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) 1991-2001, Larry Wall
72 **** You may distribute under the terms of either the GNU General Public
73 **** License or the Artistic License, as specified in the README file.
75 * Beware that some of this code is subtly aware of the way operator
76 * precedence is structured in regular expressions. Serious changes in
77 * regular-expression syntax might require a total rethink.
80 #define PERL_IN_REGEXEC_C
85 #define RF_tainted 1 /* tainted information used? */
86 #define RF_warned 2 /* warned about big count? */
87 #define RF_evaled 4 /* Did an EVAL with setting? */
88 #define RF_utf8 8 /* String contains multibyte chars? */
90 #define UTF (PL_reg_flags & RF_utf8)
92 #define RS_init 1 /* eval environment created */
93 #define RS_set 2 /* replsv value is set */
103 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
104 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
106 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
107 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
108 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
109 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
110 #define HOPc(pos,off) ((char*)HOP(pos,off))
111 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
113 #define HOPBACK(pos, off) ( \
114 (UTF && PL_reg_match_utf8) \
115 ? reghopmaybe((U8*)pos, -off) \
116 : (pos - off >= PL_bostr) \
120 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
122 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
123 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
124 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
125 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
126 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
127 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
129 #define LOAD_UTF8_CHARCLASS(a,b) STMT_START { if (!CAT2(PL_utf8_,a)) (void)CAT2(is_utf8_, a)((U8*)b); } STMT_END
131 /* for use after a quantifier and before an EXACT-like node -- japhy */
132 #define JUMPABLE(rn) ( \
133 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
134 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
135 OP(rn) == PLUS || OP(rn) == MINMOD || \
136 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
139 #define HAS_TEXT(rn) ( \
140 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
143 #define FIND_NEXT_IMPT(rn) STMT_START { \
144 while (JUMPABLE(rn)) \
145 if (OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
146 PL_regkind[(U8)OP(rn)] == CURLY) \
147 rn = NEXTOPER(NEXTOPER(rn)); \
148 else if (OP(rn) == PLUS) \
150 else rn += NEXT_OFF(rn); \
153 static void restore_pos(pTHX_ void *arg);
156 S_regcppush(pTHX_ I32 parenfloor)
158 int retval = PL_savestack_ix;
159 #define REGCP_PAREN_ELEMS 4
160 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
163 if (paren_elems_to_push < 0)
164 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
166 #define REGCP_OTHER_ELEMS 6
167 SSCHECK(paren_elems_to_push + REGCP_OTHER_ELEMS);
168 for (p = PL_regsize; p > parenfloor; p--) {
169 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
170 SSPUSHINT(PL_regendp[p]);
171 SSPUSHINT(PL_regstartp[p]);
172 SSPUSHPTR(PL_reg_start_tmp[p]);
175 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
176 SSPUSHINT(PL_regsize);
177 SSPUSHINT(*PL_reglastparen);
178 SSPUSHINT(*PL_reglastcloseparen);
179 SSPUSHPTR(PL_reginput);
180 #define REGCP_FRAME_ELEMS 2
181 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
182 * are needed for the regexp context stack bookkeeping. */
183 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
184 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
189 /* These are needed since we do not localize EVAL nodes: */
190 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
191 " Setting an EVAL scope, savestack=%"IVdf"\n", \
192 (IV)PL_savestack_ix)); cp = PL_savestack_ix
194 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
195 PerlIO_printf(Perl_debug_log, \
196 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
197 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
207 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
209 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
210 i = SSPOPINT; /* Parentheses elements to pop. */
211 input = (char *) SSPOPPTR;
212 *PL_reglastcloseparen = SSPOPINT;
213 *PL_reglastparen = SSPOPINT;
214 PL_regsize = SSPOPINT;
216 /* Now restore the parentheses context. */
217 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
218 i > 0; i -= REGCP_PAREN_ELEMS) {
219 paren = (U32)SSPOPINT;
220 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
221 PL_regstartp[paren] = SSPOPINT;
223 if (paren <= *PL_reglastparen)
224 PL_regendp[paren] = tmps;
226 PerlIO_printf(Perl_debug_log,
227 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
228 (UV)paren, (IV)PL_regstartp[paren],
229 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
230 (IV)PL_regendp[paren],
231 (paren > *PL_reglastparen ? "(no)" : ""));
235 if (*PL_reglastparen + 1 <= PL_regnpar) {
236 PerlIO_printf(Perl_debug_log,
237 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
238 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
242 /* It would seem that the similar code in regtry()
243 * already takes care of this, and in fact it is in
244 * a better location to since this code can #if 0-ed out
245 * but the code in regtry() is needed or otherwise tests
246 * requiring null fields (pat.t#187 and split.t#{13,14}
247 * (as of patchlevel 7877) will fail. Then again,
248 * this code seems to be necessary or otherwise
249 * building DynaLoader will fail:
250 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
252 for (paren = *PL_reglastparen + 1; paren <= PL_regnpar; paren++) {
253 if (paren > PL_regsize)
254 PL_regstartp[paren] = -1;
255 PL_regendp[paren] = -1;
262 S_regcp_set_to(pTHX_ I32 ss)
264 I32 tmp = PL_savestack_ix;
266 PL_savestack_ix = ss;
268 PL_savestack_ix = tmp;
272 typedef struct re_cc_state
276 struct re_cc_state *prev;
281 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
283 #define TRYPAREN(paren, n, input) { \
286 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
287 PL_regendp[paren] = input - PL_bostr; \
290 PL_regendp[paren] = -1; \
292 if (regmatch(next)) \
295 PL_regendp[paren] = -1; \
300 * pregexec and friends
304 - pregexec - match a regexp against a string
307 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
308 char *strbeg, I32 minend, SV *screamer, U32 nosave)
309 /* strend: pointer to null at end of string */
310 /* strbeg: real beginning of string */
311 /* minend: end of match must be >=minend after stringarg. */
312 /* nosave: For optimizations. */
315 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
316 nosave ? 0 : REXEC_COPY_STR);
320 S_cache_re(pTHX_ regexp *prog)
322 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
324 PL_regprogram = prog->program;
326 PL_regnpar = prog->nparens;
327 PL_regdata = prog->data;
332 * Need to implement the following flags for reg_anch:
334 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
336 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
337 * INTUIT_AUTORITATIVE_ML
338 * INTUIT_ONCE_NOML - Intuit can match in one location only.
341 * Another flag for this function: SECOND_TIME (so that float substrs
342 * with giant delta may be not rechecked).
345 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
347 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
348 Otherwise, only SvCUR(sv) is used to get strbeg. */
350 /* XXXX We assume that strpos is strbeg unless sv. */
352 /* XXXX Some places assume that there is a fixed substring.
353 An update may be needed if optimizer marks as "INTUITable"
354 RExen without fixed substrings. Similarly, it is assumed that
355 lengths of all the strings are no more than minlen, thus they
356 cannot come from lookahead.
357 (Or minlen should take into account lookahead.) */
359 /* A failure to find a constant substring means that there is no need to make
360 an expensive call to REx engine, thus we celebrate a failure. Similarly,
361 finding a substring too deep into the string means that less calls to
362 regtry() should be needed.
364 REx compiler's optimizer found 4 possible hints:
365 a) Anchored substring;
367 c) Whether we are anchored (beginning-of-line or \G);
368 d) First node (of those at offset 0) which may distingush positions;
369 We use a)b)d) and multiline-part of c), and try to find a position in the
370 string which does not contradict any of them.
373 /* Most of decisions we do here should have been done at compile time.
374 The nodes of the REx which we used for the search should have been
375 deleted from the finite automaton. */
378 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
379 char *strend, U32 flags, re_scream_pos_data *data)
381 register I32 start_shift = 0;
382 /* Should be nonnegative! */
383 register I32 end_shift = 0;
389 register char *other_last = Nullch; /* other substr checked before this */
390 char *check_at = Nullch; /* check substr found at this pos */
392 char *i_strpos = strpos;
393 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
397 char*s = UTF ? sv_uni_display(dsv, sv, 60, 0) : strpos;
398 int len = UTF ? strlen(s) : strend - strpos;
401 PerlIO_printf(Perl_debug_log,
402 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
403 PL_colors[4],PL_colors[5],PL_colors[0],
406 (strlen(prog->precomp) > 60 ? "..." : ""),
408 (int)(len > 60 ? 60 : len),
410 (len > 60 ? "..." : "")
414 if (prog->reganch & ROPT_UTF8)
415 PL_reg_flags |= RF_utf8;
417 if (prog->minlen > CHR_DIST((U8*)strend, (U8*)strpos)) {
418 DEBUG_r(PerlIO_printf(Perl_debug_log,
419 "String too short... [re_intuit_start]\n"));
422 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
424 check = prog->check_substr;
425 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
426 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
427 || ( (prog->reganch & ROPT_ANCH_BOL)
428 && !PL_multiline ) ); /* Check after \n? */
431 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
432 | ROPT_IMPLICIT)) /* not a real BOL */
433 /* SvCUR is not set on references: SvRV and SvPVX overlap */
435 && (strpos != strbeg)) {
436 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
439 if (prog->check_offset_min == prog->check_offset_max &&
440 !(prog->reganch & ROPT_CANY_SEEN)) {
441 /* Substring at constant offset from beg-of-str... */
444 s = HOP3c(strpos, prog->check_offset_min, strend);
446 slen = SvCUR(check); /* >= 1 */
448 if ( strend - s > slen || strend - s < slen - 1
449 || (strend - s == slen && strend[-1] != '\n')) {
450 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
453 /* Now should match s[0..slen-2] */
455 if (slen && (*SvPVX(check) != *s
457 && memNE(SvPVX(check), s, slen)))) {
459 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
463 else if (*SvPVX(check) != *s
464 || ((slen = SvCUR(check)) > 1
465 && memNE(SvPVX(check), s, slen)))
467 goto success_at_start;
470 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
472 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
473 end_shift = prog->minlen - start_shift -
474 CHR_SVLEN(check) + (SvTAIL(check) != 0);
476 I32 end = prog->check_offset_max + CHR_SVLEN(check)
477 - (SvTAIL(check) != 0);
478 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
480 if (end_shift < eshift)
484 else { /* Can match at random position */
487 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
488 /* Should be nonnegative! */
489 end_shift = prog->minlen - start_shift -
490 CHR_SVLEN(check) + (SvTAIL(check) != 0);
493 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
495 Perl_croak(aTHX_ "panic: end_shift");
499 /* Find a possible match in the region s..strend by looking for
500 the "check" substring in the region corrected by start/end_shift. */
501 if (flags & REXEC_SCREAM) {
502 I32 p = -1; /* Internal iterator of scream. */
503 I32 *pp = data ? data->scream_pos : &p;
505 if (PL_screamfirst[BmRARE(check)] >= 0
506 || ( BmRARE(check) == '\n'
507 && (BmPREVIOUS(check) == SvCUR(check) - 1)
509 s = screaminstr(sv, check,
510 start_shift + (s - strbeg), end_shift, pp, 0);
514 *data->scream_olds = s;
516 else if (prog->reganch & ROPT_CANY_SEEN)
517 s = fbm_instr((U8*)(s + start_shift),
518 (U8*)(strend - end_shift),
519 check, PL_multiline ? FBMrf_MULTILINE : 0);
521 s = fbm_instr(HOP3(s, start_shift, strend),
522 HOP3(strend, -end_shift, strbeg),
523 check, PL_multiline ? FBMrf_MULTILINE : 0);
525 /* Update the count-of-usability, remove useless subpatterns,
528 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
529 (s ? "Found" : "Did not find"),
530 ((check == prog->anchored_substr) ? "anchored" : "floating"),
532 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
534 PL_colors[1], (SvTAIL(check) ? "$" : ""),
535 (s ? " at offset " : "...\n") ) );
542 /* Finish the diagnostic message */
543 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
545 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
546 Start with the other substr.
547 XXXX no SCREAM optimization yet - and a very coarse implementation
548 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
549 *always* match. Probably should be marked during compile...
550 Probably it is right to do no SCREAM here...
553 if (prog->float_substr && prog->anchored_substr) {
554 /* Take into account the "other" substring. */
555 /* XXXX May be hopelessly wrong for UTF... */
558 if (check == prog->float_substr) {
561 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
564 t = s - prog->check_offset_max;
565 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
566 && (!(prog->reganch & ROPT_UTF8)
567 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
572 t = HOP3c(t, prog->anchored_offset, strend);
573 if (t < other_last) /* These positions already checked */
575 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
578 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
579 /* On end-of-str: see comment below. */
580 s = fbm_instr((unsigned char*)t,
581 HOP3(HOP3(last1, prog->anchored_offset, strend)
582 + SvCUR(prog->anchored_substr),
583 -(SvTAIL(prog->anchored_substr)!=0), strbeg),
584 prog->anchored_substr,
585 PL_multiline ? FBMrf_MULTILINE : 0);
586 DEBUG_r(PerlIO_printf(Perl_debug_log,
587 "%s anchored substr `%s%.*s%s'%s",
588 (s ? "Found" : "Contradicts"),
590 (int)(SvCUR(prog->anchored_substr)
591 - (SvTAIL(prog->anchored_substr)!=0)),
592 SvPVX(prog->anchored_substr),
593 PL_colors[1], (SvTAIL(prog->anchored_substr) ? "$" : "")));
595 if (last1 >= last2) {
596 DEBUG_r(PerlIO_printf(Perl_debug_log,
597 ", giving up...\n"));
600 DEBUG_r(PerlIO_printf(Perl_debug_log,
601 ", trying floating at offset %ld...\n",
602 (long)(HOP3c(s1, 1, strend) - i_strpos)));
603 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
604 s = HOP3c(last, 1, strend);
608 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
609 (long)(s - i_strpos)));
610 t = HOP3c(s, -prog->anchored_offset, strbeg);
611 other_last = HOP3c(s, 1, strend);
619 else { /* Take into account the floating substring. */
623 t = HOP3c(s, -start_shift, strbeg);
625 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
626 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
627 last = HOP3c(t, prog->float_max_offset, strend);
628 s = HOP3c(t, prog->float_min_offset, strend);
631 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
632 /* fbm_instr() takes into account exact value of end-of-str
633 if the check is SvTAIL(ed). Since false positives are OK,
634 and end-of-str is not later than strend we are OK. */
635 s = fbm_instr((unsigned char*)s,
636 (unsigned char*)last + SvCUR(prog->float_substr)
637 - (SvTAIL(prog->float_substr)!=0),
638 prog->float_substr, PL_multiline ? FBMrf_MULTILINE : 0);
639 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
640 (s ? "Found" : "Contradicts"),
642 (int)(SvCUR(prog->float_substr)
643 - (SvTAIL(prog->float_substr)!=0)),
644 SvPVX(prog->float_substr),
645 PL_colors[1], (SvTAIL(prog->float_substr) ? "$" : "")));
648 DEBUG_r(PerlIO_printf(Perl_debug_log,
649 ", giving up...\n"));
652 DEBUG_r(PerlIO_printf(Perl_debug_log,
653 ", trying anchored starting at offset %ld...\n",
654 (long)(s1 + 1 - i_strpos)));
656 s = HOP3c(t, 1, strend);
660 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
661 (long)(s - i_strpos)));
662 other_last = s; /* Fix this later. --Hugo */
671 t = s - prog->check_offset_max;
672 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
673 && (!(prog->reganch & ROPT_UTF8)
674 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
676 /* Fixed substring is found far enough so that the match
677 cannot start at strpos. */
679 if (ml_anch && t[-1] != '\n') {
680 /* Eventually fbm_*() should handle this, but often
681 anchored_offset is not 0, so this check will not be wasted. */
682 /* XXXX In the code below we prefer to look for "^" even in
683 presence of anchored substrings. And we search even
684 beyond the found float position. These pessimizations
685 are historical artefacts only. */
687 while (t < strend - prog->minlen) {
689 if (t < check_at - prog->check_offset_min) {
690 if (prog->anchored_substr) {
691 /* Since we moved from the found position,
692 we definitely contradict the found anchored
693 substr. Due to the above check we do not
694 contradict "check" substr.
695 Thus we can arrive here only if check substr
696 is float. Redo checking for "other"=="fixed".
699 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
700 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
701 goto do_other_anchored;
703 /* We don't contradict the found floating substring. */
704 /* XXXX Why not check for STCLASS? */
706 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
707 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
710 /* Position contradicts check-string */
711 /* XXXX probably better to look for check-string
712 than for "\n", so one should lower the limit for t? */
713 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
714 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
715 other_last = strpos = s = t + 1;
720 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
721 PL_colors[0],PL_colors[1]));
725 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
726 PL_colors[0],PL_colors[1]));
730 ++BmUSEFUL(prog->check_substr); /* hooray/5 */
733 /* The found string does not prohibit matching at strpos,
734 - no optimization of calling REx engine can be performed,
735 unless it was an MBOL and we are not after MBOL,
736 or a future STCLASS check will fail this. */
738 /* Even in this situation we may use MBOL flag if strpos is offset
739 wrt the start of the string. */
740 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
741 && (strpos != strbeg) && strpos[-1] != '\n'
742 /* May be due to an implicit anchor of m{.*foo} */
743 && !(prog->reganch & ROPT_IMPLICIT))
748 DEBUG_r( if (ml_anch)
749 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
750 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
753 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
754 && prog->check_substr /* Could be deleted already */
755 && --BmUSEFUL(prog->check_substr) < 0
756 && prog->check_substr == prog->float_substr)
758 /* If flags & SOMETHING - do not do it many times on the same match */
759 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
760 SvREFCNT_dec(prog->check_substr);
761 prog->check_substr = Nullsv; /* disable */
762 prog->float_substr = Nullsv; /* clear */
763 check = Nullsv; /* abort */
765 /* XXXX This is a remnant of the old implementation. It
766 looks wasteful, since now INTUIT can use many
768 prog->reganch &= ~RE_USE_INTUIT;
775 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
776 if (prog->regstclass) {
777 /* minlen == 0 is possible if regstclass is \b or \B,
778 and the fixed substr is ''$.
779 Since minlen is already taken into account, s+1 is before strend;
780 accidentally, minlen >= 1 guaranties no false positives at s + 1
781 even for \b or \B. But (minlen? 1 : 0) below assumes that
782 regstclass does not come from lookahead... */
783 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
784 This leaves EXACTF only, which is dealt with in find_byclass(). */
785 U8* str = (U8*)STRING(prog->regstclass);
786 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
787 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
789 char *endpos = (prog->anchored_substr || ml_anch)
790 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
791 : (prog->float_substr
792 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
795 char *startpos = strbeg;
798 if (prog->reganch & ROPT_UTF8) {
799 PL_regdata = prog->data;
802 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
807 if (endpos == strend) {
808 DEBUG_r( PerlIO_printf(Perl_debug_log,
809 "Could not match STCLASS...\n") );
812 DEBUG_r( PerlIO_printf(Perl_debug_log,
813 "This position contradicts STCLASS...\n") );
814 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
816 /* Contradict one of substrings */
817 if (prog->anchored_substr) {
818 if (prog->anchored_substr == check) {
819 DEBUG_r( what = "anchored" );
821 s = HOP3c(t, 1, strend);
822 if (s + start_shift + end_shift > strend) {
823 /* XXXX Should be taken into account earlier? */
824 DEBUG_r( PerlIO_printf(Perl_debug_log,
825 "Could not match STCLASS...\n") );
830 DEBUG_r( PerlIO_printf(Perl_debug_log,
831 "Looking for %s substr starting at offset %ld...\n",
832 what, (long)(s + start_shift - i_strpos)) );
835 /* Have both, check_string is floating */
836 if (t + start_shift >= check_at) /* Contradicts floating=check */
837 goto retry_floating_check;
838 /* Recheck anchored substring, but not floating... */
842 DEBUG_r( PerlIO_printf(Perl_debug_log,
843 "Looking for anchored substr starting at offset %ld...\n",
844 (long)(other_last - i_strpos)) );
845 goto do_other_anchored;
847 /* Another way we could have checked stclass at the
848 current position only: */
853 DEBUG_r( PerlIO_printf(Perl_debug_log,
854 "Looking for /%s^%s/m starting at offset %ld...\n",
855 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
858 if (!prog->float_substr) /* Could have been deleted */
860 /* Check is floating subtring. */
861 retry_floating_check:
862 t = check_at - start_shift;
863 DEBUG_r( what = "floating" );
864 goto hop_and_restart;
867 DEBUG_r(PerlIO_printf(Perl_debug_log,
868 "By STCLASS: moving %ld --> %ld\n",
869 (long)(t - i_strpos), (long)(s - i_strpos))
873 DEBUG_r(PerlIO_printf(Perl_debug_log,
874 "Does not contradict STCLASS...\n");
879 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
880 PL_colors[4], (check ? "Guessed" : "Giving up"),
881 PL_colors[5], (long)(s - i_strpos)) );
884 fail_finish: /* Substring not found */
885 if (prog->check_substr) /* could be removed already */
886 BmUSEFUL(prog->check_substr) += 5; /* hooray */
888 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
889 PL_colors[4],PL_colors[5]));
893 /* We know what class REx starts with. Try to find this position... */
895 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
897 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
903 register I32 tmp = 1; /* Scratch variable? */
904 register bool do_utf8 = PL_reg_match_utf8;
906 /* We know what class it must start with. */
910 if (reginclass(c, (U8*)s, do_utf8)) {
911 if (tmp && (norun || regtry(prog, s)))
918 s += do_utf8 ? UTF8SKIP(s) : 1;
923 if (tmp && (norun || regtry(prog, s)))
935 U8 tmpbuf1[UTF8_MAXLEN*2+1];
936 U8 tmpbuf2[UTF8_MAXLEN*2+1];
938 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
939 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
941 c1 = utf8_to_uvuni(tmpbuf1, 0);
942 c2 = utf8_to_uvuni(tmpbuf2, 0);
953 c2 = PL_fold_locale[c1];
958 e = s; /* Due to minlen logic of intuit() */
964 if ( utf8_to_uvchr((U8*)s, &len) == c1
971 UV c = utf8_to_uvchr((U8*)s, &len);
972 if ( (c == c1 || c == c2) && regtry(prog, s) )
981 && (ln == 1 || !(OP(c) == EXACTF
983 : ibcmp_locale(s, m, ln)))
984 && (norun || regtry(prog, s)) )
990 if ( (*(U8*)s == c1 || *(U8*)s == c2)
991 && (ln == 1 || !(OP(c) == EXACTF
993 : ibcmp_locale(s, m, ln)))
994 && (norun || regtry(prog, s)) )
1001 PL_reg_flags |= RF_tainted;
1008 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1011 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1013 tmp = ((OP(c) == BOUND ?
1014 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1015 LOAD_UTF8_CHARCLASS(alnum,"a");
1016 while (s < strend) {
1017 if (tmp == !(OP(c) == BOUND ?
1018 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1019 isALNUM_LC_utf8((U8*)s)))
1022 if ((norun || regtry(prog, s)))
1029 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1030 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1031 while (s < strend) {
1033 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1035 if ((norun || regtry(prog, s)))
1041 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1045 PL_reg_flags |= RF_tainted;
1052 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1055 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1057 tmp = ((OP(c) == NBOUND ?
1058 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1059 LOAD_UTF8_CHARCLASS(alnum,"a");
1060 while (s < strend) {
1061 if (tmp == !(OP(c) == NBOUND ?
1062 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1063 isALNUM_LC_utf8((U8*)s)))
1065 else if ((norun || regtry(prog, s)))
1071 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1072 tmp = ((OP(c) == NBOUND ?
1073 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1074 while (s < strend) {
1076 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1078 else if ((norun || regtry(prog, s)))
1083 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1088 LOAD_UTF8_CHARCLASS(alnum,"a");
1089 while (s < strend) {
1090 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1091 if (tmp && (norun || regtry(prog, s)))
1102 while (s < strend) {
1104 if (tmp && (norun || regtry(prog, s)))
1116 PL_reg_flags |= RF_tainted;
1118 while (s < strend) {
1119 if (isALNUM_LC_utf8((U8*)s)) {
1120 if (tmp && (norun || regtry(prog, s)))
1131 while (s < strend) {
1132 if (isALNUM_LC(*s)) {
1133 if (tmp && (norun || regtry(prog, s)))
1146 LOAD_UTF8_CHARCLASS(alnum,"a");
1147 while (s < strend) {
1148 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1149 if (tmp && (norun || regtry(prog, s)))
1160 while (s < strend) {
1162 if (tmp && (norun || regtry(prog, s)))
1174 PL_reg_flags |= RF_tainted;
1176 while (s < strend) {
1177 if (!isALNUM_LC_utf8((U8*)s)) {
1178 if (tmp && (norun || regtry(prog, s)))
1189 while (s < strend) {
1190 if (!isALNUM_LC(*s)) {
1191 if (tmp && (norun || regtry(prog, s)))
1204 LOAD_UTF8_CHARCLASS(space," ");
1205 while (s < strend) {
1206 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1207 if (tmp && (norun || regtry(prog, s)))
1218 while (s < strend) {
1220 if (tmp && (norun || regtry(prog, s)))
1232 PL_reg_flags |= RF_tainted;
1234 while (s < strend) {
1235 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1236 if (tmp && (norun || regtry(prog, s)))
1247 while (s < strend) {
1248 if (isSPACE_LC(*s)) {
1249 if (tmp && (norun || regtry(prog, s)))
1262 LOAD_UTF8_CHARCLASS(space," ");
1263 while (s < strend) {
1264 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1265 if (tmp && (norun || regtry(prog, s)))
1276 while (s < strend) {
1278 if (tmp && (norun || regtry(prog, s)))
1290 PL_reg_flags |= RF_tainted;
1292 while (s < strend) {
1293 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1294 if (tmp && (norun || regtry(prog, s)))
1305 while (s < strend) {
1306 if (!isSPACE_LC(*s)) {
1307 if (tmp && (norun || regtry(prog, s)))
1320 LOAD_UTF8_CHARCLASS(digit,"0");
1321 while (s < strend) {
1322 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1323 if (tmp && (norun || regtry(prog, s)))
1334 while (s < strend) {
1336 if (tmp && (norun || regtry(prog, s)))
1348 PL_reg_flags |= RF_tainted;
1350 while (s < strend) {
1351 if (isDIGIT_LC_utf8((U8*)s)) {
1352 if (tmp && (norun || regtry(prog, s)))
1363 while (s < strend) {
1364 if (isDIGIT_LC(*s)) {
1365 if (tmp && (norun || regtry(prog, s)))
1378 LOAD_UTF8_CHARCLASS(digit,"0");
1379 while (s < strend) {
1380 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1381 if (tmp && (norun || regtry(prog, s)))
1392 while (s < strend) {
1394 if (tmp && (norun || regtry(prog, s)))
1406 PL_reg_flags |= RF_tainted;
1408 while (s < strend) {
1409 if (!isDIGIT_LC_utf8((U8*)s)) {
1410 if (tmp && (norun || regtry(prog, s)))
1421 while (s < strend) {
1422 if (!isDIGIT_LC(*s)) {
1423 if (tmp && (norun || regtry(prog, s)))
1435 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1444 - regexec_flags - match a regexp against a string
1447 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1448 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1449 /* strend: pointer to null at end of string */
1450 /* strbeg: real beginning of string */
1451 /* minend: end of match must be >=minend after stringarg. */
1452 /* data: May be used for some additional optimizations. */
1453 /* nosave: For optimizations. */
1456 register regnode *c;
1457 register char *startpos = stringarg;
1458 I32 minlen; /* must match at least this many chars */
1459 I32 dontbother = 0; /* how many characters not to try at end */
1460 /* I32 start_shift = 0; */ /* Offset of the start to find
1461 constant substr. */ /* CC */
1462 I32 end_shift = 0; /* Same for the end. */ /* CC */
1463 I32 scream_pos = -1; /* Internal iterator of scream. */
1465 SV* oreplsv = GvSV(PL_replgv);
1466 bool do_utf8 = DO_UTF8(sv);
1468 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1475 PL_regnarrate = DEBUG_r_TEST;
1478 /* Be paranoid... */
1479 if (prog == NULL || startpos == NULL) {
1480 Perl_croak(aTHX_ "NULL regexp parameter");
1484 minlen = prog->minlen;
1485 if (strend - startpos < minlen) {
1486 DEBUG_r(PerlIO_printf(Perl_debug_log,
1487 "String too short [regexec_flags]...\n"));
1491 /* Check validity of program. */
1492 if (UCHARAT(prog->program) != REG_MAGIC) {
1493 Perl_croak(aTHX_ "corrupted regexp program");
1497 PL_reg_eval_set = 0;
1500 if (prog->reganch & ROPT_UTF8)
1501 PL_reg_flags |= RF_utf8;
1503 /* Mark beginning of line for ^ and lookbehind. */
1504 PL_regbol = startpos;
1508 /* Mark end of line for $ (and such) */
1511 /* see how far we have to get to not match where we matched before */
1512 PL_regtill = startpos+minend;
1514 /* We start without call_cc context. */
1517 /* If there is a "must appear" string, look for it. */
1520 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1523 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1524 PL_reg_ganch = startpos;
1525 else if (sv && SvTYPE(sv) >= SVt_PVMG
1527 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1528 && mg->mg_len >= 0) {
1529 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1530 if (prog->reganch & ROPT_ANCH_GPOS) {
1531 if (s > PL_reg_ganch)
1536 else /* pos() not defined */
1537 PL_reg_ganch = strbeg;
1540 if (do_utf8 == (UTF!=0) &&
1541 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1542 re_scream_pos_data d;
1544 d.scream_olds = &scream_olds;
1545 d.scream_pos = &scream_pos;
1546 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1548 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1549 goto phooey; /* not present */
1554 char *s = do_utf8 ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1555 int len = do_utf8 ? strlen(s) : strend - startpos;
1558 PerlIO_printf(Perl_debug_log,
1559 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1560 PL_colors[4],PL_colors[5],PL_colors[0],
1563 (strlen(prog->precomp) > 60 ? "..." : ""),
1565 (int)(len > 60 ? 60 : len),
1567 (len > 60 ? "..." : "")
1571 /* Simplest case: anchored match need be tried only once. */
1572 /* [unless only anchor is BOL and multiline is set] */
1573 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1574 if (s == startpos && regtry(prog, startpos))
1576 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1577 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1582 dontbother = minlen - 1;
1583 end = HOP3c(strend, -dontbother, strbeg) - 1;
1584 /* for multiline we only have to try after newlines */
1585 if (prog->check_substr) {
1589 if (regtry(prog, s))
1594 if (prog->reganch & RE_USE_INTUIT) {
1595 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1606 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1607 if (regtry(prog, s))
1614 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1615 if (regtry(prog, PL_reg_ganch))
1620 /* Messy cases: unanchored match. */
1621 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1622 /* we have /x+whatever/ */
1623 /* it must be a one character string (XXXX Except UTF?) */
1624 char ch = SvPVX(prog->anchored_substr)[0];
1630 while (s < strend) {
1632 DEBUG_r( did_match = 1 );
1633 if (regtry(prog, s)) goto got_it;
1635 while (s < strend && *s == ch)
1642 while (s < strend) {
1644 DEBUG_r( did_match = 1 );
1645 if (regtry(prog, s)) goto got_it;
1647 while (s < strend && *s == ch)
1653 DEBUG_r(if (!did_match)
1654 PerlIO_printf(Perl_debug_log,
1655 "Did not find anchored character...\n")
1659 else if (do_utf8 == (UTF!=0) &&
1660 (prog->anchored_substr != Nullsv
1661 || (prog->float_substr != Nullsv
1662 && prog->float_max_offset < strend - s))) {
1663 SV *must = prog->anchored_substr
1664 ? prog->anchored_substr : prog->float_substr;
1666 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1668 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1669 char *last = HOP3c(strend, /* Cannot start after this */
1670 -(I32)(CHR_SVLEN(must)
1671 - (SvTAIL(must) != 0) + back_min), strbeg);
1672 char *last1; /* Last position checked before */
1678 last1 = HOPc(s, -1);
1680 last1 = s - 1; /* bogus */
1682 /* XXXX check_substr already used to find `s', can optimize if
1683 check_substr==must. */
1685 dontbother = end_shift;
1686 strend = HOPc(strend, -dontbother);
1687 while ( (s <= last) &&
1688 ((flags & REXEC_SCREAM)
1689 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1690 end_shift, &scream_pos, 0))
1691 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1692 (unsigned char*)strend, must,
1693 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1694 DEBUG_r( did_match = 1 );
1695 if (HOPc(s, -back_max) > last1) {
1696 last1 = HOPc(s, -back_min);
1697 s = HOPc(s, -back_max);
1700 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1702 last1 = HOPc(s, -back_min);
1706 while (s <= last1) {
1707 if (regtry(prog, s))
1713 while (s <= last1) {
1714 if (regtry(prog, s))
1720 DEBUG_r(if (!did_match)
1721 PerlIO_printf(Perl_debug_log,
1722 "Did not find %s substr `%s%.*s%s'%s...\n",
1723 ((must == prog->anchored_substr)
1724 ? "anchored" : "floating"),
1726 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1728 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1732 else if ((c = prog->regstclass)) {
1733 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1734 /* don't bother with what can't match */
1735 strend = HOPc(strend, -(minlen - 1));
1737 SV *prop = sv_newmortal();
1739 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1741 if (find_byclass(prog, c, s, strend, startpos, 0))
1743 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1747 if (prog->float_substr != Nullsv) { /* Trim the end. */
1750 if (flags & REXEC_SCREAM) {
1751 last = screaminstr(sv, prog->float_substr, s - strbeg,
1752 end_shift, &scream_pos, 1); /* last one */
1754 last = scream_olds; /* Only one occurrence. */
1758 char *little = SvPV(prog->float_substr, len);
1760 if (SvTAIL(prog->float_substr)) {
1761 if (memEQ(strend - len + 1, little, len - 1))
1762 last = strend - len + 1;
1763 else if (!PL_multiline)
1764 last = memEQ(strend - len, little, len)
1765 ? strend - len : Nullch;
1771 last = rninstr(s, strend, little, little + len);
1773 last = strend; /* matching `$' */
1777 DEBUG_r(PerlIO_printf(Perl_debug_log,
1778 "%sCan't trim the tail, match fails (should not happen)%s\n",
1779 PL_colors[4],PL_colors[5]));
1780 goto phooey; /* Should not happen! */
1782 dontbother = strend - last + prog->float_min_offset;
1784 if (minlen && (dontbother < minlen))
1785 dontbother = minlen - 1;
1786 strend -= dontbother; /* this one's always in bytes! */
1787 /* We don't know much -- general case. */
1790 if (regtry(prog, s))
1799 if (regtry(prog, s))
1801 } while (s++ < strend);
1809 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1811 if (PL_reg_eval_set) {
1812 /* Preserve the current value of $^R */
1813 if (oreplsv != GvSV(PL_replgv))
1814 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1815 restored, the value remains
1817 restore_pos(aTHX_ 0);
1820 /* make sure $`, $&, $', and $digit will work later */
1821 if ( !(flags & REXEC_NOT_FIRST) ) {
1822 if (RX_MATCH_COPIED(prog)) {
1823 Safefree(prog->subbeg);
1824 RX_MATCH_COPIED_off(prog);
1826 if (flags & REXEC_COPY_STR) {
1827 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1829 s = savepvn(strbeg, i);
1832 RX_MATCH_COPIED_on(prog);
1835 prog->subbeg = strbeg;
1836 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1843 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1844 PL_colors[4],PL_colors[5]));
1845 if (PL_reg_eval_set)
1846 restore_pos(aTHX_ 0);
1851 - regtry - try match at specific point
1853 STATIC I32 /* 0 failure, 1 success */
1854 S_regtry(pTHX_ regexp *prog, char *startpos)
1862 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1864 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1867 PL_reg_eval_set = RS_init;
1869 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1870 (IV)(PL_stack_sp - PL_stack_base));
1872 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1873 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1874 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1876 /* Apparently this is not needed, judging by wantarray. */
1877 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1878 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1881 /* Make $_ available to executed code. */
1882 if (PL_reg_sv != DEFSV) {
1883 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1888 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1889 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1890 /* prepare for quick setting of pos */
1891 sv_magic(PL_reg_sv, (SV*)0,
1892 PERL_MAGIC_regex_global, Nullch, 0);
1893 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1897 PL_reg_oldpos = mg->mg_len;
1898 SAVEDESTRUCTOR_X(restore_pos, 0);
1900 if (!PL_reg_curpm) {
1901 Newz(22,PL_reg_curpm, 1, PMOP);
1904 SV* repointer = newSViv(0);
1905 /* so we know which PL_regex_padav element is PL_reg_curpm */
1906 SvFLAGS(repointer) |= SVf_BREAK;
1907 av_push(PL_regex_padav,repointer);
1908 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1909 PL_regex_pad = AvARRAY(PL_regex_padav);
1913 PM_SETRE(PL_reg_curpm, prog);
1914 PL_reg_oldcurpm = PL_curpm;
1915 PL_curpm = PL_reg_curpm;
1916 if (RX_MATCH_COPIED(prog)) {
1917 /* Here is a serious problem: we cannot rewrite subbeg,
1918 since it may be needed if this match fails. Thus
1919 $` inside (?{}) could fail... */
1920 PL_reg_oldsaved = prog->subbeg;
1921 PL_reg_oldsavedlen = prog->sublen;
1922 RX_MATCH_COPIED_off(prog);
1925 PL_reg_oldsaved = Nullch;
1926 prog->subbeg = PL_bostr;
1927 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1929 prog->startp[0] = startpos - PL_bostr;
1930 PL_reginput = startpos;
1931 PL_regstartp = prog->startp;
1932 PL_regendp = prog->endp;
1933 PL_reglastparen = &prog->lastparen;
1934 PL_reglastcloseparen = &prog->lastcloseparen;
1935 prog->lastparen = 0;
1937 DEBUG_r(PL_reg_starttry = startpos);
1938 if (PL_reg_start_tmpl <= prog->nparens) {
1939 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1940 if(PL_reg_start_tmp)
1941 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1943 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1947 sv_setpvn(PERL_DEBUG_PAD(0), "", 0);
1948 sv_setpvn(PERL_DEBUG_PAD(1), "", 0);
1949 sv_setpvn(PERL_DEBUG_PAD(2), "", 0);
1952 /* XXXX What this code is doing here?!!! There should be no need
1953 to do this again and again, PL_reglastparen should take care of
1956 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
1957 * Actually, the code in regcppop() (which Ilya may be meaning by
1958 * PL_reglastparen), is not needed at all by the test suite
1959 * (op/regexp, op/pat, op/split), but that code is needed, oddly
1960 * enough, for building DynaLoader, or otherwise this
1961 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
1962 * will happen. Meanwhile, this code *is* needed for the
1963 * above-mentioned test suite tests to succeed. The common theme
1964 * on those tests seems to be returning null fields from matches.
1969 if (prog->nparens) {
1970 for (i = prog->nparens; i > *PL_reglastparen; i--) {
1977 if (regmatch(prog->program + 1)) {
1978 prog->endp[0] = PL_reginput - PL_bostr;
1981 REGCP_UNWIND(lastcp);
1985 #define RE_UNWIND_BRANCH 1
1986 #define RE_UNWIND_BRANCHJ 2
1990 typedef struct { /* XX: makes sense to enlarge it... */
1994 } re_unwind_generic_t;
2007 } re_unwind_branch_t;
2009 typedef union re_unwind_t {
2011 re_unwind_generic_t generic;
2012 re_unwind_branch_t branch;
2015 #define sayYES goto yes
2016 #define sayNO goto no
2017 #define sayYES_FINAL goto yes_final
2018 #define sayYES_LOUD goto yes_loud
2019 #define sayNO_FINAL goto no_final
2020 #define sayNO_SILENT goto do_no
2021 #define saySAME(x) if (x) goto yes; else goto no
2023 #define REPORT_CODE_OFF 24
2026 - regmatch - main matching routine
2028 * Conceptually the strategy is simple: check to see whether the current
2029 * node matches, call self recursively to see whether the rest matches,
2030 * and then act accordingly. In practice we make some effort to avoid
2031 * recursion, in particular by going through "ordinary" nodes (that don't
2032 * need to know whether the rest of the match failed) by a loop instead of
2035 /* [lwall] I've hoisted the register declarations to the outer block in order to
2036 * maybe save a little bit of pushing and popping on the stack. It also takes
2037 * advantage of machines that use a register save mask on subroutine entry.
2039 STATIC I32 /* 0 failure, 1 success */
2040 S_regmatch(pTHX_ regnode *prog)
2042 register regnode *scan; /* Current node. */
2043 regnode *next; /* Next node. */
2044 regnode *inner; /* Next node in internal branch. */
2045 register I32 nextchr; /* renamed nextchr - nextchar colides with
2046 function of same name */
2047 register I32 n; /* no or next */
2048 register I32 ln = 0; /* len or last */
2049 register char *s = Nullch; /* operand or save */
2050 register char *locinput = PL_reginput;
2051 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2052 int minmod = 0, sw = 0, logical = 0;
2055 I32 firstcp = PL_savestack_ix;
2057 register bool do_utf8 = PL_reg_match_utf8;
2059 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2060 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2061 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2068 /* Note that nextchr is a byte even in UTF */
2069 nextchr = UCHARAT(locinput);
2071 while (scan != NULL) {
2074 SV *prop = sv_newmortal();
2075 int docolor = *PL_colors[0];
2076 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2077 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2078 /* The part of the string before starttry has one color
2079 (pref0_len chars), between starttry and current
2080 position another one (pref_len - pref0_len chars),
2081 after the current position the third one.
2082 We assume that pref0_len <= pref_len, otherwise we
2083 decrease pref0_len. */
2084 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2085 ? (5 + taill) - l : locinput - PL_bostr;
2088 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2090 pref0_len = pref_len - (locinput - PL_reg_starttry);
2091 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2092 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2093 ? (5 + taill) - pref_len : PL_regeol - locinput);
2094 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2098 if (pref0_len > pref_len)
2099 pref0_len = pref_len;
2100 regprop(prop, scan);
2104 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2106 locinput - pref_len;
2107 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2108 char *s1 = do_utf8 ?
2109 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2110 pref_len - pref0_len, 60, 0) :
2111 locinput - pref_len + pref0_len;
2112 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2113 char *s2 = do_utf8 ?
2114 pv_uni_display(dsv2, (U8*)locinput,
2115 PL_regeol - locinput, 60, 0) :
2117 int len2 = do_utf8 ? strlen(s2) : l;
2118 PerlIO_printf(Perl_debug_log,
2119 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2120 (IV)(locinput - PL_bostr),
2127 (docolor ? "" : "> <"),
2131 15 - l - pref_len + 1,
2133 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2138 next = scan + NEXT_OFF(scan);
2144 if (locinput == PL_bostr || (PL_multiline &&
2145 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2147 /* regtill = regbol; */
2152 if (locinput == PL_bostr ||
2153 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2159 if (locinput == PL_bostr)
2163 if (locinput == PL_reg_ganch)
2173 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2178 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2180 if (PL_regeol - locinput > 1)
2184 if (PL_regeol != locinput)
2188 if (!nextchr && locinput >= PL_regeol)
2191 locinput += PL_utf8skip[nextchr];
2192 if (locinput > PL_regeol)
2194 nextchr = UCHARAT(locinput);
2197 nextchr = UCHARAT(++locinput);
2200 if (!nextchr && locinput >= PL_regeol)
2202 nextchr = UCHARAT(++locinput);
2205 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2208 locinput += PL_utf8skip[nextchr];
2209 if (locinput > PL_regeol)
2211 nextchr = UCHARAT(locinput);
2214 nextchr = UCHARAT(++locinput);
2219 if (do_utf8 != (UTF!=0)) {
2220 /* The target and the pattern have differing "utf8ness". */
2226 /* The target is utf8, the pattern is not utf8. */
2230 if (NATIVE_TO_UNI(*(U8*)s) !=
2231 utf8_to_uvchr((U8*)l, &len))
2238 /* The target is not utf8, the pattern is utf8. */
2242 if (NATIVE_TO_UNI(*((U8*)l)) !=
2243 utf8_to_uvchr((U8*)s, &len))
2250 nextchr = UCHARAT(locinput);
2253 /* The target and the pattern have the same "utf8ness". */
2254 /* Inline the first character, for speed. */
2255 if (UCHARAT(s) != nextchr)
2257 if (PL_regeol - locinput < ln)
2259 if (ln > 1 && memNE(s, locinput, ln))
2262 nextchr = UCHARAT(locinput);
2265 PL_reg_flags |= RF_tainted;
2275 U8 tmpbuf[UTF8_MAXLEN*2+1];
2280 toLOWER_utf8((U8*)l, tmpbuf, &ulen);
2281 if (memNE(s, (char*)tmpbuf, ulen))
2287 nextchr = UCHARAT(locinput);
2291 /* Inline the first character, for speed. */
2292 if (UCHARAT(s) != nextchr &&
2293 UCHARAT(s) != ((OP(scan) == EXACTF)
2294 ? PL_fold : PL_fold_locale)[nextchr])
2296 if (PL_regeol - locinput < ln)
2298 if (ln > 1 && (OP(scan) == EXACTF
2299 ? ibcmp(s, locinput, ln)
2300 : ibcmp_locale(s, locinput, ln)))
2303 nextchr = UCHARAT(locinput);
2307 if (!reginclass(scan, (U8*)locinput, do_utf8))
2309 if (locinput >= PL_regeol)
2311 locinput += PL_utf8skip[nextchr];
2312 nextchr = UCHARAT(locinput);
2316 nextchr = UCHARAT(locinput);
2317 if (!reginclass(scan, (U8*)locinput, do_utf8))
2319 if (!nextchr && locinput >= PL_regeol)
2321 nextchr = UCHARAT(++locinput);
2325 PL_reg_flags |= RF_tainted;
2331 LOAD_UTF8_CHARCLASS(alnum,"a");
2332 if (!(OP(scan) == ALNUM
2333 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2334 : isALNUM_LC_utf8((U8*)locinput)))
2338 locinput += PL_utf8skip[nextchr];
2339 nextchr = UCHARAT(locinput);
2342 if (!(OP(scan) == ALNUM
2343 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2345 nextchr = UCHARAT(++locinput);
2348 PL_reg_flags |= RF_tainted;
2351 if (!nextchr && locinput >= PL_regeol)
2354 LOAD_UTF8_CHARCLASS(alnum,"a");
2355 if (OP(scan) == NALNUM
2356 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2357 : isALNUM_LC_utf8((U8*)locinput))
2361 locinput += PL_utf8skip[nextchr];
2362 nextchr = UCHARAT(locinput);
2365 if (OP(scan) == NALNUM
2366 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2368 nextchr = UCHARAT(++locinput);
2372 PL_reg_flags |= RF_tainted;
2376 /* was last char in word? */
2378 if (locinput == PL_bostr)
2381 U8 *r = reghop((U8*)locinput, -1);
2383 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2385 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2386 ln = isALNUM_uni(ln);
2387 LOAD_UTF8_CHARCLASS(alnum,"a");
2388 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2391 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2392 n = isALNUM_LC_utf8((U8*)locinput);
2396 ln = (locinput != PL_bostr) ?
2397 UCHARAT(locinput - 1) : '\n';
2398 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2400 n = isALNUM(nextchr);
2403 ln = isALNUM_LC(ln);
2404 n = isALNUM_LC(nextchr);
2407 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2408 OP(scan) == BOUNDL))
2412 PL_reg_flags |= RF_tainted;
2418 if (UTF8_IS_CONTINUED(nextchr)) {
2419 LOAD_UTF8_CHARCLASS(space," ");
2420 if (!(OP(scan) == SPACE
2421 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2422 : isSPACE_LC_utf8((U8*)locinput)))
2426 locinput += PL_utf8skip[nextchr];
2427 nextchr = UCHARAT(locinput);
2430 if (!(OP(scan) == SPACE
2431 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2433 nextchr = UCHARAT(++locinput);
2436 if (!(OP(scan) == SPACE
2437 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2439 nextchr = UCHARAT(++locinput);
2443 PL_reg_flags |= RF_tainted;
2446 if (!nextchr && locinput >= PL_regeol)
2449 LOAD_UTF8_CHARCLASS(space," ");
2450 if (OP(scan) == NSPACE
2451 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2452 : isSPACE_LC_utf8((U8*)locinput))
2456 locinput += PL_utf8skip[nextchr];
2457 nextchr = UCHARAT(locinput);
2460 if (OP(scan) == NSPACE
2461 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2463 nextchr = UCHARAT(++locinput);
2466 PL_reg_flags |= RF_tainted;
2472 LOAD_UTF8_CHARCLASS(digit,"0");
2473 if (!(OP(scan) == DIGIT
2474 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2475 : isDIGIT_LC_utf8((U8*)locinput)))
2479 locinput += PL_utf8skip[nextchr];
2480 nextchr = UCHARAT(locinput);
2483 if (!(OP(scan) == DIGIT
2484 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2486 nextchr = UCHARAT(++locinput);
2489 PL_reg_flags |= RF_tainted;
2492 if (!nextchr && locinput >= PL_regeol)
2495 LOAD_UTF8_CHARCLASS(digit,"0");
2496 if (OP(scan) == NDIGIT
2497 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2498 : isDIGIT_LC_utf8((U8*)locinput))
2502 locinput += PL_utf8skip[nextchr];
2503 nextchr = UCHARAT(locinput);
2506 if (OP(scan) == NDIGIT
2507 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2509 nextchr = UCHARAT(++locinput);
2512 LOAD_UTF8_CHARCLASS(mark,"~");
2513 if (locinput >= PL_regeol ||
2514 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2516 locinput += PL_utf8skip[nextchr];
2517 while (locinput < PL_regeol &&
2518 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2519 locinput += UTF8SKIP(locinput);
2520 if (locinput > PL_regeol)
2522 nextchr = UCHARAT(locinput);
2525 PL_reg_flags |= RF_tainted;
2529 n = ARG(scan); /* which paren pair */
2530 ln = PL_regstartp[n];
2531 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2532 if (*PL_reglastparen < n || ln == -1)
2533 sayNO; /* Do not match unless seen CLOSEn. */
2534 if (ln == PL_regendp[n])
2538 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2540 char *e = PL_bostr + PL_regendp[n];
2542 * Note that we can't do the "other character" lookup trick as
2543 * in the 8-bit case (no pun intended) because in Unicode we
2544 * have to map both upper and title case to lower case.
2546 if (OP(scan) == REFF) {
2547 STRLEN ulen1, ulen2;
2548 U8 tmpbuf1[UTF8_MAXLEN*2+1];
2549 U8 tmpbuf2[UTF8_MAXLEN*2+1];
2553 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2554 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2555 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2562 nextchr = UCHARAT(locinput);
2566 /* Inline the first character, for speed. */
2567 if (UCHARAT(s) != nextchr &&
2569 (UCHARAT(s) != ((OP(scan) == REFF
2570 ? PL_fold : PL_fold_locale)[nextchr]))))
2572 ln = PL_regendp[n] - ln;
2573 if (locinput + ln > PL_regeol)
2575 if (ln > 1 && (OP(scan) == REF
2576 ? memNE(s, locinput, ln)
2578 ? ibcmp(s, locinput, ln)
2579 : ibcmp_locale(s, locinput, ln))))
2582 nextchr = UCHARAT(locinput);
2593 OP_4tree *oop = PL_op;
2594 COP *ocurcop = PL_curcop;
2595 SV **ocurpad = PL_curpad;
2599 PL_op = (OP_4tree*)PL_regdata->data[n];
2600 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2601 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2602 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2606 CALLRUNOPS(aTHX); /* Scalar context. */
2609 ret = Nullsv; /* protect against empty (?{}) blocks. */
2617 PL_curpad = ocurpad;
2618 PL_curcop = ocurcop;
2620 if (logical == 2) { /* Postponed subexpression. */
2622 MAGIC *mg = Null(MAGIC*);
2624 CHECKPOINT cp, lastcp;
2626 if(SvROK(ret) || SvRMAGICAL(ret)) {
2627 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2630 mg = mg_find(sv, PERL_MAGIC_qr);
2633 re = (regexp *)mg->mg_obj;
2634 (void)ReREFCNT_inc(re);
2638 char *t = SvPV(ret, len);
2640 char *oprecomp = PL_regprecomp;
2641 I32 osize = PL_regsize;
2642 I32 onpar = PL_regnpar;
2645 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2647 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2648 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2650 PL_regprecomp = oprecomp;
2655 PerlIO_printf(Perl_debug_log,
2656 "Entering embedded `%s%.60s%s%s'\n",
2660 (strlen(re->precomp) > 60 ? "..." : ""))
2663 state.prev = PL_reg_call_cc;
2664 state.cc = PL_regcc;
2665 state.re = PL_reg_re;
2669 cp = regcppush(0); /* Save *all* the positions. */
2672 state.ss = PL_savestack_ix;
2673 *PL_reglastparen = 0;
2674 *PL_reglastcloseparen = 0;
2675 PL_reg_call_cc = &state;
2676 PL_reginput = locinput;
2678 /* XXXX This is too dramatic a measure... */
2681 if (regmatch(re->program + 1)) {
2682 /* Even though we succeeded, we need to restore
2683 global variables, since we may be wrapped inside
2684 SUSPEND, thus the match may be not finished yet. */
2686 /* XXXX Do this only if SUSPENDed? */
2687 PL_reg_call_cc = state.prev;
2688 PL_regcc = state.cc;
2689 PL_reg_re = state.re;
2690 cache_re(PL_reg_re);
2692 /* XXXX This is too dramatic a measure... */
2695 /* These are needed even if not SUSPEND. */
2701 REGCP_UNWIND(lastcp);
2703 PL_reg_call_cc = state.prev;
2704 PL_regcc = state.cc;
2705 PL_reg_re = state.re;
2706 cache_re(PL_reg_re);
2708 /* XXXX This is too dramatic a measure... */
2718 sv_setsv(save_scalar(PL_replgv), ret);
2722 n = ARG(scan); /* which paren pair */
2723 PL_reg_start_tmp[n] = locinput;
2728 n = ARG(scan); /* which paren pair */
2729 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2730 PL_regendp[n] = locinput - PL_bostr;
2731 if (n > *PL_reglastparen)
2732 *PL_reglastparen = n;
2733 *PL_reglastcloseparen = n;
2736 n = ARG(scan); /* which paren pair */
2737 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2740 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2742 next = NEXTOPER(NEXTOPER(scan));
2744 next = scan + ARG(scan);
2745 if (OP(next) == IFTHEN) /* Fake one. */
2746 next = NEXTOPER(NEXTOPER(next));
2750 logical = scan->flags;
2752 /*******************************************************************
2753 PL_regcc contains infoblock about the innermost (...)* loop, and
2754 a pointer to the next outer infoblock.
2756 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2758 1) After matching X, regnode for CURLYX is processed;
2760 2) This regnode creates infoblock on the stack, and calls
2761 regmatch() recursively with the starting point at WHILEM node;
2763 3) Each hit of WHILEM node tries to match A and Z (in the order
2764 depending on the current iteration, min/max of {min,max} and
2765 greediness). The information about where are nodes for "A"
2766 and "Z" is read from the infoblock, as is info on how many times "A"
2767 was already matched, and greediness.
2769 4) After A matches, the same WHILEM node is hit again.
2771 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2772 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2773 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2774 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2775 of the external loop.
2777 Currently present infoblocks form a tree with a stem formed by PL_curcc
2778 and whatever it mentions via ->next, and additional attached trees
2779 corresponding to temporarily unset infoblocks as in "5" above.
2781 In the following picture infoblocks for outer loop of
2782 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2783 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2784 infoblocks are drawn below the "reset" infoblock.
2786 In fact in the picture below we do not show failed matches for Z and T
2787 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2788 more obvious *why* one needs to *temporary* unset infoblocks.]
2790 Matched REx position InfoBlocks Comment
2794 Y A)*?Z)*?T x <- O <- I
2795 YA )*?Z)*?T x <- O <- I
2796 YA A)*?Z)*?T x <- O <- I
2797 YAA )*?Z)*?T x <- O <- I
2798 YAA Z)*?T x <- O # Temporary unset I
2801 YAAZ Y(A)*?Z)*?T x <- O
2804 YAAZY (A)*?Z)*?T x <- O
2807 YAAZY A)*?Z)*?T x <- O <- I
2810 YAAZYA )*?Z)*?T x <- O <- I
2813 YAAZYA Z)*?T x <- O # Temporary unset I
2819 YAAZYAZ T x # Temporary unset O
2826 *******************************************************************/
2829 CHECKPOINT cp = PL_savestack_ix;
2830 /* No need to save/restore up to this paren */
2831 I32 parenfloor = scan->flags;
2833 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2835 cc.oldcc = PL_regcc;
2837 /* XXXX Probably it is better to teach regpush to support
2838 parenfloor > PL_regsize... */
2839 if (parenfloor > *PL_reglastparen)
2840 parenfloor = *PL_reglastparen; /* Pessimization... */
2841 cc.parenfloor = parenfloor;
2843 cc.min = ARG1(scan);
2844 cc.max = ARG2(scan);
2845 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2849 PL_reginput = locinput;
2850 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2852 PL_regcc = cc.oldcc;
2858 * This is really hard to understand, because after we match
2859 * what we're trying to match, we must make sure the rest of
2860 * the REx is going to match for sure, and to do that we have
2861 * to go back UP the parse tree by recursing ever deeper. And
2862 * if it fails, we have to reset our parent's current state
2863 * that we can try again after backing off.
2866 CHECKPOINT cp, lastcp;
2867 CURCUR* cc = PL_regcc;
2868 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2870 n = cc->cur + 1; /* how many we know we matched */
2871 PL_reginput = locinput;
2874 PerlIO_printf(Perl_debug_log,
2875 "%*s %ld out of %ld..%ld cc=%lx\n",
2876 REPORT_CODE_OFF+PL_regindent*2, "",
2877 (long)n, (long)cc->min,
2878 (long)cc->max, (long)cc)
2881 /* If degenerate scan matches "", assume scan done. */
2883 if (locinput == cc->lastloc && n >= cc->min) {
2884 PL_regcc = cc->oldcc;
2888 PerlIO_printf(Perl_debug_log,
2889 "%*s empty match detected, try continuation...\n",
2890 REPORT_CODE_OFF+PL_regindent*2, "")
2892 if (regmatch(cc->next))
2900 /* First just match a string of min scans. */
2904 cc->lastloc = locinput;
2905 if (regmatch(cc->scan))
2908 cc->lastloc = lastloc;
2913 /* Check whether we already were at this position.
2914 Postpone detection until we know the match is not
2915 *that* much linear. */
2916 if (!PL_reg_maxiter) {
2917 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2918 PL_reg_leftiter = PL_reg_maxiter;
2920 if (PL_reg_leftiter-- == 0) {
2921 I32 size = (PL_reg_maxiter + 7)/8;
2922 if (PL_reg_poscache) {
2923 if (PL_reg_poscache_size < size) {
2924 Renew(PL_reg_poscache, size, char);
2925 PL_reg_poscache_size = size;
2927 Zero(PL_reg_poscache, size, char);
2930 PL_reg_poscache_size = size;
2931 Newz(29, PL_reg_poscache, size, char);
2934 PerlIO_printf(Perl_debug_log,
2935 "%sDetected a super-linear match, switching on caching%s...\n",
2936 PL_colors[4], PL_colors[5])
2939 if (PL_reg_leftiter < 0) {
2940 I32 o = locinput - PL_bostr, b;
2942 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2945 if (PL_reg_poscache[o] & (1<<b)) {
2947 PerlIO_printf(Perl_debug_log,
2948 "%*s already tried at this position...\n",
2949 REPORT_CODE_OFF+PL_regindent*2, "")
2953 PL_reg_poscache[o] |= (1<<b);
2957 /* Prefer next over scan for minimal matching. */
2960 PL_regcc = cc->oldcc;
2963 cp = regcppush(cc->parenfloor);
2965 if (regmatch(cc->next)) {
2967 sayYES; /* All done. */
2969 REGCP_UNWIND(lastcp);
2975 if (n >= cc->max) { /* Maximum greed exceeded? */
2976 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2977 && !(PL_reg_flags & RF_warned)) {
2978 PL_reg_flags |= RF_warned;
2979 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2980 "Complex regular subexpression recursion",
2987 PerlIO_printf(Perl_debug_log,
2988 "%*s trying longer...\n",
2989 REPORT_CODE_OFF+PL_regindent*2, "")
2991 /* Try scanning more and see if it helps. */
2992 PL_reginput = locinput;
2994 cc->lastloc = locinput;
2995 cp = regcppush(cc->parenfloor);
2997 if (regmatch(cc->scan)) {
3001 REGCP_UNWIND(lastcp);
3004 cc->lastloc = lastloc;
3008 /* Prefer scan over next for maximal matching. */
3010 if (n < cc->max) { /* More greed allowed? */
3011 cp = regcppush(cc->parenfloor);
3013 cc->lastloc = locinput;
3015 if (regmatch(cc->scan)) {
3019 REGCP_UNWIND(lastcp);
3020 regcppop(); /* Restore some previous $<digit>s? */
3021 PL_reginput = locinput;
3023 PerlIO_printf(Perl_debug_log,
3024 "%*s failed, try continuation...\n",
3025 REPORT_CODE_OFF+PL_regindent*2, "")
3028 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3029 && !(PL_reg_flags & RF_warned)) {
3030 PL_reg_flags |= RF_warned;
3031 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3032 "Complex regular subexpression recursion",
3036 /* Failed deeper matches of scan, so see if this one works. */
3037 PL_regcc = cc->oldcc;
3040 if (regmatch(cc->next))
3046 cc->lastloc = lastloc;
3051 next = scan + ARG(scan);
3054 inner = NEXTOPER(NEXTOPER(scan));
3057 inner = NEXTOPER(scan);
3061 if (OP(next) != c1) /* No choice. */
3062 next = inner; /* Avoid recursion. */
3064 I32 lastparen = *PL_reglastparen;
3066 re_unwind_branch_t *uw;
3068 /* Put unwinding data on stack */
3069 unwind1 = SSNEWt(1,re_unwind_branch_t);
3070 uw = SSPTRt(unwind1,re_unwind_branch_t);
3073 uw->type = ((c1 == BRANCH)
3075 : RE_UNWIND_BRANCHJ);
3076 uw->lastparen = lastparen;
3078 uw->locinput = locinput;
3079 uw->nextchr = nextchr;
3081 uw->regindent = ++PL_regindent;
3084 REGCP_SET(uw->lastcp);
3086 /* Now go into the first branch */
3099 /* We suppose that the next guy does not need
3100 backtracking: in particular, it is of constant length,
3101 and has no parenths to influence future backrefs. */
3102 ln = ARG1(scan); /* min to match */
3103 n = ARG2(scan); /* max to match */
3104 paren = scan->flags;
3106 if (paren > PL_regsize)
3108 if (paren > *PL_reglastparen)
3109 *PL_reglastparen = paren;
3111 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3113 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3114 PL_reginput = locinput;
3117 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3119 /* if we matched something zero-length we don't need to
3120 backtrack - capturing parens are already defined, so
3121 the caveat in the maximal case doesn't apply
3123 XXXX if ln == 0, we can redo this check first time
3124 through the following loop
3127 n = ln; /* don't backtrack */
3128 locinput = PL_reginput;
3129 if (HAS_TEXT(next) || JUMPABLE(next)) {
3130 regnode *text_node = next;
3132 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3134 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3136 if (PL_regkind[(U8)OP(text_node)] == REF) {
3138 n = ARG(text_node); /* which paren pair */
3139 ln = PL_regstartp[n];
3140 /* assume yes if we haven't seen CLOSEn */
3142 *PL_reglastparen < n ||
3149 c1 = *(PL_bostr + ln);
3151 else { c1 = (U8)*STRING(text_node); }
3152 if (OP(next) == EXACTF)
3154 else if (OP(text_node) == EXACTFL)
3155 c2 = PL_fold_locale[c1];
3164 /* This may be improved if l == 0. */
3165 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3166 /* If it could work, try it. */
3168 UCHARAT(PL_reginput) == c1 ||
3169 UCHARAT(PL_reginput) == c2)
3173 PL_regstartp[paren] =
3174 HOPc(PL_reginput, -l) - PL_bostr;
3175 PL_regendp[paren] = PL_reginput - PL_bostr;
3178 PL_regendp[paren] = -1;
3182 REGCP_UNWIND(lastcp);
3184 /* Couldn't or didn't -- move forward. */
3185 PL_reginput = locinput;
3186 if (regrepeat_hard(scan, 1, &l)) {
3188 locinput = PL_reginput;
3195 n = regrepeat_hard(scan, n, &l);
3196 /* if we matched something zero-length we don't need to
3197 backtrack, unless the minimum count is zero and we
3198 are capturing the result - in that case the capture
3199 being defined or not may affect later execution
3201 if (n != 0 && l == 0 && !(paren && ln == 0))
3202 ln = n; /* don't backtrack */
3203 locinput = PL_reginput;
3205 PerlIO_printf(Perl_debug_log,
3206 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3207 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3211 if (HAS_TEXT(next) || JUMPABLE(next)) {
3212 regnode *text_node = next;
3214 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3216 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3218 if (PL_regkind[(U8)OP(text_node)] == REF) {
3220 n = ARG(text_node); /* which paren pair */
3221 ln = PL_regstartp[n];
3222 /* assume yes if we haven't seen CLOSEn */
3224 *PL_reglastparen < n ||
3231 c1 = *(PL_bostr + ln);
3233 else { c1 = (U8)*STRING(text_node); }
3235 if (OP(text_node) == EXACTF)
3237 else if (OP(text_node) == EXACTFL)
3238 c2 = PL_fold_locale[c1];
3249 /* If it could work, try it. */
3251 UCHARAT(PL_reginput) == c1 ||
3252 UCHARAT(PL_reginput) == c2)
3255 PerlIO_printf(Perl_debug_log,
3256 "%*s trying tail with n=%"IVdf"...\n",
3257 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3261 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3262 PL_regendp[paren] = PL_reginput - PL_bostr;
3265 PL_regendp[paren] = -1;
3269 REGCP_UNWIND(lastcp);
3271 /* Couldn't or didn't -- back up. */
3273 locinput = HOPc(locinput, -l);
3274 PL_reginput = locinput;
3281 paren = scan->flags; /* Which paren to set */
3282 if (paren > PL_regsize)
3284 if (paren > *PL_reglastparen)
3285 *PL_reglastparen = paren;
3286 ln = ARG1(scan); /* min to match */
3287 n = ARG2(scan); /* max to match */
3288 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3292 ln = ARG1(scan); /* min to match */
3293 n = ARG2(scan); /* max to match */
3294 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3299 scan = NEXTOPER(scan);
3305 scan = NEXTOPER(scan);
3309 * Lookahead to avoid useless match attempts
3310 * when we know what character comes next.
3314 * Used to only do .*x and .*?x, but now it allows
3315 * for )'s, ('s and (?{ ... })'s to be in the way
3316 * of the quantifier and the EXACT-like node. -- japhy
3319 if (HAS_TEXT(next) || JUMPABLE(next)) {
3321 regnode *text_node = next;
3323 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3325 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3327 if (PL_regkind[(U8)OP(text_node)] == REF) {
3329 n = ARG(text_node); /* which paren pair */
3330 ln = PL_regstartp[n];
3331 /* assume yes if we haven't seen CLOSEn */
3333 *PL_reglastparen < n ||
3338 goto assume_ok_easy;
3340 s = (U8*)PL_bostr + ln;
3342 else { s = (U8*)STRING(text_node); }
3346 if (OP(text_node) == EXACTF)
3348 else if (OP(text_node) == EXACTFL)
3349 c2 = PL_fold_locale[c1];
3352 if (OP(text_node) == EXACTF) {
3353 STRLEN ulen1, ulen2;
3354 U8 tmpbuf1[UTF8_MAXLEN*2+1];
3355 U8 tmpbuf2[UTF8_MAXLEN*2+1];
3357 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3358 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3360 c1 = utf8_to_uvuni(tmpbuf1, 0);
3361 c2 = utf8_to_uvuni(tmpbuf2, 0);
3364 c2 = c1 = utf8_to_uvchr(s, NULL);
3372 PL_reginput = locinput;
3376 if (ln && regrepeat(scan, ln) < ln)
3378 locinput = PL_reginput;
3381 char *e; /* Should not check after this */
3382 char *old = locinput;
3384 if (n == REG_INFTY) {
3387 while (UTF8_IS_CONTINUATION(*(U8*)e))
3393 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3397 e = locinput + n - ln;
3403 /* Find place 'next' could work */
3406 while (locinput <= e &&
3407 UCHARAT(locinput) != c1)
3410 while (locinput <= e
3411 && UCHARAT(locinput) != c1
3412 && UCHARAT(locinput) != c2)
3415 count = locinput - old;
3422 utf8_to_uvchr((U8*)locinput, &len) != c1;
3427 for (count = 0; locinput <= e; count++) {
3428 UV c = utf8_to_uvchr((U8*)locinput, &len);
3429 if (c == c1 || c == c2)
3437 /* PL_reginput == old now */
3438 if (locinput != old) {
3439 ln = 1; /* Did some */
3440 if (regrepeat(scan, count) < count)
3443 /* PL_reginput == locinput now */
3444 TRYPAREN(paren, ln, locinput);
3445 PL_reginput = locinput; /* Could be reset... */
3446 REGCP_UNWIND(lastcp);
3447 /* Couldn't or didn't -- move forward. */
3450 locinput += UTF8SKIP(locinput);
3456 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3460 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3462 c = UCHARAT(PL_reginput);
3463 /* If it could work, try it. */
3464 if (c == c1 || c == c2)
3466 TRYPAREN(paren, n, PL_reginput);
3467 REGCP_UNWIND(lastcp);
3470 /* If it could work, try it. */
3471 else if (c1 == -1000)
3473 TRYPAREN(paren, n, PL_reginput);
3474 REGCP_UNWIND(lastcp);
3476 /* Couldn't or didn't -- move forward. */
3477 PL_reginput = locinput;
3478 if (regrepeat(scan, 1)) {
3480 locinput = PL_reginput;
3488 n = regrepeat(scan, n);
3489 locinput = PL_reginput;
3490 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3491 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3492 ln = n; /* why back off? */
3493 /* ...because $ and \Z can match before *and* after
3494 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3495 We should back off by one in this case. */
3496 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3505 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3507 c = UCHARAT(PL_reginput);
3509 /* If it could work, try it. */
3510 if (c1 == -1000 || c == c1 || c == c2)
3512 TRYPAREN(paren, n, PL_reginput);
3513 REGCP_UNWIND(lastcp);
3515 /* Couldn't or didn't -- back up. */
3517 PL_reginput = locinput = HOPc(locinput, -1);
3525 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3527 c = UCHARAT(PL_reginput);
3529 /* If it could work, try it. */
3530 if (c1 == -1000 || c == c1 || c == c2)
3532 TRYPAREN(paren, n, PL_reginput);
3533 REGCP_UNWIND(lastcp);
3535 /* Couldn't or didn't -- back up. */
3537 PL_reginput = locinput = HOPc(locinput, -1);
3544 if (PL_reg_call_cc) {
3545 re_cc_state *cur_call_cc = PL_reg_call_cc;
3546 CURCUR *cctmp = PL_regcc;
3547 regexp *re = PL_reg_re;
3548 CHECKPOINT cp, lastcp;
3550 cp = regcppush(0); /* Save *all* the positions. */
3552 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3554 PL_reginput = locinput; /* Make position available to
3556 cache_re(PL_reg_call_cc->re);
3557 PL_regcc = PL_reg_call_cc->cc;
3558 PL_reg_call_cc = PL_reg_call_cc->prev;
3559 if (regmatch(cur_call_cc->node)) {
3560 PL_reg_call_cc = cur_call_cc;
3564 REGCP_UNWIND(lastcp);
3566 PL_reg_call_cc = cur_call_cc;
3572 PerlIO_printf(Perl_debug_log,
3573 "%*s continuation failed...\n",
3574 REPORT_CODE_OFF+PL_regindent*2, "")
3578 if (locinput < PL_regtill) {
3579 DEBUG_r(PerlIO_printf(Perl_debug_log,
3580 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3582 (long)(locinput - PL_reg_starttry),
3583 (long)(PL_regtill - PL_reg_starttry),
3585 sayNO_FINAL; /* Cannot match: too short. */
3587 PL_reginput = locinput; /* put where regtry can find it */
3588 sayYES_FINAL; /* Success! */
3590 PL_reginput = locinput; /* put where regtry can find it */
3591 sayYES_LOUD; /* Success! */
3594 PL_reginput = locinput;
3599 s = HOPBACKc(locinput, scan->flags);
3605 PL_reginput = locinput;
3610 s = HOPBACKc(locinput, scan->flags);
3616 PL_reginput = locinput;
3619 inner = NEXTOPER(NEXTOPER(scan));
3620 if (regmatch(inner) != n) {
3635 if (OP(scan) == SUSPEND) {
3636 locinput = PL_reginput;
3637 nextchr = UCHARAT(locinput);
3642 next = scan + ARG(scan);
3647 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3648 PTR2UV(scan), OP(scan));
3649 Perl_croak(aTHX_ "regexp memory corruption");
3656 * We get here only if there's trouble -- normally "case END" is
3657 * the terminating point.
3659 Perl_croak(aTHX_ "corrupted regexp pointers");
3665 PerlIO_printf(Perl_debug_log,
3666 "%*s %scould match...%s\n",
3667 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3671 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3672 PL_colors[4],PL_colors[5]));
3678 #if 0 /* Breaks $^R */
3686 PerlIO_printf(Perl_debug_log,
3687 "%*s %sfailed...%s\n",
3688 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3694 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3697 case RE_UNWIND_BRANCH:
3698 case RE_UNWIND_BRANCHJ:
3700 re_unwind_branch_t *uwb = &(uw->branch);
3701 I32 lastparen = uwb->lastparen;
3703 REGCP_UNWIND(uwb->lastcp);
3704 for (n = *PL_reglastparen; n > lastparen; n--)
3706 *PL_reglastparen = n;
3707 scan = next = uwb->next;
3709 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3710 ? BRANCH : BRANCHJ) ) { /* Failure */
3717 /* Have more choice yet. Reuse the same uwb. */
3719 if ((n = (uwb->type == RE_UNWIND_BRANCH
3720 ? NEXT_OFF(next) : ARG(next))))
3723 next = NULL; /* XXXX Needn't unwinding in this case... */
3725 next = NEXTOPER(scan);
3726 if (uwb->type == RE_UNWIND_BRANCHJ)
3727 next = NEXTOPER(next);
3728 locinput = uwb->locinput;
3729 nextchr = uwb->nextchr;
3731 PL_regindent = uwb->regindent;
3738 Perl_croak(aTHX_ "regexp unwind memory corruption");
3749 - regrepeat - repeatedly match something simple, report how many
3752 * [This routine now assumes that it will only match on things of length 1.
3753 * That was true before, but now we assume scan - reginput is the count,
3754 * rather than incrementing count on every character. [Er, except utf8.]]
3757 S_regrepeat(pTHX_ regnode *p, I32 max)
3759 register char *scan;
3761 register char *loceol = PL_regeol;
3762 register I32 hardcount = 0;
3763 register bool do_utf8 = PL_reg_match_utf8;
3766 if (max != REG_INFTY && max < loceol - scan)
3767 loceol = scan + max;
3772 while (scan < loceol && hardcount < max && *scan != '\n') {
3773 scan += UTF8SKIP(scan);
3777 while (scan < loceol && *scan != '\n')
3787 case EXACT: /* length of string is 1 */
3789 while (scan < loceol && UCHARAT(scan) == c)
3792 case EXACTF: /* length of string is 1 */
3794 while (scan < loceol &&
3795 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3798 case EXACTFL: /* length of string is 1 */
3799 PL_reg_flags |= RF_tainted;
3801 while (scan < loceol &&
3802 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3808 while (hardcount < max && scan < loceol &&
3809 reginclass(p, (U8*)scan, do_utf8)) {
3810 scan += UTF8SKIP(scan);
3814 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3821 LOAD_UTF8_CHARCLASS(alnum,"a");
3822 while (hardcount < max && scan < loceol &&
3823 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3824 scan += UTF8SKIP(scan);
3828 while (scan < loceol && isALNUM(*scan))
3833 PL_reg_flags |= RF_tainted;
3836 while (hardcount < max && scan < loceol &&
3837 isALNUM_LC_utf8((U8*)scan)) {
3838 scan += UTF8SKIP(scan);
3842 while (scan < loceol && isALNUM_LC(*scan))
3849 LOAD_UTF8_CHARCLASS(alnum,"a");
3850 while (hardcount < max && scan < loceol &&
3851 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3852 scan += UTF8SKIP(scan);
3856 while (scan < loceol && !isALNUM(*scan))
3861 PL_reg_flags |= RF_tainted;
3864 while (hardcount < max && scan < loceol &&
3865 !isALNUM_LC_utf8((U8*)scan)) {
3866 scan += UTF8SKIP(scan);
3870 while (scan < loceol && !isALNUM_LC(*scan))
3877 LOAD_UTF8_CHARCLASS(space," ");
3878 while (hardcount < max && scan < loceol &&
3880 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3881 scan += UTF8SKIP(scan);
3885 while (scan < loceol && isSPACE(*scan))
3890 PL_reg_flags |= RF_tainted;
3893 while (hardcount < max && scan < loceol &&
3894 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3895 scan += UTF8SKIP(scan);
3899 while (scan < loceol && isSPACE_LC(*scan))
3906 LOAD_UTF8_CHARCLASS(space," ");
3907 while (hardcount < max && scan < loceol &&
3909 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3910 scan += UTF8SKIP(scan);
3914 while (scan < loceol && !isSPACE(*scan))
3919 PL_reg_flags |= RF_tainted;
3922 while (hardcount < max && scan < loceol &&
3923 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3924 scan += UTF8SKIP(scan);
3928 while (scan < loceol && !isSPACE_LC(*scan))
3935 LOAD_UTF8_CHARCLASS(digit,"0");
3936 while (hardcount < max && scan < loceol &&
3937 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3938 scan += UTF8SKIP(scan);
3942 while (scan < loceol && isDIGIT(*scan))
3949 LOAD_UTF8_CHARCLASS(digit,"0");
3950 while (hardcount < max && scan < loceol &&
3951 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3952 scan += UTF8SKIP(scan);
3956 while (scan < loceol && !isDIGIT(*scan))
3960 default: /* Called on something of 0 width. */
3961 break; /* So match right here or not at all. */
3967 c = scan - PL_reginput;
3972 SV *prop = sv_newmortal();
3975 PerlIO_printf(Perl_debug_log,
3976 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3977 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3984 - regrepeat_hard - repeatedly match something, report total lenth and length
3986 * The repeater is supposed to have constant length.
3990 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
3992 register char *scan = Nullch;
3993 register char *start;
3994 register char *loceol = PL_regeol;
3996 I32 count = 0, res = 1;
4001 start = PL_reginput;
4002 if (PL_reg_match_utf8) {
4003 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4006 while (start < PL_reginput) {
4008 start += UTF8SKIP(start);
4019 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4021 *lp = l = PL_reginput - start;
4022 if (max != REG_INFTY && l*max < loceol - scan)
4023 loceol = scan + l*max;
4036 - regclass_swash - prepare the utf8 swash
4040 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
4045 if (PL_regdata && PL_regdata->count) {
4048 if (PL_regdata->what[n] == 's') {
4049 SV *rv = (SV*)PL_regdata->data[n];
4050 AV *av = (AV*)SvRV((SV*)rv);
4053 si = *av_fetch(av, 0, FALSE);
4054 a = av_fetch(av, 1, FALSE);
4058 else if (si && doinit) {
4059 sw = swash_init("utf8", "", si, 1, 0);
4060 (void)av_store(av, 1, sw);
4072 - reginclass - determine if a character falls into a character class
4076 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4078 char flags = ANYOF_FLAGS(n);
4083 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4085 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4086 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4087 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4090 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4093 SV *sw = regclass_swash(n, TRUE, 0);
4096 if (swash_fetch(sw, p, do_utf8))
4098 else if (flags & ANYOF_FOLD) {
4100 U8 tmpbuf[UTF8_MAXLEN*2+1];
4102 toLOWER_utf8(p, tmpbuf, &ulen);
4103 if (swash_fetch(sw, tmpbuf, do_utf8))
4109 if (!match && c < 256) {
4110 if (ANYOF_BITMAP_TEST(n, c))
4112 else if (flags & ANYOF_FOLD) {
4115 if (flags & ANYOF_LOCALE) {
4116 PL_reg_flags |= RF_tainted;
4117 f = PL_fold_locale[c];
4121 if (f != c && ANYOF_BITMAP_TEST(n, f))
4125 if (!match && (flags & ANYOF_CLASS)) {
4126 PL_reg_flags |= RF_tainted;
4128 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4129 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4130 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4131 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4132 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4133 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4134 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4135 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4136 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4137 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4138 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4139 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4140 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4141 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4142 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4143 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4144 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4145 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4146 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4147 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4148 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4149 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4150 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4151 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4152 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4153 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4154 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4155 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4156 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4157 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4158 ) /* How's that for a conditional? */
4165 return (flags & ANYOF_INVERT) ? !match : match;
4169 S_reghop(pTHX_ U8 *s, I32 off)
4171 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4175 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4178 while (off-- && s < lim) {
4179 /* XXX could check well-formedness here */
4187 if (UTF8_IS_CONTINUED(*s)) {
4188 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4191 /* XXX could check well-formedness here */
4199 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4201 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4205 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4208 while (off-- && s < lim) {
4209 /* XXX could check well-formedness here */
4219 if (UTF8_IS_CONTINUED(*s)) {
4220 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4223 /* XXX could check well-formedness here */
4235 restore_pos(pTHX_ void *arg)
4237 if (PL_reg_eval_set) {
4238 if (PL_reg_oldsaved) {
4239 PL_reg_re->subbeg = PL_reg_oldsaved;
4240 PL_reg_re->sublen = PL_reg_oldsavedlen;
4241 RX_MATCH_COPIED_on(PL_reg_re);
4243 PL_reg_magic->mg_len = PL_reg_oldpos;
4244 PL_reg_eval_set = 0;
4245 PL_curpm = PL_reg_oldcurpm;