X-Git-Url: https://perl5.git.perl.org/perl5.git/blobdiff_plain/8ac853655d9b744749adcb9687c13d99cdd6e9fb..d37908890678cc520478783c272a10631e81a5d0:/regexec.c diff --git a/regexec.c b/regexec.c index 9e8d451..d4e1fc4 100644 --- a/regexec.c +++ b/regexec.c @@ -19,6 +19,26 @@ * with the POSIX routines of the same names. */ +#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 */ +# ifndef 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 +/* *These* symbols are masked to allow static link. */ +# define Perl_pregexec my_pregexec +#endif + /*SUPPRESS 112*/ /* * pregcomp and pregexec -- regsub and regerror are not used in perl @@ -42,7 +62,7 @@ * **** Alterations to Henry's code are... **** - **** Copyright (c) 1991-1997, Larry Wall + **** Copyright (c) 1991-1998, Larry Wall **** **** You may distribute under the terms of either the GNU General Public **** License or the Artistic License, as specified in the README file. @@ -53,61 +73,93 @@ */ #include "EXTERN.h" #include "perl.h" +typedef MAGIC *my_magic; + #include "regcomp.h" +#define RF_tainted 1 /* tainted information used? */ +#define RF_warned 2 /* warned about big count? */ +#define RF_evaled 4 /* Did an EVAL with setting? */ +#define RF_utf8 8 /* String contains multibyte chars? */ + +#define UTF (PL_reg_flags & RF_utf8) + +#define RS_init 1 /* eval environment created */ +#define RS_set 2 /* replsv value is set */ + #ifndef STATIC #define STATIC static #endif -#ifdef DEBUGGING -static I32 regnarrate = 0; -static char* regprogram = 0; -#endif +#ifndef PERL_OBJECT +typedef I32 CHECKPOINT; -/* Current curly descriptor */ -typedef struct curcur CURCUR; -struct curcur { - int parenfloor; /* how far back to strip paren data */ - int cur; /* how many instances of scan we've matched */ - int min; /* the minimal number of scans to match */ - int max; /* the maximal number of scans to match */ - int minmod; /* whether to work our way up or down */ - char * scan; /* the thing to match */ - char * next; /* what has to match after it */ - char * lastloc; /* where we started matching this scan */ - CURCUR * oldcc; /* current curly before we started this one */ -}; - -static CURCUR* regcc; +/* + * Forwards. + */ -typedef I32 CHECKPOINT; +static I32 regmatch _((regnode *prog)); +static I32 regrepeat _((regnode *p, I32 max)); +static I32 regrepeat_hard _((regnode *p, I32 max, I32 *lp)); +static I32 regtry _((regexp *prog, char *startpos)); +static bool reginclass _((char *p, I32 c)); +static bool reginclassutf8 _((regnode *f, U8* p)); static CHECKPOINT regcppush _((I32 parenfloor)); static char * regcppop _((void)); +#endif + +#define REGINCLASS(p,c) (*(p) ? reginclass(p,c) : ANYOF_TEST(p,c)) +#define REGINCLASSUTF8(f,p) (ARG1(f) ? reginclassutf8(f,p) : swash_fetch((SV*)PL_regdata->data[ARG2(f)],p)) -static CHECKPOINT +#define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv)) +#define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b) + +#ifndef PERL_OBJECT +static U8 * reghop _((U8 *pos, I32 off)); +static U8 * reghopmaybe _((U8 *pos, I32 off)); +#endif +#define reghop_c(pos,off) ((char*)reghop((U8*)pos, off)) +#define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off)) +#define HOP(pos,off) (UTF ? reghop((U8*)pos, off) : (U8*)(pos + off)) +#define HOPMAYBE(pos,off) (UTF ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off)) +#define HOPc(pos,off) ((char*)HOP(pos,off)) +#define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off)) + +STATIC CHECKPOINT regcppush(I32 parenfloor) { dTHR; - int retval = savestack_ix; - int i = (regsize - parenfloor) * 3; + int retval = PL_savestack_ix; + int i = (PL_regsize - parenfloor) * 4; int p; SSCHECK(i + 5); - for (p = regsize; p > parenfloor; p--) { - SSPUSHPTR(regendp[p]); - SSPUSHPTR(regstartp[p]); + for (p = PL_regsize; p > parenfloor; p--) { + SSPUSHPTR(PL_regendp[p]); + SSPUSHPTR(PL_regstartp[p]); + SSPUSHPTR(PL_reg_start_tmp[p]); SSPUSHINT(p); } - SSPUSHINT(regsize); - SSPUSHINT(*reglastparen); - SSPUSHPTR(reginput); + SSPUSHINT(PL_regsize); + SSPUSHINT(*PL_reglastparen); + SSPUSHPTR(PL_reginput); SSPUSHINT(i + 3); SSPUSHINT(SAVEt_REGCONTEXT); return retval; } -static char * +/* These are needed since we do not localize EVAL nodes: */ +# define REGCP_SET DEBUG_r(PerlIO_printf(Perl_debug_log, \ + " Setting an EVAL scope, savestack=%i\n", \ + PL_savestack_ix)); lastcp = PL_savestack_ix + +# define REGCP_UNWIND DEBUG_r(lastcp != PL_savestack_ix ? \ + PerlIO_printf(Perl_debug_log, \ + " Clearing an EVAL scope, savestack=%i..%i\n", \ + lastcp, PL_savestack_ix) : 0); regcpblow(lastcp) + +STATIC char * regcppop(void) { dTHR; @@ -118,98 +170,126 @@ regcppop(void) assert(i == SAVEt_REGCONTEXT); i = SSPOPINT; input = (char *) SSPOPPTR; - *reglastparen = SSPOPINT; - regsize = SSPOPINT; - for (i -= 3; i > 0; i -= 3) { + *PL_reglastparen = SSPOPINT; + PL_regsize = SSPOPINT; + for (i -= 3; i > 0; i -= 4) { paren = (U32)SSPOPINT; - regstartp[paren] = (char *) SSPOPPTR; + PL_reg_start_tmp[paren] = (char *) SSPOPPTR; + PL_regstartp[paren] = (char *) SSPOPPTR; tmps = (char*)SSPOPPTR; - if (paren <= *reglastparen) - regendp[paren] = tmps; + if (paren <= *PL_reglastparen) + PL_regendp[paren] = tmps; + DEBUG_r( + PerlIO_printf(Perl_debug_log, + " restoring \\%d to %d(%d)..%d%s\n", + paren, PL_regstartp[paren] - PL_regbol, + PL_reg_start_tmp[paren] - PL_regbol, + PL_regendp[paren] - PL_regbol, + (paren > *PL_reglastparen ? "(no)" : "")); + ); } - for (paren = *reglastparen + 1; paren <= regnpar; paren++) { - if (paren > regsize) - regstartp[paren] = Nullch; - regendp[paren] = Nullch; + DEBUG_r( + if (*PL_reglastparen + 1 <= PL_regnpar) { + PerlIO_printf(Perl_debug_log, + " restoring \\%d..\\%d to undef\n", + *PL_reglastparen + 1, PL_regnpar); + } + ); + for (paren = *PL_reglastparen + 1; paren <= PL_regnpar; paren++) { + if (paren > PL_regsize) + PL_regstartp[paren] = Nullch; + PL_regendp[paren] = Nullch; } return input; } -/* After a successful match in WHILEM, we want to restore paren matches - * that have been overwritten by a failed match attempt in the process - * of reaching this success. We do this by restoring regstartp[i] - * wherever regendp[i] has not changed; if OPEN is changed to modify - * regendp[], the '== endp' test below should be changed to match. - * This corrects the error of: - * 0 > length [ "foobar" =~ / ( (foo) | (bar) )* /x ]->[1] - */ -static void -regcppartblow(I32 base) +STATIC char * +regcp_set_to(I32 ss) { - dTHR; - I32 i = SSPOPINT; - U32 paren; - char *startp; - char *endp; - assert(i == SAVEt_REGCONTEXT); - i = SSPOPINT; - /* input, lastparen, size */ - SSPOPPTR; SSPOPINT; SSPOPINT; - for (i -= 3; i > 0; i -= 3) { - paren = (U32)SSPOPINT; - startp = (char *) SSPOPPTR; - endp = (char *) SSPOPPTR; - if (paren <= *reglastparen && regendp[paren] == endp) - regstartp[paren] = startp; - } - assert(savestack_ix == base); + I32 tmp = PL_savestack_ix; + + PL_savestack_ix = ss; + regcppop(); + PL_savestack_ix = tmp; } -#define regcpblow(cp) leave_scope(cp) +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) /* * pregexec and friends */ /* - * Forwards. + - pregexec - match a regexp against a string */ +I32 +pregexec(register regexp *prog, char *stringarg, register char *strend, + char *strbeg, I32 minend, SV *screamer, U32 nosave) +/* strend: pointer to null at end of string */ +/* strbeg: real beginning of string */ +/* minend: end of match must be >=minend after stringarg. */ +/* nosave: For optimizations. */ +{ + return + regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, + nosave ? 0 : REXEC_COPY_STR); +} -static I32 regmatch _((char *prog)); -static I32 regrepeat _((char *p, I32 max)); -static I32 regtry _((regexp *prog, char *startpos)); -static bool reginclass _((char *p, I32 c)); - -static bool regtainted; /* tainted information used? */ - +STATIC void +cache_re(regexp *prog) +{ + PL_regprecomp = prog->precomp; /* Needed for FAIL. */ +#ifdef DEBUGGING + PL_regprogram = prog->program; +#endif + PL_regnpar = prog->nparens; + PL_regdata = prog->data; + PL_reg_re = prog; +} + /* - - pregexec - match a regexp against a string + - regexec_flags - match a regexp against a string */ I32 -pregexec(register regexp *prog, char *stringarg, register char *strend, char *strbeg, I32 minend, SV *screamer, I32 safebase) - - - /* pointer to null at end of string */ - /* real beginning of string */ - /* end of match must be at least minend after stringarg */ - - /* no need to remember string in subbase */ +regexec_flags(register regexp *prog, char *stringarg, register char *strend, + char *strbeg, I32 minend, SV *screamer, void *data, U32 flags) +/* strend: pointer to null at end of string */ +/* strbeg: real beginning of string */ +/* minend: end of match must be >=minend after stringarg. */ +/* data: May be used for some additional optimizations. */ +/* nosave: For optimizations. */ { + dTHR; register char *s; - register char *c; + register regnode *c; register char *startpos = stringarg; register I32 tmp; - I32 minlen = 0; /* must match at least this many chars */ + I32 minlen; /* must match at least this many chars */ I32 dontbother = 0; /* how many characters not to try at end */ CURCUR cc; + I32 start_shift = 0; /* Offset of the start to find + constant substr. */ /* CC */ + I32 end_shift = 0; /* Same for the end. */ /* CC */ + I32 scream_pos = -1; /* Internal iterator of scream. */ + char *scream_olds; + SV* oreplsv = GvSV(PL_replgv); cc.cur = 0; cc.oldcc = 0; - regcc = &cc; + PL_regcc = &cc; + cache_re(prog); #ifdef DEBUGGING - regnarrate = debug & 512; - regprogram = prog->program; + PL_regnarrate = PL_debug & 512; #endif /* Be paranoid... */ @@ -218,68 +298,104 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st return 0; } + minlen = prog->minlen; + if (strend - startpos < minlen) goto phooey; + if (startpos == strbeg) /* is ^ valid at stringarg? */ - regprev = '\n'; + PL_regprev = '\n'; else { - regprev = stringarg[-1]; - if (!multiline && regprev == '\n') - regprev = '\0'; /* force ^ to NOT match */ + PL_regprev = (U32)stringarg[-1]; + if (!PL_multiline && PL_regprev == '\n') + PL_regprev = '\0'; /* force ^ to NOT match */ } - regprecomp = prog->precomp; /* Check validity of program. */ if (UCHARAT(prog->program) != MAGIC) { FAIL("corrupted regexp program"); } - regnpar = prog->nparens; - regtainted = FALSE; + PL_reg_flags = 0; + PL_reg_eval_set = 0; + + if (prog->reganch & ROPT_UTF8) + PL_reg_flags |= RF_utf8; + + /* Mark beginning of line for ^ and lookbehind. */ + PL_regbol = startpos; + PL_bostr = strbeg; + + /* 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; /* If there is a "must appear" string, look for it. */ s = startpos; - if (prog->regmust != Nullsv && + if (!(flags & REXEC_CHECKED) + && prog->check_substr != Nullsv && !(prog->reganch & ROPT_ANCH_GPOS) && - (!(prog->reganch & ROPT_ANCH_BOL) - || (multiline && prog->regback >= 0)) ) + (!(prog->reganch & (ROPT_ANCH_BOL | ROPT_ANCH_MBOL)) + || (PL_multiline && prog->check_substr == prog->anchored_substr)) ) { - if (stringarg == strbeg && screamer) { - if (screamfirst[BmRARE(prog->regmust)] >= 0) - s = screaminstr(screamer,prog->regmust); + char *t; + start_shift = prog->check_offset_min; /* okay to underestimate on CC */ + /* Should be nonnegative! */ + end_shift = minlen - start_shift - CHR_SVLEN(prog->check_substr); + if (screamer) { + if (PL_screamfirst[BmRARE(prog->check_substr)] >= 0) + s = screaminstr(screamer, prog->check_substr, + start_shift + (stringarg - strbeg), + end_shift, &scream_pos, 0); else s = Nullch; + scream_olds = s; } else - s = fbm_instr((unsigned char*)s, (unsigned char*)strend, - prog->regmust); + s = fbm_instr((unsigned char*)s + start_shift, + (unsigned char*)strend - end_shift, + prog->check_substr, 0); if (!s) { - ++BmUSEFUL(prog->regmust); /* hooray */ + ++BmUSEFUL(prog->check_substr); /* hooray */ goto phooey; /* not present */ } - else if (prog->regback >= 0) { - s -= prog->regback; - if (s < startpos) - s = startpos; - minlen = prog->regback + SvCUR(prog->regmust); + else if (s - stringarg > prog->check_offset_max && + (UTF + ? ((t = reghopmaybe_c(s, -(prog->check_offset_max))) && t >= stringarg) + : (t = s - prog->check_offset_max) != 0 + ) + ) + { + ++BmUSEFUL(prog->check_substr); /* hooray/2 */ + s = t; } - else if (!prog->naughty && --BmUSEFUL(prog->regmust) < 0) { /* boo */ - SvREFCNT_dec(prog->regmust); - prog->regmust = Nullsv; /* disable regmust */ + else if (!(prog->reganch & ROPT_NAUGHTY) + && --BmUSEFUL(prog->check_substr) < 0 + && prog->check_substr == prog->float_substr) { /* boo */ + SvREFCNT_dec(prog->check_substr); + prog->check_substr = Nullsv; /* disable */ + prog->float_substr = Nullsv; /* clear */ s = startpos; } - else { + else s = startpos; - minlen = SvCUR(prog->regmust); - } } - /* Mark beginning of line for ^ . */ - regbol = startpos; - - /* Mark end of line for $ (and such) */ - regeol = strend; - - /* see how far we have to get to not match where we matched before */ - regtill = startpos+minend; + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "%sMatching%s `%s%.60s%s%s' against `%s%.*s%s%s'\n", + PL_colors[4],PL_colors[5],PL_colors[0], + prog->precomp, + PL_colors[1], + (strlen(prog->precomp) > 60 ? "..." : ""), + PL_colors[0], + (strend - startpos > 60 ? 60 : strend - startpos), + startpos, PL_colors[1], + (strend - startpos > 60 ? "..." : "")) + ); /* Simplest case: anchored match need be tried only once. */ /* [unless only anchor is BOL and multiline is set] */ @@ -287,16 +403,17 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st if (regtry(prog, startpos)) goto got_it; else if (!(prog->reganch & ROPT_ANCH_GPOS) && - (multiline || (prog->reganch & ROPT_IMPLICIT))) + (PL_multiline || (prog->reganch & ROPT_IMPLICIT) + || (prog->reganch & ROPT_ANCH_MBOL))) { if (minlen) dontbother = minlen - 1; - strend -= dontbother; + strend = HOPc(strend, -dontbother); /* for multiline we only have to try after newlines */ if (s > startpos) s--; while (s < strend) { - if (*s++ == '\n') { + if (*s++ == '\n') { /* don't need utf8skip here */ if (s < strend && regtry(prog, s)) goto got_it; } @@ -306,14 +423,25 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st } /* Messy cases: unanchored match. */ - if (prog->regstart) { - if (prog->reganch & ROPT_SKIP) { /* we have /x+whatever/ */ - /* it must be a one character string */ - char ch = SvPVX(prog->regstart)[0]; + if (prog->anchored_substr && prog->reganch & ROPT_SKIP) { + /* we have /x+whatever/ */ + /* it must be a one character string */ + char ch = SvPVX(prog->anchored_substr)[0]; + if (UTF) { while (s < strend) { if (*s == ch) { - if (regtry(prog, s)) - goto got_it; + if (regtry(prog, s)) goto got_it; + s += UTF8SKIP(s); + while (s < strend && *s == ch) + s += UTF8SKIP(s); + } + s += UTF8SKIP(s); + } + } + else { + while (s < strend) { + if (*s == ch) { + if (regtry(prog, s)) goto got_it; s++; while (s < strend && *s == ch) s++; @@ -321,41 +449,92 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st s++; } } - else if (SvTYPE(prog->regstart) == SVt_PVBM) { - /* We know what string it must start with. */ - while ((s = fbm_instr((unsigned char*)s, - (unsigned char*)strend, prog->regstart)) != NULL) - { - if (regtry(prog, s)) - goto got_it; - s++; + } + /*SUPPRESS 560*/ + else if (prog->anchored_substr != Nullsv + || (prog->float_substr != Nullsv + && prog->float_max_offset < strend - s)) { + SV *must = prog->anchored_substr + ? prog->anchored_substr : prog->float_substr; + I32 back_max = + prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset; + I32 back_min = + prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset; + I32 delta = back_max - back_min; + char *last = HOPc(strend, 0-(CHR_SVLEN(must) + back_min)); /* Cannot start after this */ + char *last1; /* Last position checked before */ + + if (s > PL_bostr) + last1 = HOPc(s, -1); + else + last1 = s - 1; /* bogus */ + + /* XXXX check_substr already used to find `s', can optimize if + check_substr==must. */ + scream_pos = -1; + dontbother = end_shift; + strend = HOPc(strend, -dontbother); + while ( (s <= last) && + (screamer + ? (s = screaminstr(screamer, must, HOPc(s, back_min) - strbeg, + end_shift, &scream_pos, 0)) + : (s = fbm_instr((unsigned char*)HOP(s, back_min), + (unsigned char*)strend, must, 0))) ) { + if (HOPc(s, -back_max) > last1) { + last1 = HOPc(s, -back_min); + s = HOPc(s, -back_max); } - } - else { /* Optimized fbm_instr: */ - c = SvPVX(prog->regstart); - while ((s = ninstr(s, strend, c, c + SvCUR(prog->regstart))) != NULL) - { - if (regtry(prog, s)) - goto got_it; - s++; + else { + char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; + + last1 = HOPc(s, -back_min); + s = t; + } + if (UTF) { + while (s <= last1) { + if (regtry(prog, s)) + goto got_it; + s += UTF8SKIP(s); + } + } + else { + while (s <= last1) { + if (regtry(prog, s)) + goto got_it; + s++; + } } } goto phooey; } - /*SUPPRESS 560*/ - if (c = prog->regstclass) { + else if (c = prog->regstclass) { I32 doevery = (prog->reganch & ROPT_SKIP) == 0; + char *cc; if (minlen) dontbother = minlen - 1; - strend -= dontbother; /* don't bother with what can't match */ + strend = HOPc(strend, -dontbother); /* don't bother with what can't match */ tmp = 1; /* We know what class it must start with. */ switch (OP(c)) { + case ANYOFUTF8: + cc = (char *) OPERAND(c); + while (s < strend) { + if (REGINCLASSUTF8(c, (U8*)s)) { + if (tmp && regtry(prog, s)) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + break; case ANYOF: - c = OPERAND(c); + cc = (char *) OPERAND(c); while (s < strend) { - if (reginclass(c, *s)) { + if (REGINCLASS(cc, *s)) { if (tmp && regtry(prog, s)) goto got_it; else @@ -367,12 +546,14 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st } break; case BOUNDL: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case BOUND: - if (minlen) - dontbother++,strend--; - tmp = (s != startpos) ? UCHARAT(s - 1) : regprev; + if (minlen) { + dontbother++; + strend -= 1; + } + tmp = (s != startpos) ? UCHARAT(s - 1) : PL_regprev; tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0); while (s < strend) { if (tmp == !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) { @@ -385,13 +566,39 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st if ((minlen || tmp) && regtry(prog,s)) goto got_it; break; + case BOUNDLUTF8: + PL_reg_flags |= RF_tainted; + /* FALL THROUGH */ + case BOUNDUTF8: + if (minlen) { + dontbother++; + strend = reghop_c(strend, -1); + } + tmp = (I32)(s != startpos) ? utf8_to_uv(reghop((U8*)s, -1), 0) : PL_regprev; + tmp = ((OP(c) == BOUND ? isALNUM_uni(tmp) : isALNUM_LC_uni(tmp)) != 0); + while (s < strend) { + if (tmp == !(OP(c) == BOUND ? + swash_fetch(PL_utf8_alnum, (U8*)s) : + isALNUM_LC_utf8((U8*)s))) + { + tmp = !tmp; + if (regtry(prog, s)) + goto got_it; + } + s += UTF8SKIP(s); + } + if ((minlen || tmp) && regtry(prog,s)) + goto got_it; + break; case NBOUNDL: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case NBOUND: - if (minlen) - dontbother++,strend--; - tmp = (s != startpos) ? UCHARAT(s - 1) : regprev; + if (minlen) { + dontbother++; + strend -= 1; + } + tmp = (s != startpos) ? UCHARAT(s - 1) : PL_regprev; tmp = ((OP(c) == NBOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0); while (s < strend) { if (tmp == !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s))) @@ -403,6 +610,28 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st if ((minlen || !tmp) && regtry(prog,s)) goto got_it; break; + case NBOUNDLUTF8: + PL_reg_flags |= RF_tainted; + /* FALL THROUGH */ + case NBOUNDUTF8: + if (minlen) { + dontbother++; + strend = reghop_c(strend, -1); + } + tmp = (I32)(s != startpos) ? utf8_to_uv(reghop((U8*)s, -1), 0) : PL_regprev; + tmp = ((OP(c) == NBOUND ? isALNUM_uni(tmp) : isALNUM_LC_uni(tmp)) != 0); + while (s < strend) { + if (tmp == !(OP(c) == NBOUND ? + swash_fetch(PL_utf8_alnum, (U8*)s) : + isALNUM_LC_utf8((U8*)s))) + tmp = !tmp; + else if (regtry(prog, s)) + goto got_it; + s += UTF8SKIP(s); + } + if ((minlen || !tmp) && regtry(prog,s)) + goto got_it; + break; case ALNUM: while (s < strend) { if (isALNUM(*s)) { @@ -416,8 +645,21 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st s++; } break; + case ALNUMUTF8: + while (s < strend) { + if (swash_fetch(PL_utf8_alnum, (U8*)s)) { + if (tmp && regtry(prog, s)) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + break; case ALNUML: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; while (s < strend) { if (isALNUM_LC(*s)) { if (tmp && regtry(prog, s)) @@ -430,6 +672,20 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st s++; } break; + case ALNUMLUTF8: + PL_reg_flags |= RF_tainted; + while (s < strend) { + if (isALNUM_LC_utf8((U8*)s)) { + if (tmp && regtry(prog, s)) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + break; case NALNUM: while (s < strend) { if (!isALNUM(*s)) { @@ -443,8 +699,21 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st s++; } break; + case NALNUMUTF8: + while (s < strend) { + if (!swash_fetch(PL_utf8_alnum, (U8*)s)) { + if (tmp && regtry(prog, s)) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + break; case NALNUML: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; while (s < strend) { if (!isALNUM_LC(*s)) { if (tmp && regtry(prog, s)) @@ -457,6 +726,20 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st s++; } break; + case NALNUMLUTF8: + PL_reg_flags |= RF_tainted; + while (s < strend) { + if (!isALNUM_LC_utf8((U8*)s)) { + if (tmp && regtry(prog, s)) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + break; case SPACE: while (s < strend) { if (isSPACE(*s)) { @@ -470,8 +753,21 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st s++; } break; + case SPACEUTF8: + while (s < strend) { + if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s)) { + if (tmp && regtry(prog, s)) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + break; case SPACEL: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; while (s < strend) { if (isSPACE_LC(*s)) { if (tmp && regtry(prog, s)) @@ -484,6 +780,20 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st s++; } break; + case SPACELUTF8: + PL_reg_flags |= RF_tainted; + while (s < strend) { + if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) { + if (tmp && regtry(prog, s)) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + break; case NSPACE: while (s < strend) { if (!isSPACE(*s)) { @@ -497,8 +807,21 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st s++; } break; + case NSPACEUTF8: + while (s < strend) { + if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s))) { + if (tmp && regtry(prog, s)) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + break; case NSPACEL: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; while (s < strend) { if (!isSPACE_LC(*s)) { if (tmp && regtry(prog, s)) @@ -511,6 +834,20 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st s++; } break; + case NSPACELUTF8: + PL_reg_flags |= RF_tainted; + while (s < strend) { + if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) { + if (tmp && regtry(prog, s)) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + break; case DIGIT: while (s < strend) { if (isDIGIT(*s)) { @@ -524,6 +861,19 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st s++; } break; + case DIGITUTF8: + while (s < strend) { + if (swash_fetch(PL_utf8_digit,(U8*)s)) { + if (tmp && regtry(prog, s)) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + break; case NDIGIT: while (s < strend) { if (!isDIGIT(*s)) { @@ -537,38 +887,84 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, char *st s++; } break; + case NDIGITUTF8: + while (s < strend) { + if (!swash_fetch(PL_utf8_digit,(U8*)s)) { + if (tmp && regtry(prog, s)) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + break; } } else { - if (minlen) + dontbother = 0; + if (prog->float_substr != Nullsv) { /* Trim the end. */ + char *last; + I32 oldpos = scream_pos; + + if (screamer) { + last = screaminstr(screamer, prog->float_substr, s - strbeg, + end_shift, &scream_pos, 1); /* last one */ + if (!last) { + last = scream_olds; /* Only one occurence. */ + } + } + else { + STRLEN len; + char *little = SvPV(prog->float_substr, len); + if (len) + last = rninstr(s, strend, little, little + len); + else + last = strend; /* matching `$' */ + } + if (last == NULL) goto phooey; /* Should not happen! */ + dontbother = strend - last + prog->float_min_offset; + } + if (minlen && (dontbother < minlen)) dontbother = minlen - 1; - strend -= dontbother; + strend -= dontbother; /* this one's always in bytes! */ /* We don't know much -- general case. */ - do { - if (regtry(prog, s)) - goto got_it; - } while (s++ < strend); + if (UTF) { + for (;;) { + if (regtry(prog, s)) + goto got_it; + if (s >= strend) + break; + s += UTF8SKIP(s); + }; + } + else { + do { + if (regtry(prog, s)) + goto got_it; + } while (s++ < strend); + } } /* Failure. */ goto phooey; got_it: - strend += dontbother; /* uncheat */ prog->subbeg = strbeg; - prog->subend = strend; - prog->exec_tainted = regtainted; + prog->subend = PL_regeol; /* strend may have been modified */ + RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted); /* make sure $`, $&, $', and $digit will work later */ - if (strbeg != prog->subbase) { - if (safebase) { + if (strbeg != prog->subbase) { /* second+ //g match. */ + if (!(flags & REXEC_COPY_STR)) { if (prog->subbase) { Safefree(prog->subbase); prog->subbase = Nullch; } } else { - I32 i = strend - startpos + (stringarg - strbeg); + I32 i = PL_regeol - startpos + (stringarg - strbeg); s = savepvn(strbeg, i); Safefree(prog->subbase); prog->subbase = s; @@ -583,6 +979,12 @@ got_it: } } } + /* Preserve the current value of $^R */ + if (oreplsv != GvSV(PL_replgv)) { + sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is + restored, the value remains + the same. */ + } return 1; phooey: @@ -592,19 +994,42 @@ phooey: /* - regtry - try match at specific point */ -static I32 /* 0 failure, 1 success */ +STATIC I32 /* 0 failure, 1 success */ regtry(regexp *prog, char *startpos) { + dTHR; register I32 i; register char **sp; register char **ep; - - reginput = startpos; - regstartp = prog->startp; - regendp = prog->endp; - reglastparen = &prog->lastparen; + CHECKPOINT lastcp; + + if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) { + PL_reg_eval_set = RS_init; + DEBUG_r(DEBUG_s( + PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %i\n", + PL_stack_sp - PL_stack_base); + )); + SAVEINT(cxstack[cxstack_ix].blk_oldsp); + cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base; + /* Otherwise OP_NEXTSTATE will free whatever on stack now. */ + SAVETMPS; + /* Apparently this is not needed, judging by wantarray. */ + /* SAVEINT(cxstack[cxstack_ix].blk_gimme); + cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ + } + PL_reginput = startpos; + PL_regstartp = prog->startp; + PL_regendp = prog->endp; + PL_reglastparen = &prog->lastparen; prog->lastparen = 0; - regsize = 0; + PL_regsize = 0; + if (PL_reg_start_tmpl <= prog->nparens) { + PL_reg_start_tmpl = prog->nparens*3/2 + 3; + if(PL_reg_start_tmp) + Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*); + else + New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*); + } sp = prog->startp; ep = prog->endp; @@ -614,13 +1039,14 @@ regtry(regexp *prog, char *startpos) *ep++ = NULL; } } - if (regmatch(prog->program + 1) && reginput >= regtill) { + REGCP_SET; + if (regmatch(prog->program + 1)) { prog->startp[0] = startpos; - prog->endp[0] = reginput; + prog->endp[0] = PL_reginput; return 1; } - else - return 0; + REGCP_UNWIND; + return 0; } /* @@ -637,243 +1063,502 @@ regtry(regexp *prog, char *startpos) * maybe save a little bit of pushing and popping on the stack. It also takes * advantage of machines that use a register save mask on subroutine entry. */ -static I32 /* 0 failure, 1 success */ -regmatch(char *prog) +STATIC I32 /* 0 failure, 1 success */ +regmatch(regnode *prog) { - register char *scan; /* Current node. */ - char *next; /* Next node. */ - register I32 nextchar; + dTHR; + register regnode *scan; /* Current node. */ + regnode *next; /* Next node. */ + regnode *inner; /* Next node in internal branch. */ + register I32 nextchr; /* renamed nextchr - nextchar colides with + function of same name */ register I32 n; /* no or next */ register I32 ln; /* len or last */ register char *s; /* operand or save */ - register char *locinput = reginput; - register I32 c1, c2; /* case fold search */ - int minmod = 0; + register char *locinput = PL_reginput; + register I32 c1, c2, paren; /* case fold search, parenth */ + int minmod = 0, sw = 0, logical = 0; #ifdef DEBUGGING - static int regindent = 0; - regindent++; + PL_regindent++; #endif - nextchar = UCHARAT(locinput); + /* Note that nextchr is a byte even in UTF */ + nextchr = UCHARAT(locinput); scan = prog; while (scan != NULL) { +#define sayNO_L (logical ? (logical = 0, sw = 0, goto cont) : sayNO) #ifdef DEBUGGING -#define sayYES goto yes -#define sayNO goto no -#define saySAME(x) if (x) goto yes; else goto no - if (regnarrate) { - SV *prop = sv_newmortal(); - regprop(prop, scan); - PerlIO_printf(Perl_debug_log, "%*s%2ld%-8.8s\t<%.10s>\n", - regindent*2, "", (long)(scan - regprogram), - SvPVX(prop), locinput); - } +# define sayYES goto yes +# define sayNO goto no +# define saySAME(x) if (x) goto yes; else goto no +# define REPORT_CODE_OFF 24 #else -#define sayYES return 1 -#define sayNO return 0 -#define saySAME(x) return x +# define sayYES return 1 +# define sayNO return 0 +# define saySAME(x) return x #endif - -#ifdef REGALIGN - next = scan + NEXT(scan); + DEBUG_r( { + SV *prop = sv_newmortal(); + int docolor = *PL_colors[0]; + int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ + int l = (PL_regeol - locinput > taill ? taill : PL_regeol - locinput); + int pref_len = (locinput - PL_bostr > (5 + taill) - l + ? (5 + taill) - l : locinput - PL_bostr); + int pref0_len = pref_len - (locinput - PL_reginput); + + if (l + pref_len < (5 + taill) && l < PL_regeol - locinput) + l = ( PL_regeol - locinput > (5 + taill) - pref_len + ? (5 + taill) - pref_len : PL_regeol - locinput); + if (pref0_len < 0) + pref0_len = 0; + regprop(prop, scan); + PerlIO_printf(Perl_debug_log, + "%4i <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3d:%*s%s\n", + locinput - PL_bostr, + PL_colors[4], pref0_len, + locinput - pref_len, PL_colors[5], + PL_colors[2], pref_len - pref0_len, + locinput - pref_len + pref0_len, PL_colors[3], + (docolor ? "" : "> <"), + PL_colors[0], l, locinput, PL_colors[1], + 15 - l - pref_len + 1, + "", + scan - PL_regprogram, PL_regindent*2, "", + SvPVX(prop)); + } ); + + next = scan + NEXT_OFF(scan); if (next == scan) next = NULL; -#else - next = regnext(scan); -#endif switch (OP(scan)) { case BOL: - if (locinput == regbol - ? regprev == '\n' - : ((nextchar || locinput < regeol) && locinput[-1] == '\n') ) + if (locinput == PL_bostr + ? PL_regprev == '\n' + : (PL_multiline && + (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') ) { /* regtill = regbol; */ break; } sayNO; case MBOL: - if (locinput == regbol - ? regprev == '\n' - : ((nextchar || locinput < regeol) && locinput[-1] == '\n') ) + if (locinput == PL_bostr + ? PL_regprev == '\n' + : ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n') ) { break; } sayNO; case SBOL: - if (locinput == regbol && regprev == '\n') + if (locinput == PL_regbol && PL_regprev == '\n') break; sayNO; case GPOS: - if (locinput == regbol) + if (locinput == PL_regbol) break; sayNO; case EOL: - if (multiline) + if (PL_multiline) goto meol; else goto seol; case MEOL: meol: - if ((nextchar || locinput < regeol) && nextchar != '\n') + if ((nextchr || locinput < PL_regeol) && nextchr != '\n') sayNO; break; case SEOL: seol: - if ((nextchar || locinput < regeol) && nextchar != '\n') + if ((nextchr || locinput < PL_regeol) && nextchr != '\n') + sayNO; + if (PL_regeol - locinput > 1) sayNO; - if (regeol - locinput > 1) + break; + case EOS: + if (PL_regeol != locinput) + sayNO; + break; + case SANYUTF8: + if (nextchr & 0x80) { + locinput += utf8skip[nextchr]; + if (locinput > PL_regeol) + sayNO; + nextchr = UCHARAT(locinput); + break; + } + if (!nextchr && locinput >= PL_regeol) sayNO; + nextchr = UCHARAT(++locinput); break; case SANY: - if (!nextchar && locinput >= regeol) + if (!nextchr && locinput >= PL_regeol) sayNO; - nextchar = UCHARAT(++locinput); + nextchr = UCHARAT(++locinput); + break; + case ANYUTF8: + if (nextchr & 0x80) { + locinput += utf8skip[nextchr]; + if (locinput > PL_regeol) + sayNO; + nextchr = UCHARAT(locinput); + break; + } + if (!nextchr && locinput >= PL_regeol || nextchr == '\n') + sayNO; + nextchr = UCHARAT(++locinput); break; case ANY: - if (!nextchar && locinput >= regeol || nextchar == '\n') + if (!nextchr && locinput >= PL_regeol || nextchr == '\n') sayNO; - nextchar = UCHARAT(++locinput); + nextchr = UCHARAT(++locinput); break; case EXACT: - s = OPERAND(scan); - ln = *s++; + s = (char *) OPERAND(scan); + ln = UCHARAT(s++); /* Inline the first character, for speed. */ - if (UCHARAT(s) != nextchar) + if (UCHARAT(s) != nextchr) sayNO; - if (regeol - locinput < ln) + if (PL_regeol - locinput < ln) sayNO; if (ln > 1 && memNE(s, locinput, ln)) sayNO; locinput += ln; - nextchar = UCHARAT(locinput); + nextchr = UCHARAT(locinput); break; case EXACTFL: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case EXACTF: - s = OPERAND(scan); - ln = *s++; + s = (char *) OPERAND(scan); + ln = UCHARAT(s++); + + if (UTF) { + char *l = locinput; + char *e = s + ln; + c1 = OP(scan) == EXACTF; + while (s < e) { + if (l >= PL_regeol) + sayNO; + if (utf8_to_uv((U8*)s, 0) != (c1 ? + toLOWER_utf8((U8*)l) : + toLOWER_LC_utf8((U8*)l))) + { + sayNO; + } + s += UTF8SKIP(s); + l += UTF8SKIP(l); + } + locinput = l; + nextchr = UCHARAT(locinput); + break; + } + /* Inline the first character, for speed. */ - if (UCHARAT(s) != nextchar && + if (UCHARAT(s) != nextchr && UCHARAT(s) != ((OP(scan) == EXACTF) - ? fold : fold_locale)[nextchar]) + ? fold : fold_locale)[nextchr]) sayNO; - if (regeol - locinput < ln) + if (PL_regeol - locinput < ln) sayNO; if (ln > 1 && (OP(scan) == EXACTF ? ibcmp(s, locinput, ln) : ibcmp_locale(s, locinput, ln))) sayNO; locinput += ln; - nextchar = UCHARAT(locinput); + nextchr = UCHARAT(locinput); + break; + case ANYOFUTF8: + s = (char *) OPERAND(scan); + if (!REGINCLASSUTF8(scan, (U8*)locinput)) + sayNO; + if (locinput >= PL_regeol) + sayNO; + locinput += utf8skip[nextchr]; + nextchr = UCHARAT(locinput); break; case ANYOF: - s = OPERAND(scan); - if (nextchar < 0) - nextchar = UCHARAT(locinput); - if (!reginclass(s, nextchar)) + s = (char *) OPERAND(scan); + if (nextchr < 0) + nextchr = UCHARAT(locinput); + if (!REGINCLASS(s, nextchr)) sayNO; - if (!nextchar && locinput >= regeol) + if (!nextchr && locinput >= PL_regeol) sayNO; - nextchar = UCHARAT(++locinput); + nextchr = UCHARAT(++locinput); break; case ALNUML: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case ALNUM: - if (!nextchar) + if (!nextchr) sayNO; if (!(OP(scan) == ALNUM - ? isALNUM(nextchar) : isALNUM_LC(nextchar))) + ? isALNUM(nextchr) : isALNUM_LC(nextchr))) sayNO; - nextchar = UCHARAT(++locinput); + nextchr = UCHARAT(++locinput); + break; + case ALNUMLUTF8: + PL_reg_flags |= RF_tainted; + /* FALL THROUGH */ + case ALNUMUTF8: + if (!nextchr) + sayNO; + if (nextchr & 0x80) { + if (!(OP(scan) == ALNUMUTF8 + ? swash_fetch(PL_utf8_alnum, (U8*)locinput) + : isALNUM_LC_utf8((U8*)locinput))) + { + sayNO; + } + locinput += utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; + } + if (!(OP(scan) == ALNUMUTF8 + ? isALNUM(nextchr) : isALNUM_LC(nextchr))) + sayNO; + nextchr = UCHARAT(++locinput); break; case NALNUML: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case NALNUM: - if (!nextchar && locinput >= regeol) + if (!nextchr && locinput >= PL_regeol) sayNO; if (OP(scan) == NALNUM - ? isALNUM(nextchar) : isALNUM_LC(nextchar)) + ? isALNUM(nextchr) : isALNUM_LC(nextchr)) + sayNO; + nextchr = UCHARAT(++locinput); + break; + case NALNUMLUTF8: + PL_reg_flags |= RF_tainted; + /* FALL THROUGH */ + case NALNUMUTF8: + if (!nextchr && locinput >= PL_regeol) + sayNO; + if (nextchr & 0x80) { + if (OP(scan) == NALNUMUTF8 + ? swash_fetch(PL_utf8_alnum, (U8*)locinput) + : isALNUM_LC_utf8((U8*)locinput)) + { + sayNO; + } + locinput += utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; + } + if (OP(scan) == NALNUMUTF8 + ? isALNUM(nextchr) : isALNUM_LC(nextchr)) sayNO; - nextchar = UCHARAT(++locinput); + nextchr = UCHARAT(++locinput); break; case BOUNDL: case NBOUNDL: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case BOUND: case NBOUND: /* was last char in word? */ - ln = (locinput != regbol) ? UCHARAT(locinput - 1) : regprev; + ln = (locinput != PL_regbol) ? UCHARAT(locinput - 1) : PL_regprev; if (OP(scan) == BOUND || OP(scan) == NBOUND) { ln = isALNUM(ln); - n = isALNUM(nextchar); + n = isALNUM(nextchr); } else { ln = isALNUM_LC(ln); - n = isALNUM_LC(nextchar); + n = isALNUM_LC(nextchr); } if (((!ln) == (!n)) == (OP(scan) == BOUND || OP(scan) == BOUNDL)) sayNO; break; + case BOUNDLUTF8: + case NBOUNDLUTF8: + PL_reg_flags |= RF_tainted; + /* FALL THROUGH */ + case BOUNDUTF8: + case NBOUNDUTF8: + /* was last char in word? */ + ln = (locinput != PL_regbol) + ? utf8_to_uv(reghop((U8*)locinput, -1), 0) : PL_regprev; + if (OP(scan) == BOUNDUTF8 || OP(scan) == NBOUNDUTF8) { + ln = isALNUM_uni(ln); + n = swash_fetch(PL_utf8_alnum, (U8*)locinput); + } + else { + ln = isALNUM_LC_uni(ln); + n = isALNUM_LC_utf8((U8*)locinput); + } + if (((!ln) == (!n)) == (OP(scan) == BOUNDUTF8 || OP(scan) == BOUNDLUTF8)) + sayNO; + break; case SPACEL: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case SPACE: - if (!nextchar && locinput >= regeol) + if (!nextchr && locinput >= PL_regeol) sayNO; if (!(OP(scan) == SPACE - ? isSPACE(nextchar) : isSPACE_LC(nextchar))) + ? isSPACE(nextchr) : isSPACE_LC(nextchr))) + sayNO; + nextchr = UCHARAT(++locinput); + break; + case SPACELUTF8: + PL_reg_flags |= RF_tainted; + /* FALL THROUGH */ + case SPACEUTF8: + if (!nextchr && locinput >= PL_regeol) sayNO; - nextchar = UCHARAT(++locinput); + if (nextchr & 0x80) { + if (!(OP(scan) == SPACEUTF8 + ? swash_fetch(PL_utf8_space,(U8*)locinput) + : isSPACE_LC_utf8((U8*)locinput))) + { + sayNO; + } + locinput += utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; + } + if (!(OP(scan) == SPACEUTF8 + ? isSPACE(nextchr) : isSPACE_LC(nextchr))) + sayNO; + nextchr = UCHARAT(++locinput); break; case NSPACEL: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case NSPACE: - if (!nextchar) + if (!nextchr) sayNO; if (OP(scan) == SPACE - ? isSPACE(nextchar) : isSPACE_LC(nextchar)) + ? isSPACE(nextchr) : isSPACE_LC(nextchr)) sayNO; - nextchar = UCHARAT(++locinput); + nextchr = UCHARAT(++locinput); + break; + case NSPACELUTF8: + PL_reg_flags |= RF_tainted; + /* FALL THROUGH */ + case NSPACEUTF8: + if (!nextchr) + sayNO; + if (nextchr & 0x80) { + if (OP(scan) == NSPACEUTF8 + ? swash_fetch(PL_utf8_space,(U8*)locinput) + : isSPACE_LC_utf8((U8*)locinput)) + { + sayNO; + } + locinput += utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; + } + if (OP(scan) == NSPACEUTF8 + ? isSPACE(nextchr) : isSPACE_LC(nextchr)) + sayNO; + nextchr = UCHARAT(++locinput); break; case DIGIT: - if (!isDIGIT(nextchar)) + if (!isDIGIT(nextchr)) sayNO; - nextchar = UCHARAT(++locinput); + nextchr = UCHARAT(++locinput); + break; + case DIGITUTF8: + if (nextchr & 0x80) { + if (!(swash_fetch(PL_utf8_digit,(U8*)locinput))) + sayNO; + locinput += utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; + } + if (!isDIGIT(nextchr)) + sayNO; + nextchr = UCHARAT(++locinput); break; case NDIGIT: - if (!nextchar && locinput >= regeol) + if (!nextchr && locinput >= PL_regeol) + sayNO; + if (isDIGIT(nextchr)) sayNO; - if (isDIGIT(nextchar)) + nextchr = UCHARAT(++locinput); + break; + case NDIGITUTF8: + if (!nextchr && locinput >= PL_regeol) + sayNO; + if (nextchr & 0x80) { + if (swash_fetch(PL_utf8_digit,(U8*)locinput)) + sayNO; + locinput += utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; + } + if (isDIGIT(nextchr)) sayNO; - nextchar = UCHARAT(++locinput); + nextchr = UCHARAT(++locinput); + break; + case CLUMP: + if (locinput >= PL_regeol || swash_fetch(PL_utf8_mark,(U8*)locinput)) + sayNO; + locinput += utf8skip[nextchr]; + while (locinput < PL_regeol && swash_fetch(PL_utf8_mark,(U8*)locinput)) + locinput += UTF8SKIP(locinput); + if (locinput > PL_regeol) + sayNO; + nextchr = UCHARAT(locinput); break; case REFFL: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; /* FALL THROUGH */ - case REF: + case REF: case REFF: - n = ARG1(scan); /* which paren pair */ - s = regstartp[n]; - if (!s) - sayNO; - if (!regendp[n]) - sayNO; - if (s == regendp[n]) + n = ARG(scan); /* which paren pair */ + s = PL_regstartp[n]; + if (*PL_reglastparen < n || !s) + sayNO; /* Do not match unless seen CLOSEn. */ + if (s == PL_regendp[n]) break; + + if (UTF && OP(scan) != REF) { /* REF can do byte comparison */ + char *l = locinput; + char *e = PL_regendp[n]; + /* + * Note that we can't do the "other character" lookup trick as + * in the 8-bit case (no pun intended) because in Unicode we + * have to map both upper and title case to lower case. + */ + if (OP(scan) == REFF) { + while (s < e) { + if (l >= PL_regeol) + sayNO; + if (toLOWER_utf8((U8*)s) != toLOWER_utf8((U8*)l)) + sayNO; + s += UTF8SKIP(s); + l += UTF8SKIP(l); + } + } + else { + while (s < e) { + if (l >= PL_regeol) + sayNO; + if (toLOWER_LC_utf8((U8*)s) != toLOWER_LC_utf8((U8*)l)) + sayNO; + s += UTF8SKIP(s); + l += UTF8SKIP(l); + } + } + locinput = l; + nextchr = UCHARAT(locinput); + break; + } + /* Inline the first character, for speed. */ - if (UCHARAT(s) != nextchar && + if (UCHARAT(s) != nextchr && (OP(scan) == REF || (UCHARAT(s) != ((OP(scan) == REFF - ? fold : fold_locale)[nextchar])))) + ? fold : fold_locale)[nextchr])))) sayNO; - ln = regendp[n] - s; - if (locinput + ln > regeol) + ln = PL_regendp[n] - s; + if (locinput + ln > PL_regeol) sayNO; if (ln > 1 && (OP(scan) == REF ? memNE(s, locinput, ln) @@ -882,43 +1567,169 @@ regmatch(char *prog) : ibcmp_locale(s, locinput, ln)))) sayNO; locinput += ln; - nextchar = UCHARAT(locinput); + nextchr = UCHARAT(locinput); break; case NOTHING: + case TAIL: break; case BACK: break; + case EVAL: + { + dSP; + OP_4tree *oop = PL_op; + COP *ocurcop = PL_curcop; + SV **ocurpad = PL_curpad; + SV *ret; + + n = ARG(scan); + PL_op = (OP_4tree*)PL_regdata->data[n]; + DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%x\n", PL_op) ); + PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 1]); + + CALLRUNOPS(); /* Scalar context. */ + SPAGAIN; + ret = POPs; + PUTBACK; + + PL_op = oop; + PL_curpad = ocurpad; + PL_curcop = ocurcop; + if (logical) { + if (logical == 2) { /* Postponed subexpression. */ + regexp *re; + my_magic mg = Null(my_magic); + re_cc_state state; + CURCUR cctmp; + CHECKPOINT cp, lastcp; + + if(SvROK(ret) || SvRMAGICAL(ret)) { + SV *sv = SvROK(ret) ? SvRV(ret) : ret; + + if(SvMAGICAL(sv)) + mg = mg_find(sv, 'r'); + } + if (mg) { + re = (regexp *)mg->mg_obj; + ReREFCNT_inc(re); + } + else { + STRLEN len; + char *t = SvPV(ret, len); + PMOP pm; + char *oprecomp = PL_regprecomp; + I32 osize = PL_regsize; + I32 onpar = PL_regnpar; + + pm.op_pmflags = 0; + re = CALLREGCOMP(t, t + len, &pm); + if (!(SvFLAGS(ret) + & (SVs_TEMP | SVs_PADTMP | SVf_READONLY))) + sv_magic(ret,(SV*)ReREFCNT_inc(re),'r',0,0); + PL_regprecomp = oprecomp; + PL_regsize = osize; + PL_regnpar = onpar; + } + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "Entering embedded `%s%.60s%s%s'\n", + PL_colors[0], + re->precomp, + PL_colors[1], + (strlen(re->precomp) > 60 ? "..." : "")) + ); + state.node = next; + state.prev = PL_reg_call_cc; + state.cc = PL_regcc; + state.re = PL_reg_re; + + cctmp.cur = 0; + cctmp.oldcc = 0; + PL_regcc = &cctmp; + + cp = regcppush(0); /* Save *all* the positions. */ + REGCP_SET; + cache_re(re); + state.ss = PL_savestack_ix; + *PL_reglastparen = 0; + PL_reg_call_cc = &state; + PL_reginput = locinput; + if (regmatch(re->program + 1)) { + ReREFCNT_dec(re); + regcpblow(cp); + sayYES; + } + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "%*s failed...\n", + REPORT_CODE_OFF+PL_regindent*2, "") + ); + ReREFCNT_dec(re); + REGCP_UNWIND; + regcppop(); + PL_reg_call_cc = state.prev; + PL_regcc = state.cc; + PL_reg_re = state.re; + cache_re(PL_reg_re); + sayNO; + } + sw = SvTRUE(ret); + logical = 0; + } + else + sv_setsv(save_scalar(PL_replgv), ret); + break; + } case OPEN: - n = ARG1(scan); /* which paren pair */ - regstartp[n] = locinput; - if (n > regsize) - regsize = n; + n = ARG(scan); /* which paren pair */ + PL_reg_start_tmp[n] = locinput; + if (n > PL_regsize) + PL_regsize = n; break; case CLOSE: - n = ARG1(scan); /* which paren pair */ - regendp[n] = locinput; - if (n > *reglastparen) - *reglastparen = n; + n = ARG(scan); /* which paren pair */ + PL_regstartp[n] = PL_reg_start_tmp[n]; + PL_regendp[n] = locinput; + if (n > *PL_reglastparen) + *PL_reglastparen = n; + break; + case GROUPP: + n = ARG(scan); /* which paren pair */ + sw = (*PL_reglastparen >= n && PL_regendp[n] != NULL); + break; + case IFTHEN: + if (sw) + next = NEXTOPER(NEXTOPER(scan)); + else { + next = scan + ARG(scan); + if (OP(next) == IFTHEN) /* Fake one. */ + next = NEXTOPER(NEXTOPER(next)); + } + break; + case LOGICAL: + logical = scan->flags; break; case CURLYX: { - dTHR; CURCUR cc; - CHECKPOINT cp = savestack_ix; - cc.oldcc = regcc; - regcc = &cc; - cc.parenfloor = *reglastparen; + CHECKPOINT cp = PL_savestack_ix; + + if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ + next += ARG(next); + cc.oldcc = PL_regcc; + PL_regcc = &cc; + cc.parenfloor = *PL_reglastparen; cc.cur = -1; cc.min = ARG1(scan); cc.max = ARG2(scan); - cc.scan = NEXTOPER(scan) + 4; + cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS; cc.next = next; cc.minmod = minmod; cc.lastloc = 0; - reginput = locinput; + PL_reginput = locinput; n = regmatch(PREVOPER(next)); /* start on the WHILEM */ regcpblow(cp); - regcc = cc.oldcc; + PL_regcc = cc.oldcc; saySAME(n); } /* NOT REACHED */ @@ -932,26 +1743,40 @@ regmatch(char *prog) * that we can try again after backing off. */ - CHECKPOINT cp; - CURCUR* cc = regcc; + CHECKPOINT cp, lastcp; + CURCUR* cc = PL_regcc; + char *lastloc = cc->lastloc; /* Detection of 0-len. */ + n = cc->cur + 1; /* how many we know we matched */ - reginput = locinput; + PL_reginput = locinput; -#ifdef DEBUGGING - if (regnarrate) - PerlIO_printf(Perl_debug_log, "%*s %ld %lx\n", regindent*2, "", - (long)n, (long)cc); -#endif + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "%*s %ld out of %ld..%ld cc=%lx\n", + REPORT_CODE_OFF+PL_regindent*2, "", + (long)n, (long)cc->min, + (long)cc->max, (long)cc) + ); /* If degenerate scan matches "", assume scan done. */ if (locinput == cc->lastloc && n >= cc->min) { - regcc = cc->oldcc; - ln = regcc->cur; + PL_regcc = cc->oldcc; + ln = PL_regcc->cur; + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "%*s empty match detected, try continuation...\n", + REPORT_CODE_OFF+PL_regindent*2, "") + ); if (regmatch(cc->next)) sayYES; - regcc->cur = ln; - regcc = cc; + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "%*s failed...\n", + REPORT_CODE_OFF+PL_regindent*2, "") + ); + PL_regcc->cur = ln; + PL_regcc = cc; sayNO; } @@ -963,37 +1788,66 @@ regmatch(char *prog) if (regmatch(cc->scan)) sayYES; cc->cur = n - 1; + cc->lastloc = lastloc; + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "%*s failed...\n", + REPORT_CODE_OFF+PL_regindent*2, "") + ); sayNO; } /* Prefer next over scan for minimal matching. */ if (cc->minmod) { - regcc = cc->oldcc; - ln = regcc->cur; + PL_regcc = cc->oldcc; + ln = PL_regcc->cur; cp = regcppush(cc->parenfloor); + REGCP_SET; if (regmatch(cc->next)) { - regcppartblow(cp); + regcpblow(cp); sayYES; /* All done. */ } + REGCP_UNWIND; regcppop(); - regcc->cur = ln; - regcc = cc; - - if (n >= cc->max) /* Maximum greed exceeded? */ + PL_regcc->cur = ln; + PL_regcc = cc; + + if (n >= cc->max) { /* Maximum greed exceeded? */ + if (ckWARN(WARN_UNSAFE) && n >= REG_INFTY + && !(PL_reg_flags & RF_warned)) { + PL_reg_flags |= RF_warned; + warner(WARN_UNSAFE, "%s limit (%d) exceeded", + "Complex regular subexpression recursion", + REG_INFTY - 1); + } sayNO; + } + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "%*s trying longer...\n", + REPORT_CODE_OFF+PL_regindent*2, "") + ); /* Try scanning more and see if it helps. */ - reginput = locinput; + PL_reginput = locinput; cc->cur = n; cc->lastloc = locinput; cp = regcppush(cc->parenfloor); + REGCP_SET; if (regmatch(cc->scan)) { - regcppartblow(cp); + regcpblow(cp); sayYES; } + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "%*s failed...\n", + REPORT_CODE_OFF+PL_regindent*2, "") + ); + REGCP_UNWIND; regcppop(); cc->cur = n - 1; + cc->lastloc = lastloc; sayNO; } @@ -1003,48 +1857,81 @@ regmatch(char *prog) cp = regcppush(cc->parenfloor); cc->cur = n; cc->lastloc = locinput; + REGCP_SET; if (regmatch(cc->scan)) { - regcppartblow(cp); + regcpblow(cp); sayYES; } + REGCP_UNWIND; regcppop(); /* Restore some previous $s? */ - reginput = locinput; + PL_reginput = locinput; + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "%*s failed, try continuation...\n", + REPORT_CODE_OFF+PL_regindent*2, "") + ); + } + if (ckWARN(WARN_UNSAFE) && n >= REG_INFTY + && !(PL_reg_flags & RF_warned)) { + PL_reg_flags |= RF_warned; + warner(WARN_UNSAFE, "%s limit (%d) exceeded", + "Complex regular subexpression recursion", + REG_INFTY - 1); } /* Failed deeper matches of scan, so see if this one works. */ - regcc = cc->oldcc; - ln = regcc->cur; + PL_regcc = cc->oldcc; + ln = PL_regcc->cur; if (regmatch(cc->next)) sayYES; - regcc->cur = ln; - regcc = cc; + DEBUG_r( + PerlIO_printf(Perl_debug_log, "%*s failed...\n", + REPORT_CODE_OFF+PL_regindent*2, "") + ); + PL_regcc->cur = ln; + PL_regcc = cc; cc->cur = n - 1; + cc->lastloc = lastloc; sayNO; } /* NOT REACHED */ - case BRANCH: { - if (OP(next) != BRANCH) /* No choice. */ - next = NEXTOPER(scan);/* Avoid recursion. */ + case BRANCHJ: + next = scan + ARG(scan); + if (next == scan) + next = NULL; + inner = NEXTOPER(NEXTOPER(scan)); + goto do_branch; + case BRANCH: + inner = NEXTOPER(scan); + do_branch: + { + CHECKPOINT lastcp; + c1 = OP(scan); + if (OP(next) != c1) /* No choice. */ + next = inner; /* Avoid recursion. */ else { - int lastparen = *reglastparen; + int lastparen = *PL_reglastparen; + + REGCP_SET; do { - reginput = locinput; - if (regmatch(NEXTOPER(scan))) + PL_reginput = locinput; + if (regmatch(inner)) sayYES; - for (n = *reglastparen; n > lastparen; n--) - regendp[n] = 0; - *reglastparen = n; - -#ifdef REGALIGN + REGCP_UNWIND; + for (n = *PL_reglastparen; n > lastparen; n--) + PL_regendp[n] = 0; + *PL_reglastparen = n; + scan = next; /*SUPPRESS 560*/ - if (n = NEXT(scan)) - scan += n; + if (n = (c1 == BRANCH ? NEXT_OFF(next) : ARG(next))) + next += n; else - scan = NULL; -#else - scan = regnext(scan); -#endif - } while (scan != NULL && OP(scan) == BRANCH); + next = NULL; + inner = NEXTOPER(scan); + if (c1 == BRANCHJ) { + inner = NEXTOPER(inner); + } + } while (scan != NULL && OP(scan) == c1); sayNO; /* NOTREACHED */ } @@ -1053,25 +1940,171 @@ regmatch(char *prog) case MINMOD: minmod = 1; break; + case CURLYM: + { + I32 l = 0; + CHECKPOINT lastcp; + + /* We suppose that the next guy does not need + backtracking: in particular, it is of constant length, + and has no parenths to influence future backrefs. */ + 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 > *PL_reglastparen) + *PL_reglastparen = paren; + } + scan = NEXTOPER(scan) + NODE_STEP_REGNODE; + if (paren) + scan += NEXT_OFF(scan); /* Skip former OPEN. */ + PL_reginput = locinput; + if (minmod) { + minmod = 0; + if (ln && regrepeat_hard(scan, ln, &l) < ln) + sayNO; + if (ln && l == 0 && n >= ln + /* In fact, this is tricky. If paren, then the + fact that we did/didnot match may influence + future execution. */ + && !(paren && ln == 0)) + ln = n; + locinput = PL_reginput; + if (regkind[(U8)OP(next)] == EXACT) { + c1 = UCHARAT(OPERAND(next) + 1); + if (OP(next) == EXACTF) + c2 = fold[c1]; + else if (OP(next) == EXACTFL) + c2 = fold_locale[c1]; + else + c2 = c1; + } + else + c1 = c2 = -1000; + REGCP_SET; + /* This may be improved if l == 0. */ + while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */ + /* If it could work, try it. */ + if (c1 == -1000 || + UCHARAT(PL_reginput) == c1 || + UCHARAT(PL_reginput) == c2) + { + if (paren) { + if (n) { + PL_regstartp[paren] = HOPc(PL_reginput, -l); + PL_regendp[paren] = PL_reginput; + } + else + PL_regendp[paren] = NULL; + } + if (regmatch(next)) + sayYES; + REGCP_UNWIND; + } + /* Couldn't or didn't -- move forward. */ + PL_reginput = locinput; + if (regrepeat_hard(scan, 1, &l)) { + ln++; + locinput = PL_reginput; + } + else + sayNO; + } + } + else { + n = regrepeat_hard(scan, n, &l); + if (n != 0 && l == 0 + /* In fact, this is tricky. If paren, then the + fact that we did/didnot match may influence + future execution. */ + && !(paren && ln == 0)) + ln = n; + locinput = PL_reginput; + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "%*s matched %ld times, len=%ld...\n", + REPORT_CODE_OFF+PL_regindent*2, "", n, l) + ); + if (n >= ln) { + if (regkind[(U8)OP(next)] == EXACT) { + c1 = UCHARAT(OPERAND(next) + 1); + if (OP(next) == EXACTF) + c2 = fold[c1]; + else if (OP(next) == EXACTFL) + c2 = fold_locale[c1]; + else + c2 = c1; + } + else + c1 = c2 = -1000; + } + REGCP_SET; + while (n >= ln) { + /* If it could work, try it. */ + if (c1 == -1000 || + UCHARAT(PL_reginput) == c1 || + UCHARAT(PL_reginput) == c2) + { + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "%*s trying tail with n=%ld...\n", + REPORT_CODE_OFF+PL_regindent*2, "", n) + ); + if (paren) { + if (n) { + PL_regstartp[paren] = HOPc(PL_reginput, -l); + PL_regendp[paren] = PL_reginput; + } + else + PL_regendp[paren] = NULL; + } + if (regmatch(next)) + sayYES; + REGCP_UNWIND; + } + /* Couldn't or didn't -- back up. */ + n--; + locinput = HOPc(locinput, -l); + PL_reginput = locinput; + } + } + sayNO; + break; + } + case CURLYN: + paren = scan->flags; /* Which paren to set */ + if (paren > PL_regsize) + PL_regsize = paren; + if (paren > *PL_reglastparen) + *PL_reglastparen = paren; + 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 */ n = ARG2(scan); /* max to match */ - scan = NEXTOPER(scan) + 4; + scan = NEXTOPER(scan) + NODE_STEP_REGNODE; goto repeat; case STAR: ln = 0; - n = 32767; + n = REG_INFTY; scan = NEXTOPER(scan); + paren = 0; goto repeat; case PLUS: + ln = 1; + n = REG_INFTY; + scan = NEXTOPER(scan); + paren = 0; + repeat: /* * Lookahead to avoid useless match attempts * when we know what character comes next. */ - ln = 1; - n = 32767; - scan = NEXTOPER(scan); - repeat: if (regkind[(U8)OP(next)] == EXACT) { c1 = UCHARAT(OPERAND(next) + 1); if (OP(next) == EXACTF) @@ -1083,69 +2116,191 @@ regmatch(char *prog) } else c1 = c2 = -1000; - reginput = locinput; + PL_reginput = locinput; if (minmod) { + CHECKPOINT lastcp; minmod = 0; if (ln && regrepeat(scan, ln) < ln) sayNO; - while (n >= ln || (n == 32767 && ln > 0)) { /* ln overflow ? */ + locinput = PL_reginput; + REGCP_SET; + while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */ /* If it could work, try it. */ if (c1 == -1000 || - UCHARAT(reginput) == c1 || - UCHARAT(reginput) == c2) + UCHARAT(PL_reginput) == c1 || + UCHARAT(PL_reginput) == c2) { + if (paren) { + if (n) { + PL_regstartp[paren] = HOPc(PL_reginput, -1); + PL_regendp[paren] = PL_reginput; + } + else + PL_regendp[paren] = NULL; + } if (regmatch(next)) sayYES; + REGCP_UNWIND; } - /* Couldn't or didn't -- back up. */ - reginput = locinput + ln; + /* Couldn't or didn't -- move forward. */ + PL_reginput = locinput; if (regrepeat(scan, 1)) { ln++; - reginput = locinput + ln; + locinput = PL_reginput; } else sayNO; } } else { + CHECKPOINT lastcp; n = regrepeat(scan, n); + locinput = PL_reginput; if (ln < n && regkind[(U8)OP(next)] == EOL && - (!multiline || OP(next) == SEOL)) + (!PL_multiline || OP(next) == SEOL)) ln = n; /* why back off? */ - while (n >= ln) { - /* If it could work, try it. */ - if (c1 == -1000 || - UCHARAT(reginput) == c1 || - UCHARAT(reginput) == c2) - { - if (regmatch(next)) - sayYES; + REGCP_SET; + if (paren) { + while (n >= ln) { + /* If it could work, try it. */ + if (c1 == -1000 || + UCHARAT(PL_reginput) == c1 || + UCHARAT(PL_reginput) == c2) + { + if (paren && n) { + if (n) { + PL_regstartp[paren] = HOPc(PL_reginput, -1); + PL_regendp[paren] = PL_reginput; + } + else + PL_regendp[paren] = NULL; + } + if (regmatch(next)) + sayYES; + REGCP_UNWIND; + } + /* Couldn't or didn't -- back up. */ + n--; + PL_reginput = locinput = HOPc(locinput, -1); + } + } + else { + while (n >= ln) { + /* If it could work, try it. */ + if (c1 == -1000 || + UCHARAT(PL_reginput) == c1 || + UCHARAT(PL_reginput) == c2) + { + if (regmatch(next)) + sayYES; + REGCP_UNWIND; + } + /* Couldn't or didn't -- back up. */ + n--; + PL_reginput = locinput = HOPc(locinput, -1); } - /* Couldn't or didn't -- back up. */ - n--; - reginput = locinput + n; } } sayNO; - case SUCCEED: + break; case END: - reginput = locinput; /* put where regtry can find it */ - sayYES; /* Success! */ - case IFMATCH: - reginput = locinput; - scan = NEXTOPER(scan); - if (!regmatch(scan)) + if (PL_reg_call_cc) { + re_cc_state *cur_call_cc = PL_reg_call_cc; + CURCUR *cctmp = PL_regcc; + regexp *re = PL_reg_re; + CHECKPOINT cp, lastcp; + + cp = regcppush(0); /* Save *all* the positions. */ + REGCP_SET; + regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of + the caller. */ + PL_reginput = locinput; /* Make position available to + the callcc. */ + cache_re(PL_reg_call_cc->re); + PL_regcc = PL_reg_call_cc->cc; + PL_reg_call_cc = PL_reg_call_cc->prev; + if (regmatch(cur_call_cc->node)) { + PL_reg_call_cc = cur_call_cc; + regcpblow(cp); + sayYES; + } + REGCP_UNWIND; + regcppop(); + PL_reg_call_cc = cur_call_cc; + PL_regcc = cctmp; + PL_reg_re = re; + cache_re(re); + + DEBUG_r( + PerlIO_printf(Perl_debug_log, + "%*s continuation failed...\n", + REPORT_CODE_OFF+PL_regindent*2, "") + ); sayNO; - break; + } + if (locinput < PL_regtill) + sayNO; /* Cannot match: too short. */ + /* Fall through */ + case SUCCEED: + PL_reginput = locinput; /* put where regtry can find it */ + sayYES; /* Success! */ + case SUSPEND: + n = 1; + PL_reginput = locinput; + goto do_ifmatch; case UNLESSM: - reginput = locinput; - scan = NEXTOPER(scan); - if (regmatch(scan)) - sayNO; + n = 0; + if (scan->flags) { + s = HOPMAYBEc(locinput, -scan->flags); + if (!s) + goto say_yes; + PL_reginput = s; + } + else + PL_reginput = locinput; + goto do_ifmatch; + case IFMATCH: + n = 1; + if (scan->flags) { + s = HOPMAYBEc(locinput, -scan->flags); + if (!s) + goto say_no; + PL_reginput = s; + } + else + PL_reginput = locinput; + + do_ifmatch: + inner = NEXTOPER(NEXTOPER(scan)); + if (regmatch(inner) != 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. */ + case LONGJMP: + do_longjump: + next = scan + ARG(scan); + if (next == scan) + next = NULL; break; default: PerlIO_printf(PerlIO_stderr(), "%lx %d\n", - (unsigned long)scan, scan[1]); + (unsigned long)scan, OP(scan)); FAIL("regexp memory corruption"); } scan = next; @@ -1161,13 +2316,13 @@ regmatch(char *prog) yes: #ifdef DEBUGGING - regindent--; + PL_regindent--; #endif return 1; no: #ifdef DEBUGGING - regindent--; + PL_regindent--; #endif return 0; } @@ -1178,20 +2333,22 @@ no: /* * [This routine now assumes that it will only match on things of length 1. * That was true before, but now we assume scan - reginput is the count, - * rather than incrementing count on every character.] + * rather than incrementing count on every character. [Er, except utf8.]] */ -static I32 -regrepeat(char *p, I32 max) +STATIC I32 +regrepeat(regnode *p, I32 max) { + dTHR; register char *scan; register char *opnd; register I32 c; - register char *loceol = regeol; + register char *loceol = PL_regeol; + register I32 hardcount = 0; - scan = reginput; - if (max != 32767 && max < loceol - scan) + scan = PL_reginput; + if (max != REG_INFTY && max < loceol - scan) loceol = scan + max; - opnd = OPERAND(p); + opnd = (char *) OPERAND(p); switch (OP(p)) { case ANY: while (scan < loceol && *scan != '\n') @@ -1200,6 +2357,20 @@ regrepeat(char *p, I32 max) case SANY: scan = loceol; break; + case ANYUTF8: + loceol = PL_regeol; + while (scan < loceol && *scan != '\n') { + scan += UTF8SKIP(scan); + hardcount++; + } + break; + case SANYUTF8: + loceol = PL_regeol; + while (scan < loceol) { + scan += UTF8SKIP(scan); + hardcount++; + } + break; case EXACT: /* length of string is 1 */ c = UCHARAT(++opnd); while (scan < loceol && UCHARAT(scan) == c) @@ -1212,97 +2383,247 @@ regrepeat(char *p, I32 max) scan++; break; case EXACTFL: /* length of string is 1 */ - regtainted = TRUE; + PL_reg_flags |= RF_tainted; c = UCHARAT(++opnd); while (scan < loceol && (UCHARAT(scan) == c || UCHARAT(scan) == fold_locale[c])) scan++; break; + case ANYOFUTF8: + loceol = PL_regeol; + while (scan < loceol && REGINCLASSUTF8(p, (U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + break; case ANYOF: - while (scan < loceol && reginclass(opnd, *scan)) + while (scan < loceol && REGINCLASS(opnd, *scan)) scan++; break; case ALNUM: while (scan < loceol && isALNUM(*scan)) scan++; break; + case ALNUMUTF8: + loceol = PL_regeol; + while (scan < loceol && swash_fetch(PL_utf8_alnum, (U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + break; case ALNUML: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; while (scan < loceol && isALNUM_LC(*scan)) scan++; break; + case ALNUMLUTF8: + PL_reg_flags |= RF_tainted; + loceol = PL_regeol; + while (scan < loceol && isALNUM_LC_utf8((U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + break; + break; case NALNUM: while (scan < loceol && !isALNUM(*scan)) scan++; break; + case NALNUMUTF8: + loceol = PL_regeol; + while (scan < loceol && !swash_fetch(PL_utf8_alnum, (U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + break; case NALNUML: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; while (scan < loceol && !isALNUM_LC(*scan)) scan++; break; + case NALNUMLUTF8: + PL_reg_flags |= RF_tainted; + loceol = PL_regeol; + while (scan < loceol && !isALNUM_LC_utf8((U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + break; case SPACE: while (scan < loceol && isSPACE(*scan)) scan++; break; + case SPACEUTF8: + loceol = PL_regeol; + while (scan < loceol && (*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan))) { + scan += UTF8SKIP(scan); + hardcount++; + } + break; case SPACEL: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; while (scan < loceol && isSPACE_LC(*scan)) scan++; break; + case SPACELUTF8: + PL_reg_flags |= RF_tainted; + loceol = PL_regeol; + while (scan < loceol && (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) { + scan += UTF8SKIP(scan); + hardcount++; + } + break; case NSPACE: while (scan < loceol && !isSPACE(*scan)) scan++; break; + case NSPACEUTF8: + loceol = PL_regeol; + while (scan < loceol && !(*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan))) { + scan += UTF8SKIP(scan); + hardcount++; + } + break; case NSPACEL: - regtainted = TRUE; + PL_reg_flags |= RF_tainted; while (scan < loceol && !isSPACE_LC(*scan)) scan++; break; + case NSPACELUTF8: + PL_reg_flags |= RF_tainted; + loceol = PL_regeol; + while (scan < loceol && !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) { + scan += UTF8SKIP(scan); + hardcount++; + } + break; case DIGIT: while (scan < loceol && isDIGIT(*scan)) scan++; break; + case DIGITUTF8: + loceol = PL_regeol; + while (scan < loceol && swash_fetch(PL_utf8_digit,(U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + break; + break; case NDIGIT: while (scan < loceol && !isDIGIT(*scan)) scan++; break; + case NDIGITUTF8: + loceol = PL_regeol; + while (scan < loceol && !swash_fetch(PL_utf8_digit,(U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + break; default: /* Called on something of 0 width. */ break; /* So match right here or not at all. */ } - c = scan - reginput; - reginput = scan; + if (hardcount) + c = hardcount; + else + c = scan - PL_reginput; + PL_reginput = scan; + DEBUG_r( + { + SV *prop = sv_newmortal(); + + regprop(prop, p); + PerlIO_printf(Perl_debug_log, + "%*s %s can match %ld times out of %ld...\n", + REPORT_CODE_OFF+1, "", SvPVX(prop),c,max); + }); + return(c); } /* + - regrepeat_hard - repeatedly match something, report total lenth and length + * + * The repeater is supposed to have constant length. + */ + +STATIC I32 +regrepeat_hard(regnode *p, I32 max, I32 *lp) +{ + dTHR; + register char *scan; + register char *start; + register char *loceol = PL_regeol; + I32 l = 0; + I32 count = 0, res = 1; + + if (!max) + return 0; + + start = PL_reginput; + if (UTF) { + while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) { + if (!count++) { + l = 0; + while (start < PL_reginput) { + l++; + start += UTF8SKIP(start); + } + *lp = l; + if (l == 0) + return max; + } + if (count == max) + return count; + } + } + else { + while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) { + if (!count++) { + *lp = l = PL_reginput - start; + if (max != REG_INFTY && l*max < loceol - scan) + loceol = scan + l*max; + if (l == 0) + return max; + } + } + } + if (!res) + PL_reginput = scan; + + return count; +} + +/* - regclass - determine if a character falls into a character class */ -static bool +STATIC bool reginclass(register char *p, register I32 c) { + dTHR; char flags = *p; bool match = FALSE; c &= 0xFF; - if (p[1 + (c >> 3)] & (1 << (c & 7))) + if (ANYOF_TEST(p, c)) match = TRUE; else if (flags & ANYOF_FOLD) { I32 cf; if (flags & ANYOF_LOCALE) { - regtainted = TRUE; + PL_reg_flags |= RF_tainted; cf = fold_locale[c]; } else cf = fold[c]; - if (p[1 + (cf >> 3)] & (1 << (cf & 7))) + if (ANYOF_TEST(p, cf)) match = TRUE; } if (!match && (flags & ANYOF_ISA)) { - regtainted = TRUE; + PL_reg_flags |= RF_tainted; if (((flags & ANYOF_ALNUML) && isALNUM_LC(c)) || ((flags & ANYOF_NALNUML) && !isALNUM_LC(c)) || @@ -1313,33 +2634,93 @@ reginclass(register char *p, register I32 c) } } - return match ^ ((flags & ANYOF_INVERT) != 0); + return (flags & ANYOF_INVERT) ? !match : match; } -/* - - regnext - dig the "next" pointer out of a node - * - * [Note, when REGALIGN is defined there are two places in regmatch() - * that bypass this code for speed.] - */ -char * -regnext(register char *p) -{ - register I32 offset; +STATIC bool +reginclassutf8(regnode *f, U8 *p) +{ + dTHR; + char flags = ARG1(f); + bool match = FALSE; + SV *sv = (SV*)PL_regdata->data[ARG2(f)]; - if (p == ®dummy) - return(NULL); + if (swash_fetch(sv, p)) + match = TRUE; + else if (flags & ANYOF_FOLD) { + I32 cf; + U8 tmpbuf[10]; + if (flags & ANYOF_LOCALE) { + PL_reg_flags |= RF_tainted; + uv_to_utf8(tmpbuf, toLOWER_LC_utf8(p)); + } + else + uv_to_utf8(tmpbuf, toLOWER_utf8(p)); + if (swash_fetch(sv, tmpbuf)) + match = TRUE; + } - offset = NEXT(p); - if (offset == 0) - return(NULL); + if (!match && (flags & ANYOF_ISA)) { + PL_reg_flags |= RF_tainted; -#ifdef REGALIGN - return(p+offset); -#else - if (OP(p) == BACK) - return(p-offset); - else - return(p+offset); -#endif + if (((flags & ANYOF_ALNUML) && isALNUM_LC_utf8(p)) || + ((flags & ANYOF_NALNUML) && !isALNUM_LC_utf8(p)) || + ((flags & ANYOF_SPACEL) && isSPACE_LC_utf8(p)) || + ((flags & ANYOF_NSPACEL) && !isSPACE_LC_utf8(p))) + { + match = TRUE; + } + } + + return (flags & ANYOF_INVERT) ? !match : match; +} + +STATIC U8 * +reghop(U8 *s, I32 off) +{ + dTHR; + if (off >= 0) { + while (off-- && s < (U8*)PL_regeol) + s += UTF8SKIP(s); + } + else { + while (off++) { + if (s > (U8*)PL_bostr) { + s--; + if (*s & 0x80) { + while (s > (U8*)PL_bostr && (*s & 0xc0) == 0x80) + s--; + } /* XXX could check well-formedness here */ + } + } + } + return s; +} + +STATIC U8 * +reghopmaybe(U8* s, I32 off) +{ + dTHR; + if (off >= 0) { + while (off-- && s < (U8*)PL_regeol) + s += UTF8SKIP(s); + if (off >= 0) + return 0; + } + else { + while (off++) { + if (s > (U8*)PL_bostr) { + s--; + if (*s & 0x80) { + while (s > (U8*)PL_bostr && (*s & 0xc0) == 0x80) + s--; + } /* XXX could check well-formedness here */ + } + else + break; + } + if (off <= 0) + return 0; + } + return s; }