[perl5.git] / toke.c

/*    toke.c
 *
 *    Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
 *    2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 by Larry Wall and others
 *
 *    You may distribute under the terms of either the GNU General Public
 *    License or the Artistic License, as specified in the README file.
 *
 */

/*
 *  'It all comes from here, the stench and the peril.'    --Frodo
 *
 *     [p.719 of _The Lord of the Rings_, IV/ix: "Shelob's Lair"]
 */

/*
 * This file is the lexer for Perl.  It's closely linked to the
 * parser, perly.y.
 *
 * The main routine is yylex(), which returns the next token.
 */

/*
=head1 Lexer interface
This is the lower layer of the Perl parser, managing characters and tokens.

=for apidoc AmU|yy_parser *|PL_parser

Pointer to a structure encapsulating the state of the parsing operation
currently in progress.  The pointer can be locally changed to perform
a nested parse without interfering with the state of an outer parse.
Individual members of C<PL_parser> have their own documentation.

=cut
*/

#include "EXTERN.h"
#define PERL_IN_TOKE_C
#include "perl.h"
#include "dquote_inline.h"

#define new_constant(a,b,c,d,e,f,g)     \
        S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)

#define pl_yylval       (PL_parser->yylval)

/* XXX temporary backwards compatibility */
#define PL_lex_brackets         (PL_parser->lex_brackets)
#define PL_lex_allbrackets      (PL_parser->lex_allbrackets)
#define PL_lex_fakeeof          (PL_parser->lex_fakeeof)
#define PL_lex_brackstack       (PL_parser->lex_brackstack)
#define PL_lex_casemods         (PL_parser->lex_casemods)
#define PL_lex_casestack        (PL_parser->lex_casestack)
#define PL_lex_dojoin           (PL_parser->lex_dojoin)
#define PL_lex_formbrack        (PL_parser->lex_formbrack)
#define PL_lex_inpat            (PL_parser->lex_inpat)
#define PL_lex_inwhat           (PL_parser->lex_inwhat)
#define PL_lex_op               (PL_parser->lex_op)
#define PL_lex_repl             (PL_parser->lex_repl)
#define PL_lex_starts           (PL_parser->lex_starts)
#define PL_lex_stuff            (PL_parser->lex_stuff)
#define PL_multi_start          (PL_parser->multi_start)
#define PL_multi_open           (PL_parser->multi_open)
#define PL_multi_close          (PL_parser->multi_close)
#define PL_preambled            (PL_parser->preambled)
#define PL_linestr              (PL_parser->linestr)
#define PL_expect               (PL_parser->expect)
#define PL_copline              (PL_parser->copline)
#define PL_bufptr               (PL_parser->bufptr)
#define PL_oldbufptr            (PL_parser->oldbufptr)
#define PL_oldoldbufptr         (PL_parser->oldoldbufptr)
#define PL_linestart            (PL_parser->linestart)
#define PL_bufend               (PL_parser->bufend)
#define PL_last_uni             (PL_parser->last_uni)
#define PL_last_lop             (PL_parser->last_lop)
#define PL_last_lop_op          (PL_parser->last_lop_op)
#define PL_lex_state            (PL_parser->lex_state)
#define PL_rsfp                 (PL_parser->rsfp)
#define PL_rsfp_filters         (PL_parser->rsfp_filters)
#define PL_in_my                (PL_parser->in_my)
#define PL_in_my_stash          (PL_parser->in_my_stash)
#define PL_tokenbuf             (PL_parser->tokenbuf)
#define PL_multi_end            (PL_parser->multi_end)
#define PL_error_count          (PL_parser->error_count)

#  define PL_nexttoke           (PL_parser->nexttoke)
#  define PL_nexttype           (PL_parser->nexttype)
#  define PL_nextval            (PL_parser->nextval)


#define SvEVALED(sv) \
    (SvTYPE(sv) >= SVt_PVNV \
    && ((XPVIV*)SvANY(sv))->xiv_u.xivu_eval_seen)

static const char* const ident_too_long = "Identifier too long";

#  define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]

#define XENUMMASK  0x3f
#define XFAKEEOF   0x40
#define XFAKEBRACK 0x80

#ifdef USE_UTF8_SCRIPTS
#   define UTF cBOOL(!IN_BYTES)
#else
#   define UTF cBOOL((PL_linestr && DO_UTF8(PL_linestr)) || ( !(PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS) && (PL_hints & HINT_UTF8)))
#endif

/* The maximum number of characters preceding the unrecognized one to display */
#define UNRECOGNIZED_PRECEDE_COUNT 10

/* In variables named $^X, these are the legal values for X.
 * 1999-02-27 mjd-perl-patch@plover.com */
#define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))

#define SPACE_OR_TAB(c) isBLANK_A(c)

#define HEXFP_PEEK(s)     \
    (((s[0] == '.') && \
      (isXDIGIT(s[1]) || isALPHA_FOLD_EQ(s[1], 'p'))) || \
     isALPHA_FOLD_EQ(s[0], 'p'))

/* LEX_* are values for PL_lex_state, the state of the lexer.
 * They are arranged oddly so that the guard on the switch statement
 * can get by with a single comparison (if the compiler is smart enough).
 *
 * These values refer to the various states within a sublex parse,
 * i.e. within a double quotish string
 */

/* #define LEX_NOTPARSING               11 is done in perl.h. */

#define LEX_NORMAL              10 /* normal code (ie not within "...")     */
#define LEX_INTERPNORMAL         9 /* code within a string, eg "$foo[$x+1]" */
#define LEX_INTERPCASEMOD        8 /* expecting a \U, \Q or \E etc          */
#define LEX_INTERPPUSH           7 /* starting a new sublex parse level     */
#define LEX_INTERPSTART          6 /* expecting the start of a $var         */

                                   /* at end of code, eg "$x" followed by:  */
#define LEX_INTERPEND            5 /* ... eg not one of [, { or ->          */
#define LEX_INTERPENDMAYBE       4 /* ... eg one of [, { or ->              */

#define LEX_INTERPCONCAT         3 /* expecting anything, eg at start of
                                        string or after \E, $foo, etc       */
#define LEX_INTERPCONST          2 /* NOT USED */
#define LEX_FORMLINE             1 /* expecting a format line               */


#ifdef DEBUGGING
static const char* const lex_state_names[] = {
    "KNOWNEXT",
    "FORMLINE",
    "INTERPCONST",
    "INTERPCONCAT",
    "INTERPENDMAYBE",
    "INTERPEND",
    "INTERPSTART",
    "INTERPPUSH",
    "INTERPCASEMOD",
    "INTERPNORMAL",
    "NORMAL"
};
#endif

#include "keywords.h"

/* CLINE is a macro that ensures PL_copline has a sane value */

#define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))

/*
 * Convenience functions to return different tokens and prime the
 * lexer for the next token.  They all take an argument.
 *
 * TOKEN        : generic token (used for '(', DOLSHARP, etc)
 * OPERATOR     : generic operator
 * AOPERATOR    : assignment operator
 * PREBLOCK     : beginning the block after an if, while, foreach, ...
 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
 * PREREF       : *EXPR where EXPR is not a simple identifier
 * TERM         : expression term
 * POSTDEREF    : postfix dereference (->$* ->@[...] etc.)
 * LOOPX        : loop exiting command (goto, last, dump, etc)
 * FTST         : file test operator
 * FUN0         : zero-argument function
 * FUN0OP       : zero-argument function, with its op created in this file
 * FUN1         : not used, except for not, which isn't a UNIOP
 * BOop         : bitwise or or xor
 * BAop         : bitwise and
 * BCop         : bitwise complement
 * SHop         : shift operator
 * PWop         : power operator
 * PMop         : pattern-matching operator
 * Aop          : addition-level operator
 * AopNOASSIGN  : addition-level operator that is never part of .=
 * Mop          : multiplication-level operator
 * Eop          : equality-testing operator
 * Rop          : relational operator <= != gt
 *
 * Also see LOP and lop() below.
 */

#ifdef DEBUGGING /* Serve -DT. */
#   define REPORT(retval) tokereport((I32)retval, &pl_yylval)
#else
#   define REPORT(retval) (retval)
#endif

#define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
#define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
#define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, retval))
#define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
#define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
#define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
#define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
#define POSTDEREF(f) return (PL_bufptr = s, S_postderef(aTHX_ REPORT(f),s[1]))
#define LOOPX(f) return (PL_bufptr = force_word(s,BAREWORD,TRUE,FALSE), \
                         pl_yylval.ival=f, \
                         PL_expect = PL_nexttoke ? XOPERATOR : XTERM, \
                         REPORT((int)LOOPEX))
#define FTST(f)  return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
#define FUN0(f)  return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
#define FUN0OP(f)  return (pl_yylval.opval=f, CLINE, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0OP))
#define FUN1(f)  return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
#define BOop(f)  return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)BITOROP))
#define BAop(f)  return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)BITANDOP))
#define BCop(f) return pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr = s, \
                       REPORT('~')
#define SHop(f)  return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)SHIFTOP))
#define PWop(f)  return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)POWOP))
#define PMop(f)  return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
#define Aop(f)   return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)ADDOP))
#define AopNOASSIGN(f) return (pl_yylval.ival=f, PL_bufptr=s, REPORT((int)ADDOP))
#define Mop(f)   return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)MULOP))
#define Eop(f)   return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
#define Rop(f)   return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))

/* This bit of chicanery makes a unary function followed by
 * a parenthesis into a function with one argument, highest precedence.
 * The UNIDOR macro is for unary functions that can be followed by the //
 * operator (such as C<shift // 0>).
 */
#define UNI3(f,x,have_x) { \
        pl_yylval.ival = f; \
        if (have_x) PL_expect = x; \
        PL_bufptr = s; \
        PL_last_uni = PL_oldbufptr; \
        PL_last_lop_op = (f) < 0 ? -(f) : (f); \
        if (*s == '(') \
            return REPORT( (int)FUNC1 ); \
        s = skipspace(s); \
        return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
        }
#define UNI(f)    UNI3(f,XTERM,1)
#define UNIDOR(f) UNI3(f,XTERMORDORDOR,1)
#define UNIPROTO(f,optional) { \
        if (optional) PL_last_uni = PL_oldbufptr; \
        OPERATOR(f); \
        }

#define UNIBRACK(f) UNI3(f,0,0)

/* grandfather return to old style */
#define OLDLOP(f) \
        do { \
            if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
                PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
            pl_yylval.ival = (f); \
            PL_expect = XTERM; \
            PL_bufptr = s; \
            return (int)LSTOP; \
        } while(0)

#define COPLINE_INC_WITH_HERELINES                  \
    STMT_START {                                     \
        CopLINE_inc(PL_curcop);                       \
        if (PL_parser->herelines)                      \
            CopLINE(PL_curcop) += PL_parser->herelines, \
            PL_parser->herelines = 0;                    \
    } STMT_END
/* Called after scan_str to update CopLINE(PL_curcop), but only when there
 * is no sublex_push to follow. */
#define COPLINE_SET_FROM_MULTI_END            \
    STMT_START {                               \
        CopLINE_set(PL_curcop, PL_multi_end);   \
        if (PL_multi_end != PL_multi_start)      \
            PL_parser->herelines = 0;             \
    } STMT_END


#ifdef DEBUGGING

/* how to interpret the pl_yylval associated with the token */
enum token_type {
    TOKENTYPE_NONE,
    TOKENTYPE_IVAL,
    TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
    TOKENTYPE_PVAL,
    TOKENTYPE_OPVAL
};

static struct debug_tokens {
    const int token;
    enum token_type type;
    const char *name;
} const debug_tokens[] =
{
    { ADDOP,            TOKENTYPE_OPNUM,        "ADDOP" },
    { ANDAND,           TOKENTYPE_NONE,         "ANDAND" },
    { ANDOP,            TOKENTYPE_NONE,         "ANDOP" },
    { ANONSUB,          TOKENTYPE_IVAL,         "ANONSUB" },
    { ARROW,            TOKENTYPE_NONE,         "ARROW" },
    { ASSIGNOP,         TOKENTYPE_OPNUM,        "ASSIGNOP" },
    { BITANDOP,         TOKENTYPE_OPNUM,        "BITANDOP" },
    { BITOROP,          TOKENTYPE_OPNUM,        "BITOROP" },
    { COLONATTR,        TOKENTYPE_NONE,         "COLONATTR" },
    { CONTINUE,         TOKENTYPE_NONE,         "CONTINUE" },
    { DEFAULT,          TOKENTYPE_NONE,         "DEFAULT" },
    { DO,               TOKENTYPE_NONE,         "DO" },
    { DOLSHARP,         TOKENTYPE_NONE,         "DOLSHARP" },
    { DORDOR,           TOKENTYPE_NONE,         "DORDOR" },
    { DOROP,            TOKENTYPE_OPNUM,        "DOROP" },
    { DOTDOT,           TOKENTYPE_IVAL,         "DOTDOT" },
    { ELSE,             TOKENTYPE_NONE,         "ELSE" },
    { ELSIF,            TOKENTYPE_IVAL,         "ELSIF" },
    { EQOP,             TOKENTYPE_OPNUM,        "EQOP" },
    { FOR,              TOKENTYPE_IVAL,         "FOR" },
    { FORMAT,           TOKENTYPE_NONE,         "FORMAT" },
    { FORMLBRACK,       TOKENTYPE_NONE,         "FORMLBRACK" },
    { FORMRBRACK,       TOKENTYPE_NONE,         "FORMRBRACK" },
    { FUNC,             TOKENTYPE_OPNUM,        "FUNC" },
    { FUNC0,            TOKENTYPE_OPNUM,        "FUNC0" },
    { FUNC0OP,          TOKENTYPE_OPVAL,        "FUNC0OP" },
    { FUNC0SUB,         TOKENTYPE_OPVAL,        "FUNC0SUB" },
    { FUNC1,            TOKENTYPE_OPNUM,        "FUNC1" },
    { FUNCMETH,         TOKENTYPE_OPVAL,        "FUNCMETH" },
    { GIVEN,            TOKENTYPE_IVAL,         "GIVEN" },
    { HASHBRACK,        TOKENTYPE_NONE,         "HASHBRACK" },
    { IF,               TOKENTYPE_IVAL,         "IF" },
    { LABEL,            TOKENTYPE_PVAL,         "LABEL" },
    { LOCAL,            TOKENTYPE_IVAL,         "LOCAL" },
    { LOOPEX,           TOKENTYPE_OPNUM,        "LOOPEX" },
    { LSTOP,            TOKENTYPE_OPNUM,        "LSTOP" },
    { LSTOPSUB,         TOKENTYPE_OPVAL,        "LSTOPSUB" },
    { MATCHOP,          TOKENTYPE_OPNUM,        "MATCHOP" },
    { METHOD,           TOKENTYPE_OPVAL,        "METHOD" },
    { MULOP,            TOKENTYPE_OPNUM,        "MULOP" },
    { MY,               TOKENTYPE_IVAL,         "MY" },
    { NOAMP,            TOKENTYPE_NONE,         "NOAMP" },
    { NOTOP,            TOKENTYPE_NONE,         "NOTOP" },
    { OROP,             TOKENTYPE_IVAL,         "OROP" },
    { OROR,             TOKENTYPE_NONE,         "OROR" },
    { PACKAGE,          TOKENTYPE_NONE,         "PACKAGE" },
    { PLUGEXPR,         TOKENTYPE_OPVAL,        "PLUGEXPR" },
    { PLUGSTMT,         TOKENTYPE_OPVAL,        "PLUGSTMT" },
    { PMFUNC,           TOKENTYPE_OPVAL,        "PMFUNC" },
    { POSTJOIN,         TOKENTYPE_NONE,         "POSTJOIN" },
    { POSTDEC,          TOKENTYPE_NONE,         "POSTDEC" },
    { POSTINC,          TOKENTYPE_NONE,         "POSTINC" },
    { POWOP,            TOKENTYPE_OPNUM,        "POWOP" },
    { PREDEC,           TOKENTYPE_NONE,         "PREDEC" },
    { PREINC,           TOKENTYPE_NONE,         "PREINC" },
    { PRIVATEREF,       TOKENTYPE_OPVAL,        "PRIVATEREF" },
    { QWLIST,           TOKENTYPE_OPVAL,        "QWLIST" },
    { REFGEN,           TOKENTYPE_NONE,         "REFGEN" },
    { RELOP,            TOKENTYPE_OPNUM,        "RELOP" },
    { REQUIRE,          TOKENTYPE_NONE,         "REQUIRE" },
    { SHIFTOP,          TOKENTYPE_OPNUM,        "SHIFTOP" },
    { SUB,              TOKENTYPE_NONE,         "SUB" },
    { THING,            TOKENTYPE_OPVAL,        "THING" },
    { UMINUS,           TOKENTYPE_NONE,         "UMINUS" },
    { UNIOP,            TOKENTYPE_OPNUM,        "UNIOP" },
    { UNIOPSUB,         TOKENTYPE_OPVAL,        "UNIOPSUB" },
    { UNLESS,           TOKENTYPE_IVAL,         "UNLESS" },
    { UNTIL,            TOKENTYPE_IVAL,         "UNTIL" },
    { USE,              TOKENTYPE_IVAL,         "USE" },
    { WHEN,             TOKENTYPE_IVAL,         "WHEN" },
    { WHILE,            TOKENTYPE_IVAL,         "WHILE" },
    { BAREWORD,         TOKENTYPE_OPVAL,        "BAREWORD" },
    { YADAYADA,         TOKENTYPE_IVAL,         "YADAYADA" },
    { 0,                TOKENTYPE_NONE,         NULL }
};

/* dump the returned token in rv, plus any optional arg in pl_yylval */

STATIC int
S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
{
    PERL_ARGS_ASSERT_TOKEREPORT;

    if (DEBUG_T_TEST) {
        const char *name = NULL;
        enum token_type type = TOKENTYPE_NONE;
        const struct debug_tokens *p;
        SV* const report = newSVpvs("<== ");

        for (p = debug_tokens; p->token; p++) {
            if (p->token == (int)rv) {
                name = p->name;
                type = p->type;
                break;
            }
        }
        if (name)
            Perl_sv_catpv(aTHX_ report, name);
        else if (isGRAPH(rv))
        {
            Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
            if ((char)rv == 'p')
                sv_catpvs(report, " (pending identifier)");
        }
        else if (!rv)
            sv_catpvs(report, "EOF");
        else
            Perl_sv_catpvf(aTHX_ report, "?? %" IVdf, (IV)rv);
        switch (type) {
        case TOKENTYPE_NONE:
            break;
        case TOKENTYPE_IVAL:
            Perl_sv_catpvf(aTHX_ report, "(ival=%" IVdf ")", (IV)lvalp->ival);
            break;
        case TOKENTYPE_OPNUM:
            Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
                                    PL_op_name[lvalp->ival]);
            break;
        case TOKENTYPE_PVAL:
            Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
            break;
        case TOKENTYPE_OPVAL:
            if (lvalp->opval) {
                Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
                                    PL_op_name[lvalp->opval->op_type]);
                if (lvalp->opval->op_type == OP_CONST) {
                    Perl_sv_catpvf(aTHX_ report, " %s",
                        SvPEEK(cSVOPx_sv(lvalp->opval)));
                }

            }
            else
                sv_catpvs(report, "(opval=null)");
            break;
        }
        PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
    };
    return (int)rv;
}


/* print the buffer with suitable escapes */

STATIC void
S_printbuf(pTHX_ const char *const fmt, const char *const s)
{
    SV* const tmp = newSVpvs("");

    PERL_ARGS_ASSERT_PRINTBUF;

    GCC_DIAG_IGNORE(-Wformat-nonliteral); /* fmt checked by caller */
    PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
    GCC_DIAG_RESTORE;
    SvREFCNT_dec(tmp);
}

#endif

static int
S_deprecate_commaless_var_list(pTHX) {
    PL_expect = XTERM;
    deprecate_fatal_in("5.28", "Use of comma-less variable list is deprecated");
    return REPORT(','); /* grandfather non-comma-format format */
}

/*
 * S_ao
 *
 * This subroutine looks for an '=' next to the operator that has just been
 * parsed and turns it into an ASSIGNOP if it finds one.
 */

STATIC int
S_ao(pTHX_ int toketype)
{
    if (*PL_bufptr == '=') {
        PL_bufptr++;
        if (toketype == ANDAND)
            pl_yylval.ival = OP_ANDASSIGN;
        else if (toketype == OROR)
            pl_yylval.ival = OP_ORASSIGN;
        else if (toketype == DORDOR)
            pl_yylval.ival = OP_DORASSIGN;
        toketype = ASSIGNOP;
    }
    return REPORT(toketype);
}

/*
 * S_no_op
 * When Perl expects an operator and finds something else, no_op
 * prints the warning.  It always prints "<something> found where
 * operator expected.  It prints "Missing semicolon on previous line?"
 * if the surprise occurs at the start of the line.  "do you need to
 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
 * where the compiler doesn't know if foo is a method call or a function.
 * It prints "Missing operator before end of line" if there's nothing
 * after the missing operator, or "... before <...>" if there is something
 * after the missing operator.
 *
 * PL_bufptr is expected to point to the start of the thing that was found,
 * and s after the next token or partial token.
 */

STATIC void
S_no_op(pTHX_ const char *const what, char *s)
{
    char * const oldbp = PL_bufptr;
    const bool is_first = (PL_oldbufptr == PL_linestart);

    PERL_ARGS_ASSERT_NO_OP;

    if (!s)
        s = oldbp;
    else
        PL_bufptr = s;
    yywarn(Perl_form(aTHX_ "%s found where operator expected", what), UTF ? SVf_UTF8 : 0);
    if (ckWARN_d(WARN_SYNTAX)) {
        if (is_first)
            Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                    "\t(Missing semicolon on previous line?)\n");
        else if (PL_oldoldbufptr && isIDFIRST_lazy_if_safe(PL_oldoldbufptr,
                                                           PL_bufend,
                                                           UTF))
        {
            const char *t;
            for (t = PL_oldoldbufptr;
                 (isWORDCHAR_lazy_if_safe(t, PL_bufend, UTF) || *t == ':');
                 t += UTF ? UTF8SKIP(t) : 1)
            {
                NOOP;
            }
            if (t < PL_bufptr && isSPACE(*t))
                Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                        "\t(Do you need to predeclare %" UTF8f "?)\n",
                      UTF8fARG(UTF, t - PL_oldoldbufptr, PL_oldoldbufptr));
        }
        else {
            assert(s >= oldbp);
            Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                    "\t(Missing operator before %" UTF8f "?)\n",
                     UTF8fARG(UTF, s - oldbp, oldbp));
        }
    }
    PL_bufptr = oldbp;
}

/*
 * S_missingterm
 * Complain about missing quote/regexp/heredoc terminator.
 * If it's called with NULL then it cauterizes the line buffer.
 * If we're in a delimited string and the delimiter is a control
 * character, it's reformatted into a two-char sequence like ^C.
 * This is fatal.
 */

STATIC void
S_missingterm(pTHX_ char *s)
{
    char tmpbuf[UTF8_MAXBYTES + 1];
    char q;
    bool uni = FALSE;
    SV *sv;
    if (s) {
        char * const nl = strrchr(s,'\n');
        if (nl)
            *nl = '\0';
        uni = UTF;
    }
    else if (PL_multi_close < 32) {
        *tmpbuf = '^';
        tmpbuf[1] = (char)toCTRL(PL_multi_close);
        tmpbuf[2] = '\0';
        s = tmpbuf;
    }
    else {
        if (LIKELY(PL_multi_close < 256)) {
            *tmpbuf = (char)PL_multi_close;
            tmpbuf[1] = '\0';
        }
        else {
            uni = TRUE;
            *uvchr_to_utf8((U8 *)tmpbuf, PL_multi_close) = 0;
        }
        s = tmpbuf;
    }
    q = strchr(s,'"') ? '\'' : '"';
    sv = sv_2mortal(newSVpv(s,0));
    if (uni)
        SvUTF8_on(sv);
    Perl_croak(aTHX_ "Can't find string terminator %c%" SVf
                     "%c anywhere before EOF",q,SVfARG(sv),q);
}

#include "feature.h"

/*
 * Check whether the named feature is enabled.
 */
bool
Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
{
    char he_name[8 + MAX_FEATURE_LEN] = "feature_";

    PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;

    assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);

    if (namelen > MAX_FEATURE_LEN)
        return FALSE;
    memcpy(&he_name[8], name, namelen);

    return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
                                     REFCOUNTED_HE_EXISTS));
}

/*
 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
 * utf16-to-utf8-reversed.
 */

#ifdef PERL_CR_FILTER
static void
strip_return(SV *sv)
{
    const char *s = SvPVX_const(sv);
    const char * const e = s + SvCUR(sv);

    PERL_ARGS_ASSERT_STRIP_RETURN;

    /* outer loop optimized to do nothing if there are no CR-LFs */
    while (s < e) {
        if (*s++ == '\r' && *s == '\n') {
            /* hit a CR-LF, need to copy the rest */
            char *d = s - 1;
            *d++ = *s++;
            while (s < e) {
                if (*s == '\r' && s[1] == '\n')
                    s++;
                *d++ = *s++;
            }
            SvCUR(sv) -= s - d;
            return;
        }
    }
}

STATIC I32
S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
{
    const I32 count = FILTER_READ(idx+1, sv, maxlen);
    if (count > 0 && !maxlen)
        strip_return(sv);
    return count;
}
#endif

/*
=for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags

Creates and initialises a new lexer/parser state object, supplying
a context in which to lex and parse from a new source of Perl code.
A pointer to the new state object is placed in L</PL_parser>.  An entry
is made on the save stack so that upon unwinding, the new state object
will be destroyed and the former value of L</PL_parser> will be restored.
Nothing else need be done to clean up the parsing context.

The code to be parsed comes from C<line> and C<rsfp>.  C<line>, if
non-null, provides a string (in SV form) containing code to be parsed.
A copy of the string is made, so subsequent modification of C<line>
does not affect parsing.  C<rsfp>, if non-null, provides an input stream
from which code will be read to be parsed.  If both are non-null, the
code in C<line> comes first and must consist of complete lines of input,
and C<rsfp> supplies the remainder of the source.

The C<flags> parameter is reserved for future use.  Currently it is only
used by perl internally, so extensions should always pass zero.

=cut
*/

/* LEX_START_SAME_FILTER indicates that this is not a new file, so it
   can share filters with the current parser.
   LEX_START_DONT_CLOSE indicates that the file handle wasn't opened by the
   caller, hence isn't owned by the parser, so shouldn't be closed on parser
   destruction. This is used to handle the case of defaulting to reading the
   script from the standard input because no filename was given on the command
   line (without getting confused by situation where STDIN has been closed, so
   the script handle is opened on fd 0)  */

void
Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
{
    const char *s = NULL;
    yy_parser *parser, *oparser;

    if (flags && flags & ~LEX_START_FLAGS)
        Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");

    /* create and initialise a parser */

    Newxz(parser, 1, yy_parser);
    parser->old_parser = oparser = PL_parser;
    PL_parser = parser;

    parser->stack = NULL;
    parser->stack_max1 = NULL;
    parser->ps = NULL;

    /* on scope exit, free this parser and restore any outer one */
    SAVEPARSER(parser);
    parser->saved_curcop = PL_curcop;

    /* initialise lexer state */

    parser->nexttoke = 0;
    parser->error_count = oparser ? oparser->error_count : 0;
    parser->copline = parser->preambling = NOLINE;
    parser->lex_state = LEX_NORMAL;
    parser->expect = XSTATE;
    parser->rsfp = rsfp;
    parser->rsfp_filters =
      !(flags & LEX_START_SAME_FILTER) || !oparser
        ? NULL
        : MUTABLE_AV(SvREFCNT_inc(
            oparser->rsfp_filters
             ? oparser->rsfp_filters
             : (oparser->rsfp_filters = newAV())
          ));

    Newx(parser->lex_brackstack, 120, char);
    Newx(parser->lex_casestack, 12, char);
    *parser->lex_casestack = '\0';
    Newxz(parser->lex_shared, 1, LEXSHARED);

    if (line) {
        STRLEN len;
        const U8* first_bad_char_loc;

        s = SvPV_const(line, len);

        if (SvUTF8(line) && ! is_utf8_string_loc((U8 *) s,
                                                 SvCUR(line),
                                                 &first_bad_char_loc))
        {
            _force_out_malformed_utf8_message(first_bad_char_loc,
                                              (U8 *) s + SvCUR(line),
                                              0,
                                              1 /* 1 means die */ );
            NOT_REACHED; /* NOTREACHED */
        }

        parser->linestr = flags & LEX_START_COPIED
                            ? SvREFCNT_inc_simple_NN(line)
                            : newSVpvn_flags(s, len, SvUTF8(line));
        if (!rsfp)
            sv_catpvs(parser->linestr, "\n;");
    } else {
        parser->linestr = newSVpvn("\n;", rsfp ? 1 : 2);
    }

    parser->oldoldbufptr =
        parser->oldbufptr =
        parser->bufptr =
        parser->linestart = SvPVX(parser->linestr);
    parser->bufend = parser->bufptr + SvCUR(parser->linestr);
    parser->last_lop = parser->last_uni = NULL;

    STATIC_ASSERT_STMT(FITS_IN_8_BITS(LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
                                                        |LEX_DONT_CLOSE_RSFP));
    parser->lex_flags = (U8) (flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
                                                        |LEX_DONT_CLOSE_RSFP));

    parser->in_pod = parser->filtered = 0;
}


/* delete a parser object */

void
Perl_parser_free(pTHX_  const yy_parser *parser)
{
    PERL_ARGS_ASSERT_PARSER_FREE;

    PL_curcop = parser->saved_curcop;
    SvREFCNT_dec(parser->linestr);

    if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
        PerlIO_clearerr(parser->rsfp);
    else if (parser->rsfp && (!parser->old_parser
          || (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
        PerlIO_close(parser->rsfp);
    SvREFCNT_dec(parser->rsfp_filters);
    SvREFCNT_dec(parser->lex_stuff);
    SvREFCNT_dec(parser->lex_sub_repl);

    Safefree(parser->lex_brackstack);
    Safefree(parser->lex_casestack);
    Safefree(parser->lex_shared);
    PL_parser = parser->old_parser;
    Safefree(parser);
}

void
Perl_parser_free_nexttoke_ops(pTHX_  yy_parser *parser, OPSLAB *slab)
{
    I32 nexttoke = parser->nexttoke;
    PERL_ARGS_ASSERT_PARSER_FREE_NEXTTOKE_OPS;
    while (nexttoke--) {
        if (S_is_opval_token(parser->nexttype[nexttoke] & 0xffff)
         && parser->nextval[nexttoke].opval
         && parser->nextval[nexttoke].opval->op_slabbed
         && OpSLAB(parser->nextval[nexttoke].opval) == slab) {
            op_free(parser->nextval[nexttoke].opval);
            parser->nextval[nexttoke].opval = NULL;
        }
    }
}


/*
=for apidoc AmxU|SV *|PL_parser-E<gt>linestr

Buffer scalar containing the chunk currently under consideration of the
text currently being lexed.  This is always a plain string scalar (for
which C<SvPOK> is true).  It is not intended to be used as a scalar by
normal scalar means; instead refer to the buffer directly by the pointer
variables described below.

The lexer maintains various C<char*> pointers to things in the
C<PL_parser-E<gt>linestr> buffer.  If C<PL_parser-E<gt>linestr> is ever
reallocated, all of these pointers must be updated.  Don't attempt to
do this manually, but rather use L</lex_grow_linestr> if you need to
reallocate the buffer.

The content of the text chunk in the buffer is commonly exactly one
complete line of input, up to and including a newline terminator,
but there are situations where it is otherwise.  The octets of the
buffer may be intended to be interpreted as either UTF-8 or Latin-1.
The function L</lex_bufutf8> tells you which.  Do not use the C<SvUTF8>
flag on this scalar, which may disagree with it.

For direct examination of the buffer, the variable
L</PL_parser-E<gt>bufend> points to the end of the buffer.  The current
lexing position is pointed to by L</PL_parser-E<gt>bufptr>.  Direct use
of these pointers is usually preferable to examination of the scalar
through normal scalar means.

=for apidoc AmxU|char *|PL_parser-E<gt>bufend

Direct pointer to the end of the chunk of text currently being lexed, the
end of the lexer buffer.  This is equal to C<SvPVX(PL_parser-E<gt>linestr)
+ SvCUR(PL_parser-E<gt>linestr)>.  A C<NUL> character (zero octet) is
always located at the end of the buffer, and does not count as part of
the buffer's contents.

=for apidoc AmxU|char *|PL_parser-E<gt>bufptr

Points to the current position of lexing inside the lexer buffer.
Characters around this point may be freely examined, within
the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
L</PL_parser-E<gt>bufend>.  The octets of the buffer may be intended to be
interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.

Lexing code (whether in the Perl core or not) moves this pointer past
the characters that it consumes.  It is also expected to perform some
bookkeeping whenever a newline character is consumed.  This movement
can be more conveniently performed by the function L</lex_read_to>,
which handles newlines appropriately.

Interpretation of the buffer's octets can be abstracted out by
using the slightly higher-level functions L</lex_peek_unichar> and
L</lex_read_unichar>.

=for apidoc AmxU|char *|PL_parser-E<gt>linestart

Points to the start of the current line inside the lexer buffer.
This is useful for indicating at which column an error occurred, and
not much else.  This must be updated by any lexing code that consumes
a newline; the function L</lex_read_to> handles this detail.

=cut
*/

/*
=for apidoc Amx|bool|lex_bufutf8

Indicates whether the octets in the lexer buffer
(L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
of Unicode characters.  If not, they should be interpreted as Latin-1
characters.  This is analogous to the C<SvUTF8> flag for scalars.

In UTF-8 mode, it is not guaranteed that the lexer buffer actually
contains valid UTF-8.  Lexing code must be robust in the face of invalid
encoding.

The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
is significant, but not the whole story regarding the input character
encoding.  Normally, when a file is being read, the scalar contains octets
and its C<SvUTF8> flag is off, but the octets should be interpreted as
UTF-8 if the C<use utf8> pragma is in effect.  During a string eval,
however, the scalar may have the C<SvUTF8> flag on, and in this case its
octets should be interpreted as UTF-8 unless the C<use bytes> pragma
is in effect.  This logic may change in the future; use this function
instead of implementing the logic yourself.

=cut
*/

bool
Perl_lex_bufutf8(pTHX)
{
    return UTF;
}

/*
=for apidoc Amx|char *|lex_grow_linestr|STRLEN len

Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
at least C<len> octets (including terminating C<NUL>).  Returns a
pointer to the reallocated buffer.  This is necessary before making
any direct modification of the buffer that would increase its length.
L</lex_stuff_pvn> provides a more convenient way to insert text into
the buffer.

Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
this function updates all of the lexer's variables that point directly
into the buffer.

=cut
*/

char *
Perl_lex_grow_linestr(pTHX_ STRLEN len)
{
    SV *linestr;
    char *buf;
    STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
    STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
    bool current;

    linestr = PL_parser->linestr;
    buf = SvPVX(linestr);
    if (len <= SvLEN(linestr))
        return buf;

    /* Is the lex_shared linestr SV the same as the current linestr SV?
     * Only in this case does re_eval_start need adjusting, since it
     * points within lex_shared->ls_linestr's buffer */
    current = (   !PL_parser->lex_shared->ls_linestr
               || linestr == PL_parser->lex_shared->ls_linestr);

    bufend_pos = PL_parser->bufend - buf;
    bufptr_pos = PL_parser->bufptr - buf;
    oldbufptr_pos = PL_parser->oldbufptr - buf;
    oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
    linestart_pos = PL_parser->linestart - buf;
    last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
    last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
    re_eval_start_pos = (current && PL_parser->lex_shared->re_eval_start) ?
                            PL_parser->lex_shared->re_eval_start - buf : 0;

    buf = sv_grow(linestr, len);

    PL_parser->bufend = buf + bufend_pos;
    PL_parser->bufptr = buf + bufptr_pos;
    PL_parser->oldbufptr = buf + oldbufptr_pos;
    PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
    PL_parser->linestart = buf + linestart_pos;
    if (PL_parser->last_uni)
        PL_parser->last_uni = buf + last_uni_pos;
    if (PL_parser->last_lop)
        PL_parser->last_lop = buf + last_lop_pos;
    if (current && PL_parser->lex_shared->re_eval_start)
        PL_parser->lex_shared->re_eval_start  = buf + re_eval_start_pos;
    return buf;
}

/*
=for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags

Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
reallocating the buffer if necessary.  This means that lexing code that
runs later will see the characters as if they had appeared in the input.
It is not recommended to do this as part of normal parsing, and most
uses of this facility run the risk of the inserted characters being
interpreted in an unintended manner.

The string to be inserted is represented by C<len> octets starting
at C<pv>.  These octets are interpreted as either UTF-8 or Latin-1,
according to whether the C<LEX_STUFF_UTF8> flag is set in C<flags>.
The characters are recoded for the lexer buffer, according to how the
buffer is currently being interpreted (L</lex_bufutf8>).  If a string
to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
function is more convenient.

=cut
*/

void
Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
{
    dVAR;
    char *bufptr;
    PERL_ARGS_ASSERT_LEX_STUFF_PVN;
    if (flags & ~(LEX_STUFF_UTF8))
        Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
    if (UTF) {
        if (flags & LEX_STUFF_UTF8) {
            goto plain_copy;
        } else {
            STRLEN highhalf = 0;    /* Count of variants */
            const char *p, *e = pv+len;
            for (p = pv; p != e; p++) {
                if (! UTF8_IS_INVARIANT(*p)) {
                    highhalf++;
                }
            }
            if (!highhalf)
                goto plain_copy;
            lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
            bufptr = PL_parser->bufptr;
            Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
            SvCUR_set(PL_parser->linestr,
                SvCUR(PL_parser->linestr) + len+highhalf);
            PL_parser->bufend += len+highhalf;
            for (p = pv; p != e; p++) {
                U8 c = (U8)*p;
                if (! UTF8_IS_INVARIANT(c)) {
                    *bufptr++ = UTF8_TWO_BYTE_HI(c);
                    *bufptr++ = UTF8_TWO_BYTE_LO(c);
                } else {
                    *bufptr++ = (char)c;
                }
            }
        }
    } else {
        if (flags & LEX_STUFF_UTF8) {
            STRLEN highhalf = 0;
            const char *p, *e = pv+len;
            for (p = pv; p != e; p++) {
                U8 c = (U8)*p;
                if (UTF8_IS_ABOVE_LATIN1(c)) {
                    Perl_croak(aTHX_ "Lexing code attempted to stuff "
                                "non-Latin-1 character into Latin-1 input");
                } else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e)) {
                    p++;
                    highhalf++;
                } else if (! UTF8_IS_INVARIANT(c)) {
                    _force_out_malformed_utf8_message((U8 *) p, (U8 *) e,
                                                      0,
                                                      1 /* 1 means die */ );
                    NOT_REACHED; /* NOTREACHED */
                }
            }
            if (!highhalf)
                goto plain_copy;
            lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
            bufptr = PL_parser->bufptr;
            Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
            SvCUR_set(PL_parser->linestr,
                SvCUR(PL_parser->linestr) + len-highhalf);
            PL_parser->bufend += len-highhalf;
            p = pv;
            while (p < e) {
                if (UTF8_IS_INVARIANT(*p)) {
                    *bufptr++ = *p;
                    p++;
                }
                else {
                    assert(p < e -1 );
                    *bufptr++ = EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1));
                    p += 2;
                }
            }
        } else {
          plain_copy:
            lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
            bufptr = PL_parser->bufptr;
            Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
            SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
            PL_parser->bufend += len;
            Copy(pv, bufptr, len, char);
        }
    }
}

/*
=for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags

Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
reallocating the buffer if necessary.  This means that lexing code that
runs later will see the characters as if they had appeared in the input.
It is not recommended to do this as part of normal parsing, and most
uses of this facility run the risk of the inserted characters being
interpreted in an unintended manner.

The string to be inserted is represented by octets starting at C<pv>
and continuing to the first nul.  These octets are interpreted as either
UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
in C<flags>.  The characters are recoded for the lexer buffer, according
to how the buffer is currently being interpreted (L</lex_bufutf8>).
If it is not convenient to nul-terminate a string to be inserted, the
L</lex_stuff_pvn> function is more appropriate.

=cut
*/

void
Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
{
    PERL_ARGS_ASSERT_LEX_STUFF_PV;
    lex_stuff_pvn(pv, strlen(pv), flags);
}

/*
=for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags

Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
reallocating the buffer if necessary.  This means that lexing code that
runs later will see the characters as if they had appeared in the input.
It is not recommended to do this as part of normal parsing, and most
uses of this facility run the risk of the inserted characters being
interpreted in an unintended manner.

The string to be inserted is the string value of C<sv>.  The characters
are recoded for the lexer buffer, according to how the buffer is currently
being interpreted (L</lex_bufutf8>).  If a string to be inserted is
not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
need to construct a scalar.

=cut
*/

void
Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
{
    char *pv;
    STRLEN len;
    PERL_ARGS_ASSERT_LEX_STUFF_SV;
    if (flags)
        Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
    pv = SvPV(sv, len);
    lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
}

/*
=for apidoc Amx|void|lex_unstuff|char *ptr

Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
C<ptr>.  Text following C<ptr> will be moved, and the buffer shortened.
This hides the discarded text from any lexing code that runs later,
as if the text had never appeared.

This is not the normal way to consume lexed text.  For that, use
L</lex_read_to>.

=cut
*/

void
Perl_lex_unstuff(pTHX_ char *ptr)
{
    char *buf, *bufend;
    STRLEN unstuff_len;
    PERL_ARGS_ASSERT_LEX_UNSTUFF;
    buf = PL_parser->bufptr;
    if (ptr < buf)
        Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
    if (ptr == buf)
        return;
    bufend = PL_parser->bufend;
    if (ptr > bufend)
        Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
    unstuff_len = ptr - buf;
    Move(ptr, buf, bufend+1-ptr, char);
    SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
    PL_parser->bufend = bufend - unstuff_len;
}

/*
=for apidoc Amx|void|lex_read_to|char *ptr

Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
to C<ptr>.  This advances L</PL_parser-E<gt>bufptr> to match C<ptr>,
performing the correct bookkeeping whenever a newline character is passed.
This is the normal way to consume lexed text.

Interpretation of the buffer's octets can be abstracted out by
using the slightly higher-level functions L</lex_peek_unichar> and
L</lex_read_unichar>.

=cut
*/

void
Perl_lex_read_to(pTHX_ char *ptr)
{
    char *s;
    PERL_ARGS_ASSERT_LEX_READ_TO;
    s = PL_parser->bufptr;
    if (ptr < s || ptr > PL_parser->bufend)
        Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
    for (; s != ptr; s++)
        if (*s == '\n') {
            COPLINE_INC_WITH_HERELINES;
            PL_parser->linestart = s+1;
        }
    PL_parser->bufptr = ptr;
}

/*
=for apidoc Amx|void|lex_discard_to|char *ptr

Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
up to C<ptr>.  The remaining content of the buffer will be moved, and
all pointers into the buffer updated appropriately.  C<ptr> must not
be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
it is not permitted to discard text that has yet to be lexed.

Normally it is not necessarily to do this directly, because it suffices to
use the implicit discarding behaviour of L</lex_next_chunk> and things
based on it.  However, if a token stretches across multiple lines,
and the lexing code has kept multiple lines of text in the buffer for
that purpose, then after completion of the token it would be wise to
explicitly discard the now-unneeded earlier lines, to avoid future
multi-line tokens growing the buffer without bound.

=cut
*/

void
Perl_lex_discard_to(pTHX_ char *ptr)
{
    char *buf;
    STRLEN discard_len;
    PERL_ARGS_ASSERT_LEX_DISCARD_TO;
    buf = SvPVX(PL_parser->linestr);
    if (ptr < buf)
        Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
    if (ptr == buf)
        return;
    if (ptr > PL_parser->bufptr)
        Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
    discard_len = ptr - buf;
    if (PL_parser->oldbufptr < ptr)
        PL_parser->oldbufptr = ptr;
    if (PL_parser->oldoldbufptr < ptr)
        PL_parser->oldoldbufptr = ptr;
    if (PL_parser->last_uni && PL_parser->last_uni < ptr)
        PL_parser->last_uni = NULL;
    if (PL_parser->last_lop && PL_parser->last_lop < ptr)
        PL_parser->last_lop = NULL;
    Move(ptr, buf, PL_parser->bufend+1-ptr, char);
    SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
    PL_parser->bufend -= discard_len;
    PL_parser->bufptr -= discard_len;
    PL_parser->oldbufptr -= discard_len;
    PL_parser->oldoldbufptr -= discard_len;
    if (PL_parser->last_uni)
        PL_parser->last_uni -= discard_len;
    if (PL_parser->last_lop)
        PL_parser->last_lop -= discard_len;
}

/*
=for apidoc Amx|bool|lex_next_chunk|U32 flags

Reads in the next chunk of text to be lexed, appending it to
L</PL_parser-E<gt>linestr>.  This should be called when lexing code has
looked to the end of the current chunk and wants to know more.  It is
usual, but not necessary, for lexing to have consumed the entirety of
the current chunk at this time.

If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
chunk (i.e., the current chunk has been entirely consumed), normally the
current chunk will be discarded at the same time that the new chunk is
read in.  If C<flags> has the C<LEX_KEEP_PREVIOUS> bit set, the current chunk
will not be discarded.  If the current chunk has not been entirely
consumed, then it will not be discarded regardless of the flag.

Returns true if some new text was added to the buffer, or false if the
buffer has reached the end of the input text.

=cut
*/

#define LEX_FAKE_EOF 0x80000000
#define LEX_NO_TERM  0x40000000 /* here-doc */

bool
Perl_lex_next_chunk(pTHX_ U32 flags)
{
    SV *linestr;
    char *buf;
    STRLEN old_bufend_pos, new_bufend_pos;
    STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
    STRLEN linestart_pos, last_uni_pos, last_lop_pos;
    bool got_some_for_debugger = 0;
    bool got_some;
    const U8* first_bad_char_loc;

    if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF|LEX_NO_TERM))
        Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
    if (!(flags & LEX_NO_TERM) && PL_lex_inwhat)
        return FALSE;
    linestr = PL_parser->linestr;
    buf = SvPVX(linestr);
    if (!(flags & LEX_KEEP_PREVIOUS)
          && PL_parser->bufptr == PL_parser->bufend)
    {
        old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
        linestart_pos = 0;
        if (PL_parser->last_uni != PL_parser->bufend)
            PL_parser->last_uni = NULL;
        if (PL_parser->last_lop != PL_parser->bufend)
            PL_parser->last_lop = NULL;
        last_uni_pos = last_lop_pos = 0;
        *buf = 0;
        SvCUR(linestr) = 0;
    } else {
        old_bufend_pos = PL_parser->bufend - buf;
        bufptr_pos = PL_parser->bufptr - buf;
        oldbufptr_pos = PL_parser->oldbufptr - buf;
        oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
        linestart_pos = PL_parser->linestart - buf;
        last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
        last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
    }
    if (flags & LEX_FAKE_EOF) {
        goto eof;
    } else if (!PL_parser->rsfp && !PL_parser->filtered) {
        got_some = 0;
    } else if (filter_gets(linestr, old_bufend_pos)) {
        got_some = 1;
        got_some_for_debugger = 1;
    } else if (flags & LEX_NO_TERM) {
        got_some = 0;
    } else {
        if (!SvPOK(linestr))   /* can get undefined by filter_gets */
            SvPVCLEAR(linestr);
        eof:
        /* End of real input.  Close filehandle (unless it was STDIN),
         * then add implicit termination.
         */
        if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
            PerlIO_clearerr(PL_parser->rsfp);
        else if (PL_parser->rsfp)
            (void)PerlIO_close(PL_parser->rsfp);
        PL_parser->rsfp = NULL;
        PL_parser->in_pod = PL_parser->filtered = 0;
        if (!PL_in_eval && PL_minus_p) {
            sv_catpvs(linestr,
                /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
            PL_minus_n = PL_minus_p = 0;
        } else if (!PL_in_eval && PL_minus_n) {
            sv_catpvs(linestr, /*{*/";}");
            PL_minus_n = 0;
        } else
            sv_catpvs(linestr, ";");
        got_some = 1;
    }
    buf = SvPVX(linestr);
    new_bufend_pos = SvCUR(linestr);
    PL_parser->bufend = buf + new_bufend_pos;
    PL_parser->bufptr = buf + bufptr_pos;

    if (UTF && ! is_utf8_string_loc((U8 *) PL_parser->bufptr,
                                    PL_parser->bufend - PL_parser->bufptr,
                                    &first_bad_char_loc))
    {
        _force_out_malformed_utf8_message(first_bad_char_loc,
                                          (U8 *) PL_parser->bufend,
                                          0,
                                          1 /* 1 means die */ );
        NOT_REACHED; /* NOTREACHED */
    }

    PL_parser->oldbufptr = buf + oldbufptr_pos;
    PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
    PL_parser->linestart = buf + linestart_pos;
    if (PL_parser->last_uni)
        PL_parser->last_uni = buf + last_uni_pos;
    if (PL_parser->last_lop)
        PL_parser->last_lop = buf + last_lop_pos;
    if (PL_parser->preambling != NOLINE) {
        CopLINE_set(PL_curcop, PL_parser->preambling + 1);
        PL_parser->preambling = NOLINE;
    }
    if (   got_some_for_debugger
        && PERLDB_LINE_OR_SAVESRC
        && PL_curstash != PL_debstash)
    {
        /* debugger active and we're not compiling the debugger code,
         * so store the line into the debugger's array of lines
         */
        update_debugger_info(NULL, buf+old_bufend_pos,
            new_bufend_pos-old_bufend_pos);
    }
    return got_some;
}

/*
=for apidoc Amx|I32|lex_peek_unichar|U32 flags

Looks ahead one (Unicode) character in the text currently being lexed.
Returns the codepoint (unsigned integer value) of the next character,
or -1 if lexing has reached the end of the input text.  To consume the
peeked character, use L</lex_read_unichar>.

If the next character is in (or extends into) the next chunk of input
text, the next chunk will be read in.  Normally the current chunk will be
discarded at the same time, but if C<flags> has the C<LEX_KEEP_PREVIOUS>
bit set, then the current chunk will not be discarded.

If the input is being interpreted as UTF-8 and a UTF-8 encoding error
is encountered, an exception is generated.

=cut
*/

I32
Perl_lex_peek_unichar(pTHX_ U32 flags)
{
    dVAR;
    char *s, *bufend;
    if (flags & ~(LEX_KEEP_PREVIOUS))
        Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
    s = PL_parser->bufptr;
    bufend = PL_parser->bufend;
    if (UTF) {
        U8 head;
        I32 unichar;
        STRLEN len, retlen;
        if (s == bufend) {
            if (!lex_next_chunk(flags))
                return -1;
            s = PL_parser->bufptr;
            bufend = PL_parser->bufend;
        }
        head = (U8)*s;
        if (UTF8_IS_INVARIANT(head))
            return head;
        if (UTF8_IS_START(head)) {
            len = UTF8SKIP(&head);
            while ((STRLEN)(bufend-s) < len) {
                if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
                    break;
                s = PL_parser->bufptr;
                bufend = PL_parser->bufend;
            }
        }
        unichar = utf8n_to_uvchr((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
        if (retlen == (STRLEN)-1) {
            _force_out_malformed_utf8_message((U8 *) s,
                                              (U8 *) bufend,
                                              0,
                                              1 /* 1 means die */ );
            NOT_REACHED; /* NOTREACHED */
        }
        return unichar;
    } else {
        if (s == bufend) {
            if (!lex_next_chunk(flags))
                return -1;
            s = PL_parser->bufptr;
        }
        return (U8)*s;
    }
}

/*
=for apidoc Amx|I32|lex_read_unichar|U32 flags

Reads the next (Unicode) character in the text currently being lexed.
Returns the codepoint (unsigned integer value) of the character read,
and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
if lexing has reached the end of the input text.  To non-destructively
examine the next character, use L</lex_peek_unichar> instead.

If the next character is in (or extends into) the next chunk of input
text, the next chunk will be read in.  Normally the current chunk will be
discarded at the same time, but if C<flags> has the C<LEX_KEEP_PREVIOUS>
bit set, then the current chunk will not be discarded.

If the input is being interpreted as UTF-8 and a UTF-8 encoding error
is encountered, an exception is generated.

=cut
*/

I32
Perl_lex_read_unichar(pTHX_ U32 flags)
{
    I32 c;
    if (flags & ~(LEX_KEEP_PREVIOUS))
        Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
    c = lex_peek_unichar(flags);
    if (c != -1) {
        if (c == '\n')
            COPLINE_INC_WITH_HERELINES;
        if (UTF)
            PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
        else
            ++(PL_parser->bufptr);
    }
    return c;
}

/*
=for apidoc Amx|void|lex_read_space|U32 flags

Reads optional spaces, in Perl style, in the text currently being
lexed.  The spaces may include ordinary whitespace characters and
Perl-style comments.  C<#line> directives are processed if encountered.
L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
at a non-space character (or the end of the input text).

If spaces extend into the next chunk of input text, the next chunk will
be read in.  Normally the current chunk will be discarded at the same
time, but if C<flags> has the C<LEX_KEEP_PREVIOUS> bit set, then the current
chunk will not be discarded.

=cut
*/

#define LEX_NO_INCLINE    0x40000000
#define LEX_NO_NEXT_CHUNK 0x80000000

void
Perl_lex_read_space(pTHX_ U32 flags)
{
    char *s, *bufend;
    const bool can_incline = !(flags & LEX_NO_INCLINE);
    bool need_incline = 0;
    if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK|LEX_NO_INCLINE))
        Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
    s = PL_parser->bufptr;
    bufend = PL_parser->bufend;
    while (1) {
        char c = *s;
        if (c == '#') {
            do {
                c = *++s;
            } while (!(c == '\n' || (c == 0 && s == bufend)));
        } else if (c == '\n') {
            s++;
            if (can_incline) {
                PL_parser->linestart = s;
                if (s == bufend)
                    need_incline = 1;
                else
                    incline(s);
            }
        } else if (isSPACE(c)) {
            s++;
        } else if (c == 0 && s == bufend) {
            bool got_more;
            line_t l;
            if (flags & LEX_NO_NEXT_CHUNK)
                break;
            PL_parser->bufptr = s;
            l = CopLINE(PL_curcop);
            CopLINE(PL_curcop) += PL_parser->herelines + 1;
            got_more = lex_next_chunk(flags);
            CopLINE_set(PL_curcop, l);
            s = PL_parser->bufptr;
            bufend = PL_parser->bufend;
            if (!got_more)
                break;
            if (can_incline && need_incline && PL_parser->rsfp) {
                incline(s);
                need_incline = 0;
            }
        } else if (!c) {
            s++;
        } else {
            break;
        }
    }
    PL_parser->bufptr = s;
}

/*

=for apidoc EXMp|bool|validate_proto|SV *name|SV *proto|bool warn

This function performs syntax checking on a prototype, C<proto>.
If C<warn> is true, any illegal characters or mismatched brackets
will trigger illegalproto warnings, declaring that they were
detected in the prototype for C<name>.

The return value is C<true> if this is a valid prototype, and
C<false> if it is not, regardless of whether C<warn> was C<true> or
C<false>.

Note that C<NULL> is a valid C<proto> and will always return C<true>.

=cut

 */

bool
Perl_validate_proto(pTHX_ SV *name, SV *proto, bool warn)
{
    STRLEN len, origlen;
    char *p;
    bool bad_proto = FALSE;
    bool in_brackets = FALSE;
    bool after_slash = FALSE;
    char greedy_proto = ' ';
    bool proto_after_greedy_proto = FALSE;
    bool must_be_last = FALSE;
    bool underscore = FALSE;
    bool bad_proto_after_underscore = FALSE;

    PERL_ARGS_ASSERT_VALIDATE_PROTO;

    if (!proto)
        return TRUE;

    p = SvPV(proto, len);
    origlen = len;
    for (; len--; p++) {
        if (!isSPACE(*p)) {
            if (must_be_last)
                proto_after_greedy_proto = TRUE;
            if (underscore) {
                if (!strchr(";@%", *p))
                    bad_proto_after_underscore = TRUE;
                underscore = FALSE;
            }
            if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
                bad_proto = TRUE;
            }
            else {
                if (*p == '[')
                    in_brackets = TRUE;
                else if (*p == ']')
                    in_brackets = FALSE;
                else if ((*p == '@' || *p == '%')
                         && !after_slash
                         && !in_brackets )
                {
                    must_be_last = TRUE;
                    greedy_proto = *p;
                }
                else if (*p == '_')
                    underscore = TRUE;
            }
            if (*p == '\\')
                after_slash = TRUE;
            else
                after_slash = FALSE;
        }
    }

    if (warn) {
        SV *tmpsv = newSVpvs_flags("", SVs_TEMP);
        p -= origlen;
        p = SvUTF8(proto)
            ? sv_uni_display(tmpsv, newSVpvn_flags(p, origlen, SVs_TEMP | SVf_UTF8),
                             origlen, UNI_DISPLAY_ISPRINT)
            : pv_pretty(tmpsv, p, origlen, 60, NULL, NULL, PERL_PV_ESCAPE_NONASCII);

        if (proto_after_greedy_proto)
            Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
                        "Prototype after '%c' for %" SVf " : %s",
                        greedy_proto, SVfARG(name), p);
        if (in_brackets)
            Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
                        "Missing ']' in prototype for %" SVf " : %s",
                        SVfARG(name), p);
        if (bad_proto)
            Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
                        "Illegal character in prototype for %" SVf " : %s",
                        SVfARG(name), p);
        if (bad_proto_after_underscore)
            Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
                        "Illegal character after '_' in prototype for %" SVf " : %s",
                        SVfARG(name), p);
    }

    return (! (proto_after_greedy_proto || bad_proto) );
}

/*
 * S_incline
 * This subroutine has nothing to do with tilting, whether at windmills
 * or pinball tables.  Its name is short for "increment line".  It
 * increments the current line number in CopLINE(PL_curcop) and checks
 * to see whether the line starts with a comment of the form
 *    # line 500 "foo.pm"
 * If so, it sets the current line number and file to the values in the comment.
 */

STATIC void
S_incline(pTHX_ const char *s)
{
    const char *t;
    const char *n;
    const char *e;
    line_t line_num;
    UV uv;

    PERL_ARGS_ASSERT_INCLINE;

    COPLINE_INC_WITH_HERELINES;
    if (!PL_rsfp && !PL_parser->filtered && PL_lex_state == LEX_NORMAL
     && s+1 == PL_bufend && *s == ';') {
        /* fake newline in string eval */
        CopLINE_dec(PL_curcop);
        return;
    }
    if (*s++ != '#')
        return;
    while (SPACE_OR_TAB(*s))
        s++;
    if (strEQs(s, "line"))
        s += 4;
    else
        return;
    if (SPACE_OR_TAB(*s))
        s++;
    else
        return;
    while (SPACE_OR_TAB(*s))
        s++;
    if (!isDIGIT(*s))
        return;

    n = s;
    while (isDIGIT(*s))
        s++;
    if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
        return;
    while (SPACE_OR_TAB(*s))
        s++;
    if (*s == '"' && (t = strchr(s+1, '"'))) {
        s++;
        e = t + 1;
    }
    else {
        t = s;
        while (*t && !isSPACE(*t))
            t++;
        e = t;
    }
    while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
        e++;
    if (*e != '\n' && *e != '\0')
        return;         /* false alarm */

    if (!grok_atoUV(n, &uv, &e))
        return;
    line_num = ((line_t)uv) - 1;

    if (t - s > 0) {
        const STRLEN len = t - s;

        if (!PL_rsfp && !PL_parser->filtered) {
            /* must copy *{"::_<(eval N)[oldfilename:L]"}
             * to *{"::_<newfilename"} */
            /* However, the long form of evals is only turned on by the
               debugger - usually they're "(eval %lu)" */
            GV * const cfgv = CopFILEGV(PL_curcop);
            if (cfgv) {
                char smallbuf[128];
                STRLEN tmplen2 = len;
                char *tmpbuf2;
                GV *gv2;

                if (tmplen2 + 2 <= sizeof smallbuf)
                    tmpbuf2 = smallbuf;
                else
                    Newx(tmpbuf2, tmplen2 + 2, char);

                tmpbuf2[0] = '_';
                tmpbuf2[1] = '<';

                memcpy(tmpbuf2 + 2, s, tmplen2);
                tmplen2 += 2;

                gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
                if (!isGV(gv2)) {
                    gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
                    /* adjust ${"::_<newfilename"} to store the new file name */
                    GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
                    /* The line number may differ. If that is the case,
                       alias the saved lines that are in the array.
                       Otherwise alias the whole array. */
                    if (CopLINE(PL_curcop) == line_num) {
                        GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(cfgv)));
                        GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(cfgv)));
                    }
                    else if (GvAV(cfgv)) {
                        AV * const av = GvAV(cfgv);
                        const I32 start = CopLINE(PL_curcop)+1;
                        I32 items = AvFILLp(av) - start;
                        if (items > 0) {
                            AV * const av2 = GvAVn(gv2);
                            SV **svp = AvARRAY(av) + start;
                            I32 l = (I32)line_num+1;
                            while (items--)
                                av_store(av2, l++, SvREFCNT_inc(*svp++));
                        }
                    }
                }

                if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
            }
        }
        CopFILE_free(PL_curcop);
        CopFILE_setn(PL_curcop, s, len);
    }
    CopLINE_set(PL_curcop, line_num);
}

STATIC void
S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
{
    AV *av = CopFILEAVx(PL_curcop);
    if (av) {
        SV * sv;
        if (PL_parser->preambling == NOLINE) sv = newSV_type(SVt_PVMG);
        else {
            sv = *av_fetch(av, 0, 1);
            SvUPGRADE(sv, SVt_PVMG);
        }
        if (!SvPOK(sv)) SvPVCLEAR(sv);
        if (orig_sv)
            sv_catsv(sv, orig_sv);
        else
            sv_catpvn(sv, buf, len);
        if (!SvIOK(sv)) {
            (void)SvIOK_on(sv);
            SvIV_set(sv, 0);
        }
        if (PL_parser->preambling == NOLINE)
            av_store(av, CopLINE(PL_curcop), sv);
    }
}

/*
 * skipspace
 * Called to gobble the appropriate amount and type of whitespace.
 * Skips comments as well.
 * Returns the next character after the whitespace that is skipped.
 *
 * peekspace
 * Same thing, but look ahead without incrementing line numbers or
 * adjusting PL_linestart.
 */

#define skipspace(s) skipspace_flags(s, 0)
#define peekspace(s) skipspace_flags(s, LEX_NO_INCLINE)

STATIC char *
S_skipspace_flags(pTHX_ char *s, U32 flags)
{
    PERL_ARGS_ASSERT_SKIPSPACE_FLAGS;
    if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
        while (s < PL_bufend && (SPACE_OR_TAB(*s) || !*s))
            s++;
    } else {
        STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
        PL_bufptr = s;
        lex_read_space(flags | LEX_KEEP_PREVIOUS |
                (PL_lex_inwhat || PL_lex_state == LEX_FORMLINE ?
                    LEX_NO_NEXT_CHUNK : 0));
        s = PL_bufptr;
        PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
        if (PL_linestart > PL_bufptr)
            PL_bufptr = PL_linestart;
        return s;
    }
    return s;
}

/*
 * S_check_uni
 * Check the unary operators to ensure there's no ambiguity in how they're
 * used.  An ambiguous piece of code would be:
 *     rand + 5
 * This doesn't mean rand() + 5.  Because rand() is a unary operator,
 * the +5 is its argument.
 */

STATIC void
S_check_uni(pTHX)
{
    const char *s;
    const char *t;

    if (PL_oldoldbufptr != PL_last_uni)
        return;
    while (isSPACE(*PL_last_uni))
        PL_last_uni++;
    s = PL_last_uni;
    while (isWORDCHAR_lazy_if_safe(s, PL_bufend, UTF) || *s == '-')
        s += UTF ? UTF8SKIP(s) : 1;
    if ((t = strchr(s, '(')) && t < PL_bufptr)
        return;

    Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
                     "Warning: Use of \"%" UTF8f "\" without parentheses is ambiguous",
                     UTF8fARG(UTF, (int)(s - PL_last_uni), PL_last_uni));
}

/*
 * LOP : macro to build a list operator.  Its behaviour has been replaced
 * with a subroutine, S_lop() for which LOP is just another name.
 */

#define LOP(f,x) return lop(f,x,s)

/*
 * S_lop
 * Build a list operator (or something that might be one).  The rules:
 *  - if we have a next token, then it's a list operator (no parens) for
 *    which the next token has already been parsed; e.g.,
 *       sort foo @args
 *       sort foo (@args)
 *  - if the next thing is an opening paren, then it's a function
 *  - else it's a list operator
 */

STATIC I32
S_lop(pTHX_ I32 f, U8 x, char *s)
{
    PERL_ARGS_ASSERT_LOP;

    pl_yylval.ival = f;
    CLINE;
    PL_bufptr = s;
    PL_last_lop = PL_oldbufptr;
    PL_last_lop_op = (OPCODE)f;
    if (PL_nexttoke)
        goto lstop;
    PL_expect = x;
    if (*s == '(')
        return REPORT(FUNC);
    s = skipspace(s);
    if (*s == '(')
        return REPORT(FUNC);
    else {
        lstop:
        if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
            PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
        return REPORT(LSTOP);
    }
}

/*
 * S_force_next
 * When the lexer realizes it knows the next token (for instance,
 * it is reordering tokens for the parser) then it can call S_force_next
 * to know what token to return the next time the lexer is called.  Caller
 * will need to set PL_nextval[] and possibly PL_expect to ensure
 * the lexer handles the token correctly.
 */

STATIC void
S_force_next(pTHX_ I32 type)
{
#ifdef DEBUGGING
    if (DEBUG_T_TEST) {
        PerlIO_printf(Perl_debug_log, "### forced token:\n");
        tokereport(type, &NEXTVAL_NEXTTOKE);
    }
#endif
    assert(PL_nexttoke < C_ARRAY_LENGTH(PL_nexttype));
    PL_nexttype[PL_nexttoke] = type;
    PL_nexttoke++;
}

/*
 * S_postderef
 *
 * This subroutine handles postfix deref syntax after the arrow has already
 * been emitted.  @* $* etc. are emitted as two separate token right here.
 * @[ @{ %[ %{ *{ are emitted also as two tokens, but this function emits
 * only the first, leaving yylex to find the next.
 */

static int
S_postderef(pTHX_ int const funny, char const next)
{
    assert(funny == DOLSHARP || strchr("$@%&*", funny));
    if (next == '*') {
        PL_expect = XOPERATOR;
        if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
            assert('@' == funny || '$' == funny || DOLSHARP == funny);
            PL_lex_state = LEX_INTERPEND;
            if ('@' == funny)
                force_next(POSTJOIN);
        }
        force_next(next);
        PL_bufptr+=2;
    }
    else {
        if ('@' == funny && PL_lex_state == LEX_INTERPNORMAL
         && !PL_lex_brackets)
            PL_lex_dojoin = 2;
        PL_expect = XOPERATOR;
        PL_bufptr++;
    }
    return funny;
}

void
Perl_yyunlex(pTHX)
{
    int yyc = PL_parser->yychar;
    if (yyc != YYEMPTY) {
        if (yyc) {
            NEXTVAL_NEXTTOKE = PL_parser->yylval;
            if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
                PL_lex_allbrackets--;
                PL_lex_brackets--;
                yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
            } else if (yyc == '('/*)*/) {
                PL_lex_allbrackets--;
                yyc |= (2<<24);
            }
            force_next(yyc);
        }
        PL_parser->yychar = YYEMPTY;
    }
}

STATIC SV *
S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
{
    SV * const sv = newSVpvn_utf8(start, len,
                          !IN_BYTES
                          && UTF
                          && !is_utf8_invariant_string((const U8*)start, len)
                          && is_utf8_string((const U8*)start, len));
    return sv;
}

/*
 * S_force_word
 * When the lexer knows the next thing is a word (for instance, it has
 * just seen -> and it knows that the next char is a word char, then
 * it calls S_force_word to stick the next word into the PL_nexttoke/val
 * lookahead.
 *
 * Arguments:
 *   char *start : buffer position (must be within PL_linestr)
 *   int token   : PL_next* will be this type of bare word
 *                 (e.g., METHOD,BAREWORD)
 *   int check_keyword : if true, Perl checks to make sure the word isn't
 *       a keyword (do this if the word is a label, e.g. goto FOO)
 *   int allow_pack : if true, : characters will also be allowed (require,
 *       use, etc. do this)
 */

STATIC char *
S_force_word(pTHX_ char *start, int token, int check_keyword, int allow_pack)
{
    char *s;
    STRLEN len;

    PERL_ARGS_ASSERT_FORCE_WORD;

    start = skipspace(start);
    s = start;
    if (   isIDFIRST_lazy_if_safe(s, PL_bufend, UTF)
        || (allow_pack && *s == ':' && s[1] == ':') )
    {
        s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
        if (check_keyword) {
          char *s2 = PL_tokenbuf;
          STRLEN len2 = len;
          if (allow_pack && len > 6 && strEQs(s2, "CORE::"))
            s2 += 6, len2 -= 6;
          if (keyword(s2, len2, 0))
            return start;
        }
        if (token == METHOD) {
            s = skipspace(s);
            if (*s == '(')
                PL_expect = XTERM;
            else {
                PL_expect = XOPERATOR;
            }
        }
        NEXTVAL_NEXTTOKE.opval
            = newSVOP(OP_CONST,0,
                           S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
        NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
        force_next(token);
    }
    return s;
}

/*
 * S_force_ident
 * Called when the lexer wants $foo *foo &foo etc, but the program
 * text only contains the "foo" portion.  The first argument is a pointer
 * to the "foo", and the second argument is the type symbol to prefix.
 * Forces the next token to be a "BAREWORD".
 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
 */

STATIC void
S_force_ident(pTHX_ const char *s, int kind)
{
    PERL_ARGS_ASSERT_FORCE_IDENT;

    if (s[0]) {
        const STRLEN len = s[1] ? strlen(s) : 1; /* s = "\"" see yylex */
        OP* const o = newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
                                                                UTF ? SVf_UTF8 : 0));
        NEXTVAL_NEXTTOKE.opval = o;
        force_next(BAREWORD);
        if (kind) {
            o->op_private = OPpCONST_ENTERED;
            /* XXX see note in pp_entereval() for why we forgo typo
               warnings if the symbol must be introduced in an eval.
               GSAR 96-10-12 */
            gv_fetchpvn_flags(s, len,
                              (PL_in_eval ? GV_ADDMULTI
                              : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
                              kind == '$' ? SVt_PV :
                              kind == '@' ? SVt_PVAV :
                              kind == '%' ? SVt_PVHV :
                              SVt_PVGV
                              );
        }
    }
}

static void
S_force_ident_maybe_lex(pTHX_ char pit)
{
    NEXTVAL_NEXTTOKE.ival = pit;
    force_next('p');
}

NV
Perl_str_to_version(pTHX_ SV *sv)
{
    NV retval = 0.0;
    NV nshift = 1.0;
    STRLEN len;
    const char *start = SvPV_const(sv,len);
    const char * const end = start + len;
    const bool utf = cBOOL(SvUTF8(sv));

    PERL_ARGS_ASSERT_STR_TO_VERSION;

    while (start < end) {
        STRLEN skip;
        UV n;
        if (utf)
            n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
        else {
            n = *(U8*)start;
            skip = 1;
        }
        retval += ((NV)n)/nshift;
        start += skip;
        nshift *= 1000;
    }
    return retval;
}

/*
 * S_force_version
 * Forces the next token to be a version number.
 * If the next token appears to be an invalid version number, (e.g. "v2b"),
 * and if "guessing" is TRUE, then no new token is created (and the caller
 * must use an alternative parsing method).
 */

STATIC char *
S_force_version(pTHX_ char *s, int guessing)
{
    OP *version = NULL;
    char *d;

    PERL_ARGS_ASSERT_FORCE_VERSION;

    s = skipspace(s);

    d = s;
    if (*d == 'v')
        d++;
    if (isDIGIT(*d)) {
        while (isDIGIT(*d) || *d == '_' || *d == '.')
            d++;
        if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
            SV *ver;
            s = scan_num(s, &pl_yylval);
            version = pl_yylval.opval;
            ver = cSVOPx(version)->op_sv;
            if (SvPOK(ver) && !SvNIOK(ver)) {
                SvUPGRADE(ver, SVt_PVNV);
                SvNV_set(ver, str_to_version(ver));
                SvNOK_on(ver);          /* hint that it is a version */
            }
        }
        else if (guessing) {
            return s;
        }
    }

    /* NOTE: The parser sees the package name and the VERSION swapped */
    NEXTVAL_NEXTTOKE.opval = version;
    force_next(BAREWORD);

    return s;
}

/*
 * S_force_strict_version
 * Forces the next token to be a version number using strict syntax rules.
 */

STATIC char *
S_force_strict_version(pTHX_ char *s)
{
    OP *version = NULL;
    const char *errstr = NULL;

    PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;

    while (isSPACE(*s)) /* leading whitespace */
        s++;

    if (is_STRICT_VERSION(s,&errstr)) {
        SV *ver = newSV(0);
        s = (char *)scan_version(s, ver, 0);
        version = newSVOP(OP_CONST, 0, ver);
    }
    else if ((*s != ';' && *s != '{' && *s != '}' )
             && (s = skipspace(s), (*s != ';' && *s != '{' && *s != '}' )))
    {
        PL_bufptr = s;
        if (errstr)
            yyerror(errstr); /* version required */
        return s;
    }

    /* NOTE: The parser sees the package name and the VERSION swapped */
    NEXTVAL_NEXTTOKE.opval = version;
    force_next(BAREWORD);

    return s;
}

/*
 * S_tokeq
 * Tokenize a quoted string passed in as an SV.  It finds the next
 * chunk, up to end of string or a backslash.  It may make a new
 * SV containing that chunk (if HINT_NEW_STRING is on).  It also
 * turns \\ into \.
 */

STATIC SV *
S_tokeq(pTHX_ SV *sv)
{
    char *s;
    char *send;
    char *d;
    SV *pv = sv;

    PERL_ARGS_ASSERT_TOKEQ;

    assert (SvPOK(sv));
    assert (SvLEN(sv));
    assert (!SvIsCOW(sv));
    if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1) /* <<'heredoc' */
        goto finish;
    s = SvPVX(sv);
    send = SvEND(sv);
    /* This is relying on the SV being "well formed" with a trailing '\0'  */
    while (s < send && !(*s == '\\' && s[1] == '\\'))
        s++;
    if (s == send)
        goto finish;
    d = s;
    if ( PL_hints & HINT_NEW_STRING ) {
        pv = newSVpvn_flags(SvPVX_const(pv), SvCUR(sv),
                            SVs_TEMP | SvUTF8(sv));
    }
    while (s < send) {
        if (*s == '\\') {
            if (s + 1 < send && (s[1] == '\\'))
                s++;            /* all that, just for this */
        }
        *d++ = *s++;
    }
    *d = '\0';
    SvCUR_set(sv, d - SvPVX_const(sv));
  finish:
    if ( PL_hints & HINT_NEW_STRING )
       return new_constant(NULL, 0, "q", sv, pv, "q", 1);
    return sv;
}

/*
 * Now come three functions related to double-quote context,
 * S_sublex_start, S_sublex_push, and S_sublex_done.  They're used when
 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")).  They
 * interact with PL_lex_state, and create fake ( ... ) argument lists
 * to handle functions and concatenation.
 * For example,
 *   "foo\lbar"
 * is tokenised as
 *    stringify ( const[foo] concat lcfirst ( const[bar] ) )
 */

/*
 * S_sublex_start
 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
 *
 * Pattern matching will set PL_lex_op to the pattern-matching op to
 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
 *
 * OP_CONST is easy--just make the new op and return.
 *
 * Everything else becomes a FUNC.
 *
 * Sets PL_lex_state to LEX_INTERPPUSH unless ival was OP_NULL or we
 * had an OP_CONST.  This just sets us up for a
 * call to S_sublex_push().
 */

STATIC I32
S_sublex_start(pTHX)
{
    const I32 op_type = pl_yylval.ival;

    if (op_type == OP_NULL) {
        pl_yylval.opval = PL_lex_op;
        PL_lex_op = NULL;
        return THING;
    }
    if (op_type == OP_CONST) {
        SV *sv = PL_lex_stuff;
        PL_lex_stuff = NULL;
        sv = tokeq(sv);

        if (SvTYPE(sv) == SVt_PVIV) {
            /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
            STRLEN len;
            const char * const p = SvPV_const(sv, len);
            SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
            SvREFCNT_dec(sv);
            sv = nsv;
        }
        pl_yylval.opval = newSVOP(op_type, 0, sv);
        return THING;
    }

    PL_parser->lex_super_state = PL_lex_state;
    PL_parser->lex_sub_inwhat = (U16)op_type;
    PL_parser->lex_sub_op = PL_lex_op;
    PL_lex_state = LEX_INTERPPUSH;

    PL_expect = XTERM;
    if (PL_lex_op) {
        pl_yylval.opval = PL_lex_op;
        PL_lex_op = NULL;
        return PMFUNC;
    }
    else
        return FUNC;
}

/*
 * S_sublex_push
 * Create a new scope to save the lexing state.  The scope will be
 * ended in S_sublex_done.  Returns a '(', starting the function arguments
 * to the uc, lc, etc. found before.
 * Sets PL_lex_state to LEX_INTERPCONCAT.
 */

STATIC I32
S_sublex_push(pTHX)
{
    LEXSHARED *shared;
    const bool is_heredoc = PL_multi_close == '<';
    ENTER;

    PL_lex_state = PL_parser->lex_super_state;
    SAVEI8(PL_lex_dojoin);
    SAVEI32(PL_lex_brackets);
    SAVEI32(PL_lex_allbrackets);
    SAVEI32(PL_lex_formbrack);
    SAVEI8(PL_lex_fakeeof);
    SAVEI32(PL_lex_casemods);
    SAVEI32(PL_lex_starts);
    SAVEI8(PL_lex_state);
    SAVESPTR(PL_lex_repl);
    SAVEVPTR(PL_lex_inpat);
    SAVEI16(PL_lex_inwhat);
    if (is_heredoc)
    {
        SAVECOPLINE(PL_curcop);
        SAVEI32(PL_multi_end);
        SAVEI32(PL_parser->herelines);
        PL_parser->herelines = 0;
    }
    SAVEIV(PL_multi_close);
    SAVEPPTR(PL_bufptr);
    SAVEPPTR(PL_bufend);
    SAVEPPTR(PL_oldbufptr);
    SAVEPPTR(PL_oldoldbufptr);
    SAVEPPTR(PL_last_lop);
    SAVEPPTR(PL_last_uni);
    SAVEPPTR(PL_linestart);
    SAVESPTR(PL_linestr);
    SAVEGENERICPV(PL_lex_brackstack);
    SAVEGENERICPV(PL_lex_casestack);
    SAVEGENERICPV(PL_parser->lex_shared);
    SAVEBOOL(PL_parser->lex_re_reparsing);
    SAVEI32(PL_copline);

    /* The here-doc parser needs to be able to peek into outer lexing
       scopes to find the body of the here-doc.  So we put PL_linestr and
       PL_bufptr into lex_shared, to ‘share’ those values.
     */
    PL_parser->lex_shared->ls_linestr = PL_linestr;
    PL_parser->lex_shared->ls_bufptr  = PL_bufptr;

    PL_linestr = PL_lex_stuff;
    PL_lex_repl = PL_parser->lex_sub_repl;
    PL_lex_stuff = NULL;
    PL_parser->lex_sub_repl = NULL;

    /* Arrange for PL_lex_stuff to be freed on scope exit, in case it gets
       set for an inner quote-like operator and then an error causes scope-
       popping.  We must not have a PL_lex_stuff value left dangling, as
       that breaks assumptions elsewhere.  See bug #123617.  */
    SAVEGENERICSV(PL_lex_stuff);
    SAVEGENERICSV(PL_parser->lex_sub_repl);

    PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
        = SvPVX(PL_linestr);
    PL_bufend += SvCUR(PL_linestr);
    PL_last_lop = PL_last_uni = NULL;
    SAVEFREESV(PL_linestr);
    if (PL_lex_repl) SAVEFREESV(PL_lex_repl);

    PL_lex_dojoin = FALSE;
    PL_lex_brackets = PL_lex_formbrack = 0;
    PL_lex_allbrackets = 0;
    PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
    Newx(PL_lex_brackstack, 120, char);
    Newx(PL_lex_casestack, 12, char);
    PL_lex_casemods = 0;
    *PL_lex_casestack = '\0';
    PL_lex_starts = 0;
    PL_lex_state = LEX_INTERPCONCAT;
    if (is_heredoc)
        CopLINE_set(PL_curcop, (line_t)PL_multi_start);
    PL_copline = NOLINE;

    Newxz(shared, 1, LEXSHARED);
    shared->ls_prev = PL_parser->lex_shared;
    PL_parser->lex_shared = shared;

    PL_lex_inwhat = PL_parser->lex_sub_inwhat;
    if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
    if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
        PL_lex_inpat = PL_parser->lex_sub_op;
    else
        PL_lex_inpat = NULL;

    PL_parser->lex_re_reparsing = cBOOL(PL_in_eval & EVAL_RE_REPARSING);
    PL_in_eval &= ~EVAL_RE_REPARSING;

    return '(';
}

/*
 * S_sublex_done
 * Restores lexer state after a S_sublex_push.
 */

STATIC I32
S_sublex_done(pTHX)
{
    if (!PL_lex_starts++) {
        SV * const sv = newSVpvs("");
        if (SvUTF8(PL_linestr))
            SvUTF8_on(sv);
        PL_expect = XOPERATOR;
        pl_yylval.opval = newSVOP(OP_CONST, 0, sv);
        return THING;
    }

    if (PL_lex_casemods) {              /* oops, we've got some unbalanced parens */
        PL_lex_state = LEX_INTERPCASEMOD;
        return yylex();
    }

    /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
    assert(PL_lex_inwhat != OP_TRANSR);
    if (PL_lex_repl) {
        assert (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS);
        PL_linestr = PL_lex_repl;
        PL_lex_inpat = 0;
        PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
        PL_bufend += SvCUR(PL_linestr);
        PL_last_lop = PL_last_uni = NULL;
        PL_lex_dojoin = FALSE;
        PL_lex_brackets = 0;
        PL_lex_allbrackets = 0;
        PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
        PL_lex_casemods = 0;
        *PL_lex_casestack = '\0';
        PL_lex_starts = 0;
        if (SvEVALED(PL_lex_repl)) {
            PL_lex_state = LEX_INTERPNORMAL;
            PL_lex_starts++;
            /*  we don't clear PL_lex_repl here, so that we can check later
                whether this is an evalled subst; that means we rely on the
                logic to ensure sublex_done() is called again only via the
                branch (in yylex()) that clears PL_lex_repl, else we'll loop */
        }
        else {
            PL_lex_state = LEX_INTERPCONCAT;
            PL_lex_repl = NULL;
        }
        if (SvTYPE(PL_linestr) >= SVt_PVNV) {
            CopLINE(PL_curcop) +=
                ((XPVNV*)SvANY(PL_linestr))->xnv_u.xnv_lines
                 + PL_parser->herelines;
            PL_parser->herelines = 0;
        }
        return '/';
    }
    else {
        const line_t l = CopLINE(PL_curcop);
        LEAVE;
        if (PL_multi_close == '<')
            PL_parser->herelines += l - PL_multi_end;
        PL_bufend = SvPVX(PL_linestr);
        PL_bufend += SvCUR(PL_linestr);
        PL_expect = XOPERATOR;
        return ')';
    }
}

STATIC SV*
S_get_and_check_backslash_N_name(pTHX_ const char* s, const char* const e)
{
    /* <s> points to first character of interior of \N{}, <e> to one beyond the
     * interior, hence to the "}".  Finds what the name resolves to, returning
     * an SV* containing it; NULL if no valid one found */

    SV* res = newSVpvn_flags(s, e - s, UTF ? SVf_UTF8 : 0);

    HV * table;
    SV **cvp;
    SV *cv;
    SV *rv;
    HV *stash;
    const U8* first_bad_char_loc;
    const char* backslash_ptr = s - 3; /* Points to the <\> of \N{... */

    PERL_ARGS_ASSERT_GET_AND_CHECK_BACKSLASH_N_NAME;

    if (!SvCUR(res)) {
        deprecate_fatal_in("5.28", "Unknown charname '' is deprecated");
        return res;
    }

    if (UTF && ! is_utf8_string_loc((U8 *) backslash_ptr,
                                     e - backslash_ptr,
                                     &first_bad_char_loc))
    {
        _force_out_malformed_utf8_message(first_bad_char_loc,
                                          (U8 *) PL_parser->bufend,
                                          0,
                                          0 /* 0 means don't die */ );
        yyerror_pv(Perl_form(aTHX_
            "Malformed UTF-8 character immediately after '%.*s'",
            (int) (first_bad_char_loc - (U8 *) backslash_ptr), backslash_ptr),
                   SVf_UTF8);
        return NULL;
    }

    res = new_constant( NULL, 0, "charnames", res, NULL, backslash_ptr,
                        /* include the <}> */
                        e - backslash_ptr + 1);
    if (! SvPOK(res)) {
        SvREFCNT_dec_NN(res);
        return NULL;
    }

    /* See if the charnames handler is the Perl core's, and if so, we can skip
     * the validation needed for a user-supplied one, as Perl's does its own
     * validation. */
    table = GvHV(PL_hintgv);             /* ^H */
    cvp = hv_fetchs(table, "charnames", FALSE);
    if (cvp && (cv = *cvp) && SvROK(cv) && (rv = SvRV(cv),
        SvTYPE(rv) == SVt_PVCV) && ((stash = CvSTASH(rv)) != NULL))
    {
        const char * const name = HvNAME(stash);
        if (HvNAMELEN(stash) == sizeof("_charnames")-1
         && strEQ(name, "_charnames")) {
           return res;
       }
    }

    /* Here, it isn't Perl's charname handler.  We can't rely on a
     * user-supplied handler to validate the input name.  For non-ut8 input,
     * look to see that the first character is legal.  Then loop through the
     * rest checking that each is a continuation */

    /* This code makes the reasonable assumption that the only Latin1-range
     * characters that begin a character name alias are alphabetic, otherwise
     * would have to create a isCHARNAME_BEGIN macro */

    if (! UTF) {
        if (! isALPHAU(*s)) {
            goto bad_charname;
        }
        s++;
        while (s < e) {
            if (! isCHARNAME_CONT(*s)) {
                goto bad_charname;
            }
            if (*s == ' ' && *(s-1) == ' ') {
                goto multi_spaces;
            }
            s++;
        }
    }
    else {
        /* Similarly for utf8.  For invariants can check directly; for other
         * Latin1, can calculate their code point and check; otherwise  use a
         * swash */
        if (UTF8_IS_INVARIANT(*s)) {
            if (! isALPHAU(*s)) {
                goto bad_charname;
            }
            s++;
        } else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
            if (! isALPHAU(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s+1)))) {
                goto bad_charname;
            }
            s += 2;
        }
        else {
            if (! PL_utf8_charname_begin) {
                U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
                PL_utf8_charname_begin = _core_swash_init("utf8",
                                                        "_Perl_Charname_Begin",
                                                        &PL_sv_undef,
                                                        1, 0, NULL, &flags);
            }
            if (! swash_fetch(PL_utf8_charname_begin, (U8 *) s, TRUE)) {
                goto bad_charname;
            }
            s += UTF8SKIP(s);
        }

        while (s < e) {
            if (UTF8_IS_INVARIANT(*s)) {
                if (! isCHARNAME_CONT(*s)) {
                    goto bad_charname;
                }
                if (*s == ' ' && *(s-1) == ' ') {
                    goto multi_spaces;
                }
                s++;
            }
            else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
                if (! isCHARNAME_CONT(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s+1))))
                {
                    goto bad_charname;
                }
                s += 2;
            }
            else {
                if (! PL_utf8_charname_continue) {
                    U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
                    PL_utf8_charname_continue = _core_swash_init("utf8",
                                                "_Perl_Charname_Continue",
                                                &PL_sv_undef,
                                                1, 0, NULL, &flags);
                }
                if (! swash_fetch(PL_utf8_charname_continue, (U8 *) s, TRUE)) {
                    goto bad_charname;
                }
                s += UTF8SKIP(s);
            }
        }
    }
    if (*(s-1) == ' ') {
        yyerror_pv(
            Perl_form(aTHX_
            "charnames alias definitions may not contain trailing "
            "white-space; marked by <-- HERE in %.*s<-- HERE %.*s",
            (int)(s - backslash_ptr + 1), backslash_ptr,
            (int)(e - s + 1), s + 1
            ),
        UTF ? SVf_UTF8 : 0);
        return NULL;
    }

    if (SvUTF8(res)) { /* Don't accept malformed input */
        const U8* first_bad_char_loc;
        STRLEN len;
        const char* const str = SvPV_const(res, len);
        if (! is_utf8_string_loc((U8 *) str, len, &first_bad_char_loc)) {
            _force_out_malformed_utf8_message(first_bad_char_loc,
                                              (U8 *) PL_parser->bufend,
                                              0,
                                              0 /* 0 means don't die */ );
            yyerror_pv(
              Perl_form(aTHX_
                "Malformed UTF-8 returned by %.*s immediately after '%.*s'",
                 (int) (e - backslash_ptr + 1), backslash_ptr,
                 (int) ((char *) first_bad_char_loc - str), str
              ),
              SVf_UTF8);
            return NULL;
        }
    }

    return res;

  bad_charname: {

        /* The final %.*s makes sure that should the trailing NUL be missing
         * that this print won't run off the end of the string */
        yyerror_pv(
          Perl_form(aTHX_
            "Invalid character in \\N{...}; marked by <-- HERE in %.*s<-- HERE %.*s",
            (int)(s - backslash_ptr + 1), backslash_ptr,
            (int)(e - s + 1), s + 1
          ),
          UTF ? SVf_UTF8 : 0);
        return NULL;
    }

  multi_spaces:
        yyerror_pv(
          Perl_form(aTHX_
            "charnames alias definitions may not contain a sequence of "
            "multiple spaces; marked by <-- HERE in %.*s<-- HERE %.*s",
            (int)(s - backslash_ptr + 1), backslash_ptr,
            (int)(e - s + 1), s + 1
          ),
          UTF ? SVf_UTF8 : 0);
        return NULL;
}

/*
  scan_const

  Extracts the next constant part of a pattern, double-quoted string,
  or transliteration.  This is terrifying code.

  For example, in parsing the double-quoted string "ab\x63$d", it would
  stop at the '$' and return an OP_CONST containing 'abc'.

  It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
  processing a pattern (PL_lex_inpat is true), a transliteration
  (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.

  Returns a pointer to the character scanned up to. If this is
  advanced from the start pointer supplied (i.e. if anything was
  successfully parsed), will leave an OP_CONST for the substring scanned
  in pl_yylval. Caller must intuit reason for not parsing further
  by looking at the next characters herself.

  In patterns:
    expand:
      \N{FOO}  => \N{U+hex_for_character_FOO}
      (if FOO expands to multiple characters, expands to \N{U+xx.XX.yy ...})

    pass through:
        all other \-char, including \N and \N{ apart from \N{ABC}

    stops on:
        @ and $ where it appears to be a var, but not for $ as tail anchor
        \l \L \u \U \Q \E
        (?{  or  (??{

  In transliterations:
    characters are VERY literal, except for - not at the start or end
    of the string, which indicates a range.  However some backslash sequences
    are recognized: \r, \n, and the like
                    \007 \o{}, \x{}, \N{}
    If all elements in the transliteration are below 256,
    scan_const expands the range to the full set of intermediate
    characters. If the range is in utf8, the hyphen is replaced with
    a certain range mark which will be handled by pmtrans() in op.c.

  In double-quoted strings:
    backslashes:
      all those recognized in transliterations
      deprecated backrefs: \1 (in substitution replacements)
      case and quoting: \U \Q \E
    stops on @ and $

  scan_const does *not* construct ops to handle interpolated strings.
  It stops processing as soon as it finds an embedded $ or @ variable
  and leaves it to the caller to work out what's going on.

  embedded arrays (whether in pattern or not) could be:
      @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.

  $ in double-quoted strings must be the symbol of an embedded scalar.

  $ in pattern could be $foo or could be tail anchor.  Assumption:
  it's a tail anchor if $ is the last thing in the string, or if it's
  followed by one of "()| \r\n\t"

  \1 (backreferences) are turned into $1 in substitutions

  The structure of the code is
      while (there's a character to process) {
          handle transliteration ranges
          skip regexp comments /(?#comment)/ and codes /(?{code})/
          skip #-initiated comments in //x patterns
          check for embedded arrays
          check for embedded scalars
          if (backslash) {
              deprecate \1 in substitution replacements
              handle string-changing backslashes \l \U \Q \E, etc.
              switch (what was escaped) {
                  handle \- in a transliteration (becomes a literal -)
                  if a pattern and not \N{, go treat as regular character
                  handle \132 (octal characters)
                  handle \x15 and \x{1234} (hex characters)
                  handle \N{name} (named characters, also \N{3,5} in a pattern)
                  handle \cV (control characters)
                  handle printf-style backslashes (\f, \r, \n, etc)
              } (end switch)
              continue
          } (end if backslash)
          handle regular character
    } (end while character to read)

*/

STATIC char *
S_scan_const(pTHX_ char *start)
{
    char *send = PL_bufend;             /* end of the constant */
    SV *sv = newSV(send - start);       /* sv for the constant.  See note below
                                           on sizing. */
    char *s = start;                    /* start of the constant */
    char *d = SvPVX(sv);                /* destination for copies */
    bool dorange = FALSE;               /* are we in a translit range? */
    bool didrange = FALSE;              /* did we just finish a range? */
    bool in_charclass = FALSE;          /* within /[...]/ */
    bool has_utf8 = FALSE;              /* Output constant is UTF8 */
    bool  this_utf8 = cBOOL(UTF);       /* Is the source string assumed to be
                                           UTF8?  But, this can show as true
                                           when the source isn't utf8, as for
                                           example when it is entirely composed
                                           of hex constants */
    STRLEN utf8_variant_count = 0;      /* When not in UTF-8, this counts the
                                           number of characters found so far
                                           that will expand (into 2 bytes)
                                           should we have to convert to
                                           UTF-8) */
    SV *res;                            /* result from charnames */
    STRLEN offset_to_max;   /* The offset in the output to where the range
                               high-end character is temporarily placed */

    /* Does something require special handling in tr/// ?  This avoids extra
     * work in a less likely case.  As such, khw didn't feel it was worth
     * adding any branches to the more mainline code to handle this, which
     * means that this doesn't get set in some circumstances when things like
     * \x{100} get expanded out.  As a result there needs to be extra testing
     * done in the tr code */
    bool has_above_latin1 = FALSE;

    /* Note on sizing:  The scanned constant is placed into sv, which is
     * initialized by newSV() assuming one byte of output for every byte of
     * input.  This routine expects newSV() to allocate an extra byte for a
     * trailing NUL, which this routine will append if it gets to the end of
     * the input.  There may be more bytes of input than output (eg., \N{LATIN
     * CAPITAL LETTER A}), or more output than input if the constant ends up
     * recoded to utf8, but each time a construct is found that might increase
     * the needed size, SvGROW() is called.  Its size parameter each time is
     * based on the best guess estimate at the time, namely the length used so
     * far, plus the length the current construct will occupy, plus room for
     * the trailing NUL, plus one byte for every input byte still unscanned */

    UV uv = UV_MAX; /* Initialize to weird value to try to catch any uses
                       before set */
#ifdef EBCDIC
    int backslash_N = 0;            /* ? was the character from \N{} */
    int non_portable_endpoint = 0;  /* ? In a range is an endpoint
                                       platform-specific like \x65 */
#endif

    PERL_ARGS_ASSERT_SCAN_CONST;

    assert(PL_lex_inwhat != OP_TRANSR);
    if (PL_lex_inwhat == OP_TRANS && PL_parser->lex_sub_op) {
        /* If we are doing a trans and we know we want UTF8 set expectation */
        has_utf8   = PL_parser->lex_sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
        this_utf8  = PL_parser->lex_sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
    }

    /* Protect sv from errors and fatal warnings. */
    ENTER_with_name("scan_const");
    SAVEFREESV(sv);

    while (s < send
           || dorange   /* Handle tr/// range at right edge of input */
    ) {

        /* get transliterations out of the way (they're most literal) */
        if (PL_lex_inwhat == OP_TRANS) {

            /* But there isn't any special handling necessary unless there is a
             * range, so for most cases we just drop down and handle the value
             * as any other.  There are two exceptions.
             *
             * 1.  A hyphen indicates that we are actually going to have a
             *     range.  In this case, skip the '-', set a flag, then drop
             *     down to handle what should be the end range value.
             * 2.  After we've handled that value, the next time through, that
             *     flag is set and we fix up the range.
             *
             * Ranges entirely within Latin1 are expanded out entirely, in
             * order to make the transliteration a simple table look-up.
             * Ranges that extend above Latin1 have to be done differently, so
             * there is no advantage to expanding them here, so they are
             * stored here as Min, ILLEGAL_UTF8_BYTE, Max.  The illegal byte
             * signifies a hyphen without any possible ambiguity.  On EBCDIC
             * machines, if the range is expressed as Unicode, the Latin1
             * portion is expanded out even if the range extends above
             * Latin1.  This is because each code point in it has to be
             * processed here individually to get its native translation */

            if (! dorange) {

                /* Here, we don't think we're in a range.  If the new character
                 * is not a hyphen; or if it is a hyphen, but it's too close to
                 * either edge to indicate a range, then it's a regular
                 * character. */
                if (*s != '-' || s >= send - 1 || s == start) {

                    /* A regular character.  Process like any other, but first
                     * clear any flags */
                    didrange = FALSE;
                    dorange = FALSE;
#ifdef EBCDIC
                    non_portable_endpoint = 0;
                    backslash_N = 0;
#endif
                    /* The tests here for being above Latin1 and similar ones
                     * in the following 'else' suffice to find all such
                     * occurences in the constant, except those added by a
                     * backslash escape sequence, like \x{100}.  Mostly, those
                     * set 'has_above_latin1' as appropriate */
                    if (this_utf8 && UTF8_IS_ABOVE_LATIN1(*s)) {
                        has_above_latin1 = TRUE;
                    }

                    /* Drops down to generic code to process current byte */
                }
                else {  /* Is a '-' in the context where it means a range */
                    if (didrange) { /* Something like y/A-C-Z// */
                        Perl_croak(aTHX_ "Ambiguous range in transliteration"
                                         " operator");
                    }

                    dorange = TRUE;

                    s++;    /* Skip past the hyphen */

                    /* d now points to where the end-range character will be
                     * placed.  Save it so won't have to go finding it later,
                     * and drop down to get that character.  (Actually we
                     * instead save the offset, to handle the case where a
                     * realloc in the meantime could change the actual
                     * pointer).  We'll finish processing the range the next
                     * time through the loop */
                    offset_to_max = d - SvPVX_const(sv);

                    if (this_utf8 && UTF8_IS_ABOVE_LATIN1(*s)) {
                        has_above_latin1 = TRUE;
                    }

                    /* Drops down to generic code to process current byte */
                }
            }  /* End of not a range */
            else {
                /* Here we have parsed a range.  Now must handle it.  At this
                 * point:
                 * 'sv' is a SV* that contains the output string we are
                 *      constructing.  The final two characters in that string
                 *      are the range start and range end, in order.
                 * 'd'  points to just beyond the range end in the 'sv' string,
                 *      where we would next place something
                 * 'offset_to_max' is the offset in 'sv' at which the character
                 *      (the range's maximum end point) before 'd'  begins.
                 */
                char * max_ptr = SvPVX(sv) + offset_to_max;
                char * min_ptr;
                IV range_min;
                IV range_max;   /* last character in range */
                STRLEN grow;
                Size_t offset_to_min = 0;
                Size_t extras = 0;
#ifdef EBCDIC
                bool convert_unicode;
                IV real_range_max = 0;
#endif
                /* Get the code point values of the range ends. */
                if (has_utf8) {
                    /* We know the utf8 is valid, because we just constructed
                     * it ourselves in previous loop iterations */
                    min_ptr = (char*) utf8_hop( (U8*) max_ptr, -1);
                    range_min = valid_utf8_to_uvchr( (U8*) min_ptr, NULL);
                    range_max = valid_utf8_to_uvchr( (U8*) max_ptr, NULL);

                    /* This compensates for not all code setting
                     * 'has_above_latin1', so that we don't skip stuff that
                     * should be executed */
                    if (range_max > 255) {
                        has_above_latin1 = TRUE;
                    }
                }
                else {
                    min_ptr = max_ptr - 1;
                    range_min = * (U8*) min_ptr;
                    range_max = * (U8*) max_ptr;
                }

                /* If the range is just a single code point, like tr/a-a/.../,
                 * that code point is already in the output, twice.  We can
                 * just back up over the second instance and avoid all the rest
                 * of the work.  But if it is a variant character, it's been
                 * counted twice, so decrement.  (This unlikely scenario is
                 * special cased, like the one for a range of 2 code points
                 * below, only because the main-line code below needs a range
                 * of 3 or more to work without special casing.  Might as well
                 * get it out of the way now.) */
                if (UNLIKELY(range_max == range_min)) {
                    d = max_ptr;
                    if (! has_utf8 && ! UVCHR_IS_INVARIANT(range_max)) {
                        utf8_variant_count--;
                    }
                    goto range_done;
                }

#ifdef EBCDIC
                /* On EBCDIC platforms, we may have to deal with portable
                 * ranges.  These happen if at least one range endpoint is a
                 * Unicode value (\N{...}), or if the range is a subset of
                 * [A-Z] or [a-z], and both ends are literal characters,
                 * like 'A', and not like \x{C1} */
                convert_unicode =
                               cBOOL(backslash_N)   /* \N{} forces Unicode,
                                                       hence portable range */
                    || (     ! non_portable_endpoint
                        && ((  isLOWER_A(range_min) && isLOWER_A(range_max))
                           || (isUPPER_A(range_min) && isUPPER_A(range_max))));
                if (convert_unicode) {

                    /* Special handling is needed for these portable ranges.
                     * They are defined to be in Unicode terms, which includes
                     * all the Unicode code points between the end points.
                     * Convert to Unicode to get the Unicode range.  Later we
                     * will convert each code point in the range back to
                     * native.  */
                    range_min = NATIVE_TO_UNI(range_min);
                    range_max = NATIVE_TO_UNI(range_max);
                }
#endif

                if (range_min > range_max) {
#ifdef EBCDIC
                    if (convert_unicode) {
                        /* Need to convert back to native for meaningful
                         * messages for this platform */
                        range_min = UNI_TO_NATIVE(range_min);
                        range_max = UNI_TO_NATIVE(range_max);
                    }
#endif
                    /* Use the characters themselves for the error message if
                     * ASCII printables; otherwise some visible representation
                     * of them */
                    if (isPRINT_A(range_min) && isPRINT_A(range_max)) {
                        Perl_croak(aTHX_
                         "Invalid range \"%c-%c\" in transliteration operator",
                         (char)range_min, (char)range_max);
                    }
#ifdef EBCDIC
                    else if (convert_unicode) {
        /* diag_listed_as: Invalid range "%s" in transliteration operator */
                        Perl_croak(aTHX_
                           "Invalid range \"\\N{U+%04" UVXf "}-\\N{U+%04"
                           UVXf "}\" in transliteration operator",
                           range_min, range_max);
                    }
#endif
                    else {
        /* diag_listed_as: Invalid range "%s" in transliteration operator */
                        Perl_croak(aTHX_
                           "Invalid range \"\\x{%04" UVXf "}-\\x{%04" UVXf "}\""
                           " in transliteration operator",
                           range_min, range_max);
                    }
                }

                /* If the range is exactly two code points long, they are
                 * already both in the output */
                if (UNLIKELY(range_min + 1 == range_max)) {
                    goto range_done;
                }

                /* Here the range contains at least 3 code points */

                if (has_utf8) {

                    /* If everything in the transliteration is below 256, we
                     * can avoid special handling later.  A translation table
                     * for each of those bytes is created by op.c.  So we
                     * expand out all ranges to their constituent code points.
                     * But if we've encountered something above 255, the
                     * expanding won't help, so skip doing that.  But if it's
                     * EBCDIC, we may have to look at each character below 256
                     * if we have to convert to/from Unicode values */
                    if (   has_above_latin1
#ifdef EBCDIC
                        && (range_min > 255 || ! convert_unicode)
#endif
                    ) {
                        /* Move the high character one byte to the right; then
                         * insert between it and the range begin, an illegal
                         * byte which serves to indicate this is a range (using
                         * a '-' would be ambiguous). */
                        char *e = d++;
                        while (e-- > max_ptr) {
                            *(e + 1) = *e;
                        }
                        *(e + 1) = (char) ILLEGAL_UTF8_BYTE;
                        goto range_done;
                    }

                    /* Here, we're going to expand out the range.  For EBCDIC
                     * the range can extend above 255 (not so in ASCII), so
                     * for EBCDIC, split it into the parts above and below
                     * 255/256 */
#ifdef EBCDIC
                    if (range_max > 255) {
                        real_range_max = range_max;
                        range_max = 255;
                    }
#endif
                }

                /* Here we need to expand out the string to contain each
                 * character in the range.  Grow the output to handle this.
                 * For non-UTF8, we need a byte for each code point in the
                 * range, minus the three that we've already allocated for: the
                 * hyphen, the min, and the max.  For UTF-8, we need this
                 * plus an extra byte for each code point that occupies two
                 * bytes (is variant) when in UTF-8 (except we've already
                 * allocated for the end points, including if they are
                 * variants).  For ASCII platforms and Unicode ranges on EBCDIC
                 * platforms, it's easy to calculate a precise number.  To
                 * start, we count the variants in the range, which we need
                 * elsewhere in this function anyway.  (For the case where it
                 * isn't easy to calculate, 'extras' has been initialized to 0,
                 * and the calculation is done in a loop further down.) */
#ifdef EBCDIC
                if (convert_unicode)
#endif
                {
                    /* This is executed unconditionally on ASCII, and for
                     * Unicode ranges on EBCDIC.  Under these conditions, all
                     * code points above a certain value are variant; and none
                     * under that value are.  We just need to find out how much
                     * of the range is above that value.  We don't count the
                     * end points here, as they will already have been counted
                     * as they were parsed. */
                    if (range_min >= UTF_CONTINUATION_MARK) {

                        /* The whole range is made up of variants */
                        extras = (range_max - 1) - (range_min + 1) + 1;
                    }
                    else if (range_max >= UTF_CONTINUATION_MARK) {

                        /* Only the higher portion of the range is variants */
                        extras = (range_max - 1) - UTF_CONTINUATION_MARK + 1;
                    }

                    utf8_variant_count += extras;
                }

                /* The base growth is the number of code points in the range,
                 * not including the endpoints, which have already been sized
                 * for (and output).  We don't subtract for the hyphen, as it
                 * has been parsed but not output, and the SvGROW below is
                 * based only on what's been output plus what's left to parse.
                 * */
                grow = (range_max - 1) - (range_min + 1) + 1;

                if (has_utf8) {
#ifdef EBCDIC
                    /* In some cases in EBCDIC, we haven't yet calculated a
                     * precise amount needed for the UTF-8 variants.  Just
                     * assume the worst case, that everything will expand by a
                     * byte */
                    if (! convert_unicode) {
                        grow *= 2;
                    }
                    else
#endif
                    {
                        /* Otherwise we know exactly how many variants there
                         * are in the range. */
                        grow += extras;
                    }
                }

                /* Grow, but position the output to overwrite the range min end
                 * point, because in some cases we overwrite that */
                SvCUR_set(sv, d - SvPVX_const(sv));
                offset_to_min = min_ptr - SvPVX_const(sv);

                /* See Note on sizing above. */
                d = offset_to_min + SvGROW(sv, SvCUR(sv)
                                             + (send - s)
                                             + grow
                                             + 1 /* Trailing NUL */ );

                /* Now, we can expand out the range. */
#ifdef EBCDIC
                if (convert_unicode) {
                    SSize_t i;

                    /* Recall that the min and max are now in Unicode terms, so
                     * we have to convert each character to its native
                     * equivalent */
                    if (has_utf8) {
                        for (i = range_min; i <= range_max; i++) {
                            append_utf8_from_native_byte(
                                                    LATIN1_TO_NATIVE((U8) i),
                                                    (U8 **) &d);
                        }
                    }
                    else {
                        for (i = range_min; i <= range_max; i++) {
                            *d++ = (char)LATIN1_TO_NATIVE((U8) i);
                        }
                    }
                }
                else
#endif
                /* Always gets run for ASCII, and sometimes for EBCDIC. */
                {
                    SSize_t i;

                    /* Here, no conversions are necessary, which means that the
                     * first character in the range is already in 'd' and
                     * valid, so we can skip overwriting it */
                    if (has_utf8) {
                        d += UTF8SKIP(d);
                        for (i = range_min + 1; i <= range_max; i++) {
                            append_utf8_from_native_byte((U8) i, (U8 **) &d);
                        }
                    }
                    else {
                        d++;
                        assert(range_min + 1 <= range_max);
                        for (i = range_min + 1; i < range_max; i++) {
#ifdef EBCDIC
                            /* In this case on EBCDIC, we haven't calculated
                             * the variants.  Do it here, as we go along */
                            if (! UVCHR_IS_INVARIANT(i)) {
                                utf8_variant_count++;
                            }
#endif
                            *d++ = (char)i;
                        }

                        /* The range_max is done outside the loop so as to
                         * avoid having to special case not incrementing
                         * 'utf8_variant_count' on EBCDIC (it's already been
                         * counted when originally parsed) */
                        *d++ = (char) range_max;
                    }
                }

#ifdef EBCDIC
                /* If the original range extended above 255, add in that
                 * portion. */
                if (real_range_max) {
                    *d++ = (char) UTF8_TWO_BYTE_HI(0x100);
                    *d++ = (char) UTF8_TWO_BYTE_LO(0x100);
                    if (real_range_max > 0x100) {
                        if (real_range_max > 0x101) {
                            *d++ = (char) ILLEGAL_UTF8_BYTE;
                        }
                        d = (char*)uvchr_to_utf8((U8*)d, real_range_max);
                    }
                }
#endif

              range_done:
                /* mark the range as done, and continue */
                didrange = TRUE;
                dorange = FALSE;
#ifdef EBCDIC
                non_portable_endpoint = 0;
                backslash_N = 0;
#endif
                continue;
            } /* End of is a range */
        } /* End of transliteration.  Joins main code after these else's */
        else if (*s == '[' && PL_lex_inpat && !in_charclass) {
            char *s1 = s-1;
            int esc = 0;
            while (s1 >= start && *s1-- == '\\')
                esc = !esc;
            if (!esc)
                in_charclass = TRUE;
        }
        else if (*s == ']' && PL_lex_inpat && in_charclass) {
            char *s1 = s-1;
            int esc = 0;
            while (s1 >= start && *s1-- == '\\')
                esc = !esc;
            if (!esc)
                in_charclass = FALSE;
        }
            /* skip for regexp comments /(?#comment)/, except for the last
             * char, which will be done separately.  Stop on (?{..}) and
             * friends */
        else if (*s == '(' && PL_lex_inpat && s[1] == '?' && !in_charclass) {
            if (s[2] == '#') {
                while (s+1 < send && *s != ')')
                    *d++ = *s++;
            }
            else if (!PL_lex_casemods
                     && (    s[2] == '{' /* This should match regcomp.c */
                         || (s[2] == '?' && s[3] == '{')))
            {
                break;
            }
        }
            /* likewise skip #-initiated comments in //x patterns */
        else if (*s == '#'
                 && PL_lex_inpat
                 && !in_charclass
                 && ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED)
        {
            while (s < send && *s != '\n')
                *d++ = *s++;
        }
            /* no further processing of single-quoted regex */
        else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
            goto default_action;

            /* check for embedded arrays
             * (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
             */
        else if (*s == '@' && s[1]) {
            if (UTF
               ? isIDFIRST_utf8_safe(s+1, send)
               : isWORDCHAR_A(s[1]))
            {
                break;
            }
            if (strchr(":'{$", s[1]))
                break;
            if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
                break; /* in regexp, neither @+ nor @- are interpolated */
        }
            /* check for embedded scalars.  only stop if we're sure it's a
             * variable.  */
        else if (*s == '$') {
            if (!PL_lex_inpat)  /* not a regexp, so $ must be var */
                break;
            if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
                if (s[1] == '\\') {
                    Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
                                   "Possible unintended interpolation of $\\ in regex");
                }
                break;          /* in regexp, $ might be tail anchor */
            }
        }

        /* End of else if chain - OP_TRANS rejoin rest */

        if (UNLIKELY(s >= send)) {
            assert(s == send);
            break;
        }

        /* backslashes */
        if (*s == '\\' && s+1 < send) {
            char* e;    /* Can be used for ending '}', etc. */

            s++;

            /* warn on \1 - \9 in substitution replacements, but note that \11
             * is an octal; and \19 is \1 followed by '9' */
            if (PL_lex_inwhat == OP_SUBST
                && !PL_lex_inpat
                && isDIGIT(*s)
                && *s != '0'
                && !isDIGIT(s[1]))
            {
                /* diag_listed_as: \%d better written as $%d */
                Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
                *--s = '$';
                break;
            }

            /* string-change backslash escapes */
            if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
                --s;
                break;
            }
            /* In a pattern, process \N, but skip any other backslash escapes.
             * This is because we don't want to translate an escape sequence
             * into a meta symbol and have the regex compiler use the meta
             * symbol meaning, e.g. \x{2E} would be confused with a dot.  But
             * in spite of this, we do have to process \N here while the proper
             * charnames handler is in scope.  See bugs #56444 and #62056.
             *
             * There is a complication because \N in a pattern may also stand
             * for 'match a non-nl', and not mean a charname, in which case its
             * processing should be deferred to the regex compiler.  To be a
             * charname it must be followed immediately by a '{', and not look
             * like \N followed by a curly quantifier, i.e., not something like
             * \N{3,}.  regcurly returns a boolean indicating if it is a legal
             * quantifier */
            else if (PL_lex_inpat
                    && (*s != 'N'
                        || s[1] != '{'
                        || regcurly(s + 1)))
            {
                *d++ = '\\';
                goto default_action;
            }

            switch (*s) {
            default:
                {
                    if ((isALPHANUMERIC(*s)))
                        Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
                                       "Unrecognized escape \\%c passed through",
                                       *s);
                    /* default action is to copy the quoted character */
                    goto default_action;
                }

            /* eg. \132 indicates the octal constant 0132 */
            case '0': case '1': case '2': case '3':
            case '4': case '5': case '6': case '7':
                {
                    I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
                    STRLEN len = 3;
                    uv = grok_oct(s, &len, &flags, NULL);
                    s += len;
                    if (len < 3 && s < send && isDIGIT(*s)
                        && ckWARN(WARN_MISC))
                    {
                        Perl_warner(aTHX_ packWARN(WARN_MISC),
                                    "%s", form_short_octal_warning(s, len));
                    }
                }
                goto NUM_ESCAPE_INSERT;

            /* eg. \o{24} indicates the octal constant \024 */
            case 'o':
                {
                    const char* error;

                    bool valid = grok_bslash_o(&s, &uv, &error,
                                               TRUE, /* Output warning */
                                               FALSE, /* Not strict */
                                               TRUE, /* Output warnings for
                                                         non-portables */
                                               UTF);
                    if (! valid) {
                        yyerror(error);
                        uv = 0; /* drop through to ensure range ends are set */
                    }
                    goto NUM_ESCAPE_INSERT;
                }

            /* eg. \x24 indicates the hex constant 0x24 */
            case 'x':
                {
                    const char* error;

                    bool valid = grok_bslash_x(&s, &uv, &error,
                                               TRUE, /* Output warning */
                                               FALSE, /* Not strict */
                                               TRUE,  /* Output warnings for
                                                         non-portables */
                                               UTF);
                    if (! valid) {
                        yyerror(error);
                        uv = 0; /* drop through to ensure range ends are set */
                    }
                }

              NUM_ESCAPE_INSERT:
                /* Insert oct or hex escaped character. */

                /* Here uv is the ordinal of the next character being added */
                if (UVCHR_IS_INVARIANT(uv)) {
                    *d++ = (char) uv;
                }
                else {
                    if (!has_utf8 && uv > 255) {

                        /* Here, 'uv' won't fit unless we convert to UTF-8.
                         * If we've only seen invariants so far, all we have to
                         * do is turn on the flag */
                        if (utf8_variant_count == 0) {
                            SvUTF8_on(sv);
                        }
                        else {
                            SvCUR_set(sv, d - SvPVX_const(sv));
                            SvPOK_on(sv);
                            *d = '\0';

                            sv_utf8_upgrade_flags_grow(
                                           sv,
                                           SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,

                                           /* Since we're having to grow here,
                                            * make sure we have enough room for
                                            * this escape and a NUL, so the
                                            * code immediately below won't have
                                            * to actually grow again */
                                          UVCHR_SKIP(uv)
                                        + (STRLEN)(send - s) + 1);
                            d = SvPVX(sv) + SvCUR(sv);
                        }

                        has_above_latin1 = TRUE;
                        has_utf8 = TRUE;
                    }

                    if (! has_utf8) {
                        *d++ = (char)uv;
                        utf8_variant_count++;
                    }
                    else {
                       /* Usually, there will already be enough room in 'sv'
                        * since such escapes are likely longer than any UTF-8
                        * sequence they can end up as.  This isn't the case on
                        * EBCDIC where \x{40000000} contains 12 bytes, and the
                        * UTF-8 for it contains 14.  And, we have to allow for
                        * a trailing NUL.  It probably can't happen on ASCII
                        * platforms, but be safe.  See Note on sizing above. */
                        const STRLEN needed = d - SvPVX(sv)
                                            + UVCHR_SKIP(uv)
                                            + (send - s)
                                            + 1;
                        if (UNLIKELY(needed > SvLEN(sv))) {
                            SvCUR_set(sv, d - SvPVX_const(sv));
                            d = SvCUR(sv) + SvGROW(sv, needed);
                        }

                        d = (char*)uvchr_to_utf8((U8*)d, uv);
                        if (PL_lex_inwhat == OP_TRANS
                            && PL_parser->lex_sub_op)
                        {
                            PL_parser->lex_sub_op->op_private |=
                                (PL_lex_repl ? OPpTRANS_FROM_UTF
                                             : OPpTRANS_TO_UTF);
                        }
                    }
                }
#ifdef EBCDIC
                non_portable_endpoint++;
#endif
                continue;

            case 'N':
                /* In a non-pattern \N must be like \N{U+0041}, or it can be a
                 * named character, like \N{LATIN SMALL LETTER A}, or a named
                 * sequence, like \N{LATIN CAPITAL LETTER A WITH MACRON AND
                 * GRAVE} (except y/// can't handle the latter, croaking).  For
                 * convenience all three forms are referred to as "named
                 * characters" below.
                 *
                 * For patterns, \N also can mean to match a non-newline.  Code
                 * before this 'switch' statement should already have handled
                 * this situation, and hence this code only has to deal with
                 * the named character cases.
                 *
                 * For non-patterns, the named characters are converted to
                 * their string equivalents.  In patterns, named characters are
                 * not converted to their ultimate forms for the same reasons
                 * that other escapes aren't.  Instead, they are converted to
                 * the \N{U+...} form to get the value from the charnames that
                 * is in effect right now, while preserving the fact that it
                 * was a named character, so that the regex compiler knows
                 * this.
                 *
                 * The structure of this section of code (besides checking for
                 * errors and upgrading to utf8) is:
                 *    If the named character is of the form \N{U+...}, pass it
                 *      through if a pattern; otherwise convert the code point
                 *      to utf8
                 *    Otherwise must be some \N{NAME}: convert to
                 *      \N{U+c1.c2...} if a pattern; otherwise convert to utf8
                 *
                 * Transliteration is an exception.  The conversion to utf8 is
                 * only done if the code point requires it to be representable.
                 *
                 * Here, 's' points to the 'N'; the test below is guaranteed to
                 * succeed if we are being called on a pattern, as we already
                 * know from a test above that the next character is a '{'.  A
                 * non-pattern \N must mean 'named character', which requires
                 * braces */
                s++;
                if (*s != '{') {
                    yyerror("Missing braces on \\N{}");
                    continue;
                }
                s++;

                /* If there is no matching '}', it is an error. */
                if (! (e = strchr(s, '}'))) {
                    if (! PL_lex_inpat) {
                        yyerror("Missing right brace on \\N{}");
                    } else {
                        yyerror("Missing right brace on \\N{} or unescaped left brace after \\N");
                    }
                    continue;
                }

                /* Here it looks like a named character */

                if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
                    s += 2;         /* Skip to next char after the 'U+' */
                    if (PL_lex_inpat) {

                        /* In patterns, we can have \N{U+xxxx.yyyy.zzzz...} */
                        /* Check the syntax.  */
                        const char *orig_s;
                        orig_s = s - 5;
                        if (!isXDIGIT(*s)) {
                          bad_NU:
                            yyerror(
                                "Invalid hexadecimal number in \\N{U+...}"
                            );
                            s = e + 1;
                            continue;
                        }
                        while (++s < e) {
                            if (isXDIGIT(*s))
                                continue;
                            else if ((*s == '.' || *s == '_')
                                  && isXDIGIT(s[1]))
                                continue;
                            goto bad_NU;
                        }

                        /* Pass everything through unchanged.
                         * +1 is for the '}' */
                        Copy(orig_s, d, e - orig_s + 1, char);
                        d += e - orig_s + 1;
                    }
                    else {  /* Not a pattern: convert the hex to string */
                        I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
                                | PERL_SCAN_SILENT_ILLDIGIT
                                | PERL_SCAN_DISALLOW_PREFIX;
                        STRLEN len = e - s;
                        uv = grok_hex(s, &len, &flags, NULL);
                        if (len == 0 || (len != (STRLEN)(e - s)))
                            goto bad_NU;

                         /* For non-tr///, if the destination is not in utf8,
                          * unconditionally recode it to be so.  This is
                          * because \N{} implies Unicode semantics, and scalars
                          * have to be in utf8 to guarantee those semantics.
                          * tr/// doesn't care about Unicode rules, so no need
                          * there to upgrade to UTF-8 for small enough code
                          * points */
                        if (! has_utf8 && (   uv > 0xFF
                                           || PL_lex_inwhat != OP_TRANS))
                        {
                            /* See Note on sizing above.  */
                            const STRLEN extra = OFFUNISKIP(uv) + (send - e) + 1;

                            SvCUR_set(sv, d - SvPVX_const(sv));
                            SvPOK_on(sv);
                            *d = '\0';

                            if (utf8_variant_count == 0) {
                                SvUTF8_on(sv);
                                d = SvCUR(sv) + SvGROW(sv, SvCUR(sv) + extra);
                            }
                            else {
                                sv_utf8_upgrade_flags_grow(
                                               sv,
                                               SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
                                               extra);
                                d = SvPVX(sv) + SvCUR(sv);
                            }

                            has_utf8 = TRUE;
                            has_above_latin1 = TRUE;
                        }

                        /* Add the (Unicode) code point to the output. */
                        if (! has_utf8 || OFFUNI_IS_INVARIANT(uv)) {
                            *d++ = (char) LATIN1_TO_NATIVE(uv);
                        }
                        else {
                            d = (char*) uvoffuni_to_utf8_flags((U8*)d, uv, 0);
                        }
                    }
                }
                else /* Here is \N{NAME} but not \N{U+...}. */
                     if ((res = get_and_check_backslash_N_name(s, e)))
                {
                    STRLEN len;
                    const char *str = SvPV_const(res, len);
                    if (PL_lex_inpat) {

                        if (! len) { /* The name resolved to an empty string */
                            Copy("\\N{}", d, 4, char);
                            d += 4;
                        }
                        else {
                            /* In order to not lose information for the regex
                            * compiler, pass the result in the specially made
                            * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
                            * the code points in hex of each character
                            * returned by charnames */

                            const char *str_end = str + len;
                            const STRLEN off = d - SvPVX_const(sv);

                            if (! SvUTF8(res)) {
                                /* For the non-UTF-8 case, we can determine the
                                 * exact length needed without having to parse
                                 * through the string.  Each character takes up
                                 * 2 hex digits plus either a trailing dot or
                                 * the "}" */
                                const char initial_text[] = "\\N{U+";
                                const STRLEN initial_len = sizeof(initial_text)
                                                           - 1;
                                d = off + SvGROW(sv, off
                                                    + 3 * len

                                                    /* +1 for trailing NUL */
                                                    + initial_len + 1

                                                    + (STRLEN)(send - e));
                                Copy(initial_text, d, initial_len, char);
                                d += initial_len;
                                while (str < str_end) {
                                    char hex_string[4];
                                    int len =
                                        my_snprintf(hex_string,
                                                  sizeof(hex_string),
                                                  "%02X.",

                                                  /* The regex compiler is
                                                   * expecting Unicode, not
                                                   * native */
                                                  NATIVE_TO_LATIN1(*str));
                                    PERL_MY_SNPRINTF_POST_GUARD(len,
                                                           sizeof(hex_string));
                                    Copy(hex_string, d, 3, char);
                                    d += 3;
                                    str++;
                                }
                                d--;    /* Below, we will overwrite the final
                                           dot with a right brace */
                            }
                            else {
                                STRLEN char_length; /* cur char's byte length */

                                /* and the number of bytes after this is
                                 * translated into hex digits */
                                STRLEN output_length;

                                /* 2 hex per byte; 2 chars for '\N'; 2 chars
                                 * for max('U+', '.'); and 1 for NUL */
                                char hex_string[2 * UTF8_MAXBYTES + 5];

                                /* Get the first character of the result. */
                                U32 uv = utf8n_to_uvchr((U8 *) str,
                                                        len,
                                                        &char_length,
                                                        UTF8_ALLOW_ANYUV);
                                /* Convert first code point to Unicode hex,
                                 * including the boiler plate before it. */
                                output_length =
                                    my_snprintf(hex_string, sizeof(hex_string),
                                             "\\N{U+%X",
                                             (unsigned int) NATIVE_TO_UNI(uv));

                                /* Make sure there is enough space to hold it */
                                d = off + SvGROW(sv, off
                                                    + output_length
                                                    + (STRLEN)(send - e)
                                                    + 2);       /* '}' + NUL */
                                /* And output it */
                                Copy(hex_string, d, output_length, char);
                                d += output_length;

                                /* For each subsequent character, append dot and
                                * its Unicode code point in hex */
                                while ((str += char_length) < str_end) {
                                    const STRLEN off = d - SvPVX_const(sv);
                                    U32 uv = utf8n_to_uvchr((U8 *) str,
                                                            str_end - str,
                                                            &char_length,
                                                            UTF8_ALLOW_ANYUV);
                                    output_length =
                                        my_snprintf(hex_string,
                                             sizeof(hex_string),
                                             ".%X",
                                             (unsigned int) NATIVE_TO_UNI(uv));

                                    d = off + SvGROW(sv, off
                                                        + output_length
                                                        + (STRLEN)(send - e)
                                                        + 2);   /* '}' +  NUL */
                                    Copy(hex_string, d, output_length, char);
                                    d += output_length;
                                }
                            }

                            *d++ = '}'; /* Done.  Add the trailing brace */
                        }
                    }
                    else { /* Here, not in a pattern.  Convert the name to a
                            * string. */

                        if (PL_lex_inwhat == OP_TRANS) {
                            str = SvPV_const(res, len);
                            if (len > ((SvUTF8(res))
                                       ? UTF8SKIP(str)
                                       : 1U))
                            {
                                yyerror(Perl_form(aTHX_
                                    "%.*s must not be a named sequence"
                                    " in transliteration operator",
                                        /*  +1 to include the "}" */
                                    (int) (e + 1 - start), start));
                                goto end_backslash_N;
                            }

                            if (SvUTF8(res) && UTF8_IS_ABOVE_LATIN1(*str)) {
                                has_above_latin1 = TRUE;
                            }

                        }
                        else if (! SvUTF8(res)) {
                            /* Make sure \N{} return is UTF-8.  This is because
                             * \N{} implies Unicode semantics, and scalars have
                             * to be in utf8 to guarantee those semantics; but
                             * not needed in tr/// */
                            sv_utf8_upgrade_flags(res, 0);
                            str = SvPV_const(res, len);
                        }

                         /* Upgrade destination to be utf8 if this new
                          * component is */
                        if (! has_utf8 && SvUTF8(res)) {
                            /* See Note on sizing above.  */
                            const STRLEN extra = len + (send - s) + 1;

                            SvCUR_set(sv, d - SvPVX_const(sv));
                            SvPOK_on(sv);
                            *d = '\0';

                            if (utf8_variant_count == 0) {
                                SvUTF8_on(sv);
                                d = SvCUR(sv) + SvGROW(sv, SvCUR(sv) + extra);
                            }
                            else {
                                sv_utf8_upgrade_flags_grow(sv,
                                                SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
                                                extra);
                                d = SvPVX(sv) + SvCUR(sv);
                            }
                            has_utf8 = TRUE;
                        } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */

                            /* See Note on sizing above.  (NOTE: SvCUR() is not
                             * set correctly here). */
                            const STRLEN extra = len + (send - e) + 1;
                            const STRLEN off = d - SvPVX_const(sv);
                            d = off + SvGROW(sv, off + extra);
                        }
                        Copy(str, d, len, char);
                        d += len;
                    }

                    SvREFCNT_dec(res);

                } /* End \N{NAME} */

              end_backslash_N:
#ifdef EBCDIC
                backslash_N++; /* \N{} is defined to be Unicode */
#endif
                s = e + 1;  /* Point to just after the '}' */
                continue;

            /* \c is a control character */
            case 'c':
                s++;
                if (s < send) {
                    *d++ = grok_bslash_c(*s++, 1);
                }
                else {
                    yyerror("Missing control char name in \\c");
                }
#ifdef EBCDIC
                non_portable_endpoint++;
#endif
                continue;

            /* printf-style backslashes, formfeeds, newlines, etc */
            case 'b':
                *d++ = '\b';
                break;
            case 'n':
                *d++ = '\n';
                break;
            case 'r':
                *d++ = '\r';
                break;
            case 'f':
                *d++ = '\f';
                break;
            case 't':
                *d++ = '\t';
                break;
            case 'e':
                *d++ = ESC_NATIVE;
                break;
            case 'a':
                *d++ = '\a';
                break;
            } /* end switch */

            s++;
            continue;
        } /* end if (backslash) */

    default_action:
        /* Just copy the input to the output, though we may have to convert
         * to/from UTF-8.
         *
         * If the input has the same representation in UTF-8 as not, it will be
         * a single byte, and we don't care about UTF8ness; just copy the byte */
        if (NATIVE_BYTE_IS_INVARIANT((U8)(*s))) {
            *d++ = *s++;
        }
        else if (! this_utf8 && ! has_utf8) {
            /* If neither source nor output is UTF-8, is also a single byte,
             * just copy it; but this byte counts should we later have to
             * convert to UTF-8 */
            *d++ = *s++;
            utf8_variant_count++;
        }
        else if (this_utf8 && has_utf8) {   /* Both UTF-8, can just copy */
            const STRLEN len = UTF8SKIP(s);

            /* We expect the source to have already been checked for
             * malformedness */
            assert(isUTF8_CHAR((U8 *) s, (U8 *) send));

            Copy(s, d, len, U8);
            d += len;
            s += len;
        }
        else { /* UTF8ness matters and doesn't match, need to convert */
            STRLEN len = 1;
            const UV nextuv   = (this_utf8)
                                ? utf8n_to_uvchr((U8*)s, send - s, &len, 0)
                                : (UV) ((U8) *s);
            STRLEN need = UVCHR_SKIP(nextuv);

            if (!has_utf8) {
                SvCUR_set(sv, d - SvPVX_const(sv));
                SvPOK_on(sv);
                *d = '\0';

                /* See Note on sizing above. */
                need += (STRLEN)(send - s) + 1;

                if (utf8_variant_count == 0) {
                    SvUTF8_on(sv);
                    d = SvCUR(sv) + SvGROW(sv, SvCUR(sv) + need);
                }
                else {
                    sv_utf8_upgrade_flags_grow(sv,
                                               SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
                                               need);
                    d = SvPVX(sv) + SvCUR(sv);
                }
                has_utf8 = TRUE;
            } else if (need > len) {
                /* encoded value larger than old, may need extra space (NOTE:
                 * SvCUR() is not set correctly here).   See Note on sizing
                 * above.  */
                const STRLEN extra = need + (send - s) + 1;
                const STRLEN off = d - SvPVX_const(sv);
                d = off + SvGROW(sv, off + extra);
            }
            s += len;

            d = (char*)uvchr_to_utf8((U8*)d, nextuv);
        }
    } /* while loop to process each character */

    /* terminate the string and set up the sv */
    *d = '\0';
    SvCUR_set(sv, d - SvPVX_const(sv));
    if (SvCUR(sv) >= SvLEN(sv))
        Perl_croak(aTHX_ "panic: constant overflowed allocated space, %" UVuf
                   " >= %" UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));

    SvPOK_on(sv);
    if (has_utf8) {
        SvUTF8_on(sv);
        if (PL_lex_inwhat == OP_TRANS && PL_parser->lex_sub_op) {
            PL_parser->lex_sub_op->op_private |=
                    (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
        }
    }

    /* shrink the sv if we allocated more than we used */
    if (SvCUR(sv) + 5 < SvLEN(sv)) {
        SvPV_shrink_to_cur(sv);
    }

    /* return the substring (via pl_yylval) only if we parsed anything */
    if (s > start) {
        char *s2 = start;
        for (; s2 < s; s2++) {
            if (*s2 == '\n')
                COPLINE_INC_WITH_HERELINES;
        }
        SvREFCNT_inc_simple_void_NN(sv);
        if (   (PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ))
            && ! PL_parser->lex_re_reparsing)
        {
            const char *const key = PL_lex_inpat ? "qr" : "q";
            const STRLEN keylen = PL_lex_inpat ? 2 : 1;
            const char *type;
            STRLEN typelen;

            if (PL_lex_inwhat == OP_TRANS) {
                type = "tr";
                typelen = 2;
            } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
                type = "s";
                typelen = 1;
            } else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'') {
                type = "q";
                typelen = 1;
            } else  {
                type = "qq";
                typelen = 2;
            }

            sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
                                type, typelen);
        }
        pl_yylval.opval = newSVOP(OP_CONST, 0, sv);
    }
    LEAVE_with_name("scan_const");
    return s;
}

/* S_intuit_more
 * Returns TRUE if there's more to the expression (e.g., a subscript),
 * FALSE otherwise.
 *
 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
 *
 * ->[ and ->{ return TRUE
 * ->$* ->$#* ->@* ->@[ ->@{ return TRUE if postderef_qq is enabled
 * { and [ outside a pattern are always subscripts, so return TRUE
 * if we're outside a pattern and it's not { or [, then return FALSE
 * if we're in a pattern and the first char is a {
 *   {4,5} (any digits around the comma) returns FALSE
 * if we're in a pattern and the first char is a [
 *   [] returns FALSE
 *   [SOMETHING] has a funky algorithm to decide whether it's a
 *      character class or not.  It has to deal with things like
 *      /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
 * anything else returns TRUE
 */

/* This is the one truly awful dwimmer necessary to conflate C and sed. */

STATIC int
S_intuit_more(pTHX_ char *s)
{
    PERL_ARGS_ASSERT_INTUIT_MORE;

    if (PL_lex_brackets)
        return TRUE;
    if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
        return TRUE;
    if (*s == '-' && s[1] == '>'
     && FEATURE_POSTDEREF_QQ_IS_ENABLED
     && ( (s[2] == '$' && (s[3] == '*' || (s[3] == '#' && s[4] == '*')))
        ||(s[2] == '@' && strchr("*[{",s[3])) ))
        return TRUE;
    if (*s != '{' && *s != '[')
        return FALSE;
    if (!PL_lex_inpat)
        return TRUE;

    /* In a pattern, so maybe we have {n,m}. */
    if (*s == '{') {
        if (regcurly(s)) {
            return FALSE;
        }
        return TRUE;
    }

    /* On the other hand, maybe we have a character class */

    s++;
    if (*s == ']' || *s == '^')
        return FALSE;
    else {
        /* this is terrifying, and it works */
        int weight;
        char seen[256];
        const char * const send = strchr(s,']');
        unsigned char un_char, last_un_char;
        char tmpbuf[sizeof PL_tokenbuf * 4];

        if (!send)              /* has to be an expression */
            return TRUE;
        weight = 2;             /* let's weigh the evidence */

        if (*s == '$')
            weight -= 3;
        else if (isDIGIT(*s)) {
            if (s[1] != ']') {
                if (isDIGIT(s[1]) && s[2] == ']')
                    weight -= 10;
            }
            else
                weight -= 100;
        }
        Zero(seen,256,char);
        un_char = 255;
        for (; s < send; s++) {
            last_un_char = un_char;
            un_char = (unsigned char)*s;
            switch (*s) {
            case '@':
            case '&':
            case '$':
                weight -= seen[un_char] * 10;
                if (isWORDCHAR_lazy_if_safe(s+1, PL_bufend, UTF)) {
                    int len;
                    scan_ident(s, tmpbuf, sizeof tmpbuf, FALSE);
                    len = (int)strlen(tmpbuf);
                    if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
                                                    UTF ? SVf_UTF8 : 0, SVt_PV))
                        weight -= 100;
                    else
                        weight -= 10;
                }
                else if (*s == '$'
                         && s[1]
                         && strchr("[#!%*<>()-=",s[1]))
                {
                    if (/*{*/ strchr("])} =",s[2]))
                        weight -= 10;
                    else
                        weight -= 1;
                }
                break;
            case '\\':
                un_char = 254;
                if (s[1]) {
                    if (strchr("wds]",s[1]))
                        weight += 100;
                    else if (seen[(U8)'\''] || seen[(U8)'"'])
                        weight += 1;
                    else if (strchr("rnftbxcav",s[1]))
                        weight += 40;
                    else if (isDIGIT(s[1])) {
                        weight += 40;
                        while (s[1] && isDIGIT(s[1]))
                            s++;
                    }
                }
                else
                    weight += 100;
                break;
            case '-':
                if (s[1] == '\\')
                    weight += 50;
                if (strchr("aA01! ",last_un_char))
                    weight += 30;
                if (strchr("zZ79~",s[1]))
                    weight += 30;
                if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
                    weight -= 5;        /* cope with negative subscript */
                break;
            default:
                if (!isWORDCHAR(last_un_char)
                    && !(last_un_char == '$' || last_un_char == '@'
                         || last_un_char == '&')
                    && isALPHA(*s) && s[1] && isALPHA(s[1])) {
                    char *d = s;
                    while (isALPHA(*s))
                        s++;