X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=regexec.c;h=41b91cafebb2f6eee4fbb8225c2b00de65c46f49;hb=c7cffa0b862541c21ec66e4efd2d80e4f0a2f142;hp=0c4329dda35058ee503c90b6b965d2fb8a8631ae;hpb=ad64d0ecd555e97c5a216efca1ec5a96b7fd0b34;p=p5sagit%2Fp5-mst-13.2.git diff --git a/regexec.c b/regexec.c index 0c4329d..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 @@ -1003,15 +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 ); \ @@ -1050,6 +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) \ @@ -1726,28 +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; - - PERL_ARGS_ASSERT_SWAP_MATCH_BUFF; - - 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 @@ -1777,7 +1763,7 @@ 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; @@ -1835,29 +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; @@ -1922,7 +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; } @@ -2156,6 +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) @@ -2201,10 +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; } @@ -2246,7 +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) @@ -2586,7 +2597,7 @@ S_debug_start_match(pTHX_ const REGEXP *prog, const bool do_utf8, 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); @@ -2833,6 +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) @@ -2999,7 +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); @@ -3218,6 +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; @@ -3325,6 +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; @@ -3404,47 +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 */ @@ -3464,17 +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; @@ -3700,6 +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]; @@ -3725,6 +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); @@ -3878,9 +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; @@ -3896,6 +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; @@ -4391,7 +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 */ @@ -4458,8 +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 */ @@ -4551,7 +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 */ } @@ -4898,6 +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; @@ -4979,6 +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 */ @@ -5762,7 +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) { @@ -5771,8 +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; @@ -5781,13 +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)