X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=regexec.c;h=5d9e8ac8faff174781752dacbdcac6e2b083bbec;hb=698828ad8b31daca4e44400fa4bef1ac74a9a8f5;hp=fa891c86f8f79692bca07dc029eec9815cba26ae;hpb=b2a3d41024cc23941660c6f6cac48502a4284422;p=p5sagit%2Fp5-mst-13.2.git diff --git a/regexec.c b/regexec.c index fa891c8..5d9e8ac 100644 --- a/regexec.c +++ b/regexec.c @@ -39,6 +39,7 @@ /* *These* symbols are masked to allow static link. */ # define Perl_pregexec my_pregexec # define Perl_reginitcolors my_reginitcolors +# define Perl_regclass_swash my_regclass_swash # define PERL_NO_GET_CONTEXT #endif @@ -66,7 +67,7 @@ * **** Alterations to Henry's code are... **** - **** Copyright (c) 1991-1999, Larry Wall + **** Copyright (c) 1991-2001, 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. @@ -105,69 +106,85 @@ * Forwards. */ -#define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c) : ANYOF_BITMAP_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) +#define CHR_DIST(a,b) (DO_UTF8(PL_reg_sv) ? utf8_distance(a,b) : a - b) #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off)) #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off)) -#define HOP(pos,off) (UTF ? reghop((U8*)pos, off) : (U8*)(pos + off)) -#define HOPMAYBE(pos,off) (UTF ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off)) +#define HOP(pos,off) (DO_UTF8(PL_reg_sv) ? reghop((U8*)pos, off) : (U8*)(pos + off)) +#define HOPMAYBE(pos,off) (DO_UTF8(PL_reg_sv) ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off)) #define HOPc(pos,off) ((char*)HOP(pos,off)) #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off)) +#define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim)) +#define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim)) +#define HOP3(pos,off,lim) (DO_UTF8(PL_reg_sv) ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off)) +#define HOPMAYBE3(pos,off,lim) (DO_UTF8(PL_reg_sv) ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off)) +#define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim)) +#define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim)) + static void restore_pos(pTHXo_ void *arg); STATIC CHECKPOINT S_regcppush(pTHX_ I32 parenfloor) { - dTHR; int retval = PL_savestack_ix; - int i = (PL_regsize - parenfloor) * 4; +#define REGCP_PAREN_ELEMS 4 + int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS; int p; - SSCHECK(i + 5); +#define REGCP_OTHER_ELEMS 5 + SSCHECK(paren_elems_to_push + REGCP_OTHER_ELEMS); for (p = PL_regsize; p > parenfloor; p--) { +/* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */ SSPUSHINT(PL_regendp[p]); SSPUSHINT(PL_regstartp[p]); SSPUSHPTR(PL_reg_start_tmp[p]); SSPUSHINT(p); } +/* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */ SSPUSHINT(PL_regsize); SSPUSHINT(*PL_reglastparen); SSPUSHPTR(PL_reginput); - SSPUSHINT(i + 3); - SSPUSHINT(SAVEt_REGCONTEXT); +#define REGCP_FRAME_ELEMS 2 +/* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and + * are needed for the regexp context stack bookkeeping. */ + SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS); + SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */ + return retval; } /* These are needed since we do not localize EVAL nodes: */ -# define REGCP_SET DEBUG_r(PerlIO_printf(Perl_debug_log, \ +# define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \ " Setting an EVAL scope, savestack=%"IVdf"\n", \ - (IV)PL_savestack_ix)); lastcp = PL_savestack_ix + (IV)PL_savestack_ix)); cp = PL_savestack_ix -# define REGCP_UNWIND DEBUG_r(lastcp != PL_savestack_ix ? \ +# define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \ PerlIO_printf(Perl_debug_log, \ " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \ - (IV)lastcp, (IV)PL_savestack_ix) : 0); regcpblow(lastcp) + (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp) STATIC char * S_regcppop(pTHX) { - dTHR; - I32 i = SSPOPINT; + I32 i; U32 paren = 0; char *input; I32 tmps; - assert(i == SAVEt_REGCONTEXT); + + /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */ i = SSPOPINT; + assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */ + i = SSPOPINT; /* Parentheses elements to pop. */ input = (char *) SSPOPPTR; *PL_reglastparen = SSPOPINT; PL_regsize = SSPOPINT; - for (i -= 3; i > 0; i -= 4) { + + /* Now restore the parentheses context. */ + for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS); + i > 0; i -= REGCP_PAREN_ELEMS) { paren = (U32)SSPOPINT; PL_reg_start_tmp[paren] = (char *) SSPOPPTR; PL_regstartp[paren] = SSPOPINT; @@ -190,18 +207,29 @@ S_regcppop(pTHX) (IV)(*PL_reglastparen + 1), (IV)PL_regnpar); } ); +#if 1 + /* It would seem that the similar code in regtry() + * already takes care of this, and in fact it is in + * a better location to since this code can #if 0-ed out + * but the code in regtry() is needed or otherwise tests + * requiring null fields (pat.t#187 and split.t#{13,14} + * (as of patchlevel 7877) will fail. Then again, + * this code seems to be necessary or otherwise + * building DynaLoader will fail: + * "Error: '*' not in typemap in DynaLoader.xs, line 164" + * --jhi */ for (paren = *PL_reglastparen + 1; paren <= PL_regnpar; paren++) { if (paren > PL_regsize) PL_regstartp[paren] = -1; PL_regendp[paren] = -1; } +#endif return input; } STATIC char * S_regcp_set_to(pTHX_ I32 ss) { - dTHR; I32 tmp = PL_savestack_ix; PL_savestack_ix = ss; @@ -219,7 +247,23 @@ typedef struct re_cc_state regexp *re; } re_cc_state; -#define regcpblow(cp) LEAVE_SCOPE(cp) +#define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */ + +#define TRYPAREN(paren, n, input) { \ + if (paren) { \ + if (n) { \ + PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \ + PL_regendp[paren] = input - PL_bostr; \ + } \ + else \ + PL_regendp[paren] = -1; \ + } \ + if (regmatch(next)) \ + sayYES; \ + if (paren && n) \ + PL_regendp[paren] = -1; \ +} + /* * pregexec and friends @@ -244,7 +288,6 @@ Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *stren STATIC void S_cache_re(pTHX_ regexp *prog) { - dTHR; PL_regprecomp = prog->precomp; /* Needed for FAIL. */ #ifdef DEBUGGING PL_regprogram = prog->program; @@ -275,6 +318,13 @@ S_cache_re(pTHX_ regexp *prog) /* XXXX We assume that strpos is strbeg unless sv. */ +/* XXXX Some places assume that there is a fixed substring. + An update may be needed if optimizer marks as "INTUITable" + RExen without fixed substrings. Similarly, it is assumed that + lengths of all the strings are no more than minlen, thus they + cannot come from lookahead. + (Or minlen should take into account lookahead.) */ + /* A failure to find a constant substring means that there is no need to make an expensive call to REx engine, thus we celebrate a failure. Similarly, finding a substring too deep into the string means that less calls to @@ -285,10 +335,14 @@ S_cache_re(pTHX_ regexp *prog) b) Fixed substring; c) Whether we are anchored (beginning-of-line or \G); d) First node (of those at offset 0) which may distingush positions; - We use 'a', 'b', multiline-part of 'c', and try to find a position in the + We use a)b)d) and multiline-part of c), and try to find a position in the string which does not contradict any of them. */ +/* Most of decisions we do here should have been done at compile time. + The nodes of the REx which we used for the search should have been + deleted from the finite automaton. */ + char * Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, char *strend, U32 flags, re_scream_pos_data *data) @@ -298,10 +352,12 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, register I32 end_shift; register char *s; register SV *check; + char *strbeg; char *t; I32 ml_anch; char *tmp; - register char *other_last = Nullch; + register char *other_last = Nullch; /* other substr checked before this */ + char *check_at; /* check substr found at this pos */ #ifdef DEBUGGING char *i_strpos = strpos; #endif @@ -319,28 +375,37 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, (strend - strpos > 60 ? "..." : "")) ); - if (prog->minlen > strend - strpos) { + if (prog->reganch & ROPT_UTF8) + PL_reg_flags |= RF_utf8; + + if (prog->minlen > CHR_DIST((U8*)strend, (U8*)strpos)) { DEBUG_r(PerlIO_printf(Perl_debug_log, "String too short...\n")); goto fail; } + strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos; + PL_regeol = strend; + check = prog->check_substr; if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */ ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE) || ( (prog->reganch & ROPT_ANCH_BOL) && !PL_multiline ) ); /* Check after \n? */ - if ((prog->check_offset_min == prog->check_offset_max) && !ml_anch) { + if (!ml_anch) { + if ( !(prog->reganch & ROPT_ANCH_GPOS) /* Checked by the caller */ + /* SvCUR is not set on references: SvRV and SvPVX overlap */ + && sv && !SvROK(sv) + && (strpos != strbeg)) { + DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); + goto fail; + } + if (prog->check_offset_min == prog->check_offset_max && + !(prog->reganch & ROPT_SANY_SEEN)) { /* Substring at constant offset from beg-of-str... */ I32 slen; - if ( !(prog->reganch & ROPT_ANCH_GPOS) /* Checked by the caller */ - && (sv && (strpos + SvCUR(sv) != strend)) ) { - DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n")); - goto fail; - } - PL_regeol = strend; /* Used in HOP() */ - s = HOPc(strpos, prog->check_offset_min); - if (SvTAIL(prog->check_substr)) { - slen = SvCUR(prog->check_substr); /* >= 1 */ + s = HOP3c(strpos, prog->check_offset_min, strend); + if (SvTAIL(check)) { + slen = SvCUR(check); /* >= 1 */ if ( strend - s > slen || strend - s < slen - 1 || (strend - s == slen && strend[-1] != '\n')) { @@ -349,30 +414,30 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } /* Now should match s[0..slen-2] */ slen--; - if (slen && (*SvPVX(prog->check_substr) != *s + if (slen && (*SvPVX(check) != *s || (slen > 1 - && memNE(SvPVX(prog->check_substr), s, slen)))) { + && memNE(SvPVX(check), s, slen)))) { report_neq: DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n")); goto fail_finish; } } - else if (*SvPVX(prog->check_substr) != *s - || ((slen = SvCUR(prog->check_substr)) > 1 - && memNE(SvPVX(prog->check_substr), s, slen))) + else if (*SvPVX(check) != *s + || ((slen = SvCUR(check)) > 1 + && memNE(SvPVX(check), s, slen))) goto report_neq; goto success_at_start; + } } /* Match is anchored, but substr is not anchored wrt beg-of-str. */ s = strpos; start_shift = prog->check_offset_min; /* okay to underestimate on CC */ - /* Should be nonnegative! */ end_shift = prog->minlen - start_shift - - CHR_SVLEN(prog->check_substr) + (SvTAIL(prog->check_substr) != 0); + CHR_SVLEN(check) + (SvTAIL(check) != 0); if (!ml_anch) { - I32 end = prog->check_offset_max + CHR_SVLEN(prog->check_substr) - - (SvTAIL(prog->check_substr) != 0); - I32 eshift = strend - s - end; + I32 end = prog->check_offset_max + CHR_SVLEN(check) + - (SvTAIL(check) != 0); + I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end; if (end_shift < eshift) end_shift = eshift; @@ -384,7 +449,7 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, start_shift = prog->check_offset_min; /* okay to underestimate on CC */ /* Should be nonnegative! */ end_shift = prog->minlen - start_shift - - CHR_SVLEN(prog->check_substr) + (SvTAIL(prog->check_substr) != 0); + CHR_SVLEN(check) + (SvTAIL(check) != 0); } #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */ @@ -392,12 +457,10 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, Perl_croak(aTHX_ "panic: end_shift"); #endif - check = prog->check_substr; restart: /* Find a possible match in the region s..strend by looking for the "check" substring in the region corrected by start/end_shift. */ if (flags & REXEC_SCREAM) { - char *strbeg = SvPVX(sv); /* XXXX Assume PV_force() on SCREAM! */ I32 p = -1; /* Internal iterator of scream. */ I32 *pp = data ? data->scream_pos : &p; @@ -412,9 +475,13 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, if (data) *data->scream_olds = s; } + else if (prog->reganch & ROPT_SANY_SEEN) + s = fbm_instr((U8*)(s + start_shift), + (U8*)(strend - end_shift), + check, PL_multiline ? FBMrf_MULTILINE : 0); else - s = fbm_instr((unsigned char*)s + start_shift, - (unsigned char*)strend - end_shift, + s = fbm_instr(HOP3(s, start_shift, strend), + HOP3(strend, -end_shift, strbeg), check, PL_multiline ? FBMrf_MULTILINE : 0); /* Update the count-of-usability, remove useless subpatterns, @@ -432,6 +499,8 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, if (!s) goto fail_finish; + check_at = s; + /* Finish the diagnostic message */ DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) ); @@ -447,38 +516,37 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, /* Take into account the "other" substring. */ /* XXXX May be hopelessly wrong for UTF... */ if (!other_last) - other_last = strpos - 1; + other_last = strpos; if (check == prog->float_substr) { do_other_anchored: { - char *last = s - start_shift, *last1, *last2; + char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2; char *s1 = s; - tmp = PL_bostr; t = s - prog->check_offset_max; if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ && (!(prog->reganch & ROPT_UTF8) - || (PL_bostr = strpos, /* Used in regcopmaybe() */ - (t = reghopmaybe_c(s, -(prog->check_offset_max))) + || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos)) && t > strpos))) /* EMPTY */; else t = strpos; - t += prog->anchored_offset; - if (t <= other_last) - t = other_last + 1; - PL_bostr = tmp; - last2 = last1 = strend - prog->minlen; + t = HOP3c(t, prog->anchored_offset, strend); + if (t < other_last) /* These positions already checked */ + t = other_last; + last2 = last1 = HOP3c(strend, -prog->minlen, strbeg); if (last < last1) last1 = last; /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ /* On end-of-str: see comment below. */ s = fbm_instr((unsigned char*)t, - (unsigned char*)last1 + prog->anchored_offset - + SvCUR(prog->anchored_substr) - - (SvTAIL(prog->anchored_substr)!=0), - prog->anchored_substr, PL_multiline ? FBMrf_MULTILINE : 0); - DEBUG_r(PerlIO_printf(Perl_debug_log, "%s anchored substr `%s%.*s%s'%s", + HOP3(HOP3(last1, prog->anchored_offset, strend) + + SvCUR(prog->anchored_substr), + -(SvTAIL(prog->anchored_substr)!=0), strbeg), + prog->anchored_substr, + PL_multiline ? FBMrf_MULTILINE : 0); + DEBUG_r(PerlIO_printf(Perl_debug_log, + "%s anchored substr `%s%.*s%s'%s", (s ? "Found" : "Contradicts"), PL_colors[0], (int)(SvCUR(prog->anchored_substr) @@ -493,17 +561,16 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } DEBUG_r(PerlIO_printf(Perl_debug_log, ", trying floating at offset %ld...\n", - (long)(s1 + 1 - i_strpos))); - PL_regeol = strend; /* Used in HOP() */ - other_last = last1 + prog->anchored_offset; - s = HOPc(last, 1); + (long)(HOP3c(s1, 1, strend) - i_strpos))); + other_last = HOP3c(last1, prog->anchored_offset+1, strend); + s = HOP3c(last, 1, strend); goto restart; } else { DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", (long)(s - i_strpos))); - t = s - prog->anchored_offset; - other_last = s - 1; + t = HOP3c(s, -prog->anchored_offset, strbeg); + other_last = HOP3c(s, 1, strend); s = s1; if (t == strpos) goto try_at_start; @@ -515,13 +582,14 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, char *last, *last1; char *s1 = s; - t = s - start_shift; - last1 = last = strend - prog->minlen + prog->float_min_offset; - if (last - t > prog->float_max_offset) - last = t + prog->float_max_offset; - s = t + prog->float_min_offset; - if (s <= other_last) - s = other_last + 1; + t = HOP3c(s, -start_shift, strbeg); + last1 = last = + HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg); + if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset) + last = HOP3c(t, prog->float_max_offset, strend); + s = HOP3c(t, prog->float_min_offset, strend); + if (s < other_last) + s = other_last; /* XXXX It is not documented what units *_offsets are in. Assume bytes. */ /* fbm_instr() takes into account exact value of end-of-str if the check is SvTAIL(ed). Since false positives are OK, @@ -547,14 +615,13 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, ", trying anchored starting at offset %ld...\n", (long)(s1 + 1 - i_strpos))); other_last = last; - PL_regeol = strend; /* Used in HOP() */ - s = HOPc(t, 1); + s = HOP3c(t, 1, strend); goto restart; } else { DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n", (long)(s - i_strpos))); - other_last = s - 1; + other_last = s; /* Fix this later. --Hugo */ s = s1; if (t == strpos) goto try_at_start; @@ -564,13 +631,10 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, } t = s - prog->check_offset_max; - tmp = PL_bostr; if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */ && (!(prog->reganch & ROPT_UTF8) - || (PL_bostr = strpos, /* Used in regcopmaybe() */ - ((t = reghopmaybe_c(s, -(prog->check_offset_max))) - && t > strpos)))) { - PL_bostr = tmp; + || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos)) + && t > strpos))) { /* Fixed substring is found far enough so that the match cannot start at strpos. */ try_at_offset: @@ -584,9 +648,10 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, find_anchor: while (t < strend - prog->minlen) { if (*t == '\n') { - if (t < s - prog->check_offset_min) { + if (t < check_at - prog->check_offset_min) { if (prog->anchored_substr) { - /* We definitely contradict the found anchored + /* Since we moved from the found position, + we definitely contradict the found anchored substr. Due to the above check we do not contradict "check" substr. Thus we can arrive here only if check substr @@ -597,14 +662,19 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset))); goto do_other_anchored; } + /* We don't contradict the found floating substring. */ + /* XXXX Why not check for STCLASS? */ s = t + 1; DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n", PL_colors[0],PL_colors[1], (long)(s - i_strpos))); goto set_useful; } - DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting at offset %ld...\n", + /* Position contradicts check-string */ + /* XXXX probably better to look for check-string + than for "\n", so one should lower the limit for t? */ + DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n", PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos))); - strpos = s = t + 1; + other_last = strpos = s = t + 1; goto restart; } t++; @@ -613,819 +683,1008 @@ Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos, PL_colors[0],PL_colors[1])); goto fail_finish; } + else { + DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n", + PL_colors[0],PL_colors[1])); + } s = t; set_useful: ++BmUSEFUL(prog->check_substr); /* hooray/5 */ } else { - PL_bostr = tmp; - /* The found string does not prohibit matching at beg-of-str + /* The found string does not prohibit matching at strpos, - no optimization of calling REx engine can be performed, - unless it was an MBOL and we are not after MBOL. */ + unless it was an MBOL and we are not after MBOL, + or a future STCLASS check will fail this. */ try_at_start: /* Even in this situation we may use MBOL flag if strpos is offset wrt the start of the string. */ - if (ml_anch && sv - && (strpos + SvCUR(sv) != strend) && strpos[-1] != '\n') { + if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */ + && (strpos != strbeg) && strpos[-1] != '\n' + /* May be due to an implicit anchor of m{.*foo} */ + && !(prog->reganch & ROPT_IMPLICIT)) + { t = strpos; goto find_anchor; } DEBUG_r( if (ml_anch) - PerlIO_printf(Perl_debug_log, "Does not contradict /%s^%s/m...\n", - PL_colors[0],PL_colors[1]); + PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n", + (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]); ); success_at_start: if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */ + && prog->check_substr /* Could be deleted already */ && --BmUSEFUL(prog->check_substr) < 0 - && prog->check_substr == prog->float_substr) { /* boo */ + && prog->check_substr == prog->float_substr) + { /* If flags & SOMETHING - do not do it many times on the same match */ + DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n")); SvREFCNT_dec(prog->check_substr); prog->check_substr = Nullsv; /* disable */ prog->float_substr = Nullsv; /* clear */ + check = Nullsv; /* abort */ s = strpos; /* XXXX This is a remnant of the old implementation. It looks wasteful, since now INTUIT can use many - other heuristics too. */ + other heuristics. */ prog->reganch &= ~RE_USE_INTUIT; } else s = strpos; } - DEBUG_r(PerlIO_printf(Perl_debug_log, "%sGuessed:%s match at offset %ld\n", - PL_colors[4], PL_colors[5], (long)(s - i_strpos)) ); + /* Last resort... */ + /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */ + if (prog->regstclass) { + /* minlen == 0 is possible if regstclass is \b or \B, + and the fixed substr is ''$. + Since minlen is already taken into account, s+1 is before strend; + accidentally, minlen >= 1 guaranties no false positives at s + 1 + even for \b or \B. But (minlen? 1 : 0) below assumes that + regstclass does not come from lookahead... */ + /* If regstclass takes bytelength more than 1: If charlength==1, OK. + This leaves EXACTF only, which is dealt with in find_byclass(). */ + U8* str = (U8*)STRING(prog->regstclass); + int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT + ? CHR_DIST(str+STR_LEN(prog->regstclass), str) + : 1); + char *endpos = (prog->anchored_substr || ml_anch) + ? HOP3c(s, (prog->minlen ? cl_l : 0), strend) + : (prog->float_substr + ? HOP3c(HOP3c(check_at, -start_shift, strbeg), + cl_l, strend) + : strend); + char *startpos = strbeg; + + t = s; + if (prog->reganch & ROPT_UTF8) { + PL_regdata = prog->data; + PL_bostr = startpos; + } + s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1); + if (!s) { +#ifdef DEBUGGING + char *what; +#endif + if (endpos == strend) { + DEBUG_r( PerlIO_printf(Perl_debug_log, + "Could not match STCLASS...\n") ); + goto fail; + } + DEBUG_r( PerlIO_printf(Perl_debug_log, + "This position contradicts STCLASS...\n") ); + if ((prog->reganch & ROPT_ANCH) && !ml_anch) + goto fail; + /* Contradict one of substrings */ + if (prog->anchored_substr) { + if (prog->anchored_substr == check) { + DEBUG_r( what = "anchored" ); + hop_and_restart: + s = HOP3c(t, 1, strend); + if (s + start_shift + end_shift > strend) { + /* XXXX Should be taken into account earlier? */ + DEBUG_r( PerlIO_printf(Perl_debug_log, + "Could not match STCLASS...\n") ); + goto fail; + } + if (!check) + goto giveup; + DEBUG_r( PerlIO_printf(Perl_debug_log, + "Looking for %s substr starting at offset %ld...\n", + what, (long)(s + start_shift - i_strpos)) ); + goto restart; + } + /* Have both, check_string is floating */ + if (t + start_shift >= check_at) /* Contradicts floating=check */ + goto retry_floating_check; + /* Recheck anchored substring, but not floating... */ + s = check_at; + if (!check) + goto giveup; + DEBUG_r( PerlIO_printf(Perl_debug_log, + "Looking for anchored substr starting at offset %ld...\n", + (long)(other_last - i_strpos)) ); + goto do_other_anchored; + } + /* Another way we could have checked stclass at the + current position only: */ + if (ml_anch) { + s = t = t + 1; + if (!check) + goto giveup; + DEBUG_r( PerlIO_printf(Perl_debug_log, + "Looking for /%s^%s/m starting at offset %ld...\n", + PL_colors[0],PL_colors[1], (long)(t - i_strpos)) ); + goto try_at_offset; + } + if (!prog->float_substr) /* Could have been deleted */ + goto fail; + /* Check is floating subtring. */ + retry_floating_check: + t = check_at - start_shift; + DEBUG_r( what = "floating" ); + goto hop_and_restart; + } + DEBUG_r( if (t != s) + PerlIO_printf(Perl_debug_log, + "By STCLASS: moving %ld --> %ld\n", + (long)(t - i_strpos), (long)(s - i_strpos)); + else + PerlIO_printf(Perl_debug_log, + "Does not contradict STCLASS...\n") ); + } + giveup: + DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n", + PL_colors[4], (check ? "Guessed" : "Giving up"), + PL_colors[5], (long)(s - i_strpos)) ); return s; fail_finish: /* Substring not found */ - BmUSEFUL(prog->check_substr) += 5; /* hooray */ + if (prog->check_substr) /* could be removed already */ + BmUSEFUL(prog->check_substr) += 5; /* hooray */ fail: DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n", PL_colors[4],PL_colors[5])); return Nullch; } -/* - - regexec_flags - match a regexp against a string - */ -I32 -Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend, - 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. */ -/* data: May be used for some additional optimizations. */ -/* nosave: For optimizations. */ +/* We know what class REx starts with. Try to find this position... */ +STATIC char * +S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun) { - dTHR; - register char *s; - register regnode *c; - register char *startpos = stringarg; - register I32 tmp; - I32 minlen; /* must match at least this many chars */ - I32 dontbother = 0; /* how many characters not to try at end */ - 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); - - PL_regcc = 0; - - cache_re(prog); -#ifdef DEBUGGING - PL_regnarrate = PL_debug & 512; -#endif - - /* Be paranoid... */ - if (prog == NULL || startpos == NULL) { - Perl_croak(aTHX_ "NULL regexp parameter"); - return 0; - } - - minlen = prog->minlen; - if (strend - startpos < minlen) goto phooey; - - if (startpos == strbeg) /* is ^ valid at stringarg? */ - PL_regprev = '\n'; - else { - 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) != REG_MAGIC) { - Perl_croak(aTHX_ "corrupted regexp program"); - } - - PL_reg_flags = 0; - PL_reg_eval_set = 0; - PL_reg_maxiter = 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 (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */ - MAGIC *mg; + I32 doevery = (prog->reganch & ROPT_SKIP) == 0; + char *m; + STRLEN ln; + unsigned int c1; + unsigned int c2; + char *e; + register I32 tmp = 1; /* Scratch variable? */ + register bool do_utf8 = DO_UTF8(PL_reg_sv); - if (flags & REXEC_IGNOREPOS) /* Means: check only at start */ - PL_reg_ganch = startpos; - else if (sv && SvTYPE(sv) >= SVt_PVMG - && SvMAGIC(sv) - && (mg = mg_find(sv, 'g')) && mg->mg_len >= 0) { - PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */ - if (prog->reganch & ROPT_ANCH_GPOS) { - if (s > PL_reg_ganch) - goto phooey; - s = PL_reg_ganch; + /* We know what class it must start with. */ + switch (OP(c)) { + case ANYOF: + while (s < strend) { + if (reginclass(c, (U8*)s, do_utf8)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += do_utf8 ? UTF8SKIP(s) : 1; } - } - else /* pos() not defined */ - PL_reg_ganch = strbeg; - } - - if (!(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) { - re_scream_pos_data d; - - d.scream_olds = &scream_olds; - d.scream_pos = &scream_pos; - s = re_intuit_start(prog, sv, s, strend, flags, &d); - if (!s) - goto phooey; /* not present */ - } - - DEBUG_r( if (!PL_colorset) reginitcolors() ); - DEBUG_r(PerlIO_printf(Perl_debug_log, - "%sMatching REx%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], - (int)(strend - startpos > 60 ? 60 : strend - startpos), - startpos, PL_colors[1], - (strend - startpos > 60 ? "..." : "")) - ); + break; + case EXACTF: + m = STRING(c); + ln = STR_LEN(c); + if (UTF) { + c1 = to_utf8_lower((U8*)m); + c2 = to_utf8_upper((U8*)m); + } + else { + c1 = *(U8*)m; + c2 = PL_fold[c1]; + } + goto do_exactf; + case EXACTFL: + m = STRING(c); + ln = STR_LEN(c); + c1 = *(U8*)m; + c2 = PL_fold_locale[c1]; + do_exactf: + e = strend - ln; - /* Simplest case: anchored match need be tried only once. */ - /* [unless only anchor is BOL and multiline is set] */ - if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) { - if (s == startpos && regtry(prog, startpos)) - goto got_it; - else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT) - || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */ - { - char *end; + if (norun && e < s) + e = s; /* Due to minlen logic of intuit() */ - if (minlen) - dontbother = minlen - 1; - end = HOPc(strend, -dontbother) - 1; - /* for multiline we only have to try after newlines */ - if (prog->check_substr) { - if (s == startpos) - goto after_try; - while (1) { - if (regtry(prog, s)) - goto got_it; - after_try: - if (s >= end) - goto phooey; - if (prog->reganch & RE_USE_INTUIT) { - s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL); - if (!s) - goto phooey; + if (do_utf8) { + STRLEN len; + if (c1 == c2) + while (s <= e) { + if ( utf8_to_uv_simple((U8*)s, &len) == c1 + && regtry(prog, s) ) + goto got_it; + s += len; } - else + else + while (s <= e) { + UV c = utf8_to_uv_simple((U8*)s, &len); + if ( (c == c1 || c == c2) && regtry(prog, s) ) + goto got_it; + s += len; + } + } + else { + if (c1 == c2) + while (s <= e) { + if ( *(U8*)s == c1 + && (ln == 1 || !(OP(c) == EXACTF + ? ibcmp(s, m, ln) + : ibcmp_locale(s, m, ln))) + && (norun || regtry(prog, s)) ) + goto got_it; s++; - } - } else { - if (s > startpos) - s--; - while (s < end) { - if (*s++ == '\n') { /* don't need PL_utf8skip here */ - if (regtry(prog, s)) + } + else + while (s <= e) { + if ( (*(U8*)s == c1 || *(U8*)s == c2) + && (ln == 1 || !(OP(c) == EXACTF + ? ibcmp(s, m, ln) + : ibcmp_locale(s, m, ln))) + && (norun || regtry(prog, s)) ) goto got_it; + s++; } - } } - } - goto phooey; - } else if (prog->reganch & ROPT_ANCH_GPOS) { - if (regtry(prog, PL_reg_ganch)) - goto got_it; - goto phooey; - } - - /* Messy cases: unanchored match. */ - if (prog->anchored_substr && prog->reganch & ROPT_SKIP) { - /* we have /x+whatever/ */ - /* it must be a one character string (XXXX Except UTF?) */ - char ch = SvPVX(prog->anchored_substr)[0]; - 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++; - } - } - } - /*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, /* Cannot start after this */ - -(I32)(CHR_SVLEN(must) - - (SvTAIL(must) != 0) + back_min)); - 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) && - ((flags & REXEC_SCREAM) - ? (s = screaminstr(sv, must, HOPc(s, back_min) - strbeg, - end_shift, &scream_pos, 0)) - : (s = fbm_instr((unsigned char*)HOP(s, back_min), - (unsigned char*)strend, must, - PL_multiline ? FBMrf_MULTILINE : 0))) ) { - 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 = HOPc(s, -back_min); - s = t; - } - if (UTF) { - while (s <= last1) { - if (regtry(prog, s)) - goto got_it; + break; + case BOUNDL: + PL_reg_flags |= RF_tainted; + /* FALL THROUGH */ + case BOUND: + if (do_utf8) { + if (s == startpos) + tmp = '\n'; + else { + U8 *r = reghop3((U8*)s, -1, (U8*)startpos); + + tmp = (I32)utf8_to_uv(r, s - (char*)r, 0, 0); + } + 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 ((norun || 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) { - I32 doevery = (prog->reganch & ROPT_SKIP) == 0; - char *m; - int ln; - int c1; - int c2; - char *e; - - if (minlen) - dontbother = minlen - 1; - 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: - 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: - while (s < strend) { - if (REGINCLASS(c, *s)) { - if (tmp && regtry(prog, s)) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; - } - break; - case EXACTF: - m = STRING(c); - ln = STR_LEN(c); - c1 = *m; - c2 = PL_fold[c1]; - goto do_exactf; - case EXACTFL: - m = STRING(c); - ln = STR_LEN(c); - c1 = *m; - c2 = PL_fold_locale[c1]; - do_exactf: - e = strend - ln; - - /* Here it is NOT UTF! */ - if (c1 == c2) { - while (s <= e) { - if ( *s == c1 - && (ln == 1 || (OP(c) == EXACTF - ? ibcmp(s, m, ln) - : ibcmp_locale(s, m, ln))) - && regtry(prog, s) ) - goto got_it; - s++; - } - } else { - while (s <= e) { - if ( (*s == c1 || *s == c2) - && (ln == 1 || (OP(c) == EXACTF - ? ibcmp(s, m, ln) - : ibcmp_locale(s, m, ln))) - && regtry(prog, s) ) - goto got_it; + tmp = (s != startpos) ? UCHARAT(s - 1) : '\n'; + tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0); + while (s < strend) { + if (tmp == + !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) { + tmp = !tmp; + if ((norun || regtry(prog, s))) + goto got_it; + } s++; } } + if ((!prog->minlen && tmp) && (norun || regtry(prog, s))) + goto got_it; break; - case BOUNDL: + case NBOUNDL: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ - case BOUND: - 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))) { - tmp = !tmp; - if (regtry(prog, s)) + case NBOUND: + if (do_utf8) { + if (s == startpos) + tmp = '\n'; + else { + U8 *r = reghop3((U8*)s, -1, (U8*)startpos); + + tmp = (I32)utf8_to_uv(r, s - (char*)r, 0, 0); + } + 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 ((norun || regtry(prog, s))) goto got_it; + s += UTF8SKIP(s); } - s++; - } - 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)) + else { + tmp = (s != startpos) ? UCHARAT(s - 1) : '\n'; + tmp = ((OP(c) == NBOUND ? + isALNUM(tmp) : isALNUM_LC(tmp)) != 0); + while (s < strend) { + if (tmp == + !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s))) + tmp = !tmp; + else if ((norun || regtry(prog, s))) goto got_it; + s++; } - 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 -= 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))) - tmp = !tmp; - else if (regtry(prog, s)) - goto got_it; - s++; - } - 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)) + if ((!prog->minlen && !tmp) && (norun || regtry(prog, s))) goto got_it; break; case ALNUM: - while (s < strend) { - if (isALNUM(*s)) { - if (tmp && regtry(prog, s)) - goto got_it; + if (do_utf8) { + while (s < strend) { + if (swash_fetch(PL_utf8_alnum, (U8*)s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s += UTF8SKIP(s); } - else - tmp = 1; - s++; } - break; - case ALNUMUTF8: - while (s < strend) { - if (swash_fetch(PL_utf8_alnum, (U8*)s)) { - if (tmp && regtry(prog, s)) - goto got_it; + else { + while (s < strend) { + if (isALNUM(*s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s++; } - else - tmp = 1; - s += UTF8SKIP(s); } break; case ALNUML: PL_reg_flags |= RF_tainted; - while (s < strend) { - if (isALNUM_LC(*s)) { - if (tmp && regtry(prog, s)) - goto got_it; + if (do_utf8) { + while (s < strend) { + if (isALNUM_LC_utf8((U8*)s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s += UTF8SKIP(s); } - else - tmp = 1; - 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 { + while (s < strend) { + if (isALNUM_LC(*s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s++; } - else - tmp = 1; - s += UTF8SKIP(s); } break; case NALNUM: - while (s < strend) { - if (!isALNUM(*s)) { - if (tmp && regtry(prog, s)) - goto got_it; + if (do_utf8) { + while (s < strend) { + if (!swash_fetch(PL_utf8_alnum, (U8*)s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s += UTF8SKIP(s); } - else - tmp = 1; - s++; } - break; - case NALNUMUTF8: - while (s < strend) { - if (!swash_fetch(PL_utf8_alnum, (U8*)s)) { - if (tmp && regtry(prog, s)) - goto got_it; + else { + while (s < strend) { + if (!isALNUM(*s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s++; } - else - tmp = 1; - s += UTF8SKIP(s); } break; case NALNUML: PL_reg_flags |= RF_tainted; - while (s < strend) { - if (!isALNUM_LC(*s)) { - if (tmp && regtry(prog, s)) - goto got_it; + if (do_utf8) { + while (s < strend) { + if (!isALNUM_LC_utf8((U8*)s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s += UTF8SKIP(s); } - else - tmp = 1; - 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 { + while (s < strend) { + if (!isALNUM_LC(*s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s++; } - else - tmp = 1; - s += UTF8SKIP(s); } break; case SPACE: - while (s < strend) { - if (isSPACE(*s)) { - if (tmp && regtry(prog, s)) - goto got_it; + if (do_utf8) { + while (s < strend) { + if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s += UTF8SKIP(s); } - else - tmp = 1; - 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 { + while (s < strend) { + if (isSPACE(*s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s++; } - else - tmp = 1; - s += UTF8SKIP(s); } break; case SPACEL: PL_reg_flags |= RF_tainted; - while (s < strend) { - if (isSPACE_LC(*s)) { - if (tmp && regtry(prog, s)) - goto got_it; + if (do_utf8) { + while (s < strend) { + if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s += UTF8SKIP(s); } - else - tmp = 1; - 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 { + while (s < strend) { + if (isSPACE_LC(*s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s++; } - else - tmp = 1; - s += UTF8SKIP(s); } break; case NSPACE: - while (s < strend) { - if (!isSPACE(*s)) { - if (tmp && regtry(prog, s)) - goto got_it; + if (do_utf8) { + while (s < strend) { + if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s))) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s += UTF8SKIP(s); } - else - tmp = 1; - 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 { + while (s < strend) { + if (!isSPACE(*s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s++; } - else - tmp = 1; - s += UTF8SKIP(s); } break; case NSPACEL: PL_reg_flags |= RF_tainted; - while (s < strend) { - if (!isSPACE_LC(*s)) { - if (tmp && regtry(prog, s)) - goto got_it; + if (do_utf8) { + while (s < strend) { + if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s += UTF8SKIP(s); + } + } + else { + while (s < strend) { + if (!isSPACE_LC(*s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s++; } - else - tmp = 1; - s++; } break; - case NSPACELUTF8: + case DIGIT: + if (do_utf8) { + while (s < strend) { + if (swash_fetch(PL_utf8_digit,(U8*)s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + } + else { + while (s < strend) { + if (isDIGIT(*s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s++; + } + } + break; + case DIGITL: PL_reg_flags |= RF_tainted; - while (s < strend) { - if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) { - if (tmp && regtry(prog, s)) - goto got_it; + if (do_utf8) { + while (s < strend) { + if (isDIGIT_LC_utf8((U8*)s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } else - tmp = doevery; + tmp = 1; + s += UTF8SKIP(s); + } + } + else { + while (s < strend) { + if (isDIGIT_LC(*s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s++; } - else - tmp = 1; - s += UTF8SKIP(s); } break; - case DIGIT: - while (s < strend) { - if (isDIGIT(*s)) { - if (tmp && regtry(prog, s)) + case NDIGIT: + if (do_utf8) { + while (s < strend) { + if (!swash_fetch(PL_utf8_digit,(U8*)s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + } + else { + while (s < strend) { + if (!isDIGIT(*s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s++; + } + } + break; + case NDIGITL: + PL_reg_flags |= RF_tainted; + if (do_utf8) { + while (s < strend) { + if (!isDIGIT_LC_utf8((U8*)s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s += UTF8SKIP(s); + } + } + else { + while (s < strend) { + if (!isDIGIT_LC(*s)) { + if (tmp && (norun || regtry(prog, s))) + goto got_it; + else + tmp = doevery; + } + else + tmp = 1; + s++; + } + } + break; + default: + Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); + break; + } + return 0; + got_it: + return s; +} + +/* + - regexec_flags - match a regexp against a string + */ +I32 +Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend, + 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. */ +/* data: May be used for some additional optimizations. */ +/* nosave: For optimizations. */ +{ + register char *s; + register regnode *c; + register char *startpos = stringarg; + I32 minlen; /* must match at least this many chars */ + I32 dontbother = 0; /* how many characters not to try at end */ + /* 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); + bool do_utf8 = DO_UTF8(sv); + + PL_regcc = 0; + + cache_re(prog); +#ifdef DEBUGGING + PL_regnarrate = PL_debug & 512; +#endif + + /* Be paranoid... */ + if (prog == NULL || startpos == NULL) { + Perl_croak(aTHX_ "NULL regexp parameter"); + return 0; + } + + minlen = prog->minlen; + if (do_utf8) { + if (!(prog->reganch & ROPT_SANY_SEEN)) + if (utf8_distance((U8*)strend, (U8*)startpos) < minlen) goto phooey; + } + else { + if (strend - startpos < minlen) goto phooey; + } + + if (startpos == strbeg) /* is ^ valid at stringarg? */ + PL_regprev = '\n'; + else { + if (prog->reganch & ROPT_UTF8 && do_utf8) { + U8 *s = reghop3((U8*)stringarg, -1, (U8*)strbeg); + PL_regprev = utf8_to_uv(s, (U8*)stringarg - s, NULL, 0); + } + else + 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) != REG_MAGIC) { + Perl_croak(aTHX_ "corrupted regexp program"); + } + + PL_reg_flags = 0; + PL_reg_eval_set = 0; + PL_reg_maxiter = 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 (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */ + MAGIC *mg; + + if (flags & REXEC_IGNOREPOS) /* Means: check only at start */ + PL_reg_ganch = startpos; + else if (sv && SvTYPE(sv) >= SVt_PVMG + && SvMAGIC(sv) + && (mg = mg_find(sv, 'g')) && mg->mg_len >= 0) { + PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */ + if (prog->reganch & ROPT_ANCH_GPOS) { + if (s > PL_reg_ganch) + goto phooey; + s = PL_reg_ganch; + } + } + else /* pos() not defined */ + PL_reg_ganch = strbeg; + } + + if (do_utf8 == (UTF!=0) && + !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) { + re_scream_pos_data d; + + d.scream_olds = &scream_olds; + d.scream_pos = &scream_pos; + s = re_intuit_start(prog, sv, s, strend, flags, &d); + if (!s) + goto phooey; /* not present */ + } + + DEBUG_r( if (!PL_colorset) reginitcolors() ); + DEBUG_r(PerlIO_printf(Perl_debug_log, + "%sMatching REx%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], + (int)(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] */ + if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) { + if (s == startpos && regtry(prog, startpos)) + goto got_it; + else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT) + || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */ + { + char *end; + + if (minlen) + dontbother = minlen - 1; + end = HOP3c(strend, -dontbother, strbeg) - 1; + /* for multiline we only have to try after newlines */ + if (prog->check_substr) { + if (s == startpos) + goto after_try; + while (1) { + if (regtry(prog, s)) goto got_it; + after_try: + if (s >= end) + goto phooey; + if (prog->reganch & RE_USE_INTUIT) { + s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL); + if (!s) + goto phooey; + } else - tmp = doevery; - } - else - tmp = 1; - s++; + s++; + } + } else { + if (s > startpos) + s--; + while (s < end) { + if (*s++ == '\n') { /* don't need PL_utf8skip here */ + if (regtry(prog, s)) + goto got_it; + } + } } - break; - case DIGITUTF8: + } + goto phooey; + } else if (prog->reganch & ROPT_ANCH_GPOS) { + if (regtry(prog, PL_reg_ganch)) + goto got_it; + goto phooey; + } + + /* Messy cases: unanchored match. */ + if (prog->anchored_substr && prog->reganch & ROPT_SKIP) { + /* we have /x+whatever/ */ + /* it must be a one character string (XXXX Except UTF?) */ + char ch = SvPVX(prog->anchored_substr)[0]; +#ifdef DEBUGGING + int did_match = 0; +#endif + + if (do_utf8) { while (s < strend) { - if (swash_fetch(PL_utf8_digit,(U8*)s)) { - if (tmp && regtry(prog, s)) - goto got_it; - else - tmp = doevery; + if (*s == ch) { + DEBUG_r( did_match = 1 ); + if (regtry(prog, s)) goto got_it; + s += UTF8SKIP(s); + while (s < strend && *s == ch) + s += UTF8SKIP(s); } - else - tmp = 1; s += UTF8SKIP(s); } - break; - case DIGITL: - PL_reg_flags |= RF_tainted; + } + else { while (s < strend) { - if (isDIGIT_LC(*s)) { - if (tmp && regtry(prog, s)) - goto got_it; - else - tmp = doevery; + if (*s == ch) { + DEBUG_r( did_match = 1 ); + if (regtry(prog, s)) goto got_it; + s++; + while (s < strend && *s == ch) + s++; } - else - tmp = 1; s++; } - break; - case DIGITLUTF8: - PL_reg_flags |= RF_tainted; - while (s < strend) { - if (isDIGIT_LC_utf8((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)) { - if (tmp && regtry(prog, s)) - goto got_it; - else - tmp = doevery; - } - else - tmp = 1; - s++; + } + DEBUG_r(did_match || + PerlIO_printf(Perl_debug_log, + "Did not find anchored character...\n")); + } + /*SUPPRESS 560*/ + else if (do_utf8 == (UTF!=0) && + (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; + char *last = HOP3c(strend, /* Cannot start after this */ + -(I32)(CHR_SVLEN(must) + - (SvTAIL(must) != 0) + back_min), strbeg); + char *last1; /* Last position checked before */ +#ifdef DEBUGGING + int did_match = 0; +#endif + + 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) && + ((flags & REXEC_SCREAM) + ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg, + end_shift, &scream_pos, 0)) + : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend), + (unsigned char*)strend, must, + PL_multiline ? FBMrf_MULTILINE : 0))) ) { + DEBUG_r( did_match = 1 ); + if (HOPc(s, -back_max) > last1) { + last1 = HOPc(s, -back_min); + s = HOPc(s, -back_max); } - 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); + else { + char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1; + + last1 = HOPc(s, -back_min); + s = t; } - break; - case NDIGITL: - PL_reg_flags |= RF_tainted; - while (s < strend) { - if (!isDIGIT_LC(*s)) { - if (tmp && regtry(prog, s)) + if (do_utf8) { + while (s <= last1) { + if (regtry(prog, s)) goto got_it; - else - tmp = doevery; + s += UTF8SKIP(s); } - else - tmp = 1; - s++; } - break; - case NDIGITLUTF8: - PL_reg_flags |= RF_tainted; - while (s < strend) { - if (!isDIGIT_LC_utf8((U8*)s)) { - if (tmp && regtry(prog, s)) + else { + while (s <= last1) { + if (regtry(prog, s)) goto got_it; - else - tmp = doevery; + s++; } - else - tmp = 1; - s += UTF8SKIP(s); } - break; - default: - Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c)); - break; } + DEBUG_r(did_match || + PerlIO_printf(Perl_debug_log, "Did not find %s substr `%s%.*s%s'%s...\n", + ((must == prog->anchored_substr) + ? "anchored" : "floating"), + PL_colors[0], + (int)(SvCUR(must) - (SvTAIL(must)!=0)), + SvPVX(must), + PL_colors[1], (SvTAIL(must) ? "$" : ""))); + goto phooey; + } + else if ((c = prog->regstclass)) { + if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT) + /* don't bother with what can't match */ + strend = HOPc(strend, -(minlen - 1)); + DEBUG_r({ + SV *prop = sv_newmortal(); + regprop(prop, c); + PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), s); + }); + if (find_byclass(prog, c, s, strend, startpos, 0)) + goto got_it; + DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n")); } else { dontbother = 0; if (prog->float_substr != Nullsv) { /* Trim the end. */ char *last; - I32 oldpos = scream_pos; if (flags & REXEC_SCREAM) { last = screaminstr(sv, prog->float_substr, s - strbeg, end_shift, &scream_pos, 1); /* last one */ if (!last) - last = scream_olds; /* Only one occurence. */ + last = scream_olds; /* Only one occurrence. */ } else { STRLEN len; @@ -1447,14 +1706,19 @@ Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char * last = strend; /* matching `$' */ } } - if (last == NULL) goto phooey; /* Should not happen! */ + if (last == NULL) { + DEBUG_r(PerlIO_printf(Perl_debug_log, + "%sCan't trim the tail, match fails (should not happen)%s\n", + PL_colors[4],PL_colors[5])); + goto phooey; /* Should not happen! */ + } dontbother = strend - last + prog->float_min_offset; } if (minlen && (dontbother < minlen)) dontbother = minlen - 1; strend -= dontbother; /* this one's always in bytes! */ /* We don't know much -- general case. */ - if (UTF) { + if (do_utf8) { for (;;) { if (regtry(prog, s)) goto got_it; @@ -1509,6 +1773,8 @@ got_it: return 1; phooey: + DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n", + PL_colors[4],PL_colors[5])); if (PL_reg_eval_set) restore_pos(aTHXo_ 0); return 0; @@ -1520,12 +1786,14 @@ phooey: STATIC I32 /* 0 failure, 1 success */ S_regtry(pTHX_ regexp *prog, char *startpos) { - dTHR; register I32 i; register I32 *sp; register I32 *ep; CHECKPOINT lastcp; +#ifdef DEBUGGING + PL_regindent = 0; /* XXXX Not good when matches are reenterable... */ +#endif if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) { MAGIC *mg; @@ -1534,12 +1802,12 @@ S_regtry(pTHX_ regexp *prog, char *startpos) PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n", (IV)(PL_stack_sp - PL_stack_base)); )); - SAVEINT(cxstack[cxstack_ix].blk_oldsp); + SAVEI32(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); + /* SAVEI8(cxstack[cxstack_ix].blk_gimme); cxstack[cxstack_ix].blk_gimme = G_SCALAR; */ if (PL_reg_sv) { @@ -1562,7 +1830,7 @@ S_regtry(pTHX_ regexp *prog, char *startpos) SAVEDESTRUCTOR_X(restore_pos, 0); } if (!PL_reg_curpm) - New(22,PL_reg_curpm, 1, PMOP); + Newz(22,PL_reg_curpm, 1, PMOP); PL_reg_curpm->op_pmregexp = prog; PL_reg_oldcurpm = PL_curpm; PL_curpm = PL_reg_curpm; @@ -1597,24 +1865,67 @@ S_regtry(pTHX_ regexp *prog, char *startpos) /* 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! */ + this! --ilya*/ + + /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code. + * Actually, the code in regcppop() (which Ilya may be meaning by + * PL_reglastparen), is not needed at all by the test suite + * (op/regexp, op/pat, op/split), but that code is needed, oddly + * enough, for building DynaLoader, or otherwise this + * "Error: '*' not in typemap in DynaLoader.xs, line 164" + * will happen. Meanwhile, this code *is* needed for the + * above-mentioned test suite tests to succeed. The common theme + * on those tests seems to be returning null fields from matches. + * --jhi */ +#if 1 sp = prog->startp; ep = prog->endp; if (prog->nparens) { - for (i = prog->nparens; i >= 1; i--) { + for (i = prog->nparens; i > *PL_reglastparen; i--) { *++sp = -1; *++ep = -1; } } - REGCP_SET; +#endif + REGCP_SET(lastcp); if (regmatch(prog->program + 1)) { prog->endp[0] = PL_reginput - PL_bostr; return 1; } - REGCP_UNWIND; + REGCP_UNWIND(lastcp); return 0; } +#define RE_UNWIND_BRANCH 1 +#define RE_UNWIND_BRANCHJ 2 + +union re_unwind_t; + +typedef struct { /* XX: makes sense to enlarge it... */ + I32 type; + I32 prev; + CHECKPOINT lastcp; +} re_unwind_generic_t; + +typedef struct { + I32 type; + I32 prev; + CHECKPOINT lastcp; + I32 lastparen; + regnode *next; + char *locinput; + I32 nextchr; +#ifdef DEBUGGING + int regindent; +#endif +} re_unwind_branch_t; + +typedef union re_unwind_t { + I32 type; + re_unwind_generic_t generic; + re_unwind_branch_t branch; +} re_unwind_t; + /* - regmatch - main matching routine * @@ -1632,7 +1943,6 @@ S_regtry(pTHX_ regexp *prog, char *startpos) STATIC I32 /* 0 failure, 1 success */ S_regmatch(pTHX_ regnode *prog) { - dTHR; register regnode *scan; /* Current node. */ regnode *next; /* Next node. */ regnode *inner; /* Next node in internal branch. */ @@ -1644,6 +1954,10 @@ S_regmatch(pTHX_ regnode *prog) register char *locinput = PL_reginput; register I32 c1, c2, paren; /* case fold search, parenth */ int minmod = 0, sw = 0, logical = 0; + I32 unwind = 0; + I32 firstcp = PL_savestack_ix; + register bool do_utf8 = DO_UTF8(PL_reg_sv); + #ifdef DEBUGGING PL_regindent++; #endif @@ -1653,7 +1967,7 @@ S_regmatch(pTHX_ regnode *prog) scan = prog; while (scan != NULL) { #define sayNO_L (logical ? (logical = 0, sw = 0, goto cont) : sayNO) -#ifdef DEBUGGING +#if 1 # define sayYES goto yes # define sayNO goto no # define sayYES_FINAL goto yes_final @@ -1675,20 +1989,25 @@ S_regmatch(pTHX_ regnode *prog) 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 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); + int pref_len = (locinput - PL_bostr) > (5 + taill) - l + ? (5 + taill) - l : locinput - PL_bostr; + int pref0_len; + while (UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len))) + pref_len++; + 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); + while (UTF8_IS_CONTINUATION(*(U8*)(locinput + l))) + l--; if (pref0_len < 0) pref0_len = 0; if (pref0_len > pref_len) @@ -1733,7 +2052,7 @@ S_regmatch(pTHX_ regnode *prog) } sayNO; case SBOL: - if (locinput == PL_regbol && PL_regprev == '\n') + if (locinput == PL_bostr) break; sayNO; case GPOS: @@ -1761,43 +2080,52 @@ S_regmatch(pTHX_ 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 ANYUTF8: - if (nextchr & 0x80) { + case REG_ANY: + if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n') + sayNO; + if (do_utf8) { 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); + else + nextchr = UCHARAT(++locinput); break; case EXACT: s = STRING(scan); ln = STR_LEN(scan); + if (do_utf8 != (UTF!=0)) { + char *l = locinput; + char *e = s + ln; + STRLEN len; + if (do_utf8) + while (s < e) { + if (l >= PL_regeol) + sayNO; + if (*((U8*)s) != utf8_to_uv_simple((U8*)l, &len)) + sayNO; + s++; + l += len; + } + else + while (s < e) { + if (l >= PL_regeol) + sayNO; + if (*((U8*)l) != utf8_to_uv_simple((U8*)s, &len)) + sayNO; + s += len; + l++; + } + locinput = l; + nextchr = UCHARAT(locinput); + break; + } /* Inline the first character, for speed. */ if (UCHARAT(s) != nextchr) sayNO; @@ -1815,20 +2143,19 @@ S_regmatch(pTHX_ regnode *prog) s = STRING(scan); ln = STR_LEN(scan); - if (UTF) { + if (do_utf8) { char *l = locinput; - char *e = s + ln; + char *e; + 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))) - { + if (l >= PL_regeol) { sayNO; } - s += UTF8SKIP(s); + if ((UTF ? utf8_to_uv((U8*)s, e - s, 0, 0) : *((U8*)s)) != + (c1 ? toLOWER_utf8((U8*)l) : toLOWER_LC_utf8((U8*)l))) + sayNO; + s += UTF ? UTF8SKIP(s) : 1; l += UTF8SKIP(l); } locinput = l; @@ -1850,42 +2177,33 @@ S_regmatch(pTHX_ regnode *prog) locinput += ln; nextchr = UCHARAT(locinput); break; - case ANYOFUTF8: - if (!REGINCLASSUTF8(scan, (U8*)locinput)) - sayNO; - if (locinput >= PL_regeol) - sayNO; - locinput += PL_utf8skip[nextchr]; - nextchr = UCHARAT(locinput); - break; case ANYOF: - if (nextchr < 0) + if (do_utf8) { + if (!reginclass(scan, (U8*)locinput, do_utf8)) + sayNO; + if (locinput >= PL_regeol) + sayNO; + locinput += PL_utf8skip[nextchr]; nextchr = UCHARAT(locinput); - if (!REGINCLASS(scan, nextchr)) - sayNO; - if (!nextchr && locinput >= PL_regeol) - sayNO; - nextchr = UCHARAT(++locinput); - break; - case ALNUML: - PL_reg_flags |= RF_tainted; - /* FALL THROUGH */ - case ALNUM: - if (!nextchr) - sayNO; - if (!(OP(scan) == ALNUM - ? isALNUM(nextchr) : isALNUM_LC(nextchr))) - sayNO; - nextchr = UCHARAT(++locinput); + } + else { + if (nextchr < 0) + nextchr = UCHARAT(locinput); + if (!reginclass(scan, (U8*)locinput, do_utf8)) + sayNO; + if (!nextchr && locinput >= PL_regeol) + sayNO; + nextchr = UCHARAT(++locinput); + } break; - case ALNUMLUTF8: + case ALNUML: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ - case ALNUMUTF8: + case ALNUM: if (!nextchr) sayNO; - if (nextchr & 0x80) { - if (!(OP(scan) == ALNUMUTF8 + if (do_utf8) { + if (!(OP(scan) == ALNUM ? swash_fetch(PL_utf8_alnum, (U8*)locinput) : isALNUM_LC_utf8((U8*)locinput))) { @@ -1895,7 +2213,7 @@ S_regmatch(pTHX_ regnode *prog) nextchr = UCHARAT(locinput); break; } - if (!(OP(scan) == ALNUMUTF8 + if (!(OP(scan) == ALNUM ? isALNUM(nextchr) : isALNUM_LC(nextchr))) sayNO; nextchr = UCHARAT(++locinput); @@ -1906,19 +2224,8 @@ S_regmatch(pTHX_ regnode *prog) 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 NALNUMUTF8: - if (!nextchr && locinput >= PL_regeol) - sayNO; - if (nextchr & 0x80) { - if (OP(scan) == NALNUMUTF8 + if (do_utf8) { + if (OP(scan) == NALNUM ? swash_fetch(PL_utf8_alnum, (U8*)locinput) : isALNUM_LC_utf8((U8*)locinput)) { @@ -1928,7 +2235,7 @@ S_regmatch(pTHX_ regnode *prog) nextchr = UCHARAT(locinput); break; } - if (OP(scan) == NALNUMUTF8 + if (OP(scan) == NALNUM ? isALNUM(nextchr) : isALNUM_LC(nextchr)) sayNO; nextchr = UCHARAT(++locinput); @@ -1940,91 +2247,78 @@ S_regmatch(pTHX_ regnode *prog) case BOUND: case NBOUND: /* was last char in word? */ - ln = (locinput != PL_regbol) ? UCHARAT(locinput - 1) : PL_regprev; - 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 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); + if (do_utf8) { + if (locinput == PL_regbol) + ln = PL_regprev; + else { + U8 *r = reghop((U8*)locinput, -1); + + ln = utf8_to_uv(r, s - (char*)r, 0, 0); + } + if (OP(scan) == BOUND || OP(scan) == NBOUND) { + ln = isALNUM_uni(ln); + n = swash_fetch(PL_utf8_alnum, (U8*)locinput); + } + else { + ln = isALNUM_LC_uni(ln); + n = isALNUM_LC_utf8((U8*)locinput); + } } else { - ln = isALNUM_LC_uni(ln); - n = isALNUM_LC_utf8((U8*)locinput); + ln = (locinput != PL_regbol) ? + UCHARAT(locinput - 1) : PL_regprev; + 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) == BOUNDUTF8 || OP(scan) == BOUNDLUTF8)) - sayNO; + if (((!ln) == (!n)) == (OP(scan) == BOUND || + OP(scan) == BOUNDL)) + sayNO; break; case SPACEL: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case SPACE: - if (!nextchr && locinput >= PL_regeol) - sayNO; - if (!(OP(scan) == SPACE - ? 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) + if (!nextchr) sayNO; - if (nextchr & 0x80) { - if (!(OP(scan) == SPACEUTF8 - ? swash_fetch(PL_utf8_space,(U8*)locinput) - : isSPACE_LC_utf8((U8*)locinput))) - { - sayNO; + if (do_utf8) { + if (UTF8_IS_CONTINUED(nextchr)) { + if (!(OP(scan) == SPACE + ? swash_fetch(PL_utf8_space, (U8*)locinput) + : isSPACE_LC_utf8((U8*)locinput))) + { + sayNO; + } + locinput += PL_utf8skip[nextchr]; + nextchr = UCHARAT(locinput); + break; } - locinput += PL_utf8skip[nextchr]; - 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) == SPACEUTF8 - ? isSPACE(nextchr) : isSPACE_LC(nextchr))) - sayNO; - nextchr = UCHARAT(++locinput); break; case NSPACEL: PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case NSPACE: - if (!nextchr) - sayNO; - if (OP(scan) == SPACE - ? isSPACE(nextchr) : isSPACE_LC(nextchr)) - sayNO; - nextchr = UCHARAT(++locinput); - break; - case NSPACELUTF8: - PL_reg_flags |= RF_tainted; - /* FALL THROUGH */ - case NSPACEUTF8: - if (!nextchr) + if (!nextchr && locinput >= PL_regeol) sayNO; - if (nextchr & 0x80) { - if (OP(scan) == NSPACEUTF8 - ? swash_fetch(PL_utf8_space,(U8*)locinput) + if (do_utf8) { + if (OP(scan) == NSPACE + ? swash_fetch(PL_utf8_space, (U8*)locinput) : isSPACE_LC_utf8((U8*)locinput)) { sayNO; @@ -2033,7 +2327,7 @@ S_regmatch(pTHX_ regnode *prog) nextchr = UCHARAT(locinput); break; } - if (OP(scan) == NSPACEUTF8 + if (OP(scan) == NSPACE ? isSPACE(nextchr) : isSPACE_LC(nextchr)) sayNO; nextchr = UCHARAT(++locinput); @@ -2042,23 +2336,12 @@ S_regmatch(pTHX_ regnode *prog) PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case DIGIT: - if (!nextchr && locinput >= PL_regeol) - sayNO; - if (!(OP(scan) == DIGIT - ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))) - sayNO; - nextchr = UCHARAT(++locinput); - break; - case DIGITLUTF8: - PL_reg_flags |= RF_tainted; - /* FALL THROUGH */ - case DIGITUTF8: if (!nextchr) sayNO; - if (nextchr & 0x80) { - if (OP(scan) == NDIGITUTF8 - ? swash_fetch(PL_utf8_digit,(U8*)locinput) - : isDIGIT_LC_utf8((U8*)locinput)) + if (do_utf8) { + if (!(OP(scan) == DIGIT + ? swash_fetch(PL_utf8_digit, (U8*)locinput) + : isDIGIT_LC_utf8((U8*)locinput))) { sayNO; } @@ -2066,7 +2349,8 @@ S_regmatch(pTHX_ regnode *prog) nextchr = UCHARAT(locinput); break; } - if (!isDIGIT(nextchr)) + if (!(OP(scan) == DIGIT + ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))) sayNO; nextchr = UCHARAT(++locinput); break; @@ -2074,27 +2358,21 @@ S_regmatch(pTHX_ regnode *prog) PL_reg_flags |= RF_tainted; /* FALL THROUGH */ case NDIGIT: - if (!nextchr) - sayNO; - if (OP(scan) == DIGIT - ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)) - sayNO; - nextchr = UCHARAT(++locinput); - break; - case NDIGITLUTF8: - PL_reg_flags |= RF_tainted; - /* FALL THROUGH */ - case NDIGITUTF8: if (!nextchr && locinput >= PL_regeol) sayNO; - if (nextchr & 0x80) { - if (swash_fetch(PL_utf8_digit,(U8*)locinput)) + if (do_utf8) { + if (OP(scan) == NDIGIT + ? swash_fetch(PL_utf8_digit, (U8*)locinput) + : isDIGIT_LC_utf8((U8*)locinput)) + { sayNO; + } locinput += PL_utf8skip[nextchr]; nextchr = UCHARAT(locinput); break; } - if (isDIGIT(nextchr)) + if (OP(scan) == NDIGIT + ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)) sayNO; nextchr = UCHARAT(++locinput); break; @@ -2122,7 +2400,7 @@ S_regmatch(pTHX_ regnode *prog) break; s = PL_bostr + ln; - if (UTF && OP(scan) != REF) { /* REF can do byte comparison */ + if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */ char *l = locinput; char *e = PL_bostr + PL_regendp[n]; /* @@ -2251,7 +2529,7 @@ S_regmatch(pTHX_ regnode *prog) PL_regcc = 0; cp = regcppush(0); /* Save *all* the positions. */ - REGCP_SET; + REGCP_SET(lastcp); cache_re(re); state.ss = PL_savestack_ix; *PL_reglastparen = 0; @@ -2281,7 +2559,7 @@ S_regmatch(pTHX_ regnode *prog) sayYES; } ReREFCNT_dec(re); - REGCP_UNWIND; + REGCP_UNWIND(lastcp); regcppop(); PL_reg_call_cc = state.prev; PL_regcc = state.cc; @@ -2408,12 +2686,18 @@ S_regmatch(pTHX_ regnode *prog) case CURLYX: { CURCUR cc; CHECKPOINT cp = PL_savestack_ix; + /* No need to save/restore up to this paren */ + I32 parenfloor = scan->flags; if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */ next += ARG(next); cc.oldcc = PL_regcc; PL_regcc = &cc; - cc.parenfloor = *PL_reglastparen; + /* XXXX Probably it is better to teach regpush to support + parenfloor > PL_regsize... */ + if (parenfloor > *PL_reglastparen) + parenfloor = *PL_reglastparen; /* Pessimization... */ + cc.parenfloor = parenfloor; cc.cur = -1; cc.min = ARG1(scan); cc.max = ARG2(scan); @@ -2536,22 +2820,22 @@ S_regmatch(pTHX_ regnode *prog) if (PL_regcc) ln = PL_regcc->cur; cp = regcppush(cc->parenfloor); - REGCP_SET; + REGCP_SET(lastcp); if (regmatch(cc->next)) { regcpblow(cp); sayYES; /* All done. */ } - REGCP_UNWIND; + REGCP_UNWIND(lastcp); regcppop(); if (PL_regcc) PL_regcc->cur = ln; PL_regcc = cc; if (n >= cc->max) { /* Maximum greed exceeded? */ - if (ckWARN(WARN_UNSAFE) && n >= REG_INFTY + if (ckWARN(WARN_REGEXP) && n >= REG_INFTY && !(PL_reg_flags & RF_warned)) { PL_reg_flags |= RF_warned; - Perl_warner(aTHX_ WARN_UNSAFE, "%s limit (%d) exceeded", + Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded", "Complex regular subexpression recursion", REG_INFTY - 1); } @@ -2568,12 +2852,12 @@ S_regmatch(pTHX_ regnode *prog) cc->cur = n; cc->lastloc = locinput; cp = regcppush(cc->parenfloor); - REGCP_SET; + REGCP_SET(lastcp); if (regmatch(cc->scan)) { regcpblow(cp); sayYES; } - REGCP_UNWIND; + REGCP_UNWIND(lastcp); regcppop(); cc->cur = n - 1; cc->lastloc = lastloc; @@ -2586,12 +2870,12 @@ S_regmatch(pTHX_ regnode *prog) cp = regcppush(cc->parenfloor); cc->cur = n; cc->lastloc = locinput; - REGCP_SET; + REGCP_SET(lastcp); if (regmatch(cc->scan)) { regcpblow(cp); sayYES; } - REGCP_UNWIND; + REGCP_UNWIND(lastcp); regcppop(); /* Restore some previous $s? */ PL_reginput = locinput; DEBUG_r( @@ -2600,10 +2884,10 @@ S_regmatch(pTHX_ regnode *prog) REPORT_CODE_OFF+PL_regindent*2, "") ); } - if (ckWARN(WARN_UNSAFE) && n >= REG_INFTY + if (ckWARN(WARN_REGEXP) && n >= REG_INFTY && !(PL_reg_flags & RF_warned)) { PL_reg_flags |= RF_warned; - Perl_warner(aTHX_ WARN_UNSAFE, "%s limit (%d) exceeded", + Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded", "Complex regular subexpression recursion", REG_INFTY - 1); } @@ -2637,30 +2921,30 @@ S_regmatch(pTHX_ regnode *prog) if (OP(next) != c1) /* No choice. */ next = inner; /* Avoid recursion. */ else { - int lastparen = *PL_reglastparen; + I32 lastparen = *PL_reglastparen; + I32 unwind1; + re_unwind_branch_t *uw; + + /* Put unwinding data on stack */ + unwind1 = SSNEWt(1,re_unwind_branch_t); + uw = SSPTRt(unwind1,re_unwind_branch_t); + uw->prev = unwind; + unwind = unwind1; + uw->type = ((c1 == BRANCH) + ? RE_UNWIND_BRANCH + : RE_UNWIND_BRANCHJ); + uw->lastparen = lastparen; + uw->next = next; + uw->locinput = locinput; + uw->nextchr = nextchr; +#ifdef DEBUGGING + uw->regindent = ++PL_regindent; +#endif - REGCP_SET; - do { - PL_reginput = locinput; - if (regmatch(inner)) - sayYES; - REGCP_UNWIND; - for (n = *PL_reglastparen; n > lastparen; n--) - PL_regendp[n] = -1; - *PL_reglastparen = n; - scan = next; - /*SUPPRESS 560*/ - if (n = (c1 == BRANCH ? NEXT_OFF(next) : ARG(next))) - next += n; - else - next = NULL; - inner = NEXTOPER(scan); - if (c1 == BRANCHJ) { - inner = NEXTOPER(inner); - } - } while (scan != NULL && OP(scan) == c1); - sayNO; - /* NOTREACHED */ + REGCP_SET(uw->lastcp); + + /* Now go into the first branch */ + next = inner; } } break; @@ -2710,7 +2994,7 @@ S_regmatch(pTHX_ regnode *prog) } else c1 = c2 = -1000; - REGCP_SET; + REGCP_SET(lastcp); /* This may be improved if l == 0. */ while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */ /* If it could work, try it. */ @@ -2729,7 +3013,7 @@ S_regmatch(pTHX_ regnode *prog) } if (regmatch(next)) sayYES; - REGCP_UNWIND; + REGCP_UNWIND(lastcp); } /* Couldn't or didn't -- move forward. */ PL_reginput = locinput; @@ -2769,7 +3053,7 @@ S_regmatch(pTHX_ regnode *prog) else c1 = c2 = -1000; } - REGCP_SET; + REGCP_SET(lastcp); while (n >= ln) { /* If it could work, try it. */ if (c1 == -1000 || @@ -2791,7 +3075,7 @@ S_regmatch(pTHX_ regnode *prog) } if (regmatch(next)) sayYES; - REGCP_UNWIND; + REGCP_UNWIND(lastcp); } /* Couldn't or didn't -- back up. */ n--; @@ -2835,13 +3119,23 @@ S_regmatch(pTHX_ regnode *prog) * when we know what character comes next. */ if (PL_regkind[(U8)OP(next)] == EXACT) { - c1 = (U8)*STRING(next); - if (OP(next) == EXACTF) - c2 = PL_fold[c1]; - else if (OP(next) == EXACTFL) - c2 = PL_fold_locale[c1]; - else - c2 = c1; + U8 *s = (U8*)STRING(next); + if (!UTF) { + c2 = c1 = *s; + if (OP(next) == EXACTF) + c2 = PL_fold[c1]; + else if (OP(next) == EXACTFL) + c2 = PL_fold_locale[c1]; + } + else { /* UTF */ + if (OP(next) == EXACTF) { + c1 = to_utf8_lower(s); + c2 = to_utf8_upper(s); + } + else { + c2 = c1 = utf8_to_uv_simple(s, NULL); + } + } } else c1 = c2 = -1000; @@ -2852,68 +3146,95 @@ S_regmatch(pTHX_ regnode *prog) if (ln && regrepeat(scan, ln) < ln) sayNO; locinput = PL_reginput; - REGCP_SET; + REGCP_SET(lastcp); if (c1 != -1000) { - char *e = locinput + n - ln; /* Should not check after this */ + char *e; /* Should not check after this */ char *old = locinput; - if (e >= PL_regeol || (n == REG_INFTY)) + if (n == REG_INFTY) { e = PL_regeol - 1; + if (do_utf8) + while (UTF8_IS_CONTINUATION(*(U8*)e)) + e--; + } + else if (do_utf8) { + int m = n - ln; + for (e = locinput; + m >0 && e + UTF8SKIP(e) <= PL_regeol; m--) + e += UTF8SKIP(e); + } + else { + e = locinput + n - ln; + if (e >= PL_regeol) + e = PL_regeol - 1; + } while (1) { + int count; /* Find place 'next' could work */ - if (c1 == c2) { - while (locinput <= e && *locinput != c1) - locinput++; - } else { - while (locinput <= e - && *locinput != c1 - && *locinput != c2) - locinput++; + if (!do_utf8) { + if (c1 == c2) { + while (locinput <= e && *locinput != c1) + locinput++; + } else { + while (locinput <= e + && *locinput != c1 + && *locinput != c2) + locinput++; + } + count = locinput - old; + } + else { + STRLEN len; + if (c1 == c2) { + for (count = 0; + locinput <= e && + utf8_to_uv_simple((U8*)locinput, &len) != c1; + count++) + locinput += len; + + } else { + for (count = 0; locinput <= e; count++) { + UV c = utf8_to_uv_simple((U8*)locinput, &len); + if (c == c1 || c == c2) + break; + locinput += len; + } + } } if (locinput > e) sayNO; /* PL_reginput == old now */ if (locinput != old) { ln = 1; /* Did some */ - if (regrepeat(scan, locinput - old) < - locinput - old) + if (regrepeat(scan, count) < count) sayNO; } /* PL_reginput == locinput now */ - if (paren) { - if (ln) { - PL_regstartp[paren] = HOPc(locinput, -1) - PL_bostr; - PL_regendp[paren] = locinput - PL_bostr; - } - else - PL_regendp[paren] = -1; - } - if (regmatch(next)) - sayYES; + TRYPAREN(paren, ln, locinput); PL_reginput = locinput; /* Could be reset... */ - REGCP_UNWIND; + REGCP_UNWIND(lastcp); /* Couldn't or didn't -- move forward. */ - old = locinput++; + old = locinput; + if (do_utf8) + locinput += UTF8SKIP(locinput); + else + locinput++; } } else while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */ + UV c; + if (c1 != -1000) { + if (do_utf8) + c = utf8_to_uv_simple((U8*)PL_reginput, NULL); + else + c = UCHARAT(PL_reginput); + } /* If it could work, try it. */ - if (c1 == -1000 || - UCHARAT(PL_reginput) == c1 || - UCHARAT(PL_reginput) == c2) + if (c1 == -1000 || c == c1 || c == c2) { - if (paren) { - if (n) { - PL_regstartp[paren] = HOPc(PL_reginput, -1) - PL_bostr; - PL_regendp[paren] = PL_reginput - PL_bostr; - } - else - PL_regendp[paren] = -1; - } - if (regmatch(next)) - sayYES; - REGCP_UNWIND; + TRYPAREN(paren, n, PL_reginput); + REGCP_UNWIND(lastcp); } /* Couldn't or didn't -- move forward. */ PL_reginput = locinput; @@ -2930,27 +3251,29 @@ S_regmatch(pTHX_ regnode *prog) n = regrepeat(scan, n); locinput = PL_reginput; if (ln < n && PL_regkind[(U8)OP(next)] == EOL && - (!PL_multiline || OP(next) == SEOL)) + (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) { ln = n; /* why back off? */ - REGCP_SET; + /* ...because $ and \Z can match before *and* after + newline at the end. Consider "\n\n" =~ /\n+\Z\n/. + We should back off by one in this case. */ + if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS) + ln--; + } + REGCP_SET(lastcp); if (paren) { + UV c; while (n >= ln) { + if (c1 != -1000) { + if (do_utf8) + c = utf8_to_uv_simple((U8*)PL_reginput, NULL); + else + c = UCHARAT(PL_reginput); + } /* If it could work, try it. */ - if (c1 == -1000 || - UCHARAT(PL_reginput) == c1 || - UCHARAT(PL_reginput) == c2) + if (c1 == -1000 || c == c1 || c == c2) { - if (paren && n) { - if (n) { - PL_regstartp[paren] = HOPc(PL_reginput, -1) - PL_bostr; - PL_regendp[paren] = PL_reginput - PL_bostr; - } - else - PL_regendp[paren] = -1; - } - if (regmatch(next)) - sayYES; - REGCP_UNWIND; + TRYPAREN(paren, n, PL_reginput); + REGCP_UNWIND(lastcp); } /* Couldn't or didn't -- back up. */ n--; @@ -2958,15 +3281,19 @@ S_regmatch(pTHX_ regnode *prog) } } else { + UV c; while (n >= ln) { + if (c1 != -1000) { + if (do_utf8) + c = utf8_to_uv_simple((U8*)PL_reginput, NULL); + else + c = UCHARAT(PL_reginput); + } /* If it could work, try it. */ - if (c1 == -1000 || - UCHARAT(PL_reginput) == c1 || - UCHARAT(PL_reginput) == c2) + if (c1 == -1000 || c == c1 || c == c2) { - if (regmatch(next)) - sayYES; - REGCP_UNWIND; + TRYPAREN(paren, n, PL_reginput); + REGCP_UNWIND(lastcp); } /* Couldn't or didn't -- back up. */ n--; @@ -2984,7 +3311,7 @@ S_regmatch(pTHX_ regnode *prog) CHECKPOINT cp, lastcp; cp = regcppush(0); /* Save *all* the positions. */ - REGCP_SET; + REGCP_SET(lastcp); regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of the caller. */ PL_reginput = locinput; /* Make position available to @@ -2997,7 +3324,7 @@ S_regmatch(pTHX_ regnode *prog) regcpblow(cp); sayYES; } - REGCP_UNWIND; + REGCP_UNWIND(lastcp); regcppop(); PL_reg_call_cc = cur_call_cc; PL_regcc = cctmp; @@ -3104,6 +3431,7 @@ S_regmatch(pTHX_ regnode *prog) PTR2UV(scan), OP(scan)); Perl_croak(aTHX_ "regexp memory corruption"); } + reenter: scan = next; } @@ -3129,6 +3457,11 @@ yes: #ifdef DEBUGGING PL_regindent--; #endif + +#if 0 /* Breaks $^R */ + if (unwind) + regcpblow(firstcp); +#endif return 1; no: @@ -3140,6 +3473,55 @@ no: goto do_no; no_final: do_no: + if (unwind) { + re_unwind_t *uw = SSPTRt(unwind,re_unwind_t); + + switch (uw->type) { + case RE_UNWIND_BRANCH: + case RE_UNWIND_BRANCHJ: + { + re_unwind_branch_t *uwb = &(uw->branch); + I32 lastparen = uwb->lastparen; + + REGCP_UNWIND(uwb->lastcp); + for (n = *PL_reglastparen; n > lastparen; n--) + PL_regendp[n] = -1; + *PL_reglastparen = n; + scan = next = uwb->next; + if ( !scan || + OP(scan) != (uwb->type == RE_UNWIND_BRANCH + ? BRANCH : BRANCHJ) ) { /* Failure */ + unwind = uwb->prev; +#ifdef DEBUGGING + PL_regindent--; +#endif + goto do_no; + } + /* Have more choice yet. Reuse the same uwb. */ + /*SUPPRESS 560*/ + if ((n = (uwb->type == RE_UNWIND_BRANCH + ? NEXT_OFF(next) : ARG(next)))) + next += n; + else + next = NULL; /* XXXX Needn't unwinding in this case... */ + uwb->next = next; + next = NEXTOPER(scan); + if (uwb->type == RE_UNWIND_BRANCHJ) + next = NEXTOPER(next); + locinput = uwb->locinput; + nextchr = uwb->nextchr; +#ifdef DEBUGGING + PL_regindent = uwb->regindent; +#endif + + goto reenter; + } + /* NOT REACHED */ + default: + Perl_croak(aTHX_ "regexp unwind memory corruption"); + } + /* NOT REACHED */ + } #ifdef DEBUGGING PL_regindent--; #endif @@ -3157,37 +3539,31 @@ do_no: STATIC I32 S_regrepeat(pTHX_ regnode *p, I32 max) { - dTHR; register char *scan; register I32 c; register char *loceol = PL_regeol; register I32 hardcount = 0; + register bool do_utf8 = DO_UTF8(PL_reg_sv); scan = PL_reginput; if (max != REG_INFTY && max < loceol - scan) loceol = scan + max; switch (OP(p)) { case REG_ANY: - while (scan < loceol && *scan != '\n') - scan++; + if (do_utf8) { + loceol = PL_regeol; + while (scan < loceol && hardcount < max && *scan != '\n') { + scan += UTF8SKIP(scan); + hardcount++; + } + } else { + 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 = (U8)*STRING(p); while (scan < loceol && UCHARAT(scan) == c) @@ -3206,135 +3582,151 @@ S_regrepeat(pTHX_ regnode *p, I32 max) (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(p, *scan)) - scan++; + if (do_utf8) { + loceol = PL_regeol; + while (hardcount < max && scan < loceol && + reginclass(p, (U8*)scan, do_utf8)) { + scan += UTF8SKIP(scan); + hardcount++; + } + } else { + while (scan < loceol && reginclass(p, (U8*)scan, do_utf8)) + 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++; + if (do_utf8) { + loceol = PL_regeol; + while (hardcount < max && scan < loceol && + swash_fetch(PL_utf8_alnum, (U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + } else { + while (scan < loceol && isALNUM(*scan)) + scan++; } 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++; + if (do_utf8) { + loceol = PL_regeol; + while (hardcount < max && scan < loceol && + isALNUM_LC_utf8((U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + } else { + while (scan < loceol && isALNUM_LC(*scan)) + scan++; } 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++; + if (do_utf8) { + loceol = PL_regeol; + while (hardcount < max && scan < loceol && + !swash_fetch(PL_utf8_alnum, (U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + } else { + while (scan < loceol && !isALNUM(*scan)) + scan++; } 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++; + if (do_utf8) { + loceol = PL_regeol; + while (hardcount < max && scan < loceol && + !isALNUM_LC_utf8((U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + } else { + while (scan < loceol && !isALNUM_LC(*scan)) + scan++; } 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++; + if (do_utf8) { + loceol = PL_regeol; + while (hardcount < max && scan < loceol && + (*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan))) { + scan += UTF8SKIP(scan); + hardcount++; + } + } else { + while (scan < loceol && isSPACE(*scan)) + scan++; } 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++; + if (do_utf8) { + loceol = PL_regeol; + while (hardcount < max && scan < loceol && + (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) { + scan += UTF8SKIP(scan); + hardcount++; + } + } else { + while (scan < loceol && isSPACE_LC(*scan)) + scan++; } 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++; + if (do_utf8) { + loceol = PL_regeol; + while (hardcount < max && scan < loceol && + !(*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan))) { + scan += UTF8SKIP(scan); + hardcount++; + } + } else { + while (scan < loceol && !isSPACE(*scan)) + scan++; + break; } - 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++; + if (do_utf8) { + loceol = PL_regeol; + while (hardcount < max && scan < loceol && + !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) { + scan += UTF8SKIP(scan); + hardcount++; + } + } else { + while (scan < loceol && !isSPACE_LC(*scan)) + scan++; } 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++; + if (do_utf8) { + loceol = PL_regeol; + while (hardcount < max && scan < loceol && + swash_fetch(PL_utf8_digit,(U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + } else { + while (scan < loceol && isDIGIT(*scan)) + scan++; } 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++; + if (do_utf8) { + loceol = PL_regeol; + while (hardcount < max && scan < loceol && + !swash_fetch(PL_utf8_digit,(U8*)scan)) { + scan += UTF8SKIP(scan); + hardcount++; + } + } else { + while (scan < loceol && !isDIGIT(*scan)) + scan++; } break; default: /* Called on something of 0 width. */ @@ -3369,7 +3761,6 @@ S_regrepeat(pTHX_ regnode *p, I32 max) STATIC I32 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp) { - dTHR; register char *scan; register char *start; register char *loceol = PL_regeol; @@ -3380,7 +3771,7 @@ S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp) return 0; start = PL_reginput; - if (UTF) { + if (DO_UTF8(PL_reg_sv)) { while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) { if (!count++) { l = 0; @@ -3414,93 +3805,141 @@ S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp) } /* - - reginclass - determine if a character falls into a character class - */ +- regclass_swash - prepare the utf8 swash +*/ -STATIC bool -S_reginclass(pTHX_ register regnode *p, register I32 c) +SV * +Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp) { - dTHR; - char flags = ANYOF_FLAGS(p); - bool match = FALSE; + SV *sw = NULL; + SV *si = NULL; - c &= 0xFF; - if (ANYOF_BITMAP_TEST(p, c)) - match = TRUE; - else if (flags & ANYOF_FOLD) { - I32 cf; - if (flags & ANYOF_LOCALE) { - PL_reg_flags |= RF_tainted; - cf = PL_fold_locale[c]; - } - else - cf = PL_fold[c]; - if (ANYOF_BITMAP_TEST(p, cf)) - match = TRUE; - } + if (PL_regdata && PL_regdata->count) { + U32 n = ARG(node); - if (!match && (flags & ANYOF_CLASS)) { - PL_reg_flags |= RF_tainted; - if ( - (ANYOF_CLASS_TEST(p, ANYOF_ALNUM) && isALNUM_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NALNUM) && !isALNUM_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_SPACE) && isSPACE_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NSPACE) && !isSPACE_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_DIGIT) && isDIGIT_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NDIGIT) && !isDIGIT_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_ALNUMC) && isALNUMC_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NALNUMC) && !isALNUMC_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_ALPHA) && isALPHA_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NALPHA) && !isALPHA_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_ASCII) && isASCII(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NASCII) && !isASCII(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_CNTRL) && isCNTRL_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NCNTRL) && !isCNTRL_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_GRAPH) && isGRAPH_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NGRAPH) && !isGRAPH_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_LOWER) && isLOWER_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NLOWER) && !isLOWER_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_PRINT) && isPRINT_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NPRINT) && !isPRINT_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_PUNCT) && isPUNCT_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NPUNCT) && !isPUNCT_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_UPPER) && isUPPER_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NUPPER) && !isUPPER_LC(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_XDIGIT) && isXDIGIT(c)) || - (ANYOF_CLASS_TEST(p, ANYOF_NXDIGIT) && !isXDIGIT(c)) - ) /* How's that for a conditional? */ - { - match = TRUE; + if (PL_regdata->what[n] == 's') { + SV *rv = (SV*)PL_regdata->data[n]; + AV *av = (AV*)SvRV((SV*)rv); + SV **a; + + si = *av_fetch(av, 0, FALSE); + a = av_fetch(av, 1, FALSE); + + if (a) + sw = *a; + else if (si && doinit) { + sw = swash_init("utf8", "", si, 1, 0); + (void)av_store(av, 1, sw); + } } } + + if (initsvp) + *initsvp = si; - return (flags & ANYOF_INVERT) ? !match : match; + return sw; } +/* + - reginclass - determine if a character falls into a character class + */ + STATIC bool -S_reginclassutf8(pTHX_ regnode *f, U8 *p) -{ - dTHR; - char flags = ARG1(f); +S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8) +{ + char flags = ANYOF_FLAGS(n); 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)); + UV c; + STRLEN len; + + if (do_utf8) + c = utf8_to_uv_simple(p, &len); + else + c = *p; + + if (do_utf8 || (flags & ANYOF_UNICODE)) { + if (do_utf8 && !ANYOF_RUNTIME(n)) { + if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c)) + match = TRUE; } - else - uv_to_utf8(tmpbuf, toLOWER_utf8(p)); - if (swash_fetch(sv, tmpbuf)) + if (do_utf8 && flags & ANYOF_UNICODE_ALL && c >= 256) match = TRUE; + if (!match) { + SV *sw = regclass_swash(n, TRUE, 0); + + if (sw) { + if (swash_fetch(sw, p)) + match = TRUE; + else if (flags & ANYOF_FOLD) { + U8 tmpbuf[UTF8_MAXLEN+1]; + + 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(sw, tmpbuf)) + match = TRUE; + } + } + } } + if (!match && c < 256) { + if (ANYOF_BITMAP_TEST(n, c)) + match = TRUE; + else if (flags & ANYOF_FOLD) { + I32 f; - /* UTF8 combined with ANYOF_CLASS is ill-defined. */ + if (flags & ANYOF_LOCALE) { + PL_reg_flags |= RF_tainted; + f = PL_fold_locale[c]; + } + else + f = PL_fold[c]; + if (f != c && ANYOF_BITMAP_TEST(n, f)) + match = TRUE; + } + + if (!match && (flags & ANYOF_CLASS)) { + PL_reg_flags |= RF_tainted; + if ( + (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) || + (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c)) + ) /* How's that for a conditional? */ + { + match = TRUE; + } + } + } return (flags & ANYOF_INVERT) ? !match : match; } @@ -3508,19 +3947,27 @@ S_reginclassutf8(pTHX_ regnode *f, U8 *p) STATIC U8 * S_reghop(pTHX_ U8 *s, I32 off) { - dTHR; + return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)); +} + +STATIC U8 * +S_reghop3(pTHX_ U8 *s, I32 off, U8* lim) +{ if (off >= 0) { - while (off-- && s < (U8*)PL_regeol) + while (off-- && s < lim) { + /* XXX could check well-formedness here */ s += UTF8SKIP(s); + } } else { while (off++) { - if (s > (U8*)PL_bostr) { + if (s > lim) { s--; - if (*s & 0x80) { - while (s > (U8*)PL_bostr && (*s & 0xc0) == 0x80) + if (UTF8_IS_CONTINUED(*s)) { + while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s)) s--; - } /* XXX could check well-formedness here */ + } + /* XXX could check well-formedness here */ } } } @@ -3528,23 +3975,31 @@ S_reghop(pTHX_ U8 *s, I32 off) } STATIC U8 * -S_reghopmaybe(pTHX_ U8* s, I32 off) +S_reghopmaybe(pTHX_ U8 *s, I32 off) +{ + return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)); +} + +STATIC U8 * +S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim) { - dTHR; if (off >= 0) { - while (off-- && s < (U8*)PL_regeol) + while (off-- && s < lim) { + /* XXX could check well-formedness here */ s += UTF8SKIP(s); + } if (off >= 0) return 0; } else { while (off++) { - if (s > (U8*)PL_bostr) { + if (s > lim) { s--; - if (*s & 0x80) { - while (s > (U8*)PL_bostr && (*s & 0xc0) == 0x80) + if (UTF8_IS_CONTINUED(*s)) { + while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s)) s--; - } /* XXX could check well-formedness here */ + } + /* XXX could check well-formedness here */ } else break; @@ -3562,7 +4017,6 @@ S_reghopmaybe(pTHX_ U8* s, I32 off) static void restore_pos(pTHXo_ void *arg) { - dTHR; if (PL_reg_eval_set) { if (PL_reg_oldsaved) { PL_reg_re->subbeg = PL_reg_oldsaved; @@ -3574,4 +4028,3 @@ restore_pos(pTHXo_ void *arg) PL_curpm = PL_reg_oldcurpm; } } -