X-Git-Url: https://perl5.git.perl.org/perl5.git/blobdiff_plain/9f7f391326e967b539b86ed051c163bbf8f6e7de..93d343c6c59d1d6b5e6307c31585c77ab830183d:/regexec.c diff --git a/regexec.c b/regexec.c index 9cb15b8..34fd1ac 100644 --- a/regexec.c +++ b/regexec.c @@ -107,7 +107,7 @@ #define STATIC static #endif -#define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c))) +#define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c))) /* * Forwards. @@ -213,11 +213,10 @@ S_regcppush(pTHX_ I32 parenfloor) (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp) STATIC char * -S_regcppop(pTHX) +S_regcppop(pTHX_ const regexp *rex) { dVAR; I32 i; - U32 paren = 0; char *input; GET_RE_DEBUG_FLAGS_DECL; @@ -235,7 +234,7 @@ S_regcppop(pTHX) for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS); i > 0; i -= REGCP_PAREN_ELEMS) { I32 tmps; - paren = (U32)SSPOPINT; + U32 paren = (U32)SSPOPINT; PL_reg_start_tmp[paren] = (char *) SSPOPPTR; PL_regstartp[paren] = SSPOPINT; tmps = SSPOPINT; @@ -251,10 +250,10 @@ S_regcppop(pTHX) ); } DEBUG_EXECUTE_r( - if ((I32)(*PL_reglastparen + 1) <= PL_regnpar) { + if ((I32)(*PL_reglastparen + 1) <= rex->nparens) { PerlIO_printf(Perl_debug_log, " restoring \\%"IVdf"..\\%"IVdf" to undef\n", - (IV)(*PL_reglastparen + 1), (IV)PL_regnpar); + (IV)(*PL_reglastparen + 1), (IV)rex->nparens); } ); #if 1 @@ -268,10 +267,10 @@ S_regcppop(pTHX) * building DynaLoader will fail: * "Error: '*' not in typemap in DynaLoader.xs, line 164" * --jhi */ - for (paren = *PL_reglastparen + 1; (I32)paren <= PL_regnpar; paren++) { - if ((I32)paren > PL_regsize) - PL_regstartp[paren] = -1; - PL_regendp[paren] = -1; + for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) { + if (i > PL_regsize) + PL_regstartp[i] = -1; + PL_regendp[i] = -1; } #endif return input; @@ -316,18 +315,6 @@ Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *stren nosave ? 0 : REXEC_COPY_STR); } -STATIC void -S_cache_re(pTHX_ regexp *prog) -{ - dVAR; - PL_regprecomp = prog->precomp; /* Needed for FAIL. */ -#ifdef DEBUGGING - PL_regprogram = prog->program; -#endif - PL_regnpar = prog->nparens; - PL_regdata = prog->data; - PL_reg_re = prog; -} /* * Need to implement the following flags for reg_anch: @@ -852,8 +839,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, : strend); t = s; - cache_re(prog); - s = find_byclass(prog, prog->regstclass, s, endpos, 1); + s = find_byclass(prog, prog->regstclass, s, endpos, NULL); if (!s) { #ifdef DEBUGGING const char *what = NULL; @@ -945,8 +931,11 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } /* We know what class REx starts with. Try to find this position... */ +/* if reginfo is NULL, its a dryrun */ + STATIC char * -S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 norun) +S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, const char +*strend, const regmatch_info *reginfo) { dVAR; const I32 doevery = (prog->reganch & ROPT_SKIP) == 0; @@ -967,9 +956,9 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 while (s + (uskip = UTF8SKIP(s)) <= strend) { if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) || !UTF8_IS_INVARIANT((U8)s[0]) ? - reginclass(c, (U8*)s, 0, do_utf8) : - REGINCLASS(c, (U8*)s)) { - if (tmp && (norun || regtry(prog, s))) + reginclass(prog, c, (U8*)s, 0, do_utf8) : + REGINCLASS(prog, c, (U8*)s)) { + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -983,12 +972,12 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 while (s < strend) { STRLEN skip = 1; - if (REGINCLASS(c, (U8*)s) || + if (REGINCLASS(prog, c, (U8*)s) || (ANYOF_FOLD_SHARP_S(c, s, strend) && /* The assignment of 2 is intentional: * for the folded sharp s, the skip is 2. */ (skip = SHARP_S_SKIP))) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1001,7 +990,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 break; case CANY: while (s < strend) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1046,7 +1035,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 do_exactf: e = HOP3c(strend, -((I32)lnc), s); - if (norun && e < s) + if (!reginfo && e < s) e = s; /* Due to minlen logic of intuit() */ /* The idea in the EXACTF* cases is to first find the @@ -1075,7 +1064,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 && (ln == len || ibcmp_utf8(s, (char **)0, 0, do_utf8, m, (char **)0, ln, (bool)UTF)) - && (norun || regtry(prog, s)) ) + && (!reginfo || regtry(reginfo, s)) ) goto got_it; else { U8 foldbuf[UTF8_MAXBYTES_CASE+1]; @@ -1088,7 +1077,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 (char **)0, foldlen, do_utf8, m, (char **)0, ln, (bool)UTF)) - && (norun || regtry(prog, s)) ) + && (!reginfo || regtry(reginfo, s)) ) goto got_it; } s += len; @@ -1114,7 +1103,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 && (ln == len || ibcmp_utf8(s, (char **)0, 0, do_utf8, m, (char **)0, ln, (bool)UTF)) - && (norun || regtry(prog, s)) ) + && (!reginfo || regtry(reginfo, s)) ) goto got_it; else { U8 foldbuf[UTF8_MAXBYTES_CASE+1]; @@ -1127,7 +1116,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 (char **)0, foldlen, do_utf8, m, (char **)0, ln, (bool)UTF)) - && (norun || regtry(prog, s)) ) + && (!reginfo || regtry(reginfo, s)) ) goto got_it; } s += len; @@ -1141,7 +1130,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 && (ln == 1 || !(OP(c) == EXACTF ? ibcmp(s, m, ln) : ibcmp_locale(s, m, ln))) - && (norun || regtry(prog, s)) ) + && (!reginfo || regtry(reginfo, s)) ) goto got_it; s++; } @@ -1151,7 +1140,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 && (ln == 1 || !(OP(c) == EXACTF ? ibcmp(s, m, ln) : ibcmp_locale(s, m, ln))) - && (norun || regtry(prog, s)) ) + && (!reginfo || regtry(reginfo, s)) ) goto got_it; s++; } @@ -1177,7 +1166,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 isALNUM_LC_utf8((U8*)s))) { tmp = !tmp; - if ((norun || regtry(prog, s))) + if ((!reginfo || regtry(reginfo, s))) goto got_it; } s += uskip; @@ -1190,13 +1179,13 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 if (tmp == !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) { tmp = !tmp; - if ((norun || regtry(prog, s))) + if ((!reginfo || regtry(reginfo, s))) goto got_it; } s++; } } - if ((!prog->minlen && tmp) && (norun || regtry(prog, s))) + if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, s))) goto got_it; break; case NBOUNDL: @@ -1218,7 +1207,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) : isALNUM_LC_utf8((U8*)s))) tmp = !tmp; - else if ((norun || regtry(prog, s))) + else if ((!reginfo || regtry(reginfo, s))) goto got_it; s += uskip; } @@ -1231,12 +1220,12 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 if (tmp == !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s))) tmp = !tmp; - else if ((norun || regtry(prog, s))) + else if ((!reginfo || regtry(reginfo, s))) goto got_it; s++; } } - if ((!prog->minlen && !tmp) && (norun || regtry(prog, s))) + if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, s))) goto got_it; break; case ALNUM: @@ -1244,7 +1233,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 LOAD_UTF8_CHARCLASS_ALNUM(); while (s + (uskip = UTF8SKIP(s)) <= strend) { if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1257,7 +1246,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 else { while (s < strend) { if (isALNUM(*s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1273,7 +1262,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 if (do_utf8) { while (s + (uskip = UTF8SKIP(s)) <= strend) { if (isALNUM_LC_utf8((U8*)s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1286,7 +1275,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 else { while (s < strend) { if (isALNUM_LC(*s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1302,7 +1291,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 LOAD_UTF8_CHARCLASS_ALNUM(); while (s + (uskip = UTF8SKIP(s)) <= strend) { if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1315,7 +1304,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 else { while (s < strend) { if (!isALNUM(*s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1331,7 +1320,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 if (do_utf8) { while (s + (uskip = UTF8SKIP(s)) <= strend) { if (!isALNUM_LC_utf8((U8*)s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1344,7 +1333,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 else { while (s < strend) { if (!isALNUM_LC(*s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1360,7 +1349,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 LOAD_UTF8_CHARCLASS_SPACE(); while (s + (uskip = UTF8SKIP(s)) <= strend) { if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1373,7 +1362,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 else { while (s < strend) { if (isSPACE(*s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1389,7 +1378,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 if (do_utf8) { while (s + (uskip = UTF8SKIP(s)) <= strend) { if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1402,7 +1391,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 else { while (s < strend) { if (isSPACE_LC(*s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1418,7 +1407,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 LOAD_UTF8_CHARCLASS_SPACE(); while (s + (uskip = UTF8SKIP(s)) <= strend) { if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1431,7 +1420,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 else { while (s < strend) { if (!isSPACE(*s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1447,7 +1436,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 if (do_utf8) { while (s + (uskip = UTF8SKIP(s)) <= strend) { if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1460,7 +1449,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 else { while (s < strend) { if (!isSPACE_LC(*s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1476,7 +1465,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 LOAD_UTF8_CHARCLASS_DIGIT(); while (s + (uskip = UTF8SKIP(s)) <= strend) { if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1489,7 +1478,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 else { while (s < strend) { if (isDIGIT(*s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1505,7 +1494,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 if (do_utf8) { while (s + (uskip = UTF8SKIP(s)) <= strend) { if (isDIGIT_LC_utf8((U8*)s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1518,7 +1507,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 else { while (s < strend) { if (isDIGIT_LC(*s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1534,7 +1523,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 LOAD_UTF8_CHARCLASS_DIGIT(); while (s + (uskip = UTF8SKIP(s)) <= strend) { if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1547,7 +1536,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 else { while (s < strend) { if (!isDIGIT(*s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1563,7 +1552,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 if (do_utf8) { while (s + (uskip = UTF8SKIP(s)) <= strend) { if (!isDIGIT_LC_utf8((U8*)s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1576,7 +1565,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 else { while (s < strend) { if (!isDIGIT_LC(*s)) { - if (tmp && (norun || regtry(prog, s))) + if (tmp && (!reginfo || regtry(reginfo, s))) goto got_it; else tmp = doevery; @@ -1619,21 +1608,16 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * char *scream_olds = NULL; SV* oreplsv = GvSV(PL_replgv); const bool do_utf8 = DO_UTF8(sv); - const I32 multiline = prog->reganch & PMf_MULTILINE; + I32 multiline; #ifdef DEBUGGING - SV * const dsv0 = PERL_DEBUG_PAD_ZERO(0); - SV * const dsv1 = PERL_DEBUG_PAD_ZERO(1); + SV* dsv0; + SV* dsv1; #endif + regmatch_info reginfo; /* create some info to pass to regtry etc */ GET_RE_DEBUG_FLAGS_DECL; PERL_UNUSED_ARG(data); - RX_MATCH_UTF8_set(prog,do_utf8); - - cache_re(prog); -#ifdef DEBUGGING - PL_regnarrate = DEBUG_r_TEST; -#endif /* Be paranoid... */ if (prog == NULL || startpos == NULL) { @@ -1641,6 +1625,16 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * return 0; } + multiline = prog->reganch & PMf_MULTILINE; + reginfo.prog = prog; + +#ifdef DEBUGGING + dsv0 = PERL_DEBUG_PAD_ZERO(0); + dsv1 = PERL_DEBUG_PAD_ZERO(1); +#endif + + RX_MATCH_UTF8_set(prog, do_utf8); + minlen = prog->minlen; if (strend - startpos < minlen) { DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, @@ -1661,40 +1655,37 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * PL_reg_flags |= RF_utf8; /* Mark beginning of line for ^ and lookbehind. */ - PL_regbol = startpos; + reginfo.bol = startpos; /* XXX not used ??? */ PL_bostr = strbeg; - PL_reg_sv = sv; + reginfo.sv = sv; /* Mark end of line for $ (and such) */ PL_regeol = strend; /* see how far we have to get to not match where we matched before */ - PL_regtill = startpos+minend; - - /* We start without call_cc context. */ - PL_reg_call_cc = 0; + reginfo.till = startpos+minend; /* If there is a "must appear" string, look for it. */ s = startpos; - if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */ + if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to set reginfo->ganch */ MAGIC *mg; if (flags & REXEC_IGNOREPOS) /* Means: check only at start */ - PL_reg_ganch = startpos; + reginfo.ganch = startpos; else if (sv && SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv) && (mg = mg_find(sv, PERL_MAGIC_regex_global)) && mg->mg_len >= 0) { - PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */ + reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ if (prog->reganch & ROPT_ANCH_GPOS) { - if (s > PL_reg_ganch) + if (s > reginfo.ganch) goto phooey; - s = PL_reg_ganch; + s = reginfo.ganch; } } else /* pos() not defined */ - PL_reg_ganch = strbeg; + reginfo.ganch = strbeg; } if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { @@ -1736,7 +1727,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * /* Simplest case: anchored match need be tried only once. */ /* [unless only anchor is BOL and multiline is set] */ if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) { - if (s == startpos && regtry(prog, startpos)) + if (s == startpos && regtry(®info, startpos)) goto got_it; else if (multiline || (prog->reganch & ROPT_IMPLICIT) || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */ @@ -1751,7 +1742,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * if (s == startpos) goto after_try; while (1) { - if (regtry(prog, s)) + if (regtry(®info, s)) goto got_it; after_try: if (s >= end) @@ -1769,7 +1760,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * s--; while (s < end) { if (*s++ == '\n') { /* don't need PL_utf8skip here */ - if (regtry(prog, s)) + if (regtry(®info, s)) goto got_it; } } @@ -1777,7 +1768,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * } goto phooey; } else if (prog->reganch & ROPT_ANCH_GPOS) { - if (regtry(prog, PL_reg_ganch)) + if (regtry(®info, reginfo.ganch)) goto got_it; goto phooey; } @@ -1798,7 +1789,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * while (s < strend) { if (*s == ch) { DEBUG_EXECUTE_r( did_match = 1 ); - if (regtry(prog, s)) goto got_it; + if (regtry(®info, s)) goto got_it; s += UTF8SKIP(s); while (s < strend && *s == ch) s += UTF8SKIP(s); @@ -1810,7 +1801,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * while (s < strend) { if (*s == ch) { DEBUG_EXECUTE_r( did_match = 1 ); - if (regtry(prog, s)) goto got_it; + if (regtry(®info, s)) goto got_it; s++; while (s < strend && *s == ch) s++; @@ -1888,14 +1879,14 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * } if (do_utf8) { while (s <= last1) { - if (regtry(prog, s)) + if (regtry(®info, s)) goto got_it; s += UTF8SKIP(s); } } else { while (s <= last1) { - if (regtry(prog, s)) + if (regtry(®info, s)) goto got_it; s++; } @@ -1927,7 +1918,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * int len0; int len1; - regprop(prop, c); + regprop(prog, prop, c); s0 = UTF ? pv_uni_display(dsv0, (U8*)SvPVX_const(prop), SvCUR(prop), 60, UNI_DISPLAY_REGEX) : @@ -1941,7 +1932,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * len0, len0, s0, len1, len1, s1); }); - if (find_byclass(prog, c, s, strend, 0)) + if (find_byclass(prog, c, s, strend, ®info)) goto got_it; DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n")); } @@ -1999,7 +1990,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * /* We don't know much -- general case. */ if (do_utf8) { for (;;) { - if (regtry(prog, s)) + if (regtry(®info, s)) goto got_it; if (s >= strend) break; @@ -2008,7 +1999,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * } else { do { - if (regtry(prog, s)) + if (regtry(®info, s)) goto got_it; } while (s++ < strend); } @@ -2026,7 +2017,7 @@ got_it: sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is restored, the value remains the same. */ - restore_pos(aTHX_ 0); + restore_pos(aTHX_ prog); } /* make sure $`, $&, $', and $digit will work later */ @@ -2066,7 +2057,7 @@ phooey: DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", PL_colors[4], PL_colors[5])); if (PL_reg_eval_set) - restore_pos(aTHX_ 0); + restore_pos(aTHX_ prog); return 0; } @@ -2074,13 +2065,13 @@ phooey: - regtry - try match at specific point */ STATIC I32 /* 0 failure, 1 success */ -S_regtry(pTHX_ regexp *prog, char *startpos) +S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos) { dVAR; - register I32 i; register I32 *sp; register I32 *ep; CHECKPOINT lastcp; + regexp *prog = reginfo->prog; GET_RE_DEBUG_FLAGS_DECL; #ifdef DEBUGGING @@ -2102,27 +2093,27 @@ S_regtry(pTHX_ regexp *prog, char *startpos) /* SAVEI8(cxstack[cxstack_ix].blk_gimme); cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ - if (PL_reg_sv) { + if (reginfo->sv) { /* Make $_ available to executed code. */ - if (PL_reg_sv != DEFSV) { + if (reginfo->sv != DEFSV) { SAVE_DEFSV; - DEFSV = PL_reg_sv; + DEFSV = reginfo->sv; } - if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv) - && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) { + if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) + && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { /* prepare for quick setting of pos */ #ifdef PERL_OLD_COPY_ON_WRITE if (SvIsCOW(sv)) sv_force_normal_flags(sv, 0); #endif - mg = sv_magicext(PL_reg_sv, (SV*)0, PERL_MAGIC_regex_global, + mg = sv_magicext(reginfo->sv, (SV*)0, PERL_MAGIC_regex_global, &PL_vtbl_mglob, NULL, 0); mg->mg_len = -1; } PL_reg_magic = mg; PL_reg_oldpos = mg->mg_len; - SAVEDESTRUCTOR_X(restore_pos, 0); + SAVEDESTRUCTOR_X(restore_pos, prog); } if (!PL_reg_curpm) { Newxz(PL_reg_curpm, 1, PMOP); @@ -2192,6 +2183,7 @@ S_regtry(pTHX_ regexp *prog, char *startpos) sp = prog->startp; ep = prog->endp; if (prog->nparens) { + register I32 i; for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) { *++sp = -1; *++ep = -1; @@ -2199,7 +2191,7 @@ S_regtry(pTHX_ regexp *prog, char *startpos) } #endif REGCP_SET(lastcp); - if (regmatch(prog->program + 1)) { + if (regmatch(reginfo, prog->program + 1)) { prog->endp[0] = PL_reginput - PL_bostr; return 1; } @@ -2226,6 +2218,7 @@ typedef struct { regnode *next; char *locinput; I32 nextchr; + int minmod; #ifdef DEBUGGING int regindent; #endif @@ -2250,7 +2243,7 @@ typedef union re_unwind_t { #define POSCACHE_SEEN 1 /* we know what we're caching */ #define POSCACHE_START 2 /* the real cache: this bit maps to pos 0 */ #define CACHEsayYES STMT_START { \ - if (st->whilem.cache_offset | st->whilem.cache_bit) { \ + if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) { \ if (!(PL_reg_poscache[0] & (1<whilem.cache_offset] &= ~(1<whilem.cache_bit); \ + PL_reg_poscache[st->u.whilem.cache_offset] &= ~(1<u.whilem.cache_bit); \ } \ } \ sayYES; \ } STMT_END #define CACHEsayNO STMT_START { \ - if (st->whilem.cache_offset | st->whilem.cache_bit) { \ + if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) { \ if (!(PL_reg_poscache[0] & (1<whilem.cache_offset] &= ~(1<whilem.cache_bit); \ + PL_reg_poscache[st->u.whilem.cache_offset] &= ~(1<u.whilem.cache_bit); \ } \ } \ sayNO; \ @@ -2319,6 +2312,44 @@ S_push_slab(pTHX) goto start_recurse; \ resume_point_##where: + +/* push a new regex state. Set newst to point to it */ + +#define PUSH_STATE(newst, resume) \ + depth++; \ + DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "PUSH STATE(%d)\n", depth)); \ + st->scan = scan; \ + st->next = next; \ + st->n = n; \ + st->locinput = locinput; \ + st->resume_state = resume; \ + newst = st+1; \ + if (newst > &(PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1])) \ + newst = S_push_slab(aTHX); \ + PL_regmatch_state = newst; \ + newst->cc = 0; \ + newst->minmod = 0; \ + newst->sw = 0; \ + newst->logical = 0; \ + newst->unwind = 0; \ + locinput = PL_reginput; \ + nextchr = UCHARAT(locinput); + +#define POP_STATE \ + DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "POP STATE(%d)\n", depth)); \ + depth--; \ + st--; \ + if (st < &PL_regmatch_slab->states[0]) { \ + PL_regmatch_slab = PL_regmatch_slab->prev; \ + st = &PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]; \ + } \ + PL_regmatch_state = st; \ + scan = st->scan; \ + next = st->next; \ + n = st->n; \ + locinput = st->locinput; \ + nextchr = UCHARAT(locinput); + /* - regmatch - main matching routine * @@ -2426,12 +2457,14 @@ S_push_slab(pTHX) STATIC I32 /* 0 failure, 1 success */ -S_regmatch(pTHX_ regnode *prog) +S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) { dVAR; register const bool do_utf8 = PL_reg_match_utf8; const U32 uniflags = UTF8_ALLOW_DEFAULT; + regexp *rex = reginfo->prog; + regmatch_slab *orig_slab; regmatch_state *orig_state; @@ -2449,6 +2482,8 @@ S_regmatch(pTHX_ regnode *prog) bool result; /* return value of S_regmatch */ regnode *inner; /* Next node in internal branch. */ int depth = 0; /* depth of recursion */ + regmatch_state *newst; /* when pushing a state, this is the new one */ + regmatch_state *cur_eval = NULL; /* most recent (??{}) state */ #ifdef DEBUGGING SV *re_debug_flags = NULL; @@ -2511,7 +2546,7 @@ S_regmatch(pTHX_ regnode *prog) pref0_len = 0; if (pref0_len > pref_len) pref0_len = pref_len; - regprop(prop, scan); + regprop(rex, prop, scan); { const char * const s0 = do_utf8 && OP(scan) != CANY ? @@ -2545,7 +2580,7 @@ S_regmatch(pTHX_ regnode *prog) PL_colors[1], 15 - l - pref_len + 1, "", - (IV)(scan - PL_regprogram), PL_regindent*2, "", + (IV)(scan - rex->program), PL_regindent*2, "", SvPVX_const(prop)); } }); @@ -2558,7 +2593,7 @@ S_regmatch(pTHX_ regnode *prog) case BOL: if (locinput == PL_bostr) { - /* regtill = regbol; */ + /* reginfo->till = reginfo->bol; */ break; } sayNO; @@ -2574,7 +2609,7 @@ S_regmatch(pTHX_ regnode *prog) break; sayNO; case GPOS: - if (locinput == PL_reg_ganch) + if (locinput == reginfo->ganch) break; sayNO; case EOL: @@ -2653,14 +2688,14 @@ S_regmatch(pTHX_ regnode *prog) /* what trie are we using right now */ reg_trie_data *trie - = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ]; - st->trie.accepted = 0; /* how many accepting states we have seen */ + = (reg_trie_data*)rex->data->data[ ARG( scan ) ]; + st->u.trie.accepted = 0; /* how many accepting states we have seen */ result = 0; while ( state && uc <= (U8*)PL_regeol ) { if (trie->states[ state ].wordnum) { - if (!st->trie.accepted ) { + if (!st->u.trie.accepted ) { ENTER; SAVETMPS; bufflen = TRIE_INITAL_ACCEPT_BUFFLEN; @@ -2670,22 +2705,22 @@ S_regmatch(pTHX_ regnode *prog) sizeof(reg_trie_accepted)); SvPOK_on(sv_accept_buff); sv_2mortal(sv_accept_buff); - st->trie.accept_buff = + st->u.trie.accept_buff = (reg_trie_accepted*)SvPV_nolen(sv_accept_buff ); } else { - if (st->trie.accepted >= bufflen) { + if (st->u.trie.accepted >= bufflen) { bufflen *= 2; - st->trie.accept_buff =(reg_trie_accepted*) + st->u.trie.accept_buff =(reg_trie_accepted*) SvGROW(sv_accept_buff, bufflen * sizeof(reg_trie_accepted)); } SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff) + sizeof(reg_trie_accepted)); } - st->trie.accept_buff[st->trie.accepted].wordnum = trie->states[state].wordnum; - st->trie.accept_buff[st->trie.accepted].endpos = uc; - ++st->trie.accepted; + st->u.trie.accept_buff[st->u.trie.accepted].wordnum = trie->states[state].wordnum; + st->u.trie.accept_buff[st->u.trie.accepted].endpos = uc; + ++st->u.trie.accepted; } base = trie->states[ state ].trans.base; @@ -2694,7 +2729,7 @@ S_regmatch(pTHX_ regnode *prog) PerlIO_printf( Perl_debug_log, "%*s %sState: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ", REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], - (UV)state, (UV)base, (UV)st->trie.accepted ); + (UV)state, (UV)base, (UV)st->u.trie.accepted ); ); if ( base ) { @@ -2761,7 +2796,7 @@ S_regmatch(pTHX_ regnode *prog) charid, uvc, (UV)state, PL_colors[5] ); ); } - if (!st->trie.accepted ) + if (!st->u.trie.accepted ) sayNO; /* @@ -2777,17 +2812,17 @@ S_regmatch(pTHX_ regnode *prog) have been tried. */ - if ( st->trie.accepted == 1 ) { + if ( st->u.trie.accepted == 1 ) { DEBUG_EXECUTE_r({ - SV **tmp = av_fetch( trie->words, st->trie.accept_buff[ 0 ].wordnum-1, 0 ); + SV ** const tmp = av_fetch( trie->words, st->u.trie.accept_buff[ 0 ].wordnum-1, 0 ); PerlIO_printf( Perl_debug_log, "%*s %sonly one match : #%d <%s>%s\n", REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], - st->trie.accept_buff[ 0 ].wordnum, + st->u.trie.accept_buff[ 0 ].wordnum, tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr", PL_colors[5] ); }); - PL_reginput = (char *)st->trie.accept_buff[ 0 ].endpos; + PL_reginput = (char *)st->u.trie.accept_buff[ 0 ].endpos; /* in this case we free tmps/leave before we call regmatch as we wont be using accept_buff again. */ FREETMPS; @@ -2797,40 +2832,42 @@ S_regmatch(pTHX_ regnode *prog) } else { DEBUG_EXECUTE_r( PerlIO_printf( Perl_debug_log,"%*s %sgot %"IVdf" possible matches%s\n", - REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], (IV)st->trie.accepted, + REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], (IV)st->u.trie.accepted, PL_colors[5] ); ); - while ( !result && st->trie.accepted-- ) { + while ( !result && st->u.trie.accepted-- ) { U32 best = 0; U32 cur; - for( cur = 1 ; cur <= st->trie.accepted ; cur++ ) { + for( cur = 1 ; cur <= st->u.trie.accepted ; cur++ ) { DEBUG_TRIE_EXECUTE_r( PerlIO_printf( Perl_debug_log, "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n", REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], - (IV)best, st->trie.accept_buff[ best ].wordnum, (IV)cur, - st->trie.accept_buff[ cur ].wordnum, PL_colors[5] ); + (IV)best, st->u.trie.accept_buff[ best ].wordnum, (IV)cur, + st->u.trie.accept_buff[ cur ].wordnum, PL_colors[5] ); ); - if (st->trie.accept_buff[cur].wordnum < - st->trie.accept_buff[best].wordnum) + if (st->u.trie.accept_buff[cur].wordnum < + st->u.trie.accept_buff[best].wordnum) best = cur; } DEBUG_EXECUTE_r({ - SV ** const tmp = av_fetch( trie->words, st->trie.accept_buff[ best ].wordnum - 1, 0 ); + reg_trie_data * const trie = (reg_trie_data*) + rex->data->data[ARG(scan)]; + SV ** const tmp = av_fetch( trie->words, st->u.trie.accept_buff[ best ].wordnum - 1, 0 ); PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at 0x%p%s\n", REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], - st->trie.accept_buff[best].wordnum, + st->u.trie.accept_buff[best].wordnum, tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",scan, PL_colors[5] ); }); - if ( besttrie.accepted ) { - reg_trie_accepted tmp = st->trie.accept_buff[ best ]; - st->trie.accept_buff[ best ] = st->trie.accept_buff[ st->trie.accepted ]; - st->trie.accept_buff[ st->trie.accepted ] = tmp; - best = st->trie.accepted; + if ( bestu.trie.accepted ) { + reg_trie_accepted tmp = st->u.trie.accept_buff[ best ]; + st->u.trie.accept_buff[ best ] = st->u.trie.accept_buff[ st->u.trie.accepted ]; + st->u.trie.accept_buff[ st->u.trie.accepted ] = tmp; + best = st->u.trie.accepted; } - PL_reginput = (char *)st->trie.accept_buff[ best ].endpos; + PL_reginput = (char *)st->u.trie.accept_buff[ best ].endpos; /* as far as I can tell we only need the SAVETMPS/FREETMPS @@ -2958,7 +2995,7 @@ S_regmatch(pTHX_ regnode *prog) if (do_utf8) { STRLEN inclasslen = PL_regeol - locinput; - if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8)) + if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8)) sayNO_ANYOF; if (locinput >= PL_regeol) sayNO; @@ -2969,7 +3006,7 @@ S_regmatch(pTHX_ regnode *prog) else { if (nextchr < 0) nextchr = UCHARAT(locinput); - if (!REGINCLASS(scan, (U8*)locinput)) + if (!REGINCLASS(rex, scan, (U8*)locinput)) sayNO_ANYOF; if (!nextchr && locinput >= PL_regeol) sayNO; @@ -3263,21 +3300,21 @@ S_regmatch(pTHX_ regnode *prog) break; case EVAL: { - dSP; - OP_4tree * const oop = PL_op; - COP * const ocurcop = PL_curcop; - PAD *old_comppad; SV *ret; - struct regexp * const oreg = PL_reg_re; - - n = ARG(scan); - PL_op = (OP_4tree*)PL_regdata->data[n]; - DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); - PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]); - PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr; - { + /* execute the code in the {...} */ + dSP; SV ** const before = SP; + OP_4tree * const oop = PL_op; + COP * const ocurcop = PL_curcop; + PAD *old_comppad; + + n = ARG(scan); + PL_op = (OP_4tree*)rex->data->data[n]; + DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); + PAD_SAVE_LOCAL(old_comppad, (PAD*)rex->data->data[n + 2]); + PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr; + CALLRUNOPS(aTHX); /* Scalar context. */ SPAGAIN; if (SP == before) @@ -3286,19 +3323,24 @@ S_regmatch(pTHX_ regnode *prog) ret = POPs; PUTBACK; } + + PL_op = oop; + PAD_RESTORE_LOCAL(old_comppad); + PL_curcop = ocurcop; + if (!st->logical) { + /* /(?{...})/ */ + sv_setsv(save_scalar(PL_replgv), ret); + break; + } } + if (st->logical == 2) { /* Postponed subexpression: /(??{...})/ */ + regexp *re; + { + /* extract RE object from returned value; compiling if + * necessary */ - PL_op = oop; - PAD_RESTORE_LOCAL(old_comppad); - PL_curcop = ocurcop; - if (st->logical) { - if (st->logical == 2) { /* Postponed subexpression. */ - regexp *re; MAGIC *mg = NULL; - re_cc_state state; - int toggleutf; - register SV *sv; - + SV *sv; if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret))) mg = mg_find(sv, PERL_MAGIC_qr); else if (SvSMAGICAL(ret)) { @@ -3316,95 +3358,62 @@ S_regmatch(pTHX_ regnode *prog) STRLEN len; const char * const t = SvPV_const(ret, len); PMOP pm; - char * const oprecomp = PL_regprecomp; const I32 osize = PL_regsize; - const I32 onpar = PL_regnpar; Zero(&pm, 1, PMOP); - if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8; + if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8; re = CALLREGCOMP(aTHX_ (char*)t, (char*)t + len, &pm); if (!(SvFLAGS(ret) & (SVs_TEMP | SVs_PADTMP | SVf_READONLY | SVs_GMG))) sv_magic(ret,(SV*)ReREFCNT_inc(re), PERL_MAGIC_qr,0,0); - PL_regprecomp = oprecomp; PL_regsize = osize; - PL_regnpar = onpar; } - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "Entering embedded \"%s%.60s%s%s\"\n", - PL_colors[0], - re->precomp, - PL_colors[1], - (strlen(re->precomp) > 60 ? "..." : "")) - ); - state.node = next; - state.prev = PL_reg_call_cc; - state.cc = st->cc; - state.re = PL_reg_re; + } - st->cc = 0; - - st->eval.cp = regcppush(0); /* Save *all* the positions. */ - REGCP_SET(st->eval.lastcp); - cache_re(re); - state.ss = PL_savestack_ix; - *PL_reglastparen = 0; - *PL_reglastcloseparen = 0; - PL_reg_call_cc = &state; - PL_reginput = locinput; - toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^ - ((re->reganch & ROPT_UTF8) != 0); - if (toggleutf) PL_reg_flags ^= RF_utf8; - - /* XXXX This is too dramatic a measure... */ - PL_reg_maxiter = 0; - - /* XXX the only recursion left in regmatch() */ - if (regmatch(re->program + 1)) { - /* Even though we succeeded, we need to restore - global variables, since we may be wrapped inside - SUSPEND, thus the match may be not finished yet. */ - - /* XXXX Do this only if SUSPENDed? */ - PL_reg_call_cc = state.prev; - st->cc = state.cc; - PL_reg_re = state.re; - cache_re(PL_reg_re); - if (toggleutf) PL_reg_flags ^= RF_utf8; - - /* XXXX This is too dramatic a measure... */ - PL_reg_maxiter = 0; - - /* These are needed even if not SUSPEND. */ - ReREFCNT_dec(re); - regcpblow(st->eval.cp); - sayYES; - } - ReREFCNT_dec(re); - REGCP_UNWIND(st->eval.lastcp); - regcppop(); - PL_reg_call_cc = state.prev; - st->cc = state.cc; - PL_reg_re = state.re; - cache_re(PL_reg_re); - if (toggleutf) PL_reg_flags ^= RF_utf8; - - /* XXXX This is too dramatic a measure... */ - PL_reg_maxiter = 0; + /* run the pattern returned from (??{...}) */ + + DEBUG_EXECUTE_r( + PerlIO_printf(Perl_debug_log, + "Entering embedded \"%s%.60s%s%s\"\n", + PL_colors[0], + re->precomp, + PL_colors[1], + (strlen(re->precomp) > 60 ? "..." : "")) + ); + + st->u.eval.cp = regcppush(0); /* Save *all* the positions. */ + REGCP_SET(st->u.eval.lastcp); + *PL_reglastparen = 0; + *PL_reglastcloseparen = 0; + PL_reginput = locinput; + + /* XXXX This is too dramatic a measure... */ + PL_reg_maxiter = 0; - st->logical = 0; - sayNO; - } - st->sw = SvTRUE(ret); st->logical = 0; + st->u.eval.toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^ + ((re->reganch & ROPT_UTF8) != 0); + if (st->u.eval.toggleutf) PL_reg_flags ^= RF_utf8; + st->u.eval.prev_rex = rex; + rex = re; + + st->u.eval.prev_eval = cur_eval; + st->u.eval.prev_slab = PL_regmatch_slab; + st->u.eval.depth = depth; + cur_eval = st; + PUSH_STATE(newst, resume_EVAL); + st = newst; + + /* now continue from first node in postoned RE */ + next = re->program + 1; + break; + /* NOTREACHED */ } - else { - sv_setsv(save_scalar(PL_replgv), ret); - cache_re(oreg); - } + /* /(?(?{...})X|Y)/ */ + st->sw = SvTRUE(ret); + st->logical = 0; break; } case OPEN: @@ -3439,20 +3448,21 @@ S_regmatch(pTHX_ regnode *prog) st->logical = scan->flags; break; /******************************************************************* - cc contains infoblock about the innermost (...)* loop, and - a pointer to the next outer infoblock. + cc points to the regmatch_state associated with the most recent CURLYX. + This struct contains info about the innermost (...)* loop (an + "infoblock"), and a pointer to the next outer cc. Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM): 1) After matching Y, regnode for CURLYX is processed; - 2) This regnode mallocs an infoblock, and calls regmatch() recursively + 2) This regnode populates cc, and calls regmatch() recursively with the starting point at WHILEM node; 3) Each hit of WHILEM node tries to match A and Z (in the order depending on the current iteration, min/max of {min,max} and greediness). The information about where are nodes for "A" - and "Z" is read from the infoblock, as is info on how many times "A" + and "Z" is read from cc, as is info on how many times "A" was already matched, and greediness. 4) After A matches, the same WHILEM node is hit again. @@ -3463,7 +3473,7 @@ S_regmatch(pTHX_ regnode *prog) as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node of the external loop. - Currently present infoblocks form a tree with a stem formed by PL_curcc + Currently present infoblocks form a tree with a stem formed by st->cc and whatever it mentions via ->next, and additional attached trees corresponding to temporarily unset infoblocks as in "5" above. @@ -3518,34 +3528,32 @@ S_regmatch(pTHX_ regnode *prog) /* No need to save/restore up to this paren */ I32 parenfloor = scan->flags; - { - CURCUR *newcc; - Newx(newcc, 1, CURCUR); - st->curlyx.savecc = st->cc; - newcc->oldcc = st->cc; - st->cc = newcc; - } - st->curlyx.cp = PL_savestack_ix; - if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ + if (next && (OP(PREVOPER(next)) == NOTHING)) /* LONGJMP */ next += ARG(next); /* XXXX Probably it is better to teach regpush to support parenfloor > PL_regsize... */ if (parenfloor > (I32)*PL_reglastparen) parenfloor = *PL_reglastparen; /* Pessimization... */ - st->cc->parenfloor = parenfloor; - st->cc->cur = -1; - st->cc->min = ARG1(scan); - st->cc->max = ARG2(scan); - st->cc->scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS; - st->cc->next = next; - st->cc->minmod = st->minmod; - st->cc->lastloc = 0; + + st->u.curlyx.cp = PL_savestack_ix; + st->u.curlyx.outercc = st->cc; + st->cc = st; + /* these fields contain the state of the current curly. + * they are accessed by subsequent WHILEMs; + * cur and lastloc are also updated by WHILEM */ + st->u.curlyx.parenfloor = parenfloor; + st->u.curlyx.cur = -1; /* this will be updated by WHILEM */ + st->u.curlyx.min = ARG1(scan); + st->u.curlyx.max = ARG2(scan); + st->u.curlyx.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS; + st->u.curlyx.lastloc = 0; + /* st->next and st->minmod are also read by WHILEM */ + PL_reginput = locinput; REGMATCH(PREVOPER(next), CURLYX); /* start on the WHILEM */ /*** all unsaved local vars undefined at this point */ - regcpblow(st->curlyx.cp); - Safefree(st->cc); - st->cc = st->curlyx.savecc; + regcpblow(st->u.curlyx.cp); + st->cc = st->u.curlyx.outercc; saySAME(result); } /* NOTREACHED */ @@ -3559,54 +3567,54 @@ S_regmatch(pTHX_ regnode *prog) * that we can try again after backing off. */ - st->whilem.lastloc = st->cc->lastloc; /* Detection of 0-len. */ - st->whilem.cache_offset = 0; - st->whilem.cache_bit = 0; + st->u.whilem.lastloc = st->cc->u.curlyx.lastloc; /* Detection of 0-len. */ + st->u.whilem.cache_offset = 0; + st->u.whilem.cache_bit = 0; - n = st->cc->cur + 1; /* how many we know we matched */ + n = st->cc->u.curlyx.cur + 1; /* how many we know we matched */ PL_reginput = locinput; DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, "%*s %ld out of %ld..%ld cc=%"UVxf"\n", REPORT_CODE_OFF+PL_regindent*2, "", - (long)n, (long)st->cc->min, - (long)st->cc->max, PTR2UV(st->cc)) + (long)n, (long)st->cc->u.curlyx.min, + (long)st->cc->u.curlyx.max, PTR2UV(st->cc)) ); /* If degenerate scan matches "", assume scan done. */ - if (locinput == st->cc->lastloc && n >= st->cc->min) { - st->whilem.savecc = st->cc; - st->cc = st->cc->oldcc; + if (locinput == st->cc->u.curlyx.lastloc && n >= st->cc->u.curlyx.min) { + st->u.whilem.savecc = st->cc; + st->cc = st->cc->u.curlyx.outercc; if (st->cc) - st->ln = st->cc->cur; + st->ln = st->cc->u.curlyx.cur; DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, "%*s empty match detected, try continuation...\n", REPORT_CODE_OFF+PL_regindent*2, "") ); - REGMATCH(st->whilem.savecc->next, WHILEM1); + REGMATCH(st->u.whilem.savecc->next, WHILEM1); /*** all unsaved local vars undefined at this point */ - st->cc = st->whilem.savecc; + st->cc = st->u.whilem.savecc; if (result) sayYES; - if (st->cc->oldcc) - st->cc->oldcc->cur = st->ln; + if (st->cc->u.curlyx.outercc) + st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln; sayNO; } /* First just match a string of min scans. */ - if (n < st->cc->min) { - st->cc->cur = n; - st->cc->lastloc = locinput; - REGMATCH(st->cc->scan, WHILEM2); + if (n < st->cc->u.curlyx.min) { + st->cc->u.curlyx.cur = n; + st->cc->u.curlyx.lastloc = locinput; + REGMATCH(st->cc->u.curlyx.scan, WHILEM2); /*** all unsaved local vars undefined at this point */ if (result) sayYES; - st->cc->cur = n - 1; - st->cc->lastloc = st->whilem.lastloc; + st->cc->u.curlyx.cur = n - 1; + st->cc->u.curlyx.lastloc = st->u.whilem.lastloc; sayNO; } @@ -3638,13 +3646,13 @@ S_regmatch(pTHX_ regnode *prog) ); } if (PL_reg_leftiter < 0) { - st->whilem.cache_offset = locinput - PL_bostr; + st->u.whilem.cache_offset = locinput - PL_bostr; - st->whilem.cache_offset = (scan->flags & 0xf) - 1 + POSCACHE_START - + st->whilem.cache_offset * (scan->flags>>4); - st->whilem.cache_bit = st->whilem.cache_offset % 8; - st->whilem.cache_offset /= 8; - if (PL_reg_poscache[st->whilem.cache_offset] & (1<whilem.cache_bit)) { + st->u.whilem.cache_offset = (scan->flags & 0xf) - 1 + POSCACHE_START + + st->u.whilem.cache_offset * (scan->flags>>4); + st->u.whilem.cache_bit = st->u.whilem.cache_offset % 8; + st->u.whilem.cache_offset /= 8; + if (PL_reg_poscache[st->u.whilem.cache_offset] & (1<u.whilem.cache_bit)) { DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, "%*s already tried at this position...\n", @@ -3657,32 +3665,32 @@ S_regmatch(pTHX_ regnode *prog) /* cache records failure */ sayNO_SILENT; } - PL_reg_poscache[st->whilem.cache_offset] |= (1<whilem.cache_bit); + PL_reg_poscache[st->u.whilem.cache_offset] |= (1<u.whilem.cache_bit); } } /* Prefer next over scan for minimal matching. */ if (st->cc->minmod) { - st->whilem.savecc = st->cc; - st->cc = st->cc->oldcc; + st->u.whilem.savecc = st->cc; + st->cc = st->cc->u.curlyx.outercc; if (st->cc) - st->ln = st->cc->cur; - st->whilem.cp = regcppush(st->whilem.savecc->parenfloor); - REGCP_SET(st->whilem.lastcp); - REGMATCH(st->whilem.savecc->next, WHILEM3); + st->ln = st->cc->u.curlyx.cur; + st->u.whilem.cp = regcppush(st->u.whilem.savecc->u.curlyx.parenfloor); + REGCP_SET(st->u.whilem.lastcp); + REGMATCH(st->u.whilem.savecc->next, WHILEM3); /*** all unsaved local vars undefined at this point */ - st->cc = st->whilem.savecc; + st->cc = st->u.whilem.savecc; if (result) { - regcpblow(st->whilem.cp); + regcpblow(st->u.whilem.cp); CACHEsayYES; /* All done. */ } - REGCP_UNWIND(st->whilem.lastcp); - regcppop(); - if (st->cc->oldcc) - st->cc->oldcc->cur = st->ln; + REGCP_UNWIND(st->u.whilem.lastcp); + regcppop(rex); + if (st->cc->u.curlyx.outercc) + st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln; - if (n >= st->cc->max) { /* Maximum greed exceeded? */ + if (n >= st->cc->u.curlyx.max) { /* Maximum greed exceeded? */ if (ckWARN(WARN_REGEXP) && n >= REG_INFTY && !(PL_reg_flags & RF_warned)) { PL_reg_flags |= RF_warned; @@ -3700,38 +3708,38 @@ S_regmatch(pTHX_ regnode *prog) ); /* Try scanning more and see if it helps. */ PL_reginput = locinput; - st->cc->cur = n; - st->cc->lastloc = locinput; - st->whilem.cp = regcppush(st->cc->parenfloor); - REGCP_SET(st->whilem.lastcp); - REGMATCH(st->cc->scan, WHILEM4); + st->cc->u.curlyx.cur = n; + st->cc->u.curlyx.lastloc = locinput; + st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor); + REGCP_SET(st->u.whilem.lastcp); + REGMATCH(st->cc->u.curlyx.scan, WHILEM4); /*** all unsaved local vars undefined at this point */ if (result) { - regcpblow(st->whilem.cp); + regcpblow(st->u.whilem.cp); CACHEsayYES; } - REGCP_UNWIND(st->whilem.lastcp); - regcppop(); - st->cc->cur = n - 1; - st->cc->lastloc = st->whilem.lastloc; + REGCP_UNWIND(st->u.whilem.lastcp); + regcppop(rex); + st->cc->u.curlyx.cur = n - 1; + st->cc->u.curlyx.lastloc = st->u.whilem.lastloc; CACHEsayNO; } /* Prefer scan over next for maximal matching. */ - if (n < st->cc->max) { /* More greed allowed? */ - st->whilem.cp = regcppush(st->cc->parenfloor); - st->cc->cur = n; - st->cc->lastloc = locinput; - REGCP_SET(st->whilem.lastcp); - REGMATCH(st->cc->scan, WHILEM5); + if (n < st->cc->u.curlyx.max) { /* More greed allowed? */ + st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor); + st->cc->u.curlyx.cur = n; + st->cc->u.curlyx.lastloc = locinput; + REGCP_SET(st->u.whilem.lastcp); + REGMATCH(st->cc->u.curlyx.scan, WHILEM5); /*** all unsaved local vars undefined at this point */ if (result) { - regcpblow(st->whilem.cp); + regcpblow(st->u.whilem.cp); CACHEsayYES; } - REGCP_UNWIND(st->whilem.lastcp); - regcppop(); /* Restore some previous $s? */ + REGCP_UNWIND(st->u.whilem.lastcp); + regcppop(rex); /* Restore some previous $s? */ PL_reginput = locinput; DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, @@ -3748,19 +3756,19 @@ S_regmatch(pTHX_ regnode *prog) } /* Failed deeper matches of scan, so see if this one works. */ - st->whilem.savecc = st->cc; - st->cc = st->cc->oldcc; + st->u.whilem.savecc = st->cc; + st->cc = st->cc->u.curlyx.outercc; if (st->cc) - st->ln = st->cc->cur; - REGMATCH(st->whilem.savecc->next, WHILEM6); + st->ln = st->cc->u.curlyx.cur; + REGMATCH(st->u.whilem.savecc->next, WHILEM6); /*** all unsaved local vars undefined at this point */ - st->cc = st->whilem.savecc; + st->cc = st->u.whilem.savecc; if (result) CACHEsayYES; - if (st->cc->oldcc) - st->cc->oldcc->cur = st->ln; - st->cc->cur = n - 1; - st->cc->lastloc = st->whilem.lastloc; + if (st->cc->u.curlyx.outercc) + st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln; + st->cc->u.curlyx.cur = n - 1; + st->cc->u.curlyx.lastloc = st->u.whilem.lastloc; CACHEsayNO; } /* NOTREACHED */ @@ -3776,7 +3784,7 @@ S_regmatch(pTHX_ regnode *prog) { I32 type; type = OP(scan); - if (OP(next) != type) /* No choice. */ + if (!next || OP(next) != type) /* No choice. */ next = inner; /* Avoid recursion. */ else { const I32 lastparen = *PL_reglastparen; @@ -3793,6 +3801,7 @@ S_regmatch(pTHX_ regnode *prog) uw->next = next; uw->locinput = locinput; uw->nextchr = nextchr; + uw->minmod = st->minmod; #ifdef DEBUGGING uw->regindent = ++PL_regindent; #endif @@ -3809,45 +3818,45 @@ S_regmatch(pTHX_ regnode *prog) break; case CURLYM: { - st->curlym.l = st->curlym.matches = 0; + st->u.curlym.l = st->u.curlym.matches = 0; /* We suppose that the next guy does not need backtracking: in particular, it is of constant non-zero length, and has no parenths to influence future backrefs. */ st->ln = ARG1(scan); /* min to match */ n = ARG2(scan); /* max to match */ - st->curlym.paren = scan->flags; - if (st->curlym.paren) { - if (st->curlym.paren > PL_regsize) - PL_regsize = st->curlym.paren; - if (st->curlym.paren > (I32)*PL_reglastparen) - *PL_reglastparen = st->curlym.paren; + st->u.curlym.paren = scan->flags; + if (st->u.curlym.paren) { + if (st->u.curlym.paren > PL_regsize) + PL_regsize = st->u.curlym.paren; + if (st->u.curlym.paren > (I32)*PL_reglastparen) + *PL_reglastparen = st->u.curlym.paren; } scan = NEXTOPER(scan) + NODE_STEP_REGNODE; - if (st->curlym.paren) + if (st->u.curlym.paren) scan += NEXT_OFF(scan); /* Skip former OPEN. */ PL_reginput = locinput; - st->curlym.maxwanted = st->minmod ? st->ln : n; - if (st->curlym.maxwanted) { - while (PL_reginput < PL_regeol && st->curlym.matches < st->curlym.maxwanted) { + st->u.curlym.maxwanted = st->minmod ? st->ln : n; + if (st->u.curlym.maxwanted) { + while (PL_reginput < PL_regeol && st->u.curlym.matches < st->u.curlym.maxwanted) { REGMATCH(scan, CURLYM1); /*** all unsaved local vars undefined at this point */ if (!result) break; - /* on first match, determine length, curlym.l */ - if (!st->curlym.matches++) { + /* on first match, determine length, u.curlym.l */ + if (!st->u.curlym.matches++) { if (PL_reg_match_utf8) { char *s = locinput; while (s < PL_reginput) { - st->curlym.l++; + st->u.curlym.l++; s += UTF8SKIP(s); } } else { - st->curlym.l = PL_reginput - locinput; + st->u.curlym.l = PL_reginput - locinput; } - if (st->curlym.l == 0) { - st->curlym.matches = st->curlym.maxwanted; + if (st->u.curlym.l == 0) { + st->u.curlym.matches = st->u.curlym.maxwanted; break; } } @@ -3859,52 +3868,52 @@ S_regmatch(pTHX_ regnode *prog) if (st->minmod) { st->minmod = 0; - if (st->ln && st->curlym.matches < st->ln) + if (st->ln && st->u.curlym.matches < st->ln) sayNO; if (HAS_TEXT(next) || JUMPABLE(next)) { regnode *text_node = next; if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node); - if (! HAS_TEXT(text_node)) st->curlym.c1 = st->curlym.c2 = -1000; + if (! HAS_TEXT(text_node)) st->u.curlym.c1 = st->u.curlym.c2 = -1000; else { if (PL_regkind[(U8)OP(text_node)] == REF) { - st->curlym.c1 = st->curlym.c2 = -1000; + st->u.curlym.c1 = st->u.curlym.c2 = -1000; goto assume_ok_MM; } - else { st->curlym.c1 = (U8)*STRING(text_node); } + else { st->u.curlym.c1 = (U8)*STRING(text_node); } if (OP(text_node) == EXACTF || OP(text_node) == REFF) - st->curlym.c2 = PL_fold[st->curlym.c1]; + st->u.curlym.c2 = PL_fold[st->u.curlym.c1]; else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL) - st->curlym.c2 = PL_fold_locale[st->curlym.c1]; + st->u.curlym.c2 = PL_fold_locale[st->u.curlym.c1]; else - st->curlym.c2 = st->curlym.c1; + st->u.curlym.c2 = st->u.curlym.c1; } } else - st->curlym.c1 = st->curlym.c2 = -1000; + st->u.curlym.c1 = st->u.curlym.c2 = -1000; assume_ok_MM: - REGCP_SET(st->curlym.lastcp); + REGCP_SET(st->u.curlym.lastcp); while (n >= st->ln || (n == REG_INFTY && st->ln > 0)) { /* ln overflow ? */ /* If it could work, try it. */ - if (st->curlym.c1 == -1000 || - UCHARAT(PL_reginput) == st->curlym.c1 || - UCHARAT(PL_reginput) == st->curlym.c2) + if (st->u.curlym.c1 == -1000 || + UCHARAT(PL_reginput) == st->u.curlym.c1 || + UCHARAT(PL_reginput) == st->u.curlym.c2) { - if (st->curlym.paren) { + if (st->u.curlym.paren) { if (st->ln) { - PL_regstartp[st->curlym.paren] = - HOPc(PL_reginput, -st->curlym.l) - PL_bostr; - PL_regendp[st->curlym.paren] = PL_reginput - PL_bostr; + PL_regstartp[st->u.curlym.paren] = + HOPc(PL_reginput, -st->u.curlym.l) - PL_bostr; + PL_regendp[st->u.curlym.paren] = PL_reginput - PL_bostr; } else - PL_regendp[st->curlym.paren] = -1; + PL_regendp[st->u.curlym.paren] = -1; } REGMATCH(next, CURLYM2); /*** all unsaved local vars undefined at this point */ if (result) sayYES; - REGCP_UNWIND(st->curlym.lastcp); + REGCP_UNWIND(st->u.curlym.lastcp); } /* Couldn't or didn't -- move forward. */ PL_reginput = locinput; @@ -3923,65 +3932,65 @@ S_regmatch(pTHX_ regnode *prog) PerlIO_printf(Perl_debug_log, "%*s matched %"IVdf" times, len=%"IVdf"...\n", (int)(REPORT_CODE_OFF+PL_regindent*2), "", - (IV) st->curlym.matches, (IV)st->curlym.l) + (IV) st->u.curlym.matches, (IV)st->u.curlym.l) ); - if (st->curlym.matches >= st->ln) { + if (st->u.curlym.matches >= st->ln) { if (HAS_TEXT(next) || JUMPABLE(next)) { regnode *text_node = next; if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node); - if (! HAS_TEXT(text_node)) st->curlym.c1 = st->curlym.c2 = -1000; + if (! HAS_TEXT(text_node)) st->u.curlym.c1 = st->u.curlym.c2 = -1000; else { if (PL_regkind[(U8)OP(text_node)] == REF) { - st->curlym.c1 = st->curlym.c2 = -1000; + st->u.curlym.c1 = st->u.curlym.c2 = -1000; goto assume_ok_REG; } - else { st->curlym.c1 = (U8)*STRING(text_node); } + else { st->u.curlym.c1 = (U8)*STRING(text_node); } if (OP(text_node) == EXACTF || OP(text_node) == REFF) - st->curlym.c2 = PL_fold[st->curlym.c1]; + st->u.curlym.c2 = PL_fold[st->u.curlym.c1]; else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL) - st->curlym.c2 = PL_fold_locale[st->curlym.c1]; + st->u.curlym.c2 = PL_fold_locale[st->u.curlym.c1]; else - st->curlym.c2 = st->curlym.c1; + st->u.curlym.c2 = st->u.curlym.c1; } } else - st->curlym.c1 = st->curlym.c2 = -1000; + st->u.curlym.c1 = st->u.curlym.c2 = -1000; } assume_ok_REG: - REGCP_SET(st->curlym.lastcp); - while (st->curlym.matches >= st->ln) { + REGCP_SET(st->u.curlym.lastcp); + while (st->u.curlym.matches >= st->ln) { /* If it could work, try it. */ - if (st->curlym.c1 == -1000 || - UCHARAT(PL_reginput) == st->curlym.c1 || - UCHARAT(PL_reginput) == st->curlym.c2) + if (st->u.curlym.c1 == -1000 || + UCHARAT(PL_reginput) == st->u.curlym.c1 || + UCHARAT(PL_reginput) == st->u.curlym.c2) { DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, "%*s trying tail with matches=%"IVdf"...\n", (int)(REPORT_CODE_OFF+PL_regindent*2), - "", (IV)st->curlym.matches) + "", (IV)st->u.curlym.matches) ); - if (st->curlym.paren) { - if (st->curlym.matches) { - PL_regstartp[st->curlym.paren] - = HOPc(PL_reginput, -st->curlym.l) - PL_bostr; - PL_regendp[st->curlym.paren] = PL_reginput - PL_bostr; + if (st->u.curlym.paren) { + if (st->u.curlym.matches) { + PL_regstartp[st->u.curlym.paren] + = HOPc(PL_reginput, -st->u.curlym.l) - PL_bostr; + PL_regendp[st->u.curlym.paren] = PL_reginput - PL_bostr; } else - PL_regendp[st->curlym.paren] = -1; + PL_regendp[st->u.curlym.paren] = -1; } REGMATCH(next, CURLYM4); /*** all unsaved local vars undefined at this point */ if (result) sayYES; - REGCP_UNWIND(st->curlym.lastcp); + REGCP_UNWIND(st->u.curlym.lastcp); } /* Couldn't or didn't -- back up. */ - st->curlym.matches--; - locinput = HOPc(locinput, -st->curlym.l); + st->u.curlym.matches--; + locinput = HOPc(locinput, -st->u.curlym.l); PL_reginput = locinput; } } @@ -3990,17 +3999,17 @@ S_regmatch(pTHX_ regnode *prog) break; } case CURLYN: - st->plus.paren = scan->flags; /* Which paren to set */ - if (st->plus.paren > PL_regsize) - PL_regsize = st->plus.paren; - if (st->plus.paren > (I32)*PL_reglastparen) - *PL_reglastparen = st->plus.paren; + st->u.plus.paren = scan->flags; /* Which paren to set */ + if (st->u.plus.paren > PL_regsize) + PL_regsize = st->u.plus.paren; + if (st->u.plus.paren > (I32)*PL_reglastparen) + *PL_reglastparen = st->u.plus.paren; st->ln = ARG1(scan); /* min to match */ n = ARG2(scan); /* max to match */ scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE); goto repeat; case CURLY: - st->plus.paren = 0; + st->u.plus.paren = 0; st->ln = ARG1(scan); /* min to match */ n = ARG2(scan); /* max to match */ scan = NEXTOPER(scan) + NODE_STEP_REGNODE; @@ -4009,13 +4018,13 @@ S_regmatch(pTHX_ regnode *prog) st->ln = 0; n = REG_INFTY; scan = NEXTOPER(scan); - st->plus.paren = 0; + st->u.plus.paren = 0; goto repeat; case PLUS: st->ln = 1; n = REG_INFTY; scan = NEXTOPER(scan); - st->plus.paren = 0; + st->u.plus.paren = 0; repeat: /* * Lookahead to avoid useless match attempts @@ -4034,20 +4043,20 @@ S_regmatch(pTHX_ regnode *prog) if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node); - if (! HAS_TEXT(text_node)) st->plus.c1 = st->plus.c2 = -1000; + if (! HAS_TEXT(text_node)) st->u.plus.c1 = st->u.plus.c2 = -1000; else { if (PL_regkind[(U8)OP(text_node)] == REF) { - st->plus.c1 = st->plus.c2 = -1000; + st->u.plus.c1 = st->u.plus.c2 = -1000; goto assume_ok_easy; } else { s = (U8*)STRING(text_node); } if (!UTF) { - st->plus.c2 = st->plus.c1 = *s; + st->u.plus.c2 = st->u.plus.c1 = *s; if (OP(text_node) == EXACTF || OP(text_node) == REFF) - st->plus.c2 = PL_fold[st->plus.c1]; + st->u.plus.c2 = PL_fold[st->u.plus.c1]; else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL) - st->plus.c2 = PL_fold_locale[st->plus.c1]; + st->u.plus.c2 = PL_fold_locale[st->u.plus.c1]; } else { /* UTF */ if (OP(text_node) == EXACTF || OP(text_node) == REFF) { @@ -4058,117 +4067,117 @@ S_regmatch(pTHX_ regnode *prog) to_utf8_lower((U8*)s, tmpbuf1, &ulen1); to_utf8_upper((U8*)s, tmpbuf2, &ulen2); - st->plus.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0, + st->u.plus.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0, uniflags); - st->plus.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0, + st->u.plus.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0, uniflags); } else { - st->plus.c2 = st->plus.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0, + st->u.plus.c2 = st->u.plus.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0, uniflags); } } } } else - st->plus.c1 = st->plus.c2 = -1000; + st->u.plus.c1 = st->u.plus.c2 = -1000; assume_ok_easy: PL_reginput = locinput; if (st->minmod) { st->minmod = 0; - if (st->ln && regrepeat(scan, st->ln) < st->ln) + if (st->ln && regrepeat(rex, scan, st->ln) < st->ln) sayNO; locinput = PL_reginput; - REGCP_SET(st->plus.lastcp); - if (st->plus.c1 != -1000) { - st->plus.old = locinput; - st->plus.count = 0; + REGCP_SET(st->u.plus.lastcp); + if (st->u.plus.c1 != -1000) { + st->u.plus.old = locinput; + st->u.plus.count = 0; if (n == REG_INFTY) { - st->plus.e = PL_regeol - 1; + st->u.plus.e = PL_regeol - 1; if (do_utf8) - while (UTF8_IS_CONTINUATION(*(U8*)st->plus.e)) - st->plus.e--; + while (UTF8_IS_CONTINUATION(*(U8*)st->u.plus.e)) + st->u.plus.e--; } else if (do_utf8) { int m = n - st->ln; - for (st->plus.e = locinput; - m >0 && st->plus.e + UTF8SKIP(st->plus.e) <= PL_regeol; m--) - st->plus.e += UTF8SKIP(st->plus.e); + for (st->u.plus.e = locinput; + m >0 && st->u.plus.e + UTF8SKIP(st->u.plus.e) <= PL_regeol; m--) + st->u.plus.e += UTF8SKIP(st->u.plus.e); } else { - st->plus.e = locinput + n - st->ln; - if (st->plus.e >= PL_regeol) - st->plus.e = PL_regeol - 1; + st->u.plus.e = locinput + n - st->ln; + if (st->u.plus.e >= PL_regeol) + st->u.plus.e = PL_regeol - 1; } while (1) { /* Find place 'next' could work */ if (!do_utf8) { - if (st->plus.c1 == st->plus.c2) { - while (locinput <= st->plus.e && - UCHARAT(locinput) != st->plus.c1) + if (st->u.plus.c1 == st->u.plus.c2) { + while (locinput <= st->u.plus.e && + UCHARAT(locinput) != st->u.plus.c1) locinput++; } else { - while (locinput <= st->plus.e - && UCHARAT(locinput) != st->plus.c1 - && UCHARAT(locinput) != st->plus.c2) + while (locinput <= st->u.plus.e + && UCHARAT(locinput) != st->u.plus.c1 + && UCHARAT(locinput) != st->u.plus.c2) locinput++; } - st->plus.count = locinput - st->plus.old; + st->u.plus.count = locinput - st->u.plus.old; } else { - if (st->plus.c1 == st->plus.c2) { + if (st->u.plus.c1 == st->u.plus.c2) { STRLEN len; /* count initialised to * utf8_distance(old, locinput) */ - while (locinput <= st->plus.e && + while (locinput <= st->u.plus.e && utf8n_to_uvchr((U8*)locinput, UTF8_MAXBYTES, &len, - uniflags) != (UV)st->plus.c1) { + uniflags) != (UV)st->u.plus.c1) { locinput += len; - st->plus.count++; + st->u.plus.count++; } } else { STRLEN len; /* count initialised to * utf8_distance(old, locinput) */ - while (locinput <= st->plus.e) { + while (locinput <= st->u.plus.e) { UV c = utf8n_to_uvchr((U8*)locinput, UTF8_MAXBYTES, &len, uniflags); - if (c == (UV)st->plus.c1 || c == (UV)st->plus.c2) + if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2) break; locinput += len; - st->plus.count++; + st->u.plus.count++; } } } - if (locinput > st->plus.e) + if (locinput > st->u.plus.e) sayNO; /* PL_reginput == old now */ - if (locinput != st->plus.old) { + if (locinput != st->u.plus.old) { st->ln = 1; /* Did some */ - if (regrepeat(scan, st->plus.count) < st->plus.count) + if (regrepeat(rex, scan, st->u.plus.count) < st->u.plus.count) sayNO; } /* PL_reginput == locinput now */ - TRYPAREN(st->plus.paren, st->ln, locinput, PLUS1); + TRYPAREN(st->u.plus.paren, st->ln, locinput, PLUS1); /*** all unsaved local vars undefined at this point */ PL_reginput = locinput; /* Could be reset... */ - REGCP_UNWIND(st->plus.lastcp); + REGCP_UNWIND(st->u.plus.lastcp); /* Couldn't or didn't -- move forward. */ - st->plus.old = locinput; + st->u.plus.old = locinput; if (do_utf8) locinput += UTF8SKIP(locinput); else locinput++; - st->plus.count = 1; + st->u.plus.count = 1; } } else while (n >= st->ln || (n == REG_INFTY && st->ln > 0)) { /* ln overflow ? */ UV c; - if (st->plus.c1 != -1000) { + if (st->u.plus.c1 != -1000) { if (do_utf8) c = utf8n_to_uvchr((U8*)PL_reginput, UTF8_MAXBYTES, 0, @@ -4176,23 +4185,23 @@ S_regmatch(pTHX_ regnode *prog) else c = UCHARAT(PL_reginput); /* If it could work, try it. */ - if (c == (UV)st->plus.c1 || c == (UV)st->plus.c2) + if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2) { - TRYPAREN(st->plus.paren, st->ln, PL_reginput, PLUS2); + TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS2); /*** all unsaved local vars undefined at this point */ - REGCP_UNWIND(st->plus.lastcp); + REGCP_UNWIND(st->u.plus.lastcp); } } /* If it could work, try it. */ - else if (st->plus.c1 == -1000) + else if (st->u.plus.c1 == -1000) { - TRYPAREN(st->plus.paren, st->ln, PL_reginput, PLUS3); + TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS3); /*** all unsaved local vars undefined at this point */ - REGCP_UNWIND(st->plus.lastcp); + REGCP_UNWIND(st->u.plus.lastcp); } /* Couldn't or didn't -- move forward. */ PL_reginput = locinput; - if (regrepeat(scan, 1)) { + if (regrepeat(rex, scan, 1)) { st->ln++; locinput = PL_reginput; } @@ -4201,7 +4210,7 @@ S_regmatch(pTHX_ regnode *prog) } } else { - n = regrepeat(scan, n); + n = regrepeat(rex, scan, n); locinput = PL_reginput; if (st->ln < n && PL_regkind[(U8)OP(next)] == EOL && (OP(next) != MEOL || @@ -4214,11 +4223,11 @@ S_regmatch(pTHX_ regnode *prog) if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS) st->ln--; } - REGCP_SET(st->plus.lastcp); + REGCP_SET(st->u.plus.lastcp); { UV c = 0; while (n >= st->ln) { - if (st->plus.c1 != -1000) { + if (st->u.plus.c1 != -1000) { if (do_utf8) c = utf8n_to_uvchr((U8*)PL_reginput, UTF8_MAXBYTES, 0, @@ -4227,11 +4236,11 @@ S_regmatch(pTHX_ regnode *prog) c = UCHARAT(PL_reginput); } /* If it could work, try it. */ - if (st->plus.c1 == -1000 || c == (UV)st->plus.c1 || c == (UV)st->plus.c2) + if (st->u.plus.c1 == -1000 || c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2) { - TRYPAREN(st->plus.paren, n, PL_reginput, PLUS4); + TRYPAREN(st->u.plus.paren, n, PL_reginput, PLUS4); /*** all unsaved local vars undefined at this point */ - REGCP_UNWIND(st->plus.lastcp); + REGCP_UNWIND(st->u.plus.lastcp); } /* Couldn't or didn't -- back up. */ n--; @@ -4242,56 +4251,54 @@ S_regmatch(pTHX_ regnode *prog) sayNO; break; case END: - if (PL_reg_call_cc) { - st->end.cur_call_cc = PL_reg_call_cc; - st->end.end_re = PL_reg_re; + if (cur_eval) { + /* we have successfully completed the execution of a + * postponed re. Pop all states back to the last EVAL + * then continue with the node following the (??{...}) + */ + + /* this simulates a POP_STATE, except that it pops several + * levels, and doesn't restore locinput */ + + st = cur_eval; + PL_regmatch_slab = st->u.eval.prev_slab; + cur_eval = st->u.eval.prev_eval; + depth = st->u.eval.depth; - /* Save *all* the positions. */ - st->end.cp = regcppush(0); - REGCP_SET(st->end.lastcp); + PL_regmatch_state = st; + scan = st->scan; + next = st->next; + n = st->n; - /* Restore parens of the caller. */ + if (st->u.eval.toggleutf) + PL_reg_flags ^= RF_utf8; + ReREFCNT_dec(rex); + rex = st->u.eval.prev_rex; + /* XXXX This is too dramatic a measure... */ + PL_reg_maxiter = 0; + + /* Restore parens of the caller without popping the + * savestack */ { I32 tmp = PL_savestack_ix; - PL_savestack_ix = PL_reg_call_cc->ss; - regcppop(); + PL_savestack_ix = st->u.eval.lastcp; + regcppop(rex); PL_savestack_ix = tmp; } - /* Make position available to the callcc. */ - PL_reginput = locinput; - cache_re(PL_reg_call_cc->re); - st->end.savecc = st->cc; - st->cc = PL_reg_call_cc->cc; - PL_reg_call_cc = PL_reg_call_cc->prev; - REGMATCH(st->end.cur_call_cc->node, END); - /*** all unsaved local vars undefined at this point */ - if (result) { - PL_reg_call_cc = st->end.cur_call_cc; - regcpblow(st->end.cp); - sayYES; - } - REGCP_UNWIND(st->end.lastcp); - regcppop(); - PL_reg_call_cc = st->end.cur_call_cc; - st->cc = st->end.savecc; - PL_reg_re = st->end.end_re; - cache_re(st->end.end_re); + PL_reginput = locinput; + /* resume at node following the (??{...}) */ + break; - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s continuation failed...\n", - REPORT_CODE_OFF+PL_regindent*2, "") - ); - sayNO_SILENT; } - if (locinput < PL_regtill) { + + if (locinput < reginfo->till) { DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n", PL_colors[4], (long)(locinput - PL_reg_starttry), - (long)(PL_regtill - PL_reg_starttry), + (long)(reginfo->till - PL_reg_starttry), PL_colors[5])); sayNO_FINAL; /* Cannot match: too short. */ } @@ -4387,7 +4394,6 @@ S_regmatch(pTHX_ regnode *prog) oldst->next = next; oldst->n = n; oldst->locinput = locinput; - oldst->reg_call_cc = PL_reg_call_cc; st->cc = oldst->cc; locinput = PL_reginput; @@ -4402,6 +4408,8 @@ S_regmatch(pTHX_ regnode *prog) } } + + /* * We get here only if there's trouble -- normally "case END" is * the terminating point. @@ -4426,7 +4434,56 @@ yes: #endif result = 1; - goto exit_level; + /* XXX this is duplicate(ish) code to that in the do_no section. + * eventually a yes should just pop the whole stack */ + if (depth) { + /* restore previous state and re-enter */ + POP_STATE; + + switch (st->resume_state) { + case resume_TRIE1: + goto resume_point_TRIE1; + case resume_TRIE2: + goto resume_point_TRIE2; + case resume_EVAL: + break; + case resume_CURLYX: + goto resume_point_CURLYX; + case resume_WHILEM1: + goto resume_point_WHILEM1; + case resume_WHILEM2: + goto resume_point_WHILEM2; + case resume_WHILEM3: + goto resume_point_WHILEM3; + case resume_WHILEM4: + goto resume_point_WHILEM4; + case resume_WHILEM5: + goto resume_point_WHILEM5; + case resume_WHILEM6: + goto resume_point_WHILEM6; + case resume_CURLYM1: + goto resume_point_CURLYM1; + case resume_CURLYM2: + goto resume_point_CURLYM2; + case resume_CURLYM3: + goto resume_point_CURLYM3; + case resume_CURLYM4: + goto resume_point_CURLYM4; + case resume_IFMATCH: + goto resume_point_IFMATCH; + case resume_PLUS1: + goto resume_point_PLUS1; + case resume_PLUS2: + goto resume_point_PLUS2; + case resume_PLUS3: + goto resume_point_PLUS3; + case resume_PLUS4: + goto resume_point_PLUS4; + default: + Perl_croak(aTHX_ "regexp resume memory corruption"); + } + } + goto final_exit; no: DEBUG_EXECUTE_r( @@ -4452,6 +4509,7 @@ do_no: PL_regendp[n] = -1; *PL_reglastparen = n; scan = next = uwb->next; + st->minmod = uwb->minmod; if ( !scan || OP(scan) != (uwb->type == RE_UNWIND_BRANCH ? BRANCH : BRANCHJ) ) { /* Failure */ @@ -4485,33 +4543,37 @@ do_no: } /* NOTREACHED */ } + #ifdef DEBUGGING PL_regindent--; #endif result = 0; -exit_level: - - if (depth--) { - /* restore previous state and re-enter */ - st--; - if (st < &PL_regmatch_slab->states[0]) { - PL_regmatch_slab = PL_regmatch_slab->prev; - st = &PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]; - } - PL_regmatch_state = st; - - PL_reg_call_cc = st->reg_call_cc; - scan = st->scan; - next = st->next; - n = st->n; - locinput = st->locinput; - nextchr = UCHARAT(locinput); + if (depth) { + /* there's a previous state to backtrack to */ + POP_STATE; switch (st->resume_state) { case resume_TRIE1: goto resume_point_TRIE1; case resume_TRIE2: goto resume_point_TRIE2; + case resume_EVAL: + /* we have failed an (??{...}). Restore state to the outer re + * then re-throw the failure */ + if (st->u.eval.toggleutf) + PL_reg_flags ^= RF_utf8; + ReREFCNT_dec(rex); + rex = st->u.eval.prev_rex; + cur_eval = st->u.eval.prev_eval; + + /* XXXX This is too dramatic a measure... */ + PL_reg_maxiter = 0; + + PL_reginput = locinput; + REGCP_UNWIND(st->u.eval.lastcp); + regcppop(rex); + goto do_no; + case resume_CURLYX: goto resume_point_CURLYX; case resume_WHILEM1: @@ -4544,13 +4606,13 @@ exit_level: goto resume_point_PLUS3; case resume_PLUS4: goto resume_point_PLUS4; - case resume_END: - goto resume_point_END; default: Perl_croak(aTHX_ "regexp resume memory corruption"); } - /* NOTREACHED */ } + +final_exit: + /* restore original high-water mark */ PL_regmatch_slab = orig_slab; PL_regmatch_state = orig_state; @@ -4562,6 +4624,7 @@ exit_level: while (sl) { osl = sl; sl = sl->next; + Safefree(osl); } } @@ -4578,7 +4641,7 @@ exit_level: * rather than incrementing count on every character. [Er, except utf8.]] */ STATIC I32 -S_regrepeat(pTHX_ const regnode *p, I32 max) +S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max) { dVAR; register char *scan; @@ -4591,7 +4654,7 @@ S_regrepeat(pTHX_ const regnode *p, I32 max) if (max == REG_INFTY) max = I32_MAX; else if (max < loceol - scan) - loceol = scan + max; + loceol = scan + max; switch (OP(p)) { case REG_ANY: if (do_utf8) { @@ -4641,12 +4704,12 @@ S_regrepeat(pTHX_ const regnode *p, I32 max) if (do_utf8) { loceol = PL_regeol; while (hardcount < max && scan < loceol && - reginclass(p, (U8*)scan, 0, do_utf8)) { + reginclass(prog, p, (U8*)scan, 0, do_utf8)) { scan += UTF8SKIP(scan); hardcount++; } } else { - while (scan < loceol && REGINCLASS(p, (U8*)scan)) + while (scan < loceol && REGINCLASS(prog, p, (U8*)scan)) scan++; } break; @@ -4807,7 +4870,7 @@ S_regrepeat(pTHX_ const regnode *p, I32 max) SV * const prop = sv_newmortal(); GET_RE_DEBUG_FLAGS; DEBUG_EXECUTE_r({ - regprop(prop, p); + regprop(prog, prop, p); PerlIO_printf(Perl_debug_log, "%*s %s can match %"IVdf" times out of %"IVdf"...\n", REPORT_CODE_OFF+1, "", SvPVX_const(prop),(IV)c,(IV)max); @@ -4823,18 +4886,19 @@ S_regrepeat(pTHX_ const regnode *p, I32 max) */ SV * -Perl_regclass_swash(pTHX_ register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) +Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool doinit, SV** listsvp, SV **altsvp) { dVAR; SV *sw = NULL; SV *si = NULL; SV *alt = NULL; + const struct reg_data *data = prog ? prog->data : NULL; - if (PL_regdata && PL_regdata->count) { + if (data && data->count) { const U32 n = ARG(node); - if (PL_regdata->what[n] == 's') { - SV * const rv = (SV*)PL_regdata->data[n]; + if (data->what[n] == 's') { + SV * const rv = (SV*)data->data[n]; AV * const av = (AV*)SvRV((SV*)rv); SV **const ary = AvARRAY(av); SV **a, **b; @@ -4876,7 +4940,7 @@ Perl_regclass_swash(pTHX_ register const regnode* node, bool doinit, SV** listsv */ STATIC bool -S_reginclass(pTHX_ register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8) +S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8) { dVAR; const char flags = ANYOF_FLAGS(n); @@ -4905,7 +4969,7 @@ S_reginclass(pTHX_ register const regnode *n, register const U8* p, STRLEN* lenp match = TRUE; if (!match) { AV *av; - SV * const sw = regclass_swash(n, TRUE, 0, (SV**)&av); + SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av); if (sw) { if (swash_fetch(sw, p, do_utf8)) @@ -5058,15 +5122,15 @@ static void restore_pos(pTHX_ void *arg) { dVAR; - PERL_UNUSED_ARG(arg); + regexp * const rex = (regexp *)arg; if (PL_reg_eval_set) { if (PL_reg_oldsaved) { - PL_reg_re->subbeg = PL_reg_oldsaved; - PL_reg_re->sublen = PL_reg_oldsavedlen; + rex->subbeg = PL_reg_oldsaved; + rex->sublen = PL_reg_oldsavedlen; #ifdef PERL_OLD_COPY_ON_WRITE - PL_reg_re->saved_copy = PL_nrs; + rex->saved_copy = PL_nrs; #endif - RX_MATCH_COPIED_on(PL_reg_re); + RX_MATCH_COPIED_on(rex); } PL_reg_magic->mg_len = PL_reg_oldpos; PL_reg_eval_set = 0; @@ -5078,8 +5142,8 @@ STATIC void S_to_utf8_substr(pTHX_ register regexp *prog) { if (prog->float_substr && !prog->float_utf8) { - SV* sv; - prog->float_utf8 = sv = newSVsv(prog->float_substr); + SV* const sv = newSVsv(prog->float_substr); + prog->float_utf8 = sv; sv_utf8_upgrade(sv); if (SvTAIL(prog->float_substr)) SvTAIL_on(sv); @@ -5087,8 +5151,8 @@ S_to_utf8_substr(pTHX_ register regexp *prog) prog->check_utf8 = sv; } if (prog->anchored_substr && !prog->anchored_utf8) { - SV* sv; - prog->anchored_utf8 = sv = newSVsv(prog->anchored_substr); + SV* const sv = newSVsv(prog->anchored_substr); + prog->anchored_utf8 = sv; sv_utf8_upgrade(sv); if (SvTAIL(prog->anchored_substr)) SvTAIL_on(sv); @@ -5102,8 +5166,8 @@ S_to_byte_substr(pTHX_ register regexp *prog) { dVAR; if (prog->float_utf8 && !prog->float_substr) { - SV* sv; - prog->float_substr = sv = newSVsv(prog->float_utf8); + SV* sv = newSVsv(prog->float_utf8); + prog->float_substr = sv; if (sv_utf8_downgrade(sv, TRUE)) { if (SvTAIL(prog->float_utf8)) SvTAIL_on(sv); @@ -5115,8 +5179,8 @@ S_to_byte_substr(pTHX_ register regexp *prog) prog->check_substr = sv; } if (prog->anchored_utf8 && !prog->anchored_substr) { - SV* sv; - prog->anchored_substr = sv = newSVsv(prog->anchored_utf8); + SV* sv = newSVsv(prog->anchored_utf8); + prog->anchored_substr = sv; if (sv_utf8_downgrade(sv, TRUE)) { if (SvTAIL(prog->anchored_utf8)) SvTAIL_on(sv);