 *PL_reglastparen = 0;
2935 *PL_reglastcloseparen = 0;
2936 PL_reg_call_cc = &state;
2937 PL_reginput = locinput;
2938 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
2939 ((re->reganch & ROPT_UTF8) != 0);
2940 if (toggleutf) PL_reg_flags ^= RF_utf8;
2942 /* XXXX This is too dramatic a measure... */
2945 if (regmatch(re->program + 1)) {
2946 /* Even though we succeeded, we need to restore
2947 global variables, since we may be wrapped inside
2948 SUSPEND, thus the match may be not finished yet. */
2950 /* XXXX Do this only if SUSPENDed? */
2951 PL_reg_call_cc = state.prev;
2952 PL_regcc = state.cc;
2953 PL_reg_re = state.re;
2954 cache_re(PL_reg_re);
2955 if (toggleutf) PL_reg_flags ^= RF_utf8;
2957 /* XXXX This is too dramatic a measure... */
2960 /* These are needed even if not SUSPEND. */
2966 REGCP_UNWIND(lastcp);
2968 PL_reg_call_cc = state.prev;
2969 PL_regcc = state.cc;
2970 PL_reg_re = state.re;
2971 cache_re(PL_reg_re);
2972 if (toggleutf) PL_reg_flags ^= RF_utf8;
2974 /* XXXX This is too dramatic a measure... */
2984 sv_setsv(save_scalar(PL_replgv), ret);
2990 n = ARG(scan); /* which paren pair */
2991 PL_reg_start_tmp[n] = locinput;
2996 n = ARG(scan); /* which paren pair */
2997 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2998 PL_regendp[n] = locinput - PL_bostr;
2999 if (n > (I32)*PL_reglastparen)
3000 *PL_reglastparen = n;
3001 *PL_reglastcloseparen = n;
3004 n = ARG(scan); /* which paren pair */
3005 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
3008 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3010 next = NEXTOPER(NEXTOPER(scan));
3012 next = scan + ARG(scan);
3013 if (OP(next) == IFTHEN) /* Fake one. */
3014 next = NEXTOPER(NEXTOPER(next));
3018 logical = scan->flags;
3020 /*******************************************************************
3021 PL_regcc contains infoblock about the innermost (...)* loop, and
3022 a pointer to the next outer infoblock.
3024 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3026 1) After matching X, regnode for CURLYX is processed;
3028 2) This regnode creates infoblock on the stack, and calls
3029 regmatch() recursively with the starting point at WHILEM node;
3031 3) Each hit of WHILEM node tries to match A and Z (in the order
3032 depending on the current iteration, min/max of {min,max} and
3033 greediness). The information about where are nodes for "A"
3034 and "Z" is read from the infoblock, as is info on how many times "A"
3035 was already matched, and greediness.
3037 4) After A matches, the same WHILEM node is hit again.
3039 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3040 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3041 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3042 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3043 of the external loop.
3045 Currently present infoblocks form a tree with a stem formed by PL_curcc
3046 and whatever it mentions via ->next, and additional attached trees
3047 corresponding to temporarily unset infoblocks as in "5" above.
3049 In the following picture infoblocks for outer loop of
3050 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3051 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3052 infoblocks are drawn below the "reset" infoblock.
3054 In fact in the picture below we do not show failed matches for Z and T
3055 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3056 more obvious *why* one needs to *temporary* unset infoblocks.]
3058 Matched REx position InfoBlocks Comment
3062 Y A)*?Z)*?T x <- O <- I
3063 YA )*?Z)*?T x <- O <- I
3064 YA A)*?Z)*?T x <- O <- I
3065 YAA )*?Z)*?T x <- O <- I
3066 YAA Z)*?T x <- O # Temporary unset I
3069 YAAZ Y(A)*?Z)*?T x <- O
3072 YAAZY (A)*?Z)*?T x <- O
3075 YAAZY A)*?Z)*?T x <- O <- I
3078 YAAZYA )*?Z)*?T x <- O <- I
3081 YAAZYA Z)*?T x <- O # Temporary unset I
3087 YAAZYAZ T x # Temporary unset O
3094 *******************************************************************/
3097 CHECKPOINT cp = PL_savestack_ix;
3098 /* No need to save/restore up to this paren */
3099 I32 parenfloor = scan->flags;
3101 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3103 cc.oldcc = PL_regcc;
3105 /* XXXX Probably it is better to teach regpush to support
3106 parenfloor > PL_regsize... */
3107 if (parenfloor > (I32)*PL_reglastparen)
3108 parenfloor = *PL_reglastparen; /* Pessimization... */
3109 cc.parenfloor = parenfloor;
3111 cc.min = ARG1(scan);
3112 cc.max = ARG2(scan);
3113 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3117 PL_reginput = locinput;
3118 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3120 PL_regcc = cc.oldcc;
3126 * This is really hard to understand, because after we match
3127 * what we're trying to match, we must make sure the rest of
3128 * the REx is going to match for sure, and to do that we have
3129 * to go back UP the parse tree by recursing ever deeper. And
3130 * if it fails, we have to reset our parent's current state
3131 * that we can try again after backing off.
3134 CHECKPOINT cp, lastcp;
3135 CURCUR* cc = PL_regcc;
3136 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3138 n = cc->cur + 1; /* how many we know we matched */
3139 PL_reginput = locinput;
3142 PerlIO_printf(Perl_debug_log,
3143 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3144 REPORT_CODE_OFF+PL_regindent*2, "",
3145 (long)n, (long)cc->min,
3146 (long)cc->max, PTR2UV(cc))
3149 /* If degenerate scan matches "", assume scan done. */
3151 if (locinput == cc->lastloc && n >= cc->min) {
3152 PL_regcc = cc->oldcc;
3156 PerlIO_printf(Perl_debug_log,
3157 "%*s empty match detected, try continuation...\n",
3158 REPORT_CODE_OFF+PL_regindent*2, "")
3160 if (regmatch(cc->next))
3168 /* First just match a string of min scans. */
3172 cc->lastloc = locinput;
3173 if (regmatch(cc->scan))
3176 cc->lastloc = lastloc;
3181 /* Check whether we already were at this position.
3182 Postpone detection until we know the match is not
3183 *that* much linear. */
3184 if (!PL_reg_maxiter) {
3185 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3186 PL_reg_leftiter = PL_reg_maxiter;
3188 if (PL_reg_leftiter-- == 0) {
3189 I32 size = (PL_reg_maxiter + 7)/8;
3190 if (PL_reg_poscache) {
3191 if ((I32)PL_reg_poscache_size < size) {
3192 Renew(PL_reg_poscache, size, char);
3193 PL_reg_poscache_size = size;
3195 Zero(PL_reg_poscache, size, char);
3198 PL_reg_poscache_size = size;
3199 Newz(29, PL_reg_poscache, size, char);
3202 PerlIO_printf(Perl_debug_log,
3203 "%sDetected a super-linear match, switching on caching%s...\n",
3204 PL_colors[4], PL_colors[5])
3207 if (PL_reg_leftiter < 0) {
3208 I32 o = locinput - PL_bostr, b;
3210 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3213 if (PL_reg_poscache[o] & (1<<b)) {
3215 PerlIO_printf(Perl_debug_log,
3216 "%*s already tried at this position...\n",
3217 REPORT_CODE_OFF+PL_regindent*2, "")
3219 if (PL_reg_flags & RF_false)
3224 PL_reg_poscache[o] |= (1<<b);
3228 /* Prefer next over scan for minimal matching. */
3231 PL_regcc = cc->oldcc;
3234 cp = regcppush(cc->parenfloor);
3236 if (regmatch(cc->next)) {
3238 sayYES; /* All done. */
3240 REGCP_UNWIND(lastcp);
3246 if (n >= cc->max) { /* Maximum greed exceeded? */
3247 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3248 && !(PL_reg_flags & RF_warned)) {
3249 PL_reg_flags |= RF_warned;
3250 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3251 "Complex regular subexpression recursion",
3258 PerlIO_printf(Perl_debug_log,
3259 "%*s trying longer...\n",
3260 REPORT_CODE_OFF+PL_regindent*2, "")
3262 /* Try scanning more and see if it helps. */
3263 PL_reginput = locinput;
3265 cc->lastloc = locinput;
3266 cp = regcppush(cc->parenfloor);
3268 if (regmatch(cc->scan)) {
3272 REGCP_UNWIND(lastcp);
3275 cc->lastloc = lastloc;
3279 /* Prefer scan over next for maximal matching. */
3281 if (n < cc->max) { /* More greed allowed? */
3282 cp = regcppush(cc->parenfloor);
3284 cc->lastloc = locinput;
3286 if (regmatch(cc->scan)) {
3290 REGCP_UNWIND(lastcp);
3291 regcppop(); /* Restore some previous $<digit>s? */
3292 PL_reginput = locinput;
3294 PerlIO_printf(Perl_debug_log,
3295 "%*s failed, try continuation...\n",
3296 REPORT_CODE_OFF+PL_regindent*2, "")
3299 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3300 && !(PL_reg_flags & RF_warned)) {
3301 PL_reg_flags |= RF_warned;
3302 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3303 "Complex regular subexpression recursion",
3307 /* Failed deeper matches of scan, so see if this one works. */
3308 PL_regcc = cc->oldcc;
3311 if (regmatch(cc->next))
3317 cc->lastloc = lastloc;
3322 next = scan + ARG(scan);
3325 inner = NEXTOPER(NEXTOPER(scan));
3328 inner = NEXTOPER(scan);
3332 if (OP(next) != c1) /* No choice. */
3333 next = inner; /* Avoid recursion. */
3335 I32 lastparen = *PL_reglastparen;
3337 re_unwind_branch_t *uw;
3339 /* Put unwinding data on stack */
3340 unwind1 = SSNEWt(1,re_unwind_branch_t);
3341 uw = SSPTRt(unwind1,re_unwind_branch_t);
3344 uw->type = ((c1 == BRANCH)
3346 : RE_UNWIND_BRANCHJ);
3347 uw->lastparen = lastparen;
3349 uw->locinput = locinput;
3350 uw->nextchr = nextchr;
3352 uw->regindent = ++PL_regindent;
3355 REGCP_SET(uw->lastcp);
3357 /* Now go into the first branch */
3370 /* We suppose that the next guy does not need
3371 backtracking: in particular, it is of constant length,
3372 and has no parenths to influence future backrefs. */
3373 ln = ARG1(scan); /* min to match */
3374 n = ARG2(scan); /* max to match */
3375 paren = scan->flags;
3377 if (paren > PL_regsize)
3379 if (paren > (I32)*PL_reglastparen)
3380 *PL_reglastparen = paren;
3382 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3384 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3385 PL_reginput = locinput;
3388 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3390 /* if we matched something zero-length we don't need to
3391 backtrack - capturing parens are already defined, so
3392 the caveat in the maximal case doesn't apply
3394 XXXX if ln == 0, we can redo this check first time
3395 through the following loop
3398 n = ln; /* don't backtrack */
3399 locinput = PL_reginput;
3400 if (HAS_TEXT(next) || JUMPABLE(next)) {
3401 regnode *text_node = next;
3403 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3405 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3407 if (PL_regkind[(U8)OP(text_node)] == REF) {
3411 else { c1 = (U8)*STRING(text_node); }
3412 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3414 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3415 c2 = PL_fold_locale[c1];
3424 /* This may be improved if l == 0. */
3425 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3426 /* If it could work, try it. */
3428 UCHARAT(PL_reginput) == c1 ||
3429 UCHARAT(PL_reginput) == c2)
3433 PL_regstartp[paren] =
3434 HOPc(PL_reginput, -l) - PL_bostr;
3435 PL_regendp[paren] = PL_reginput - PL_bostr;
3438 PL_regendp[paren] = -1;
3442 REGCP_UNWIND(lastcp);
3444 /* Couldn't or didn't -- move forward. */
3445 PL_reginput = locinput;
3446 if (regrepeat_hard(scan, 1, &l)) {
3448 locinput = PL_reginput;
3455 n = regrepeat_hard(scan, n, &l);
3456 /* if we matched something zero-length we don't need to
3457 backtrack, unless the minimum count is zero and we
3458 are capturing the result - in that case the capture
3459 being defined or not may affect later execution
3461 if (n != 0 && l == 0 && !(paren && ln == 0))
3462 ln = n; /* don't backtrack */
3463 locinput = PL_reginput;
3465 PerlIO_printf(Perl_debug_log,
3466 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3467 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3471 if (HAS_TEXT(next) || JUMPABLE(next)) {
3472 regnode *text_node = next;
3474 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3476 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3478 if (PL_regkind[(U8)OP(text_node)] == REF) {
3482 else { c1 = (U8)*STRING(text_node); }
3484 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3486 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3487 c2 = PL_fold_locale[c1];
3498 /* If it could work, try it. */
3500 UCHARAT(PL_reginput) == c1 ||
3501 UCHARAT(PL_reginput) == c2)
3504 PerlIO_printf(Perl_debug_log,
3505 "%*s trying tail with n=%"IVdf"...\n",
3506 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3510 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3511 PL_regendp[paren] = PL_reginput - PL_bostr;
3514 PL_regendp[paren] = -1;
3518 REGCP_UNWIND(lastcp);
3520 /* Couldn't or didn't -- back up. */
3522 locinput = HOPc(locinput, -l);
3523 PL_reginput = locinput;
3530 paren = scan->flags; /* Which paren to set */
3531 if (paren > PL_regsize)
3533 if (paren > (I32)*PL_reglastparen)
3534 *PL_reglastparen = paren;
3535 ln = ARG1(scan); /* min to match */
3536 n = ARG2(scan); /* max to match */
3537 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3541 ln = ARG1(scan); /* min to match */
3542 n = ARG2(scan); /* max to match */
3543 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3548 scan = NEXTOPER(scan);
3554 scan = NEXTOPER(scan);
3558 * Lookahead to avoid useless match attempts
3559 * when we know what character comes next.
3563 * Used to only do .*x and .*?x, but now it allows
3564 * for )'s, ('s and (?{ ... })'s to be in the way
3565 * of the quantifier and the EXACT-like node. -- japhy
3568 if (HAS_TEXT(next) || JUMPABLE(next)) {
3570 regnode *text_node = next;
3572 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3574 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3576 if (PL_regkind[(U8)OP(text_node)] == REF) {
3578 goto assume_ok_easy;
3580 else { s = (U8*)STRING(text_node); }
3584 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3586 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3587 c2 = PL_fold_locale[c1];
3590 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3591 STRLEN ulen1, ulen2;
3592 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3593 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3595 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3596 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3598 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXLEN, 0,
3600 0 : UTF8_ALLOW_ANY);
3601 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXLEN, 0,
3603 0 : UTF8_ALLOW_ANY);
3606 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXLEN, 0,
3608 0 : UTF8_ALLOW_ANY);
3616 PL_reginput = locinput;
3620 if (ln && regrepeat(scan, ln) < ln)
3622 locinput = PL_reginput;
3625 char *e; /* Should not check after this */
3626 char *old = locinput;
3629 if (n == REG_INFTY) {
3632 while (UTF8_IS_CONTINUATION(*(U8*)e))
3638 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3642 e = locinput + n - ln;
3647 /* Find place 'next' could work */
3650 while (locinput <= e &&
3651 UCHARAT(locinput) != c1)
3654 while (locinput <= e
3655 && UCHARAT(locinput) != c1
3656 && UCHARAT(locinput) != c2)
3659 count = locinput - old;
3664 /* count initialised to
3665 * utf8_distance(old, locinput) */
3666 while (locinput <= e &&
3667 utf8n_to_uvchr((U8*)locinput,
3670 0 : UTF8_ALLOW_ANY) != (UV)c1) {
3675 /* count initialised to
3676 * utf8_distance(old, locinput) */
3677 while (locinput <= e) {
3678 UV c = utf8n_to_uvchr((U8*)locinput,
3681 0 : UTF8_ALLOW_ANY);
3682 if (c == (UV)c1 || c == (UV)c2)
3691 /* PL_reginput == old now */
3692 if (locinput != old) {
3693 ln = 1; /* Did some */
3694 if (regrepeat(scan, count) < count)
3697 /* PL_reginput == locinput now */
3698 TRYPAREN(paren, ln, locinput);
3699 PL_reginput = locinput; /* Could be reset... */
3700 REGCP_UNWIND(lastcp);
3701 /* Couldn't or didn't -- move forward. */
3704 locinput += UTF8SKIP(locinput);
3711 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3715 c = utf8n_to_uvchr((U8*)PL_reginput,
3718 0 : UTF8_ALLOW_ANY);
3720 c = UCHARAT(PL_reginput);
3721 /* If it could work, try it. */
3722 if (c == (UV)c1 || c == (UV)c2)
3724 TRYPAREN(paren, ln, PL_reginput);
3725 REGCP_UNWIND(lastcp);
3728 /* If it could work, try it. */
3729 else if (c1 == -1000)
3731 TRYPAREN(paren, ln, PL_reginput);
3732 REGCP_UNWIND(lastcp);
3734 /* Couldn't or didn't -- move forward. */
3735 PL_reginput = locinput;
3736 if (regrepeat(scan, 1)) {
3738 locinput = PL_reginput;
3746 n = regrepeat(scan, n);
3747 locinput = PL_reginput;
3748 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3749 ((!PL_multiline && OP(next) != MEOL) ||
3750 OP(next) == SEOL || OP(next) == EOS))
3752 ln = n; /* why back off? */
3753 /* ...because $ and \Z can match before *and* after
3754 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3755 We should back off by one in this case. */
3756 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3765 c = utf8n_to_uvchr((U8*)PL_reginput,
3768 0 : UTF8_ALLOW_ANY);
3770 c = UCHARAT(PL_reginput);
3772 /* If it could work, try it. */
3773 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3775 TRYPAREN(paren, n, PL_reginput);
3776 REGCP_UNWIND(lastcp);
3778 /* Couldn't or didn't -- back up. */
3780 PL_reginput = locinput = HOPc(locinput, -1);
3788 c = utf8n_to_uvchr((U8*)PL_reginput,
3791 0 : UTF8_ALLOW_ANY);
3793 c = UCHARAT(PL_reginput);
3795 /* If it could work, try it. */
3796 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3798 TRYPAREN(paren, n, PL_reginput);
3799 REGCP_UNWIND(lastcp);
3801 /* Couldn't or didn't -- back up. */
3803 PL_reginput = locinput = HOPc(locinput, -1);
3810 if (PL_reg_call_cc) {
3811 re_cc_state *cur_call_cc = PL_reg_call_cc;
3812 CURCUR *cctmp = PL_regcc;
3813 regexp *re = PL_reg_re;
3814 CHECKPOINT cp, lastcp;
3816 cp = regcppush(0); /* Save *all* the positions. */
3818 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3820 PL_reginput = locinput; /* Make position available to
3822 cache_re(PL_reg_call_cc->re);
3823 PL_regcc = PL_reg_call_cc->cc;
3824 PL_reg_call_cc = PL_reg_call_cc->prev;
3825 if (regmatch(cur_call_cc->node)) {
3826 PL_reg_call_cc = cur_call_cc;
3830 REGCP_UNWIND(lastcp);
3832 PL_reg_call_cc = cur_call_cc;
3838 PerlIO_printf(Perl_debug_log,
3839 "%*s continuation failed...\n",
3840 REPORT_CODE_OFF+PL_regindent*2, "")
3844 if (locinput < PL_regtill) {
3845 DEBUG_r(PerlIO_printf(Perl_debug_log,
3846 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3848 (long)(locinput - PL_reg_starttry),
3849 (long)(PL_regtill - PL_reg_starttry),
3851 sayNO_FINAL; /* Cannot match: too short. */
3853 PL_reginput = locinput; /* put where regtry can find it */
3854 sayYES_FINAL; /* Success! */
3856 PL_reginput = locinput; /* put where regtry can find it */
3857 sayYES_LOUD; /* Success! */
3860 PL_reginput = locinput;
3865 s = HOPBACKc(locinput, scan->flags);
3871 PL_reginput = locinput;
3872 PL_reg_flags ^= RF_false;
3877 s = HOPBACKc(locinput, scan->flags);
3883 PL_reginput = locinput;
3886 inner = NEXTOPER(NEXTOPER(scan));
3887 if (regmatch(inner) != n) {
3889 PL_reg_flags ^= RF_false;
3900 PL_reg_flags ^= RF_false;
3906 if (OP(scan) == SUSPEND) {
3907 locinput = PL_reginput;
3908 nextchr = UCHARAT(locinput);
3913 next = scan + ARG(scan);
3918 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3919 PTR2UV(scan), OP(scan));
3920 Perl_croak(aTHX_ "regexp memory corruption");
3927 * We get here only if there's trouble -- normally "case END" is
3928 * the terminating point.
3930 Perl_croak(aTHX_ "corrupted regexp pointers");
3936 PerlIO_printf(Perl_debug_log,
3937 "%*s %scould match...%s\n",
3938 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3942 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3943 PL_colors[4],PL_colors[5]));
3949 #if 0 /* Breaks $^R */
3957 PerlIO_printf(Perl_debug_log,
3958 "%*s %sfailed...%s\n",
3959 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3965 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3968 case RE_UNWIND_BRANCH:
3969 case RE_UNWIND_BRANCHJ:
3971 re_unwind_branch_t *uwb = &(uw->branch);
3972 I32 lastparen = uwb->lastparen;
3974 REGCP_UNWIND(uwb->lastcp);
3975 for (n = *PL_reglastparen; n > lastparen; n--)
3977 *PL_reglastparen = n;
3978 scan = next = uwb->next;
3980 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3981 ? BRANCH : BRANCHJ) ) { /* Failure */
3988 /* Have more choice yet. Reuse the same uwb. */
3990 if ((n = (uwb->type == RE_UNWIND_BRANCH
3991 ? NEXT_OFF(next) : ARG(next))))
3994 next = NULL; /* XXXX Needn't unwinding in this case... */
3996 next = NEXTOPER(scan);
3997 if (uwb->type == RE_UNWIND_BRANCHJ)
3998 next = NEXTOPER(next);
3999 locinput = uwb->locinput;
4000 nextchr = uwb->nextchr;
4002 PL_regindent = uwb->regindent;
4009 Perl_croak(aTHX_ "regexp unwind memory corruption");
4020 - regrepeat - repeatedly match something simple, report how many
4023 * [This routine now assumes that it will only match on things of length 1.
4024 * That was true before, but now we assume scan - reginput is the count,
4025 * rather than incrementing count on every character. [Er, except utf8.]]
4028 S_regrepeat(pTHX_ regnode *p, I32 max)
4030 register char *scan;
4032 register char *loceol = PL_regeol;
4033 register I32 hardcount = 0;
4034 register bool do_utf8 = PL_reg_match_utf8;
4037 if (max == REG_INFTY)
4039 else if (max < loceol - scan)
4040 loceol = scan + max;
4045 while (scan < loceol && hardcount < max && *scan != '\n') {
4046 scan += UTF8SKIP(scan);
4050 while (scan < loceol && *scan != '\n')
4057 while (scan < loceol && hardcount < max) {
4058 scan += UTF8SKIP(scan);
4068 case EXACT: /* length of string is 1 */
4070 while (scan < loceol && UCHARAT(scan) == c)
4073 case EXACTF: /* length of string is 1 */
4075 while (scan < loceol &&
4076 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4079 case EXACTFL: /* length of string is 1 */
4080 PL_reg_flags |= RF_tainted;
4082 while (scan < loceol &&
4083 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4089 while (hardcount < max && scan < loceol &&
4090 reginclass(p, (U8*)scan, 0, do_utf8)) {
4091 scan += UTF8SKIP(scan);
4095 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4102 LOAD_UTF8_CHARCLASS(alnum,"a");
4103 while (hardcount < max && scan < loceol &&
4104 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4105 scan += UTF8SKIP(scan);
4109 while (scan < loceol && isALNUM(*scan))
4114 PL_reg_flags |= RF_tainted;
4117 while (hardcount < max && scan < loceol &&
4118 isALNUM_LC_utf8((U8*)scan)) {
4119 scan += UTF8SKIP(scan);
4123 while (scan < loceol && isALNUM_LC(*scan))
4130 LOAD_UTF8_CHARCLASS(alnum,"a");
4131 while (hardcount < max && scan < loceol &&
4132 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4133 scan += UTF8SKIP(scan);
4137 while (scan < loceol && !isALNUM(*scan))
4142 PL_reg_flags |= RF_tainted;
4145 while (hardcount < max && scan < loceol &&
4146 !isALNUM_LC_utf8((U8*)scan)) {
4147 scan += UTF8SKIP(scan);
4151 while (scan < loceol && !isALNUM_LC(*scan))
4158 LOAD_UTF8_CHARCLASS(space," ");
4159 while (hardcount < max && scan < loceol &&
4161 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4162 scan += UTF8SKIP(scan);
4166 while (scan < loceol && isSPACE(*scan))
4171 PL_reg_flags |= RF_tainted;
4174 while (hardcount < max && scan < loceol &&
4175 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4176 scan += UTF8SKIP(scan);
4180 while (scan < loceol && isSPACE_LC(*scan))
4187 LOAD_UTF8_CHARCLASS(space," ");
4188 while (hardcount < max && scan < loceol &&
4190 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4191 scan += UTF8SKIP(scan);
4195 while (scan < loceol && !isSPACE(*scan))
4200 PL_reg_flags |= RF_tainted;
4203 while (hardcount < max && scan < loceol &&
4204 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4205 scan += UTF8SKIP(scan);
4209 while (scan < loceol && !isSPACE_LC(*scan))
4216 LOAD_UTF8_CHARCLASS(digit,"0");
4217 while (hardcount < max && scan < loceol &&
4218 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4219 scan += UTF8SKIP(scan);
4223 while (scan < loceol && isDIGIT(*scan))
4230 LOAD_UTF8_CHARCLASS(digit,"0");
4231 while (hardcount < max && scan < loceol &&
4232 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4233 scan += UTF8SKIP(scan);
4237 while (scan < loceol && !isDIGIT(*scan))
4241 default: /* Called on something of 0 width. */
4242 break; /* So match right here or not at all. */
4248 c = scan - PL_reginput;
4253 SV *prop = sv_newmortal();
4256 PerlIO_printf(Perl_debug_log,
4257 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4258 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4265 - regrepeat_hard - repeatedly match something, report total lenth and length
4267 * The repeater is supposed to have constant length.
4271 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4273 register char *scan = Nullch;
4274 register char *start;
4275 register char *loceol = PL_regeol;
4277 I32 count = 0, res = 1;
4282 start = PL_reginput;
4283 if (PL_reg_match_utf8) {
4284 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4287 while (start < PL_reginput) {
4289 start += UTF8SKIP(start);
4300 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4302 *lp = l = PL_reginput - start;
4303 if (max != REG_INFTY && l*max < loceol - scan)
4304 loceol = scan + l*max;
4317 - regclass_swash - prepare the utf8 swash
4321 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4327 if (PL_regdata && PL_regdata->count) {
4330 if (PL_regdata->what[n] == 's') {
4331 SV *rv = (SV*)PL_regdata->data[n];
4332 AV *av = (AV*)SvRV((SV*)rv);
4333 SV **ary = AvARRAY(av);
4336 /* See the end of regcomp.c:S_reglass() for
4337 * documentation of these array elements. */
4340 a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0;
4341 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4345 else if (si && doinit) {
4346 sw = swash_init("utf8", "", si, 1, 0);
4347 (void)av_store(av, 1, sw);
4363 - reginclass - determine if a character falls into a character class
4365 The n is the ANYOF regnode, the p is the target string, lenp
4366 is pointer to the maximum length of how far to go in the p
4367 (if the lenp is zero, UTF8SKIP(p) is used),
4368 do_utf8 tells whether the target string is in UTF-8.
4373 S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8)
4375 char flags = ANYOF_FLAGS(n);
4381 if (do_utf8 && !UTF8_IS_INVARIANT(c))
4382 c = utf8n_to_uvchr(p, UTF8_MAXLEN, &len,
4383 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
4385 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4386 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4389 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4390 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4393 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4397 SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4400 if (swash_fetch(sw, p, do_utf8))
4402 else if (flags & ANYOF_FOLD) {
4403 if (!match && lenp && av) {
4406 for (i = 0; i <= av_len(av); i++) {
4407 SV* sv = *av_fetch(av, i, FALSE);
4409 char *s = SvPV(sv, len);
4411 if (len <= plen && memEQ(s, (char*)p, len)) {
4419 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4422 to_utf8_fold(p, tmpbuf, &tmplen);
4423 if (swash_fetch(sw, tmpbuf, do_utf8))
4429 if (match && lenp && *lenp == 0)
4430 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4432 if (!match && c < 256) {
4433 if (ANYOF_BITMAP_TEST(n, c))
4435 else if (flags & ANYOF_FOLD) {
4438 if (flags & ANYOF_LOCALE) {
4439 PL_reg_flags |= RF_tainted;
4440 f = PL_fold_locale[c];
4444 if (f != c && ANYOF_BITMAP_TEST(n, f))
4448 if (!match && (flags & ANYOF_CLASS)) {
4449 PL_reg_flags |= RF_tainted;
4451 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4452 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4453 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4454 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4455 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4456 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4457 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4458 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4459 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4460 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4461 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4462 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4463 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4464 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4465 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4466 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4467 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4468 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4469 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4470 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4471 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4472 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4473 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4474 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4475 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4476 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4477 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4478 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4479 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4480 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4481 ) /* How's that for a conditional? */
4488 return (flags & ANYOF_INVERT) ? !match : match;
4492 S_reghop(pTHX_ U8 *s, I32 off)
4494 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4498 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4501 while (off-- && s < lim) {
4502 /* XXX could check well-formedness here */
4510 if (UTF8_IS_CONTINUED(*s)) {
4511 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4514 /* XXX could check well-formedness here */
4522 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4524 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4528 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4531 while (off-- && s < lim) {
4532 /* XXX could check well-formedness here */
4542 if (UTF8_IS_CONTINUED(*s)) {
4543 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4546 /* XXX could check well-formedness here */
4558 restore_pos(pTHX_ void *arg)
4560 if (PL_reg_eval_set) {
4561 if (PL_reg_oldsaved) {
4562 PL_reg_re->subbeg = PL_reg_oldsaved;
4563 PL_reg_re->sublen = PL_reg_oldsavedlen;
4564 #ifdef PERL_COPY_ON_WRITE
4565 PL_reg_re->saved_copy = PL_nrs;
4567 RX_MATCH_COPIED_on(PL_reg_re);
4569 PL_reg_magic->mg_len = PL_reg_oldpos;
4570 PL_reg_eval_set = 0;
4571 PL_curpm = PL_reg_oldcurpm;
4576 S_to_utf8_substr(pTHX_ register regexp *prog)
4579 if (prog->float_substr && !prog->float_utf8) {
4580 prog->float_utf8 = sv = NEWSV(117, 0);
4581 SvSetSV(sv, prog->float_substr);
4582 sv_utf8_upgrade(sv);
4583 if (SvTAIL(prog->float_substr))
4585 if (prog->float_substr == prog->check_substr)
4586 prog->check_utf8 = sv;
4588 if (prog->anchored_substr && !prog->anchored_utf8) {
4589 prog->anchored_utf8 = sv = NEWSV(118, 0);
4590 SvSetSV(sv, prog->anchored_substr);
4591 sv_utf8_upgrade(sv);
4592 if (SvTAIL(prog->anchored_substr))
4594 if (prog->anchored_substr == prog->check_substr)
4595 prog->check_utf8 = sv;
4600 S_to_byte_substr(pTHX_ register regexp *prog)
4603 if (prog->float_utf8 && !prog->float_substr) {
4604 prog->float_substr = sv = NEWSV(117, 0);
4605 SvSetSV(sv, prog->float_utf8);
4606 if (sv_utf8_downgrade(sv, TRUE)) {
4607 if (SvTAIL(prog->float_utf8))
4611 prog->float_substr = sv = &PL_sv_undef;
4613 if (prog->float_utf8 == prog->check_utf8)
4614 prog->check_substr = sv;
4616 if (prog->anchored_utf8 && !prog->anchored_substr) {
4617 prog->anchored_substr = sv = NEWSV(118, 0);
4618 SvSetSV(sv, prog->anchored_utf8);
4619 if (sv_utf8_downgrade(sv, TRUE)) {
4620 if (SvTAIL(prog->anchored_utf8))
4624 prog->anchored_substr = sv = &PL_sv_undef;
4626 if (prog->anchored_utf8 == prog->check_utf8)
4627 prog->check_substr = sv;