X-Git-Url: https://perl5.git.perl.org/perl5.git/blobdiff_plain/d2f13c598f9812cd520097e22d819d1b9f7b0cb6..0bdb90953749ee86ab7a033ca1aea5852a9bb804:/regexec.c diff --git a/regexec.c b/regexec.c index 0f1c9e1..41b91ca 100644 --- a/regexec.c +++ b/regexec.c @@ -2,7 +2,11 @@ */ /* - * "One Ring to rule them all, One Ring to find them..." + * One Ring to rule them all, One Ring to find them + & + * [p.v of _The Lord of the Rings_, opening poem] + * [p.50 of _The Lord of the Rings_, I/iii: "The Shadow of the Past"] + * [p.254 of _The Lord of the Rings_, II/ii: "The Council of Elrond"] */ /* This file contains functions for executing a regular expression. See @@ -56,7 +60,8 @@ **** Alterations to Henry's code are... **** **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, - **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007 by Larry Wall and others + **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 + **** by Larry Wall and others **** **** You may distribute under the terms of either the GNU General Public **** License or the Artistic License, as specified in the README file. @@ -237,9 +242,10 @@ S_regcppop(pTHX_ const regexp *rex) dVAR; U32 i; char *input; - GET_RE_DEBUG_FLAGS_DECL; + PERL_ARGS_ASSERT_REGCPPOP; + /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ i = SSPOPINT; assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ @@ -314,6 +320,8 @@ Perl_pregexec(pTHX_ REGEXP * const prog, char* stringarg, register char *strend, /* minend: end of match must be >=minend after stringarg. */ /* nosave: For optimizations. */ { + PERL_ARGS_ASSERT_PREGEXEC; + return regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL, nosave ? 0 : REXEC_COPY_STR); @@ -392,9 +400,10 @@ Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos, #ifdef DEBUGGING const char * const i_strpos = strpos; #endif - GET_RE_DEBUG_FLAGS_DECL; + PERL_ARGS_ASSERT_RE_INTUIT_START; + RX_MATCH_UTF8_set(rx,do_utf8); if (RX_UTF8(rx)) { @@ -998,15 +1007,16 @@ Perl_re_intuit_start(pTHX_ REGEXP * const rx, SV *sv, char *strpos, #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, \ uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ + UV uvc_unfolded = 0; \ switch (trie_type) { \ case trie_utf8_fold: \ if ( foldlen>0 ) { \ - uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ + uvc_unfolded = uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \ foldlen -= len; \ uscan += len; \ len=0; \ } else { \ - uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ + uvc_unfolded = uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \ uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \ foldlen -= UNISKIP( uvc ); \ uscan = foldbuf + UNISKIP( uvc ); \ @@ -1045,6 +1055,9 @@ uvc, charid, foldlen, foldbuf, uniflags) STMT_START { \ charid = (U16)SvIV(*svpp); \ } \ } \ + if (!charid && trie_type == trie_utf8_fold && !UTF) { \ + charid = trie->charmap[uvc_unfolded]; \ + } \ } STMT_END #define REXEC_FBC_EXACTISH_CHECK(CoNd) \ @@ -1181,6 +1194,8 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, register I32 tmp = 1; /* Scratch variable? */ register const bool do_utf8 = PL_reg_match_utf8; RXi_GET_DECL(prog,progi); + + PERL_ARGS_ASSERT_FIND_BYCLASS; /* We know what class it must start with. */ switch (OP(c)) { @@ -1509,7 +1524,7 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, = (reg_ac_data*)progi->data->data[ ARG( c ) ]; reg_trie_data *trie = (reg_trie_data*)progi->data->data[ aho->trie ]; - HV *widecharmap = (HV*) progi->data->data[ aho->trie + 1 ]; + HV *widecharmap = MUTABLE_HV(progi->data->data[ aho->trie + 1 ]); const char *last_start = strend - trie->minlen; #ifdef DEBUGGING @@ -1719,25 +1734,6 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, return s; } -static void -S_swap_match_buff (pTHX_ regexp *prog) { - regexp_paren_pair *t; - - if (!prog->swap) { - /* We have to be careful. If the previous successful match - was from this regex we don't want a subsequent paritally - successful match to clobber the old results. - So when we detect this possibility we add a swap buffer - to the re, and switch the buffer each match. If we fail - we switch it back, otherwise we leave it swapped. - */ - Newxz(prog->swap, (prog->nparens + 1), regexp_paren_pair); - } - t = prog->swap; - prog->swap = prog->offs; - prog->offs = t; -} - /* - regexec_flags - match a regexp against a string @@ -1767,10 +1763,10 @@ Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *stre I32 multiline; RXi_GET_DECL(prog,progi); regmatch_info reginfo; /* create some info to pass to regtry etc */ - bool swap_on_fail = 0; - + regexp_paren_pair *swap = NULL; GET_RE_DEBUG_FLAGS_DECL; + PERL_ARGS_ASSERT_REGEXEC_FLAGS; PERL_UNUSED_ARG(data); /* Be paranoid... */ @@ -1825,29 +1821,50 @@ Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *stre if (prog->extflags & RXf_GPOS_SEEN) { /* Need to set reginfo->ganch */ MAGIC *mg; - - if (flags & REXEC_IGNOREPOS) /* Means: check only at start */ + if (flags & REXEC_IGNOREPOS){ /* Means: check only at start */ reginfo.ganch = startpos + prog->gofs; - else if (sv && SvTYPE(sv) >= SVt_PVMG + DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, + "GPOS IGNOREPOS: reginfo.ganch = startpos + %"UVxf"\n",(UV)prog->gofs)); + } else if (sv && SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv) && (mg = mg_find(sv, PERL_MAGIC_regex_global)) && mg->mg_len >= 0) { reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */ + DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, + "GPOS MAGIC: reginfo.ganch = strbeg + %"IVdf"\n",(IV)mg->mg_len)); + if (prog->extflags & RXf_ANCH_GPOS) { if (s > reginfo.ganch) goto phooey; s = reginfo.ganch - prog->gofs; + DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, + "GPOS ANCH_GPOS: s = ganch - %"UVxf"\n",(UV)prog->gofs)); + if (s < strbeg) + goto phooey; } } else if (data) { reginfo.ganch = strbeg + PTR2UV(data); - } else /* pos() not defined */ + DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, + "GPOS DATA: reginfo.ganch= strbeg + %"UVxf"\n",PTR2UV(data))); + + } else { /* pos() not defined */ reginfo.ganch = strbeg; + DEBUG_GPOS_r(PerlIO_printf(Perl_debug_log, + "GPOS: reginfo.ganch = strbeg\n")); + } } if (PL_curpm && (PM_GETRE(PL_curpm) == rx)) { - swap_on_fail = 1; - swap_match_buff(prog); /* do we need a save destructor here for - eval dies? */ + /* We have to be careful. If the previous successful match + was from this regex we don't want a subsequent partially + successful match to clobber the old results. + So when we detect this possibility we add a swap buffer + to the re, and switch the buffer each match. If we fail + we switch it back, otherwise we leave it swapped. + */ + swap = prog->offs; + /* do we need a save destructor here for eval dies? */ + Newxz(prog->offs, (prog->nparens + 1), regexp_paren_pair); } if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) { re_scream_pos_data d; @@ -1912,7 +1929,8 @@ Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *stre is bogus -- we set it above, when prog->extflags & RXf_GPOS_SEEN and we only enter this block when the same bit is set. */ char *tmp_s = reginfo.ganch - prog->gofs; - if (regtry(®info, &tmp_s)) + + if (tmp_s >= strbeg && regtry(®info, &tmp_s)) goto got_it; goto phooey; } @@ -2146,6 +2164,7 @@ Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, register char *stre goto phooey; got_it: + Safefree(swap); RX_MATCH_TAINTED_set(rx, PL_reg_flags & RF_tainted); if (PL_reg_eval_set) @@ -2191,10 +2210,12 @@ phooey: PL_colors[4], PL_colors[5])); if (PL_reg_eval_set) restore_pos(aTHX_ prog); - if (swap_on_fail) + if (swap) { /* we failed :-( roll it back */ - swap_match_buff(prog); - + Safefree(prog->offs); + prog->offs = swap; + } + return 0; } @@ -2211,6 +2232,9 @@ S_regtry(pTHX_ regmatch_info *reginfo, char **startpos) regexp *const prog = (struct regexp *)SvANY(rx); RXi_GET_DECL(prog,progi); GET_RE_DEBUG_FLAGS_DECL; + + PERL_ARGS_ASSERT_REGTRY; + reginfo->cutpoint=NULL; if ((prog->extflags & RXf_EVAL_SEEN) && !PL_reg_eval_set) { @@ -2233,7 +2257,7 @@ S_regtry(pTHX_ regmatch_info *reginfo, char **startpos) /* Make $_ available to executed code. */ if (reginfo->sv != DEFSV) { SAVE_DEFSV; - DEFSV = reginfo->sv; + DEFSV_set(reginfo->sv); } if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv) @@ -2255,15 +2279,24 @@ S_regtry(pTHX_ regmatch_info *reginfo, char **startpos) Newxz(PL_reg_curpm, 1, PMOP); #ifdef USE_ITHREADS { - SV* const repointer = newSViv(0); - /* so we know which PL_regex_padav element is PL_reg_curpm */ - SvFLAGS(repointer) |= SVf_BREAK; - av_push(PL_regex_padav,repointer); + SV* const repointer = &PL_sv_undef; + /* this regexp is also owned by the new PL_reg_curpm, which + will try to free it. */ + av_push(PL_regex_padav, repointer); PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav); PL_regex_pad = AvARRAY(PL_regex_padav); } #endif } +#ifdef USE_ITHREADS + /* It seems that non-ithreads works both with and without this code. + So for efficiency reasons it seems best not to have the code + compiled when it is not needed. */ + /* This is safe against NULLs: */ + ReREFCNT_dec(PM_GETRE(PL_reg_curpm)); + /* PM_reg_curpm owns a reference to this regexp. */ + ReREFCNT_inc(rx); +#endif PM_SETRE(PL_reg_curpm, rx); PL_reg_oldcurpm = PL_curpm; PL_curpm = PL_reg_curpm; @@ -2557,11 +2590,14 @@ S_debug_start_match(pTHX_ const REGEXP *prog, const bool do_utf8, const char *start, const char *end, const char *blurb) { const bool utf8_pat = RX_UTF8(prog) ? 1 : 0; + + PERL_ARGS_ASSERT_DEBUG_START_MATCH; + if (!PL_colorset) reginitcolors(); { RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0), - RX_PRECOMP(prog), RX_PRELEN(prog), 60); + RX_PRECOMP_const(prog), RX_PRELEN(prog), 60); RE_PV_QUOTED_DECL(s1, do_utf8, PERL_DEBUG_PAD_ZERO(1), start, end - start, 60); @@ -2600,6 +2636,8 @@ S_dump_exec_pos(pTHX_ const char *locinput, ? (5 + taill) - l : locinput - loc_bostr; int pref0_len; + PERL_ARGS_ASSERT_DUMP_EXEC_POS; + while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) pref_len++; pref0_len = pref_len - (locinput - loc_reg_starttry); @@ -2650,11 +2688,15 @@ S_dump_exec_pos(pTHX_ const char *locinput, * or 0 if non of the buffers matched. */ STATIC I32 -S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) { +S_reg_check_named_buff_matched(pTHX_ const regexp *rex, const regnode *scan) +{ I32 n; RXi_GET_DECL(rex,rexi); - SV *sv_dat=(SV*)rexi->data->data[ ARG( scan ) ]; + SV *sv_dat= MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); I32 *nums=(I32*)SvPVX(sv_dat); + + PERL_ARGS_ASSERT_REG_CHECK_NAMED_BUFF_MATCHED; + for ( n=0; n= nums[n] && PL_regoffs[nums[n]].end != -1) @@ -2698,16 +2740,12 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) dVAR; register const bool do_utf8 = PL_reg_match_utf8; const U32 uniflags = UTF8_ALLOW_DEFAULT; - REGEXP *rex_sv = reginfo->prog; regexp *rex = (struct regexp *)SvANY(rex_sv); RXi_GET_DECL(rex,rexi); - I32 oldsave; - /* the current state. This is a cached copy of PL_regmatch_state */ register regmatch_state *st; - /* cache heavy used fields of st in registers */ register regnode *scan; register regnode *next; @@ -2722,13 +2760,11 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) const U32 max_nochange_depth = (3 * rex->nparens > MAX_RECURSE_EVAL_NOCHANGE_DEPTH) ? 3 * rex->nparens : MAX_RECURSE_EVAL_NOCHANGE_DEPTH; - regmatch_state *yes_state = NULL; /* state to pop to on success of subpattern */ /* mark_state piggy backs on the yes_state logic so that when we unwind the stack on success we can update the mark_state as we go */ regmatch_state *mark_state = NULL; /* last mark state we have seen */ - regmatch_state *cur_eval = NULL; /* most recent EVAL_AB state */ struct regmatch_state *cur_curlyx = NULL; /* most recent curlyx */ U32 state_num; @@ -2741,10 +2777,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) during a successfull match */ U32 lastopen = 0; /* last open we saw */ bool has_cutgroup = RX_HAS_CUTGROUP(rex) ? 1 : 0; - SV* const oreplsv = GvSV(PL_replgv); - - /* these three flags are set by various ops to signal information to * the very next op. They have a useful lifetime of exactly one loop * iteration, and are not preserved or restored by state pushes/pops @@ -2759,11 +2792,12 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) false: plain (?=foo) true: used as a condition: (?(?=foo)) */ - #ifdef DEBUGGING GET_RE_DEBUG_FLAGS_DECL; #endif + PERL_ARGS_ASSERT_REGMATCH; + DEBUG_OPTIMISE_r( DEBUG_EXECUTE_r({ PerlIO_printf(Perl_debug_log,"regmatch start\n"); })); @@ -2810,6 +2844,11 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) state_num = OP(scan); reenter_switch: + + assert(PL_reglastparen == &rex->lastparen); + assert(PL_reglastcloseparen == &rex->lastcloseparen); + assert(PL_regoffs == rex->offs); + switch (state_num) { case BOL: if (locinput == PL_bostr) @@ -2919,7 +2958,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) /* what trie are we using right now */ reg_trie_data * const trie = (reg_trie_data*)rexi->data->data[ ARG( scan ) ]; - HV * widecharmap = (HV *)rexi->data->data[ ARG( scan ) + 1 ]; + HV * widecharmap = MUTABLE_HV(rexi->data->data[ ARG( scan ) + 1 ]); U32 state = trie->startstate; if (trie->bitmap && trie_type != trie_utf8_fold && @@ -2976,7 +3015,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) if ( got_wordnum ) { if ( ! ST.accepted ) { ENTER; - /* SAVETMPS; */ /* XXX is this necessary? dmq */ + SAVETMPS; /* XXX is this necessary? dmq */ bufflen = TRIE_INITAL_ACCEPT_BUFFLEN; sv_accept_buff=newSV(bufflen * sizeof(reg_trie_accepted) - 1); @@ -3076,7 +3115,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) /* only one choice left - just continue */ DEBUG_EXECUTE_r({ AV *const trie_words - = (AV *) rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]; + = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]); SV ** const tmp = av_fetch( trie_words, ST.accept_buff[ 0 ].wordnum-1, 0 ); SV *sv= tmp ? sv_newmortal() : NULL; @@ -3158,7 +3197,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) DEBUG_EXECUTE_r({ AV *const trie_words - = (AV *) rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]; + = MUTABLE_AV(rexi->data->data[ARG(ST.me)+TRIE_WORDS_OFFSET]); SV ** const tmp = av_fetch( trie_words, ST.accept_buff[ best ].wordnum - 1, 0 ); regnode *nextop=(!ST.jump || !ST.jump[ST.accept_buff[best].wordnum]) ? @@ -3195,6 +3234,9 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) } /* NOTREACHED */ case TRIE_next: + /* we dont want to throw this away, see bug 57042*/ + if (oreplsv != GvSV(PL_replgv)) + sv_setsv(oreplsv, GvSV(PL_replgv)); FREETMPS; LEAVE; sayYES; @@ -3302,6 +3344,47 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) nextchr = UCHARAT(locinput); break; } + case BOUNDL: + case NBOUNDL: + PL_reg_flags |= RF_tainted; + /* FALL THROUGH */ + case BOUND: + case NBOUND: + /* was last char in word? */ + if (do_utf8) { + if (locinput == PL_bostr) + ln = '\n'; + else { + const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); + + ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); + } + if (OP(scan) == BOUND || OP(scan) == NBOUND) { + ln = isALNUM_uni(ln); + LOAD_UTF8_CHARCLASS_ALNUM(); + n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8); + } + else { + ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln)); + n = isALNUM_LC_utf8((U8*)locinput); + } + } + else { + ln = (locinput != PL_bostr) ? + UCHARAT(locinput - 1) : '\n'; + if (OP(scan) == BOUND || OP(scan) == NBOUND) { + ln = isALNUM(ln); + n = isALNUM(nextchr); + } + else { + ln = isALNUM_LC(ln); + n = isALNUM_LC(nextchr); + } + } + if (((!ln) == (!n)) == (OP(scan) == BOUND || + OP(scan) == BOUNDL)) + sayNO; + break; case ANYOF: if (do_utf8) { STRLEN inclasslen = PL_regeol - locinput; @@ -3381,47 +3464,6 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) sayNO; nextchr = UCHARAT(++locinput); break; - case BOUNDL: - case NBOUNDL: - PL_reg_flags |= RF_tainted; - /* FALL THROUGH */ - case BOUND: - case NBOUND: - /* was last char in word? */ - if (do_utf8) { - if (locinput == PL_bostr) - ln = '\n'; - else { - const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr); - - ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags); - } - if (OP(scan) == BOUND || OP(scan) == NBOUND) { - ln = isALNUM_uni(ln); - LOAD_UTF8_CHARCLASS_ALNUM(); - n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8); - } - else { - ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln)); - n = isALNUM_LC_utf8((U8*)locinput); - } - } - else { - ln = (locinput != PL_bostr) ? - UCHARAT(locinput - 1) : '\n'; - if (OP(scan) == BOUND || OP(scan) == NBOUND) { - ln = isALNUM(ln); - n = isALNUM(nextchr); - } - else { - ln = isALNUM_LC(ln); - n = isALNUM_LC(nextchr); - } - } - if (((!ln) == (!n)) == (OP(scan) == BOUND || - OP(scan) == BOUNDL)) - sayNO; - break; case SPACEL: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ @@ -3441,17 +3483,11 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) nextchr = UCHARAT(locinput); break; } - if (!(OP(scan) == SPACE - ? isSPACE(nextchr) : isSPACE_LC(nextchr))) - sayNO; - nextchr = UCHARAT(++locinput); - } - else { - if (!(OP(scan) == SPACE - ? isSPACE(nextchr) : isSPACE_LC(nextchr))) - sayNO; - nextchr = UCHARAT(++locinput); } + if (!(OP(scan) == SPACE + ? isSPACE(nextchr) : isSPACE_LC(nextchr))) + sayNO; + nextchr = UCHARAT(++locinput); break; case NSPACEL: PL_reg_flags |= RF_tainted; @@ -3677,6 +3713,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) OP_4tree * const oop = PL_op; COP * const ocurcop = PL_curcop; PAD *old_comppad; + char *saved_regeol = PL_regeol; n = ARG(scan); PL_op = (OP_4tree*)rexi->data->data[n]; @@ -3702,6 +3739,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) PL_op = oop; PAD_RESTORE_LOCAL(old_comppad); PL_curcop = ocurcop; + PL_regeol = saved_regeol; if (!logical) { /* /(?{...})/ */ sv_setsv(save_scalar(PL_replgv), ret); @@ -3774,7 +3812,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) /* This isn't a first class regexp. Instead, it's caching a regexp onto an existing, Perl visible scalar. */ - sv_magic(ret, (SV*) rx, PERL_MAGIC_qr, 0, 0); + sv_magic(ret, MUTABLE_SV(rx), PERL_MAGIC_qr, 0, 0); } PL_regsize = osize; } @@ -3855,9 +3893,12 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) regcpblow(ST.cp); cur_eval = ST.prev_eval; cur_curlyx = ST.prev_curlyx; - + + /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ PL_reglastparen = &rex->lastparen; PL_reglastcloseparen = &rex->lastcloseparen; + /* also update PL_regoffs */ + PL_regoffs = rex->offs; /* XXXX This is too dramatic a measure... */ PL_reg_maxiter = 0; @@ -3873,6 +3914,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, regnode *prog) SETREX(rex_sv,ST.prev_rex); rex = (struct regexp *)SvANY(rex_sv); rexi = RXi_GET(rex); + /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ PL_reglastparen = &rex->lastparen; PL_reglastcloseparen = &rex->lastcloseparen; @@ -4320,7 +4362,7 @@ NULL case CUTGROUP: PL_reginput = locinput; sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL : - (SV*)rexi->data->data[ ARG( scan ) ]; + MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); PUSH_STATE_GOTO(CUTGROUP_next,next); /* NOTREACHED */ case CUTGROUP_next_fail: @@ -4368,7 +4410,7 @@ NULL case CURLYM: /* /A{m,n}B/ where A is fixed-length */ /* This is an optimisation of CURLYX that enables us to push - * only a single backtracking state, no matter now many matches + * only a single backtracking state, no matter how many matches * there are in {m,n}. It relies on the pattern being constant * length, with no parens to influence future backrefs */ @@ -4435,8 +4477,11 @@ NULL cur_eval->u.eval.close_paren == (U32)ST.me->flags) goto fake_end; - if ( ST.count < (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)) ) - goto curlym_do_A; /* try to match another A */ + { + I32 max = (ST.minmod ? ARG1(ST.me) : ARG2(ST.me)); + if ( max == REG_INFTY || ST.count < max ) + goto curlym_do_A; /* try to match another A */ + } goto curlym_do_B; /* try to match B */ case CURLYM_A_fail: /* just failed to match an A */ @@ -4528,7 +4573,8 @@ NULL case CURLYM_B_fail: /* just failed to match a B */ REGCP_UNWIND(ST.cp); if (ST.minmod) { - if (ST.count == ARG2(ST.me) /* max */) + I32 max = ARG2(ST.me); + if (max != REG_INFTY && ST.count == max) sayNO; goto curlym_do_A; /* try to match a further A */ } @@ -4875,6 +4921,11 @@ NULL cur_curlyx = cur_eval->u.eval.prev_curlyx; ReREFCNT_inc(rex_sv); st->u.eval.cp = regcppush(0); /* Save *all* the positions. */ + + /* rex was changed so update the pointer in PL_reglastparen and PL_reglastcloseparen */ + PL_reglastparen = &rex->lastparen; + PL_reglastcloseparen = &rex->lastcloseparen; + REGCP_SET(st->u.eval.lastcp); PL_reginput = locinput; @@ -4956,6 +5007,8 @@ NULL do_ifmatch: ST.me = scan; ST.logical = logical; + logical = 0; /* XXX: reset state of logical once it has been saved into ST */ + /* execute body of (?...A) */ PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan))); /* NOTREACHED */ @@ -4995,7 +5048,7 @@ NULL case PRUNE: PL_reginput = locinput; if (!scan->flags) - sv_yes_mark = sv_commit = (SV*)rexi->data->data[ ARG( scan ) ]; + sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); PUSH_STATE_GOTO(COMMIT_next,next); /* NOTREACHED */ case COMMIT_next_fail: @@ -5009,7 +5062,7 @@ NULL case MARKPOINT: ST.prev_mark = mark_state; ST.mark_name = sv_commit = sv_yes_mark - = (SV*)rexi->data->data[ ARG( scan ) ]; + = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); mark_state = st; ST.mark_loc = PL_reginput = locinput; PUSH_YES_STATE_GOTO(MARKPOINT_next,next); @@ -5050,7 +5103,7 @@ NULL otherwise do nothing. Meaning we need to scan */ regmatch_state *cur = mark_state; - SV *find = (SV*)rexi->data->data[ ARG( scan ) ]; + SV *find = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); while (cur) { if ( sv_eq( cur->u.mark.mark_name, @@ -5339,6 +5392,8 @@ S_regrepeat(pTHX_ const regexp *prog, const regnode *p, I32 max, int depth) PERL_UNUSED_ARG(depth); #endif + PERL_ARGS_ASSERT_REGREPEAT; + scan = PL_reginput; if (max == REG_INFTY) max = I32_MAX; @@ -5654,12 +5709,14 @@ Perl_regclass_swash(pTHX_ const regexp *prog, register const regnode* node, bool RXi_GET_DECL(prog,progi); const struct reg_data * const data = prog ? progi->data : NULL; + PERL_ARGS_ASSERT_REGCLASS_SWASH; + if (data && data->count) { const U32 n = ARG(node); if (data->what[n] == 's') { - SV * const rv = (SV*)data->data[n]; - AV * const av = (AV*)SvRV((SV*)rv); + SV * const rv = MUTABLE_SV(data->data[n]); + AV * const av = MUTABLE_AV(SvRV(rv)); SV **const ary = AvARRAY(av); SV **a, **b; @@ -5710,6 +5767,8 @@ S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const STRLEN len = 0; STRLEN plen; + PERL_ARGS_ASSERT_REGINCLASS; + if (do_utf8 && !UTF8_IS_INVARIANT(c)) { c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len, (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) | UTF8_CHECK_ONLY); @@ -5733,7 +5792,14 @@ S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const SV * const sw = regclass_swash(prog, n, TRUE, 0, (SV**)&av); if (sw) { - if (swash_fetch(sw, p, do_utf8)) + U8 * utf8_p; + if (do_utf8) { + utf8_p = (U8 *) p; + } else { + STRLEN len = 1; + utf8_p = bytes_to_utf8(p, &len); + } + if (swash_fetch(sw, utf8_p, 1)) match = TRUE; else if (flags & ANYOF_FOLD) { if (!match && lenp && av) { @@ -5742,8 +5808,7 @@ S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const SV* const sv = *av_fetch(av, i, FALSE); STRLEN len; const char * const s = SvPV_const(sv, len); - - if (len <= plen && memEQ(s, (char*)p, len)) { + if (len <= plen && memEQ(s, (char*)utf8_p, len)) { *lenp = len; match = TRUE; break; @@ -5752,13 +5817,16 @@ S_reginclass(pTHX_ const regexp *prog, register const regnode *n, register const } if (!match) { U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; - STRLEN tmplen; - to_utf8_fold(p, tmpbuf, &tmplen); - if (swash_fetch(sw, tmpbuf, do_utf8)) + STRLEN tmplen; + to_utf8_fold(utf8_p, tmpbuf, &tmplen); + if (swash_fetch(sw, tmpbuf, 1)) match = TRUE; } } + + /* If we allocated a string above, free it */ + if (! do_utf8) Safefree(utf8_p); } } if (match && lenp && *lenp == 0) @@ -5827,6 +5895,9 @@ STATIC U8 * S_reghop3(U8 *s, I32 off, const U8* lim) { dVAR; + + PERL_ARGS_ASSERT_REGHOP3; + if (off >= 0) { while (off-- && s < lim) { /* XXX could check well-formedness here */ @@ -5855,6 +5926,9 @@ STATIC U8 * S_reghop4(U8 *s, I32 off, const U8* llim, const U8* rlim) { dVAR; + + PERL_ARGS_ASSERT_REGHOP4; + if (off >= 0) { while (off-- && s < rlim) { /* XXX could check well-formedness here */ @@ -5879,6 +5953,9 @@ STATIC U8 * S_reghopmaybe3(U8* s, I32 off, const U8* lim) { dVAR; + + PERL_ARGS_ASSERT_REGHOPMAYBE3; + if (off >= 0) { while (off-- && s < lim) { /* XXX could check well-formedness here */ @@ -5926,6 +6003,9 @@ STATIC void S_to_utf8_substr(pTHX_ register regexp *prog) { int i = 1; + + PERL_ARGS_ASSERT_TO_UTF8_SUBSTR; + do { if (prog->substrs->data[i].substr && !prog->substrs->data[i].utf8_substr) { @@ -5955,6 +6035,9 @@ S_to_byte_substr(pTHX_ register regexp *prog) { dVAR; int i = 1; + + PERL_ARGS_ASSERT_TO_BYTE_SUBSTR; + do { if (prog->substrs->data[i].utf8_substr && !prog->substrs->data[i].substr) {