More consistent names for all the parameter name/declaration methods

[p5sagit/p5-mst-13.2.git] / toke.c

/*    toke.c
 *
 *    Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
 *    2000, 2001, 2002, 2003, 2004, 2005, 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
 */

/*
 * 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.
 */

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

#define yychar  (*PL_yycharp)
#define yylval  (*PL_yylvalp)

static char ident_too_long[] = "Identifier too long";
static char c_without_g[] = "Use of /c modifier is meaningless without /g";
static char c_in_subst[] = "Use of /c modifier is meaningless in s///";

static void restore_rsfp(pTHX_ void *f);
#ifndef PERL_NO_UTF16_FILTER
static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
#endif

#define XFAKEBRACK 128
#define XENUMMASK 127

#ifdef USE_UTF8_SCRIPTS
#   define UTF (!IN_BYTES)
#else
#   define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
#endif

/* 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)))

/* On MacOS, respect nonbreaking spaces */
#ifdef MACOS_TRADITIONAL
#define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
#else
#define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
#endif

/* 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).
 */

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

#define LEX_NORMAL              10
#define LEX_INTERPNORMAL         9
#define LEX_INTERPCASEMOD        8
#define LEX_INTERPPUSH           7
#define LEX_INTERPSTART          6
#define LEX_INTERPEND            5
#define LEX_INTERPENDMAYBE       4
#define LEX_INTERPCONCAT         3
#define LEX_INTERPCONST          2
#define LEX_FORMLINE             1
#define LEX_KNOWNEXT             0

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

#ifdef ff_next
#undef ff_next
#endif

#include "keywords.h"

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

#ifdef CLINE
#undef CLINE
#endif
#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
 * LOOPX        : loop exiting command (goto, last, dump, etc)
 * FTST         : file test operator
 * FUN0         : zero-argument function
 * FUN1         : not used, except for not, which isn't a UNIOP
 * BOop         : bitwise or or xor
 * BAop         : bitwise and
 * SHop         : shift operator
 * PWop         : power operator
 * PMop         : pattern-matching operator
 * Aop          : addition-level operator
 * 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(s,(int)retval)
#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, REPORT(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 LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
#define FTST(f)  return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
#define FUN0(f)  return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
#define FUN1(f)  return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
#define BOop(f)  return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
#define BAop(f)  return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
#define SHop(f)  return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
#define PWop(f)  return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
#define PMop(f)  return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
#define Aop(f)   return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
#define Mop(f)   return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
#define Eop(f)   return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
#define Rop(f)   return (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 UNI2(f,x) return ( \
        yylval.ival = f, \
        PL_expect = x, \
        PL_bufptr = s, \
        PL_last_uni = PL_oldbufptr, \
        PL_last_lop_op = f, \
        REPORT( \
            (*s == '(' || (s = skipspace(s), *s == '(')  \
            ? (int)FUNC1 : (int)UNIOP)))
#define UNI(f)    UNI2(f,XTERM)
#define UNIDOR(f) UNI2(f,XTERMORDORDOR)

#define UNIBRACK(f) return ( \
        yylval.ival = f, \
        PL_bufptr = s, \
        PL_last_uni = PL_oldbufptr, \
        REPORT( \
            (*s == '(' || (s = skipspace(s), *s == '(') \
        ? (int)FUNC1 : (int)UNIOP)))

/* grandfather return to old style */
#define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)

#ifdef DEBUGGING

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

static struct debug_tokens { int token, type; char *name;} 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" },
    { 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" },
    { FUNC,             TOKENTYPE_OPNUM,        "FUNC" },
    { FUNC0,            TOKENTYPE_OPNUM,        "FUNC0" },
    { FUNC0SUB,         TOKENTYPE_OPVAL,        "FUNC0SUB" },
    { FUNC1,            TOKENTYPE_OPNUM,        "FUNC1" },
    { FUNCMETH,         TOKENTYPE_OPVAL,        "FUNCMETH" },
    { 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" },
    { MYSUB,            TOKENTYPE_NONE,         "MYSUB" },
    { NOAMP,            TOKENTYPE_NONE,         "NOAMP" },
    { NOTOP,            TOKENTYPE_NONE,         "NOTOP" },
    { OROP,             TOKENTYPE_IVAL,         "OROP" },
    { OROR,             TOKENTYPE_NONE,         "OROR" },
    { PACKAGE,          TOKENTYPE_NONE,         "PACKAGE" },
    { PMFUNC,           TOKENTYPE_OPVAL,        "PMFUNC" },
    { POSTDEC,          TOKENTYPE_NONE,         "POSTDEC" },
    { POSTINC,          TOKENTYPE_NONE,         "POSTINC" },
    { POWOP,            TOKENTYPE_OPNUM,        "POWOP" },
    { PREDEC,           TOKENTYPE_NONE,         "PREDEC" },
    { PREINC,           TOKENTYPE_NONE,         "PREINC" },
    { PRIVATEREF,       TOKENTYPE_OPVAL,        "PRIVATEREF" },
    { REFGEN,           TOKENTYPE_NONE,         "REFGEN" },
    { RELOP,            TOKENTYPE_OPNUM,        "RELOP" },
    { 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" },
    { WHILE,            TOKENTYPE_IVAL,         "WHILE" },
    { WORD,             TOKENTYPE_OPVAL,        "WORD" },
    { 0,                TOKENTYPE_NONE,         0 }
};

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

STATIC int
S_tokereport(pTHX_ char* s, I32 rv)
{
    if (DEBUG_T_TEST) {
        char *name = Nullch;
        enum token_type type = TOKENTYPE_NONE;
        struct debug_tokens *p;
        SV* report = newSVpvn("<== ", 4);

        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 ((char)rv > ' ' && (char)rv < '~')
            Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
        else if (!rv)
            Perl_sv_catpv(aTHX_ report, "EOF");
        else
            Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
        switch (type) {
        case TOKENTYPE_NONE:
        case TOKENTYPE_GVVAL: /* doesn't appear to be used */
            break;
        case TOKENTYPE_IVAL:
            Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", yylval.ival);
            break;
        case TOKENTYPE_OPNUM:
            Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
                                    PL_op_name[yylval.ival]);
            break;
        case TOKENTYPE_PVAL:
            Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
            break;
        case TOKENTYPE_OPVAL:
            if (yylval.opval)
                Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
                                    PL_op_name[yylval.opval->op_type]);
            else
                Perl_sv_catpv(aTHX_ report, "(opval=null)");
            break;
        }
        Perl_sv_catpvf(aTHX_ report, " at line %d [", CopLINE(PL_curcop));
        if (s - PL_bufptr > 0)
            sv_catpvn(report, PL_bufptr, s - PL_bufptr);
        else {
            if (PL_oldbufptr && *PL_oldbufptr)
                sv_catpv(report, PL_tokenbuf);
        }
        PerlIO_printf(Perl_debug_log, "### %s]\n", SvPV_nolen(report));
    };
    return (int)rv;
}

#endif

/*
 * S_ao
 *
 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
 */

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

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

    if (!s)
        s = oldbp;
    else
        PL_bufptr = s;
    yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
    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(PL_oldoldbufptr,UTF)) {
            char *t;
            for (t = PL_oldoldbufptr; *t && (isALNUM_lazy_if(t,UTF) || *t == ':'); t++) ;
            if (t < PL_bufptr && isSPACE(*t))
                Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                        "\t(Do you need to predeclare %.*s?)\n",
                    t - PL_oldoldbufptr, PL_oldoldbufptr);
        }
        else {
            assert(s >= oldbp);
            Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                    "\t(Missing operator before %.*s?)\n", s - oldbp, oldbp);
        }
    }
    PL_bufptr = oldbp;
}

/*
 * S_missingterm
 * Complain about missing quote/regexp/heredoc terminator.
 * If it's called with (char *)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[3];
    char q;
    if (s) {
        char *nl = strrchr(s,'\n');
        if (nl)
            *nl = '\0';
    }
    else if (
#ifdef EBCDIC
        iscntrl(PL_multi_close)
#else
        PL_multi_close < 32 || PL_multi_close == 127
#endif
        ) {
        *tmpbuf = '^';
        tmpbuf[1] = toCTRL(PL_multi_close);
        s = "\\n";
        tmpbuf[2] = '\0';
        s = tmpbuf;
    }
    else {
        *tmpbuf = (char)PL_multi_close;
        tmpbuf[1] = '\0';
        s = tmpbuf;
    }
    q = strchr(s,'"') ? '\'' : '"';
    Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
}

/*
 * Perl_deprecate
 */

void
Perl_deprecate(pTHX_ char *s)
{
    if (ckWARN(WARN_DEPRECATED))
        Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
}

void
Perl_deprecate_old(pTHX_ char *s)
{
    /* This function should NOT be called for any new deprecated warnings */
    /* Use Perl_deprecate instead                                         */
    /*                                                                    */
    /* It is here to maintain backward compatibility with the pre-5.8     */
    /* warnings category hierarchy. The "deprecated" category used to     */
    /* live under the "syntax" category. It is now a top-level category   */
    /* in its own right.                                                  */

    if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
        Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX), 
                        "Use of %s is deprecated", s);
}

/*
 * depcom
 * Deprecate a comma-less variable list.
 */

STATIC void
S_depcom(pTHX)
{
    deprecate_old("comma-less variable list");
}

/*
 * 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)
{
    register char *s = SvPVX(sv);
    register char *e = s + SvCUR(sv);
    /* 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 */
            register 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)
{
    I32 count = FILTER_READ(idx+1, sv, maxlen);
    if (count > 0 && !maxlen)
        strip_return(sv);
    return count;
}
#endif

/*
 * Perl_lex_start
 * Initialize variables.  Uses the Perl save_stack to save its state (for
 * recursive calls to the parser).
 */

void
Perl_lex_start(pTHX_ SV *line)
{
    char *s;
    STRLEN len;

    SAVEI32(PL_lex_dojoin);
    SAVEI32(PL_lex_brackets);
    SAVEI32(PL_lex_casemods);
    SAVEI32(PL_lex_starts);
    SAVEI32(PL_lex_state);
    SAVEVPTR(PL_lex_inpat);
    SAVEI32(PL_lex_inwhat);
    if (PL_lex_state == LEX_KNOWNEXT) {
        I32 toke = PL_nexttoke;
        while (--toke >= 0) {
            SAVEI32(PL_nexttype[toke]);
            SAVEVPTR(PL_nextval[toke]);
        }
        SAVEI32(PL_nexttoke);
    }
    SAVECOPLINE(PL_curcop);
    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);
    SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
    SAVESPTR(PL_lex_stuff);
    SAVEI32(PL_lex_defer);
    SAVEI32(PL_sublex_info.sub_inwhat);
    SAVESPTR(PL_lex_repl);
    SAVEINT(PL_expect);
    SAVEINT(PL_lex_expect);

    PL_lex_state = LEX_NORMAL;
    PL_lex_defer = 0;
    PL_expect = XSTATE;
    PL_lex_brackets = 0;
    New(899, PL_lex_brackstack, 120, char);
    New(899, PL_lex_casestack, 12, char);
    PL_lex_casemods = 0;
    *PL_lex_casestack = '\0';
    PL_lex_dojoin = 0;
    PL_lex_starts = 0;
    PL_lex_stuff = Nullsv;
    PL_lex_repl = Nullsv;
    PL_lex_inpat = 0;
    PL_nexttoke = 0;
    PL_lex_inwhat = 0;
    PL_sublex_info.sub_inwhat = 0;
    PL_linestr = line;
    if (SvREADONLY(PL_linestr))
        PL_linestr = sv_2mortal(newSVsv(PL_linestr));
    s = SvPV(PL_linestr, len);
    if (!len || s[len-1] != ';') {
        if (!(SvFLAGS(PL_linestr) & SVs_TEMP))
            PL_linestr = sv_2mortal(newSVsv(PL_linestr));
        sv_catpvn(PL_linestr, "\n;", 2);
    }
    SvTEMP_off(PL_linestr);
    PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
    PL_bufend = PL_bufptr + SvCUR(PL_linestr);
    PL_last_lop = PL_last_uni = Nullch;
    PL_rsfp = 0;
}

/*
 * Perl_lex_end
 * Finalizer for lexing operations.  Must be called when the parser is
 * done with the lexer.
 */

void
Perl_lex_end(pTHX)
{
    PL_doextract = FALSE;
}

/*
 * 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_ char *s)
{
    char *t;
    char *n;
    char *e;
    char ch;

    CopLINE_inc(PL_curcop);
    if (*s++ != '#')
        return;
    while (SPACE_OR_TAB(*s)) s++;
    if (strnEQ(s, "line", 4))
        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++;
    while (SPACE_OR_TAB(*s))
        s++;
    if (*s == '"' && (t = strchr(s+1, '"'))) {
        s++;
        e = t + 1;
    }
    else {
        for (t = s; !isSPACE(*t); t++) ;
        e = t;
    }
    while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
        e++;
    if (*e != '\n' && *e != '\0')
        return;         /* false alarm */

    ch = *t;
    *t = '\0';
    if (t - s > 0) {
        CopFILE_free(PL_curcop);
        CopFILE_set(PL_curcop, s);
    }
    *t = ch;
    CopLINE_set(PL_curcop, atoi(n)-1);
}

/*
 * S_skipspace
 * Called to gobble the appropriate amount and type of whitespace.
 * Skips comments as well.
 */

STATIC char *
S_skipspace(pTHX_ register char *s)
{
    if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
        while (s < PL_bufend && SPACE_OR_TAB(*s))
            s++;
        return s;
    }
    for (;;) {
        STRLEN prevlen;
        SSize_t oldprevlen, oldoldprevlen;
        SSize_t oldloplen = 0, oldunilen = 0;
        while (s < PL_bufend && isSPACE(*s)) {
            if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
                incline(s);
        }

        /* comment */
        if (s < PL_bufend && *s == '#') {
            while (s < PL_bufend && *s != '\n')
                s++;
            if (s < PL_bufend) {
                s++;
                if (PL_in_eval && !PL_rsfp) {
                    incline(s);
                    continue;
                }
            }
        }

        /* only continue to recharge the buffer if we're at the end
         * of the buffer, we're not reading from a source filter, and
         * we're in normal lexing mode
         */
        if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
                PL_lex_state == LEX_FORMLINE)
            return s;

        /* try to recharge the buffer */
        if ((s = filter_gets(PL_linestr, PL_rsfp,
                             (prevlen = SvCUR(PL_linestr)))) == Nullch)
        {
            /* end of file.  Add on the -p or -n magic */
            if (PL_minus_p) {
                sv_setpv(PL_linestr,
                         ";}continue{print or die qq(-p destination: $!\\n);}");
                PL_minus_n = PL_minus_p = 0;
            }
            else if (PL_minus_n) {
                sv_setpvn(PL_linestr, ";}", 2);
                PL_minus_n = 0;
            }
            else
                sv_setpvn(PL_linestr,";", 1);

            /* reset variables for next time we lex */
            PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
                = SvPVX(PL_linestr);
            PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
            PL_last_lop = PL_last_uni = Nullch;

            /* Close the filehandle.  Could be from -P preprocessor,
             * STDIN, or a regular file.  If we were reading code from
             * STDIN (because the commandline held no -e or filename)
             * then we don't close it, we reset it so the code can
             * read from STDIN too.
             */

            if (PL_preprocess && !PL_in_eval)
                (void)PerlProc_pclose(PL_rsfp);
            else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
                PerlIO_clearerr(PL_rsfp);
            else
                (void)PerlIO_close(PL_rsfp);
            PL_rsfp = Nullfp;
            return s;
        }

        /* not at end of file, so we only read another line */
        /* make corresponding updates to old pointers, for yyerror() */
        oldprevlen = PL_oldbufptr - PL_bufend;
        oldoldprevlen = PL_oldoldbufptr - PL_bufend;
        if (PL_last_uni)
            oldunilen = PL_last_uni - PL_bufend;
        if (PL_last_lop)
            oldloplen = PL_last_lop - PL_bufend;
        PL_linestart = PL_bufptr = s + prevlen;
        PL_bufend = s + SvCUR(PL_linestr);
        s = PL_bufptr;
        PL_oldbufptr = s + oldprevlen;
        PL_oldoldbufptr = s + oldoldprevlen;
        if (PL_last_uni)
            PL_last_uni = s + oldunilen;
        if (PL_last_lop)
            PL_last_lop = s + oldloplen;
        incline(s);

        /* debugger active and we're not compiling the debugger code,
         * so store the line into the debugger's array of lines
         */
        if (PERLDB_LINE && PL_curstash != PL_debstash) {
            SV *sv = NEWSV(85,0);

            sv_upgrade(sv, SVt_PVMG);
            sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
            (void)SvIOK_on(sv);
            SvIVX(sv) = 0;
            av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
        }
    }
}

/*
 * 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)
{
    char *s;
    char *t;

    if (PL_oldoldbufptr != PL_last_uni)
        return;
    while (isSPACE(*PL_last_uni))
        PL_last_uni++;
    for (s = PL_last_uni; isALNUM_lazy_if(s,UTF) || *s == '-'; s++) ;
    if ((t = strchr(s, '(')) && t < PL_bufptr)
        return;
    if (ckWARN_d(WARN_AMBIGUOUS)){
        char ch = *s;
        *s = '\0';
        Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
                   "Warning: Use of \"%s\" without parentheses is ambiguous",
                   PL_last_uni);
        *s = ch;
    }
}

/*
 * 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 [why?]
 *  - 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, int x, char *s)
{
    yylval.ival = f;
    CLINE;
    PL_expect = x;
    PL_bufptr = s;
    PL_last_lop = PL_oldbufptr;
    PL_last_lop_op = (OPCODE)f;
    if (PL_nexttoke)
        return REPORT(LSTOP);
    if (*s == '(')
        return REPORT(FUNC);
    s = skipspace(s);
    if (*s == '(')
        return REPORT(FUNC);
    else
        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)
{
    PL_nexttype[PL_nexttoke] = type;
    PL_nexttoke++;
    if (PL_lex_state != LEX_KNOWNEXT) {
        PL_lex_defer = PL_lex_state;
        PL_lex_expect = PL_expect;
        PL_lex_state = LEX_KNOWNEXT;
    }
}

STATIC SV *
S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
{
    SV *sv = newSVpvn(start,len);
    if (UTF && !IN_BYTES && is_utf8_string((U8*)start, len))
        SvUTF8_on(sv);
    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_next 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,WORD)
 *   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)
 *   int allow_initial_tick : used by the "sub" lexer only.
 */

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

    start = skipspace(start);
    s = start;
    if (isIDFIRST_lazy_if(s,UTF) ||
        (allow_pack && *s == ':') ||
        (allow_initial_tick && *s == '\'') )
    {
        s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
        if (check_keyword && keyword(PL_tokenbuf, len))
            return start;
        if (token == METHOD) {
            s = skipspace(s);
            if (*s == '(')
                PL_expect = XTERM;
            else {
                PL_expect = XOPERATOR;
            }
        }
        PL_nextval[PL_nexttoke].opval
            = (OP*)newSVOP(OP_CONST,0,
                           S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
        PL_nextval[PL_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 "WORD".
 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
 */

STATIC void
S_force_ident(pTHX_ register char *s, int kind)
{
    if (s && *s) {
        OP* o = (OP*)newSVOP(OP_CONST, 0, newSVpv(s,0));
        PL_nextval[PL_nexttoke].opval = o;
        force_next(WORD);
        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_fetchpv(s, PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : TRUE,
                kind == '$' ? SVt_PV :
                kind == '@' ? SVt_PVAV :
                kind == '%' ? SVt_PVHV :
                              SVt_PVGV
                );
        }
    }
}

NV
Perl_str_to_version(pTHX_ SV *sv)
{
    NV retval = 0.0;
    NV nshift = 1.0;
    STRLEN len;
    char *start = SvPVx(sv,len);
    bool utf = SvUTF8(sv) ? TRUE : FALSE;
    char *end = start + len;
    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 = Nullop;
    char *d;

    s = skipspace(s);

    d = s;
    if (*d == 'v')
        d++;
    if (isDIGIT(*d)) {
        while (isDIGIT(*d) || *d == '_' || *d == '.')
            d++;
        if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
            SV *ver;
            s = scan_num(s, &yylval);
            version = yylval.opval;
            ver = cSVOPx(version)->op_sv;
            if (SvPOK(ver) && !SvNIOK(ver)) {
                (void)SvUPGRADE(ver, SVt_PVNV);
                SvNVX(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 */
    PL_nextval[PL_nexttoke].opval = version;
    force_next(WORD);

    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)
{
    register char *s;
    register char *send;
    register char *d;
    STRLEN len = 0;
    SV *pv = sv;

    if (!SvLEN(sv))
        goto finish;

    s = SvPV_force(sv, len);
    if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
        goto finish;
    send = s + len;
    while (s < send && *s != '\\')
        s++;
    if (s == send)
        goto finish;
    d = s;
    if ( PL_hints & HINT_NEW_STRING ) {
        pv = sv_2mortal(newSVpvn(SvPVX(pv), len));
        if (SvUTF8(sv))
            SvUTF8_on(pv);
    }
    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(sv));
  finish:
    if ( PL_hints & HINT_NEW_STRING )
       return new_constant(NULL, 0, "q", sv, pv, "q");
    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.
 * They assume that whoever calls them will be setting up a fake
 * join call, because each subthing puts a ',' after it.  This lets
 *   "lower \luPpEr"
 * become
 *  join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
 *
 * (I'm not sure whether the spurious commas at the end of lcfirst's
 * arguments and join's arguments are created or not).
 */

/*
 * S_sublex_start
 * Assumes that 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 yylval.ival is OP_NULL, PMFUNC otherwise).
 *
 * OP_CONST and OP_READLINE are 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 or OP_READLINE).  This just sets us up for a
 * call to S_sublex_push().
 */

STATIC I32
S_sublex_start(pTHX)
{
    register I32 op_type = yylval.ival;

    if (op_type == OP_NULL) {
        yylval.opval = PL_lex_op;
        PL_lex_op = Nullop;
        return THING;
    }
    if (op_type == OP_CONST || op_type == OP_READLINE) {
        SV *sv = tokeq(PL_lex_stuff);

        if (SvTYPE(sv) == SVt_PVIV) {
            /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
            STRLEN len;
            char *p;
            SV *nsv;

            p = SvPV(sv, len);
            nsv = newSVpvn(p, len);
            if (SvUTF8(sv))
                SvUTF8_on(nsv);
            SvREFCNT_dec(sv);
            sv = nsv;
        }
        yylval.opval = (OP*)newSVOP(op_type, 0, sv);
        PL_lex_stuff = Nullsv;
        /* Allow <FH> // "foo" */
        if (op_type == OP_READLINE)
            PL_expect = XTERMORDORDOR;
        return THING;
    }

    PL_sublex_info.super_state = PL_lex_state;
    PL_sublex_info.sub_inwhat = op_type;
    PL_sublex_info.sub_op = PL_lex_op;
    PL_lex_state = LEX_INTERPPUSH;

    PL_expect = XTERM;
    if (PL_lex_op) {
        yylval.opval = PL_lex_op;
        PL_lex_op = Nullop;
        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)
{
    ENTER;

    PL_lex_state = PL_sublex_info.super_state;
    SAVEI32(PL_lex_dojoin);
    SAVEI32(PL_lex_brackets);
    SAVEI32(PL_lex_casemods);
    SAVEI32(PL_lex_starts);
    SAVEI32(PL_lex_state);
    SAVEVPTR(PL_lex_inpat);
    SAVEI32(PL_lex_inwhat);
    SAVECOPLINE(PL_curcop);
    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);

    PL_linestr = PL_lex_stuff;
    PL_lex_stuff = Nullsv;

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

    PL_lex_dojoin = FALSE;
    PL_lex_brackets = 0;
    New(899, PL_lex_brackstack, 120, char);
    New(899, PL_lex_casestack, 12, char);
    PL_lex_casemods = 0;
    *PL_lex_casestack = '\0';
    PL_lex_starts = 0;
    PL_lex_state = LEX_INTERPCONCAT;
    CopLINE_set(PL_curcop, (line_t)PL_multi_start);

    PL_lex_inwhat = PL_sublex_info.sub_inwhat;
    if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
        PL_lex_inpat = PL_sublex_info.sub_op;
    else
        PL_lex_inpat = Nullop;

    return '(';
}

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

STATIC I32
S_sublex_done(pTHX)
{
    if (!PL_lex_starts++) {
        SV *sv = newSVpvn("",0);
        if (SvUTF8(PL_linestr))
            SvUTF8_on(sv);
        PL_expect = XOPERATOR;
        yylval.opval = (OP*)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/) */
    if (PL_lex_repl && (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 = Nullch;
        SAVEFREESV(PL_linestr);
        PL_lex_dojoin = FALSE;
        PL_lex_brackets = 0;
        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 = Nullsv;
        }
        return ',';
    }
    else {
        LEAVE;
        PL_bufend = SvPVX(PL_linestr);
        PL_bufend += SvCUR(PL_linestr);
        PL_expect = XOPERATOR;
        PL_sublex_info.sub_inwhat = 0;
        return ')';
    }
}

/*
  scan_const

  Extracts a pattern, double-quoted string, or transliteration.  This
  is terrifying code.

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

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

  In patterns:
    backslashes:
      double-quoted style: \r and \n
      regexp special ones: \D \s
      constants: \x3
      backrefs: \1 (deprecated in substitution replacements)
      case and quoting: \U \Q \E
    stops on @ and $, but not for $ as tail anchor

  In transliterations:
    characters are VERY literal, except for - not at the start or end
    of the string, which indicates a range.  scan_const expands the
    range to the full set of intermediate characters.

  In double-quoted strings:
    backslashes:
      double-quoted style: \r and \n
      constants: \x3
      backrefs: \1 (deprecated)
      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.

  @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @::foo.

  $ 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 ")| \n\t"

  \1 (backreferences) are turned into $1

  The structure of the code is
      while (there's a character to process) {
          handle transliteration ranges
          skip regexp comments
          skip # initiated comments in //x patterns
          check for embedded @foo
          check for embedded scalars
          if (backslash) {
              leave intact backslashes from leave (below)
              deprecate \1 in strings and sub replacements
              handle string-changing backslashes \l \U \Q \E, etc.
              switch (what was escaped) {
                  handle - in a transliteration (becomes a literal -)
                  handle \132 octal characters
                  handle 0x15 hex characters
                  handle \cV (control V)
                  handle printf backslashes (\f, \r, \n, etc)
              } (end switch)
          } (end if backslash)
    } (end while character to read)
                
*/

STATIC char *
S_scan_const(pTHX_ char *start)
{
    register char *send = PL_bufend;            /* end of the constant */
    SV *sv = NEWSV(93, send - start);           /* sv for the constant */
    register char *s = start;                   /* start of the constant */
    register 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? */
    I32  has_utf8 = FALSE;                      /* Output constant is UTF8 */
    I32  this_utf8 = UTF;                       /* The source string is assumed to be UTF8 */
    UV uv;

    const char *leaveit =       /* set of acceptably-backslashed characters */
        PL_lex_inpat
            ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxz0123456789[{]} \t\n\r\f\v#"
            : "";

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


    while (s < send || dorange) {
        /* get transliterations out of the way (they're most literal) */
        if (PL_lex_inwhat == OP_TRANS) {
            /* expand a range A-Z to the full set of characters.  AIE! */
            if (dorange) {
                I32 i;                          /* current expanded character */
                I32 min;                        /* first character in range */
                I32 max;                        /* last character in range */

                if (has_utf8) {
                    char *c = (char*)utf8_hop((U8*)d, -1);
                    char *e = d++;
                    while (e-- > c)
                        *(e + 1) = *e;
                    *c = (char)UTF_TO_NATIVE(0xff);
                    /* mark the range as done, and continue */
                    dorange = FALSE;
                    didrange = TRUE;
                    continue;
                }

                i = d - SvPVX(sv);              /* remember current offset */
                SvGROW(sv, SvLEN(sv) + 256);    /* never more than 256 chars in a range */
                d = SvPVX(sv) + i;              /* refresh d after realloc */
                d -= 2;                         /* eat the first char and the - */

                min = (U8)*d;                   /* first char in range */
                max = (U8)d[1];                 /* last char in range  */

                if (min > max) {
                    Perl_croak(aTHX_
                               "Invalid range \"%c-%c\" in transliteration operator",
                               (char)min, (char)max);
                }

#ifdef EBCDIC
                if ((isLOWER(min) && isLOWER(max)) ||
                    (isUPPER(min) && isUPPER(max))) {
                    if (isLOWER(min)) {
                        for (i = min; i <= max; i++)
                            if (isLOWER(i))
                                *d++ = NATIVE_TO_NEED(has_utf8,i);
                    } else {
                        for (i = min; i <= max; i++)
                            if (isUPPER(i))
                                *d++ = NATIVE_TO_NEED(has_utf8,i);
                    }
                }
                else
#endif
                    for (i = min; i <= max; i++)
                        *d++ = (char)i;

                /* mark the range as done, and continue */
                dorange = FALSE;
                didrange = TRUE;
                continue;
            }

            /* range begins (ignore - as first or last char) */
            else if (*s == '-' && s+1 < send  && s != start) {
                if (didrange) {
                    Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
                }
                if (has_utf8) {
                    *d++ = (char)UTF_TO_NATIVE(0xff);   /* use illegal utf8 byte--see pmtrans */
                    s++;
                    continue;
                }
                dorange = TRUE;
                s++;
            }
            else {
                didrange = FALSE;
            }
        }

        /* if we get here, we're not doing a transliteration */

        /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
           except for the last char, which will be done separately. */
        else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
            if (s[2] == '#') {
                while (s+1 < send && *s != ')')
                    *d++ = NATIVE_TO_NEED(has_utf8,*s++);
            }
            else if (s[2] == '{' /* This should match regcomp.c */
                     || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
            {
                I32 count = 1;
                char *regparse = s + (s[2] == '{' ? 3 : 4);
                char c;

                while (count && (c = *regparse)) {
                    if (c == '\\' && regparse[1])
                        regparse++;
                    else if (c == '{')
                        count++;
                    else if (c == '}')
                        count--;
                    regparse++;
                }
                if (*regparse != ')')
                    regparse--;         /* Leave one char for continuation. */
                while (s < regparse)
                    *d++ = NATIVE_TO_NEED(has_utf8,*s++);
            }
        }

        /* likewise skip #-initiated comments in //x patterns */
        else if (*s == '#' && PL_lex_inpat &&
          ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
            while (s+1 < send && *s != '\n')
                *d++ = NATIVE_TO_NEED(has_utf8,*s++);
        }

        /* check for embedded arrays
           (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
           */
        else if (*s == '@' && s[1]
                 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
            break;

        /* 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]))
                break;          /* in regexp, $ might be tail anchor */
        }

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

        /* backslashes */
        if (*s == '\\' && s+1 < send) {
            s++;

            /* some backslashes we leave behind */
            if (*leaveit && *s && strchr(leaveit, *s)) {
                *d++ = NATIVE_TO_NEED(has_utf8,'\\');
                *d++ = NATIVE_TO_NEED(has_utf8,*s++);
                continue;
            }

            /* deprecate \1 in strings and substitution replacements */
            if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
                isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
            {
                if (ckWARN(WARN_SYNTAX))
                    Perl_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("lLuUEQ", *s)) {
                --s;
                break;
            }

            /* if we get here, it's either a quoted -, or a digit */
            switch (*s) {

            /* quoted - in transliterations */
            case '-':
                if (PL_lex_inwhat == OP_TRANS) {
                    *d++ = *s++;
                    continue;
                }
                /* FALL THROUGH */
            default:
                {
                    if (ckWARN(WARN_MISC) &&
                        isALNUM(*s) && 
                        *s != '_')
                        Perl_warner(aTHX_ packWARN(WARN_MISC),
                               "Unrecognized escape \\%c passed through",
                               *s);
                    /* default action is to copy the quoted character */
                    goto default_action;
                }

            /* \132 indicates an octal constant */
            case '0': case '1': case '2': case '3':
            case '4': case '5': case '6': case '7':
                {
                    I32 flags = 0;
                    STRLEN len = 3;
                    uv = grok_oct(s, &len, &flags, NULL);
                    s += len;
                }
                goto NUM_ESCAPE_INSERT;

            /* \x24 indicates a hex constant */
            case 'x':
                ++s;
                if (*s == '{') {
                    char* e = strchr(s, '}');
                    I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
                      PERL_SCAN_DISALLOW_PREFIX;
                    STRLEN len;

                    ++s;
                    if (!e) {
                        yyerror("Missing right brace on \\x{}");
                        continue;
                    }
                    len = e - s;
                    uv = grok_hex(s, &len, &flags, NULL);
                    s = e + 1;
                }
                else {
                    {
                        STRLEN len = 2;
                        I32 flags = PERL_SCAN_DISALLOW_PREFIX;
                        uv = grok_hex(s, &len, &flags, NULL);
                        s += len;
                    }
                }

              NUM_ESCAPE_INSERT:
                /* Insert oct or hex escaped character.
                 * There will always enough room in sv since such
                 * escapes will be longer than any UTF-8 sequence
                 * they can end up as. */
                
                /* We need to map to chars to ASCII before doing the tests
                   to cover EBCDIC
                */
                if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
                    if (!has_utf8 && uv > 255) {
                        /* Might need to recode whatever we have
                         * accumulated so far if it contains any
                         * hibit chars.
                         *
                         * (Can't we keep track of that and avoid
                         *  this rescan? --jhi)
                         */
                        int hicount = 0;
                        U8 *c;
                        for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
                            if (!NATIVE_IS_INVARIANT(*c)) {
                                hicount++;
                            }
                        }
                        if (hicount) {
                            STRLEN offset = d - SvPVX(sv);
                            U8 *src, *dst;
                            d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
                            src = (U8 *)d - 1;
                            dst = src+hicount;
                            d  += hicount;
                            while (src >= (U8 *)SvPVX(sv)) {
                                if (!NATIVE_IS_INVARIANT(*src)) {
                                    U8 ch = NATIVE_TO_ASCII(*src);
                                    *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
                                    *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
                                }
                                else {
                                    *dst-- = *src;
                                }
                                src--;
                            }
                        }
                    }

                    if (has_utf8 || uv > 255) {
                        d = (char*)uvchr_to_utf8((U8*)d, uv);
                        has_utf8 = TRUE;
                        if (PL_lex_inwhat == OP_TRANS &&
                            PL_sublex_info.sub_op) {
                            PL_sublex_info.sub_op->op_private |=
                                (PL_lex_repl ? OPpTRANS_FROM_UTF
                                             : OPpTRANS_TO_UTF);
                        }
                    }
                    else {
                        *d++ = (char)uv;
                    }
                }
                else {
                    *d++ = (char) uv;
                }
                continue;

            /* \N{LATIN SMALL LETTER A} is a named character */
            case 'N':
                ++s;
                if (*s == '{') {
                    char* e = strchr(s, '}');
                    SV *res;
                    STRLEN len;
                    char *str;

                    if (!e) {
                        yyerror("Missing right brace on \\N{}");
                        e = s - 1;
                        goto cont_scan;
                    }
                    if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
                        /* \N{U+...} */
                        I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
                          PERL_SCAN_DISALLOW_PREFIX;
                        s += 3;
                        len = e - s;
                        uv = grok_hex(s, &len, &flags, NULL);
                        s = e + 1;
                        goto NUM_ESCAPE_INSERT;
                    }
                    res = newSVpvn(s + 1, e - s - 1);
                    res = new_constant( Nullch, 0, "charnames",
                                        res, Nullsv, "\\N{...}" );
                    if (has_utf8)
                        sv_utf8_upgrade(res);
                    str = SvPV(res,len);
#ifdef EBCDIC_NEVER_MIND
                    /* charnames uses pack U and that has been
                     * recently changed to do the below uni->native
                     * mapping, so this would be redundant (and wrong,
                     * the code point would be doubly converted).
                     * But leave this in just in case the pack U change
                     * gets revoked, but the semantics is still
                     * desireable for charnames. --jhi */
                    {
                         UV uv = utf8_to_uvchr((U8*)str, 0);

                         if (uv < 0x100) {
                              U8 tmpbuf[UTF8_MAXBYTES+1], *d;

                              d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
                              sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
                              str = SvPV(res, len);
                         }
                    }
#endif
                    if (!has_utf8 && SvUTF8(res)) {
                        char *ostart = SvPVX(sv);
                        SvCUR_set(sv, d - ostart);
                        SvPOK_on(sv);
                        *d = '\0';
                        sv_utf8_upgrade(sv);
                        /* this just broke our allocation above... */
                        SvGROW(sv, (STRLEN)(send - start));
                        d = SvPVX(sv) + SvCUR(sv);
                        has_utf8 = TRUE;
                    }
                    if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
                        char *odest = SvPVX(sv);

                        SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
                        d = SvPVX(sv) + (d - odest);
                    }
                    Copy(str, d, len, char);
                    d += len;
                    SvREFCNT_dec(res);
                  cont_scan:
                    s = e + 1;
                }
                else
                    yyerror("Missing braces on \\N{}");
                continue;

            /* \c is a control character */
            case 'c':
                s++;
                if (s < send) {
                    U8 c = *s++;
#ifdef EBCDIC
                    if (isLOWER(c))
                        c = toUPPER(c);
#endif
                    *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
                }
                else {
                    yyerror("Missing control char name in \\c");
                }
                continue;

            /* printf-style backslashes, formfeeds, newlines, etc */
            case 'b':
                *d++ = NATIVE_TO_NEED(has_utf8,'\b');
                break;
            case 'n':
                *d++ = NATIVE_TO_NEED(has_utf8,'\n');
                break;
            case 'r':
                *d++ = NATIVE_TO_NEED(has_utf8,'\r');
                break;
            case 'f':
                *d++ = NATIVE_TO_NEED(has_utf8,'\f');
                break;
            case 't':
                *d++ = NATIVE_TO_NEED(has_utf8,'\t');
                break;
            case 'e':
                *d++ = ASCII_TO_NEED(has_utf8,'\033');
                break;
            case 'a':
                *d++ = ASCII_TO_NEED(has_utf8,'\007');
                break;
            } /* end switch */

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

    default_action:
        /* If we started with encoded form, or already know we want it
           and then encode the next character */
        if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
            STRLEN len  = 1;
            UV uv       = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
            STRLEN need = UNISKIP(NATIVE_TO_UNI(uv));
            s += len;
            if (need > len) {
                /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
                STRLEN off = d - SvPVX(sv);
                d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
            }
            d = (char*)uvchr_to_utf8((U8*)d, uv);
            has_utf8 = TRUE;
        }
        else {
            *d++ = NATIVE_TO_NEED(has_utf8,*s++);
        }
    } /* while loop to process each character */

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

    SvPOK_on(sv);
    if (PL_encoding && !has_utf8) {
        sv_recode_to_utf8(sv, PL_encoding);
        if (SvUTF8(sv))
            has_utf8 = TRUE;
    }
    if (has_utf8) {
        SvUTF8_on(sv);
        if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
            PL_sublex_info.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)) {
        SvLEN_set(sv, SvCUR(sv) + 1);
        Renew(SvPVX(sv), SvLEN(sv), char);
    }

    /* return the substring (via yylval) only if we parsed anything */
    if (s > PL_bufptr) {
        if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
            sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
                              sv, Nullsv,
                              ( PL_lex_inwhat == OP_TRANS
                                ? "tr"
                                : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
                                    ? "s"
                                    : "qq")));
        yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
    } else
        SvREFCNT_dec(sv);
    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
 * { 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_ register char *s)
{
    if (PL_lex_brackets)
        return TRUE;
    if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
        return TRUE;
    if (*s != '{' && *s != '[')
        return FALSE;
    if (!PL_lex_inpat)
        return TRUE;

    /* In a pattern, so maybe we have {n,m}. */
    if (*s == '{') {
        s++;
        if (!isDIGIT(*s))
            return TRUE;
        while (isDIGIT(*s))
            s++;
        if (*s == ',')
            s++;
        while (isDIGIT(*s))
            s++;
        if (*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 = 2;         /* let's weigh the evidence */
        char seen[256];
        unsigned char un_char = 255, last_un_char;
        char *send = strchr(s,']');
        char tmpbuf[sizeof PL_tokenbuf * 4];

        if (!send)              /* has to be an expression */
            return TRUE;

        Zero(seen,256,char);
        if (*s == '$')
            weight -= 3;
        else if (isDIGIT(*s)) {
            if (s[1] != ']') {
                if (isDIGIT(s[1]) && s[2] == ']')
                    weight -= 10;
            }
            else
                weight -= 100;
        }
        for (; s < send; s++) {
            last_un_char = un_char;
            un_char = (unsigned char)*s;
            switch (*s) {
            case '@':
            case '&':
            case '$':
                weight -= seen[un_char] * 10;
                if (isALNUM_lazy_if(s+1,UTF)) {
                    scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
                    if ((int)strlen(tmpbuf) > 1 && gv_fetchpv(tmpbuf,FALSE, SVt_PV))
                        weight -= 100;
                    else
                        weight -= 10;
                }
                else if (*s == '$' && s[1] &&
                  strchr("[#!%*<>()-=",s[1])) {
                    if (/*{*/ strchr("])} =",s[2]))
                        weight -= 10;
                    else
                        weight -= 1;
                }
                break;
            case '\\':
                un_char = 254;
                if (s[1]) {
                    if (strchr("wds]",s[1]))
                        weight += 100;
                    else if (seen['\''] || seen['"'])
                        weight += 1;
                    else if (strchr("rnftbxcav",s[1]))
                        weight += 40;
                    else if (isDIGIT(s[1])) {
                        weight += 40;
                        while (s[1] && isDIGIT(s[1]))
                            s++;
                    }
                }
                else
                    weight += 100;
                break;
            case '-':
                if (s[1] == '\\')
                    weight += 50;
                if (strchr("aA01! ",last_un_char))
                    weight += 30;
                if (strchr("zZ79~",s[1]))
                    weight += 30;
                if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
                    weight -= 5;        /* cope with negative subscript */
                break;
            default:
                if (!isALNUM(last_un_char)
                    && !(last_un_char == '$' || last_un_char == '@'
                         || last_un_char == '&')
                    && isALPHA(*s) && s[1] && isALPHA(s[1])) {
                    char *d = tmpbuf;
                    while (isALPHA(*s))
                        *d++ = *s++;
                    *d = '\0';
                    if (keyword(tmpbuf, d - tmpbuf))
                        weight -= 150;
                }
                if (un_char == last_un_char + 1)
                    weight += 5;
                weight -= seen[un_char];
                break;
            }
            seen[un_char]++;
        }
        if (weight >= 0)        /* probably a character class */
            return FALSE;
    }

    return TRUE;
}

/*
 * S_intuit_method
 *
 * Does all the checking to disambiguate
 *   foo bar
 * between foo(bar) and bar->foo.  Returns 0 if not a method, otherwise
 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
 *
 * First argument is the stuff after the first token, e.g. "bar".
 *
 * Not a method if bar is a filehandle.
 * Not a method if foo is a subroutine prototyped to take a filehandle.
 * Not a method if it's really "Foo $bar"
 * Method if it's "foo $bar"
 * Not a method if it's really "print foo $bar"
 * Method if it's really "foo package::" (interpreted as package->foo)
 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
 * Not a method if bar is a filehandle or package, but is quoted with
 *   =>
 */

STATIC int
S_intuit_method(pTHX_ char *start, GV *gv)
{
    char *s = start + (*start == '$');
    char tmpbuf[sizeof PL_tokenbuf];
    STRLEN len;
    GV* indirgv;

    if (gv) {
        CV *cv;
        if (GvIO(gv))
            return 0;
        if ((cv = GvCVu(gv))) {
            char *proto = SvPVX(cv);
            if (proto) {
                if (*proto == ';')
                    proto++;
                if (*proto == '*')
                    return 0;
            }
        } else
            gv = 0;
    }
    s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
    /* start is the beginning of the possible filehandle/object,
     * and s is the end of it
     * tmpbuf is a copy of it
     */

    if (*start == '$') {
        if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
            return 0;
        s = skipspace(s);
        PL_bufptr = start;
        PL_expect = XREF;
        return *s == '(' ? FUNCMETH : METHOD;
    }
    if (!keyword(tmpbuf, len)) {
        if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
            len -= 2;
            tmpbuf[len] = '\0';
            goto bare_package;
        }
        indirgv = gv_fetchpv(tmpbuf, FALSE, SVt_PVCV);
        if (indirgv && GvCVu(indirgv))
            return 0;
        /* filehandle or package name makes it a method */
        if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
            s = skipspace(s);
            if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
                return 0;       /* no assumptions -- "=>" quotes bearword */
      bare_package:
            PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST, 0,
                                                   newSVpvn(tmpbuf,len));
            PL_nextval[PL_nexttoke].opval->op_private = OPpCONST_BARE;
            PL_expect = XTERM;
            force_next(WORD);
            PL_bufptr = s;
            return *s == '(' ? FUNCMETH : METHOD;
        }
    }
    return 0;
}

/*
 * S_incl_perldb
 * Return a string of Perl code to load the debugger.  If PERL5DB
 * is set, it will return the contents of that, otherwise a
 * compile-time require of perl5db.pl.
 */

STATIC char*
S_incl_perldb(pTHX)
{
    if (PL_perldb) {
        char *pdb = PerlEnv_getenv("PERL5DB");

        if (pdb)
            return pdb;
        SETERRNO(0,SS_NORMAL);
        return "BEGIN { require 'perl5db.pl' }";
    }
    return "";
}


/* Encoded script support. filter_add() effectively inserts a
 * 'pre-processing' function into the current source input stream.
 * Note that the filter function only applies to the current source file
 * (e.g., it will not affect files 'require'd or 'use'd by this one).
 *
 * The datasv parameter (which may be NULL) can be used to pass
 * private data to this instance of the filter. The filter function
 * can recover the SV using the FILTER_DATA macro and use it to
 * store private buffers and state information.
 *
 * The supplied datasv parameter is upgraded to a PVIO type
 * and the IoDIRP/IoANY field is used to store the function pointer,
 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
 * private use must be set using malloc'd pointers.
 */

SV *
Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
{
    if (!funcp)
        return Nullsv;

    if (!PL_rsfp_filters)
        PL_rsfp_filters = newAV();
    if (!datasv)
        datasv = NEWSV(255,0);
    if (!SvUPGRADE(datasv, SVt_PVIO))
        Perl_die(aTHX_ "Can't upgrade filter_add data to SVt_PVIO");
    IoANY(datasv) = (void *)funcp; /* stash funcp into spare field */
    IoFLAGS(datasv) |= IOf_FAKE_DIRP;
    DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
                          (void*)funcp, SvPV_nolen(datasv)));
    av_unshift(PL_rsfp_filters, 1);
    av_store(PL_rsfp_filters, 0, datasv) ;
    return(datasv);
}


/* Delete most recently added instance of this filter function. */
void
Perl_filter_del(pTHX_ filter_t funcp)
{
    SV *datasv;
    DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p", (void*)funcp));
    if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
        return;
    /* if filter is on top of stack (usual case) just pop it off */
    datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
    if (IoANY(datasv) == (void *)funcp) {
        IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
        IoANY(datasv) = (void *)NULL;
        sv_free(av_pop(PL_rsfp_filters));

        return;
    }
    /* we need to search for the correct entry and clear it     */
    Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
}


/* Invoke the idxth filter function for the current rsfp.        */
/* maxlen 0 = read one text line */
I32
Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
{
    filter_t funcp;
    SV *datasv = NULL;

    if (!PL_rsfp_filters)
        return -1;
    if (idx > AvFILLp(PL_rsfp_filters)) {       /* Any more filters?    */
        /* Provide a default input filter to make life easy.    */
        /* Note that we append to the line. This is handy.      */
        DEBUG_P(PerlIO_printf(Perl_debug_log,
                              "filter_read %d: from rsfp\n", idx));
        if (maxlen) {
            /* Want a block */
            int len ;
            int old_len = SvCUR(buf_sv) ;

            /* ensure buf_sv is large enough */
            SvGROW(buf_sv, (STRLEN)(old_len + maxlen)) ;
            if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len, maxlen)) <= 0){
                if (PerlIO_error(PL_rsfp))
                    return -1;          /* error */
                else
                    return 0 ;          /* end of file */
            }
            SvCUR_set(buf_sv, old_len + len) ;
        } else {
            /* Want a line */
            if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
                if (PerlIO_error(PL_rsfp))
                    return -1;          /* error */
                else
                    return 0 ;          /* end of file */
            }
        }
        return SvCUR(buf_sv);
    }
    /* Skip this filter slot if filter has been deleted */
    if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
        DEBUG_P(PerlIO_printf(Perl_debug_log,
                              "filter_read %d: skipped (filter deleted)\n",
                              idx));
        return FILTER_READ(idx+1, buf_sv, maxlen); /* recurse */
    }
    /* Get function pointer hidden within datasv        */
    funcp = (filter_t)IoANY(datasv);
    DEBUG_P(PerlIO_printf(Perl_debug_log,
                          "filter_read %d: via function %p (%s)\n",
                          idx, (void*)funcp, SvPV_nolen(datasv)));
    /* Call function. The function is expected to       */
    /* call "FILTER_READ(idx+1, buf_sv)" first.         */
    /* Return: <0:error, =0:eof, >0:not eof             */
    return (*funcp)(aTHX_ idx, buf_sv, maxlen);
}

STATIC char *
S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
{
#ifdef PERL_CR_FILTER
    if (!PL_rsfp_filters) {
        filter_add(S_cr_textfilter,NULL);
    }
#endif
    if (PL_rsfp_filters) {
        if (!append)
            SvCUR_set(sv, 0);   /* start with empty line        */
        if (FILTER_READ(0, sv, 0) > 0)
            return ( SvPVX(sv) ) ;
        else
            return Nullch ;
    }
    else
        return (sv_gets(sv, fp, append));
}

STATIC HV *
S_find_in_my_stash(pTHX_ char *pkgname, I32 len)
{
    GV *gv;

    if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
        return PL_curstash;

    if (len > 2 &&
        (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
        (gv = gv_fetchpv(pkgname, FALSE, SVt_PVHV)))
    {
        return GvHV(gv);                        /* Foo:: */
    }

    /* use constant CLASS => 'MyClass' */
    if ((gv = gv_fetchpv(pkgname, FALSE, SVt_PVCV))) {
        SV *sv;
        if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
            pkgname = SvPV_nolen(sv);
        }
    }

    return gv_stashpv(pkgname, FALSE);
}

#ifdef DEBUGGING
    static char* exp_name[] =
        { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
          "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
        };
#endif

/*
  yylex

  Works out what to call the token just pulled out of the input
  stream.  The yacc parser takes care of taking the ops we return and
  stitching them into a tree.

  Returns:
    PRIVATEREF

  Structure:
      if read an identifier
          if we're in a my declaration
              croak if they tried to say my($foo::bar)
              build the ops for a my() declaration
          if it's an access to a my() variable
              are we in a sort block?
                  croak if my($a); $a <=> $b
              build ops for access to a my() variable
          if in a dq string, and they've said @foo and we can't find @foo
              croak
          build ops for a bareword
      if we already built the token before, use it.
*/


#ifdef __SC__
#pragma segment Perl_yylex
#endif
int
Perl_yylex(pTHX)
{
    register char *s = PL_bufptr;
    register char *d;
    register I32 tmp;
    STRLEN len;
    GV *gv = Nullgv;
    GV **gvp = 0;
    bool bof = FALSE;
    I32 orig_keyword = 0;

    DEBUG_T( {
        PerlIO_printf(Perl_debug_log, "### LEX_%s\n",
                                        lex_state_names[PL_lex_state]);
    } );
    /* check if there's an identifier for us to look at */
    if (PL_pending_ident)
        return REPORT(S_pending_ident(aTHX));

    /* no identifier pending identification */

    switch (PL_lex_state) {
#ifdef COMMENTARY
    case LEX_NORMAL:            /* Some compilers will produce faster */
    case LEX_INTERPNORMAL:      /* code if we comment these out. */
        break;
#endif

    /* when we've already built the next token, just pull it out of the queue */
    case LEX_KNOWNEXT:
        PL_nexttoke--;
        yylval = PL_nextval[PL_nexttoke];
        if (!PL_nexttoke) {
            PL_lex_state = PL_lex_defer;
            PL_expect = PL_lex_expect;
            PL_lex_defer = LEX_NORMAL;
        }
        DEBUG_T({ PerlIO_printf(Perl_debug_log,
              "### Next token after '%s' was known, type %"IVdf"\n", PL_bufptr,
              (IV)PL_nexttype[PL_nexttoke]); });

        return REPORT(PL_nexttype[PL_nexttoke]);

    /* interpolated case modifiers like \L \U, including \Q and \E.
       when we get here, PL_bufptr is at the \
    */
    case LEX_INTERPCASEMOD:
#ifdef DEBUGGING
        if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
            Perl_croak(aTHX_ "panic: INTERPCASEMOD");
#endif
        /* handle \E or end of string */
        if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
            char oldmod;

            /* if at a \E */
            if (PL_lex_casemods) {
                oldmod = PL_lex_casestack[--PL_lex_casemods];
                PL_lex_casestack[PL_lex_casemods] = '\0';

                if (PL_bufptr != PL_bufend
                    && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
                    PL_bufptr += 2;
                    PL_lex_state = LEX_INTERPCONCAT;
                }
                return REPORT(')');
            }
            if (PL_bufptr != PL_bufend)
                PL_bufptr += 2;
            PL_lex_state = LEX_INTERPCONCAT;
            return yylex();
        }
        else {
            DEBUG_T({ PerlIO_printf(Perl_debug_log,
              "### Saw case modifier at '%s'\n", PL_bufptr); });
            s = PL_bufptr + 1;
            if (s[1] == '\\' && s[2] == 'E') {
                PL_bufptr = s + 3;
                PL_lex_state = LEX_INTERPCONCAT;
                return yylex();
            }
            else {
                if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
                    tmp = *s, *s = s[2], s[2] = (char)tmp;      /* misordered... */
                if ((*s == 'L' || *s == 'U') &&
                    (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
                    PL_lex_casestack[--PL_lex_casemods] = '\0';
                    return REPORT(')');
                }
                if (PL_lex_casemods > 10)
                    Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
                PL_lex_casestack[PL_lex_casemods++] = *s;
                PL_lex_casestack[PL_lex_casemods] = '\0';
                PL_lex_state = LEX_INTERPCONCAT;
                PL_nextval[PL_nexttoke].ival = 0;
                force_next('(');
                if (*s == 'l')
                    PL_nextval[PL_nexttoke].ival = OP_LCFIRST;
                else if (*s == 'u')
                    PL_nextval[PL_nexttoke].ival = OP_UCFIRST;
                else if (*s == 'L')
                    PL_nextval[PL_nexttoke].ival = OP_LC;
                else if (*s == 'U')
                    PL_nextval[PL_nexttoke].ival = OP_UC;
                else if (*s == 'Q')
                    PL_nextval[PL_nexttoke].ival = OP_QUOTEMETA;
                else
                    Perl_croak(aTHX_ "panic: yylex");
                PL_bufptr = s + 1;
            }
            force_next(FUNC);
            if (PL_lex_starts) {
                s = PL_bufptr;
                PL_lex_starts = 0;
                /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
                if (PL_lex_casemods == 1 && PL_lex_inpat)
                    OPERATOR(',');
                else
                    Aop(OP_CONCAT);
            }
            else
                return yylex();
        }

    case LEX_INTERPPUSH:
        return REPORT(sublex_push());

    case LEX_INTERPSTART:
        if (PL_bufptr == PL_bufend)
            return REPORT(sublex_done());
        DEBUG_T({ PerlIO_printf(Perl_debug_log,
              "### Interpolated variable at '%s'\n", PL_bufptr); });
        PL_expect = XTERM;
        PL_lex_dojoin = (*PL_bufptr == '@');
        PL_lex_state = LEX_INTERPNORMAL;
        if (PL_lex_dojoin) {
            PL_nextval[PL_nexttoke].ival = 0;
            force_next(',');
            force_ident("\"", '$');
            PL_nextval[PL_nexttoke].ival = 0;
            force_next('$');
            PL_nextval[PL_nexttoke].ival = 0;
            force_next('(');
            PL_nextval[PL_nexttoke].ival = OP_JOIN;     /* emulate join($", ...) */
            force_next(FUNC);
        }
        if (PL_lex_starts++) {
            s = PL_bufptr;
            /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
            if (!PL_lex_casemods && PL_lex_inpat)
                OPERATOR(',');
            else
                Aop(OP_CONCAT);
        }
        return yylex();

    case LEX_INTERPENDMAYBE:
        if (intuit_more(PL_bufptr)) {
            PL_lex_state = LEX_INTERPNORMAL;    /* false alarm, more expr */
            break;
        }
        /* FALL THROUGH */

    case LEX_INTERPEND:
        if (PL_lex_dojoin) {
            PL_lex_dojoin = FALSE;
            PL_lex_state = LEX_INTERPCONCAT;
            return REPORT(')');
        }
        if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
            && SvEVALED(PL_lex_repl))
        {
            if (PL_bufptr != PL_bufend)
                Perl_croak(aTHX_ "Bad evalled substitution pattern");
            PL_lex_repl = Nullsv;
        }
        /* FALLTHROUGH */
    case LEX_INTERPCONCAT:
#ifdef DEBUGGING
        if (PL_lex_brackets)
            Perl_croak(aTHX_ "panic: INTERPCONCAT");
#endif
        if (PL_bufptr == PL_bufend)
            return REPORT(sublex_done());

        if (SvIVX(PL_linestr) == '\'') {
            SV *sv = newSVsv(PL_linestr);
            if (!PL_lex_inpat)
                sv = tokeq(sv);
            else if ( PL_hints & HINT_NEW_RE )
                sv = new_constant(NULL, 0, "qr", sv, sv, "q");
            yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
            s = PL_bufend;
        }
        else {
            s = scan_const(PL_bufptr);
            if (*s == '\\')
                PL_lex_state = LEX_INTERPCASEMOD;
            else
                PL_lex_state = LEX_INTERPSTART;
        }

        if (s != PL_bufptr) {
            PL_nextval[PL_nexttoke] = yylval;
            PL_expect = XTERM;
            force_next(THING);
            if (PL_lex_starts++) {
                /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
                if (!PL_lex_casemods && PL_lex_inpat)
                    OPERATOR(',');
                else
                    Aop(OP_CONCAT);
            }
            else {
                PL_bufptr = s;
                return yylex();
            }
        }

        return yylex();
    case LEX_FORMLINE:
        PL_lex_state = LEX_NORMAL;
        s = scan_formline(PL_bufptr);
        if (!PL_lex_formbrack)
            goto rightbracket;
        OPERATOR(';');
    }

    s = PL_bufptr;
    PL_oldoldbufptr = PL_oldbufptr;
    PL_oldbufptr = s;
    DEBUG_T( {
        PerlIO_printf(Perl_debug_log, "### Tokener expecting %s at [%s]\n",
                      exp_name[PL_expect], s);
    } );

  retry:
    switch (*s) {
    default:
        if (isIDFIRST_lazy_if(s,UTF))
            goto keylookup;
        Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
    case 4:
    case 26:
        goto fake_eof;                  /* emulate EOF on ^D or ^Z */
    case 0:
        if (!PL_rsfp) {
            PL_last_uni = 0;
            PL_last_lop = 0;
            if (PL_lex_brackets) {
                if (PL_lex_formbrack)
                    yyerror("Format not terminated");
                else
                    yyerror("Missing right curly or square bracket");
            }
            DEBUG_T( { PerlIO_printf(Perl_debug_log,
                        "### Tokener got EOF\n");
            } );
            TOKEN(0);
        }
        if (s++ < PL_bufend)
            goto retry;                 /* ignore stray nulls */
        PL_last_uni = 0;
        PL_last_lop = 0;
        if (!PL_in_eval && !PL_preambled) {
            PL_preambled = TRUE;
            sv_setpv(PL_linestr,incl_perldb());
            if (SvCUR(PL_linestr))
                sv_catpvn(PL_linestr,";", 1);
            if (PL_preambleav){
                while(AvFILLp(PL_preambleav) >= 0) {
                    SV *tmpsv = av_shift(PL_preambleav);
                    sv_catsv(PL_linestr, tmpsv);
                    sv_catpvn(PL_linestr, ";", 1);
                    sv_free(tmpsv);
                }
                sv_free((SV*)PL_preambleav);
                PL_preambleav = NULL;
            }
            if (PL_minus_n || PL_minus_p) {
                sv_catpv(PL_linestr, "LINE: while (<>) {");
                if (PL_minus_l)
                    sv_catpv(PL_linestr,"chomp;");
                if (PL_minus_a) {
                    if (PL_minus_F) {
                        if ((*PL_splitstr == '/' || *PL_splitstr == '\''
                             || *PL_splitstr == '"')
                              && strchr(PL_splitstr + 1, *PL_splitstr))
                            Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
                        else {
                            /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
                               bytes can be used as quoting characters.  :-) */
                            /* The count here deliberately includes the NUL
                               that terminates the C string constant.  This
                               embeds the opening NUL into the string.  */
                            sv_catpvn(PL_linestr, "our @F=split(q", 15);
                            s = PL_splitstr;
                            do {
                                /* Need to \ \s  */
                                if (*s == '\\')
                                    sv_catpvn(PL_linestr, s, 1);
                                sv_catpvn(PL_linestr, s, 1);
                            } while (*s++);
                            /* This loop will embed the trailing NUL of
                               PL_linestr as the last thing it does before
                               terminating.  */
                            sv_catpvn(PL_linestr, ");", 2);
                        }
                    }
                    else
                        sv_catpv(PL_linestr,"our @F=split(' ');");
                }
            }
            sv_catpvn(PL_linestr, "\n", 1);
            PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
            PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
            PL_last_lop = PL_last_uni = Nullch;
            if (PERLDB_LINE && PL_curstash != PL_debstash) {
                SV *sv = NEWSV(85,0);

                sv_upgrade(sv, SVt_PVMG);
                sv_setsv(sv,PL_linestr);
                (void)SvIOK_on(sv);
                SvIVX(sv) = 0;
                av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
            }
            goto retry;
        }
        do {
            bof = PL_rsfp ? TRUE : FALSE;
            if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == Nullch) {
              fake_eof:
                if (PL_rsfp) {
                    if (PL_preprocess && !PL_in_eval)
                        (void)PerlProc_pclose(PL_rsfp);
                    else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
                        PerlIO_clearerr(PL_rsfp);
                    else
                        (void)PerlIO_close(PL_rsfp);
                    PL_rsfp = Nullfp;
                    PL_doextract = FALSE;
                }
                if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
                    sv_setpv(PL_linestr,PL_minus_p
                             ? ";}continue{print;}" : ";}");
                    PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
                    PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
                    PL_last_lop = PL_last_uni = Nullch;
                    PL_minus_n = PL_minus_p = 0;
                    goto retry;
                }
                PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
                PL_last_lop = PL_last_uni = Nullch;
                sv_setpv(PL_linestr,"");
                TOKEN(';');     /* not infinite loop because rsfp is NULL now */
            }
            /* If it looks like the start of a BOM or raw UTF-16,
             * check if it in fact is. */
            else if (bof &&
                     (*s == 0 ||
                      *(U8*)s == 0xEF ||
                      *(U8*)s >= 0xFE ||
                      s[1] == 0)) {
#ifdef PERLIO_IS_STDIO
#  ifdef __GNU_LIBRARY__
#    if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
#      define FTELL_FOR_PIPE_IS_BROKEN
#    endif
#  else
#    ifdef __GLIBC__
#      if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
#        define FTELL_FOR_PIPE_IS_BROKEN
#      endif
#    endif
#  endif
#endif
#ifdef FTELL_FOR_PIPE_IS_BROKEN
                /* This loses the possibility to detect the bof
                 * situation on perl -P when the libc5 is being used.
                 * Workaround?  Maybe attach some extra state to PL_rsfp?
                 */
                if (!PL_preprocess)
                    bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
#else
                bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
#endif
                if (bof) {
                    PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
                    s = swallow_bom((U8*)s);
                }
            }
            if (PL_doextract) {
                /* Incest with pod. */
                if (*s == '=' && strnEQ(s, "=cut", 4)) {
                    sv_setpv(PL_linestr, "");
                    PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
                    PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
                    PL_last_lop = PL_last_uni = Nullch;
                    PL_doextract = FALSE;
                }
            }
            incline(s);
        } while (PL_doextract);
        PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
        if (PERLDB_LINE && PL_curstash != PL_debstash) {
            SV *sv = NEWSV(85,0);

            sv_upgrade(sv, SVt_PVMG);
            sv_setsv(sv,PL_linestr);
            (void)SvIOK_on(sv);
            SvIVX(sv) = 0;
            av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
        }
        PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
        PL_last_lop = PL_last_uni = Nullch;
        if (CopLINE(PL_curcop) == 1) {
            while (s < PL_bufend && isSPACE(*s))
                s++;
            if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
                s++;
            d = Nullch;
            if (!PL_in_eval) {
                if (*s == '#' && *(s+1) == '!')
                    d = s + 2;
#ifdef ALTERNATE_SHEBANG
                else {
                    static char as[] = ALTERNATE_SHEBANG;
                    if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
                        d = s + (sizeof(as) - 1);
                }
#endif /* ALTERNATE_SHEBANG */
            }
            if (d) {
                char *ipath;
                char *ipathend;

                while (isSPACE(*d))
                    d++;
                ipath = d;
                while (*d && !isSPACE(*d))
                    d++;
                ipathend = d;

#ifdef ARG_ZERO_IS_SCRIPT
                if (ipathend > ipath) {
                    /*
                     * HP-UX (at least) sets argv[0] to the script name,
                     * which makes $^X incorrect.  And Digital UNIX and Linux,
                     * at least, set argv[0] to the basename of the Perl
                     * interpreter. So, having found "#!", we'll set it right.
                     */
                    SV *x = GvSV(gv_fetchpv("\030", TRUE, SVt_PV)); /* $^X */
                    assert(SvPOK(x) || SvGMAGICAL(x));
                    if (sv_eq(x, CopFILESV(PL_curcop))) {
                        sv_setpvn(x, ipath, ipathend - ipath);
                        SvSETMAGIC(x);
                    }
                    else {
                        STRLEN blen;
                        STRLEN llen;
                        char *bstart = SvPV(CopFILESV(PL_curcop),blen);
                        char *lstart = SvPV(x,llen);
                        if (llen < blen) {
                            bstart += blen - llen;
                            if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
                                sv_setpvn(x, ipath, ipathend - ipath);
                                SvSETMAGIC(x);
                            }
                        }
                    }
                    TAINT_NOT;  /* $^X is always tainted, but that's OK */
                }
#endif /* ARG_ZERO_IS_SCRIPT */

                /*
                 * Look for options.
                 */
                d = instr(s,"perl -");
                if (!d) {
                    d = instr(s,"perl");
#if defined(DOSISH)
                    /* avoid getting into infinite loops when shebang
                     * line contains "Perl" rather than "perl" */
                    if (!d) {
                        for (d = ipathend-4; d >= ipath; --d) {
                            if ((*d == 'p' || *d == 'P')
                                && !ibcmp(d, "perl", 4))
                            {
                                break;
                            }
                        }
                        if (d < ipath)
                            d = Nullch;
                    }
#endif
                }
#ifdef ALTERNATE_SHEBANG
                /*
                 * If the ALTERNATE_SHEBANG on this system starts with a
                 * character that can be part of a Perl expression, then if
                 * we see it but not "perl", we're probably looking at the
                 * start of Perl code, not a request to hand off to some
                 * other interpreter.  Similarly, if "perl" is there, but
                 * not in the first 'word' of the line, we assume the line
                 * contains the start of the Perl program.
                 */
                if (d && *s != '#') {
                    char *c = ipath;
                    while (*c && !strchr("; \t\r\n\f\v#", *c))
                        c++;
                    if (c < d)
                        d = Nullch;     /* "perl" not in first word; ignore */
                    else
                        *s = '#';       /* Don't try to parse shebang line */
                }
#endif /* ALTERNATE_SHEBANG */
#ifndef MACOS_TRADITIONAL
                if (!d &&
                    *s == '#' &&
                    ipathend > ipath &&
                    !PL_minus_c &&
                    !instr(s,"indir") &&
                    instr(PL_origargv[0],"perl"))
                {
                    char **newargv;

                    *ipathend = '\0';
                    s = ipathend + 1;
                    while (s < PL_bufend && isSPACE(*s))
                        s++;
                    if (s < PL_bufend) {
                        Newz(899,newargv,PL_origargc+3,char*);
                        newargv[1] = s;
                        while (s < PL_bufend && !isSPACE(*s))
                            s++;
                        *s = '\0';
                        Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
                    }
                    else
                        newargv = PL_origargv;
                    newargv[0] = ipath;
                    PERL_FPU_PRE_EXEC
                    PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
                    PERL_FPU_POST_EXEC
                    Perl_croak(aTHX_ "Can't exec %s", ipath);
                }
#endif
                if (d) {
                    U32 oldpdb = PL_perldb;
                    bool oldn = PL_minus_n;
                    bool oldp = PL_minus_p;

                    while (*d && !isSPACE(*d)) d++;
                    while (SPACE_OR_TAB(*d)) d++;

                    if (*d++ == '-') {
                        bool switches_done = PL_doswitches;
                        do {
                            if (*d == 'M' || *d == 'm') {
                                char *m = d;
                                while (*d && !isSPACE(*d)) d++;
                                Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
                                      (int)(d - m), m);
                            }
                            d = moreswitches(d);
                        } while (d);
                        if (PL_doswitches && !switches_done) {
                            int argc = PL_origargc;
                            char **argv = PL_origargv;
                            do {
                                argc--,argv++;
                            } while (argc && argv[0][0] == '-' && argv[0][1]);
                            init_argv_symbols(argc,argv);
                        }
                        if ((PERLDB_LINE && !oldpdb) ||
                            ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
                              /* if we have already added "LINE: while (<>) {",
                                 we must not do it again */
                        {
                            sv_setpv(PL_linestr, "");
                            PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
                            PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
                            PL_last_lop = PL_last_uni = Nullch;
                            PL_preambled = FALSE;
                            if (PERLDB_LINE)
                                (void)gv_fetchfile(PL_origfilename);
                            goto retry;
                        }
                        if (PL_doswitches && !switches_done) {
                            int argc = PL_origargc;
                            char **argv = PL_origargv;
                            do {
                                argc--,argv++;
                            } while (argc && argv[0][0] == '-' && argv[0][1]);
                            init_argv_symbols(argc,argv);
                        }
                    }
                }
            }
        }
        if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
            PL_bufptr = s;
            PL_lex_state = LEX_FORMLINE;
            return yylex();
        }
        goto retry;
    case '\r':
#ifdef PERL_STRICT_CR
        Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
        Perl_croak(aTHX_
      "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
#endif
    case ' ': case '\t': case '\f': case 013:
#ifdef MACOS_TRADITIONAL
    case '\312':
#endif
        s++;
        goto retry;
    case '#':
    case '\n':
        if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
            if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
                /* handle eval qq[#line 1 "foo"\n ...] */
                CopLINE_dec(PL_curcop);
                incline(s);
            }
            d = PL_bufend;
            while (s < d && *s != '\n')
                s++;
            if (s < d)
                s++;
            else if (s > d) /* Found by Ilya: feed random input to Perl. */
              Perl_croak(aTHX_ "panic: input overflow");
            incline(s);
            if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
                PL_bufptr = s;
                PL_lex_state = LEX_FORMLINE;
                return yylex();
            }
        }
        else {
            *s = '\0';
            PL_bufend = s;
        }
        goto retry;
    case '-':
        if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
            I32 ftst = 0;

            s++;
            PL_bufptr = s;
            tmp = *s++;

            while (s < PL_bufend && SPACE_OR_TAB(*s))
                s++;

            if (strnEQ(s,"=>",2)) {
                s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
                DEBUG_T( { PerlIO_printf(Perl_debug_log,
                            "### Saw unary minus before =>, forcing word '%s'\n", s);
                } );
                OPERATOR('-');          /* unary minus */
            }
            PL_last_uni = PL_oldbufptr;
            switch (tmp) {
            case 'r': ftst = OP_FTEREAD;        break;
            case 'w': ftst = OP_FTEWRITE;       break;
            case 'x': ftst = OP_FTEEXEC;        break;
            case 'o': ftst = OP_FTEOWNED;       break;
            case 'R': ftst = OP_FTRREAD;        break;
            case 'W': ftst = OP_FTRWRITE;       break;
            case 'X': ftst = OP_FTREXEC;        break;
            case 'O': ftst = OP_FTROWNED;       break;
            case 'e': ftst = OP_FTIS;           break;
            case 'z': ftst = OP_FTZERO;         break;
            case 's': ftst = OP_FTSIZE;         break;
            case 'f': ftst = OP_FTFILE;         break;
            case 'd': ftst = OP_FTDIR;          break;
            case 'l': ftst = OP_FTLINK;         break;
            case 'p': ftst = OP_FTPIPE;         break;
            case 'S': ftst = OP_FTSOCK;         break;
            case 'u': ftst = OP_FTSUID;         break;
            case 'g': ftst = OP_FTSGID;         break;
            case 'k': ftst = OP_FTSVTX;         break;
            case 'b': ftst = OP_FTBLK;          break;
            case 'c': ftst = OP_FTCHR;          break;
            case 't': ftst = OP_FTTTY;          break;
            case 'T': ftst = OP_FTTEXT;         break;
            case 'B': ftst = OP_FTBINARY;       break;
            case 'M': case 'A': case 'C':
                gv_fetchpv("\024",TRUE, SVt_PV);
                switch (tmp) {
                case 'M': ftst = OP_FTMTIME;    break;
                case 'A': ftst = OP_FTATIME;    break;
                case 'C': ftst = OP_FTCTIME;    break;
                default:                        break;
                }
                break;
            default:
                break;
            }
            if (ftst) {
                PL_last_lop_op = (OPCODE)ftst;
                DEBUG_T( { PerlIO_printf(Perl_debug_log,
                        "### Saw file test %c\n", (int)ftst);
                } );
                FTST(ftst);
            }
            else {
                /* Assume it was a minus followed by a one-letter named
                 * subroutine call (or a -bareword), then. */
                DEBUG_T( { PerlIO_printf(Perl_debug_log,
                        "### '-%c' looked like a file test but was not\n",
                        (int) tmp);
                } );
                s = --PL_bufptr;
            }
        }
        tmp = *s++;
        if (*s == tmp) {
            s++;
            if (PL_expect == XOPERATOR)
                TERM(POSTDEC);
            else
                OPERATOR(PREDEC);
        }
        else if (*s == '>') {
            s++;
            s = skipspace(s);
            if (isIDFIRST_lazy_if(s,UTF)) {
                s = force_word(s,METHOD,FALSE,TRUE,FALSE);
                TOKEN(ARROW);
            }
            else if (*s == '$')
                OPERATOR(ARROW);
            else
                TERM(ARROW);
        }
        if (PL_expect == XOPERATOR)
            Aop(OP_SUBTRACT);
        else {
            if (isSPACE(*s) || !isSPACE(*PL_bufptr))
                check_uni();
            OPERATOR('-');              /* unary minus */
        }

    case '+':
        tmp = *s++;
        if (*s == tmp) {
            s++;
            if (PL_expect == XOPERATOR)
                TERM(POSTINC);
            else
                OPERATOR(PREINC);
        }
        if (PL_expect == XOPERATOR)
            Aop(OP_ADD);
        else {
            if (isSPACE(*s) || !isSPACE(*PL_bufptr))
                check_uni();
            OPERATOR('+');
        }

    case '*':
        if (PL_expect != XOPERATOR) {
            s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
            PL_expect = XOPERATOR;
            force_ident(PL_tokenbuf, '*');
            if (!*PL_tokenbuf)
                PREREF('*');
            TERM('*');
        }
        s++;
        if (*s == '*') {
            s++;
            PWop(OP_POW);
        }
        Mop(OP_MULTIPLY);

    case '%':
        if (PL_expect == XOPERATOR) {
            ++s;
            Mop(OP_MODULO);
        }
        PL_tokenbuf[0] = '%';
        s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
        if (!PL_tokenbuf[1]) {
            PREREF('%');
        }
        PL_pending_ident = '%';
        TERM('%');

    case '^':
        s++;
        BOop(OP_BIT_XOR);
    case '[':
        PL_lex_brackets++;
        /* FALL THROUGH */
    case '~':
    case ',':
        tmp = *s++;
        OPERATOR(tmp);
    case ':':
        if (s[1] == ':') {
            len = 0;
            goto just_a_word;
        }
        s++;
        switch (PL_expect) {
            OP *attrs;
        case XOPERATOR:
            if (!PL_in_my || PL_lex_state != LEX_NORMAL)
                break;
            PL_bufptr = s;      /* update in case we back off */
            goto grabattrs;
        case XATTRBLOCK:
            PL_expect = XBLOCK;
            goto grabattrs;
        case XATTRTERM:
            PL_expect = XTERMBLOCK;
         grabattrs:
            s = skipspace(s);
            attrs = Nullop;
            while (isIDFIRST_lazy_if(s,UTF)) {
                d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
                if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
                    if (tmp < 0) tmp = -tmp;
                    switch (tmp) {
                    case KEY_or:
                    case KEY_and:
                    case KEY_err:
                    case KEY_for:
                    case KEY_unless:
                    case KEY_if:
                    case KEY_while:
                    case KEY_until:
                        goto got_attrs;
                    default:
                        break;
                    }
                }
                if (*d == '(') {
                    d = scan_str(d,TRUE,TRUE);
                    if (!d) {
                        /* MUST advance bufptr here to avoid bogus
                           "at end of line" context messages from yyerror().
                         */
                        PL_bufptr = s + len;
                        yyerror("Unterminated attribute parameter in attribute list");
                        if (attrs)
                            op_free(attrs);
                        return REPORT(0);       /* EOF indicator */
                    }
                }
                if (PL_lex_stuff) {
                    SV *sv = newSVpvn(s, len);
                    sv_catsv(sv, PL_lex_stuff);
                    attrs = append_elem(OP_LIST, attrs,
                                        newSVOP(OP_CONST, 0, sv));
                    SvREFCNT_dec(PL_lex_stuff);
                    PL_lex_stuff = Nullsv;
                }
                else {
                    if (len == 6 && strnEQ(s, "unique", len)) {
                        if (PL_in_my == KEY_our)
#ifdef USE_ITHREADS
                            GvUNIQUE_on(cGVOPx_gv(yylval.opval));
#else
                            ; /* skip to avoid loading attributes.pm */
#endif
                        else 
                            Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
                    }

                    /* NOTE: any CV attrs applied here need to be part of
                       the CVf_BUILTIN_ATTRS define in cv.h! */
                    else if (!PL_in_my && len == 6 && strnEQ(s, "lvalue", len))
                        CvLVALUE_on(PL_compcv);
                    else if (!PL_in_my && len == 6 && strnEQ(s, "locked", len))
                        CvLOCKED_on(PL_compcv);
                    else if (!PL_in_my && len == 6 && strnEQ(s, "method", len))
                        CvMETHOD_on(PL_compcv);
                    else if (!PL_in_my && len == 9 && strnEQ(s, "assertion", len))
                        CvASSERTION_on(PL_compcv);
                    /* After we've set the flags, it could be argued that
                       we don't need to do the attributes.pm-based setting
                       process, and shouldn't bother appending recognized
                       flags.  To experiment with that, uncomment the
                       following "else".  (Note that's already been
                       uncommented.  That keeps the above-applied built-in
                       attributes from being intercepted (and possibly
                       rejected) by a package's attribute routines, but is
                       justified by the performance win for the common case
                       of applying only built-in attributes.) */
                    else
                        attrs = append_elem(OP_LIST, attrs,
                                            newSVOP(OP_CONST, 0,
                                                    newSVpvn(s, len)));
                }
                s = skipspace(d);
                if (*s == ':' && s[1] != ':')
                    s = skipspace(s+1);
                else if (s == d)
                    break;      /* require real whitespace or :'s */
            }
            tmp = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
            if (*s != ';' && *s != '}' && *s != tmp && (tmp != '=' || *s != ')')) {
                char q = ((*s == '\'') ? '"' : '\'');
                /* If here for an expression, and parsed no attrs, back off. */
                if (tmp == '=' && !attrs) {
                    s = PL_bufptr;
                    break;
                }
                /* MUST advance bufptr here to avoid bogus "at end of line"
                   context messages from yyerror().
                 */
                PL_bufptr = s;
                if (!*s)
                    yyerror("Unterminated attribute list");
                else
                    yyerror(Perl_form(aTHX_ "Invalid separator character %c%c%c in attribute list",
                                      q, *s, q));
                if (attrs)
                    op_free(attrs);
                OPERATOR(':');
            }
        got_attrs:
            if (attrs) {
                PL_nextval[PL_nexttoke].opval = attrs;
                force_next(THING);
            }
            TOKEN(COLONATTR);
        }
        OPERATOR(':');
    case '(':
        s++;
        if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
            PL_oldbufptr = PL_oldoldbufptr;             /* allow print(STDOUT 123) */
        else
            PL_expect = XTERM;
        s = skipspace(s);
        TOKEN('(');
    case ';':
        CLINE;
        tmp = *s++;
        OPERATOR(tmp);
    case ')':
        tmp = *s++;
        s = skipspace(s);
        if (*s == '{')
            PREBLOCK(tmp);
        TERM(tmp);
    case ']':
        s++;
        if (PL_lex_brackets <= 0)
            yyerror("Unmatched right square bracket");
        else
            --PL_lex_brackets;
        if (PL_lex_state == LEX_INTERPNORMAL) {
            if (PL_lex_brackets == 0) {
                if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
                    PL_lex_state = LEX_INTERPEND;
            }
        }
        TERM(']');
    case '{':
      leftbracket:
        s++;
        if (PL_lex_brackets > 100) {
            Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
        }
        switch (PL_expect) {
        case XTERM:
            if (PL_lex_formbrack) {
                s--;
                PRETERMBLOCK(DO);
            }
            if (PL_oldoldbufptr == PL_last_lop)
                PL_lex_brackstack[PL_lex_brackets++] = XTERM;
            else
                PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
            OPERATOR(HASHBRACK);
        case XOPERATOR:
            while (s < PL_bufend && SPACE_OR_TAB(*s))
                s++;
            d = s;
            PL_tokenbuf[0] = '\0';
            if (d < PL_bufend && *d == '-') {
                PL_tokenbuf[0] = '-';
                d++;
                while (d < PL_bufend && SPACE_OR_TAB(*d))
                    d++;
            }
            if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
                d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
                              FALSE, &len);
                while (d < PL_bufend && SPACE_OR_TAB(*d))
                    d++;
                if (*d == '}') {
                    char minus = (PL_tokenbuf[0] == '-');
                    s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
                    if (minus)
                        force_next('-');
                }
            }
            /* FALL THROUGH */
        case XATTRBLOCK:
        case XBLOCK:
            PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
            PL_expect = XSTATE;
            break;
        case XATTRTERM:
        case XTERMBLOCK:
            PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
            PL_expect = XSTATE;
            break;
        default: {
                char *t;
                if (PL_oldoldbufptr == PL_last_lop)
                    PL_lex_brackstack[PL_lex_brackets++] = XTERM;
                else
                    PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
                s = skipspace(s);
                if (*s == '}') {
                    if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
                        PL_expect = XTERM;
                        /* This hack is to get the ${} in the message. */
                        PL_bufptr = s+1;
                        yyerror("syntax error");
                        break;
                    }
                    OPERATOR(HASHBRACK);
                }
                /* This hack serves to disambiguate a pair of curlies
                 * as being a block or an anon hash.  Normally, expectation
                 * determines that, but in cases where we're not in a
                 * position to expect anything in particular (like inside
                 * eval"") we have to resolve the ambiguity.  This code
                 * covers the case where the first term in the curlies is a
                 * quoted string.  Most other cases need to be explicitly
                 * disambiguated by prepending a `+' before the opening
                 * curly in order to force resolution as an anon hash.
                 *
                 * XXX should probably propagate the outer expectation
                 * into eval"" to rely less on this hack, but that could
                 * potentially break current behavior of eval"".
                 * GSAR 97-07-21
                 */
                t = s;
                if (*s == '\'' || *s == '"' || *s == '`') {
                    /* common case: get past first string, handling escapes */
                    for (t++; t < PL_bufend && *t != *s;)
                        if (*t++ == '\\' && (*t == '\\' || *t == *s))
                            t++;
                    t++;
                }
                else if (*s == 'q') {
                    if (++t < PL_bufend
                        && (!isALNUM(*t)
                            || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
                                && !isALNUM(*t))))
                    {
                        /* skip q//-like construct */
                        char *tmps;
                        char open, close, term;
                        I32 brackets = 1;

                        while (t < PL_bufend && isSPACE(*t))
                            t++;
                        /* check for q => */
                        if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
                            OPERATOR(HASHBRACK);
                        }
                        term = *t;
                        open = term;
                        if (term && (tmps = strchr("([{< )]}> )]}>",term)))
                            term = tmps[5];
                        close = term;
                        if (open == close)
                            for (t++; t < PL_bufend; t++) {
                                if (*t == '\\' && t+1 < PL_bufend && open != '\\')
                                    t++;
                                else if (*t == open)
                                    break;
                            }
                        else {
                            for (t++; t < PL_bufend; t++) {
                                if (*t == '\\' && t+1 < PL_bufend)
                                    t++;
                                else if (*t == close && --brackets <= 0)
                                    break;
                                else if (*t == open)
                                    brackets++;
                            }
                        }
                        t++;
                    }
                    else
                        /* skip plain q word */
                        while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
                             t += UTF8SKIP(t);
                }
                else if (isALNUM_lazy_if(t,UTF)) {
                    t += UTF8SKIP(t);
                    while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
                         t += UTF8SKIP(t);
                }
                while (t < PL_bufend && isSPACE(*t))
                    t++;
                /* if comma follows first term, call it an anon hash */
                /* XXX it could be a comma expression with loop modifiers */
                if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
                                   || (*t == '=' && t[1] == '>')))
                    OPERATOR(HASHBRACK);
                if (PL_expect == XREF)
                    PL_expect = XTERM;
                else {
                    PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
                    PL_expect = XSTATE;
                }
            }
            break;
        }
        yylval.ival = CopLINE(PL_curcop);
        if (isSPACE(*s) || *s == '#')
            PL_copline = NOLINE;   /* invalidate current command line number */
        TOKEN('{');
    case '}':
      rightbracket:
        s++;
        if (PL_lex_brackets <= 0)
            yyerror("Unmatched right curly bracket");
        else
            PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
        if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
            PL_lex_formbrack = 0;
        if (PL_lex_state == LEX_INTERPNORMAL) {
            if (PL_lex_brackets == 0) {
                if (PL_expect & XFAKEBRACK) {
                    PL_expect &= XENUMMASK;
                    PL_lex_state = LEX_INTERPEND;
                    PL_bufptr = s;
                    return yylex();     /* ignore fake brackets */
                }
                if (*s == '-' && s[1] == '>')
                    PL_lex_state = LEX_INTERPENDMAYBE;
                else if (*s != '[' && *s != '{')
                    PL_lex_state = LEX_INTERPEND;
            }
        }
        if (PL_expect & XFAKEBRACK) {
            PL_expect &= XENUMMASK;
            PL_bufptr = s;
            return yylex();             /* ignore fake brackets */
        }
        force_next('}');
        TOKEN(';');
    case '&':
        s++;
        tmp = *s++;
        if (tmp == '&')
            AOPERATOR(ANDAND);
        s--;
        if (PL_expect == XOPERATOR) {
            if (ckWARN(WARN_SEMICOLON)
                && isIDFIRST_lazy_if(s,UTF) && PL_bufptr == PL_linestart)
            {
                CopLINE_dec(PL_curcop);
                Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
                CopLINE_inc(PL_curcop);
            }
            BAop(OP_BIT_AND);
        }

        s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
        if (*PL_tokenbuf) {
            PL_expect = XOPERATOR;
            force_ident(PL_tokenbuf, '&');
        }
        else
            PREREF('&');
        yylval.ival = (OPpENTERSUB_AMPER<<8);
        TERM('&');

    case '|':
        s++;
        tmp = *s++;
        if (tmp == '|')
            AOPERATOR(OROR);
        s--;
        BOop(OP_BIT_OR);
    case '=':
        s++;
        tmp = *s++;
        if (tmp == '=')
            Eop(OP_EQ);
        if (tmp == '>')
            OPERATOR(',');
        if (tmp == '~')
            PMop(OP_MATCH);
        if (ckWARN(WARN_SYNTAX) && tmp && isSPACE(*s) && strchr("+-*/%.^&|<",tmp))
            Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Reversed %c= operator",(int)tmp);
        s--;
        if (PL_expect == XSTATE && isALPHA(tmp) &&
                (s == PL_linestart+1 || s[-2] == '\n') )
        {
            if (PL_in_eval && !PL_rsfp) {
                d = PL_bufend;
                while (s < d) {
                    if (*s++ == '\n') {
                        incline(s);
                        if (strnEQ(s,"=cut",4)) {
                            s = strchr(s,'\n');
                            if (s)
                                s++;
                            else
                                s = d;
                            incline(s);
                            goto retry;
                        }
                    }
                }
                goto retry;
            }
            s = PL_bufend;
            PL_doextract = TRUE;
            goto retry;
        }
        if (PL_lex_brackets < PL_lex_formbrack) {
            char *t;
#ifdef PERL_STRICT_CR
            for (t = s; SPACE_OR_TAB(*t); t++) ;
#else
            for (t = s; SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
#endif
            if (*t == '\n' || *t == '#') {
                s--;
                PL_expect = XBLOCK;
                goto leftbracket;
            }
        }
        yylval.ival = 0;
        OPERATOR(ASSIGNOP);
    case '!':
        s++;
        tmp = *s++;
        if (tmp == '=') {
            /* was this !=~ where !~ was meant?
             * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */

            if (*s == '~' && ckWARN(WARN_SYNTAX)) {
                char *t = s+1;

                while (t < PL_bufend && isSPACE(*t))
                    ++t;

                if (*t == '/' || *t == '?' ||
                    ((*t == 'm' || *t == 's' || *t == 'y') && !isALNUM(t[1])) ||
                    (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
                    Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                                "!=~ should be !~");
            }
            Eop(OP_NE);
        }
        if (tmp == '~')
            PMop(OP_NOT);
        s--;
        OPERATOR('!');
    case '<':
        if (PL_expect != XOPERATOR) {
            if (s[1] != '<' && !strchr(s,'>'))
                check_uni();
            if (s[1] == '<')
                s = scan_heredoc(s);
            else
                s = scan_inputsymbol(s);
            TERM(sublex_start());
        }
        s++;
        tmp = *s++;
        if (tmp == '<')
            SHop(OP_LEFT_SHIFT);
        if (tmp == '=') {
            tmp = *s++;
            if (tmp == '>')
                Eop(OP_NCMP);
            s--;
            Rop(OP_LE);
        }
        s--;
        Rop(OP_LT);
    case '>':
        s++;
        tmp = *s++;
        if (tmp == '>')
            SHop(OP_RIGHT_SHIFT);
        if (tmp == '=')
            Rop(OP_GE);
        s--;
        Rop(OP_GT);

    case '$':
        CLINE;

        if (PL_expect == XOPERATOR) {
            if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
                PL_expect = XTERM;
                depcom();
                return REPORT(','); /* grandfather non-comma-format format */
            }
        }

        if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
            PL_tokenbuf[0] = '@';
            s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
                           sizeof PL_tokenbuf - 1, FALSE);
            if (PL_expect == XOPERATOR)
                no_op("Array length", s);
            if (!PL_tokenbuf[1])
                PREREF(DOLSHARP);
            PL_expect = XOPERATOR;
            PL_pending_ident = '#';
            TOKEN(DOLSHARP);
        }

        PL_tokenbuf[0] = '$';
        s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
                       sizeof PL_tokenbuf - 1, FALSE);
        if (PL_expect == XOPERATOR)
            no_op("Scalar", s);
        if (!PL_tokenbuf[1]) {
            if (s == PL_bufend)
                yyerror("Final $ should be \\$ or $name");
            PREREF('$');
        }

        /* This kludge not intended to be bulletproof. */
        if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
            yylval.opval = newSVOP(OP_CONST, 0,
                                   newSViv(PL_compiling.cop_arybase));
            yylval.opval->op_private = OPpCONST_ARYBASE;
            TERM(THING);
        }

        d = s;
        tmp = (I32)*s;
        if (PL_lex_state == LEX_NORMAL)
            s = skipspace(s);

        if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
            char *t;
            if (*s == '[') {
                PL_tokenbuf[0] = '@';
                if (ckWARN(WARN_SYNTAX)) {
                    for(t = s + 1;
                        isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$';
                        t++) ;
                    if (*t++ == ',') {
                        PL_bufptr = skipspace(PL_bufptr);
                        while (t < PL_bufend && *t != ']')
                            t++;
                        Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                                "Multidimensional syntax %.*s not supported",
                                (t - PL_bufptr) + 1, PL_bufptr);
                    }
                }
            }
            else if (*s == '{') {
                PL_tokenbuf[0] = '%';
                if (ckWARN(WARN_SYNTAX) && strEQ(PL_tokenbuf+1, "SIG") &&
                    (t = strchr(s, '}')) && (t = strchr(t, '=')))
                {
                    char tmpbuf[sizeof PL_tokenbuf];
                    STRLEN len;
                    for (t++; isSPACE(*t); t++) ;
                    if (isIDFIRST_lazy_if(t,UTF)) {
                        t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE, &len);
                        for (; isSPACE(*t); t++) ;
                        if (*t == ';' && get_cv(tmpbuf, FALSE))
                            Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                                "You need to quote \"%s\"", tmpbuf);
                    }
                }
            }
        }

        PL_expect = XOPERATOR;
        if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
            bool islop = (PL_last_lop == PL_oldoldbufptr);
            if (!islop || PL_last_lop_op == OP_GREPSTART)
                PL_expect = XOPERATOR;
            else if (strchr("$@\"'`q", *s))
                PL_expect = XTERM;              /* e.g. print $fh "foo" */
            else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
                PL_expect = XTERM;              /* e.g. print $fh &sub */
            else if (isIDFIRST_lazy_if(s,UTF)) {
                char tmpbuf[sizeof PL_tokenbuf];
                scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
                if ((tmp = keyword(tmpbuf, len))) {
                    /* binary operators exclude handle interpretations */
                    switch (tmp) {
                    case -KEY_x:
                    case -KEY_eq:
                    case -KEY_ne:
                    case -KEY_gt:
                    case -KEY_lt:
                    case -KEY_ge:
                    case -KEY_le:
                    case -KEY_cmp:
                        break;
                    default:
                        PL_expect = XTERM;      /* e.g. print $fh length() */
                        break;
                    }
                }
                else {
                    PL_expect = XTERM;          /* e.g. print $fh subr() */
                }
            }
            else if (isDIGIT(*s))
                PL_expect = XTERM;              /* e.g. print $fh 3 */
            else if (*s == '.' && isDIGIT(s[1]))
                PL_expect = XTERM;              /* e.g. print $fh .3 */
            else if ((*s == '?' || *s == '-' || *s == '+')
                     && !isSPACE(s[1]) && s[1] != '=')
                PL_expect = XTERM;              /* e.g. print $fh -1 */
            else if (*s == '/' && !isSPACE(s[1]) && s[1] != '=' && s[1] != '/')
                PL_expect = XTERM;              /* e.g. print $fh /.../
                                                 XXX except DORDOR operator */
            else if (*s == '<' && s[1] == '<' && !isSPACE(s[2]) && s[2] != '=')
                PL_expect = XTERM;              /* print $fh <<"EOF" */
        }
        PL_pending_ident = '$';
        TOKEN('$');

    case '@':
        if (PL_expect == XOPERATOR)
            no_op("Array", s);
        PL_tokenbuf[0] = '@';
        s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
        if (!PL_tokenbuf[1]) {
            PREREF('@');
        }
        if (PL_lex_state == LEX_NORMAL)
            s = skipspace(s);
        if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
            if (*s == '{')
                PL_tokenbuf[0] = '%';

            /* Warn about @ where they meant $. */
            if (ckWARN(WARN_SYNTAX)) {
                if (*s == '[' || *s == '{') {
                    char *t = s + 1;
                    while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
                        t++;
                    if (*t == '}' || *t == ']') {
                        t++;
                        PL_bufptr = skipspace(PL_bufptr);
                        Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                            "Scalar value %.*s better written as $%.*s",
                            t-PL_bufptr, PL_bufptr, t-PL_bufptr-1, PL_bufptr+1);
                    }
                }
            }
        }
        PL_pending_ident = '@';
        TERM('@');

     case '/':                  /* may be division, defined-or, or pattern */
        if (PL_expect == XTERMORDORDOR && s[1] == '/') {
            s += 2;
            AOPERATOR(DORDOR);
        }
     case '?':                  /* may either be conditional or pattern */
         if(PL_expect == XOPERATOR) {
             tmp = *s++;
             if(tmp == '?') {
                  OPERATOR('?');
             }
             else {
                 tmp = *s++;
                 if(tmp == '/') {
                     /* A // operator. */
                    AOPERATOR(DORDOR);
                 }
                 else {
                     s--;
                     Mop(OP_DIVIDE);
                 }
             }
         }
         else {
             /* Disable warning on "study /blah/" */
             if (PL_oldoldbufptr == PL_last_uni
              && (*PL_last_uni != 's' || s - PL_last_uni < 5
                  || memNE(PL_last_uni, "study", 5)
                  || isALNUM_lazy_if(PL_last_uni+5,UTF)
              ))
                 check_uni();
             s = scan_pat(s,OP_MATCH);
             TERM(sublex_start());
         }

    case '.':
        if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
#ifdef PERL_STRICT_CR
            && s[1] == '\n'
#else
            && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
#endif
            && (s == PL_linestart || s[-1] == '\n') )
        {
            PL_lex_formbrack = 0;
            PL_expect = XSTATE;
            goto rightbracket;
        }
        if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
            tmp = *s++;
            if (*s == tmp) {
                s++;
                if (*s == tmp) {
                    s++;
                    yylval.ival = OPf_SPECIAL;
                }
                else
                    yylval.ival = 0;
                OPERATOR(DOTDOT);
            }
            if (PL_expect != XOPERATOR)
                check_uni();
            Aop(OP_CONCAT);
        }
        /* FALL THROUGH */
    case '0': case '1': case '2': case '3': case '4':
    case '5': case '6': case '7': case '8': case '9':
        s = scan_num(s, &yylval);
        DEBUG_T( { PerlIO_printf(Perl_debug_log,
                    "### Saw number in '%s'\n", s);
        } );
        if (PL_expect == XOPERATOR)
            no_op("Number",s);
        TERM(THING);

    case '\'':
        s = scan_str(s,FALSE,FALSE);
        DEBUG_T( { PerlIO_printf(Perl_debug_log,
                    "### Saw string before '%s'\n", s);
        } );
        if (PL_expect == XOPERATOR) {
            if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
                PL_expect = XTERM;
                depcom();
                return REPORT(','); /* grandfather non-comma-format format */
            }
            else
                no_op("String",s);
        }
        if (!s)
            missingterm((char*)0);
        yylval.ival = OP_CONST;
        TERM(sublex_start());

    case '"':
        s = scan_str(s,FALSE,FALSE);
        DEBUG_T( { PerlIO_printf(Perl_debug_log,
                    "### Saw string before '%s'\n", s);
        } );
        if (PL_expect == XOPERATOR) {
            if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
                PL_expect = XTERM;
                depcom();
                return REPORT(','); /* grandfather non-comma-format format */
            }
            else
                no_op("String",s);
        }
        if (!s)
            missingterm((char*)0);
        yylval.ival = OP_CONST;
        for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
            if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
                yylval.ival = OP_STRINGIFY;
                break;
            }
        }
        TERM(sublex_start());

    case '`':
        s = scan_str(s,FALSE,FALSE);
        DEBUG_T( { PerlIO_printf(Perl_debug_log,
                    "### Saw backtick string before '%s'\n", s);
        } );
        if (PL_expect == XOPERATOR)
            no_op("Backticks",s);
        if (!s)
            missingterm((char*)0);
        yylval.ival = OP_BACKTICK;
        set_csh();
        TERM(sublex_start());

    case '\\':
        s++;
        if (ckWARN(WARN_SYNTAX) && PL_lex_inwhat && isDIGIT(*s))
            Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
                        *s, *s);
        if (PL_expect == XOPERATOR)
            no_op("Backslash",s);
        OPERATOR(REFGEN);

    case 'v':
        if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
            char *start = s;
            start++;
            start++;
            while (isDIGIT(*start) || *start == '_')
                start++;
            if (*start == '.' && isDIGIT(start[1])) {
                s = scan_num(s, &yylval);
                TERM(THING);
            }
            /* avoid v123abc() or $h{v1}, allow C<print v10;> */
            else if (!isALPHA(*start) && (PL_expect == XTERM
                        || PL_expect == XREF || PL_expect == XSTATE
                        || PL_expect == XTERMORDORDOR)) {
                char c = *start;
                GV *gv;
                *start = '\0';
                gv = gv_fetchpv(s, FALSE, SVt_PVCV);
                *start = c;
                if (!gv) {
                    s = scan_num(s, &yylval);
                    TERM(THING);
                }
            }
        }
        goto keylookup;
    case 'x':
        if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
            s++;
            Mop(OP_REPEAT);
        }
        goto keylookup;

    case '_':
    case 'a': case 'A':
    case 'b': case 'B':
    case 'c': case 'C':
    case 'd': case 'D':
    case 'e': case 'E':
    case 'f': case 'F':
    case 'g': case 'G':
    case 'h': case 'H':
    case 'i': case 'I':
    case 'j': case 'J':
    case 'k': case 'K':
    case 'l': case 'L':
    case 'm': case 'M':
    case 'n': case 'N':
    case 'o': case 'O':
    case 'p': case 'P':
    case 'q': case 'Q':
    case 'r': case 'R':
    case 's': case 'S':
    case 't': case 'T':
    case 'u': case 'U':
              case 'V':
    case 'w': case 'W':
              case 'X':
    case 'y': case 'Y':
    case 'z': case 'Z':

      keylookup: {
        orig_keyword = 0;
        gv = Nullgv;
        gvp = 0;

        PL_bufptr = s;
        s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);

        /* Some keywords can be followed by any delimiter, including ':' */
        tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
               (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
                             (PL_tokenbuf[0] == 'q' &&
                              strchr("qwxr", PL_tokenbuf[1])))));

        /* x::* is just a word, unless x is "CORE" */
        if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
            goto just_a_word;

        d = s;
        while (d < PL_bufend && isSPACE(*d))
                d++;    /* no comments skipped here, or s### is misparsed */

        /* Is this a label? */
        if (!tmp && PL_expect == XSTATE
              && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
            s = d + 1;
            yylval.pval = savepv(PL_tokenbuf);
            CLINE;
            TOKEN(LABEL);
        }

        /* Check for keywords */
        tmp = keyword(PL_tokenbuf, len);

        /* Is this a word before a => operator? */
        if (*d == '=' && d[1] == '>') {
            CLINE;
            yylval.opval
                = (OP*)newSVOP(OP_CONST, 0,
                               S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
            yylval.opval->op_private = OPpCONST_BARE;
            TERM(WORD);
        }

        if (tmp < 0) {                  /* second-class keyword? */
            GV *ogv = Nullgv;   /* override (winner) */
            GV *hgv = Nullgv;   /* hidden (loser) */
            if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
                CV *cv;
                if ((gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV)) &&
                    (cv = GvCVu(gv)))
                {
                    if (GvIMPORTED_CV(gv))
                        ogv = gv;
                    else if (! CvMETHOD(cv))
                        hgv = gv;
                }
                if (!ogv &&
                    (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
                    (gv = *gvp) != (GV*)&PL_sv_undef &&
                    GvCVu(gv) && GvIMPORTED_CV(gv))
                {
                    ogv = gv;
                }
            }
            if (ogv) {
                orig_keyword = tmp;
                tmp = 0;                /* overridden by import or by GLOBAL */
            }
            else if (gv && !gvp
                     && -tmp==KEY_lock  /* XXX generalizable kludge */
                     && GvCVu(gv)
                     && !hv_fetch(GvHVn(PL_incgv), "Thread.pm", 9, FALSE))
            {
                tmp = 0;                /* any sub overrides "weak" keyword */
            }
            else if (gv && !gvp
                    && tmp == -KEY_err
                    && GvCVu(gv)
                    && PL_expect != XOPERATOR
                    && PL_expect != XTERMORDORDOR)
            {
                /* any sub overrides the "err" keyword, except when really an
                 * operator is expected */
                tmp = 0;
            }
            else {                      /* no override */
                tmp = -tmp;
                if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
                    Perl_warner(aTHX_ packWARN(WARN_MISC),
                            "dump() better written as CORE::dump()");
                }
                gv = Nullgv;
                gvp = 0;
                if (ckWARN(WARN_AMBIGUOUS) && hgv
                    && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
                    Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
                        "Ambiguous call resolved as CORE::%s(), %s",
                         GvENAME(hgv), "qualify as such or use &");
            }
        }

      reserved_word:
        switch (tmp) {

        default:                        /* not a keyword */
          just_a_word: {
                SV *sv;
                int pkgname = 0;
                char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);

                /* Get the rest if it looks like a package qualifier */

                if (*s == '\'' || (*s == ':' && s[1] == ':')) {
                    STRLEN morelen;
                    s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
                                  TRUE, &morelen);
                    if (!morelen)
                        Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
                                *s == '\'' ? "'" : "::");
                    len += morelen;
                    pkgname = 1;
                }

                if (PL_expect == XOPERATOR) {
                    if (PL_bufptr == PL_linestart) {
                        CopLINE_dec(PL_curcop);
                        Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
                        CopLINE_inc(PL_curcop);
                    }
                    else
                        no_op("Bareword",s);
                }

                /* Look for a subroutine with this name in current package,
                   unless name is "Foo::", in which case Foo is a bearword
                   (and a package name). */

                if (len > 2 &&
                    PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
                {
                    if (ckWARN(WARN_BAREWORD) && ! gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVHV))
                        Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
                            "Bareword \"%s\" refers to nonexistent package",
                             PL_tokenbuf);
                    len -= 2;
                    PL_tokenbuf[len] = '\0';
                    gv = Nullgv;
                    gvp = 0;
                }
                else {
                    len = 0;
                    if (!gv)
                        gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV);
                }

                /* if we saw a global override before, get the right name */

                if (gvp) {
                    sv = newSVpvn("CORE::GLOBAL::",14);
                    sv_catpv(sv,PL_tokenbuf);
                }
                else {
                    /* If len is 0, newSVpv does strlen(), which is correct.
                       If len is non-zero, then it will be the true length,
                       and so the scalar will be created correctly.  */
                    sv = newSVpv(PL_tokenbuf,len);
                }

                /* Presume this is going to be a bareword of some sort. */

                CLINE;
                yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
                yylval.opval->op_private = OPpCONST_BARE;
                /* UTF-8 package name? */
                if (UTF && !IN_BYTES &&
                    is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
                    SvUTF8_on(sv);

                /* And if "Foo::", then that's what it certainly is. */

                if (len)
                    goto safe_bareword;

                /* See if it's the indirect object for a list operator. */

                if (PL_oldoldbufptr &&
                    PL_oldoldbufptr < PL_bufptr &&
                    (PL_oldoldbufptr == PL_last_lop
                     || PL_oldoldbufptr == PL_last_uni) &&
                    /* NO SKIPSPACE BEFORE HERE! */
                    (PL_expect == XREF ||
                     ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
                {
                    bool immediate_paren = *s == '(';

                    /* (Now we can afford to cross potential line boundary.) */
                    s = skipspace(s);

                    /* Two barewords in a row may indicate method call. */

                    if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') && (tmp=intuit_method(s,gv)))
                        return REPORT(tmp);

                    /* If not a declared subroutine, it's an indirect object. */
                    /* (But it's an indir obj regardless for sort.) */

                    if ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
                         ((!gv || !GvCVu(gv)) &&
                        (PL_last_lop_op != OP_MAPSTART &&
                         PL_last_lop_op != OP_GREPSTART))))
                    {
                        PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
                        goto bareword;
                    }
                }

                PL_expect = XOPERATOR;
                s = skipspace(s);

                /* Is this a word before a => operator? */
                if (*s == '=' && s[1] == '>' && !pkgname) {
                    CLINE;
                    sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
                    if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
                      SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
                    TERM(WORD);
                }

                /* If followed by a paren, it's certainly a subroutine. */
                if (*s == '(') {
                    CLINE;
                    if (gv && GvCVu(gv)) {
                        for (d = s + 1; SPACE_OR_TAB(*d); d++) ;
                        if (*d == ')' && (sv = cv_const_sv(GvCV(gv)))) {
                            s = d + 1;
                            goto its_constant;
                        }
                    }
                    PL_nextval[PL_nexttoke].opval = yylval.opval;
                    PL_expect = XOPERATOR;
                    force_next(WORD);
                    yylval.ival = 0;
                    TOKEN('&');
                }

                /* If followed by var or block, call it a method (unless sub) */

                if ((*s == '$' || *s == '{') && (!gv || !GvCVu(gv))) {
                    PL_last_lop = PL_oldbufptr;
                    PL_last_lop_op = OP_METHOD;
                    PREBLOCK(METHOD);
                }

                /* If followed by a bareword, see if it looks like indir obj. */

                if (!orig_keyword
                        && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
                        && (tmp = intuit_method(s,gv)))
                    return REPORT(tmp);

                /* Not a method, so call it a subroutine (if defined) */

                if (gv && GvCVu(gv)) {
                    CV* cv;
                    if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
                        Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
                                "Ambiguous use of -%s resolved as -&%s()",
                                PL_tokenbuf, PL_tokenbuf);
                    /* Check for a constant sub */
                    cv = GvCV(gv);
                    if ((sv = cv_const_sv(cv))) {
                  its_constant:
                        SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
                        ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc(sv);
                        yylval.opval->op_private = 0;
                        TOKEN(WORD);
                    }

                    /* Resolve to GV now. */
                    op_free(yylval.opval);
                    yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
                    yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
                    PL_last_lop = PL_oldbufptr;
                    PL_last_lop_op = OP_ENTERSUB;
                    /* Is there a prototype? */
                    if (SvPOK(cv)) {
                        STRLEN len;
                        char *proto = SvPV((SV*)cv, len);
                        if (!len)
                            TERM(FUNC0SUB);
                        if (*proto == '$' && proto[1] == '\0')
                            OPERATOR(UNIOPSUB);
                        while (*proto == ';')
                            proto++;
                        if (*proto == '&' && *s == '{') {
                            sv_setpv(PL_subname, PL_curstash ? 
                                        "__ANON__" : "__ANON__::__ANON__");
                            PREBLOCK(LSTOPSUB);
                        }
                    }
                    PL_nextval[PL_nexttoke].opval = yylval.opval;
                    PL_expect = XTERM;
                    force_next(WORD);
                    TOKEN(NOAMP);
                }

                /* Call it a bare word */

                if (PL_hints & HINT_STRICT_SUBS)
                    yylval.opval->op_private |= OPpCONST_STRICT;
                else {
                bareword:
                    if (ckWARN(WARN_RESERVED)) {
                        if (lastchar != '-') {
                            for (d = PL_tokenbuf; *d && isLOWER(*d); d++) ;
                            if (!*d && !gv_stashpv(PL_tokenbuf,FALSE))
                                Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
                                       PL_tokenbuf);
                        }
                    }
                }

            safe_bareword:
                if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
                    && ckWARN_d(WARN_AMBIGUOUS)) {
                    Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
                        "Operator or semicolon missing before %c%s",
                        lastchar, PL_tokenbuf);
                    Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
                        "Ambiguous use of %c resolved as operator %c",
                        lastchar, lastchar);
                }
                TOKEN(WORD);
            }

        case KEY___FILE__:
            yylval.opval = (OP*)newSVOP(OP_CONST, 0,
                                        newSVpv(CopFILE(PL_curcop),0));
            TERM(THING);

        case KEY___LINE__:
            yylval.opval = (OP*)newSVOP(OP_CONST, 0,
                                    Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
            TERM(THING);

        case KEY___PACKAGE__:
            yylval.opval = (OP*)newSVOP(OP_CONST, 0,
                                        (PL_curstash
                                         ? newSVpv(HvNAME(PL_curstash), 0)
                                         : &PL_sv_undef));
            TERM(THING);

        case KEY___DATA__:
        case KEY___END__: {
            GV *gv;

            /*SUPPRESS 560*/
            if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
                char *pname = "main";
                if (PL_tokenbuf[2] == 'D')
                    pname = HvNAME(PL_curstash ? PL_curstash : PL_defstash);
                gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), TRUE, SVt_PVIO);
                GvMULTI_on(gv);
                if (!GvIO(gv))
                    GvIOp(gv) = newIO();
                IoIFP(GvIOp(gv)) = PL_rsfp;
#if defined(HAS_FCNTL) && defined(F_SETFD)
                {
                    int fd = PerlIO_fileno(PL_rsfp);
                    fcntl(fd,F_SETFD,fd >= 3);
                }
#endif
                /* Mark this internal pseudo-handle as clean */
                IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
                if (PL_preprocess)
                    IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
                else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
                    IoTYPE(GvIOp(gv)) = IoTYPE_STD;
                else
                    IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
#if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
                /* if the script was opened in binmode, we need to revert
                 * it to text mode for compatibility; but only iff it has CRs
                 * XXX this is a questionable hack at best. */
                if (PL_bufend-PL_bufptr > 2
                    && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
                {
                    Off_t loc = 0;
                    if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
                        loc = PerlIO_tell(PL_rsfp);
                        (void)PerlIO_seek(PL_rsfp, 0L, 0);
                    }
#ifdef NETWARE
                        if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
#else
                    if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
#endif  /* NETWARE */
#ifdef PERLIO_IS_STDIO /* really? */
#  if defined(__BORLANDC__)
                        /* XXX see note in do_binmode() */
                        ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
#  endif
#endif
                        if (loc > 0)
                            PerlIO_seek(PL_rsfp, loc, 0);
                    }
                }
#endif
#ifdef PERLIO_LAYERS
                if (!IN_BYTES) {
                    if (UTF)
                        PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
                    else if (PL_encoding) {
                        SV *name;
                        dSP;
                        ENTER;
                        SAVETMPS;
                        PUSHMARK(sp);
                        EXTEND(SP, 1);
                        XPUSHs(PL_encoding);
                        PUTBACK;
                        call_method("name", G_SCALAR);
                        SPAGAIN;
                        name = POPs;
                        PUTBACK;
                        PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, 
                                            Perl_form(aTHX_ ":encoding(%"SVf")",
                                                      name));
                        FREETMPS;
                        LEAVE;
                    }
                }
#endif
                PL_rsfp = Nullfp;
            }
            goto fake_eof;
        }

        case KEY_AUTOLOAD:
        case KEY_DESTROY:
        case KEY_BEGIN:
        case KEY_CHECK:
        case KEY_INIT:
        case KEY_END:
            if (PL_expect == XSTATE) {
                s = PL_bufptr;
                goto really_sub;
            }
            goto just_a_word;

        case KEY_CORE:
            if (*s == ':' && s[1] == ':') {
                s += 2;
                d = s;
                s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
                if (!(tmp = keyword(PL_tokenbuf, len)))
                    Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
                if (tmp < 0)
                    tmp = -tmp;
                goto reserved_word;
            }
            goto just_a_word;

        case KEY_abs:
            UNI(OP_ABS);

        case KEY_alarm:
            UNI(OP_ALARM);

        case KEY_accept:
            LOP(OP_ACCEPT,XTERM);

        case KEY_and:
            OPERATOR(ANDOP);

        case KEY_atan2:
            LOP(OP_ATAN2,XTERM);

        case KEY_bind:
            LOP(OP_BIND,XTERM);

        case KEY_binmode:
            LOP(OP_BINMODE,XTERM);

        case KEY_bless:
            LOP(OP_BLESS,XTERM);

        case KEY_chop:
            UNI(OP_CHOP);

        case KEY_continue:
            PREBLOCK(CONTINUE);

        case KEY_chdir:
            (void)gv_fetchpv("ENV",TRUE, SVt_PVHV);     /* may use HOME */
            UNI(OP_CHDIR);

        case KEY_close:
            UNI(OP_CLOSE);

        case KEY_closedir:
            UNI(OP_CLOSEDIR);

        case KEY_cmp:
            Eop(OP_SCMP);

        case KEY_caller:
            UNI(OP_CALLER);

        case KEY_crypt:
#ifdef FCRYPT
            if (!PL_cryptseen) {
                PL_cryptseen = TRUE;
                init_des();
            }
#endif
            LOP(OP_CRYPT,XTERM);

        case KEY_chmod:
            LOP(OP_CHMOD,XTERM);

        case KEY_chown:
            LOP(OP_CHOWN,XTERM);

        case KEY_connect:
            LOP(OP_CONNECT,XTERM);

        case KEY_chr:
            UNI(OP_CHR);

        case KEY_cos:
            UNI(OP_COS);

        case KEY_chroot:
            UNI(OP_CHROOT);

        case KEY_do:
            s = skipspace(s);
            if (*s == '{')
                PRETERMBLOCK(DO);
            if (*s != '\'')
                s = force_word(s,WORD,TRUE,TRUE,FALSE);
            OPERATOR(DO);

        case KEY_die:
            PL_hints |= HINT_BLOCK_SCOPE;
            LOP(OP_DIE,XTERM);

        case KEY_defined:
            UNI(OP_DEFINED);

        case KEY_delete:
            UNI(OP_DELETE);

        case KEY_dbmopen:
            gv_fetchpv("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
            LOP(OP_DBMOPEN,XTERM);

        case KEY_dbmclose:
            UNI(OP_DBMCLOSE);

        case KEY_dump:
            s = force_word(s,WORD,TRUE,FALSE,FALSE);
            LOOPX(OP_DUMP);

        case KEY_else:
            PREBLOCK(ELSE);

        case KEY_elsif:
            yylval.ival = CopLINE(PL_curcop);
            OPERATOR(ELSIF);

        case KEY_eq:
            Eop(OP_SEQ);

        case KEY_exists:
            UNI(OP_EXISTS);
        
        case KEY_exit:
            UNI(OP_EXIT);

        case KEY_eval:
            s = skipspace(s);
            PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
            UNIBRACK(OP_ENTEREVAL);

        case KEY_eof:
            UNI(OP_EOF);

        case KEY_err:
            OPERATOR(DOROP);

        case KEY_exp:
            UNI(OP_EXP);

        case KEY_each:
            UNI(OP_EACH);

        case KEY_exec:
            set_csh();
            LOP(OP_EXEC,XREF);

        case KEY_endhostent:
            FUN0(OP_EHOSTENT);

        case KEY_endnetent:
            FUN0(OP_ENETENT);

        case KEY_endservent:
            FUN0(OP_ESERVENT);

        case KEY_endprotoent:
            FUN0(OP_EPROTOENT);

        case KEY_endpwent:
            FUN0(OP_EPWENT);

        case KEY_endgrent:
            FUN0(OP_EGRENT);

        case KEY_for:
        case KEY_foreach:
            yylval.ival = CopLINE(PL_curcop);
            s = skipspace(s);
            if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
                char *p = s;
                if ((PL_bufend - p) >= 3 &&
                    strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
                    p += 2;
                else if ((PL_bufend - p) >= 4 &&
                    strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
                    p += 3;
                p = skipspace(p);
                if (isIDFIRST_lazy_if(p,UTF)) {
                    p = scan_ident(p, PL_bufend,
                        PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
                    p = skipspace(p);
                }
                if (*p != '$')
                    Perl_croak(aTHX_ "Missing $ on loop variable");
            }
            OPERATOR(FOR);

        case KEY_formline:
            LOP(OP_FORMLINE,XTERM);

        case KEY_fork:
            FUN0(OP_FORK);

        case KEY_fcntl:
            LOP(OP_FCNTL,XTERM);

        case KEY_fileno:
            UNI(OP_FILENO);

        case KEY_flock:
            LOP(OP_FLOCK,XTERM);

        case KEY_gt:
            Rop(OP_SGT);

        case KEY_ge:
            Rop(OP_SGE);

        case KEY_grep:
            LOP(OP_GREPSTART, XREF);

        case KEY_goto:
            s = force_word(s,WORD,TRUE,FALSE,FALSE);
            LOOPX(OP_GOTO);

        case KEY_gmtime:
            UNI(OP_GMTIME);

        case KEY_getc:
            UNIDOR(OP_GETC);

        case KEY_getppid:
            FUN0(OP_GETPPID);

        case KEY_getpgrp:
            UNI(OP_GETPGRP);

        case KEY_getpriority:
            LOP(OP_GETPRIORITY,XTERM);

        case KEY_getprotobyname:
            UNI(OP_GPBYNAME);

        case KEY_getprotobynumber:
            LOP(OP_GPBYNUMBER,XTERM);

        case KEY_getprotoent:
            FUN0(OP_GPROTOENT);

        case KEY_getpwent:
            FUN0(OP_GPWENT);

        case KEY_getpwnam:
            UNI(OP_GPWNAM);

        case KEY_getpwuid:
            UNI(OP_GPWUID);

        case KEY_getpeername:
            UNI(OP_GETPEERNAME);

        case KEY_gethostbyname:
            UNI(OP_GHBYNAME);

        case KEY_gethostbyaddr:
            LOP(OP_GHBYADDR,XTERM);

        case KEY_gethostent:
            FUN0(OP_GHOSTENT);

        case KEY_getnetbyname:
            UNI(OP_GNBYNAME);

        case KEY_getnetbyaddr:
            LOP(OP_GNBYADDR,XTERM);

        case KEY_getnetent:
            FUN0(OP_GNETENT);

        case KEY_getservbyname:
            LOP(OP_GSBYNAME,XTERM);

        case KEY_getservbyport:
            LOP(OP_GSBYPORT,XTERM);

        case KEY_getservent:
            FUN0(OP_GSERVENT);

        case KEY_getsockname:
            UNI(OP_GETSOCKNAME);

        case KEY_getsockopt:
            LOP(OP_GSOCKOPT,XTERM);

        case KEY_getgrent:
            FUN0(OP_GGRENT);

        case KEY_getgrnam:
            UNI(OP_GGRNAM);

        case KEY_getgrgid:
            UNI(OP_GGRGID);

        case KEY_getlogin:
            FUN0(OP_GETLOGIN);

        case KEY_glob:
            set_csh();
            LOP(OP_GLOB,XTERM);

        case KEY_hex:
            UNI(OP_HEX);

        case KEY_if:
            yylval.ival = CopLINE(PL_curcop);
            OPERATOR(IF);

        case KEY_index:
            LOP(OP_INDEX,XTERM);

        case KEY_int:
            UNI(OP_INT);

        case KEY_ioctl:
            LOP(OP_IOCTL,XTERM);

        case KEY_join:
            LOP(OP_JOIN,XTERM);

        case KEY_keys:
            UNI(OP_KEYS);

        case KEY_kill:
            LOP(OP_KILL,XTERM);

        case KEY_last:
            s = force_word(s,WORD,TRUE,FALSE,FALSE);
            LOOPX(OP_LAST);
        
        case KEY_lc:
            UNI(OP_LC);

        case KEY_lcfirst:
            UNI(OP_LCFIRST);

        case KEY_local:
            yylval.ival = 0;
            OPERATOR(LOCAL);

        case KEY_length:
            UNI(OP_LENGTH);

        case KEY_lt:
            Rop(OP_SLT);

        case KEY_le:
            Rop(OP_SLE);

        case KEY_localtime:
            UNI(OP_LOCALTIME);

        case KEY_log:
            UNI(OP_LOG);

        case KEY_link:
            LOP(OP_LINK,XTERM);

        case KEY_listen:
            LOP(OP_LISTEN,XTERM);

        case KEY_lock:
            UNI(OP_LOCK);

        case KEY_lstat:
            UNI(OP_LSTAT);

        case KEY_m:
            s = scan_pat(s,OP_MATCH);
            TERM(sublex_start());

        case KEY_map:
            LOP(OP_MAPSTART, XREF);

        case KEY_mkdir:
            LOP(OP_MKDIR,XTERM);

        case KEY_msgctl:
            LOP(OP_MSGCTL,XTERM);

        case KEY_msgget:
            LOP(OP_MSGGET,XTERM);

        case KEY_msgrcv:
            LOP(OP_MSGRCV,XTERM);

        case KEY_msgsnd:
            LOP(OP_MSGSND,XTERM);

        case KEY_our:
        case KEY_my:
            PL_in_my = tmp;
            s = skipspace(s);
            if (isIDFIRST_lazy_if(s,UTF)) {
                s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
                if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
                    goto really_sub;
                PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
                if (!PL_in_my_stash) {
                    char tmpbuf[1024];
                    PL_bufptr = s;
                    sprintf(tmpbuf, "No such class %.1000s", PL_tokenbuf);
                    yyerror(tmpbuf);
                }
            }
            yylval.ival = 1;
            OPERATOR(MY);

        case KEY_next:
            s = force_word(s,WORD,TRUE,FALSE,FALSE);
            LOOPX(OP_NEXT);

        case KEY_ne:
            Eop(OP_SNE);

        case KEY_no:
            if (PL_expect != XSTATE)
                yyerror("\"no\" not allowed in expression");
            s = force_word(s,WORD,FALSE,TRUE,FALSE);
            s = force_version(s, FALSE);
            yylval.ival = 0;
            OPERATOR(USE);

        case KEY_not:
            if (*s == '(' || (s = skipspace(s), *s == '('))
                FUN1(OP_NOT);
            else
                OPERATOR(NOTOP);

        case KEY_open:
            s = skipspace(s);
            if (isIDFIRST_lazy_if(s,UTF)) {
                char *t;
                for (d = s; isALNUM_lazy_if(d,UTF); d++) ;
                for (t=d; *t && isSPACE(*t); t++) ;
                if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
                    /* [perl #16184] */
                    && !(t[0] == '=' && t[1] == '>')
                ) {
                    Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
                           "Precedence problem: open %.*s should be open(%.*s)",
                            d - s, s, d - s, s);
                }
            }
            LOP(OP_OPEN,XTERM);

        case KEY_or:
            yylval.ival = OP_OR;
            OPERATOR(OROP);

        case KEY_ord:
            UNI(OP_ORD);

        case KEY_oct:
            UNI(OP_OCT);

        case KEY_opendir:
            LOP(OP_OPEN_DIR,XTERM);

        case KEY_print:
            checkcomma(s,PL_tokenbuf,"filehandle");
            LOP(OP_PRINT,XREF);

        case KEY_printf:
            checkcomma(s,PL_tokenbuf,"filehandle");
            LOP(OP_PRTF,XREF);

        case KEY_prototype:
            UNI(OP_PROTOTYPE);

        case KEY_push:
            LOP(OP_PUSH,XTERM);

        case KEY_pop:
            UNIDOR(OP_POP);

        case KEY_pos:
            UNIDOR(OP_POS);
        
        case KEY_pack:
            LOP(OP_PACK,XTERM);

        case KEY_package:
            s = force_word(s,WORD,FALSE,TRUE,FALSE);
            OPERATOR(PACKAGE);

        case KEY_pipe:
            LOP(OP_PIPE_OP,XTERM);

        case KEY_q:
            s = scan_str(s,FALSE,FALSE);
            if (!s)
                missingterm((char*)0);
            yylval.ival = OP_CONST;
            TERM(sublex_start());

        case KEY_quotemeta:
            UNI(OP_QUOTEMETA);

        case KEY_qw:
            s = scan_str(s,FALSE,FALSE);
            if (!s)
                missingterm((char*)0);
            force_next(')');
            if (SvCUR(PL_lex_stuff)) {
                OP *words = Nullop;
                int warned = 0;
                d = SvPV_force(PL_lex_stuff, len);
                while (len) {
                    SV *sv;
                    for (; isSPACE(*d) && len; --len, ++d) ;
                    if (len) {
                        char *b = d;
                        if (!warned && ckWARN(WARN_QW)) {
                            for (; !isSPACE(*d) && len; --len, ++d) {
                                if (*d == ',') {
                                    Perl_warner(aTHX_ packWARN(WARN_QW),
                                        "Possible attempt to separate words with commas");
                                    ++warned;
                                }
                                else if (*d == '#') {
                                    Perl_warner(aTHX_ packWARN(WARN_QW),
                                        "Possible attempt to put comments in qw() list");
                                    ++warned;
                                }
                            }
                        }
                        else {
                            for (; !isSPACE(*d) && len; --len, ++d) ;
                        }
                        sv = newSVpvn(b, d-b);
                        if (DO_UTF8(PL_lex_stuff))
                            SvUTF8_on(sv);
                        words = append_elem(OP_LIST, words,
                                            newSVOP(OP_CONST, 0, tokeq(sv)));
                    }
                }
                if (words) {
                    PL_nextval[PL_nexttoke].opval = words;
                    force_next(THING);
                }
            }
            if (PL_lex_stuff) {
                SvREFCNT_dec(PL_lex_stuff);
                PL_lex_stuff = Nullsv;
            }
            PL_expect = XTERM;
            TOKEN('(');

        case KEY_qq:
            s = scan_str(s,FALSE,FALSE);
            if (!s)
                missingterm((char*)0);
            yylval.ival = OP_STRINGIFY;
            if (SvIVX(PL_lex_stuff) == '\'')
                SvIVX(PL_lex_stuff) = 0;        /* qq'$foo' should intepolate */
            TERM(sublex_start());

        case KEY_qr:
            s = scan_pat(s,OP_QR);
            TERM(sublex_start());

        case KEY_qx:
            s = scan_str(s,FALSE,FALSE);
            if (!s)
                missingterm((char*)0);
            yylval.ival = OP_BACKTICK;
            set_csh();
            TERM(sublex_start());

        case KEY_return:
            OLDLOP(OP_RETURN);

        case KEY_require:
            s = skipspace(s);
            if (isDIGIT(*s)) {
                s = force_version(s, FALSE);
            }
            else if (*s != 'v' || !isDIGIT(s[1])
                    || (s = force_version(s, TRUE), *s == 'v'))
            {
                *PL_tokenbuf = '\0';
                s = force_word(s,WORD,TRUE,TRUE,FALSE);
                if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
                    gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
                else if (*s == '<')
                    yyerror("<> should be quotes");
            }
            UNI(OP_REQUIRE);

        case KEY_reset:
            UNI(OP_RESET);

        case KEY_redo:
            s = force_word(s,WORD,TRUE,FALSE,FALSE);
            LOOPX(OP_REDO);

        case KEY_rename:
            LOP(OP_RENAME,XTERM);

        case KEY_rand:
            UNI(OP_RAND);

        case KEY_rmdir:
            UNI(OP_RMDIR);

        case KEY_rindex:
            LOP(OP_RINDEX,XTERM);

        case KEY_read:
            LOP(OP_READ,XTERM);

        case KEY_readdir:
            UNI(OP_READDIR);

        case KEY_readline:
            set_csh();
            UNIDOR(OP_READLINE);

        case KEY_readpipe:
            set_csh();
            UNI(OP_BACKTICK);

        case KEY_rewinddir:
            UNI(OP_REWINDDIR);

        case KEY_recv:
            LOP(OP_RECV,XTERM);

        case KEY_reverse:
            LOP(OP_REVERSE,XTERM);

        case KEY_readlink:
            UNIDOR(OP_READLINK);

        case KEY_ref:
            UNI(OP_REF);

        case KEY_s:
            s = scan_subst(s);
            if (yylval.opval)
                TERM(sublex_start());
            else
                TOKEN(1);       /* force error */

        case KEY_chomp:
            UNI(OP_CHOMP);
        
        case KEY_scalar:
            UNI(OP_SCALAR);

        case KEY_select:
            LOP(OP_SELECT,XTERM);

        case KEY_seek:
            LOP(OP_SEEK,XTERM);

        case KEY_semctl:
            LOP(OP_SEMCTL,XTERM);

        case KEY_semget:
            LOP(OP_SEMGET,XTERM);

        case KEY_semop:
            LOP(OP_SEMOP,XTERM);

        case KEY_send:
            LOP(OP_SEND,XTERM);

        case KEY_setpgrp:
            LOP(OP_SETPGRP,XTERM);

        case KEY_setpriority:
            LOP(OP_SETPRIORITY,XTERM);

        case KEY_sethostent:
            UNI(OP_SHOSTENT);

        case KEY_setnetent:
            UNI(OP_SNETENT);

        case KEY_setservent:
            UNI(OP_SSERVENT);

        case KEY_setprotoent:
            UNI(OP_SPROTOENT);

        case KEY_setpwent:
            FUN0(OP_SPWENT);

        case KEY_setgrent:
            FUN0(OP_SGRENT);

        case KEY_seekdir:
            LOP(OP_SEEKDIR,XTERM);

        case KEY_setsockopt:
            LOP(OP_SSOCKOPT,XTERM);

        case KEY_shift:
            UNIDOR(OP_SHIFT);

        case KEY_shmctl:
            LOP(OP_SHMCTL,XTERM);

        case KEY_shmget:
            LOP(OP_SHMGET,XTERM);

        case KEY_shmread:
            LOP(OP_SHMREAD,XTERM);

        case KEY_shmwrite:
            LOP(OP_SHMWRITE,XTERM);

        case KEY_shutdown:
            LOP(OP_SHUTDOWN,XTERM);

        case KEY_sin:
            UNI(OP_SIN);

        case KEY_sleep:
            UNI(OP_SLEEP);

        case KEY_socket:
            LOP(OP_SOCKET,XTERM);

        case KEY_socketpair:
            LOP(OP_SOCKPAIR,XTERM);

        case KEY_sort:
            checkcomma(s,PL_tokenbuf,"subroutine name");
            s = skipspace(s);
            if (*s == ';' || *s == ')')         /* probably a close */
                Perl_croak(aTHX_ "sort is now a reserved word");
            PL_expect = XTERM;
            s = force_word(s,WORD,TRUE,TRUE,FALSE);
            LOP(OP_SORT,XREF);

        case KEY_split:
            LOP(OP_SPLIT,XTERM);

        case KEY_sprintf:
            LOP(OP_SPRINTF,XTERM);

        case KEY_splice:
            LOP(OP_SPLICE,XTERM);

        case KEY_sqrt:
            UNI(OP_SQRT);

        case KEY_srand:
            UNI(OP_SRAND);

        case KEY_stat:
            UNI(OP_STAT);

        case KEY_study:
            UNI(OP_STUDY);

        case KEY_substr:
            LOP(OP_SUBSTR,XTERM);

        case KEY_format:
        case KEY_sub:
          really_sub:
            {
                char tmpbuf[sizeof PL_tokenbuf];
                SSize_t tboffset = 0;
                expectation attrful;
                bool have_name, have_proto, bad_proto;
                int key = tmp;

                s = skipspace(s);

                if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
                    (*s == ':' && s[1] == ':'))
                {
                    PL_expect = XBLOCK;
                    attrful = XATTRBLOCK;
                    /* remember buffer pos'n for later force_word */
                    tboffset = s - PL_oldbufptr;
                    d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
                    if (strchr(tmpbuf, ':'))
                        sv_setpv(PL_subname, tmpbuf);
                    else {
                        sv_setsv(PL_subname,PL_curstname);
                        sv_catpvn(PL_subname,"::",2);
                        sv_catpvn(PL_subname,tmpbuf,len);
                    }
                    s = skipspace(d);
                    have_name = TRUE;
                }
                else {
                    if (key == KEY_my)
                        Perl_croak(aTHX_ "Missing name in \"my sub\"");
                    PL_expect = XTERMBLOCK;
                    attrful = XATTRTERM;
                    sv_setpv(PL_subname,"?");
                    have_name = FALSE;
                }

                if (key == KEY_format) {
                    if (*s == '=')
                        PL_lex_formbrack = PL_lex_brackets + 1;
                    if (have_name)
                        (void) force_word(PL_oldbufptr + tboffset, WORD,
                                          FALSE, TRUE, TRUE);
                    OPERATOR(FORMAT);
                }

                /* Look for a prototype */
                if (*s == '(') {
                    char *p;

                    s = scan_str(s,FALSE,FALSE);
                    if (!s)
                        Perl_croak(aTHX_ "Prototype not terminated");
                    /* strip spaces and check for bad characters */
                    d = SvPVX(PL_lex_stuff);
                    tmp = 0;
                    bad_proto = FALSE;
                    for (p = d; *p; ++p) {
                        if (!isSPACE(*p)) {
                            d[tmp++] = *p;
                            if (!strchr("$@%*;[]&\\", *p))
                                bad_proto = TRUE;
                        }
                    }
                    d[tmp] = '\0';
                    if (bad_proto && ckWARN(WARN_SYNTAX))
                        Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                                    "Illegal character in prototype for %"SVf" : %s",
                                    PL_subname, d);
                    SvCUR(PL_lex_stuff) = tmp;
                    have_proto = TRUE;

                    s = skipspace(s);
                }
                else
                    have_proto = FALSE;

                if (*s == ':' && s[1] != ':')
                    PL_expect = attrful;
                else if (*s != '{' && key == KEY_sub) {
                    if (!have_name)
                        Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
                    else if (*s != ';')
                        Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, PL_subname);
                }

                if (have_proto) {
                    PL_nextval[PL_nexttoke].opval =
                        (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
                    PL_lex_stuff = Nullsv;
                    force_next(THING);
                }
                if (!have_name) {
                    sv_setpv(PL_subname,
                        PL_curstash ? "__ANON__" : "__ANON__::__ANON__");
                    TOKEN(ANONSUB);
                }
                (void) force_word(PL_oldbufptr + tboffset, WORD,
                                  FALSE, TRUE, TRUE);
                if (key == KEY_my)
                    TOKEN(MYSUB);
                TOKEN(SUB);
            }

        case KEY_system:
            set_csh();
            LOP(OP_SYSTEM,XREF);

        case KEY_symlink:
            LOP(OP_SYMLINK,XTERM);

        case KEY_syscall:
            LOP(OP_SYSCALL,XTERM);

        case KEY_sysopen:
            LOP(OP_SYSOPEN,XTERM);

        case KEY_sysseek:
            LOP(OP_SYSSEEK,XTERM);

        case KEY_sysread:
            LOP(OP_SYSREAD,XTERM);

        case KEY_syswrite:
            LOP(OP_SYSWRITE,XTERM);

        case KEY_tr:
            s = scan_trans(s);
            TERM(sublex_start());

        case KEY_tell:
            UNI(OP_TELL);

        case KEY_telldir:
            UNI(OP_TELLDIR);

        case KEY_tie:
            LOP(OP_TIE,XTERM);

        case KEY_tied:
            UNI(OP_TIED);

        case KEY_time:
            FUN0(OP_TIME);

        case KEY_times:
            FUN0(OP_TMS);

        case KEY_truncate:
            LOP(OP_TRUNCATE,XTERM);

        case KEY_uc:
            UNI(OP_UC);

        case KEY_ucfirst:
            UNI(OP_UCFIRST);

        case KEY_untie:
            UNI(OP_UNTIE);

        case KEY_until:
            yylval.ival = CopLINE(PL_curcop);
            OPERATOR(UNTIL);

        case KEY_unless:
            yylval.ival = CopLINE(PL_curcop);
            OPERATOR(UNLESS);

        case KEY_unlink:
            LOP(OP_UNLINK,XTERM);

        case KEY_undef:
            UNIDOR(OP_UNDEF);

        case KEY_unpack:
            LOP(OP_UNPACK,XTERM);

        case KEY_utime:
            LOP(OP_UTIME,XTERM);

        case KEY_umask:
            UNIDOR(OP_UMASK);

        case KEY_unshift:
            LOP(OP_UNSHIFT,XTERM);

        case KEY_use:
            if (PL_expect != XSTATE)
                yyerror("\"use\" not allowed in expression");
            s = skipspace(s);
            if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
                s = force_version(s, TRUE);
                if (*s == ';' || (s = skipspace(s), *s == ';')) {
                    PL_nextval[PL_nexttoke].opval = Nullop;
                    force_next(WORD);
                }
                else if (*s == 'v') {
                    s = force_word(s,WORD,FALSE,TRUE,FALSE);
                    s = force_version(s, FALSE);
                }
            }
            else {
                s = force_word(s,WORD,FALSE,TRUE,FALSE);
                s = force_version(s, FALSE);
            }
            yylval.ival = 1;
            OPERATOR(USE);

        case KEY_values:
            UNI(OP_VALUES);

        case KEY_vec:
            LOP(OP_VEC,XTERM);

        case KEY_while:
            yylval.ival = CopLINE(PL_curcop);
            OPERATOR(WHILE);

        case KEY_warn:
            PL_hints |= HINT_BLOCK_SCOPE;
            LOP(OP_WARN,XTERM);

        case KEY_wait:
            FUN0(OP_WAIT);

        case KEY_waitpid:
            LOP(OP_WAITPID,XTERM);

        case KEY_wantarray:
            FUN0(OP_WANTARRAY);

        case KEY_write:
#ifdef EBCDIC
        {
            char ctl_l[2];
            ctl_l[0] = toCTRL('L');
            ctl_l[1] = '\0';
            gv_fetchpv(ctl_l,TRUE, SVt_PV);
        }
#else
            gv_fetchpv("\f",TRUE, SVt_PV);      /* Make sure $^L is defined */
#endif
            UNI(OP_ENTERWRITE);

        case KEY_x:
            if (PL_expect == XOPERATOR)
                Mop(OP_REPEAT);
            check_uni();
            goto just_a_word;

        case KEY_xor:
            yylval.ival = OP_XOR;
            OPERATOR(OROP);

        case KEY_y:
            s = scan_trans(s);
            TERM(sublex_start());
        }
    }}
}
#ifdef __SC__
#pragma segment Main
#endif

static int
S_pending_ident(pTHX)
{
    register char *d;
    register I32 tmp = 0;
    /* pit holds the identifier we read and pending_ident is reset */
    char pit = PL_pending_ident;
    PL_pending_ident = 0;

    DEBUG_T({ PerlIO_printf(Perl_debug_log,
          "### Tokener saw identifier '%s'\n", PL_tokenbuf); });

    /* if we're in a my(), we can't allow dynamics here.
       $foo'bar has already been turned into $foo::bar, so
       just check for colons.

       if it's a legal name, the OP is a PADANY.
    */
    if (PL_in_my) {
        if (PL_in_my == KEY_our) {      /* "our" is merely analogous to "my" */
            if (strchr(PL_tokenbuf,':'))
                yyerror(Perl_form(aTHX_ "No package name allowed for "
                                  "variable %s in \"our\"",
                                  PL_tokenbuf));
            tmp = allocmy(PL_tokenbuf);
        }
        else {
            if (strchr(PL_tokenbuf,':'))
                yyerror(Perl_form(aTHX_ PL_no_myglob,PL_tokenbuf));

            yylval.opval = newOP(OP_PADANY, 0);
            yylval.opval->op_targ = allocmy(PL_tokenbuf);
            return PRIVATEREF;
        }
    }

    /*
       build the ops for accesses to a my() variable.

       Deny my($a) or my($b) in a sort block, *if* $a or $b is
       then used in a comparison.  This catches most, but not
       all cases.  For instance, it catches
           sort { my($a); $a <=> $b }
       but not
           sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
       (although why you'd do that is anyone's guess).
    */

    if (!strchr(PL_tokenbuf,':')) {
        if (!PL_in_my)
            tmp = pad_findmy(PL_tokenbuf);
        if (tmp != NOT_IN_PAD) {
            /* might be an "our" variable" */
            if (PAD_COMPNAME_FLAGS(tmp) & SVpad_OUR) {
                /* build ops for a bareword */
                SV *sym = newSVpv(HvNAME(PAD_COMPNAME_OURSTASH(tmp)), 0);
                sv_catpvn(sym, "::", 2);
                sv_catpv(sym, PL_tokenbuf+1);
                yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
                yylval.opval->op_private = OPpCONST_ENTERED;
                gv_fetchsv(sym,
                    (PL_in_eval
                        ? (GV_ADDMULTI | GV_ADDINEVAL)
                        : GV_ADDMULTI
                    ),
                    ((PL_tokenbuf[0] == '$') ? SVt_PV
                     : (PL_tokenbuf[0] == '@') ? SVt_PVAV
                     : SVt_PVHV));
                return WORD;
            }

            /* if it's a sort block and they're naming $a or $b */
            if (PL_last_lop_op == OP_SORT &&
                PL_tokenbuf[0] == '$' &&
                (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
                && !PL_tokenbuf[2])
            {
                for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
                     d < PL_bufend && *d != '\n';
                     d++)
                {
                    if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
                        Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
                              PL_tokenbuf);
                    }
                }
            }

            yylval.opval = newOP(OP_PADANY, 0);
            yylval.opval->op_targ = tmp;
            return PRIVATEREF;
        }
    }

    /*
       Whine if they've said @foo in a doublequoted string,
       and @foo isn't a variable we can find in the symbol
       table.
    */
    if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
        GV *gv = gv_fetchpv(PL_tokenbuf+1, FALSE, SVt_PVAV);
        if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
             && ckWARN(WARN_AMBIGUOUS))
        {
            /* Downgraded from fatal to warning 20000522 mjd */
            Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
                        "Possible unintended interpolation of %s in string",
                         PL_tokenbuf);
        }
    }

    /* build ops for a bareword */
    yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
    yylval.opval->op_private = OPpCONST_ENTERED;
    gv_fetchpv(PL_tokenbuf+1, PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : TRUE,
               ((PL_tokenbuf[0] == '$') ? SVt_PV
                : (PL_tokenbuf[0] == '@') ? SVt_PVAV
                : SVt_PVHV));
    return WORD;
}

/* Weights are the number of occurrences of that keyword in about 135M of
   input to Perl_keyword from a lot of real perl. This routine is about 20%
   faster than the routine it replaces.  */

I32
Perl_keyword (pTHX_ char *name, I32 len) {
  /* Initially switch on the length of the name.  */
  switch (len) {
  case 1:
    /* Names all of length 1.  */
    /* m q s x y */
    /* Offset 0 gives the best switch position.  */
    switch (name[0]) {
    case 'm':
      {
        return KEY_m; /* Weight 148776 */
      }
      break;
    case 'q':
      {
        return KEY_q; /* Weight 69076 */
      }
      break;
    case 's':
      {
        return KEY_s; /* Weight 403691 */
      }
      break;
    case 'x':
      {
        return -KEY_x; /* Weight 38549 */
      }
      break;
    case 'y':
      {
        return KEY_y; /* Weight 567 */
      }
      break;
    }
    break;
  case 2:
    /* Names all of length 2.  */
    /* do eq ge gt if lc le lt my ne no or qq qr qw qx tr uc */
    /* Offset 0 gives the best switch position.  */
    switch (name[0]) {
    case 'd':
      if (name[1] == 'o') {
        return KEY_do; /* Weight 96004 */
      }
      break;
    case 'e':
      if (name[1] == 'q') {
        return -KEY_eq; /* Weight 797065 */
      }
      break;
    case 'g':
      if (name[1] == 'e') {
        return -KEY_ge; /* Weight 5666 */
      }
      if (name[1] == 't') {
        return -KEY_gt; /* Weight 897 */
      }
      break;
    case 'i':
      if (name[1] == 'f') {
        return KEY_if; /* Weight 2482605 */
      }
      break;
    case 'l':
      if (name[1] == 'c') {
        return -KEY_lc; /* Weight 38487 */
      }
      if (name[1] == 'e') {
        return -KEY_le; /* Weight 4052 */
      }
      if (name[1] == 't') {
        return -KEY_lt; /* Weight 335 */
      }
      break;
    case 'm':
      if (name[1] == 'y') {
        return KEY_my; /* Weight 3785925 */
      }
      break;
    case 'n':
      if (name[1] == 'e') {
        return -KEY_ne; /* Weight 112906 */
      }
      if (name[1] == 'o') {
        return KEY_no; /* Weight 61989 */
      }
      break;
    case 'o':
      if (name[1] == 'r') {
        return -KEY_or; /* Weight 405163 */
      }
      break;
    case 'q':
      if (name[1] == 'w') {
        return KEY_qw; /* Weight 415641 */
      }
      if (name[1] == 'q') {
        return KEY_qq; /* Weight 55149 */
      }
      if (name[1] == 'r') {
        return KEY_qr; /* Weight 28519 */
      }
      if (name[1] == 'x') {
        return KEY_qx; /* Weight 177 */
      }
      break;
    case 't':
      if (name[1] == 'r') {
        return KEY_tr; /* Weight 22665 */
      }
      break;
    case 'u':
      if (name[1] == 'c') {
        return -KEY_uc; /* Weight 16961 */
      }
      break;
    }
    break;
  case 3:
    /* Names all of length 3.  */
    /* END abs and chr cmp cos die eof err exp for hex int log map not oct ord
       our pop pos ref sin sub tie use vec xor */
    /* Offset 0 gives the best switch position.  */
    switch (*name++) {
    case 'E':
      if (name[0] == 'N' && name[1] == 'D') {
        return KEY_END; /* Weight 3565 */
      }
      break;
    case 'a':
      if (name[0] == 'n' && name[1] == 'd') {
        return -KEY_and; /* Weight 284867 */
      }
      if (name[0] == 'b' && name[1] == 's') {
        return -KEY_abs; /* Weight 7767 */
      }
      break;
    case 'c':
      if (name[0] == 'h' && name[1] == 'r') {
        return -KEY_chr; /* Weight 35654 */
      }
      if (name[0] == 'm' && name[1] == 'p') {
        return -KEY_cmp; /* Weight 6808 */
      }
      if (name[0] == 'o' && name[1] == 's') {
        return -KEY_cos; /* Weight 447 */
      }
      break;
    case 'd':
      if (name[0] == 'i' && name[1] == 'e') {
        return -KEY_die; /* Weight 192203 */
      }
      break;
    case 'e':
      if (name[0] == 'o' && name[1] == 'f') {
        return -KEY_eof; /* Weight 1618 */
      }
      if (name[0] == 'r' && name[1] == 'r') {
        return -KEY_err; /* Weight 522 */
      }
      if (name[0] == 'x' && name[1] == 'p') {
        return -KEY_exp; /* Weight 423 */
      }
      break;
    case 'f':
      if (name[0] == 'o' && name[1] == 'r') {
        return KEY_for; /* Weight 118158 */
      }
      break;
    case 'h':
      if (name[0] == 'e' && name[1] == 'x') {
        return -KEY_hex; /* Weight 3629 */
      }
      break;
    case 'i':
      if (name[0] == 'n' && name[1] == 't') {
        return -KEY_int; /* Weight 18549 */
      }
      break;
    case 'l':
      if (name[0] == 'o' && name[1] == 'g') {
        return -KEY_log;
      }
      break;
    case 'm':
      if (name[0] == 'a' && name[1] == 'p') {
        return KEY_map; /* Weight 115207 */
      }
      break;
    case 'n':
      if (name[0] == 'o' && name[1] == 't') {
        return -KEY_not; /* Weight 55868 */
      }
      break;
    case 'o':
      if (name[0] == 'u' && name[1] == 'r') {
        return KEY_our; /* Weight 194417 */
      }
      if (name[0] == 'r' && name[1] == 'd') {
        return -KEY_ord; /* Weight 22221 */
      }
      if (name[0] == 'c' && name[1] == 't') {
        return -KEY_oct; /* Weight 4195 */
      }
      break;
    case 'p':
      if (name[0] == 'o' && name[1] == 'p') {
        return -KEY_pop; /* Weight 46933 */
      }
      if (name[0] == 'o' && name[1] == 's') {
        return KEY_pos; /* Weight 5503 */
      }
      break;
    case 'r':
      if (name[0] == 'e' && name[1] == 'f') {
        return -KEY_ref; /* Weight 347102 */
      }
      break;
    case 's':
      if (name[0] == 'u' && name[1] == 'b') {
        return KEY_sub; /* Weight 2053554 */
      }
      if (name[0] == 'i' && name[1] == 'n') {
        return -KEY_sin; /* Weight 499 */
      }
      break;
    case 't':
      if (name[0] == 'i' && name[1] == 'e') {
        return KEY_tie; /* Weight 10131 */
      }
      break;
    case 'u':
      if (name[0] == 's' && name[1] == 'e') {
        return KEY_use; /* Weight 686081 */
      }
      break;
    case 'v':
      if (name[0] == 'e' && name[1] == 'c') {
        return -KEY_vec; /* Weight 110566 */
      }
      break;
    case 'x':
      if (name[0] == 'o' && name[1] == 'r') {
        return -KEY_xor; /* Weight 619 */
      }
      break;
    }
    break;
  case 4:
    /* Names all of length 4.  */
    /* CORE INIT bind chop dump each else eval exec exit fork getc glob goto
       grep join keys kill last link lock next open pack pipe push rand read
       recv redo seek send sort sqrt stat tell tied time wait warn */
    /* Offset 0 gives the best switch position.  */
    switch (*name++) {
    case 'C':
      if (!memcmp(name, "ORE", 3)) {
      /*                C         */
        return -KEY_CORE; /* Weight 47391 */
      }
      break;
    case 'I':
      if (!memcmp(name, "NIT", 3)) {
      /*                I         */
        return KEY_INIT; /* Weight 418 */
      }
      break;
    case 'b':
      if (!memcmp(name, "ind", 3)) {
      /*                b         */
        return -KEY_bind; /* Weight 290 */
      }
      break;
    case 'c':
      if (!memcmp(name, "hop", 3)) {
      /*                c         */
        return -KEY_chop; /* Weight 10172 */
      }
      break;
    case 'd':
      if (!memcmp(name, "ump", 3)) {
      /*                d         */
        return -KEY_dump; /* Weight 274 */
      }
      break;
    case 'e':
      if (!memcmp(name, "lse", 3)) {
      /*                e         */
        return KEY_else; /* Weight 527806 */
      }
      if (!memcmp(name, "val", 3)) {
      /*                e         */
        return KEY_eval; /* Weight 136977 */
      }
      if (!memcmp(name, "ach", 3)) {
      /*                e         */
        return -KEY_each; /* Weight 18414 */
      }
      if (!memcmp(name, "xit", 3)) {
      /*                e         */
        return -KEY_exit; /* Weight 8262 */
      }
      if (!memcmp(name, "xec", 3)) {
      /*                e         */
        return -KEY_exec; /* Weight 429 */
      }
      break;
    case 'f':
      if (!memcmp(name, "ork", 3)) {
      /*                f         */
        return -KEY_fork; /* Weight 327 */
      }
      break;
    case 'g':
      if (!memcmp(name, "oto", 3)) {
      /*                g         */
        return KEY_goto; /* Weight 109258 */
      }
      if (!memcmp(name, "rep", 3)) {
      /*                g         */
        return KEY_grep; /* Weight 75912 */
      }
      if (!memcmp(name, "lob", 3)) {
      /*                g         */
        return KEY_glob; /* Weight 2172 */
      }
      if (!memcmp(name, "etc", 3)) {
      /*                g         */
        return -KEY_getc; /* Weight 981 */
      }
      break;
    case 'j':
      if (!memcmp(name, "oin", 3)) {
      /*                j         */
        return -KEY_join; /* Weight 130820 */
      }
      break;
    case 'k':
      if (!memcmp(name, "eys", 3)) {
      /*                k         */
        return -KEY_keys; /* Weight 131427 */
      }
      if (!memcmp(name, "ill", 3)) {
      /*                k         */
        return -KEY_kill; /* Weight 382 */
      }
      break;
    case 'l':
      if (!memcmp(name, "ast", 3)) {
      /*                l         */
        return KEY_last; /* Weight 95078 */
      }
      if (!memcmp(name, "ock", 3)) {
      /*                l         */
        return -KEY_lock; /* Weight 4210 */
      }
      if (!memcmp(name, "ink", 3)) {
      /*                l         */
        return -KEY_link; /* Weight 425 */
      }
      break;
    case 'n':
      if (!memcmp(name, "ext", 3)) {
      /*                n         */
        return KEY_next; /* Weight 153355 */
      }
      break;
    case 'o':
      if (!memcmp(name, "pen", 3)) {
      /*                o         */
        return -KEY_open; /* Weight 39060 */
      }
      break;
    case 'p':
      if (!memcmp(name, "ush", 3)) {
      /*                p         */
        return -KEY_push; /* Weight 256975 */
      }
      if (!memcmp(name, "ack", 3)) {
      /*                p         */
        return -KEY_pack; /* Weight 14491 */
      }
      if (!memcmp(name, "ipe", 3)) {
      /*                p         */
        return -KEY_pipe; /* Weight 344 */
      }
      break;
    case 'r':
      if (!memcmp(name, "ead", 3)) {
      /*                r         */
        return -KEY_read; /* Weight 9434 */
      }
      if (!memcmp(name, "edo", 3)) {
      /*                r         */
        return KEY_redo; /* Weight 5219 */
      }
      if (!memcmp(name, "and", 3)) {
      /*                r         */
        return -KEY_rand; /* Weight 1824 */
      }
      if (!memcmp(name, "ecv", 3)) {
      /*                r         */
        return -KEY_recv; /* Weight 250 */
      }
      break;
    case 's':
      if (!memcmp(name, "tat", 3)) {
      /*                s         */
        return -KEY_stat; /* Weight 36702 */
      }
      if (!memcmp(name, "ort", 3)) {
      /*                s         */
        return KEY_sort; /* Weight 36394 */
      }
      if (!memcmp(name, "eek", 3)) {
      /*                s         */
        return -KEY_seek; /* Weight 2174 */
      }
      if (!memcmp(name, "qrt", 3)) {
      /*                s         */
        return -KEY_sqrt; /* Weight 766 */
      }
      if (!memcmp(name, "end", 3)) {
      /*                s         */
        return -KEY_send; /* Weight 496 */
      }
      break;
    case 't':
      if (!memcmp(name, "ime", 3)) {
      /*                t         */
        return -KEY_time; /* Weight 32168 */
      }
      if (!memcmp(name, "ied", 3)) {
      /*                t         */
        return KEY_tied; /* Weight 9749 */
      }
      if (!memcmp(name, "ell", 3)) {
      /*                t         */
        return -KEY_tell; /* Weight 2578 */
      }
      break;
    case 'w':
      if (!memcmp(name, "arn", 3)) {
      /*                w         */
        return -KEY_warn; /* Weight 91372 */
      }
      if (!memcmp(name, "ait", 3)) {
      /*                w         */
        return -KEY_wait;
      }
      break;
    }
    break;
  case 5:
    /* Names all of length 5.  */
    /* BEGIN CHECK alarm atan2 bless chdir chmod chomp chown close crypt elsif
       fcntl flock index ioctl local lstat mkdir print reset rmdir semop shift
       sleep split srand study times umask undef untie until utime while write
       */
    /* Offset 3 gives the best switch position.  */
    switch (name[3]) {
    case 'C':
      if (!memcmp(name, "CHECK", 5)) {
      /*                    ^       */
        return KEY_CHECK; /* Weight 538 */
      }
      break;
    case 'I':
      if (!memcmp(name, "BEGIN", 5)) {
      /*                    ^       */
        return KEY_BEGIN; /* Weight 24125 */
      }
      break;
    case 'a':
      if (!memcmp(name, "local", 5)) {
      /*                    ^       */
        return KEY_local; /* Weight 262973 */
      }
      if (!memcmp(name, "lstat", 5)) {
      /*                    ^       */
        return -KEY_lstat; /* Weight 13859 */
      }
      break;
    case 'c':
      if (!memcmp(name, "flock", 5)) {
      /*                    ^       */
        return -KEY_flock; /* Weight 260 */
      }
      break;
    case 'd':
      if (!memcmp(name, "study", 5)) {
      /*                    ^       */
        return KEY_study; /* Weight 1933 */
      }
      break;
    case 'e':
      if (!memcmp(name, "undef", 5)) {
      /*                    ^       */
        return KEY_undef; /* Weight 311156 */
      }
      if (!memcmp(name, "index", 5)) {
      /*                    ^       */
        return -KEY_index; /* Weight 51465 */
      }
      if (!memcmp(name, "sleep", 5)) {
      /*                    ^       */
        return -KEY_sleep; /* Weight 519 */
      }
      if (!memcmp(name, "times", 5)) {
      /*                    ^       */
        return -KEY_times; /* Weight 310 */
      }
      if (!memcmp(name, "reset", 5)) {
      /*                    ^       */
        return -KEY_reset; /* Weight 127 */
      }
      break;
    case 'f':
      if (!memcmp(name, "shift", 5)) {
      /*                    ^       */
        return -KEY_shift; /* Weight 904125 */
      }
      break;
    case 'i':
      if (!memcmp(name, "elsif", 5)) {
      /*                    ^       */
        return KEY_elsif; /* Weight 322365 */
      }
      if (!memcmp(name, "split", 5)) {
      /*                    ^       */
        return KEY_split; /* Weight 93678 */
      }
      if (!memcmp(name, "chdir", 5)) {
      /*                    ^       */
        return -KEY_chdir; /* Weight 20317 */
      }
      if (!memcmp(name, "mkdir", 5)) {
      /*                    ^       */
        return -KEY_mkdir; /* Weight 2951 */
      }
      if (!memcmp(name, "rmdir", 5)) {
      /*                    ^       */
        return -KEY_rmdir; /* Weight 2493 */
      }
      if (!memcmp(name, "until", 5)) {
      /*                    ^       */
        return KEY_until; /* Weight 818 */
      }
      if (!memcmp(name, "untie", 5)) {
      /*                    ^       */
        return KEY_untie; /* Weight 420 */
      }
      break;
    case 'l':
      if (!memcmp(name, "while", 5)) {
      /*                    ^       */
        return KEY_while; /* Weight 120305 */
      }
      break;
    case 'm':
      if (!memcmp(name, "chomp", 5)) {
      /*                    ^       */
        return -KEY_chomp; /* Weight 22337 */
      }
      if (!memcmp(name, "utime", 5)) {
      /*                    ^       */
        return -KEY_utime; /* Weight 3849 */
      }
      break;
    case 'n':
      if (!memcmp(name, "print", 5)) {
      /*                    ^       */
        return KEY_print; /* Weight 220904 */
      }
      if (!memcmp(name, "atan2", 5)) {
      /*                    ^       */
        return -KEY_atan2; /* Weight 350 */
      }
      if (!memcmp(name, "srand", 5)) {
      /*                    ^       */
        return -KEY_srand; /* Weight 41 */
      }
      break;
    case 'o':
      if (!memcmp(name, "chmod", 5)) {
      /*                    ^       */
        return -KEY_chmod; /* Weight 18455 */
      }
      if (!memcmp(name, "semop", 5)) {
      /*                    ^       */
        return -KEY_semop;
      }
      break;
    case 'p':
      if (!memcmp(name, "crypt", 5)) {
      /*                    ^       */
        return -KEY_crypt; /* Weight 8 */
      }
      break;
    case 'r':
      if (!memcmp(name, "alarm", 5)) {
      /*                    ^       */
        return -KEY_alarm;
      }
      break;
    case 's':
      if (!memcmp(name, "bless", 5)) {
      /*                    ^       */
        return -KEY_bless; /* Weight 62111 */
      }
      if (!memcmp(name, "close", 5)) {
      /*                    ^       */
        return -KEY_close; /* Weight 44077 */
      }
      if (!memcmp(name, "umask", 5)) {
      /*                    ^       */
        return -KEY_umask; /* Weight 1658 */
      }
      break;
    case 't':
      if (!memcmp(name, "write", 5)) {
      /*                    ^       */
        return -KEY_write; /* Weight 2525 */
      }
      if (!memcmp(name, "fcntl", 5)) {
      /*                    ^       */
        return -KEY_fcntl; /* Weight 1257 */
      }
      if (!memcmp(name, "ioctl", 5)) {
      /*                    ^       */
        return -KEY_ioctl; /* Weight 967 */
      }
      break;
    case 'w':
      if (!memcmp(name, "chown", 5)) {
      /*                    ^       */
        return -KEY_chown; /* Weight 34 */
      }
      break;
    }
    break;
  case 6:
    /* Names all of length 6.  */
    /* accept caller chroot delete elseif exists fileno format gmtime length
       listen msgctl msgget msgrcv msgsnd printf rename return rindex scalar
       select semctl semget shmctl shmget socket splice substr system unless
       unlink unpack values */
    /* Offset 3 gives the best switch position.  */
    switch (name[3]) {
    case 'a':
      if (!memcmp(name, "unpack", 6)) {
      /*                    ^        */
        return -KEY_unpack; /* Weight 7849 */
      }
      if (!memcmp(name, "rename", 6)) {
      /*                    ^        */
        return -KEY_rename; /* Weight 4918 */
      }
      break;
    case 'c':
      if (!memcmp(name, "semctl", 6)) {
      /*                    ^        */
        return -KEY_semctl; /* Weight 17 */
      }
      if (!memcmp(name, "msgctl", 6)) {
      /*                    ^        */
        return -KEY_msgctl;
      }
      if (!memcmp(name, "shmctl", 6)) {
      /*                    ^        */
        return -KEY_shmctl;
      }
      break;
    case 'd':
      if (!memcmp(name, "rindex", 6)) {
      /*                    ^        */
        return -KEY_rindex; /* Weight 5005 */
      }
      break;
    case 'e':
      if (!memcmp(name, "unless", 6)) {
      /*                    ^        */
        return KEY_unless; /* Weight 913955 */
      }
      if (!memcmp(name, "delete", 6)) {
      /*                    ^        */
        return KEY_delete; /* Weight 74966 */
      }
      if (!memcmp(name, "select", 6)) {
      /*                    ^        */
        return -KEY_select; /* Weight 12209 */
      }
      if (!memcmp(name, "fileno", 6)) {
      /*                    ^        */
        return -KEY_fileno; /* Weight 8591 */
      }
      if (!memcmp(name, "accept", 6)) {
      /*                    ^        */
        return -KEY_accept; /* Weight 233 */
      }
      if (!memcmp(name, "elseif", 6)) {
      /*                    ^        */
        /* This is somewhat hacky.  */
        if(ckWARN_d(WARN_SYNTAX))
          Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
        break;
      }
      break;
    case 'g':
      if (!memcmp(name, "length", 6)) {
      /*                    ^        */
        return -KEY_length; /* Weight 163975 */
      }
      if (!memcmp(name, "msgget", 6)) {
      /*                    ^        */
        return -KEY_msgget;
      }
      if (!memcmp(name, "semget", 6)) {
      /*                    ^        */
        return -KEY_semget;
      }
      if (!memcmp(name, "shmget", 6)) {
      /*                    ^        */
        return -KEY_shmget;
      }
      break;
    case 'i':
      if (!memcmp(name, "splice", 6)) {
      /*                    ^        */
        return -KEY_splice; /* Weight 25143 */
      }
      if (!memcmp(name, "unlink", 6)) {
      /*                    ^        */
        return -KEY_unlink; /* Weight 18616 */
      }
      if (!memcmp(name, "gmtime", 6)) {
      /*                    ^        */
        return -KEY_gmtime; /* Weight 4040 */
      }
      break;
    case 'k':
      if (!memcmp(name, "socket", 6)) {
      /*                    ^        */
        return -KEY_socket;
      }
      break;
    case 'l':
      if (!memcmp(name, "caller", 6)) {
      /*                    ^        */
        return -KEY_caller; /* Weight 148457 */
      }
      if (!memcmp(name, "scalar", 6)) {
      /*                    ^        */
        return KEY_scalar; /* Weight 43953 */
      }
      break;
    case 'm':
      if (!memcmp(name, "format", 6)) {
      /*                    ^        */
        return KEY_format; /* Weight 1735 */
      }
      break;
    case 'n':
      if (!memcmp(name, "printf", 6)) {
      /*                    ^        */
        return KEY_printf; /* Weight 6874 */
      }
      break;
    case 'o':
      if (!memcmp(name, "chroot", 6)) {
      /*                    ^        */
        return -KEY_chroot;
      }
      break;
    case 'r':
      if (!memcmp(name, "msgrcv", 6)) {
      /*                    ^        */
        return -KEY_msgrcv;
      }
      break;
    case 's':
      if (!memcmp(name, "exists", 6)) {
      /*                    ^        */
        return KEY_exists; /* Weight 145939 */
      }
      if (!memcmp(name, "substr", 6)) {
      /*                    ^        */
        return -KEY_substr; /* Weight 121344 */
      }
      if (!memcmp(name, "msgsnd", 6)) {
      /*                    ^        */
        return -KEY_msgsnd;
      }
      break;
    case 't':
      if (!memcmp(name, "system", 6)) {
      /*                    ^        */
        return -KEY_system; /* Weight 4326 */
      }
      if (!memcmp(name, "listen", 6)) {
      /*                    ^        */
        return -KEY_listen;
      }
      break;
    case 'u':
      if (!memcmp(name, "return", 6)) {
      /*                    ^        */
        return KEY_return; /* Weight 1401629 */
      }
      if (!memcmp(name, "values", 6)) {
      /*                    ^        */
        return -KEY_values; /* Weight 10110 */
      }
      break;
    }
    break;
  case 7:
    /* Names all of length 7.  */
    /* DESTROY __END__ binmode connect dbmopen defined foreach getpgrp getppid
       lcfirst opendir package readdir require reverse seekdir setpgrp shmread
       sprintf symlink syscall sysopen sysread sysseek telldir ucfirst unshift
       waitpid */
    /* Offset 3 gives the best switch position.  */
    switch (name[3]) {
    case 'N':
      if (!memcmp(name, "__END__", 7)) {
      /*                    ^         */
        return KEY___END__; /* Weight 112636 */
      }
      break;
    case 'T':
      if (!memcmp(name, "DESTROY", 7)) {
      /*                    ^         */
        return KEY_DESTROY; /* Weight 7 */
      }
      break;
    case 'c':
      if (!memcmp(name, "syscall", 7)) {
      /*                    ^         */
        return -KEY_syscall; /* Weight 560 */
      }
      break;
    case 'd':
      if (!memcmp(name, "readdir", 7)) {
      /*                    ^         */
        return -KEY_readdir; /* Weight 11716 */
      }
      break;
    case 'e':
      if (!memcmp(name, "foreach", 7)) {
      /*                    ^         */
        return KEY_foreach; /* Weight 281720 */
      }
      if (!memcmp(name, "reverse", 7)) {
      /*                    ^         */
        return -KEY_reverse; /* Weight 10571 */
      }
      break;
    case 'h':
      if (!memcmp(name, "unshift", 7)) {
      /*                    ^         */
        return -KEY_unshift; /* Weight 36504 */
      }
      break;
    case 'i':
      if (!memcmp(name, "defined", 7)) {
      /*                    ^         */
        return KEY_defined; /* Weight 694277 */
      }
      if (!memcmp(name, "sprintf", 7)) {
      /*                    ^         */
        return -KEY_sprintf; /* Weight 72704 */
      }
      if (!memcmp(name, "ucfirst", 7)) {
      /*                    ^         */
        return -KEY_ucfirst; /* Weight 1012 */
      }
      if (!memcmp(name, "lcfirst", 7)) {
      /*                    ^         */
        return -KEY_lcfirst; /* Weight 165 */
      }
      break;
    case 'k':
      if (!memcmp(name, "package", 7)) {
      /*                    ^         */
        return KEY_package; /* Weight 245661 */
      }
      if (!memcmp(name, "seekdir", 7)) {
      /*                    ^         */
        return -KEY_seekdir; /* Weight 20 */
      }
      break;
    case 'l':
      if (!memcmp(name, "symlink", 7)) {
      /*                    ^         */
        return -KEY_symlink; /* Weight 386 */
      }
      if (!memcmp(name, "telldir", 7)) {
      /*                    ^         */
        return -KEY_telldir; /* Weight 294 */
      }
      break;
    case 'm':
      if (!memcmp(name, "binmode", 7)) {
      /*                    ^         */
        return -KEY_binmode; /* Weight 12301 */
      }
      break;
    case 'n':
      if (!memcmp(name, "opendir", 7)) {
      /*                    ^         */
        return -KEY_opendir; /* Weight 9007 */
      }
      if (!memcmp(name, "connect", 7)) {
      /*                    ^         */
        return -KEY_connect; /* Weight 526 */
      }
      break;
    case 'o':
      if (!memcmp(name, "sysopen", 7)) {
      /*                    ^         */
        return -KEY_sysopen; /* Weight 1230 */
      }
      if (!memcmp(name, "dbmopen", 7)) {
      /*                    ^         */
        return -KEY_dbmopen;
      }
      break;
    case 'p':
      if (!memcmp(name, "getppid", 7)) {
      /*                    ^         */
        return -KEY_getppid; /* Weight 10 */
      }
      if (!memcmp(name, "getpgrp", 7)) {
      /*                    ^         */
        return -KEY_getpgrp;
      }
      if (!memcmp(name, "setpgrp", 7)) {
      /*                    ^         */
        return -KEY_setpgrp;
      }
      break;
    case 'r':
      if (!memcmp(name, "sysread", 7)) {
      /*                    ^         */
        return -KEY_sysread; /* Weight 3729 */
      }
      if (!memcmp(name, "shmread", 7)) {
      /*                    ^         */
        return -KEY_shmread;
      }
      break;
    case 's':
      if (!memcmp(name, "sysseek", 7)) {
      /*                    ^         */
        return -KEY_sysseek; /* Weight 721 */
      }
      break;
    case 't':
      if (!memcmp(name, "waitpid", 7)) {
      /*                    ^         */
        return -KEY_waitpid; /* Weight 414 */
      }
      break;
    case 'u':
      if (!memcmp(name, "require", 7)) {
      /*                    ^         */
        return KEY_require; /* Weight 375220 */
      }
      break;
    }
    break;
  case 8:
    /* Names all of length 8.  */
    /* AUTOLOAD __DATA__ __FILE__ __LINE__ closedir continue dbmclose endgrent
       endpwent formline getgrent getgrgid getgrnam getlogin getpwent getpwnam
       getpwuid readline readlink readpipe setgrent setpwent shmwrite shutdown
       syswrite truncate */
    /* Offset 3 gives the best switch position.  */
    switch (name[3]) {
    case 'A':
      if (!memcmp(name, "__DATA__", 8)) {
      /*                    ^          */
        return KEY___DATA__; /* Weight 395 */
      }
      break;
    case 'I':
      if (!memcmp(name, "__FILE__", 8)) {
      /*                    ^          */
        return -KEY___FILE__; /* Weight 888 */
      }
      if (!memcmp(name, "__LINE__", 8)) {
      /*                    ^          */
        return -KEY___LINE__; /* Weight 209 */
      }
      break;
    case 'O':
      if (!memcmp(name, "AUTOLOAD", 8)) {
      /*                    ^          */
        return KEY_AUTOLOAD; /* Weight 2713 */
      }
      break;
    case 'c':
      if (!memcmp(name, "dbmclose", 8)) {
      /*                    ^          */
        return -KEY_dbmclose;
      }
      break;
    case 'd':
      if (!memcmp(name, "readlink", 8)) {
      /*                    ^          */
        return -KEY_readlink; /* Weight 1537 */
      }
      if (!memcmp(name, "readline", 8)) {
      /*                    ^          */
        return -KEY_readline; /* Weight 19 */
      }
      if (!memcmp(name, "readpipe", 8)) {
      /*                    ^          */
        return -KEY_readpipe;
      }
      break;
    case 'g':
      if (!memcmp(name, "getgrgid", 8)) {
      /*                    ^          */
        return -KEY_getgrgid; /* Weight 67 */
      }
      if (!memcmp(name, "getgrnam", 8)) {
      /*                    ^          */
        return -KEY_getgrnam; /* Weight 11 */
      }
      if (!memcmp(name, "endgrent", 8)) {
      /*                    ^          */
        return -KEY_endgrent;
      }
      if (!memcmp(name, "getgrent", 8)) {
      /*                    ^          */
        return -KEY_getgrent;
      }
      if (!memcmp(name, "setgrent", 8)) {
      /*                    ^          */
        return -KEY_setgrent;
      }
      break;
    case 'l':
      if (!memcmp(name, "getlogin", 8)) {
      /*                    ^          */
        return -KEY_getlogin; /* Weight 158 */
      }
      break;
    case 'm':
      if (!memcmp(name, "formline", 8)) {
      /*                    ^          */
        return -KEY_formline; /* Weight 959 */
      }
      break;
    case 'n':
      if (!memcmp(name, "truncate", 8)) {
      /*                    ^          */
        return -KEY_truncate; /* Weight 1351 */
      }
      break;
    case 'p':
      if (!memcmp(name, "getpwuid", 8)) {
      /*                    ^          */
        return -KEY_getpwuid; /* Weight 681 */
      }
      if (!memcmp(name, "getpwnam", 8)) {
      /*                    ^          */
        return -KEY_getpwnam; /* Weight 483 */
      }
      if (!memcmp(name, "getpwent", 8)) {
      /*                    ^          */
        return -KEY_getpwent; /* Weight 12 */
      }
      if (!memcmp(name, "endpwent", 8)) {
      /*                    ^          */
        return -KEY_endpwent;
      }
      if (!memcmp(name, "setpwent", 8)) {
      /*                    ^          */
        return -KEY_setpwent;
      }
      break;
    case 's':
      if (!memcmp(name, "closedir", 8)) {
      /*                    ^          */
        return -KEY_closedir; /* Weight 11986 */
      }
      break;
    case 't':
      if (!memcmp(name, "continue", 8)) {
      /*                    ^          */
        return -KEY_continue; /* Weight 2925 */
      }
      if (!memcmp(name, "shutdown", 8)) {
      /*                    ^          */
        return -KEY_shutdown;
      }
      break;
    case 'w':
      if (!memcmp(name, "syswrite", 8)) {
      /*                    ^          */
        return -KEY_syswrite; /* Weight 4437 */
      }
      if (!memcmp(name, "shmwrite", 8)) {
      /*                    ^          */
        return -KEY_shmwrite;
      }
      break;
    }
    break;
  case 9:
    /* Names all of length 9.  */
    /* endnetent getnetent localtime prototype quotemeta rewinddir setnetent
       wantarray */
    /* Offset 0 gives the best switch position.  */
    switch (*name++) {
    case 'e':
      if (!memcmp(name, "ndnetent", 8)) {
      /*                e              */
        return -KEY_endnetent;
      }
      break;
    case 'g':
      if (!memcmp(name, "etnetent", 8)) {
      /*                g              */
        return -KEY_getnetent;
      }
      break;
    case 'l':
      if (!memcmp(name, "ocaltime", 8)) {
      /*                l              */
        return -KEY_localtime; /* Weight 7993 */
      }
      break;
    case 'p':
      if (!memcmp(name, "rototype", 8)) {
      /*                p              */
        return KEY_prototype; /* Weight 1602 */
      }
      break;
    case 'q':
      if (!memcmp(name, "uotemeta", 8)) {
      /*                q              */
        return -KEY_quotemeta; /* Weight 3120 */
      }
      break;
    case 'r':
      if (!memcmp(name, "ewinddir", 8)) {
      /*                r              */
        return -KEY_rewinddir; /* Weight 218 */
      }
      break;
    case 's':
      if (!memcmp(name, "etnetent", 8)) {
      /*                s              */
        return -KEY_setnetent; /* Weight 1 */
      }
      break;
    case 'w':
      if (!memcmp(name, "antarray", 8)) {
      /*                w              */
        return -KEY_wantarray; /* Weight 43024 */
      }
      break;
    }
    break;
  case 10:
    /* Names all of length 10.  */
    /* endhostent endservent gethostent getservent getsockopt sethostent
       setservent setsockopt socketpair */
    /* Offset 6 gives the best switch position.  */
    switch (name[6]) {
    case 'k':
      if (!memcmp(name, "setsockopt", 10)) {
      /*                       ^          */
        return -KEY_setsockopt; /* Weight 356 */
      }
      if (!memcmp(name, "getsockopt", 10)) {
      /*                       ^          */
        return -KEY_getsockopt; /* Weight 243 */
      }
      break;
    case 'p':
      if (!memcmp(name, "socketpair", 10)) {
      /*                       ^          */
        return -KEY_socketpair;
      }
      break;
    case 't':
      if (!memcmp(name, "gethostent", 10)) {
      /*                       ^          */
        return -KEY_gethostent; /* Weight 3 */
      }
      if (!memcmp(name, "endhostent", 10)) {
      /*                       ^          */
        return -KEY_endhostent;
      }
      if (!memcmp(name, "sethostent", 10)) {
      /*                       ^          */
        return -KEY_sethostent;
      }
      break;
    case 'v':
      if (!memcmp(name, "getservent", 10)) {
      /*                       ^          */
        return -KEY_getservent; /* Weight 4 */
      }
      if (!memcmp(name, "endservent", 10)) {
      /*                       ^          */
        return -KEY_endservent;
      }
      if (!memcmp(name, "setservent", 10)) {
      /*                       ^          */
        return -KEY_setservent;
      }
      break;
    }
    break;
  case 11:
    /* Names all of length 11.  */
    /* __PACKAGE__ endprotoent getpeername getpriority getprotoent getsockname
       setpriority setprotoent */
    /* Offset 5 gives the best switch position.  */
    switch (name[5]) {
    case 'K':
      if (!memcmp(name, "__PACKAGE__", 11)) {
      /*                      ^            */
        return -KEY___PACKAGE__; /* Weight 36767 */
      }
      break;
    case 'c':
      if (!memcmp(name, "getsockname", 11)) {
      /*                      ^            */
        return -KEY_getsockname; /* Weight 235 */
      }
      break;
    case 'e':
      if (!memcmp(name, "getpeername", 11)) {
      /*                      ^            */
        return -KEY_getpeername; /* Weight 713 */
      }
      break;
    case 'i':
      if (!memcmp(name, "getpriority", 11)) {
      /*                      ^            */
        return -KEY_getpriority; /* Weight 5 */
      }
      if (!memcmp(name, "setpriority", 11)) {
      /*                      ^            */
        return -KEY_setpriority; /* Weight 2 */
      }
      break;
    case 'o':
      if (!memcmp(name, "endprotoent", 11)) {
      /*                      ^            */
        return -KEY_endprotoent;
      }
      if (!memcmp(name, "getprotoent", 11)) {
      /*                      ^            */
        return -KEY_getprotoent;
      }
      if (!memcmp(name, "setprotoent", 11)) {
      /*                      ^            */
        return -KEY_setprotoent;
      }
      break;
    }
    break;
  case 12:
    /* Names all of length 12.  */
    /* getnetbyaddr getnetbyname */
    /* Offset 9 gives the best switch position.  */
    switch (name[9]) {
    case 'a':
      if (!memcmp(name, "getnetbyname", 12)) {
      /*                          ^         */
        return -KEY_getnetbyname;
      }
      break;
    case 'd':
      if (!memcmp(name, "getnetbyaddr", 12)) {
      /*                          ^         */
        return -KEY_getnetbyaddr;
      }
      break;
    }
    break;
  case 13:
    /* Names all of length 13.  */
    /* gethostbyaddr gethostbyname getservbyname getservbyport */
    /* Offset 10 gives the best switch position.  */
    switch (name[10]) {
    case 'a':
      if (!memcmp(name, "gethostbyname", 13)) {
      /*                           ^         */
        return -KEY_gethostbyname; /* Weight 970 */
      }
      if (!memcmp(name, "getservbyname", 13)) {
      /*                           ^         */
        return -KEY_getservbyname; /* Weight 299 */
      }
      break;
    case 'd':
      if (!memcmp(name, "gethostbyaddr", 13)) {
      /*                           ^         */
        return -KEY_gethostbyaddr; /* Weight 68 */
      }
      break;
    case 'o':
      if (!memcmp(name, "getservbyport", 13)) {
      /*                           ^         */
        return -KEY_getservbyport;
      }
      break;
    }
    break;
  case 14:
    if (!memcmp(name, "getprotobyname", 14)) {
      return -KEY_getprotobyname; /* Weight 755 */
    }
    break;
  case 16:
    if (!memcmp(name, "getprotobynumber", 16)) {
      return -KEY_getprotobynumber; /* Weight 232 */
    }
    break;
  }
  return 0;
}

STATIC void
S_checkcomma(pTHX_ register char *s, char *name, char *what)
{
    char *w;

    if (*s == ' ' && s[1] == '(') {     /* XXX gotta be a better way */
        if (ckWARN(WARN_SYNTAX)) {
            int level = 1;
            for (w = s+2; *w && level; w++) {
                if (*w == '(')
                    ++level;
                else if (*w == ')')
                    --level;
            }
            if (*w)
                for (; *w && isSPACE(*w); w++) ;
            if (!*w || !strchr(";|})]oaiuw!=", *w))     /* an advisory hack only... */
                Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                            "%s (...) interpreted as function",name);
        }
    }
    while (s < PL_bufend && isSPACE(*s))
        s++;
    if (*s == '(')
        s++;
    while (s < PL_bufend && isSPACE(*s))
        s++;
    if (isIDFIRST_lazy_if(s,UTF)) {
        w = s++;
        while (isALNUM_lazy_if(s,UTF))
            s++;
        while (s < PL_bufend && isSPACE(*s))
            s++;
        if (*s == ',') {
            int kw;
            *s = '\0';
            kw = keyword(w, s - w) || get_cv(w, FALSE) != 0;
            *s = ',';
            if (kw)
                return;
            Perl_croak(aTHX_ "No comma allowed after %s", what);
        }
    }
}

/* Either returns sv, or mortalizes sv and returns a new SV*.
   Best used as sv=new_constant(..., sv, ...).
   If s, pv are NULL, calls subroutine with one argument,
   and type is used with error messages only. */

STATIC SV *
S_new_constant(pTHX_ char *s, STRLEN len, const char *key, SV *sv, SV *pv,
               const char *type)
{
    dSP;
    HV *table = GvHV(PL_hintgv);                 /* ^H */
    SV *res;
    SV **cvp;
    SV *cv, *typesv;
    const char *why1, *why2, *why3;

    if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
        SV *msg;
        
        why2 = strEQ(key,"charnames")
               ? "(possibly a missing \"use charnames ...\")"
               : "";
        msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
                            (type ? type: "undef"), why2);

        /* This is convoluted and evil ("goto considered harmful")
         * but I do not understand the intricacies of all the different
         * failure modes of %^H in here.  The goal here is to make
         * the most probable error message user-friendly. --jhi */

        goto msgdone;

    report:
        msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
                            (type ? type: "undef"), why1, why2, why3);
    msgdone:
        yyerror(SvPVX(msg));
        SvREFCNT_dec(msg);
        return sv;
    }
    cvp = hv_fetch(table, key, strlen(key), FALSE);
    if (!cvp || !SvOK(*cvp)) {
        why1 = "$^H{";
        why2 = key;
        why3 = "} is not defined";
        goto report;
    }
    sv_2mortal(sv);                     /* Parent created it permanently */
    cv = *cvp;
    if (!pv && s)
        pv = sv_2mortal(newSVpvn(s, len));
    if (type && pv)
        typesv = sv_2mortal(newSVpv(type, 0));
    else
        typesv = &PL_sv_undef;

    PUSHSTACKi(PERLSI_OVERLOAD);
    ENTER ;
    SAVETMPS;

    PUSHMARK(SP) ;
    EXTEND(sp, 3);
    if (pv)
        PUSHs(pv);
    PUSHs(sv);
    if (pv)
        PUSHs(typesv);
    PUTBACK;
    call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));

    SPAGAIN ;

    /* Check the eval first */
    if (!PL_in_eval && SvTRUE(ERRSV)) {
        STRLEN n_a;
        sv_catpv(ERRSV, "Propagated");
        yyerror(SvPV(ERRSV, n_a)); /* Duplicates the message inside eval */
        (void)POPs;
        res = SvREFCNT_inc(sv);
    }
    else {
        res = POPs;
        (void)SvREFCNT_inc(res);
    }

    PUTBACK ;
    FREETMPS ;
    LEAVE ;
    POPSTACK;

    if (!SvOK(res)) {
        why1 = "Call to &{$^H{";
        why2 = key;
        why3 = "}} did not return a defined value";
        sv = res;
        goto report;
    }

    return res;
}

/* Returns a NUL terminated string, with the length of the string written to
   *slp
   */
STATIC char *
S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
{
    register char *d = dest;
    register char *e = d + destlen - 3;  /* two-character token, ending NUL */
    for (;;) {
        if (d >= e)
            Perl_croak(aTHX_ ident_too_long);
        if (isALNUM(*s))        /* UTF handled below */
            *d++ = *s++;
        else if (*s == '\'' && allow_package && isIDFIRST_lazy_if(s+1,UTF)) {
            *d++ = ':';
            *d++ = ':';
            s++;
        }
        else if (*s == ':' && s[1] == ':' && allow_package && s[2] != '$') {
            *d++ = *s++;
            *d++ = *s++;
        }
        else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
            char *t = s + UTF8SKIP(s);
            while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
                t += UTF8SKIP(t);
            if (d + (t - s) > e)
                Perl_croak(aTHX_ ident_too_long);
            Copy(s, d, t - s, char);
            d += t - s;
            s = t;
        }
        else {
            *d = '\0';
            *slp = d - dest;
            return s;
        }
    }
}

STATIC char *
S_scan_ident(pTHX_ register char *s, register char *send, char *dest, STRLEN destlen, I32 ck_uni)
{
    register char *d;
    register char *e;
    char *bracket = 0;
    char funny = *s++;

    if (isSPACE(*s))
        s = skipspace(s);
    d = dest;
    e = d + destlen - 3;        /* two-character token, ending NUL */
    if (isDIGIT(*s)) {
        while (isDIGIT(*s)) {
            if (d >= e)
                Perl_croak(aTHX_ ident_too_long);
            *d++ = *s++;
        }
    }
    else {
        for (;;) {
            if (d >= e)
                Perl_croak(aTHX_ ident_too_long);
            if (isALNUM(*s))    /* UTF handled below */
                *d++ = *s++;
            else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
                *d++ = ':';
                *d++ = ':';
                s++;
            }
            else if (*s == ':' && s[1] == ':') {
                *d++ = *s++;
                *d++ = *s++;
            }
            else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
                char *t = s + UTF8SKIP(s);
                while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
                    t += UTF8SKIP(t);
                if (d + (t - s) > e)
                    Perl_croak(aTHX_ ident_too_long);
                Copy(s, d, t - s, char);
                d += t - s;
                s = t;
            }
            else
                break;
        }
    }
    *d = '\0';
    d = dest;
    if (*d) {
        if (PL_lex_state != LEX_NORMAL)
            PL_lex_state = LEX_INTERPENDMAYBE;
        return s;
    }
    if (*s == '$' && s[1] &&
        (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
    {
        return s;
    }
    if (*s == '{') {
        bracket = s;
        s++;
    }
    else if (ck_uni)
        check_uni();
    if (s < send)
        *d = *s++;
    d[1] = '\0';
    if (*d == '^' && *s && isCONTROLVAR(*s)) {
        *d = toCTRL(*s);
        s++;
    }
    if (bracket) {
        if (isSPACE(s[-1])) {
            while (s < send) {
                char ch = *s++;
                if (!SPACE_OR_TAB(ch)) {
                    *d = ch;
                    break;
                }
            }
        }
        if (isIDFIRST_lazy_if(d,UTF)) {
            d++;
            if (UTF) {
                e = s;
                while ((e < send && isALNUM_lazy_if(e,UTF)) || *e == ':') {
                    e += UTF8SKIP(e);
                    while (e < send && UTF8_IS_CONTINUED(*e) && is_utf8_mark((U8*)e))
                        e += UTF8SKIP(e);
                }
                Copy(s, d, e - s, char);
                d += e - s;
                s = e;
            }
            else {
                while ((isALNUM(*s) || *s == ':') && d < e)
                    *d++ = *s++;
                if (d >= e)
                    Perl_croak(aTHX_ ident_too_long);
            }
            *d = '\0';
            while (s < send && SPACE_OR_TAB(*s)) s++;
            if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
                if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest)) {
                    const char *brack = *s == '[' ? "[...]" : "{...}";
                    Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
                        "Ambiguous use of %c{%s%s} resolved to %c%s%s",
                        funny, dest, brack, funny, dest, brack);
                }
                bracket++;
                PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
                return s;
            }
        }
        /* Handle extended ${^Foo} variables
         * 1999-02-27 mjd-perl-patch@plover.com */
        else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
                 && isALNUM(*s))
        {
            d++;
            while (isALNUM(*s) && d < e) {
                *d++ = *s++;
            }
            if (d >= e)
                Perl_croak(aTHX_ ident_too_long);
            *d = '\0';
        }
        if (*s == '}') {
            s++;
            if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
                PL_lex_state = LEX_INTERPEND;
                PL_expect = XREF;
            }
            if (funny == '#')
                funny = '@';
            if (PL_lex_state == LEX_NORMAL) {
                if (ckWARN(WARN_AMBIGUOUS) &&
                    (keyword(dest, d - dest) || get_cv(dest, FALSE)))
                {
                    Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
                        "Ambiguous use of %c{%s} resolved to %c%s",
                        funny, dest, funny, dest);
                }
            }
        }
        else {
            s = bracket;                /* let the parser handle it */
            *dest = '\0';
        }
    }
    else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
        PL_lex_state = LEX_INTERPEND;
    return s;
}

void
Perl_pmflag(pTHX_ U32* pmfl, int ch)
{
    if (ch == 'i')
        *pmfl |= PMf_FOLD;
    else if (ch == 'g')
        *pmfl |= PMf_GLOBAL;
    else if (ch == 'c')
        *pmfl |= PMf_CONTINUE;
    else if (ch == 'o')
        *pmfl |= PMf_KEEP;
    else if (ch == 'm')
        *pmfl |= PMf_MULTILINE;
    else if (ch == 's')
        *pmfl |= PMf_SINGLELINE;
    else if (ch == 'x')
        *pmfl |= PMf_EXTENDED;
}

STATIC char *
S_scan_pat(pTHX_ char *start, I32 type)
{
    PMOP *pm;
    char *s;

    s = scan_str(start,FALSE,FALSE);
    if (!s)
        Perl_croak(aTHX_ "Search pattern not terminated");

    pm = (PMOP*)newPMOP(type, 0);
    if (PL_multi_open == '?')
        pm->op_pmflags |= PMf_ONCE;
    if(type == OP_QR) {
        while (*s && strchr("iomsx", *s))
            pmflag(&pm->op_pmflags,*s++);
    }
    else {
        while (*s && strchr("iogcmsx", *s))
            pmflag(&pm->op_pmflags,*s++);
    }
    /* issue a warning if /c is specified,but /g is not */
    if (ckWARN(WARN_REGEXP) && 
        (pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
    {
        Perl_warner(aTHX_ packWARN(WARN_REGEXP), c_without_g);
    }

    pm->op_pmpermflags = pm->op_pmflags;

    PL_lex_op = (OP*)pm;
    yylval.ival = OP_MATCH;
    return s;
}

STATIC char *
S_scan_subst(pTHX_ char *start)
{
    register char *s;
    register PMOP *pm;
    I32 first_start;
    I32 es = 0;

    yylval.ival = OP_NULL;

    s = scan_str(start,FALSE,FALSE);

    if (!s)
        Perl_croak(aTHX_ "Substitution pattern not terminated");

    if (s[-1] == PL_multi_open)
        s--;

    first_start = PL_multi_start;
    s = scan_str(s,FALSE,FALSE);
    if (!s) {
        if (PL_lex_stuff) {
            SvREFCNT_dec(PL_lex_stuff);
            PL_lex_stuff = Nullsv;
        }
        Perl_croak(aTHX_ "Substitution replacement not terminated");
    }
    PL_multi_start = first_start;       /* so whole substitution is taken together */

    pm = (PMOP*)newPMOP(OP_SUBST, 0);
    while (*s) {
        if (*s == 'e') {
            s++;
            es++;
        }
        else if (strchr("iogcmsx", *s))
            pmflag(&pm->op_pmflags,*s++);
        else
            break;
    }

    /* /c is not meaningful with s/// */
    if (ckWARN(WARN_REGEXP) && (pm->op_pmflags & PMf_CONTINUE))
    {
        Perl_warner(aTHX_ packWARN(WARN_REGEXP), c_in_subst);
    }

    if (es) {
        SV *repl;
        PL_sublex_info.super_bufptr = s;
        PL_sublex_info.super_bufend = PL_bufend;
        PL_multi_end = 0;
        pm->op_pmflags |= PMf_EVAL;
        repl = newSVpvn("",0);
        while (es-- > 0)
            sv_catpv(repl, es ? "eval " : "do ");
        sv_catpvn(repl, "{ ", 2);
        sv_catsv(repl, PL_lex_repl);
        sv_catpvn(repl, " };", 2);
        SvEVALED_on(repl);
        SvREFCNT_dec(PL_lex_repl);
        PL_lex_repl = repl;
    }

    pm->op_pmpermflags = pm->op_pmflags;
    PL_lex_op = (OP*)pm;
    yylval.ival = OP_SUBST;
    return s;
}

STATIC char *
S_scan_trans(pTHX_ char *start)
{
    register char* s;
    OP *o;
    short *tbl;
    I32 squash;
    I32 del;
    I32 complement;

    yylval.ival = OP_NULL;

    s = scan_str(start,FALSE,FALSE);
    if (!s)
        Perl_croak(aTHX_ "Transliteration pattern not terminated");
    if (s[-1] == PL_multi_open)
        s--;

    s = scan_str(s,FALSE,FALSE);
    if (!s) {
        if (PL_lex_stuff) {
            SvREFCNT_dec(PL_lex_stuff);
            PL_lex_stuff = Nullsv;
        }
        Perl_croak(aTHX_ "Transliteration replacement not terminated");
    }

    complement = del = squash = 0;
    while (1) {
        switch (*s) {
        case 'c':
            complement = OPpTRANS_COMPLEMENT;
            break;
        case 'd':
            del = OPpTRANS_DELETE;
            break;
        case 's':
            squash = OPpTRANS_SQUASH;
            break;
        default:
            goto no_more;
        }
        s++;
    }
  no_more:

    New(803, tbl, complement&&!del?258:256, short);
    o = newPVOP(OP_TRANS, 0, (char*)tbl);
    o->op_private &= ~OPpTRANS_ALL;
    o->op_private |= del|squash|complement|
      (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
      (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF   : 0);

    PL_lex_op = o;
    yylval.ival = OP_TRANS;
    return s;
}

STATIC char *
S_scan_heredoc(pTHX_ register char *s)
{
    SV *herewas;
    I32 op_type = OP_SCALAR;
    I32 len;
    SV *tmpstr;
    char term;
    register char *d;
    register char *e;
    char *peek;
    int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));

    s += 2;
    d = PL_tokenbuf;
    e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
    if (!outer)
        *d++ = '\n';
    for (peek = s; SPACE_OR_TAB(*peek); peek++) ;
    if (*peek == '`' || *peek == '\'' || *peek =='"') {
        s = peek;
        term = *s++;
        s = delimcpy(d, e, s, PL_bufend, term, &len);
        d += len;
        if (s < PL_bufend)
            s++;
    }
    else {
        if (*s == '\\')
            s++, term = '\'';
        else
            term = '"';
        if (!isALNUM_lazy_if(s,UTF))
            deprecate_old("bare << to mean <<\"\"");
        for (; isALNUM_lazy_if(s,UTF); s++) {
            if (d < e)
                *d++ = *s;
        }
    }
    if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
        Perl_croak(aTHX_ "Delimiter for here document is too long");
    *d++ = '\n';
    *d = '\0';
    len = d - PL_tokenbuf;
#ifndef PERL_STRICT_CR
    d = strchr(s, '\r');
    if (d) {
        char *olds = s;
        s = d;
        while (s < PL_bufend) {
            if (*s == '\r') {
                *d++ = '\n';
                if (*++s == '\n')
                    s++;
            }
            else if (*s == '\n' && s[1] == '\r') {      /* \015\013 on a mac? */
                *d++ = *s++;
                s++;
            }
            else
                *d++ = *s++;
        }
        *d = '\0';
        PL_bufend = d;
        SvCUR_set(PL_linestr, PL_bufend - SvPVX(PL_linestr));
        s = olds;
    }
#endif
    d = "\n";
    if (outer || !(d=ninstr(s,PL_bufend,d,d+1)))
        herewas = newSVpvn(s,PL_bufend-s);
    else
        s--, herewas = newSVpvn(s,d-s);
    s += SvCUR(herewas);

    tmpstr = NEWSV(87,79);
    sv_upgrade(tmpstr, SVt_PVIV);
    if (term == '\'') {
        op_type = OP_CONST;
        SvIVX(tmpstr) = -1;
    }
    else if (term == '`') {
        op_type = OP_BACKTICK;
        SvIVX(tmpstr) = '\\';
    }

    CLINE;
    PL_multi_start = CopLINE(PL_curcop);
    PL_multi_open = PL_multi_close = '<';
    term = *PL_tokenbuf;
    if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
        char *bufptr = PL_sublex_info.super_bufptr;
        char *bufend = PL_sublex_info.super_bufend;
        char *olds = s - SvCUR(herewas);
        s = strchr(bufptr, '\n');
        if (!s)
            s = bufend;
        d = s;
        while (s < bufend &&
          (*s != term || memNE(s,PL_tokenbuf,len)) ) {
            if (*s++ == '\n')
                CopLINE_inc(PL_curcop);
        }
        if (s >= bufend) {
            CopLINE_set(PL_curcop, (line_t)PL_multi_start);
            missingterm(PL_tokenbuf);
        }
        sv_setpvn(herewas,bufptr,d-bufptr+1);
        sv_setpvn(tmpstr,d+1,s-d);
        s += len - 1;
        sv_catpvn(herewas,s,bufend-s);
        Copy(SvPVX(herewas),bufptr,SvCUR(herewas) + 1,char);

        s = olds;
        goto retval;
    }
    else if (!outer) {
        d = s;
        while (s < PL_bufend &&
          (*s != term || memNE(s,PL_tokenbuf,len)) ) {
            if (*s++ == '\n')
                CopLINE_inc(PL_curcop);
        }
        if (s >= PL_bufend) {
            CopLINE_set(PL_curcop, (line_t)PL_multi_start);
            missingterm(PL_tokenbuf);
        }
        sv_setpvn(tmpstr,d+1,s-d);
        s += len - 1;
        CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */

        sv_catpvn(herewas,s,PL_bufend-s);
        sv_setsv(PL_linestr,herewas);
        PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
        PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
        PL_last_lop = PL_last_uni = Nullch;
    }
    else
        sv_setpvn(tmpstr,"",0);   /* avoid "uninitialized" warning */
    while (s >= PL_bufend) {    /* multiple line string? */
        if (!outer ||
         !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
            CopLINE_set(PL_curcop, (line_t)PL_multi_start);
            missingterm(PL_tokenbuf);
        }
        CopLINE_inc(PL_curcop);
        PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
        PL_last_lop = PL_last_uni = Nullch;
#ifndef PERL_STRICT_CR
        if (PL_bufend - PL_linestart >= 2) {
            if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
                (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
            {
                PL_bufend[-2] = '\n';
                PL_bufend--;
                SvCUR_set(PL_linestr, PL_bufend - SvPVX(PL_linestr));
            }
            else if (PL_bufend[-1] == '\r')
                PL_bufend[-1] = '\n';
        }
        else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
            PL_bufend[-1] = '\n';
#endif
        if (PERLDB_LINE && PL_curstash != PL_debstash) {
            SV *sv = NEWSV(88,0);

            sv_upgrade(sv, SVt_PVMG);
            sv_setsv(sv,PL_linestr);
            (void)SvIOK_on(sv);
            SvIVX(sv) = 0;
            av_store(CopFILEAV(PL_curcop), (I32)CopLINE(PL_curcop),sv);
        }
        if (*s == term && memEQ(s,PL_tokenbuf,len)) {
            STRLEN off = PL_bufend - 1 - SvPVX(PL_linestr);
            *(SvPVX(PL_linestr) + off ) = ' ';
            sv_catsv(PL_linestr,herewas);
            PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
            s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
        }
        else {
            s = PL_bufend;
            sv_catsv(tmpstr,PL_linestr);
        }
    }
    s++;
retval:
    PL_multi_end = CopLINE(PL_curcop);
    if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
        SvLEN_set(tmpstr, SvCUR(tmpstr) + 1);
        Renew(SvPVX(tmpstr), SvLEN(tmpstr), char);
    }
    SvREFCNT_dec(herewas);
    if (!IN_BYTES) {
        if (UTF && is_utf8_string((U8*)SvPVX(tmpstr), SvCUR(tmpstr)))
            SvUTF8_on(tmpstr);
        else if (PL_encoding)
            sv_recode_to_utf8(tmpstr, PL_encoding);
    }
    PL_lex_stuff = tmpstr;
    yylval.ival = op_type;
    return s;
}

/* scan_inputsymbol
   takes: current position in input buffer
   returns: new position in input buffer
   side-effects: yylval and lex_op are set.

   This code handles:

   <>           read from ARGV
   <FH>         read from filehandle
   <pkg::FH>    read from package qualified filehandle
   <pkg'FH>     read from package qualified filehandle
   <$fh>        read from filehandle in $fh
   <*.h>        filename glob

*/

STATIC char *
S_scan_inputsymbol(pTHX_ char *start)
{
    register char *s = start;           /* current position in buffer */
    register char *d;
    register char *e;
    char *end;
    I32 len;

    d = PL_tokenbuf;                    /* start of temp holding space */
    e = PL_tokenbuf + sizeof PL_tokenbuf;       /* end of temp holding space */
    end = strchr(s, '\n');
    if (!end)
        end = PL_bufend;
    s = delimcpy(d, e, s + 1, end, '>', &len);  /* extract until > */

    /* die if we didn't have space for the contents of the <>,
       or if it didn't end, or if we see a newline
    */

    if (len >= sizeof PL_tokenbuf)
        Perl_croak(aTHX_ "Excessively long <> operator");
    if (s >= end)
        Perl_croak(aTHX_ "Unterminated <> operator");

    s++;

    /* check for <$fh>
       Remember, only scalar variables are interpreted as filehandles by
       this code.  Anything more complex (e.g., <$fh{$num}>) will be
       treated as a glob() call.
       This code makes use of the fact that except for the $ at the front,
       a scalar variable and a filehandle look the same.
    */
    if (*d == '$' && d[1]) d++;

    /* allow <Pkg'VALUE> or <Pkg::VALUE> */
    while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
        d++;

    /* If we've tried to read what we allow filehandles to look like, and
       there's still text left, then it must be a glob() and not a getline.
       Use scan_str to pull out the stuff between the <> and treat it
       as nothing more than a string.
    */

    if (d - PL_tokenbuf != len) {
        yylval.ival = OP_GLOB;
        set_csh();
        s = scan_str(start,FALSE,FALSE);
        if (!s)
           Perl_croak(aTHX_ "Glob not terminated");
        return s;
    }
    else {
        bool readline_overriden = FALSE;
        GV *gv_readline = Nullgv;
        GV **gvp;
        /* we're in a filehandle read situation */
        d = PL_tokenbuf;

        /* turn <> into <ARGV> */
        if (!len)
            Copy("ARGV",d,5,char);

        /* Check whether readline() is overriden */
        if (((gv_readline = gv_fetchpv("readline", FALSE, SVt_PVCV))
                && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
                ||
                ((gvp = (GV**)hv_fetch(PL_globalstash, "readline", 8, FALSE))
                && (gv_readline = *gvp) != (GV*)&PL_sv_undef
                && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
            readline_overriden = TRUE;

        /* if <$fh>, create the ops to turn the variable into a
           filehandle
        */
        if (*d == '$') {
            I32 tmp;

            /* try to find it in the pad for this block, otherwise find
               add symbol table ops
            */
            if ((tmp = pad_findmy(d)) != NOT_IN_PAD) {
                if (PAD_COMPNAME_FLAGS(tmp) & SVpad_OUR) {
                    SV *sym = sv_2mortal(
                            newSVpv(HvNAME(PAD_COMPNAME_OURSTASH(tmp)),0));
                    sv_catpvn(sym, "::", 2);
                    sv_catpv(sym, d+1);
                    d = SvPVX(sym);
                    goto intro_sym;
                }
                else {
                    OP *o = newOP(OP_PADSV, 0);
                    o->op_targ = tmp;
                    PL_lex_op = readline_overriden
                        ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
                                append_elem(OP_LIST, o,
                                    newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
                        : (OP*)newUNOP(OP_READLINE, 0, o);
                }
            }
            else {
                GV *gv;
                ++d;
intro_sym:
                gv = gv_fetchpv(d,
                                (PL_in_eval
                                 ? (GV_ADDMULTI | GV_ADDINEVAL)
                                 : GV_ADDMULTI),
                                SVt_PV);
                PL_lex_op = readline_overriden
                    ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
                            append_elem(OP_LIST,
                                newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
                                newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
                    : (OP*)newUNOP(OP_READLINE, 0,
                            newUNOP(OP_RV2SV, 0,
                                newGVOP(OP_GV, 0, gv)));
            }
            if (!readline_overriden)
                PL_lex_op->op_flags |= OPf_SPECIAL;
            /* we created the ops in PL_lex_op, so make yylval.ival a null op */
            yylval.ival = OP_NULL;
        }

        /* If it's none of the above, it must be a literal filehandle
           (<Foo::BAR> or <FOO>) so build a simple readline OP */
        else {
            GV *gv = gv_fetchpv(d,TRUE, SVt_PVIO);
            PL_lex_op = readline_overriden
                ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
                        append_elem(OP_LIST,
                            newGVOP(OP_GV, 0, gv),
                            newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
                : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
            yylval.ival = OP_NULL;
        }
    }

    return s;
}


/* scan_str
   takes: start position in buffer
          keep_quoted preserve \ on the embedded delimiter(s)
          keep_delims preserve the delimiters around the string
   returns: position to continue reading from buffer
   side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
        updates the read buffer.

   This subroutine pulls a string out of the input.  It is called for:
        q               single quotes           q(literal text)
        '               single quotes           'literal text'
        qq              double quotes           qq(interpolate $here please)
        "               double quotes           "interpolate $here please"
        qx              backticks               qx(/bin/ls -l)
        `               backticks               `/bin/ls -l`
        qw              quote words             @EXPORT_OK = qw( func() $spam )
        m//             regexp match            m/this/
        s///            regexp substitute       s/this/that/
        tr///           string transliterate    tr/this/that/
        y///            string transliterate    y/this/that/
        ($*@)           sub prototypes          sub foo ($)
        (stuff)         sub attr parameters     sub foo : attr(stuff)
        <>              readline or globs       <FOO>, <>, <$fh>, or <*.c>
        
   In most of these cases (all but <>, patterns and transliterate)
   yylex() calls scan_str().  m// makes yylex() call scan_pat() which
   calls scan_str().  s/// makes yylex() call scan_subst() which calls
   scan_str().  tr/// and y/// make yylex() call scan_trans() which
   calls scan_str().

   It skips whitespace before the string starts, and treats the first
   character as the delimiter.  If the delimiter is one of ([{< then
   the corresponding "close" character )]}> is used as the closing
   delimiter.  It allows quoting of delimiters, and if the string has
   balanced delimiters ([{<>}]) it allows nesting.

   On success, the SV with the resulting string is put into lex_stuff or,
   if that is already non-NULL, into lex_repl. The second case occurs only
   when parsing the RHS of the special constructs s/// and tr/// (y///).
   For convenience, the terminating delimiter character is stuffed into
   SvIVX of the SV.
*/

STATIC char *
S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
{
    SV *sv;                             /* scalar value: string */
    char *tmps;                         /* temp string, used for delimiter matching */
    register char *s = start;           /* current position in the buffer */
    register char term;                 /* terminating character */
    register char *to;                  /* current position in the sv's data */
    I32 brackets = 1;                   /* bracket nesting level */
    bool has_utf8 = FALSE;              /* is there any utf8 content? */
    I32 termcode;                       /* terminating char. code */
    U8 termstr[UTF8_MAXBYTES];          /* terminating string */
    STRLEN termlen;                     /* length of terminating string */
    char *last = NULL;                  /* last position for nesting bracket */

    /* skip space before the delimiter */
    if (isSPACE(*s))
        s = skipspace(s);

    /* mark where we are, in case we need to report errors */
    CLINE;

    /* after skipping whitespace, the next character is the terminator */
    term = *s;
    if (!UTF) {
        termcode = termstr[0] = term;
        termlen = 1;
    }
    else {
        termcode = utf8_to_uvchr((U8*)s, &termlen);
        Copy(s, termstr, termlen, U8);
        if (!UTF8_IS_INVARIANT(term))
            has_utf8 = TRUE;
    }

    /* mark where we are */
    PL_multi_start = CopLINE(PL_curcop);
    PL_multi_open = term;

    /* find corresponding closing delimiter */
    if (term && (tmps = strchr("([{< )]}> )]}>",term)))
        termcode = termstr[0] = term = tmps[5];

    PL_multi_close = term;

    /* create a new SV to hold the contents.  87 is leak category, I'm
       assuming.  79 is the SV's initial length.  What a random number. */
    sv = NEWSV(87,79);
    sv_upgrade(sv, SVt_PVIV);
    SvIVX(sv) = termcode;
    (void)SvPOK_only(sv);               /* validate pointer */

    /* move past delimiter and try to read a complete string */
    if (keep_delims)
        sv_catpvn(sv, s, termlen);
    s += termlen;
    for (;;) {
        if (PL_encoding && !UTF) {
            bool cont = TRUE;

            while (cont) {
                int offset = s - SvPVX(PL_linestr);
                bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
                                           &offset, (char*)termstr, termlen);
                char *ns = SvPVX(PL_linestr) + offset;
                char *svlast = SvEND(sv) - 1;

                for (; s < ns; s++) {
                    if (*s == '\n' && !PL_rsfp)
                        CopLINE_inc(PL_curcop);
                }
                if (!found)
                    goto read_more_line;
                else {
                    /* handle quoted delimiters */
                    if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
                        char *t;
                        for (t = svlast-2; t >= SvPVX(sv) && *t == '\\';)
                            t--;
                        if ((svlast-1 - t) % 2) {
                            if (!keep_quoted) {
                                *(svlast-1) = term;
                                *svlast = '\0';
                                SvCUR_set(sv, SvCUR(sv) - 1);
                            }
                            continue;
                        }
                    }
                    if (PL_multi_open == PL_multi_close) {
                        cont = FALSE;
                    }
                    else {
                        char *t, *w;
                        if (!last)
                            last = SvPVX(sv);
                        for (w = t = last; t < svlast; w++, t++) {
                            /* At here, all closes are "was quoted" one,
                               so we don't check PL_multi_close. */
                            if (*t == '\\') {
                                if (!keep_quoted && *(t+1) == PL_multi_open)
                                    t++;
                                else
                                    *w++ = *t++;
                            }
                            else if (*t == PL_multi_open)
                                brackets++;

                            *w = *t;
                        }
                        if (w < t) {
                            *w++ = term;
                            *w = '\0';
                            SvCUR_set(sv, w - SvPVX(sv));
                        }
                        last = w;
                        if (--brackets <= 0)
                            cont = FALSE;
                    }
                }
            }
            if (!keep_delims) {
                SvCUR_set(sv, SvCUR(sv) - 1);
                *SvEND(sv) = '\0';
            }
            break;
        }

        /* extend sv if need be */
        SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
        /* set 'to' to the next character in the sv's string */
        to = SvPVX(sv)+SvCUR(sv);

        /* if open delimiter is the close delimiter read unbridle */
        if (PL_multi_open == PL_multi_close) {
            for (; s < PL_bufend; s++,to++) {
                /* embedded newlines increment the current line number */
                if (*s == '\n' && !PL_rsfp)
                    CopLINE_inc(PL_curcop);
                /* handle quoted delimiters */
                if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
                    if (!keep_quoted && s[1] == term)
                        s++;
                /* any other quotes are simply copied straight through */
                    else
                        *to++ = *s++;
                }
                /* terminate when run out of buffer (the for() condition), or
                   have found the terminator */
                else if (*s == term) {
                    if (termlen == 1)
                        break;
                    if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
                        break;
                }
                else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
                    has_utf8 = TRUE;
                *to = *s;
            }
        }
        
        /* if the terminator isn't the same as the start character (e.g.,
           matched brackets), we have to allow more in the quoting, and
           be prepared for nested brackets.
        */
        else {
            /* read until we run out of string, or we find the terminator */
            for (; s < PL_bufend; s++,to++) {
                /* embedded newlines increment the line count */
                if (*s == '\n' && !PL_rsfp)
                    CopLINE_inc(PL_curcop);
                /* backslashes can escape the open or closing characters */
                if (*s == '\\' && s+1 < PL_bufend) {
                    if (!keep_quoted &&
                        ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
                        s++;
                    else
                        *to++ = *s++;
                }
                /* allow nested opens and closes */
                else if (*s == PL_multi_close && --brackets <= 0)
                    break;
                else if (*s == PL_multi_open)
                    brackets++;
                else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
                    has_utf8 = TRUE;
                *to = *s;
            }
        }
        /* terminate the copied string and update the sv's end-of-string */
        *to = '\0';
        SvCUR_set(sv, to - SvPVX(sv));

        /*
         * this next chunk reads more into the buffer if we're not done yet
         */

        if (s < PL_bufend)
            break;              /* handle case where we are done yet :-) */

#ifndef PERL_STRICT_CR
        if (to - SvPVX(sv) >= 2) {
            if ((to[-2] == '\r' && to[-1] == '\n') ||
                (to[-2] == '\n' && to[-1] == '\r'))
            {
                to[-2] = '\n';
                to--;
                SvCUR_set(sv, to - SvPVX(sv));
            }
            else if (to[-1] == '\r')
                to[-1] = '\n';
        }
        else if (to - SvPVX(sv) == 1 && to[-1] == '\r')
            to[-1] = '\n';
#endif
        
     read_more_line:
        /* if we're out of file, or a read fails, bail and reset the current
           line marker so we can report where the unterminated string began
        */
        if (!PL_rsfp ||
         !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
            sv_free(sv);
            CopLINE_set(PL_curcop, (line_t)PL_multi_start);
            return Nullch;
        }
        /* we read a line, so increment our line counter */
        CopLINE_inc(PL_curcop);

        /* update debugger info */
        if (PERLDB_LINE && PL_curstash != PL_debstash) {
            SV *sv = NEWSV(88,0);

            sv_upgrade(sv, SVt_PVMG);
            sv_setsv(sv,PL_linestr);
            (void)SvIOK_on(sv);
            SvIVX(sv) = 0;
            av_store(CopFILEAV(PL_curcop), (I32)CopLINE(PL_curcop), sv);
        }

        /* having changed the buffer, we must update PL_bufend */
        PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
        PL_last_lop = PL_last_uni = Nullch;
    }

    /* at this point, we have successfully read the delimited string */

    if (!PL_encoding || UTF) {
        if (keep_delims)
            sv_catpvn(sv, s, termlen);
        s += termlen;
    }
    if (has_utf8 || PL_encoding)
        SvUTF8_on(sv);

    PL_multi_end = CopLINE(PL_curcop);

    /* if we allocated too much space, give some back */
    if (SvCUR(sv) + 5 < SvLEN(sv)) {
        SvLEN_set(sv, SvCUR(sv) + 1);
        Renew(SvPVX(sv), SvLEN(sv), char);
    }

    /* decide whether this is the first or second quoted string we've read
       for this op
    */

    if (PL_lex_stuff)
        PL_lex_repl = sv;
    else
        PL_lex_stuff = sv;
    return s;
}

/*
  scan_num
  takes: pointer to position in buffer
  returns: pointer to new position in buffer
  side-effects: builds ops for the constant in yylval.op

  Read a number in any of the formats that Perl accepts:

  \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*)      12 12.34 12.
  \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*)                     .34
  0b[01](_?[01])*
  0[0-7](_?[0-7])*
  0x[0-9A-Fa-f](_?[0-9A-Fa-f])*

  Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
  thing it reads.

  If it reads a number without a decimal point or an exponent, it will
  try converting the number to an integer and see if it can do so
  without loss of precision.
*/

char *
Perl_scan_num(pTHX_ char *start, YYSTYPE* lvalp)
{
    register char *s = start;           /* current position in buffer */
    register char *d;                   /* destination in temp buffer */
    register char *e;                   /* end of temp buffer */
    NV nv;                              /* number read, as a double */
    SV *sv = Nullsv;                    /* place to put the converted number */
    bool floatit;                       /* boolean: int or float? */
    char *lastub = 0;                   /* position of last underbar */
    static char number_too_long[] = "Number too long";

    /* We use the first character to decide what type of number this is */

    switch (*s) {
    default:
      Perl_croak(aTHX_ "panic: scan_num");

    /* if it starts with a 0, it could be an octal number, a decimal in
       0.13 disguise, or a hexadecimal number, or a binary number. */
    case '0':
        {
          /* variables:
             u          holds the "number so far"
             shift      the power of 2 of the base
                        (hex == 4, octal == 3, binary == 1)
             overflowed was the number more than we can hold?

             Shift is used when we add a digit.  It also serves as an "are
             we in octal/hex/binary?" indicator to disallow hex characters
             when in octal mode.
           */
            NV n = 0.0;
            UV u = 0;
            I32 shift;
            bool overflowed = FALSE;
            bool just_zero  = TRUE;     /* just plain 0 or binary number? */
            static NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
            static char* bases[5] = { "", "binary", "", "octal",
                                      "hexadecimal" };
            static char* Bases[5] = { "", "Binary", "", "Octal",
                                      "Hexadecimal" };
            static char *maxima[5] = { "",
                                       "0b11111111111111111111111111111111",
                                       "",
                                       "037777777777",
                                       "0xffffffff" };
            char *base, *Base, *max;

            /* check for hex */
            if (s[1] == 'x') {
                shift = 4;
                s += 2;
                just_zero = FALSE;
            } else if (s[1] == 'b') {
                shift = 1;
                s += 2;
                just_zero = FALSE;
            }
            /* check for a decimal in disguise */
            else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
                goto decimal;
            /* so it must be octal */
            else {
                shift = 3;
                s++;
            }

            if (*s == '_') {
               if (ckWARN(WARN_SYNTAX))
                   Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                               "Misplaced _ in number");
               lastub = s++;
            }

            base = bases[shift];
            Base = Bases[shift];
            max  = maxima[shift];

            /* read the rest of the number */
            for (;;) {
                /* x is used in the overflow test,
                   b is the digit we're adding on. */
                UV x, b;

                switch (*s) {

                /* if we don't mention it, we're done */
                default:
                    goto out;

                /* _ are ignored -- but warned about if consecutive */
                case '_':
                    if (ckWARN(WARN_SYNTAX) && lastub && s == lastub + 1)
                        Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                                    "Misplaced _ in number");
                    lastub = s++;
                    break;

                /* 8 and 9 are not octal */
                case '8': case '9':
                    if (shift == 3)
                        yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
                    /* FALL THROUGH */

                /* octal digits */
                case '2': case '3': case '4':
                case '5': case '6': case '7':
                    if (shift == 1)
                        yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
                    /* FALL THROUGH */

                case '0': case '1':
                    b = *s++ & 15;              /* ASCII digit -> value of digit */
                    goto digit;

                /* hex digits */
                case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
                case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
                    /* make sure they said 0x */
                    if (shift != 4)
                        goto out;
                    b = (*s++ & 7) + 9;

                    /* Prepare to put the digit we have onto the end
                       of the number so far.  We check for overflows.
                    */

                  digit:
                    just_zero = FALSE;
                    if (!overflowed) {
                        x = u << shift; /* make room for the digit */

                        if ((x >> shift) != u
                            && !(PL_hints & HINT_NEW_BINARY)) {
                            overflowed = TRUE;
                            n = (NV) u;
                            if (ckWARN_d(WARN_OVERFLOW))
                                Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
                                            "Integer overflow in %s number",
                                            base);
                        } else
                            u = x | b;          /* add the digit to the end */
                    }
                    if (overflowed) {
                        n *= nvshift[shift];
                        /* If an NV has not enough bits in its
                         * mantissa to represent an UV this summing of
                         * small low-order numbers is a waste of time
                         * (because the NV cannot preserve the
                         * low-order bits anyway): we could just
                         * remember when did we overflow and in the
                         * end just multiply n by the right
                         * amount. */
                        n += (NV) b;
                    }
                    break;
                }
            }

          /* if we get here, we had success: make a scalar value from
             the number.
          */
          out:

            /* final misplaced underbar check */
            if (s[-1] == '_') {
                if (ckWARN(WARN_SYNTAX))
                    Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
            }

            sv = NEWSV(92,0);
            if (overflowed) {
                if (ckWARN(WARN_PORTABLE) && n > 4294967295.0)
                    Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
                                "%s number > %s non-portable",
                                Base, max);
                sv_setnv(sv, n);
            }
            else {
#if UVSIZE > 4
                if (ckWARN(WARN_PORTABLE) && u > 0xffffffff)
                    Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
                                "%s number > %s non-portable",
                                Base, max);
#endif
                sv_setuv(sv, u);
            }
            if (just_zero && (PL_hints & HINT_NEW_INTEGER))
                sv = new_constant(start, s - start, "integer", 
                                  sv, Nullsv, NULL);
            else if (PL_hints & HINT_NEW_BINARY)
                sv = new_constant(start, s - start, "binary", sv, Nullsv, NULL);
        }
        break;

    /*
      handle decimal numbers.
      we're also sent here when we read a 0 as the first digit
    */
    case '1': case '2': case '3': case '4': case '5':
    case '6': case '7': case '8': case '9': case '.':
      decimal:
        d = PL_tokenbuf;
        e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
        floatit = FALSE;

        /* read next group of digits and _ and copy into d */
        while (isDIGIT(*s) || *s == '_') {
            /* skip underscores, checking for misplaced ones
               if -w is on
            */
            if (*s == '_') {
                if (ckWARN(WARN_SYNTAX) && lastub && s == lastub + 1)
                    Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                                "Misplaced _ in number");
                lastub = s++;
            }
            else {
                /* check for end of fixed-length buffer */
                if (d >= e)
                    Perl_croak(aTHX_ number_too_long);
                /* if we're ok, copy the character */
                *d++ = *s++;
            }
        }

        /* final misplaced underbar check */
        if (lastub && s == lastub + 1) {
            if (ckWARN(WARN_SYNTAX))
                Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
        }

        /* read a decimal portion if there is one.  avoid
           3..5 being interpreted as the number 3. followed
           by .5
        */
        if (*s == '.' && s[1] != '.') {
            floatit = TRUE;
            *d++ = *s++;

            if (*s == '_') {
                if (ckWARN(WARN_SYNTAX))
                    Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                                "Misplaced _ in number");
                lastub = s;
            }

            /* copy, ignoring underbars, until we run out of digits.
            */
            for (; isDIGIT(*s) || *s == '_'; s++) {
                /* fixed length buffer check */
                if (d >= e)
                    Perl_croak(aTHX_ number_too_long);
                if (*s == '_') {
                   if (ckWARN(WARN_SYNTAX) && lastub && s == lastub + 1)
                       Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                                   "Misplaced _ in number");
                   lastub = s;
                }
                else
                    *d++ = *s;
            }
            /* fractional part ending in underbar? */
            if (s[-1] == '_') {
                if (ckWARN(WARN_SYNTAX))
                    Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                                "Misplaced _ in number");
            }
            if (*s == '.' && isDIGIT(s[1])) {
                /* oops, it's really a v-string, but without the "v" */
                s = start;
                goto vstring;
            }
        }

        /* read exponent part, if present */
        if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
            floatit = TRUE;
            s++;

            /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
            *d++ = 'e';         /* At least some Mach atof()s don't grok 'E' */

            /* stray preinitial _ */
            if (*s == '_') {
                if (ckWARN(WARN_SYNTAX))
                    Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                                "Misplaced _ in number");
                lastub = s++;
            }

            /* allow positive or negative exponent */
            if (*s == '+' || *s == '-')
                *d++ = *s++;

            /* stray initial _ */
            if (*s == '_') {
                if (ckWARN(WARN_SYNTAX))
                    Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                                "Misplaced _ in number");
                lastub = s++;
            }

            /* read digits of exponent */
            while (isDIGIT(*s) || *s == '_') {
                if (isDIGIT(*s)) {
                    if (d >= e)
                        Perl_croak(aTHX_ number_too_long);
                    *d++ = *s++;
                }
                else {
                   if (ckWARN(WARN_SYNTAX) &&
                       ((lastub && s == lastub + 1) ||
                        (!isDIGIT(s[1]) && s[1] != '_')))
                       Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
                                   "Misplaced _ in number");
                   lastub = s++;
                }
            }
        }


        /* make an sv from the string */
        sv = NEWSV(92,0);

        /*
           We try to do an integer conversion first if no characters
           indicating "float" have been found.
         */

        if (!floatit) {
            UV uv;
            int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);

            if (flags == IS_NUMBER_IN_UV) {
              if (uv <= IV_MAX)
                sv_setiv(sv, uv); /* Prefer IVs over UVs. */
              else
                sv_setuv(sv, uv);
            } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
              if (uv <= (UV) IV_MIN)
                sv_setiv(sv, -(IV)uv);
              else
                floatit = TRUE;
            } else
              floatit = TRUE;
        }
        if (floatit) {
            /* terminate the string */
            *d = '\0';
            nv = Atof(PL_tokenbuf);
            sv_setnv(sv, nv);
        }

        if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
                       (PL_hints & HINT_NEW_INTEGER) )
            sv = new_constant(PL_tokenbuf, d - PL_tokenbuf,
                              (floatit ? "float" : "integer"),
                              sv, Nullsv, NULL);
        break;

    /* if it starts with a v, it could be a v-string */
    case 'v':
vstring:
                sv = NEWSV(92,5); /* preallocate storage space */
                s = scan_vstring(s,sv);
        break;
    }

    /* make the op for the constant and return */

    if (sv)
        lvalp->opval = newSVOP(OP_CONST, 0, sv);
    else
        lvalp->opval = Nullop;

    return s;
}

STATIC char *
S_scan_formline(pTHX_ register char *s)
{
    register char *eol;
    register char *t;
    SV *stuff = newSVpvn("",0);
    bool needargs = FALSE;
    bool eofmt = FALSE;

    while (!needargs) {
        if (*s == '.') {
            /*SUPPRESS 530*/
#ifdef PERL_STRICT_CR
            for (t = s+1;SPACE_OR_TAB(*t); t++) ;
#else
            for (t = s+1;SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
#endif
            if (*t == '\n' || t == PL_bufend) {
                eofmt = TRUE;
                break;
            }
        }
        if (PL_in_eval && !PL_rsfp) {
            eol = memchr(s,'\n',PL_bufend-s);
            if (!eol++)
                eol = PL_bufend;
        }
        else
            eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
        if (*s != '#') {
            for (t = s; t < eol; t++) {
                if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
                    needargs = FALSE;
                    goto enough;        /* ~~ must be first line in formline */
                }
                if (*t == '@' || *t == '^')
                    needargs = TRUE;
            }
            if (eol > s) {
                sv_catpvn(stuff, s, eol-s);
#ifndef PERL_STRICT_CR
                if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
                    char *end = SvPVX(stuff) + SvCUR(stuff);
                    end[-2] = '\n';
                    end[-1] = '\0';
                    SvCUR(stuff)--;
                }
#endif
            }
            else
              break;
        }
        s = eol;
        if (PL_rsfp) {
            s = filter_gets(PL_linestr, PL_rsfp, 0);
            PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
            PL_bufend = PL_bufptr + SvCUR(PL_linestr);
            PL_last_lop = PL_last_uni = Nullch;
            if (!s) {
                s = PL_bufptr;
                break;
            }
        }
        incline(s);
    }
  enough:
    if (SvCUR(stuff)) {
        PL_expect = XTERM;
        if (needargs) {
            PL_lex_state = LEX_NORMAL;
            PL_nextval[PL_nexttoke].ival = 0;
            force_next(',');
        }
        else
            PL_lex_state = LEX_FORMLINE;
        if (!IN_BYTES) {
            if (UTF && is_utf8_string((U8*)SvPVX(stuff), SvCUR(stuff)))
                SvUTF8_on(stuff);
            else if (PL_encoding)
                sv_recode_to_utf8(stuff, PL_encoding);
        }
        PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST, 0, stuff);
        force_next(THING);
        PL_nextval[PL_nexttoke].ival = OP_FORMLINE;
        force_next(LSTOP);
    }
    else {
        SvREFCNT_dec(stuff);
        if (eofmt)
            PL_lex_formbrack = 0;
        PL_bufptr = s;
    }
    return s;
}

STATIC void
S_set_csh(pTHX)
{
#ifdef CSH
    if (!PL_cshlen)
        PL_cshlen = strlen(PL_cshname);
#endif
}

I32
Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
{
    I32 oldsavestack_ix = PL_savestack_ix;
    CV* outsidecv = PL_compcv;

    if (PL_compcv) {
        assert(SvTYPE(PL_compcv) == SVt_PVCV);
    }
    SAVEI32(PL_subline);
    save_item(PL_subname);
    SAVESPTR(PL_compcv);

    PL_compcv = (CV*)NEWSV(1104,0);
    sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
    CvFLAGS(PL_compcv) |= flags;

    PL_subline = CopLINE(PL_curcop);
    CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
    CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc(outsidecv);
    CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;

    return oldsavestack_ix;
}

#ifdef __SC__
#pragma segment Perl_yylex
#endif
int
Perl_yywarn(pTHX_ char *s)
{
    PL_in_eval |= EVAL_WARNONLY;
    yyerror(s);
    PL_in_eval &= ~EVAL_WARNONLY;
    return 0;
}

int
Perl_yyerror(pTHX_ char *s)
{
    char *where = NULL;
    char *context = NULL;
    int contlen = -1;
    SV *msg;

    if (!yychar || (yychar == ';' && !PL_rsfp))
        where = "at EOF";
    else if (PL_bufptr > PL_oldoldbufptr && PL_bufptr - PL_oldoldbufptr < 200 &&
      PL_oldoldbufptr != PL_oldbufptr && PL_oldbufptr != PL_bufptr) {
        /*
                Only for NetWare:
                The code below is removed for NetWare because it abends/crashes on NetWare
                when the script has error such as not having the closing quotes like:
                    if ($var eq "value)
                Checking of white spaces is anyway done in NetWare code.
        */
#ifndef NETWARE
        while (isSPACE(*PL_oldoldbufptr))
            PL_oldoldbufptr++;
#endif
        context = PL_oldoldbufptr;
        contlen = PL_bufptr - PL_oldoldbufptr;
    }
    else if (PL_bufptr > PL_oldbufptr && PL_bufptr - PL_oldbufptr < 200 &&
      PL_oldbufptr != PL_bufptr) {
        /*
                Only for NetWare:
                The code below is removed for NetWare because it abends/crashes on NetWare
                when the script has error such as not having the closing quotes like:
                    if ($var eq "value)
                Checking of white spaces is anyway done in NetWare code.
        */
#ifndef NETWARE
        while (isSPACE(*PL_oldbufptr))
            PL_oldbufptr++;
#endif
        context = PL_oldbufptr;
        contlen = PL_bufptr - PL_oldbufptr;
    }
    else if (yychar > 255)
        where = "next token ???";
    else if (yychar == -2) { /* YYEMPTY */
        if (PL_lex_state == LEX_NORMAL ||
           (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
            where = "at end of line";
        else if (PL_lex_inpat)
            where = "within pattern";
        else
            where = "within string";
    }
    else {
        SV *where_sv = sv_2mortal(newSVpvn("next char ", 10));
        if (yychar < 32)
            Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
        else if (isPRINT_LC(yychar))
            Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
        else
            Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
        where = SvPVX(where_sv);
    }
    msg = sv_2mortal(newSVpv(s, 0));
    Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
        OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
    if (context)
        Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
    else
        Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
    if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
        Perl_sv_catpvf(aTHX_ msg,
        "  (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
                (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
        PL_multi_end = 0;
    }
    if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
        Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, msg);
    else
        qerror(msg);
    if (PL_error_count >= 10) {
        if (PL_in_eval && SvCUR(ERRSV))
            Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
            ERRSV, OutCopFILE(PL_curcop));
        else
            Perl_croak(aTHX_ "%s has too many errors.\n",
            OutCopFILE(PL_curcop));
    }
    PL_in_my = 0;
    PL_in_my_stash = Nullhv;
    return 0;
}
#ifdef __SC__
#pragma segment Main
#endif

STATIC char*
S_swallow_bom(pTHX_ U8 *s)
{
    STRLEN slen;
    slen = SvCUR(PL_linestr);
    switch (s[0]) {
    case 0xFF:
        if (s[1] == 0xFE) {
            /* UTF-16 little-endian? (or UTF32-LE?) */
            if (s[2] == 0 && s[3] == 0)  /* UTF-32 little-endian */
                Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
#ifndef PERL_NO_UTF16_FILTER
            if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
            s += 2;
        utf16le:
            if (PL_bufend > (char*)s) {
                U8 *news;
                I32 newlen;

                filter_add(utf16rev_textfilter, NULL);
                New(898, news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
                utf16_to_utf8_reversed(s, news,
                                       PL_bufend - (char*)s - 1,
                                       &newlen);
                sv_setpvn(PL_linestr, (const char*)news, newlen);
                Safefree(news);
                SvUTF8_on(PL_linestr);
                s = (U8*)SvPVX(PL_linestr);
                PL_bufend = SvPVX(PL_linestr) + newlen;
            }
#else
            Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
#endif
        }
        break;
    case 0xFE:
        if (s[1] == 0xFF) {   /* UTF-16 big-endian? */
#ifndef PERL_NO_UTF16_FILTER
            if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
            s += 2;
        utf16be:
            if (PL_bufend > (char *)s) {
                U8 *news;
                I32 newlen;

                filter_add(utf16_textfilter, NULL);
                New(898, news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
                utf16_to_utf8(s, news,
                              PL_bufend - (char*)s,
                              &newlen);
                sv_setpvn(PL_linestr, (const char*)news, newlen);
                Safefree(news);
                SvUTF8_on(PL_linestr);
                s = (U8*)SvPVX(PL_linestr);
                PL_bufend = SvPVX(PL_linestr) + newlen;
            }
#else
            Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
#endif
        }
        break;
    case 0xEF:
        if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
            if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
            s += 3;                      /* UTF-8 */
        }
        break;
    case 0:
        if (slen > 3) {
             if (s[1] == 0) {
                  if (s[2] == 0xFE && s[3] == 0xFF) {
                       /* UTF-32 big-endian */
                       Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
                  }
             }
             else if (s[2] == 0 && s[3] != 0) {
                  /* Leading bytes
                   * 00 xx 00 xx
                   * are a good indicator of UTF-16BE. */
                  if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
                  goto utf16be;
             }
        }
    default:
         if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
                  /* Leading bytes
                   * xx 00 xx 00
                   * are a good indicator of UTF-16LE. */
              if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
              goto utf16le;
         }
    }
    return (char*)s;
}

/*
 * restore_rsfp
 * Restore a source filter.
 */

static void
restore_rsfp(pTHX_ void *f)
{
    PerlIO *fp = (PerlIO*)f;

    if (PL_rsfp == PerlIO_stdin())
        PerlIO_clearerr(PL_rsfp);
    else if (PL_rsfp && (PL_rsfp != fp))
        PerlIO_close(PL_rsfp);
    PL_rsfp = fp;
}

#ifndef PERL_NO_UTF16_FILTER
static I32
utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
{
    STRLEN old = SvCUR(sv);
    I32 count = FILTER_READ(idx+1, sv, maxlen);
    DEBUG_P(PerlIO_printf(Perl_debug_log,
                          "utf16_textfilter(%p): %d %d (%d)\n",
                          utf16_textfilter, idx, maxlen, (int) count));
    if (count) {
        U8* tmps;
        I32 newlen;
        New(898, tmps, SvCUR(sv) * 3 / 2 + 1, U8);
        Copy(SvPVX(sv), tmps, old, char);
        utf16_to_utf8((U8*)SvPVX(sv) + old, tmps + old,
                      SvCUR(sv) - old, &newlen);
        sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
    }
    DEBUG_P({sv_dump(sv);});
    return SvCUR(sv);
}

static I32
utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
{
    STRLEN old = SvCUR(sv);
    I32 count = FILTER_READ(idx+1, sv, maxlen);
    DEBUG_P(PerlIO_printf(Perl_debug_log,
                          "utf16rev_textfilter(%p): %d %d (%d)\n",
                          utf16rev_textfilter, idx, maxlen, (int) count));
    if (count) {
        U8* tmps;
        I32 newlen;
        New(898, tmps, SvCUR(sv) * 3 / 2 + 1, U8);
        Copy(SvPVX(sv), tmps, old, char);
        utf16_to_utf8((U8*)SvPVX(sv) + old, tmps + old,
                      SvCUR(sv) - old, &newlen);
        sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
    }
    DEBUG_P({ sv_dump(sv); });
    return count;
}
#endif

/*
Returns a pointer to the next character after the parsed
vstring, as well as updating the passed in sv.

Function must be called like

        sv = NEWSV(92,5);
        s = scan_vstring(s,sv);

The sv should already be large enough to store the vstring
passed in, for performance reasons.

*/

char *
Perl_scan_vstring(pTHX_ char *s, SV *sv)
{
    char *pos = s;
    char *start = s;
    if (*pos == 'v') pos++;  /* get past 'v' */
    while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
        pos++;
    if ( *pos != '.') {
        /* this may not be a v-string if followed by => */
        char *next = pos;
        while (next < PL_bufend && isSPACE(*next))
            ++next;
        if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
            /* return string not v-string */
            sv_setpvn(sv,(char *)s,pos-s);
            return pos;
        }
    }

    if (!isALPHA(*pos)) {
        UV rev;
        U8 tmpbuf[UTF8_MAXBYTES+1];
        U8 *tmpend;

        if (*s == 'v') s++;  /* get past 'v' */

        sv_setpvn(sv, "", 0);

        for (;;) {
            rev = 0;
            {
                /* this is atoi() that tolerates underscores */
                char *end = pos;
                UV mult = 1;
                while (--end >= s) {
                    UV orev;
                    if (*end == '_')
                        continue;
                    orev = rev;
                    rev += (*end - '0') * mult;
                    mult *= 10;
                    if (orev > rev && ckWARN_d(WARN_OVERFLOW))
                        Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
                                    "Integer overflow in decimal number");
                }
            }
#ifdef EBCDIC
            if (rev > 0x7FFFFFFF)
                 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
#endif
            /* Append native character for the rev point */
            tmpend = uvchr_to_utf8(tmpbuf, rev);
            sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
            if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
                 SvUTF8_on(sv);
            if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
                 s = ++pos;
            else {
                 s = pos;
                 break;
            }
            while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
                 pos++;
        }
        SvPOK_on(sv);
        sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
        SvRMAGICAL_on(sv);
    }
    return s;
}