1N/A/* regcomp.c
1N/A */
1N/A
1N/A/*
1N/A * "A fair jaw-cracker dwarf-language must be." --Samwise Gamgee
1N/A */
1N/A
1N/A/* NOTE: this is derived from Henry Spencer's regexp code, and should not
1N/A * confused with the original package (see point 3 below). Thanks, Henry!
1N/A */
1N/A
1N/A/* Additional note: this code is very heavily munged from Henry's version
1N/A * in places. In some spots I've traded clarity for efficiency, so don't
1N/A * blame Henry for some of the lack of readability.
1N/A */
1N/A
1N/A/* The names of the functions have been changed from regcomp and
1N/A * regexec to pregcomp and pregexec in order to avoid conflicts
1N/A * with the POSIX routines of the same names.
1N/A*/
1N/A
1N/A#ifdef PERL_EXT_RE_BUILD
1N/A/* need to replace pregcomp et al, so enable that */
1N/A# ifndef PERL_IN_XSUB_RE
1N/A# define PERL_IN_XSUB_RE
1N/A# endif
1N/A/* need access to debugger hooks */
1N/A# if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
1N/A# define DEBUGGING
1N/A# endif
1N/A#endif
1N/A
1N/A#ifdef PERL_IN_XSUB_RE
1N/A/* We *really* need to overwrite these symbols: */
1N/A# define Perl_pregcomp my_regcomp
1N/A# define Perl_regdump my_regdump
1N/A# define Perl_regprop my_regprop
1N/A# define Perl_pregfree my_regfree
1N/A# define Perl_re_intuit_string my_re_intuit_string
1N/A/* *These* symbols are masked to allow static link. */
1N/A# define Perl_regnext my_regnext
1N/A# define Perl_save_re_context my_save_re_context
1N/A# define Perl_reginitcolors my_reginitcolors
1N/A
1N/A# define PERL_NO_GET_CONTEXT
1N/A#endif
1N/A
1N/A/*SUPPRESS 112*/
1N/A/*
1N/A * pregcomp and pregexec -- regsub and regerror are not used in perl
1N/A *
1N/A * Copyright (c) 1986 by University of Toronto.
1N/A * Written by Henry Spencer. Not derived from licensed software.
1N/A *
1N/A * Permission is granted to anyone to use this software for any
1N/A * purpose on any computer system, and to redistribute it freely,
1N/A * subject to the following restrictions:
1N/A *
1N/A * 1. The author is not responsible for the consequences of use of
1N/A * this software, no matter how awful, even if they arise
1N/A * from defects in it.
1N/A *
1N/A * 2. The origin of this software must not be misrepresented, either
1N/A * by explicit claim or by omission.
1N/A *
1N/A * 3. Altered versions must be plainly marked as such, and must not
1N/A * be misrepresented as being the original software.
1N/A *
1N/A *
1N/A **** Alterations to Henry's code are...
1N/A ****
1N/A **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
1N/A **** 2000, 2001, 2002, 2003, by Larry Wall and others
1N/A ****
1N/A **** You may distribute under the terms of either the GNU General Public
1N/A **** License or the Artistic License, as specified in the README file.
1N/A
1N/A *
1N/A * Beware that some of this code is subtly aware of the way operator
1N/A * precedence is structured in regular expressions. Serious changes in
1N/A * regular-expression syntax might require a total rethink.
1N/A */
1N/A#include "EXTERN.h"
1N/A#define PERL_IN_REGCOMP_C
1N/A#include "perl.h"
1N/A
1N/A#ifndef PERL_IN_XSUB_RE
1N/A# include "INTERN.h"
1N/A#endif
1N/A
1N/A#define REG_COMP_C
1N/A#include "regcomp.h"
1N/A
1N/A#ifdef op
1N/A#undef op
1N/A#endif /* op */
1N/A
1N/A#ifdef MSDOS
1N/A# if defined(BUGGY_MSC6)
1N/A /* MSC 6.00A breaks on op/regexp.t test 85 unless we turn this off */
1N/A# pragma optimize("a",off)
1N/A /* But MSC 6.00A is happy with 'w', for aliases only across function calls*/
1N/A# pragma optimize("w",on )
1N/A# endif /* BUGGY_MSC6 */
1N/A#endif /* MSDOS */
1N/A
1N/A#ifndef STATIC
1N/A#define STATIC static
1N/A#endif
1N/A
1N/Atypedef struct RExC_state_t {
1N/A U32 flags; /* are we folding, multilining? */
1N/A char *precomp; /* uncompiled string. */
1N/A regexp *rx;
1N/A char *start; /* Start of input for compile */
1N/A char *end; /* End of input for compile */
1N/A char *parse; /* Input-scan pointer. */
1N/A I32 whilem_seen; /* number of WHILEM in this expr */
1N/A regnode *emit_start; /* Start of emitted-code area */
1N/A regnode *emit; /* Code-emit pointer; &regdummy = don't = compiling */
1N/A I32 naughty; /* How bad is this pattern? */
1N/A I32 sawback; /* Did we see \1, ...? */
1N/A U32 seen;
1N/A I32 size; /* Code size. */
1N/A I32 npar; /* () count. */
1N/A I32 extralen;
1N/A I32 seen_zerolen;
1N/A I32 seen_evals;
1N/A I32 utf8; /* whether the pattern is utf8 or not */
1N/A I32 orig_utf8; /* whether the pattern was originally in utf8 */
1N/A /* XXX use this for future optimisation of case
1N/A * where pattern must be upgraded to utf8. */
1N/A#if ADD_TO_REGEXEC
1N/A char *starttry; /* -Dr: where regtry was called. */
1N/A#define RExC_starttry (pRExC_state->starttry)
1N/A#endif
1N/A} RExC_state_t;
1N/A
1N/A#define RExC_flags (pRExC_state->flags)
1N/A#define RExC_precomp (pRExC_state->precomp)
1N/A#define RExC_rx (pRExC_state->rx)
1N/A#define RExC_start (pRExC_state->start)
1N/A#define RExC_end (pRExC_state->end)
1N/A#define RExC_parse (pRExC_state->parse)
1N/A#define RExC_whilem_seen (pRExC_state->whilem_seen)
1N/A#define RExC_offsets (pRExC_state->rx->offsets) /* I am not like the others */
1N/A#define RExC_emit (pRExC_state->emit)
1N/A#define RExC_emit_start (pRExC_state->emit_start)
1N/A#define RExC_naughty (pRExC_state->naughty)
1N/A#define RExC_sawback (pRExC_state->sawback)
1N/A#define RExC_seen (pRExC_state->seen)
1N/A#define RExC_size (pRExC_state->size)
1N/A#define RExC_npar (pRExC_state->npar)
1N/A#define RExC_extralen (pRExC_state->extralen)
1N/A#define RExC_seen_zerolen (pRExC_state->seen_zerolen)
1N/A#define RExC_seen_evals (pRExC_state->seen_evals)
1N/A#define RExC_utf8 (pRExC_state->utf8)
1N/A#define RExC_orig_utf8 (pRExC_state->orig_utf8)
1N/A
1N/A#define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
1N/A#define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
1N/A ((*s) == '{' && regcurly(s)))
1N/A
1N/A#ifdef SPSTART
1N/A#undef SPSTART /* dratted cpp namespace... */
1N/A#endif
1N/A/*
1N/A * Flags to be passed up and down.
1N/A */
1N/A#define WORST 0 /* Worst case. */
1N/A#define HASWIDTH 0x1 /* Known to match non-null strings. */
1N/A#define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */
1N/A#define SPSTART 0x4 /* Starts with * or +. */
1N/A#define TRYAGAIN 0x8 /* Weeded out a declaration. */
1N/A
1N/A/* Length of a variant. */
1N/A
1N/Atypedef struct scan_data_t {
1N/A I32 len_min;
1N/A I32 len_delta;
1N/A I32 pos_min;
1N/A I32 pos_delta;
1N/A SV *last_found;
1N/A I32 last_end; /* min value, <0 unless valid. */
1N/A I32 last_start_min;
1N/A I32 last_start_max;
1N/A SV **longest; /* Either &l_fixed, or &l_float. */
1N/A SV *longest_fixed;
1N/A I32 offset_fixed;
1N/A SV *longest_float;
1N/A I32 offset_float_min;
1N/A I32 offset_float_max;
1N/A I32 flags;
1N/A I32 whilem_c;
1N/A I32 *last_closep;
1N/A struct regnode_charclass_class *start_class;
1N/A} scan_data_t;
1N/A
1N/A/*
1N/A * Forward declarations for pregcomp()'s friends.
1N/A */
1N/A
1N/Astatic scan_data_t zero_scan_data = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1N/A 0, 0, 0, 0, 0, 0};
1N/A
1N/A#define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
1N/A#define SF_BEFORE_SEOL 0x1
1N/A#define SF_BEFORE_MEOL 0x2
1N/A#define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
1N/A#define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
1N/A
1N/A#ifdef NO_UNARY_PLUS
1N/A# define SF_FIX_SHIFT_EOL (0+2)
1N/A# define SF_FL_SHIFT_EOL (0+4)
1N/A#else
1N/A# define SF_FIX_SHIFT_EOL (+2)
1N/A# define SF_FL_SHIFT_EOL (+4)
1N/A#endif
1N/A
1N/A#define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
1N/A#define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
1N/A
1N/A#define SF_FL_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL)
1N/A#define SF_FL_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */
1N/A#define SF_IS_INF 0x40
1N/A#define SF_HAS_PAR 0x80
1N/A#define SF_IN_PAR 0x100
1N/A#define SF_HAS_EVAL 0x200
1N/A#define SCF_DO_SUBSTR 0x400
1N/A#define SCF_DO_STCLASS_AND 0x0800
1N/A#define SCF_DO_STCLASS_OR 0x1000
1N/A#define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
1N/A#define SCF_WHILEM_VISITED_POS 0x2000
1N/A
1N/A#define UTF (RExC_utf8 != 0)
1N/A#define LOC ((RExC_flags & PMf_LOCALE) != 0)
1N/A#define FOLD ((RExC_flags & PMf_FOLD) != 0)
1N/A
1N/A#define OOB_UNICODE 12345678
1N/A#define OOB_NAMEDCLASS -1
1N/A
1N/A#define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
1N/A#define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b)
1N/A
1N/A
1N/A/* length of regex to show in messages that don't mark a position within */
1N/A#define RegexLengthToShowInErrorMessages 127
1N/A
1N/A/*
1N/A * If MARKER[12] are adjusted, be sure to adjust the constants at the top
1N/A * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
1N/A * op/pragma/warn/regcomp.
1N/A */
1N/A#define MARKER1 "<-- HERE" /* marker as it appears in the description */
1N/A#define MARKER2 " <-- HERE " /* marker as it appears within the regex */
1N/A
1N/A#define REPORT_LOCATION " in regex; marked by " MARKER1 " in m/%.*s" MARKER2 "%s/"
1N/A
1N/A/*
1N/A * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
1N/A * arg. Show regex, up to a maximum length. If it's too long, chop and add
1N/A * "...".
1N/A */
1N/A#define FAIL(msg) STMT_START { \
1N/A char *ellipses = ""; \
1N/A IV len = RExC_end - RExC_precomp; \
1N/A \
1N/A if (!SIZE_ONLY) \
1N/A SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
1N/A if (len > RegexLengthToShowInErrorMessages) { \
1N/A /* chop 10 shorter than the max, to ensure meaning of "..." */ \
1N/A len = RegexLengthToShowInErrorMessages - 10; \
1N/A ellipses = "..."; \
1N/A } \
1N/A Perl_croak(aTHX_ "%s in regex m/%.*s%s/", \
1N/A msg, (int)len, RExC_precomp, ellipses); \
1N/A} STMT_END
1N/A
1N/A/*
1N/A * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
1N/A * args. Show regex, up to a maximum length. If it's too long, chop and add
1N/A * "...".
1N/A */
1N/A#define FAIL2(pat,msg) STMT_START { \
1N/A char *ellipses = ""; \
1N/A IV len = RExC_end - RExC_precomp; \
1N/A \
1N/A if (!SIZE_ONLY) \
1N/A SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
1N/A if (len > RegexLengthToShowInErrorMessages) { \
1N/A /* chop 10 shorter than the max, to ensure meaning of "..." */ \
1N/A len = RegexLengthToShowInErrorMessages - 10; \
1N/A ellipses = "..."; \
1N/A } \
1N/A S_re_croak2(aTHX_ pat, " in regex m/%.*s%s/", \
1N/A msg, (int)len, RExC_precomp, ellipses); \
1N/A} STMT_END
1N/A
1N/A
1N/A/*
1N/A * Simple_vFAIL -- like FAIL, but marks the current location in the scan
1N/A */
1N/A#define Simple_vFAIL(m) STMT_START { \
1N/A IV offset = RExC_parse - RExC_precomp; \
1N/A Perl_croak(aTHX_ "%s" REPORT_LOCATION, \
1N/A m, (int)offset, RExC_precomp, RExC_precomp + offset); \
1N/A} STMT_END
1N/A
1N/A/*
1N/A * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
1N/A */
1N/A#define vFAIL(m) STMT_START { \
1N/A if (!SIZE_ONLY) \
1N/A SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
1N/A Simple_vFAIL(m); \
1N/A} STMT_END
1N/A
1N/A/*
1N/A * Like Simple_vFAIL(), but accepts two arguments.
1N/A */
1N/A#define Simple_vFAIL2(m,a1) STMT_START { \
1N/A IV offset = RExC_parse - RExC_precomp; \
1N/A S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, \
1N/A (int)offset, RExC_precomp, RExC_precomp + offset); \
1N/A} STMT_END
1N/A
1N/A/*
1N/A * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
1N/A */
1N/A#define vFAIL2(m,a1) STMT_START { \
1N/A if (!SIZE_ONLY) \
1N/A SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
1N/A Simple_vFAIL2(m, a1); \
1N/A} STMT_END
1N/A
1N/A
1N/A/*
1N/A * Like Simple_vFAIL(), but accepts three arguments.
1N/A */
1N/A#define Simple_vFAIL3(m, a1, a2) STMT_START { \
1N/A IV offset = RExC_parse - RExC_precomp; \
1N/A S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, \
1N/A (int)offset, RExC_precomp, RExC_precomp + offset); \
1N/A} STMT_END
1N/A
1N/A/*
1N/A * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
1N/A */
1N/A#define vFAIL3(m,a1,a2) STMT_START { \
1N/A if (!SIZE_ONLY) \
1N/A SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
1N/A Simple_vFAIL3(m, a1, a2); \
1N/A} STMT_END
1N/A
1N/A/*
1N/A * Like Simple_vFAIL(), but accepts four arguments.
1N/A */
1N/A#define Simple_vFAIL4(m, a1, a2, a3) STMT_START { \
1N/A IV offset = RExC_parse - RExC_precomp; \
1N/A S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, a3, \
1N/A (int)offset, RExC_precomp, RExC_precomp + offset); \
1N/A} STMT_END
1N/A
1N/A/*
1N/A * Like Simple_vFAIL(), but accepts five arguments.
1N/A */
1N/A#define Simple_vFAIL5(m, a1, a2, a3, a4) STMT_START { \
1N/A IV offset = RExC_parse - RExC_precomp; \
1N/A S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, a3, a4, \
1N/A (int)offset, RExC_precomp, RExC_precomp + offset); \
1N/A} STMT_END
1N/A
1N/A
1N/A#define vWARN(loc,m) STMT_START { \
1N/A IV offset = loc - RExC_precomp; \
1N/A Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s" REPORT_LOCATION, \
1N/A m, (int)offset, RExC_precomp, RExC_precomp + offset); \
1N/A} STMT_END
1N/A
1N/A#define vWARNdep(loc,m) STMT_START { \
1N/A IV offset = loc - RExC_precomp; \
1N/A Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
1N/A "%s" REPORT_LOCATION, \
1N/A m, (int)offset, RExC_precomp, RExC_precomp + offset); \
1N/A} STMT_END
1N/A
1N/A
1N/A#define vWARN2(loc, m, a1) STMT_START { \
1N/A IV offset = loc - RExC_precomp; \
1N/A Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
1N/A a1, (int)offset, RExC_precomp, RExC_precomp + offset); \
1N/A} STMT_END
1N/A
1N/A#define vWARN3(loc, m, a1, a2) STMT_START { \
1N/A IV offset = loc - RExC_precomp; \
1N/A Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
1N/A a1, a2, (int)offset, RExC_precomp, RExC_precomp + offset); \
1N/A} STMT_END
1N/A
1N/A#define vWARN4(loc, m, a1, a2, a3) STMT_START { \
1N/A IV offset = loc - RExC_precomp; \
1N/A Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
1N/A a1, a2, a3, (int)offset, RExC_precomp, RExC_precomp + offset); \
1N/A} STMT_END
1N/A
1N/A#define vWARN5(loc, m, a1, a2, a3, a4) STMT_START { \
1N/A IV offset = loc - RExC_precomp; \
1N/A Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
1N/A a1, a2, a3, a4, (int)offset, RExC_precomp, RExC_precomp + offset); \
1N/A} STMT_END
1N/A
1N/A
1N/A/* Allow for side effects in s */
1N/A#define REGC(c,s) STMT_START { \
1N/A if (!SIZE_ONLY) *(s) = (c); else (void)(s); \
1N/A} STMT_END
1N/A
1N/A/* Macros for recording node offsets. 20001227 mjd@plover.com
1N/A * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in
1N/A * element 2*n-1 of the array. Element #2n holds the byte length node #n.
1N/A * Element 0 holds the number n.
1N/A */
1N/A
1N/A#define MJD_OFFSET_DEBUG(x)
1N/A/* #define MJD_OFFSET_DEBUG(x) Perl_warn_nocontext x */
1N/A
1N/A
1N/A#define Set_Node_Offset_To_R(node,byte) STMT_START { \
1N/A if (! SIZE_ONLY) { \
1N/A MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \
1N/A __LINE__, (node), (byte))); \
1N/A if((node) < 0) { \
1N/A Perl_croak(aTHX_ "value of node is %d in Offset macro", node); \
1N/A } else { \
1N/A RExC_offsets[2*(node)-1] = (byte); \
1N/A } \
1N/A } \
1N/A} STMT_END
1N/A
1N/A#define Set_Node_Offset(node,byte) \
1N/A Set_Node_Offset_To_R((node)-RExC_emit_start, (byte)-RExC_start)
1N/A#define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse)
1N/A
1N/A#define Set_Node_Length_To_R(node,len) STMT_START { \
1N/A if (! SIZE_ONLY) { \
1N/A MJD_OFFSET_DEBUG(("** (%d) size of node %d is %d.\n", \
1N/A __LINE__, (node), (len))); \
1N/A if((node) < 0) { \
1N/A Perl_croak(aTHX_ "value of node is %d in Length macro", node); \
1N/A } else { \
1N/A RExC_offsets[2*(node)] = (len); \
1N/A } \
1N/A } \
1N/A} STMT_END
1N/A
1N/A#define Set_Node_Length(node,len) \
1N/A Set_Node_Length_To_R((node)-RExC_emit_start, len)
1N/A#define Set_Cur_Node_Length(len) Set_Node_Length(RExC_emit, len)
1N/A#define Set_Node_Cur_Length(node) \
1N/A Set_Node_Length(node, RExC_parse - parse_start)
1N/A
1N/A/* Get offsets and lengths */
1N/A#define Node_Offset(n) (RExC_offsets[2*((n)-RExC_emit_start)-1])
1N/A#define Node_Length(n) (RExC_offsets[2*((n)-RExC_emit_start)])
1N/A
1N/Astatic void clear_re(pTHX_ void *r);
1N/A
1N/A/* Mark that we cannot extend a found fixed substring at this point.
1N/A Updata the longest found anchored substring and the longest found
1N/A floating substrings if needed. */
1N/A
1N/ASTATIC void
1N/AS_scan_commit(pTHX_ RExC_state_t *pRExC_state, scan_data_t *data)
1N/A{
1N/A STRLEN l = CHR_SVLEN(data->last_found);
1N/A STRLEN old_l = CHR_SVLEN(*data->longest);
1N/A
1N/A if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
1N/A SvSetMagicSV(*data->longest, data->last_found);
1N/A if (*data->longest == data->longest_fixed) {
1N/A data->offset_fixed = l ? data->last_start_min : data->pos_min;
1N/A if (data->flags & SF_BEFORE_EOL)
1N/A data->flags
1N/A |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL);
1N/A else
1N/A data->flags &= ~SF_FIX_BEFORE_EOL;
1N/A }
1N/A else {
1N/A data->offset_float_min = l ? data->last_start_min : data->pos_min;
1N/A data->offset_float_max = (l
1N/A ? data->last_start_max
1N/A : data->pos_min + data->pos_delta);
1N/A if ((U32)data->offset_float_max > (U32)I32_MAX)
1N/A data->offset_float_max = I32_MAX;
1N/A if (data->flags & SF_BEFORE_EOL)
1N/A data->flags
1N/A |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL);
1N/A else
1N/A data->flags &= ~SF_FL_BEFORE_EOL;
1N/A }
1N/A }
1N/A SvCUR_set(data->last_found, 0);
1N/A {
1N/A SV * sv = data->last_found;
1N/A MAGIC *mg =
1N/A SvUTF8(sv) && SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
1N/A if (mg && mg->mg_len > 0)
1N/A mg->mg_len = 0;
1N/A }
1N/A data->last_end = -1;
1N/A data->flags &= ~SF_BEFORE_EOL;
1N/A}
1N/A
1N/A/* Can match anything (initialization) */
1N/ASTATIC void
1N/AS_cl_anything(pTHX_ RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
1N/A{
1N/A ANYOF_CLASS_ZERO(cl);
1N/A ANYOF_BITMAP_SETALL(cl);
1N/A cl->flags = ANYOF_EOS|ANYOF_UNICODE_ALL;
1N/A if (LOC)
1N/A cl->flags |= ANYOF_LOCALE;
1N/A}
1N/A
1N/A/* Can match anything (initialization) */
1N/ASTATIC int
1N/AS_cl_is_anything(pTHX_ struct regnode_charclass_class *cl)
1N/A{
1N/A int value;
1N/A
1N/A for (value = 0; value <= ANYOF_MAX; value += 2)
1N/A if (ANYOF_CLASS_TEST(cl, value) && ANYOF_CLASS_TEST(cl, value + 1))
1N/A return 1;
1N/A if (!(cl->flags & ANYOF_UNICODE_ALL))
1N/A return 0;
1N/A if (!ANYOF_BITMAP_TESTALLSET(cl))
1N/A return 0;
1N/A return 1;
1N/A}
1N/A
1N/A/* Can match anything (initialization) */
1N/ASTATIC void
1N/AS_cl_init(pTHX_ RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
1N/A{
1N/A Zero(cl, 1, struct regnode_charclass_class);
1N/A cl->type = ANYOF;
1N/A cl_anything(pRExC_state, cl);
1N/A}
1N/A
1N/ASTATIC void
1N/AS_cl_init_zero(pTHX_ RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
1N/A{
1N/A Zero(cl, 1, struct regnode_charclass_class);
1N/A cl->type = ANYOF;
1N/A cl_anything(pRExC_state, cl);
1N/A if (LOC)
1N/A cl->flags |= ANYOF_LOCALE;
1N/A}
1N/A
1N/A/* 'And' a given class with another one. Can create false positives */
1N/A/* We assume that cl is not inverted */
1N/ASTATIC void
1N/AS_cl_and(pTHX_ struct regnode_charclass_class *cl,
1N/A struct regnode_charclass_class *and_with)
1N/A{
1N/A if (!(and_with->flags & ANYOF_CLASS)
1N/A && !(cl->flags & ANYOF_CLASS)
1N/A && (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
1N/A && !(and_with->flags & ANYOF_FOLD)
1N/A && !(cl->flags & ANYOF_FOLD)) {
1N/A int i;
1N/A
1N/A if (and_with->flags & ANYOF_INVERT)
1N/A for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
1N/A cl->bitmap[i] &= ~and_with->bitmap[i];
1N/A else
1N/A for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
1N/A cl->bitmap[i] &= and_with->bitmap[i];
1N/A } /* XXXX: logic is complicated otherwise, leave it along for a moment. */
1N/A if (!(and_with->flags & ANYOF_EOS))
1N/A cl->flags &= ~ANYOF_EOS;
1N/A
1N/A if (cl->flags & ANYOF_UNICODE_ALL && and_with->flags & ANYOF_UNICODE &&
1N/A !(and_with->flags & ANYOF_INVERT)) {
1N/A cl->flags &= ~ANYOF_UNICODE_ALL;
1N/A cl->flags |= ANYOF_UNICODE;
1N/A ARG_SET(cl, ARG(and_with));
1N/A }
1N/A if (!(and_with->flags & ANYOF_UNICODE_ALL) &&
1N/A !(and_with->flags & ANYOF_INVERT))
1N/A cl->flags &= ~ANYOF_UNICODE_ALL;
1N/A if (!(and_with->flags & (ANYOF_UNICODE|ANYOF_UNICODE_ALL)) &&
1N/A !(and_with->flags & ANYOF_INVERT))
1N/A cl->flags &= ~ANYOF_UNICODE;
1N/A}
1N/A
1N/A/* 'OR' a given class with another one. Can create false positives */
1N/A/* We assume that cl is not inverted */
1N/ASTATIC void
1N/AS_cl_or(pTHX_ RExC_state_t *pRExC_state, struct regnode_charclass_class *cl, struct regnode_charclass_class *or_with)
1N/A{
1N/A if (or_with->flags & ANYOF_INVERT) {
1N/A /* We do not use
1N/A * (B1 | CL1) | (!B2 & !CL2) = (B1 | !B2 & !CL2) | (CL1 | (!B2 & !CL2))
1N/A * <= (B1 | !B2) | (CL1 | !CL2)
1N/A * which is wasteful if CL2 is small, but we ignore CL2:
1N/A * (B1 | CL1) | (!B2 & !CL2) <= (B1 | CL1) | !B2 = (B1 | !B2) | CL1
1N/A * XXXX Can we handle case-fold? Unclear:
1N/A * (OK1(i) | OK1(i')) | !(OK1(i) | OK1(i')) =
1N/A * (OK1(i) | OK1(i')) | (!OK1(i) & !OK1(i'))
1N/A */
1N/A if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
1N/A && !(or_with->flags & ANYOF_FOLD)
1N/A && !(cl->flags & ANYOF_FOLD) ) {
1N/A int i;
1N/A
1N/A for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
1N/A cl->bitmap[i] |= ~or_with->bitmap[i];
1N/A } /* XXXX: logic is complicated otherwise */
1N/A else {
1N/A cl_anything(pRExC_state, cl);
1N/A }
1N/A } else {
1N/A /* (B1 | CL1) | (B2 | CL2) = (B1 | B2) | (CL1 | CL2)) */
1N/A if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
1N/A && (!(or_with->flags & ANYOF_FOLD)
1N/A || (cl->flags & ANYOF_FOLD)) ) {
1N/A int i;
1N/A
1N/A /* OR char bitmap and class bitmap separately */
1N/A for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
1N/A cl->bitmap[i] |= or_with->bitmap[i];
1N/A if (or_with->flags & ANYOF_CLASS) {
1N/A for (i = 0; i < ANYOF_CLASSBITMAP_SIZE; i++)
1N/A cl->classflags[i] |= or_with->classflags[i];
1N/A cl->flags |= ANYOF_CLASS;
1N/A }
1N/A }
1N/A else { /* XXXX: logic is complicated, leave it along for a moment. */
1N/A cl_anything(pRExC_state, cl);
1N/A }
1N/A }
1N/A if (or_with->flags & ANYOF_EOS)
1N/A cl->flags |= ANYOF_EOS;
1N/A
1N/A if (cl->flags & ANYOF_UNICODE && or_with->flags & ANYOF_UNICODE &&
1N/A ARG(cl) != ARG(or_with)) {
1N/A cl->flags |= ANYOF_UNICODE_ALL;
1N/A cl->flags &= ~ANYOF_UNICODE;
1N/A }
1N/A if (or_with->flags & ANYOF_UNICODE_ALL) {
1N/A cl->flags |= ANYOF_UNICODE_ALL;
1N/A cl->flags &= ~ANYOF_UNICODE;
1N/A }
1N/A}
1N/A
1N/A/*
1N/A * There are strange code-generation bugs caused on sparc64 by gcc-2.95.2.
1N/A * These need to be revisited when a newer toolchain becomes available.
1N/A */
1N/A#if defined(__sparc64__) && defined(__GNUC__)
1N/A# if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 96)
1N/A# undef SPARC64_GCC_WORKAROUND
1N/A# define SPARC64_GCC_WORKAROUND 1
1N/A# endif
1N/A#endif
1N/A
1N/A/* REx optimizer. Converts nodes into quickier variants "in place".
1N/A Finds fixed substrings. */
1N/A
1N/A/* Stops at toplevel WHILEM as well as at `last'. At end *scanp is set
1N/A to the position after last scanned or to NULL. */
1N/A
1N/ASTATIC I32
1N/AS_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 flags)
1N/A /* scanp: Start here (read-write). */
1N/A /* deltap: Write maxlen-minlen here. */
1N/A /* last: Stop before this one. */
1N/A{
1N/A I32 min = 0, pars = 0, code;
1N/A regnode *scan = *scanp, *next;
1N/A I32 delta = 0;
1N/A int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF);
1N/A int is_inf_internal = 0; /* The studied chunk is infinite */
1N/A I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
1N/A scan_data_t data_fake;
1N/A struct regnode_charclass_class and_with; /* Valid if flags & SCF_DO_STCLASS_OR */
1N/A
1N/A while (scan && OP(scan) != END && scan < last) {
1N/A /* Peephole optimizer: */
1N/A
1N/A if (PL_regkind[(U8)OP(scan)] == EXACT) {
1N/A /* Merge several consecutive EXACTish nodes into one. */
1N/A regnode *n = regnext(scan);
1N/A U32 stringok = 1;
1N/A#ifdef DEBUGGING
1N/A regnode *stop = scan;
1N/A#endif
1N/A
1N/A next = scan + NODE_SZ_STR(scan);
1N/A /* Skip NOTHING, merge EXACT*. */
1N/A while (n &&
1N/A ( PL_regkind[(U8)OP(n)] == NOTHING ||
1N/A (stringok && (OP(n) == OP(scan))))
1N/A && NEXT_OFF(n)
1N/A && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX) {
1N/A if (OP(n) == TAIL || n > next)
1N/A stringok = 0;
1N/A if (PL_regkind[(U8)OP(n)] == NOTHING) {
1N/A NEXT_OFF(scan) += NEXT_OFF(n);
1N/A next = n + NODE_STEP_REGNODE;
1N/A#ifdef DEBUGGING
1N/A if (stringok)
1N/A stop = n;
1N/A#endif
1N/A n = regnext(n);
1N/A }
1N/A else if (stringok) {
1N/A int oldl = STR_LEN(scan);
1N/A regnode *nnext = regnext(n);
1N/A
1N/A if (oldl + STR_LEN(n) > U8_MAX)
1N/A break;
1N/A NEXT_OFF(scan) += NEXT_OFF(n);
1N/A STR_LEN(scan) += STR_LEN(n);
1N/A next = n + NODE_SZ_STR(n);
1N/A /* Now we can overwrite *n : */
1N/A Move(STRING(n), STRING(scan) + oldl, STR_LEN(n), char);
1N/A#ifdef DEBUGGING
1N/A stop = next - 1;
1N/A#endif
1N/A n = nnext;
1N/A }
1N/A }
1N/A
1N/A if (UTF && OP(scan) == EXACTF && STR_LEN(scan) >= 6) {
1N/A/*
1N/A Two problematic code points in Unicode casefolding of EXACT nodes:
1N/A
1N/A U+0390 - GREEK SMALL LETTER IOTA WITH DIALYTIKA AND TONOS
1N/A U+03B0 - GREEK SMALL LETTER UPSILON WITH DIALYTIKA AND TONOS
1N/A
1N/A which casefold to
1N/A
1N/A Unicode UTF-8
1N/A
1N/A U+03B9 U+0308 U+0301 0xCE 0xB9 0xCC 0x88 0xCC 0x81
1N/A U+03C5 U+0308 U+0301 0xCF 0x85 0xCC 0x88 0xCC 0x81
1N/A
1N/A This means that in case-insensitive matching (or "loose matching",
1N/A as Unicode calls it), an EXACTF of length six (the UTF-8 encoded byte
1N/A length of the above casefolded versions) can match a target string
1N/A of length two (the byte length of UTF-8 encoded U+0390 or U+03B0).
1N/A This would rather mess up the minimum length computation.
1N/A
1N/A What we'll do is to look for the tail four bytes, and then peek
1N/A at the preceding two bytes to see whether we need to decrease
1N/A the minimum length by four (six minus two).
1N/A
1N/A Thanks to the design of UTF-8, there cannot be false matches:
1N/A A sequence of valid UTF-8 bytes cannot be a subsequence of
1N/A another valid sequence of UTF-8 bytes.
1N/A
1N/A*/
1N/A char *s0 = STRING(scan), *s, *t;
1N/A char *s1 = s0 + STR_LEN(scan) - 1, *s2 = s1 - 4;
1N/A char *t0 = "\xcc\x88\xcc\x81";
1N/A char *t1 = t0 + 3;
1N/A
1N/A for (s = s0 + 2;
1N/A s < s2 && (t = ninstr(s, s1, t0, t1));
1N/A s = t + 4) {
1N/A if (((U8)t[-1] == 0xB9 && (U8)t[-2] == 0xCE) ||
1N/A ((U8)t[-1] == 0x85 && (U8)t[-2] == 0xCF))
1N/A min -= 4;
1N/A }
1N/A }
1N/A
1N/A#ifdef DEBUGGING
1N/A /* Allow dumping */
1N/A n = scan + NODE_SZ_STR(scan);
1N/A while (n <= stop) {
1N/A if (PL_regkind[(U8)OP(n)] != NOTHING || OP(n) == NOTHING) {
1N/A OP(n) = OPTIMIZED;
1N/A NEXT_OFF(n) = 0;
1N/A }
1N/A n++;
1N/A }
1N/A#endif
1N/A }
1N/A /* Follow the next-chain of the current node and optimize
1N/A away all the NOTHINGs from it. */
1N/A if (OP(scan) != CURLYX) {
1N/A int max = (reg_off_by_arg[OP(scan)]
1N/A ? I32_MAX
1N/A /* I32 may be smaller than U16 on CRAYs! */
1N/A : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
1N/A int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
1N/A int noff;
1N/A regnode *n = scan;
1N/A
1N/A /* Skip NOTHING and LONGJMP. */
1N/A while ((n = regnext(n))
1N/A && ((PL_regkind[(U8)OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
1N/A || ((OP(n) == LONGJMP) && (noff = ARG(n))))
1N/A && off + noff < max)
1N/A off += noff;
1N/A if (reg_off_by_arg[OP(scan)])
1N/A ARG(scan) = off;
1N/A else
1N/A NEXT_OFF(scan) = off;
1N/A }
1N/A /* The principal pseudo-switch. Cannot be a switch, since we
1N/A look into several different things. */
1N/A if (OP(scan) == BRANCH || OP(scan) == BRANCHJ
1N/A || OP(scan) == IFTHEN || OP(scan) == SUSPEND) {
1N/A next = regnext(scan);
1N/A code = OP(scan);
1N/A
1N/A if (OP(next) == code || code == IFTHEN || code == SUSPEND) {
1N/A I32 max1 = 0, min1 = I32_MAX, num = 0;
1N/A struct regnode_charclass_class accum;
1N/A
1N/A if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */
1N/A scan_commit(pRExC_state, data); /* Cannot merge strings after this. */
1N/A if (flags & SCF_DO_STCLASS)
1N/A cl_init_zero(pRExC_state, &accum);
1N/A while (OP(scan) == code) {
1N/A I32 deltanext, minnext, f = 0, fake;
1N/A struct regnode_charclass_class this_class;
1N/A
1N/A num++;
1N/A data_fake.flags = 0;
1N/A if (data) {
1N/A data_fake.whilem_c = data->whilem_c;
1N/A data_fake.last_closep = data->last_closep;
1N/A }
1N/A else
1N/A data_fake.last_closep = &fake;
1N/A next = regnext(scan);
1N/A scan = NEXTOPER(scan);
1N/A if (code != BRANCH)
1N/A scan = NEXTOPER(scan);
1N/A if (flags & SCF_DO_STCLASS) {
1N/A cl_init(pRExC_state, &this_class);
1N/A data_fake.start_class = &this_class;
1N/A f = SCF_DO_STCLASS_AND;
1N/A }
1N/A if (flags & SCF_WHILEM_VISITED_POS)
1N/A f |= SCF_WHILEM_VISITED_POS;
1N/A /* we suppose the run is continuous, last=next...*/
1N/A minnext = study_chunk(pRExC_state, &scan, &deltanext,
1N/A next, &data_fake, f);
1N/A if (min1 > minnext)
1N/A min1 = minnext;
1N/A if (max1 < minnext + deltanext)
1N/A max1 = minnext + deltanext;
1N/A if (deltanext == I32_MAX)
1N/A is_inf = is_inf_internal = 1;
1N/A scan = next;
1N/A if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
1N/A pars++;
1N/A if (data && (data_fake.flags & SF_HAS_EVAL))
1N/A data->flags |= SF_HAS_EVAL;
1N/A if (data)
1N/A data->whilem_c = data_fake.whilem_c;
1N/A if (flags & SCF_DO_STCLASS)
1N/A cl_or(pRExC_state, &accum, &this_class);
1N/A if (code == SUSPEND)
1N/A break;
1N/A }
1N/A if (code == IFTHEN && num < 2) /* Empty ELSE branch */
1N/A min1 = 0;
1N/A if (flags & SCF_DO_SUBSTR) {
1N/A data->pos_min += min1;
1N/A data->pos_delta += max1 - min1;
1N/A if (max1 != min1 || is_inf)
1N/A data->longest = &(data->longest_float);
1N/A }
1N/A min += min1;
1N/A delta += max1 - min1;
1N/A if (flags & SCF_DO_STCLASS_OR) {
1N/A cl_or(pRExC_state, data->start_class, &accum);
1N/A if (min1) {
1N/A cl_and(data->start_class, &and_with);
1N/A flags &= ~SCF_DO_STCLASS;
1N/A }
1N/A }
1N/A else if (flags & SCF_DO_STCLASS_AND) {
1N/A if (min1) {
1N/A cl_and(data->start_class, &accum);
1N/A flags &= ~SCF_DO_STCLASS;
1N/A }
1N/A else {
1N/A /* Switch to OR mode: cache the old value of
1N/A * data->start_class */
1N/A StructCopy(data->start_class, &and_with,
1N/A struct regnode_charclass_class);
1N/A flags &= ~SCF_DO_STCLASS_AND;
1N/A StructCopy(&accum, data->start_class,
1N/A struct regnode_charclass_class);
1N/A flags |= SCF_DO_STCLASS_OR;
1N/A data->start_class->flags |= ANYOF_EOS;
1N/A }
1N/A }
1N/A }
1N/A else if (code == BRANCHJ) /* single branch is optimized. */
1N/A scan = NEXTOPER(NEXTOPER(scan));
1N/A else /* single branch is optimized. */
1N/A scan = NEXTOPER(scan);
1N/A continue;
1N/A }
1N/A else if (OP(scan) == EXACT) {
1N/A I32 l = STR_LEN(scan);
1N/A UV uc = *((U8*)STRING(scan));
1N/A if (UTF) {
1N/A U8 *s = (U8*)STRING(scan);
1N/A l = utf8_length(s, s + l);
1N/A uc = utf8_to_uvchr(s, NULL);
1N/A }
1N/A min += l;
1N/A if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */
1N/A /* The code below prefers earlier match for fixed
1N/A offset, later match for variable offset. */
1N/A if (data->last_end == -1) { /* Update the start info. */
1N/A data->last_start_min = data->pos_min;
1N/A data->last_start_max = is_inf
1N/A ? I32_MAX : data->pos_min + data->pos_delta;
1N/A }
1N/A sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan));
1N/A {
1N/A SV * sv = data->last_found;
1N/A MAGIC *mg = SvUTF8(sv) && SvMAGICAL(sv) ?
1N/A mg_find(sv, PERL_MAGIC_utf8) : NULL;
1N/A if (mg && mg->mg_len >= 0)
1N/A mg->mg_len += utf8_length((U8*)STRING(scan),
1N/A (U8*)STRING(scan)+STR_LEN(scan));
1N/A }
1N/A if (UTF)
1N/A SvUTF8_on(data->last_found);
1N/A data->last_end = data->pos_min + l;
1N/A data->pos_min += l; /* As in the first entry. */
1N/A data->flags &= ~SF_BEFORE_EOL;
1N/A }
1N/A if (flags & SCF_DO_STCLASS_AND) {
1N/A /* Check whether it is compatible with what we know already! */
1N/A int compat = 1;
1N/A
1N/A if (uc >= 0x100 ||
1N/A (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
1N/A && !ANYOF_BITMAP_TEST(data->start_class, uc)
1N/A && (!(data->start_class->flags & ANYOF_FOLD)
1N/A || !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc])))
1N/A )
1N/A compat = 0;
1N/A ANYOF_CLASS_ZERO(data->start_class);
1N/A ANYOF_BITMAP_ZERO(data->start_class);
1N/A if (compat)
1N/A ANYOF_BITMAP_SET(data->start_class, uc);
1N/A data->start_class->flags &= ~ANYOF_EOS;
1N/A if (uc < 0x100)
1N/A data->start_class->flags &= ~ANYOF_UNICODE_ALL;
1N/A }
1N/A else if (flags & SCF_DO_STCLASS_OR) {
1N/A /* false positive possible if the class is case-folded */
1N/A if (uc < 0x100)
1N/A ANYOF_BITMAP_SET(data->start_class, uc);
1N/A else
1N/A data->start_class->flags |= ANYOF_UNICODE_ALL;
1N/A data->start_class->flags &= ~ANYOF_EOS;
1N/A cl_and(data->start_class, &and_with);
1N/A }
1N/A flags &= ~SCF_DO_STCLASS;
1N/A }
1N/A else if (PL_regkind[(U8)OP(scan)] == EXACT) { /* But OP != EXACT! */
1N/A I32 l = STR_LEN(scan);
1N/A UV uc = *((U8*)STRING(scan));
1N/A
1N/A /* Search for fixed substrings supports EXACT only. */
1N/A if (flags & SCF_DO_SUBSTR)
1N/A scan_commit(pRExC_state, data);
1N/A if (UTF) {
1N/A U8 *s = (U8 *)STRING(scan);
1N/A l = utf8_length(s, s + l);
1N/A uc = utf8_to_uvchr(s, NULL);
1N/A }
1N/A min += l;
1N/A if (data && (flags & SCF_DO_SUBSTR))
1N/A data->pos_min += l;
1N/A if (flags & SCF_DO_STCLASS_AND) {
1N/A /* Check whether it is compatible with what we know already! */
1N/A int compat = 1;
1N/A
1N/A if (uc >= 0x100 ||
1N/A (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
1N/A && !ANYOF_BITMAP_TEST(data->start_class, uc)
1N/A && !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc])))
1N/A compat = 0;
1N/A ANYOF_CLASS_ZERO(data->start_class);
1N/A ANYOF_BITMAP_ZERO(data->start_class);
1N/A if (compat) {
1N/A ANYOF_BITMAP_SET(data->start_class, uc);
1N/A data->start_class->flags &= ~ANYOF_EOS;
1N/A data->start_class->flags |= ANYOF_FOLD;
1N/A if (OP(scan) == EXACTFL)
1N/A data->start_class->flags |= ANYOF_LOCALE;
1N/A }
1N/A }
1N/A else if (flags & SCF_DO_STCLASS_OR) {
1N/A if (data->start_class->flags & ANYOF_FOLD) {
1N/A /* false positive possible if the class is case-folded.
1N/A Assume that the locale settings are the same... */
1N/A if (uc < 0x100)
1N/A ANYOF_BITMAP_SET(data->start_class, uc);
1N/A data->start_class->flags &= ~ANYOF_EOS;
1N/A }
1N/A cl_and(data->start_class, &and_with);
1N/A }
1N/A flags &= ~SCF_DO_STCLASS;
1N/A }
1N/A else if (strchr((char*)PL_varies,OP(scan))) {
1N/A I32 mincount, maxcount, minnext, deltanext, fl = 0;
1N/A I32 f = flags, pos_before = 0;
1N/A regnode *oscan = scan;
1N/A struct regnode_charclass_class this_class;
1N/A struct regnode_charclass_class *oclass = NULL;
1N/A I32 next_is_eval = 0;
1N/A
1N/A switch (PL_regkind[(U8)OP(scan)]) {
1N/A case WHILEM: /* End of (?:...)* . */
1N/A scan = NEXTOPER(scan);
1N/A goto finish;
1N/A case PLUS:
1N/A if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
1N/A next = NEXTOPER(scan);
1N/A if (OP(next) == EXACT || (flags & SCF_DO_STCLASS)) {
1N/A mincount = 1;
1N/A maxcount = REG_INFTY;
1N/A next = regnext(scan);
1N/A scan = NEXTOPER(scan);
1N/A goto do_curly;
1N/A }
1N/A }
1N/A if (flags & SCF_DO_SUBSTR)
1N/A data->pos_min++;
1N/A min++;
1N/A /* Fall through. */
1N/A case STAR:
1N/A if (flags & SCF_DO_STCLASS) {
1N/A mincount = 0;
1N/A maxcount = REG_INFTY;
1N/A next = regnext(scan);
1N/A scan = NEXTOPER(scan);
1N/A goto do_curly;
1N/A }
1N/A is_inf = is_inf_internal = 1;
1N/A scan = regnext(scan);
1N/A if (flags & SCF_DO_SUBSTR) {
1N/A scan_commit(pRExC_state, data); /* Cannot extend fixed substrings */
1N/A data->longest = &(data->longest_float);
1N/A }
1N/A goto optimize_curly_tail;
1N/A case CURLY:
1N/A mincount = ARG1(scan);
1N/A maxcount = ARG2(scan);
1N/A next = regnext(scan);
1N/A if (OP(scan) == CURLYX) {
1N/A I32 lp = (data ? *(data->last_closep) : 0);
1N/A
1N/A scan->flags = ((lp <= U8_MAX) ? lp : U8_MAX);
1N/A }
1N/A scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
1N/A next_is_eval = (OP(scan) == EVAL);
1N/A do_curly:
1N/A if (flags & SCF_DO_SUBSTR) {
1N/A if (mincount == 0) scan_commit(pRExC_state,data); /* Cannot extend fixed substrings */
1N/A pos_before = data->pos_min;
1N/A }
1N/A if (data) {
1N/A fl = data->flags;
1N/A data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
1N/A if (is_inf)
1N/A data->flags |= SF_IS_INF;
1N/A }
1N/A if (flags & SCF_DO_STCLASS) {
1N/A cl_init(pRExC_state, &this_class);
1N/A oclass = data->start_class;
1N/A data->start_class = &this_class;
1N/A f |= SCF_DO_STCLASS_AND;
1N/A f &= ~SCF_DO_STCLASS_OR;
1N/A }
1N/A /* These are the cases when once a subexpression
1N/A fails at a particular position, it cannot succeed
1N/A even after backtracking at the enclosing scope.
1N/A
1N/A XXXX what if minimal match and we are at the
1N/A initial run of {n,m}? */
1N/A if ((mincount != maxcount - 1) && (maxcount != REG_INFTY))
1N/A f &= ~SCF_WHILEM_VISITED_POS;
1N/A
1N/A /* This will finish on WHILEM, setting scan, or on NULL: */
1N/A minnext = study_chunk(pRExC_state, &scan, &deltanext, last, data,
1N/A mincount == 0
1N/A ? (f & ~SCF_DO_SUBSTR) : f);
1N/A
1N/A if (flags & SCF_DO_STCLASS)
1N/A data->start_class = oclass;
1N/A if (mincount == 0 || minnext == 0) {
1N/A if (flags & SCF_DO_STCLASS_OR) {
1N/A cl_or(pRExC_state, data->start_class, &this_class);
1N/A }
1N/A else if (flags & SCF_DO_STCLASS_AND) {
1N/A /* Switch to OR mode: cache the old value of
1N/A * data->start_class */
1N/A StructCopy(data->start_class, &and_with,
1N/A struct regnode_charclass_class);
1N/A flags &= ~SCF_DO_STCLASS_AND;
1N/A StructCopy(&this_class, data->start_class,
1N/A struct regnode_charclass_class);
1N/A flags |= SCF_DO_STCLASS_OR;
1N/A data->start_class->flags |= ANYOF_EOS;
1N/A }
1N/A } else { /* Non-zero len */
1N/A if (flags & SCF_DO_STCLASS_OR) {
1N/A cl_or(pRExC_state, data->start_class, &this_class);
1N/A cl_and(data->start_class, &and_with);
1N/A }
1N/A else if (flags & SCF_DO_STCLASS_AND)
1N/A cl_and(data->start_class, &this_class);
1N/A flags &= ~SCF_DO_STCLASS;
1N/A }
1N/A if (!scan) /* It was not CURLYX, but CURLY. */
1N/A scan = next;
1N/A if (ckWARN(WARN_REGEXP)
1N/A /* ? quantifier ok, except for (?{ ... }) */
1N/A && (next_is_eval || !(mincount == 0 && maxcount == 1))
1N/A && (minnext == 0) && (deltanext == 0)
1N/A && data && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
1N/A && maxcount <= REG_INFTY/3) /* Complement check for big count */
1N/A {
1N/A vWARN(RExC_parse,
1N/A "Quantifier unexpected on zero-length expression");
1N/A }
1N/A
1N/A min += minnext * mincount;
1N/A is_inf_internal |= ((maxcount == REG_INFTY
1N/A && (minnext + deltanext) > 0)
1N/A || deltanext == I32_MAX);
1N/A is_inf |= is_inf_internal;
1N/A delta += (minnext + deltanext) * maxcount - minnext * mincount;
1N/A
1N/A /* Try powerful optimization CURLYX => CURLYN. */
1N/A if ( OP(oscan) == CURLYX && data
1N/A && data->flags & SF_IN_PAR
1N/A && !(data->flags & SF_HAS_EVAL)
1N/A && !deltanext && minnext == 1 ) {
1N/A /* Try to optimize to CURLYN. */
1N/A regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS;
1N/A regnode *nxt1 = nxt;
1N/A#ifdef DEBUGGING
1N/A regnode *nxt2;
1N/A#endif
1N/A
1N/A /* Skip open. */
1N/A nxt = regnext(nxt);
1N/A if (!strchr((char*)PL_simple,OP(nxt))
1N/A && !(PL_regkind[(U8)OP(nxt)] == EXACT
1N/A && STR_LEN(nxt) == 1))
1N/A goto nogo;
1N/A#ifdef DEBUGGING
1N/A nxt2 = nxt;
1N/A#endif
1N/A nxt = regnext(nxt);
1N/A if (OP(nxt) != CLOSE)
1N/A goto nogo;
1N/A /* Now we know that nxt2 is the only contents: */
1N/A oscan->flags = (U8)ARG(nxt);
1N/A OP(oscan) = CURLYN;
1N/A OP(nxt1) = NOTHING; /* was OPEN. */
1N/A#ifdef DEBUGGING
1N/A OP(nxt1 + 1) = OPTIMIZED; /* was count. */
1N/A NEXT_OFF(nxt1+ 1) = 0; /* just for consistancy. */
1N/A NEXT_OFF(nxt2) = 0; /* just for consistancy with CURLY. */
1N/A OP(nxt) = OPTIMIZED; /* was CLOSE. */
1N/A OP(nxt + 1) = OPTIMIZED; /* was count. */
1N/A NEXT_OFF(nxt+ 1) = 0; /* just for consistancy. */
1N/A#endif
1N/A }
1N/A nogo:
1N/A
1N/A /* Try optimization CURLYX => CURLYM. */
1N/A if ( OP(oscan) == CURLYX && data
1N/A && !(data->flags & SF_HAS_PAR)
1N/A && !(data->flags & SF_HAS_EVAL)
1N/A && !deltanext ) {
1N/A /* XXXX How to optimize if data == 0? */
1N/A /* Optimize to a simpler form. */
1N/A regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */
1N/A regnode *nxt2;
1N/A
1N/A OP(oscan) = CURLYM;
1N/A while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/
1N/A && (OP(nxt2) != WHILEM))
1N/A nxt = nxt2;
1N/A OP(nxt2) = SUCCEED; /* Whas WHILEM */
1N/A /* Need to optimize away parenths. */
1N/A if (data->flags & SF_IN_PAR) {
1N/A /* Set the parenth number. */
1N/A regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/
1N/A
1N/A if (OP(nxt) != CLOSE)
1N/A FAIL("Panic opt close");
1N/A oscan->flags = (U8)ARG(nxt);
1N/A OP(nxt1) = OPTIMIZED; /* was OPEN. */
1N/A OP(nxt) = OPTIMIZED; /* was CLOSE. */
1N/A#ifdef DEBUGGING
1N/A OP(nxt1 + 1) = OPTIMIZED; /* was count. */
1N/A OP(nxt + 1) = OPTIMIZED; /* was count. */
1N/A NEXT_OFF(nxt1 + 1) = 0; /* just for consistancy. */
1N/A NEXT_OFF(nxt + 1) = 0; /* just for consistancy. */
1N/A#endif
1N/A#if 0
1N/A while ( nxt1 && (OP(nxt1) != WHILEM)) {
1N/A regnode *nnxt = regnext(nxt1);
1N/A
1N/A if (nnxt == nxt) {
1N/A if (reg_off_by_arg[OP(nxt1)])
1N/A ARG_SET(nxt1, nxt2 - nxt1);
1N/A else if (nxt2 - nxt1 < U16_MAX)
1N/A NEXT_OFF(nxt1) = nxt2 - nxt1;
1N/A else
1N/A OP(nxt) = NOTHING; /* Cannot beautify */
1N/A }
1N/A nxt1 = nnxt;
1N/A }
1N/A#endif
1N/A /* Optimize again: */
1N/A study_chunk(pRExC_state, &nxt1, &deltanext, nxt,
1N/A NULL, 0);
1N/A }
1N/A else
1N/A oscan->flags = 0;
1N/A }
1N/A else if ((OP(oscan) == CURLYX)
1N/A && (flags & SCF_WHILEM_VISITED_POS)
1N/A /* See the comment on a similar expression above.
1N/A However, this time it not a subexpression
1N/A we care about, but the expression itself. */
1N/A && (maxcount == REG_INFTY)
1N/A && data && ++data->whilem_c < 16) {
1N/A /* This stays as CURLYX, we can put the count/of pair. */
1N/A /* Find WHILEM (as in regexec.c) */
1N/A regnode *nxt = oscan + NEXT_OFF(oscan);
1N/A
1N/A if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */
1N/A nxt += ARG(nxt);
1N/A PREVOPER(nxt)->flags = (U8)(data->whilem_c
1N/A | (RExC_whilem_seen << 4)); /* On WHILEM */
1N/A }
1N/A if (data && fl & (SF_HAS_PAR|SF_IN_PAR))
1N/A pars++;
1N/A if (flags & SCF_DO_SUBSTR) {
1N/A SV *last_str = Nullsv;
1N/A int counted = mincount != 0;
1N/A
1N/A if (data->last_end > 0 && mincount != 0) { /* Ends with a string. */
1N/A#if defined(SPARC64_GCC_WORKAROUND)
1N/A I32 b = 0;
1N/A STRLEN l = 0;
1N/A char *s = NULL;
1N/A I32 old = 0;
1N/A
1N/A if (pos_before >= data->last_start_min)
1N/A b = pos_before;
1N/A else
1N/A b = data->last_start_min;
1N/A
1N/A l = 0;
1N/A s = SvPV(data->last_found, l);
1N/A old = b - data->last_start_min;
1N/A
1N/A#else
1N/A I32 b = pos_before >= data->last_start_min
1N/A ? pos_before : data->last_start_min;
1N/A STRLEN l;
1N/A char *s = SvPV(data->last_found, l);
1N/A I32 old = b - data->last_start_min;
1N/A#endif
1N/A
1N/A if (UTF)
1N/A old = utf8_hop((U8*)s, old) - (U8*)s;
1N/A
1N/A l -= old;
1N/A /* Get the added string: */
1N/A last_str = newSVpvn(s + old, l);
1N/A if (UTF)
1N/A SvUTF8_on(last_str);
1N/A if (deltanext == 0 && pos_before == b) {
1N/A /* What was added is a constant string */
1N/A if (mincount > 1) {
1N/A SvGROW(last_str, (mincount * l) + 1);
1N/A repeatcpy(SvPVX(last_str) + l,
1N/A SvPVX(last_str), l, mincount - 1);
1N/A SvCUR(last_str) *= mincount;
1N/A /* Add additional parts. */
1N/A SvCUR_set(data->last_found,
1N/A SvCUR(data->last_found) - l);
1N/A sv_catsv(data->last_found, last_str);
1N/A {
1N/A SV * sv = data->last_found;
1N/A MAGIC *mg =
1N/A SvUTF8(sv) && SvMAGICAL(sv) ?
1N/A mg_find(sv, PERL_MAGIC_utf8) : NULL;
1N/A if (mg && mg->mg_len >= 0)
1N/A mg->mg_len += CHR_SVLEN(last_str);
1N/A }
1N/A data->last_end += l * (mincount - 1);
1N/A }
1N/A } else {
1N/A /* start offset must point into the last copy */
1N/A data->last_start_min += minnext * (mincount - 1);
1N/A data->last_start_max += is_inf ? I32_MAX
1N/A : (maxcount - 1) * (minnext + data->pos_delta);
1N/A }
1N/A }
1N/A /* It is counted once already... */
1N/A data->pos_min += minnext * (mincount - counted);
1N/A data->pos_delta += - counted * deltanext +
1N/A (minnext + deltanext) * maxcount - minnext * mincount;
1N/A if (mincount != maxcount) {
1N/A /* Cannot extend fixed substrings found inside
1N/A the group. */
1N/A scan_commit(pRExC_state,data);
1N/A if (mincount && last_str) {
1N/A sv_setsv(data->last_found, last_str);
1N/A data->last_end = data->pos_min;
1N/A data->last_start_min =
1N/A data->pos_min - CHR_SVLEN(last_str);
1N/A data->last_start_max = is_inf
1N/A ? I32_MAX
1N/A : data->pos_min + data->pos_delta
1N/A - CHR_SVLEN(last_str);
1N/A }
1N/A data->longest = &(data->longest_float);
1N/A }
1N/A SvREFCNT_dec(last_str);
1N/A }
1N/A if (data && (fl & SF_HAS_EVAL))
1N/A data->flags |= SF_HAS_EVAL;
1N/A optimize_curly_tail:
1N/A if (OP(oscan) != CURLYX) {
1N/A while (PL_regkind[(U8)OP(next = regnext(oscan))] == NOTHING
1N/A && NEXT_OFF(next))
1N/A NEXT_OFF(oscan) += NEXT_OFF(next);
1N/A }
1N/A continue;
1N/A default: /* REF and CLUMP only? */
1N/A if (flags & SCF_DO_SUBSTR) {
1N/A scan_commit(pRExC_state,data); /* Cannot expect anything... */
1N/A data->longest = &(data->longest_float);
1N/A }
1N/A is_inf = is_inf_internal = 1;
1N/A if (flags & SCF_DO_STCLASS_OR)
1N/A cl_anything(pRExC_state, data->start_class);
1N/A flags &= ~SCF_DO_STCLASS;
1N/A break;
1N/A }
1N/A }
1N/A else if (strchr((char*)PL_simple,OP(scan))) {
1N/A int value = 0;
1N/A
1N/A if (flags & SCF_DO_SUBSTR) {
1N/A scan_commit(pRExC_state,data);
1N/A data->pos_min++;
1N/A }
1N/A min++;
1N/A if (flags & SCF_DO_STCLASS) {
1N/A data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
1N/A
1N/A /* Some of the logic below assumes that switching
1N/A locale on will only add false positives. */
1N/A switch (PL_regkind[(U8)OP(scan)]) {
1N/A case SANY:
1N/A default:
1N/A do_default:
1N/A /* Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d", OP(scan)); */
1N/A if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
1N/A cl_anything(pRExC_state, data->start_class);
1N/A break;
1N/A case REG_ANY:
1N/A if (OP(scan) == SANY)
1N/A goto do_default;
1N/A if (flags & SCF_DO_STCLASS_OR) { /* Everything but \n */
1N/A value = (ANYOF_BITMAP_TEST(data->start_class,'\n')
1N/A || (data->start_class->flags & ANYOF_CLASS));
1N/A cl_anything(pRExC_state, data->start_class);
1N/A }
1N/A if (flags & SCF_DO_STCLASS_AND || !value)
1N/A ANYOF_BITMAP_CLEAR(data->start_class,'\n');
1N/A break;
1N/A case ANYOF:
1N/A if (flags & SCF_DO_STCLASS_AND)
1N/A cl_and(data->start_class,
1N/A (struct regnode_charclass_class*)scan);
1N/A else
1N/A cl_or(pRExC_state, data->start_class,
1N/A (struct regnode_charclass_class*)scan);
1N/A break;
1N/A case ALNUM:
1N/A if (flags & SCF_DO_STCLASS_AND) {
1N/A if (!(data->start_class->flags & ANYOF_LOCALE)) {
1N/A ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
1N/A for (value = 0; value < 256; value++)
1N/A if (!isALNUM(value))
1N/A ANYOF_BITMAP_CLEAR(data->start_class, value);
1N/A }
1N/A }
1N/A else {
1N/A if (data->start_class->flags & ANYOF_LOCALE)
1N/A ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isALNUM(value))
1N/A ANYOF_BITMAP_SET(data->start_class, value);
1N/A }
1N/A }
1N/A break;
1N/A case ALNUML:
1N/A if (flags & SCF_DO_STCLASS_AND) {
1N/A if (data->start_class->flags & ANYOF_LOCALE)
1N/A ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
1N/A }
1N/A else {
1N/A ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
1N/A data->start_class->flags |= ANYOF_LOCALE;
1N/A }
1N/A break;
1N/A case NALNUM:
1N/A if (flags & SCF_DO_STCLASS_AND) {
1N/A if (!(data->start_class->flags & ANYOF_LOCALE)) {
1N/A ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
1N/A for (value = 0; value < 256; value++)
1N/A if (isALNUM(value))
1N/A ANYOF_BITMAP_CLEAR(data->start_class, value);
1N/A }
1N/A }
1N/A else {
1N/A if (data->start_class->flags & ANYOF_LOCALE)
1N/A ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isALNUM(value))
1N/A ANYOF_BITMAP_SET(data->start_class, value);
1N/A }
1N/A }
1N/A break;
1N/A case NALNUML:
1N/A if (flags & SCF_DO_STCLASS_AND) {
1N/A if (data->start_class->flags & ANYOF_LOCALE)
1N/A ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
1N/A }
1N/A else {
1N/A data->start_class->flags |= ANYOF_LOCALE;
1N/A ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
1N/A }
1N/A break;
1N/A case SPACE:
1N/A if (flags & SCF_DO_STCLASS_AND) {
1N/A if (!(data->start_class->flags & ANYOF_LOCALE)) {
1N/A ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
1N/A for (value = 0; value < 256; value++)
1N/A if (!isSPACE(value))
1N/A ANYOF_BITMAP_CLEAR(data->start_class, value);
1N/A }
1N/A }
1N/A else {
1N/A if (data->start_class->flags & ANYOF_LOCALE)
1N/A ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isSPACE(value))
1N/A ANYOF_BITMAP_SET(data->start_class, value);
1N/A }
1N/A }
1N/A break;
1N/A case SPACEL:
1N/A if (flags & SCF_DO_STCLASS_AND) {
1N/A if (data->start_class->flags & ANYOF_LOCALE)
1N/A ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
1N/A }
1N/A else {
1N/A data->start_class->flags |= ANYOF_LOCALE;
1N/A ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
1N/A }
1N/A break;
1N/A case NSPACE:
1N/A if (flags & SCF_DO_STCLASS_AND) {
1N/A if (!(data->start_class->flags & ANYOF_LOCALE)) {
1N/A ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
1N/A for (value = 0; value < 256; value++)
1N/A if (isSPACE(value))
1N/A ANYOF_BITMAP_CLEAR(data->start_class, value);
1N/A }
1N/A }
1N/A else {
1N/A if (data->start_class->flags & ANYOF_LOCALE)
1N/A ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isSPACE(value))
1N/A ANYOF_BITMAP_SET(data->start_class, value);
1N/A }
1N/A }
1N/A break;
1N/A case NSPACEL:
1N/A if (flags & SCF_DO_STCLASS_AND) {
1N/A if (data->start_class->flags & ANYOF_LOCALE) {
1N/A ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
1N/A for (value = 0; value < 256; value++)
1N/A if (!isSPACE(value))
1N/A ANYOF_BITMAP_CLEAR(data->start_class, value);
1N/A }
1N/A }
1N/A else {
1N/A data->start_class->flags |= ANYOF_LOCALE;
1N/A ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
1N/A }
1N/A break;
1N/A case DIGIT:
1N/A if (flags & SCF_DO_STCLASS_AND) {
1N/A ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NDIGIT);
1N/A for (value = 0; value < 256; value++)
1N/A if (!isDIGIT(value))
1N/A ANYOF_BITMAP_CLEAR(data->start_class, value);
1N/A }
1N/A else {
1N/A if (data->start_class->flags & ANYOF_LOCALE)
1N/A ANYOF_CLASS_SET(data->start_class,ANYOF_DIGIT);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isDIGIT(value))
1N/A ANYOF_BITMAP_SET(data->start_class, value);
1N/A }
1N/A }
1N/A break;
1N/A case NDIGIT:
1N/A if (flags & SCF_DO_STCLASS_AND) {
1N/A ANYOF_CLASS_CLEAR(data->start_class,ANYOF_DIGIT);
1N/A for (value = 0; value < 256; value++)
1N/A if (isDIGIT(value))
1N/A ANYOF_BITMAP_CLEAR(data->start_class, value);
1N/A }
1N/A else {
1N/A if (data->start_class->flags & ANYOF_LOCALE)
1N/A ANYOF_CLASS_SET(data->start_class,ANYOF_NDIGIT);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isDIGIT(value))
1N/A ANYOF_BITMAP_SET(data->start_class, value);
1N/A }
1N/A }
1N/A break;
1N/A }
1N/A if (flags & SCF_DO_STCLASS_OR)
1N/A cl_and(data->start_class, &and_with);
1N/A flags &= ~SCF_DO_STCLASS;
1N/A }
1N/A }
1N/A else if (PL_regkind[(U8)OP(scan)] == EOL && flags & SCF_DO_SUBSTR) {
1N/A data->flags |= (OP(scan) == MEOL
1N/A ? SF_BEFORE_MEOL
1N/A : SF_BEFORE_SEOL);
1N/A }
1N/A else if ( PL_regkind[(U8)OP(scan)] == BRANCHJ
1N/A /* Lookbehind, or need to calculate parens/evals/stclass: */
1N/A && (scan->flags || data || (flags & SCF_DO_STCLASS))
1N/A && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) {
1N/A /* Lookahead/lookbehind */
1N/A I32 deltanext, minnext, fake = 0;
1N/A regnode *nscan;
1N/A struct regnode_charclass_class intrnl;
1N/A int f = 0;
1N/A
1N/A data_fake.flags = 0;
1N/A if (data) {
1N/A data_fake.whilem_c = data->whilem_c;
1N/A data_fake.last_closep = data->last_closep;
1N/A }
1N/A else
1N/A data_fake.last_closep = &fake;
1N/A if ( flags & SCF_DO_STCLASS && !scan->flags
1N/A && OP(scan) == IFMATCH ) { /* Lookahead */
1N/A cl_init(pRExC_state, &intrnl);
1N/A data_fake.start_class = &intrnl;
1N/A f |= SCF_DO_STCLASS_AND;
1N/A }
1N/A if (flags & SCF_WHILEM_VISITED_POS)
1N/A f |= SCF_WHILEM_VISITED_POS;
1N/A next = regnext(scan);
1N/A nscan = NEXTOPER(NEXTOPER(scan));
1N/A minnext = study_chunk(pRExC_state, &nscan, &deltanext, last, &data_fake, f);
1N/A if (scan->flags) {
1N/A if (deltanext) {
1N/A vFAIL("Variable length lookbehind not implemented");
1N/A }
1N/A else if (minnext > U8_MAX) {
1N/A vFAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
1N/A }
1N/A scan->flags = (U8)minnext;
1N/A }
1N/A if (data && data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
1N/A pars++;
1N/A if (data && (data_fake.flags & SF_HAS_EVAL))
1N/A data->flags |= SF_HAS_EVAL;
1N/A if (data)
1N/A data->whilem_c = data_fake.whilem_c;
1N/A if (f & SCF_DO_STCLASS_AND) {
1N/A int was = (data->start_class->flags & ANYOF_EOS);
1N/A
1N/A cl_and(data->start_class, &intrnl);
1N/A if (was)
1N/A data->start_class->flags |= ANYOF_EOS;
1N/A }
1N/A }
1N/A else if (OP(scan) == OPEN) {
1N/A pars++;
1N/A }
1N/A else if (OP(scan) == CLOSE) {
1N/A if ((I32)ARG(scan) == is_par) {
1N/A next = regnext(scan);
1N/A
1N/A if ( next && (OP(next) != WHILEM) && next < last)
1N/A is_par = 0; /* Disable optimization */
1N/A }
1N/A if (data)
1N/A *(data->last_closep) = ARG(scan);
1N/A }
1N/A else if (OP(scan) == EVAL) {
1N/A if (data)
1N/A data->flags |= SF_HAS_EVAL;
1N/A }
1N/A else if (OP(scan) == LOGICAL && scan->flags == 2) { /* Embedded follows */
1N/A if (flags & SCF_DO_SUBSTR) {
1N/A scan_commit(pRExC_state,data);
1N/A data->longest = &(data->longest_float);
1N/A }
1N/A is_inf = is_inf_internal = 1;
1N/A if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
1N/A cl_anything(pRExC_state, data->start_class);
1N/A flags &= ~SCF_DO_STCLASS;
1N/A }
1N/A /* Else: zero-length, ignore. */
1N/A scan = regnext(scan);
1N/A }
1N/A
1N/A finish:
1N/A *scanp = scan;
1N/A *deltap = is_inf_internal ? I32_MAX : delta;
1N/A if (flags & SCF_DO_SUBSTR && is_inf)
1N/A data->pos_delta = I32_MAX - data->pos_min;
1N/A if (is_par > U8_MAX)
1N/A is_par = 0;
1N/A if (is_par && pars==1 && data) {
1N/A data->flags |= SF_IN_PAR;
1N/A data->flags &= ~SF_HAS_PAR;
1N/A }
1N/A else if (pars && data) {
1N/A data->flags |= SF_HAS_PAR;
1N/A data->flags &= ~SF_IN_PAR;
1N/A }
1N/A if (flags & SCF_DO_STCLASS_OR)
1N/A cl_and(data->start_class, &and_with);
1N/A return min;
1N/A}
1N/A
1N/ASTATIC I32
1N/AS_add_data(pTHX_ RExC_state_t *pRExC_state, I32 n, char *s)
1N/A{
1N/A if (RExC_rx->data) {
1N/A Renewc(RExC_rx->data,
1N/A sizeof(*RExC_rx->data) + sizeof(void*) * (RExC_rx->data->count + n - 1),
1N/A char, struct reg_data);
1N/A Renew(RExC_rx->data->what, RExC_rx->data->count + n, U8);
1N/A RExC_rx->data->count += n;
1N/A }
1N/A else {
1N/A Newc(1207, RExC_rx->data, sizeof(*RExC_rx->data) + sizeof(void*) * (n - 1),
1N/A char, struct reg_data);
1N/A New(1208, RExC_rx->data->what, n, U8);
1N/A RExC_rx->data->count = n;
1N/A }
1N/A Copy(s, RExC_rx->data->what + RExC_rx->data->count - n, n, U8);
1N/A return RExC_rx->data->count - n;
1N/A}
1N/A
1N/Avoid
1N/APerl_reginitcolors(pTHX)
1N/A{
1N/A int i = 0;
1N/A char *s = PerlEnv_getenv("PERL_RE_COLORS");
1N/A
1N/A if (s) {
1N/A PL_colors[0] = s = savepv(s);
1N/A while (++i < 6) {
1N/A s = strchr(s, '\t');
1N/A if (s) {
1N/A *s = '\0';
1N/A PL_colors[i] = ++s;
1N/A }
1N/A else
1N/A PL_colors[i] = s = "";
1N/A }
1N/A } else {
1N/A while (i < 6)
1N/A PL_colors[i++] = "";
1N/A }
1N/A PL_colorset = 1;
1N/A}
1N/A
1N/A
1N/A/*
1N/A - pregcomp - compile a regular expression into internal code
1N/A *
1N/A * We can't allocate space until we know how big the compiled form will be,
1N/A * but we can't compile it (and thus know how big it is) until we've got a
1N/A * place to put the code. So we cheat: we compile it twice, once with code
1N/A * generation turned off and size counting turned on, and once "for real".
1N/A * This also means that we don't allocate space until we are sure that the
1N/A * thing really will compile successfully, and we never have to move the
1N/A * code and thus invalidate pointers into it. (Note that it has to be in
1N/A * one piece because free() must be able to free it all.) [NB: not true in perl]
1N/A *
1N/A * Beware that the optimization-preparation code in here knows about some
1N/A * of the structure of the compiled regexp. [I'll say.]
1N/A */
1N/Aregexp *
1N/APerl_pregcomp(pTHX_ char *exp, char *xend, PMOP *pm)
1N/A{
1N/A register regexp *r;
1N/A regnode *scan;
1N/A regnode *first;
1N/A I32 flags;
1N/A I32 minlen = 0;
1N/A I32 sawplus = 0;
1N/A I32 sawopen = 0;
1N/A scan_data_t data;
1N/A RExC_state_t RExC_state;
1N/A RExC_state_t *pRExC_state = &RExC_state;
1N/A
1N/A if (exp == NULL)
1N/A FAIL("NULL regexp argument");
1N/A
1N/A RExC_utf8 = RExC_orig_utf8 = pm->op_pmdynflags & PMdf_CMP_UTF8;
1N/A
1N/A DEBUG_r({
1N/A if (!PL_colorset) reginitcolors();
1N/A PerlIO_printf(Perl_debug_log, "%sCompiling REx%s `%s%*s%s'\n",
1N/A PL_colors[4],PL_colors[5],PL_colors[0],
1N/A (int)(xend - exp), exp, PL_colors[1]);
1N/A });
1N/A
1N/Aredo_first_pass:
1N/A RExC_precomp = exp;
1N/A RExC_flags = pm->op_pmflags;
1N/A RExC_sawback = 0;
1N/A
1N/A RExC_seen = 0;
1N/A RExC_seen_zerolen = *exp == '^' ? -1 : 0;
1N/A RExC_seen_evals = 0;
1N/A RExC_extralen = 0;
1N/A
1N/A /* First pass: determine size, legality. */
1N/A RExC_parse = exp;
1N/A RExC_start = exp;
1N/A RExC_end = xend;
1N/A RExC_naughty = 0;
1N/A RExC_npar = 1;
1N/A RExC_size = 0L;
1N/A RExC_emit = &PL_regdummy;
1N/A RExC_whilem_seen = 0;
1N/A#if 0 /* REGC() is (currently) a NOP at the first pass.
1N/A * Clever compilers notice this and complain. --jhi */
1N/A REGC((U8)REG_MAGIC, (char*)RExC_emit);
1N/A#endif
1N/A if (reg(pRExC_state, 0, &flags) == NULL) {
1N/A RExC_precomp = Nullch;
1N/A return(NULL);
1N/A }
1N/A
1N/A if (RExC_utf8 && !RExC_orig_utf8) {
1N/A /* It's possible to write a regexp in ascii that represents unicode
1N/A codepoints outside of the byte range, such as via \x{100}. If we
1N/A detect such a sequence we have to convert the entire pattern to utf8
1N/A and then recompile, as our sizing calculation will have been based
1N/A on 1 byte == 1 character, but we will need to use utf8 to encode
1N/A at least some part of the pattern, and therefore must convert the whole
1N/A thing.
1N/A XXX: somehow figure out how to make this less expensive...
1N/A -- dmq */
1N/A STRLEN len = xend-exp;
1N/A DEBUG_r(PerlIO_printf(Perl_debug_log,
1N/A "UTF8 mismatch! Converting to utf8 for resizing and compile\n"));
1N/A exp = (char*)Perl_bytes_to_utf8(aTHX_ (U8*)exp, &len);
1N/A xend = exp + len;
1N/A RExC_orig_utf8 = RExC_utf8;
1N/A SAVEFREEPV(exp);
1N/A goto redo_first_pass;
1N/A }
1N/A DEBUG_r(PerlIO_printf(Perl_debug_log, "size %"IVdf" ", (IV)RExC_size));
1N/A
1N/A /* Small enough for pointer-storage convention?
1N/A If extralen==0, this means that we will not need long jumps. */
1N/A if (RExC_size >= 0x10000L && RExC_extralen)
1N/A RExC_size += RExC_extralen;
1N/A else
1N/A RExC_extralen = 0;
1N/A if (RExC_whilem_seen > 15)
1N/A RExC_whilem_seen = 15;
1N/A
1N/A /* Allocate space and initialize. */
1N/A Newc(1001, r, sizeof(regexp) + (unsigned)RExC_size * sizeof(regnode),
1N/A char, regexp);
1N/A if (r == NULL)
1N/A FAIL("Regexp out of space");
1N/A
1N/A#ifdef DEBUGGING
1N/A /* avoid reading uninitialized memory in DEBUGGING code in study_chunk() */
1N/A Zero(r, sizeof(regexp) + (unsigned)RExC_size * sizeof(regnode), char);
1N/A#endif
1N/A r->refcnt = 1;
1N/A r->prelen = xend - exp;
1N/A r->precomp = savepvn(RExC_precomp, r->prelen);
1N/A r->subbeg = NULL;
1N/A r->reganch = pm->op_pmflags & PMf_COMPILETIME;
1N/A r->nparens = RExC_npar - 1; /* set early to validate backrefs */
1N/A
1N/A r->substrs = 0; /* Useful during FAIL. */
1N/A r->startp = 0; /* Useful during FAIL. */
1N/A r->endp = 0; /* Useful during FAIL. */
1N/A
1N/A Newz(1304, r->offsets, 2*RExC_size+1, U32); /* MJD 20001228 */
1N/A if (r->offsets) {
1N/A r->offsets[0] = RExC_size;
1N/A }
1N/A DEBUG_r(PerlIO_printf(Perl_debug_log,
1N/A "%s %"UVuf" bytes for offset annotations.\n",
1N/A r->offsets ? "Got" : "Couldn't get",
1N/A (UV)((2*RExC_size+1) * sizeof(U32))));
1N/A
1N/A RExC_rx = r;
1N/A
1N/A /* Second pass: emit code. */
1N/A RExC_flags = pm->op_pmflags; /* don't let top level (?i) bleed */
1N/A RExC_parse = exp;
1N/A RExC_end = xend;
1N/A RExC_naughty = 0;
1N/A RExC_npar = 1;
1N/A RExC_emit_start = r->program;
1N/A RExC_emit = r->program;
1N/A /* Store the count of eval-groups for security checks: */
1N/A RExC_emit->next_off = (U16)((RExC_seen_evals > U16_MAX) ? U16_MAX : RExC_seen_evals);
1N/A REGC((U8)REG_MAGIC, (char*) RExC_emit++);
1N/A r->data = 0;
1N/A if (reg(pRExC_state, 0, &flags) == NULL)
1N/A return(NULL);
1N/A
1N/A /* Dig out information for optimizations. */
1N/A r->reganch = pm->op_pmflags & PMf_COMPILETIME; /* Again? */
1N/A pm->op_pmflags = RExC_flags;
1N/A if (UTF)
1N/A r->reganch |= ROPT_UTF8; /* Unicode in it? */
1N/A r->regstclass = NULL;
1N/A if (RExC_naughty >= 10) /* Probably an expensive pattern. */
1N/A r->reganch |= ROPT_NAUGHTY;
1N/A scan = r->program + 1; /* First BRANCH. */
1N/A
1N/A /* XXXX To minimize changes to RE engine we always allocate
1N/A 3-units-long substrs field. */
1N/A Newz(1004, r->substrs, 1, struct reg_substr_data);
1N/A
1N/A StructCopy(&zero_scan_data, &data, scan_data_t);
1N/A /* XXXX Should not we check for something else? Usually it is OPEN1... */
1N/A if (OP(scan) != BRANCH) { /* Only one top-level choice. */
1N/A I32 fake;
1N/A STRLEN longest_float_length, longest_fixed_length;
1N/A struct regnode_charclass_class ch_class;
1N/A int stclass_flag;
1N/A I32 last_close = 0;
1N/A
1N/A first = scan;
1N/A /* Skip introductions and multiplicators >= 1. */
1N/A while ((OP(first) == OPEN && (sawopen = 1)) ||
1N/A /* An OR of *one* alternative - should not happen now. */
1N/A (OP(first) == BRANCH && OP(regnext(first)) != BRANCH) ||
1N/A (OP(first) == PLUS) ||
1N/A (OP(first) == MINMOD) ||
1N/A /* An {n,m} with n>0 */
1N/A (PL_regkind[(U8)OP(first)] == CURLY && ARG1(first) > 0) ) {
1N/A if (OP(first) == PLUS)
1N/A sawplus = 1;
1N/A else
1N/A first += regarglen[(U8)OP(first)];
1N/A first = NEXTOPER(first);
1N/A }
1N/A
1N/A /* Starting-point info. */
1N/A again:
1N/A if (PL_regkind[(U8)OP(first)] == EXACT) {
1N/A if (OP(first) == EXACT)
1N/A ; /* Empty, get anchored substr later. */
1N/A else if ((OP(first) == EXACTF || OP(first) == EXACTFL))
1N/A r->regstclass = first;
1N/A }
1N/A else if (strchr((char*)PL_simple,OP(first)))
1N/A r->regstclass = first;
1N/A else if (PL_regkind[(U8)OP(first)] == BOUND ||
1N/A PL_regkind[(U8)OP(first)] == NBOUND)
1N/A r->regstclass = first;
1N/A else if (PL_regkind[(U8)OP(first)] == BOL) {
1N/A r->reganch |= (OP(first) == MBOL
1N/A ? ROPT_ANCH_MBOL
1N/A : (OP(first) == SBOL
1N/A ? ROPT_ANCH_SBOL
1N/A : ROPT_ANCH_BOL));
1N/A first = NEXTOPER(first);
1N/A goto again;
1N/A }
1N/A else if (OP(first) == GPOS) {
1N/A r->reganch |= ROPT_ANCH_GPOS;
1N/A first = NEXTOPER(first);
1N/A goto again;
1N/A }
1N/A else if (!sawopen && (OP(first) == STAR &&
1N/A PL_regkind[(U8)OP(NEXTOPER(first))] == REG_ANY) &&
1N/A !(r->reganch & ROPT_ANCH) )
1N/A {
1N/A /* turn .* into ^.* with an implied $*=1 */
1N/A int type = OP(NEXTOPER(first));
1N/A
1N/A if (type == REG_ANY)
1N/A type = ROPT_ANCH_MBOL;
1N/A else
1N/A type = ROPT_ANCH_SBOL;
1N/A
1N/A r->reganch |= type | ROPT_IMPLICIT;
1N/A first = NEXTOPER(first);
1N/A goto again;
1N/A }
1N/A if (sawplus && (!sawopen || !RExC_sawback)
1N/A && !(RExC_seen & REG_SEEN_EVAL)) /* May examine pos and $& */
1N/A /* x+ must match at the 1st pos of run of x's */
1N/A r->reganch |= ROPT_SKIP;
1N/A
1N/A /* Scan is after the zeroth branch, first is atomic matcher. */
1N/A DEBUG_r(PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
1N/A (IV)(first - scan + 1)));
1N/A /*
1N/A * If there's something expensive in the r.e., find the
1N/A * longest literal string that must appear and make it the
1N/A * regmust. Resolve ties in favor of later strings, since
1N/A * the regstart check works with the beginning of the r.e.
1N/A * and avoiding duplication strengthens checking. Not a
1N/A * strong reason, but sufficient in the absence of others.
1N/A * [Now we resolve ties in favor of the earlier string if
1N/A * it happens that c_offset_min has been invalidated, since the
1N/A * earlier string may buy us something the later one won't.]
1N/A */
1N/A minlen = 0;
1N/A
1N/A data.longest_fixed = newSVpvn("",0);
1N/A data.longest_float = newSVpvn("",0);
1N/A data.last_found = newSVpvn("",0);
1N/A data.longest = &(data.longest_fixed);
1N/A first = scan;
1N/A if (!r->regstclass) {
1N/A cl_init(pRExC_state, &ch_class);
1N/A data.start_class = &ch_class;
1N/A stclass_flag = SCF_DO_STCLASS_AND;
1N/A } else /* XXXX Check for BOUND? */
1N/A stclass_flag = 0;
1N/A data.last_closep = &last_close;
1N/A
1N/A minlen = study_chunk(pRExC_state, &first, &fake, scan + RExC_size, /* Up to end */
1N/A &data, SCF_DO_SUBSTR | SCF_WHILEM_VISITED_POS | stclass_flag);
1N/A if ( RExC_npar == 1 && data.longest == &(data.longest_fixed)
1N/A && data.last_start_min == 0 && data.last_end > 0
1N/A && !RExC_seen_zerolen
1N/A && (!(RExC_seen & REG_SEEN_GPOS) || (r->reganch & ROPT_ANCH_GPOS)))
1N/A r->reganch |= ROPT_CHECK_ALL;
1N/A scan_commit(pRExC_state, &data);
1N/A SvREFCNT_dec(data.last_found);
1N/A
1N/A longest_float_length = CHR_SVLEN(data.longest_float);
1N/A if (longest_float_length
1N/A || (data.flags & SF_FL_BEFORE_EOL
1N/A && (!(data.flags & SF_FL_BEFORE_MEOL)
1N/A || (RExC_flags & PMf_MULTILINE)))) {
1N/A int t;
1N/A
1N/A if (SvCUR(data.longest_fixed) /* ok to leave SvCUR */
1N/A && data.offset_fixed == data.offset_float_min
1N/A && SvCUR(data.longest_fixed) == SvCUR(data.longest_float))
1N/A goto remove_float; /* As in (a)+. */
1N/A
1N/A if (SvUTF8(data.longest_float)) {
1N/A r->float_utf8 = data.longest_float;
1N/A r->float_substr = Nullsv;
1N/A } else {
1N/A r->float_substr = data.longest_float;
1N/A r->float_utf8 = Nullsv;
1N/A }
1N/A r->float_min_offset = data.offset_float_min;
1N/A r->float_max_offset = data.offset_float_max;
1N/A t = (data.flags & SF_FL_BEFORE_EOL /* Can't have SEOL and MULTI */
1N/A && (!(data.flags & SF_FL_BEFORE_MEOL)
1N/A || (RExC_flags & PMf_MULTILINE)));
1N/A fbm_compile(data.longest_float, t ? FBMcf_TAIL : 0);
1N/A }
1N/A else {
1N/A remove_float:
1N/A r->float_substr = r->float_utf8 = Nullsv;
1N/A SvREFCNT_dec(data.longest_float);
1N/A longest_float_length = 0;
1N/A }
1N/A
1N/A longest_fixed_length = CHR_SVLEN(data.longest_fixed);
1N/A if (longest_fixed_length
1N/A || (data.flags & SF_FIX_BEFORE_EOL /* Cannot have SEOL and MULTI */
1N/A && (!(data.flags & SF_FIX_BEFORE_MEOL)
1N/A || (RExC_flags & PMf_MULTILINE)))) {
1N/A int t;
1N/A
1N/A if (SvUTF8(data.longest_fixed)) {
1N/A r->anchored_utf8 = data.longest_fixed;
1N/A r->anchored_substr = Nullsv;
1N/A } else {
1N/A r->anchored_substr = data.longest_fixed;
1N/A r->anchored_utf8 = Nullsv;
1N/A }
1N/A r->anchored_offset = data.offset_fixed;
1N/A t = (data.flags & SF_FIX_BEFORE_EOL /* Can't have SEOL and MULTI */
1N/A && (!(data.flags & SF_FIX_BEFORE_MEOL)
1N/A || (RExC_flags & PMf_MULTILINE)));
1N/A fbm_compile(data.longest_fixed, t ? FBMcf_TAIL : 0);
1N/A }
1N/A else {
1N/A r->anchored_substr = r->anchored_utf8 = Nullsv;
1N/A SvREFCNT_dec(data.longest_fixed);
1N/A longest_fixed_length = 0;
1N/A }
1N/A if (r->regstclass
1N/A && (OP(r->regstclass) == REG_ANY || OP(r->regstclass) == SANY))
1N/A r->regstclass = NULL;
1N/A if ((!(r->anchored_substr || r->anchored_utf8) || r->anchored_offset)
1N/A && stclass_flag
1N/A && !(data.start_class->flags & ANYOF_EOS)
1N/A && !cl_is_anything(data.start_class))
1N/A {
1N/A I32 n = add_data(pRExC_state, 1, "f");
1N/A
1N/A New(1006, RExC_rx->data->data[n], 1,
1N/A struct regnode_charclass_class);
1N/A StructCopy(data.start_class,
1N/A (struct regnode_charclass_class*)RExC_rx->data->data[n],
1N/A struct regnode_charclass_class);
1N/A r->regstclass = (regnode*)RExC_rx->data->data[n];
1N/A r->reganch &= ~ROPT_SKIP; /* Used in find_byclass(). */
1N/A PL_regdata = r->data; /* for regprop() */
1N/A DEBUG_r({ SV *sv = sv_newmortal();
1N/A regprop(sv, (regnode*)data.start_class);
1N/A PerlIO_printf(Perl_debug_log,
1N/A "synthetic stclass `%s'.\n",
1N/A SvPVX(sv));});
1N/A }
1N/A
1N/A /* A temporary algorithm prefers floated substr to fixed one to dig more info. */
1N/A if (longest_fixed_length > longest_float_length) {
1N/A r->check_substr = r->anchored_substr;
1N/A r->check_utf8 = r->anchored_utf8;
1N/A r->check_offset_min = r->check_offset_max = r->anchored_offset;
1N/A if (r->reganch & ROPT_ANCH_SINGLE)
1N/A r->reganch |= ROPT_NOSCAN;
1N/A }
1N/A else {
1N/A r->check_substr = r->float_substr;
1N/A r->check_utf8 = r->float_utf8;
1N/A r->check_offset_min = data.offset_float_min;
1N/A r->check_offset_max = data.offset_float_max;
1N/A }
1N/A /* XXXX Currently intuiting is not compatible with ANCH_GPOS.
1N/A This should be changed ASAP! */
1N/A if ((r->check_substr || r->check_utf8) && !(r->reganch & ROPT_ANCH_GPOS)) {
1N/A r->reganch |= RE_USE_INTUIT;
1N/A if (SvTAIL(r->check_substr ? r->check_substr : r->check_utf8))
1N/A r->reganch |= RE_INTUIT_TAIL;
1N/A }
1N/A }
1N/A else {
1N/A /* Several toplevels. Best we can is to set minlen. */
1N/A I32 fake;
1N/A struct regnode_charclass_class ch_class;
1N/A I32 last_close = 0;
1N/A
1N/A DEBUG_r(PerlIO_printf(Perl_debug_log, "\n"));
1N/A scan = r->program + 1;
1N/A cl_init(pRExC_state, &ch_class);
1N/A data.start_class = &ch_class;
1N/A data.last_closep = &last_close;
1N/A minlen = study_chunk(pRExC_state, &scan, &fake, scan + RExC_size, &data, SCF_DO_STCLASS_AND|SCF_WHILEM_VISITED_POS);
1N/A r->check_substr = r->check_utf8 = r->anchored_substr = r->anchored_utf8
1N/A = r->float_substr = r->float_utf8 = Nullsv;
1N/A if (!(data.start_class->flags & ANYOF_EOS)
1N/A && !cl_is_anything(data.start_class))
1N/A {
1N/A I32 n = add_data(pRExC_state, 1, "f");
1N/A
1N/A New(1006, RExC_rx->data->data[n], 1,
1N/A struct regnode_charclass_class);
1N/A StructCopy(data.start_class,
1N/A (struct regnode_charclass_class*)RExC_rx->data->data[n],
1N/A struct regnode_charclass_class);
1N/A r->regstclass = (regnode*)RExC_rx->data->data[n];
1N/A r->reganch &= ~ROPT_SKIP; /* Used in find_byclass(). */
1N/A DEBUG_r({ SV* sv = sv_newmortal();
1N/A regprop(sv, (regnode*)data.start_class);
1N/A PerlIO_printf(Perl_debug_log,
1N/A "synthetic stclass `%s'.\n",
1N/A SvPVX(sv));});
1N/A }
1N/A }
1N/A
1N/A r->minlen = minlen;
1N/A if (RExC_seen & REG_SEEN_GPOS)
1N/A r->reganch |= ROPT_GPOS_SEEN;
1N/A if (RExC_seen & REG_SEEN_LOOKBEHIND)
1N/A r->reganch |= ROPT_LOOKBEHIND_SEEN;
1N/A if (RExC_seen & REG_SEEN_EVAL)
1N/A r->reganch |= ROPT_EVAL_SEEN;
1N/A if (RExC_seen & REG_SEEN_CANY)
1N/A r->reganch |= ROPT_CANY_SEEN;
1N/A Newz(1002, r->startp, RExC_npar, I32);
1N/A Newz(1002, r->endp, RExC_npar, I32);
1N/A PL_regdata = r->data; /* for regprop() */
1N/A DEBUG_r(regdump(r));
1N/A return(r);
1N/A}
1N/A
1N/A/*
1N/A - reg - regular expression, i.e. main body or parenthesized thing
1N/A *
1N/A * Caller must absorb opening parenthesis.
1N/A *
1N/A * Combining parenthesis handling with the base level of regular expression
1N/A * is a trifle forced, but the need to tie the tails of the branches to what
1N/A * follows makes it hard to avoid.
1N/A */
1N/ASTATIC regnode *
1N/AS_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp)
1N/A /* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */
1N/A{
1N/A register regnode *ret; /* Will be the head of the group. */
1N/A register regnode *br;
1N/A register regnode *lastbr;
1N/A register regnode *ender = 0;
1N/A register I32 parno = 0;
1N/A I32 flags, oregflags = RExC_flags, have_branch = 0, open = 0;
1N/A
1N/A /* for (?g), (?gc), and (?o) warnings; warning
1N/A about (?c) will warn about (?g) -- japhy */
1N/A
1N/A I32 wastedflags = 0x00,
1N/A wasted_o = 0x01,
1N/A wasted_g = 0x02,
1N/A wasted_gc = 0x02 | 0x04,
1N/A wasted_c = 0x04;
1N/A
1N/A char * parse_start = RExC_parse; /* MJD */
1N/A char *oregcomp_parse = RExC_parse;
1N/A char c;
1N/A
1N/A *flagp = 0; /* Tentatively. */
1N/A
1N/A
1N/A /* Make an OPEN node, if parenthesized. */
1N/A if (paren) {
1N/A if (*RExC_parse == '?') { /* (?...) */
1N/A U32 posflags = 0, negflags = 0;
1N/A U32 *flagsp = &posflags;
1N/A int logical = 0;
1N/A char *seqstart = RExC_parse;
1N/A
1N/A RExC_parse++;
1N/A paren = *RExC_parse++;
1N/A ret = NULL; /* For look-ahead/behind. */
1N/A switch (paren) {
1N/A case '<': /* (?<...) */
1N/A RExC_seen |= REG_SEEN_LOOKBEHIND;
1N/A if (*RExC_parse == '!')
1N/A paren = ',';
1N/A if (*RExC_parse != '=' && *RExC_parse != '!')
1N/A goto unknown;
1N/A RExC_parse++;
1N/A case '=': /* (?=...) */
1N/A case '!': /* (?!...) */
1N/A RExC_seen_zerolen++;
1N/A case ':': /* (?:...) */
1N/A case '>': /* (?>...) */
1N/A break;
1N/A case '$': /* (?$...) */
1N/A case '@': /* (?@...) */
1N/A vFAIL2("Sequence (?%c...) not implemented", (int)paren);
1N/A break;
1N/A case '#': /* (?#...) */
1N/A while (*RExC_parse && *RExC_parse != ')')
1N/A RExC_parse++;
1N/A if (*RExC_parse != ')')
1N/A FAIL("Sequence (?#... not terminated");
1N/A nextchar(pRExC_state);
1N/A *flagp = TRYAGAIN;
1N/A return NULL;
1N/A case 'p': /* (?p...) */
1N/A if (SIZE_ONLY && ckWARN2(WARN_DEPRECATED, WARN_REGEXP))
1N/A vWARNdep(RExC_parse, "(?p{}) is deprecated - use (??{})");
1N/A /* FALL THROUGH*/
1N/A case '?': /* (??...) */
1N/A logical = 1;
1N/A if (*RExC_parse != '{')
1N/A goto unknown;
1N/A paren = *RExC_parse++;
1N/A /* FALL THROUGH */
1N/A case '{': /* (?{...}) */
1N/A {
1N/A I32 count = 1, n = 0;
1N/A char c;
1N/A char *s = RExC_parse;
1N/A SV *sv;
1N/A OP_4tree *sop, *rop;
1N/A
1N/A RExC_seen_zerolen++;
1N/A RExC_seen |= REG_SEEN_EVAL;
1N/A while (count && (c = *RExC_parse)) {
1N/A if (c == '\\' && RExC_parse[1])
1N/A RExC_parse++;
1N/A else if (c == '{')
1N/A count++;
1N/A else if (c == '}')
1N/A count--;
1N/A RExC_parse++;
1N/A }
1N/A if (*RExC_parse != ')')
1N/A {
1N/A RExC_parse = s;
1N/A vFAIL("Sequence (?{...}) not terminated or not {}-balanced");
1N/A }
1N/A if (!SIZE_ONLY) {
1N/A PAD *pad;
1N/A
1N/A if (RExC_parse - 1 - s)
1N/A sv = newSVpvn(s, RExC_parse - 1 - s);
1N/A else
1N/A sv = newSVpvn("", 0);
1N/A
1N/A ENTER;
1N/A Perl_save_re_context(aTHX);
1N/A rop = sv_compile_2op(sv, &sop, "re", &pad);
1N/A sop->op_private |= OPpREFCOUNTED;
1N/A /* re_dup will OpREFCNT_inc */
1N/A OpREFCNT_set(sop, 1);
1N/A LEAVE;
1N/A
1N/A n = add_data(pRExC_state, 3, "nop");
1N/A RExC_rx->data->data[n] = (void*)rop;
1N/A RExC_rx->data->data[n+1] = (void*)sop;
1N/A RExC_rx->data->data[n+2] = (void*)pad;
1N/A SvREFCNT_dec(sv);
1N/A }
1N/A else { /* First pass */
1N/A if (PL_reginterp_cnt < ++RExC_seen_evals
1N/A && IN_PERL_RUNTIME)
1N/A /* No compiled RE interpolated, has runtime
1N/A components ===> unsafe. */
1N/A FAIL("Eval-group not allowed at runtime, use re 'eval'");
1N/A if (PL_tainting && PL_tainted)
1N/A FAIL("Eval-group in insecure regular expression");
1N/A }
1N/A
1N/A nextchar(pRExC_state);
1N/A if (logical) {
1N/A ret = reg_node(pRExC_state, LOGICAL);
1N/A if (!SIZE_ONLY)
1N/A ret->flags = 2;
1N/A regtail(pRExC_state, ret, reganode(pRExC_state, EVAL, n));
1N/A /* deal with the length of this later - MJD */
1N/A return ret;
1N/A }
1N/A ret = reganode(pRExC_state, EVAL, n);
1N/A Set_Node_Length(ret, RExC_parse - parse_start + 1);
1N/A Set_Node_Offset(ret, parse_start);
1N/A return ret;
1N/A }
1N/A case '(': /* (?(?{...})...) and (?(?=...)...) */
1N/A {
1N/A if (RExC_parse[0] == '?') { /* (?(?...)) */
1N/A if (RExC_parse[1] == '=' || RExC_parse[1] == '!'
1N/A || RExC_parse[1] == '<'
1N/A || RExC_parse[1] == '{') { /* Lookahead or eval. */
1N/A I32 flag;
1N/A
1N/A ret = reg_node(pRExC_state, LOGICAL);
1N/A if (!SIZE_ONLY)
1N/A ret->flags = 1;
1N/A regtail(pRExC_state, ret, reg(pRExC_state, 1, &flag));
1N/A goto insert_if;
1N/A }
1N/A }
1N/A else if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
1N/A /* (?(1)...) */
1N/A parno = atoi(RExC_parse++);
1N/A
1N/A while (isDIGIT(*RExC_parse))
1N/A RExC_parse++;
1N/A ret = reganode(pRExC_state, GROUPP, parno);
1N/A
1N/A if ((c = *nextchar(pRExC_state)) != ')')
1N/A vFAIL("Switch condition not recognized");
1N/A insert_if:
1N/A regtail(pRExC_state, ret, reganode(pRExC_state, IFTHEN, 0));
1N/A br = regbranch(pRExC_state, &flags, 1);
1N/A if (br == NULL)
1N/A br = reganode(pRExC_state, LONGJMP, 0);
1N/A else
1N/A regtail(pRExC_state, br, reganode(pRExC_state, LONGJMP, 0));
1N/A c = *nextchar(pRExC_state);
1N/A if (flags&HASWIDTH)
1N/A *flagp |= HASWIDTH;
1N/A if (c == '|') {
1N/A lastbr = reganode(pRExC_state, IFTHEN, 0); /* Fake one for optimizer. */
1N/A regbranch(pRExC_state, &flags, 1);
1N/A regtail(pRExC_state, ret, lastbr);
1N/A if (flags&HASWIDTH)
1N/A *flagp |= HASWIDTH;
1N/A c = *nextchar(pRExC_state);
1N/A }
1N/A else
1N/A lastbr = NULL;
1N/A if (c != ')')
1N/A vFAIL("Switch (?(condition)... contains too many branches");
1N/A ender = reg_node(pRExC_state, TAIL);
1N/A regtail(pRExC_state, br, ender);
1N/A if (lastbr) {
1N/A regtail(pRExC_state, lastbr, ender);
1N/A regtail(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender);
1N/A }
1N/A else
1N/A regtail(pRExC_state, ret, ender);
1N/A return ret;
1N/A }
1N/A else {
1N/A vFAIL2("Unknown switch condition (?(%.2s", RExC_parse);
1N/A }
1N/A }
1N/A case 0:
1N/A RExC_parse--; /* for vFAIL to print correctly */
1N/A vFAIL("Sequence (? incomplete");
1N/A break;
1N/A default:
1N/A --RExC_parse;
1N/A parse_flags: /* (?i) */
1N/A while (*RExC_parse && strchr("iogcmsx", *RExC_parse)) {
1N/A /* (?g), (?gc) and (?o) are useless here
1N/A and must be globally applied -- japhy */
1N/A
1N/A if (*RExC_parse == 'o' || *RExC_parse == 'g') {
1N/A if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
1N/A I32 wflagbit = *RExC_parse == 'o' ? wasted_o : wasted_g;
1N/A if (! (wastedflags & wflagbit) ) {
1N/A wastedflags |= wflagbit;
1N/A vWARN5(
1N/A RExC_parse + 1,
1N/A "Useless (%s%c) - %suse /%c modifier",
1N/A flagsp == &negflags ? "?-" : "?",
1N/A *RExC_parse,
1N/A flagsp == &negflags ? "don't " : "",
1N/A *RExC_parse
1N/A );
1N/A }
1N/A }
1N/A }
1N/A else if (*RExC_parse == 'c') {
1N/A if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
1N/A if (! (wastedflags & wasted_c) ) {
1N/A wastedflags |= wasted_gc;
1N/A vWARN3(
1N/A RExC_parse + 1,
1N/A "Useless (%sc) - %suse /gc modifier",
1N/A flagsp == &negflags ? "?-" : "?",
1N/A flagsp == &negflags ? "don't " : ""
1N/A );
1N/A }
1N/A }
1N/A }
1N/A else { pmflag(flagsp, *RExC_parse); }
1N/A
1N/A ++RExC_parse;
1N/A }
1N/A if (*RExC_parse == '-') {
1N/A flagsp = &negflags;
1N/A wastedflags = 0; /* reset so (?g-c) warns twice */
1N/A ++RExC_parse;
1N/A goto parse_flags;
1N/A }
1N/A RExC_flags |= posflags;
1N/A RExC_flags &= ~negflags;
1N/A if (*RExC_parse == ':') {
1N/A RExC_parse++;
1N/A paren = ':';
1N/A break;
1N/A }
1N/A unknown:
1N/A if (*RExC_parse != ')') {
1N/A RExC_parse++;
1N/A vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
1N/A }
1N/A nextchar(pRExC_state);
1N/A *flagp = TRYAGAIN;
1N/A return NULL;
1N/A }
1N/A }
1N/A else { /* (...) */
1N/A parno = RExC_npar;
1N/A RExC_npar++;
1N/A ret = reganode(pRExC_state, OPEN, parno);
1N/A Set_Node_Length(ret, 1); /* MJD */
1N/A Set_Node_Offset(ret, RExC_parse); /* MJD */
1N/A open = 1;
1N/A }
1N/A }
1N/A else /* ! paren */
1N/A ret = NULL;
1N/A
1N/A /* Pick up the branches, linking them together. */
1N/A parse_start = RExC_parse; /* MJD */
1N/A br = regbranch(pRExC_state, &flags, 1);
1N/A /* branch_len = (paren != 0); */
1N/A
1N/A if (br == NULL)
1N/A return(NULL);
1N/A if (*RExC_parse == '|') {
1N/A if (!SIZE_ONLY && RExC_extralen) {
1N/A reginsert(pRExC_state, BRANCHJ, br);
1N/A }
1N/A else { /* MJD */
1N/A reginsert(pRExC_state, BRANCH, br);
1N/A Set_Node_Length(br, paren != 0);
1N/A Set_Node_Offset_To_R(br-RExC_emit_start, parse_start-RExC_start);
1N/A }
1N/A have_branch = 1;
1N/A if (SIZE_ONLY)
1N/A RExC_extralen += 1; /* For BRANCHJ-BRANCH. */
1N/A }
1N/A else if (paren == ':') {
1N/A *flagp |= flags&SIMPLE;
1N/A }
1N/A if (open) { /* Starts with OPEN. */
1N/A regtail(pRExC_state, ret, br); /* OPEN -> first. */
1N/A }
1N/A else if (paren != '?') /* Not Conditional */
1N/A ret = br;
1N/A *flagp |= flags & (SPSTART | HASWIDTH);
1N/A lastbr = br;
1N/A while (*RExC_parse == '|') {
1N/A if (!SIZE_ONLY && RExC_extralen) {
1N/A ender = reganode(pRExC_state, LONGJMP,0);
1N/A regtail(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender); /* Append to the previous. */
1N/A }
1N/A if (SIZE_ONLY)
1N/A RExC_extralen += 2; /* Account for LONGJMP. */
1N/A nextchar(pRExC_state);
1N/A br = regbranch(pRExC_state, &flags, 0);
1N/A
1N/A if (br == NULL)
1N/A return(NULL);
1N/A regtail(pRExC_state, lastbr, br); /* BRANCH -> BRANCH. */
1N/A lastbr = br;
1N/A if (flags&HASWIDTH)
1N/A *flagp |= HASWIDTH;
1N/A *flagp |= flags&SPSTART;
1N/A }
1N/A
1N/A if (have_branch || paren != ':') {
1N/A /* Make a closing node, and hook it on the end. */
1N/A switch (paren) {
1N/A case ':':
1N/A ender = reg_node(pRExC_state, TAIL);
1N/A break;
1N/A case 1:
1N/A ender = reganode(pRExC_state, CLOSE, parno);
1N/A Set_Node_Offset(ender,RExC_parse+1); /* MJD */
1N/A Set_Node_Length(ender,1); /* MJD */
1N/A break;
1N/A case '<':
1N/A case ',':
1N/A case '=':
1N/A case '!':
1N/A *flagp &= ~HASWIDTH;
1N/A /* FALL THROUGH */
1N/A case '>':
1N/A ender = reg_node(pRExC_state, SUCCEED);
1N/A break;
1N/A case 0:
1N/A ender = reg_node(pRExC_state, END);
1N/A break;
1N/A }
1N/A regtail(pRExC_state, lastbr, ender);
1N/A
1N/A if (have_branch) {
1N/A /* Hook the tails of the branches to the closing node. */
1N/A for (br = ret; br != NULL; br = regnext(br)) {
1N/A regoptail(pRExC_state, br, ender);
1N/A }
1N/A }
1N/A }
1N/A
1N/A {
1N/A char *p;
1N/A static char parens[] = "=!<,>";
1N/A
1N/A if (paren && (p = strchr(parens, paren))) {
1N/A U8 node = ((p - parens) % 2) ? UNLESSM : IFMATCH;
1N/A int flag = (p - parens) > 1;
1N/A
1N/A if (paren == '>')
1N/A node = SUSPEND, flag = 0;
1N/A reginsert(pRExC_state, node,ret);
1N/A Set_Node_Cur_Length(ret);
1N/A Set_Node_Offset(ret, parse_start + 1);
1N/A ret->flags = flag;
1N/A regtail(pRExC_state, ret, reg_node(pRExC_state, TAIL));
1N/A }
1N/A }
1N/A
1N/A /* Check for proper termination. */
1N/A if (paren) {
1N/A RExC_flags = oregflags;
1N/A if (RExC_parse >= RExC_end || *nextchar(pRExC_state) != ')') {
1N/A RExC_parse = oregcomp_parse;
1N/A vFAIL("Unmatched (");
1N/A }
1N/A }
1N/A else if (!paren && RExC_parse < RExC_end) {
1N/A if (*RExC_parse == ')') {
1N/A RExC_parse++;
1N/A vFAIL("Unmatched )");
1N/A }
1N/A else
1N/A FAIL("Junk on end of regexp"); /* "Can't happen". */
1N/A /* NOTREACHED */
1N/A }
1N/A
1N/A return(ret);
1N/A}
1N/A
1N/A/*
1N/A - regbranch - one alternative of an | operator
1N/A *
1N/A * Implements the concatenation operator.
1N/A */
1N/ASTATIC regnode *
1N/AS_regbranch(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, I32 first)
1N/A{
1N/A register regnode *ret;
1N/A register regnode *chain = NULL;
1N/A register regnode *latest;
1N/A I32 flags = 0, c = 0;
1N/A
1N/A if (first)
1N/A ret = NULL;
1N/A else {
1N/A if (!SIZE_ONLY && RExC_extralen)
1N/A ret = reganode(pRExC_state, BRANCHJ,0);
1N/A else {
1N/A ret = reg_node(pRExC_state, BRANCH);
1N/A Set_Node_Length(ret, 1);
1N/A }
1N/A }
1N/A
1N/A if (!first && SIZE_ONLY)
1N/A RExC_extralen += 1; /* BRANCHJ */
1N/A
1N/A *flagp = WORST; /* Tentatively. */
1N/A
1N/A RExC_parse--;
1N/A nextchar(pRExC_state);
1N/A while (RExC_parse < RExC_end && *RExC_parse != '|' && *RExC_parse != ')') {
1N/A flags &= ~TRYAGAIN;
1N/A latest = regpiece(pRExC_state, &flags);
1N/A if (latest == NULL) {
1N/A if (flags & TRYAGAIN)
1N/A continue;
1N/A return(NULL);
1N/A }
1N/A else if (ret == NULL)
1N/A ret = latest;
1N/A *flagp |= flags&HASWIDTH;
1N/A if (chain == NULL) /* First piece. */
1N/A *flagp |= flags&SPSTART;
1N/A else {
1N/A RExC_naughty++;
1N/A regtail(pRExC_state, chain, latest);
1N/A }
1N/A chain = latest;
1N/A c++;
1N/A }
1N/A if (chain == NULL) { /* Loop ran zero times. */
1N/A chain = reg_node(pRExC_state, NOTHING);
1N/A if (ret == NULL)
1N/A ret = chain;
1N/A }
1N/A if (c == 1) {
1N/A *flagp |= flags&SIMPLE;
1N/A }
1N/A
1N/A return(ret);
1N/A}
1N/A
1N/A/*
1N/A - regpiece - something followed by possible [*+?]
1N/A *
1N/A * Note that the branching code sequences used for ? and the general cases
1N/A * of * and + are somewhat optimized: they use the same NOTHING node as
1N/A * both the endmarker for their branch list and the body of the last branch.
1N/A * It might seem that this node could be dispensed with entirely, but the
1N/A * endmarker role is not redundant.
1N/A */
1N/ASTATIC regnode *
1N/AS_regpiece(pTHX_ RExC_state_t *pRExC_state, I32 *flagp)
1N/A{
1N/A register regnode *ret;
1N/A register char op;
1N/A register char *next;
1N/A I32 flags;
1N/A char *origparse = RExC_parse;
1N/A char *maxpos;
1N/A I32 min;
1N/A I32 max = REG_INFTY;
1N/A char *parse_start;
1N/A
1N/A ret = regatom(pRExC_state, &flags);
1N/A if (ret == NULL) {
1N/A if (flags & TRYAGAIN)
1N/A *flagp |= TRYAGAIN;
1N/A return(NULL);
1N/A }
1N/A
1N/A op = *RExC_parse;
1N/A
1N/A if (op == '{' && regcurly(RExC_parse)) {
1N/A parse_start = RExC_parse; /* MJD */
1N/A next = RExC_parse + 1;
1N/A maxpos = Nullch;
1N/A while (isDIGIT(*next) || *next == ',') {
1N/A if (*next == ',') {
1N/A if (maxpos)
1N/A break;
1N/A else
1N/A maxpos = next;
1N/A }
1N/A next++;
1N/A }
1N/A if (*next == '}') { /* got one */
1N/A if (!maxpos)
1N/A maxpos = next;
1N/A RExC_parse++;
1N/A min = atoi(RExC_parse);
1N/A if (*maxpos == ',')
1N/A maxpos++;
1N/A else
1N/A maxpos = RExC_parse;
1N/A max = atoi(maxpos);
1N/A if (!max && *maxpos != '0')
1N/A max = REG_INFTY; /* meaning "infinity" */
1N/A else if (max >= REG_INFTY)
1N/A vFAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
1N/A RExC_parse = next;
1N/A nextchar(pRExC_state);
1N/A
1N/A do_curly:
1N/A if ((flags&SIMPLE)) {
1N/A RExC_naughty += 2 + RExC_naughty / 2;
1N/A reginsert(pRExC_state, CURLY, ret);
1N/A Set_Node_Offset(ret, parse_start+1); /* MJD */
1N/A Set_Node_Cur_Length(ret);
1N/A }
1N/A else {
1N/A regnode *w = reg_node(pRExC_state, WHILEM);
1N/A
1N/A w->flags = 0;
1N/A regtail(pRExC_state, ret, w);
1N/A if (!SIZE_ONLY && RExC_extralen) {
1N/A reginsert(pRExC_state, LONGJMP,ret);
1N/A reginsert(pRExC_state, NOTHING,ret);
1N/A NEXT_OFF(ret) = 3; /* Go over LONGJMP. */
1N/A }
1N/A reginsert(pRExC_state, CURLYX,ret);
1N/A /* MJD hk */
1N/A Set_Node_Offset(ret, parse_start+1);
1N/A Set_Node_Length(ret,
1N/A op == '{' ? (RExC_parse - parse_start) : 1);
1N/A
1N/A if (!SIZE_ONLY && RExC_extralen)
1N/A NEXT_OFF(ret) = 3; /* Go over NOTHING to LONGJMP. */
1N/A regtail(pRExC_state, ret, reg_node(pRExC_state, NOTHING));
1N/A if (SIZE_ONLY)
1N/A RExC_whilem_seen++, RExC_extralen += 3;
1N/A RExC_naughty += 4 + RExC_naughty; /* compound interest */
1N/A }
1N/A ret->flags = 0;
1N/A
1N/A if (min > 0)
1N/A *flagp = WORST;
1N/A if (max > 0)
1N/A *flagp |= HASWIDTH;
1N/A if (max && max < min)
1N/A vFAIL("Can't do {n,m} with n > m");
1N/A if (!SIZE_ONLY) {
1N/A ARG1_SET(ret, (U16)min);
1N/A ARG2_SET(ret, (U16)max);
1N/A }
1N/A
1N/A goto nest_check;
1N/A }
1N/A }
1N/A
1N/A if (!ISMULT1(op)) {
1N/A *flagp = flags;
1N/A return(ret);
1N/A }
1N/A
1N/A#if 0 /* Now runtime fix should be reliable. */
1N/A
1N/A /* if this is reinstated, don't forget to put this back into perldiag:
1N/A
1N/A =item Regexp *+ operand could be empty at {#} in regex m/%s/
1N/A
1N/A (F) The part of the regexp subject to either the * or + quantifier
1N/A could match an empty string. The {#} shows in the regular
1N/A expression about where the problem was discovered.
1N/A
1N/A */
1N/A
1N/A if (!(flags&HASWIDTH) && op != '?')
1N/A vFAIL("Regexp *+ operand could be empty");
1N/A#endif
1N/A
1N/A parse_start = RExC_parse;
1N/A nextchar(pRExC_state);
1N/A
1N/A *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
1N/A
1N/A if (op == '*' && (flags&SIMPLE)) {
1N/A reginsert(pRExC_state, STAR, ret);
1N/A ret->flags = 0;
1N/A RExC_naughty += 4;
1N/A }
1N/A else if (op == '*') {
1N/A min = 0;
1N/A goto do_curly;
1N/A }
1N/A else if (op == '+' && (flags&SIMPLE)) {
1N/A reginsert(pRExC_state, PLUS, ret);
1N/A ret->flags = 0;
1N/A RExC_naughty += 3;
1N/A }
1N/A else if (op == '+') {
1N/A min = 1;
1N/A goto do_curly;
1N/A }
1N/A else if (op == '?') {
1N/A min = 0; max = 1;
1N/A goto do_curly;
1N/A }
1N/A nest_check:
1N/A if (ckWARN(WARN_REGEXP) && !SIZE_ONLY && !(flags&HASWIDTH) && max > REG_INFTY/3) {
1N/A vWARN3(RExC_parse,
1N/A "%.*s matches null string many times",
1N/A RExC_parse - origparse,
1N/A origparse);
1N/A }
1N/A
1N/A if (*RExC_parse == '?') {
1N/A nextchar(pRExC_state);
1N/A reginsert(pRExC_state, MINMOD, ret);
1N/A regtail(pRExC_state, ret, ret + NODE_STEP_REGNODE);
1N/A }
1N/A if (ISMULT2(RExC_parse)) {
1N/A RExC_parse++;
1N/A vFAIL("Nested quantifiers");
1N/A }
1N/A
1N/A return(ret);
1N/A}
1N/A
1N/A/*
1N/A - regatom - the lowest level
1N/A *
1N/A * Optimization: gobbles an entire sequence of ordinary characters so that
1N/A * it can turn them into a single node, which is smaller to store and
1N/A * faster to run. Backslashed characters are exceptions, each becoming a
1N/A * separate node; the code is simpler that way and it's not worth fixing.
1N/A *
1N/A * [Yes, it is worth fixing, some scripts can run twice the speed.] */
1N/ASTATIC regnode *
1N/AS_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp)
1N/A{
1N/A register regnode *ret = 0;
1N/A I32 flags;
1N/A char *parse_start = RExC_parse;
1N/A
1N/A *flagp = WORST; /* Tentatively. */
1N/A
1N/Atryagain:
1N/A switch (*RExC_parse) {
1N/A case '^':
1N/A RExC_seen_zerolen++;
1N/A nextchar(pRExC_state);
1N/A if (RExC_flags & PMf_MULTILINE)
1N/A ret = reg_node(pRExC_state, MBOL);
1N/A else if (RExC_flags & PMf_SINGLELINE)
1N/A ret = reg_node(pRExC_state, SBOL);
1N/A else
1N/A ret = reg_node(pRExC_state, BOL);
1N/A Set_Node_Length(ret, 1); /* MJD */
1N/A break;
1N/A case '$':
1N/A nextchar(pRExC_state);
1N/A if (*RExC_parse)
1N/A RExC_seen_zerolen++;
1N/A if (RExC_flags & PMf_MULTILINE)
1N/A ret = reg_node(pRExC_state, MEOL);
1N/A else if (RExC_flags & PMf_SINGLELINE)
1N/A ret = reg_node(pRExC_state, SEOL);
1N/A else
1N/A ret = reg_node(pRExC_state, EOL);
1N/A Set_Node_Length(ret, 1); /* MJD */
1N/A break;
1N/A case '.':
1N/A nextchar(pRExC_state);
1N/A if (RExC_flags & PMf_SINGLELINE)
1N/A ret = reg_node(pRExC_state, SANY);
1N/A else
1N/A ret = reg_node(pRExC_state, REG_ANY);
1N/A *flagp |= HASWIDTH|SIMPLE;
1N/A RExC_naughty++;
1N/A Set_Node_Length(ret, 1); /* MJD */
1N/A break;
1N/A case '[':
1N/A {
1N/A char *oregcomp_parse = ++RExC_parse;
1N/A ret = regclass(pRExC_state);
1N/A if (*RExC_parse != ']') {
1N/A RExC_parse = oregcomp_parse;
1N/A vFAIL("Unmatched [");
1N/A }
1N/A nextchar(pRExC_state);
1N/A *flagp |= HASWIDTH|SIMPLE;
1N/A Set_Node_Length(ret, RExC_parse - oregcomp_parse + 1); /* MJD */
1N/A break;
1N/A }
1N/A case '(':
1N/A nextchar(pRExC_state);
1N/A ret = reg(pRExC_state, 1, &flags);
1N/A if (ret == NULL) {
1N/A if (flags & TRYAGAIN) {
1N/A if (RExC_parse == RExC_end) {
1N/A /* Make parent create an empty node if needed. */
1N/A *flagp |= TRYAGAIN;
1N/A return(NULL);
1N/A }
1N/A goto tryagain;
1N/A }
1N/A return(NULL);
1N/A }
1N/A *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE);
1N/A break;
1N/A case '|':
1N/A case ')':
1N/A if (flags & TRYAGAIN) {
1N/A *flagp |= TRYAGAIN;
1N/A return NULL;
1N/A }
1N/A vFAIL("Internal urp");
1N/A /* Supposed to be caught earlier. */
1N/A break;
1N/A case '{':
1N/A if (!regcurly(RExC_parse)) {
1N/A RExC_parse++;
1N/A goto defchar;
1N/A }
1N/A /* FALL THROUGH */
1N/A case '?':
1N/A case '+':
1N/A case '*':
1N/A RExC_parse++;
1N/A vFAIL("Quantifier follows nothing");
1N/A break;
1N/A case '\\':
1N/A switch (*++RExC_parse) {
1N/A case 'A':
1N/A RExC_seen_zerolen++;
1N/A ret = reg_node(pRExC_state, SBOL);
1N/A *flagp |= SIMPLE;
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 'G':
1N/A ret = reg_node(pRExC_state, GPOS);
1N/A RExC_seen |= REG_SEEN_GPOS;
1N/A *flagp |= SIMPLE;
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 'Z':
1N/A ret = reg_node(pRExC_state, SEOL);
1N/A *flagp |= SIMPLE;
1N/A RExC_seen_zerolen++; /* Do not optimize RE away */
1N/A nextchar(pRExC_state);
1N/A break;
1N/A case 'z':
1N/A ret = reg_node(pRExC_state, EOS);
1N/A *flagp |= SIMPLE;
1N/A RExC_seen_zerolen++; /* Do not optimize RE away */
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 'C':
1N/A ret = reg_node(pRExC_state, CANY);
1N/A RExC_seen |= REG_SEEN_CANY;
1N/A *flagp |= HASWIDTH|SIMPLE;
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 'X':
1N/A ret = reg_node(pRExC_state, CLUMP);
1N/A *flagp |= HASWIDTH;
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 'w':
1N/A ret = reg_node(pRExC_state, (U8)(LOC ? ALNUML : ALNUM));
1N/A *flagp |= HASWIDTH|SIMPLE;
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 'W':
1N/A ret = reg_node(pRExC_state, (U8)(LOC ? NALNUML : NALNUM));
1N/A *flagp |= HASWIDTH|SIMPLE;
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 'b':
1N/A RExC_seen_zerolen++;
1N/A RExC_seen |= REG_SEEN_LOOKBEHIND;
1N/A ret = reg_node(pRExC_state, (U8)(LOC ? BOUNDL : BOUND));
1N/A *flagp |= SIMPLE;
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 'B':
1N/A RExC_seen_zerolen++;
1N/A RExC_seen |= REG_SEEN_LOOKBEHIND;
1N/A ret = reg_node(pRExC_state, (U8)(LOC ? NBOUNDL : NBOUND));
1N/A *flagp |= SIMPLE;
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 's':
1N/A ret = reg_node(pRExC_state, (U8)(LOC ? SPACEL : SPACE));
1N/A *flagp |= HASWIDTH|SIMPLE;
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 'S':
1N/A ret = reg_node(pRExC_state, (U8)(LOC ? NSPACEL : NSPACE));
1N/A *flagp |= HASWIDTH|SIMPLE;
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 'd':
1N/A ret = reg_node(pRExC_state, DIGIT);
1N/A *flagp |= HASWIDTH|SIMPLE;
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 'D':
1N/A ret = reg_node(pRExC_state, NDIGIT);
1N/A *flagp |= HASWIDTH|SIMPLE;
1N/A nextchar(pRExC_state);
1N/A Set_Node_Length(ret, 2); /* MJD */
1N/A break;
1N/A case 'p':
1N/A case 'P':
1N/A {
1N/A char* oldregxend = RExC_end;
1N/A char* parse_start = RExC_parse - 2;
1N/A
1N/A if (RExC_parse[1] == '{') {
1N/A /* a lovely hack--pretend we saw [\pX] instead */
1N/A RExC_end = strchr(RExC_parse, '}');
1N/A if (!RExC_end) {
1N/A U8 c = (U8)*RExC_parse;
1N/A RExC_parse += 2;
1N/A RExC_end = oldregxend;
1N/A vFAIL2("Missing right brace on \\%c{}", c);
1N/A }
1N/A RExC_end++;
1N/A }
1N/A else {
1N/A RExC_end = RExC_parse + 2;
1N/A if (RExC_end > oldregxend)
1N/A RExC_end = oldregxend;
1N/A }
1N/A RExC_parse--;
1N/A
1N/A ret = regclass(pRExC_state);
1N/A
1N/A RExC_end = oldregxend;
1N/A RExC_parse--;
1N/A
1N/A Set_Node_Offset(ret, parse_start + 2);
1N/A Set_Node_Cur_Length(ret);
1N/A nextchar(pRExC_state);
1N/A *flagp |= HASWIDTH|SIMPLE;
1N/A }
1N/A break;
1N/A case 'n':
1N/A case 'r':
1N/A case 't':
1N/A case 'f':
1N/A case 'e':
1N/A case 'a':
1N/A case 'x':
1N/A case 'c':
1N/A case '0':
1N/A goto defchar;
1N/A case '1': case '2': case '3': case '4':
1N/A case '5': case '6': case '7': case '8': case '9':
1N/A {
1N/A I32 num = atoi(RExC_parse);
1N/A
1N/A if (num > 9 && num >= RExC_npar)
1N/A goto defchar;
1N/A else {
1N/A char * parse_start = RExC_parse - 1; /* MJD */
1N/A while (isDIGIT(*RExC_parse))
1N/A RExC_parse++;
1N/A
1N/A if (!SIZE_ONLY && num > (I32)RExC_rx->nparens)
1N/A vFAIL("Reference to nonexistent group");
1N/A RExC_sawback = 1;
1N/A ret = reganode(pRExC_state,
1N/A (U8)(FOLD ? (LOC ? REFFL : REFF) : REF),
1N/A num);
1N/A *flagp |= HASWIDTH;
1N/A
1N/A /* override incorrect value set in reganode MJD */
1N/A Set_Node_Offset(ret, parse_start+1);
1N/A Set_Node_Cur_Length(ret); /* MJD */
1N/A RExC_parse--;
1N/A nextchar(pRExC_state);
1N/A }
1N/A }
1N/A break;
1N/A case '\0':
1N/A if (RExC_parse >= RExC_end)
1N/A FAIL("Trailing \\");
1N/A /* FALL THROUGH */
1N/A default:
1N/A /* Do not generate `unrecognized' warnings here, we fall
1N/A back into the quick-grab loop below */
1N/A parse_start--;
1N/A goto defchar;
1N/A }
1N/A break;
1N/A
1N/A case '#':
1N/A if (RExC_flags & PMf_EXTENDED) {
1N/A while (RExC_parse < RExC_end && *RExC_parse != '\n') RExC_parse++;
1N/A if (RExC_parse < RExC_end)
1N/A goto tryagain;
1N/A }
1N/A /* FALL THROUGH */
1N/A
1N/A default: {
1N/A register STRLEN len;
1N/A register UV ender;
1N/A register char *p;
1N/A char *oldp, *s;
1N/A STRLEN numlen;
1N/A STRLEN foldlen;
1N/A U8 tmpbuf[UTF8_MAXLEN_FOLD+1], *foldbuf;
1N/A
1N/A parse_start = RExC_parse - 1;
1N/A
1N/A RExC_parse++;
1N/A
1N/A defchar:
1N/A ender = 0;
1N/A ret = reg_node(pRExC_state,
1N/A (U8)(FOLD ? (LOC ? EXACTFL : EXACTF) : EXACT));
1N/A s = STRING(ret);
1N/A for (len = 0, p = RExC_parse - 1;
1N/A len < 127 && p < RExC_end;
1N/A len++)
1N/A {
1N/A oldp = p;
1N/A
1N/A if (RExC_flags & PMf_EXTENDED)
1N/A p = regwhite(p, RExC_end);
1N/A switch (*p) {
1N/A case '^':
1N/A case '$':
1N/A case '.':
1N/A case '[':
1N/A case '(':
1N/A case ')':
1N/A case '|':
1N/A goto loopdone;
1N/A case '\\':
1N/A switch (*++p) {
1N/A case 'A':
1N/A case 'C':
1N/A case 'X':
1N/A case 'G':
1N/A case 'Z':
1N/A case 'z':
1N/A case 'w':
1N/A case 'W':
1N/A case 'b':
1N/A case 'B':
1N/A case 's':
1N/A case 'S':
1N/A case 'd':
1N/A case 'D':
1N/A case 'p':
1N/A case 'P':
1N/A --p;
1N/A goto loopdone;
1N/A case 'n':
1N/A ender = '\n';
1N/A p++;
1N/A break;
1N/A case 'r':
1N/A ender = '\r';
1N/A p++;
1N/A break;
1N/A case 't':
1N/A ender = '\t';
1N/A p++;
1N/A break;
1N/A case 'f':
1N/A ender = '\f';
1N/A p++;
1N/A break;
1N/A case 'e':
1N/A ender = ASCII_TO_NATIVE('\033');
1N/A p++;
1N/A break;
1N/A case 'a':
1N/A ender = ASCII_TO_NATIVE('\007');
1N/A p++;
1N/A break;
1N/A case 'x':
1N/A if (*++p == '{') {
1N/A char* e = strchr(p, '}');
1N/A
1N/A if (!e) {
1N/A RExC_parse = p + 1;
1N/A vFAIL("Missing right brace on \\x{}");
1N/A }
1N/A else {
1N/A I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
1N/A | PERL_SCAN_DISALLOW_PREFIX;
1N/A numlen = e - p - 1;
1N/A ender = grok_hex(p + 1, &numlen, &flags, NULL);
1N/A if (ender > 0xff)
1N/A RExC_utf8 = 1;
1N/A p = e + 1;
1N/A }
1N/A }
1N/A else {
1N/A I32 flags = PERL_SCAN_DISALLOW_PREFIX;
1N/A numlen = 2;
1N/A ender = grok_hex(p, &numlen, &flags, NULL);
1N/A p += numlen;
1N/A }
1N/A break;
1N/A case 'c':
1N/A p++;
1N/A ender = UCHARAT(p++);
1N/A ender = toCTRL(ender);
1N/A break;
1N/A case '0': case '1': case '2': case '3':case '4':
1N/A case '5': case '6': case '7': case '8':case '9':
1N/A if (*p == '0' ||
1N/A (isDIGIT(p[1]) && atoi(p) >= RExC_npar) ) {
1N/A I32 flags = 0;
1N/A numlen = 3;
1N/A ender = grok_oct(p, &numlen, &flags, NULL);
1N/A p += numlen;
1N/A }
1N/A else {
1N/A --p;
1N/A goto loopdone;
1N/A }
1N/A break;
1N/A case '\0':
1N/A if (p >= RExC_end)
1N/A FAIL("Trailing \\");
1N/A /* FALL THROUGH */
1N/A default:
1N/A if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(*p))
1N/A vWARN2(p + 1, "Unrecognized escape \\%c passed through", UCHARAT(p));
1N/A goto normal_default;
1N/A }
1N/A break;
1N/A default:
1N/A normal_default:
1N/A if (UTF8_IS_START(*p) && UTF) {
1N/A ender = utf8n_to_uvchr((U8*)p, RExC_end - p,
1N/A &numlen, 0);
1N/A p += numlen;
1N/A }
1N/A else
1N/A ender = *p++;
1N/A break;
1N/A }
1N/A if (RExC_flags & PMf_EXTENDED)
1N/A p = regwhite(p, RExC_end);
1N/A if (UTF && FOLD) {
1N/A /* Prime the casefolded buffer. */
1N/A ender = toFOLD_uni(ender, tmpbuf, &foldlen);
1N/A }
1N/A if (ISMULT2(p)) { /* Back off on ?+*. */
1N/A if (len)
1N/A p = oldp;
1N/A else if (UTF) {
1N/A STRLEN unilen;
1N/A
1N/A if (FOLD) {
1N/A /* Emit all the Unicode characters. */
1N/A for (foldbuf = tmpbuf;
1N/A foldlen;
1N/A foldlen -= numlen) {
1N/A ender = utf8_to_uvchr(foldbuf, &numlen);
1N/A if (numlen > 0) {
1N/A reguni(pRExC_state, ender, s, &unilen);
1N/A s += unilen;
1N/A len += unilen;
1N/A /* In EBCDIC the numlen
1N/A * and unilen can differ. */
1N/A foldbuf += numlen;
1N/A if (numlen >= foldlen)
1N/A break;
1N/A }
1N/A else
1N/A break; /* "Can't happen." */
1N/A }
1N/A }
1N/A else {
1N/A reguni(pRExC_state, ender, s, &unilen);
1N/A if (unilen > 0) {
1N/A s += unilen;
1N/A len += unilen;
1N/A }
1N/A }
1N/A }
1N/A else {
1N/A len++;
1N/A REGC((char)ender, s++);
1N/A }
1N/A break;
1N/A }
1N/A if (UTF) {
1N/A STRLEN unilen;
1N/A
1N/A if (FOLD) {
1N/A /* Emit all the Unicode characters. */
1N/A for (foldbuf = tmpbuf;
1N/A foldlen;
1N/A foldlen -= numlen) {
1N/A ender = utf8_to_uvchr(foldbuf, &numlen);
1N/A if (numlen > 0) {
1N/A reguni(pRExC_state, ender, s, &unilen);
1N/A len += unilen;
1N/A s += unilen;
1N/A /* In EBCDIC the numlen
1N/A * and unilen can differ. */
1N/A foldbuf += numlen;
1N/A if (numlen >= foldlen)
1N/A break;
1N/A }
1N/A else
1N/A break;
1N/A }
1N/A }
1N/A else {
1N/A reguni(pRExC_state, ender, s, &unilen);
1N/A if (unilen > 0) {
1N/A s += unilen;
1N/A len += unilen;
1N/A }
1N/A }
1N/A len--;
1N/A }
1N/A else
1N/A REGC((char)ender, s++);
1N/A }
1N/A loopdone:
1N/A RExC_parse = p - 1;
1N/A Set_Node_Cur_Length(ret); /* MJD */
1N/A nextchar(pRExC_state);
1N/A {
1N/A /* len is STRLEN which is unsigned, need to copy to signed */
1N/A IV iv = len;
1N/A if (iv < 0)
1N/A vFAIL("Internal disaster");
1N/A }
1N/A if (len > 0)
1N/A *flagp |= HASWIDTH;
1N/A if (len == 1 && UNI_IS_INVARIANT(ender))
1N/A *flagp |= SIMPLE;
1N/A if (!SIZE_ONLY)
1N/A STR_LEN(ret) = len;
1N/A if (SIZE_ONLY)
1N/A RExC_size += STR_SZ(len);
1N/A else
1N/A RExC_emit += STR_SZ(len);
1N/A }
1N/A break;
1N/A }
1N/A
1N/A /* If the encoding pragma is in effect recode the text of
1N/A * any EXACT-kind nodes. */
1N/A if (PL_encoding && PL_regkind[(U8)OP(ret)] == EXACT) {
1N/A STRLEN oldlen = STR_LEN(ret);
1N/A SV *sv = sv_2mortal(newSVpvn(STRING(ret), oldlen));
1N/A
1N/A if (RExC_utf8)
1N/A SvUTF8_on(sv);
1N/A if (sv_utf8_downgrade(sv, TRUE)) {
1N/A char *s = sv_recode_to_utf8(sv, PL_encoding);
1N/A STRLEN newlen = SvCUR(sv);
1N/A
1N/A if (SvUTF8(sv))
1N/A RExC_utf8 = 1;
1N/A if (!SIZE_ONLY) {
1N/A DEBUG_r(PerlIO_printf(Perl_debug_log, "recode %*s to %*s\n",
1N/A (int)oldlen, STRING(ret),
1N/A (int)newlen, s));
1N/A Copy(s, STRING(ret), newlen, char);
1N/A STR_LEN(ret) += newlen - oldlen;
1N/A RExC_emit += STR_SZ(newlen) - STR_SZ(oldlen);
1N/A } else
1N/A RExC_size += STR_SZ(newlen) - STR_SZ(oldlen);
1N/A }
1N/A }
1N/A
1N/A return(ret);
1N/A}
1N/A
1N/ASTATIC char *
1N/AS_regwhite(pTHX_ char *p, char *e)
1N/A{
1N/A while (p < e) {
1N/A if (isSPACE(*p))
1N/A ++p;
1N/A else if (*p == '#') {
1N/A do {
1N/A p++;
1N/A } while (p < e && *p != '\n');
1N/A }
1N/A else
1N/A break;
1N/A }
1N/A return p;
1N/A}
1N/A
1N/A/* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]].
1N/A Character classes ([:foo:]) can also be negated ([:^foo:]).
1N/A Returns a named class id (ANYOF_XXX) if successful, -1 otherwise.
1N/A Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed,
1N/A but trigger failures because they are currently unimplemented. */
1N/A
1N/A#define POSIXCC_DONE(c) ((c) == ':')
1N/A#define POSIXCC_NOTYET(c) ((c) == '=' || (c) == '.')
1N/A#define POSIXCC(c) (POSIXCC_DONE(c) || POSIXCC_NOTYET(c))
1N/A
1N/ASTATIC I32
1N/AS_regpposixcc(pTHX_ RExC_state_t *pRExC_state, I32 value)
1N/A{
1N/A char *posixcc = 0;
1N/A I32 namedclass = OOB_NAMEDCLASS;
1N/A
1N/A if (value == '[' && RExC_parse + 1 < RExC_end &&
1N/A /* I smell either [: or [= or [. -- POSIX has been here, right? */
1N/A POSIXCC(UCHARAT(RExC_parse))) {
1N/A char c = UCHARAT(RExC_parse);
1N/A char* s = RExC_parse++;
1N/A
1N/A while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != c)
1N/A RExC_parse++;
1N/A if (RExC_parse == RExC_end)
1N/A /* Grandfather lone [:, [=, [. */
1N/A RExC_parse = s;
1N/A else {
1N/A char* t = RExC_parse++; /* skip over the c */
1N/A
1N/A if (UCHARAT(RExC_parse) == ']') {
1N/A RExC_parse++; /* skip over the ending ] */
1N/A posixcc = s + 1;
1N/A if (*s == ':') {
1N/A I32 complement = *posixcc == '^' ? *posixcc++ : 0;
1N/A I32 skip = 5; /* the most common skip */
1N/A
1N/A switch (*posixcc) {
1N/A case 'a':
1N/A if (strnEQ(posixcc, "alnum", 5))
1N/A namedclass =
1N/A complement ? ANYOF_NALNUMC : ANYOF_ALNUMC;
1N/A else if (strnEQ(posixcc, "alpha", 5))
1N/A namedclass =
1N/A complement ? ANYOF_NALPHA : ANYOF_ALPHA;
1N/A else if (strnEQ(posixcc, "ascii", 5))
1N/A namedclass =
1N/A complement ? ANYOF_NASCII : ANYOF_ASCII;
1N/A break;
1N/A case 'b':
1N/A if (strnEQ(posixcc, "blank", 5))
1N/A namedclass =
1N/A complement ? ANYOF_NBLANK : ANYOF_BLANK;
1N/A break;
1N/A case 'c':
1N/A if (strnEQ(posixcc, "cntrl", 5))
1N/A namedclass =
1N/A complement ? ANYOF_NCNTRL : ANYOF_CNTRL;
1N/A break;
1N/A case 'd':
1N/A if (strnEQ(posixcc, "digit", 5))
1N/A namedclass =
1N/A complement ? ANYOF_NDIGIT : ANYOF_DIGIT;
1N/A break;
1N/A case 'g':
1N/A if (strnEQ(posixcc, "graph", 5))
1N/A namedclass =
1N/A complement ? ANYOF_NGRAPH : ANYOF_GRAPH;
1N/A break;
1N/A case 'l':
1N/A if (strnEQ(posixcc, "lower", 5))
1N/A namedclass =
1N/A complement ? ANYOF_NLOWER : ANYOF_LOWER;
1N/A break;
1N/A case 'p':
1N/A if (strnEQ(posixcc, "print", 5))
1N/A namedclass =
1N/A complement ? ANYOF_NPRINT : ANYOF_PRINT;
1N/A else if (strnEQ(posixcc, "punct", 5))
1N/A namedclass =
1N/A complement ? ANYOF_NPUNCT : ANYOF_PUNCT;
1N/A break;
1N/A case 's':
1N/A if (strnEQ(posixcc, "space", 5))
1N/A namedclass =
1N/A complement ? ANYOF_NPSXSPC : ANYOF_PSXSPC;
1N/A break;
1N/A case 'u':
1N/A if (strnEQ(posixcc, "upper", 5))
1N/A namedclass =
1N/A complement ? ANYOF_NUPPER : ANYOF_UPPER;
1N/A break;
1N/A case 'w': /* this is not POSIX, this is the Perl \w */
1N/A if (strnEQ(posixcc, "word", 4)) {
1N/A namedclass =
1N/A complement ? ANYOF_NALNUM : ANYOF_ALNUM;
1N/A skip = 4;
1N/A }
1N/A break;
1N/A case 'x':
1N/A if (strnEQ(posixcc, "xdigit", 6)) {
1N/A namedclass =
1N/A complement ? ANYOF_NXDIGIT : ANYOF_XDIGIT;
1N/A skip = 6;
1N/A }
1N/A break;
1N/A }
1N/A if (namedclass == OOB_NAMEDCLASS ||
1N/A posixcc[skip] != ':' ||
1N/A posixcc[skip+1] != ']')
1N/A {
1N/A Simple_vFAIL3("POSIX class [:%.*s:] unknown",
1N/A t - s - 1, s + 1);
1N/A }
1N/A } else if (!SIZE_ONLY) {
1N/A /* [[=foo=]] and [[.foo.]] are still future. */
1N/A
1N/A /* adjust RExC_parse so the warning shows after
1N/A the class closes */
1N/A while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse) != ']')
1N/A RExC_parse++;
1N/A Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
1N/A }
1N/A } else {
1N/A /* Maternal grandfather:
1N/A * "[:" ending in ":" but not in ":]" */
1N/A RExC_parse = s;
1N/A }
1N/A }
1N/A }
1N/A
1N/A return namedclass;
1N/A}
1N/A
1N/ASTATIC void
1N/AS_checkposixcc(pTHX_ RExC_state_t *pRExC_state)
1N/A{
1N/A if (!SIZE_ONLY && POSIXCC(UCHARAT(RExC_parse))) {
1N/A char *s = RExC_parse;
1N/A char c = *s++;
1N/A
1N/A while(*s && isALNUM(*s))
1N/A s++;
1N/A if (*s && c == *s && s[1] == ']') {
1N/A if (ckWARN(WARN_REGEXP))
1N/A vWARN3(s+2,
1N/A "POSIX syntax [%c %c] belongs inside character classes",
1N/A c, c);
1N/A
1N/A /* [[=foo=]] and [[.foo.]] are still future. */
1N/A if (POSIXCC_NOTYET(c)) {
1N/A /* adjust RExC_parse so the error shows after
1N/A the class closes */
1N/A while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse++) != ']')
1N/A ;
1N/A Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
1N/A }
1N/A }
1N/A }
1N/A}
1N/A
1N/ASTATIC regnode *
1N/AS_regclass(pTHX_ RExC_state_t *pRExC_state)
1N/A{
1N/A register UV value;
1N/A register UV nextvalue;
1N/A register IV prevvalue = OOB_UNICODE;
1N/A register IV range = 0;
1N/A register regnode *ret;
1N/A STRLEN numlen;
1N/A IV namedclass;
1N/A char *rangebegin = 0;
1N/A bool need_class = 0;
1N/A SV *listsv = Nullsv;
1N/A register char *e;
1N/A UV n;
1N/A bool optimize_invert = TRUE;
1N/A AV* unicode_alternate = 0;
1N/A#ifdef EBCDIC
1N/A UV literal_endpoint = 0;
1N/A#endif
1N/A
1N/A ret = reganode(pRExC_state, ANYOF, 0);
1N/A
1N/A if (!SIZE_ONLY)
1N/A ANYOF_FLAGS(ret) = 0;
1N/A
1N/A if (UCHARAT(RExC_parse) == '^') { /* Complement of range. */
1N/A RExC_naughty++;
1N/A RExC_parse++;
1N/A if (!SIZE_ONLY)
1N/A ANYOF_FLAGS(ret) |= ANYOF_INVERT;
1N/A }
1N/A
1N/A if (SIZE_ONLY)
1N/A RExC_size += ANYOF_SKIP;
1N/A else {
1N/A RExC_emit += ANYOF_SKIP;
1N/A if (FOLD)
1N/A ANYOF_FLAGS(ret) |= ANYOF_FOLD;
1N/A if (LOC)
1N/A ANYOF_FLAGS(ret) |= ANYOF_LOCALE;
1N/A ANYOF_BITMAP_ZERO(ret);
1N/A listsv = newSVpvn("# comment\n", 10);
1N/A }
1N/A
1N/A nextvalue = RExC_parse < RExC_end ? UCHARAT(RExC_parse) : 0;
1N/A
1N/A if (!SIZE_ONLY && POSIXCC(nextvalue))
1N/A checkposixcc(pRExC_state);
1N/A
1N/A /* allow 1st char to be ] (allowing it to be - is dealt with later) */
1N/A if (UCHARAT(RExC_parse) == ']')
1N/A goto charclassloop;
1N/A
1N/A while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != ']') {
1N/A
1N/A charclassloop:
1N/A
1N/A namedclass = OOB_NAMEDCLASS; /* initialize as illegal */
1N/A
1N/A if (!range)
1N/A rangebegin = RExC_parse;
1N/A if (UTF) {
1N/A value = utf8n_to_uvchr((U8*)RExC_parse,
1N/A RExC_end - RExC_parse,
1N/A &numlen, 0);
1N/A RExC_parse += numlen;
1N/A }
1N/A else
1N/A value = UCHARAT(RExC_parse++);
1N/A nextvalue = RExC_parse < RExC_end ? UCHARAT(RExC_parse) : 0;
1N/A if (value == '[' && POSIXCC(nextvalue))
1N/A namedclass = regpposixcc(pRExC_state, value);
1N/A else if (value == '\\') {
1N/A if (UTF) {
1N/A value = utf8n_to_uvchr((U8*)RExC_parse,
1N/A RExC_end - RExC_parse,
1N/A &numlen, 0);
1N/A RExC_parse += numlen;
1N/A }
1N/A else
1N/A value = UCHARAT(RExC_parse++);
1N/A /* Some compilers cannot handle switching on 64-bit integer
1N/A * values, therefore value cannot be an UV. Yes, this will
1N/A * be a problem later if we want switch on Unicode.
1N/A * A similar issue a little bit later when switching on
1N/A * namedclass. --jhi */
1N/A switch ((I32)value) {
1N/A case 'w': namedclass = ANYOF_ALNUM; break;
1N/A case 'W': namedclass = ANYOF_NALNUM; break;
1N/A case 's': namedclass = ANYOF_SPACE; break;
1N/A case 'S': namedclass = ANYOF_NSPACE; break;
1N/A case 'd': namedclass = ANYOF_DIGIT; break;
1N/A case 'D': namedclass = ANYOF_NDIGIT; break;
1N/A case 'p':
1N/A case 'P':
1N/A if (RExC_parse >= RExC_end)
1N/A vFAIL2("Empty \\%c{}", (U8)value);
1N/A if (*RExC_parse == '{') {
1N/A U8 c = (U8)value;
1N/A e = strchr(RExC_parse++, '}');
1N/A if (!e)
1N/A vFAIL2("Missing right brace on \\%c{}", c);
1N/A while (isSPACE(UCHARAT(RExC_parse)))
1N/A RExC_parse++;
1N/A if (e == RExC_parse)
1N/A vFAIL2("Empty \\%c{}", c);
1N/A n = e - RExC_parse;
1N/A while (isSPACE(UCHARAT(RExC_parse + n - 1)))
1N/A n--;
1N/A }
1N/A else {
1N/A e = RExC_parse;
1N/A n = 1;
1N/A }
1N/A if (!SIZE_ONLY) {
1N/A if (UCHARAT(RExC_parse) == '^') {
1N/A RExC_parse++;
1N/A n--;
1N/A value = value == 'p' ? 'P' : 'p'; /* toggle */
1N/A while (isSPACE(UCHARAT(RExC_parse))) {
1N/A RExC_parse++;
1N/A n--;
1N/A }
1N/A }
1N/A if (value == 'p')
1N/A Perl_sv_catpvf(aTHX_ listsv,
1N/A "+utf8::%.*s\n", (int)n, RExC_parse);
1N/A else
1N/A Perl_sv_catpvf(aTHX_ listsv,
1N/A "!utf8::%.*s\n", (int)n, RExC_parse);
1N/A }
1N/A RExC_parse = e + 1;
1N/A ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
1N/A continue;
1N/A case 'n': value = '\n'; break;
1N/A case 'r': value = '\r'; break;
1N/A case 't': value = '\t'; break;
1N/A case 'f': value = '\f'; break;
1N/A case 'b': value = '\b'; break;
1N/A case 'e': value = ASCII_TO_NATIVE('\033');break;
1N/A case 'a': value = ASCII_TO_NATIVE('\007');break;
1N/A case 'x':
1N/A if (*RExC_parse == '{') {
1N/A I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
1N/A | PERL_SCAN_DISALLOW_PREFIX;
1N/A e = strchr(RExC_parse++, '}');
1N/A if (!e)
1N/A vFAIL("Missing right brace on \\x{}");
1N/A
1N/A numlen = e - RExC_parse;
1N/A value = grok_hex(RExC_parse, &numlen, &flags, NULL);
1N/A RExC_parse = e + 1;
1N/A }
1N/A else {
1N/A I32 flags = PERL_SCAN_DISALLOW_PREFIX;
1N/A numlen = 2;
1N/A value = grok_hex(RExC_parse, &numlen, &flags, NULL);
1N/A RExC_parse += numlen;
1N/A }
1N/A break;
1N/A case 'c':
1N/A value = UCHARAT(RExC_parse++);
1N/A value = toCTRL(value);
1N/A break;
1N/A case '0': case '1': case '2': case '3': case '4':
1N/A case '5': case '6': case '7': case '8': case '9':
1N/A {
1N/A I32 flags = 0;
1N/A numlen = 3;
1N/A value = grok_oct(--RExC_parse, &numlen, &flags, NULL);
1N/A RExC_parse += numlen;
1N/A break;
1N/A }
1N/A default:
1N/A if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(value))
1N/A vWARN2(RExC_parse,
1N/A "Unrecognized escape \\%c in character class passed through",
1N/A (int)value);
1N/A break;
1N/A }
1N/A } /* end of \blah */
1N/A#ifdef EBCDIC
1N/A else
1N/A literal_endpoint++;
1N/A#endif
1N/A
1N/A if (namedclass > OOB_NAMEDCLASS) { /* this is a named class \blah */
1N/A
1N/A if (!SIZE_ONLY && !need_class)
1N/A ANYOF_CLASS_ZERO(ret);
1N/A
1N/A need_class = 1;
1N/A
1N/A /* a bad range like a-\d, a-[:digit:] ? */
1N/A if (range) {
1N/A if (!SIZE_ONLY) {
1N/A if (ckWARN(WARN_REGEXP))
1N/A vWARN4(RExC_parse,
1N/A "False [] range \"%*.*s\"",
1N/A RExC_parse - rangebegin,
1N/A RExC_parse - rangebegin,
1N/A rangebegin);
1N/A if (prevvalue < 256) {
1N/A ANYOF_BITMAP_SET(ret, prevvalue);
1N/A ANYOF_BITMAP_SET(ret, '-');
1N/A }
1N/A else {
1N/A ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
1N/A Perl_sv_catpvf(aTHX_ listsv,
1N/A "%04"UVxf"\n%04"UVxf"\n", (UV)prevvalue, (UV) '-');
1N/A }
1N/A }
1N/A
1N/A range = 0; /* this was not a true range */
1N/A }
1N/A
1N/A if (!SIZE_ONLY) {
1N/A if (namedclass > OOB_NAMEDCLASS)
1N/A optimize_invert = FALSE;
1N/A /* Possible truncation here but in some 64-bit environments
1N/A * the compiler gets heartburn about switch on 64-bit values.
1N/A * A similar issue a little earlier when switching on value.
1N/A * --jhi */
1N/A switch ((I32)namedclass) {
1N/A case ANYOF_ALNUM:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_ALNUM);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isALNUM(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsWord\n");
1N/A break;
1N/A case ANYOF_NALNUM:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NALNUM);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isALNUM(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsWord\n");
1N/A break;
1N/A case ANYOF_ALNUMC:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_ALNUMC);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isALNUMC(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsAlnum\n");
1N/A break;
1N/A case ANYOF_NALNUMC:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NALNUMC);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isALNUMC(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsAlnum\n");
1N/A break;
1N/A case ANYOF_ALPHA:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_ALPHA);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isALPHA(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsAlpha\n");
1N/A break;
1N/A case ANYOF_NALPHA:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NALPHA);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isALPHA(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsAlpha\n");
1N/A break;
1N/A case ANYOF_ASCII:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_ASCII);
1N/A else {
1N/A#ifndef EBCDIC
1N/A for (value = 0; value < 128; value++)
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A#else /* EBCDIC */
1N/A for (value = 0; value < 256; value++) {
1N/A if (isASCII(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A#endif /* EBCDIC */
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsASCII\n");
1N/A break;
1N/A case ANYOF_NASCII:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NASCII);
1N/A else {
1N/A#ifndef EBCDIC
1N/A for (value = 128; value < 256; value++)
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A#else /* EBCDIC */
1N/A for (value = 0; value < 256; value++) {
1N/A if (!isASCII(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A#endif /* EBCDIC */
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsASCII\n");
1N/A break;
1N/A case ANYOF_BLANK:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_BLANK);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isBLANK(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsBlank\n");
1N/A break;
1N/A case ANYOF_NBLANK:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NBLANK);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isBLANK(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsBlank\n");
1N/A break;
1N/A case ANYOF_CNTRL:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_CNTRL);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isCNTRL(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsCntrl\n");
1N/A break;
1N/A case ANYOF_NCNTRL:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NCNTRL);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isCNTRL(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsCntrl\n");
1N/A break;
1N/A case ANYOF_DIGIT:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_DIGIT);
1N/A else {
1N/A /* consecutive digits assumed */
1N/A for (value = '0'; value <= '9'; value++)
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsDigit\n");
1N/A break;
1N/A case ANYOF_NDIGIT:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NDIGIT);
1N/A else {
1N/A /* consecutive digits assumed */
1N/A for (value = 0; value < '0'; value++)
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A for (value = '9' + 1; value < 256; value++)
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsDigit\n");
1N/A break;
1N/A case ANYOF_GRAPH:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_GRAPH);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isGRAPH(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsGraph\n");
1N/A break;
1N/A case ANYOF_NGRAPH:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NGRAPH);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isGRAPH(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsGraph\n");
1N/A break;
1N/A case ANYOF_LOWER:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_LOWER);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isLOWER(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsLower\n");
1N/A break;
1N/A case ANYOF_NLOWER:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NLOWER);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isLOWER(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsLower\n");
1N/A break;
1N/A case ANYOF_PRINT:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_PRINT);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isPRINT(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsPrint\n");
1N/A break;
1N/A case ANYOF_NPRINT:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NPRINT);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isPRINT(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsPrint\n");
1N/A break;
1N/A case ANYOF_PSXSPC:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_PSXSPC);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isPSXSPC(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsSpace\n");
1N/A break;
1N/A case ANYOF_NPSXSPC:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NPSXSPC);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isPSXSPC(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsSpace\n");
1N/A break;
1N/A case ANYOF_PUNCT:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_PUNCT);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isPUNCT(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsPunct\n");
1N/A break;
1N/A case ANYOF_NPUNCT:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NPUNCT);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isPUNCT(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsPunct\n");
1N/A break;
1N/A case ANYOF_SPACE:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_SPACE);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isSPACE(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsSpacePerl\n");
1N/A break;
1N/A case ANYOF_NSPACE:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NSPACE);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isSPACE(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsSpacePerl\n");
1N/A break;
1N/A case ANYOF_UPPER:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_UPPER);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isUPPER(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsUpper\n");
1N/A break;
1N/A case ANYOF_NUPPER:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NUPPER);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isUPPER(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsUpper\n");
1N/A break;
1N/A case ANYOF_XDIGIT:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_XDIGIT);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (isXDIGIT(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsXDigit\n");
1N/A break;
1N/A case ANYOF_NXDIGIT:
1N/A if (LOC)
1N/A ANYOF_CLASS_SET(ret, ANYOF_NXDIGIT);
1N/A else {
1N/A for (value = 0; value < 256; value++)
1N/A if (!isXDIGIT(value))
1N/A ANYOF_BITMAP_SET(ret, value);
1N/A }
1N/A Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsXDigit\n");
1N/A break;
1N/A default:
1N/A vFAIL("Invalid [::] class");
1N/A break;
1N/A }
1N/A if (LOC)
1N/A ANYOF_FLAGS(ret) |= ANYOF_CLASS;
1N/A continue;
1N/A }
1N/A } /* end of namedclass \blah */
1N/A
1N/A if (range) {
1N/A if (prevvalue > (IV)value) /* b-a */ {
1N/A Simple_vFAIL4("Invalid [] range \"%*.*s\"",
1N/A RExC_parse - rangebegin,
1N/A RExC_parse - rangebegin,
1N/A rangebegin);
1N/A range = 0; /* not a valid range */
1N/A }
1N/A }
1N/A else {
1N/A prevvalue = value; /* save the beginning of the range */
1N/A if (*RExC_parse == '-' && RExC_parse+1 < RExC_end &&
1N/A RExC_parse[1] != ']') {
1N/A RExC_parse++;
1N/A
1N/A /* a bad range like \w-, [:word:]- ? */
1N/A if (namedclass > OOB_NAMEDCLASS) {
1N/A if (ckWARN(WARN_REGEXP))
1N/A vWARN4(RExC_parse,
1N/A "False [] range \"%*.*s\"",
1N/A RExC_parse - rangebegin,
1N/A RExC_parse - rangebegin,
1N/A rangebegin);
1N/A if (!SIZE_ONLY)
1N/A ANYOF_BITMAP_SET(ret, '-');
1N/A } else
1N/A range = 1; /* yeah, it's a range! */
1N/A continue; /* but do it the next time */
1N/A }
1N/A }
1N/A
1N/A /* now is the next time */
1N/A if (!SIZE_ONLY) {
1N/A IV i;
1N/A
1N/A if (prevvalue < 256) {
1N/A IV ceilvalue = value < 256 ? value : 255;
1N/A
1N/A#ifdef EBCDIC
1N/A /* In EBCDIC [\x89-\x91] should include
1N/A * the \x8e but [i-j] should not. */
1N/A if (literal_endpoint == 2 &&
1N/A ((isLOWER(prevvalue) && isLOWER(ceilvalue)) ||
1N/A (isUPPER(prevvalue) && isUPPER(ceilvalue))))
1N/A {
1N/A if (isLOWER(prevvalue)) {
1N/A for (i = prevvalue; i <= ceilvalue; i++)
1N/A if (isLOWER(i))
1N/A ANYOF_BITMAP_SET(ret, i);
1N/A } else {
1N/A for (i = prevvalue; i <= ceilvalue; i++)
1N/A if (isUPPER(i))
1N/A ANYOF_BITMAP_SET(ret, i);
1N/A }
1N/A }
1N/A else
1N/A#endif
1N/A for (i = prevvalue; i <= ceilvalue; i++)
1N/A ANYOF_BITMAP_SET(ret, i);
1N/A }
1N/A if (value > 255 || UTF) {
1N/A UV prevnatvalue = NATIVE_TO_UNI(prevvalue);
1N/A UV natvalue = NATIVE_TO_UNI(value);
1N/A
1N/A ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
1N/A if (prevnatvalue < natvalue) { /* what about > ? */
1N/A Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\t%04"UVxf"\n",
1N/A prevnatvalue, natvalue);
1N/A }
1N/A else if (prevnatvalue == natvalue) {
1N/A Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n", natvalue);
1N/A if (FOLD) {
1N/A U8 foldbuf[UTF8_MAXLEN_FOLD+1];
1N/A STRLEN foldlen;
1N/A UV f = to_uni_fold(natvalue, foldbuf, &foldlen);
1N/A
1N/A /* If folding and foldable and a single
1N/A * character, insert also the folded version
1N/A * to the charclass. */
1N/A if (f != value) {
1N/A if (foldlen == (STRLEN)UNISKIP(f))
1N/A Perl_sv_catpvf(aTHX_ listsv,
1N/A "%04"UVxf"\n", f);
1N/A else {
1N/A /* Any multicharacter foldings
1N/A * require the following transform:
1N/A * [ABCDEF] -> (?:[ABCabcDEFd]|pq|rst)
1N/A * where E folds into "pq" and F folds
1N/A * into "rst", all other characters
1N/A * fold to single characters. We save
1N/A * away these multicharacter foldings,
1N/A * to be later saved as part of the
1N/A * additional "s" data. */
1N/A SV *sv;
1N/A
1N/A if (!unicode_alternate)
1N/A unicode_alternate = newAV();
1N/A sv = newSVpvn((char*)foldbuf, foldlen);
1N/A SvUTF8_on(sv);
1N/A av_push(unicode_alternate, sv);
1N/A }
1N/A }
1N/A
1N/A /* If folding and the value is one of the Greek
1N/A * sigmas insert a few more sigmas to make the
1N/A * folding rules of the sigmas to work right.
1N/A * Note that not all the possible combinations
1N/A * are handled here: some of them are handled
1N/A * by the standard folding rules, and some of
1N/A * them (literal or EXACTF cases) are handled
1N/A * during runtime in regexec.c:S_find_byclass(). */
1N/A if (value == UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA) {
1N/A Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n",
1N/A (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA);
1N/A Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n",
1N/A (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA);
1N/A }
1N/A else if (value == UNICODE_GREEK_CAPITAL_LETTER_SIGMA)
1N/A Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n",
1N/A (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA);
1N/A }
1N/A }
1N/A }
1N/A#ifdef EBCDIC
1N/A literal_endpoint = 0;
1N/A#endif
1N/A }
1N/A
1N/A range = 0; /* this range (if it was one) is done now */
1N/A }
1N/A
1N/A if (need_class) {
1N/A ANYOF_FLAGS(ret) |= ANYOF_LARGE;
1N/A if (SIZE_ONLY)
1N/A RExC_size += ANYOF_CLASS_ADD_SKIP;
1N/A else
1N/A RExC_emit += ANYOF_CLASS_ADD_SKIP;
1N/A }
1N/A
1N/A /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */
1N/A if (!SIZE_ONLY &&
1N/A /* If the only flag is folding (plus possibly inversion). */
1N/A ((ANYOF_FLAGS(ret) & (ANYOF_FLAGS_ALL ^ ANYOF_INVERT)) == ANYOF_FOLD)
1N/A ) {
1N/A for (value = 0; value < 256; ++value) {
1N/A if (ANYOF_BITMAP_TEST(ret, value)) {
1N/A UV fold = PL_fold[value];
1N/A
1N/A if (fold != value)
1N/A ANYOF_BITMAP_SET(ret, fold);
1N/A }
1N/A }
1N/A ANYOF_FLAGS(ret) &= ~ANYOF_FOLD;
1N/A }
1N/A
1N/A /* optimize inverted simple patterns (e.g. [^a-z]) */
1N/A if (!SIZE_ONLY && optimize_invert &&
1N/A /* If the only flag is inversion. */
1N/A (ANYOF_FLAGS(ret) & ANYOF_FLAGS_ALL) == ANYOF_INVERT) {
1N/A for (value = 0; value < ANYOF_BITMAP_SIZE; ++value)
1N/A ANYOF_BITMAP(ret)[value] ^= ANYOF_FLAGS_ALL;
1N/A ANYOF_FLAGS(ret) = ANYOF_UNICODE_ALL;
1N/A }
1N/A
1N/A if (!SIZE_ONLY) {
1N/A AV *av = newAV();
1N/A SV *rv;
1N/A
1N/A /* The 0th element stores the character class description
1N/A * in its textual form: used later (regexec.c:Perl_regclass_swash())
1N/A * to initialize the appropriate swash (which gets stored in
1N/A * the 1st element), and also useful for dumping the regnode.
1N/A * The 2nd element stores the multicharacter foldings,
1N/A * used later (regexec.c:S_reginclass()). */
1N/A av_store(av, 0, listsv);
1N/A av_store(av, 1, NULL);
1N/A av_store(av, 2, (SV*)unicode_alternate);
1N/A rv = newRV_noinc((SV*)av);
1N/A n = add_data(pRExC_state, 1, "s");
1N/A RExC_rx->data->data[n] = (void*)rv;
1N/A ARG_SET(ret, n);
1N/A }
1N/A
1N/A return ret;
1N/A}
1N/A
1N/ASTATIC char*
1N/AS_nextchar(pTHX_ RExC_state_t *pRExC_state)
1N/A{
1N/A char* retval = RExC_parse++;
1N/A
1N/A for (;;) {
1N/A if (*RExC_parse == '(' && RExC_parse[1] == '?' &&
1N/A RExC_parse[2] == '#') {
1N/A while (*RExC_parse != ')') {
1N/A if (RExC_parse == RExC_end)
1N/A FAIL("Sequence (?#... not terminated");
1N/A RExC_parse++;
1N/A }
1N/A RExC_parse++;
1N/A continue;
1N/A }
1N/A if (RExC_flags & PMf_EXTENDED) {
1N/A if (isSPACE(*RExC_parse)) {
1N/A RExC_parse++;
1N/A continue;
1N/A }
1N/A else if (*RExC_parse == '#') {
1N/A while (RExC_parse < RExC_end)
1N/A if (*RExC_parse++ == '\n') break;
1N/A continue;
1N/A }
1N/A }
1N/A return retval;
1N/A }
1N/A}
1N/A
1N/A/*
1N/A- reg_node - emit a node
1N/A*/
1N/ASTATIC regnode * /* Location. */
1N/AS_reg_node(pTHX_ RExC_state_t *pRExC_state, U8 op)
1N/A{
1N/A register regnode *ret;
1N/A register regnode *ptr;
1N/A
1N/A ret = RExC_emit;
1N/A if (SIZE_ONLY) {
1N/A SIZE_ALIGN(RExC_size);
1N/A RExC_size += 1;
1N/A return(ret);
1N/A }
1N/A
1N/A NODE_ALIGN_FILL(ret);
1N/A ptr = ret;
1N/A FILL_ADVANCE_NODE(ptr, op);
1N/A if (RExC_offsets) { /* MJD */
1N/A MJD_OFFSET_DEBUG(("%s:%u: (op %s) %s %u <- %u (len %u) (max %u).\n",
1N/A "reg_node", __LINE__,
1N/A reg_name[op],
1N/A RExC_emit - RExC_emit_start > RExC_offsets[0]
1N/A ? "Overwriting end of array!\n" : "OK",
1N/A RExC_emit - RExC_emit_start,
1N/A RExC_parse - RExC_start,
1N/A RExC_offsets[0]));
1N/A Set_Node_Offset(RExC_emit, RExC_parse + (op == END));
1N/A }
1N/A
1N/A RExC_emit = ptr;
1N/A
1N/A return(ret);
1N/A}
1N/A
1N/A/*
1N/A- reganode - emit a node with an argument
1N/A*/
1N/ASTATIC regnode * /* Location. */
1N/AS_reganode(pTHX_ RExC_state_t *pRExC_state, U8 op, U32 arg)
1N/A{
1N/A register regnode *ret;
1N/A register regnode *ptr;
1N/A
1N/A ret = RExC_emit;
1N/A if (SIZE_ONLY) {
1N/A SIZE_ALIGN(RExC_size);
1N/A RExC_size += 2;
1N/A return(ret);
1N/A }
1N/A
1N/A NODE_ALIGN_FILL(ret);
1N/A ptr = ret;
1N/A FILL_ADVANCE_NODE_ARG(ptr, op, arg);
1N/A if (RExC_offsets) { /* MJD */
1N/A MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s %u <- %u (max %u).\n",
1N/A "reganode",
1N/A __LINE__,
1N/A reg_name[op],
1N/A RExC_emit - RExC_emit_start > RExC_offsets[0] ?
1N/A "Overwriting end of array!\n" : "OK",
1N/A RExC_emit - RExC_emit_start,
1N/A RExC_parse - RExC_start,
1N/A RExC_offsets[0]));
1N/A Set_Cur_Node_Offset;
1N/A }
1N/A
1N/A RExC_emit = ptr;
1N/A
1N/A return(ret);
1N/A}
1N/A
1N/A/*
1N/A- reguni - emit (if appropriate) a Unicode character
1N/A*/
1N/ASTATIC void
1N/AS_reguni(pTHX_ RExC_state_t *pRExC_state, UV uv, char* s, STRLEN* lenp)
1N/A{
1N/A *lenp = SIZE_ONLY ? UNISKIP(uv) : (uvchr_to_utf8((U8*)s, uv) - (U8*)s);
1N/A}
1N/A
1N/A/*
1N/A- reginsert - insert an operator in front of already-emitted operand
1N/A*
1N/A* Means relocating the operand.
1N/A*/
1N/ASTATIC void
1N/AS_reginsert(pTHX_ RExC_state_t *pRExC_state, U8 op, regnode *opnd)
1N/A{
1N/A register regnode *src;
1N/A register regnode *dst;
1N/A register regnode *place;
1N/A register int offset = regarglen[(U8)op];
1N/A
1N/A/* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
1N/A
1N/A if (SIZE_ONLY) {
1N/A RExC_size += NODE_STEP_REGNODE + offset;
1N/A return;
1N/A }
1N/A
1N/A src = RExC_emit;
1N/A RExC_emit += NODE_STEP_REGNODE + offset;
1N/A dst = RExC_emit;
1N/A while (src > opnd) {
1N/A StructCopy(--src, --dst, regnode);
1N/A if (RExC_offsets) { /* MJD 20010112 */
1N/A MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s copy %u -> %u (max %u).\n",
1N/A "reg_insert",
1N/A __LINE__,
1N/A reg_name[op],
1N/A dst - RExC_emit_start > RExC_offsets[0]
1N/A ? "Overwriting end of array!\n" : "OK",
1N/A src - RExC_emit_start,
1N/A dst - RExC_emit_start,
1N/A RExC_offsets[0]));
1N/A Set_Node_Offset_To_R(dst-RExC_emit_start, Node_Offset(src));
1N/A Set_Node_Length_To_R(dst-RExC_emit_start, Node_Length(src));
1N/A }
1N/A }
1N/A
1N/A
1N/A place = opnd; /* Op node, where operand used to be. */
1N/A if (RExC_offsets) { /* MJD */
1N/A MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s %u <- %u (max %u).\n",
1N/A "reginsert",
1N/A __LINE__,
1N/A reg_name[op],
1N/A place - RExC_emit_start > RExC_offsets[0]
1N/A ? "Overwriting end of array!\n" : "OK",
1N/A place - RExC_emit_start,
1N/A RExC_parse - RExC_start,
1N/A RExC_offsets[0]));
1N/A Set_Node_Offset(place, RExC_parse);
1N/A Set_Node_Length(place, 1);
1N/A }
1N/A src = NEXTOPER(place);
1N/A FILL_ADVANCE_NODE(place, op);
1N/A Zero(src, offset, regnode);
1N/A}
1N/A
1N/A/*
1N/A- regtail - set the next-pointer at the end of a node chain of p to val.
1N/A*/
1N/ASTATIC void
1N/AS_regtail(pTHX_ RExC_state_t *pRExC_state, regnode *p, regnode *val)
1N/A{
1N/A register regnode *scan;
1N/A register regnode *temp;
1N/A
1N/A if (SIZE_ONLY)
1N/A return;
1N/A
1N/A /* Find last node. */
1N/A scan = p;
1N/A for (;;) {
1N/A temp = regnext(scan);
1N/A if (temp == NULL)
1N/A break;
1N/A scan = temp;
1N/A }
1N/A
1N/A if (reg_off_by_arg[OP(scan)]) {
1N/A ARG_SET(scan, val - scan);
1N/A }
1N/A else {
1N/A NEXT_OFF(scan) = val - scan;
1N/A }
1N/A}
1N/A
1N/A/*
1N/A- regoptail - regtail on operand of first argument; nop if operandless
1N/A*/
1N/ASTATIC void
1N/AS_regoptail(pTHX_ RExC_state_t *pRExC_state, regnode *p, regnode *val)
1N/A{
1N/A /* "Operandless" and "op != BRANCH" are synonymous in practice. */
1N/A if (p == NULL || SIZE_ONLY)
1N/A return;
1N/A if (PL_regkind[(U8)OP(p)] == BRANCH) {
1N/A regtail(pRExC_state, NEXTOPER(p), val);
1N/A }
1N/A else if ( PL_regkind[(U8)OP(p)] == BRANCHJ) {
1N/A regtail(pRExC_state, NEXTOPER(NEXTOPER(p)), val);
1N/A }
1N/A else
1N/A return;
1N/A}
1N/A
1N/A/*
1N/A - regcurly - a little FSA that accepts {\d+,?\d*}
1N/A */
1N/ASTATIC I32
1N/AS_regcurly(pTHX_ register char *s)
1N/A{
1N/A if (*s++ != '{')
1N/A return FALSE;
1N/A if (!isDIGIT(*s))
1N/A return FALSE;
1N/A while (isDIGIT(*s))
1N/A s++;
1N/A if (*s == ',')
1N/A s++;
1N/A while (isDIGIT(*s))
1N/A s++;
1N/A if (*s != '}')
1N/A return FALSE;
1N/A return TRUE;
1N/A}
1N/A
1N/A
1N/A#ifdef DEBUGGING
1N/A
1N/ASTATIC regnode *
1N/AS_dumpuntil(pTHX_ regnode *start, regnode *node, regnode *last, SV* sv, I32 l)
1N/A{
1N/A register U8 op = EXACT; /* Arbitrary non-END op. */
1N/A register regnode *next;
1N/A
1N/A while (op != END && (!last || node < last)) {
1N/A /* While that wasn't END last time... */
1N/A
1N/A NODE_ALIGN(node);
1N/A op = OP(node);
1N/A if (op == CLOSE)
1N/A l--;
1N/A next = regnext(node);
1N/A /* Where, what. */
1N/A if (OP(node) == OPTIMIZED)
1N/A goto after_print;
1N/A regprop(sv, node);
1N/A PerlIO_printf(Perl_debug_log, "%4"IVdf":%*s%s", (IV)(node - start),
1N/A (int)(2*l + 1), "", SvPVX(sv));
1N/A if (next == NULL) /* Next ptr. */
1N/A PerlIO_printf(Perl_debug_log, "(0)");
1N/A else
1N/A PerlIO_printf(Perl_debug_log, "(%"IVdf")", (IV)(next - start));
1N/A (void)PerlIO_putc(Perl_debug_log, '\n');
1N/A after_print:
1N/A if (PL_regkind[(U8)op] == BRANCHJ) {
1N/A register regnode *nnode = (OP(next) == LONGJMP
1N/A ? regnext(next)
1N/A : next);
1N/A if (last && nnode > last)
1N/A nnode = last;
1N/A node = dumpuntil(start, NEXTOPER(NEXTOPER(node)), nnode, sv, l + 1);
1N/A }
1N/A else if (PL_regkind[(U8)op] == BRANCH) {
1N/A node = dumpuntil(start, NEXTOPER(node), next, sv, l + 1);
1N/A }
1N/A else if ( op == CURLY) { /* `next' might be very big: optimizer */
1N/A node = dumpuntil(start, NEXTOPER(node) + EXTRA_STEP_2ARGS,
1N/A NEXTOPER(node) + EXTRA_STEP_2ARGS + 1, sv, l + 1);
1N/A }
1N/A else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) {
1N/A node = dumpuntil(start, NEXTOPER(node) + EXTRA_STEP_2ARGS,
1N/A next, sv, l + 1);
1N/A }
1N/A else if ( op == PLUS || op == STAR) {
1N/A node = dumpuntil(start, NEXTOPER(node), NEXTOPER(node) + 1, sv, l + 1);
1N/A }
1N/A else if (op == ANYOF) {
1N/A /* arglen 1 + class block */
1N/A node += 1 + ((ANYOF_FLAGS(node) & ANYOF_LARGE)
1N/A ? ANYOF_CLASS_SKIP : ANYOF_SKIP);
1N/A node = NEXTOPER(node);
1N/A }
1N/A else if (PL_regkind[(U8)op] == EXACT) {
1N/A /* Literal string, where present. */
1N/A node += NODE_SZ_STR(node) - 1;
1N/A node = NEXTOPER(node);
1N/A }
1N/A else {
1N/A node = NEXTOPER(node);
1N/A node += regarglen[(U8)op];
1N/A }
1N/A if (op == CURLYX || op == OPEN)
1N/A l++;
1N/A else if (op == WHILEM)
1N/A l--;
1N/A }
1N/A return node;
1N/A}
1N/A
1N/A#endif /* DEBUGGING */
1N/A
1N/A/*
1N/A - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
1N/A */
1N/Avoid
1N/APerl_regdump(pTHX_ regexp *r)
1N/A{
1N/A#ifdef DEBUGGING
1N/A SV *sv = sv_newmortal();
1N/A
1N/A (void)dumpuntil(r->program, r->program + 1, NULL, sv, 0);
1N/A
1N/A /* Header fields of interest. */
1N/A if (r->anchored_substr)
1N/A PerlIO_printf(Perl_debug_log,
1N/A "anchored `%s%.*s%s'%s at %"IVdf" ",
1N/A PL_colors[0],
1N/A (int)(SvCUR(r->anchored_substr) - (SvTAIL(r->anchored_substr)!=0)),
1N/A SvPVX(r->anchored_substr),
1N/A PL_colors[1],
1N/A SvTAIL(r->anchored_substr) ? "$" : "",
1N/A (IV)r->anchored_offset);
1N/A else if (r->anchored_utf8)
1N/A PerlIO_printf(Perl_debug_log,
1N/A "anchored utf8 `%s%.*s%s'%s at %"IVdf" ",
1N/A PL_colors[0],
1N/A (int)(SvCUR(r->anchored_utf8) - (SvTAIL(r->anchored_utf8)!=0)),
1N/A SvPVX(r->anchored_utf8),
1N/A PL_colors[1],
1N/A SvTAIL(r->anchored_utf8) ? "$" : "",
1N/A (IV)r->anchored_offset);
1N/A if (r->float_substr)
1N/A PerlIO_printf(Perl_debug_log,
1N/A "floating `%s%.*s%s'%s at %"IVdf"..%"UVuf" ",
1N/A PL_colors[0],
1N/A (int)(SvCUR(r->float_substr) - (SvTAIL(r->float_substr)!=0)),
1N/A SvPVX(r->float_substr),
1N/A PL_colors[1],
1N/A SvTAIL(r->float_substr) ? "$" : "",
1N/A (IV)r->float_min_offset, (UV)r->float_max_offset);
1N/A else if (r->float_utf8)
1N/A PerlIO_printf(Perl_debug_log,
1N/A "floating utf8 `%s%.*s%s'%s at %"IVdf"..%"UVuf" ",
1N/A PL_colors[0],
1N/A (int)(SvCUR(r->float_utf8) - (SvTAIL(r->float_utf8)!=0)),
1N/A SvPVX(r->float_utf8),
1N/A PL_colors[1],
1N/A SvTAIL(r->float_utf8) ? "$" : "",
1N/A (IV)r->float_min_offset, (UV)r->float_max_offset);
1N/A if (r->check_substr || r->check_utf8)
1N/A PerlIO_printf(Perl_debug_log,
1N/A r->check_substr == r->float_substr
1N/A && r->check_utf8 == r->float_utf8
1N/A ? "(checking floating" : "(checking anchored");
1N/A if (r->reganch & ROPT_NOSCAN)
1N/A PerlIO_printf(Perl_debug_log, " noscan");
1N/A if (r->reganch & ROPT_CHECK_ALL)
1N/A PerlIO_printf(Perl_debug_log, " isall");
1N/A if (r->check_substr || r->check_utf8)
1N/A PerlIO_printf(Perl_debug_log, ") ");
1N/A
1N/A if (r->regstclass) {
1N/A regprop(sv, r->regstclass);
1N/A PerlIO_printf(Perl_debug_log, "stclass `%s' ", SvPVX(sv));
1N/A }
1N/A if (r->reganch & ROPT_ANCH) {
1N/A PerlIO_printf(Perl_debug_log, "anchored");
1N/A if (r->reganch & ROPT_ANCH_BOL)
1N/A PerlIO_printf(Perl_debug_log, "(BOL)");
1N/A if (r->reganch & ROPT_ANCH_MBOL)
1N/A PerlIO_printf(Perl_debug_log, "(MBOL)");
1N/A if (r->reganch & ROPT_ANCH_SBOL)
1N/A PerlIO_printf(Perl_debug_log, "(SBOL)");
1N/A if (r->reganch & ROPT_ANCH_GPOS)
1N/A PerlIO_printf(Perl_debug_log, "(GPOS)");
1N/A PerlIO_putc(Perl_debug_log, ' ');
1N/A }
1N/A if (r->reganch & ROPT_GPOS_SEEN)
1N/A PerlIO_printf(Perl_debug_log, "GPOS ");
1N/A if (r->reganch & ROPT_SKIP)
1N/A PerlIO_printf(Perl_debug_log, "plus ");
1N/A if (r->reganch & ROPT_IMPLICIT)
1N/A PerlIO_printf(Perl_debug_log, "implicit ");
1N/A PerlIO_printf(Perl_debug_log, "minlen %ld ", (long) r->minlen);
1N/A if (r->reganch & ROPT_EVAL_SEEN)
1N/A PerlIO_printf(Perl_debug_log, "with eval ");
1N/A PerlIO_printf(Perl_debug_log, "\n");
1N/A if (r->offsets) {
1N/A U32 i;
1N/A U32 len = r->offsets[0];
1N/A PerlIO_printf(Perl_debug_log, "Offsets: [%"UVuf"]\n\t", (UV)r->offsets[0]);
1N/A for (i = 1; i <= len; i++)
1N/A PerlIO_printf(Perl_debug_log, "%"UVuf"[%"UVuf"] ",
1N/A (UV)r->offsets[i*2-1],
1N/A (UV)r->offsets[i*2]);
1N/A PerlIO_printf(Perl_debug_log, "\n");
1N/A }
1N/A#endif /* DEBUGGING */
1N/A}
1N/A
1N/A#ifdef DEBUGGING
1N/A
1N/ASTATIC void
1N/AS_put_byte(pTHX_ SV *sv, int c)
1N/A{
1N/A if (isCNTRL(c) || c == 255 || !isPRINT(c))
1N/A Perl_sv_catpvf(aTHX_ sv, "\\%o", c);
1N/A else if (c == '-' || c == ']' || c == '\\' || c == '^')
1N/A Perl_sv_catpvf(aTHX_ sv, "\\%c", c);
1N/A else
1N/A Perl_sv_catpvf(aTHX_ sv, "%c", c);
1N/A}
1N/A
1N/A#endif /* DEBUGGING */
1N/A
1N/A/*
1N/A- regprop - printable representation of opcode
1N/A*/
1N/Avoid
1N/APerl_regprop(pTHX_ SV *sv, regnode *o)
1N/A{
1N/A#ifdef DEBUGGING
1N/A register int k;
1N/A
1N/A sv_setpvn(sv, "", 0);
1N/A if (OP(o) >= reg_num) /* regnode.type is unsigned */
1N/A /* It would be nice to FAIL() here, but this may be called from
1N/A regexec.c, and it would be hard to supply pRExC_state. */
1N/A Perl_croak(aTHX_ "Corrupted regexp opcode");
1N/A sv_catpv(sv, (char*)reg_name[OP(o)]); /* Take off const! */
1N/A
1N/A k = PL_regkind[(U8)OP(o)];
1N/A
1N/A if (k == EXACT) {
1N/A SV *dsv = sv_2mortal(newSVpvn("", 0));
1N/A /* Using is_utf8_string() is a crude hack but it may
1N/A * be the best for now since we have no flag "this EXACTish
1N/A * node was UTF-8" --jhi */
1N/A bool do_utf8 = is_utf8_string((U8*)STRING(o), STR_LEN(o));
1N/A char *s = do_utf8 ?
1N/A pv_uni_display(dsv, (U8*)STRING(o), STR_LEN(o), 60,
1N/A UNI_DISPLAY_REGEX) :
1N/A STRING(o);
1N/A int len = do_utf8 ?
1N/A strlen(s) :
1N/A STR_LEN(o);
1N/A Perl_sv_catpvf(aTHX_ sv, " <%s%.*s%s>",
1N/A PL_colors[0],
1N/A len, s,
1N/A PL_colors[1]);
1N/A }
1N/A else if (k == CURLY) {
1N/A if (OP(o) == CURLYM || OP(o) == CURLYN || OP(o) == CURLYX)
1N/A Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */
1N/A Perl_sv_catpvf(aTHX_ sv, " {%d,%d}", ARG1(o), ARG2(o));
1N/A }
1N/A else if (k == WHILEM && o->flags) /* Ordinal/of */
1N/A Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4);
1N/A else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP )
1N/A Perl_sv_catpvf(aTHX_ sv, "%d", (int)ARG(o)); /* Parenth number */
1N/A else if (k == LOGICAL)
1N/A Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* 2: embedded, otherwise 1 */
1N/A else if (k == ANYOF) {
1N/A int i, rangestart = -1;
1N/A U8 flags = ANYOF_FLAGS(o);
1N/A const char * const anyofs[] = { /* Should be synchronized with
1N/A * ANYOF_ #xdefines in regcomp.h */
1N/A "\\w",
1N/A "\\W",
1N/A "\\s",
1N/A "\\S",
1N/A "\\d",
1N/A "\\D",
1N/A "[:alnum:]",
1N/A "[:^alnum:]",
1N/A "[:alpha:]",
1N/A "[:^alpha:]",
1N/A "[:ascii:]",
1N/A "[:^ascii:]",
1N/A "[:ctrl:]",
1N/A "[:^ctrl:]",
1N/A "[:graph:]",
1N/A "[:^graph:]",
1N/A "[:lower:]",
1N/A "[:^lower:]",
1N/A "[:print:]",
1N/A "[:^print:]",
1N/A "[:punct:]",
1N/A "[:^punct:]",
1N/A "[:upper:]",
1N/A "[:^upper:]",
1N/A "[:xdigit:]",
1N/A "[:^xdigit:]",
1N/A "[:space:]",
1N/A "[:^space:]",
1N/A "[:blank:]",
1N/A "[:^blank:]"
1N/A };
1N/A
1N/A if (flags & ANYOF_LOCALE)
1N/A sv_catpv(sv, "{loc}");
1N/A if (flags & ANYOF_FOLD)
1N/A sv_catpv(sv, "{i}");
1N/A Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
1N/A if (flags & ANYOF_INVERT)
1N/A sv_catpv(sv, "^");
1N/A for (i = 0; i <= 256; i++) {
1N/A if (i < 256 && ANYOF_BITMAP_TEST(o,i)) {
1N/A if (rangestart == -1)
1N/A rangestart = i;
1N/A } else if (rangestart != -1) {
1N/A if (i <= rangestart + 3)
1N/A for (; rangestart < i; rangestart++)
1N/A put_byte(sv, rangestart);
1N/A else {
1N/A put_byte(sv, rangestart);
1N/A sv_catpv(sv, "-");
1N/A put_byte(sv, i - 1);
1N/A }
1N/A rangestart = -1;
1N/A }
1N/A }
1N/A
1N/A if (o->flags & ANYOF_CLASS)
1N/A for (i = 0; i < sizeof(anyofs)/sizeof(char*); i++)
1N/A if (ANYOF_CLASS_TEST(o,i))
1N/A sv_catpv(sv, anyofs[i]);
1N/A
1N/A if (flags & ANYOF_UNICODE)
1N/A sv_catpv(sv, "{unicode}");
1N/A else if (flags & ANYOF_UNICODE_ALL)
1N/A sv_catpv(sv, "{unicode_all}");
1N/A
1N/A {
1N/A SV *lv;
1N/A SV *sw = regclass_swash(o, FALSE, &lv, 0);
1N/A
1N/A if (lv) {
1N/A if (sw) {
1N/A U8 s[UTF8_MAXLEN+1];
1N/A
1N/A for (i = 0; i <= 256; i++) { /* just the first 256 */
1N/A U8 *e = uvchr_to_utf8(s, i);
1N/A
1N/A if (i < 256 && swash_fetch(sw, s, TRUE)) {
1N/A if (rangestart == -1)
1N/A rangestart = i;
1N/A } else if (rangestart != -1) {
1N/A U8 *p;
1N/A
1N/A if (i <= rangestart + 3)
1N/A for (; rangestart < i; rangestart++) {
1N/A for(e = uvchr_to_utf8(s, rangestart), p = s; p < e; p++)
1N/A put_byte(sv, *p);
1N/A }
1N/A else {
1N/A for (e = uvchr_to_utf8(s, rangestart), p = s; p < e; p++)
1N/A put_byte(sv, *p);
1N/A sv_catpv(sv, "-");
1N/A for (e = uvchr_to_utf8(s, i - 1), p = s; p < e; p++)
1N/A put_byte(sv, *p);
1N/A }
1N/A rangestart = -1;
1N/A }
1N/A }
1N/A
1N/A sv_catpv(sv, "..."); /* et cetera */
1N/A }
1N/A
1N/A {
1N/A char *s = savepv(SvPVX(lv));
1N/A char *origs = s;
1N/A
1N/A while(*s && *s != '\n') s++;
1N/A
1N/A if (*s == '\n') {
1N/A char *t = ++s;
1N/A
1N/A while (*s) {
1N/A if (*s == '\n')
1N/A *s = ' ';
1N/A s++;
1N/A }
1N/A if (s[-1] == ' ')
1N/A s[-1] = 0;
1N/A
1N/A sv_catpv(sv, t);
1N/A }
1N/A
1N/A Safefree(origs);
1N/A }
1N/A }
1N/A }
1N/A
1N/A Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
1N/A }
1N/A else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH))
1N/A Perl_sv_catpvf(aTHX_ sv, "[-%d]", o->flags);
1N/A#endif /* DEBUGGING */
1N/A}
1N/A
1N/ASV *
1N/APerl_re_intuit_string(pTHX_ regexp *prog)
1N/A{ /* Assume that RE_INTUIT is set */
1N/A DEBUG_r(
1N/A { STRLEN n_a;
1N/A char *s = SvPV(prog->check_substr
1N/A ? prog->check_substr : prog->check_utf8, n_a);
1N/A
1N/A if (!PL_colorset) reginitcolors();
1N/A PerlIO_printf(Perl_debug_log,
1N/A "%sUsing REx %ssubstr:%s `%s%.60s%s%s'\n",
1N/A PL_colors[4],
1N/A prog->check_substr ? "" : "utf8 ",
1N/A PL_colors[5],PL_colors[0],
1N/A s,
1N/A PL_colors[1],
1N/A (strlen(s) > 60 ? "..." : ""));
1N/A } );
1N/A
1N/A return prog->check_substr ? prog->check_substr : prog->check_utf8;
1N/A}
1N/A
1N/Avoid
1N/APerl_pregfree(pTHX_ struct regexp *r)
1N/A{
1N/A#ifdef DEBUGGING
1N/A SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1N/A#endif
1N/A
1N/A if (!r || (--r->refcnt > 0))
1N/A return;
1N/A DEBUG_r({
1N/A int len;
1N/A char *s;
1N/A
1N/A s = (r->reganch & ROPT_UTF8) ? pv_uni_display(dsv, (U8*)r->precomp,
1N/A r->prelen, 60, UNI_DISPLAY_REGEX)
1N/A : pv_display(dsv, r->precomp, r->prelen, 0, 60);
1N/A len = SvCUR(dsv);
1N/A if (!PL_colorset)
1N/A reginitcolors();
1N/A PerlIO_printf(Perl_debug_log,
1N/A "%sFreeing REx:%s `%s%*.*s%s%s'\n",
1N/A PL_colors[4],PL_colors[5],PL_colors[0],
1N/A len, len, s,
1N/A PL_colors[1],
1N/A len > 60 ? "..." : "");
1N/A });
1N/A
1N/A if (r->precomp)
1N/A Safefree(r->precomp);
1N/A if (r->offsets) /* 20010421 MJD */
1N/A Safefree(r->offsets);
1N/A if (RX_MATCH_COPIED(r))
1N/A Safefree(r->subbeg);
1N/A if (r->substrs) {
1N/A if (r->anchored_substr)
1N/A SvREFCNT_dec(r->anchored_substr);
1N/A if (r->anchored_utf8)
1N/A SvREFCNT_dec(r->anchored_utf8);
1N/A if (r->float_substr)
1N/A SvREFCNT_dec(r->float_substr);
1N/A if (r->float_utf8)
1N/A SvREFCNT_dec(r->float_utf8);
1N/A Safefree(r->substrs);
1N/A }
1N/A if (r->data) {
1N/A int n = r->data->count;
1N/A PAD* new_comppad = NULL;
1N/A PAD* old_comppad;
1N/A
1N/A while (--n >= 0) {
1N/A /* If you add a ->what type here, update the comment in regcomp.h */
1N/A switch (r->data->what[n]) {
1N/A case 's':
1N/A SvREFCNT_dec((SV*)r->data->data[n]);
1N/A break;
1N/A case 'f':
1N/A Safefree(r->data->data[n]);
1N/A break;
1N/A case 'p':
1N/A new_comppad = (AV*)r->data->data[n];
1N/A break;
1N/A case 'o':
1N/A if (new_comppad == NULL)
1N/A Perl_croak(aTHX_ "panic: pregfree comppad");
1N/A PAD_SAVE_LOCAL(old_comppad,
1N/A /* Watch out for global destruction's random ordering. */
1N/A (SvTYPE(new_comppad) == SVt_PVAV) ?
1N/A new_comppad : Null(PAD *)
1N/A );
1N/A if (!OpREFCNT_dec((OP_4tree*)r->data->data[n])) {
1N/A op_free((OP_4tree*)r->data->data[n]);
1N/A }
1N/A
1N/A PAD_RESTORE_LOCAL(old_comppad);
1N/A SvREFCNT_dec((SV*)new_comppad);
1N/A new_comppad = NULL;
1N/A break;
1N/A case 'n':
1N/A break;
1N/A default:
1N/A Perl_croak(aTHX_ "panic: regfree data code '%c'", r->data->what[n]);
1N/A }
1N/A }
1N/A Safefree(r->data->what);
1N/A Safefree(r->data);
1N/A }
1N/A Safefree(r->startp);
1N/A Safefree(r->endp);
1N/A Safefree(r);
1N/A}
1N/A
1N/A/*
1N/A - regnext - dig the "next" pointer out of a node
1N/A *
1N/A * [Note, when REGALIGN is defined there are two places in regmatch()
1N/A * that bypass this code for speed.]
1N/A */
1N/Aregnode *
1N/APerl_regnext(pTHX_ register regnode *p)
1N/A{
1N/A register I32 offset;
1N/A
1N/A if (p == &PL_regdummy)
1N/A return(NULL);
1N/A
1N/A offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
1N/A if (offset == 0)
1N/A return(NULL);
1N/A
1N/A return(p+offset);
1N/A}
1N/A
1N/ASTATIC void
1N/AS_re_croak2(pTHX_ const char* pat1,const char* pat2,...)
1N/A{
1N/A va_list args;
1N/A STRLEN l1 = strlen(pat1);
1N/A STRLEN l2 = strlen(pat2);
1N/A char buf[512];
1N/A SV *msv;
1N/A char *message;
1N/A
1N/A if (l1 > 510)
1N/A l1 = 510;
1N/A if (l1 + l2 > 510)
1N/A l2 = 510 - l1;
1N/A Copy(pat1, buf, l1 , char);
1N/A Copy(pat2, buf + l1, l2 , char);
1N/A buf[l1 + l2] = '\n';
1N/A buf[l1 + l2 + 1] = '\0';
1N/A#ifdef I_STDARG
1N/A /* ANSI variant takes additional second argument */
1N/A va_start(args, pat2);
1N/A#else
1N/A va_start(args);
1N/A#endif
1N/A msv = vmess(buf, &args);
1N/A va_end(args);
1N/A message = SvPV(msv,l1);
1N/A if (l1 > 512)
1N/A l1 = 512;
1N/A Copy(message, buf, l1 , char);
1N/A buf[l1-1] = '\0'; /* Overwrite \n */
1N/A Perl_croak(aTHX_ "%s", buf);
1N/A}
1N/A
1N/A/* XXX Here's a total kludge. But we need to re-enter for swash routines. */
1N/A
1N/Avoid
1N/APerl_save_re_context(pTHX)
1N/A{
1N/A SAVEI32(PL_reg_flags); /* from regexec.c */
1N/A SAVEPPTR(PL_bostr);
1N/A SAVEPPTR(PL_reginput); /* String-input pointer. */
1N/A SAVEPPTR(PL_regbol); /* Beginning of input, for ^ check. */
1N/A SAVEPPTR(PL_regeol); /* End of input, for $ check. */
1N/A SAVEVPTR(PL_regstartp); /* Pointer to startp array. */
1N/A SAVEVPTR(PL_regendp); /* Ditto for endp. */
1N/A SAVEVPTR(PL_reglastparen); /* Similarly for lastparen. */
1N/A SAVEVPTR(PL_reglastcloseparen); /* Similarly for lastcloseparen. */
1N/A SAVEPPTR(PL_regtill); /* How far we are required to go. */
1N/A SAVEGENERICPV(PL_reg_start_tmp); /* from regexec.c */
1N/A PL_reg_start_tmp = 0;
1N/A SAVEI32(PL_reg_start_tmpl); /* from regexec.c */
1N/A PL_reg_start_tmpl = 0;
1N/A SAVEVPTR(PL_regdata);
1N/A SAVEI32(PL_reg_eval_set); /* from regexec.c */
1N/A SAVEI32(PL_regnarrate); /* from regexec.c */
1N/A SAVEVPTR(PL_regprogram); /* from regexec.c */
1N/A SAVEINT(PL_regindent); /* from regexec.c */
1N/A SAVEVPTR(PL_regcc); /* from regexec.c */
1N/A SAVEVPTR(PL_curcop);
1N/A SAVEVPTR(PL_reg_call_cc); /* from regexec.c */
1N/A SAVEVPTR(PL_reg_re); /* from regexec.c */
1N/A SAVEPPTR(PL_reg_ganch); /* from regexec.c */
1N/A SAVESPTR(PL_reg_sv); /* from regexec.c */
1N/A SAVEBOOL(PL_reg_match_utf8); /* from regexec.c */
1N/A SAVEVPTR(PL_reg_magic); /* from regexec.c */
1N/A SAVEI32(PL_reg_oldpos); /* from regexec.c */
1N/A SAVEVPTR(PL_reg_oldcurpm); /* from regexec.c */
1N/A SAVEVPTR(PL_reg_curpm); /* from regexec.c */
1N/A SAVEPPTR(PL_reg_oldsaved); /* old saved substr during match */
1N/A PL_reg_oldsaved = Nullch;
1N/A SAVEI32(PL_reg_oldsavedlen); /* old length of saved substr during match */
1N/A PL_reg_oldsavedlen = 0;
1N/A SAVEI32(PL_reg_maxiter); /* max wait until caching pos */
1N/A PL_reg_maxiter = 0;
1N/A SAVEI32(PL_reg_leftiter); /* wait until caching pos */
1N/A PL_reg_leftiter = 0;
1N/A SAVEGENERICPV(PL_reg_poscache); /* cache of pos of WHILEM */
1N/A PL_reg_poscache = Nullch;
1N/A SAVEI32(PL_reg_poscache_size); /* size of pos cache of WHILEM */
1N/A PL_reg_poscache_size = 0;
1N/A SAVEPPTR(PL_regprecomp); /* uncompiled string. */
1N/A SAVEI32(PL_regnpar); /* () count. */
1N/A SAVEI32(PL_regsize); /* from regexec.c */
1N/A
1N/A {
1N/A /* Save $1..$n (#18107: UTF-8 s/(\w+)/uc($1)/e); AMS 20021106. */
1N/A U32 i;
1N/A GV *mgv;
1N/A REGEXP *rx;
1N/A char digits[16];
1N/A
1N/A if (PL_curpm && (rx = PM_GETRE(PL_curpm))) {
1N/A for (i = 1; i <= rx->nparens; i++) {
1N/A sprintf(digits, "%lu", (long)i);
1N/A if ((mgv = gv_fetchpv(digits, FALSE, SVt_PV)))
1N/A save_scalar(mgv);
1N/A }
1N/A }
1N/A }
1N/A
1N/A#ifdef DEBUGGING
1N/A SAVEPPTR(PL_reg_starttry); /* from regexec.c */
1N/A#endif
1N/A}
1N/A
1N/Astatic void
1N/Aclear_re(pTHX_ void *r)
1N/A{
1N/A ReREFCNT_dec((regexp *)r);
1N/A}
1N/A