X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=regexec.c;h=f53567e9a6c92521a46914d32b2fd65f2bc16e63;hb=47cadb06ccb7dbe87fdda243544a1ecb06fb104f;hp=2c10383854e648ce6066b62a09c2a76d4c99b074;hpb=533c011aecf9bca2c9ad025efccd7b74ad222cda;p=p5sagit%2Fp5-mst-13.2.git diff --git a/regexec.c b/regexec.c index 2c10383..f53567e 100644 --- a/regexec.c +++ b/regexec.c @@ -37,6 +37,7 @@ # define Perl_regprop my_regprop /* *These* symbols are masked to allow static link. */ # define Perl_pregexec my_pregexec +# define Perl_reginitcolors my_reginitcolors #endif /*SUPPRESS 112*/ @@ -62,7 +63,7 @@ * **** 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. @@ -73,11 +74,15 @@ */ #include "EXTERN.h" #include "perl.h" + #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 */ @@ -99,10 +104,30 @@ 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)); +static char * regcp_set_to _((I32 ss)); +static void cache_re _((regexp *prog)); +static void restore_pos _((void *arg)); #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)) + +#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) @@ -181,6 +206,27 @@ regcppop(void) return input; } +STATIC char * +regcp_set_to(I32 ss) +{ + dTHR; + I32 tmp = PL_savestack_ix; + + PL_savestack_ix = ss; + regcppop(); + PL_savestack_ix = tmp; + return Nullch; +} + +typedef struct re_cc_state +{ + I32 ss; + regnode *node; + struct re_cc_state *prev; + CURCUR *cc; + regexp *re; +} re_cc_state; + #define regcpblow(cp) LEAVE_SCOPE(cp) /* @@ -202,13 +248,38 @@ pregexec(register regexp *prog, char *stringarg, register char *strend, regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, nosave ? 0 : REXEC_COPY_STR); } - + +STATIC void +cache_re(regexp *prog) +{ + dTHR; + 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; +} + +STATIC void +restore_pos(void *arg) +{ + dTHR; + if (PL_reg_eval_set) { + PL_reg_magic->mg_len = PL_reg_oldpos; + PL_reg_eval_set = 0; + PL_curpm = PL_reg_oldcurpm; + } +} + + /* - regexec_flags - match a regexp against a string */ I32 regexec_flags(register regexp *prog, char *stringarg, register char *strend, - char *strbeg, I32 minend, SV *screamer, void *data, U32 flags) + char *strbeg, I32 minend, SV *sv, 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. */ @@ -224,8 +295,8 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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. */ - I32 end_shift = 0; /* Same for the end. */ + 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); @@ -234,10 +305,9 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, cc.oldcc = 0; PL_regcc = &cc; - PL_regprecomp = prog->precomp; /* Needed for error messages. */ + cache_re(prog); #ifdef DEBUGGING PL_regnarrate = PL_debug & 512; - PL_regprogram = prog->program; #endif /* Be paranoid... */ @@ -252,20 +322,36 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, if (startpos == strbeg) /* is ^ valid at stringarg? */ PL_regprev = '\n'; else { - PL_regprev = stringarg[-1]; + PL_regprev = (U32)stringarg[-1]; if (!PL_multiline && PL_regprev == '\n') PL_regprev = '\0'; /* force ^ to NOT match */ } /* Check validity of program. */ - if (UCHARAT(prog->program) != MAGIC) { - FAIL("corrupted regexp program"); + if (UCHARAT(prog->program) != REG_MAGIC) { + croak("corrupted regexp program"); } - PL_regnpar = prog->nparens; 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; + PL_reg_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; + /* If there is a "must appear" string, look for it. */ s = startpos; if (!(flags & REXEC_CHECKED) @@ -274,12 +360,13 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, (!(prog->reganch & (ROPT_ANCH_BOL | ROPT_ANCH_MBOL)) || (PL_multiline && prog->check_substr == prog->anchored_substr)) ) { - start_shift = prog->check_offset_min; + char *t; + start_shift = prog->check_offset_min; /* okay to underestimate on CC */ /* Should be nonnegative! */ - end_shift = minlen - start_shift - SvCUR(prog->check_substr); - if (screamer) { + end_shift = minlen - start_shift - CHR_SVLEN(prog->check_substr); + if (flags & REXEC_SCREAM) { if (PL_screamfirst[BmRARE(prog->check_substr)] >= 0) - s = screaminstr(screamer, prog->check_substr, + s = screaminstr(sv, prog->check_substr, start_shift + (stringarg - strbeg), end_shift, &scream_pos, 0); else @@ -293,62 +380,78 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, if (!s) { ++BmUSEFUL(prog->check_substr); /* hooray */ goto phooey; /* not present */ - } else if ((s - stringarg) > prog->check_offset_max) { + } + 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 -= prog->check_offset_max; - } else if (!prog->naughty + s = t; + } + 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 s = startpos; + } + else + s = startpos; } - /* 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; - - DEBUG_r( - PerlIO_printf(Perl_debug_log, - "Matching `%.60s%s' against `%.*s%s'\n", - prog->precomp, + DEBUG_r(if (!PL_colorset) reginitcolors()); + 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, + startpos, PL_colors[1], (strend - startpos > 60 ? "..." : "")) ); + if (prog->reganch & ROPT_GPOS_SEEN) { + MAGIC *mg; + + if (!(flags & REXEC_IGNOREPOS) && sv && SvTYPE(sv) >= SVt_PVMG + && SvMAGIC(sv) && (mg = mg_find(sv, 'g')) && mg->mg_len >= 0) + PL_reg_ganch = strbeg + mg->mg_len; + else + PL_reg_ganch = startpos; + } + /* Simplest case: anchored match need be tried only once. */ /* [unless only anchor is BOL and multiline is set] */ - if (prog->reganch & ROPT_ANCH) { + if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) { if (regtry(prog, startpos)) goto got_it; - else if (!(prog->reganch & ROPT_ANCH_GPOS) && - (PL_multiline || (prog->reganch & ROPT_IMPLICIT) - || (prog->reganch & ROPT_ANCH_MBOL))) + else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT) + || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */ { 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 PL_utf8skip here */ if (s < strend && regtry(prog, s)) goto got_it; } } } goto phooey; + } else if (prog->reganch & ROPT_ANCH_GPOS) { + if (regtry(prog, PL_reg_ganch)) + goto got_it; + goto phooey; } /* Messy cases: unanchored match. */ @@ -356,14 +459,27 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, /* we have /x+whatever/ */ /* it must be a one character string */ char ch = SvPVX(prog->anchored_substr)[0]; - while (s < strend) { - if (*s == ch) { - if (regtry(prog, s)) goto got_it; - s++; - while (s < strend && *s == ch) + if (UTF) { + while (s < strend) { + if (*s == ch) { + 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++; + } + s++; } - s++; } } /*SUPPRESS 560*/ @@ -377,50 +493,80 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, I32 back_min = prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset; I32 delta = back_max - back_min; - char *last = strend - SvCUR(must) - back_min; /* Cannot start after this */ - char *last1 = s - 1; /* Last position checked before */ + 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 -= dontbother; + strend = HOPc(strend, -dontbother); while ( (s <= last) && - (screamer - ? (s = screaminstr(screamer, must, s + back_min - strbeg, + ((flags & REXEC_SCREAM) + ? (s = screaminstr(sv, must, HOPc(s, back_min) - strbeg, end_shift, &scream_pos, 0)) - : (s = fbm_instr((unsigned char*)s + back_min, + : (s = fbm_instr((unsigned char*)HOP(s, back_min), (unsigned char*)strend, must, 0))) ) { - if (s - back_max > last1) { - last1 = s - back_min; - s = s - back_max; - } else { - char *t = last1 + 1; + if (HOPc(s, -back_max) > last1) { + last1 = HOPc(s, -back_min); + s = HOPc(s, -back_max); + } + else { + char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; - last1 = s - back_min; + last1 = HOPc(s, -back_min); s = t; } - while (s <= last1) { - if (regtry(prog, s)) - goto got_it; - s++; + 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; - } else if (c = prog->regstclass) { + } + else if (c = prog->regstclass) { I32 doevery = (prog->reganch & ROPT_SKIP) == 0; - char *Class; + 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: - Class = (char *) OPERAND(c); + cc = (char *) OPERAND(c); while (s < strend) { - if (REGINCLASS(Class, *s)) { + if (REGINCLASS(cc, *s)) { if (tmp && regtry(prog, s)) goto got_it; else @@ -435,8 +581,10 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case BOUND: - if (minlen) - dontbother++,strend--; + 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) { @@ -450,12 +598,38 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case NBOUND: - if (minlen) - dontbother++,strend--; + 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) { @@ -468,6 +642,28 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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)) { @@ -481,6 +677,19 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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: PL_reg_flags |= RF_tainted; while (s < strend) { @@ -495,6 +704,20 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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)) { @@ -508,6 +731,19 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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: PL_reg_flags |= RF_tainted; while (s < strend) { @@ -522,6 +758,20 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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)) { @@ -535,6 +785,19 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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: PL_reg_flags |= RF_tainted; while (s < strend) { @@ -549,6 +812,20 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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)) { @@ -562,6 +839,19 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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: PL_reg_flags |= RF_tainted; while (s < strend) { @@ -576,6 +866,20 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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)) { @@ -589,6 +893,19 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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)) { @@ -602,6 +919,19 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, 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 { @@ -610,37 +940,51 @@ regexec_flags(register regexp *prog, char *stringarg, register char *strend, char *last; I32 oldpos = scream_pos; - if (screamer) { - last = screaminstr(screamer, prog->float_substr, s - strbeg, + if (flags & REXEC_SCREAM) { + last = screaminstr(sv, prog->float_substr, s - strbeg, end_shift, &scream_pos, 1); /* last one */ if (!last) { last = scream_olds; /* Only one occurence. */ } - } else { + } + else { STRLEN len; char *little = SvPV(prog->float_substr, len); - last = rninstr(s, strend, little, little + 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 - 1; + 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->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 */ @@ -652,7 +996,7 @@ got_it: } } else { - I32 i = strend - startpos + (stringarg - strbeg); + I32 i = PL_regeol - startpos + (stringarg - strbeg); s = savepvn(strbeg, i); Safefree(prog->subbase); prog->subbase = s; @@ -667,15 +1011,20 @@ 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. */ + if (PL_reg_eval_set) { + /* 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. */ + restore_pos(0); } + return 1; phooey: + if (PL_reg_eval_set) + restore_pos(0); return 0; } @@ -692,6 +1041,8 @@ regtry(regexp *prog, char *startpos) CHECKPOINT lastcp; if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) { + MAGIC *mg; + PL_reg_eval_set = RS_init; DEBUG_r(DEBUG_s( PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %i\n", @@ -704,13 +1055,42 @@ regtry(regexp *prog, char *startpos) /* Apparently this is not needed, judging by wantarray. */ /* SAVEINT(cxstack[cxstack_ix].blk_gimme); cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ + + if (PL_reg_sv) { + /* Make $_ available to executed code. */ + if (PL_reg_sv != DEFSV) { + /* SAVE_DEFSV does *not* suffice here for USE_THREADS */ + SAVESPTR(DEFSV); + DEFSV = PL_reg_sv; + } + + if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv) + && (mg = mg_find(PL_reg_sv, 'g')))) { + /* prepare for quick setting of pos */ + sv_magic(PL_reg_sv, (SV*)0, 'g', Nullch, 0); + mg = mg_find(PL_reg_sv, 'g'); + mg->mg_len = -1; + } + PL_reg_magic = mg; + PL_reg_oldpos = mg->mg_len; + SAVEDESTRUCTOR(restore_pos, 0); + } + if (!PL_reg_curpm) + New(22,PL_reg_curpm, 1, PMOP); + PL_reg_curpm->op_pmregexp = prog; + PL_reg_oldcurpm = PL_curpm; + PL_curpm = PL_reg_curpm; + prog->subbeg = PL_bostr; + prog->subend = PL_regeol; /* strend may have been modified */ } + prog->startp[0] = startpos; PL_reginput = startpos; PL_regstartp = prog->startp; PL_regendp = prog->endp; PL_reglastparen = &prog->lastparen; prog->lastparen = 0; PL_regsize = 0; + DEBUG_r(PL_reg_starttry = startpos); if (PL_reg_start_tmpl <= prog->nparens) { PL_reg_start_tmpl = prog->nparens*3/2 + 3; if(PL_reg_start_tmp) @@ -719,18 +1099,19 @@ regtry(regexp *prog, char *startpos) New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*); } + /* XXXX What this code is doing here?!!! There should be no need + to do this again and again, PL_reglastparen should take care of + this! */ sp = prog->startp; ep = prog->endp; - PL_regdata = prog->data; if (prog->nparens) { - for (i = prog->nparens; i >= 0; i--) { - *sp++ = NULL; - *ep++ = NULL; + for (i = prog->nparens; i >= 1; i--) { + *++sp = NULL; + *++ep = NULL; } } REGCP_SET; if (regmatch(prog->program + 1)) { - prog->startp[0] = startpos; prog->endp[0] = PL_reginput; return 1; } @@ -771,6 +1152,7 @@ regmatch(regnode *prog) PL_regindent++; #endif + /* Note that nextchr is a byte even in UTF */ nextchr = UCHARAT(locinput); scan = prog; while (scan != NULL) { @@ -790,17 +1172,31 @@ regmatch(regnode *prog) int docolor = *PL_colors[0]; int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */ int l = (PL_regeol - locinput > taill ? taill : PL_regeol - locinput); + /* The part of the string before starttry has one color + (pref0_len chars), between starttry and current + position another one (pref_len - pref0_len chars), + after the current position the third one. + We assume that pref0_len <= pref_len, otherwise we + decrease pref0_len. */ int pref_len = (locinput - PL_bostr > (5 + taill) - l ? (5 + taill) - l : locinput - PL_bostr); + int pref0_len = pref_len - (locinput - PL_reg_starttry); 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; + if (pref0_len > pref_len) + pref0_len = pref_len; regprop(prop, scan); PerlIO_printf(Perl_debug_log, - "%4i <%s%.*s%s%s%s%.*s%s>%*s|%3d:%*s%s\n", + "%4i <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3d:%*s%s\n", locinput - PL_bostr, - PL_colors[2], pref_len, locinput - pref_len, PL_colors[3], + 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, @@ -837,7 +1233,7 @@ regmatch(regnode *prog) break; sayNO; case GPOS: - if (locinput == PL_regbol) + if (locinput == PL_reg_ganch) break; sayNO; case EOL: @@ -861,12 +1257,36 @@ regmatch(regnode *prog) if (PL_regeol != locinput) sayNO; break; + case SANYUTF8: + if (nextchr & 0x80) { + locinput += PL_utf8skip[nextchr]; + if (locinput > PL_regeol) + sayNO; + nextchr = UCHARAT(locinput); + break; + } + if (!nextchr && locinput >= PL_regeol) + sayNO; + nextchr = UCHARAT(++locinput); + break; case SANY: if (!nextchr && locinput >= PL_regeol) sayNO; nextchr = UCHARAT(++locinput); break; - case ANY: + case ANYUTF8: + if (nextchr & 0x80) { + locinput += PL_utf8skip[nextchr]; + if (locinput > PL_regeol) + sayNO; + nextchr = UCHARAT(locinput); + break; + } + if (!nextchr && locinput >= PL_regeol || nextchr == '\n') + sayNO; + nextchr = UCHARAT(++locinput); + break; + case REG_ANY: if (!nextchr && locinput >= PL_regeol || nextchr == '\n') sayNO; nextchr = UCHARAT(++locinput); @@ -890,10 +1310,32 @@ regmatch(regnode *prog) case EXACTF: 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) != nextchr && UCHARAT(s) != ((OP(scan) == EXACTF) - ? fold : fold_locale)[nextchr]) + ? PL_fold : PL_fold_locale)[nextchr]) sayNO; if (PL_regeol - locinput < ln) sayNO; @@ -904,6 +1346,15 @@ regmatch(regnode *prog) locinput += ln; nextchr = UCHARAT(locinput); break; + case ANYOFUTF8: + s = (char *) OPERAND(scan); + if (!REGINCLASSUTF8(scan, (U8*)locinput)) + sayNO; + if (locinput >= PL_regeol) + sayNO; + locinput += PL_utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; case ANYOF: s = (char *) OPERAND(scan); if (nextchr < 0) @@ -925,13 +1376,57 @@ regmatch(regnode *prog) sayNO; nextchr = UCHARAT(++locinput); break; - case NALNUML: + 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 += PL_utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; + } + if (!(OP(scan) == ALNUMUTF8 + ? isALNUM(nextchr) : isALNUM_LC(nextchr))) + sayNO; + nextchr = UCHARAT(++locinput); + break; + case NALNUML: + PL_reg_flags |= RF_tainted; + /* FALL THROUGH */ + case NALNUM: + if (!nextchr && locinput >= PL_regeol) + sayNO; + if (OP(scan) == NALNUM + ? isALNUM(nextchr) : isALNUM_LC(nextchr)) + sayNO; + nextchr = UCHARAT(++locinput); + break; + case NALNUMLUTF8: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ - case NALNUM: + case NALNUMUTF8: if (!nextchr && locinput >= PL_regeol) sayNO; - if (OP(scan) == NALNUM + if (nextchr & 0x80) { + if (OP(scan) == NALNUMUTF8 + ? swash_fetch(PL_utf8_alnum, (U8*)locinput) + : isALNUM_LC_utf8((U8*)locinput)) + { + sayNO; + } + locinput += PL_utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; + } + if (OP(scan) == NALNUMUTF8 ? isALNUM(nextchr) : isALNUM_LC(nextchr)) sayNO; nextchr = UCHARAT(++locinput); @@ -955,6 +1450,26 @@ regmatch(regnode *prog) 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: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ @@ -966,6 +1481,28 @@ regmatch(regnode *prog) sayNO; nextchr = UCHARAT(++locinput); break; + case SPACELUTF8: + PL_reg_flags |= RF_tainted; + /* FALL THROUGH */ + case SPACEUTF8: + if (!nextchr && locinput >= PL_regeol) + sayNO; + if (nextchr & 0x80) { + if (!(OP(scan) == SPACEUTF8 + ? swash_fetch(PL_utf8_space,(U8*)locinput) + : isSPACE_LC_utf8((U8*)locinput))) + { + sayNO; + } + locinput += PL_utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; + } + if (!(OP(scan) == SPACEUTF8 + ? isSPACE(nextchr) : isSPACE_LC(nextchr))) + sayNO; + nextchr = UCHARAT(++locinput); + break; case NSPACEL: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ @@ -977,11 +1514,45 @@ regmatch(regnode *prog) sayNO; 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 += PL_utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; + } + if (OP(scan) == NSPACEUTF8 + ? isSPACE(nextchr) : isSPACE_LC(nextchr)) + sayNO; + nextchr = UCHARAT(++locinput); + break; case DIGIT: if (!isDIGIT(nextchr)) sayNO; nextchr = UCHARAT(++locinput); break; + case DIGITUTF8: + if (nextchr & 0x80) { + if (!(swash_fetch(PL_utf8_digit,(U8*)locinput))) + sayNO; + locinput += PL_utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; + } + if (!isDIGIT(nextchr)) + sayNO; + nextchr = UCHARAT(++locinput); + break; case NDIGIT: if (!nextchr && locinput >= PL_regeol) sayNO; @@ -989,6 +1560,30 @@ regmatch(regnode *prog) sayNO; 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 += PL_utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; + } + if (isDIGIT(nextchr)) + sayNO; + nextchr = UCHARAT(++locinput); + break; + case CLUMP: + if (locinput >= PL_regeol || swash_fetch(PL_utf8_mark,(U8*)locinput)) + sayNO; + locinput += PL_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: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ @@ -1000,11 +1595,45 @@ regmatch(regnode *prog) 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) != nextchr && (OP(scan) == REF || (UCHARAT(s) != ((OP(scan) == REFF - ? fold : fold_locale)[nextchr])))) + ? PL_fold : PL_fold_locale)[nextchr])))) sayNO; ln = PL_regendp[n] - s; if (locinput + ln > PL_regeol) @@ -1036,20 +1665,100 @@ regmatch(regnode *prog) 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]); + PL_reg_magic->mg_len = locinput - PL_bostr; + PL_regendp[0] = locinput; CALLRUNOPS(); /* Scalar context. */ SPAGAIN; ret = POPs; PUTBACK; - if (logical) { - logical = 0; - sw = SvTRUE(ret); - } else - sv_setsv(save_scalar(PL_replgv), ret); PL_op = oop; PL_curpad = ocurpad; PL_curcop = ocurcop; + if (logical) { + if (logical == 2) { /* Postponed subexpression. */ + regexp *re; + MAGIC *mg = Null(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; + (void)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: @@ -1079,7 +1788,7 @@ regmatch(regnode *prog) } break; case LOGICAL: - logical = 1; + logical = scan->flags; break; case CURLYX: { CURCUR cc; @@ -1185,11 +1894,12 @@ regmatch(regnode *prog) PL_regcc = cc; if (n >= cc->max) { /* Maximum greed exceeded? */ - if (PL_dowarn && n >= REG_INFTY + if (ckWARN(WARN_UNSAFE) && n >= REG_INFTY && !(PL_reg_flags & RF_warned)) { PL_reg_flags |= RF_warned; - warn("Complex regular subexpression recursion " - "limit (%d) exceeded", REG_INFTY - 1); + warner(WARN_UNSAFE, "%s limit (%d) exceeded", + "Complex regular subexpression recursion", + REG_INFTY - 1); } sayNO; } @@ -1241,9 +1951,12 @@ regmatch(regnode *prog) REPORT_CODE_OFF+PL_regindent*2, "") ); } - if (PL_dowarn && n >= REG_INFTY && !(PL_reg_flags & RF_warned)) { + if (ckWARN(WARN_UNSAFE) && n >= REG_INFTY + && !(PL_reg_flags & RF_warned)) { PL_reg_flags |= RF_warned; - warn("count exceeded %d", REG_INFTY - 1); + 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. */ @@ -1339,15 +2052,16 @@ regmatch(regnode *prog) && !(paren && ln == 0)) ln = n; locinput = PL_reginput; - if (regkind[(U8)OP(next)] == EXACT) { + if (PL_regkind[(U8)OP(next)] == EXACT) { c1 = UCHARAT(OPERAND(next) + 1); if (OP(next) == EXACTF) - c2 = fold[c1]; + c2 = PL_fold[c1]; else if (OP(next) == EXACTFL) - c2 = fold_locale[c1]; + c2 = PL_fold_locale[c1]; else c2 = c1; - } else + } + else c1 = c2 = -1000; REGCP_SET; /* This may be improved if l == 0. */ @@ -1359,9 +2073,10 @@ regmatch(regnode *prog) { if (paren) { if (n) { - PL_regstartp[paren] = PL_reginput - l; + PL_regstartp[paren] = HOPc(PL_reginput, -l); PL_regendp[paren] = PL_reginput; - } else + } + else PL_regendp[paren] = NULL; } if (regmatch(next)) @@ -1377,7 +2092,8 @@ regmatch(regnode *prog) else sayNO; } - } else { + } + else { n = regrepeat_hard(scan, n, &l); if (n != 0 && l == 0 /* In fact, this is tricky. If paren, then the @@ -1392,15 +2108,16 @@ regmatch(regnode *prog) REPORT_CODE_OFF+PL_regindent*2, "", n, l) ); if (n >= ln) { - if (regkind[(U8)OP(next)] == EXACT) { + if (PL_regkind[(U8)OP(next)] == EXACT) { c1 = UCHARAT(OPERAND(next) + 1); if (OP(next) == EXACTF) - c2 = fold[c1]; + c2 = PL_fold[c1]; else if (OP(next) == EXACTFL) - c2 = fold_locale[c1]; + c2 = PL_fold_locale[c1]; else c2 = c1; - } else + } + else c1 = c2 = -1000; } REGCP_SET; @@ -1409,26 +2126,27 @@ regmatch(regnode *prog) if (c1 == -1000 || UCHARAT(PL_reginput) == c1 || UCHARAT(PL_reginput) == c2) - { - DEBUG_r( + { + 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] = PL_reginput - l; - PL_regendp[paren] = PL_reginput; - } else - PL_regendp[paren] = NULL; + ); + if (paren) { + if (n) { + PL_regstartp[paren] = HOPc(PL_reginput, -l); + PL_regendp[paren] = PL_reginput; } - if (regmatch(next)) - sayYES; - REGCP_UNWIND; + else + PL_regendp[paren] = NULL; } + if (regmatch(next)) + sayYES; + REGCP_UNWIND; + } /* Couldn't or didn't -- back up. */ n--; - locinput -= l; + locinput = HOPc(locinput, -l); PL_reginput = locinput; } } @@ -1467,12 +2185,12 @@ regmatch(regnode *prog) * Lookahead to avoid useless match attempts * when we know what character comes next. */ - if (regkind[(U8)OP(next)] == EXACT) { + if (PL_regkind[(U8)OP(next)] == EXACT) { c1 = UCHARAT(OPERAND(next) + 1); if (OP(next) == EXACTF) - c2 = fold[c1]; + c2 = PL_fold[c1]; else if (OP(next) == EXACTFL) - c2 = fold_locale[c1]; + c2 = PL_fold_locale[c1]; else c2 = c1; } @@ -1484,7 +2202,52 @@ regmatch(regnode *prog) minmod = 0; if (ln && regrepeat(scan, ln) < ln) sayNO; + locinput = PL_reginput; REGCP_SET; + if (c1 != -1000) { + char *e = locinput + n - ln; /* Should not check after this */ + char *old = locinput; + + if (e >= PL_regeol || (n == REG_INFTY)) + e = PL_regeol - 1; + while (1) { + /* Find place 'next' could work */ + if (c1 == c2) { + while (locinput <= e && *locinput != c1) + locinput++; + } else { + while (locinput <= e + && *locinput != c1 + && *locinput != c2) + locinput++; + } + if (locinput > e) + sayNO; + /* PL_reginput == old now */ + if (locinput != old) { + ln = 1; /* Did some */ + if (regrepeat(scan, locinput - old) < + locinput - old) + sayNO; + } + /* PL_reginput == locinput now */ + if (paren) { + if (ln) { + PL_regstartp[paren] = HOPc(locinput, -1); + PL_regendp[paren] = locinput; + } + else + PL_regendp[paren] = NULL; + } + if (regmatch(next)) + sayYES; + PL_reginput = locinput; /* Could be reset... */ + REGCP_UNWIND; + /* Couldn't or didn't -- move forward. */ + old = locinput++; + } + } + else while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */ /* If it could work, try it. */ if (c1 == -1000 || @@ -1493,9 +2256,10 @@ regmatch(regnode *prog) { if (paren) { if (n) { - PL_regstartp[paren] = PL_reginput - 1; + PL_regstartp[paren] = HOPc(PL_reginput, -1); PL_regendp[paren] = PL_reginput; - } else + } + else PL_regendp[paren] = NULL; } if (regmatch(next)) @@ -1503,18 +2267,20 @@ regmatch(regnode *prog) REGCP_UNWIND; } /* Couldn't or didn't -- move forward. */ - PL_reginput = locinput + ln; + PL_reginput = locinput; if (regrepeat(scan, 1)) { ln++; - PL_reginput = locinput + ln; - } else + locinput = PL_reginput; + } + else sayNO; } } else { CHECKPOINT lastcp; n = regrepeat(scan, n); - if (ln < n && regkind[(U8)OP(next)] == EOL && + locinput = PL_reginput; + if (ln < n && PL_regkind[(U8)OP(next)] == EOL && (!PL_multiline || OP(next) == SEOL)) ln = n; /* why back off? */ REGCP_SET; @@ -1527,9 +2293,10 @@ regmatch(regnode *prog) { if (paren && n) { if (n) { - PL_regstartp[paren] = PL_reginput - 1; + PL_regstartp[paren] = HOPc(PL_reginput, -1); PL_regendp[paren] = PL_reginput; - } else + } + else PL_regendp[paren] = NULL; } if (regmatch(next)) @@ -1538,9 +2305,10 @@ regmatch(regnode *prog) } /* Couldn't or didn't -- back up. */ n--; - PL_reginput = locinput + n; + PL_reginput = locinput = HOPc(locinput, -1); } - } else { + } + else { while (n >= ln) { /* If it could work, try it. */ if (c1 == -1000 || @@ -1553,13 +2321,47 @@ regmatch(regnode *prog) } /* Couldn't or didn't -- back up. */ n--; - PL_reginput = locinput + n; + PL_reginput = locinput = HOPc(locinput, -1); } } } sayNO; break; case END: + if (PL_reg_call_cc) { + re_cc_state *cur_call_cc = PL_reg_call_cc; + CURCUR *cctmp = PL_regcc; + regexp *re = PL_reg_re; + CHECKPOINT cp, lastcp; + + cp = regcppush(0); /* Save *all* the positions. */ + REGCP_SET; + 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; + } if (locinput < PL_regtill) sayNO; /* Cannot match: too short. */ /* Fall through */ @@ -1568,18 +2370,51 @@ regmatch(regnode *prog) sayYES; /* Success! */ case SUSPEND: n = 1; + PL_reginput = locinput; goto do_ifmatch; case UNLESSM: n = 0; - if (locinput < PL_bostr + scan->flags) - goto say_yes; + if (scan->flags) { + if (UTF) { /* XXXX This is absolutely + broken, we read before + start of string. */ + s = HOPMAYBEc(locinput, -scan->flags); + if (!s) + goto say_yes; + PL_reginput = s; + } + else { + if (locinput < PL_bostr + scan->flags) + goto say_yes; + PL_reginput = locinput - scan->flags; + goto do_ifmatch; + } + } + else + PL_reginput = locinput; goto do_ifmatch; case IFMATCH: n = 1; - if (locinput < PL_bostr + scan->flags) - goto say_no; + if (scan->flags) { + if (UTF) { /* XXXX This is absolutely + broken, we read before + start of string. */ + s = HOPMAYBEc(locinput, -scan->flags); + if (!s || s < PL_bostr) + goto say_no; + PL_reginput = s; + } + else { + if (locinput < PL_bostr + scan->flags) + goto say_no; + PL_reginput = locinput - scan->flags; + goto do_ifmatch; + } + } + else + PL_reginput = locinput; + do_ifmatch: - PL_reginput = locinput - scan->flags; inner = NEXTOPER(NEXTOPER(scan)); if (regmatch(inner) != n) { say_no: @@ -1587,7 +2422,8 @@ regmatch(regnode *prog) logical = 0; sw = 0; goto do_longjump; - } else + } + else sayNO; } say_yes: @@ -1609,7 +2445,7 @@ regmatch(regnode *prog) default: PerlIO_printf(PerlIO_stderr(), "%lx %d\n", (unsigned long)scan, OP(scan)); - FAIL("regexp memory corruption"); + croak("regexp memory corruption"); } scan = next; } @@ -1618,7 +2454,7 @@ regmatch(regnode *prog) * We get here only if there's trouble -- normally "case END" is * the terminating point. */ - FAIL("corrupted regexp pointers"); + croak("corrupted regexp pointers"); /*NOTREACHED*/ sayNO; @@ -1641,7 +2477,7 @@ 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(regnode *p, I32 max) @@ -1651,19 +2487,34 @@ regrepeat(regnode *p, I32 max) register char *opnd; register I32 c; register char *loceol = PL_regeol; + register I32 hardcount = 0; scan = PL_reginput; if (max != REG_INFTY && max < loceol - scan) loceol = scan + max; opnd = (char *) OPERAND(p); switch (OP(p)) { - case ANY: + case REG_ANY: while (scan < loceol && *scan != '\n') scan++; break; 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) @@ -1672,16 +2523,23 @@ regrepeat(regnode *p, I32 max) case EXACTF: /* length of string is 1 */ c = UCHARAT(++opnd); while (scan < loceol && - (UCHARAT(scan) == c || UCHARAT(scan) == fold[c])) + (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c])) scan++; break; case EXACTFL: /* length of string is 1 */ PL_reg_flags |= RF_tainted; c = UCHARAT(++opnd); while (scan < loceol && - (UCHARAT(scan) == c || UCHARAT(scan) == fold_locale[c])) + (UCHARAT(scan) == c || UCHARAT(scan) == PL_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)) scan++; @@ -1690,51 +2548,130 @@ regrepeat(regnode *p, I32 max) 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: 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: 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: 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: 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 - PL_reginput; + if (hardcount) + c = hardcount; + else + c = scan - PL_reginput; PL_reginput = scan; DEBUG_r( @@ -1763,29 +2700,48 @@ regrepeat_hard(regnode *p, I32 max, I32 *lp) register char *scan; register char *start; register char *loceol = PL_regeol; - I32 l = -1; + I32 l = 0; + I32 count = 0, res = 1; + + if (!max) + return 0; start = PL_reginput; - while (PL_reginput < loceol && (scan = PL_reginput, regmatch(p))) { - if (l == -1) { - *lp = l = PL_reginput - start; - if (max != REG_INFTY && l*max < loceol - scan) - loceol = scan + l*max; - if (l == 0) { - return max; + 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; } } - if (PL_reginput < loceol) + 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; - else - scan = PL_reginput; - return (scan - start)/l; + return count; } /* - - regclass - determine if a character falls into a character class + - reginclass - determine if a character falls into a character class */ STATIC bool @@ -1802,10 +2758,10 @@ reginclass(register char *p, register I32 c) I32 cf; if (flags & ANYOF_LOCALE) { PL_reg_flags |= RF_tainted; - cf = fold_locale[c]; + cf = PL_fold_locale[c]; } else - cf = fold[c]; + cf = PL_fold[c]; if (ANYOF_TEST(p, cf)) match = TRUE; } @@ -1825,5 +2781,90 @@ reginclass(register char *p, register I32 c) return (flags & ANYOF_INVERT) ? !match : match; } +STATIC bool +reginclassutf8(regnode *f, U8 *p) +{ + dTHR; + char flags = ARG1(f); + bool match = FALSE; + SV *sv = (SV*)PL_regdata->data[ARG2(f)]; + + 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; + } + + if (!match && (flags & ANYOF_ISA)) { + PL_reg_flags |= RF_tainted; + + 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; +}