* 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
*
**** Alterations to Henry's code are...
****
- **** Copyright (c) 1991-1994, 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.
* 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"
+#ifdef op
+#undef op
+#endif /* op */
+
#ifdef MSDOS
# if defined(BUGGY_MSC6)
/* MSC 6.00A breaks on op/regexp.t test 85 unless we turn this off */
* Flags to be passed up and down.
*/
#define WORST 0 /* Worst case. */
-#define HASWIDTH 0x1 /* Known never to match null string. */
+#define HASWIDTH 0x1 /* Known to match non-null strings. */
#define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */
#define SPSTART 0x4 /* Starts with * or +. */
#define TRYAGAIN 0x8 /* Weeded out a declaration. */
+/* Length of a variant. */
+
+typedef struct scan_data_t {
+ I32 len_min;
+ I32 len_delta;
+ I32 pos_min;
+ I32 pos_delta;
+ SV *last_found;
+ I32 last_end; /* min value, <0 unless valid. */
+ I32 last_start_min;
+ I32 last_start_max;
+ SV **longest; /* Either &l_fixed, or &l_float. */
+ SV *longest_fixed;
+ I32 offset_fixed;
+ SV *longest_float;
+ I32 offset_float_min;
+ I32 offset_float_max;
+ I32 flags;
+ I32 whilem_c;
+ struct regnode_charclass_class *start_class;
+} scan_data_t;
+
/*
* Forward declarations for pregcomp()'s friends.
*/
-static char *reg _((I32, I32 *));
-static char *reganode _((char, unsigned short));
-static char *regatom _((I32 *));
-static char *regbranch _((I32 *));
-static void regc _((char));
-static char *regclass _((void));
-STATIC I32 regcurly _((char *));
-static char *regnode _((char));
-static char *regpiece _((I32 *));
-static void reginsert _((char, char *));
-static void regoptail _((char *, char *));
-static void regset _((char *, I32, I32));
-static void regtail _((char *, char *));
-static char* nextchar _((void));
+static scan_data_t zero_scan_data = { 0, 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
+#define SF_BEFORE_MEOL 0x2
+#define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
+#define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
+
+#ifdef NO_UNARY_PLUS
+# define SF_FIX_SHIFT_EOL (0+2)
+# define SF_FL_SHIFT_EOL (0+4)
+#else
+# define SF_FIX_SHIFT_EOL (+2)
+# define SF_FL_SHIFT_EOL (+4)
+#endif
+
+#define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
+#define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
+
+#define SF_FL_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL)
+#define SF_FL_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */
+#define SF_IS_INF 0x40
+#define SF_HAS_PAR 0x80
+#define SF_IN_PAR 0x100
+#define SF_HAS_EVAL 0x200
+#define SCF_DO_SUBSTR 0x400
+#define SCF_DO_STCLASS_AND 0x0800
+#define SCF_DO_STCLASS_OR 0x1000
+#define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
+
+#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);
+
+/* Mark that we cannot extend a found fixed substring at this point.
+ Updata the longest found anchored substring and the longest found
+ floating substrings if needed. */
+
+STATIC void
+S_scan_commit(pTHX_ scan_data_t *data)
+{
+ 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);
+ if (*data->longest == data->longest_fixed) {
+ data->offset_fixed = l ? data->last_start_min : data->pos_min;
+ if (data->flags & SF_BEFORE_EOL)
+ data->flags
+ |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL);
+ else
+ data->flags &= ~SF_FIX_BEFORE_EOL;
+ }
+ else {
+ data->offset_float_min = l ? data->last_start_min : data->pos_min;
+ data->offset_float_max = (l
+ ? data->last_start_max
+ : data->pos_min + data->pos_delta);
+ if (data->flags & SF_BEFORE_EOL)
+ data->flags
+ |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL);
+ else
+ data->flags &= ~SF_FL_BEFORE_EOL;
+ }
+ }
+ SvCUR_set(data->last_found, 0);
+ data->last_end = -1;
+ data->flags &= ~SF_BEFORE_EOL;
+}
+
+/* Can match anything (initialization) */
+STATIC void
+S_cl_anything(pTHX_ struct regnode_charclass_class *cl)
+{
+ int value;
+
+ ANYOF_CLASS_ZERO(cl);
+ for (value = 0; value < 256; ++value)
+ ANYOF_BITMAP_SET(cl, value);
+ cl->flags = ANYOF_EOS;
+ if (LOC)
+ cl->flags |= ANYOF_LOCALE;
+}
+
+/* Can match anything (initialization) */
+STATIC int
+S_cl_is_anything(pTHX_ struct regnode_charclass_class *cl)
+{
+ int value;
+
+ for (value = 0; value < ANYOF_MAX; value += 2)
+ if (ANYOF_CLASS_TEST(cl, value) && ANYOF_CLASS_TEST(cl, value + 1))
+ return 1;
+ for (value = 0; value < 256; ++value)
+ if (!ANYOF_BITMAP_TEST(cl, value))
+ return 0;
+ return 1;
+}
+
+/* Can match anything (initialization) */
+STATIC void
+S_cl_init(pTHX_ struct regnode_charclass_class *cl)
+{
+ cl->type = ANYOF;
+ cl_anything(cl);
+}
+
+STATIC void
+S_cl_init_zero(pTHX_ struct regnode_charclass_class *cl)
+{
+ cl->type = ANYOF;
+ cl_anything(cl);
+ ANYOF_CLASS_ZERO(cl);
+ ANYOF_BITMAP_ZERO(cl);
+ if (LOC)
+ cl->flags |= ANYOF_LOCALE;
+}
+
+/* 'And' a given class with another one. Can create false positives */
+/* We assume that cl is not inverted */
+STATIC void
+S_cl_and(pTHX_ struct regnode_charclass_class *cl,
+ struct regnode_charclass_class *and_with)
+{
+ int value;
+
+ if (!(and_with->flags & ANYOF_CLASS)
+ && !(cl->flags & ANYOF_CLASS)
+ && (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
+ && !(and_with->flags & ANYOF_FOLD)
+ && !(cl->flags & ANYOF_FOLD)) {
+ int i;
+
+ if (and_with->flags & ANYOF_INVERT)
+ for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
+ cl->bitmap[i] &= ~and_with->bitmap[i];
+ else
+ for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
+ cl->bitmap[i] &= and_with->bitmap[i];
+ } /* XXXX: logic is complicated otherwise, leave it along for a moment. */
+ if (!(and_with->flags & ANYOF_EOS))
+ cl->flags &= ~ANYOF_EOS;
+}
+
+/* 'OR' a given class with another one. Can create false positives */
+/* We assume that cl is not inverted */
+STATIC void
+S_cl_or(pTHX_ struct regnode_charclass_class *cl, struct regnode_charclass_class *or_with)
+{
+ int value;
+
+ if (or_with->flags & ANYOF_INVERT) {
+ /* We do not use
+ * (B1 | CL1) | (!B2 & !CL2) = (B1 | !B2 & !CL2) | (CL1 | (!B2 & !CL2))
+ * <= (B1 | !B2) | (CL1 | !CL2)
+ * which is wasteful if CL2 is small, but we ignore CL2:
+ * (B1 | CL1) | (!B2 & !CL2) <= (B1 | CL1) | !B2 = (B1 | !B2) | CL1
+ * XXXX Can we handle case-fold? Unclear:
+ * (OK1(i) | OK1(i')) | !(OK1(i) | OK1(i')) =
+ * (OK1(i) | OK1(i')) | (!OK1(i) & !OK1(i'))
+ */
+ if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
+ && !(or_with->flags & ANYOF_FOLD)
+ && !(cl->flags & ANYOF_FOLD) ) {
+ int i;
+
+ for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
+ cl->bitmap[i] |= ~or_with->bitmap[i];
+ } /* XXXX: logic is complicated otherwise */
+ else {
+ cl_anything(cl);
+ }
+ } else {
+ /* (B1 | CL1) | (B2 | CL2) = (B1 | B2) | (CL1 | CL2)) */
+ if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
+ && (!(or_with->flags & ANYOF_FOLD)
+ || (cl->flags & ANYOF_FOLD)) ) {
+ int i;
+
+ /* OR char bitmap and class bitmap separately */
+ for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
+ cl->bitmap[i] |= or_with->bitmap[i];
+ if (or_with->flags & ANYOF_CLASS) {
+ for (i = 0; i < ANYOF_CLASSBITMAP_SIZE; i++)
+ cl->classflags[i] |= or_with->classflags[i];
+ cl->flags |= ANYOF_CLASS;
+ }
+ }
+ else { /* XXXX: logic is complicated, leave it along for a moment. */
+ cl_anything(cl);
+ }
+ }
+ if (or_with->flags & ANYOF_EOS)
+ cl->flags |= ANYOF_EOS;
+}
+
+/* REx optimizer. Converts nodes into quickier variants "in place".
+ Finds fixed substrings. */
+
+/* Stops at toplevel WHILEM as well as at `last'. At end *scanp is set
+ to the position after last scanned or to NULL. */
+
+STATIC I32
+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;
+ struct regnode_charclass_class and_with; /* Valid if flags & SCF_DO_STCLASS_OR */
+
+ while (scan && OP(scan) != END && scan < last) {
+ /* Peephole optimizer: */
+
+ if (PL_regkind[(U8)OP(scan)] == EXACT) {
+ /* Merge several consecutive EXACTish nodes into one. */
+ regnode *n = regnext(scan);
+ U32 stringok = 1;
+#ifdef DEBUGGING
+ regnode *stop = scan;
+#endif
+
+ next = scan + NODE_SZ_STR(scan);
+ /* Skip NOTHING, merge EXACT*. */
+ while (n &&
+ ( 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 (PL_regkind[(U8)OP(n)] == NOTHING) {
+ NEXT_OFF(scan) += NEXT_OFF(n);
+ next = n + NODE_STEP_REGNODE;
+#ifdef DEBUGGING
+ if (stringok)
+ stop = n;
+#endif
+ n = regnext(n);
+ }
+ else {
+ int oldl = STR_LEN(scan);
+ regnode *nnext = regnext(n);
+
+ if (oldl + STR_LEN(n) > U8_MAX)
+ break;
+ NEXT_OFF(scan) += NEXT_OFF(n);
+ STR_LEN(scan) += STR_LEN(n);
+ next = n + NODE_SZ_STR(n);
+ /* Now we can overwrite *n : */
+ Move(STRING(n), STRING(scan) + oldl,
+ STR_LEN(n), char);
+#ifdef DEBUGGING
+ if (stringok)
+ stop = next - 1;
+#endif
+ n = nnext;
+ }
+ }
+#ifdef DEBUGGING
+ /* Allow dumping */
+ n = scan + NODE_SZ_STR(scan);
+ while (n <= stop) {
+ if (PL_regkind[(U8)OP(n)] != NOTHING || OP(n) == NOTHING) {
+ OP(n) = OPTIMIZED;
+ NEXT_OFF(n) = 0;
+ }
+ n++;
+ }
+#endif
+ }
+ /* Follow the next-chain of the current node and optimize
+ away all the NOTHINGs from it. */
+ if (OP(scan) != CURLYX) {
+ 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))
+ && ((PL_regkind[(U8)OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
+ || ((OP(n) == LONGJMP) && (noff = ARG(n))))
+ && off + noff < max)
+ off += noff;
+ if (reg_off_by_arg[OP(scan)])
+ ARG(scan) = off;
+ else
+ NEXT_OFF(scan) = off;
+ }
+ /* The principal pseudo-switch. Cannot be a switch, since we
+ look into several different things. */
+ if (OP(scan) == BRANCH || OP(scan) == BRANCHJ
+ || OP(scan) == IFTHEN || OP(scan) == SUSPEND) {
+ next = regnext(scan);
+ code = OP(scan);
+
+ if (OP(next) == code || code == IFTHEN || code == SUSPEND) {
+ I32 max1 = 0, min1 = I32_MAX, num = 0;
+ struct regnode_charclass_class accum;
+
+ if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */
+ scan_commit(data); /* Cannot merge strings after this. */
+ if (flags & SCF_DO_STCLASS)
+ cl_init_zero(&accum);
+ while (OP(scan) == code) {
+ I32 deltanext, minnext, f = 0;
+ struct regnode_charclass_class this_class;
+
+ num++;
+ data_fake.flags = 0;
+ if (data)
+ data_fake.whilem_c = data->whilem_c;
+ next = regnext(scan);
+ scan = NEXTOPER(scan);
+ if (code != BRANCH)
+ scan = NEXTOPER(scan);
+ if (flags & SCF_DO_STCLASS) {
+ cl_init(&this_class);
+ data_fake.start_class = &this_class;
+ f = SCF_DO_STCLASS_AND;
+ }
+ /* we suppose the run is continuous, last=next...*/
+ minnext = study_chunk(&scan, &deltanext, next,
+ &data_fake, f);
+ if (min1 > minnext)
+ min1 = minnext;
+ if (max1 < minnext + deltanext)
+ max1 = minnext + deltanext;
+ if (deltanext == I32_MAX)
+ 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 (flags & SCF_DO_STCLASS)
+ cl_or(&accum, &this_class);
+ if (code == SUSPEND)
+ break;
+ }
+ if (code == IFTHEN && num < 2) /* Empty ELSE branch */
+ min1 = 0;
+ if (flags & SCF_DO_SUBSTR) {
+ data->pos_min += min1;
+ data->pos_delta += max1 - min1;
+ if (max1 != min1 || is_inf)
+ data->longest = &(data->longest_float);
+ }
+ min += min1;
+ delta += max1 - min1;
+ if (flags & SCF_DO_STCLASS_OR) {
+ cl_or(data->start_class, &accum);
+ if (min1) {
+ cl_and(data->start_class, &and_with);
+ flags &= ~SCF_DO_STCLASS;
+ }
+ }
+ else if (flags & SCF_DO_STCLASS_AND) {
+ if (min1) {
+ cl_and(data->start_class, &accum);
+ flags &= ~SCF_DO_STCLASS;
+ }
+ else {
+ /* Switch to OR mode: cache the old value of
+ * data->start_class */
+ StructCopy(data->start_class, &and_with,
+ struct regnode_charclass_class);
+ flags &= ~SCF_DO_STCLASS_AND;
+ StructCopy(&accum, data->start_class,
+ struct regnode_charclass_class);
+ flags |= SCF_DO_STCLASS_OR;
+ data->start_class->flags |= ANYOF_EOS;
+ }
+ }
+ }
+ 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) {
+ 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. */
+ /* The code below prefers earlier match for fixed
+ offset, later match for variable offset. */
+ if (data->last_end == -1) { /* Update the start info. */
+ data->last_start_min = data->pos_min;
+ data->last_start_max = is_inf
+ ? I32_MAX : data->pos_min + data->pos_delta;
+ }
+ 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;
+ }
+ if (flags & SCF_DO_STCLASS_AND) {
+ /* Check whether it is compatible with what we know already! */
+ int compat = 1;
+
+ if (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
+ && !ANYOF_BITMAP_TEST(data->start_class, *STRING(scan))
+ && (!(data->start_class->flags & ANYOF_FOLD)
+ || !ANYOF_BITMAP_TEST(data->start_class,
+ PL_fold[*STRING(scan)])))
+ compat = 0;
+ ANYOF_CLASS_ZERO(data->start_class);
+ ANYOF_BITMAP_ZERO(data->start_class);
+ if (compat)
+ ANYOF_BITMAP_SET(data->start_class, *STRING(scan));
+ data->start_class->flags &= ~ANYOF_EOS;
+ }
+ else if (flags & SCF_DO_STCLASS_OR) {
+ /* false positive possible if the class is case-folded */
+ ANYOF_BITMAP_SET(data->start_class, *STRING(scan));
+ data->start_class->flags &= ~ANYOF_EOS;
+ cl_and(data->start_class, &and_with);
+ }
+ flags &= ~SCF_DO_STCLASS;
+ }
+ else if (PL_regkind[(U8)OP(scan)] == EXACT) { /* But OP != EXACT! */
+ I32 l = STR_LEN(scan);
+
+ /* Search for fixed substrings supports EXACT only. */
+ if (flags & SCF_DO_SUBSTR)
+ scan_commit(data);
+ 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 += l;
+ if (flags & SCF_DO_STCLASS_AND) {
+ /* Check whether it is compatible with what we know already! */
+ int compat = 1;
+
+ if (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
+ && !ANYOF_BITMAP_TEST(data->start_class, *STRING(scan))
+ && !ANYOF_BITMAP_TEST(data->start_class,
+ PL_fold[*STRING(scan)]))
+ compat = 0;
+ ANYOF_CLASS_ZERO(data->start_class);
+ ANYOF_BITMAP_ZERO(data->start_class);
+ if (compat) {
+ ANYOF_BITMAP_SET(data->start_class, *STRING(scan));
+ data->start_class->flags &= ~ANYOF_EOS;
+ data->start_class->flags |= ANYOF_FOLD;
+ if (OP(scan) == EXACTFL)
+ data->start_class->flags |= ANYOF_LOCALE;
+ }
+ }
+ else if (flags & SCF_DO_STCLASS_OR) {
+ if (data->start_class->flags & ANYOF_FOLD) {
+ /* false positive possible if the class is case-folded.
+ Assume that the locale settings are the same... */
+ ANYOF_BITMAP_SET(data->start_class, *STRING(scan));
+ data->start_class->flags &= ~ANYOF_EOS;
+ }
+ cl_and(data->start_class, &and_with);
+ }
+ flags &= ~SCF_DO_STCLASS;
+ }
+ else if (strchr((char*)PL_varies,OP(scan))) {
+ I32 mincount, maxcount, minnext, deltanext, pos_before, fl;
+ I32 f = flags;
+ regnode *oscan = scan;
+ struct regnode_charclass_class this_class;
+ struct regnode_charclass_class *oclass = NULL;
+
+ switch (PL_regkind[(U8)OP(scan)]) {
+ case WHILEM: /* End of (?:...)* . */
+ scan = NEXTOPER(scan);
+ goto finish;
+ case PLUS:
+ if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
+ next = NEXTOPER(scan);
+ if (OP(next) == EXACT || (flags & SCF_DO_STCLASS)) {
+ mincount = 1;
+ maxcount = REG_INFTY;
+ next = regnext(scan);
+ scan = NEXTOPER(scan);
+ goto do_curly;
+ }
+ }
+ if (flags & SCF_DO_SUBSTR)
+ data->pos_min++;
+ min++;
+ /* Fall through. */
+ case STAR:
+ if (flags & SCF_DO_STCLASS) {
+ mincount = 0;
+ maxcount = REG_INFTY;
+ next = regnext(scan);
+ scan = NEXTOPER(scan);
+ goto do_curly;
+ }
+ is_inf = is_inf_internal = 1;
+ scan = regnext(scan);
+ if (flags & SCF_DO_SUBSTR) {
+ scan_commit(data); /* Cannot extend fixed substrings */
+ data->longest = &(data->longest_float);
+ }
+ goto optimize_curly_tail;
+ case CURLY:
+ mincount = ARG1(scan);
+ maxcount = ARG2(scan);
+ next = regnext(scan);
+ scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
+ do_curly:
+ if (flags & SCF_DO_SUBSTR) {
+ if (mincount == 0) scan_commit(data); /* Cannot extend fixed substrings */
+ pos_before = data->pos_min;
+ }
+ if (data) {
+ fl = data->flags;
+ data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
+ if (is_inf)
+ data->flags |= SF_IS_INF;
+ }
+ if (flags & SCF_DO_STCLASS) {
+ cl_init(&this_class);
+ oclass = data->start_class;
+ data->start_class = &this_class;
+ f |= SCF_DO_STCLASS_AND;
+ f &= ~SCF_DO_STCLASS_OR;
+ }
+
+ /* This will finish on WHILEM, setting scan, or on NULL: */
+ minnext = study_chunk(&scan, &deltanext, last, data,
+ mincount == 0
+ ? (f & ~SCF_DO_SUBSTR) : f);
+
+ if (flags & SCF_DO_STCLASS)
+ data->start_class = oclass;
+ if (mincount == 0 || minnext == 0) {
+ if (flags & SCF_DO_STCLASS_OR) {
+ cl_or(data->start_class, &this_class);
+ }
+ else if (flags & SCF_DO_STCLASS_AND) {
+ /* Switch to OR mode: cache the old value of
+ * data->start_class */
+ StructCopy(data->start_class, &and_with,
+ struct regnode_charclass_class);
+ flags &= ~SCF_DO_STCLASS_AND;
+ StructCopy(&this_class, data->start_class,
+ struct regnode_charclass_class);
+ flags |= SCF_DO_STCLASS_OR;
+ data->start_class->flags |= ANYOF_EOS;
+ }
+ } else { /* Non-zero len */
+ if (flags & SCF_DO_STCLASS_OR) {
+ cl_or(data->start_class, &this_class);
+ cl_and(data->start_class, &and_with);
+ }
+ else if (flags & SCF_DO_STCLASS_AND)
+ cl_and(data->start_class, &this_class);
+ flags &= ~SCF_DO_STCLASS;
+ }
+ if (!scan) /* It was not CURLYX, but CURLY. */
+ scan = next;
+ if (ckWARN(WARN_UNSAFE) && (minnext + deltanext == 0)
+ && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
+ && maxcount <= REG_INFTY/3) /* Complement check for big count */
+ Perl_warner(aTHX_ WARN_UNSAFE,
+ "Strange *+?{} on zero-length expression");
+ min += minnext * mincount;
+ 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. */
+ if ( OP(oscan) == CURLYX && data
+ && data->flags & SF_IN_PAR
+ && !(data->flags & SF_HAS_EVAL)
+ && !deltanext && minnext == 1 ) {
+ /* Try to optimize to CURLYN. */
+ regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS;
+ regnode *nxt1 = nxt, *nxt2;
+
+ /* Skip open. */
+ nxt = regnext(nxt);
+ if (!strchr((char*)PL_simple,OP(nxt))
+ && !(PL_regkind[(U8)OP(nxt)] == EXACT
+ && STR_LEN(nxt) == 1))
+ goto nogo;
+ nxt2 = nxt;
+ nxt = regnext(nxt);
+ if (OP(nxt) != CLOSE)
+ goto nogo;
+ /* Now we know that nxt2 is the only contents: */
+ oscan->flags = ARG(nxt);
+ OP(oscan) = CURLYN;
+ OP(nxt1) = NOTHING; /* was OPEN. */
+#ifdef DEBUGGING
+ OP(nxt1 + 1) = OPTIMIZED; /* was count. */
+ NEXT_OFF(nxt1+ 1) = 0; /* just for consistancy. */
+ NEXT_OFF(nxt2) = 0; /* just for consistancy with CURLY. */
+ OP(nxt) = OPTIMIZED; /* was CLOSE. */
+ OP(nxt + 1) = OPTIMIZED; /* was count. */
+ NEXT_OFF(nxt+ 1) = 0; /* just for consistancy. */
+#endif
+ }
+ nogo:
+
+ /* Try optimization CURLYX => CURLYM. */
+ if ( OP(oscan) == CURLYX && data
+ && !(data->flags & SF_HAS_PAR)
+ && !(data->flags & SF_HAS_EVAL)
+ && !deltanext ) {
+ /* XXXX How to optimize if data == 0? */
+ /* Optimize to a simpler form. */
+ regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */
+ regnode *nxt2;
+
+ OP(oscan) = CURLYM;
+ while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/
+ && (OP(nxt2) != WHILEM))
+ nxt = nxt2;
+ OP(nxt2) = SUCCEED; /* Whas WHILEM */
+ /* Need to optimize away parenths. */
+ if (data->flags & SF_IN_PAR) {
+ /* Set the parenth number. */
+ regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/
+
+ if (OP(nxt) != CLOSE)
+ FAIL("panic opt close");
+ oscan->flags = ARG(nxt);
+ OP(nxt1) = OPTIMIZED; /* was OPEN. */
+ OP(nxt) = OPTIMIZED; /* was CLOSE. */
+#ifdef DEBUGGING
+ OP(nxt1 + 1) = OPTIMIZED; /* was count. */
+ OP(nxt + 1) = OPTIMIZED; /* was count. */
+ NEXT_OFF(nxt1 + 1) = 0; /* just for consistancy. */
+ NEXT_OFF(nxt + 1) = 0; /* just for consistancy. */
+#endif
+#if 0
+ while ( nxt1 && (OP(nxt1) != WHILEM)) {
+ regnode *nnxt = regnext(nxt1);
+
+ if (nnxt == nxt) {
+ if (reg_off_by_arg[OP(nxt1)])
+ ARG_SET(nxt1, nxt2 - nxt1);
+ else if (nxt2 - nxt1 < U16_MAX)
+ NEXT_OFF(nxt1) = nxt2 - nxt1;
+ else
+ OP(nxt) = NOTHING; /* Cannot beautify */
+ }
+ nxt1 = nnxt;
+ }
+#endif
+ /* Optimize again: */
+ study_chunk(&nxt1, &deltanext, nxt, NULL, 0);
+ }
+ 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) {
+ SV *last_str = Nullsv;
+ int counted = mincount != 0;
+
+ if (data->last_end > 0 && mincount != 0) { /* Ends with a string. */
+ I32 b = pos_before >= data->last_start_min
+ ? 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 -= old;
+ /* Get the added string: */
+ last_str = newSVpvn(s + old, l);
+ if (deltanext == 0 && pos_before == b) {
+ /* What was added is a constant string */
+ if (mincount > 1) {
+ SvGROW(last_str, (mincount * l) + 1);
+ repeatcpy(SvPVX(last_str) + l,
+ SvPVX(last_str), l, mincount - 1);
+ SvCUR(last_str) *= mincount;
+ /* Add additional parts. */
+ SvCUR_set(data->last_found,
+ SvCUR(data->last_found) - l);
+ sv_catsv(data->last_found, last_str);
+ data->last_end += l * (mincount - 1);
+ }
+ }
+ }
+ /* It is counted once already... */
+ data->pos_min += minnext * (mincount - counted);
+ data->pos_delta += - counted * deltanext +
+ (minnext + deltanext) * maxcount - minnext * mincount;
+ if (mincount != maxcount) {
+ /* Cannot extend fixed substrings found inside
+ the group. */
+ scan_commit(data);
+ if (mincount && last_str) {
+ sv_setsv(data->last_found, last_str);
+ data->last_end = data->pos_min;
+ data->last_start_min =
+ data->pos_min - CHR_SVLEN(last_str);
+ data->last_start_max = is_inf
+ ? I32_MAX
+ : data->pos_min + data->pos_delta
+ - 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 (PL_regkind[(U8)OP(next = regnext(oscan))] == NOTHING
+ && NEXT_OFF(next))
+ NEXT_OFF(oscan) += NEXT_OFF(next);
+ }
+ continue;
+ default: /* REF and CLUMP only? */
+ if (flags & SCF_DO_SUBSTR) {
+ scan_commit(data); /* Cannot expect anything... */
+ data->longest = &(data->longest_float);
+ }
+ is_inf = is_inf_internal = 1;
+ if (flags & SCF_DO_STCLASS_OR)
+ cl_anything(data->start_class);
+ flags &= ~SCF_DO_STCLASS;
+ break;
+ }
+ }
+ else if (strchr((char*)PL_simple,OP(scan)) || PL_regkind[(U8)OP(scan)] == ANYUTF8) {
+ int value;
+
+ if (flags & SCF_DO_SUBSTR) {
+ scan_commit(data);
+ data->pos_min++;
+ }
+ min++;
+ if (flags & SCF_DO_STCLASS) {
+ data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
+
+ /* Some of the logic below assumes that switching
+ locale on will only add false positives. */
+ switch (PL_regkind[(U8)OP(scan)]) {
+ case ANYUTF8:
+ case SANY:
+ case SANYUTF8:
+ case ALNUMUTF8:
+ case ANYOFUTF8:
+ case ALNUMLUTF8:
+ case NALNUMUTF8:
+ case NALNUMLUTF8:
+ case SPACEUTF8:
+ case NSPACEUTF8:
+ case SPACELUTF8:
+ case NSPACELUTF8:
+ case DIGITUTF8:
+ case NDIGITUTF8:
+ default:
+ do_default:
+ /* Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d", OP(scan)); */
+ if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
+ cl_anything(data->start_class);
+ break;
+ case REG_ANY:
+ if (OP(scan) == SANY)
+ goto do_default;
+ if (flags & SCF_DO_STCLASS_OR) { /* Everything but \n */
+ value = (ANYOF_BITMAP_TEST(data->start_class,'\n')
+ || (data->start_class->flags & ANYOF_CLASS));
+ cl_anything(data->start_class);
+ }
+ if (flags & SCF_DO_STCLASS_AND || !value)
+ ANYOF_BITMAP_CLEAR(data->start_class,'\n');
+ break;
+ case ANYOF:
+ if (flags & SCF_DO_STCLASS_AND)
+ cl_and(data->start_class,
+ (struct regnode_charclass_class*)scan);
+ else
+ cl_or(data->start_class,
+ (struct regnode_charclass_class*)scan);
+ break;
+ case ALNUM:
+ if (flags & SCF_DO_STCLASS_AND) {
+ if (!(data->start_class->flags & ANYOF_LOCALE)) {
+ ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
+ for (value = 0; value < 256; value++)
+ if (!isALNUM(value))
+ ANYOF_BITMAP_CLEAR(data->start_class, value);
+ }
+ }
+ else {
+ if (data->start_class->flags & ANYOF_LOCALE)
+ ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isALNUM(value))
+ ANYOF_BITMAP_SET(data->start_class, value);
+ }
+ }
+ break;
+ case ALNUML:
+ if (flags & SCF_DO_STCLASS_AND) {
+ if (data->start_class->flags & ANYOF_LOCALE)
+ ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
+ }
+ else {
+ ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
+ data->start_class->flags |= ANYOF_LOCALE;
+ }
+ break;
+ case NALNUM:
+ if (flags & SCF_DO_STCLASS_AND) {
+ if (!(data->start_class->flags & ANYOF_LOCALE)) {
+ ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
+ for (value = 0; value < 256; value++)
+ if (isALNUM(value))
+ ANYOF_BITMAP_CLEAR(data->start_class, value);
+ }
+ }
+ else {
+ if (data->start_class->flags & ANYOF_LOCALE)
+ ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isALNUM(value))
+ ANYOF_BITMAP_SET(data->start_class, value);
+ }
+ }
+ break;
+ case NALNUML:
+ if (flags & SCF_DO_STCLASS_AND) {
+ if (data->start_class->flags & ANYOF_LOCALE)
+ ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
+ }
+ else {
+ data->start_class->flags |= ANYOF_LOCALE;
+ ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
+ }
+ break;
+ case SPACE:
+ if (flags & SCF_DO_STCLASS_AND) {
+ if (!(data->start_class->flags & ANYOF_LOCALE)) {
+ ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
+ for (value = 0; value < 256; value++)
+ if (!isSPACE(value))
+ ANYOF_BITMAP_CLEAR(data->start_class, value);
+ }
+ }
+ else {
+ if (data->start_class->flags & ANYOF_LOCALE)
+ ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isSPACE(value))
+ ANYOF_BITMAP_SET(data->start_class, value);
+ }
+ }
+ break;
+ case SPACEL:
+ if (flags & SCF_DO_STCLASS_AND) {
+ if (data->start_class->flags & ANYOF_LOCALE)
+ ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
+ }
+ else {
+ data->start_class->flags |= ANYOF_LOCALE;
+ ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
+ }
+ break;
+ case NSPACE:
+ if (flags & SCF_DO_STCLASS_AND) {
+ if (!(data->start_class->flags & ANYOF_LOCALE)) {
+ ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
+ for (value = 0; value < 256; value++)
+ if (isSPACE(value))
+ ANYOF_BITMAP_CLEAR(data->start_class, value);
+ }
+ }
+ else {
+ if (data->start_class->flags & ANYOF_LOCALE)
+ ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isSPACE(value))
+ ANYOF_BITMAP_SET(data->start_class, value);
+ }
+ }
+ break;
+ case NSPACEL:
+ if (flags & SCF_DO_STCLASS_AND) {
+ if (data->start_class->flags & ANYOF_LOCALE) {
+ ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
+ for (value = 0; value < 256; value++)
+ if (!isSPACE(value))
+ ANYOF_BITMAP_CLEAR(data->start_class, value);
+ }
+ }
+ else {
+ data->start_class->flags |= ANYOF_LOCALE;
+ ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
+ }
+ break;
+ case DIGIT:
+ if (flags & SCF_DO_STCLASS_AND) {
+ ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NDIGIT);
+ for (value = 0; value < 256; value++)
+ if (!isDIGIT(value))
+ ANYOF_BITMAP_CLEAR(data->start_class, value);
+ }
+ else {
+ if (data->start_class->flags & ANYOF_LOCALE)
+ ANYOF_CLASS_SET(data->start_class,ANYOF_DIGIT);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isDIGIT(value))
+ ANYOF_BITMAP_SET(data->start_class, value);
+ }
+ }
+ break;
+ case NDIGIT:
+ if (flags & SCF_DO_STCLASS_AND) {
+ ANYOF_CLASS_CLEAR(data->start_class,ANYOF_DIGIT);
+ for (value = 0; value < 256; value++)
+ if (isDIGIT(value))
+ ANYOF_BITMAP_CLEAR(data->start_class, value);
+ }
+ else {
+ if (data->start_class->flags & ANYOF_LOCALE)
+ ANYOF_CLASS_SET(data->start_class,ANYOF_NDIGIT);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isDIGIT(value))
+ ANYOF_BITMAP_SET(data->start_class, value);
+ }
+ }
+ break;
+ }
+ if (flags & SCF_DO_STCLASS_OR)
+ cl_and(data->start_class, &and_with);
+ flags &= ~SCF_DO_STCLASS;
+ }
+ }
+ 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 ( PL_regkind[(U8)OP(scan)] == BRANCHJ
+ /* Lookbehind, or need to calculate parens/evals/stclass: */
+ && (scan->flags || data || (flags & SCF_DO_STCLASS))
+ && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) {
+ /* Lookahead/lookbehind */
+ I32 deltanext, minnext;
+ regnode *nscan;
+ struct regnode_charclass_class intrnl;
+ int f = 0;
+
+ data_fake.flags = 0;
+ if (data)
+ data_fake.whilem_c = data->whilem_c;
+ if ( flags & SCF_DO_STCLASS && !scan->flags
+ && OP(scan) == IFMATCH ) { /* Lookahead */
+ cl_init(&intrnl);
+ data_fake.start_class = &intrnl;
+ f = SCF_DO_STCLASS_AND;
+ }
+ next = regnext(scan);
+ nscan = NEXTOPER(NEXTOPER(scan));
+ minnext = study_chunk(&nscan, &deltanext, last, &data_fake, f);
+ if (scan->flags) {
+ if (deltanext) {
+ FAIL("variable length lookbehind not implemented");
+ }
+ else if (minnext > U8_MAX) {
+ FAIL2("lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
+ }
+ scan->flags = minnext;
+ }
+ if (data && 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 (f) {
+ int was = (data->start_class->flags & ANYOF_EOS);
+
+ cl_and(data->start_class, &intrnl);
+ if (was)
+ data->start_class->flags |= ANYOF_EOS;
+ }
+ }
+ else if (OP(scan) == OPEN) {
+ pars++;
+ }
+ 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) {
+ 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;
+ if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
+ cl_anything(data->start_class);
+ }
+ /* Else: zero-length, ignore. */
+ scan = regnext(scan);
+ }
+
+ finish:
+ *scanp = scan;
+ *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)
+ is_par = 0;
+ if (is_par && pars==1 && data) {
+ data->flags |= SF_IN_PAR;
+ data->flags &= ~SF_HAS_PAR;
+ }
+ else if (pars && data) {
+ data->flags |= SF_HAS_PAR;
+ data->flags &= ~SF_IN_PAR;
+ }
+ if (flags & SCF_DO_STCLASS_OR)
+ cl_and(data->start_class, &and_with);
+ return min;
+}
+
+STATIC I32
+S_add_data(pTHX_ I32 n, char *s)
+{
+ 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(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, 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++] = "";
+ }
+ PL_colorset = 1;
+}
/*
- pregcomp - compile a regular expression into internal code
* of the structure of the compiled regexp. [I'll say.]
*/
regexp *
-pregcomp(exp,xend,pm)
-char* exp;
-char* xend;
-PMOP* pm;
+Perl_pregcomp(pTHX_ char *exp, char *xend, PMOP *pm)
{
- I32 fold = pm->op_pmflags & PMf_FOLD;
+ dTHR;
register regexp *r;
- register char *scan;
- register SV *longish;
- SV *longest;
- register I32 len;
- register char *first;
+ regnode *scan;
+ SV **longest;
+ SV *longest_fixed;
+ SV *longest_float;
+ regnode *first;
I32 flags;
- I32 backish;
- I32 backest;
- I32 curback;
I32 minlen = 0;
I32 sawplus = 0;
I32 sawopen = 0;
+ scan_data_t data;
if (exp == NULL)
- croak("NULL regexp argument");
+ FAIL("NULL regexp argument");
+
+ 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],
+ (int)(xend - exp), PL_regprecomp, PL_colors[1]));
+ PL_regflags = pm->op_pmflags;
+ PL_regsawback = 0;
+
+ PL_regseen = 0;
+ PL_seen_zerolen = *exp == '^' ? -1 : 0;
+ PL_seen_evals = 0;
+ PL_extralen = 0;
/* First pass: determine size, legality. */
- regflags = pm->op_pmflags;
- regparse = exp;
- regxend = xend;
- regprecomp = savepvn(exp,xend-exp);
- regnaughty = 0;
- regsawback = 0;
- regnpar = 1;
- regsize = 0L;
- regcode = ®dummy;
- regc((char)MAGIC);
+ 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 %"IVdf" ", (IV)PL_regsize));
- /* Small enough for pointer-storage convention? */
- if (regsize >= 32767L) /* Probably could be 65535L. */
- FAIL("regexp too big");
+ /* Small enough for pointer-storage convention?
+ If extralen==0, this means that we will not need long jumps. */
+ if (PL_regsize >= 0x10000L && PL_extralen)
+ PL_regsize += PL_extralen;
+ else
+ PL_extralen = 0;
+ if (PL_reg_whilem_seen > 15)
+ PL_reg_whilem_seen = 15;
- /* Allocate space. */
- Newc(1001, r, sizeof(regexp) + (unsigned)regsize, char, regexp);
+ /* Allocate space and initialize. */
+ 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 = 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. */
- r->prelen = xend-exp;
- r->precomp = regprecomp;
- r->subbeg = r->subbase = NULL;
- regnaughty = 0;
- regparse = exp;
- regnpar = 1;
- regcode = r->program;
- regc((char)MAGIC);
+ 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;
- fold = pm->op_pmflags & PMf_FOLD;
- r->regstart = Nullsv; /* Worst-case defaults. */
- r->reganch = 0;
- r->regmust = Nullsv;
- r->regback = -1;
- r->regstclass = Nullch;
- r->naughty = regnaughty >= 10; /* Probably an expensive pattern. */
- scan = r->program+1; /* First BRANCH. */
- if (OP(regnext(scan)) == END) {/* Only one top-level choice. */
- scan = NEXTOPER(scan);
+ r->reganch = pm->op_pmflags & PMf_COMPILETIME; /* Again? */
+ pm->op_pmflags = PL_regflags;
+ if (UTF)
+ r->reganch |= ROPT_UTF8;
+ r->regstclass = NULL;
+ 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);
+ /* XXXX Should not we check for something else? Usually it is OPEN1... */
+ if (OP(scan) != BRANCH) { /* Only one top-level choice. */
+ I32 fake;
+ STRLEN longest_float_length, longest_fixed_length;
+ struct regnode_charclass_class ch_class;
+ int stclass_flag;
first = scan;
+ /* Skip introductions and multiplicators >= 1. */
while ((OP(first) == OPEN && (sawopen = 1)) ||
+ /* An OR of *one* alternative - should not happen now. */
(OP(first) == BRANCH && OP(regnext(first)) != BRANCH) ||
(OP(first) == PLUS) ||
(OP(first) == MINMOD) ||
- (regkind[(U8)OP(first)] == CURLY && ARG1(first) > 0) ) {
+ /* An {n,m} with n>0 */
+ (PL_regkind[(U8)OP(first)] == CURLY && ARG1(first) > 0) ) {
if (OP(first) == PLUS)
sawplus = 1;
else
/* Starting-point info. */
again:
- if (OP(first) == EXACTLY) {
- r->regstart = newSVpv(OPERAND(first)+1,*OPERAND(first));
- if (SvCUR(r->regstart) > !(sawstudy|fold))
- fbm_compile(r->regstart,fold);
- else
- sv_upgrade(r->regstart, SVt_PVBM);
+ if (PL_regkind[(U8)OP(first)] == EXACT) {
+ if (OP(first) == EXACT); /* Empty, get anchored substr later. */
+ else if ((OP(first) == EXACTF || OP(first) == EXACTFL)
+ && !UTF)
+ r->regstclass = first;
}
- else if (strchr(simple+2,OP(first)))
+ else if (strchr((char*)PL_simple,OP(first)))
r->regstclass = first;
- else if (OP(first) == BOUND || 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 = ROPT_ANCH;
+ 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;
+ }
+ else if (OP(first) == GPOS) {
+ r->reganch |= ROPT_ANCH_GPOS;
first = NEXTOPER(first);
- goto again;
+ 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 | 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;
+ 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;
- DEBUG_r(fprintf(Perl_debug_log,"first %d next %d offset %d\n",
- OP(first), OP(NEXTOPER(first)), first - scan));
+ /* Scan is after the zeroth branch, first is atomic matcher. */
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
+ (IV)(first - scan + 1)));
/*
* If there's something expensive in the r.e., find the
* longest literal string that must appear and make it the
* and avoiding duplication strengthens checking. Not a
* strong reason, but sufficient in the absence of others.
* [Now we resolve ties in favor of the earlier string if
- * it happens that curback has been invalidated, since the
+ * it happens that c_offset_min has been invalidated, since the
* earlier string may buy us something the later one won't.]
*/
- longish = newSVpv("",0);
- longest = newSVpv("",0);
- len = 0;
minlen = 0;
- curback = 0;
- backish = 0;
- backest = 0;
- while (OP(scan) != END) {
- if (OP(scan) == BRANCH) {
- if (OP(regnext(scan)) == BRANCH) {
- curback = -30000;
- while (OP(scan) == BRANCH)
- scan = regnext(scan);
- }
- else /* single branch is ok */
- scan = NEXTOPER(scan);
- continue;
- }
- if (OP(scan) == UNLESSM) {
- curback = -30000;
- scan = regnext(scan);
- continue;
- }
- if (OP(scan) == EXACTLY) {
- char *t;
-
- first = scan;
- while (OP(t = regnext(scan)) == CLOSE)
- scan = t;
- minlen += *OPERAND(first);
- if (curback - backish == len) {
- sv_catpvn(longish, OPERAND(first)+1,
- *OPERAND(first));
- len += *OPERAND(first);
- curback += *OPERAND(first);
- first = regnext(scan);
- }
- else if (*OPERAND(first) >= len + (curback >= 0)) {
- len = *OPERAND(first);
- sv_setpvn(longish, OPERAND(first)+1,len);
- backish = curback;
- curback += len;
- first = regnext(scan);
- }
- else
- curback += *OPERAND(first);
- }
- else if (strchr(varies,OP(scan))) {
- curback = -30000;
- len = 0;
- if (SvCUR(longish) > SvCUR(longest)) {
- sv_setsv(longest,longish);
- backest = backish;
- }
- sv_setpvn(longish,"",0);
- if (OP(scan) == PLUS && strchr(simple,OP(NEXTOPER(scan))))
- minlen++;
- else if (regkind[(U8)OP(scan)] == CURLY &&
- strchr(simple,OP(NEXTOPER(scan)+4)))
- minlen += ARG1(scan);
- }
- else if (strchr(simple,OP(scan))) {
- curback++;
- minlen++;
- len = 0;
- if (SvCUR(longish) > SvCUR(longest)) {
- sv_setsv(longest,longish);
- backest = backish;
- }
- sv_setpvn(longish,"",0);
- }
- scan = regnext(scan);
+
+ data.longest_fixed = newSVpvn("",0);
+ data.longest_float = newSVpvn("",0);
+ data.last_found = newSVpvn("",0);
+ data.longest = &(data.longest_fixed);
+ first = scan;
+ if (!r->regstclass) {
+ cl_init(&ch_class);
+ data.start_class = &ch_class;
+ stclass_flag = SCF_DO_STCLASS_AND;
+ } else /* XXXX Check for BOUND? */
+ stclass_flag = 0;
+
+ minlen = study_chunk(&first, &fake, scan + PL_regsize, /* Up to end */
+ &data, SCF_DO_SUBSTR | stclass_flag);
+ if ( PL_regnpar == 1 && data.longest == &(data.longest_fixed)
+ && data.last_start_min == 0 && data.last_end > 0
+ && !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 = CHR_SVLEN(data.longest_float);
+ if (longest_float_length
+ || (data.flags & SF_FL_BEFORE_EOL
+ && (!(data.flags & SF_FL_BEFORE_MEOL)
+ || (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;
+ 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;
}
- /* Prefer earlier on tie, unless we can tail match latter */
+ 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)
+ || (PL_regflags & PMf_MULTILINE)))) {
+ int t;
- if (SvCUR(longish) + (regkind[(U8)OP(first)] == EOL) >
- SvCUR(longest))
- {
- sv_setsv(longest,longish);
- backest = backish;
+ r->anchored_substr = data.longest_fixed;
+ r->anchored_offset = data.offset_fixed;
+ 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
- sv_setpvn(longish,"",0);
- if (SvCUR(longest)
- &&
- (!r->regstart
- ||
- !fbm_instr((unsigned char*) SvPVX(r->regstart),
- (unsigned char *) SvPVX(r->regstart)
- + SvCUR(r->regstart),
- longest)
- )
- )
- {
- r->regmust = longest;
- if (backest < 0)
- backest = -1;
- r->regback = backest;
- if (SvCUR(longest) > !(sawstudy || fold ||
- regkind[(U8)OP(first)]==EOL))
- fbm_compile(r->regmust,fold);
- (void)SvUPGRADE(r->regmust, SVt_PVBM);
- BmUSEFUL(r->regmust) = 100;
- if (regkind[(U8)OP(first)] == EOL && SvCUR(longish))
- SvTAIL_on(r->regmust);
+ else {
+ r->anchored_substr = Nullsv;
+ SvREFCNT_dec(data.longest_fixed);
+ longest_fixed_length = 0;
+ }
+ if (r->regstclass
+ && (OP(r->regstclass) == REG_ANY || OP(r->regstclass) == ANYUTF8
+ || OP(r->regstclass) == SANYUTF8 || OP(r->regstclass) == SANY))
+ r->regstclass = NULL;
+ if ((!r->anchored_substr || r->anchored_offset) && stclass_flag
+ && !(data.start_class->flags & ANYOF_EOS)
+ && !cl_is_anything(data.start_class)) {
+ SV *sv;
+ I32 n = add_data(1, "f");
+
+ New(1006, PL_regcomp_rx->data->data[n], 1,
+ struct regnode_charclass_class);
+ StructCopy(data.start_class,
+ (struct regnode_charclass_class*)PL_regcomp_rx->data->data[n],
+ struct regnode_charclass_class);
+ r->regstclass = (regnode*)PL_regcomp_rx->data->data[n];
+ r->reganch &= ~ROPT_SKIP; /* Used in find_byclass(). */
+ DEBUG_r((sv = sv_newmortal(),
+ regprop(sv, (regnode*)data.start_class),
+ PerlIO_printf(Perl_debug_log, "synthetic stclass.\n",
+ SvPVX(sv))));
+ }
+
+ /* A temporary algorithm prefers floated substr to fixed one to dig more info. */
+ if (longest_fixed_length > longest_float_length) {
+ r->check_substr = r->anchored_substr;
+ r->check_offset_min = r->check_offset_max = r->anchored_offset;
+ if (r->reganch & ROPT_ANCH_SINGLE)
+ r->reganch |= ROPT_NOSCAN;
}
else {
- SvREFCNT_dec(longest);
- longest = Nullsv;
+ r->check_substr = r->float_substr;
+ r->check_offset_min = data.offset_float_min;
+ r->check_offset_max = data.offset_float_max;
+ }
+ /* 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;
+ struct regnode_charclass_class ch_class;
+
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "\n"));
+ scan = r->program + 1;
+ cl_init(&ch_class);
+ data.start_class = &ch_class;
+ minlen = study_chunk(&scan, &fake, scan + PL_regsize, &data, SCF_DO_STCLASS_AND);
+ r->check_substr = r->anchored_substr = r->float_substr = Nullsv;
+ if (!(data.start_class->flags & ANYOF_EOS)
+ && !cl_is_anything(data.start_class)) {
+ SV *sv;
+ I32 n = add_data(1, "f");
+
+ New(1006, PL_regcomp_rx->data->data[n], 1,
+ struct regnode_charclass_class);
+ StructCopy(data.start_class,
+ (struct regnode_charclass_class*)PL_regcomp_rx->data->data[n],
+ struct regnode_charclass_class);
+ r->regstclass = (regnode*)PL_regcomp_rx->data->data[n];
+ r->reganch &= ~ROPT_SKIP; /* Used in find_byclass(). */
+ DEBUG_r((sv = sv_newmortal(),
+ regprop(sv, (regnode*)data.start_class),
+ PerlIO_printf(Perl_debug_log, "synthetic stclass.\n",
+ SvPVX(sv))));
}
- SvREFCNT_dec(longish);
}
- r->do_folding = fold;
- r->nparens = regnpar - 1;
r->minlen = minlen;
- Newz(1002, r->startp, regnpar, char*);
- Newz(1002, r->endp, regnpar, char*);
+ if (PL_regseen & REG_SEEN_GPOS)
+ r->reganch |= ROPT_GPOS_SEEN;
+ if (PL_regseen & REG_SEEN_LOOKBEHIND)
+ r->reganch |= ROPT_LOOKBEHIND_SEEN;
+ if (PL_regseen & REG_SEEN_EVAL)
+ r->reganch |= ROPT_EVAL_SEEN;
+ Newz(1002, r->startp, PL_regnpar, I32);
+ Newz(1002, r->endp, PL_regnpar, I32);
DEBUG_r(regdump(r));
return(r);
}
* is a trifle forced, but the need to tie the tails of the branches to what
* follows makes it hard to avoid.
*/
-static char *
-reg(paren, flagp)
-I32 paren; /* Parenthesized? */
-I32 *flagp;
+STATIC regnode *
+S_reg(pTHX_ I32 paren, I32 *flagp)
+ /* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */
{
- register char *ret;
- register char *br;
- register char *ender = 0;
+ 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;
+ I32 flags, oregflags = PL_regflags, have_branch = 0, open = 0;
+ char c;
- *flagp = HASWIDTH; /* Tentatively. */
+ *flagp = 0; /* Tentatively. */
/* Make an OPEN node, if parenthesized. */
if (paren) {
- if (*regparse == '?') {
- regparse++;
- paren = *regparse++;
- ret = NULL;
+ 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 ':':
+ case '<':
+ PL_regseen |= REG_SEEN_LOOKBEHIND;
+ if (*PL_regcomp_parse == '!')
+ paren = ',';
+ if (*PL_regcomp_parse != '=' && *PL_regcomp_parse != '!')
+ goto unknown;
+ PL_regcomp_parse++;
case '=':
case '!':
+ PL_seen_zerolen++;
+ case ':':
+ case '>':
break;
case '$':
case '@':
- croak("Sequence (?%c...) not implemented", paren);
+ FAIL2("Sequence (?%c...) not implemented", (int)paren);
break;
case '#':
- while (*regparse && *regparse != ')')
- regparse++;
- if (*regparse != ')')
- croak("Sequence (?#... not terminated");
+ 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 = PL_regcomp_parse;
+ SV *sv;
+ OP_4tree *sop, *rop;
+
+ 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--;
+ PL_regcomp_parse++;
+ }
+ if (*PL_regcomp_parse != ')')
+ FAIL("Sequence (?{...}) not terminated or not {}-balanced");
+ if (!SIZE_ONLY) {
+ AV *av;
+
+ if (PL_regcomp_parse - 1 - s)
+ sv = newSVpvn(s, PL_regcomp_parse - 1 - s);
+ else
+ sv = newSVpvn("", 0);
+
+ rop = sv_compile_2op(sv, &sop, "re", &av);
+
+ 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 (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 (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 (PL_regcomp_parse[0] >= '1' && PL_regcomp_parse[0] <= '9' ) {
+ parno = atoi(PL_regcomp_parse++);
+
+ while (isDIGIT(*PL_regcomp_parse))
+ PL_regcomp_parse++;
+ ret = reganode(GROUPP, parno);
+ if ((c = *nextchar()) != ')')
+ FAIL2("Switch (?(number%c not recognized", c);
+ insert_if:
+ regtail(ret, reganode(IFTHEN, 0));
+ br = regbranch(&flags, 1);
+ if (br == NULL)
+ br = reganode(LONGJMP, 0);
+ 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
+ lastbr = NULL;
+ if (c != ')')
+ FAIL("Switch (?(condition)... contains too many branches");
+ ender = reg_node(TAIL);
+ regtail(br, ender);
+ if (lastbr) {
+ regtail(lastbr, ender);
+ regtail(NEXTOPER(NEXTOPER(lastbr)), ender);
+ }
+ else
+ regtail(ret, ender);
+ return ret;
+ }
+ else {
+ FAIL2("Unknown condition for (?(%.2s", PL_regcomp_parse);
+ }
+ }
+ case 0:
+ FAIL("Sequence (? incomplete");
+ break;
default:
- --regparse;
- while (*regparse && strchr("iogmsx", *regparse))
- pmflag(®flags, *regparse++);
- if (*regparse != ')')
- croak("Sequence (?%c...) not recognized", *regparse);
+ --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 (*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);
+ br = regbranch(&flags, 1);
if (br == NULL)
return(NULL);
- if (ret != NULL)
- regtail(ret, br); /* OPEN -> first. */
- else
+ if (*PL_regcomp_parse == '|') {
+ if (!SIZE_ONLY && PL_extralen) {
+ reginsert(BRANCHJ, br);
+ }
+ else
+ reginsert(BRANCH, br);
+ have_branch = 1;
+ if (SIZE_ONLY)
+ 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 */
ret = br;
- if (!(flags&HASWIDTH))
- *flagp &= ~HASWIDTH;
+ if (flags&HASWIDTH)
+ *flagp |= HASWIDTH;
*flagp |= flags&SPSTART;
- while (*regparse == '|') {
+ lastbr = br;
+ 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)
+ PL_extralen += 2; /* Account for LONGJMP. */
nextchar();
- br = regbranch(&flags);
+ br = regbranch(&flags, 0);
if (br == NULL)
return(NULL);
- regtail(ret, br); /* BRANCH -> BRANCH. */
- if (!(flags&HASWIDTH))
- *flagp &= ~HASWIDTH;
+ regtail(lastbr, br); /* BRANCH -> BRANCH. */
+ lastbr = br;
+ if (flags&HASWIDTH)
+ *flagp |= HASWIDTH;
*flagp |= flags&SPSTART;
}
- /* Make a closing node, and hook it on the end. */
- switch (paren) {
- case ':':
- ender = regnode(NOTHING);
- break;
- case 1:
- ender = reganode(CLOSE, parno);
- break;
- case '=':
- case '!':
- ender = regnode(SUCCEED);
- *flagp &= ~HASWIDTH;
- break;
- case 0:
- ender = regnode(END);
- break;
+ if (have_branch || paren != ':') {
+ /* Make a closing node, and hook it on the end. */
+ switch (paren) {
+ case ':':
+ ender = reg_node(TAIL);
+ break;
+ case 1:
+ ender = reganode(CLOSE, parno);
+ break;
+ case '<':
+ case ',':
+ case '=':
+ case '!':
+ *flagp &= ~HASWIDTH;
+ /* FALL THROUGH */
+ case '>':
+ ender = reg_node(SUCCEED);
+ break;
+ case 0:
+ ender = reg_node(END);
+ break;
+ }
+ regtail(lastbr, ender);
+
+ if (have_branch) {
+ /* Hook the tails of the branches to the closing node. */
+ for (br = ret; br != NULL; br = regnext(br)) {
+ regoptail(br, ender);
+ }
+ }
}
- regtail(ret, ender);
- /* Hook the tails of the branches to the closing node. */
- for (br = ret; br != NULL; br = regnext(br))
- regoptail(br, ender);
+ {
+ char *p;
+ static char parens[] = "=!<,>";
- if (paren == '=') {
- reginsert(IFMATCH,ret);
- regtail(ret, regnode(NOTHING));
- }
- else if (paren == '!') {
- reginsert(UNLESSM,ret);
- regtail(ret, regnode(NOTHING));
+ if (paren && (p = strchr(parens, paren))) {
+ int node = ((p - parens) % 2) ? UNLESSM : IFMATCH;
+ int flag = (p - parens) > 1;
+
+ if (paren == '>')
+ node = SUSPEND, flag = 0;
+ reginsert(node,ret);
+ ret->flags = flag;
+ regtail(ret, reg_node(TAIL));
+ }
}
/* Check for proper termination. */
- if (paren && (regparse >= regxend || *nextchar() != ')')) {
- FAIL("unmatched () in regexp");
- } else if (!paren && regparse < regxend) {
- if (*regparse == ')') {
+ 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 */
}
*
* Implements the concatenation operator.
*/
-static char *
-regbranch(flagp)
-I32 *flagp;
+STATIC regnode *
+S_regbranch(pTHX_ I32 *flagp, I32 first)
{
- register char *ret;
- register char *chain;
- register char *latest;
- I32 flags = 0;
+ dTHR;
+ register regnode *ret;
+ register regnode *chain = NULL;
+ register regnode *latest;
+ I32 flags = 0, c = 0;
- *flagp = WORST; /* Tentatively. */
+ if (first)
+ ret = NULL;
+ else {
+ if (!SIZE_ONLY && PL_extralen)
+ ret = reganode(BRANCHJ,0);
+ else
+ ret = reg_node(BRANCH);
+ }
+
+ if (!first && SIZE_ONLY)
+ PL_extralen += 1; /* BRANCHJ */
+
+ *flagp = WORST; /* Tentatively. */
- ret = regnode(BRANCH);
- chain = NULL;
- regparse--;
+ PL_regcomp_parse--;
nextchar();
- while (regparse < regxend && *regparse != '|' && *regparse != ')') {
+ while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != '|' && *PL_regcomp_parse != ')') {
flags &= ~TRYAGAIN;
latest = regpiece(&flags);
if (latest == NULL) {
continue;
return(NULL);
}
+ else if (ret == NULL)
+ ret = latest;
*flagp |= flags&HASWIDTH;
- if (chain == NULL) /* First piece. */
+ if (chain == NULL) /* First piece. */
*flagp |= flags&SPSTART;
else {
- regnaughty++;
+ PL_regnaughty++;
regtail(chain, latest);
}
chain = latest;
+ c++;
+ }
+ if (chain == NULL) { /* Loop ran zero times. */
+ chain = reg_node(NOTHING);
+ if (ret == NULL)
+ ret = chain;
+ }
+ if (c == 1) {
+ *flagp |= flags&SIMPLE;
}
- if (chain == NULL) /* Loop ran zero times. */
- (void) regnode(NOTHING);
return(ret);
}
* It might seem that this node could be dispensed with entirely, but the
* endmarker role is not redundant.
*/
-static char *
-regpiece(flagp)
-I32 *flagp;
+STATIC regnode *
+S_regpiece(pTHX_ I32 *flagp)
{
- register char *ret;
+ dTHR;
+ register regnode *ret;
register char op;
register char *next;
I32 flags;
- char *origparse = regparse;
+ char *origparse = PL_regcomp_parse;
char *maxpos;
I32 min;
- I32 max = 32767;
+ I32 max = REG_INFTY;
ret = regatom(&flags);
if (ret == NULL) {
return(NULL);
}
- op = *regparse;
- if (op == '(' && regparse[1] == '?' && regparse[2] == '#') {
- while (op && op != ')')
- op = *++regparse;
- if (op) {
- nextchar();
- op = *regparse;
- }
- }
+ op = *PL_regcomp_parse;
- if (op == '{' && regcurly(regparse)) {
- next = regparse + 1;
+ if (op == '{' && regcurly(PL_regcomp_parse)) {
+ next = PL_regcomp_parse + 1;
maxpos = Nullch;
while (isDIGIT(*next) || *next == ',') {
if (*next == ',') {
if (*next == '}') { /* got one */
if (!maxpos)
maxpos = next;
- regparse++;
- min = atoi(regparse);
+ PL_regcomp_parse++;
+ min = atoi(PL_regcomp_parse);
if (*maxpos == ',')
maxpos++;
else
- maxpos = regparse;
+ maxpos = PL_regcomp_parse;
max = atoi(maxpos);
if (!max && *maxpos != '0')
- max = 32767; /* meaning "infinity" */
- regparse = next;
+ max = REG_INFTY; /* meaning "infinity" */
+ else if (max >= REG_INFTY)
+ FAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
+ 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, regnode(WHILEM));
+ 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);
- regtail(ret, regnode(NOTHING));
+ if (!SIZE_ONLY && PL_extralen)
+ NEXT_OFF(ret) = 3; /* Go over NOTHING to LONGJMP. */
+ regtail(ret, reg_node(NOTHING));
+ if (SIZE_ONLY)
+ PL_reg_whilem_seen++, PL_extralen += 3;
+ PL_regnaughty += 4 + PL_regnaughty; /* compound interest */
}
+ ret->flags = 0;
if (min > 0)
- *flagp = (WORST|HASWIDTH);
+ *flagp = WORST;
+ if (max > 0)
+ *flagp |= HASWIDTH;
if (max && max < min)
- croak("Can't do {n,m} with n > m");
- if (regcode != ®dummy) {
-#ifdef REGALIGN
- *(unsigned short *)(ret+3) = min;
- *(unsigned short *)(ret+5) = max;
-#else
- ret[3] = min >> 8; ret[4] = min & 0377;
- ret[5] = max >> 8; ret[6] = max & 0377;
-#endif
+ FAIL("Can't do {n,m} with n > m");
+ if (!SIZE_ONLY) {
+ ARG1_SET(ret, min);
+ ARG2_SET(ret, max);
}
goto nest_check;
*flagp = flags;
return(ret);
}
+
+#if 0 /* Now runtime fix should be reliable. */
+ if (!(flags&HASWIDTH) && op != '?')
+ FAIL("regexp *+ operand could be empty");
+#endif
+
nextchar();
- *flagp = (op != '+') ? (WORST|SPSTART) : (WORST|HASWIDTH);
+ *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
if (op == '*' && (flags&SIMPLE)) {
reginsert(STAR, ret);
- regnaughty += 4;
+ ret->flags = 0;
+ PL_regnaughty += 4;
}
else if (op == '*') {
min = 0;
goto do_curly;
- } else if (op == '+' && (flags&SIMPLE)) {
+ }
+ else if (op == '+' && (flags&SIMPLE)) {
reginsert(PLUS, ret);
- regnaughty += 3;
+ ret->flags = 0;
+ 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 && regcode != ®dummy && !(flags&HASWIDTH) && max > 10000) {
- warn("%.*s matches null string many times",
- regparse - 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 (*regparse == '?') {
+ if (*PL_regcomp_parse == '?') {
nextchar();
reginsert(MINMOD, ret);
-#ifdef REGALIGN
- regtail(ret, ret + 4);
-#else
- regtail(ret, ret + 3);
-#endif
+ regtail(ret, ret + NODE_STEP_REGNODE);
}
- if (ISMULT2(regparse))
+ if (ISMULT2(PL_regcomp_parse))
FAIL("nested *?+ in regexp");
return(ret);
*
* [Yes, it is worth fixing, some scripts can run twice the speed.]
*/
-static char *
-regatom(flagp)
-I32 *flagp;
+STATIC regnode *
+S_regatom(pTHX_ I32 *flagp)
{
- register char *ret = 0;
+ dTHR;
+ register regnode *ret = 0;
I32 flags;
*flagp = WORST; /* Tentatively. */
tryagain:
- switch (*regparse) {
+ switch (*PL_regcomp_parse) {
case '^':
+ PL_seen_zerolen++;
nextchar();
- if (regflags & PMf_MULTILINE)
- ret = regnode(MBOL);
- else if (regflags & PMf_SINGLELINE)
- ret = regnode(SBOL);
+ if (PL_regflags & PMf_MULTILINE)
+ ret = reg_node(MBOL);
+ else if (PL_regflags & PMf_SINGLELINE)
+ ret = reg_node(SBOL);
else
- ret = regnode(BOL);
+ ret = reg_node(BOL);
break;
case '$':
+ if (PL_regcomp_parse[1])
+ PL_seen_zerolen++;
nextchar();
- if (regflags & PMf_MULTILINE)
- ret = regnode(MEOL);
- else if (regflags & PMf_SINGLELINE)
- ret = regnode(SEOL);
+ if (PL_regflags & PMf_MULTILINE)
+ ret = reg_node(MEOL);
+ else if (PL_regflags & PMf_SINGLELINE)
+ ret = reg_node(SEOL);
else
- ret = regnode(EOL);
+ ret = reg_node(EOL);
break;
case '.':
nextchar();
- if (regflags & PMf_SINGLELINE)
- ret = regnode(SANY);
- else
- ret = regnode(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 '[':
- regparse++;
- ret = regclass();
+ PL_regcomp_parse++;
+ ret = (UTF ? regclassutf8() : regclass());
+ if (*PL_regcomp_parse != ']')
+ FAIL("unmatched [] in regexp");
+ nextchar();
*flagp |= HASWIDTH|SIMPLE;
break;
case '(':
goto tryagain;
return(NULL);
}
- *flagp |= flags&(HASWIDTH|SPSTART);
+ *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE);
break;
case '|':
case ')':
*flagp |= TRYAGAIN;
return NULL;
}
- croak("internal urp in regexp at /%s/", regparse);
+ FAIL2("internal urp in regexp at /%s/", PL_regcomp_parse);
/* Supposed to be caught earlier. */
break;
+ case '{':
+ if (!regcurly(PL_regcomp_parse)) {
+ PL_regcomp_parse++;
+ goto defchar;
+ }
+ /* FALL THROUGH */
case '?':
case '+':
case '*':
- FAIL("?+* follows nothing in regexp");
+ FAIL("?+*{} follows nothing in regexp");
break;
case '\\':
- switch (*++regparse) {
+ switch (*++PL_regcomp_parse) {
case 'A':
- ret = regnode(SBOL);
+ PL_seen_zerolen++;
+ ret = reg_node(SBOL);
*flagp |= SIMPLE;
nextchar();
break;
case 'G':
- ret = regnode(GBOL);
+ ret = reg_node(GPOS);
+ PL_regseen |= REG_SEEN_GPOS;
*flagp |= SIMPLE;
nextchar();
break;
case 'Z':
- ret = regnode(SEOL);
+ ret = reg_node(SEOL);
+ *flagp |= SIMPLE;
+ nextchar();
+ break;
+ case 'z':
+ ret = reg_node(EOS);
*flagp |= SIMPLE;
+ 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 = regnode(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 = regnode(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':
- ret = regnode(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':
- ret = regnode(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 = regnode(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 = regnode(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 = regnode(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 = regnode(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':
case '1': case '2': case '3': case '4':
case '5': case '6': case '7': case '8': case '9':
{
- I32 num = atoi(regparse);
+ I32 num = atoi(PL_regcomp_parse);
- if (num > 9 && num >= regnpar)
+ if (num > 9 && num >= PL_regnpar)
goto defchar;
else {
- regsawback = 1;
- ret = reganode(REF, num);
+ 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(*regparse))
- regparse++;
- regparse--;
+ while (isDIGIT(*PL_regcomp_parse))
+ PL_regcomp_parse++;
+ PL_regcomp_parse--;
nextchar();
}
}
break;
case '\0':
- if (regparse >= 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 (regparse < regxend && *regparse != '\n') regparse++;
- if (regparse < 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 char ender;
+ register UV ender;
register char *p;
- char *oldp;
+ char *oldp, *s;
I32 numlen;
- regparse++;
+ PL_regcomp_parse++;
defchar:
- ret = regnode(EXACTLY);
- regc(0); /* save spot for len */
- for (len = 0, p = regparse - 1;
- len < 127 && p < regxend;
+ ret = reg_node(FOLD
+ ? (LOC ? EXACTFL : EXACTF)
+ : EXACT);
+ s = STRING(ret);
+ for (len = 0, p = PL_regcomp_parse - 1;
+ len < 127 && p < PL_regxend;
len++)
{
oldp = p;
+
+ if (PL_regflags & PMf_EXTENDED)
+ p = regwhite(p, PL_regxend);
switch (*p) {
case '^':
case '$':
case 'A':
case 'G':
case 'Z':
+ case 'z':
case 'w':
case 'W':
case 'b':
case 'S':
case 'd':
case 'D':
+ case 'p':
+ case 'P':
--p;
goto loopdone;
case 'n':
p++;
break;
case 'e':
- ender = '\033';
+#ifdef ASCIIish
+ ender = '\033';
+#else
+ ender = '\047';
+#endif
p++;
break;
case 'a':
- ender = '\007';
+#ifdef ASCIIish
+ ender = '\007';
+#else
+ ender = '\057';
+#endif
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 = (UV)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 = (UV)scan_hex(p, 2, &numlen);
+ p += numlen;
+ }
break;
case 'c':
p++;
- ender = *p++;
- if (isLOWER(ender))
- ender = toUPPER(ender);
- ender ^= 64;
+ ender = UCHARAT(p++);
+ ender = toCTRL(ender);
break;
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) ) {
- ender = scan_oct(p, 3, &numlen);
+ (isDIGIT(p[1]) && atoi(p) >= PL_regnpar) ) {
+ ender = (UV)scan_oct(p, 3, &numlen);
p += numlen;
}
else {
}
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;
- case '#':
- if (regflags & PMf_EXTENDED) {
- while (p < regxend && *p != '\n') p++;
- }
- /* FALL THROUGH */
- case ' ': case '\t': case '\n': case '\r': case '\f': case '\v':
- if (regflags & PMf_EXTENDED) {
- p++;
- len--;
- continue;
- }
- /* FALL THROUGH */
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_FOLD && isUPPER(ender))
- ender = toLOWER(ender);
+ 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);
+ REGC(ender, s++);
+ }
+ break;
+ }
+ if (ender >= 0x80 && UTF) {
+ reguni(ender, s, &numlen);
+ s += numlen;
+ len += numlen - 1;
+ }
+ else
+ REGC(ender, s++);
+ }
+ loopdone:
+ PL_regcomp_parse = p - 1;
+ nextchar();
+ if (len < 0)
+ FAIL("internal disaster in regexp");
+ if (len > 0)
+ *flagp |= HASWIDTH;
+ if (len == 1)
+ *flagp |= SIMPLE;
+ if (!SIZE_ONLY)
+ STR_LEN(ret) = len;
+ if (SIZE_ONLY)
+ PL_regsize += STR_SZ(len);
+ else
+ PL_regcode += STR_SZ(len);
+ }
+ break;
+ }
+
+ return(ret);
+}
+
+STATIC char *
+S_regwhite(pTHX_ char *p, char *e)
+{
+ while (p < e) {
+ if (isSPACE(*p))
+ ++p;
+ else if (*p == '#') {
+ do {
+ p++;
+ } while (p < e && *p != '\n');
+ }
+ else
+ break;
+ }
+ 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 = OOB_NAMEDCLASS;
+
+ 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 (!SIZE_ONLY && ckWARN(WARN_UNSAFE) &&
+ (*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 *
+S_regclass(pTHX)
+{
+ dTHR;
+ register UV value;
+ register I32 lastvalue = OOB_CHAR8;
+ register I32 range = 0;
+ register regnode *ret;
+ register I32 def;
+ I32 numlen;
+ I32 namedclass;
+ char *rangebegin;
+ bool need_class = 0;
+
+ ret = reg_node(ANYOF);
+ if (SIZE_ONLY)
+ PL_regsize += ANYOF_SKIP;
+ else {
+ ret->flags = 0;
+ ANYOF_BITMAP_ZERO(ret);
+ PL_regcode += ANYOF_SKIP;
+ if (FOLD)
+ ANYOF_FLAGS(ret) |= ANYOF_FOLD;
+ if (LOC)
+ ANYOF_FLAGS(ret) |= ANYOF_LOCALE;
+ }
+ if (*PL_regcomp_parse == '^') { /* Complement of range. */
+ PL_regnaughty++;
+ PL_regcomp_parse++;
+ if (!SIZE_ONLY)
+ ANYOF_FLAGS(ret) |= ANYOF_INVERT;
+ }
+
+ if (!SIZE_ONLY && ckWARN(WARN_UNSAFE))
+ 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 = 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;
+#ifdef ASCIIish
+ case 'e': value = '\033'; break;
+ case 'a': value = '\007'; break;
+#else
+ case 'e': value = '\047'; break;
+ case 'a': value = '\057'; break;
+#endif
+ case 'x':
+ value = (UV)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 = (UV)scan_oct(--PL_regcomp_parse, 3, &numlen);
+ PL_regcomp_parse += numlen;
+ break;
+ default:
+ if (!SIZE_ONLY && ckWARN(WARN_UNSAFE) && isALPHA(value))
+ Perl_warner(aTHX_ WARN_UNSAFE,
+ "/%.127s/: Unrecognized escape \\%c in character class passed through",
+ PL_regprecomp,
+ (int)value);
+ break;
+ }
+ }
+ if (namedclass > OOB_NAMEDCLASS) {
+ if (!need_class && !SIZE_ONLY)
+ ANYOF_CLASS_ZERO(ret);
+ need_class = 1;
+ 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);
+ ANYOF_BITMAP_SET(ret, lastvalue);
+ ANYOF_BITMAP_SET(ret, '-');
+ }
+ range = 0; /* this is not a true range */
+ }
+ if (!SIZE_ONLY) {
+ switch (namedclass) {
+ case ANYOF_ALNUM:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_ALNUM);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isALNUM(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_NALNUM:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NALNUM);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isALNUM(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_SPACE:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_SPACE);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isSPACE(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_NSPACE:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NSPACE);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isSPACE(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_DIGIT:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_DIGIT);
+ else {
+ for (value = '0'; value <= '9'; value++)
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_NDIGIT:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NDIGIT);
+ else {
+ for (value = 0; value < '0'; value++)
+ ANYOF_BITMAP_SET(ret, value);
+ for (value = '9' + 1; value < 256; value++)
+ ANYOF_BITMAP_SET(ret, value);
}
break;
+ case ANYOF_NALNUMC:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NALNUMC);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isALNUMC(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_ALNUMC:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_ALNUMC);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isALNUMC(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_ALPHA:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_ALPHA);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isALPHA(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_NALPHA:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NALPHA);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isALPHA(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_ASCII:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_ASCII);
+ else {
+#ifdef ASCIIish
+ for (value = 0; value < 128; value++)
+ ANYOF_BITMAP_SET(ret, value);
+#else /* EBCDIC */
+ for (value = 0; value < 256; value++)
+ if (isASCII(value))
+ ANYOF_BITMAP_SET(ret, value);
+#endif /* EBCDIC */
+ }
+ break;
+ case ANYOF_NASCII:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NASCII);
+ else {
+#ifdef ASCIIish
+ for (value = 128; value < 256; value++)
+ ANYOF_BITMAP_SET(ret, value);
+#else /* EBCDIC */
+ for (value = 0; value < 256; value++)
+ if (!isASCII(value))
+ ANYOF_BITMAP_SET(ret, value);
+#endif /* EBCDIC */
+ }
+ break;
+ case ANYOF_CNTRL:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_CNTRL);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isCNTRL(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ lastvalue = OOB_CHAR8;
+ break;
+ case ANYOF_NCNTRL:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NCNTRL);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isCNTRL(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_GRAPH:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_GRAPH);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isGRAPH(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_NGRAPH:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NGRAPH);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isGRAPH(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_LOWER:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_LOWER);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isLOWER(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_NLOWER:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NLOWER);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isLOWER(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_PRINT:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_PRINT);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isPRINT(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_NPRINT:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NPRINT);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isPRINT(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_PUNCT:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_PUNCT);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isPUNCT(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_NPUNCT:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NPUNCT);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isPUNCT(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_UPPER:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_UPPER);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isUPPER(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_NUPPER:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NUPPER);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isUPPER(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_XDIGIT:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_XDIGIT);
+ else {
+ for (value = 0; value < 256; value++)
+ if (isXDIGIT(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ case ANYOF_NXDIGIT:
+ if (LOC)
+ ANYOF_CLASS_SET(ret, ANYOF_NXDIGIT);
+ else {
+ for (value = 0; value < 256; value++)
+ if (!isXDIGIT(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+ break;
+ default:
+ FAIL("invalid [::] class in regexp");
+ break;
+ }
+ if (LOC)
+ ANYOF_FLAGS(ret) |= ANYOF_CLASS;
+ 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 {
+ 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(ret, '-');
+ } 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(ret, i);
+ } else {
+ for (i = lastvalue; i <= value; i++)
+ if (isUPPER(i))
+ ANYOF_BITMAP_SET(ret, i);
}
- regc(ender);
}
- loopdone:
- regparse = p - 1;
- nextchar();
- if (len < 0)
- FAIL("internal disaster in regexp");
- if (len > 0)
- *flagp |= HASWIDTH;
- if (len == 1)
- *flagp |= SIMPLE;
- if (regcode != ®dummy)
- *OPERAND(ret) = len;
- regc('\0');
+ else
+#endif
+ for ( ; lastvalue <= value; lastvalue++)
+ ANYOF_BITMAP_SET(ret, lastvalue);
+ }
+ range = 0;
+ }
+ if (need_class) {
+ if (SIZE_ONLY)
+ PL_regsize += ANYOF_CLASS_ADD_SKIP;
+ else
+ PL_regcode += ANYOF_CLASS_ADD_SKIP;
+ }
+ /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */
+ if (!SIZE_ONLY &&
+ (ANYOF_FLAGS(ret) & (ANYOF_FLAGS_ALL ^ ANYOF_INVERT)) == ANYOF_FOLD) {
+ for (value = 0; value < 256; ++value) {
+ if (ANYOF_BITMAP_TEST(ret, value)) {
+ I32 cf = PL_fold[value];
+ ANYOF_BITMAP_SET(ret, cf);
+ }
}
- break;
+ ANYOF_FLAGS(ret) &= ~ANYOF_FOLD;
}
-
- return(ret);
-}
-
-static void
-regset(bits,def,c)
-char *bits;
-I32 def;
-register I32 c;
-{
- if (regcode == ®dummy)
- return;
- c &= 255;
- if (def)
- bits[c >> 3] &= ~(1 << (c & 7));
- else
- bits[c >> 3] |= (1 << (c & 7));
+ /* optimize inverted simple patterns (e.g. [^a-z]) */
+ if (!SIZE_ONLY && (ANYOF_FLAGS(ret) & ANYOF_FLAGS_ALL) == ANYOF_INVERT) {
+ for (value = 0; value < ANYOF_BITMAP_SIZE; ++value)
+ ANYOF_BITMAP(ret)[value] ^= ANYOF_FLAGS_ALL;
+ ANYOF_FLAGS(ret) = 0;
+ }
+ return ret;
}
-static char *
-regclass()
+STATIC regnode *
+S_regclassutf8(pTHX)
{
- register char *bits;
- register I32 class;
- register I32 lastclass = 1234;
+ dTHR;
+ register char *e;
+ register UV value;
+ register U32 lastvalue = OOB_UTF8;
register I32 range = 0;
- register char *ret;
- register I32 def;
+ register regnode *ret;
I32 numlen;
+ I32 n;
+ SV *listsv;
+ U8 flags = 0;
+ I32 namedclass;
+ char *rangebegin;
- ret = regnode(ANYOF);
- if (*regparse == '^') { /* Complement of range. */
- regnaughty++;
- regparse++;
- def = 0;
- } else {
- def = 255;
+ 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);
}
- bits = regcode;
- for (class = 0; class < 32; class++)
- regc(def);
- if (*regparse == ']' || *regparse == '-')
+
+ if (!SIZE_ONLY && ckWARN(WARN_UNSAFE))
+ checkposixcc();
+
+ if (*PL_regcomp_parse == ']' || *PL_regcomp_parse == '-')
goto skipcond; /* allow 1st char to be ] or - */
- while (regparse < regxend && *regparse != ']') {
+
+ while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != ']') {
skipcond:
- class = UCHARAT(regparse++);
- if (class == '\\') {
- class = UCHARAT(regparse++);
- switch (class) {
- case 'w':
- for (class = 0; class < 256; class++)
- if (isALNUM(class))
- regset(bits,def,class);
- lastclass = 1234;
- continue;
- case 'W':
- for (class = 0; class < 256; class++)
- if (!isALNUM(class))
- regset(bits,def,class);
- lastclass = 1234;
- continue;
- case 's':
- for (class = 0; class < 256; class++)
- if (isSPACE(class))
- regset(bits,def,class);
- lastclass = 1234;
- continue;
- case 'S':
- for (class = 0; class < 256; class++)
- if (!isSPACE(class))
- regset(bits,def,class);
- lastclass = 1234;
- continue;
- case 'd':
- for (class = '0'; class <= '9'; class++)
- regset(bits,def,class);
- lastclass = 1234;
- continue;
- case 'D':
- for (class = 0; class < '0'; class++)
- regset(bits,def,class);
- for (class = '9' + 1; class < 256; class++)
- regset(bits,def,class);
- lastclass = 1234;
+ 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;
+ }
+ if (!SIZE_ONLY) {
+ 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);
+ }
+ 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;
+#ifdef ASCIIish
+ case 'e': value = '\033'; break;
+ case 'a': value = '\007'; break;
+#else
+ case 'e': value = '\047'; break;
+ case 'a': value = '\057'; break;
+#endif
case 'x':
- class = scan_hex(regparse, 2, &numlen);
- regparse += numlen;
+ if (*PL_regcomp_parse == '{') {
+ e = strchr(PL_regcomp_parse++, '}');
+ if (!e)
+ FAIL("Missing right brace on \\x{}");
+ value = (UV)scan_hex(PL_regcomp_parse,
+ e - PL_regcomp_parse,
+ &numlen);
+ PL_regcomp_parse = e + 1;
+ }
+ else {
+ value = (UV)scan_hex(PL_regcomp_parse, 2, &numlen);
+ PL_regcomp_parse += numlen;
+ }
break;
case 'c':
- class = *regparse++;
- if (isLOWER(class))
- class = toUPPER(class);
- class ^= 64;
+ 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(--regparse, 3, &numlen);
- regparse += numlen;
+ value = (UV)scan_oct(--PL_regcomp_parse, 3, &numlen);
+ PL_regcomp_parse += numlen;
+ break;
+ default:
+ if (!SIZE_ONLY && ckWARN(WARN_UNSAFE) && isALPHA(value))
+ Perl_warner(aTHX_ WARN_UNSAFE,
+ "/%.127s/: Unrecognized escape \\%c in character class passed through",
+ PL_regprecomp,
+ (int)value);
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 (*regparse == '-' && regparse+1 < regxend &&
- regparse[1] != ']') {
- regparse++;
- 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 */
}
}
- for ( ; lastclass <= class; lastclass++) {
- regset(bits,def,lastclass);
- if (regflags & PMf_FOLD && isUPPER(lastclass))
- regset(bits,def,toLOWER(lastclass));
- }
- lastclass = class;
+ /* 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;
}
- if (*regparse != ']')
- FAIL("unmatched [] in regexp");
- nextchar();
+
+ 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()
+STATIC char*
+S_nextchar(pTHX)
{
- char* retval = regparse++;
+ dTHR;
+ char* retval = PL_regcomp_parse++;
for (;;) {
- if (*regparse == '(' && regparse[1] == '?' &&
- regparse[2] == '#') {
- while (*regparse && *regparse != ')')
- regparse++;
- regparse++;
+ 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(*regparse)) {
- regparse++;
+ if (PL_regflags & PMf_EXTENDED) {
+ if (isSPACE(*PL_regcomp_parse)) {
+ PL_regcomp_parse++;
continue;
}
- else if (*regparse == '#') {
- while (*regparse && *regparse != '\n')
- regparse++;
- regparse++;
+ else if (*PL_regcomp_parse == '#') {
+ while (*PL_regcomp_parse && *PL_regcomp_parse != '\n')
+ PL_regcomp_parse++;
+ PL_regcomp_parse++;
continue;
}
}
}
/*
-- regnode - emit a node
+- reg_node - emit a node
*/
-#ifdef CAN_PROTOTYPE
-static char * /* Location. */
-regnode(char op)
-#else
-static char * /* Location. */
-regnode(op)
-char op;
-#endif
+STATIC regnode * /* Location. */
+S_reg_node(pTHX_ U8 op)
{
- register char *ret;
- register char *ptr;
-
- ret = regcode;
- if (ret == ®dummy) {
-#ifdef REGALIGN
- if (!(regsize & 1))
- regsize++;
-#endif
- regsize += 3;
+ dTHR;
+ register regnode *ret;
+ register regnode *ptr;
+
+ ret = PL_regcode;
+ if (SIZE_ONLY) {
+ SIZE_ALIGN(PL_regsize);
+ PL_regsize += 1;
return(ret);
}
-#ifdef REGALIGN
-#ifndef lint
- if (!((long)ret & 1))
- *ret++ = 127;
-#endif
-#endif
+ NODE_ALIGN_FILL(ret);
ptr = ret;
- *ptr++ = op;
- *ptr++ = '\0'; /* Null "next" pointer. */
- *ptr++ = '\0';
- regcode = ptr;
+ FILL_ADVANCE_NODE(ptr, op);
+ PL_regcode = ptr;
return(ret);
}
/*
- reganode - emit a node with an argument
*/
-#ifdef CAN_PROTOTYPE
-static char * /* Location. */
-reganode(char op, unsigned short arg)
-#else
-static char * /* Location. */
-reganode(op, arg)
-char op;
-unsigned short arg;
-#endif
+STATIC regnode * /* Location. */
+S_reganode(pTHX_ U8 op, U32 arg)
{
- register char *ret;
- register char *ptr;
-
- ret = regcode;
- if (ret == ®dummy) {
-#ifdef REGALIGN
- if (!(regsize & 1))
- regsize++;
-#endif
- regsize += 5;
+ dTHR;
+ register regnode *ret;
+ register regnode *ptr;
+
+ ret = PL_regcode;
+ if (SIZE_ONLY) {
+ SIZE_ALIGN(PL_regsize);
+ PL_regsize += 2;
return(ret);
}
-#ifdef REGALIGN
-#ifndef lint
- if (!((long)ret & 1))
- *ret++ = 127;
-#endif
-#endif
+ NODE_ALIGN_FILL(ret);
ptr = ret;
- *ptr++ = op;
- *ptr++ = '\0'; /* Null "next" pointer. */
- *ptr++ = '\0';
-#ifdef REGALIGN
- *(unsigned short *)(ret+3) = arg;
-#else
- ret[3] = arg >> 8; ret[4] = arg & 0377;
-#endif
- ptr += 2;
- regcode = ptr;
+ FILL_ADVANCE_NODE_ARG(ptr, op, arg);
+ PL_regcode = ptr;
return(ret);
}
/*
-- regc - emit (if appropriate) a byte of code
+- reguni - emit (if appropriate) a Unicode character
*/
-#ifdef CAN_PROTOTYPE
-static void
-regc(char b)
-#else
-static void
-regc(b)
-char b;
-#endif
+STATIC void
+S_reguni(pTHX_ UV uv, char* s, I32* lenp)
{
- if (regcode != ®dummy)
- *regcode++ = b;
+ dTHR;
+ if (SIZE_ONLY) {
+ U8 tmpbuf[10];
+ *lenp = uv_to_utf8(tmpbuf, uv) - tmpbuf;
+ }
else
- regsize++;
+ *lenp = uv_to_utf8((U8*)s, uv) - (U8*)s;
+
}
/*
*
* Means relocating the operand.
*/
-#ifdef CAN_PROTOTYPE
-static void
-reginsert(char op, char *opnd)
-#else
-static void
-reginsert(op, opnd)
-char op;
-char *opnd;
-#endif
+STATIC void
+S_reginsert(pTHX_ U8 op, regnode *opnd)
{
- register char *src;
- register char *dst;
- register char *place;
- register int offset = (regkind[(U8)op] == CURLY ? 4 : 0);
-
- if (regcode == ®dummy) {
-#ifdef REGALIGN
- regsize += 4 + offset;
-#else
- regsize += 3 + offset;
-#endif
+ dTHR;
+ register regnode *src;
+ register regnode *dst;
+ register regnode *place;
+ register int offset = regarglen[(U8)op];
+
+/* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
+
+ if (SIZE_ONLY) {
+ PL_regsize += NODE_STEP_REGNODE + offset;
return;
}
- src = regcode;
-#ifdef REGALIGN
- regcode += 4 + offset;
-#else
- regcode += 3 + offset;
-#endif
- dst = regcode;
+ src = PL_regcode;
+ PL_regcode += NODE_STEP_REGNODE + offset;
+ dst = PL_regcode;
while (src > opnd)
- *--dst = *--src;
+ StructCopy(--src, --dst, regnode);
place = opnd; /* Op node, where operand used to be. */
- *place++ = op;
- *place++ = '\0';
- *place++ = '\0';
- while (offset-- > 0)
- *place++ = '\0';
-#ifdef REGALIGN
- *place++ = '\177';
-#endif
+ src = NEXTOPER(place);
+ FILL_ADVANCE_NODE(place, op);
+ Zero(src, offset, regnode);
}
/*
-- regtail - set the next-pointer at the end of a node chain
+- regtail - set the next-pointer at the end of a node chain of p to val.
*/
-static void
-regtail(p, val)
-char *p;
-char *val;
+STATIC void
+S_regtail(pTHX_ regnode *p, regnode *val)
{
- register char *scan;
- register char *temp;
+ dTHR;
+ register regnode *scan;
+ register regnode *temp;
register I32 offset;
- if (p == ®dummy)
+ if (SIZE_ONLY)
return;
/* Find last node. */
scan = temp;
}
-#ifdef REGALIGN
- offset = val - scan;
-#ifndef lint
- *(short*)(scan+1) = offset;
-#else
- offset = offset;
-#endif
-#else
- if (OP(scan) == BACK)
- offset = scan - val;
- else
- offset = val - scan;
- *(scan+1) = (offset>>8)&0377;
- *(scan+2) = offset&0377;
-#endif
+ if (reg_off_by_arg[OP(scan)]) {
+ ARG_SET(scan, val - scan);
+ }
+ else {
+ NEXT_OFF(scan) = val - scan;
+ }
}
/*
- regoptail - regtail on operand of first argument; nop if operandless
*/
-static void
-regoptail(p, val)
-char *p;
-char *val;
+STATIC void
+S_regoptail(pTHX_ regnode *p, regnode *val)
{
+ dTHR;
/* "Operandless" and "op != BRANCH" are synonymous in practice. */
- if (p == NULL || p == ®dummy || regkind[(U8)OP(p)] != BRANCH)
+ if (p == NULL || SIZE_ONLY)
+ return;
+ if (PL_regkind[(U8)OP(p)] == BRANCH) {
+ regtail(NEXTOPER(p), val);
+ }
+ else if ( PL_regkind[(U8)OP(p)] == BRANCHJ) {
+ regtail(NEXTOPER(NEXTOPER(p)), val);
+ }
+ else
return;
- regtail(NEXTOPER(p), val);
}
/*
- regcurly - a little FSA that accepts {\d+,?\d*}
*/
STATIC I32
-regcurly(s)
-register char *s;
+S_regcurly(pTHX_ register char *s)
{
if (*s++ != '{')
return FALSE;
return TRUE;
}
-#ifdef DEBUGGING
-/*
- - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
- */
-void
-regdump(r)
-regexp *r;
+STATIC regnode *
+S_dumpuntil(pTHX_ regnode *start, regnode *node, regnode *last, SV* sv, I32 l)
{
- register char *s;
- register char op = EXACTLY; /* Arbitrary non-END op. */
- register char *next;
+#ifdef DEBUGGING
+ register U8 op = EXACT; /* Arbitrary non-END op. */
+ register regnode *next, *onode;
+ while (op != END && (!last || node < last)) {
+ /* While that wasn't END last time... */
- s = r->program + 1;
- while (op != END) { /* While that wasn't END last time... */
-#ifdef REGALIGN
- if (!((long)s & 1))
- s++;
-#endif
- op = OP(s);
- fprintf(Perl_debug_log,"%2d%s", s-r->program, regprop(s)); /* Where, what. */
- next = regnext(s);
- s += regarglen[(U8)op];
+ NODE_ALIGN(node);
+ op = OP(node);
+ if (op == CLOSE)
+ l--;
+ next = regnext(node);
+ /* Where, what. */
+ if (OP(node) == OPTIMIZED)
+ goto after_print;
+ regprop(sv, node);
+ PerlIO_printf(Perl_debug_log, "%4"IVdf":%*s%s", (IV)(node - start),
+ (int)(2*l + 1), "", SvPVX(sv));
if (next == NULL) /* Next ptr. */
- fprintf(Perl_debug_log,"(0)");
+ PerlIO_printf(Perl_debug_log, "(0)");
else
- fprintf(Perl_debug_log,"(%d)", (s-r->program)+(next-s));
- s += 3;
- if (op == ANYOF) {
- s += 32;
+ PerlIO_printf(Perl_debug_log, "(%"IVdf")", (IV)(next - start));
+ (void)PerlIO_putc(Perl_debug_log, '\n');
+ after_print:
+ 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);
}
- if (op == EXACTLY) {
- /* Literal string, where present. */
- s++;
- (void)putc(' ', Perl_debug_log);
- (void)putc('<', Perl_debug_log);
- while (*s != '\0') {
- (void)putc(*s, Perl_debug_log);
- s++;
- }
- (void)putc('>', Perl_debug_log);
- s++;
+ else if (PL_regkind[(U8)op] == BRANCH) {
+ node = dumpuntil(start, NEXTOPER(node), next, sv, l + 1);
}
- (void)putc('\n', Perl_debug_log);
+ 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 (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) {
+ node = dumpuntil(start, NEXTOPER(node), NEXTOPER(node) + 1, sv, l + 1);
+ }
+ else if (op == ANYOF) {
+ node = NEXTOPER(node);
+ node += ANYOF_SKIP;
+ }
+ else if (PL_regkind[(U8)op] == EXACT) {
+ /* Literal string, where present. */
+ node += NODE_SZ_STR(node) - 1;
+ node = NEXTOPER(node);
+ }
+ else {
+ node = NEXTOPER(node);
+ node += regarglen[(U8)op];
+ }
+ if (op == CURLYX || op == OPEN)
+ l++;
+ else if (op == WHILEM)
+ l--;
}
+#endif /* DEBUGGING */
+ return node;
+}
+
+/*
+ - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
+ */
+void
+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->regstart)
- fprintf(Perl_debug_log,"start `%s' ", SvPVX(r->regstart));
- if (r->regstclass)
- fprintf(Perl_debug_log,"stclass `%s' ", regprop(r->regstclass));
- if (r->reganch & ROPT_ANCH)
- fprintf(Perl_debug_log,"anchored ");
+ if (r->anchored_substr)
+ PerlIO_printf(Perl_debug_log,
+ "anchored `%s%.*s%s'%s at %"IVdf" ",
+ PL_colors[0],
+ (int)(SvCUR(r->anchored_substr) - (SvTAIL(r->anchored_substr)!=0)),
+ SvPVX(r->anchored_substr),
+ PL_colors[1],
+ SvTAIL(r->anchored_substr) ? "$" : "",
+ (IV)r->anchored_offset);
+ if (r->float_substr)
+ PerlIO_printf(Perl_debug_log,
+ "floating `%s%.*s%s'%s at %"IVdf"..%"UVuf" ",
+ PL_colors[0],
+ (int)(SvCUR(r->float_substr) - (SvTAIL(r->float_substr)!=0)),
+ SvPVX(r->float_substr),
+ PL_colors[1],
+ SvTAIL(r->float_substr) ? "$" : "",
+ (IV)r->float_min_offset, (UV)r->float_max_offset);
+ if (r->check_substr)
+ PerlIO_printf(Perl_debug_log,
+ r->check_substr == r->float_substr
+ ? "(checking floating" : "(checking anchored");
+ if (r->reganch & ROPT_NOSCAN)
+ PerlIO_printf(Perl_debug_log, " noscan");
+ if (r->reganch & ROPT_CHECK_ALL)
+ PerlIO_printf(Perl_debug_log, " isall");
+ if (r->check_substr)
+ PerlIO_printf(Perl_debug_log, ") ");
+
+ if (r->regstclass) {
+ regprop(sv, r->regstclass);
+ PerlIO_printf(Perl_debug_log, "stclass `%s' ", SvPVX(sv));
+ }
+ if (r->reganch & ROPT_ANCH) {
+ PerlIO_printf(Perl_debug_log, "anchored");
+ if (r->reganch & ROPT_ANCH_BOL)
+ 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, ' ');
+ }
+ if (r->reganch & ROPT_GPOS_SEEN)
+ PerlIO_printf(Perl_debug_log, "GPOS ");
if (r->reganch & ROPT_SKIP)
- fprintf(Perl_debug_log,"plus ");
+ PerlIO_printf(Perl_debug_log, "plus ");
if (r->reganch & ROPT_IMPLICIT)
- fprintf(Perl_debug_log,"implicit ");
- if (r->regmust != NULL)
- fprintf(Perl_debug_log,"must have \"%s\" back %ld ", SvPVX(r->regmust),
- (long) r->regback);
- fprintf(Perl_debug_log, "minlen %ld ", (long) r->minlen);
- fprintf(Perl_debug_log,"\n");
+ PerlIO_printf(Perl_debug_log, "implicit ");
+ PerlIO_printf(Perl_debug_log, "minlen %ld ", (long) r->minlen);
+ if (r->reganch & ROPT_EVAL_SEEN)
+ PerlIO_printf(Perl_debug_log, "with eval ");
+ PerlIO_printf(Perl_debug_log, "\n");
+#endif /* DEBUGGING */
+}
+
+STATIC void
+S_put_byte(pTHX_ SV *sv, int c)
+{
+ if (c <= ' ' || c == 127 || c == 255)
+ Perl_sv_catpvf(aTHX_ sv, "\\%o", c);
+ else if (c == '-' || c == ']' || c == '\\' || c == '^')
+ Perl_sv_catpvf(aTHX_ sv, "\\%c", c);
+ else
+ Perl_sv_catpvf(aTHX_ sv, "%c", c);
}
/*
- regprop - printable representation of opcode
*/
-char *
-regprop(op)
-char *op;
+void
+Perl_regprop(pTHX_ SV *sv, regnode *o)
{
- register char *p = 0;
-
- (void) strcpy(buf, ":");
+#ifdef DEBUGGING
+ dTHR;
+ register int k;
- switch (OP(op)) {
- case BOL:
- p = "BOL";
- break;
- case MBOL:
- p = "MBOL";
- break;
- case SBOL:
- p = "SBOL";
- break;
- case EOL:
- p = "EOL";
- 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 EXACTLY:
- p = "EXACTLY";
- break;
- case NOTHING:
- p = "NOTHING";
- break;
- case BACK:
- p = "BACK";
- break;
- case END:
- p = "END";
- break;
- case ALNUM:
- p = "ALNUM";
- break;
- case NALNUM:
- p = "NALNUM";
- break;
- case BOUND:
- p = "BOUND";
- break;
- case NBOUND:
- p = "NBOUND";
- break;
- case SPACE:
- p = "SPACE";
- break;
- case NSPACE:
- p = "NSPACE";
- break;
- case DIGIT:
- p = "DIGIT";
- break;
- case NDIGIT:
- p = "NDIGIT";
- break;
- case CURLY:
- (void)sprintf(buf+strlen(buf), "CURLY {%d,%d}", ARG1(op),ARG2(op));
- p = NULL;
- break;
- case CURLYX:
- (void)sprintf(buf+strlen(buf), "CURLYX {%d,%d}", ARG1(op),ARG2(op));
- p = NULL;
- break;
- case REF:
- (void)sprintf(buf+strlen(buf), "REF%d", ARG1(op));
- p = NULL;
- break;
- case OPEN:
- (void)sprintf(buf+strlen(buf), "OPEN%d", ARG1(op));
- p = NULL;
- break;
- case CLOSE:
- (void)sprintf(buf+strlen(buf), "CLOSE%d", ARG1(op));
- p = NULL;
- break;
- case STAR:
- p = "STAR";
- break;
- case PLUS:
- p = "PLUS";
- break;
- case MINMOD:
- p = "MINMOD";
- break;
- case GBOL:
- p = "GBOL";
- break;
- case UNLESSM:
- p = "UNLESSM";
- break;
- case IFMATCH:
- p = "IFMATCH";
- break;
- case SUCCEED:
- p = "SUCCEED";
- break;
- case WHILEM:
- p = "WHILEM";
- 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));
+ }
+ 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 == ANYOF) {
+ int i, rangestart = -1;
+ const char * const out[] = { /* Should be syncronized with
+ a table in regcomp.h */
+ "\\w",
+ "\\W",
+ "\\s",
+ "\\S",
+ "\\d",
+ "\\D",
+ "[:alnum:]",
+ "[:^alnum:]",
+ "[:alpha:]",
+ "[:^alpha:]",
+ "[:ascii:]",
+ "[:^ascii:]",
+ "[:ctrl:]",
+ "[:^ctrl:]",
+ "[:graph:]",
+ "[:^graph:]",
+ "[:lower:]",
+ "[:^lower:]",
+ "[:print:]",
+ "[:^print:]",
+ "[:punct:]",
+ "[:^punct:]",
+ "[:upper:]",
+ "[:!upper:]",
+ "[:xdigit:]",
+ "[:^xdigit:]"
+ };
+
+ if (o->flags & ANYOF_LOCALE)
+ sv_catpv(sv, "{loc}");
+ if (o->flags & ANYOF_FOLD)
+ sv_catpv(sv, "{i}");
+ Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
+ if (o->flags & ANYOF_INVERT)
+ sv_catpv(sv, "^");
+ for (i = 0; i <= 256; i++) {
+ if (i < 256 && ANYOF_BITMAP_TEST(o,i)) {
+ if (rangestart == -1)
+ rangestart = i;
+ } else if (rangestart != -1) {
+ if (i <= rangestart + 3)
+ for (; rangestart < i; rangestart++)
+ put_byte(sv, rangestart);
+ else {
+ put_byte(sv, rangestart);
+ sv_catpv(sv, "-");
+ put_byte(sv, i - 1);
+ }
+ rangestart = -1;
+ }
+ }
+ if (o->flags & ANYOF_CLASS)
+ for (i = 0; i < sizeof(out)/sizeof(char*); i++)
+ if (ANYOF_CLASS_TEST(o,i))
+ sv_catpv(sv, out[i]);
+ Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
}
- if (p != NULL)
- (void) strcat(buf, p);
- return(buf);
+ 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;
}
-#endif /* DEBUGGING */
void
-pregfree(r)
-struct regexp *r;
+Perl_pregfree(pTHX_ struct regexp *r)
{
- if (!r)
+ dTHR;
+ DEBUG_r(if (!PL_colorset) reginitcolors());
+
+ if (!r || (--r->refcnt > 0))
return;
- if (r->precomp) {
+ 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);
- r->precomp = Nullch;
- }
- if (r->subbase) {
- Safefree(r->subbase);
- r->subbase = Nullch;
+ if (RX_MATCH_COPIED(r))
+ Safefree(r->subbeg);
+ if (r->substrs) {
+ if (r->anchored_substr)
+ SvREFCNT_dec(r->anchored_substr);
+ if (r->float_substr)
+ SvREFCNT_dec(r->float_substr);
+ Safefree(r->substrs);
}
- if (r->regmust) {
- SvREFCNT_dec(r->regmust);
- r->regmust = Nullsv;
- }
- if (r->regstart) {
- SvREFCNT_dec(r->regstart);
- r->regstart = Nullsv;
+ 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 'f':
+ Safefree(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;
+ default:
+ FAIL2("panic: regfree data code '%c'", r->data->what[n]);
+ }
+ }
+ Safefree(r->data->what);
+ Safefree(r->data);
}
Safefree(r->startp);
Safefree(r->endp);
Safefree(r);
}
+
+/*
+ - regnext - dig the "next" pointer out of a node
+ *
+ * [Note, when REGALIGN is defined there are two places in regmatch()
+ * that bypass this code for speed.]
+ */
+regnode *
+Perl_regnext(pTHX_ register regnode *p)
+{
+ dTHR;
+ register I32 offset;
+
+ if (p == &PL_regdummy)
+ return(NULL);
+
+ offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
+ if (offset == 0)
+ return(NULL);
+
+ return(p+offset);
+}
+
+STATIC void
+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)
+ l1 = 510;
+ if (l1 + l2 > 510)
+ l2 = 510 - l1;
+ Copy(pat1, buf, l1 , char);
+ 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);
+#else
+ va_start(args);
+#endif
+ msv = vmess(buf, &args);
+ va_end(args);
+ message = SvPV(msv,l1);
+ if (l1 > 512)
+ l1 = 512;
+ Copy(message, buf, l1 , char);
+ buf[l1] = '\0'; /* Overwrite \n */
+ 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. */
+ SAVEVPTR(PL_regstartp); /* Pointer to startp array. */
+ SAVEVPTR(PL_regendp); /* Ditto for endp. */
+ SAVEVPTR(PL_reglastparen); /* Similarly for lastparen. */
+ SAVEPPTR(PL_regtill); /* How far we are required to go. */
+ SAVEI8(PL_regprev); /* char before regbol, \n if none */
+ SAVEVPTR(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;
+ SAVEVPTR(PL_regdata);
+ SAVEI32(PL_reg_flags); /* from regexec.c */
+ SAVEI32(PL_reg_eval_set); /* from regexec.c */
+ SAVEI32(PL_regnarrate); /* from regexec.c */
+ SAVEVPTR(PL_regprogram); /* from regexec.c */
+ SAVEINT(PL_regindent); /* from regexec.c */
+ SAVEVPTR(PL_regcc); /* from regexec.c */
+ SAVEVPTR(PL_curcop);
+ SAVEVPTR(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? */
+ SAVEVPTR(PL_regcode); /* Code-emit pointer; ®dummy = don't */
+ SAVEPPTR(PL_regxend); /* End of input for compile */
+ SAVEPPTR(PL_regcomp_parse); /* Input-scan pointer. */
+ SAVEVPTR(PL_reg_call_cc); /* from regexec.c */
+ SAVEVPTR(PL_reg_re); /* from regexec.c */
+ SAVEPPTR(PL_reg_ganch); /* from regexec.c */
+ SAVESPTR(PL_reg_sv); /* from regexec.c */
+ SAVEVPTR(PL_reg_magic); /* from regexec.c */
+ SAVEI32(PL_reg_oldpos); /* from regexec.c */
+ SAVEVPTR(PL_reg_oldcurpm); /* from regexec.c */
+ SAVEVPTR(PL_reg_curpm); /* from regexec.c */
+#ifdef DEBUGGING
+ SAVEPPTR(PL_reg_starttry); /* from regexec.c */
+#endif
+}
+
+#ifdef PERL_OBJECT
+#include "XSUB.h"
+#undef this
+#define this pPerl
+#endif
+
+static void
+clear_re(pTHXo_ void *r)
+{
+ ReREFCNT_dec((regexp *)r);
+}
+
projects / p5sagit/p5-mst-13.2.git / blobdiff