X-Git-Url: https://perl5.git.perl.org/perl5.git/blobdiff_plain/b94e2f88b5f2c63d65c0fcd37652030bb9aee3ef..77cb431f8c908a31b286bd0ba26649cb6041bd72:/regexec.c diff --git a/regexec.c b/regexec.c index de95e31..4d0388d 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. @@ -116,28 +116,20 @@ #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv)) #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b) -#define reghop_c(pos,off) ((char*)reghop((U8*)pos, off)) -#define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off)) -#define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off)) -#define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off)) -#define HOPc(pos,off) ((char*)HOP(pos,off)) -#define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off)) - -#define HOPBACK(pos, off) ( \ - (PL_reg_match_utf8) \ - ? reghopmaybe((U8*)pos, -off) \ +#define HOPc(pos,off) ((char *)(PL_reg_match_utf8 \ + ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \ + : (U8*)(pos + off))) +#define HOPBACKc(pos, off) ((char*) \ + ((PL_reg_match_utf8) \ + ? reghopmaybe3((U8*)pos, -off, ((U8*)(off < 0 ? PL_regeol : PL_bostr))) \ : (pos - off >= PL_bostr) \ ? (U8*)(pos - off) \ - : (U8*)NULL \ + : (U8*)NULL) \ ) -#define HOPBACKc(pos, off) (char*)HOPBACK(pos, off) -#define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim)) #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim)) #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off)) -#define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off)) #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim)) -#define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim)) #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \ if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END @@ -178,6 +170,7 @@ static void restore_pos(pTHX_ void *arg); STATIC CHECKPOINT S_regcppush(pTHX_ I32 parenfloor) { + dVAR; const int retval = PL_savestack_ix; #define REGCP_PAREN_ELEMS 4 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; @@ -220,10 +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; @@ -241,7 +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; @@ -257,10 +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 @@ -274,38 +267,18 @@ 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; } -STATIC char * -S_regcp_set_to(pTHX_ I32 ss) -{ - const I32 tmp = PL_savestack_ix; - - PL_savestack_ix = ss; - regcppop(); - PL_savestack_ix = tmp; - return Nullch; -} - -typedef struct re_cc_state -{ - I32 ss; - regnode *node; - struct re_cc_state *prev; - CURCUR *cc; - regexp *re; -} re_cc_state; - #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ -#define TRYPAREN(paren, n, input) { \ +#define TRYPAREN(paren, n, input, where) { \ if (paren) { \ if (n) { \ PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \ @@ -314,7 +287,8 @@ typedef struct re_cc_state else \ PL_regendp[paren] = -1; \ } \ - if (regmatch(next)) \ + REGMATCH(next, where); \ + if (result) \ sayYES; \ if (paren && n) \ PL_regendp[paren] = -1; \ @@ -341,17 +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) -{ - 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: @@ -403,6 +366,7 @@ char * Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, char *strend, U32 flags, re_scream_pos_data *data) { + dVAR; register I32 start_shift = 0; /* Should be nonnegative! */ register I32 end_shift = 0; @@ -412,8 +376,8 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, char *t; const int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */ I32 ml_anch; - register char *other_last = Nullch; /* other substr checked before this */ - char *check_at = Nullch; /* check substr found at this pos */ + register char *other_last = NULL; /* other substr checked before this */ + char *check_at = NULL; /* check substr found at this pos */ const I32 multiline = prog->reganch & PMf_MULTILINE; #ifdef DEBUGGING const char * const i_strpos = strpos; @@ -839,9 +803,9 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr); if (do_utf8 ? prog->check_substr : prog->check_utf8) SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8); - prog->check_substr = prog->check_utf8 = Nullsv; /* disable */ - prog->float_substr = prog->float_utf8 = Nullsv; /* clear */ - check = Nullsv; /* abort */ + prog->check_substr = prog->check_utf8 = NULL; /* disable */ + prog->float_substr = prog->float_utf8 = NULL; /* clear */ + check = NULL; /* abort */ s = strpos; /* XXXX This is a remnant of the old implementation. It looks wasteful, since now INTUIT can use many @@ -875,8 +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; @@ -964,12 +927,15 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, fail: DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", PL_colors[4], PL_colors[5])); - return Nullch; + return NULL; } /* 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; @@ -990,9 +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; @@ -1006,12 +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; @@ -1024,7 +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; @@ -1040,7 +1006,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 U8 *sm = (U8 *) m; U8 tmpbuf1[UTF8_MAXBYTES_CASE+1]; U8 tmpbuf2[UTF8_MAXBYTES_CASE+1]; - const U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY; + const U32 uniflags = UTF8_ALLOW_DEFAULT; to_utf8_lower((U8*)m, tmpbuf1, &ulen1); to_utf8_upper((U8*)m, tmpbuf2, &ulen2); @@ -1069,7 +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 @@ -1087,7 +1053,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 UV c, f; U8 tmpbuf [UTF8_MAXBYTES+1]; STRLEN len, foldlen; - const U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY; + const U32 uniflags = UTF8_ALLOW_DEFAULT; if (c1 == c2) { /* Upper and lower of 1st char are equal - * probably not a "letter". */ @@ -1098,7 +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]; @@ -1111,7 +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; @@ -1137,7 +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]; @@ -1150,7 +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; @@ -1164,7 +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++; } @@ -1174,7 +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++; } @@ -1188,9 +1154,8 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 if (s == PL_bostr) tmp = '\n'; else { - U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr); - - tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0); + U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); + tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); } tmp = ((OP(c) == BOUND ? isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0); @@ -1201,7 +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; @@ -1214,13 +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: @@ -1231,9 +1196,8 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 if (s == PL_bostr) tmp = '\n'; else { - U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr); - - tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0); + U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); + tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT); } tmp = ((OP(c) == NBOUND ? isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0); @@ -1243,7 +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; } @@ -1256,12 +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: @@ -1269,7 +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; @@ -1282,7 +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; @@ -1298,7 +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; @@ -1311,7 +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; @@ -1327,7 +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; @@ -1340,7 +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; @@ -1356,7 +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; @@ -1369,7 +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; @@ -1385,7 +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; @@ -1398,7 +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; @@ -1414,7 +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; @@ -1427,7 +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; @@ -1443,7 +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; @@ -1456,7 +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; @@ -1472,7 +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; @@ -1485,7 +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; @@ -1501,7 +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; @@ -1514,7 +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; @@ -1530,7 +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; @@ -1543,7 +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; @@ -1559,7 +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; @@ -1572,7 +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; @@ -1588,7 +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; @@ -1601,7 +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; @@ -1633,6 +1597,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * /* data: May be used for some additional optimizations. */ /* nosave: For optimizations. */ { + dVAR; register char *s; register regnode *c; register char *startpos = stringarg; @@ -1640,26 +1605,19 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * I32 dontbother = 0; /* how many characters not to try at end */ I32 end_shift = 0; /* Same for the end. */ /* CC */ I32 scream_pos = -1; /* Internal iterator of scream. */ - char *scream_olds; + 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 *dsv0 = PERL_DEBUG_PAD_ZERO(0); - SV *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); - - PL_regcc = 0; - - cache_re(prog); -#ifdef DEBUGGING - PL_regnarrate = DEBUG_r_TEST; -#endif /* Be paranoid... */ if (prog == NULL || startpos == NULL) { @@ -1667,6 +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, @@ -1687,43 +1655,40 @@ 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 != Nullsv || prog->check_utf8 != Nullsv)) { + if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { re_scream_pos_data d; d.scream_olds = &scream_olds; @@ -1762,7 +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? */ @@ -1777,7 +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) @@ -1795,7 +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; } } @@ -1803,7 +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; } @@ -1824,7 +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); @@ -1836,7 +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++; @@ -1849,9 +1814,9 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * "Did not find anchored character...\n") ); } - else if (prog->anchored_substr != Nullsv - || prog->anchored_utf8 != Nullsv - || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) + else if (prog->anchored_substr != NULL + || prog->anchored_utf8 != NULL + || ((prog->float_substr != NULL || prog->float_utf8 != NULL) && prog->float_max_offset < strend - s)) { SV *must; I32 back_max; @@ -1907,21 +1872,21 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * s = HOPc(s, -back_max); } else { - char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; + char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; last1 = HOPc(s, -back_min); - s = t; + s = t; } 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++; } @@ -1953,7 +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) : @@ -1967,13 +1932,13 @@ 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")); } else { dontbother = 0; - if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) { + if (prog->float_substr != NULL || prog->float_utf8 != NULL) { /* Trim the end. */ char *last; SV* float_real; @@ -2000,7 +1965,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * last = strend - len + 1; else if (!multiline) last = memEQ(strend - len, little, len) - ? strend - len : Nullch; + ? strend - len : NULL; else goto find_last; } else { @@ -2025,7 +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; @@ -2034,7 +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); } @@ -2052,7 +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 */ @@ -2092,7 +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; } @@ -2100,12 +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) { - register I32 i; + dVAR; register I32 *sp; register I32 *ep; CHECKPOINT lastcp; + regexp *prog = reginfo->prog; GET_RE_DEBUG_FLAGS_DECL; #ifdef DEBUGGING @@ -2127,24 +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 */ - sv_magic(PL_reg_sv, (SV*)0, - PERL_MAGIC_regex_global, Nullch, 0); - mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global); +#ifdef PERL_OLD_COPY_ON_WRITE + if (SvIsCOW(sv)) + sv_force_normal_flags(sv, 0); +#endif + 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); @@ -2174,7 +2143,7 @@ S_regtry(pTHX_ regexp *prog, char *startpos) RX_MATCH_COPIED_off(prog); } else - PL_reg_oldsaved = Nullch; + PL_reg_oldsaved = NULL; prog->subbeg = PL_bostr; prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */ } @@ -2214,6 +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; @@ -2221,7 +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; } @@ -2248,6 +2218,7 @@ typedef struct { regnode *next; char *locinput; I32 nextchr; + int minmod; #ifdef DEBUGGING int regindent; #endif @@ -2272,7 +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 (cache_offset | cache_bit) { \ + if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) { \ if (!(PL_reg_poscache[0] & (1<u.whilem.cache_offset] &= ~(1<u.whilem.cache_bit); \ } \ } \ sayYES; \ } STMT_END #define CACHEsayNO STMT_START { \ - if (cache_offset | cache_bit) { \ + if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) { \ if (!(PL_reg_poscache[0] & (1<u.whilem.cache_offset] &= ~(1<u.whilem.cache_bit); \ } \ } \ sayNO; \ @@ -2316,57 +2287,71 @@ typedef union re_unwind_t { /* Make sure there is a test for this +1 options in re_tests */ #define TRIE_INITAL_ACCEPT_BUFFLEN 4; -#define TRIE_CHECK_STATE_IS_ACCEPTING STMT_START { \ - if ( trie->states[ state ].wordnum ) { \ - if ( !accepted ) { \ - ENTER; \ - SAVETMPS; \ - bufflen = TRIE_INITAL_ACCEPT_BUFFLEN ; \ - sv_accept_buff=NEWSV( 1234, \ - bufflen * sizeof(reg_trie_accepted) - 1 ); \ - SvCUR_set( sv_accept_buff, sizeof(reg_trie_accepted) ); \ - SvPOK_on( sv_accept_buff ); \ - sv_2mortal( sv_accept_buff ); \ - accept_buff = (reg_trie_accepted*)SvPV_nolen( sv_accept_buff );\ - } else { \ - if ( accepted >= bufflen ) { \ - bufflen *= 2; \ - 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 ) ); \ - } \ - accept_buff[ accepted ].wordnum = trie->states[ state ].wordnum; \ - accept_buff[ accepted ].endpos = uc; \ - ++accepted; \ - } } STMT_END - -#define TRIE_HANDLE_CHAR STMT_START { \ - if ( uvc < 256 ) { \ - charid = trie->charmap[ uvc ]; \ - } else { \ - charid = 0; \ - if( trie->widecharmap ) { \ - SV** svpp = (SV**)NULL; \ - svpp = hv_fetch( trie->widecharmap, (char*)&uvc, \ - sizeof( UV ), 0 ); \ - if ( svpp ) { \ - charid = (U16)SvIV( *svpp ); \ - } \ - } \ - } \ - if ( charid && \ - ( base + charid > trie->uniquecharcount ) && \ - ( base + charid - 1 - trie->uniquecharcount < trie->lasttrans) && \ - trie->trans[ base + charid - 1 - trie->uniquecharcount ].check == state ) \ - { \ - state = trie->trans[ base + charid - 1 - trie->uniquecharcount ].next; \ - } else { \ - state = 0; \ - } \ - uc += len; \ - } STMT_END +#define SLAB_FIRST(s) (&(s)->states[0]) +#define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1]) + +/* grab a new slab and return the first slot in it */ + +STATIC regmatch_state * +S_push_slab(pTHX) +{ + regmatch_slab *s = PL_regmatch_slab->next; + if (!s) { + Newx(s, 1, regmatch_slab); + s->prev = PL_regmatch_slab; + s->next = NULL; + PL_regmatch_slab->next = s; + } + PL_regmatch_slab = s; + return SLAB_FIRST(s); +} + +/* simulate a recursive call to regmatch */ + +#define REGMATCH(ns, where) \ + st->scan = scan; \ + scan = (ns); \ + st->resume_state = resume_##where; \ + 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 > SLAB_LAST(PL_regmatch_slab)) \ + 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 < SLAB_FIRST(PL_regmatch_slab)) { \ + PL_regmatch_slab = PL_regmatch_slab->prev; \ + st = SLAB_LAST(PL_regmatch_slab); \ + } \ + PL_regmatch_state = st; \ + scan = st->scan; \ + next = st->next; \ + n = st->n; \ + locinput = st->locinput; \ + nextchr = UCHARAT(locinput); /* - regmatch - main matching routine @@ -2381,57 +2366,165 @@ typedef union re_unwind_t { /* [lwall] I've hoisted the register declarations to the outer block in order to * maybe save a little bit of pushing and popping on the stack. It also takes * advantage of machines that use a register save mask on subroutine entry. + * + * This function used to be heavily recursive, but since this had the + * effect of blowing the CPU stack on complex regexes, it has been + * restructured to be iterative, and to save state onto the heap rather + * than the stack. Essentially whereever regmatch() used to be called, it + * pushes the current state, notes where to return, then jumps back into + * the main loop. + * + * Originally the structure of this function used to look something like + + S_regmatch() { + int a = 1, b = 2; + ... + while (scan != NULL) { + a++; // do stuff with a and b + ... + switch (OP(scan)) { + case FOO: { + int local = 3; + ... + if (regmatch(...)) // recurse + goto yes; + } + ... + } + } + yes: + return 1; + } + + * Now it looks something like this: + + typedef struct { + int a, b, local; + int resume_state; + } regmatch_state; + + S_regmatch() { + regmatch_state *st = new(); + int depth=0; + st->a++; // do stuff with a and b + ... + while (scan != NULL) { + ... + switch (OP(scan)) { + case FOO: { + st->local = 3; + ... + st->scan = scan; + scan = ...; + st->resume_state = resume_FOO; + goto start_recurse; // recurse + + resume_point_FOO: + if (result) + goto yes; + } + ... + } + start_recurse: + st = new(); push a new state + st->a = 1; st->b = 2; + depth++; + } + yes: + result = 1; + if (depth--) { + st = pop(); + switch (resume_state) { + case resume_FOO: + goto resume_point_FOO; + ... + } + } + return result + } + + * WARNING: this means that any line in this function that contains a + * REGMATCH() or TRYPAREN() is actually simulating a recursive call to + * regmatch() using gotos instead. Thus the values of any local variables + * not saved in the regmatch_state structure will have been lost when + * execution resumes on the next line . + * + * States (ie the st pointer) are allocated in slabs of about 4K in size. + * PL_regmatch_state always points to the currently active state, and + * PL_regmatch_slab points to the slab currently containing PL_regmatch_state. + * The first time regmatch is called, the first slab is allocated, and is + * never freed until interpreter desctruction. When the slab is full, + * a new one is allocated chained to the end. At exit from regmatch, slabs + * allocated since entry are freed. */ + + STATIC I32 /* 0 failure, 1 success */ -S_regmatch(pTHX_ regnode *prog) +S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog) { dVAR; - register regnode *scan; /* Current node. */ - regnode *next; /* Next node. */ - regnode *inner; /* Next node in internal branch. */ - register I32 nextchr; /* renamed nextchr - nextchar colides with - function of same name */ - register I32 n; /* no or next */ - register I32 ln = 0; /* len or last */ - register char *s = Nullch; /* operand or save */ - register char *locinput = PL_reginput; - register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */ - int minmod = 0, sw = 0, logical = 0; - I32 unwind = 0; - - /* used by the trie code */ - SV *sv_accept_buff = NULL; /* accepting states we have traversed */ - reg_trie_accepted *accept_buff = NULL; /* "" */ - reg_trie_data *trie; /* what trie are we using right now */ - U32 accepted = 0; /* how many accepting states we have seen*/ - -#if 0 - I32 firstcp = PL_savestack_ix; -#endif register const bool do_utf8 = PL_reg_match_utf8; -#ifdef DEBUGGING - SV *dsv0 = PERL_DEBUG_PAD_ZERO(0); - SV *dsv1 = PERL_DEBUG_PAD_ZERO(1); - SV *dsv2 = PERL_DEBUG_PAD_ZERO(2); + const U32 uniflags = UTF8_ALLOW_DEFAULT; - SV *re_debug_flags = NULL; -#endif - U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY; + regexp *rex = reginfo->prog; - GET_RE_DEBUG_FLAGS; + regmatch_slab *orig_slab; + regmatch_state *orig_state; + + /* the current state. This is a cached copy of PL_regmatch_state */ + register regmatch_state *st; + /* cache heavy used fields of st in registers */ + register regnode *scan; + register regnode *next; + register I32 n = 0; /* initialize to shut up compiler warning */ + register char *locinput = PL_reginput; + + /* these variables are NOT saved during a recusive RFEGMATCH: */ + register I32 nextchr; /* is always set to UCHARAT(locinput) */ + 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 *yes_state = NULL; /* state to pop to on success of + subpattern */ + #ifdef DEBUGGING + SV *re_debug_flags = NULL; + GET_RE_DEBUG_FLAGS; PL_regindent++; #endif + /* on first ever call to regmatch, allocate first slab */ + if (!PL_regmatch_slab) { + Newx(PL_regmatch_slab, 1, regmatch_slab); + PL_regmatch_slab->prev = NULL; + PL_regmatch_slab->next = NULL; + PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab); + } + /* remember current high-water mark for exit */ + /* XXX this should be done with SAVE* instead */ + orig_slab = PL_regmatch_slab; + orig_state = PL_regmatch_state; + + /* grab next free state slot */ + st = ++PL_regmatch_state; + if (st > SLAB_LAST(PL_regmatch_slab)) + st = PL_regmatch_state = S_push_slab(aTHX); + + st->minmod = 0; + st->sw = 0; + st->logical = 0; + st->unwind = 0; + st->cc = NULL; /* Note that nextchr is a byte even in UTF */ nextchr = UCHARAT(locinput); scan = prog; while (scan != NULL) { DEBUG_EXECUTE_r( { - SV *prop = sv_newmortal(); + SV * const prop = sv_newmortal(); const int docolor = *PL_colors[0]; const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput); @@ -2457,21 +2550,22 @@ 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 ? - pv_uni_display(dsv0, (U8*)(locinput - pref_len), + pv_uni_display(PERL_DEBUG_PAD(0), (U8*)(locinput - pref_len), pref0_len, 60, UNI_DISPLAY_REGEX) : locinput - pref_len; const int len0 = do_utf8 ? strlen(s0) : pref0_len; const char * const s1 = do_utf8 && OP(scan) != CANY ? - pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len), + pv_uni_display(PERL_DEBUG_PAD(1), + (U8*)(locinput - pref_len + pref0_len), pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) : locinput - pref_len + pref0_len; const int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len; const char * const s2 = do_utf8 && OP(scan) != CANY ? - pv_uni_display(dsv2, (U8*)locinput, + pv_uni_display(PERL_DEBUG_PAD(2), (U8*)locinput, PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) : locinput; const int len2 = do_utf8 ? strlen(s2) : l; @@ -2490,7 +2584,7 @@ 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)); } }); @@ -2503,7 +2597,7 @@ S_regmatch(pTHX_ regnode *prog) case BOL: if (locinput == PL_bostr) { - /* regtill = regbol; */ + /* reginfo->till = reginfo->bol; */ break; } sayNO; @@ -2519,7 +2613,7 @@ S_regmatch(pTHX_ regnode *prog) break; sayNO; case GPOS: - if (locinput == PL_reg_ganch) + if (locinput == reginfo->ganch) break; sayNO; case EOL: @@ -2575,12 +2669,9 @@ S_regmatch(pTHX_ regnode *prog) traverse the TRIE keeping track of all accepting states we transition through until we get to a failing node. - we use two slightly different pieces of code to handle - the traversal depending on whether its case sensitive or - not. we reuse the accept code however. (this should probably - be turned into a macro.) */ + case TRIE: case TRIEF: case TRIEFL: { @@ -2593,13 +2684,48 @@ S_regmatch(pTHX_ regnode *prog) STRLEN foldlen = 0; U8 *uscan = (U8*)NULL; STRLEN bufflen=0; - accepted = 0; - - trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ]; + SV *sv_accept_buff = NULL; + const enum { trie_plain, trie_utf8, trie_uft8_fold } + trie_type = do_utf8 ? + (OP(scan) == TRIE ? trie_utf8 : trie_uft8_fold) + : trie_plain; + + /* what trie are we using right now */ + reg_trie_data *trie + = (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 ) { - TRIE_CHECK_STATE_IS_ACCEPTING; + if (trie->states[ state ].wordnum) { + if (!st->u.trie.accepted ) { + ENTER; + SAVETMPS; + bufflen = TRIE_INITAL_ACCEPT_BUFFLEN; + sv_accept_buff=newSV(bufflen * + sizeof(reg_trie_accepted) - 1); + SvCUR_set(sv_accept_buff, + sizeof(reg_trie_accepted)); + SvPOK_on(sv_accept_buff); + sv_2mortal(sv_accept_buff); + st->u.trie.accept_buff = + (reg_trie_accepted*)SvPV_nolen(sv_accept_buff ); + } + else { + if (st->u.trie.accepted >= bufflen) { + bufflen *= 2; + 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->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; @@ -2607,12 +2733,12 @@ 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)accepted ); + (UV)state, (UV)base, (UV)st->u.trie.accepted ); ); if ( base ) { - - if ( do_utf8 ) { + switch (trie_type) { + case trie_uft8_fold: if ( foldlen>0 ) { uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); foldlen -= len; @@ -2625,14 +2751,47 @@ S_regmatch(pTHX_ regnode *prog) foldlen -= UNISKIP( uvc ); uscan = foldbuf + UNISKIP( uvc ); } - } else { + break; + case trie_utf8: + uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, + &len, uniflags ); + break; + case trie_plain: uvc = (UV)*uc; len = 1; } - TRIE_HANDLE_CHAR; + if (uvc < 256) { + charid = trie->charmap[ uvc ]; + } + else { + charid = 0; + if (trie->widecharmap) { + SV** svpp = (SV**)NULL; + svpp = hv_fetch(trie->widecharmap, + (char*)&uvc, sizeof(UV), 0); + if (svpp) + charid = (U16)SvIV(*svpp); + } + } + + if (charid && + (base + charid > trie->uniquecharcount ) + && (base + charid - 1 - trie->uniquecharcount + < trie->lasttrans) + && trie->trans[base + charid - 1 - + trie->uniquecharcount].check == state) + { + state = trie->trans[base + charid - 1 - + trie->uniquecharcount ].next; + } + else { + state = 0; + } + uc += len; - } else { + } + else { state = 0; } DEBUG_TRIE_EXECUTE_r( @@ -2641,65 +2800,8 @@ S_regmatch(pTHX_ regnode *prog) charid, uvc, (UV)state, PL_colors[5] ); ); } - if ( !accepted ) { - sayNO; - } else { - goto TrieAccept; - } - } - /* unreached codepoint: we jump into the middle of the next case - from previous if blocks */ - case TRIE: - { - U8 *uc = (U8*)locinput; - U32 state = 1; - U16 charid = 0; - U32 base = 0; - UV uvc = 0; - STRLEN len = 0; - STRLEN bufflen = 0; - accepted = 0; - - trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ]; - - while ( state && uc <= (U8*)PL_regeol ) { - - TRIE_CHECK_STATE_IS_ACCEPTING; - - base = trie->states[ state ].trans.base; - - DEBUG_TRIE_EXECUTE_r( - 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)accepted ); - ); - - if ( base ) { - - if ( do_utf8 ) { - uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); - } else { - uvc = (U32)*uc; - len = 1; - } - - TRIE_HANDLE_CHAR; - - } else { - state = 0; - } - DEBUG_TRIE_EXECUTE_r( - PerlIO_printf( Perl_debug_log, - "Charid:%3x CV:%4"UVxf" After State: %4"UVxf"%s\n", - charid, uvc, (UV)state, PL_colors[5] ); - ); - } - if ( !accepted ) { + if (!st->u.trie.accepted ) sayNO; - } - } - /* There was at least one accepting state that we @@ -2713,62 +2815,63 @@ S_regmatch(pTHX_ regnode *prog) eventually terminates once all of the accepting states have been tried. */ - TrieAccept: - { - int gotit = 0; - if ( accepted == 1 ) { + if ( st->u.trie.accepted == 1 ) { DEBUG_EXECUTE_r({ - SV **tmp = av_fetch( trie->words, 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], - 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 *)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; LEAVE; - gotit = regmatch( scan + NEXT_OFF( scan ) ); + REGMATCH(scan + NEXT_OFF(scan), TRIE1); + /*** all unsaved local vars undefined at this point */ } 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)accepted, + REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], (IV)st->u.trie.accepted, PL_colors[5] ); ); - while ( !gotit && accepted-- ) { + while ( !result && st->u.trie.accepted-- ) { U32 best = 0; U32 cur; - for( cur = 1 ; cur <= 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, accept_buff[ best ].wordnum, (IV)cur, - 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 ( accept_buff[ cur ].wordnum < accept_buff[ best ].wordnum ) - best = cur; + if (st->u.trie.accept_buff[cur].wordnum < + st->u.trie.accept_buff[best].wordnum) + best = cur; } DEBUG_EXECUTE_r({ - SV **tmp = av_fetch( trie->words, 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], - 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 ( 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 *)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 @@ -2776,27 +2879,28 @@ S_regmatch(pTHX_ regnode *prog) all until I can be sure. */ SAVETMPS; - gotit = regmatch( scan + NEXT_OFF( scan ) ) ; + REGMATCH(scan + NEXT_OFF(scan), TRIE2); + /*** all unsaved local vars undefined at this point */ FREETMPS; } FREETMPS; LEAVE; } - if ( gotit ) { + if (result) { sayYES; } else { sayNO; } } /* unreached codepoint */ - case EXACT: - s = STRING(scan); - ln = STR_LEN(scan); + case EXACT: { + char *s = STRING(scan); + st->ln = STR_LEN(scan); if (do_utf8 != UTF) { /* The target and the pattern have differing utf8ness. */ char *l = locinput; - const char *e = s + ln; + const char *e = s + st->ln; if (do_utf8) { /* The target is utf8, the pattern is not utf8. */ @@ -2834,26 +2938,27 @@ S_regmatch(pTHX_ regnode *prog) /* Inline the first character, for speed. */ if (UCHARAT(s) != nextchr) sayNO; - if (PL_regeol - locinput < ln) + if (PL_regeol - locinput < st->ln) sayNO; - if (ln > 1 && memNE(s, locinput, ln)) + if (st->ln > 1 && memNE(s, locinput, st->ln)) sayNO; - locinput += ln; + locinput += st->ln; nextchr = UCHARAT(locinput); break; + } case EXACTFL: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ - case EXACTF: - s = STRING(scan); - ln = STR_LEN(scan); + case EXACTF: { + char *s = STRING(scan); + st->ln = STR_LEN(scan); if (do_utf8 || UTF) { /* Either target or the pattern are utf8. */ char *l = locinput; char *e = PL_regeol; - if (ibcmp_utf8(s, 0, ln, (bool)UTF, + if (ibcmp_utf8(s, 0, st->ln, (bool)UTF, l, &e, 0, do_utf8)) { /* One more case for the sharp s: * pack("U0U*", 0xDF) =~ /ss/i, @@ -2861,7 +2966,7 @@ S_regmatch(pTHX_ regnode *prog) * byte sequence for the U+00DF. */ if (!(do_utf8 && toLOWER(s[0]) == 's' && - ln >= 2 && + st->ln >= 2 && toLOWER(s[1]) == 's' && (U8)l[0] == 0xC3 && e - l >= 2 && @@ -2880,20 +2985,21 @@ S_regmatch(pTHX_ regnode *prog) UCHARAT(s) != ((OP(scan) == EXACTF) ? PL_fold : PL_fold_locale)[nextchr]) sayNO; - if (PL_regeol - locinput < ln) + if (PL_regeol - locinput < st->ln) sayNO; - if (ln > 1 && (OP(scan) == EXACTF - ? ibcmp(s, locinput, ln) - : ibcmp_locale(s, locinput, ln))) + if (st->ln > 1 && (OP(scan) == EXACTF + ? ibcmp(s, locinput, st->ln) + : ibcmp_locale(s, locinput, st->ln))) sayNO; - locinput += ln; + locinput += st->ln; nextchr = UCHARAT(locinput); break; + } case ANYOF: 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; @@ -2904,7 +3010,7 @@ 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; @@ -2977,35 +3083,35 @@ S_regmatch(pTHX_ regnode *prog) /* was last char in word? */ if (do_utf8) { if (locinput == PL_bostr) - ln = '\n'; + st->ln = '\n'; else { const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); - ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0); + st->ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); } if (OP(scan) == BOUND || OP(scan) == NBOUND) { - ln = isALNUM_uni(ln); + st->ln = isALNUM_uni(st->ln); LOAD_UTF8_CHARCLASS_ALNUM(); n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8); } else { - ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln)); + st->ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(st->ln)); n = isALNUM_LC_utf8((U8*)locinput); } } else { - ln = (locinput != PL_bostr) ? + st->ln = (locinput != PL_bostr) ? UCHARAT(locinput - 1) : '\n'; if (OP(scan) == BOUND || OP(scan) == NBOUND) { - ln = isALNUM(ln); + st->ln = isALNUM(st->ln); n = isALNUM(nextchr); } else { - ln = isALNUM_LC(ln); + st->ln = isALNUM_LC(st->ln); n = isALNUM_LC(nextchr); } } - if (((!ln) == (!n)) == (OP(scan) == BOUND || + if (((!st->ln) == (!n)) == (OP(scan) == BOUND || OP(scan) == BOUNDL)) sayNO; break; @@ -3131,16 +3237,17 @@ S_regmatch(pTHX_ regnode *prog) PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case REF: - case REFF: + case REFF: { + char *s; n = ARG(scan); /* which paren pair */ - ln = PL_regstartp[n]; + st->ln = PL_regstartp[n]; PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ - if ((I32)*PL_reglastparen < n || ln == -1) + if ((I32)*PL_reglastparen < n || st->ln == -1) sayNO; /* Do not match unless seen CLOSEn. */ - if (ln == PL_regendp[n]) + if (st->ln == PL_regendp[n]) break; - s = PL_bostr + ln; + s = PL_bostr + st->ln; if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */ char *l = locinput; const char *e = PL_bostr + PL_regendp[n]; @@ -3176,18 +3283,19 @@ S_regmatch(pTHX_ regnode *prog) (UCHARAT(s) != ((OP(scan) == REFF ? PL_fold : PL_fold_locale)[nextchr])))) sayNO; - ln = PL_regendp[n] - ln; - if (locinput + ln > PL_regeol) + st->ln = PL_regendp[n] - st->ln; + if (locinput + st->ln > PL_regeol) sayNO; - if (ln > 1 && (OP(scan) == REF - ? memNE(s, locinput, ln) + if (st->ln > 1 && (OP(scan) == REF + ? memNE(s, locinput, st->ln) : (OP(scan) == REFF - ? ibcmp(s, locinput, ln) - : ibcmp_locale(s, locinput, ln)))) + ? ibcmp(s, locinput, st->ln) + : ibcmp_locale(s, locinput, st->ln)))) sayNO; - locinput += ln; + locinput += st->ln; nextchr = UCHARAT(locinput); break; + } case NOTHING: case TAIL: @@ -3196,21 +3304,21 @@ S_regmatch(pTHX_ regnode *prog) break; case EVAL: { - dSP; - OP_4tree *oop = PL_op; - COP *ocurcop = PL_curcop; - PAD *old_comppad; SV *ret; - struct regexp *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; - { - SV **before = SP; + /* 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) @@ -3219,20 +3327,24 @@ S_regmatch(pTHX_ regnode *prog) ret = POPs; PUTBACK; } - } - PL_op = oop; - PAD_RESTORE_LOCAL(old_comppad); - PL_curcop = ocurcop; - if (logical) { - if (logical == 2) { /* Postponed subexpression. */ - regexp *re; - MAGIC *mg = Null(MAGIC*); - re_cc_state state; - CHECKPOINT cp, lastcp; - int toggleutf; - register SV *sv; + 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 */ + MAGIC *mg = NULL; + SV *sv; if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret))) mg = mg_find(sv, PERL_MAGIC_qr); else if (SvSMAGICAL(ret)) { @@ -3248,96 +3360,63 @@ S_regmatch(pTHX_ regnode *prog) } else { STRLEN len; - const char *t = SvPV_const(ret, 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 = PL_regcc; - state.re = PL_reg_re; - - PL_regcc = 0; - - cp = regcppush(0); /* Save *all* the positions. */ - REGCP_SET(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; - - 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; - PL_regcc = 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(cp); - sayYES; - } - ReREFCNT_dec(re); - REGCP_UNWIND(lastcp); - regcppop(); - PL_reg_call_cc = state.prev; - PL_regcc = 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; - - logical = 0; - sayNO; } - sw = SvTRUE(ret); - logical = 0; - } - else { - sv_setsv(save_scalar(PL_replgv), ret); - cache_re(oreg); + + /* 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; + 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; + + /* resume to current state on success */ + st->u.yes.prev_yes_state = yes_state; + yes_state = st; + PUSH_STATE(newst, resume_EVAL); + st = newst; + + /* now continue from first node in postoned RE */ + next = re->program + 1; + break; + /* NOTREACHED */ } + /* /(?(?{...})X|Y)/ */ + st->sw = SvTRUE(ret); + st->logical = 0; break; } case OPEN: @@ -3356,11 +3435,11 @@ S_regmatch(pTHX_ regnode *prog) break; case GROUPP: n = ARG(scan); /* which paren pair */ - sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1); + st->sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1); break; case IFTHEN: PL_reg_leftiter = PL_reg_maxiter; /* Void cache */ - if (sw) + if (st->sw) next = NEXTOPER(NEXTOPER(scan)); else { next = scan + ARG(scan); @@ -3369,38 +3448,39 @@ S_regmatch(pTHX_ regnode *prog) } break; case LOGICAL: - logical = scan->flags; + st->logical = scan->flags; break; /******************************************************************* - PL_regcc 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 X, regnode for CURLYX is processed; + 1) After matching Y, regnode for CURLYX is processed; - 2) This regnode creates infoblock on the stack, and calls - regmatch() recursively with the starting point at WHILEM node; + 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. - 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX + 5) Each time WHILEM is hit, cc is the infoblock created by CURLYX of the same pair. Thus when WHILEM tries to match Z, it temporarily - resets PL_regcc, since this Y(A)*Z can be a part of some other loop: + resets cc, since this Y(A)*Z can be a part of some other loop: 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. - In the following picture infoblocks for outer loop of + In the following picture, infoblocks for outer loop of (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block is denoted by x. The matched string is YAAZYAZT. Temporarily postponed infoblocks are drawn below the "reset" infoblock. @@ -3446,35 +3526,40 @@ S_regmatch(pTHX_ regnode *prog) O I,I *******************************************************************/ + case CURLYX: { - CURCUR cc; - CHECKPOINT cp = PL_savestack_ix; /* No need to save/restore up to this paren */ I32 parenfloor = scan->flags; if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ next += ARG(next); - cc.oldcc = PL_regcc; - PL_regcc = &cc; /* XXXX Probably it is better to teach regpush to support parenfloor > PL_regsize... */ if (parenfloor > (I32)*PL_reglastparen) parenfloor = *PL_reglastparen; /* Pessimization... */ - cc.parenfloor = parenfloor; - cc.cur = -1; - cc.min = ARG1(scan); - cc.max = ARG2(scan); - cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS; - cc.next = next; - cc.minmod = minmod; - 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; - n = regmatch(PREVOPER(next)); /* start on the WHILEM */ - regcpblow(cp); - PL_regcc = cc.oldcc; - saySAME(n); + REGMATCH(PREVOPER(next), CURLYX); /* start on the WHILEM */ + /*** all unsaved local vars undefined at this point */ + regcpblow(st->u.curlyx.cp); + st->cc = st->u.curlyx.outercc; + saySAME(result); } - /* NOT REACHED */ + /* NOTREACHED */ case WHILEM: { /* * This is really hard to understand, because after we match @@ -3485,50 +3570,54 @@ S_regmatch(pTHX_ regnode *prog) * that we can try again after backing off. */ - CHECKPOINT cp, lastcp; - CURCUR* cc = PL_regcc; - char *lastloc = cc->lastloc; /* Detection of 0-len. */ - I32 cache_offset = 0, 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 = 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)cc->min, - (long)cc->max, PTR2UV(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 == cc->lastloc && n >= cc->min) { - PL_regcc = cc->oldcc; - if (PL_regcc) - ln = PL_regcc->cur; + 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->u.curlyx.cur; DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, "%*s empty match detected, try continuation...\n", REPORT_CODE_OFF+PL_regindent*2, "") ); - if (regmatch(cc->next)) + REGMATCH(st->u.whilem.savecc->next, WHILEM1); + /*** all unsaved local vars undefined at this point */ + st->cc = st->u.whilem.savecc; + if (result) sayYES; - if (PL_regcc) - PL_regcc->cur = ln; - PL_regcc = cc; + 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 < cc->min) { - cc->cur = n; - cc->lastloc = locinput; - if (regmatch(cc->scan)) + 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; - cc->cur = n - 1; - cc->lastloc = lastloc; + st->cc->u.curlyx.cur = n - 1; + st->cc->u.curlyx.lastloc = st->u.whilem.lastloc; sayNO; } @@ -3560,13 +3649,13 @@ S_regmatch(pTHX_ regnode *prog) ); } if (PL_reg_leftiter < 0) { - cache_offset = locinput - PL_bostr; + st->u.whilem.cache_offset = locinput - PL_bostr; - cache_offset = (scan->flags & 0xf) - 1 + POSCACHE_START - + cache_offset * (scan->flags>>4); - cache_bit = cache_offset % 8; - cache_offset /= 8; - if (PL_reg_poscache[cache_offset] & (1<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", @@ -3579,29 +3668,32 @@ S_regmatch(pTHX_ regnode *prog) /* cache records failure */ sayNO_SILENT; } - PL_reg_poscache[cache_offset] |= (1<u.whilem.cache_offset] |= (1<u.whilem.cache_bit); } } /* Prefer next over scan for minimal matching. */ - if (cc->minmod) { - PL_regcc = cc->oldcc; - if (PL_regcc) - ln = PL_regcc->cur; - cp = regcppush(cc->parenfloor); - REGCP_SET(lastcp); - if (regmatch(cc->next)) { - regcpblow(cp); + if (st->cc->minmod) { + st->u.whilem.savecc = st->cc; + st->cc = st->cc->u.curlyx.outercc; + if (st->cc) + 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->u.whilem.savecc; + if (result) { + regcpblow(st->u.whilem.cp); CACHEsayYES; /* All done. */ } - REGCP_UNWIND(lastcp); - regcppop(); - if (PL_regcc) - PL_regcc->cur = ln; - PL_regcc = cc; + 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 >= 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; @@ -3619,34 +3711,38 @@ S_regmatch(pTHX_ regnode *prog) ); /* Try scanning more and see if it helps. */ PL_reginput = locinput; - cc->cur = n; - cc->lastloc = locinput; - cp = regcppush(cc->parenfloor); - REGCP_SET(lastcp); - if (regmatch(cc->scan)) { - regcpblow(cp); + 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->u.whilem.cp); CACHEsayYES; } - REGCP_UNWIND(lastcp); - regcppop(); - cc->cur = n - 1; - cc->lastloc = 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 < cc->max) { /* More greed allowed? */ - cp = regcppush(cc->parenfloor); - cc->cur = n; - cc->lastloc = locinput; - REGCP_SET(lastcp); - if (regmatch(cc->scan)) { - regcpblow(cp); + 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->u.whilem.cp); CACHEsayYES; } - REGCP_UNWIND(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, @@ -3663,19 +3759,22 @@ S_regmatch(pTHX_ regnode *prog) } /* Failed deeper matches of scan, so see if this one works. */ - PL_regcc = cc->oldcc; - if (PL_regcc) - ln = PL_regcc->cur; - if (regmatch(cc->next)) + st->u.whilem.savecc = st->cc; + st->cc = st->cc->u.curlyx.outercc; + if (st->cc) + 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->u.whilem.savecc; + if (result) CACHEsayYES; - if (PL_regcc) - PL_regcc->cur = ln; - PL_regcc = cc; - cc->cur = n - 1; - cc->lastloc = 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; } - /* NOT REACHED */ + /* NOTREACHED */ case BRANCHJ: next = scan + ARG(scan); if (next == scan) @@ -3686,26 +3785,26 @@ S_regmatch(pTHX_ regnode *prog) inner = NEXTOPER(scan); do_branch: { - c1 = OP(scan); - if (OP(next) != c1) /* No choice. */ + I32 type; + type = OP(scan); + if (!next || OP(next) != type) /* No choice. */ next = inner; /* Avoid recursion. */ else { const I32 lastparen = *PL_reglastparen; - I32 unwind1; - re_unwind_branch_t *uw; - /* Put unwinding data on stack */ - unwind1 = SSNEWt(1,re_unwind_branch_t); - uw = SSPTRt(unwind1,re_unwind_branch_t); - uw->prev = unwind; - unwind = unwind1; - uw->type = ((c1 == BRANCH) + const I32 unwind1 = SSNEWt(1,re_unwind_branch_t); + re_unwind_branch_t * const uw = SSPTRt(unwind1,re_unwind_branch_t); + + uw->prev = st->unwind; + st->unwind = unwind1; + uw->type = ((type == BRANCH) ? RE_UNWIND_BRANCH : RE_UNWIND_BRANCHJ); uw->lastparen = lastparen; uw->next = next; uw->locinput = locinput; uw->nextchr = nextchr; + uw->minmod = st->minmod; #ifdef DEBUGGING uw->regindent = ++PL_regindent; #endif @@ -3718,81 +3817,113 @@ S_regmatch(pTHX_ regnode *prog) } break; case MINMOD: - minmod = 1; + st->minmod = 1; break; case CURLYM: { - I32 l = 0; - CHECKPOINT lastcp; + 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. */ - ln = ARG1(scan); /* min to match */ + st->ln = ARG1(scan); /* min to match */ n = ARG2(scan); /* max to match */ - paren = scan->flags; - if (paren) { - if (paren > PL_regsize) - PL_regsize = paren; - if (paren > (I32)*PL_reglastparen) - *PL_reglastparen = 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 (paren) + if (st->u.curlym.paren) scan += NEXT_OFF(scan); /* Skip former OPEN. */ PL_reginput = locinput; - if (minmod) { - minmod = 0; - if (ln && regrepeat_hard(scan, ln, &l) < ln) + 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, u.curlym.l */ + if (!st->u.curlym.matches++) { + if (PL_reg_match_utf8) { + char *s = locinput; + while (s < PL_reginput) { + st->u.curlym.l++; + s += UTF8SKIP(s); + } + } + else { + st->u.curlym.l = PL_reginput - locinput; + } + if (st->u.curlym.l == 0) { + st->u.curlym.matches = st->u.curlym.maxwanted; + break; + } + } + locinput = PL_reginput; + } + } + + PL_reginput = locinput; + + if (st->minmod) { + st->minmod = 0; + if (st->ln && st->u.curlym.matches < st->ln) sayNO; - locinput = PL_reginput; 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)) c1 = 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) { - c1 = c2 = -1000; + st->u.curlym.c1 = st->u.curlym.c2 = -1000; goto assume_ok_MM; } - else { c1 = (U8)*STRING(text_node); } + else { st->u.curlym.c1 = (U8)*STRING(text_node); } if (OP(text_node) == EXACTF || OP(text_node) == REFF) - c2 = PL_fold[c1]; + st->u.curlym.c2 = PL_fold[st->u.curlym.c1]; else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL) - c2 = PL_fold_locale[c1]; + st->u.curlym.c2 = PL_fold_locale[st->u.curlym.c1]; else - c2 = c1; + st->u.curlym.c2 = st->u.curlym.c1; } } else - c1 = c2 = -1000; + st->u.curlym.c1 = st->u.curlym.c2 = -1000; assume_ok_MM: - REGCP_SET(lastcp); - while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */ + 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 (c1 == -1000 || - UCHARAT(PL_reginput) == c1 || - UCHARAT(PL_reginput) == c2) + if (st->u.curlym.c1 == -1000 || + UCHARAT(PL_reginput) == st->u.curlym.c1 || + UCHARAT(PL_reginput) == st->u.curlym.c2) { - if (paren) { - if (ln) { - PL_regstartp[paren] = - HOPc(PL_reginput, -l) - PL_bostr; - PL_regendp[paren] = PL_reginput - PL_bostr; + if (st->u.curlym.paren) { + if (st->ln) { + 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[paren] = -1; + PL_regendp[st->u.curlym.paren] = -1; } - if (regmatch(next)) + REGMATCH(next, CURLYM2); + /*** all unsaved local vars undefined at this point */ + if (result) sayYES; - REGCP_UNWIND(lastcp); + REGCP_UNWIND(st->u.curlym.lastcp); } /* Couldn't or didn't -- move forward. */ PL_reginput = locinput; - if (regrepeat_hard(scan, 1, &l)) { - ln++; + REGMATCH(scan, CURLYM3); + /*** all unsaved local vars undefined at this point */ + if (result) { + st->ln++; locinput = PL_reginput; } else @@ -3800,100 +3931,103 @@ S_regmatch(pTHX_ regnode *prog) } } else { - n = regrepeat_hard(scan, n, &l); - locinput = PL_reginput; DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, - "%*s matched %"IVdf" times, len=%"IVdf"...\n", - (int)(REPORT_CODE_OFF+PL_regindent*2), "", - (IV) n, (IV)l) + "%*s matched %"IVdf" times, len=%"IVdf"...\n", + (int)(REPORT_CODE_OFF+PL_regindent*2), "", + (IV) st->u.curlym.matches, (IV)st->u.curlym.l) ); - if (n >= 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)) c1 = 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) { - c1 = c2 = -1000; + st->u.curlym.c1 = st->u.curlym.c2 = -1000; goto assume_ok_REG; } - else { c1 = (U8)*STRING(text_node); } + else { st->u.curlym.c1 = (U8)*STRING(text_node); } if (OP(text_node) == EXACTF || OP(text_node) == REFF) - c2 = PL_fold[c1]; + st->u.curlym.c2 = PL_fold[st->u.curlym.c1]; else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL) - c2 = PL_fold_locale[c1]; + st->u.curlym.c2 = PL_fold_locale[st->u.curlym.c1]; else - c2 = c1; + st->u.curlym.c2 = st->u.curlym.c1; } } else - c1 = c2 = -1000; + st->u.curlym.c1 = st->u.curlym.c2 = -1000; } assume_ok_REG: - REGCP_SET(lastcp); - while (n >= ln) { + REGCP_SET(st->u.curlym.lastcp); + while (st->u.curlym.matches >= st->ln) { /* If it could work, try it. */ - if (c1 == -1000 || - UCHARAT(PL_reginput) == c1 || - UCHARAT(PL_reginput) == 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 n=%"IVdf"...\n", - (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n) + PerlIO_printf(Perl_debug_log, + "%*s trying tail with matches=%"IVdf"...\n", + (int)(REPORT_CODE_OFF+PL_regindent*2), + "", (IV)st->u.curlym.matches) ); - if (paren) { - if (n) { - PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr; - PL_regendp[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[paren] = -1; + PL_regendp[st->u.curlym.paren] = -1; } - if (regmatch(next)) + REGMATCH(next, CURLYM4); + /*** all unsaved local vars undefined at this point */ + if (result) sayYES; - REGCP_UNWIND(lastcp); + REGCP_UNWIND(st->u.curlym.lastcp); } /* Couldn't or didn't -- back up. */ - n--; - locinput = HOPc(locinput, -l); + st->u.curlym.matches--; + locinput = HOPc(locinput, -st->u.curlym.l); PL_reginput = locinput; } } sayNO; + /* NOTREACHED */ break; } case CURLYN: - paren = scan->flags; /* Which paren to set */ - if (paren > PL_regsize) - PL_regsize = paren; - if (paren > (I32)*PL_reglastparen) - *PL_reglastparen = paren; - ln = ARG1(scan); /* min to match */ + 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: - paren = 0; - ln = ARG1(scan); /* min to match */ + st->u.plus.paren = 0; + st->ln = ARG1(scan); /* min to match */ n = ARG2(scan); /* max to match */ scan = NEXTOPER(scan) + NODE_STEP_REGNODE; goto repeat; case STAR: - ln = 0; + st->ln = 0; n = REG_INFTY; scan = NEXTOPER(scan); - paren = 0; + st->u.plus.paren = 0; goto repeat; case PLUS: - ln = 1; + st->ln = 1; n = REG_INFTY; scan = NEXTOPER(scan); - paren = 0; + st->u.plus.paren = 0; repeat: /* * Lookahead to avoid useless match attempts @@ -3912,20 +4046,20 @@ S_regmatch(pTHX_ regnode *prog) if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node); - if (! HAS_TEXT(text_node)) c1 = 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) { - c1 = c2 = -1000; + st->u.plus.c1 = st->u.plus.c2 = -1000; goto assume_ok_easy; } else { s = (U8*)STRING(text_node); } if (!UTF) { - c2 = c1 = *s; + st->u.plus.c2 = st->u.plus.c1 = *s; if (OP(text_node) == EXACTF || OP(text_node) == REFF) - c2 = PL_fold[c1]; + st->u.plus.c2 = PL_fold[st->u.plus.c1]; else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL) - c2 = PL_fold_locale[c1]; + st->u.plus.c2 = PL_fold_locale[st->u.plus.c1]; } else { /* UTF */ if (OP(text_node) == EXACTF || OP(text_node) == REFF) { @@ -3936,118 +4070,117 @@ S_regmatch(pTHX_ regnode *prog) to_utf8_lower((U8*)s, tmpbuf1, &ulen1); to_utf8_upper((U8*)s, tmpbuf2, &ulen2); - c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0, + st->u.plus.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0, uniflags); - c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0, + st->u.plus.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0, uniflags); } else { - c2 = 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 - c1 = c2 = -1000; + st->u.plus.c1 = st->u.plus.c2 = -1000; assume_ok_easy: PL_reginput = locinput; - if (minmod) { - CHECKPOINT lastcp; - minmod = 0; - if (ln && regrepeat(scan, ln) < ln) + if (st->minmod) { + st->minmod = 0; + if (st->ln && regrepeat(rex, scan, st->ln) < st->ln) sayNO; locinput = PL_reginput; - REGCP_SET(lastcp); - if (c1 != -1000) { - char *e; /* Should not check after this */ - char *old = locinput; - int 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) { - e = PL_regeol - 1; + st->u.plus.e = PL_regeol - 1; if (do_utf8) - while (UTF8_IS_CONTINUATION(*(U8*)e)) - e--; + while (UTF8_IS_CONTINUATION(*(U8*)st->u.plus.e)) + st->u.plus.e--; } else if (do_utf8) { - int m = n - ln; - for (e = locinput; - m >0 && e + UTF8SKIP(e) <= PL_regeol; m--) - e += UTF8SKIP(e); + int m = n - st->ln; + 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 { - e = locinput + n - ln; - if (e >= PL_regeol) - 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 (c1 == c2) { - while (locinput <= e && - UCHARAT(locinput) != 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 <= e - && UCHARAT(locinput) != c1 - && UCHARAT(locinput) != c2) + while (locinput <= st->u.plus.e + && UCHARAT(locinput) != st->u.plus.c1 + && UCHARAT(locinput) != st->u.plus.c2) locinput++; } - count = locinput - old; + st->u.plus.count = locinput - st->u.plus.old; } else { - if (c1 == c2) { + if (st->u.plus.c1 == st->u.plus.c2) { STRLEN len; /* count initialised to * utf8_distance(old, locinput) */ - while (locinput <= e && + while (locinput <= st->u.plus.e && utf8n_to_uvchr((U8*)locinput, UTF8_MAXBYTES, &len, - uniflags) != (UV)c1) { + uniflags) != (UV)st->u.plus.c1) { locinput += len; - count++; + st->u.plus.count++; } } else { STRLEN len; /* count initialised to * utf8_distance(old, locinput) */ - while (locinput <= e) { + while (locinput <= st->u.plus.e) { UV c = utf8n_to_uvchr((U8*)locinput, UTF8_MAXBYTES, &len, uniflags); - if (c == (UV)c1 || c == (UV)c2) + if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2) break; locinput += len; - count++; + st->u.plus.count++; } } } - if (locinput > e) + if (locinput > st->u.plus.e) sayNO; /* PL_reginput == old now */ - if (locinput != old) { - ln = 1; /* Did some */ - if (regrepeat(scan, count) < count) + if (locinput != st->u.plus.old) { + st->ln = 1; /* Did some */ + if (regrepeat(rex, scan, st->u.plus.count) < st->u.plus.count) sayNO; } /* PL_reginput == locinput now */ - TRYPAREN(paren, ln, locinput); + 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(lastcp); + REGCP_UNWIND(st->u.plus.lastcp); /* Couldn't or didn't -- move forward. */ - old = locinput; + st->u.plus.old = locinput; if (do_utf8) locinput += UTF8SKIP(locinput); else locinput++; - count = 1; + st->u.plus.count = 1; } } else - while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */ + while (n >= st->ln || (n == REG_INFTY && st->ln > 0)) { /* ln overflow ? */ UV c; - if (c1 != -1000) { + if (st->u.plus.c1 != -1000) { if (do_utf8) c = utf8n_to_uvchr((U8*)PL_reginput, UTF8_MAXBYTES, 0, @@ -4055,22 +4188,24 @@ S_regmatch(pTHX_ regnode *prog) else c = UCHARAT(PL_reginput); /* If it could work, try it. */ - if (c == (UV)c1 || c == (UV)c2) + if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2) { - TRYPAREN(paren, ln, PL_reginput); - REGCP_UNWIND(lastcp); + TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS2); + /*** all unsaved local vars undefined at this point */ + REGCP_UNWIND(st->u.plus.lastcp); } } /* If it could work, try it. */ - else if (c1 == -1000) + else if (st->u.plus.c1 == -1000) { - TRYPAREN(paren, ln, PL_reginput); - REGCP_UNWIND(lastcp); + TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS3); + /*** all unsaved local vars undefined at this point */ + REGCP_UNWIND(st->u.plus.lastcp); } /* Couldn't or didn't -- move forward. */ PL_reginput = locinput; - if (regrepeat(scan, 1)) { - ln++; + if (regrepeat(rex, scan, 1)) { + st->ln++; locinput = PL_reginput; } else @@ -4078,47 +4213,24 @@ S_regmatch(pTHX_ regnode *prog) } } else { - CHECKPOINT lastcp; - n = regrepeat(scan, n); + n = regrepeat(rex, scan, n); locinput = PL_reginput; - if (ln < n && PL_regkind[(U8)OP(next)] == EOL && + if (st->ln < n && PL_regkind[(U8)OP(next)] == EOL && (OP(next) != MEOL || OP(next) == SEOL || OP(next) == EOS)) { - ln = n; /* why back off? */ + st->ln = n; /* why back off? */ /* ...because $ and \Z can match before *and* after newline at the end. Consider "\n\n" =~ /\n+\Z\n/. We should back off by one in this case. */ if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS) - ln--; + st->ln--; } - REGCP_SET(lastcp); - if (paren) { - UV c = 0; - while (n >= ln) { - if (c1 != -1000) { - if (do_utf8) - c = utf8n_to_uvchr((U8*)PL_reginput, - UTF8_MAXBYTES, 0, - uniflags); - else - c = UCHARAT(PL_reginput); - } - /* If it could work, try it. */ - if (c1 == -1000 || c == (UV)c1 || c == (UV)c2) - { - TRYPAREN(paren, n, PL_reginput); - REGCP_UNWIND(lastcp); - } - /* Couldn't or didn't -- back up. */ - n--; - PL_reginput = locinput = HOPc(locinput, -1); - } - } - else { + REGCP_SET(st->u.plus.lastcp); + { UV c = 0; - while (n >= ln) { - if (c1 != -1000) { + while (n >= st->ln) { + if (st->u.plus.c1 != -1000) { if (do_utf8) c = utf8n_to_uvchr((U8*)PL_reginput, UTF8_MAXBYTES, 0, @@ -4127,10 +4239,11 @@ S_regmatch(pTHX_ regnode *prog) c = UCHARAT(PL_reginput); } /* If it could work, try it. */ - if (c1 == -1000 || c == (UV)c1 || c == (UV)c2) + if (st->u.plus.c1 == -1000 || c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2) { - TRYPAREN(paren, n, PL_reginput); - REGCP_UNWIND(lastcp); + TRYPAREN(st->u.plus.paren, n, PL_reginput, PLUS4); + /*** all unsaved local vars undefined at this point */ + REGCP_UNWIND(st->u.plus.lastcp); } /* Couldn't or didn't -- back up. */ n--; @@ -4141,46 +4254,12 @@ S_regmatch(pTHX_ regnode *prog) sayNO; break; case END: - if (PL_reg_call_cc) { - re_cc_state *cur_call_cc = PL_reg_call_cc; - CURCUR *cctmp = PL_regcc; - regexp *re = PL_reg_re; - CHECKPOINT cp, lastcp; - - cp = regcppush(0); /* Save *all* the positions. */ - REGCP_SET(lastcp); - regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of - the caller. */ - PL_reginput = locinput; /* Make position available to - the callcc. */ - cache_re(PL_reg_call_cc->re); - PL_regcc = PL_reg_call_cc->cc; - PL_reg_call_cc = PL_reg_call_cc->prev; - if (regmatch(cur_call_cc->node)) { - PL_reg_call_cc = cur_call_cc; - regcpblow(cp); - sayYES; - } - REGCP_UNWIND(lastcp); - regcppop(); - PL_reg_call_cc = cur_call_cc; - PL_regcc = cctmp; - PL_reg_re = re; - cache_re(re); - - 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. */ } @@ -4196,7 +4275,7 @@ S_regmatch(pTHX_ regnode *prog) case UNLESSM: n = 0; if (scan->flags) { - s = HOPBACKc(locinput, scan->flags); + char * const s = HOPBACKc(locinput, scan->flags); if (!s) goto say_yes; PL_reginput = s; @@ -4207,7 +4286,7 @@ S_regmatch(pTHX_ regnode *prog) case IFMATCH: n = 1; if (scan->flags) { - s = HOPBACKc(locinput, scan->flags); + char * const s = HOPBACKc(locinput, scan->flags); if (!s) goto say_no; PL_reginput = s; @@ -4216,21 +4295,22 @@ S_regmatch(pTHX_ regnode *prog) PL_reginput = locinput; do_ifmatch: - inner = NEXTOPER(NEXTOPER(scan)); - if (regmatch(inner) != n) { + REGMATCH(NEXTOPER(NEXTOPER(scan)), IFMATCH); + /*** all unsaved local vars undefined at this point */ + if (result != n) { say_no: - if (logical) { - logical = 0; - sw = 0; + if (st->logical) { + st->logical = 0; + st->sw = 0; goto do_longjump; } else sayNO; } say_yes: - if (logical) { - logical = 0; - sw = 1; + if (st->logical) { + st->logical = 0; + st->sw = 1; } if (OP(scan) == SUSPEND) { locinput = PL_reginput; @@ -4248,10 +4328,49 @@ S_regmatch(pTHX_ regnode *prog) PTR2UV(scan), OP(scan)); Perl_croak(aTHX_ "regexp memory corruption"); } + reenter: scan = next; + continue; + /* NOTREACHED */ + + /* simulate recursively calling regmatch(), but without actually + * recursing - ie save the current state on the heap rather than on + * the stack, then re-enter the loop. This avoids complex regexes + * blowing the processor stack */ + + start_recurse: + { + /* push new state */ + regmatch_state *oldst = st; + + depth++; + + /* grab the next free state slot */ + st++; + if (st > SLAB_LAST(PL_regmatch_slab)) + st = S_push_slab(aTHX); + PL_regmatch_state = st; + + oldst->next = next; + oldst->n = n; + oldst->locinput = locinput; + + st->cc = oldst->cc; + locinput = PL_reginput; + nextchr = UCHARAT(locinput); + st->minmod = 0; + st->sw = 0; + st->logical = 0; + st->unwind = 0; +#ifdef DEBUGGING + PL_regindent++; +#endif + } } + + /* * We get here only if there's trouble -- normally "case END" is * the terminating point. @@ -4268,6 +4387,52 @@ yes_loud: ); goto yes; yes_final: + + if (yes_state) { + /* we have successfully completed a subexpression, but we must now + * pop to the state marked by yes_state and continue from there */ + + assert(st != yes_state); + while (yes_state < SLAB_FIRST(PL_regmatch_slab) + || yes_state > SLAB_LAST(PL_regmatch_slab)) + { + /* not in this slab, pop slab */ + depth -= (st - SLAB_FIRST(PL_regmatch_slab) + 1); + PL_regmatch_slab = PL_regmatch_slab->prev; + st = SLAB_LAST(PL_regmatch_slab); + } + depth -= (st - yes_state); + DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "POP STATE TO (%d)\n", depth)); \ + st = yes_state; + yes_state = st->u.yes.prev_yes_state; + PL_regmatch_state = st; + + switch (st->resume_state) { + case resume_EVAL: + 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 = st->u.eval.lastcp; + regcppop(rex); + PL_savestack_ix = tmp; + } + PL_reginput = locinput; + /* continue at the node following the (??{...}) */ + next = st->next; + goto reenter; + + default: + Perl_croak(aTHX_ "unexpected yes reume state"); + } + } + DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n", PL_colors[4], PL_colors[5])); yes: @@ -4275,11 +4440,58 @@ yes: PL_regindent--; #endif -#if 0 /* Breaks $^R */ - if (unwind) - regcpblow(firstcp); -#endif - return 1; + result = 1; + /* XXX this is duplicate(ish) code to that in the do_no section. + * eventually a yes should just pop the stack back to the current + * yes_state */ + 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_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; + + case resume_EVAL: + default: + Perl_croak(aTHX_ "regexp resume memory corruption"); + } + } + goto final_exit; no: DEBUG_EXECUTE_r( @@ -4290,14 +4502,14 @@ no: goto do_no; no_final: do_no: - if (unwind) { - re_unwind_t *uw = SSPTRt(unwind,re_unwind_t); + if (st->unwind) { + re_unwind_t * const uw = SSPTRt(st->unwind,re_unwind_t); switch (uw->type) { case RE_UNWIND_BRANCH: case RE_UNWIND_BRANCHJ: { - re_unwind_branch_t *uwb = &(uw->branch); + re_unwind_branch_t * const uwb = &(uw->branch); const I32 lastparen = uwb->lastparen; REGCP_UNWIND(uwb->lastcp); @@ -4305,10 +4517,11 @@ 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 */ - unwind = uwb->prev; + st->unwind = uwb->prev; #ifdef DEBUGGING PL_regindent--; #endif @@ -4332,16 +4545,99 @@ do_no: goto reenter; } - /* NOT REACHED */ + /* NOTREACHED */ default: Perl_croak(aTHX_ "regexp unwind memory corruption"); } - /* NOT REACHED */ + /* NOTREACHED */ } + #ifdef DEBUGGING PL_regindent--; #endif - return 0; + result = 0; + + 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; + yes_state = st->u.yes.prev_yes_state; + + /* 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: + 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"); + } + } + +final_exit: + + /* restore original high-water mark */ + PL_regmatch_slab = orig_slab; + PL_regmatch_state = orig_state; + + /* free all slabs above current one */ + if (orig_slab->next) { + regmatch_slab *osl, *sl = orig_slab->next; + orig_slab->next = NULL; + while (sl) { + osl = sl; + sl = sl->next; + Safefree(osl); + } + } + + return result; + } /* @@ -4353,7 +4649,7 @@ do_no: * 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; @@ -4366,7 +4662,7 @@ 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) { @@ -4416,12 +4712,12 @@ 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; @@ -4579,10 +4875,10 @@ S_regrepeat(pTHX_ const regnode *p, I32 max) DEBUG_r({ SV *re_debug_flags = NULL; - SV *prop = sv_newmortal(); + 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); @@ -4592,74 +4888,25 @@ S_regrepeat(pTHX_ const regnode *p, I32 max) return(c); } -/* - - regrepeat_hard - repeatedly match something, report total lenth and length - * - * The repeater is supposed to have constant non-zero length. - */ - -STATIC I32 -S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp) -{ - register char *scan = Nullch; - register char *start; - register char *loceol = PL_regeol; - I32 l = 0; - I32 count = 0, res = 1; - - if (!max) - return 0; - - start = PL_reginput; - if (PL_reg_match_utf8) { - while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) { - if (!count++) { - l = 0; - while (start < PL_reginput) { - l++; - start += UTF8SKIP(start); - } - *lp = l; - if (l == 0) - return max; - } - if (count == max) - return count; - } - } - else { - while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) { - if (!count++) { - *lp = l = PL_reginput - start; - if (max != REG_INFTY && l*max < loceol - scan) - loceol = scan + l*max; - if (l == 0) - return max; - } - } - } - if (!res) - PL_reginput = scan; - - return count; -} /* - regclass_swash - prepare the utf8 swash */ 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; @@ -4701,7 +4948,7 @@ 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); @@ -4712,8 +4959,8 @@ S_reginclass(pTHX_ register const regnode *n, register const U8* p, STRLEN* lenp if (do_utf8 && !UTF8_IS_INVARIANT(c)) { c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len, - ckWARN(WARN_UTF8) ? UTF8_CHECK_ONLY : - UTF8_ALLOW_ANYUV|UTF8_CHECK_ONLY); + (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) | UTF8_CHECK_ONLY); + /* see [perl #37836] for UTF8_ALLOW_ANYUV */ if (len == (STRLEN)-1) Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)"); } @@ -4730,7 +4977,7 @@ 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)) @@ -4824,14 +5071,9 @@ S_reginclass(pTHX_ register const regnode *n, register const U8* p, STRLEN* lenp } STATIC U8 * -S_reghop(pTHX_ U8 *s, I32 off) -{ - return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)); -} - -STATIC U8 * -S_reghop3(pTHX_ U8 *s, I32 off, U8* lim) +S_reghop3(U8 *s, I32 off, U8* lim) { + dVAR; if (off >= 0) { while (off-- && s < lim) { /* XXX could check well-formedness here */ @@ -4854,14 +5096,9 @@ S_reghop3(pTHX_ U8 *s, I32 off, U8* lim) } STATIC U8 * -S_reghopmaybe(pTHX_ U8 *s, I32 off) -{ - return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)); -} - -STATIC U8 * -S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim) +S_reghopmaybe3(U8* s, I32 off, U8* lim) { + dVAR; if (off >= 0) { while (off-- && s < lim) { /* XXX could check well-formedness here */ @@ -4892,15 +5129,16 @@ S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim) static void restore_pos(pTHX_ void *arg) { - PERL_UNUSED_ARG(arg); + dVAR; + 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; @@ -4912,8 +5150,8 @@ 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); @@ -4921,8 +5159,8 @@ 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); @@ -4934,9 +5172,10 @@ S_to_utf8_substr(pTHX_ register regexp *prog) STATIC void 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); @@ -4948,8 +5187,8 @@ 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);