X-Git-Url: https://perl5.git.perl.org/perl5.git/blobdiff_plain/b2f4e957e3a5b840ae5e6b1d807b03ef82e766dc..205681b5d1e56b5ec94a53a7a80c51b866099763:/regexec.c diff --git a/regexec.c b/regexec.c index 77e5b55..e74ca18 100644 --- a/regexec.c +++ b/regexec.c @@ -536,12 +536,10 @@ S_isFOO_utf8_lc(pTHX_ const U8 classnum, const U8* character) case _CC_ENUM_BLANK: return is_HORIZWS_high(character); case _CC_ENUM_XDIGIT: return is_XDIGIT_high(character); case _CC_ENUM_VERTSPACE: return is_VERTWS_high(character); - default: return 0; /* Things like CNTRL are always - below 256 */ + default: break; } - assert(0); /* NOTREACHED */ - return FALSE; + return FALSE; /* Things like CNTRL are always below 256 */ } /* @@ -705,6 +703,7 @@ Perl_re_intuit_start(pTHX_ goto fail; } + RX_MATCH_UTF8_set(rx,utf8_target); reginfo->is_utf8_target = cBOOL(utf8_target); reginfo->info_aux = NULL; reginfo->strbeg = strbeg; @@ -774,7 +773,7 @@ Perl_re_intuit_start(pTHX_ * be too fiddly (e.g. REXEC_IGNOREPOS). */ if ( strpos != strbeg - && (prog->intflags & (PREGf_ANCH_BOL|PREGf_ANCH_SBOL))) + && (prog->intflags & PREGf_ANCH_SBOL)) { DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " Not at start...\n")); @@ -898,7 +897,7 @@ Perl_re_intuit_start(pTHX_ /* If the regex is absolutely anchored to either the start of the - * string (BOL,SBOL) or to pos() (ANCH_GPOS), then + * string (SBOL) or to pos() (ANCH_GPOS), then * check_offset_max represents an upper bound on the string where * the substr could start. For the ANCH_GPOS case, we assume that * the caller of intuit will have already set strpos to @@ -1502,6 +1501,10 @@ STMT_START { } \ } STMT_END +#define DUMP_EXEC_POS(li,s,doutf8) \ + dump_exec_pos(li,s,(reginfo->strend),(reginfo->strbeg), \ + startpos, doutf8) + #define REXEC_FBC_EXACTISH_SCAN(COND) \ STMT_START { \ while (s <= e) { \ @@ -1529,34 +1532,30 @@ STMT_START { \ } \ } STMT_END -#define REXEC_FBC_UTF8_CLASS_SCAN(COND) \ -REXEC_FBC_UTF8_SCAN( \ - if (COND) { \ - if (tmp && (reginfo->intuit || regtry(reginfo, &s))) \ - goto got_it; \ - else \ - tmp = doevery; \ - } \ - else \ - tmp = 1; \ +#define REXEC_FBC_UTF8_CLASS_SCAN(COND) \ +REXEC_FBC_UTF8_SCAN( /* Loops while (s < strend) */ \ + if (COND) { \ + if (tmp && (reginfo->intuit || regtry(reginfo, &s))) \ + goto got_it; \ + else \ + tmp = doevery; \ + } \ + else \ + tmp = 1; \ ) -#define REXEC_FBC_CLASS_SCAN(COND) \ -REXEC_FBC_SCAN( \ - if (COND) { \ - if (tmp && (reginfo->intuit || regtry(reginfo, &s))) \ - goto got_it; \ - else \ - tmp = doevery; \ - } \ - else \ - tmp = 1; \ +#define REXEC_FBC_CLASS_SCAN(COND) \ +REXEC_FBC_SCAN( /* Loops while (s < strend) */ \ + if (COND) { \ + if (tmp && (reginfo->intuit || regtry(reginfo, &s))) \ + goto got_it; \ + else \ + tmp = doevery; \ + } \ + else \ + tmp = 1; \ ) -#define REXEC_FBC_TRYIT \ -if ((reginfo->intuit || regtry(reginfo, &s))) \ - goto got_it - #define REXEC_FBC_CSCAN(CONDUTF8,COND) \ if (utf8_target) { \ REXEC_FBC_UTF8_CLASS_SCAN(CONDUTF8); \ @@ -1564,97 +1563,152 @@ if ((reginfo->intuit || regtry(reginfo, &s))) \ else { \ REXEC_FBC_CLASS_SCAN(COND); \ } - -#define DUMP_EXEC_POS(li,s,doutf8) \ - dump_exec_pos(li,s,(reginfo->strend),(reginfo->strbeg), \ - startpos, doutf8) - +/* The three macros below are slightly different versions of the same logic. + * + * The first is for /a and /aa when the target string is UTF-8. This can only + * match ascii, but it must advance based on UTF-8. The other two handle the + * non-UTF-8 and the more generic UTF-8 cases. In all three, we are looking + * for the boundary (or non-boundary) between a word and non-word character. + * The utf8 and non-utf8 cases have the same logic, but the details must be + * different. Find the "wordness" of the character just prior to this one, and + * compare it with the wordness of this one. If they differ, we have a + * boundary. At the beginning of the string, pretend that the previous + * character was a new-line. + * + * All these macros uncleanly have side-effects with each other and outside + * variables. So far it's been too much trouble to clean-up + * + * TEST_NON_UTF8 is the macro or function to call to test if its byte input is + * a word character or not. + * IF_SUCCESS is code to do if it finds that we are at a boundary between + * word/non-word + * IF_FAIL is code to do if we aren't at a boundary between word/non-word + * + * Exactly one of the two IF_FOO parameters is a no-op, depending on whether we + * are looking for a boundary or for a non-boundary. If we are looking for a + * boundary, we want IF_FAIL to be the no-op, and for IF_SUCCESS to go out and + * see if this tentative match actually works, and if so, to quit the loop + * here. And vice-versa if we are looking for a non-boundary. + * + * 'tmp' below in the next three macros in the REXEC_FBC_SCAN and + * REXEC_FBC_UTF8_SCAN loops is a loop invariant, a bool giving the return of + * TEST_NON_UTF8(s-1). To see this, note that that's what it is defined to be + * at entry to the loop, and to get to the IF_FAIL branch, tmp must equal + * TEST_NON_UTF8(s), and in the opposite branch, IF_SUCCESS, tmp is that + * complement. But in that branch we complement tmp, meaning that at the + * bottom of the loop tmp is always going to be equal to TEST_NON_UTF8(s), + * which means at the top of the loop in the next iteration, it is + * TEST_NON_UTF8(s-1) */ #define FBC_UTF8_A(TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ - tmp = (s != reginfo->strbeg) ? UCHARAT(s - 1) : '\n'; \ - tmp = TEST_NON_UTF8(tmp); \ - REXEC_FBC_UTF8_SCAN( \ - if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ - tmp = !tmp; \ - IF_SUCCESS; \ - } \ - else { \ - IF_FAIL; \ - } \ - ); \ - -#define FBC_UTF8(TEST1_UTF8, TEST2_UTF8, IF_SUCCESS, IF_FAIL) \ - if (s == reginfo->strbeg) { \ - tmp = '\n'; \ - } \ - else { \ - U8 * const r = reghop3((U8*)s, -1, (U8*)reginfo->strbeg); \ - tmp = utf8n_to_uvchr(r, (U8*) reginfo->strend - r, \ + tmp = (s != reginfo->strbeg) ? UCHARAT(s - 1) : '\n'; \ + tmp = TEST_NON_UTF8(tmp); \ + REXEC_FBC_UTF8_SCAN( /* advances s while s < strend */ \ + if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ + tmp = !tmp; \ + IF_SUCCESS; /* Is a boundary if values for s-1 and s differ */ \ + } \ + else { \ + IF_FAIL; \ + } \ + ); \ + +/* Like FBC_UTF8_A, but TEST_UV is a macro which takes a UV as its input, and + * TEST_UTF8 is a macro that for the same input code points returns identically + * to TEST_UV, but takes a pointer to a UTF-8 encoded string instead */ +#define FBC_UTF8(TEST_UV, TEST_UTF8, IF_SUCCESS, IF_FAIL) \ + if (s == reginfo->strbeg) { \ + tmp = '\n'; \ + } \ + else { /* Back-up to the start of the previous character */ \ + U8 * const r = reghop3((U8*)s, -1, (U8*)reginfo->strbeg); \ + tmp = utf8n_to_uvchr(r, (U8*) reginfo->strend - r, \ 0, UTF8_ALLOW_DEFAULT); \ - } \ - tmp = TEST1_UTF8; \ - LOAD_UTF8_CHARCLASS_ALNUM(); \ - REXEC_FBC_UTF8_SCAN( \ - if (tmp == ! (TEST2_UTF8)) { \ - tmp = !tmp; \ - IF_SUCCESS; \ - } \ - else { \ - IF_FAIL; \ - } \ - ); \ + } \ + tmp = TEST_UV(tmp); \ + LOAD_UTF8_CHARCLASS_ALNUM(); \ + REXEC_FBC_UTF8_SCAN( /* advances s while s < strend */ \ + if (tmp == ! (TEST_UTF8((U8 *) s))) { \ + tmp = !tmp; \ + IF_SUCCESS; \ + } \ + else { \ + IF_FAIL; \ + } \ + ); -/* The only difference between the BOUND and NBOUND cases is that - * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in - * NBOUND. This is accomplished by passing it in either the if or else clause, - * with the other one being empty */ -#define FBC_BOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ - FBC_BOUND_COMMON(FBC_UTF8(TEST1_UTF8, TEST2_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) - -#define FBC_BOUND_A(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ - FBC_BOUND_COMMON(FBC_UTF8_A(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) - -#define FBC_NBOUND(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ - FBC_BOUND_COMMON(FBC_UTF8(TEST1_UTF8, TEST2_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) - -#define FBC_NBOUND_A(TEST_NON_UTF8, TEST1_UTF8, TEST2_UTF8) \ - FBC_BOUND_COMMON(FBC_UTF8_A(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) - - -/* Common to the BOUND and NBOUND cases. Unfortunately the UTF8 tests need to - * be passed in completely with the variable name being tested, which isn't - * such a clean interface, but this is easier to read than it was before. We - * are looking for the boundary (or non-boundary between a word and non-word - * character. The utf8 and non-utf8 cases have the same logic, but the details - * must be different. Find the "wordness" of the character just prior to this - * one, and compare it with the wordness of this one. If they differ, we have - * a boundary. At the beginning of the string, pretend that the previous - * character was a new-line */ +/* Like the above two macros. UTF8_CODE is the complete code for handling + * UTF-8. Common to the BOUND and NBOUND cases, set-up by the FBC_BOUND, etc + * macros below */ #define FBC_BOUND_COMMON(UTF8_CODE, TEST_NON_UTF8, IF_SUCCESS, IF_FAIL) \ if (utf8_target) { \ - UTF8_CODE \ + UTF8_CODE \ } \ else { /* Not utf8 */ \ tmp = (s != reginfo->strbeg) ? UCHARAT(s - 1) : '\n'; \ tmp = TEST_NON_UTF8(tmp); \ - REXEC_FBC_SCAN( \ + REXEC_FBC_SCAN( /* advances s while s < strend */ \ if (tmp == ! TEST_NON_UTF8((U8) *s)) { \ - tmp = !tmp; \ IF_SUCCESS; \ + tmp = !tmp; \ } \ else { \ IF_FAIL; \ } \ ); \ } \ - if ((!prog->minlen && tmp) && (reginfo->intuit || regtry(reginfo, &s))) \ - goto got_it; + /* Here, things have been set up by the previous code so that tmp is the \ + * return of TEST_NON_UTF(s-1) or TEST_UTF8(s-1) (depending on the \ + * utf8ness of the target). We also have to check if this matches against \ + * the EOS, which we treat as a \n (which is the same value in both UTF-8 \ + * or non-UTF8, so can use the non-utf8 test condition even for a UTF-8 \ + * string */ \ + if (tmp == ! TEST_NON_UTF8('\n')) { \ + IF_SUCCESS; \ + } \ + else { \ + IF_FAIL; \ + } + +/* This is the macro to use when we want to see if something that looks like it + * could match, actually does, and if so exits the loop */ +#define REXEC_FBC_TRYIT \ + if ((reginfo->intuit || regtry(reginfo, &s))) \ + goto got_it + +/* The only difference between the BOUND and NBOUND cases is that + * REXEC_FBC_TRYIT is called when matched in BOUND, and when non-matched in + * NBOUND. This is accomplished by passing it as either the if or else clause, + * with the other one being empty (PLACEHOLDER is defined as empty). + * + * The TEST_FOO parameters are for operating on different forms of input, but + * all should be ones that return identically for the same underlying code + * points */ +#define FBC_BOUND(TEST_NON_UTF8, TEST_UV, TEST_UTF8) \ + FBC_BOUND_COMMON( \ + FBC_UTF8(TEST_UV, TEST_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), \ + TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) + +#define FBC_BOUND_A(TEST_NON_UTF8, TEST_UV, TEST_UTF8) \ + FBC_BOUND_COMMON( \ + FBC_UTF8_A(TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER), \ + TEST_NON_UTF8, REXEC_FBC_TRYIT, PLACEHOLDER) + +#define FBC_NBOUND(TEST_NON_UTF8, TEST_UV, TEST_UTF8) \ + FBC_BOUND_COMMON( \ + FBC_UTF8(TEST_UV, TEST_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), \ + TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) + +#define FBC_NBOUND_A(TEST_NON_UTF8, TEST_UV, TEST_UTF8) \ + FBC_BOUND_COMMON( \ + FBC_UTF8_A(TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT), \ + TEST_NON_UTF8, PLACEHOLDER, REXEC_FBC_TRYIT) + /* We know what class REx starts with. Try to find this position... */ /* if reginfo->intuit, its a dryrun */ /* annoyingly all the vars in this routine have different names from their counterparts in regmatch. /grrr */ - STATIC char * S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, const char *strend, regmatch_info *reginfo) @@ -1843,45 +1897,30 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, } break; } + case BOUNDL: - FBC_BOUND(isWORDCHAR_LC, - isWORDCHAR_LC_uvchr(tmp), - isWORDCHAR_LC_utf8((U8*)s)); + FBC_BOUND(isWORDCHAR_LC, isWORDCHAR_LC_uvchr, isWORDCHAR_LC_utf8); break; case NBOUNDL: - FBC_NBOUND(isWORDCHAR_LC, - isWORDCHAR_LC_uvchr(tmp), - isWORDCHAR_LC_utf8((U8*)s)); + FBC_NBOUND(isWORDCHAR_LC, isWORDCHAR_LC_uvchr, isWORDCHAR_LC_utf8); break; case BOUND: - FBC_BOUND(isWORDCHAR, - isWORDCHAR_uni(tmp), - cBOOL(swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)s, utf8_target))); + FBC_BOUND(isWORDCHAR, isWORDCHAR_uni, isWORDCHAR_utf8); break; case BOUNDA: - FBC_BOUND_A(isWORDCHAR_A, - isWORDCHAR_A(tmp), - isWORDCHAR_A((U8*)s)); + FBC_BOUND_A(isWORDCHAR_A, isWORDCHAR_A, isWORDCHAR_A); break; case NBOUND: - FBC_NBOUND(isWORDCHAR, - isWORDCHAR_uni(tmp), - cBOOL(swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)s, utf8_target))); + FBC_NBOUND(isWORDCHAR, isWORDCHAR_uni, isWORDCHAR_utf8); break; case NBOUNDA: - FBC_NBOUND_A(isWORDCHAR_A, - isWORDCHAR_A(tmp), - isWORDCHAR_A((U8*)s)); + FBC_NBOUND_A(isWORDCHAR_A, isWORDCHAR_A, isWORDCHAR_A); break; case BOUNDU: - FBC_BOUND(isWORDCHAR_L1, - isWORDCHAR_uni(tmp), - cBOOL(swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)s, utf8_target))); + FBC_BOUND(isWORDCHAR_L1, isWORDCHAR_uni, isWORDCHAR_utf8); break; case NBOUNDU: - FBC_NBOUND(isWORDCHAR_L1, - isWORDCHAR_uni(tmp), - cBOOL(swash_fetch(PL_utf8_swash_ptrs[_CC_WORDCHAR], (U8*)s, utf8_target))); + FBC_NBOUND(isWORDCHAR_L1, isWORDCHAR_uni, isWORDCHAR_utf8); break; case LNBREAK: REXEC_FBC_CSCAN(is_LNBREAK_utf8_safe(s, strend), @@ -2579,6 +2618,7 @@ Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, char *strend, } RX_MATCH_TAINTED_off(rx); + RX_MATCH_UTF8_set(rx, utf8_target); reginfo->prog = rx; /* Yes, sorry that this is confusing. */ reginfo->intuit = 0; @@ -2599,7 +2639,6 @@ Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, char *strend, magic belonging to this SV. Not newSVsv, either, as it does not COW. */ - assert(!IS_PADGV(sv)); reginfo->sv = newSV(0); SvSetSV_nosteal(reginfo->sv, sv); SAVEFREESV(reginfo->sv); @@ -2677,7 +2716,7 @@ Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, char *strend, } /* Simplest case: anchored match need be tried only once. */ - /* [unless only anchor is BOL and multiline is set] */ + /* [unless only anchor is MBOL - implying multiline is set] */ if (prog->intflags & (PREGf_ANCH & ~PREGf_ANCH_GPOS)) { if (s == startpos && regtry(reginfo, &s)) goto got_it; @@ -2931,7 +2970,7 @@ Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, char *strend, } DEBUG_EXECUTE_r({ SV * const prop = sv_newmortal(); - regprop(prog, prop, c, reginfo); + regprop(prog, prop, c, reginfo, NULL); { RE_PV_QUOTED_DECL(quoted,utf8_target,PERL_DEBUG_PAD_ZERO(1), s,strend-s,60); @@ -3037,7 +3076,8 @@ Perl_regexec_flags(pTHX_ REGEXP * const rx, char *stringarg, char *strend, * and replaced it with this one. Yves */ DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, - "String does not contain required substring, cannot match.\n" + "%sString does not contain required substring, cannot match.%s\n", + PL_colors[4], PL_colors[5] )); goto phooey; } @@ -3099,8 +3139,6 @@ got_it: if (RXp_PAREN_NAMES(prog)) (void)hv_iterinit(RXp_PAREN_NAMES(prog)); - RX_MATCH_UTF8_set(rx, utf8_target); - /* make sure $`, $&, $', and $digit will work later */ if ( !(flags & REXEC_NOT_FIRST) ) S_reg_set_capture_string(aTHX_ rx, @@ -3596,8 +3634,8 @@ S_setup_EXACTISH_ST_c1_c2(pTHX_ const regnode * const text_node, int *c1p, const bool utf8_target = reginfo->is_utf8_target; - UV c1 = CHRTEST_NOT_A_CP_1; - UV c2 = CHRTEST_NOT_A_CP_2; + UV c1 = (UV)CHRTEST_NOT_A_CP_1; + UV c2 = (UV)CHRTEST_NOT_A_CP_2; bool use_chrtest_void = FALSE; const bool is_utf8_pat = reginfo->is_utf8_pat; @@ -3953,7 +3991,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) SV * const prop = sv_newmortal(); regnode *rnext=regnext(scan); DUMP_EXEC_POS( locinput, scan, utf8_target ); - regprop(rex, prop, scan, reginfo); + regprop(rex, prop, scan, reginfo, NULL); PerlIO_printf(Perl_debug_log, "%3"IVdf":%*s%s(%"IVdf")\n", @@ -3975,8 +4013,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) assert(nextchr < 256 && (nextchr >= 0 || nextchr == NEXTCHR_EOS)); switch (state_num) { - case BOL: /* /^../ */ - case SBOL: /* /^../s */ + case SBOL: /* /^../ and /\A../ */ if (locinput == reginfo->strbeg) break; sayNO; @@ -3999,21 +4036,22 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) st->u.keeper.val = rex->offs[0].start; rex->offs[0].start = locinput - reginfo->strbeg; PUSH_STATE_GOTO(KEEPS_next, next, locinput); - assert(0); /*NOTREACHED*/ + /* NOTREACHED */ + assert(0); + case KEEPS_next_fail: /* rollback the start point change */ rex->offs[0].start = st->u.keeper.val; sayNO_SILENT; - assert(0); /*NOTREACHED*/ + /* NOTREACHED */ + assert(0); case MEOL: /* /..$/m */ if (!NEXTCHR_IS_EOS && nextchr != '\n') sayNO; break; - case EOL: /* /..$/ */ - /* FALLTHROUGH */ - case SEOL: /* /..$/s */ + case SEOL: /* /..$/ */ if (!NEXTCHR_IS_EOS && nextchr != '\n') sayNO; if (reginfo->strend - locinput > 1) @@ -4055,7 +4093,8 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) REPORT_CODE_OFF+depth*2, "", PL_colors[4], PL_colors[5]) ); sayNO_SILENT; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); } /* FALLTHROUGH */ case TRIE: /* (ab|cd) */ @@ -4242,7 +4281,8 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) ); goto trie_first_try; /* jump into the fail handler */ }} - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case TRIE_next_fail: /* we failed - try next alternative */ { @@ -4356,7 +4396,8 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) if (ST.accepted > 1 || has_cutgroup) { PUSH_STATE_GOTO(TRIE_next, scan, (char*)uc); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); } /* only one choice left - just continue */ DEBUG_EXECUTE_r({ @@ -4380,7 +4421,8 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) locinput = (char*)uc; continue; /* execute rest of RE */ - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); } #undef ST @@ -5147,9 +5189,6 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) case TAIL: /* placeholder while compiling (A|B|C) */ break; - case BACK: /* ??? doesn't appear to be used ??? */ - break; - #undef ST #define ST st->u.eval { @@ -5189,7 +5228,8 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) /* and then jump to the code we share with EVAL */ goto eval_recurse_doit; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */ if (cur_eval && cur_eval->locinput==locinput) { @@ -5278,7 +5318,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) assert(o->op_targ == OP_LEAVE); o = cUNOPo->op_first; assert(o->op_type == OP_ENTER); - o = o->op_sibling; + o = OP_SIBLING(o); } if (o->op_type != OP_STUB) { @@ -5405,7 +5445,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) re_sv = rex->engine->op_comp(aTHX_ &ret, 1, NULL, rex->engine, NULL, NULL, /* copy /msix etc to inner pattern */ - scan->flags, + ARG2L(scan), pm_flags); if (!(SvFLAGS(ret) @@ -5470,7 +5510,8 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) cur_eval = st; /* now continue from first node in postoned RE */ PUSH_YES_STATE_GOTO(EVAL_AB, startpoint, locinput); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); } case EVAL_AB: /* cleanup after a successful (??{A})B */ @@ -5579,7 +5620,7 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) } } goto fake_end; - /*NOTREACHED*/ + /* NOTREACHED */ case GROUPP: /* (?(1)) */ n = ARG(scan); /* which paren pair */ @@ -5729,19 +5770,22 @@ NULL ST.lastloc = NULL; /* this will be updated by WHILEM */ PUSH_YES_STATE_GOTO(CURLYX_end, PREVOPER(next), locinput); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); } case CURLYX_end: /* just finished matching all of A*B */ cur_curlyx = ST.prev_curlyx; sayYES; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case CURLYX_end_fail: /* just failed to match all of A*B */ regcpblow(ST.cp); cur_curlyx = ST.prev_curlyx; sayNO; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); #undef ST @@ -5751,11 +5795,14 @@ NULL { /* see the discussion above about CURLYX/WHILEM */ I32 n; - int min = ARG1(cur_curlyx->u.curlyx.me); - int max = ARG2(cur_curlyx->u.curlyx.me); - regnode *A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS; + int min, max; + regnode *A; assert(cur_curlyx); /* keep Coverity happy */ + + min = ARG1(cur_curlyx->u.curlyx.me); + max = ARG2(cur_curlyx->u.curlyx.me); + A = NEXTOPER(cur_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS; n = ++cur_curlyx->u.curlyx.count; /* how many A's matched */ ST.save_lastloc = cur_curlyx->u.curlyx.lastloc; ST.cache_offset = 0; @@ -5776,7 +5823,8 @@ NULL REGCP_SET(ST.lastcp); PUSH_STATE_GOTO(WHILEM_A_pre, A, locinput); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); } /* If degenerate A matches "", assume A done. */ @@ -5888,7 +5936,8 @@ NULL REGCP_SET(ST.lastcp); PUSH_YES_STATE_GOTO(WHILEM_B_min, ST.save_curlyx->u.curlyx.B, locinput); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); } /* Prefer A over B for maximal matching. */ @@ -5899,24 +5948,28 @@ NULL cur_curlyx->u.curlyx.lastloc = locinput; REGCP_SET(ST.lastcp); PUSH_STATE_GOTO(WHILEM_A_max, A, locinput); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); } goto do_whilem_B_max; } - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case WHILEM_B_min: /* just matched B in a minimal match */ case WHILEM_B_max: /* just matched B in a maximal match */ cur_curlyx = ST.save_curlyx; sayYES; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case WHILEM_B_max_fail: /* just failed to match B in a maximal match */ cur_curlyx = ST.save_curlyx; cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; cur_curlyx->u.curlyx.count--; CACHEsayNO; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case WHILEM_A_min_fail: /* just failed to match A in a minimal match */ /* FALLTHROUGH */ @@ -5926,7 +5979,8 @@ NULL cur_curlyx->u.curlyx.lastloc = ST.save_lastloc; cur_curlyx->u.curlyx.count--; CACHEsayNO; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case WHILEM_A_max_fail: /* just failed to match A in a maximal match */ REGCP_UNWIND(ST.lastcp); @@ -5952,7 +6006,8 @@ NULL cur_curlyx = cur_curlyx->u.curlyx.prev_curlyx; PUSH_YES_STATE_GOTO(WHILEM_B_max, ST.save_curlyx->u.curlyx.B, locinput); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case WHILEM_B_min_fail: /* just failed to match B in a minimal match */ cur_curlyx = ST.save_curlyx; @@ -5986,7 +6041,8 @@ NULL PUSH_STATE_GOTO(WHILEM_A_min, /*A*/ NEXTOPER(ST.save_curlyx->u.curlyx.me) + EXTRA_STEP_2ARGS, locinput); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); #undef ST #define ST st->u.branch @@ -6011,13 +6067,15 @@ NULL } else { PUSH_STATE_GOTO(BRANCH_next, scan, locinput); } - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case CUTGROUP: /* /(*THEN)/ */ sv_yes_mark = st->u.mark.mark_name = scan->flags ? NULL : MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); PUSH_STATE_GOTO(CUTGROUP_next, next, locinput); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case CUTGROUP_next_fail: do_cutgroup = 1; @@ -6025,11 +6083,13 @@ NULL if (st->u.mark.mark_name) sv_commit = st->u.mark.mark_name; sayNO; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case BRANCH_next: sayYES; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case BRANCH_next_fail: /* that branch failed; try the next, if any */ if (do_cutgroup) { @@ -6051,7 +6111,8 @@ NULL sayNO_SILENT; } continue; /* execute next BRANCH[J] op */ - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case MINMOD: /* next op will be non-greedy, e.g. A*? */ minmod = 1; @@ -6095,7 +6156,8 @@ NULL curlym_do_A: /* execute the A in /A{m,n}B/ */ PUSH_YES_STATE_GOTO(CURLYM_A, ST.A, locinput); /* match A */ - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case CURLYM_A: /* we've just matched an A */ ST.count++; @@ -6145,6 +6207,7 @@ NULL /* calculate c1 and c2 for possible match of 1st char * following curly */ ST.c1 = ST.c2 = CHRTEST_VOID; + assert(ST.B); if (HAS_TEXT(ST.B) || JUMPABLE(ST.B)) { regnode *text_node = ST.B; if (! HAS_TEXT(text_node)) @@ -6230,7 +6293,8 @@ NULL } PUSH_STATE_GOTO(CURLYM_B, ST.B, locinput); /* match B */ - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case CURLYM_B_fail: /* just failed to match a B */ REGCP_UNWIND(ST.cp); @@ -6408,8 +6472,8 @@ NULL REGCP_SET(ST.cp); goto curly_try_B_max; } - assert(0); /* NOTREACHED */ - + /* NOTREACHED */ + assert(0); case CURLY_B_min_known_fail: /* failed to find B in a non-greedy match where c1,c2 valid */ @@ -6484,8 +6548,8 @@ NULL } PUSH_STATE_GOTO(CURLY_B_min_known, ST.B, locinput); } - assert(0); /* NOTREACHED */ - + /* NOTREACHED */ + assert(0); case CURLY_B_min_fail: /* failed to find B in a non-greedy match where c1,c2 invalid */ @@ -6517,8 +6581,8 @@ NULL } } sayNO; - assert(0); /* NOTREACHED */ - + /* NOTREACHED */ + assert(0); curly_try_B_max: /* a successful greedy match: now try to match B */ @@ -6548,7 +6612,8 @@ NULL if (ST.c1 == CHRTEST_VOID || could_match) { CURLY_SETPAREN(ST.paren, ST.count); PUSH_STATE_GOTO(CURLY_B_max, ST.B, locinput); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); } } /* FALLTHROUGH */ @@ -6667,7 +6732,8 @@ NULL /* execute body of (?...A) */ PUSH_YES_STATE_GOTO(IFMATCH_A, NEXTOPER(NEXTOPER(scan)), newstart); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); } case IFMATCH_A_fail: /* body of (?...A) failed */ @@ -6707,7 +6773,8 @@ NULL if (!scan->flags) sv_yes_mark = sv_commit = MUTABLE_SV(rexi->data->data[ ARG( scan ) ]); PUSH_STATE_GOTO(COMMIT_next, next, locinput); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case COMMIT_next_fail: no_final = 1; @@ -6715,7 +6782,8 @@ NULL case OPFAIL: /* (*FAIL) */ sayNO; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); #define ST st->u.mark case MARKPOINT: /* (*MARK:foo) */ @@ -6725,12 +6793,14 @@ NULL mark_state = st; ST.mark_loc = locinput; PUSH_YES_STATE_GOTO(MARKPOINT_next, next, locinput); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case MARKPOINT_next: mark_state = ST.prev_mark; sayYES; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case MARKPOINT_next_fail: if (popmark && sv_eq(ST.mark_name,popmark)) @@ -6751,7 +6821,8 @@ NULL sv_yes_mark = mark_state ? mark_state->u.mark.mark_name : NULL; sayNO; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); case SKIP: /* (*SKIP) */ if (scan->flags) { @@ -6796,7 +6867,8 @@ NULL } no_final = 1; sayNO; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); #undef ST case LNBREAK: /* \R */ @@ -6830,7 +6902,8 @@ NULL /* switch break jumps here */ scan = next; /* prepare to execute the next op and ... */ continue; /* ... jump back to the top, reusing st */ - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); push_yes_state: /* push a state that backtracks on success */ @@ -6873,7 +6946,8 @@ NULL locinput = pushinput; st = newst; continue; - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); } } @@ -6882,7 +6956,7 @@ NULL * the terminating point. */ Perl_croak(aTHX_ "corrupted regexp pointers"); - /*NOTREACHED*/ + /* NOTREACHED */ sayNO; yes: @@ -7509,7 +7583,8 @@ S_regrepeat(pTHX_ regexp *prog, char **startposp, const regnode *p, default: Perl_croak(aTHX_ "panic: regrepeat() called with unrecognized node type %d='%s'", OP(p), PL_reg_name[OP(p)]); - assert(0); /* NOTREACHED */ + /* NOTREACHED */ + assert(0); } @@ -7523,7 +7598,7 @@ S_regrepeat(pTHX_ regexp *prog, char **startposp, const regnode *p, GET_RE_DEBUG_FLAGS_DECL; DEBUG_EXECUTE_r({ SV * const prop = sv_newmortal(); - regprop(prog, prop, p, reginfo); + regprop(prog, prop, p, reginfo, NULL); PerlIO_printf(Perl_debug_log, "%*s %s can match %"IVdf" times out of %"IVdf"...\n", REPORT_CODE_OFF + depth*2, "", SvPVX_const(prop),(IV)c,(IV)max); @@ -7549,121 +7624,9 @@ Perl_regclass_swash(pTHX_ const regexp *prog, const regnode* node, bool doinit, *altsvp = NULL; } - return newSVsv(_get_regclass_nonbitmap_data(prog, node, doinit, listsvp, NULL)); + return newSVsv(_get_regclass_nonbitmap_data(prog, node, doinit, listsvp, NULL, NULL)); } -SV * -Perl__get_regclass_nonbitmap_data(pTHX_ const regexp *prog, - const regnode* node, - bool doinit, - SV** listsvp, - SV** only_utf8_locale_ptr) -{ - /* For internal core use only. - * Returns the swash for the input 'node' in the regex 'prog'. - * If is 'true', will attempt to create the swash if not already - * done. - * If is non-null, will return the printable contents of the - * swash. This can be used to get debugging information even before the - * swash exists, by calling this function with 'doinit' set to false, in - * which case the components that will be used to eventually create the - * swash are returned (in a printable form). - * Tied intimately to how regcomp.c sets up the data structure */ - - SV *sw = NULL; - SV *si = NULL; /* Input swash initialization string */ - SV* invlist = NULL; - - RXi_GET_DECL(prog,progi); - const struct reg_data * const data = prog ? progi->data : NULL; - - PERL_ARGS_ASSERT__GET_REGCLASS_NONBITMAP_DATA; - - assert(ANYOF_FLAGS(node) - & (ANYOF_UTF8|ANYOF_NONBITMAP_NON_UTF8|ANYOF_LOC_FOLD)); - - if (data && data->count) { - const U32 n = ARG(node); - - if (data->what[n] == 's') { - SV * const rv = MUTABLE_SV(data->data[n]); - AV * const av = MUTABLE_AV(SvRV(rv)); - SV **const ary = AvARRAY(av); - U8 swash_init_flags = _CORE_SWASH_INIT_ACCEPT_INVLIST; - - si = *ary; /* ary[0] = the string to initialize the swash with */ - - /* Elements 3 and 4 are either both present or both absent. [3] is - * any inversion list generated at compile time; [4] indicates if - * that inversion list has any user-defined properties in it. */ - if (av_tindex(av) >= 2) { - if (only_utf8_locale_ptr - && ary[2] - && ary[2] != &PL_sv_undef) - { - *only_utf8_locale_ptr = ary[2]; - } - else { - assert(only_utf8_locale_ptr); - *only_utf8_locale_ptr = NULL; - } - - if (av_tindex(av) >= 3) { - invlist = ary[3]; - if (SvUV(ary[4])) { - swash_init_flags |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY; - } - } - else { - invlist = NULL; - } - } - - /* Element [1] is reserved for the set-up swash. If already there, - * return it; if not, create it and store it there */ - if (ary[1] && SvROK(ary[1])) { - sw = ary[1]; - } - else if (doinit && ((si && si != &PL_sv_undef) - || (invlist && invlist != &PL_sv_undef))) { - assert(si); - sw = _core_swash_init("utf8", /* the utf8 package */ - "", /* nameless */ - si, - 1, /* binary */ - 0, /* not from tr/// */ - invlist, - &swash_init_flags); - (void)av_store(av, 1, sw); - } - } - } - - /* If requested, return a printable version of what this swash matches */ - if (listsvp) { - SV* matches_string = newSVpvs(""); - - /* The swash should be used, if possible, to get the data, as it - * contains the resolved data. But this function can be called at - * compile-time, before everything gets resolved, in which case we - * return the currently best available information, which is the string - * that will eventually be used to do that resolving, 'si' */ - if ((! sw || (invlist = _get_swash_invlist(sw)) == NULL) - && (si && si != &PL_sv_undef)) - { - sv_catsv(matches_string, si); - } - - /* Add the inversion list to whatever we have. This may have come from - * the swash, or from an input parameter */ - if (invlist) { - sv_catsv(matches_string, _invlist_contents(invlist)); - } - *listsvp = matches_string; - } - - return sw; -} #endif /* !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) */ /* @@ -7706,22 +7669,25 @@ S_reginclass(pTHX_ regexp * const prog, const regnode * const n, const U8* const } /* If this character is potentially in the bitmap, check it */ - if (c < 256) { + if (c < NUM_ANYOF_CODE_POINTS) { if (ANYOF_BITMAP_TEST(n, c)) match = TRUE; - else if (flags & ANYOF_NON_UTF8_NON_ASCII_ALL - && ! utf8_target - && ! isASCII(c)) + else if ((flags & ANYOF_MATCHES_ALL_NON_UTF8_NON_ASCII) + && ! utf8_target + && ! isASCII(c)) { match = TRUE; } else if (flags & ANYOF_LOCALE_FLAGS) { - if (flags & ANYOF_LOC_FOLD) { - if (ANYOF_BITMAP_TEST(n, PL_fold_locale[c])) { - match = TRUE; - } + if ((flags & ANYOF_LOC_FOLD) + && c < 256 + && ANYOF_BITMAP_TEST(n, PL_fold_locale[c])) + { + match = TRUE; } - if (! match && ANYOF_POSIXL_TEST_ANY_SET(n)) { + else if (ANYOF_POSIXL_TEST_ANY_SET(n) + && c < 256 + ) { /* The data structure is arranged so bits 0, 2, 4, ... are set * if the class includes the Posix character class given by @@ -7774,18 +7740,20 @@ S_reginclass(pTHX_ regexp * const prog, const regnode * const n, const U8* const /* If the bitmap didn't (or couldn't) match, and something outside the * bitmap could match, try that. */ if (!match) { - if (c >= 256 && (flags & ANYOF_ABOVE_LATIN1_ALL)) { - match = TRUE; /* Everything above 255 matches */ + if (c >= NUM_ANYOF_CODE_POINTS + && (flags & ANYOF_MATCHES_ALL_ABOVE_BITMAP)) + { + match = TRUE; /* Everything above the bitmap matches */ } - else if ((flags & ANYOF_NONBITMAP_NON_UTF8) - || (utf8_target && (flags & ANYOF_UTF8)) + else if ((flags & ANYOF_HAS_NONBITMAP_NON_UTF8_MATCHES) + || (utf8_target && (flags & ANYOF_HAS_UTF8_NONBITMAP_MATCHES)) || ((flags & ANYOF_LOC_FOLD) && IN_UTF8_CTYPE_LOCALE - && ARG(n) != ANYOF_NONBITMAP_EMPTY)) + && ARG(n) != ANYOF_ONLY_HAS_BITMAP)) { SV* only_utf8_locale = NULL; SV * const sw = _get_regclass_nonbitmap_data(prog, n, TRUE, 0, - &only_utf8_locale); + &only_utf8_locale, NULL); if (sw) { U8 utf8_buffer[2]; U8 * utf8_p;