X-Git-Url: https://perl5.git.perl.org/perl5.git/blobdiff_plain/821b33a5de9c8f8170754d88edfcb4cdcdededf7..25685dc7c95e47b94f62c769e3db898f82b0eb99:/regcomp.c diff --git a/regcomp.c b/regcomp.c index ca4da9c..0dafdd0 100644 --- a/regcomp.c +++ b/regcomp.c @@ -19,6 +19,32 @@ * with the POSIX routines of the same names. */ +#ifdef PERL_EXT_RE_BUILD +/* need to replace pregcomp et al, so enable that */ +# ifndef PERL_IN_XSUB_RE +# define PERL_IN_XSUB_RE +# endif +/* need access to debugger hooks */ +# if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING) +# define DEBUGGING +# endif +#endif + +#ifdef PERL_IN_XSUB_RE +/* We *really* need to overwrite these symbols: */ +# define Perl_pregcomp my_regcomp +# define Perl_regdump my_regdump +# define Perl_regprop my_regprop +# define Perl_pregfree my_regfree +# define Perl_re_intuit_string my_re_intuit_string +/* *These* symbols are masked to allow static link. */ +# define Perl_regnext my_regnext +# define Perl_save_re_context my_save_re_context +# define Perl_reginitcolors my_reginitcolors + +# define PERL_NO_GET_CONTEXT +#endif + /*SUPPRESS 112*/ /* * pregcomp and pregexec -- regsub and regerror are not used in perl @@ -43,7 +69,7 @@ * **** Alterations to Henry's code are... **** - **** Copyright (c) 1991-1997, Larry Wall + **** Copyright (c) 1991-1999, 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. @@ -54,8 +80,16 @@ * regular-expression syntax might require a total rethink. */ #include "EXTERN.h" +#define PERL_IN_REGCOMP_C #include "perl.h" -#include "INTERN.h" + +#ifdef PERL_IN_XSUB_RE +# if defined(PERL_CAPI) || defined(PERL_OBJECT) +# include "XSUB.h" +# endif +#else +# include "INTERN.h" +#endif #define REG_COMP_C #include "regcomp.h" @@ -98,32 +132,9 @@ #define SPSTART 0x4 /* Starts with * or +. */ #define TRYAGAIN 0x8 /* Weeded out a declaration. */ -/* - * Forward declarations for pregcomp()'s friends. - */ - -#ifndef PERL_OBJECT -static regnode *reg _((I32, I32 *)); -static regnode *reganode _((U8, U32)); -static regnode *regatom _((I32 *)); -static regnode *regbranch _((I32 *, I32)); -static void regc _((U8, char *)); -static regnode *regclass _((void)); -STATIC I32 regcurly _((char *)); -static regnode *reg_node _((U8)); -static regnode *regpiece _((I32 *)); -static void reginsert _((U8, regnode *)); -static void regoptail _((regnode *, regnode *)); -static void regtail _((regnode *, regnode *)); -static char* regwhite _((char *, char *)); -static char* nextchar _((void)); -static void re_croak2 _((const char* pat1,const char* pat2,...)) __attribute__((noreturn)); -#endif - /* Length of a variant. */ -#ifndef PERL_OBJECT -typedef struct { +typedef struct scan_data_t { I32 len_min; I32 len_delta; I32 pos_min; @@ -139,10 +150,15 @@ typedef struct { I32 offset_float_min; I32 offset_float_max; I32 flags; + I32 whilem_c; } scan_data_t; -#endif -static scan_data_t zero_scan_data = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; +/* + * Forward declarations for pregcomp()'s friends. + */ + +static scan_data_t zero_scan_data = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0 }; #define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL) #define SF_BEFORE_SEOL 0x1 @@ -169,11 +185,29 @@ static scan_data_t zero_scan_data = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }; #define SF_HAS_EVAL 0x200 #define SCF_DO_SUBSTR 0x400 +#define RF_utf8 8 +#define UTF (PL_reg_flags & RF_utf8) +#define LOC (PL_regflags & PMf_LOCALE) +#define FOLD (PL_regflags & PMf_FOLD) + +#define OOB_CHAR8 1234 +#define OOB_UTF8 123456 +#define OOB_NAMEDCLASS -1 + +#define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv)) +#define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b) + +/* Allow for side effects in s */ +#define REGC(c,s) STMT_START { if (!SIZE_ONLY) *(s) = (c); else (s);} STMT_END + +static void clear_re(pTHXo_ void *r); + STATIC void -scan_commit(scan_data_t *data) +S_scan_commit(pTHX_ scan_data_t *data) { - STRLEN l = SvCUR(data->last_found); - STRLEN old_l = SvCUR(*data->longest); + dTHR; + STRLEN l = CHR_SVLEN(data->last_found); + STRLEN old_l = CHR_SVLEN(*data->longest); if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) { sv_setsv(*data->longest, data->last_found); @@ -184,7 +218,8 @@ scan_commit(scan_data_t *data) |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL); else data->flags &= ~SF_FIX_BEFORE_EOL; - } else { + } + else { data->offset_float_min = l ? data->last_start_min : data->pos_min; data->offset_float_max = (l ? data->last_start_max @@ -205,38 +240,40 @@ scan_commit(scan_data_t *data) to the position after last scanned or to NULL. */ STATIC I32 -study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 flags) +S_study_chunk(pTHX_ regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 flags) /* scanp: Start here (read-write). */ /* deltap: Write maxlen-minlen here. */ /* last: Stop before this one. */ { + dTHR; I32 min = 0, pars = 0, code; regnode *scan = *scanp, *next; I32 delta = 0; int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF); + int is_inf_internal = 0; /* The studied chunk is infinite */ I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0; scan_data_t data_fake; while (scan && OP(scan) != END && scan < last) { /* Peephole optimizer: */ - if (regkind[(U8)OP(scan)] == EXACT) { + if (PL_regkind[(U8)OP(scan)] == EXACT) { regnode *n = regnext(scan); U32 stringok = 1; #ifdef DEBUGGING regnode *stop = scan; #endif - next = scan + (*OPERAND(scan) + 2 - 1)/sizeof(regnode) + 2; + next = scan + NODE_SZ_STR(scan); /* Skip NOTHING, merge EXACT*. */ while (n && - ( regkind[(U8)OP(n)] == NOTHING || + ( PL_regkind[(U8)OP(n)] == NOTHING || (stringok && (OP(n) == OP(scan)))) && NEXT_OFF(n) && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX) { if (OP(n) == TAIL || n > next) stringok = 0; - if (regkind[(U8)OP(n)] == NOTHING) { + if (PL_regkind[(U8)OP(n)] == NOTHING) { NEXT_OFF(scan) += NEXT_OFF(n); next = n + NODE_STEP_REGNODE; #ifdef DEBUGGING @@ -244,18 +281,19 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 stop = n; #endif n = regnext(n); - } else { - int oldl = *OPERAND(scan); + } + else { + int oldl = STR_LEN(scan); regnode *nnext = regnext(n); - if (oldl + *OPERAND(n) > U8_MAX) + if (oldl + STR_LEN(n) > U8_MAX) break; NEXT_OFF(scan) += NEXT_OFF(n); - *OPERAND(scan) += *OPERAND(n); - next = n + (*OPERAND(n) + 2 - 1)/sizeof(regnode) + 2; + STR_LEN(scan) += STR_LEN(n); + next = n + NODE_SZ_STR(n); /* Now we can overwrite *n : */ - Move(OPERAND(n) + 1, OPERAND(scan) + oldl + 1, - *OPERAND(n) + 1, char); + Move(STRING(n), STRING(scan) + oldl, + STR_LEN(n), char); #ifdef DEBUGGING if (stringok) stop = next - 1; @@ -265,13 +303,13 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 } #ifdef DEBUGGING /* Allow dumping */ - n = scan + (*OPERAND(scan) + 2 - 1)/sizeof(regnode) + 2; + n = scan + NODE_SZ_STR(scan); while (n <= stop) { /* Purify reports a benign UMR here sometimes, because we * don't initialize the OP() slot of a node when that node * is occupied by just the trailing null of the string in * an EXACT node */ - if (regkind[(U8)OP(n)] != NOTHING || OP(n) == NOTHING) { + if (PL_regkind[(U8)OP(n)] != NOTHING || OP(n) == NOTHING) { OP(n) = OPTIMIZED; NEXT_OFF(n) = 0; } @@ -281,14 +319,17 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 } if (OP(scan) != CURLYX) { - int max = (reg_off_by_arg[OP(scan)] ? I32_MAX : U16_MAX); + int max = (reg_off_by_arg[OP(scan)] + ? I32_MAX + /* I32 may be smaller than U16 on CRAYs! */ + : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX)); int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan)); int noff; regnode *n = scan; /* Skip NOTHING and LONGJMP. */ while ((n = regnext(n)) - && ((regkind[(U8)OP(n)] == NOTHING && (noff = NEXT_OFF(n))) + && ((PL_regkind[(U8)OP(n)] == NOTHING && (noff = NEXT_OFF(n))) || ((OP(n) == LONGJMP) && (noff = ARG(n)))) && off + noff < max) off += noff; @@ -312,6 +353,8 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 num++; data_fake.flags = 0; + if (data) + data_fake.whilem_c = data->whilem_c; next = regnext(scan); scan = NEXTOPER(scan); if (code != BRANCH) @@ -324,12 +367,14 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 if (max1 < minnext + deltanext) max1 = minnext + deltanext; if (deltanext == I32_MAX) - is_inf = 1; + is_inf = is_inf_internal = 1; scan = next; if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR)) pars++; if (data && (data_fake.flags & SF_HAS_EVAL)) data->flags |= SF_HAS_EVAL; + if (data) + data->whilem_c = data_fake.whilem_c; if (code == SUSPEND) break; } @@ -343,16 +388,27 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 } min += min1; delta += max1 - min1; - } else if (code == BRANCHJ) /* single branch is optimized. */ + } + else if (code == BRANCHJ) /* single branch is optimized. */ scan = NEXTOPER(NEXTOPER(scan)); else /* single branch is optimized. */ scan = NEXTOPER(scan); continue; - } else if (OP(scan) == EXACT) { - min += *OPERAND(scan); + } + else if (OP(scan) == EXACT) { + I32 l = STR_LEN(scan); + if (UTF) { + unsigned char *s = (unsigned char *)STRING(scan); + unsigned char *e = s + l; + I32 newl = 0; + while (s < e) { + newl++; + s += UTF8SKIP(s); + } + l = newl; + } + min += l; if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */ - I32 l = *OPERAND(scan); - /* The code below prefers earlier match for fixed offset, later match for variable offset. */ if (data->last_end == -1) { /* Update the start info. */ @@ -360,22 +416,35 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 data->last_start_max = is_inf ? I32_MAX : data->pos_min + data->pos_delta; } - sv_catpvn(data->last_found, (char *)(OPERAND(scan)+1), l); + sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan)); data->last_end = data->pos_min + l; data->pos_min += l; /* As in the first entry. */ data->flags &= ~SF_BEFORE_EOL; } - } else if (regkind[(U8)OP(scan)] == EXACT) { + } + else if (PL_regkind[(U8)OP(scan)] == EXACT) { + I32 l = STR_LEN(scan); if (flags & SCF_DO_SUBSTR) scan_commit(data); - min += *OPERAND(scan); + if (UTF) { + unsigned char *s = (unsigned char *)STRING(scan); + unsigned char *e = s + l; + I32 newl = 0; + while (s < e) { + newl++; + s += UTF8SKIP(s); + } + l = newl; + } + min += l; if (data && (flags & SCF_DO_SUBSTR)) - data->pos_min += *OPERAND(scan); - } else if (strchr(varies,OP(scan))) { + data->pos_min += l; + } + else if (strchr((char*)PL_varies,OP(scan))) { I32 mincount, maxcount, minnext, deltanext, pos_before, fl; regnode *oscan = scan; - switch (regkind[(U8)OP(scan)]) { + switch (PL_regkind[(U8)OP(scan)]) { case WHILEM: scan = NEXTOPER(scan); goto finish; @@ -395,7 +464,7 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 min++; /* Fall through. */ case STAR: - is_inf = 1; + is_inf = is_inf_internal = 1; scan = regnext(scan); if (flags & SCF_DO_SUBSTR) { scan_commit(data); @@ -424,13 +493,16 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 ? (flags & ~SCF_DO_SUBSTR) : flags); if (!scan) /* It was not CURLYX, but CURLY. */ scan = next; - if (dowarn && (minnext + deltanext == 0) + if (ckWARN(WARN_UNSAFE) && (minnext + deltanext == 0) && !(data->flags & (SF_HAS_PAR|SF_IN_PAR)) - && maxcount <= 10000) /* Complement check for big count */ - warn("Strange *+?{} on zero-length expression"); + && maxcount <= REG_INFTY/3) /* Complement check for big count */ + Perl_warner(aTHX_ WARN_UNSAFE, + "Strange *+?{} on zero-length expression"); min += minnext * mincount; - is_inf |= (maxcount == REG_INFTY && (minnext + deltanext) > 0 - || deltanext == I32_MAX); + is_inf_internal |= (maxcount == REG_INFTY + && (minnext + deltanext) > 0 + || deltanext == I32_MAX); + is_inf |= is_inf_internal; delta += (minnext + deltanext) * maxcount - minnext * mincount; /* Try powerful optimization CURLYX => CURLYN. */ @@ -444,9 +516,9 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 /* Skip open. */ nxt = regnext(nxt); - if (!strchr(simple,OP(nxt)) - && !(regkind[(U8)OP(nxt)] == EXACT - && *OPERAND(nxt) == 1)) + if (!strchr((char*)PL_simple,OP(nxt)) + && !(PL_regkind[(U8)OP(nxt)] == EXACT + && STR_LEN(nxt) == 1)) goto nogo; nxt2 = nxt; nxt = regnext(nxt); @@ -515,9 +587,20 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 #endif /* Optimize again: */ study_chunk(&nxt1, &deltanext, nxt, NULL, 0); - } else + } + else oscan->flags = 0; } + else if (OP(oscan) == CURLYX && data && ++data->whilem_c < 16) { + /* This stays as CURLYX, and can put the count/of pair. */ + /* Find WHILEM (as in regexec.c) */ + regnode *nxt = oscan + NEXT_OFF(oscan); + + if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */ + nxt += ARG(nxt); + PREVOPER(nxt)->flags = data->whilem_c + | (PL_reg_whilem_seen << 4); /* On WHILEM */ + } if (data && fl & (SF_HAS_PAR|SF_IN_PAR)) pars++; if (flags & SCF_DO_SUBSTR) { @@ -529,10 +612,14 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 ? pos_before : data->last_start_min; STRLEN l; char *s = SvPV(data->last_found, l); + I32 old = b - data->last_start_min; + + if (UTF) + old = utf8_hop((U8*)s, old) - (U8*)s; - l -= b - data->last_start_min; + l -= old; /* Get the added string: */ - last_str = newSVpv(s + b - data->last_start_min, l); + last_str = newSVpvn(s + old, l); if (deltanext == 0 && pos_before == b) { /* What was added is a constant string */ if (mincount > 1) { @@ -558,20 +645,21 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 sv_setsv(data->last_found, last_str); data->last_end = data->pos_min; data->last_start_min = - data->pos_min - SvCUR(last_str); + data->pos_min - CHR_SVLEN(last_str); data->last_start_max = is_inf ? I32_MAX : data->pos_min + data->pos_delta - - SvCUR(last_str); + - CHR_SVLEN(last_str); } data->longest = &(data->longest_float); } + SvREFCNT_dec(last_str); } if (data && (fl & SF_HAS_EVAL)) data->flags |= SF_HAS_EVAL; optimize_curly_tail: if (OP(oscan) != CURLYX) { - while (regkind[(U8)OP(next = regnext(oscan))] == NOTHING + while (PL_regkind[(U8)OP(next = regnext(oscan))] == NOTHING && NEXT_OFF(next)) NEXT_OFF(oscan) += NEXT_OFF(next); } @@ -581,34 +669,40 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 scan_commit(data); data->longest = &(data->longest_float); } - is_inf = 1; + is_inf = is_inf_internal = 1; break; } - } else if (strchr(simple,OP(scan))) { + } + else if (strchr((char*)PL_simple,OP(scan)) || PL_regkind[(U8)OP(scan)] == ANYUTF8) { if (flags & SCF_DO_SUBSTR) { scan_commit(data); data->pos_min++; } min++; - } else if (regkind[(U8)OP(scan)] == EOL && flags & SCF_DO_SUBSTR) { + } + else if (PL_regkind[(U8)OP(scan)] == EOL && flags & SCF_DO_SUBSTR) { data->flags |= (OP(scan) == MEOL ? SF_BEFORE_MEOL : SF_BEFORE_SEOL); - } else if (regkind[(U8)OP(scan)] == BRANCHJ + } + else if (PL_regkind[(U8)OP(scan)] == BRANCHJ && (scan->flags || data) && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) { I32 deltanext, minnext; regnode *nscan; data_fake.flags = 0; + if (data) + data_fake.whilem_c = data->whilem_c; next = regnext(scan); nscan = NEXTOPER(NEXTOPER(scan)); minnext = study_chunk(&nscan, &deltanext, last, &data_fake, 0); if (scan->flags) { if (deltanext) { FAIL("variable length lookbehind not implemented"); - } else if (minnext > U8_MAX) { - FAIL2("lookbehind longer than %d not implemented", U8_MAX); + } + else if (minnext > U8_MAX) { + FAIL2("lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX); } scan->flags = minnext; } @@ -616,24 +710,36 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 pars++; if (data && (data_fake.flags & SF_HAS_EVAL)) data->flags |= SF_HAS_EVAL; - } else if (OP(scan) == OPEN) { + if (data) + data->whilem_c = data_fake.whilem_c; + } + else if (OP(scan) == OPEN) { pars++; - } else if (OP(scan) == CLOSE && ARG(scan) == is_par) { + } + else if (OP(scan) == CLOSE && ARG(scan) == is_par) { next = regnext(scan); if ( next && (OP(next) != WHILEM) && next < last) is_par = 0; /* Disable optimization */ - } else if (OP(scan) == EVAL) { + } + else if (OP(scan) == EVAL) { if (data) data->flags |= SF_HAS_EVAL; } + else if (OP(scan) == LOGICAL && scan->flags == 2) { /* Embedded */ + if (flags & SCF_DO_SUBSTR) { + scan_commit(data); + data->longest = &(data->longest_float); + } + is_inf = is_inf_internal = 1; + } /* Else: zero-length, ignore. */ scan = regnext(scan); } finish: *scanp = scan; - *deltap = is_inf ? I32_MAX : delta; + *deltap = is_inf_internal ? I32_MAX : delta; if (flags & SCF_DO_SUBSTR && is_inf) data->pos_delta = I32_MAX - data->pos_min; if (is_par > U8_MAX) @@ -641,7 +747,8 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 if (is_par && pars==1 && data) { data->flags |= SF_IN_PAR; data->flags &= ~SF_HAS_PAR; - } else if (pars && data) { + } + else if (pars && data) { data->flags |= SF_HAS_PAR; data->flags &= ~SF_IN_PAR; } @@ -649,22 +756,49 @@ study_chunk(regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 } STATIC I32 -add_data(I32 n, char *s) +S_add_data(pTHX_ I32 n, char *s) { - if (regcomp_rx->data) { - Renewc(regcomp_rx->data, - sizeof(*regcomp_rx->data) + sizeof(void*) * (regcomp_rx->data->count + n - 1), + dTHR; + if (PL_regcomp_rx->data) { + Renewc(PL_regcomp_rx->data, + sizeof(*PL_regcomp_rx->data) + sizeof(void*) * (PL_regcomp_rx->data->count + n - 1), char, struct reg_data); - Renew(regcomp_rx->data->what, regcomp_rx->data->count + n, U8); - regcomp_rx->data->count += n; - } else { - Newc(1207, regcomp_rx->data, sizeof(*regcomp_rx->data) + sizeof(void*) * (n - 1), + Renew(PL_regcomp_rx->data->what, PL_regcomp_rx->data->count + n, U8); + PL_regcomp_rx->data->count += n; + } + else { + Newc(1207, PL_regcomp_rx->data, sizeof(*PL_regcomp_rx->data) + sizeof(void*) * (n - 1), char, struct reg_data); - New(1208, regcomp_rx->data->what, n, U8); - regcomp_rx->data->count = n; + New(1208, PL_regcomp_rx->data->what, n, U8); + PL_regcomp_rx->data->count = n; + } + Copy(s, PL_regcomp_rx->data->what + PL_regcomp_rx->data->count - n, n, U8); + return PL_regcomp_rx->data->count - n; +} + +void +Perl_reginitcolors(pTHX) +{ + dTHR; + int i = 0; + char *s = PerlEnv_getenv("PERL_RE_COLORS"); + + if (s) { + PL_colors[0] = s = savepv(s); + while (++i < 6) { + s = strchr(s, '\t'); + if (s) { + *s = '\0'; + PL_colors[i] = ++s; + } + else + PL_colors[i] = s = ""; + } + } else { + while (i < 6) + PL_colors[i++] = ""; } - Copy(s, regcomp_rx->data->what + regcomp_rx->data->count - n, n, U8); - return regcomp_rx->data->count - n; + PL_colorset = 1; } /* @@ -683,8 +817,9 @@ add_data(I32 n, char *s) * of the structure of the compiled regexp. [I'll say.] */ regexp * -pregcomp(char *exp, char *xend, PMOP *pm) +Perl_pregcomp(pTHX_ char *exp, char *xend, PMOP *pm) { + dTHR; register regexp *r; regnode *scan; SV **longest; @@ -695,113 +830,111 @@ pregcomp(char *exp, char *xend, PMOP *pm) I32 minlen = 0; I32 sawplus = 0; I32 sawopen = 0; + scan_data_t data; if (exp == NULL) FAIL("NULL regexp argument"); - regprecomp = savepvn(exp, xend - exp); - DEBUG_r(PerlIO_printf(Perl_debug_log, "compiling RE `%*s'\n", - xend - exp, regprecomp)); - regflags = pm->op_pmflags; - regsawback = 0; + if (PL_curcop == &PL_compiling ? (PL_hints & HINT_UTF8) : IN_UTF8) + PL_reg_flags |= RF_utf8; + else + PL_reg_flags = 0; + + PL_regprecomp = savepvn(exp, xend - exp); + DEBUG_r(if (!PL_colorset) reginitcolors()); + DEBUG_r(PerlIO_printf(Perl_debug_log, "%sCompiling REx%s `%s%*s%s'\n", + PL_colors[4],PL_colors[5],PL_colors[0], + xend - exp, PL_regprecomp, PL_colors[1])); + PL_regflags = pm->op_pmflags; + PL_regsawback = 0; - regseen = 0; - seen_zerolen = *exp == '^' ? -1 : 0; - extralen = 0; + PL_regseen = 0; + PL_seen_zerolen = *exp == '^' ? -1 : 0; + PL_seen_evals = 0; + PL_extralen = 0; /* First pass: determine size, legality. */ - regcomp_parse = exp; - regxend = xend; - regnaughty = 0; - regnpar = 1; - regsize = 0L; - regcode = ®dummy; - regc((U8)MAGIC, (char*)regcode); + PL_regcomp_parse = exp; + PL_regxend = xend; + PL_regnaughty = 0; + PL_regnpar = 1; + PL_regsize = 0L; + PL_regcode = &PL_regdummy; + PL_reg_whilem_seen = 0; + REGC((U8)REG_MAGIC, (char*)PL_regcode); if (reg(0, &flags) == NULL) { - Safefree(regprecomp); - regprecomp = Nullch; + Safefree(PL_regprecomp); + PL_regprecomp = Nullch; return(NULL); } - DEBUG_r(PerlIO_printf(Perl_debug_log, "size %d ", regsize)); - - DEBUG_r( - if (!colorset) { - int i = 0; - char *s = PerlEnv_getenv("TERMCAP_COLORS"); - - colorset = 1; - if (s) { - colors[0] = s = savepv(s); - while (++i < 4) { - s = strchr(s, '\t'); - if (!s) - FAIL("Not enough TABs in TERMCAP_COLORS"); - *s = '\0'; - colors[i] = ++s; - } - } else { - while (i < 4) - colors[i++] = ""; - } - /* Reset colors: */ - PerlIO_printf(Perl_debug_log, "%s%s%s%s", - colors[0],colors[1],colors[2],colors[3]); - } - ); + DEBUG_r(PerlIO_printf(Perl_debug_log, "size %d ", PL_regsize)); /* Small enough for pointer-storage convention? If extralen==0, this means that we will not need long jumps. */ - if (regsize >= 0x10000L && extralen) - regsize += extralen; + if (PL_regsize >= 0x10000L && PL_extralen) + PL_regsize += PL_extralen; else - extralen = 0; + PL_extralen = 0; + if (PL_reg_whilem_seen > 15) + PL_reg_whilem_seen = 15; /* Allocate space and initialize. */ - Newc(1001, r, sizeof(regexp) + (unsigned)regsize * sizeof(regnode), + Newc(1001, r, sizeof(regexp) + (unsigned)PL_regsize * sizeof(regnode), char, regexp); if (r == NULL) FAIL("regexp out of space"); r->refcnt = 1; r->prelen = xend - exp; - r->precomp = regprecomp; - r->subbeg = r->subbase = NULL; - regcomp_rx = r; + r->precomp = PL_regprecomp; + r->subbeg = NULL; + r->reganch = pm->op_pmflags & PMf_COMPILETIME; + r->nparens = PL_regnpar - 1; /* set early to validate backrefs */ + + r->substrs = 0; /* Useful during FAIL. */ + r->startp = 0; /* Useful during FAIL. */ + r->endp = 0; /* Useful during FAIL. */ + + PL_regcomp_rx = r; /* Second pass: emit code. */ - regcomp_parse = exp; - regxend = xend; - regnaughty = 0; - regnpar = 1; - regcode = r->program; - regc((U8)MAGIC, (char*) regcode++); + PL_regcomp_parse = exp; + PL_regxend = xend; + PL_regnaughty = 0; + PL_regnpar = 1; + PL_regcode = r->program; + /* Store the count of eval-groups for security checks: */ + PL_regcode->next_off = ((PL_seen_evals > U16_MAX) ? U16_MAX : PL_seen_evals); + REGC((U8)REG_MAGIC, (char*) PL_regcode++); r->data = 0; if (reg(0, &flags) == NULL) return(NULL); /* Dig out information for optimizations. */ - pm->op_pmflags = regflags; - r->reganch = 0; + r->reganch = pm->op_pmflags & PMf_COMPILETIME; /* Again? */ + pm->op_pmflags = PL_regflags; + if (UTF) + r->reganch |= ROPT_UTF8; r->regstclass = NULL; - r->naughty = regnaughty >= 10; /* Probably an expensive pattern. */ + if (PL_regnaughty >= 10) /* Probably an expensive pattern. */ + r->reganch |= ROPT_NAUGHTY; scan = r->program + 1; /* First BRANCH. */ /* XXXX To minimize changes to RE engine we always allocate 3-units-long substrs field. */ Newz(1004, r->substrs, 1, struct reg_substr_data); + StructCopy(&zero_scan_data, &data, scan_data_t); if (OP(scan) != BRANCH) { /* Only one top-level choice. */ - scan_data_t data; I32 fake; STRLEN longest_float_length, longest_fixed_length; - StructCopy(&zero_scan_data, &data, scan_data_t); first = scan; /* Skip introductions and multiplicators >= 1. */ while ((OP(first) == OPEN && (sawopen = 1)) || (OP(first) == BRANCH && OP(regnext(first)) != BRANCH) || (OP(first) == PLUS) || (OP(first) == MINMOD) || - (regkind[(U8)OP(first)] == CURLY && ARG1(first) > 0) ) { + (PL_regkind[(U8)OP(first)] == CURLY && ARG1(first) > 0) ) { if (OP(first) == PLUS) sawplus = 1; else @@ -812,13 +945,17 @@ pregcomp(char *exp, char *xend, PMOP *pm) /* Starting-point info. */ again: if (OP(first) == EXACT); /* Empty, get anchored substr later. */ - else if (strchr(simple+2,OP(first))) + else if (strchr((char*)PL_simple+4,OP(first))) r->regstclass = first; - else if (regkind[(U8)OP(first)] == BOUND || - regkind[(U8)OP(first)] == NBOUND) + else if (PL_regkind[(U8)OP(first)] == BOUND || + PL_regkind[(U8)OP(first)] == NBOUND) r->regstclass = first; - else if (regkind[(U8)OP(first)] == BOL) { - r->reganch |= (OP(first) == MBOL ? ROPT_ANCH_MBOL: ROPT_ANCH_BOL); + else if (PL_regkind[(U8)OP(first)] == BOL) { + r->reganch |= (OP(first) == MBOL + ? ROPT_ANCH_MBOL + : (OP(first) == SBOL + ? ROPT_ANCH_SBOL + : ROPT_ANCH_BOL)); first = NEXTOPER(first); goto again; } @@ -828,16 +965,25 @@ pregcomp(char *exp, char *xend, PMOP *pm) goto again; } else if ((OP(first) == STAR && - regkind[(U8)OP(NEXTOPER(first))] == ANY) && + PL_regkind[(U8)OP(NEXTOPER(first))] == REG_ANY) && !(r->reganch & ROPT_ANCH) ) { /* turn .* into ^.* with an implied $*=1 */ - r->reganch |= ROPT_ANCH_BOL | ROPT_IMPLICIT; + int type = OP(NEXTOPER(first)); + + if (type == REG_ANY || type == ANYUTF8) + type = ROPT_ANCH_MBOL; + else + type = ROPT_ANCH_SBOL; + + r->reganch |= type | ROPT_IMPLICIT; first = NEXTOPER(first); goto again; } - if (sawplus && (!sawopen || !regsawback)) - r->reganch |= ROPT_SKIP; /* x+ must match 1st of run */ + if (sawplus && (!sawopen || !PL_regsawback) + && !(PL_regseen & REG_SEEN_EVAL)) /* May examine pos and $& */ + /* x+ must match at the 1st pos of run of x's */ + r->reganch |= ROPT_SKIP; /* Scan is after the zeroth branch, first is atomic matcher. */ DEBUG_r(PerlIO_printf(Perl_debug_log, "first at %d\n", @@ -855,61 +1001,64 @@ pregcomp(char *exp, char *xend, PMOP *pm) */ minlen = 0; - data.longest_fixed = newSVpv("",0); - data.longest_float = newSVpv("",0); - data.last_found = newSVpv("",0); + data.longest_fixed = newSVpvn("",0); + data.longest_float = newSVpvn("",0); + data.last_found = newSVpvn("",0); data.longest = &(data.longest_fixed); first = scan; - minlen = study_chunk(&first, &fake, scan + regsize, /* Up to end */ + minlen = study_chunk(&first, &fake, scan + PL_regsize, /* Up to end */ &data, SCF_DO_SUBSTR); - if ( regnpar == 1 && data.longest == &(data.longest_fixed) + if ( PL_regnpar == 1 && data.longest == &(data.longest_fixed) && data.last_start_min == 0 && data.last_end > 0 - && !seen_zerolen - && (!(regseen & REG_SEEN_GPOS) || (r->reganch & ROPT_ANCH_GPOS))) + && !PL_seen_zerolen + && (!(PL_regseen & REG_SEEN_GPOS) || (r->reganch & ROPT_ANCH_GPOS))) r->reganch |= ROPT_CHECK_ALL; scan_commit(&data); SvREFCNT_dec(data.last_found); - longest_float_length = SvCUR(data.longest_float); + longest_float_length = CHR_SVLEN(data.longest_float); if (longest_float_length || (data.flags & SF_FL_BEFORE_EOL && (!(data.flags & SF_FL_BEFORE_MEOL) - || (regflags & PMf_MULTILINE)))) { - if (SvCUR(data.longest_fixed) - && data.offset_fixed == data.offset_float_min) - goto remove; /* Like in (a)+. */ - + || (PL_regflags & PMf_MULTILINE)))) { + int t; + + if (SvCUR(data.longest_fixed) /* ok to leave SvCUR */ + && data.offset_fixed == data.offset_float_min + && SvCUR(data.longest_fixed) == SvCUR(data.longest_float)) + goto remove_float; /* As in (a)+. */ + r->float_substr = data.longest_float; r->float_min_offset = data.offset_float_min; r->float_max_offset = data.offset_float_max; - fbm_compile(r->float_substr, 0); - BmUSEFUL(r->float_substr) = 100; - if (data.flags & SF_FL_BEFORE_EOL /* Cannot have SEOL and MULTI */ - && (!(data.flags & SF_FL_BEFORE_MEOL) - || (regflags & PMf_MULTILINE))) - SvTAIL_on(r->float_substr); - } else { - remove: + t = (data.flags & SF_FL_BEFORE_EOL /* Can't have SEOL and MULTI */ + && (!(data.flags & SF_FL_BEFORE_MEOL) + || (PL_regflags & PMf_MULTILINE))); + fbm_compile(r->float_substr, t ? FBMcf_TAIL : 0); + } + else { + remove_float: r->float_substr = Nullsv; SvREFCNT_dec(data.longest_float); longest_float_length = 0; } - longest_fixed_length = SvCUR(data.longest_fixed); + longest_fixed_length = CHR_SVLEN(data.longest_fixed); if (longest_fixed_length || (data.flags & SF_FIX_BEFORE_EOL /* Cannot have SEOL and MULTI */ && (!(data.flags & SF_FIX_BEFORE_MEOL) - || (regflags & PMf_MULTILINE)))) { + || (PL_regflags & PMf_MULTILINE)))) { + int t; + r->anchored_substr = data.longest_fixed; r->anchored_offset = data.offset_fixed; - fbm_compile(r->anchored_substr, 0); - BmUSEFUL(r->anchored_substr) = 100; - if (data.flags & SF_FIX_BEFORE_EOL /* Cannot have SEOL and MULTI */ - && (!(data.flags & SF_FIX_BEFORE_MEOL) - || (regflags & PMf_MULTILINE))) - SvTAIL_on(r->anchored_substr); - } else { + t = (data.flags & SF_FIX_BEFORE_EOL /* Can't have SEOL and MULTI */ + && (!(data.flags & SF_FIX_BEFORE_MEOL) + || (PL_regflags & PMf_MULTILINE))); + fbm_compile(r->anchored_substr, t ? FBMcf_TAIL : 0); + } + else { r->anchored_substr = Nullsv; SvREFCNT_dec(data.longest_fixed); longest_fixed_length = 0; @@ -921,31 +1070,39 @@ pregcomp(char *exp, char *xend, PMOP *pm) r->check_offset_min = r->check_offset_max = r->anchored_offset; if (r->reganch & ROPT_ANCH_SINGLE) r->reganch |= ROPT_NOSCAN; - } else { + } + else { r->check_substr = r->float_substr; r->check_offset_min = data.offset_float_min; r->check_offset_max = data.offset_float_max; } - } else { + /* XXXX Currently intuiting is not compatible with ANCH_GPOS. + This should be changed ASAP! */ + if (r->check_substr && !(r->reganch & ROPT_ANCH_GPOS)) { + r->reganch |= RE_USE_INTUIT; + if (SvTAIL(r->check_substr)) + r->reganch |= RE_INTUIT_TAIL; + } + } + else { /* Several toplevels. Best we can is to set minlen. */ I32 fake; DEBUG_r(PerlIO_printf(Perl_debug_log, "\n")); scan = r->program + 1; - minlen = study_chunk(&scan, &fake, scan + regsize, NULL, 0); + minlen = study_chunk(&scan, &fake, scan + PL_regsize, &data, 0); r->check_substr = r->anchored_substr = r->float_substr = Nullsv; } - r->nparens = regnpar - 1; r->minlen = minlen; - if (regseen & REG_SEEN_GPOS) + if (PL_regseen & REG_SEEN_GPOS) r->reganch |= ROPT_GPOS_SEEN; - if (regseen & REG_SEEN_LOOKBEHIND) + if (PL_regseen & REG_SEEN_LOOKBEHIND) r->reganch |= ROPT_LOOKBEHIND_SEEN; - if (regseen & REG_SEEN_EVAL) + if (PL_regseen & REG_SEEN_EVAL) r->reganch |= ROPT_EVAL_SEEN; - Newz(1002, r->startp, regnpar, char*); - Newz(1002, r->endp, regnpar, char*); + Newz(1002, r->startp, PL_regnpar, I32); + Newz(1002, r->endp, PL_regnpar, I32); DEBUG_r(regdump(r)); return(r); } @@ -960,36 +1117,41 @@ pregcomp(char *exp, char *xend, PMOP *pm) * follows makes it hard to avoid. */ STATIC regnode * -reg(I32 paren, I32 *flagp) +S_reg(pTHX_ I32 paren, I32 *flagp) /* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */ { + dTHR; register regnode *ret; /* Will be the head of the group. */ register regnode *br; register regnode *lastbr; register regnode *ender = 0; register I32 parno = 0; - I32 flags, oregflags = regflags, have_branch = 0, open = 0; + I32 flags, oregflags = PL_regflags, have_branch = 0, open = 0; char c; *flagp = 0; /* Tentatively. */ /* Make an OPEN node, if parenthesized. */ if (paren) { - if (*regcomp_parse == '?') { - regcomp_parse++; - paren = *regcomp_parse++; + if (*PL_regcomp_parse == '?') { + U16 posflags = 0, negflags = 0; + U16 *flagsp = &posflags; + int logical = 0; + + PL_regcomp_parse++; + paren = *PL_regcomp_parse++; ret = NULL; /* For look-ahead/behind. */ switch (paren) { case '<': - regseen |= REG_SEEN_LOOKBEHIND; - if (*regcomp_parse == '!') + PL_regseen |= REG_SEEN_LOOKBEHIND; + if (*PL_regcomp_parse == '!') paren = ','; - if (*regcomp_parse != '=' && *regcomp_parse != '!') + if (*PL_regcomp_parse != '=' && *PL_regcomp_parse != '!') goto unknown; - regcomp_parse++; + PL_regcomp_parse++; case '=': case '!': - seen_zerolen++; + PL_seen_zerolen++; case ':': case '>': break; @@ -998,75 +1160,95 @@ reg(I32 paren, I32 *flagp) FAIL2("Sequence (?%c...) not implemented", (int)paren); break; case '#': - while (*regcomp_parse && *regcomp_parse != ')') - regcomp_parse++; - if (*regcomp_parse != ')') + while (*PL_regcomp_parse && *PL_regcomp_parse != ')') + PL_regcomp_parse++; + if (*PL_regcomp_parse != ')') FAIL("Sequence (?#... not terminated"); nextchar(); *flagp = TRYAGAIN; return NULL; + case 'p': + logical = 1; + paren = *PL_regcomp_parse++; + /* FALL THROUGH */ case '{': { dTHR; I32 count = 1, n = 0; char c; - char *s = regcomp_parse; + char *s = PL_regcomp_parse; SV *sv; OP_4tree *sop, *rop; - seen_zerolen++; - regseen |= REG_SEEN_EVAL; - while (count && (c = *regcomp_parse)) { - if (c == '\\' && regcomp_parse[1]) - regcomp_parse++; + PL_seen_zerolen++; + PL_regseen |= REG_SEEN_EVAL; + while (count && (c = *PL_regcomp_parse)) { + if (c == '\\' && PL_regcomp_parse[1]) + PL_regcomp_parse++; else if (c == '{') count++; else if (c == '}') count--; - regcomp_parse++; + PL_regcomp_parse++; } - if (*regcomp_parse != ')') + if (*PL_regcomp_parse != ')') FAIL("Sequence (?{...}) not terminated or not {}-balanced"); if (!SIZE_ONLY) { AV *av; - if (regcomp_parse - 1 - s) - sv = newSVpv(s, regcomp_parse - 1 - s); + if (PL_regcomp_parse - 1 - s) + sv = newSVpvn(s, PL_regcomp_parse - 1 - s); else - sv = newSVpv("", 0); + sv = newSVpvn("", 0); rop = sv_compile_2op(sv, &sop, "re", &av); - n = add_data(3, "nso"); - regcomp_rx->data->data[n] = (void*)rop; - regcomp_rx->data->data[n+1] = (void*)av; - regcomp_rx->data->data[n+2] = (void*)sop; + n = add_data(3, "nop"); + PL_regcomp_rx->data->data[n] = (void*)rop; + PL_regcomp_rx->data->data[n+1] = (void*)sop; + PL_regcomp_rx->data->data[n+2] = (void*)av; SvREFCNT_dec(sv); - } else { /* First pass */ - if (tainted) + } + else { /* First pass */ + if (PL_reginterp_cnt < ++PL_seen_evals + && PL_curcop != &PL_compiling) + /* No compiled RE interpolated, has runtime + components ===> unsafe. */ + FAIL("Eval-group not allowed at runtime, use re 'eval'"); + if (PL_tainted) FAIL("Eval-group in insecure regular expression"); } nextchar(); + if (logical) { + ret = reg_node(LOGICAL); + if (!SIZE_ONLY) + ret->flags = 2; + regtail(ret, reganode(EVAL, n)); + return ret; + } return reganode(EVAL, n); } case '(': { - if (regcomp_parse[0] == '?') { - if (regcomp_parse[1] == '=' || regcomp_parse[1] == '!' - || regcomp_parse[1] == '<' - || regcomp_parse[1] == '{') { /* Lookahead or eval. */ + if (PL_regcomp_parse[0] == '?') { + if (PL_regcomp_parse[1] == '=' || PL_regcomp_parse[1] == '!' + || PL_regcomp_parse[1] == '<' + || PL_regcomp_parse[1] == '{') { /* Lookahead or eval. */ I32 flag; ret = reg_node(LOGICAL); + if (!SIZE_ONLY) + ret->flags = 1; regtail(ret, reg(1, &flag)); goto insert_if; } - } else if (regcomp_parse[0] >= '1' && regcomp_parse[0] <= '9' ) { - parno = atoi(regcomp_parse++); + } + else if (PL_regcomp_parse[0] >= '1' && PL_regcomp_parse[0] <= '9' ) { + parno = atoi(PL_regcomp_parse++); - while (isDIGIT(*regcomp_parse)) - regcomp_parse++; + while (isDIGIT(*PL_regcomp_parse)) + PL_regcomp_parse++; ret = reganode(GROUPP, parno); if ((c = *nextchar()) != ')') FAIL2("Switch (?(number%c not recognized", c); @@ -1078,12 +1260,17 @@ reg(I32 paren, I32 *flagp) else regtail(br, reganode(LONGJMP, 0)); c = *nextchar(); + if (flags&HASWIDTH) + *flagp |= HASWIDTH; if (c == '|') { lastbr = reganode(IFTHEN, 0); /* Fake one for optimizer. */ regbranch(&flags, 1); regtail(ret, lastbr); + if (flags&HASWIDTH) + *flagp |= HASWIDTH; c = *nextchar(); - } else + } + else lastbr = NULL; if (c != ')') FAIL("Switch (?(condition)... contains too many branches"); @@ -1092,70 +1279,89 @@ reg(I32 paren, I32 *flagp) if (lastbr) { regtail(lastbr, ender); regtail(NEXTOPER(NEXTOPER(lastbr)), ender); - } else + } + else regtail(ret, ender); return ret; - } else { - FAIL2("Unknown condition for (?(%.2s", regcomp_parse); + } + else { + FAIL2("Unknown condition for (?(%.2s", PL_regcomp_parse); } } case 0: FAIL("Sequence (? incomplete"); break; default: - --regcomp_parse; - while (*regcomp_parse && strchr("iogcmsx", *regcomp_parse)) { - if (*regcomp_parse != 'o') - pmflag(®flags, *regcomp_parse); - ++regcomp_parse; + --PL_regcomp_parse; + parse_flags: + while (*PL_regcomp_parse && strchr("iogcmsx", *PL_regcomp_parse)) { + if (*PL_regcomp_parse != 'o') + pmflag(flagsp, *PL_regcomp_parse); + ++PL_regcomp_parse; } + if (*PL_regcomp_parse == '-') { + flagsp = &negflags; + ++PL_regcomp_parse; + goto parse_flags; + } + PL_regflags |= posflags; + PL_regflags &= ~negflags; + if (*PL_regcomp_parse == ':') { + PL_regcomp_parse++; + paren = ':'; + break; + } unknown: - if (*regcomp_parse != ')') - FAIL2("Sequence (?%c...) not recognized", *regcomp_parse); + if (*PL_regcomp_parse != ')') + FAIL2("Sequence (?%c...) not recognized", *PL_regcomp_parse); nextchar(); *flagp = TRYAGAIN; return NULL; } } else { - parno = regnpar; - regnpar++; + parno = PL_regnpar; + PL_regnpar++; ret = reganode(OPEN, parno); open = 1; } - } else + } + else ret = NULL; /* Pick up the branches, linking them together. */ br = regbranch(&flags, 1); if (br == NULL) return(NULL); - if (*regcomp_parse == '|') { - if (!SIZE_ONLY && extralen) { + if (*PL_regcomp_parse == '|') { + if (!SIZE_ONLY && PL_extralen) { reginsert(BRANCHJ, br); - } else + } + else reginsert(BRANCH, br); have_branch = 1; if (SIZE_ONLY) - extralen += 1; /* For BRANCHJ-BRANCH. */ - } else if (paren == ':') { + PL_extralen += 1; /* For BRANCHJ-BRANCH. */ + } + else if (paren == ':') { *flagp |= flags&SIMPLE; } if (open) { /* Starts with OPEN. */ regtail(ret, br); /* OPEN -> first. */ - } else if (paren != '?') /* Not Conditional */ + } + else if (paren != '?') /* Not Conditional */ ret = br; if (flags&HASWIDTH) *flagp |= HASWIDTH; *flagp |= flags&SPSTART; lastbr = br; - while (*regcomp_parse == '|') { - if (!SIZE_ONLY && extralen) { + while (*PL_regcomp_parse == '|') { + if (!SIZE_ONLY && PL_extralen) { ender = reganode(LONGJMP,0); regtail(NEXTOPER(NEXTOPER(lastbr)), ender); /* Append to the previous. */ } if (SIZE_ONLY) - extralen += 2; /* Account for LONGJMP. */ + PL_extralen += 2; /* Account for LONGJMP. */ nextchar(); br = regbranch(&flags, 0); if (br == NULL) @@ -1216,18 +1422,20 @@ reg(I32 paren, I32 *flagp) } /* Check for proper termination. */ - if (paren && (regcomp_parse >= regxend || *nextchar() != ')')) { - FAIL("unmatched () in regexp"); - } else if (!paren && regcomp_parse < regxend) { - if (*regcomp_parse == ')') { + if (paren) { + PL_regflags = oregflags; + if (PL_regcomp_parse >= PL_regxend || *nextchar() != ')') { FAIL("unmatched () in regexp"); - } else + } + } + else if (!paren && PL_regcomp_parse < PL_regxend) { + if (*PL_regcomp_parse == ')') { + FAIL("unmatched () in regexp"); + } + else FAIL("junk on end of regexp"); /* "Can't happen". */ /* NOTREACHED */ } - if (paren != 0) { - regflags = oregflags; - } return(ret); } @@ -1238,8 +1446,9 @@ reg(I32 paren, I32 *flagp) * Implements the concatenation operator. */ STATIC regnode * -regbranch(I32 *flagp, I32 first) +S_regbranch(pTHX_ I32 *flagp, I32 first) { + dTHR; register regnode *ret; register regnode *chain = NULL; register regnode *latest; @@ -1248,33 +1457,34 @@ regbranch(I32 *flagp, I32 first) if (first) ret = NULL; else { - if (!SIZE_ONLY && extralen) + if (!SIZE_ONLY && PL_extralen) ret = reganode(BRANCHJ,0); else ret = reg_node(BRANCH); } if (!first && SIZE_ONLY) - extralen += 1; /* BRANCHJ */ + PL_extralen += 1; /* BRANCHJ */ *flagp = WORST; /* Tentatively. */ - regcomp_parse--; + PL_regcomp_parse--; nextchar(); - while (regcomp_parse < regxend && *regcomp_parse != '|' && *regcomp_parse != ')') { + while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != '|' && *PL_regcomp_parse != ')') { flags &= ~TRYAGAIN; latest = regpiece(&flags); if (latest == NULL) { if (flags & TRYAGAIN) continue; return(NULL); - } else if (ret == NULL) + } + else if (ret == NULL) ret = latest; *flagp |= flags&HASWIDTH; if (chain == NULL) /* First piece. */ *flagp |= flags&SPSTART; else { - regnaughty++; + PL_regnaughty++; regtail(chain, latest); } chain = latest; @@ -1302,13 +1512,14 @@ regbranch(I32 *flagp, I32 first) * endmarker role is not redundant. */ STATIC regnode * -regpiece(I32 *flagp) +S_regpiece(pTHX_ I32 *flagp) { + dTHR; register regnode *ret; register char op; register char *next; I32 flags; - char *origparse = regcomp_parse; + char *origparse = PL_regcomp_parse; char *maxpos; I32 min; I32 max = REG_INFTY; @@ -1320,10 +1531,10 @@ regpiece(I32 *flagp) return(NULL); } - op = *regcomp_parse; + op = *PL_regcomp_parse; - if (op == '{' && regcurly(regcomp_parse)) { - next = regcomp_parse + 1; + if (op == '{' && regcurly(PL_regcomp_parse)) { + next = PL_regcomp_parse + 1; maxpos = Nullch; while (isDIGIT(*next) || *next == ',') { if (*next == ',') { @@ -1337,39 +1548,42 @@ regpiece(I32 *flagp) if (*next == '}') { /* got one */ if (!maxpos) maxpos = next; - regcomp_parse++; - min = atoi(regcomp_parse); + PL_regcomp_parse++; + min = atoi(PL_regcomp_parse); if (*maxpos == ',') maxpos++; else - maxpos = regcomp_parse; + maxpos = PL_regcomp_parse; max = atoi(maxpos); if (!max && *maxpos != '0') max = REG_INFTY; /* meaning "infinity" */ else if (max >= REG_INFTY) FAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1); - regcomp_parse = next; + PL_regcomp_parse = next; nextchar(); do_curly: if ((flags&SIMPLE)) { - regnaughty += 2 + regnaughty / 2; + PL_regnaughty += 2 + PL_regnaughty / 2; reginsert(CURLY, ret); } else { - regnaughty += 4 + regnaughty; /* compound interest */ - regtail(ret, reg_node(WHILEM)); - if (!SIZE_ONLY && extralen) { + regnode *w = reg_node(WHILEM); + + w->flags = 0; + regtail(ret, w); + if (!SIZE_ONLY && PL_extralen) { reginsert(LONGJMP,ret); reginsert(NOTHING,ret); NEXT_OFF(ret) = 3; /* Go over LONGJMP. */ } reginsert(CURLYX,ret); - if (!SIZE_ONLY && extralen) + if (!SIZE_ONLY && PL_extralen) NEXT_OFF(ret) = 3; /* Go over NOTHING to LONGJMP. */ regtail(ret, reg_node(NOTHING)); if (SIZE_ONLY) - extralen += 3; + PL_reg_whilem_seen++, PL_extralen += 3; + PL_regnaughty += 4 + PL_regnaughty; /* compound interest */ } ret->flags = 0; @@ -1405,35 +1619,37 @@ regpiece(I32 *flagp) if (op == '*' && (flags&SIMPLE)) { reginsert(STAR, ret); ret->flags = 0; - regnaughty += 4; + PL_regnaughty += 4; } else if (op == '*') { min = 0; goto do_curly; - } else if (op == '+' && (flags&SIMPLE)) { + } + else if (op == '+' && (flags&SIMPLE)) { reginsert(PLUS, ret); ret->flags = 0; - regnaughty += 3; + PL_regnaughty += 3; } else if (op == '+') { min = 1; goto do_curly; - } else if (op == '?') { + } + else if (op == '?') { min = 0; max = 1; goto do_curly; } nest_check: - if (dowarn && !SIZE_ONLY && !(flags&HASWIDTH) && max > 10000) { - warn("%.*s matches null string many times", - regcomp_parse - origparse, origparse); + if (ckWARN(WARN_UNSAFE) && !SIZE_ONLY && !(flags&HASWIDTH) && max > REG_INFTY/3) { + Perl_warner(aTHX_ WARN_UNSAFE, "%.*s matches null string many times", + PL_regcomp_parse - origparse, origparse); } - if (*regcomp_parse == '?') { + if (*PL_regcomp_parse == '?') { nextchar(); reginsert(MINMOD, ret); regtail(ret, ret + NODE_STEP_REGNODE); } - if (ISMULT2(regcomp_parse)) + if (ISMULT2(PL_regcomp_parse)) FAIL("nested *?+ in regexp"); return(ret); @@ -1450,48 +1666,61 @@ regpiece(I32 *flagp) * [Yes, it is worth fixing, some scripts can run twice the speed.] */ STATIC regnode * -regatom(I32 *flagp) +S_regatom(pTHX_ I32 *flagp) { + dTHR; register regnode *ret = 0; I32 flags; *flagp = WORST; /* Tentatively. */ tryagain: - switch (*regcomp_parse) { + switch (*PL_regcomp_parse) { case '^': - seen_zerolen++; + PL_seen_zerolen++; nextchar(); - if (regflags & PMf_MULTILINE) + if (PL_regflags & PMf_MULTILINE) ret = reg_node(MBOL); - else if (regflags & PMf_SINGLELINE) + else if (PL_regflags & PMf_SINGLELINE) ret = reg_node(SBOL); else ret = reg_node(BOL); break; case '$': - if (regcomp_parse[1]) - seen_zerolen++; + if (PL_regcomp_parse[1]) + PL_seen_zerolen++; nextchar(); - if (regflags & PMf_MULTILINE) + if (PL_regflags & PMf_MULTILINE) ret = reg_node(MEOL); - else if (regflags & PMf_SINGLELINE) + else if (PL_regflags & PMf_SINGLELINE) ret = reg_node(SEOL); else ret = reg_node(EOL); break; case '.': nextchar(); - if (regflags & PMf_SINGLELINE) - ret = reg_node(SANY); - else - ret = reg_node(ANY); - regnaughty++; - *flagp |= HASWIDTH|SIMPLE; + if (UTF) { + if (PL_regflags & PMf_SINGLELINE) + ret = reg_node(SANYUTF8); + else + ret = reg_node(ANYUTF8); + *flagp |= HASWIDTH; + } + else { + if (PL_regflags & PMf_SINGLELINE) + ret = reg_node(SANY); + else + ret = reg_node(REG_ANY); + *flagp |= HASWIDTH|SIMPLE; + } + PL_regnaughty++; break; case '[': - regcomp_parse++; - ret = regclass(); + PL_regcomp_parse++; + ret = (UTF ? regclassutf8() : regclass()); + if (*PL_regcomp_parse != ']') + FAIL("unmatched [] in regexp"); + nextchar(); *flagp |= HASWIDTH|SIMPLE; break; case '(': @@ -1510,12 +1739,12 @@ tryagain: *flagp |= TRYAGAIN; return NULL; } - FAIL2("internal urp in regexp at /%s/", regcomp_parse); + FAIL2("internal urp in regexp at /%s/", PL_regcomp_parse); /* Supposed to be caught earlier. */ break; case '{': - if (!regcurly(regcomp_parse)) { - regcomp_parse++; + if (!regcurly(PL_regcomp_parse)) { + PL_regcomp_parse++; goto defchar; } /* FALL THROUGH */ @@ -1525,16 +1754,16 @@ tryagain: FAIL("?+*{} follows nothing in regexp"); break; case '\\': - switch (*++regcomp_parse) { + switch (*++PL_regcomp_parse) { case 'A': - seen_zerolen++; + PL_seen_zerolen++; ret = reg_node(SBOL); *flagp |= SIMPLE; nextchar(); break; case 'G': ret = reg_node(GPOS); - regseen |= REG_SEEN_GPOS; + PL_regseen |= REG_SEEN_GPOS; *flagp |= SIMPLE; nextchar(); break; @@ -1546,50 +1775,121 @@ tryagain: case 'z': ret = reg_node(EOS); *flagp |= SIMPLE; - seen_zerolen++; /* Do not optimize RE away */ + PL_seen_zerolen++; /* Do not optimize RE away */ + nextchar(); + break; + case 'C': + ret = reg_node(SANY); + *flagp |= HASWIDTH|SIMPLE; + nextchar(); + break; + case 'X': + ret = reg_node(CLUMP); + *flagp |= HASWIDTH; nextchar(); + if (UTF && !PL_utf8_mark) + is_utf8_mark((U8*)"~"); /* preload table */ break; case 'w': - ret = reg_node((regflags & PMf_LOCALE) ? ALNUML : ALNUM); + ret = reg_node( + UTF + ? (LOC ? ALNUMLUTF8 : ALNUMUTF8) + : (LOC ? ALNUML : ALNUM)); *flagp |= HASWIDTH|SIMPLE; nextchar(); + if (UTF && !PL_utf8_alnum) + is_utf8_alnum((U8*)"a"); /* preload table */ break; case 'W': - ret = reg_node((regflags & PMf_LOCALE) ? NALNUML : NALNUM); + ret = reg_node( + UTF + ? (LOC ? NALNUMLUTF8 : NALNUMUTF8) + : (LOC ? NALNUML : NALNUM)); *flagp |= HASWIDTH|SIMPLE; nextchar(); + if (UTF && !PL_utf8_alnum) + is_utf8_alnum((U8*)"a"); /* preload table */ break; case 'b': - seen_zerolen++; - ret = reg_node((regflags & PMf_LOCALE) ? BOUNDL : BOUND); + PL_seen_zerolen++; + PL_regseen |= REG_SEEN_LOOKBEHIND; + ret = reg_node( + UTF + ? (LOC ? BOUNDLUTF8 : BOUNDUTF8) + : (LOC ? BOUNDL : BOUND)); *flagp |= SIMPLE; nextchar(); + if (UTF && !PL_utf8_alnum) + is_utf8_alnum((U8*)"a"); /* preload table */ break; case 'B': - seen_zerolen++; - ret = reg_node((regflags & PMf_LOCALE) ? NBOUNDL : NBOUND); + PL_seen_zerolen++; + PL_regseen |= REG_SEEN_LOOKBEHIND; + ret = reg_node( + UTF + ? (LOC ? NBOUNDLUTF8 : NBOUNDUTF8) + : (LOC ? NBOUNDL : NBOUND)); *flagp |= SIMPLE; nextchar(); + if (UTF && !PL_utf8_alnum) + is_utf8_alnum((U8*)"a"); /* preload table */ break; case 's': - ret = reg_node((regflags & PMf_LOCALE) ? SPACEL : SPACE); + ret = reg_node( + UTF + ? (LOC ? SPACELUTF8 : SPACEUTF8) + : (LOC ? SPACEL : SPACE)); *flagp |= HASWIDTH|SIMPLE; nextchar(); + if (UTF && !PL_utf8_space) + is_utf8_space((U8*)" "); /* preload table */ break; case 'S': - ret = reg_node((regflags & PMf_LOCALE) ? NSPACEL : NSPACE); + ret = reg_node( + UTF + ? (LOC ? NSPACELUTF8 : NSPACEUTF8) + : (LOC ? NSPACEL : NSPACE)); *flagp |= HASWIDTH|SIMPLE; nextchar(); + if (UTF && !PL_utf8_space) + is_utf8_space((U8*)" "); /* preload table */ break; case 'd': - ret = reg_node(DIGIT); + ret = reg_node(UTF ? DIGITUTF8 : DIGIT); *flagp |= HASWIDTH|SIMPLE; nextchar(); + if (UTF && !PL_utf8_digit) + is_utf8_digit((U8*)"1"); /* preload table */ break; case 'D': - ret = reg_node(NDIGIT); + ret = reg_node(UTF ? NDIGITUTF8 : NDIGIT); *flagp |= HASWIDTH|SIMPLE; nextchar(); + if (UTF && !PL_utf8_digit) + is_utf8_digit((U8*)"1"); /* preload table */ + break; + case 'p': + case 'P': + { /* a lovely hack--pretend we saw [\pX] instead */ + char* oldregxend = PL_regxend; + + if (PL_regcomp_parse[1] == '{') { + PL_regxend = strchr(PL_regcomp_parse, '}'); + if (!PL_regxend) + FAIL("Missing right brace on \\p{}"); + PL_regxend++; + } + else + PL_regxend = PL_regcomp_parse + 2; + PL_regcomp_parse--; + + ret = regclassutf8(); + + PL_regxend = oldregxend; + PL_regcomp_parse--; + nextchar(); + *flagp |= HASWIDTH|SIMPLE; + } break; case 'n': case 'r': @@ -1604,63 +1904,66 @@ tryagain: case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': { - I32 num = atoi(regcomp_parse); + I32 num = atoi(PL_regcomp_parse); - if (num > 9 && num >= regnpar) + if (num > 9 && num >= PL_regnpar) goto defchar; else { - regsawback = 1; - ret = reganode((regflags & PMf_FOLD) - ? ((regflags & PMf_LOCALE) ? REFFL : REFF) + if (!SIZE_ONLY && num > PL_regcomp_rx->nparens) + FAIL("reference to nonexistent group"); + PL_regsawback = 1; + ret = reganode(FOLD + ? (LOC ? REFFL : REFF) : REF, num); *flagp |= HASWIDTH; - while (isDIGIT(*regcomp_parse)) - regcomp_parse++; - regcomp_parse--; + while (isDIGIT(*PL_regcomp_parse)) + PL_regcomp_parse++; + PL_regcomp_parse--; nextchar(); } } break; case '\0': - if (regcomp_parse >= regxend) + if (PL_regcomp_parse >= PL_regxend) FAIL("trailing \\ in regexp"); /* FALL THROUGH */ default: + /* Do not generate `unrecognized' warnings here, we fall + back into the quick-grab loop below */ goto defchar; } break; case '#': - if (regflags & PMf_EXTENDED) { - while (regcomp_parse < regxend && *regcomp_parse != '\n') regcomp_parse++; - if (regcomp_parse < regxend) + if (PL_regflags & PMf_EXTENDED) { + while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != '\n') PL_regcomp_parse++; + if (PL_regcomp_parse < PL_regxend) goto tryagain; } /* FALL THROUGH */ default: { register I32 len; - register U8 ender; + register UV ender; register char *p; char *oldp, *s; I32 numlen; - regcomp_parse++; + PL_regcomp_parse++; defchar: - ret = reg_node((regflags & PMf_FOLD) - ? ((regflags & PMf_LOCALE) ? EXACTFL : EXACTF) + ret = reg_node(FOLD + ? (LOC ? EXACTFL : EXACTF) : EXACT); - s = (char *) OPERAND(ret); - regc(0, s++); /* save spot for len */ - for (len = 0, p = regcomp_parse - 1; - len < 127 && p < regxend; + s = STRING(ret); + for (len = 0, p = PL_regcomp_parse - 1; + len < 127 && p < PL_regxend; len++) { oldp = p; - if (regflags & PMf_EXTENDED) - p = regwhite(p, regxend); + if (PL_regflags & PMf_EXTENDED) + p = regwhite(p, PL_regxend); switch (*p) { case '^': case '$': @@ -1684,6 +1987,8 @@ tryagain: case 'S': case 'd': case 'D': + case 'p': + case 'P': --p; goto loopdone; case 'n': @@ -1711,8 +2016,26 @@ tryagain: p++; break; case 'x': - ender = scan_hex(++p, 2, &numlen); - p += numlen; + if (*++p == '{') { + char* e = strchr(p, '}'); + + if (!e) + FAIL("Missing right brace on \\x{}"); + else if (UTF) { + ender = scan_hex(p + 1, e - p, &numlen); + if (numlen + len >= 127) { /* numlen is generous */ + p--; + goto loopdone; + } + p = e + 1; + } + else + FAIL("Can't use \\x{} without 'use utf8' declaration"); + } + else { + ender = scan_hex(p, 2, &numlen); + p += numlen; + } break; case 'c': p++; @@ -1722,7 +2045,7 @@ tryagain: case '0': case '1': case '2': case '3':case '4': case '5': case '6': case '7': case '8':case '9': if (*p == '0' || - (isDIGIT(p[1]) && atoi(p) >= regnpar) ) { + (isDIGIT(p[1]) && atoi(p) >= PL_regnpar) ) { ender = scan_oct(p, 3, &numlen); p += numlen; } @@ -1732,33 +2055,60 @@ tryagain: } break; case '\0': - if (p >= regxend) + if (p >= PL_regxend) FAIL("trailing \\ in regexp"); /* FALL THROUGH */ default: - ender = *p++; - break; + if (!SIZE_ONLY && ckWARN(WARN_UNSAFE) && isALPHA(*p)) + Perl_warner(aTHX_ WARN_UNSAFE, + "/%.127s/: Unrecognized escape \\%c passed through", + PL_regprecomp, + *p); + goto normal_default; } break; default: - ender = *p++; + normal_default: + if ((*p & 0xc0) == 0xc0 && UTF) { + ender = utf8_to_uv((U8*)p, &numlen); + p += numlen; + } + else + ender = *p++; break; } - if (regflags & PMf_EXTENDED) - p = regwhite(p, regxend); + if (PL_regflags & PMf_EXTENDED) + p = regwhite(p, PL_regxend); + if (UTF && FOLD) { + if (LOC) + ender = toLOWER_LC_uni(ender); + else + ender = toLOWER_uni(ender); + } if (ISMULT2(p)) { /* Back off on ?+*. */ if (len) p = oldp; + else if (ender >= 0x80 && UTF) { + reguni(ender, s, &numlen); + s += numlen; + len += numlen; + } else { len++; - regc(ender, s++); + REGC(ender, s++); } break; } - regc(ender, s++); + if (ender >= 0x80 && UTF) { + reguni(ender, s, &numlen); + s += numlen; + len += numlen - 1; + } + else + REGC(ender, s++); } loopdone: - regcomp_parse = p - 1; + PL_regcomp_parse = p - 1; nextchar(); if (len < 0) FAIL("internal disaster in regexp"); @@ -1767,13 +2117,11 @@ tryagain: if (len == 1) *flagp |= SIMPLE; if (!SIZE_ONLY) - *OPERAND(ret) = len; - regc('\0', s++); - if (SIZE_ONLY) { - regsize += (len + 2 + sizeof(regnode) - 1) / sizeof(regnode); - } else { - regcode += (len + 2 + sizeof(regnode) - 1) / sizeof(regnode); - } + STR_LEN(ret) = len; + if (SIZE_ONLY) + PL_regsize += STR_SZ(len); + else + PL_regcode += STR_SZ(len); } break; } @@ -1782,7 +2130,7 @@ tryagain: } STATIC char * -regwhite(char *p, char *e) +S_regwhite(pTHX_ char *p, char *e) { while (p < e) { if (isSPACE(*p)) @@ -1798,232 +2146,821 @@ regwhite(char *p, char *e) return p; } +/* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]]. + Character classes ([:foo:]) can also be negated ([:^foo:]). + Returns a named class id (ANYOF_XXX) if successful, -1 otherwise. + Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed, + but trigger warnings because they are currently unimplemented. */ +STATIC I32 +S_regpposixcc(pTHX_ I32 value) +{ + dTHR; + char *posixcc = 0; + I32 namedclass = -1; + + if (value == '[' && PL_regcomp_parse + 1 < PL_regxend && + /* I smell either [: or [= or [. -- POSIX has been here, right? */ + (*PL_regcomp_parse == ':' || + *PL_regcomp_parse == '=' || + *PL_regcomp_parse == '.')) { + char c = *PL_regcomp_parse; + char* s = PL_regcomp_parse++; + + while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != c) + PL_regcomp_parse++; + if (PL_regcomp_parse == PL_regxend) + /* Grandfather lone [:, [=, [. */ + PL_regcomp_parse = s; + else { + char* t = PL_regcomp_parse++; /* skip over the c */ + + if (*PL_regcomp_parse == ']') { + PL_regcomp_parse++; /* skip over the ending ] */ + posixcc = s + 1; + if (*s == ':') { + I32 complement = *posixcc == '^' ? *posixcc++ : 0; + I32 skip = 5; /* the most common skip */ + + switch (*posixcc) { + case 'a': + if (strnEQ(posixcc, "alnum", 5)) + namedclass = + complement ? ANYOF_NALNUMC : ANYOF_ALNUMC; + else if (strnEQ(posixcc, "alpha", 5)) + namedclass = + complement ? ANYOF_NALPHA : ANYOF_ALPHA; + else if (strnEQ(posixcc, "ascii", 5)) + namedclass = + complement ? ANYOF_NASCII : ANYOF_ASCII; + break; + case 'c': + if (strnEQ(posixcc, "cntrl", 5)) + namedclass = + complement ? ANYOF_NCNTRL : ANYOF_CNTRL; + break; + case 'd': + if (strnEQ(posixcc, "digit", 5)) + namedclass = + complement ? ANYOF_NDIGIT : ANYOF_DIGIT; + break; + case 'g': + if (strnEQ(posixcc, "graph", 5)) + namedclass = + complement ? ANYOF_NGRAPH : ANYOF_GRAPH; + break; + case 'l': + if (strnEQ(posixcc, "lower", 5)) + namedclass = + complement ? ANYOF_NLOWER : ANYOF_LOWER; + break; + case 'p': + if (strnEQ(posixcc, "print", 5)) + namedclass = + complement ? ANYOF_NPRINT : ANYOF_PRINT; + else if (strnEQ(posixcc, "punct", 5)) + namedclass = + complement ? ANYOF_NPUNCT : ANYOF_PUNCT; + break; + case 's': + if (strnEQ(posixcc, "space", 5)) + namedclass = + complement ? ANYOF_NSPACE : ANYOF_SPACE; + case 'u': + if (strnEQ(posixcc, "upper", 5)) + namedclass = + complement ? ANYOF_NUPPER : ANYOF_UPPER; + break; + case 'w': /* this is not POSIX, this is the Perl \w */ + if (strnEQ(posixcc, "word", 4)) { + namedclass = + complement ? ANYOF_NALNUM : ANYOF_ALNUM; + skip = 4; + } + break; + case 'x': + if (strnEQ(posixcc, "xdigit", 6)) { + namedclass = + complement ? ANYOF_NXDIGIT : ANYOF_XDIGIT; + skip = 6; + } + break; + } + if ((namedclass == OOB_NAMEDCLASS || + !(posixcc + skip + 2 < PL_regxend && + (posixcc[skip] == ':' && + posixcc[skip + 1] == ']')))) + Perl_croak(aTHX_ "Character class [:%.*s:] unknown", + t - s - 1, s + 1); + } else if (ckWARN(WARN_UNSAFE) && !SIZE_ONLY) + /* [[=foo=]] and [[.foo.]] are still future. */ + Perl_warner(aTHX_ WARN_UNSAFE, + "Character class syntax [%c %c] is reserved for future extensions", c, c); + } else { + /* Maternal grandfather: + * "[:" ending in ":" but not in ":]" */ + PL_regcomp_parse = s; + } + } + } + + return namedclass; +} + +STATIC void +S_checkposixcc(pTHX) +{ + if (ckWARN(WARN_UNSAFE) && !SIZE_ONLY && + (*PL_regcomp_parse == ':' || + *PL_regcomp_parse == '=' || + *PL_regcomp_parse == '.')) { + char *s = PL_regcomp_parse; + char c = *s++; + + while(*s && isALNUM(*s)) + s++; + if (*s && c == *s && s[1] == ']') { + Perl_warner(aTHX_ WARN_UNSAFE, + "Character class syntax [%c %c] belongs inside character classes", c, c); + if (c == '=' || c == '.') + Perl_warner(aTHX_ WARN_UNSAFE, + "Character class syntax [%c %c] is reserved for future extensions", c, c); + } + } +} + STATIC regnode * -regclass(void) +S_regclass(pTHX) { + dTHR; register char *opnd, *s; - register I32 Class; - register I32 lastclass = 1234; + register I32 value; + register I32 lastvalue = OOB_CHAR8; register I32 range = 0; register regnode *ret; register I32 def; I32 numlen; + I32 namedclass; + char *rangebegin; - s = opnd = (char *) OPERAND(regcode); + s = opnd = MASK(PL_regcode); ret = reg_node(ANYOF); - for (Class = 0; Class < 33; Class++) - regc(0, s++); - if (*regcomp_parse == '^') { /* Complement of range. */ - regnaughty++; - regcomp_parse++; + for (value = 0; value < ANYOF_SIZE; value++) + REGC(0, s++); + if (*PL_regcomp_parse == '^') { /* Complement of range. */ + PL_regnaughty++; + PL_regcomp_parse++; if (!SIZE_ONLY) - *opnd |= ANYOF_INVERT; + ANYOF_FLAGS(opnd) |= ANYOF_INVERT; } if (!SIZE_ONLY) { - regcode += ANY_SKIP; - if (regflags & PMf_FOLD) - *opnd |= ANYOF_FOLD; - if (regflags & PMf_LOCALE) - *opnd |= ANYOF_LOCALE; - } else { - regsize += ANY_SKIP; + PL_regcode += ANY_SKIP; + if (FOLD) + ANYOF_FLAGS(opnd) |= ANYOF_FOLD; + if (LOC) + ANYOF_FLAGS(opnd) |= ANYOF_LOCALE; } - if (*regcomp_parse == ']' || *regcomp_parse == '-') + else { + PL_regsize += ANY_SKIP; + } + + checkposixcc(); + + if (*PL_regcomp_parse == ']' || *PL_regcomp_parse == '-') goto skipcond; /* allow 1st char to be ] or - */ - while (regcomp_parse < regxend && *regcomp_parse != ']') { + while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != ']') { skipcond: - Class = UCHARAT(regcomp_parse++); - if (Class == '[' && regcomp_parse + 1 < regxend && - /* I smell either [: or [= or [. -- POSIX has been here, right? */ - (*regcomp_parse == ':' || *regcomp_parse == '=' || *regcomp_parse == '.')) { - char posixccc = *regcomp_parse; - char* posixccs = regcomp_parse++; - - while (regcomp_parse < regxend && *regcomp_parse != posixccc) - regcomp_parse++; - if (regcomp_parse == regxend) - /* Grandfather lone [:, [=, [. */ - regcomp_parse = posixccs; - else { - regcomp_parse++; /* skip over the posixccc */ - if (*regcomp_parse == ']') { - /* Not Implemented Yet. - * (POSIX Extended Character Classes, that is) - * The text between e.g. [: and :] would start - * at posixccs + 1 and stop at regcomp_parse - 2. */ - if (dowarn && !SIZE_ONLY) - warn("Character class syntax [%c %c] is reserved for future extensions", posixccc, posixccc); - regcomp_parse++; /* skip over the ending ] */ - } + namedclass = OOB_NAMEDCLASS; + if (!range) + rangebegin = PL_regcomp_parse; + value = UCHARAT(PL_regcomp_parse++); + if (value == '[') + namedclass = regpposixcc(value); + else if (value == '\\') { + value = UCHARAT(PL_regcomp_parse++); + switch (value) { + case 'w': namedclass = ANYOF_ALNUM; break; + case 'W': namedclass = ANYOF_NALNUM; break; + case 's': namedclass = ANYOF_SPACE; break; + case 'S': namedclass = ANYOF_NSPACE; break; + case 'd': namedclass = ANYOF_DIGIT; break; + case 'D': namedclass = ANYOF_NDIGIT; break; + case 'n': value = '\n'; break; + case 'r': value = '\r'; break; + case 't': value = '\t'; break; + case 'f': value = '\f'; break; + case 'b': value = '\b'; break; + case 'e': value = '\033'; break; + case 'a': value = '\007'; break; + case 'x': + value = scan_hex(PL_regcomp_parse, 2, &numlen); + PL_regcomp_parse += numlen; + break; + case 'c': + value = UCHARAT(PL_regcomp_parse++); + value = toCTRL(value); + break; + case '0': case '1': case '2': case '3': case '4': + case '5': case '6': case '7': case '8': case '9': + value = scan_oct(--PL_regcomp_parse, 3, &numlen); + PL_regcomp_parse += numlen; + break; } } - if (Class == '\\') { - Class = UCHARAT(regcomp_parse++); - switch (Class) { - case 'w': + if (namedclass > OOB_NAMEDCLASS) { + if (range) { /* a-\d, a-[:digit:] */ if (!SIZE_ONLY) { - if (regflags & PMf_LOCALE) - *opnd |= ANYOF_ALNUML; + if (ckWARN(WARN_UNSAFE)) + Perl_warner(aTHX_ WARN_UNSAFE, + "/%.127s/: false [] range \"%*.*s\" in regexp", + PL_regprecomp, + PL_regcomp_parse - rangebegin, + PL_regcomp_parse - rangebegin, + rangebegin); + ANYOF_BITMAP_SET(opnd, lastvalue); + ANYOF_BITMAP_SET(opnd, '-'); + } + range = 0; /* this is not a true range */ + } + if (!SIZE_ONLY) { + switch (namedclass) { + case ANYOF_ALNUM: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_ALNUM); else { - for (Class = 0; Class < 256; Class++) - if (isALNUM(Class)) - ANYOF_SET(opnd, Class); + for (value = 0; value < 256; value++) + if (isALNUM(value)) + ANYOF_BITMAP_SET(opnd, value); } - } - lastclass = 1234; - continue; - case 'W': - if (!SIZE_ONLY) { - if (regflags & PMf_LOCALE) - *opnd |= ANYOF_NALNUML; + break; + case ANYOF_NALNUM: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NALNUM); else { - for (Class = 0; Class < 256; Class++) - if (!isALNUM(Class)) - ANYOF_SET(opnd, Class); + for (value = 0; value < 256; value++) + if (!isALNUM(value)) + ANYOF_BITMAP_SET(opnd, value); } - } - lastclass = 1234; - continue; - case 's': - if (!SIZE_ONLY) { - if (regflags & PMf_LOCALE) - *opnd |= ANYOF_SPACEL; + break; + case ANYOF_SPACE: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_SPACE); else { - for (Class = 0; Class < 256; Class++) - if (isSPACE(Class)) - ANYOF_SET(opnd, Class); + for (value = 0; value < 256; value++) + if (isSPACE(value)) + ANYOF_BITMAP_SET(opnd, value); } - } - lastclass = 1234; - continue; - case 'S': - if (!SIZE_ONLY) { - if (regflags & PMf_LOCALE) - *opnd |= ANYOF_NSPACEL; + break; + case ANYOF_NSPACE: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NSPACE); + else { + for (value = 0; value < 256; value++) + if (!isSPACE(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_DIGIT: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_DIGIT); + else { + for (value = '0'; value <= '9'; value++) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_NDIGIT: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NDIGIT); + else { + for (value = 0; value < '0'; value++) + ANYOF_BITMAP_SET(opnd, value); + for (value = '9' + 1; value < 256; value++) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_NALNUMC: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NALNUMC); + else { + for (value = 0; value < 256; value++) + if (!isALNUMC(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_ALNUMC: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_ALNUMC); + else { + for (value = 0; value < 256; value++) + if (isALNUMC(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_ALPHA: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_ALPHA); + else { + for (value = 0; value < 256; value++) + if (isALPHA(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_NALPHA: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NALPHA); + else { + for (value = 0; value < 256; value++) + if (!isALPHA(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_ASCII: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_ASCII); + else { + for (value = 0; value < 128; value++) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_NASCII: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NASCII); + else { + for (value = 128; value < 256; value++) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_CNTRL: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_CNTRL); + else { + for (value = 0; value < 256; value++) + if (isCNTRL(value)) + ANYOF_BITMAP_SET(opnd, value); + } + lastvalue = OOB_CHAR8; + break; + case ANYOF_NCNTRL: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NCNTRL); + else { + for (value = 0; value < 256; value++) + if (!isCNTRL(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_GRAPH: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_GRAPH); + else { + for (value = 0; value < 256; value++) + if (isGRAPH(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_NGRAPH: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NGRAPH); + else { + for (value = 0; value < 256; value++) + if (!isGRAPH(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_LOWER: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_LOWER); + else { + for (value = 0; value < 256; value++) + if (isLOWER(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_NLOWER: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NLOWER); + else { + for (value = 0; value < 256; value++) + if (!isLOWER(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_PRINT: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_PRINT); + else { + for (value = 0; value < 256; value++) + if (isPRINT(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_NPRINT: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NPRINT); + else { + for (value = 0; value < 256; value++) + if (!isPRINT(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_PUNCT: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_PUNCT); + else { + for (value = 0; value < 256; value++) + if (isPUNCT(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_NPUNCT: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NPUNCT); + else { + for (value = 0; value < 256; value++) + if (!isPUNCT(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_UPPER: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_UPPER); + else { + for (value = 0; value < 256; value++) + if (isUPPER(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_NUPPER: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NUPPER); + else { + for (value = 0; value < 256; value++) + if (!isUPPER(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_XDIGIT: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_XDIGIT); + else { + for (value = 0; value < 256; value++) + if (isXDIGIT(value)) + ANYOF_BITMAP_SET(opnd, value); + } + break; + case ANYOF_NXDIGIT: + if (LOC) + ANYOF_CLASS_SET(opnd, ANYOF_NXDIGIT); else { - for (Class = 0; Class < 256; Class++) - if (!isSPACE(Class)) - ANYOF_SET(opnd, Class); + for (value = 0; value < 256; value++) + if (!isXDIGIT(value)) + ANYOF_BITMAP_SET(opnd, value); } + break; + default: + FAIL("invalid [::] class in regexp"); + break; } - lastclass = 1234; + if (LOC) + ANYOF_FLAGS(opnd) |= ANYOF_CLASS; continue; - case 'd': - if (!SIZE_ONLY) { - for (Class = '0'; Class <= '9'; Class++) - ANYOF_SET(opnd, Class); + } + } + if (range) { + if (lastvalue > value) /* b-a */ { + Perl_croak(aTHX_ + "/%.127s/: invalid [] range \"%*.*s\" in regexp", + PL_regprecomp, + PL_regcomp_parse - rangebegin, + PL_regcomp_parse - rangebegin, + rangebegin); + } + range = 0; + } + else { + lastvalue = value; + if (*PL_regcomp_parse == '-' && PL_regcomp_parse+1 < PL_regxend && + PL_regcomp_parse[1] != ']') { + PL_regcomp_parse++; + if (namedclass > OOB_NAMEDCLASS) { /* \w-, [:word:]- */ + if (ckWARN(WARN_UNSAFE)) + Perl_warner(aTHX_ WARN_UNSAFE, + "/%.127s/: false [] range \"%*.*s\" in regexp", + PL_regprecomp, + PL_regcomp_parse - rangebegin, + PL_regcomp_parse - rangebegin, + rangebegin); + if (!SIZE_ONLY) + ANYOF_BITMAP_SET(opnd, '-'); + } else + range = 1; + continue; /* do it next time */ + } + } + /* now is the next time */ + if (!SIZE_ONLY) { +#ifndef ASCIIish /* EBCDIC, for example. */ + if ((isLOWER(lastvalue) && isLOWER(value)) || + (isUPPER(lastvalue) && isUPPER(value))) + { + I32 i; + if (isLOWER(lastvalue)) { + for (i = lastvalue; i <= value; i++) + if (isLOWER(i)) + ANYOF_BITMAP_SET(opnd, i); + } else { + for (i = lastvalue; i <= value; i++) + if (isUPPER(i)) + ANYOF_BITMAP_SET(opnd, i); + } + } + else +#endif + for ( ; lastvalue <= value; lastvalue++) + ANYOF_BITMAP_SET(opnd, lastvalue); + } + range = 0; + } + /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */ + if (!SIZE_ONLY && + (ANYOF_FLAGS(opnd) & (ANYOF_FLAGS_ALL ^ ANYOF_INVERT)) == ANYOF_FOLD) { + for (value = 0; value < 256; ++value) { + if (ANYOF_BITMAP_TEST(opnd, value)) { + I32 cf = PL_fold[value]; + ANYOF_BITMAP_SET(opnd, cf); + } + } + ANYOF_FLAGS(opnd) &= ~ANYOF_FOLD; + } + /* optimize inverted simple patterns (e.g. [^a-z]) */ + if (!SIZE_ONLY && (ANYOF_FLAGS(opnd) & ANYOF_FLAGS_ALL) == ANYOF_INVERT) { + for (value = 0; value < ANYOF_BITMAP_SIZE; ++value) + opnd[ANYOF_BITMAP_OFFSET + value] ^= ANYOF_FLAGS_ALL; + ANYOF_FLAGS(opnd) = 0; + } + return ret; +} + +STATIC regnode * +S_regclassutf8(pTHX) +{ + dTHR; + register char *opnd, *e; + register U32 value; + register U32 lastvalue = OOB_UTF8; + register I32 range = 0; + register regnode *ret; + I32 numlen; + I32 n; + SV *listsv; + U8 flags = 0; + I32 namedclass; + char *rangebegin; + + if (*PL_regcomp_parse == '^') { /* Complement of range. */ + PL_regnaughty++; + PL_regcomp_parse++; + if (!SIZE_ONLY) + flags |= ANYOF_INVERT; + } + if (!SIZE_ONLY) { + if (FOLD) + flags |= ANYOF_FOLD; + if (LOC) + flags |= ANYOF_LOCALE; + listsv = newSVpvn("# comment\n",10); + } + + checkposixcc(); + + if (*PL_regcomp_parse == ']' || *PL_regcomp_parse == '-') + goto skipcond; /* allow 1st char to be ] or - */ + + while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != ']') { + skipcond: + namedclass = OOB_NAMEDCLASS; + if (!range) + rangebegin = PL_regcomp_parse; + value = utf8_to_uv((U8*)PL_regcomp_parse, &numlen); + PL_regcomp_parse += numlen; + if (value == '[') + namedclass = regpposixcc(value); + else if (value == '\\') { + value = utf8_to_uv((U8*)PL_regcomp_parse, &numlen); + PL_regcomp_parse += numlen; + switch (value) { + case 'w': namedclass = ANYOF_ALNUM; break; + case 'W': namedclass = ANYOF_NALNUM; break; + case 's': namedclass = ANYOF_SPACE; break; + case 'S': namedclass = ANYOF_NSPACE; break; + case 'd': namedclass = ANYOF_DIGIT; break; + case 'D': namedclass = ANYOF_NDIGIT; break; + case 'p': + case 'P': + if (*PL_regcomp_parse == '{') { + e = strchr(PL_regcomp_parse++, '}'); + if (!e) + FAIL("Missing right brace on \\p{}"); + n = e - PL_regcomp_parse; + } + else { + e = PL_regcomp_parse; + n = 1; } - lastclass = 1234; - continue; - case 'D': if (!SIZE_ONLY) { - for (Class = 0; Class < '0'; Class++) - ANYOF_SET(opnd, Class); - for (Class = '9' + 1; Class < 256; Class++) - ANYOF_SET(opnd, Class); + if (value == 'p') + Perl_sv_catpvf(aTHX_ listsv, + "+utf8::%.*s\n", n, PL_regcomp_parse); + else + Perl_sv_catpvf(aTHX_ listsv, + "!utf8::%.*s\n", n, PL_regcomp_parse); } - lastclass = 1234; + PL_regcomp_parse = e + 1; + lastvalue = OOB_UTF8; continue; - case 'n': - Class = '\n'; - break; - case 'r': - Class = '\r'; - break; - case 't': - Class = '\t'; - break; - case 'f': - Class = '\f'; - break; - case 'b': - Class = '\b'; - break; - case 'e': - Class = '\033'; - break; - case 'a': - Class = '\007'; - break; + case 'n': value = '\n'; break; + case 'r': value = '\r'; break; + case 't': value = '\t'; break; + case 'f': value = '\f'; break; + case 'b': value = '\b'; break; + case 'e': value = '\033'; break; + case 'a': value = '\007'; break; case 'x': - Class = scan_hex(regcomp_parse, 2, &numlen); - regcomp_parse += numlen; + if (*PL_regcomp_parse == '{') { + e = strchr(PL_regcomp_parse++, '}'); + if (!e) + FAIL("Missing right brace on \\x{}"); + value = scan_hex(PL_regcomp_parse, + e - PL_regcomp_parse, + &numlen); + PL_regcomp_parse = e + 1; + } + else { + value = scan_hex(PL_regcomp_parse, 2, &numlen); + PL_regcomp_parse += numlen; + } break; case 'c': - Class = UCHARAT(regcomp_parse++); - Class = toCTRL(Class); + value = UCHARAT(PL_regcomp_parse++); + value = toCTRL(value); break; case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': - Class = scan_oct(--regcomp_parse, 3, &numlen); - regcomp_parse += numlen; + value = scan_oct(--PL_regcomp_parse, 3, &numlen); + PL_regcomp_parse += numlen; break; } } - if (range) { - if (lastclass > Class) - FAIL("invalid [] range in regexp"); + if (namedclass > OOB_NAMEDCLASS) { + if (range) { /* a-\d, a-[:digit:] */ + if (!SIZE_ONLY) { + if (ckWARN(WARN_UNSAFE)) + Perl_warner(aTHX_ WARN_UNSAFE, + "/%.127s/: false [] range \"%*.*s\" in regexp", + PL_regprecomp, + PL_regcomp_parse - rangebegin, + PL_regcomp_parse - rangebegin, + rangebegin); + Perl_sv_catpvf(aTHX_ listsv, + /* 0x002D is Unicode for '-' */ + "%04"UVxf"\n002D\n", (UV)lastvalue); + } + range = 0; + } + if (!SIZE_ONLY) { + switch (namedclass) { + case ANYOF_ALNUM: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsWord\n"); break; + case ANYOF_NALNUM: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsWord\n"); break; + case ANYOF_ALNUMC: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsAlnum\n"); break; + case ANYOF_NALNUMC: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsAlnum\n"); break; + case ANYOF_ALPHA: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsAlpha\n"); break; + case ANYOF_NALPHA: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsAlpha\n"); break; + case ANYOF_ASCII: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsASCII\n"); break; + case ANYOF_NASCII: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsASCII\n"); break; + case ANYOF_CNTRL: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsCntrl\n"); break; + case ANYOF_NCNTRL: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsCntrl\n"); break; + case ANYOF_GRAPH: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsGraph\n"); break; + case ANYOF_NGRAPH: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsGraph\n"); break; + case ANYOF_DIGIT: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsDigit\n"); break; + case ANYOF_NDIGIT: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsDigit\n"); break; + case ANYOF_LOWER: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsLower\n"); break; + case ANYOF_NLOWER: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsLower\n"); break; + case ANYOF_PRINT: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsPrint\n"); break; + case ANYOF_NPRINT: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsPrint\n"); break; + case ANYOF_PUNCT: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsPunct\n"); break; + case ANYOF_NPUNCT: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsPunct\n"); break; + case ANYOF_SPACE: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsSpace\n"); break; + case ANYOF_NSPACE: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsSpace\n"); break; + case ANYOF_UPPER: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsUpper\n"); break; + case ANYOF_NUPPER: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsUpper\n"); break; + case ANYOF_XDIGIT: + Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsXDigit\n"); break; + case ANYOF_NXDIGIT: + Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsXDigit\n"); break; + } + continue; + } + } + if (range) { + if (lastvalue > value) { /* b-a */ + Perl_croak(aTHX_ + "/%.127s/: invalid [] range \"%*.*s\" in regexp", + PL_regprecomp, + PL_regcomp_parse - rangebegin, + PL_regcomp_parse - rangebegin, + rangebegin); + } range = 0; } else { - lastclass = Class; - if (*regcomp_parse == '-' && regcomp_parse+1 < regxend && - regcomp_parse[1] != ']') { - regcomp_parse++; - range = 1; + lastvalue = value; + if (*PL_regcomp_parse == '-' && PL_regcomp_parse+1 < PL_regxend && + PL_regcomp_parse[1] != ']') { + PL_regcomp_parse++; + if (namedclass > OOB_NAMEDCLASS) { /* \w-, [:word:]- */ + if (ckWARN(WARN_UNSAFE)) + Perl_warner(aTHX_ WARN_UNSAFE, + "/%.127s/: false [] range \"%*.*s\" in regexp", + PL_regprecomp, + PL_regcomp_parse - rangebegin, + PL_regcomp_parse - rangebegin, + rangebegin); + if (!SIZE_ONLY) + Perl_sv_catpvf(aTHX_ listsv, + /* 0x002D is Unicode for '-' */ + "002D\n"); + } else + range = 1; continue; /* do it next time */ } } - if (!SIZE_ONLY) { - for ( ; lastclass <= Class; lastclass++) - ANYOF_SET(opnd, lastclass); - } - lastclass = Class; - } - if (*regcomp_parse != ']') - FAIL("unmatched [] in regexp"); - nextchar(); - /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */ - if (!SIZE_ONLY && (*opnd & (0xFF ^ ANYOF_INVERT)) == ANYOF_FOLD) { - for (Class = 0; Class < 256; ++Class) { - if (ANYOF_TEST(opnd, Class)) { - I32 cf = fold[Class]; - ANYOF_SET(opnd, cf); - } - } - *opnd &= ~ANYOF_FOLD; + /* now is the next time */ + if (!SIZE_ONLY) + Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\t%04"UVxf"\n", + (UV)lastvalue, (UV)value); + range = 0; } - /* optimize inverted simple patterns (e.g. [^a-z]) */ - if (!SIZE_ONLY && (*opnd & 0xFF) == ANYOF_INVERT) { - for (Class = 0; Class < 32; ++Class) - opnd[1 + Class] ^= 0xFF; - *opnd = 0; + + ret = reganode(ANYOFUTF8, 0); + + if (!SIZE_ONLY) { + SV *rv = swash_init("utf8", "", listsv, 1, 0); + SvREFCNT_dec(listsv); + n = add_data(1,"s"); + PL_regcomp_rx->data->data[n] = (void*)rv; + ARG1_SET(ret, flags); + ARG2_SET(ret, n); } + return ret; } STATIC char* -nextchar(void) +S_nextchar(pTHX) { - char* retval = regcomp_parse++; + dTHR; + char* retval = PL_regcomp_parse++; for (;;) { - if (*regcomp_parse == '(' && regcomp_parse[1] == '?' && - regcomp_parse[2] == '#') { - while (*regcomp_parse && *regcomp_parse != ')') - regcomp_parse++; - regcomp_parse++; + if (*PL_regcomp_parse == '(' && PL_regcomp_parse[1] == '?' && + PL_regcomp_parse[2] == '#') { + while (*PL_regcomp_parse && *PL_regcomp_parse != ')') + PL_regcomp_parse++; + PL_regcomp_parse++; continue; } - if (regflags & PMf_EXTENDED) { - if (isSPACE(*regcomp_parse)) { - regcomp_parse++; + if (PL_regflags & PMf_EXTENDED) { + if (isSPACE(*PL_regcomp_parse)) { + PL_regcomp_parse++; continue; } - else if (*regcomp_parse == '#') { - while (*regcomp_parse && *regcomp_parse != '\n') - regcomp_parse++; - regcomp_parse++; + else if (*PL_regcomp_parse == '#') { + while (*PL_regcomp_parse && *PL_regcomp_parse != '\n') + PL_regcomp_parse++; + PL_regcomp_parse++; continue; } } @@ -2035,22 +2972,23 @@ nextchar(void) - reg_node - emit a node */ STATIC regnode * /* Location. */ -reg_node(U8 op) +S_reg_node(pTHX_ U8 op) { + dTHR; register regnode *ret; register regnode *ptr; - ret = regcode; + ret = PL_regcode; if (SIZE_ONLY) { - SIZE_ALIGN(regsize); - regsize += 1; + SIZE_ALIGN(PL_regsize); + PL_regsize += 1; return(ret); } NODE_ALIGN_FILL(ret); ptr = ret; FILL_ADVANCE_NODE(ptr, op); - regcode = ptr; + PL_regcode = ptr; return(ret); } @@ -2059,34 +2997,41 @@ reg_node(U8 op) - reganode - emit a node with an argument */ STATIC regnode * /* Location. */ -reganode(U8 op, U32 arg) +S_reganode(pTHX_ U8 op, U32 arg) { + dTHR; register regnode *ret; register regnode *ptr; - ret = regcode; + ret = PL_regcode; if (SIZE_ONLY) { - SIZE_ALIGN(regsize); - regsize += 2; + SIZE_ALIGN(PL_regsize); + PL_regsize += 2; return(ret); } NODE_ALIGN_FILL(ret); ptr = ret; FILL_ADVANCE_NODE_ARG(ptr, op, arg); - regcode = ptr; + PL_regcode = ptr; return(ret); } /* -- regc - emit (if appropriate) a byte of code +- reguni - emit (if appropriate) a Unicode character */ STATIC void -regc(U8 b, char* s) +S_reguni(pTHX_ UV uv, char* s, I32* lenp) { - if (!SIZE_ONLY) - *s = b; + dTHR; + if (SIZE_ONLY) { + U8 tmpbuf[10]; + *lenp = uv_to_utf8(tmpbuf, uv) - tmpbuf; + } + else + *lenp = uv_to_utf8((U8*)s, uv) - (U8*)s; + } /* @@ -2095,23 +3040,24 @@ regc(U8 b, char* s) * Means relocating the operand. */ STATIC void -reginsert(U8 op, regnode *opnd) +S_reginsert(pTHX_ U8 op, regnode *opnd) { + dTHR; register regnode *src; register regnode *dst; register regnode *place; register int offset = regarglen[(U8)op]; -/* (regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */ +/* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */ if (SIZE_ONLY) { - regsize += NODE_STEP_REGNODE + offset; + PL_regsize += NODE_STEP_REGNODE + offset; return; } - src = regcode; - regcode += NODE_STEP_REGNODE + offset; - dst = regcode; + src = PL_regcode; + PL_regcode += NODE_STEP_REGNODE + offset; + dst = PL_regcode; while (src > opnd) StructCopy(--src, --dst, regnode); @@ -2125,8 +3071,9 @@ reginsert(U8 op, regnode *opnd) - regtail - set the next-pointer at the end of a node chain of p to val. */ STATIC void -regtail(regnode *p, regnode *val) +S_regtail(pTHX_ regnode *p, regnode *val) { + dTHR; register regnode *scan; register regnode *temp; register I32 offset; @@ -2145,7 +3092,8 @@ regtail(regnode *p, regnode *val) if (reg_off_by_arg[OP(scan)]) { ARG_SET(scan, val - scan); - } else { + } + else { NEXT_OFF(scan) = val - scan; } } @@ -2154,16 +3102,19 @@ regtail(regnode *p, regnode *val) - regoptail - regtail on operand of first argument; nop if operandless */ STATIC void -regoptail(regnode *p, regnode *val) +S_regoptail(pTHX_ regnode *p, regnode *val) { + dTHR; /* "Operandless" and "op != BRANCH" are synonymous in practice. */ if (p == NULL || SIZE_ONLY) return; - if (regkind[(U8)OP(p)] == BRANCH) { + if (PL_regkind[(U8)OP(p)] == BRANCH) { regtail(NEXTOPER(p), val); - } else if ( regkind[(U8)OP(p)] == BRANCHJ) { + } + else if ( PL_regkind[(U8)OP(p)] == BRANCHJ) { regtail(NEXTOPER(NEXTOPER(p)), val); - } else + } + else return; } @@ -2171,7 +3122,7 @@ regoptail(regnode *p, regnode *val) - regcurly - a little FSA that accepts {\d+,?\d*} */ STATIC I32 -regcurly(register char *s) +S_regcurly(pTHX_ register char *s) { if (*s++ != '{') return FALSE; @@ -2190,10 +3141,10 @@ regcurly(register char *s) STATIC regnode * -dumpuntil(regnode *start, regnode *node, regnode *last, SV* sv, I32 l) +S_dumpuntil(pTHX_ regnode *start, regnode *node, regnode *last, SV* sv, I32 l) { #ifdef DEBUGGING - register char op = EXACT; /* Arbitrary non-END op. */ + register U8 op = EXACT; /* Arbitrary non-END op. */ register regnode *next, *onode; while (op != END && (!last || node < last)) { @@ -2208,7 +3159,7 @@ dumpuntil(regnode *start, regnode *node, regnode *last, SV* sv, I32 l) if (OP(node) == OPTIMIZED) goto after_print; regprop(sv, node); - PerlIO_printf(Perl_debug_log, "%4d%*s%s", node - start, + PerlIO_printf(Perl_debug_log, "%4d:%*s%s", node - start, 2*l + 1, "", SvPVX(sv)); if (next == NULL) /* Next ptr. */ PerlIO_printf(Perl_debug_log, "(0)"); @@ -2216,31 +3167,38 @@ dumpuntil(regnode *start, regnode *node, regnode *last, SV* sv, I32 l) PerlIO_printf(Perl_debug_log, "(%d)", next - start); (void)PerlIO_putc(Perl_debug_log, '\n'); after_print: - if (regkind[(U8)op] == BRANCHJ) { + if (PL_regkind[(U8)op] == BRANCHJ) { register regnode *nnode = (OP(next) == LONGJMP ? regnext(next) : next); if (last && nnode > last) nnode = last; node = dumpuntil(start, NEXTOPER(NEXTOPER(node)), nnode, sv, l + 1); - } else if (regkind[(U8)op] == BRANCH) { + } + else if (PL_regkind[(U8)op] == BRANCH) { node = dumpuntil(start, NEXTOPER(node), next, sv, l + 1); - } else if ( op == CURLY) { /* `next' might be very big: optimizer */ + } + else if ( op == CURLY) { /* `next' might be very big: optimizer */ node = dumpuntil(start, NEXTOPER(node) + EXTRA_STEP_2ARGS, NEXTOPER(node) + EXTRA_STEP_2ARGS + 1, sv, l + 1); - } else if (regkind[(U8)op] == CURLY && op != CURLYX) { + } + else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) { node = dumpuntil(start, NEXTOPER(node) + EXTRA_STEP_2ARGS, next, sv, l + 1); - } else if ( op == PLUS || op == STAR) { + } + else if ( op == PLUS || op == STAR) { node = dumpuntil(start, NEXTOPER(node), NEXTOPER(node) + 1, sv, l + 1); - } else if (op == ANYOF) { + } + else if (op == ANYOF) { node = NEXTOPER(node); node += ANY_SKIP; - } else if (regkind[(U8)op] == EXACT) { + } + else if (PL_regkind[(U8)op] == EXACT) { /* Literal string, where present. */ - node += ((*OPERAND(node)) + 2 + sizeof(regnode) - 1) / sizeof(regnode); + node += NODE_SZ_STR(node) - 1; node = NEXTOPER(node); - } else { + } + else { node = NEXTOPER(node); node += regarglen[(U8)op]; } @@ -2257,26 +3215,29 @@ dumpuntil(regnode *start, regnode *node, regnode *last, SV* sv, I32 l) - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form */ void -regdump(regexp *r) +Perl_regdump(pTHX_ regexp *r) { #ifdef DEBUGGING + dTHR; SV *sv = sv_newmortal(); (void)dumpuntil(r->program, r->program + 1, NULL, sv, 0); /* Header fields of interest. */ if (r->anchored_substr) - PerlIO_printf(Perl_debug_log, "anchored `%s%s%s'%s at %d ", - colors[0], + PerlIO_printf(Perl_debug_log, "anchored `%s%.*s%s'%s at %d ", + PL_colors[0], + SvCUR(r->anchored_substr) - (SvTAIL(r->anchored_substr)!=0), SvPVX(r->anchored_substr), - colors[1], + PL_colors[1], SvTAIL(r->anchored_substr) ? "$" : "", r->anchored_offset); if (r->float_substr) - PerlIO_printf(Perl_debug_log, "floating `%s%s%s'%s at %d..%u ", - colors[0], - SvPVX(r->float_substr), - colors[1], + PerlIO_printf(Perl_debug_log, "floating `%s%.*s%s'%s at %d..%u ", + PL_colors[0], + SvCUR(r->float_substr) - (SvTAIL(r->float_substr)!=0), + SvPVX(r->float_substr), + PL_colors[1], SvTAIL(r->float_substr) ? "$" : "", r->float_min_offset, r->float_max_offset); if (r->check_substr) @@ -2300,6 +3261,8 @@ regdump(regexp *r) PerlIO_printf(Perl_debug_log, "(BOL)"); if (r->reganch & ROPT_ANCH_MBOL) PerlIO_printf(Perl_debug_log, "(MBOL)"); + if (r->reganch & ROPT_ANCH_SBOL) + PerlIO_printf(Perl_debug_log, "(SBOL)"); if (r->reganch & ROPT_ANCH_GPOS) PerlIO_printf(Perl_debug_log, "(GPOS)"); PerlIO_putc(Perl_debug_log, ' '); @@ -2321,205 +3284,76 @@ regdump(regexp *r) - regprop - printable representation of opcode */ void -regprop(SV *sv, regnode *o) +Perl_regprop(pTHX_ SV *sv, regnode *o) { #ifdef DEBUGGING - register char *p = 0; + dTHR; + register int k; - sv_setpv(sv, ":"); - switch (OP(o)) { - case BOL: - p = "BOL"; - break; - case MBOL: - p = "MBOL"; - break; - case SBOL: - p = "SBOL"; - break; - case EOL: - p = "EOL"; - break; - case EOS: - p = "EOS"; - break; - case MEOL: - p = "MEOL"; - break; - case SEOL: - p = "SEOL"; - break; - case ANY: - p = "ANY"; - break; - case SANY: - p = "SANY"; - break; - case ANYOF: - p = "ANYOF"; - break; - case BRANCH: - p = "BRANCH"; - break; - case EXACT: - sv_catpvf(sv, "EXACT <%s%s%s>", colors[0], OPERAND(o) + 1, colors[1]); - break; - case EXACTF: - sv_catpvf(sv, "EXACTF <%s%s%s>", colors[0], OPERAND(o) + 1, colors[1]); - break; - case EXACTFL: - sv_catpvf(sv, "EXACTFL <%s%s%s>", colors[0], OPERAND(o) + 1, colors[1]); - break; - case NOTHING: - p = "NOTHING"; - break; - case TAIL: - p = "TAIL"; - break; - case BACK: - p = "BACK"; - break; - case END: - p = "END"; - break; - case BOUND: - p = "BOUND"; - break; - case BOUNDL: - p = "BOUNDL"; - break; - case NBOUND: - p = "NBOUND"; - break; - case NBOUNDL: - p = "NBOUNDL"; - break; - case CURLY: - sv_catpvf(sv, "CURLY {%d,%d}", ARG1(o), ARG2(o)); - break; - case CURLYM: - sv_catpvf(sv, "CURLYM[%d] {%d,%d}", o->flags, ARG1(o), ARG2(o)); - break; - case CURLYN: - sv_catpvf(sv, "CURLYN[%d] {%d,%d}", o->flags, ARG1(o), ARG2(o)); - break; - case CURLYX: - sv_catpvf(sv, "CURLYX {%d,%d}", ARG1(o), ARG2(o)); - break; - case REF: - sv_catpvf(sv, "REF%d", ARG(o)); - break; - case REFF: - sv_catpvf(sv, "REFF%d", ARG(o)); - break; - case REFFL: - sv_catpvf(sv, "REFFL%d", ARG(o)); - break; - case OPEN: - sv_catpvf(sv, "OPEN%d", ARG(o)); - break; - case CLOSE: - sv_catpvf(sv, "CLOSE%d", ARG(o)); - p = NULL; - break; - case STAR: - p = "STAR"; - break; - case PLUS: - p = "PLUS"; - break; - case MINMOD: - p = "MINMOD"; - break; - case GPOS: - p = "GPOS"; - break; - case UNLESSM: - sv_catpvf(sv, "UNLESSM[-%d]", o->flags); - break; - case IFMATCH: - sv_catpvf(sv, "IFMATCH[-%d]", o->flags); - break; - case SUCCEED: - p = "SUCCEED"; - break; - case WHILEM: - p = "WHILEM"; - break; - case DIGIT: - p = "DIGIT"; - break; - case NDIGIT: - p = "NDIGIT"; - break; - case ALNUM: - p = "ALNUM"; - break; - case NALNUM: - p = "NALNUM"; - break; - case SPACE: - p = "SPACE"; - break; - case NSPACE: - p = "NSPACE"; - break; - case ALNUML: - p = "ALNUML"; - break; - case NALNUML: - p = "NALNUML"; - break; - case SPACEL: - p = "SPACEL"; - break; - case NSPACEL: - p = "NSPACEL"; - break; - case EVAL: - p = "EVAL"; - break; - case LONGJMP: - p = "LONGJMP"; - break; - case BRANCHJ: - p = "BRANCHJ"; - break; - case IFTHEN: - p = "IFTHEN"; - break; - case GROUPP: - sv_catpvf(sv, "GROUPP%d", ARG(o)); - break; - case LOGICAL: - p = "LOGICAL"; - break; - case SUSPEND: - p = "SUSPEND"; - break; - case RENUM: - p = "RENUM"; - break; - case OPTIMIZED: - p = "OPTIMIZED"; - break; - default: + sv_setpvn(sv, "", 0); + if (OP(o) >= reg_num) /* regnode.type is unsigned */ FAIL("corrupted regexp opcode"); + sv_catpv(sv, (char*)reg_name[OP(o)]); /* Take off const! */ + + k = PL_regkind[(U8)OP(o)]; + + if (k == EXACT) + Perl_sv_catpvf(aTHX_ sv, " <%s%.*s%s>", PL_colors[0], + STR_LEN(o), STRING(o), PL_colors[1]); + else if (k == CURLY) { + if (OP(o) == CURLYM || OP(o) == CURLYN) + Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */ + Perl_sv_catpvf(aTHX_ sv, " {%d,%d}", ARG1(o), ARG2(o)); } - if (p) - sv_catpv(sv, p); + else if (k == WHILEM && o->flags) /* Ordinal/of */ + Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4); + else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP ) + Perl_sv_catpvf(aTHX_ sv, "%d", ARG(o)); /* Parenth number */ + else if (k == LOGICAL) + Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* 2: embedded, otherwise 1 */ + else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH)) + Perl_sv_catpvf(aTHX_ sv, "[-%d]", o->flags); #endif /* DEBUGGING */ } +SV * +Perl_re_intuit_string(pTHX_ regexp *prog) +{ /* Assume that RE_INTUIT is set */ + DEBUG_r( + { STRLEN n_a; + char *s = SvPV(prog->check_substr,n_a); + + if (!PL_colorset) reginitcolors(); + PerlIO_printf(Perl_debug_log, + "%sUsing REx substr:%s `%s%.60s%s%s'\n", + PL_colors[4],PL_colors[5],PL_colors[0], + s, + PL_colors[1], + (strlen(s) > 60 ? "..." : "")); + } ); + + return prog->check_substr; +} + void -pregfree(struct regexp *r) +Perl_pregfree(pTHX_ struct regexp *r) { + dTHR; + DEBUG_r(if (!PL_colorset) reginitcolors()); + if (!r || (--r->refcnt > 0)) return; + DEBUG_r(PerlIO_printf(Perl_debug_log, + "%sFreeing REx:%s `%s%.60s%s%s'\n", + PL_colors[4],PL_colors[5],PL_colors[0], + r->precomp, + PL_colors[1], + (strlen(r->precomp) > 60 ? "..." : ""))); + if (r->precomp) Safefree(r->precomp); - if (r->subbase) - Safefree(r->subbase); + if (RX_MATCH_COPIED(r)) + Safefree(r->subbeg); if (r->substrs) { if (r->anchored_substr) SvREFCNT_dec(r->anchored_substr); @@ -2529,13 +3363,30 @@ pregfree(struct regexp *r) } if (r->data) { int n = r->data->count; + AV* new_comppad = NULL; + AV* old_comppad; + SV** old_curpad; + while (--n >= 0) { switch (r->data->what[n]) { case 's': SvREFCNT_dec((SV*)r->data->data[n]); break; + case 'p': + new_comppad = (AV*)r->data->data[n]; + break; case 'o': + if (new_comppad == NULL) + Perl_croak(aTHX_ "panic: pregfree comppad"); + old_comppad = PL_comppad; + old_curpad = PL_curpad; + PL_comppad = new_comppad; + PL_curpad = AvARRAY(new_comppad); op_free((OP_4tree*)r->data->data[n]); + PL_comppad = old_comppad; + PL_curpad = old_curpad; + SvREFCNT_dec((SV*)new_comppad); + new_comppad = NULL; break; case 'n': break; @@ -2558,11 +3409,12 @@ pregfree(struct regexp *r) * that bypass this code for speed.] */ regnode * -regnext(register regnode *p) +Perl_regnext(pTHX_ register regnode *p) { + dTHR; register I32 offset; - if (p == ®dummy) + if (p == &PL_regdummy) return(NULL); offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p)); @@ -2573,12 +3425,13 @@ regnext(register regnode *p) } STATIC void -re_croak2(const char* pat1,const char* pat2,...) +S_re_croak2(pTHX_ const char* pat1,const char* pat2,...) { va_list args; STRLEN l1 = strlen(pat1); STRLEN l2 = strlen(pat2); char buf[512]; + SV *msv; char *message; if (l1 > 510) @@ -2589,13 +3442,84 @@ re_croak2(const char* pat1,const char* pat2,...) Copy(pat2, buf + l1, l2 , char); buf[l1 + l2] = '\n'; buf[l1 + l2 + 1] = '\0'; +#ifdef I_STDARG + /* ANSI variant takes additional second argument */ va_start(args, pat2); - message = mess(buf, &args); +#else + va_start(args); +#endif + msv = vmess(buf, &args); va_end(args); - l1 = strlen(message); + message = SvPV(msv,l1); if (l1 > 512) l1 = 512; Copy(message, buf, l1 , char); buf[l1] = '\0'; /* Overwrite \n */ - croak("%s", buf); + Perl_croak(aTHX_ "%s", buf); } + +/* XXX Here's a total kludge. But we need to re-enter for swash routines. */ + +void +Perl_save_re_context(pTHX) +{ + dTHR; + SAVEPPTR(PL_bostr); + SAVEPPTR(PL_regprecomp); /* uncompiled string. */ + SAVEI32(PL_regnpar); /* () count. */ + SAVEI32(PL_regsize); /* Code size. */ + SAVEI16(PL_regflags); /* are we folding, multilining? */ + SAVEPPTR(PL_reginput); /* String-input pointer. */ + SAVEPPTR(PL_regbol); /* Beginning of input, for ^ check. */ + SAVEPPTR(PL_regeol); /* End of input, for $ check. */ + SAVESPTR(PL_regstartp); /* Pointer to startp array. */ + SAVESPTR(PL_regendp); /* Ditto for endp. */ + SAVESPTR(PL_reglastparen); /* Similarly for lastparen. */ + SAVEPPTR(PL_regtill); /* How far we are required to go. */ + SAVEI32(PL_regprev); /* char before regbol, \n if none */ + SAVESPTR(PL_reg_start_tmp); /* from regexec.c */ + PL_reg_start_tmp = 0; + SAVEFREEPV(PL_reg_start_tmp); + SAVEI32(PL_reg_start_tmpl); /* from regexec.c */ + PL_reg_start_tmpl = 0; + SAVESPTR(PL_regdata); + SAVEI32(PL_reg_flags); /* from regexec.c */ + SAVEI32(PL_reg_eval_set); /* from regexec.c */ + SAVEI32(PL_regnarrate); /* from regexec.c */ + SAVESPTR(PL_regprogram); /* from regexec.c */ + SAVEINT(PL_regindent); /* from regexec.c */ + SAVESPTR(PL_regcc); /* from regexec.c */ + SAVESPTR(PL_curcop); + SAVESPTR(PL_regcomp_rx); /* from regcomp.c */ + SAVEI32(PL_regseen); /* from regcomp.c */ + SAVEI32(PL_regsawback); /* Did we see \1, ...? */ + SAVEI32(PL_regnaughty); /* How bad is this pattern? */ + SAVESPTR(PL_regcode); /* Code-emit pointer; ®dummy = don't */ + SAVEPPTR(PL_regxend); /* End of input for compile */ + SAVEPPTR(PL_regcomp_parse); /* Input-scan pointer. */ + SAVESPTR(PL_reg_call_cc); /* from regexec.c */ + SAVESPTR(PL_reg_re); /* from regexec.c */ + SAVEPPTR(PL_reg_ganch); /* from regexec.c */ + SAVESPTR(PL_reg_sv); /* from regexec.c */ + SAVESPTR(PL_reg_magic); /* from regexec.c */ + SAVEI32(PL_reg_oldpos); /* from regexec.c */ + SAVESPTR(PL_reg_oldcurpm); /* from regexec.c */ + SAVESPTR(PL_reg_curpm); /* from regexec.c */ +#ifdef DEBUGGING + SAVEPPTR(PL_reg_starttry); /* from regexec.c */ +#endif +} + +#ifdef PERL_OBJECT +#define NO_XSLOCKS +#include "XSUB.h" +#undef this +#define this pPerl +#endif + +static void +clear_re(pTHXo_ void *r) +{ + ReREFCNT_dec((regexp *)r); +} +