X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=regexec.c;h=747b61be3bfb6383d24d20a75045700907255474;hb=ce5e94717f361c3fd6b9b0fb704412d30f3ccf66;hp=f5fdd8d64118807c987cef259baf8ae7afd4dff9;hpb=7087a21c096179886f18cec5311cc7e897850eef;p=p5sagit%2Fp5-mst-13.2.git diff --git a/regexec.c b/regexec.c index f5fdd8d..747b61b 100644 --- a/regexec.c +++ b/regexec.c @@ -31,28 +31,7 @@ */ #ifdef PERL_EXT_RE_BUILD -/* need to replace pregcomp et al, so enable that */ -# ifndef PERL_IN_XSUB_RE -# define PERL_IN_XSUB_RE -# endif -/* need access to debugger hooks */ -# if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING) -# define DEBUGGING -# endif -#endif - -#ifdef PERL_IN_XSUB_RE -/* We *really* need to overwrite these symbols: */ -# define Perl_regexec_flags my_regexec -# define Perl_regdump my_regdump -# define Perl_regprop my_regprop -# define Perl_re_intuit_start my_re_intuit_start -/* *These* symbols are masked to allow static link. */ -# define Perl_pregexec my_pregexec -# define Perl_reginitcolors my_reginitcolors -# define Perl_regclass_swash my_regclass_swash - -# define PERL_NO_GET_CONTEXT +#include "re_top.h" #endif /* @@ -91,7 +70,11 @@ #define PERL_IN_REGEXEC_C #include "perl.h" -#include "regcomp.h" +#ifdef PERL_IN_XSUB_RE +# include "re_comp.h" +#else +# include "regcomp.h" +#endif #define RF_tainted 1 /* tainted information used? */ #define RF_warned 2 /* warned about big count? */ @@ -107,7 +90,7 @@ #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 +99,20 @@ #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 @@ -221,11 +196,10 @@ 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; @@ -243,7 +217,7 @@ 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; @@ -259,10 +233,10 @@ 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 @@ -276,24 +250,15 @@ 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; } -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, where) { \ @@ -317,6 +282,7 @@ typedef struct re_cc_state * pregexec and friends */ +#ifndef PERL_IN_XSUB_RE /* - pregexec - match a regexp against a string */ @@ -332,19 +298,7 @@ Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *stren regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, nosave ? 0 : REXEC_COPY_STR); } - -STATIC void -S_cache_re(pTHX_ regexp *prog) -{ - dVAR; - PL_regprecomp = prog->precomp; /* Needed for FAIL. */ -#ifdef DEBUGGING - PL_regprogram = prog->program; #endif - PL_regnpar = prog->nparens; - PL_regdata = prog->data; - PL_reg_re = prog; -} /* * Need to implement the following flags for reg_anch: @@ -869,8 +823,7 @@ 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; @@ -962,8 +915,11 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } /* We know what class REx starts with. Try to find this position... */ +/* if reginfo is NULL, its a dryrun */ + STATIC char * -S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 norun) +S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, const char +*strend, const regmatch_info *reginfo) { dVAR; const I32 doevery = (prog->reganch & ROPT_SKIP) == 0; @@ -984,9 +940,9 @@ 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; @@ -1000,12 +956,12 @@ 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; @@ -1018,7 +974,7 @@ 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; @@ -1034,7 +990,7 @@ 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); @@ -1063,7 +1019,7 @@ 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 @@ -1081,7 +1037,7 @@ 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". */ @@ -1092,7 +1048,7 @@ 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]; @@ -1105,7 +1061,7 @@ 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; @@ -1131,7 +1087,7 @@ 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]; @@ -1144,7 +1100,7 @@ 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; @@ -1158,7 +1114,7 @@ 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++; } @@ -1168,7 +1124,7 @@ 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++; } @@ -1183,7 +1139,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 tmp = '\n'; else { U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); - tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0); + 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); @@ -1194,7 +1150,7 @@ 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; @@ -1207,13 +1163,13 @@ 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: @@ -1225,7 +1181,7 @@ S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 tmp = '\n'; else { U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr); - tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0); + 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); @@ -1235,7 +1191,7 @@ 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; } @@ -1248,12 +1204,12 @@ 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: @@ -1261,7 +1217,7 @@ 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; @@ -1274,7 +1230,7 @@ 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; @@ -1290,7 +1246,7 @@ 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; @@ -1303,7 +1259,7 @@ 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; @@ -1319,7 +1275,7 @@ 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; @@ -1332,7 +1288,7 @@ 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; @@ -1348,7 +1304,7 @@ 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; @@ -1361,7 +1317,7 @@ 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; @@ -1377,7 +1333,7 @@ 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; @@ -1390,7 +1346,7 @@ 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; @@ -1406,7 +1362,7 @@ 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; @@ -1419,7 +1375,7 @@ 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; @@ -1435,7 +1391,7 @@ 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; @@ -1448,7 +1404,7 @@ 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; @@ -1464,7 +1420,7 @@ 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; @@ -1477,7 +1433,7 @@ 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; @@ -1493,7 +1449,7 @@ 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; @@ -1506,7 +1462,7 @@ 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; @@ -1522,7 +1478,7 @@ 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; @@ -1535,7 +1491,7 @@ 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; @@ -1551,7 +1507,7 @@ 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; @@ -1564,7 +1520,7 @@ 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; @@ -1580,7 +1536,7 @@ 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; @@ -1593,7 +1549,7 @@ 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; @@ -1636,21 +1592,16 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * char *scream_olds = NULL; SV* oreplsv = GvSV(PL_replgv); const bool do_utf8 = DO_UTF8(sv); - const I32 multiline = prog->reganch & PMf_MULTILINE; + I32 multiline; #ifdef DEBUGGING - SV * const dsv0 = PERL_DEBUG_PAD_ZERO(0); - SV * const dsv1 = PERL_DEBUG_PAD_ZERO(1); + SV* dsv0; + SV* dsv1; #endif + regmatch_info reginfo; /* create some info to pass to regtry etc */ GET_RE_DEBUG_FLAGS_DECL; PERL_UNUSED_ARG(data); - RX_MATCH_UTF8_set(prog,do_utf8); - - cache_re(prog); -#ifdef DEBUGGING - PL_regnarrate = DEBUG_r_TEST; -#endif /* Be paranoid... */ if (prog == NULL || startpos == NULL) { @@ -1658,6 +1609,16 @@ 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, @@ -1678,40 +1639,37 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * PL_reg_flags |= RF_utf8; /* Mark beginning of line for ^ and lookbehind. */ - PL_regbol = startpos; + reginfo.bol = startpos; /* XXX not used ??? */ PL_bostr = strbeg; - PL_reg_sv = sv; + reginfo.sv = sv; /* Mark end of line for $ (and such) */ PL_regeol = strend; /* see how far we have to get to not match where we matched before */ - PL_regtill = startpos+minend; - - /* We start without call_cc context. */ - PL_reg_call_cc = 0; + reginfo.till = startpos+minend; /* If there is a "must appear" string, look for it. */ s = startpos; - if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */ + if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to set reginfo->ganch */ MAGIC *mg; if (flags & REXEC_IGNOREPOS) /* Means: check only at start */ - PL_reg_ganch = startpos; + reginfo.ganch = startpos; else if (sv && SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv) && (mg = mg_find(sv, PERL_MAGIC_regex_global)) && mg->mg_len >= 0) { - PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */ + reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ if (prog->reganch & ROPT_ANCH_GPOS) { - if (s > PL_reg_ganch) + if (s > reginfo.ganch) goto phooey; - s = PL_reg_ganch; + s = reginfo.ganch; } } else /* pos() not defined */ - PL_reg_ganch = strbeg; + reginfo.ganch = strbeg; } if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { @@ -1753,7 +1711,7 @@ 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? */ @@ -1768,7 +1726,7 @@ 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) @@ -1786,7 +1744,7 @@ 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; } } @@ -1794,7 +1752,7 @@ 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; } @@ -1815,7 +1773,7 @@ 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); @@ -1827,7 +1785,7 @@ 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++; @@ -1898,21 +1856,21 @@ 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++; } @@ -1944,7 +1902,7 @@ 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) : @@ -1958,7 +1916,7 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * len0, len0, s0, len1, len1, s1); }); - if (find_byclass(prog, c, s, strend, 0)) + if (find_byclass(prog, c, s, strend, ®info)) goto got_it; DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n")); } @@ -2016,7 +1974,7 @@ 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; @@ -2025,7 +1983,7 @@ 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); } @@ -2043,7 +2001,7 @@ 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 */ @@ -2083,7 +2041,7 @@ 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; } @@ -2091,13 +2049,13 @@ phooey: - regtry - try match at specific point */ STATIC I32 /* 0 failure, 1 success */ -S_regtry(pTHX_ regexp *prog, char *startpos) +S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos) { dVAR; - register I32 i; register I32 *sp; register I32 *ep; CHECKPOINT lastcp; + regexp *prog = reginfo->prog; GET_RE_DEBUG_FLAGS_DECL; #ifdef DEBUGGING @@ -2119,27 +2077,27 @@ S_regtry(pTHX_ regexp *prog, char *startpos) /* SAVEI8(cxstack[cxstack_ix].blk_gimme); cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ - if (PL_reg_sv) { + if (reginfo->sv) { /* Make $_ available to executed code. */ - if (PL_reg_sv != DEFSV) { + if (reginfo->sv != DEFSV) { SAVE_DEFSV; - DEFSV = PL_reg_sv; + DEFSV = reginfo->sv; } - if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv) - && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) { + if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) + && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) { /* prepare for quick setting of pos */ #ifdef PERL_OLD_COPY_ON_WRITE if (SvIsCOW(sv)) sv_force_normal_flags(sv, 0); #endif - mg = sv_magicext(PL_reg_sv, (SV*)0, PERL_MAGIC_regex_global, + mg = sv_magicext(reginfo->sv, (SV*)0, PERL_MAGIC_regex_global, &PL_vtbl_mglob, NULL, 0); mg->mg_len = -1; } PL_reg_magic = mg; PL_reg_oldpos = mg->mg_len; - SAVEDESTRUCTOR_X(restore_pos, 0); + SAVEDESTRUCTOR_X(restore_pos, prog); } if (!PL_reg_curpm) { Newxz(PL_reg_curpm, 1, PMOP); @@ -2209,6 +2167,7 @@ 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; @@ -2216,7 +2175,7 @@ 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; } @@ -2243,6 +2202,7 @@ typedef struct { regnode *next; char *locinput; I32 nextchr; + int minmod; #ifdef DEBUGGING int regindent; #endif @@ -2258,7 +2218,6 @@ typedef union re_unwind_t { #define sayNO goto no #define sayNO_ANYOF goto no_anyof #define sayYES_FINAL goto yes_final -#define sayYES_LOUD goto yes_loud #define sayNO_FINAL goto no_final #define sayNO_SILENT goto do_no #define saySAME(x) if (x) goto yes; else goto no @@ -2266,34 +2225,46 @@ typedef union re_unwind_t { #define POSCACHE_SUCCESS 0 /* caching success rather than failure */ #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 (!(PL_reg_poscache[0] & (1<u.whilem.cache_offset | st->u.whilem.cache_bit) { \ + if (!(PL_reg_poscache[0] & (1<u.whilem.cache_offset] |= (1<u.whilem.cache_bit); \ + } \ + else if (PL_reg_poscache[0] & (1<u.whilem.cache_offset] |= (1<u.whilem.cache_bit); \ + } \ + else { \ /* cache records failure, but this is success */ \ DEBUG_r( \ PerlIO_printf(Perl_debug_log, \ "%*s (remove success from failure cache)\n", \ REPORT_CODE_OFF+PL_regindent*2, "") \ ); \ - PL_reg_poscache[cache_offset] &= ~(1<u.whilem.cache_offset] &= ~(1<u.whilem.cache_bit); \ } \ } \ sayYES; \ } STMT_END + #define CACHEsayNO STMT_START { \ - if (cache_offset | cache_bit) { \ - if (!(PL_reg_poscache[0] & (1<u.whilem.cache_offset | st->u.whilem.cache_bit) { \ + if (!(PL_reg_poscache[0] & (1<u.whilem.cache_offset] |= (1<u.whilem.cache_bit); \ + } \ + else if (!(PL_reg_poscache[0] & (1<u.whilem.cache_offset] |= (1<u.whilem.cache_bit); \ + } \ + else { \ /* cache records success, but this is failure */ \ DEBUG_r( \ PerlIO_printf(Perl_debug_log, \ "%*s (remove failure from success cache)\n", \ REPORT_CODE_OFF+PL_regindent*2, "") \ ); \ - PL_reg_poscache[cache_offset] &= ~(1<u.whilem.cache_offset] &= ~(1<u.whilem.cache_bit); \ } \ } \ sayNO; \ @@ -2311,70 +2282,77 @@ typedef union re_unwind_t { /* Make sure there is a test for this +1 options in re_tests */ #define TRIE_INITAL_ACCEPT_BUFFLEN 4; +/* this value indiciates that the c1/c2 "next char" test should be skipped */ +#define CHRTEST_VOID -1000 + +#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) +{ +#if PERL_VERSION < 9 + dMY_CXT; +#endif + 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) \ - new_scan = (ns); \ - resume_state = resume_##where; \ + st->scan = scan; \ + scan = (ns); \ + st->resume_state = resume_##where; \ goto start_recurse; \ resume_point_##where: -typedef enum { - resume_TRIE1, - resume_TRIE2, - resume_CURLYX, - resume_WHILEM1, - resume_WHILEM2, - resume_WHILEM3, - resume_WHILEM4, - resume_WHILEM5, - resume_WHILEM6, - resume_CURLYM1, - resume_CURLYM2, - resume_CURLYM3, - resume_CURLYM4, - resume_IFMATCH, - resume_PLUS1, - resume_PLUS2, - resume_PLUS3, - resume_PLUS4, - resume_END -} resume_states; - - -struct regmatch_state { - struct regmatch_state *prev_state; - resume_states resume_state; - regnode *scan; - regnode *next; - bool minmod; - bool sw; - int logical; - I32 unwind; - CURCUR *cc; - char *locinput; - I32 n; - I32 ln; - I32 c1, c2, paren; - CHECKPOINT cp, lastcp; - CURCUR *oldcc; - char *lastloc; - I32 cache_offset, cache_bit; - I32 curlym_l; - I32 matches; - I32 maxwanted; - char *e; - char *old; - int count; - re_cc_state *cur_call_cc; - regexp *end_re; - reg_trie_accepted *accept_buff; - U32 accepted; - - re_cc_state *reg_call_cc; /* saved value of PL_reg_call_cc */ -}; +/* 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 @@ -2403,6 +2381,7 @@ struct regmatch_state { int a = 1, b = 2; ... while (scan != NULL) { + a++; // do stuff with a and b ... switch (OP(scan)) { case FOO: { @@ -2420,40 +2399,42 @@ struct regmatch_state { * Now it looks something like this: - struct regmatch_state { + typedef struct { int a, b, local; int resume_state; - }; + } regmatch_state; S_regmatch() { - int a = 1, b = 2; - int local; + regmatch_state *st = new(); + int depth=0; + st->a++; // do stuff with a and b ... while (scan != NULL) { ... switch (OP(scan)) { case FOO: { - local = 3; + st->local = 3; ... - resume_state = resume_FOO; - new_scan = ...; - goto start_recurse; - resume_point_FOO: + st->scan = scan; + scan = ...; + st->resume_state = resume_FOO; + goto start_recurse; // recurse - if (result) // recurse + resume_point_FOO: + if (result) goto yes; } ... } - start_recurse: - ...push a, b, local onto the heap - a = 1; b = 2; - scan = new_scan; + start_recurse: + st = new(); push a new state + st->a = 1; st->b = 2; + depth++; } - yes: + yes: result = 1; - if (states pushed on heap) { - ... restore a, b, local from heap + if (depth--) { + st = pop(); switch (resume_state) { case resume_FOO: goto resume_point_FOO; @@ -2468,68 +2449,49 @@ struct regmatch_state { * 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) { +#if PERL_VERSION < 9 + dMY_CXT; +#endif dVAR; register const bool do_utf8 = PL_reg_match_utf8; - const U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY; - - /************************************************************ - * the following variabes are saved and restored on each fake - * recursive call to regmatch: - * - * The first ones contain state that needs to be maintained - * across the main while loop: */ - - struct regmatch_state *prev_state = NULL; /* stack of pushed states */ - resume_states resume_state; /* where to jump to on return */ - register regnode *scan; /* Current node. */ - regnode *next; /* Next node. */ - bool minmod = 0; /* the next {n.m} is a {n,m}? */ - bool sw = 0; /* the condition value in (?(cond)a|b) */ - int logical = 0; - I32 unwind = 0; /* savestack index of current unwind block */ - CURCUR *cc = NULL; /* current innermost curly struct */ - register char *locinput = PL_reginput; + const U32 uniflags = UTF8_ALLOW_DEFAULT; - /* while the rest of these are local to an individual branch, and - * have only been hoisted into this outer scope to allow for saving and - * restoration - thus they can be safely reused in other branches. - * Note that they are only initialized here to silence compiler - * warnings :-( */ - - register I32 n = 0; /* no or next */ - register I32 ln = 0; /* len or last */ - register I32 c1 = 0, c2 = 0, paren = 0; /* case fold search, parenth */ - CHECKPOINT cp = 0; /* remember current savestack indexes */ - CHECKPOINT lastcp = 0; - CURCUR *oldcc = NULL; /* tmp copy of cc */ - char *lastloc = NULL; /* Detection of 0-len. */ - I32 cache_offset = 0; - I32 cache_bit = 0; - I32 curlym_l = 0; - I32 matches = 0; - I32 maxwanted = 0; - char *e = NULL; - char *old = NULL; - int count = 0; - re_cc_state *cur_call_cc = NULL; - regexp *end_re = NULL; - reg_trie_accepted *accept_buff = NULL; - U32 accepted = 0; /* how many accepting states we have seen */ - - /************************************************************ - * these variables are NOT saved: */ + regexp *rex = reginfo->prog; - register I32 nextchr; /* is always set to UCHARAT(locinput) */ - regnode *new_scan; /* node to begin exedcution at when recursing */ - bool result; /* return value of S_regmatch */ + 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; - regnode *inner; /* Next node in internal branch. */ + /* 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; @@ -2537,6 +2499,29 @@ S_regmatch(pTHX_ regnode *prog) 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; @@ -2569,7 +2554,7 @@ S_regmatch(pTHX_ regnode *prog) pref0_len = 0; if (pref0_len > pref_len) pref0_len = pref_len; - regprop(prop, scan); + regprop(rex, prop, scan); { const char * const s0 = do_utf8 && OP(scan) != CANY ? @@ -2603,7 +2588,7 @@ 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)); } }); @@ -2616,7 +2601,7 @@ S_regmatch(pTHX_ regnode *prog) case BOL: if (locinput == PL_bostr) { - /* regtill = regbol; */ + /* reginfo->till = reginfo->bol; */ break; } sayNO; @@ -2632,7 +2617,7 @@ S_regmatch(pTHX_ regnode *prog) break; sayNO; case GPOS: - if (locinput == PL_reg_ganch) + if (locinput == reginfo->ganch) break; sayNO; case EOL: @@ -2711,14 +2696,14 @@ S_regmatch(pTHX_ regnode *prog) /* what trie are we using right now */ reg_trie_data *trie - = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ]; - accepted = 0; /* how many accepting states we have seen */ + = (reg_trie_data*)rex->data->data[ ARG( scan ) ]; + st->u.trie.accepted = 0; /* how many accepting states we have seen */ result = 0; while ( state && uc <= (U8*)PL_regeol ) { if (trie->states[ state ].wordnum) { - if (!accepted ) { + if (!st->u.trie.accepted ) { ENTER; SAVETMPS; bufflen = TRIE_INITAL_ACCEPT_BUFFLEN; @@ -2728,22 +2713,22 @@ S_regmatch(pTHX_ regnode *prog) sizeof(reg_trie_accepted)); SvPOK_on(sv_accept_buff); sv_2mortal(sv_accept_buff); - accept_buff = + st->u.trie.accept_buff = (reg_trie_accepted*)SvPV_nolen(sv_accept_buff ); } else { - if (accepted >= bufflen) { + if (st->u.trie.accepted >= bufflen) { bufflen *= 2; - accept_buff =(reg_trie_accepted*) + st->u.trie.accept_buff =(reg_trie_accepted*) SvGROW(sv_accept_buff, bufflen * sizeof(reg_trie_accepted)); } SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff) + sizeof(reg_trie_accepted)); } - accept_buff[accepted].wordnum = trie->states[state].wordnum; - accept_buff[accepted].endpos = uc; - ++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; @@ -2752,7 +2737,7 @@ S_regmatch(pTHX_ regnode *prog) PerlIO_printf( Perl_debug_log, "%*s %sState: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ", REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], - (UV)state, (UV)base, (UV)accepted ); + (UV)state, (UV)base, (UV)st->u.trie.accepted ); ); if ( base ) { @@ -2819,7 +2804,7 @@ S_regmatch(pTHX_ regnode *prog) charid, uvc, (UV)state, PL_colors[5] ); ); } - if (!accepted ) + if (!st->u.trie.accepted ) sayNO; /* @@ -2835,17 +2820,17 @@ S_regmatch(pTHX_ regnode *prog) have been tried. */ - 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; @@ -2855,39 +2840,42 @@ S_regmatch(pTHX_ regnode *prog) } else { DEBUG_EXECUTE_r( PerlIO_printf( Perl_debug_log,"%*s %sgot %"IVdf" possible matches%s\n", - REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], (IV)accepted, + REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], (IV)st->u.trie.accepted, PL_colors[5] ); ); - while ( !result && 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 ** const 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 @@ -2912,11 +2900,11 @@ S_regmatch(pTHX_ regnode *prog) /* unreached codepoint */ case EXACT: { char *s = STRING(scan); - ln = STR_LEN(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. */ @@ -2954,11 +2942,11 @@ 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; } @@ -2967,14 +2955,14 @@ S_regmatch(pTHX_ regnode *prog) /* FALL THROUGH */ case EXACTF: { char *s = STRING(scan); - ln = STR_LEN(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, @@ -2982,7 +2970,7 @@ 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 && @@ -3001,13 +2989,13 @@ 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; } @@ -3015,7 +3003,7 @@ S_regmatch(pTHX_ regnode *prog) if (do_utf8) { STRLEN inclasslen = PL_regeol - locinput; - if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8)) + if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8)) sayNO_ANYOF; if (locinput >= PL_regeol) sayNO; @@ -3026,7 +3014,7 @@ 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; @@ -3099,35 +3087,35 @@ 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; @@ -3256,14 +3244,14 @@ S_regmatch(pTHX_ regnode *prog) 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]; @@ -3299,16 +3287,16 @@ 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; } @@ -3320,21 +3308,21 @@ S_regmatch(pTHX_ regnode *prog) break; case EVAL: { - dSP; - OP_4tree * const oop = PL_op; - COP * const ocurcop = PL_curcop; - PAD *old_comppad; SV *ret; - struct regexp * const oreg = PL_reg_re; - - n = ARG(scan); - PL_op = (OP_4tree*)PL_regdata->data[n]; - DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); - PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]); - PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr; - { + /* execute the code in the {...} */ + dSP; SV ** const before = SP; + OP_4tree * const oop = PL_op; + COP * const ocurcop = PL_curcop; + PAD *old_comppad; + + n = ARG(scan); + PL_op = (OP_4tree*)rex->data->data[n]; + DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) ); + PAD_SAVE_LOCAL(old_comppad, (PAD*)rex->data->data[n + 2]); + PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr; + CALLRUNOPS(aTHX); /* Scalar context. */ SPAGAIN; if (SP == before) @@ -3343,19 +3331,24 @@ S_regmatch(pTHX_ regnode *prog) ret = POPs; PUTBACK; } + + PL_op = oop; + PAD_RESTORE_LOCAL(old_comppad); + PL_curcop = ocurcop; + if (!st->logical) { + /* /(?{...})/ */ + sv_setsv(save_scalar(PL_replgv), ret); + break; + } } + if (st->logical == 2) { /* Postponed subexpression: /(??{...})/ */ + regexp *re; + { + /* extract RE object from returned value; compiling if + * necessary */ - PL_op = oop; - PAD_RESTORE_LOCAL(old_comppad); - PL_curcop = ocurcop; - if (logical) { - if (logical == 2) { /* Postponed subexpression. */ - regexp *re; MAGIC *mg = NULL; - re_cc_state state; - int toggleutf; - register SV *sv; - + SV *sv; if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret))) mg = mg_find(sv, PERL_MAGIC_qr); else if (SvSMAGICAL(ret)) { @@ -3373,95 +3366,61 @@ S_regmatch(pTHX_ regnode *prog) STRLEN len; const char * const t = SvPV_const(ret, len); PMOP pm; - char * const oprecomp = PL_regprecomp; const I32 osize = PL_regsize; - const I32 onpar = PL_regnpar; Zero(&pm, 1, PMOP); - if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8; + if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8; re = CALLREGCOMP(aTHX_ (char*)t, (char*)t + len, &pm); if (!(SvFLAGS(ret) & (SVs_TEMP | SVs_PADTMP | SVf_READONLY | SVs_GMG))) sv_magic(ret,(SV*)ReREFCNT_inc(re), PERL_MAGIC_qr,0,0); - PL_regprecomp = oprecomp; PL_regsize = osize; - PL_regnpar = onpar; - } - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "Entering embedded \"%s%.60s%s%s\"\n", - PL_colors[0], - re->precomp, - PL_colors[1], - (strlen(re->precomp) > 60 ? "..." : "")) - ); - state.node = next; - state.prev = PL_reg_call_cc; - state.cc = cc; - state.re = PL_reg_re; - - cc = 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; - - /* XXX the only recursion left in regmatch() */ - if (regmatch(re->program + 1)) { - /* Even though we succeeded, we need to restore - global variables, since we may be wrapped inside - SUSPEND, thus the match may be not finished yet. */ - - /* XXXX Do this only if SUSPENDed? */ - PL_reg_call_cc = state.prev; - cc = state.cc; - PL_reg_re = state.re; - cache_re(PL_reg_re); - if (toggleutf) PL_reg_flags ^= RF_utf8; - - /* XXXX This is too dramatic a measure... */ - PL_reg_maxiter = 0; - - /* These are needed even if not SUSPEND. */ - ReREFCNT_dec(re); - regcpblow(cp); - sayYES; } - ReREFCNT_dec(re); - REGCP_UNWIND(lastcp); - regcppop(); - PL_reg_call_cc = state.prev; - cc = state.cc; - PL_reg_re = state.re; - cache_re(PL_reg_re); - if (toggleutf) PL_reg_flags ^= RF_utf8; - - /* XXXX This is too dramatic a measure... */ - PL_reg_maxiter = 0; - - 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: @@ -3480,11 +3439,11 @@ 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); @@ -3493,23 +3452,24 @@ S_regmatch(pTHX_ regnode *prog) } break; case LOGICAL: - logical = scan->flags; + st->logical = scan->flags; break; /******************************************************************* - cc contains infoblock about the innermost (...)* loop, and - a pointer to the next outer infoblock. + cc points to the regmatch_state associated with the most recent CURLYX. + This struct contains info about the innermost (...)* loop (an + "infoblock"), and a pointer to the next outer cc. Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM): 1) After matching Y, regnode for CURLYX is processed; - 2) This regnode mallocs an infoblock, and calls regmatch() recursively + 2) This regnode populates cc, and calls regmatch() recursively with the starting point at WHILEM node; 3) Each hit of WHILEM node tries to match A and Z (in the order depending on the current iteration, min/max of {min,max} and greediness). The information about where are nodes for "A" - and "Z" is read from the infoblock, as is info on how many times "A" + and "Z" is read from cc, as is info on how many times "A" was already matched, and greediness. 4) After A matches, the same WHILEM node is hit again. @@ -3520,7 +3480,7 @@ S_regmatch(pTHX_ regnode *prog) as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node of the external loop. - Currently present infoblocks form a tree with a stem formed by PL_curcc + Currently present infoblocks form a tree with a stem formed by st->cc and whatever it mentions via ->next, and additional attached trees corresponding to temporarily unset infoblocks as in "5" above. @@ -3575,34 +3535,44 @@ S_regmatch(pTHX_ regnode *prog) /* No need to save/restore up to this paren */ I32 parenfloor = scan->flags; - { - CURCUR *newcc; - Newx(newcc, 1, CURCUR); - oldcc = cc; - newcc->oldcc = cc; - cc = newcc; - } - cp = PL_savestack_ix; + /* Dave says: + + CURLYX and WHILEM are always paired: they're the moral + equivalent of pp_enteriter anbd pp_iter. + + The only time next could be null is if the node tree is + corrupt. This was mentioned on p5p a few days ago. + + See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2006-04/msg00556.html + So we'll assert that this is true: + */ + assert(next); if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ next += ARG(next); /* XXXX Probably it is better to teach regpush to support parenfloor > PL_regsize... */ if (parenfloor > (I32)*PL_reglastparen) parenfloor = *PL_reglastparen; /* Pessimization... */ - 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; REGMATCH(PREVOPER(next), CURLYX); /* start on the WHILEM */ /*** all unsaved local vars undefined at this point */ - regcpblow(cp); - Safefree(cc); - cc = oldcc; + regcpblow(st->u.curlyx.cp); + st->cc = st->u.curlyx.outercc; saySAME(result); } /* NOTREACHED */ @@ -3616,54 +3586,64 @@ S_regmatch(pTHX_ regnode *prog) * that we can try again after backing off. */ - lastloc = cc->lastloc; /* Detection of 0-len. */ - cache_offset = 0; - cache_bit = 0; + /* Dave says: + + st->cc gets initialised by CURLYX ready for use by WHILEM. + So again, unless somethings been corrupted, st->cc cannot + be null at that point in WHILEM. + + See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2006-04/msg00556.html + So we'll assert that this is true: + */ + assert(st->cc); + 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) { - oldcc = cc; - cc = cc->oldcc; - if (cc) - ln = cc->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, "") ); - REGMATCH(oldcc->next, WHILEM1); + REGMATCH(st->u.whilem.savecc->next, WHILEM1); /*** all unsaved local vars undefined at this point */ - cc = oldcc; + st->cc = st->u.whilem.savecc; if (result) sayYES; - if (cc->oldcc) - cc->oldcc->cur = ln; + if (st->cc->u.curlyx.outercc) + st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln; sayNO; } /* First just match a string of min scans. */ - if (n < cc->min) { - cc->cur = n; - cc->lastloc = locinput; - REGMATCH(cc->scan, WHILEM2); + if (n < st->cc->u.curlyx.min) { + st->cc->u.curlyx.cur = n; + st->cc->u.curlyx.lastloc = locinput; + REGMATCH(st->cc->u.curlyx.scan, WHILEM2); /*** all unsaved local vars undefined at this point */ if (result) sayYES; - cc->cur = n - 1; - cc->lastloc = lastloc; + st->cc->u.curlyx.cur = n - 1; + st->cc->u.curlyx.lastloc = st->u.whilem.lastloc; sayNO; } @@ -3695,13 +3675,13 @@ 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", @@ -3714,32 +3694,31 @@ S_regmatch(pTHX_ regnode *prog) /* cache records failure */ sayNO_SILENT; } - PL_reg_poscache[cache_offset] |= (1<minmod) { - oldcc = cc; - cc = cc->oldcc; - if (cc) - ln = cc->cur; - cp = regcppush(oldcc->parenfloor); - REGCP_SET(lastcp); - REGMATCH(oldcc->next, WHILEM3); + 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 */ - cc = oldcc; + st->cc = st->u.whilem.savecc; if (result) { - regcpblow(cp); + regcpblow(st->u.whilem.cp); CACHEsayYES; /* All done. */ } - REGCP_UNWIND(lastcp); - regcppop(); - if (cc->oldcc) - cc->oldcc->cur = ln; + REGCP_UNWIND(st->u.whilem.lastcp); + regcppop(rex); + if (st->cc->u.curlyx.outercc) + st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln; - if (n >= 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; @@ -3757,38 +3736,38 @@ 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); - REGMATCH(cc->scan, WHILEM4); + st->cc->u.curlyx.cur = n; + st->cc->u.curlyx.lastloc = locinput; + st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor); + REGCP_SET(st->u.whilem.lastcp); + REGMATCH(st->cc->u.curlyx.scan, WHILEM4); /*** all unsaved local vars undefined at this point */ if (result) { - regcpblow(cp); + 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); - REGMATCH(cc->scan, WHILEM5); + if (n < st->cc->u.curlyx.max) { /* More greed allowed? */ + st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor); + st->cc->u.curlyx.cur = n; + st->cc->u.curlyx.lastloc = locinput; + REGCP_SET(st->u.whilem.lastcp); + REGMATCH(st->cc->u.curlyx.scan, WHILEM5); /*** all unsaved local vars undefined at this point */ if (result) { - regcpblow(cp); + 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, @@ -3805,19 +3784,19 @@ S_regmatch(pTHX_ regnode *prog) } /* Failed deeper matches of scan, so see if this one works. */ - oldcc = cc; - cc = cc->oldcc; - if (cc) - ln = cc->cur; - REGMATCH(oldcc->next, WHILEM6); + 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 */ - cc = oldcc; + st->cc = st->u.whilem.savecc; if (result) CACHEsayYES; - if (cc->oldcc) - cc->oldcc->cur = ln; - 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; } /* NOTREACHED */ @@ -3831,8 +3810,9 @@ 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; @@ -3840,15 +3820,16 @@ S_regmatch(pTHX_ regnode *prog) const I32 unwind1 = SSNEWt(1,re_unwind_branch_t); re_unwind_branch_t * const uw = SSPTRt(unwind1,re_unwind_branch_t); - uw->prev = unwind; - unwind = unwind1; - uw->type = ((c1 == BRANCH) + 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 @@ -3861,183 +3842,149 @@ S_regmatch(pTHX_ regnode *prog) } break; case MINMOD: - minmod = 1; + st->minmod = 1; break; case CURLYM: { - curlym_l = matches = 0; + st->u.curlym.l = st->u.curlym.matches = 0; /* We suppose that the next guy does not need backtracking: in particular, it is of constant non-zero length, and has no parenths to influence future backrefs. */ - 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; - maxwanted = minmod ? ln : n; - if (maxwanted) { - while (PL_reginput < PL_regeol && matches < maxwanted) { - REGMATCH(scan, CURLYM1); - /*** all unsaved local vars undefined at this point */ - if (!result) - break; - /* on first match, determine length, curlym_l */ - if (!matches++) { - if (PL_reg_match_utf8) { - char *s = locinput; - while (s < PL_reginput) { - curlym_l++; - s += UTF8SKIP(s); - } - } - else { - curlym_l = PL_reginput - locinput; - } - if (curlym_l == 0) { - matches = maxwanted; - break; + st->u.curlym.maxwanted = st->minmod ? st->ln : n; + while (PL_reginput < PL_regeol && st->u.curlym.matches < st->u.curlym.maxwanted) { + /* resume to current state on success */ + st->u.yes.prev_yes_state = yes_state; + yes_state = st; + REGMATCH(scan, CURLYM1); + yes_state = st->u.yes.prev_yes_state; + /*** 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); } } - locinput = PL_reginput; + 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->u.curlym.matches < st->ln) { + st->minmod = 0; + sayNO; + } - if (minmod) { - minmod = 0; - if (ln && matches < ln) - sayNO; - if (HAS_TEXT(next) || JUMPABLE(next)) { - regnode *text_node = next; - - if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node); + DEBUG_EXECUTE_r( + PerlIO_printf(Perl_debug_log, + "%*s matched %"IVdf" times, len=%"IVdf"...\n", + (int)(REPORT_CODE_OFF+PL_regindent*2), "", + (IV) st->u.curlym.matches, (IV)st->u.curlym.l) + ); + + /* calculate c1 and c1 for possible match of 1st char + * following curly */ + st->u.curlym.c1 = st->u.curlym.c2 = CHRTEST_VOID; + 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) + && PL_regkind[(U8)OP(text_node)] != REF) + { + st->u.curlym.c1 = (U8)*STRING(text_node); + st->u.curlym.c2 = + (OP(text_node) == EXACTF || OP(text_node) == REFF) + ? PL_fold[st->u.curlym.c1] + : (OP(text_node) == EXACTFL || OP(text_node) == REFFL) + ? PL_fold_locale[st->u.curlym.c1] + : st->u.curlym.c1; + } + } - if (! HAS_TEXT(text_node)) c1 = c2 = -1000; - else { - if (PL_regkind[(U8)OP(text_node)] == REF) { - c1 = c2 = -1000; - goto assume_ok_MM; + REGCP_SET(st->u.curlym.lastcp); + + st->u.curlym.minmod = st->minmod; + st->minmod = 0; + while (st->u.curlym.matches >= st->ln + && (st->u.curlym.matches <= n + /* for REG_INFTY, ln could overflow to negative */ + || (n == REG_INFTY && st->u.curlym.matches >= 0))) + { + /* If it could work, try it. */ + if (st->u.curlym.c1 == CHRTEST_VOID || + UCHARAT(PL_reginput) == st->u.curlym.c1 || + UCHARAT(PL_reginput) == st->u.curlym.c2) + { + DEBUG_EXECUTE_r( + PerlIO_printf(Perl_debug_log, + "%*s trying tail with matches=%"IVdf"...\n", + (int)(REPORT_CODE_OFF+PL_regindent*2), + "", (IV)st->u.curlym.matches) + ); + 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 { c1 = (U8)*STRING(text_node); } - if (OP(text_node) == EXACTF || OP(text_node) == REFF) - c2 = PL_fold[c1]; - else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL) - c2 = PL_fold_locale[c1]; else - c2 = c1; + PL_regendp[st->u.curlym.paren] = -1; } + /* resume to current state on success */ + st->u.yes.prev_yes_state = yes_state; + yes_state = st; + REGMATCH(next, CURLYM2); + yes_state = st->u.yes.prev_yes_state; + /*** all unsaved local vars undefined at this point */ + if (result) + /* XXX tmp sayYES; */ + sayYES_FINAL; + REGCP_UNWIND(st->u.curlym.lastcp); } - else - c1 = c2 = -1000; - assume_ok_MM: - REGCP_SET(lastcp); - while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */ - /* If it could work, try it. */ - if (c1 == -1000 || - UCHARAT(PL_reginput) == c1 || - UCHARAT(PL_reginput) == c2) - { - if (paren) { - if (ln) { - PL_regstartp[paren] = - HOPc(PL_reginput, -curlym_l) - PL_bostr; - PL_regendp[paren] = PL_reginput - PL_bostr; - } - else - PL_regendp[paren] = -1; - } - REGMATCH(next, CURLYM2); - /*** all unsaved local vars undefined at this point */ - if (result) - sayYES; - REGCP_UNWIND(lastcp); - } - /* Couldn't or didn't -- move forward. */ + /* Couldn't or didn't -- move forward/backward. */ + if (st->u.curlym.minmod) { PL_reginput = locinput; + /* resume to current state on success */ + st->u.yes.prev_yes_state = yes_state; + yes_state = st; REGMATCH(scan, CURLYM3); + yes_state = st->u.yes.prev_yes_state; /*** all unsaved local vars undefined at this point */ if (result) { - ln++; + st->u.curlym.matches++; locinput = PL_reginput; } else sayNO; } - } - else { - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s matched %"IVdf" times, len=%"IVdf"...\n", - (int)(REPORT_CODE_OFF+PL_regindent*2), "", - (IV) matches, (IV)curlym_l) - ); - if (matches >= 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; - else { - if (PL_regkind[(U8)OP(text_node)] == REF) { - c1 = c2 = -1000; - goto assume_ok_REG; - } - else { c1 = (U8)*STRING(text_node); } - - if (OP(text_node) == EXACTF || OP(text_node) == REFF) - c2 = PL_fold[c1]; - else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL) - c2 = PL_fold_locale[c1]; - else - c2 = c1; - } - } - else - c1 = c2 = -1000; - } - assume_ok_REG: - REGCP_SET(lastcp); - while (matches >= ln) { - /* If it could work, try it. */ - if (c1 == -1000 || - UCHARAT(PL_reginput) == c1 || - UCHARAT(PL_reginput) == c2) - { - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s trying tail with matches=%"IVdf"...\n", - (int)(REPORT_CODE_OFF+PL_regindent*2), - "", (IV)matches) - ); - if (paren) { - if (matches) { - PL_regstartp[paren] - = HOPc(PL_reginput, -curlym_l) - PL_bostr; - PL_regendp[paren] = PL_reginput - PL_bostr; - } - else - PL_regendp[paren] = -1; - } - REGMATCH(next, CURLYM4); - /*** all unsaved local vars undefined at this point */ - if (result) - sayYES; - REGCP_UNWIND(lastcp); - } - /* Couldn't or didn't -- back up. */ - matches--; - locinput = HOPc(locinput, -curlym_l); + else { + st->u.curlym.matches--; + locinput = HOPc(locinput, -st->u.curlym.l); PL_reginput = locinput; } } @@ -4046,32 +3993,32 @@ S_regmatch(pTHX_ regnode *prog) 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 @@ -4090,20 +4037,21 @@ 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 = CHRTEST_VOID; else { if (PL_regkind[(U8)OP(text_node)] == REF) { - c1 = c2 = -1000; + st->u.plus.c1 = st->u.plus.c2 = CHRTEST_VOID; 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) { @@ -4114,117 +4062,117 @@ 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 = CHRTEST_VOID; assume_ok_easy: PL_reginput = locinput; - if (minmod) { - 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) { - old = locinput; - count = 0; + REGCP_SET(st->u.plus.lastcp); + if (st->u.plus.c1 != CHRTEST_VOID) { + 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) { - UV c = utf8n_to_uvchr((U8*)locinput, + while (locinput <= st->u.plus.e) { + STRLEN len; + const 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, PLUS1); + TRYPAREN(st->u.plus.paren, st->ln, locinput, PLUS1); /*** all unsaved local vars undefined at this point */ PL_reginput = locinput; /* Could be reset... */ - REGCP_UNWIND(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 != CHRTEST_VOID) { if (do_utf8) c = utf8n_to_uvchr((U8*)PL_reginput, UTF8_MAXBYTES, 0, @@ -4232,24 +4180,24 @@ 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, PLUS2); + TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS2); /*** all unsaved local vars undefined at this point */ - REGCP_UNWIND(lastcp); + REGCP_UNWIND(st->u.plus.lastcp); } } /* If it could work, try it. */ - else if (c1 == -1000) + else if (st->u.plus.c1 == CHRTEST_VOID) { - TRYPAREN(paren, ln, PL_reginput, PLUS3); + TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS3); /*** all unsaved local vars undefined at this point */ - REGCP_UNWIND(lastcp); + 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 @@ -4257,24 +4205,24 @@ S_regmatch(pTHX_ regnode *prog) } } else { - 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); + REGCP_SET(st->u.plus.lastcp); { UV c = 0; - while (n >= ln) { - if (c1 != -1000) { + while (n >= st->ln) { + if (st->u.plus.c1 != CHRTEST_VOID) { if (do_utf8) c = utf8n_to_uvchr((U8*)PL_reginput, UTF8_MAXBYTES, 0, @@ -4283,11 +4231,11 @@ 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 == CHRTEST_VOID || c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2) { - TRYPAREN(paren, n, PL_reginput, PLUS4); + TRYPAREN(st->u.plus.paren, n, PL_reginput, PLUS4); /*** all unsaved local vars undefined at this point */ - REGCP_UNWIND(lastcp); + REGCP_UNWIND(st->u.plus.lastcp); } /* Couldn't or didn't -- back up. */ n--; @@ -4298,115 +4246,68 @@ S_regmatch(pTHX_ regnode *prog) sayNO; break; case END: - if (PL_reg_call_cc) { - cur_call_cc = PL_reg_call_cc; - end_re = PL_reg_re; - - /* Save *all* the positions. */ - cp = regcppush(0); - REGCP_SET(lastcp); - - /* Restore parens of the caller. */ - { - I32 tmp = PL_savestack_ix; - PL_savestack_ix = PL_reg_call_cc->ss; - regcppop(); - PL_savestack_ix = tmp; - } - - /* Make position available to the callcc. */ - PL_reginput = locinput; - - cache_re(PL_reg_call_cc->re); - oldcc = cc; - cc = PL_reg_call_cc->cc; - PL_reg_call_cc = PL_reg_call_cc->prev; - REGMATCH(cur_call_cc->node, END); - /*** all unsaved local vars undefined at this point */ - if (result) { - PL_reg_call_cc = cur_call_cc; - regcpblow(cp); - sayYES; - } - REGCP_UNWIND(lastcp); - regcppop(); - PL_reg_call_cc = cur_call_cc; - cc = oldcc; - PL_reg_re = end_re; - cache_re(end_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. */ } PL_reginput = locinput; /* put where regtry can find it */ sayYES_FINAL; /* Success! */ - case SUCCEED: + + case SUCCEED: /* successful SUSPEND/UNLESSM/IFMATCH/CURLYM */ + DEBUG_EXECUTE_r( + PerlIO_printf(Perl_debug_log, + "%*s %ssubpattern success...%s\n", + REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])); PL_reginput = locinput; /* put where regtry can find it */ - sayYES_LOUD; /* Success! */ - case SUSPEND: - n = 1; + sayYES_FINAL; /* Success! */ + + case SUSPEND: /* (?>FOO) */ + st->u.ifmatch.wanted = 1; PL_reginput = locinput; goto do_ifmatch; - case UNLESSM: - n = 0; - if (scan->flags) { - char *s = HOPBACKc(locinput, scan->flags); - if (!s) - goto say_yes; - PL_reginput = s; - } - else - PL_reginput = locinput; - goto do_ifmatch; - case IFMATCH: - n = 1; + + case UNLESSM: /* -ve lookaround: (?!FOO), or with flags, (?u.ifmatch.wanted = 0; + goto ifmatch_trivial_fail_test; + + case IFMATCH: /* +ve lookaround: (?=FOO), or with flags, (?<=foo) */ + st->u.ifmatch.wanted = 1; + ifmatch_trivial_fail_test: if (scan->flags) { - char *s = HOPBACKc(locinput, scan->flags); - if (!s) - goto say_no; + char * const s = HOPBACKc(locinput, scan->flags); + if (!s) { + /* trivial fail */ + if (st->logical) { + st->logical = 0; + st->sw = 1 - st->u.ifmatch.wanted; + } + else if (st->u.ifmatch.wanted) + sayNO; + next = scan + ARG(scan); + if (next == scan) + next = NULL; + break; + } PL_reginput = s; } else PL_reginput = locinput; do_ifmatch: - REGMATCH(NEXTOPER(NEXTOPER(scan)), IFMATCH); - /*** all unsaved local vars undefined at this point */ - if (result != n) { - say_no: - if (logical) { - logical = 0; - sw = 0; - goto do_longjump; - } - else - sayNO; - } - say_yes: - if (logical) { - logical = 0; - sw = 1; - } - if (OP(scan) == SUSPEND) { - locinput = PL_reginput; - nextchr = UCHARAT(locinput); - } - /* FALL THROUGH. */ + /* resume to current state on success */ + st->u.yes.prev_yes_state = yes_state; + yes_state = st; + PUSH_STATE(newst, resume_IFMATCH); + st = newst; + next = NEXTOPER(NEXTOPER(scan)); + break; + case LONGJMP: - do_longjump: next = scan + ARG(scan); if (next == scan) next = NULL; @@ -4429,57 +4330,36 @@ S_regmatch(pTHX_ regnode *prog) start_recurse: { - /* save existing local variables */ - struct regmatch_state *p; - - Newx(p, 1, struct regmatch_state); - p->prev_state = prev_state; - prev_state = p; - p->resume_state = resume_state; - p->scan = scan; - p->next = next; - p->minmod = minmod; - p->sw = sw; - p->logical = logical; - p->unwind = unwind; - p->cc = cc; - p->locinput = locinput; - p->n = n; - p->ln = ln; - p->c1 = c1; - p->c2 = c2; - p->paren = paren; - p->cp = cp; - p->lastcp = lastcp; - p->oldcc = oldcc; - p->lastloc = lastloc; - p->cache_offset = cache_offset; - p->cache_bit = cache_bit; - p->curlym_l = curlym_l; - p->matches = matches; - p->maxwanted = maxwanted; - p->e = e; - p->old = old; - p->count = count; - p->cur_call_cc = cur_call_cc; - p->end_re = end_re; - p->accept_buff = accept_buff; - p->accepted = accepted; - p->reg_call_cc = PL_reg_call_cc; - - scan = new_scan; + /* 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); - minmod = 0; - sw = 0; - logical = 0; - unwind = 0; + 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. @@ -4488,14 +4368,90 @@ S_regmatch(pTHX_ regnode *prog) /*NOTREACHED*/ sayNO; -yes_loud: - DEBUG_EXECUTE_r( - PerlIO_printf(Perl_debug_log, - "%*s %scould match...%s\n", - REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5]) - ); - 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 */ + + /*XXX tmp for CURLYM*/ + regmatch_slab * const oslab = PL_regmatch_slab; + regmatch_state * const ost = st; + regmatch_state * const oys = yes_state; + int odepth = depth; + + 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 */ + { + const 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; + + case resume_IFMATCH: + if (st->logical) { + st->logical = 0; + st->sw = st->u.ifmatch.wanted; + } + else if (!st->u.ifmatch.wanted) + sayNO; + + if (OP(st->scan) == SUSPEND) + locinput = PL_reginput; + else { + locinput = PL_reginput = st->locinput; + nextchr = UCHARAT(locinput); + } + next = st->scan + ARG(st->scan); + if (next == st->scan) + next = NULL; + goto reenter; + + /* XXX tmp don't handle yes_state yet */ + case resume_CURLYM1: + case resume_CURLYM2: + case resume_CURLYM3: + PL_regmatch_slab =oslab; + st = ost; + PL_regmatch_state = st; + depth = odepth; + yes_state = oys; + DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "XXX revering a CURLYM\n")); + goto yes; + + 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: @@ -4504,7 +4460,54 @@ yes: #endif result = 1; - goto exit_level; + /* XXX this is duplicate(ish) code to that in the do_no section. + * eventually a yes should just pop the 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_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_IFMATCH: + case resume_EVAL: + default: + Perl_croak(aTHX_ "regexp resume memory corruption"); + } + } + goto final_exit; no: DEBUG_EXECUTE_r( @@ -4515,8 +4518,8 @@ no: goto do_no; no_final: do_no: - if (unwind) { - re_unwind_t * const 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: @@ -4530,10 +4533,11 @@ 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 @@ -4563,55 +4567,37 @@ do_no: } /* NOTREACHED */ } + #ifdef DEBUGGING PL_regindent--; #endif result = 0; -exit_level: - if (prev_state) { - /* restore previous state and re-enter */ - struct regmatch_state *p = prev_state; - resume_state = p->resume_state; - scan = p->scan; - next = p->next; - minmod = p->minmod; - sw = p->sw; - logical = p->logical; - unwind = p->unwind; - cc = p->cc; - locinput = p->locinput; - nextchr = UCHARAT(locinput); - n = p->n; - ln = p->ln; - c1 = p->c1; - c2 = p->c2; - paren = p->paren; - cp = p->cp; - lastcp = p->lastcp; - oldcc = p->oldcc; - lastloc = p->lastloc; - cache_offset = p->cache_offset; - cache_bit = p->cache_bit; - curlym_l = p->curlym_l; - matches = p->matches; - maxwanted = p->maxwanted; - e = p->e; - old = p->old; - count = p->count; - cur_call_cc = p->cur_call_cc; - end_re = p->end_re; - accept_buff = p->accept_buff; - accepted = p->accepted; - PL_reg_call_cc = p->reg_call_cc; - prev_state = p->prev_state; - Safefree(p); - - switch (resume_state) { + 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: @@ -4632,10 +4618,23 @@ exit_level: 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; + yes_state = st->u.yes.prev_yes_state; + if (st->logical) { + st->logical = 0; + st->sw = !st->u.ifmatch.wanted; + } + else if (st->u.ifmatch.wanted) + sayNO; + + assert(OP(scan) != SUSPEND); /* XXX DAPM tmp */ + locinput = PL_reginput = st->locinput; + nextchr = UCHARAT(locinput); + next = scan + ARG(scan); + if (next == scan) + next = NULL; + goto reenter; + case resume_PLUS1: goto resume_point_PLUS1; case resume_PLUS2: @@ -4644,13 +4643,28 @@ exit_level: goto resume_point_PLUS3; case resume_PLUS4: goto resume_point_PLUS4; - case resume_END: - goto resume_point_END; default: Perl_croak(aTHX_ "regexp resume memory corruption"); } - /* NOTREACHED */ } + +final_exit: + + /* restore original high-water mark */ + PL_regmatch_slab = orig_slab; + PL_regmatch_state = orig_state; + + /* free all slabs above current one */ + if (orig_slab->next) { + regmatch_slab *sl = orig_slab->next; + orig_slab->next = NULL; + while (sl) { + regmatch_slab * const osl = sl; + sl = sl->next; + Safefree(osl); + } + } + return result; } @@ -4664,7 +4678,7 @@ exit_level: * rather than incrementing count on every character. [Er, except utf8.]] */ STATIC I32 -S_regrepeat(pTHX_ const regnode *p, I32 max) +S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max) { dVAR; register char *scan; @@ -4677,7 +4691,7 @@ 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) { @@ -4727,12 +4741,12 @@ 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; @@ -4893,7 +4907,7 @@ S_regrepeat(pTHX_ const regnode *p, I32 max) SV * const prop = sv_newmortal(); GET_RE_DEBUG_FLAGS; DEBUG_EXECUTE_r({ - regprop(prop, p); + regprop(prog, prop, p); PerlIO_printf(Perl_debug_log, "%*s %s can match %"IVdf" times out of %"IVdf"...\n", REPORT_CODE_OFF+1, "", SvPVX_const(prop),(IV)c,(IV)max); @@ -4904,23 +4918,25 @@ S_regrepeat(pTHX_ const regnode *p, I32 max) } +#ifndef PERL_IN_XSUB_RE /* - 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; @@ -4950,6 +4966,7 @@ Perl_regclass_swash(pTHX_ register const regnode* node, bool doinit, SV** listsv return sw; } +#endif /* - reginclass - determine if a character falls into a character class @@ -4962,7 +4979,7 @@ 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); @@ -4973,8 +4990,8 @@ 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)"); } @@ -4991,7 +5008,7 @@ 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)) @@ -5085,13 +5102,6 @@ S_reginclass(pTHX_ register const regnode *n, register const U8* p, STRLEN* lenp } STATIC U8 * -S_reghop(pTHX_ U8 *s, I32 off) -{ - dVAR; - return S_reghop3(s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)); -} - -STATIC U8 * S_reghop3(U8 *s, I32 off, U8* lim) { dVAR; @@ -5117,13 +5127,6 @@ S_reghop3(U8 *s, I32 off, U8* lim) } STATIC U8 * -S_reghopmaybe(pTHX_ U8 *s, I32 off) -{ - dVAR; - return S_reghopmaybe3(s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)); -} - -STATIC U8 * S_reghopmaybe3(U8* s, I32 off, U8* lim) { dVAR; @@ -5158,15 +5161,15 @@ static void restore_pos(pTHX_ void *arg) { dVAR; - PERL_UNUSED_ARG(arg); + regexp * const rex = (regexp *)arg; if (PL_reg_eval_set) { if (PL_reg_oldsaved) { - PL_reg_re->subbeg = PL_reg_oldsaved; - PL_reg_re->sublen = PL_reg_oldsavedlen; + rex->subbeg = PL_reg_oldsaved; + rex->sublen = PL_reg_oldsavedlen; #ifdef PERL_OLD_COPY_ON_WRITE - PL_reg_re->saved_copy = PL_nrs; + rex->saved_copy = PL_nrs; #endif - RX_MATCH_COPIED_on(PL_reg_re); + RX_MATCH_COPIED_on(rex); } PL_reg_magic->mg_len = PL_reg_oldpos; PL_reg_eval_set = 0; @@ -5178,8 +5181,8 @@ 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); @@ -5187,8 +5190,8 @@ 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); @@ -5202,8 +5205,8 @@ S_to_byte_substr(pTHX_ register regexp *prog) { dVAR; if (prog->float_utf8 && !prog->float_substr) { - SV* sv; - prog->float_substr = sv = newSVsv(prog->float_utf8); + SV* sv = newSVsv(prog->float_utf8); + prog->float_substr = sv; if (sv_utf8_downgrade(sv, TRUE)) { if (SvTAIL(prog->float_utf8)) SvTAIL_on(sv); @@ -5215,8 +5218,8 @@ 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);