3 * Copyright (c) 1991-2001, Larry Wall
5 * You may distribute under the terms of either the GNU General Public
6 * License or the Artistic License, as specified in the README file.
11 * "It all comes from here, the stench and the peril." --Frodo
15 * This file is the lexer for Perl. It's closely linked to the
18 * The main routine is yylex(), which returns the next token.
22 #define PERL_IN_TOKE_C
25 #define yychar PL_yychar
26 #define yylval PL_yylval
28 static char ident_too_long[] = "Identifier too long";
30 static void restore_rsfp(pTHXo_ void *f);
31 #ifndef PERL_NO_UTF16_FILTER
32 static I32 utf16_textfilter(pTHXo_ int idx, SV *sv, int maxlen);
33 static I32 utf16rev_textfilter(pTHXo_ int idx, SV *sv, int maxlen);
36 #define XFAKEBRACK 128
39 /*#define UTF (SvUTF8(PL_linestr) && !(PL_hints & HINT_BYTE))*/
40 #define UTF (PL_hints & HINT_UTF8)
42 /* In variables name $^X, these are the legal values for X.
43 * 1999-02-27 mjd-perl-patch@plover.com */
44 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
46 /* On MacOS, respect nonbreaking spaces */
47 #ifdef MACOS_TRADITIONAL
48 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
50 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
53 /* LEX_* are values for PL_lex_state, the state of the lexer.
54 * They are arranged oddly so that the guard on the switch statement
55 * can get by with a single comparison (if the compiler is smart enough).
58 /* #define LEX_NOTPARSING 11 is done in perl.h. */
61 #define LEX_INTERPNORMAL 9
62 #define LEX_INTERPCASEMOD 8
63 #define LEX_INTERPPUSH 7
64 #define LEX_INTERPSTART 6
65 #define LEX_INTERPEND 5
66 #define LEX_INTERPENDMAYBE 4
67 #define LEX_INTERPCONCAT 3
68 #define LEX_INTERPCONST 2
69 #define LEX_FORMLINE 1
70 #define LEX_KNOWNEXT 0
78 # define YYMAXLEVEL 100
80 YYSTYPE* yylval_pointer[YYMAXLEVEL];
81 int* yychar_pointer[YYMAXLEVEL];
85 # define yylval (*yylval_pointer[yyactlevel])
86 # define yychar (*yychar_pointer[yyactlevel])
87 # define PERL_YYLEX_PARAM yylval_pointer[yyactlevel],yychar_pointer[yyactlevel]
89 # define yylex() Perl_yylex_r(aTHX_ yylval_pointer[yyactlevel],yychar_pointer[yyactlevel])
94 /* CLINE is a macro that ensures PL_copline has a sane value */
99 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
102 * Convenience functions to return different tokens and prime the
103 * lexer for the next token. They all take an argument.
105 * TOKEN : generic token (used for '(', DOLSHARP, etc)
106 * OPERATOR : generic operator
107 * AOPERATOR : assignment operator
108 * PREBLOCK : beginning the block after an if, while, foreach, ...
109 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
110 * PREREF : *EXPR where EXPR is not a simple identifier
111 * TERM : expression term
112 * LOOPX : loop exiting command (goto, last, dump, etc)
113 * FTST : file test operator
114 * FUN0 : zero-argument function
115 * FUN1 : not used, except for not, which isn't a UNIOP
116 * BOop : bitwise or or xor
118 * SHop : shift operator
119 * PWop : power operator
120 * PMop : pattern-matching operator
121 * Aop : addition-level operator
122 * Mop : multiplication-level operator
123 * Eop : equality-testing operator
124 * Rop : relational operator <= != gt
126 * Also see LOP and lop() below.
129 #define TOKEN(retval) return (PL_bufptr = s,(int)retval)
130 #define OPERATOR(retval) return (PL_expect = XTERM,PL_bufptr = s,(int)retval)
131 #define AOPERATOR(retval) return ao((PL_expect = XTERM,PL_bufptr = s,(int)retval))
132 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s,(int)retval)
133 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s,(int)retval)
134 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s,(int)retval)
135 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR,PL_bufptr = s,(int)retval)
136 #define LOOPX(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LOOPEX)
137 #define FTST(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)UNIOP)
138 #define FUN0(f) return(yylval.ival = f,PL_expect = XOPERATOR,PL_bufptr = s,(int)FUNC0)
139 #define FUN1(f) return(yylval.ival = f,PL_expect = XOPERATOR,PL_bufptr = s,(int)FUNC1)
140 #define BOop(f) return ao((yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)BITOROP))
141 #define BAop(f) return ao((yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)BITANDOP))
142 #define SHop(f) return ao((yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)SHIFTOP))
143 #define PWop(f) return ao((yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)POWOP))
144 #define PMop(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)MATCHOP)
145 #define Aop(f) return ao((yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)ADDOP))
146 #define Mop(f) return ao((yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)MULOP))
147 #define Eop(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)EQOP)
148 #define Rop(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)RELOP)
150 /* This bit of chicanery makes a unary function followed by
151 * a parenthesis into a function with one argument, highest precedence.
153 #define UNI(f) return(yylval.ival = f, \
156 PL_last_uni = PL_oldbufptr, \
157 PL_last_lop_op = f, \
158 (*s == '(' || (s = skipspace(s), *s == '(') ? (int)FUNC1 : (int)UNIOP) )
160 #define UNIBRACK(f) return(yylval.ival = f, \
162 PL_last_uni = PL_oldbufptr, \
163 (*s == '(' || (s = skipspace(s), *s == '(') ? (int)FUNC1 : (int)UNIOP) )
165 /* grandfather return to old style */
166 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
171 * This subroutine detects &&= and ||= and turns an ANDAND or OROR
172 * into an OP_ANDASSIGN or OP_ORASSIGN
176 S_ao(pTHX_ int toketype)
178 if (*PL_bufptr == '=') {
180 if (toketype == ANDAND)
181 yylval.ival = OP_ANDASSIGN;
182 else if (toketype == OROR)
183 yylval.ival = OP_ORASSIGN;
191 * When Perl expects an operator and finds something else, no_op
192 * prints the warning. It always prints "<something> found where
193 * operator expected. It prints "Missing semicolon on previous line?"
194 * if the surprise occurs at the start of the line. "do you need to
195 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
196 * where the compiler doesn't know if foo is a method call or a function.
197 * It prints "Missing operator before end of line" if there's nothing
198 * after the missing operator, or "... before <...>" if there is something
199 * after the missing operator.
203 S_no_op(pTHX_ char *what, char *s)
205 char *oldbp = PL_bufptr;
206 bool is_first = (PL_oldbufptr == PL_linestart);
212 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
214 Perl_warn(aTHX_ "\t(Missing semicolon on previous line?)\n");
215 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
217 for (t = PL_oldoldbufptr; *t && (isALNUM_lazy_if(t,UTF) || *t == ':'); t++) ;
218 if (t < PL_bufptr && isSPACE(*t))
219 Perl_warn(aTHX_ "\t(Do you need to predeclare %.*s?)\n",
220 t - PL_oldoldbufptr, PL_oldoldbufptr);
224 Perl_warn(aTHX_ "\t(Missing operator before %.*s?)\n", s - oldbp, oldbp);
231 * Complain about missing quote/regexp/heredoc terminator.
232 * If it's called with (char *)NULL then it cauterizes the line buffer.
233 * If we're in a delimited string and the delimiter is a control
234 * character, it's reformatted into a two-char sequence like ^C.
239 S_missingterm(pTHX_ char *s)
244 char *nl = strrchr(s,'\n');
250 iscntrl(PL_multi_close)
252 PL_multi_close < 32 || PL_multi_close == 127
256 tmpbuf[1] = toCTRL(PL_multi_close);
262 *tmpbuf = PL_multi_close;
266 q = strchr(s,'"') ? '\'' : '"';
267 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
275 Perl_deprecate(pTHX_ char *s)
277 if (ckWARN(WARN_DEPRECATED))
278 Perl_warner(aTHX_ WARN_DEPRECATED, "Use of %s is deprecated", s);
283 * Deprecate a comma-less variable list.
289 deprecate("comma-less variable list");
293 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
294 * utf16-to-utf8-reversed.
297 #ifdef PERL_CR_FILTER
301 register char *s = SvPVX(sv);
302 register char *e = s + SvCUR(sv);
303 /* outer loop optimized to do nothing if there are no CR-LFs */
305 if (*s++ == '\r' && *s == '\n') {
306 /* hit a CR-LF, need to copy the rest */
307 register char *d = s - 1;
310 if (*s == '\r' && s[1] == '\n')
321 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
323 I32 count = FILTER_READ(idx+1, sv, maxlen);
324 if (count > 0 && !maxlen)
332 * Initialize variables. Uses the Perl save_stack to save its state (for
333 * recursive calls to the parser).
337 Perl_lex_start(pTHX_ SV *line)
342 SAVEI32(PL_lex_dojoin);
343 SAVEI32(PL_lex_brackets);
344 SAVEI32(PL_lex_casemods);
345 SAVEI32(PL_lex_starts);
346 SAVEI32(PL_lex_state);
347 SAVEVPTR(PL_lex_inpat);
348 SAVEI32(PL_lex_inwhat);
349 if (PL_lex_state == LEX_KNOWNEXT) {
350 I32 toke = PL_nexttoke;
351 while (--toke >= 0) {
352 SAVEI32(PL_nexttype[toke]);
353 SAVEVPTR(PL_nextval[toke]);
355 SAVEI32(PL_nexttoke);
357 SAVECOPLINE(PL_curcop);
360 SAVEPPTR(PL_oldbufptr);
361 SAVEPPTR(PL_oldoldbufptr);
362 SAVEPPTR(PL_linestart);
363 SAVESPTR(PL_linestr);
364 SAVEPPTR(PL_lex_brackstack);
365 SAVEPPTR(PL_lex_casestack);
366 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
367 SAVESPTR(PL_lex_stuff);
368 SAVEI32(PL_lex_defer);
369 SAVEI32(PL_sublex_info.sub_inwhat);
370 SAVESPTR(PL_lex_repl);
372 SAVEINT(PL_lex_expect);
374 PL_lex_state = LEX_NORMAL;
378 New(899, PL_lex_brackstack, 120, char);
379 New(899, PL_lex_casestack, 12, char);
380 SAVEFREEPV(PL_lex_brackstack);
381 SAVEFREEPV(PL_lex_casestack);
383 *PL_lex_casestack = '\0';
386 PL_lex_stuff = Nullsv;
387 PL_lex_repl = Nullsv;
391 PL_sublex_info.sub_inwhat = 0;
393 if (SvREADONLY(PL_linestr))
394 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
395 s = SvPV(PL_linestr, len);
396 if (len && s[len-1] != ';') {
397 if (!(SvFLAGS(PL_linestr) & SVs_TEMP))
398 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
399 sv_catpvn(PL_linestr, "\n;", 2);
401 SvTEMP_off(PL_linestr);
402 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
403 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
405 PL_rs = newSVpvn("\n", 1);
411 * Finalizer for lexing operations. Must be called when the parser is
412 * done with the lexer.
418 PL_doextract = FALSE;
423 * This subroutine has nothing to do with tilting, whether at windmills
424 * or pinball tables. Its name is short for "increment line". It
425 * increments the current line number in CopLINE(PL_curcop) and checks
426 * to see whether the line starts with a comment of the form
427 * # line 500 "foo.pm"
428 * If so, it sets the current line number and file to the values in the comment.
432 S_incline(pTHX_ char *s)
439 CopLINE_inc(PL_curcop);
442 while (SPACE_OR_TAB(*s)) s++;
443 if (strnEQ(s, "line", 4))
447 if (*s == ' ' || *s == '\t')
451 while (SPACE_OR_TAB(*s)) s++;
457 while (SPACE_OR_TAB(*s))
459 if (*s == '"' && (t = strchr(s+1, '"'))) {
464 for (t = s; !isSPACE(*t); t++) ;
467 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
469 if (*e != '\n' && *e != '\0')
470 return; /* false alarm */
476 Safefree(CopFILE(PL_curcop));
478 SvREFCNT_dec(CopFILEGV(PL_curcop));
480 CopFILE_set(PL_curcop, s);
483 CopLINE_set(PL_curcop, atoi(n)-1);
488 * Called to gobble the appropriate amount and type of whitespace.
489 * Skips comments as well.
493 S_skipspace(pTHX_ register char *s)
495 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
496 while (s < PL_bufend && SPACE_OR_TAB(*s))
502 SSize_t oldprevlen, oldoldprevlen;
503 SSize_t oldloplen, oldunilen;
504 while (s < PL_bufend && isSPACE(*s)) {
505 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
510 if (s < PL_bufend && *s == '#') {
511 while (s < PL_bufend && *s != '\n')
515 if (PL_in_eval && !PL_rsfp) {
522 /* only continue to recharge the buffer if we're at the end
523 * of the buffer, we're not reading from a source filter, and
524 * we're in normal lexing mode
526 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
527 PL_lex_state == LEX_FORMLINE)
530 /* try to recharge the buffer */
531 if ((s = filter_gets(PL_linestr, PL_rsfp,
532 (prevlen = SvCUR(PL_linestr)))) == Nullch)
534 /* end of file. Add on the -p or -n magic */
535 if (PL_minus_n || PL_minus_p) {
536 sv_setpv(PL_linestr,PL_minus_p ?
537 ";}continue{print or die qq(-p destination: $!\\n)" :
539 sv_catpv(PL_linestr,";}");
540 PL_minus_n = PL_minus_p = 0;
543 sv_setpv(PL_linestr,";");
545 /* reset variables for next time we lex */
546 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
548 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
550 /* Close the filehandle. Could be from -P preprocessor,
551 * STDIN, or a regular file. If we were reading code from
552 * STDIN (because the commandline held no -e or filename)
553 * then we don't close it, we reset it so the code can
554 * read from STDIN too.
557 if (PL_preprocess && !PL_in_eval)
558 (void)PerlProc_pclose(PL_rsfp);
559 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
560 PerlIO_clearerr(PL_rsfp);
562 (void)PerlIO_close(PL_rsfp);
567 /* not at end of file, so we only read another line */
568 /* make corresponding updates to old pointers, for yyerror() */
569 oldprevlen = PL_oldbufptr - PL_bufend;
570 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
572 oldunilen = PL_last_uni - PL_bufend;
574 oldloplen = PL_last_lop - PL_bufend;
575 PL_linestart = PL_bufptr = s + prevlen;
576 PL_bufend = s + SvCUR(PL_linestr);
578 PL_oldbufptr = s + oldprevlen;
579 PL_oldoldbufptr = s + oldoldprevlen;
581 PL_last_uni = s + oldunilen;
583 PL_last_lop = s + oldloplen;
586 /* debugger active and we're not compiling the debugger code,
587 * so store the line into the debugger's array of lines
589 if (PERLDB_LINE && PL_curstash != PL_debstash) {
590 SV *sv = NEWSV(85,0);
592 sv_upgrade(sv, SVt_PVMG);
593 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
594 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
601 * Check the unary operators to ensure there's no ambiguity in how they're
602 * used. An ambiguous piece of code would be:
604 * This doesn't mean rand() + 5. Because rand() is a unary operator,
605 * the +5 is its argument.
614 if (PL_oldoldbufptr != PL_last_uni)
616 while (isSPACE(*PL_last_uni))
618 for (s = PL_last_uni; isALNUM_lazy_if(s,UTF) || *s == '-'; s++) ;
619 if ((t = strchr(s, '(')) && t < PL_bufptr)
621 if (ckWARN_d(WARN_AMBIGUOUS)){
624 Perl_warner(aTHX_ WARN_AMBIGUOUS,
625 "Warning: Use of \"%s\" without parens is ambiguous",
631 /* workaround to replace the UNI() macro with a function. Only the
632 * hints/uts.sh file mentions this. Other comments elsewhere in the
633 * source indicate Microport Unix might need it too.
639 #define UNI(f) return uni(f,s)
642 S_uni(pTHX_ I32 f, char *s)
647 PL_last_uni = PL_oldbufptr;
658 #endif /* CRIPPLED_CC */
661 * LOP : macro to build a list operator. Its behaviour has been replaced
662 * with a subroutine, S_lop() for which LOP is just another name.
665 #define LOP(f,x) return lop(f,x,s)
669 * Build a list operator (or something that might be one). The rules:
670 * - if we have a next token, then it's a list operator [why?]
671 * - if the next thing is an opening paren, then it's a function
672 * - else it's a list operator
676 S_lop(pTHX_ I32 f, int x, char *s)
682 PL_last_lop = PL_oldbufptr;
697 * When the lexer realizes it knows the next token (for instance,
698 * it is reordering tokens for the parser) then it can call S_force_next
699 * to know what token to return the next time the lexer is called. Caller
700 * will need to set PL_nextval[], and possibly PL_expect to ensure the lexer
701 * handles the token correctly.
705 S_force_next(pTHX_ I32 type)
707 PL_nexttype[PL_nexttoke] = type;
709 if (PL_lex_state != LEX_KNOWNEXT) {
710 PL_lex_defer = PL_lex_state;
711 PL_lex_expect = PL_expect;
712 PL_lex_state = LEX_KNOWNEXT;
718 * When the lexer knows the next thing is a word (for instance, it has
719 * just seen -> and it knows that the next char is a word char, then
720 * it calls S_force_word to stick the next word into the PL_next lookahead.
723 * char *start : buffer position (must be within PL_linestr)
724 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
725 * int check_keyword : if true, Perl checks to make sure the word isn't
726 * a keyword (do this if the word is a label, e.g. goto FOO)
727 * int allow_pack : if true, : characters will also be allowed (require,
729 * int allow_initial_tick : used by the "sub" lexer only.
733 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
738 start = skipspace(start);
740 if (isIDFIRST_lazy_if(s,UTF) ||
741 (allow_pack && *s == ':') ||
742 (allow_initial_tick && *s == '\'') )
744 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
745 if (check_keyword && keyword(PL_tokenbuf, len))
747 if (token == METHOD) {
752 PL_expect = XOPERATOR;
755 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST,0, newSVpv(PL_tokenbuf,0));
756 PL_nextval[PL_nexttoke].opval->op_private |= OPpCONST_BARE;
764 * Called when the lexer wants $foo *foo &foo etc, but the program
765 * text only contains the "foo" portion. The first argument is a pointer
766 * to the "foo", and the second argument is the type symbol to prefix.
767 * Forces the next token to be a "WORD".
768 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
772 S_force_ident(pTHX_ register char *s, int kind)
775 OP* o = (OP*)newSVOP(OP_CONST, 0, newSVpv(s,0));
776 PL_nextval[PL_nexttoke].opval = o;
779 o->op_private = OPpCONST_ENTERED;
780 /* XXX see note in pp_entereval() for why we forgo typo
781 warnings if the symbol must be introduced in an eval.
783 gv_fetchpv(s, PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : TRUE,
784 kind == '$' ? SVt_PV :
785 kind == '@' ? SVt_PVAV :
786 kind == '%' ? SVt_PVHV :
794 Perl_str_to_version(pTHX_ SV *sv)
799 char *start = SvPVx(sv,len);
800 bool utf = SvUTF8(sv) ? TRUE : FALSE;
801 char *end = start + len;
802 while (start < end) {
806 n = utf8_to_uv((U8*)start, len, &skip, 0);
811 retval += ((NV)n)/nshift;
820 * Forces the next token to be a version number.
824 S_force_version(pTHX_ char *s)
826 OP *version = Nullop;
835 for (; isDIGIT(*d) || *d == '_' || *d == '.'; d++);
836 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
838 s = scan_num(s, &yylval);
839 version = yylval.opval;
840 ver = cSVOPx(version)->op_sv;
841 if (SvPOK(ver) && !SvNIOK(ver)) {
842 (void)SvUPGRADE(ver, SVt_PVNV);
843 SvNVX(ver) = str_to_version(ver);
844 SvNOK_on(ver); /* hint that it is a version */
849 /* NOTE: The parser sees the package name and the VERSION swapped */
850 PL_nextval[PL_nexttoke].opval = version;
858 * Tokenize a quoted string passed in as an SV. It finds the next
859 * chunk, up to end of string or a backslash. It may make a new
860 * SV containing that chunk (if HINT_NEW_STRING is on). It also
865 S_tokeq(pTHX_ SV *sv)
876 s = SvPV_force(sv, len);
877 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
880 while (s < send && *s != '\\')
885 if ( PL_hints & HINT_NEW_STRING )
886 pv = sv_2mortal(newSVpvn(SvPVX(pv), len));
889 if (s + 1 < send && (s[1] == '\\'))
890 s++; /* all that, just for this */
895 SvCUR_set(sv, d - SvPVX(sv));
897 if ( PL_hints & HINT_NEW_STRING )
898 return new_constant(NULL, 0, "q", sv, pv, "q");
903 * Now come three functions related to double-quote context,
904 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
905 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
906 * interact with PL_lex_state, and create fake ( ... ) argument lists
907 * to handle functions and concatenation.
908 * They assume that whoever calls them will be setting up a fake
909 * join call, because each subthing puts a ',' after it. This lets
912 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
914 * (I'm not sure whether the spurious commas at the end of lcfirst's
915 * arguments and join's arguments are created or not).
920 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
922 * Pattern matching will set PL_lex_op to the pattern-matching op to
923 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
925 * OP_CONST and OP_READLINE are easy--just make the new op and return.
927 * Everything else becomes a FUNC.
929 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
930 * had an OP_CONST or OP_READLINE). This just sets us up for a
931 * call to S_sublex_push().
937 register I32 op_type = yylval.ival;
939 if (op_type == OP_NULL) {
940 yylval.opval = PL_lex_op;
944 if (op_type == OP_CONST || op_type == OP_READLINE) {
945 SV *sv = tokeq(PL_lex_stuff);
947 if (SvTYPE(sv) == SVt_PVIV) {
948 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
954 nsv = newSVpvn(p, len);
960 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
961 PL_lex_stuff = Nullsv;
965 PL_sublex_info.super_state = PL_lex_state;
966 PL_sublex_info.sub_inwhat = op_type;
967 PL_sublex_info.sub_op = PL_lex_op;
968 PL_lex_state = LEX_INTERPPUSH;
972 yylval.opval = PL_lex_op;
982 * Create a new scope to save the lexing state. The scope will be
983 * ended in S_sublex_done. Returns a '(', starting the function arguments
984 * to the uc, lc, etc. found before.
985 * Sets PL_lex_state to LEX_INTERPCONCAT.
993 PL_lex_state = PL_sublex_info.super_state;
994 SAVEI32(PL_lex_dojoin);
995 SAVEI32(PL_lex_brackets);
996 SAVEI32(PL_lex_casemods);
997 SAVEI32(PL_lex_starts);
998 SAVEI32(PL_lex_state);
999 SAVEVPTR(PL_lex_inpat);
1000 SAVEI32(PL_lex_inwhat);
1001 SAVECOPLINE(PL_curcop);
1002 SAVEPPTR(PL_bufptr);
1003 SAVEPPTR(PL_oldbufptr);
1004 SAVEPPTR(PL_oldoldbufptr);
1005 SAVEPPTR(PL_linestart);
1006 SAVESPTR(PL_linestr);
1007 SAVEPPTR(PL_lex_brackstack);
1008 SAVEPPTR(PL_lex_casestack);
1010 PL_linestr = PL_lex_stuff;
1011 PL_lex_stuff = Nullsv;
1013 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1014 = SvPVX(PL_linestr);
1015 PL_bufend += SvCUR(PL_linestr);
1016 SAVEFREESV(PL_linestr);
1018 PL_lex_dojoin = FALSE;
1019 PL_lex_brackets = 0;
1020 New(899, PL_lex_brackstack, 120, char);
1021 New(899, PL_lex_casestack, 12, char);
1022 SAVEFREEPV(PL_lex_brackstack);
1023 SAVEFREEPV(PL_lex_casestack);
1024 PL_lex_casemods = 0;
1025 *PL_lex_casestack = '\0';
1027 PL_lex_state = LEX_INTERPCONCAT;
1028 CopLINE_set(PL_curcop, PL_multi_start);
1030 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1031 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1032 PL_lex_inpat = PL_sublex_info.sub_op;
1034 PL_lex_inpat = Nullop;
1041 * Restores lexer state after a S_sublex_push.
1047 if (!PL_lex_starts++) {
1048 SV *sv = newSVpvn("",0);
1049 if (SvUTF8(PL_linestr))
1051 PL_expect = XOPERATOR;
1052 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1056 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1057 PL_lex_state = LEX_INTERPCASEMOD;
1061 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1062 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1063 PL_linestr = PL_lex_repl;
1065 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1066 PL_bufend += SvCUR(PL_linestr);
1067 SAVEFREESV(PL_linestr);
1068 PL_lex_dojoin = FALSE;
1069 PL_lex_brackets = 0;
1070 PL_lex_casemods = 0;
1071 *PL_lex_casestack = '\0';
1073 if (SvEVALED(PL_lex_repl)) {
1074 PL_lex_state = LEX_INTERPNORMAL;
1076 /* we don't clear PL_lex_repl here, so that we can check later
1077 whether this is an evalled subst; that means we rely on the
1078 logic to ensure sublex_done() is called again only via the
1079 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1082 PL_lex_state = LEX_INTERPCONCAT;
1083 PL_lex_repl = Nullsv;
1089 PL_bufend = SvPVX(PL_linestr);
1090 PL_bufend += SvCUR(PL_linestr);
1091 PL_expect = XOPERATOR;
1092 PL_sublex_info.sub_inwhat = 0;
1100 Extracts a pattern, double-quoted string, or transliteration. This
1103 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1104 processing a pattern (PL_lex_inpat is true), a transliteration
1105 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1107 Returns a pointer to the character scanned up to. Iff this is
1108 advanced from the start pointer supplied (ie if anything was
1109 successfully parsed), will leave an OP for the substring scanned
1110 in yylval. Caller must intuit reason for not parsing further
1111 by looking at the next characters herself.
1115 double-quoted style: \r and \n
1116 regexp special ones: \D \s
1118 backrefs: \1 (deprecated in substitution replacements)
1119 case and quoting: \U \Q \E
1120 stops on @ and $, but not for $ as tail anchor
1122 In transliterations:
1123 characters are VERY literal, except for - not at the start or end
1124 of the string, which indicates a range. scan_const expands the
1125 range to the full set of intermediate characters.
1127 In double-quoted strings:
1129 double-quoted style: \r and \n
1131 backrefs: \1 (deprecated)
1132 case and quoting: \U \Q \E
1135 scan_const does *not* construct ops to handle interpolated strings.
1136 It stops processing as soon as it finds an embedded $ or @ variable
1137 and leaves it to the caller to work out what's going on.
1139 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @:foo.
1141 $ in pattern could be $foo or could be tail anchor. Assumption:
1142 it's a tail anchor if $ is the last thing in the string, or if it's
1143 followed by one of ")| \n\t"
1145 \1 (backreferences) are turned into $1
1147 The structure of the code is
1148 while (there's a character to process) {
1149 handle transliteration ranges
1150 skip regexp comments
1151 skip # initiated comments in //x patterns
1152 check for embedded @foo
1153 check for embedded scalars
1155 leave intact backslashes from leave (below)
1156 deprecate \1 in strings and sub replacements
1157 handle string-changing backslashes \l \U \Q \E, etc.
1158 switch (what was escaped) {
1159 handle - in a transliteration (becomes a literal -)
1160 handle \132 octal characters
1161 handle 0x15 hex characters
1162 handle \cV (control V)
1163 handle printf backslashes (\f, \r, \n, etc)
1165 } (end if backslash)
1166 } (end while character to read)
1171 S_scan_const(pTHX_ char *start)
1173 register char *send = PL_bufend; /* end of the constant */
1174 SV *sv = NEWSV(93, send - start); /* sv for the constant */
1175 register char *s = start; /* start of the constant */
1176 register char *d = SvPVX(sv); /* destination for copies */
1177 bool dorange = FALSE; /* are we in a translit range? */
1178 bool didrange = FALSE; /* did we just finish a range? */
1179 bool has_utf8 = (PL_linestr && SvUTF8(PL_linestr));
1180 /* the constant is UTF8 */
1183 I32 utf = (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op)
1184 ? (PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF))
1186 I32 this_utf8 = (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op)
1187 ? (PL_sublex_info.sub_op->op_private & (PL_lex_repl ?
1188 OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF))
1190 const char *leaveit = /* set of acceptably-backslashed characters */
1192 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxcz0123456789[{]} \t\n\r\f\v#"
1195 while (s < send || dorange) {
1196 /* get transliterations out of the way (they're most literal) */
1197 if (PL_lex_inwhat == OP_TRANS) {
1198 /* expand a range A-Z to the full set of characters. AIE! */
1200 I32 i; /* current expanded character */
1201 I32 min; /* first character in range */
1202 I32 max; /* last character in range */
1204 i = d - SvPVX(sv); /* remember current offset */
1205 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1206 d = SvPVX(sv) + i; /* refresh d after realloc */
1207 d -= 2; /* eat the first char and the - */
1209 min = (U8)*d; /* first char in range */
1210 max = (U8)d[1]; /* last char in range */
1214 "Invalid [] range \"%c-%c\" in transliteration operator",
1215 (char)min, (char)max);
1219 if ((isLOWER(min) && isLOWER(max)) ||
1220 (isUPPER(min) && isUPPER(max))) {
1222 for (i = min; i <= max; i++)
1226 for (i = min; i <= max; i++)
1233 for (i = min; i <= max; i++)
1236 /* mark the range as done, and continue */
1242 /* range begins (ignore - as first or last char) */
1243 else if (*s == '-' && s+1 < send && s != start) {
1245 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1248 *d++ = (char)0xff; /* use illegal utf8 byte--see pmtrans */
1260 /* if we get here, we're not doing a transliteration */
1262 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1263 except for the last char, which will be done separately. */
1264 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1266 while (s < send && *s != ')')
1269 else if (s[2] == '{' /* This should match regcomp.c */
1270 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1273 char *regparse = s + (s[2] == '{' ? 3 : 4);
1276 while (count && (c = *regparse)) {
1277 if (c == '\\' && regparse[1])
1285 if (*regparse != ')') {
1286 regparse--; /* Leave one char for continuation. */
1287 yyerror("Sequence (?{...}) not terminated or not {}-balanced");
1289 while (s < regparse)
1294 /* likewise skip #-initiated comments in //x patterns */
1295 else if (*s == '#' && PL_lex_inpat &&
1296 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1297 while (s+1 < send && *s != '\n')
1301 /* check for embedded arrays
1302 (@foo, @:foo, @'foo, @{foo}, @$foo, @+, @-)
1304 else if (*s == '@' && s[1]
1305 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1308 /* check for embedded scalars. only stop if we're sure it's a
1311 else if (*s == '$') {
1312 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1314 if (s + 1 < send && !strchr("()| \n\t", s[1]))
1315 break; /* in regexp, $ might be tail anchor */
1319 if (*s == '\\' && s+1 < send) {
1322 /* some backslashes we leave behind */
1323 if (*leaveit && *s && strchr(leaveit, *s)) {
1329 /* deprecate \1 in strings and substitution replacements */
1330 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1331 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1333 if (ckWARN(WARN_SYNTAX))
1334 Perl_warner(aTHX_ WARN_SYNTAX, "\\%c better written as $%c", *s, *s);
1339 /* string-change backslash escapes */
1340 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1345 /* if we get here, it's either a quoted -, or a digit */
1348 /* quoted - in transliterations */
1350 if (PL_lex_inwhat == OP_TRANS) {
1357 if (ckWARN(WARN_MISC) && isALNUM(*s))
1358 Perl_warner(aTHX_ WARN_MISC,
1359 "Unrecognized escape \\%c passed through",
1361 /* default action is to copy the quoted character */
1366 /* \132 indicates an octal constant */
1367 case '0': case '1': case '2': case '3':
1368 case '4': case '5': case '6': case '7':
1370 STRLEN len = 0; /* disallow underscores */
1371 uv = (UV)scan_oct(s, 3, &len);
1374 goto NUM_ESCAPE_INSERT;
1376 /* \x24 indicates a hex constant */
1380 char* e = strchr(s, '}');
1382 yyerror("Missing right brace on \\x{}");
1386 STRLEN len = 1; /* allow underscores */
1387 uv = (UV)scan_hex(s + 1, e - s - 1, &len);
1393 STRLEN len = 0; /* disallow underscores */
1394 uv = (UV)scan_hex(s, 2, &len);
1400 /* Insert oct or hex escaped character.
1401 * There will always enough room in sv since such
1402 * escapes will be longer than any UT-F8 sequence
1403 * they can end up as. */
1405 /* This spot is wrong for EBCDIC. Characters like
1406 * the lowercase letters and digits are >127 in EBCDIC,
1407 * so here they would need to be mapped to the Unicode
1408 * repertoire. --jhi */
1411 if (!has_utf8 && uv > 255) {
1412 /* Might need to recode whatever we have
1413 * accumulated so far if it contains any
1416 * (Can't we keep track of that and avoid
1417 * this rescan? --jhi)
1422 for (c = SvPVX(sv); c < d; c++) {
1423 if (UTF8_IS_CONTINUED(*c))
1427 char *old_pvx = SvPVX(sv);
1431 SvCUR(sv) + hicount + 1) +
1439 if (UTF8_IS_CONTINUED(*src)) {
1440 *dst-- = UTF8_EIGHT_BIT_LO(*src);
1441 *dst-- = UTF8_EIGHT_BIT_HI(*src--);
1450 if (has_utf8 || uv > 255) {
1451 d = (char*)uv_to_utf8((U8*)d, uv);
1463 /* \N{latin small letter a} is a named character */
1467 char* e = strchr(s, '}');
1473 yyerror("Missing right brace on \\N{}");
1477 res = newSVpvn(s + 1, e - s - 1);
1478 res = new_constant( Nullch, 0, "charnames",
1479 res, Nullsv, "\\N{...}" );
1480 str = SvPV(res,len);
1481 if (!has_utf8 && SvUTF8(res)) {
1482 char *ostart = SvPVX(sv);
1483 SvCUR_set(sv, d - ostart);
1486 sv_utf8_upgrade(sv);
1487 /* this just broke our allocation above... */
1488 SvGROW(sv, send - start);
1489 d = SvPVX(sv) + SvCUR(sv);
1492 if (len > e - s + 4) {
1493 char *odest = SvPVX(sv);
1495 SvGROW(sv, (SvCUR(sv) + len - (e - s + 4)));
1496 d = SvPVX(sv) + (d - odest);
1498 Copy(str, d, len, char);
1505 yyerror("Missing braces on \\N{}");
1508 /* \c is a control character */
1525 /* printf-style backslashes, formfeeds, newlines, etc */
1543 *d++ = '\047'; /* CP 1047 */
1546 *d++ = '\057'; /* CP 1047 */
1560 } /* end if (backslash) */
1562 /* (now in tr/// code again) */
1564 if (UTF8_IS_CONTINUED(*s) && (this_utf8 || has_utf8)) {
1565 STRLEN len = (STRLEN) -1;
1568 uv = utf8_to_uv((U8*)s, send - s, &len, 0);
1570 if (len == (STRLEN)-1) {
1571 /* Illegal UTF8 (a high-bit byte), make it valid. */
1572 char *old_pvx = SvPVX(sv);
1573 /* need space for one extra char (NOTE: SvCUR() not set here) */
1574 d = SvGROW(sv, SvLEN(sv) + 1) + (d - old_pvx);
1575 d = (char*)uv_to_utf8((U8*)d, (U8)*s++);
1586 } /* while loop to process each character */
1588 /* terminate the string and set up the sv */
1590 SvCUR_set(sv, d - SvPVX(sv));
1595 /* shrink the sv if we allocated more than we used */
1596 if (SvCUR(sv) + 5 < SvLEN(sv)) {
1597 SvLEN_set(sv, SvCUR(sv) + 1);
1598 Renew(SvPVX(sv), SvLEN(sv), char);
1601 /* return the substring (via yylval) only if we parsed anything */
1602 if (s > PL_bufptr) {
1603 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
1604 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
1606 ( PL_lex_inwhat == OP_TRANS
1608 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
1611 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1618 * Returns TRUE if there's more to the expression (e.g., a subscript),
1621 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
1623 * ->[ and ->{ return TRUE
1624 * { and [ outside a pattern are always subscripts, so return TRUE
1625 * if we're outside a pattern and it's not { or [, then return FALSE
1626 * if we're in a pattern and the first char is a {
1627 * {4,5} (any digits around the comma) returns FALSE
1628 * if we're in a pattern and the first char is a [
1630 * [SOMETHING] has a funky algorithm to decide whether it's a
1631 * character class or not. It has to deal with things like
1632 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
1633 * anything else returns TRUE
1636 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
1639 S_intuit_more(pTHX_ register char *s)
1641 if (PL_lex_brackets)
1643 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
1645 if (*s != '{' && *s != '[')
1650 /* In a pattern, so maybe we have {n,m}. */
1667 /* On the other hand, maybe we have a character class */
1670 if (*s == ']' || *s == '^')
1673 /* this is terrifying, and it works */
1674 int weight = 2; /* let's weigh the evidence */
1676 unsigned char un_char = 255, last_un_char;
1677 char *send = strchr(s,']');
1678 char tmpbuf[sizeof PL_tokenbuf * 4];
1680 if (!send) /* has to be an expression */
1683 Zero(seen,256,char);
1686 else if (isDIGIT(*s)) {
1688 if (isDIGIT(s[1]) && s[2] == ']')
1694 for (; s < send; s++) {
1695 last_un_char = un_char;
1696 un_char = (unsigned char)*s;
1701 weight -= seen[un_char] * 10;
1702 if (isALNUM_lazy_if(s+1,UTF)) {
1703 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
1704 if ((int)strlen(tmpbuf) > 1 && gv_fetchpv(tmpbuf,FALSE, SVt_PV))
1709 else if (*s == '$' && s[1] &&
1710 strchr("[#!%*<>()-=",s[1])) {
1711 if (/*{*/ strchr("])} =",s[2]))
1720 if (strchr("wds]",s[1]))
1722 else if (seen['\''] || seen['"'])
1724 else if (strchr("rnftbxcav",s[1]))
1726 else if (isDIGIT(s[1])) {
1728 while (s[1] && isDIGIT(s[1]))
1738 if (strchr("aA01! ",last_un_char))
1740 if (strchr("zZ79~",s[1]))
1742 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
1743 weight -= 5; /* cope with negative subscript */
1746 if (!isALNUM(last_un_char) && !strchr("$@&",last_un_char) &&
1747 isALPHA(*s) && s[1] && isALPHA(s[1])) {
1752 if (keyword(tmpbuf, d - tmpbuf))
1755 if (un_char == last_un_char + 1)
1757 weight -= seen[un_char];
1762 if (weight >= 0) /* probably a character class */
1772 * Does all the checking to disambiguate
1774 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
1775 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
1777 * First argument is the stuff after the first token, e.g. "bar".
1779 * Not a method if bar is a filehandle.
1780 * Not a method if foo is a subroutine prototyped to take a filehandle.
1781 * Not a method if it's really "Foo $bar"
1782 * Method if it's "foo $bar"
1783 * Not a method if it's really "print foo $bar"
1784 * Method if it's really "foo package::" (interpreted as package->foo)
1785 * Not a method if bar is known to be a subroutne ("sub bar; foo bar")
1786 * Not a method if bar is a filehandle or package, but is quoted with
1791 S_intuit_method(pTHX_ char *start, GV *gv)
1793 char *s = start + (*start == '$');
1794 char tmpbuf[sizeof PL_tokenbuf];
1802 if ((cv = GvCVu(gv))) {
1803 char *proto = SvPVX(cv);
1813 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
1814 /* start is the beginning of the possible filehandle/object,
1815 * and s is the end of it
1816 * tmpbuf is a copy of it
1819 if (*start == '$') {
1820 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
1825 return *s == '(' ? FUNCMETH : METHOD;
1827 if (!keyword(tmpbuf, len)) {
1828 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
1833 indirgv = gv_fetchpv(tmpbuf, FALSE, SVt_PVCV);
1834 if (indirgv && GvCVu(indirgv))
1836 /* filehandle or package name makes it a method */
1837 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
1839 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
1840 return 0; /* no assumptions -- "=>" quotes bearword */
1842 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST, 0,
1843 newSVpvn(tmpbuf,len));
1844 PL_nextval[PL_nexttoke].opval->op_private = OPpCONST_BARE;
1848 return *s == '(' ? FUNCMETH : METHOD;
1856 * Return a string of Perl code to load the debugger. If PERL5DB
1857 * is set, it will return the contents of that, otherwise a
1858 * compile-time require of perl5db.pl.
1865 char *pdb = PerlEnv_getenv("PERL5DB");
1869 SETERRNO(0,SS$_NORMAL);
1870 return "BEGIN { require 'perl5db.pl' }";
1876 /* Encoded script support. filter_add() effectively inserts a
1877 * 'pre-processing' function into the current source input stream.
1878 * Note that the filter function only applies to the current source file
1879 * (e.g., it will not affect files 'require'd or 'use'd by this one).
1881 * The datasv parameter (which may be NULL) can be used to pass
1882 * private data to this instance of the filter. The filter function
1883 * can recover the SV using the FILTER_DATA macro and use it to
1884 * store private buffers and state information.
1886 * The supplied datasv parameter is upgraded to a PVIO type
1887 * and the IoDIRP/IoANY field is used to store the function pointer,
1888 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
1889 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
1890 * private use must be set using malloc'd pointers.
1894 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
1899 if (!PL_rsfp_filters)
1900 PL_rsfp_filters = newAV();
1902 datasv = NEWSV(255,0);
1903 if (!SvUPGRADE(datasv, SVt_PVIO))
1904 Perl_die(aTHX_ "Can't upgrade filter_add data to SVt_PVIO");
1905 IoANY(datasv) = (void *)funcp; /* stash funcp into spare field */
1906 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
1907 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
1908 funcp, SvPV_nolen(datasv)));
1909 av_unshift(PL_rsfp_filters, 1);
1910 av_store(PL_rsfp_filters, 0, datasv) ;
1915 /* Delete most recently added instance of this filter function. */
1917 Perl_filter_del(pTHX_ filter_t funcp)
1920 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p", funcp));
1921 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
1923 /* if filter is on top of stack (usual case) just pop it off */
1924 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
1925 if (IoANY(datasv) == (void *)funcp) {
1926 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
1927 IoANY(datasv) = (void *)NULL;
1928 sv_free(av_pop(PL_rsfp_filters));
1932 /* we need to search for the correct entry and clear it */
1933 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
1937 /* Invoke the n'th filter function for the current rsfp. */
1939 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
1942 /* 0 = read one text line */
1947 if (!PL_rsfp_filters)
1949 if (idx > AvFILLp(PL_rsfp_filters)){ /* Any more filters? */
1950 /* Provide a default input filter to make life easy. */
1951 /* Note that we append to the line. This is handy. */
1952 DEBUG_P(PerlIO_printf(Perl_debug_log,
1953 "filter_read %d: from rsfp\n", idx));
1957 int old_len = SvCUR(buf_sv) ;
1959 /* ensure buf_sv is large enough */
1960 SvGROW(buf_sv, old_len + maxlen) ;
1961 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len, maxlen)) <= 0){
1962 if (PerlIO_error(PL_rsfp))
1963 return -1; /* error */
1965 return 0 ; /* end of file */
1967 SvCUR_set(buf_sv, old_len + len) ;
1970 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
1971 if (PerlIO_error(PL_rsfp))
1972 return -1; /* error */
1974 return 0 ; /* end of file */
1977 return SvCUR(buf_sv);
1979 /* Skip this filter slot if filter has been deleted */
1980 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef){
1981 DEBUG_P(PerlIO_printf(Perl_debug_log,
1982 "filter_read %d: skipped (filter deleted)\n",
1984 return FILTER_READ(idx+1, buf_sv, maxlen); /* recurse */
1986 /* Get function pointer hidden within datasv */
1987 funcp = (filter_t)IoANY(datasv);
1988 DEBUG_P(PerlIO_printf(Perl_debug_log,
1989 "filter_read %d: via function %p (%s)\n",
1990 idx, funcp, SvPV_nolen(datasv)));
1991 /* Call function. The function is expected to */
1992 /* call "FILTER_READ(idx+1, buf_sv)" first. */
1993 /* Return: <0:error, =0:eof, >0:not eof */
1994 return (*funcp)(aTHXo_ idx, buf_sv, maxlen);
1998 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2000 #ifdef PERL_CR_FILTER
2001 if (!PL_rsfp_filters) {
2002 filter_add(S_cr_textfilter,NULL);
2005 if (PL_rsfp_filters) {
2008 SvCUR_set(sv, 0); /* start with empty line */
2009 if (FILTER_READ(0, sv, 0) > 0)
2010 return ( SvPVX(sv) ) ;
2015 return (sv_gets(sv, fp, append));
2019 S_find_in_my_stash(pTHX_ char *pkgname, I32 len)
2023 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2027 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2028 (gv = gv_fetchpv(pkgname, FALSE, SVt_PVHV)))
2030 return GvHV(gv); /* Foo:: */
2033 /* use constant CLASS => 'MyClass' */
2034 if ((gv = gv_fetchpv(pkgname, FALSE, SVt_PVCV))) {
2036 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2037 pkgname = SvPV_nolen(sv);
2041 return gv_stashpv(pkgname, FALSE);
2045 static char* exp_name[] =
2046 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2047 "ATTRTERM", "TERMBLOCK"
2054 Works out what to call the token just pulled out of the input
2055 stream. The yacc parser takes care of taking the ops we return and
2056 stitching them into a tree.
2062 if read an identifier
2063 if we're in a my declaration
2064 croak if they tried to say my($foo::bar)
2065 build the ops for a my() declaration
2066 if it's an access to a my() variable
2067 are we in a sort block?
2068 croak if my($a); $a <=> $b
2069 build ops for access to a my() variable
2070 if in a dq string, and they've said @foo and we can't find @foo
2072 build ops for a bareword
2073 if we already built the token before, use it.
2076 #ifdef USE_PURE_BISON
2078 #pragma segment Perl_yylex_r
2081 Perl_yylex_r(pTHX_ YYSTYPE *lvalp, int *lcharp)
2086 yylval_pointer[yyactlevel] = lvalp;
2087 yychar_pointer[yyactlevel] = lcharp;
2088 if (yyactlevel >= YYMAXLEVEL)
2089 Perl_croak(aTHX_ "panic: YYMAXLEVEL");
2091 r = Perl_yylex(aTHX);
2101 #pragma segment Perl_yylex
2114 /* check if there's an identifier for us to look at */
2115 if (PL_pending_ident) {
2116 /* pit holds the identifier we read and pending_ident is reset */
2117 char pit = PL_pending_ident;
2118 PL_pending_ident = 0;
2120 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2121 "### Tokener saw identifier '%s'\n", PL_tokenbuf); })
2123 /* if we're in a my(), we can't allow dynamics here.
2124 $foo'bar has already been turned into $foo::bar, so
2125 just check for colons.
2127 if it's a legal name, the OP is a PADANY.
2130 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
2131 if (strchr(PL_tokenbuf,':'))
2132 yyerror(Perl_form(aTHX_ "No package name allowed for "
2133 "variable %s in \"our\"",
2135 tmp = pad_allocmy(PL_tokenbuf);
2138 if (strchr(PL_tokenbuf,':'))
2139 yyerror(Perl_form(aTHX_ PL_no_myglob,PL_tokenbuf));
2141 yylval.opval = newOP(OP_PADANY, 0);
2142 yylval.opval->op_targ = pad_allocmy(PL_tokenbuf);
2148 build the ops for accesses to a my() variable.
2150 Deny my($a) or my($b) in a sort block, *if* $a or $b is
2151 then used in a comparison. This catches most, but not
2152 all cases. For instance, it catches
2153 sort { my($a); $a <=> $b }
2155 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
2156 (although why you'd do that is anyone's guess).
2159 if (!strchr(PL_tokenbuf,':')) {
2161 /* Check for single character per-thread SVs */
2162 if (PL_tokenbuf[0] == '$' && PL_tokenbuf[2] == '\0'
2163 && !isALPHA(PL_tokenbuf[1]) /* Rule out obvious non-threadsvs */
2164 && (tmp = find_threadsv(&PL_tokenbuf[1])) != NOT_IN_PAD)
2166 yylval.opval = newOP(OP_THREADSV, 0);
2167 yylval.opval->op_targ = tmp;
2170 #endif /* USE_THREADS */
2171 if ((tmp = pad_findmy(PL_tokenbuf)) != NOT_IN_PAD) {
2172 SV *namesv = AvARRAY(PL_comppad_name)[tmp];
2173 /* might be an "our" variable" */
2174 if (SvFLAGS(namesv) & SVpad_OUR) {
2175 /* build ops for a bareword */
2176 SV *sym = newSVpv(HvNAME(GvSTASH(namesv)),0);
2177 sv_catpvn(sym, "::", 2);
2178 sv_catpv(sym, PL_tokenbuf+1);
2179 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
2180 yylval.opval->op_private = OPpCONST_ENTERED;
2181 gv_fetchpv(SvPVX(sym),
2183 ? (GV_ADDMULTI | GV_ADDINEVAL)
2186 ((PL_tokenbuf[0] == '$') ? SVt_PV
2187 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
2192 /* if it's a sort block and they're naming $a or $b */
2193 if (PL_last_lop_op == OP_SORT &&
2194 PL_tokenbuf[0] == '$' &&
2195 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
2198 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
2199 d < PL_bufend && *d != '\n';
2202 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
2203 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
2209 yylval.opval = newOP(OP_PADANY, 0);
2210 yylval.opval->op_targ = tmp;
2216 Whine if they've said @foo in a doublequoted string,
2217 and @foo isn't a variable we can find in the symbol
2220 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
2221 GV *gv = gv_fetchpv(PL_tokenbuf+1, FALSE, SVt_PVAV);
2222 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
2223 && ckWARN(WARN_AMBIGUOUS))
2225 /* Downgraded from fatal to warning 20000522 mjd */
2226 Perl_warner(aTHX_ WARN_AMBIGUOUS,
2227 "Possible unintended interpolation of %s in string",
2232 /* build ops for a bareword */
2233 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
2234 yylval.opval->op_private = OPpCONST_ENTERED;
2235 gv_fetchpv(PL_tokenbuf+1, PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : TRUE,
2236 ((PL_tokenbuf[0] == '$') ? SVt_PV
2237 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
2242 /* no identifier pending identification */
2244 switch (PL_lex_state) {
2246 case LEX_NORMAL: /* Some compilers will produce faster */
2247 case LEX_INTERPNORMAL: /* code if we comment these out. */
2251 /* when we've already built the next token, just pull it out of the queue */
2254 yylval = PL_nextval[PL_nexttoke];
2256 PL_lex_state = PL_lex_defer;
2257 PL_expect = PL_lex_expect;
2258 PL_lex_defer = LEX_NORMAL;
2260 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2261 "### Next token after '%s' was known, type %"IVdf"\n", PL_bufptr,
2262 (IV)PL_nexttype[PL_nexttoke]); })
2264 return(PL_nexttype[PL_nexttoke]);
2266 /* interpolated case modifiers like \L \U, including \Q and \E.
2267 when we get here, PL_bufptr is at the \
2269 case LEX_INTERPCASEMOD:
2271 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
2272 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
2274 /* handle \E or end of string */
2275 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
2279 if (PL_lex_casemods) {
2280 oldmod = PL_lex_casestack[--PL_lex_casemods];
2281 PL_lex_casestack[PL_lex_casemods] = '\0';
2283 if (PL_bufptr != PL_bufend && strchr("LUQ", oldmod)) {
2285 PL_lex_state = LEX_INTERPCONCAT;
2289 if (PL_bufptr != PL_bufend)
2291 PL_lex_state = LEX_INTERPCONCAT;
2295 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2296 "### Saw case modifier at '%s'\n", PL_bufptr); })
2298 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
2299 tmp = *s, *s = s[2], s[2] = tmp; /* misordered... */
2300 if (strchr("LU", *s) &&
2301 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U')))
2303 PL_lex_casestack[--PL_lex_casemods] = '\0';
2306 if (PL_lex_casemods > 10) {
2307 char* newlb = Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
2308 if (newlb != PL_lex_casestack) {
2310 PL_lex_casestack = newlb;
2313 PL_lex_casestack[PL_lex_casemods++] = *s;
2314 PL_lex_casestack[PL_lex_casemods] = '\0';
2315 PL_lex_state = LEX_INTERPCONCAT;
2316 PL_nextval[PL_nexttoke].ival = 0;
2319 PL_nextval[PL_nexttoke].ival = OP_LCFIRST;
2321 PL_nextval[PL_nexttoke].ival = OP_UCFIRST;
2323 PL_nextval[PL_nexttoke].ival = OP_LC;
2325 PL_nextval[PL_nexttoke].ival = OP_UC;
2327 PL_nextval[PL_nexttoke].ival = OP_QUOTEMETA;
2329 Perl_croak(aTHX_ "panic: yylex");
2332 if (PL_lex_starts) {
2341 case LEX_INTERPPUSH:
2342 return sublex_push();
2344 case LEX_INTERPSTART:
2345 if (PL_bufptr == PL_bufend)
2346 return sublex_done();
2347 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2348 "### Interpolated variable at '%s'\n", PL_bufptr); })
2350 PL_lex_dojoin = (*PL_bufptr == '@');
2351 PL_lex_state = LEX_INTERPNORMAL;
2352 if (PL_lex_dojoin) {
2353 PL_nextval[PL_nexttoke].ival = 0;
2356 PL_nextval[PL_nexttoke].opval = newOP(OP_THREADSV, 0);
2357 PL_nextval[PL_nexttoke].opval->op_targ = find_threadsv("\"");
2358 force_next(PRIVATEREF);
2360 force_ident("\"", '$');
2361 #endif /* USE_THREADS */
2362 PL_nextval[PL_nexttoke].ival = 0;
2364 PL_nextval[PL_nexttoke].ival = 0;
2366 PL_nextval[PL_nexttoke].ival = OP_JOIN; /* emulate join($", ...) */
2369 if (PL_lex_starts++) {
2375 case LEX_INTERPENDMAYBE:
2376 if (intuit_more(PL_bufptr)) {
2377 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
2383 if (PL_lex_dojoin) {
2384 PL_lex_dojoin = FALSE;
2385 PL_lex_state = LEX_INTERPCONCAT;
2388 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
2389 && SvEVALED(PL_lex_repl))
2391 if (PL_bufptr != PL_bufend)
2392 Perl_croak(aTHX_ "Bad evalled substitution pattern");
2393 PL_lex_repl = Nullsv;
2396 case LEX_INTERPCONCAT:
2398 if (PL_lex_brackets)
2399 Perl_croak(aTHX_ "panic: INTERPCONCAT");
2401 if (PL_bufptr == PL_bufend)
2402 return sublex_done();
2404 if (SvIVX(PL_linestr) == '\'') {
2405 SV *sv = newSVsv(PL_linestr);
2408 else if ( PL_hints & HINT_NEW_RE )
2409 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
2410 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2414 s = scan_const(PL_bufptr);
2416 PL_lex_state = LEX_INTERPCASEMOD;
2418 PL_lex_state = LEX_INTERPSTART;
2421 if (s != PL_bufptr) {
2422 PL_nextval[PL_nexttoke] = yylval;
2425 if (PL_lex_starts++)
2435 PL_lex_state = LEX_NORMAL;
2436 s = scan_formline(PL_bufptr);
2437 if (!PL_lex_formbrack)
2443 PL_oldoldbufptr = PL_oldbufptr;
2446 PerlIO_printf(Perl_debug_log, "### Tokener expecting %s at %s\n",
2447 exp_name[PL_expect], s);
2453 if (isIDFIRST_lazy_if(s,UTF))
2455 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
2458 goto fake_eof; /* emulate EOF on ^D or ^Z */
2463 if (PL_lex_brackets)
2464 yyerror("Missing right curly or square bracket");
2465 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2466 "### Tokener got EOF\n");
2470 if (s++ < PL_bufend)
2471 goto retry; /* ignore stray nulls */
2474 if (!PL_in_eval && !PL_preambled) {
2475 PL_preambled = TRUE;
2476 sv_setpv(PL_linestr,incl_perldb());
2477 if (SvCUR(PL_linestr))
2478 sv_catpv(PL_linestr,";");
2480 while(AvFILLp(PL_preambleav) >= 0) {
2481 SV *tmpsv = av_shift(PL_preambleav);
2482 sv_catsv(PL_linestr, tmpsv);
2483 sv_catpv(PL_linestr, ";");
2486 sv_free((SV*)PL_preambleav);
2487 PL_preambleav = NULL;
2489 if (PL_minus_n || PL_minus_p) {
2490 sv_catpv(PL_linestr, "LINE: while (<>) {");
2492 sv_catpv(PL_linestr,"chomp;");
2494 GV* gv = gv_fetchpv("::F", TRUE, SVt_PVAV);
2496 GvIMPORTED_AV_on(gv);
2498 if (strchr("/'\"", *PL_splitstr)
2499 && strchr(PL_splitstr + 1, *PL_splitstr))
2500 Perl_sv_catpvf(aTHX_ PL_linestr, "@F=split(%s);", PL_splitstr);
2503 s = "'~#\200\1'"; /* surely one char is unused...*/
2504 while (s[1] && strchr(PL_splitstr, *s)) s++;
2506 Perl_sv_catpvf(aTHX_ PL_linestr, "@F=split(%s%c",
2507 "q" + (delim == '\''), delim);
2508 for (s = PL_splitstr; *s; s++) {
2510 sv_catpvn(PL_linestr, "\\", 1);
2511 sv_catpvn(PL_linestr, s, 1);
2513 Perl_sv_catpvf(aTHX_ PL_linestr, "%c);", delim);
2517 sv_catpv(PL_linestr,"@F=split(' ');");
2520 sv_catpv(PL_linestr, "\n");
2521 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2522 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2523 if (PERLDB_LINE && PL_curstash != PL_debstash) {
2524 SV *sv = NEWSV(85,0);
2526 sv_upgrade(sv, SVt_PVMG);
2527 sv_setsv(sv,PL_linestr);
2528 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
2533 bof = PL_rsfp ? TRUE : FALSE;
2535 #ifdef PERLIO_IS_STDIO
2536 # ifdef __GNU_LIBRARY__
2537 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
2538 # define FTELL_FOR_PIPE_IS_BROKEN
2542 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
2543 # define FTELL_FOR_PIPE_IS_BROKEN
2548 #ifdef FTELL_FOR_PIPE_IS_BROKEN
2549 /* This loses the possibility to detect the bof
2550 * situation on perl -P when the libc5 is being used.
2551 * Workaround? Maybe attach some extra state to PL_rsfp?
2554 bof = PerlIO_tell(PL_rsfp) == 0;
2556 bof = PerlIO_tell(PL_rsfp) == 0;
2559 s = filter_gets(PL_linestr, PL_rsfp, 0);
2563 if (PL_preprocess && !PL_in_eval)
2564 (void)PerlProc_pclose(PL_rsfp);
2565 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
2566 PerlIO_clearerr(PL_rsfp);
2568 (void)PerlIO_close(PL_rsfp);
2570 PL_doextract = FALSE;
2572 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
2573 sv_setpv(PL_linestr,PL_minus_p ? ";}continue{print" : "");
2574 sv_catpv(PL_linestr,";}");
2575 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2576 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2577 PL_minus_n = PL_minus_p = 0;
2580 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2581 sv_setpv(PL_linestr,"");
2582 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
2584 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2585 s = swallow_bom((U8*)s);
2588 if (*s == '#' && s[1] == '!' && instr(s,"perl"))
2589 PL_doextract = FALSE;
2591 /* Incest with pod. */
2592 if (*s == '=' && strnEQ(s, "=cut", 4)) {
2593 sv_setpv(PL_linestr, "");
2594 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2595 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2596 PL_doextract = FALSE;
2600 } while (PL_doextract);
2601 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
2602 if (PERLDB_LINE && PL_curstash != PL_debstash) {
2603 SV *sv = NEWSV(85,0);
2605 sv_upgrade(sv, SVt_PVMG);
2606 sv_setsv(sv,PL_linestr);
2607 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
2609 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2610 if (CopLINE(PL_curcop) == 1) {
2611 while (s < PL_bufend && isSPACE(*s))
2613 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
2617 if (*s == '#' && *(s+1) == '!')
2619 #ifdef ALTERNATE_SHEBANG
2621 static char as[] = ALTERNATE_SHEBANG;
2622 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
2623 d = s + (sizeof(as) - 1);
2625 #endif /* ALTERNATE_SHEBANG */
2634 while (*d && !isSPACE(*d))
2638 #ifdef ARG_ZERO_IS_SCRIPT
2639 if (ipathend > ipath) {
2641 * HP-UX (at least) sets argv[0] to the script name,
2642 * which makes $^X incorrect. And Digital UNIX and Linux,
2643 * at least, set argv[0] to the basename of the Perl
2644 * interpreter. So, having found "#!", we'll set it right.
2646 SV *x = GvSV(gv_fetchpv("\030", TRUE, SVt_PV));
2647 assert(SvPOK(x) || SvGMAGICAL(x));
2648 if (sv_eq(x, CopFILESV(PL_curcop))) {
2649 sv_setpvn(x, ipath, ipathend - ipath);
2652 TAINT_NOT; /* $^X is always tainted, but that's OK */
2654 #endif /* ARG_ZERO_IS_SCRIPT */
2659 d = instr(s,"perl -");
2661 d = instr(s,"perl");
2663 /* avoid getting into infinite loops when shebang
2664 * line contains "Perl" rather than "perl" */
2666 for (d = ipathend-4; d >= ipath; --d) {
2667 if ((*d == 'p' || *d == 'P')
2668 && !ibcmp(d, "perl", 4))
2678 #ifdef ALTERNATE_SHEBANG
2680 * If the ALTERNATE_SHEBANG on this system starts with a
2681 * character that can be part of a Perl expression, then if
2682 * we see it but not "perl", we're probably looking at the
2683 * start of Perl code, not a request to hand off to some
2684 * other interpreter. Similarly, if "perl" is there, but
2685 * not in the first 'word' of the line, we assume the line
2686 * contains the start of the Perl program.
2688 if (d && *s != '#') {
2690 while (*c && !strchr("; \t\r\n\f\v#", *c))
2693 d = Nullch; /* "perl" not in first word; ignore */
2695 *s = '#'; /* Don't try to parse shebang line */
2697 #endif /* ALTERNATE_SHEBANG */
2698 #ifndef MACOS_TRADITIONAL
2703 !instr(s,"indir") &&
2704 instr(PL_origargv[0],"perl"))
2710 while (s < PL_bufend && isSPACE(*s))
2712 if (s < PL_bufend) {
2713 Newz(899,newargv,PL_origargc+3,char*);
2715 while (s < PL_bufend && !isSPACE(*s))
2718 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
2721 newargv = PL_origargv;
2723 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
2724 Perl_croak(aTHX_ "Can't exec %s", ipath);
2728 U32 oldpdb = PL_perldb;
2729 bool oldn = PL_minus_n;
2730 bool oldp = PL_minus_p;
2732 while (*d && !isSPACE(*d)) d++;
2733 while (SPACE_OR_TAB(*d)) d++;
2737 if (*d == 'M' || *d == 'm') {
2739 while (*d && !isSPACE(*d)) d++;
2740 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
2743 d = moreswitches(d);
2745 if ((PERLDB_LINE && !oldpdb) ||
2746 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
2747 /* if we have already added "LINE: while (<>) {",
2748 we must not do it again */
2750 sv_setpv(PL_linestr, "");
2751 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2752 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2753 PL_preambled = FALSE;
2755 (void)gv_fetchfile(PL_origfilename);
2762 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
2764 PL_lex_state = LEX_FORMLINE;
2769 #ifdef PERL_STRICT_CR
2770 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
2772 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
2774 case ' ': case '\t': case '\f': case 013:
2775 #ifdef MACOS_TRADITIONAL
2782 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
2783 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
2784 /* handle eval qq[#line 1 "foo"\n ...] */
2785 CopLINE_dec(PL_curcop);
2789 while (s < d && *s != '\n')
2794 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
2796 PL_lex_state = LEX_FORMLINE;
2806 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
2813 while (s < PL_bufend && SPACE_OR_TAB(*s))
2816 if (strnEQ(s,"=>",2)) {
2817 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
2818 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2819 "### Saw unary minus before =>, forcing word '%s'\n", s);
2821 OPERATOR('-'); /* unary minus */
2823 PL_last_uni = PL_oldbufptr;
2825 case 'r': ftst = OP_FTEREAD; break;
2826 case 'w': ftst = OP_FTEWRITE; break;
2827 case 'x': ftst = OP_FTEEXEC; break;
2828 case 'o': ftst = OP_FTEOWNED; break;
2829 case 'R': ftst = OP_FTRREAD; break;
2830 case 'W': ftst = OP_FTRWRITE; break;
2831 case 'X': ftst = OP_FTREXEC; break;
2832 case 'O': ftst = OP_FTROWNED; break;
2833 case 'e': ftst = OP_FTIS; break;
2834 case 'z': ftst = OP_FTZERO; break;
2835 case 's': ftst = OP_FTSIZE; break;
2836 case 'f': ftst = OP_FTFILE; break;
2837 case 'd': ftst = OP_FTDIR; break;
2838 case 'l': ftst = OP_FTLINK; break;
2839 case 'p': ftst = OP_FTPIPE; break;
2840 case 'S': ftst = OP_FTSOCK; break;
2841 case 'u': ftst = OP_FTSUID; break;
2842 case 'g': ftst = OP_FTSGID; break;
2843 case 'k': ftst = OP_FTSVTX; break;
2844 case 'b': ftst = OP_FTBLK; break;
2845 case 'c': ftst = OP_FTCHR; break;
2846 case 't': ftst = OP_FTTTY; break;
2847 case 'T': ftst = OP_FTTEXT; break;
2848 case 'B': ftst = OP_FTBINARY; break;
2849 case 'M': case 'A': case 'C':
2850 gv_fetchpv("\024",TRUE, SVt_PV);
2852 case 'M': ftst = OP_FTMTIME; break;
2853 case 'A': ftst = OP_FTATIME; break;
2854 case 'C': ftst = OP_FTCTIME; break;
2862 PL_last_lop_op = ftst;
2863 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2864 "### Saw file test %c\n", (int)ftst);
2869 /* Assume it was a minus followed by a one-letter named
2870 * subroutine call (or a -bareword), then. */
2871 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2872 "### %c looked like a file test but was not\n",
2881 if (PL_expect == XOPERATOR)
2886 else if (*s == '>') {
2889 if (isIDFIRST_lazy_if(s,UTF)) {
2890 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
2898 if (PL_expect == XOPERATOR)
2901 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
2903 OPERATOR('-'); /* unary minus */
2910 if (PL_expect == XOPERATOR)
2915 if (PL_expect == XOPERATOR)
2918 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
2924 if (PL_expect != XOPERATOR) {
2925 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
2926 PL_expect = XOPERATOR;
2927 force_ident(PL_tokenbuf, '*');
2940 if (PL_expect == XOPERATOR) {
2944 PL_tokenbuf[0] = '%';
2945 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
2946 if (!PL_tokenbuf[1]) {
2948 yyerror("Final % should be \\% or %name");
2951 PL_pending_ident = '%';
2970 switch (PL_expect) {
2973 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
2975 PL_bufptr = s; /* update in case we back off */
2981 PL_expect = XTERMBLOCK;
2985 while (isIDFIRST_lazy_if(s,UTF)) {
2986 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
2987 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
2988 if (tmp < 0) tmp = -tmp;
3003 d = scan_str(d,TRUE,TRUE);
3006 SvREFCNT_dec(PL_lex_stuff);
3007 PL_lex_stuff = Nullsv;
3009 /* MUST advance bufptr here to avoid bogus
3010 "at end of line" context messages from yyerror().
3012 PL_bufptr = s + len;
3013 yyerror("Unterminated attribute parameter in attribute list");
3016 return 0; /* EOF indicator */
3020 SV *sv = newSVpvn(s, len);
3021 sv_catsv(sv, PL_lex_stuff);
3022 attrs = append_elem(OP_LIST, attrs,
3023 newSVOP(OP_CONST, 0, sv));
3024 SvREFCNT_dec(PL_lex_stuff);
3025 PL_lex_stuff = Nullsv;
3028 if (!PL_in_my && len == 6 && strnEQ(s, "lvalue", len))
3029 CvLVALUE_on(PL_compcv);
3030 else if (!PL_in_my && len == 6 && strnEQ(s, "locked", len))
3031 CvLOCKED_on(PL_compcv);
3032 else if (!PL_in_my && len == 6 && strnEQ(s, "method", len))
3033 CvMETHOD_on(PL_compcv);
3034 /* After we've set the flags, it could be argued that
3035 we don't need to do the attributes.pm-based setting
3036 process, and shouldn't bother appending recognized
3037 flags. To experiment with that, uncomment the
3038 following "else": */
3040 attrs = append_elem(OP_LIST, attrs,
3041 newSVOP(OP_CONST, 0,
3045 if (*s == ':' && s[1] != ':')
3048 break; /* require real whitespace or :'s */
3050 tmp = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
3051 if (*s != ';' && *s != tmp && (tmp != '=' || *s != ')')) {
3052 char q = ((*s == '\'') ? '"' : '\'');
3053 /* If here for an expression, and parsed no attrs, back off. */
3054 if (tmp == '=' && !attrs) {
3058 /* MUST advance bufptr here to avoid bogus "at end of line"
3059 context messages from yyerror().
3063 yyerror("Unterminated attribute list");
3065 yyerror(Perl_form(aTHX_ "Invalid separator character %c%c%c in attribute list",
3073 PL_nextval[PL_nexttoke].opval = attrs;
3081 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
3082 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
3098 if (PL_lex_brackets <= 0)
3099 yyerror("Unmatched right square bracket");
3102 if (PL_lex_state == LEX_INTERPNORMAL) {
3103 if (PL_lex_brackets == 0) {
3104 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
3105 PL_lex_state = LEX_INTERPEND;
3112 if (PL_lex_brackets > 100) {
3113 char* newlb = Renew(PL_lex_brackstack, PL_lex_brackets + 1, char);
3114 if (newlb != PL_lex_brackstack) {
3116 PL_lex_brackstack = newlb;
3119 switch (PL_expect) {
3121 if (PL_lex_formbrack) {
3125 if (PL_oldoldbufptr == PL_last_lop)
3126 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
3128 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3129 OPERATOR(HASHBRACK);
3131 while (s < PL_bufend && SPACE_OR_TAB(*s))
3134 PL_tokenbuf[0] = '\0';
3135 if (d < PL_bufend && *d == '-') {
3136 PL_tokenbuf[0] = '-';
3138 while (d < PL_bufend && SPACE_OR_TAB(*d))
3141 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
3142 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
3144 while (d < PL_bufend && SPACE_OR_TAB(*d))
3147 char minus = (PL_tokenbuf[0] == '-');
3148 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
3149 if (UTF && !IN_BYTE && is_utf8_string((U8*)PL_tokenbuf, 0) &&
3150 PL_nextval[PL_nexttoke-1].opval)
3151 SvUTF8_on(((SVOP*)PL_nextval[PL_nexttoke-1].opval)->op_sv);
3159 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
3164 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3169 if (PL_oldoldbufptr == PL_last_lop)
3170 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
3172 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3175 OPERATOR(HASHBRACK);
3176 /* This hack serves to disambiguate a pair of curlies
3177 * as being a block or an anon hash. Normally, expectation
3178 * determines that, but in cases where we're not in a
3179 * position to expect anything in particular (like inside
3180 * eval"") we have to resolve the ambiguity. This code
3181 * covers the case where the first term in the curlies is a
3182 * quoted string. Most other cases need to be explicitly
3183 * disambiguated by prepending a `+' before the opening
3184 * curly in order to force resolution as an anon hash.
3186 * XXX should probably propagate the outer expectation
3187 * into eval"" to rely less on this hack, but that could
3188 * potentially break current behavior of eval"".
3192 if (*s == '\'' || *s == '"' || *s == '`') {
3193 /* common case: get past first string, handling escapes */
3194 for (t++; t < PL_bufend && *t != *s;)
3195 if (*t++ == '\\' && (*t == '\\' || *t == *s))
3199 else if (*s == 'q') {
3202 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
3206 char open, close, term;
3209 while (t < PL_bufend && isSPACE(*t))
3213 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
3217 for (t++; t < PL_bufend; t++) {
3218 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
3220 else if (*t == open)
3224 for (t++; t < PL_bufend; t++) {
3225 if (*t == '\\' && t+1 < PL_bufend)
3227 else if (*t == close && --brackets <= 0)
3229 else if (*t == open)
3235 else if (isALNUM_lazy_if(t,UTF)) {
3237 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
3240 while (t < PL_bufend && isSPACE(*t))
3242 /* if comma follows first term, call it an anon hash */
3243 /* XXX it could be a comma expression with loop modifiers */
3244 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
3245 || (*t == '=' && t[1] == '>')))
3246 OPERATOR(HASHBRACK);
3247 if (PL_expect == XREF)
3250 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
3256 yylval.ival = CopLINE(PL_curcop);
3257 if (isSPACE(*s) || *s == '#')
3258 PL_copline = NOLINE; /* invalidate current command line number */
3263 if (PL_lex_brackets <= 0)
3264 yyerror("Unmatched right curly bracket");
3266 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
3267 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
3268 PL_lex_formbrack = 0;
3269 if (PL_lex_state == LEX_INTERPNORMAL) {
3270 if (PL_lex_brackets == 0) {
3271 if (PL_expect & XFAKEBRACK) {
3272 PL_expect &= XENUMMASK;
3273 PL_lex_state = LEX_INTERPEND;
3275 return yylex(); /* ignore fake brackets */
3277 if (*s == '-' && s[1] == '>')
3278 PL_lex_state = LEX_INTERPENDMAYBE;
3279 else if (*s != '[' && *s != '{')
3280 PL_lex_state = LEX_INTERPEND;
3283 if (PL_expect & XFAKEBRACK) {
3284 PL_expect &= XENUMMASK;
3286 return yylex(); /* ignore fake brackets */
3296 if (PL_expect == XOPERATOR) {
3297 if (ckWARN(WARN_SEMICOLON)
3298 && isIDFIRST_lazy_if(s,UTF) && PL_bufptr == PL_linestart)
3300 CopLINE_dec(PL_curcop);
3301 Perl_warner(aTHX_ WARN_SEMICOLON, PL_warn_nosemi);
3302 CopLINE_inc(PL_curcop);
3307 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3309 PL_expect = XOPERATOR;
3310 force_ident(PL_tokenbuf, '&');
3314 yylval.ival = (OPpENTERSUB_AMPER<<8);
3333 if (ckWARN(WARN_SYNTAX) && tmp && isSPACE(*s) && strchr("+-*/%.^&|<",tmp))
3334 Perl_warner(aTHX_ WARN_SYNTAX, "Reversed %c= operator",(int)tmp);
3336 if (PL_expect == XSTATE && isALPHA(tmp) &&
3337 (s == PL_linestart+1 || s[-2] == '\n') )
3339 if (PL_in_eval && !PL_rsfp) {
3344 if (strnEQ(s,"=cut",4)) {
3358 PL_doextract = TRUE;
3361 if (PL_lex_brackets < PL_lex_formbrack) {
3363 #ifdef PERL_STRICT_CR
3364 for (t = s; SPACE_OR_TAB(*t); t++) ;
3366 for (t = s; SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
3368 if (*t == '\n' || *t == '#') {
3386 if (PL_expect != XOPERATOR) {
3387 if (s[1] != '<' && !strchr(s,'>'))
3390 s = scan_heredoc(s);
3392 s = scan_inputsymbol(s);
3393 TERM(sublex_start());
3398 SHop(OP_LEFT_SHIFT);
3412 SHop(OP_RIGHT_SHIFT);
3421 if (PL_expect == XOPERATOR) {
3422 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3425 return ','; /* grandfather non-comma-format format */
3429 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
3430 PL_tokenbuf[0] = '@';
3431 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
3432 sizeof PL_tokenbuf - 1, FALSE);
3433 if (PL_expect == XOPERATOR)
3434 no_op("Array length", s);
3435 if (!PL_tokenbuf[1])
3437 PL_expect = XOPERATOR;
3438 PL_pending_ident = '#';
3442 PL_tokenbuf[0] = '$';
3443 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
3444 sizeof PL_tokenbuf - 1, FALSE);
3445 if (PL_expect == XOPERATOR)
3447 if (!PL_tokenbuf[1]) {
3449 yyerror("Final $ should be \\$ or $name");
3453 /* This kludge not intended to be bulletproof. */
3454 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
3455 yylval.opval = newSVOP(OP_CONST, 0,
3456 newSViv(PL_compiling.cop_arybase));
3457 yylval.opval->op_private = OPpCONST_ARYBASE;
3463 if (PL_lex_state == LEX_NORMAL)
3466 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
3469 PL_tokenbuf[0] = '@';
3470 if (ckWARN(WARN_SYNTAX)) {
3472 isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$';
3475 PL_bufptr = skipspace(PL_bufptr);
3476 while (t < PL_bufend && *t != ']')
3478 Perl_warner(aTHX_ WARN_SYNTAX,
3479 "Multidimensional syntax %.*s not supported",
3480 (t - PL_bufptr) + 1, PL_bufptr);
3484 else if (*s == '{') {
3485 PL_tokenbuf[0] = '%';
3486 if (ckWARN(WARN_SYNTAX) && strEQ(PL_tokenbuf+1, "SIG") &&
3487 (t = strchr(s, '}')) && (t = strchr(t, '=')))
3489 char tmpbuf[sizeof PL_tokenbuf];
3491 for (t++; isSPACE(*t); t++) ;
3492 if (isIDFIRST_lazy_if(t,UTF)) {
3493 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE, &len);
3494 for (; isSPACE(*t); t++) ;
3495 if (*t == ';' && get_cv(tmpbuf, FALSE))
3496 Perl_warner(aTHX_ WARN_SYNTAX,
3497 "You need to quote \"%s\"", tmpbuf);
3503 PL_expect = XOPERATOR;
3504 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
3505 bool islop = (PL_last_lop == PL_oldoldbufptr);
3506 if (!islop || PL_last_lop_op == OP_GREPSTART)
3507 PL_expect = XOPERATOR;
3508 else if (strchr("$@\"'`q", *s))
3509 PL_expect = XTERM; /* e.g. print $fh "foo" */
3510 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
3511 PL_expect = XTERM; /* e.g. print $fh &sub */
3512 else if (isIDFIRST_lazy_if(s,UTF)) {
3513 char tmpbuf[sizeof PL_tokenbuf];
3514 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3515 if ((tmp = keyword(tmpbuf, len))) {
3516 /* binary operators exclude handle interpretations */
3528 PL_expect = XTERM; /* e.g. print $fh length() */
3533 GV *gv = gv_fetchpv(tmpbuf, FALSE, SVt_PVCV);
3534 if (gv && GvCVu(gv))
3535 PL_expect = XTERM; /* e.g. print $fh subr() */
3538 else if (isDIGIT(*s))
3539 PL_expect = XTERM; /* e.g. print $fh 3 */
3540 else if (*s == '.' && isDIGIT(s[1]))
3541 PL_expect = XTERM; /* e.g. print $fh .3 */
3542 else if (strchr("/?-+", *s) && !isSPACE(s[1]) && s[1] != '=')
3543 PL_expect = XTERM; /* e.g. print $fh -1 */
3544 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2]) && s[2] != '=')
3545 PL_expect = XTERM; /* print $fh <<"EOF" */
3547 PL_pending_ident = '$';
3551 if (PL_expect == XOPERATOR)
3553 PL_tokenbuf[0] = '@';
3554 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
3555 if (!PL_tokenbuf[1]) {
3557 yyerror("Final @ should be \\@ or @name");
3560 if (PL_lex_state == LEX_NORMAL)
3562 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
3564 PL_tokenbuf[0] = '%';
3566 /* Warn about @ where they meant $. */
3567 if (ckWARN(WARN_SYNTAX)) {
3568 if (*s == '[' || *s == '{') {
3570 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
3572 if (*t == '}' || *t == ']') {
3574 PL_bufptr = skipspace(PL_bufptr);
3575 Perl_warner(aTHX_ WARN_SYNTAX,
3576 "Scalar value %.*s better written as $%.*s",
3577 t-PL_bufptr, PL_bufptr, t-PL_bufptr-1, PL_bufptr+1);
3582 PL_pending_ident = '@';
3585 case '/': /* may either be division or pattern */
3586 case '?': /* may either be conditional or pattern */
3587 if (PL_expect != XOPERATOR) {
3588 /* Disable warning on "study /blah/" */
3589 if (PL_oldoldbufptr == PL_last_uni
3590 && (*PL_last_uni != 's' || s - PL_last_uni < 5
3591 || memNE(PL_last_uni, "study", 5)
3592 || isALNUM_lazy_if(PL_last_uni+5,UTF)))
3594 s = scan_pat(s,OP_MATCH);
3595 TERM(sublex_start());
3603 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
3604 #ifdef PERL_STRICT_CR
3607 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
3609 && (s == PL_linestart || s[-1] == '\n') )
3611 PL_lex_formbrack = 0;
3615 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
3621 yylval.ival = OPf_SPECIAL;
3627 if (PL_expect != XOPERATOR)
3632 case '0': case '1': case '2': case '3': case '4':
3633 case '5': case '6': case '7': case '8': case '9':
3634 s = scan_num(s, &yylval);
3635 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3636 "### Saw number in '%s'\n", s);
3638 if (PL_expect == XOPERATOR)
3643 s = scan_str(s,FALSE,FALSE);
3644 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3645 "### Saw string in '%s'\n", s);
3647 if (PL_expect == XOPERATOR) {
3648 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3651 return ','; /* grandfather non-comma-format format */
3657 missingterm((char*)0);
3658 yylval.ival = OP_CONST;
3659 TERM(sublex_start());
3662 s = scan_str(s,FALSE,FALSE);
3663 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3664 "### Saw string in '%s'\n", s);
3666 if (PL_expect == XOPERATOR) {
3667 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3670 return ','; /* grandfather non-comma-format format */
3676 missingterm((char*)0);
3677 yylval.ival = OP_CONST;
3678 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
3679 if (*d == '$' || *d == '@' || *d == '\\' || UTF8_IS_CONTINUED(*d)) {
3680 yylval.ival = OP_STRINGIFY;
3684 TERM(sublex_start());
3687 s = scan_str(s,FALSE,FALSE);
3688 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3689 "### Saw backtick string in '%s'\n", s);
3691 if (PL_expect == XOPERATOR)
3692 no_op("Backticks",s);
3694 missingterm((char*)0);
3695 yylval.ival = OP_BACKTICK;
3697 TERM(sublex_start());
3701 if (ckWARN(WARN_SYNTAX) && PL_lex_inwhat && isDIGIT(*s))
3702 Perl_warner(aTHX_ WARN_SYNTAX,"Can't use \\%c to mean $%c in expression",
3704 if (PL_expect == XOPERATOR)
3705 no_op("Backslash",s);
3709 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
3713 while (isDIGIT(*start) || *start == '_')
3715 if (*start == '.' && isDIGIT(start[1])) {
3716 s = scan_num(s, &yylval);
3719 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
3720 else if (!isALPHA(*start) && (PL_expect == XTERM || PL_expect == XREF)) {
3724 gv = gv_fetchpv(s, FALSE, SVt_PVCV);
3727 s = scan_num(s, &yylval);
3734 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
3773 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
3775 /* Some keywords can be followed by any delimiter, including ':' */
3776 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
3777 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
3778 (PL_tokenbuf[0] == 'q' &&
3779 strchr("qwxr", PL_tokenbuf[1])))));
3781 /* x::* is just a word, unless x is "CORE" */
3782 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
3786 while (d < PL_bufend && isSPACE(*d))
3787 d++; /* no comments skipped here, or s### is misparsed */
3789 /* Is this a label? */
3790 if (!tmp && PL_expect == XSTATE
3791 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
3793 yylval.pval = savepv(PL_tokenbuf);
3798 /* Check for keywords */
3799 tmp = keyword(PL_tokenbuf, len);
3801 /* Is this a word before a => operator? */
3802 if (*d == '=' && d[1] == '>') {
3804 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf,0));
3805 yylval.opval->op_private = OPpCONST_BARE;
3806 if (UTF && !IN_BYTE && is_utf8_string((U8*)PL_tokenbuf, len))
3807 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
3811 if (tmp < 0) { /* second-class keyword? */
3812 GV *ogv = Nullgv; /* override (winner) */
3813 GV *hgv = Nullgv; /* hidden (loser) */
3814 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
3816 if ((gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV)) &&
3819 if (GvIMPORTED_CV(gv))
3821 else if (! CvMETHOD(cv))
3825 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
3826 (gv = *gvp) != (GV*)&PL_sv_undef &&
3827 GvCVu(gv) && GvIMPORTED_CV(gv))
3833 tmp = 0; /* overridden by import or by GLOBAL */
3836 && -tmp==KEY_lock /* XXX generalizable kludge */
3838 && !hv_fetch(GvHVn(PL_incgv), "Thread.pm", 9, FALSE))
3840 tmp = 0; /* any sub overrides "weak" keyword */
3842 else { /* no override */
3846 if (ckWARN(WARN_AMBIGUOUS) && hgv
3847 && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
3848 Perl_warner(aTHX_ WARN_AMBIGUOUS,
3849 "Ambiguous call resolved as CORE::%s(), %s",
3850 GvENAME(hgv), "qualify as such or use &");
3857 default: /* not a keyword */
3860 char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
3862 /* Get the rest if it looks like a package qualifier */
3864 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
3866 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
3869 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
3870 *s == '\'' ? "'" : "::");
3874 if (PL_expect == XOPERATOR) {
3875 if (PL_bufptr == PL_linestart) {
3876 CopLINE_dec(PL_curcop);
3877 Perl_warner(aTHX_ WARN_SEMICOLON, PL_warn_nosemi);
3878 CopLINE_inc(PL_curcop);
3881 no_op("Bareword",s);
3884 /* Look for a subroutine with this name in current package,
3885 unless name is "Foo::", in which case Foo is a bearword
3886 (and a package name). */
3889 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
3891 if (ckWARN(WARN_BAREWORD) && ! gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVHV))
3892 Perl_warner(aTHX_ WARN_BAREWORD,
3893 "Bareword \"%s\" refers to nonexistent package",
3896 PL_tokenbuf[len] = '\0';
3903 gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV);
3906 /* if we saw a global override before, get the right name */
3909 sv = newSVpvn("CORE::GLOBAL::",14);
3910 sv_catpv(sv,PL_tokenbuf);
3913 sv = newSVpv(PL_tokenbuf,0);
3915 /* Presume this is going to be a bareword of some sort. */
3918 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3919 yylval.opval->op_private = OPpCONST_BARE;
3921 /* And if "Foo::", then that's what it certainly is. */
3926 /* See if it's the indirect object for a list operator. */
3928 if (PL_oldoldbufptr &&
3929 PL_oldoldbufptr < PL_bufptr &&
3930 (PL_oldoldbufptr == PL_last_lop
3931 || PL_oldoldbufptr == PL_last_uni) &&
3932 /* NO SKIPSPACE BEFORE HERE! */
3933 (PL_expect == XREF ||
3934 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
3936 bool immediate_paren = *s == '(';
3938 /* (Now we can afford to cross potential line boundary.) */
3941 /* Two barewords in a row may indicate method call. */
3943 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') && (tmp=intuit_method(s,gv)))
3946 /* If not a declared subroutine, it's an indirect object. */
3947 /* (But it's an indir obj regardless for sort.) */
3949 if ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
3950 ((!gv || !GvCVu(gv)) &&
3951 (PL_last_lop_op != OP_MAPSTART &&
3952 PL_last_lop_op != OP_GREPSTART))))
3954 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
3960 PL_expect = XOPERATOR;
3963 /* Is this a word before a => operator? */
3964 if (*s == '=' && s[1] == '>') {
3966 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
3967 if (UTF && !IN_BYTE && is_utf8_string((U8*)PL_tokenbuf, len))
3968 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
3972 /* If followed by a paren, it's certainly a subroutine. */
3975 if (gv && GvCVu(gv)) {
3976 for (d = s + 1; SPACE_OR_TAB(*d); d++) ;
3977 if (*d == ')' && (sv = cv_const_sv(GvCV(gv)))) {
3982 PL_nextval[PL_nexttoke].opval = yylval.opval;
3983 PL_expect = XOPERATOR;
3989 /* If followed by var or block, call it a method (unless sub) */
3991 if ((*s == '$' || *s == '{') && (!gv || !GvCVu(gv))) {
3992 PL_last_lop = PL_oldbufptr;
3993 PL_last_lop_op = OP_METHOD;
3997 /* If followed by a bareword, see if it looks like indir obj. */
3999 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') && (tmp = intuit_method(s,gv)))
4002 /* Not a method, so call it a subroutine (if defined) */
4004 if (gv && GvCVu(gv)) {
4006 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
4007 Perl_warner(aTHX_ WARN_AMBIGUOUS,
4008 "Ambiguous use of -%s resolved as -&%s()",
4009 PL_tokenbuf, PL_tokenbuf);
4010 /* Check for a constant sub */
4012 if ((sv = cv_const_sv(cv))) {
4014 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
4015 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc(sv);
4016 yylval.opval->op_private = 0;
4020 /* Resolve to GV now. */
4021 op_free(yylval.opval);
4022 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
4023 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
4024 PL_last_lop = PL_oldbufptr;
4025 PL_last_lop_op = OP_ENTERSUB;
4026 /* Is there a prototype? */
4029 char *proto = SvPV((SV*)cv, len);
4032 if (strEQ(proto, "$"))
4034 if (*proto == '&' && *s == '{') {
4035 sv_setpv(PL_subname,"__ANON__");
4039 PL_nextval[PL_nexttoke].opval = yylval.opval;
4045 /* Call it a bare word */
4047 if (PL_hints & HINT_STRICT_SUBS)
4048 yylval.opval->op_private |= OPpCONST_STRICT;
4051 if (ckWARN(WARN_RESERVED)) {
4052 if (lastchar != '-') {
4053 for (d = PL_tokenbuf; *d && isLOWER(*d); d++) ;
4055 Perl_warner(aTHX_ WARN_RESERVED, PL_warn_reserved,
4062 if (lastchar && strchr("*%&", lastchar) && ckWARN_d(WARN_AMBIGUOUS)) {
4063 Perl_warner(aTHX_ WARN_AMBIGUOUS,
4064 "Operator or semicolon missing before %c%s",
4065 lastchar, PL_tokenbuf);
4066 Perl_warner(aTHX_ WARN_AMBIGUOUS,
4067 "Ambiguous use of %c resolved as operator %c",
4068 lastchar, lastchar);
4074 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4075 newSVpv(CopFILE(PL_curcop),0));
4079 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4080 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
4083 case KEY___PACKAGE__:
4084 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4086 ? newSVsv(PL_curstname)
4095 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
4096 char *pname = "main";
4097 if (PL_tokenbuf[2] == 'D')
4098 pname = HvNAME(PL_curstash ? PL_curstash : PL_defstash);
4099 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), TRUE, SVt_PVIO);
4102 GvIOp(gv) = newIO();
4103 IoIFP(GvIOp(gv)) = PL_rsfp;
4104 #if defined(HAS_FCNTL) && defined(F_SETFD)
4106 int fd = PerlIO_fileno(PL_rsfp);
4107 fcntl(fd,F_SETFD,fd >= 3);
4110 /* Mark this internal pseudo-handle as clean */
4111 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
4113 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
4114 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
4115 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
4117 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
4118 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
4119 /* if the script was opened in binmode, we need to revert
4120 * it to text mode for compatibility; but only iff it has CRs
4121 * XXX this is a questionable hack at best. */
4122 if (PL_bufend-PL_bufptr > 2
4123 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
4126 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
4127 loc = PerlIO_tell(PL_rsfp);
4128 (void)PerlIO_seek(PL_rsfp, 0L, 0);
4130 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
4131 #if defined(__BORLANDC__)
4132 /* XXX see note in do_binmode() */
4133 ((FILE*)PL_rsfp)->flags |= _F_BIN;
4136 PerlIO_seek(PL_rsfp, loc, 0);
4140 #ifdef PERLIO_LAYERS
4141 if (UTF && !IN_BYTE)
4142 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
4155 if (PL_expect == XSTATE) {
4162 if (*s == ':' && s[1] == ':') {
4165 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4166 if (!(tmp = keyword(PL_tokenbuf, len)))
4167 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
4181 LOP(OP_ACCEPT,XTERM);
4187 LOP(OP_ATAN2,XTERM);
4193 LOP(OP_BINMODE,XTERM);
4196 LOP(OP_BLESS,XTERM);
4205 (void)gv_fetchpv("ENV",TRUE, SVt_PVHV); /* may use HOME */
4222 if (!PL_cryptseen) {
4223 PL_cryptseen = TRUE;
4227 LOP(OP_CRYPT,XTERM);
4230 if (ckWARN(WARN_CHMOD)) {
4231 for (d = s; d < PL_bufend && (isSPACE(*d) || *d == '('); d++) ;
4232 if (*d != '0' && isDIGIT(*d))
4233 Perl_warner(aTHX_ WARN_CHMOD,
4234 "chmod() mode argument is missing initial 0");
4236 LOP(OP_CHMOD,XTERM);
4239 LOP(OP_CHOWN,XTERM);
4242 LOP(OP_CONNECT,XTERM);
4258 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4262 PL_hints |= HINT_BLOCK_SCOPE;
4272 gv_fetchpv("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
4273 LOP(OP_DBMOPEN,XTERM);
4279 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4286 yylval.ival = CopLINE(PL_curcop);
4300 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
4301 UNIBRACK(OP_ENTEREVAL);
4316 case KEY_endhostent:
4322 case KEY_endservent:
4325 case KEY_endprotoent:
4336 yylval.ival = CopLINE(PL_curcop);
4338 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
4340 if ((PL_bufend - p) >= 3 &&
4341 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
4343 else if ((PL_bufend - p) >= 4 &&
4344 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
4347 if (isIDFIRST_lazy_if(p,UTF)) {
4348 p = scan_ident(p, PL_bufend,
4349 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4353 Perl_croak(aTHX_ "Missing $ on loop variable");
4358 LOP(OP_FORMLINE,XTERM);
4364 LOP(OP_FCNTL,XTERM);
4370 LOP(OP_FLOCK,XTERM);
4379 LOP(OP_GREPSTART, XREF);
4382 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4397 case KEY_getpriority:
4398 LOP(OP_GETPRIORITY,XTERM);
4400 case KEY_getprotobyname:
4403 case KEY_getprotobynumber:
4404 LOP(OP_GPBYNUMBER,XTERM);
4406 case KEY_getprotoent:
4418 case KEY_getpeername:
4419 UNI(OP_GETPEERNAME);
4421 case KEY_gethostbyname:
4424 case KEY_gethostbyaddr:
4425 LOP(OP_GHBYADDR,XTERM);
4427 case KEY_gethostent:
4430 case KEY_getnetbyname:
4433 case KEY_getnetbyaddr:
4434 LOP(OP_GNBYADDR,XTERM);
4439 case KEY_getservbyname:
4440 LOP(OP_GSBYNAME,XTERM);
4442 case KEY_getservbyport:
4443 LOP(OP_GSBYPORT,XTERM);
4445 case KEY_getservent:
4448 case KEY_getsockname:
4449 UNI(OP_GETSOCKNAME);
4451 case KEY_getsockopt:
4452 LOP(OP_GSOCKOPT,XTERM);
4474 yylval.ival = CopLINE(PL_curcop);
4478 LOP(OP_INDEX,XTERM);
4484 LOP(OP_IOCTL,XTERM);
4496 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4528 LOP(OP_LISTEN,XTERM);
4537 s = scan_pat(s,OP_MATCH);
4538 TERM(sublex_start());
4541 LOP(OP_MAPSTART, XREF);
4544 LOP(OP_MKDIR,XTERM);
4547 LOP(OP_MSGCTL,XTERM);
4550 LOP(OP_MSGGET,XTERM);
4553 LOP(OP_MSGRCV,XTERM);
4556 LOP(OP_MSGSND,XTERM);
4562 if (isIDFIRST_lazy_if(s,UTF)) {
4563 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
4564 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
4566 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
4567 if (!PL_in_my_stash) {
4570 sprintf(tmpbuf, "No such class %.1000s", PL_tokenbuf);
4578 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4585 if (PL_expect != XSTATE)
4586 yyerror("\"no\" not allowed in expression");
4587 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4588 s = force_version(s);
4593 if (*s == '(' || (s = skipspace(s), *s == '('))
4600 if (isIDFIRST_lazy_if(s,UTF)) {
4602 for (d = s; isALNUM_lazy_if(d,UTF); d++) ;
4604 if (strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE))
4605 Perl_warner(aTHX_ WARN_PRECEDENCE,
4606 "Precedence problem: open %.*s should be open(%.*s)",
4612 yylval.ival = OP_OR;
4622 LOP(OP_OPEN_DIR,XTERM);
4625 checkcomma(s,PL_tokenbuf,"filehandle");
4629 checkcomma(s,PL_tokenbuf,"filehandle");
4648 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4652 LOP(OP_PIPE_OP,XTERM);
4655 s = scan_str(s,FALSE,FALSE);
4657 missingterm((char*)0);
4658 yylval.ival = OP_CONST;
4659 TERM(sublex_start());
4665 s = scan_str(s,FALSE,FALSE);
4667 missingterm((char*)0);
4669 if (SvCUR(PL_lex_stuff)) {
4672 d = SvPV_force(PL_lex_stuff, len);
4675 for (; isSPACE(*d) && len; --len, ++d) ;
4678 if (!warned && ckWARN(WARN_QW)) {
4679 for (; !isSPACE(*d) && len; --len, ++d) {
4681 Perl_warner(aTHX_ WARN_QW,
4682 "Possible attempt to separate words with commas");
4685 else if (*d == '#') {
4686 Perl_warner(aTHX_ WARN_QW,
4687 "Possible attempt to put comments in qw() list");
4693 for (; !isSPACE(*d) && len; --len, ++d) ;
4695 sv = newSVpvn(b, d-b);
4696 if (DO_UTF8(PL_lex_stuff))
4698 words = append_elem(OP_LIST, words,
4699 newSVOP(OP_CONST, 0, tokeq(sv)));
4703 PL_nextval[PL_nexttoke].opval = words;
4708 SvREFCNT_dec(PL_lex_stuff);
4709 PL_lex_stuff = Nullsv;
4715 s = scan_str(s,FALSE,FALSE);
4716 if (tmp == KEY_qu && is_utf8_string((U8*)s, SvCUR(PL_lex_stuff)))
4717 SvUTF8_on(PL_lex_stuff);
4719 missingterm((char*)0);
4720 yylval.ival = OP_STRINGIFY;
4721 if (SvIVX(PL_lex_stuff) == '\'')
4722 SvIVX(PL_lex_stuff) = 0; /* qq'$foo' should intepolate */
4723 TERM(sublex_start());
4726 s = scan_pat(s,OP_QR);
4727 TERM(sublex_start());
4730 s = scan_str(s,FALSE,FALSE);
4732 missingterm((char*)0);
4733 yylval.ival = OP_BACKTICK;
4735 TERM(sublex_start());
4742 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4743 s = force_version(s);
4746 *PL_tokenbuf = '\0';
4747 s = force_word(s,WORD,TRUE,TRUE,FALSE);
4748 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
4749 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
4751 yyerror("<> should be quotes");
4759 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4763 LOP(OP_RENAME,XTERM);
4772 LOP(OP_RINDEX,XTERM);
4795 LOP(OP_REVERSE,XTERM);
4806 TERM(sublex_start());
4808 TOKEN(1); /* force error */
4817 LOP(OP_SELECT,XTERM);
4823 LOP(OP_SEMCTL,XTERM);
4826 LOP(OP_SEMGET,XTERM);
4829 LOP(OP_SEMOP,XTERM);
4835 LOP(OP_SETPGRP,XTERM);
4837 case KEY_setpriority:
4838 LOP(OP_SETPRIORITY,XTERM);
4840 case KEY_sethostent:
4846 case KEY_setservent:
4849 case KEY_setprotoent:
4859 LOP(OP_SEEKDIR,XTERM);
4861 case KEY_setsockopt:
4862 LOP(OP_SSOCKOPT,XTERM);
4868 LOP(OP_SHMCTL,XTERM);
4871 LOP(OP_SHMGET,XTERM);
4874 LOP(OP_SHMREAD,XTERM);
4877 LOP(OP_SHMWRITE,XTERM);
4880 LOP(OP_SHUTDOWN,XTERM);
4889 LOP(OP_SOCKET,XTERM);
4891 case KEY_socketpair:
4892 LOP(OP_SOCKPAIR,XTERM);
4895 checkcomma(s,PL_tokenbuf,"subroutine name");
4897 if (*s == ';' || *s == ')') /* probably a close */
4898 Perl_croak(aTHX_ "sort is now a reserved word");
4900 s = force_word(s,WORD,TRUE,TRUE,FALSE);
4904 LOP(OP_SPLIT,XTERM);
4907 LOP(OP_SPRINTF,XTERM);
4910 LOP(OP_SPLICE,XTERM);
4925 LOP(OP_SUBSTR,XTERM);
4931 char tmpbuf[sizeof PL_tokenbuf];
4933 expectation attrful;
4934 bool have_name, have_proto;
4939 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
4940 (*s == ':' && s[1] == ':'))
4943 attrful = XATTRBLOCK;
4944 /* remember buffer pos'n for later force_word */
4945 tboffset = s - PL_oldbufptr;
4946 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4947 if (strchr(tmpbuf, ':'))
4948 sv_setpv(PL_subname, tmpbuf);
4950 sv_setsv(PL_subname,PL_curstname);
4951 sv_catpvn(PL_subname,"::",2);
4952 sv_catpvn(PL_subname,tmpbuf,len);
4959 Perl_croak(aTHX_ "Missing name in \"my sub\"");
4960 PL_expect = XTERMBLOCK;
4961 attrful = XATTRTERM;
4962 sv_setpv(PL_subname,"?");
4966 if (key == KEY_format) {
4968 PL_lex_formbrack = PL_lex_brackets + 1;
4970 (void) force_word(PL_oldbufptr + tboffset, WORD,
4975 /* Look for a prototype */
4979 s = scan_str(s,FALSE,FALSE);
4982 SvREFCNT_dec(PL_lex_stuff);
4983 PL_lex_stuff = Nullsv;
4984 Perl_croak(aTHX_ "Prototype not terminated");
4987 d = SvPVX(PL_lex_stuff);
4989 for (p = d; *p; ++p) {
4994 SvCUR(PL_lex_stuff) = tmp;
5002 if (*s == ':' && s[1] != ':')
5003 PL_expect = attrful;
5006 PL_nextval[PL_nexttoke].opval =
5007 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
5008 PL_lex_stuff = Nullsv;
5012 sv_setpv(PL_subname,"__ANON__");
5015 (void) force_word(PL_oldbufptr + tboffset, WORD,
5024 LOP(OP_SYSTEM,XREF);
5027 LOP(OP_SYMLINK,XTERM);
5030 LOP(OP_SYSCALL,XTERM);
5033 LOP(OP_SYSOPEN,XTERM);
5036 LOP(OP_SYSSEEK,XTERM);
5039 LOP(OP_SYSREAD,XTERM);
5042 LOP(OP_SYSWRITE,XTERM);
5046 TERM(sublex_start());
5067 LOP(OP_TRUNCATE,XTERM);
5079 yylval.ival = CopLINE(PL_curcop);
5083 yylval.ival = CopLINE(PL_curcop);
5087 LOP(OP_UNLINK,XTERM);
5093 LOP(OP_UNPACK,XTERM);
5096 LOP(OP_UTIME,XTERM);
5099 if (ckWARN(WARN_UMASK)) {
5100 for (d = s; d < PL_bufend && (isSPACE(*d) || *d == '('); d++) ;
5101 if (*d != '0' && isDIGIT(*d))
5102 Perl_warner(aTHX_ WARN_UMASK,
5103 "umask: argument is missing initial 0");
5108 LOP(OP_UNSHIFT,XTERM);
5111 if (PL_expect != XSTATE)
5112 yyerror("\"use\" not allowed in expression");
5114 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
5115 s = force_version(s);
5116 if (*s == ';' || (s = skipspace(s), *s == ';')) {
5117 PL_nextval[PL_nexttoke].opval = Nullop;
5122 s = force_word(s,WORD,FALSE,TRUE,FALSE);
5123 s = force_version(s);
5135 yylval.ival = CopLINE(PL_curcop);
5139 PL_hints |= HINT_BLOCK_SCOPE;
5146 LOP(OP_WAITPID,XTERM);
5154 static char ctl_l[2];
5156 if (ctl_l[0] == '\0')
5157 ctl_l[0] = toCTRL('L');
5158 gv_fetchpv(ctl_l,TRUE, SVt_PV);
5161 gv_fetchpv("\f",TRUE, SVt_PV); /* Make sure $^L is defined */
5166 if (PL_expect == XOPERATOR)
5172 yylval.ival = OP_XOR;
5177 TERM(sublex_start());
5182 #pragma segment Main
5186 Perl_keyword(pTHX_ register char *d, I32 len)
5191 if (strEQ(d,"__FILE__")) return -KEY___FILE__;
5192 if (strEQ(d,"__LINE__")) return -KEY___LINE__;
5193 if (strEQ(d,"__PACKAGE__")) return -KEY___PACKAGE__;
5194 if (strEQ(d,"__DATA__")) return KEY___DATA__;
5195 if (strEQ(d,"__END__")) return KEY___END__;
5199 if (strEQ(d,"AUTOLOAD")) return KEY_AUTOLOAD;
5204 if (strEQ(d,"and")) return -KEY_and;
5205 if (strEQ(d,"abs")) return -KEY_abs;
5208 if (strEQ(d,"alarm")) return -KEY_alarm;
5209 if (strEQ(d,"atan2")) return -KEY_atan2;
5212 if (strEQ(d,"accept")) return -KEY_accept;
5217 if (strEQ(d,"BEGIN")) return KEY_BEGIN;
5220 if (strEQ(d,"bless")) return -KEY_bless;
5221 if (strEQ(d,"bind")) return -KEY_bind;
5222 if (strEQ(d,"binmode")) return -KEY_binmode;
5225 if (strEQ(d,"CORE")) return -KEY_CORE;
5226 if (strEQ(d,"CHECK")) return KEY_CHECK;
5231 if (strEQ(d,"cmp")) return -KEY_cmp;
5232 if (strEQ(d,"chr")) return -KEY_chr;
5233 if (strEQ(d,"cos")) return -KEY_cos;
5236 if (strEQ(d,"chop")) return -KEY_chop;
5239 if (strEQ(d,"close")) return -KEY_close;
5240 if (strEQ(d,"chdir")) return -KEY_chdir;
5241 if (strEQ(d,"chomp")) return -KEY_chomp;
5242 if (strEQ(d,"chmod")) return -KEY_chmod;
5243 if (strEQ(d,"chown")) return -KEY_chown;
5244 if (strEQ(d,"crypt")) return -KEY_crypt;
5247 if (strEQ(d,"chroot")) return -KEY_chroot;
5248 if (strEQ(d,"caller")) return -KEY_caller;
5251 if (strEQ(d,"connect")) return -KEY_connect;
5254 if (strEQ(d,"closedir")) return -KEY_closedir;
5255 if (strEQ(d,"continue")) return -KEY_continue;
5260 if (strEQ(d,"DESTROY")) return KEY_DESTROY;
5265 if (strEQ(d,"do")) return KEY_do;
5268 if (strEQ(d,"die")) return -KEY_die;
5271 if (strEQ(d,"dump")) return -KEY_dump;
5274 if (strEQ(d,"delete")) return KEY_delete;
5277 if (strEQ(d,"defined")) return KEY_defined;
5278 if (strEQ(d,"dbmopen")) return -KEY_dbmopen;
5281 if (strEQ(d,"dbmclose")) return -KEY_dbmclose;
5286 if (strEQ(d,"END")) return KEY_END;
5291 if (strEQ(d,"eq")) return -KEY_eq;
5294 if (strEQ(d,"eof")) return -KEY_eof;
5295 if (strEQ(d,"exp")) return -KEY_exp;
5298 if (strEQ(d,"else")) return KEY_else;
5299 if (strEQ(d,"exit")) return -KEY_exit;
5300 if (strEQ(d,"eval")) return KEY_eval;
5301 if (strEQ(d,"exec")) return -KEY_exec;
5302 if (strEQ(d,"each")) return -KEY_each;
5305 if (strEQ(d,"elsif")) return KEY_elsif;
5308 if (strEQ(d,"exists")) return KEY_exists;
5309 if (strEQ(d,"elseif")) Perl_warn(aTHX_ "elseif should be elsif");
5312 if (strEQ(d,"endgrent")) return -KEY_endgrent;
5313 if (strEQ(d,"endpwent")) return -KEY_endpwent;
5316 if (strEQ(d,"endnetent")) return -KEY_endnetent;
5319 if (strEQ(d,"endhostent")) return -KEY_endhostent;
5320 if (strEQ(d,"endservent")) return -KEY_endservent;
5323 if (strEQ(d,"endprotoent")) return -KEY_endprotoent;
5330 if (strEQ(d,"for")) return KEY_for;
5333 if (strEQ(d,"fork")) return -KEY_fork;
5336 if (strEQ(d,"fcntl")) return -KEY_fcntl;
5337 if (strEQ(d,"flock")) return -KEY_flock;
5340 if (strEQ(d,"format")) return KEY_format;
5341 if (strEQ(d,"fileno")) return -KEY_fileno;
5344 if (strEQ(d,"foreach")) return KEY_foreach;
5347 if (strEQ(d,"formline")) return -KEY_formline;
5352 if (strnEQ(d,"get",3)) {
5357 if (strEQ(d,"ppid")) return -KEY_getppid;
5358 if (strEQ(d,"pgrp")) return -KEY_getpgrp;
5361 if (strEQ(d,"pwent")) return -KEY_getpwent;
5362 if (strEQ(d,"pwnam")) return -KEY_getpwnam;
5363 if (strEQ(d,"pwuid")) return -KEY_getpwuid;
5366 if (strEQ(d,"peername")) return -KEY_getpeername;
5367 if (strEQ(d,"protoent")) return -KEY_getprotoent;
5368 if (strEQ(d,"priority")) return -KEY_getpriority;
5371 if (strEQ(d,"protobyname")) return -KEY_getprotobyname;
5374 if (strEQ(d,"protobynumber"))return -KEY_getprotobynumber;
5378 else if (*d == 'h') {
5379 if (strEQ(d,"hostbyname")) return -KEY_gethostbyname;
5380 if (strEQ(d,"hostbyaddr")) return -KEY_gethostbyaddr;
5381 if (strEQ(d,"hostent")) return -KEY_gethostent;
5383 else if (*d == 'n') {
5384 if (strEQ(d,"netbyname")) return -KEY_getnetbyname;
5385 if (strEQ(d,"netbyaddr")) return -KEY_getnetbyaddr;
5386 if (strEQ(d,"netent")) return -KEY_getnetent;
5388 else if (*d == 's') {
5389 if (strEQ(d,"servbyname")) return -KEY_getservbyname;
5390 if (strEQ(d,"servbyport")) return -KEY_getservbyport;
5391 if (strEQ(d,"servent")) return -KEY_getservent;
5392 if (strEQ(d,"sockname")) return -KEY_getsockname;
5393 if (strEQ(d,"sockopt")) return -KEY_getsockopt;
5395 else if (*d == 'g') {
5396 if (strEQ(d,"grent")) return -KEY_getgrent;
5397 if (strEQ(d,"grnam")) return -KEY_getgrnam;
5398 if (strEQ(d,"grgid")) return -KEY_getgrgid;
5400 else if (*d == 'l') {
5401 if (strEQ(d,"login")) return -KEY_getlogin;
5403 else if (strEQ(d,"c")) return -KEY_getc;
5408 if (strEQ(d,"gt")) return -KEY_gt;
5409 if (strEQ(d,"ge")) return -KEY_ge;
5412 if (strEQ(d,"grep")) return KEY_grep;
5413 if (strEQ(d,"goto")) return KEY_goto;
5414 if (strEQ(d,"glob")) return KEY_glob;
5417 if (strEQ(d,"gmtime")) return -KEY_gmtime;
5422 if (strEQ(d,"hex")) return -KEY_hex;
5425 if (strEQ(d,"INIT")) return KEY_INIT;
5430 if (strEQ(d,"if")) return KEY_if;
5433 if (strEQ(d,"int")) return -KEY_int;
5436 if (strEQ(d,"index")) return -KEY_index;
5437 if (strEQ(d,"ioctl")) return -KEY_ioctl;
5442 if (strEQ(d,"join")) return -KEY_join;
5446 if (strEQ(d,"keys")) return -KEY_keys;
5447 if (strEQ(d,"kill")) return -KEY_kill;
5453 if (strEQ(d,"lt")) return -KEY_lt;
5454 if (strEQ(d,"le")) return -KEY_le;
5455 if (strEQ(d,"lc")) return -KEY_lc;
5458 if (strEQ(d,"log")) return -KEY_log;
5461 if (strEQ(d,"last")) return KEY_last;
5462 if (strEQ(d,"link")) return -KEY_link;
5463 if (strEQ(d,"lock")) return -KEY_lock;
5466 if (strEQ(d,"local")) return KEY_local;
5467 if (strEQ(d,"lstat")) return -KEY_lstat;
5470 if (strEQ(d,"length")) return -KEY_length;
5471 if (strEQ(d,"listen")) return -KEY_listen;
5474 if (strEQ(d,"lcfirst")) return -KEY_lcfirst;
5477 if (strEQ(d,"localtime")) return -KEY_localtime;
5483 case 1: return KEY_m;
5485 if (strEQ(d,"my")) return KEY_my;
5488 if (strEQ(d,"map")) return KEY_map;
5491 if (strEQ(d,"mkdir")) return -KEY_mkdir;
5494 if (strEQ(d,"msgctl")) return -KEY_msgctl;
5495 if (strEQ(d,"msgget")) return -KEY_msgget;
5496 if (strEQ(d,"msgrcv")) return -KEY_msgrcv;
5497 if (strEQ(d,"msgsnd")) return -KEY_msgsnd;
5502 if (strEQ(d,"next")) return KEY_next;
5503 if (strEQ(d,"ne")) return -KEY_ne;
5504 if (strEQ(d,"not")) return -KEY_not;
5505 if (strEQ(d,"no")) return KEY_no;
5510 if (strEQ(d,"or")) return -KEY_or;
5513 if (strEQ(d,"ord")) return -KEY_ord;
5514 if (strEQ(d,"oct")) return -KEY_oct;
5515 if (strEQ(d,"our")) return KEY_our;
5518 if (strEQ(d,"open")) return -KEY_open;
5521 if (strEQ(d,"opendir")) return -KEY_opendir;
5528 if (strEQ(d,"pop")) return -KEY_pop;
5529 if (strEQ(d,"pos")) return KEY_pos;
5532 if (strEQ(d,"push")) return -KEY_push;
5533 if (strEQ(d,"pack")) return -KEY_pack;
5534 if (strEQ(d,"pipe")) return -KEY_pipe;
5537 if (strEQ(d,"print")) return KEY_print;
5540 if (strEQ(d,"printf")) return KEY_printf;
5543 if (strEQ(d,"package")) return KEY_package;
5546 if (strEQ(d,"prototype")) return KEY_prototype;
5551 if (strEQ(d,"q")) return KEY_q;
5552 if (strEQ(d,"qr")) return KEY_qr;
5553 if (strEQ(d,"qq")) return KEY_qq;
5554 if (strEQ(d,"qu")) return KEY_qu;
5555 if (strEQ(d,"qw")) return KEY_qw;
5556 if (strEQ(d,"qx")) return KEY_qx;
5558 else if (strEQ(d,"quotemeta")) return -KEY_quotemeta;
5563 if (strEQ(d,"ref")) return -KEY_ref;
5566 if (strEQ(d,"read")) return -KEY_read;
5567 if (strEQ(d,"rand")) return -KEY_rand;
5568 if (strEQ(d,"recv")) return -KEY_recv;
5569 if (strEQ(d,"redo")) return KEY_redo;
5572 if (strEQ(d,"rmdir")) return -KEY_rmdir;
5573 if (strEQ(d,"reset")) return -KEY_reset;
5576 if (strEQ(d,"return")) return KEY_return;
5577 if (strEQ(d,"rename")) return -KEY_rename;
5578 if (strEQ(d,"rindex")) return -KEY_rindex;
5581 if (strEQ(d,"require")) return -KEY_require;
5582 if (strEQ(d,"reverse")) return -KEY_reverse;
5583 if (strEQ(d,"readdir")) return -KEY_readdir;
5586 if (strEQ(d,"readlink")) return -KEY_readlink;
5587 if (strEQ(d,"readline")) return -KEY_readline;
5588 if (strEQ(d,"readpipe")) return -KEY_readpipe;
5591 if (strEQ(d,"rewinddir")) return -KEY_rewinddir;
5597 case 0: return KEY_s;
5599 if (strEQ(d,"scalar")) return KEY_scalar;
5604 if (strEQ(d,"seek")) return -KEY_seek;
5605 if (strEQ(d,"send")) return -KEY_send;
5608 if (strEQ(d,"semop")) return -KEY_semop;
5611 if (strEQ(d,"select")) return -KEY_select;
5612 if (strEQ(d,"semctl")) return -KEY_semctl;
5613 if (strEQ(d,"semget")) return -KEY_semget;
5616 if (strEQ(d,"setpgrp")) return -KEY_setpgrp;
5617 if (strEQ(d,"seekdir")) return -KEY_seekdir;
5620 if (strEQ(d,"setpwent")) return -KEY_setpwent;
5621 if (strEQ(d,"setgrent")) return -KEY_setgrent;
5624 if (strEQ(d,"setnetent")) return -KEY_setnetent;
5627 if (strEQ(d,"setsockopt")) return -KEY_setsockopt;
5628 if (strEQ(d,"sethostent")) return -KEY_sethostent;
5629 if (strEQ(d,"setservent")) return -KEY_setservent;
5632 if (strEQ(d,"setpriority")) return -KEY_setpriority;
5633 if (strEQ(d,"setprotoent")) return -KEY_setprotoent;
5640 if (strEQ(d,"shift")) return -KEY_shift;
5643 if (strEQ(d,"shmctl")) return -KEY_shmctl;
5644 if (strEQ(d,"shmget")) return -KEY_shmget;
5647 if (strEQ(d,"shmread")) return -KEY_shmread;
5650 if (strEQ(d,"shmwrite")) return -KEY_shmwrite;
5651 if (strEQ(d,"shutdown")) return -KEY_shutdown;
5656 if (strEQ(d,"sin")) return -KEY_sin;
5659 if (strEQ(d,"sleep")) return -KEY_sleep;
5662 if (strEQ(d,"sort")) return KEY_sort;
5663 if (strEQ(d,"socket")) return -KEY_socket;
5664 if (strEQ(d,"socketpair")) return -KEY_socketpair;
5667 if (strEQ(d,"split")) return KEY_split;
5668 if (strEQ(d,"sprintf")) return -KEY_sprintf;
5669 if (strEQ(d,"splice")) return -KEY_splice;
5672 if (strEQ(d,"sqrt")) return -KEY_sqrt;
5675 if (strEQ(d,"srand")) return -KEY_srand;
5678 if (strEQ(d,"stat")) return -KEY_stat;
5679 if (strEQ(d,"study")) return KEY_study;
5682 if (strEQ(d,"substr")) return -KEY_substr;
5683 if (strEQ(d,"sub")) return KEY_sub;
5688 if (strEQ(d,"system")) return -KEY_system;
5691 if (strEQ(d,"symlink")) return -KEY_symlink;
5692 if (strEQ(d,"syscall")) return -KEY_syscall;
5693 if (strEQ(d,"sysopen")) return -KEY_sysopen;
5694 if (strEQ(d,"sysread")) return -KEY_sysread;
5695 if (strEQ(d,"sysseek")) return -KEY_sysseek;
5698 if (strEQ(d,"syswrite")) return -KEY_syswrite;
5707 if (strEQ(d,"tr")) return KEY_tr;
5710 if (strEQ(d,"tie")) return KEY_tie;
5713 if (strEQ(d,"tell")) return -KEY_tell;
5714 if (strEQ(d,"tied")) return KEY_tied;
5715 if (strEQ(d,"time")) return -KEY_time;
5718 if (strEQ(d,"times")) return -KEY_times;
5721 if (strEQ(d,"telldir")) return -KEY_telldir;
5724 if (strEQ(d,"truncate")) return -KEY_truncate;
5731 if (strEQ(d,"uc")) return -KEY_uc;
5734 if (strEQ(d,"use")) return KEY_use;
5737 if (strEQ(d,"undef")) return KEY_undef;
5738 if (strEQ(d,"until")) return KEY_until;
5739 if (strEQ(d,"untie")) return KEY_untie;
5740 if (strEQ(d,"utime")) return -KEY_utime;
5741 if (strEQ(d,"umask")) return -KEY_umask;
5744 if (strEQ(d,"unless")) return KEY_unless;
5745 if (strEQ(d,"unpack")) return -KEY_unpack;
5746 if (strEQ(d,"unlink")) return -KEY_unlink;
5749 if (strEQ(d,"unshift")) return -KEY_unshift;
5750 if (strEQ(d,"ucfirst")) return -KEY_ucfirst;
5755 if (strEQ(d,"values")) return -KEY_values;
5756 if (strEQ(d,"vec")) return -KEY_vec;
5761 if (strEQ(d,"warn")) return -KEY_warn;
5762 if (strEQ(d,"wait")) return -KEY_wait;
5765 if (strEQ(d,"while")) return KEY_while;
5766 if (strEQ(d,"write")) return -KEY_write;
5769 if (strEQ(d,"waitpid")) return -KEY_waitpid;
5772 if (strEQ(d,"wantarray")) return -KEY_wantarray;
5777 if (len == 1) return -KEY_x;
5778 if (strEQ(d,"xor")) return -KEY_xor;
5781 if (len == 1) return KEY_y;
5790 S_checkcomma(pTHX_ register char *s, char *name, char *what)
5794 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
5795 if (ckWARN(WARN_SYNTAX)) {
5797 for (w = s+2; *w && level; w++) {
5804 for (; *w && isSPACE(*w); w++) ;
5805 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
5806 Perl_warner(aTHX_ WARN_SYNTAX,
5807 "%s (...) interpreted as function",name);
5810 while (s < PL_bufend && isSPACE(*s))
5814 while (s < PL_bufend && isSPACE(*s))
5816 if (isIDFIRST_lazy_if(s,UTF)) {
5818 while (isALNUM_lazy_if(s,UTF))
5820 while (s < PL_bufend && isSPACE(*s))
5825 kw = keyword(w, s - w) || get_cv(w, FALSE) != 0;
5829 Perl_croak(aTHX_ "No comma allowed after %s", what);
5834 /* Either returns sv, or mortalizes sv and returns a new SV*.
5835 Best used as sv=new_constant(..., sv, ...).
5836 If s, pv are NULL, calls subroutine with one argument,
5837 and type is used with error messages only. */
5840 S_new_constant(pTHX_ char *s, STRLEN len, const char *key, SV *sv, SV *pv,
5844 HV *table = GvHV(PL_hintgv); /* ^H */
5848 const char *why1, *why2, *why3;
5850 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
5853 why2 = strEQ(key,"charnames")
5854 ? "(possibly a missing \"use charnames ...\")"
5856 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
5857 (type ? type: "undef"), why2);
5859 /* This is convoluted and evil ("goto considered harmful")
5860 * but I do not understand the intricacies of all the different
5861 * failure modes of %^H in here. The goal here is to make
5862 * the most probable error message user-friendly. --jhi */
5867 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
5868 (type ? type: "undef"), why1, why2, why3);
5870 yyerror(SvPVX(msg));
5874 cvp = hv_fetch(table, key, strlen(key), FALSE);
5875 if (!cvp || !SvOK(*cvp)) {
5878 why3 = "} is not defined";
5881 sv_2mortal(sv); /* Parent created it permanently */
5884 pv = sv_2mortal(newSVpvn(s, len));
5886 typesv = sv_2mortal(newSVpv(type, 0));
5888 typesv = &PL_sv_undef;
5890 PUSHSTACKi(PERLSI_OVERLOAD);
5902 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
5906 /* Check the eval first */
5907 if (!PL_in_eval && SvTRUE(ERRSV)) {
5909 sv_catpv(ERRSV, "Propagated");
5910 yyerror(SvPV(ERRSV, n_a)); /* Duplicates the message inside eval */
5912 res = SvREFCNT_inc(sv);
5916 (void)SvREFCNT_inc(res);
5925 why1 = "Call to &{$^H{";
5927 why3 = "}} did not return a defined value";
5936 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
5938 register char *d = dest;
5939 register char *e = d + destlen - 3; /* two-character token, ending NUL */
5942 Perl_croak(aTHX_ ident_too_long);
5943 if (isALNUM(*s)) /* UTF handled below */
5945 else if (*s == '\'' && allow_package && isIDFIRST_lazy_if(s+1,UTF)) {
5950 else if (*s == ':' && s[1] == ':' && allow_package && s[2] != '$') {
5954 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
5955 char *t = s + UTF8SKIP(s);
5956 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
5958 if (d + (t - s) > e)
5959 Perl_croak(aTHX_ ident_too_long);
5960 Copy(s, d, t - s, char);
5973 S_scan_ident(pTHX_ register char *s, register char *send, char *dest, STRLEN destlen, I32 ck_uni)
5983 e = d + destlen - 3; /* two-character token, ending NUL */
5985 while (isDIGIT(*s)) {
5987 Perl_croak(aTHX_ ident_too_long);
5994 Perl_croak(aTHX_ ident_too_long);
5995 if (isALNUM(*s)) /* UTF handled below */
5997 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
6002 else if (*s == ':' && s[1] == ':') {
6006 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
6007 char *t = s + UTF8SKIP(s);
6008 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
6010 if (d + (t - s) > e)
6011 Perl_croak(aTHX_ ident_too_long);
6012 Copy(s, d, t - s, char);
6023 if (PL_lex_state != LEX_NORMAL)
6024 PL_lex_state = LEX_INTERPENDMAYBE;
6027 if (*s == '$' && s[1] &&
6028 (isALNUM_lazy_if(s+1,UTF) || strchr("${", s[1]) || strnEQ(s+1,"::",2)) )
6041 if (*d == '^' && *s && isCONTROLVAR(*s)) {
6046 if (isSPACE(s[-1])) {
6049 if (!SPACE_OR_TAB(ch)) {
6055 if (isIDFIRST_lazy_if(d,UTF)) {
6059 while ((e < send && isALNUM_lazy_if(e,UTF)) || *e == ':') {
6061 while (e < send && UTF8_IS_CONTINUED(*e) && is_utf8_mark((U8*)e))
6064 Copy(s, d, e - s, char);
6069 while ((isALNUM(*s) || *s == ':') && d < e)
6072 Perl_croak(aTHX_ ident_too_long);
6075 while (s < send && SPACE_OR_TAB(*s)) s++;
6076 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
6077 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest)) {
6078 const char *brack = *s == '[' ? "[...]" : "{...}";
6079 Perl_warner(aTHX_ WARN_AMBIGUOUS,
6080 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
6081 funny, dest, brack, funny, dest, brack);
6084 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
6088 /* Handle extended ${^Foo} variables
6089 * 1999-02-27 mjd-perl-patch@plover.com */
6090 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
6094 while (isALNUM(*s) && d < e) {
6098 Perl_croak(aTHX_ ident_too_long);
6103 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets)
6104 PL_lex_state = LEX_INTERPEND;
6107 if (PL_lex_state == LEX_NORMAL) {
6108 if (ckWARN(WARN_AMBIGUOUS) &&
6109 (keyword(dest, d - dest) || get_cv(dest, FALSE)))
6111 Perl_warner(aTHX_ WARN_AMBIGUOUS,
6112 "Ambiguous use of %c{%s} resolved to %c%s",
6113 funny, dest, funny, dest);
6118 s = bracket; /* let the parser handle it */
6122 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
6123 PL_lex_state = LEX_INTERPEND;
6128 Perl_pmflag(pTHX_ U16 *pmfl, int ch)
6133 *pmfl |= PMf_GLOBAL;
6135 *pmfl |= PMf_CONTINUE;
6139 *pmfl |= PMf_MULTILINE;
6141 *pmfl |= PMf_SINGLELINE;
6143 *pmfl |= PMf_EXTENDED;
6147 S_scan_pat(pTHX_ char *start, I32 type)
6152 s = scan_str(start,FALSE,FALSE);
6155 SvREFCNT_dec(PL_lex_stuff);
6156 PL_lex_stuff = Nullsv;
6157 Perl_croak(aTHX_ "Search pattern not terminated");
6160 pm = (PMOP*)newPMOP(type, 0);
6161 if (PL_multi_open == '?')
6162 pm->op_pmflags |= PMf_ONCE;
6164 while (*s && strchr("iomsx", *s))
6165 pmflag(&pm->op_pmflags,*s++);
6168 while (*s && strchr("iogcmsx", *s))
6169 pmflag(&pm->op_pmflags,*s++);
6171 pm->op_pmpermflags = pm->op_pmflags;
6173 PL_lex_op = (OP*)pm;
6174 yylval.ival = OP_MATCH;
6179 S_scan_subst(pTHX_ char *start)
6186 yylval.ival = OP_NULL;
6188 s = scan_str(start,FALSE,FALSE);
6192 SvREFCNT_dec(PL_lex_stuff);
6193 PL_lex_stuff = Nullsv;
6194 Perl_croak(aTHX_ "Substitution pattern not terminated");
6197 if (s[-1] == PL_multi_open)
6200 first_start = PL_multi_start;
6201 s = scan_str(s,FALSE,FALSE);
6204 SvREFCNT_dec(PL_lex_stuff);
6205 PL_lex_stuff = Nullsv;
6207 SvREFCNT_dec(PL_lex_repl);
6208 PL_lex_repl = Nullsv;
6209 Perl_croak(aTHX_ "Substitution replacement not terminated");
6211 PL_multi_start = first_start; /* so whole substitution is taken together */
6213 pm = (PMOP*)newPMOP(OP_SUBST, 0);
6219 else if (strchr("iogcmsx", *s))
6220 pmflag(&pm->op_pmflags,*s++);
6227 PL_sublex_info.super_bufptr = s;
6228 PL_sublex_info.super_bufend = PL_bufend;
6230 pm->op_pmflags |= PMf_EVAL;
6231 repl = newSVpvn("",0);
6233 sv_catpv(repl, es ? "eval " : "do ");
6234 sv_catpvn(repl, "{ ", 2);
6235 sv_catsv(repl, PL_lex_repl);
6236 sv_catpvn(repl, " };", 2);
6238 SvREFCNT_dec(PL_lex_repl);
6242 pm->op_pmpermflags = pm->op_pmflags;
6243 PL_lex_op = (OP*)pm;
6244 yylval.ival = OP_SUBST;
6249 S_scan_trans(pTHX_ char *start)
6260 yylval.ival = OP_NULL;
6262 s = scan_str(start,FALSE,FALSE);
6265 SvREFCNT_dec(PL_lex_stuff);
6266 PL_lex_stuff = Nullsv;
6267 Perl_croak(aTHX_ "Transliteration pattern not terminated");
6269 if (s[-1] == PL_multi_open)
6272 s = scan_str(s,FALSE,FALSE);
6275 SvREFCNT_dec(PL_lex_stuff);
6276 PL_lex_stuff = Nullsv;
6278 SvREFCNT_dec(PL_lex_repl);
6279 PL_lex_repl = Nullsv;
6280 Perl_croak(aTHX_ "Transliteration replacement not terminated");
6283 New(803,tbl,256,short);
6284 o = newPVOP(OP_TRANS, 0, (char*)tbl);
6286 complement = del = squash = 0;
6287 while (strchr("cds", *s)) {
6289 complement = OPpTRANS_COMPLEMENT;
6291 del = OPpTRANS_DELETE;
6293 squash = OPpTRANS_SQUASH;
6296 o->op_private = del|squash|complement|
6297 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
6298 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
6301 yylval.ival = OP_TRANS;
6306 S_scan_heredoc(pTHX_ register char *s)
6309 I32 op_type = OP_SCALAR;
6316 int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
6320 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
6323 for (peek = s; SPACE_OR_TAB(*peek); peek++) ;
6324 if (*peek && strchr("`'\"",*peek)) {
6327 s = delimcpy(d, e, s, PL_bufend, term, &len);
6337 if (!isALNUM_lazy_if(s,UTF))
6338 deprecate("bare << to mean <<\"\"");
6339 for (; isALNUM_lazy_if(s,UTF); s++) {
6344 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
6345 Perl_croak(aTHX_ "Delimiter for here document is too long");
6348 len = d - PL_tokenbuf;
6349 #ifndef PERL_STRICT_CR
6350 d = strchr(s, '\r');
6354 while (s < PL_bufend) {
6360 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
6369 SvCUR_set(PL_linestr, PL_bufend - SvPVX(PL_linestr));
6374 if (outer || !(d=ninstr(s,PL_bufend,d,d+1)))
6375 herewas = newSVpvn(s,PL_bufend-s);
6377 s--, herewas = newSVpvn(s,d-s);
6378 s += SvCUR(herewas);
6380 tmpstr = NEWSV(87,79);
6381 sv_upgrade(tmpstr, SVt_PVIV);
6386 else if (term == '`') {
6387 op_type = OP_BACKTICK;
6388 SvIVX(tmpstr) = '\\';
6392 PL_multi_start = CopLINE(PL_curcop);
6393 PL_multi_open = PL_multi_close = '<';
6394 term = *PL_tokenbuf;
6395 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
6396 char *bufptr = PL_sublex_info.super_bufptr;
6397 char *bufend = PL_sublex_info.super_bufend;
6398 char *olds = s - SvCUR(herewas);
6399 s = strchr(bufptr, '\n');
6403 while (s < bufend &&
6404 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
6406 CopLINE_inc(PL_curcop);
6409 CopLINE_set(PL_curcop, PL_multi_start);
6410 missingterm(PL_tokenbuf);
6412 sv_setpvn(herewas,bufptr,d-bufptr+1);
6413 sv_setpvn(tmpstr,d+1,s-d);
6415 sv_catpvn(herewas,s,bufend-s);
6416 (void)strcpy(bufptr,SvPVX(herewas));
6423 while (s < PL_bufend &&
6424 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
6426 CopLINE_inc(PL_curcop);
6428 if (s >= PL_bufend) {
6429 CopLINE_set(PL_curcop, PL_multi_start);
6430 missingterm(PL_tokenbuf);
6432 sv_setpvn(tmpstr,d+1,s-d);
6434 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
6436 sv_catpvn(herewas,s,PL_bufend-s);
6437 sv_setsv(PL_linestr,herewas);
6438 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
6439 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
6442 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
6443 while (s >= PL_bufend) { /* multiple line string? */
6445 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
6446 CopLINE_set(PL_curcop, PL_multi_start);
6447 missingterm(PL_tokenbuf);
6449 CopLINE_inc(PL_curcop);
6450 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
6451 #ifndef PERL_STRICT_CR
6452 if (PL_bufend - PL_linestart >= 2) {
6453 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
6454 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
6456 PL_bufend[-2] = '\n';
6458 SvCUR_set(PL_linestr, PL_bufend - SvPVX(PL_linestr));
6460 else if (PL_bufend[-1] == '\r')
6461 PL_bufend[-1] = '\n';
6463 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
6464 PL_bufend[-1] = '\n';
6466 if (PERLDB_LINE && PL_curstash != PL_debstash) {
6467 SV *sv = NEWSV(88,0);
6469 sv_upgrade(sv, SVt_PVMG);
6470 sv_setsv(sv,PL_linestr);
6471 av_store(CopFILEAV(PL_curcop), (I32)CopLINE(PL_curcop),sv);
6473 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
6476 sv_catsv(PL_linestr,herewas);
6477 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
6481 sv_catsv(tmpstr,PL_linestr);
6486 PL_multi_end = CopLINE(PL_curcop);
6487 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
6488 SvLEN_set(tmpstr, SvCUR(tmpstr) + 1);
6489 Renew(SvPVX(tmpstr), SvLEN(tmpstr), char);
6491 SvREFCNT_dec(herewas);
6492 if (UTF && !IN_BYTE && is_utf8_string((U8*)SvPVX(tmpstr), SvCUR(tmpstr)))
6494 PL_lex_stuff = tmpstr;
6495 yylval.ival = op_type;
6500 takes: current position in input buffer
6501 returns: new position in input buffer
6502 side-effects: yylval and lex_op are set.
6507 <FH> read from filehandle
6508 <pkg::FH> read from package qualified filehandle
6509 <pkg'FH> read from package qualified filehandle
6510 <$fh> read from filehandle in $fh
6516 S_scan_inputsymbol(pTHX_ char *start)
6518 register char *s = start; /* current position in buffer */
6524 d = PL_tokenbuf; /* start of temp holding space */
6525 e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
6526 end = strchr(s, '\n');
6529 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
6531 /* die if we didn't have space for the contents of the <>,
6532 or if it didn't end, or if we see a newline
6535 if (len >= sizeof PL_tokenbuf)
6536 Perl_croak(aTHX_ "Excessively long <> operator");
6538 Perl_croak(aTHX_ "Unterminated <> operator");
6543 Remember, only scalar variables are interpreted as filehandles by
6544 this code. Anything more complex (e.g., <$fh{$num}>) will be
6545 treated as a glob() call.
6546 This code makes use of the fact that except for the $ at the front,
6547 a scalar variable and a filehandle look the same.
6549 if (*d == '$' && d[1]) d++;
6551 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
6552 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
6555 /* If we've tried to read what we allow filehandles to look like, and
6556 there's still text left, then it must be a glob() and not a getline.
6557 Use scan_str to pull out the stuff between the <> and treat it
6558 as nothing more than a string.
6561 if (d - PL_tokenbuf != len) {
6562 yylval.ival = OP_GLOB;
6564 s = scan_str(start,FALSE,FALSE);
6566 Perl_croak(aTHX_ "Glob not terminated");
6570 /* we're in a filehandle read situation */
6573 /* turn <> into <ARGV> */
6575 (void)strcpy(d,"ARGV");
6577 /* if <$fh>, create the ops to turn the variable into a
6583 /* try to find it in the pad for this block, otherwise find
6584 add symbol table ops
6586 if ((tmp = pad_findmy(d)) != NOT_IN_PAD) {
6587 OP *o = newOP(OP_PADSV, 0);
6589 PL_lex_op = (OP*)newUNOP(OP_READLINE, 0, o);
6592 GV *gv = gv_fetchpv(d+1,TRUE, SVt_PV);
6593 PL_lex_op = (OP*)newUNOP(OP_READLINE, 0,
6594 newUNOP(OP_RV2SV, 0,
6595 newGVOP(OP_GV, 0, gv)));
6597 PL_lex_op->op_flags |= OPf_SPECIAL;
6598 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
6599 yylval.ival = OP_NULL;
6602 /* If it's none of the above, it must be a literal filehandle
6603 (<Foo::BAR> or <FOO>) so build a simple readline OP */
6605 GV *gv = gv_fetchpv(d,TRUE, SVt_PVIO);
6606 PL_lex_op = (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
6607 yylval.ival = OP_NULL;
6616 takes: start position in buffer
6617 keep_quoted preserve \ on the embedded delimiter(s)
6618 keep_delims preserve the delimiters around the string
6619 returns: position to continue reading from buffer
6620 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
6621 updates the read buffer.
6623 This subroutine pulls a string out of the input. It is called for:
6624 q single quotes q(literal text)
6625 ' single quotes 'literal text'
6626 qq double quotes qq(interpolate $here please)
6627 " double quotes "interpolate $here please"
6628 qx backticks qx(/bin/ls -l)
6629 ` backticks `/bin/ls -l`
6630 qw quote words @EXPORT_OK = qw( func() $spam )
6631 m// regexp match m/this/
6632 s/// regexp substitute s/this/that/
6633 tr/// string transliterate tr/this/that/
6634 y/// string transliterate y/this/that/
6635 ($*@) sub prototypes sub foo ($)
6636 (stuff) sub attr parameters sub foo : attr(stuff)
6637 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
6639 In most of these cases (all but <>, patterns and transliterate)
6640 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
6641 calls scan_str(). s/// makes yylex() call scan_subst() which calls
6642 scan_str(). tr/// and y/// make yylex() call scan_trans() which
6645 It skips whitespace before the string starts, and treats the first
6646 character as the delimiter. If the delimiter is one of ([{< then
6647 the corresponding "close" character )]}> is used as the closing
6648 delimiter. It allows quoting of delimiters, and if the string has
6649 balanced delimiters ([{<>}]) it allows nesting.
6651 The lexer always reads these strings into lex_stuff, except in the
6652 case of the operators which take *two* arguments (s/// and tr///)
6653 when it checks to see if lex_stuff is full (presumably with the 1st
6654 arg to s or tr) and if so puts the string into lex_repl.
6659 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
6661 SV *sv; /* scalar value: string */
6662 char *tmps; /* temp string, used for delimiter matching */
6663 register char *s = start; /* current position in the buffer */
6664 register char term; /* terminating character */
6665 register char *to; /* current position in the sv's data */
6666 I32 brackets = 1; /* bracket nesting level */
6667 bool has_utf8 = FALSE; /* is there any utf8 content? */
6669 /* skip space before the delimiter */
6673 /* mark where we are, in case we need to report errors */
6676 /* after skipping whitespace, the next character is the terminator */
6678 if (UTF8_IS_CONTINUED(term) && UTF)
6681 /* mark where we are */
6682 PL_multi_start = CopLINE(PL_curcop);
6683 PL_multi_open = term;
6685 /* find corresponding closing delimiter */
6686 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
6688 PL_multi_close = term;
6690 /* create a new SV to hold the contents. 87 is leak category, I'm
6691 assuming. 79 is the SV's initial length. What a random number. */
6693 sv_upgrade(sv, SVt_PVIV);
6695 (void)SvPOK_only(sv); /* validate pointer */
6697 /* move past delimiter and try to read a complete string */
6699 sv_catpvn(sv, s, 1);
6702 /* extend sv if need be */
6703 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
6704 /* set 'to' to the next character in the sv's string */
6705 to = SvPVX(sv)+SvCUR(sv);
6707 /* if open delimiter is the close delimiter read unbridle */
6708 if (PL_multi_open == PL_multi_close) {
6709 for (; s < PL_bufend; s++,to++) {
6710 /* embedded newlines increment the current line number */
6711 if (*s == '\n' && !PL_rsfp)
6712 CopLINE_inc(PL_curcop);
6713 /* handle quoted delimiters */
6714 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
6715 if (!keep_quoted && s[1] == term)
6717 /* any other quotes are simply copied straight through */
6721 /* terminate when run out of buffer (the for() condition), or
6722 have found the terminator */
6723 else if (*s == term)
6725 else if (!has_utf8 && UTF8_IS_CONTINUED(*s) && UTF)
6731 /* if the terminator isn't the same as the start character (e.g.,
6732 matched brackets), we have to allow more in the quoting, and
6733 be prepared for nested brackets.
6736 /* read until we run out of string, or we find the terminator */
6737 for (; s < PL_bufend; s++,to++) {
6738 /* embedded newlines increment the line count */
6739 if (*s == '\n' && !PL_rsfp)
6740 CopLINE_inc(PL_curcop);
6741 /* backslashes can escape the open or closing characters */
6742 if (*s == '\\' && s+1 < PL_bufend) {
6744 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
6749 /* allow nested opens and closes */
6750 else if (*s == PL_multi_close && --brackets <= 0)
6752 else if (*s == PL_multi_open)
6754 else if (!has_utf8 && UTF8_IS_CONTINUED(*s) && UTF)
6759 /* terminate the copied string and update the sv's end-of-string */
6761 SvCUR_set(sv, to - SvPVX(sv));
6764 * this next chunk reads more into the buffer if we're not done yet
6768 break; /* handle case where we are done yet :-) */
6770 #ifndef PERL_STRICT_CR
6771 if (to - SvPVX(sv) >= 2) {
6772 if ((to[-2] == '\r' && to[-1] == '\n') ||
6773 (to[-2] == '\n' && to[-1] == '\r'))
6777 SvCUR_set(sv, to - SvPVX(sv));
6779 else if (to[-1] == '\r')
6782 else if (to - SvPVX(sv) == 1 && to[-1] == '\r')
6786 /* if we're out of file, or a read fails, bail and reset the current
6787 line marker so we can report where the unterminated string began
6790 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
6792 CopLINE_set(PL_curcop, PL_multi_start);
6795 /* we read a line, so increment our line counter */
6796 CopLINE_inc(PL_curcop);
6798 /* update debugger info */
6799 if (PERLDB_LINE && PL_curstash != PL_debstash) {
6800 SV *sv = NEWSV(88,0);
6802 sv_upgrade(sv, SVt_PVMG);
6803 sv_setsv(sv,PL_linestr);
6804 av_store(CopFILEAV(PL_curcop), (I32)CopLINE(PL_curcop), sv);
6807 /* having changed the buffer, we must update PL_bufend */
6808 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
6811 /* at this point, we have successfully read the delimited string */
6814 sv_catpvn(sv, s, 1);
6817 PL_multi_end = CopLINE(PL_curcop);
6820 /* if we allocated too much space, give some back */
6821 if (SvCUR(sv) + 5 < SvLEN(sv)) {
6822 SvLEN_set(sv, SvCUR(sv) + 1);
6823 Renew(SvPVX(sv), SvLEN(sv), char);
6826 /* decide whether this is the first or second quoted string we've read
6839 takes: pointer to position in buffer
6840 returns: pointer to new position in buffer
6841 side-effects: builds ops for the constant in yylval.op
6843 Read a number in any of the formats that Perl accepts:
6845 0(x[0-7A-F]+)|([0-7]+)|(b[01])
6846 [\d_]+(\.[\d_]*)?[Ee](\d+)
6848 Underbars (_) are allowed in decimal numbers. If -w is on,
6849 underbars before a decimal point must be at three digit intervals.
6851 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
6854 If it reads a number without a decimal point or an exponent, it will
6855 try converting the number to an integer and see if it can do so
6856 without loss of precision.
6860 Perl_scan_num(pTHX_ char *start, YYSTYPE* lvalp)
6862 register char *s = start; /* current position in buffer */
6863 register char *d; /* destination in temp buffer */
6864 register char *e; /* end of temp buffer */
6865 NV nv; /* number read, as a double */
6866 SV *sv = Nullsv; /* place to put the converted number */
6867 bool floatit; /* boolean: int or float? */
6868 char *lastub = 0; /* position of last underbar */
6869 static char number_too_long[] = "Number too long";
6871 /* We use the first character to decide what type of number this is */
6875 Perl_croak(aTHX_ "panic: scan_num");
6877 /* if it starts with a 0, it could be an octal number, a decimal in
6878 0.13 disguise, or a hexadecimal number, or a binary number. */
6882 u holds the "number so far"
6883 shift the power of 2 of the base
6884 (hex == 4, octal == 3, binary == 1)
6885 overflowed was the number more than we can hold?
6887 Shift is used when we add a digit. It also serves as an "are
6888 we in octal/hex/binary?" indicator to disallow hex characters
6894 bool overflowed = FALSE;
6895 static NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
6896 static char* bases[5] = { "", "binary", "", "octal",
6898 static char* Bases[5] = { "", "Binary", "", "Octal",
6900 static char *maxima[5] = { "",
6901 "0b11111111111111111111111111111111",
6905 char *base, *Base, *max;
6911 } else if (s[1] == 'b') {
6915 /* check for a decimal in disguise */
6916 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
6918 /* so it must be octal */
6922 base = bases[shift];
6923 Base = Bases[shift];
6924 max = maxima[shift];
6926 /* read the rest of the number */
6928 /* x is used in the overflow test,
6929 b is the digit we're adding on. */
6934 /* if we don't mention it, we're done */
6943 /* 8 and 9 are not octal */
6946 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
6950 case '2': case '3': case '4':
6951 case '5': case '6': case '7':
6953 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
6957 b = *s++ & 15; /* ASCII digit -> value of digit */
6961 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
6962 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
6963 /* make sure they said 0x */
6968 /* Prepare to put the digit we have onto the end
6969 of the number so far. We check for overflows.
6974 x = u << shift; /* make room for the digit */
6976 if ((x >> shift) != u
6977 && !(PL_hints & HINT_NEW_BINARY)) {
6980 if (ckWARN_d(WARN_OVERFLOW))
6981 Perl_warner(aTHX_ WARN_OVERFLOW,
6982 "Integer overflow in %s number",
6985 u = x | b; /* add the digit to the end */
6988 n *= nvshift[shift];
6989 /* If an NV has not enough bits in its
6990 * mantissa to represent an UV this summing of
6991 * small low-order numbers is a waste of time
6992 * (because the NV cannot preserve the
6993 * low-order bits anyway): we could just
6994 * remember when did we overflow and in the
6995 * end just multiply n by the right
7003 /* if we get here, we had success: make a scalar value from
7009 if (ckWARN(WARN_PORTABLE) && n > 4294967295.0)
7010 Perl_warner(aTHX_ WARN_PORTABLE,
7011 "%s number > %s non-portable",
7017 if (ckWARN(WARN_PORTABLE) && u > 0xffffffff)
7018 Perl_warner(aTHX_ WARN_PORTABLE,
7019 "%s number > %s non-portable",
7024 if (PL_hints & HINT_NEW_BINARY)
7025 sv = new_constant(start, s - start, "binary", sv, Nullsv, NULL);
7030 handle decimal numbers.
7031 we're also sent here when we read a 0 as the first digit
7033 case '1': case '2': case '3': case '4': case '5':
7034 case '6': case '7': case '8': case '9': case '.':
7037 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
7040 /* read next group of digits and _ and copy into d */
7041 while (isDIGIT(*s) || *s == '_') {
7042 /* skip underscores, checking for misplaced ones
7046 if (ckWARN(WARN_SYNTAX) && lastub && s - lastub != 3)
7047 Perl_warner(aTHX_ WARN_SYNTAX, "Misplaced _ in number");
7051 /* check for end of fixed-length buffer */
7053 Perl_croak(aTHX_ number_too_long);
7054 /* if we're ok, copy the character */
7059 /* final misplaced underbar check */
7060 if (lastub && s - lastub != 3) {
7061 if (ckWARN(WARN_SYNTAX))
7062 Perl_warner(aTHX_ WARN_SYNTAX, "Misplaced _ in number");
7065 /* read a decimal portion if there is one. avoid
7066 3..5 being interpreted as the number 3. followed
7069 if (*s == '.' && s[1] != '.') {
7073 /* copy, ignoring underbars, until we run out of
7074 digits. Note: no misplaced underbar checks!
7076 for (; isDIGIT(*s) || *s == '_'; s++) {
7077 /* fixed length buffer check */
7079 Perl_croak(aTHX_ number_too_long);
7083 if (*s == '.' && isDIGIT(s[1])) {
7084 /* oops, it's really a v-string, but without the "v" */
7090 /* read exponent part, if present */
7091 if (*s && strchr("eE",*s) && strchr("+-0123456789",s[1])) {
7095 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
7096 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
7098 /* allow positive or negative exponent */
7099 if (*s == '+' || *s == '-')
7102 /* read digits of exponent (no underbars :-) */
7103 while (isDIGIT(*s)) {
7105 Perl_croak(aTHX_ number_too_long);
7110 /* terminate the string */
7113 /* make an sv from the string */
7116 #if defined(Strtol) && defined(Strtoul)
7119 strtol/strtoll sets errno to ERANGE if the number is too big
7120 for an integer. We try to do an integer conversion first
7121 if no characters indicating "float" have been found.
7128 if (*PL_tokenbuf == '-')
7129 iv = Strtol(PL_tokenbuf, (char**)NULL, 10);
7131 uv = Strtoul(PL_tokenbuf, (char**)NULL, 10);
7133 floatit = TRUE; /* Probably just too large. */
7134 else if (*PL_tokenbuf == '-')
7136 else if (uv <= IV_MAX)
7137 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
7142 nv = Atof(PL_tokenbuf);
7147 No working strtou?ll?.
7149 Unfortunately atol() doesn't do range checks (returning
7150 LONG_MIN/LONG_MAX, and setting errno to ERANGE on overflows)
7151 everywhere [1], so we cannot use use atol() (or atoll()).
7152 If we could, they would be used, as Atol(), very much like
7153 Strtol() and Strtoul() are used above.
7155 [1] XXX Configure test needed to check for atol()
7156 (and atoll()) overflow behaviour XXX
7160 We need to do this the hard way. */
7162 nv = Atof(PL_tokenbuf);
7164 /* See if we can make do with an integer value without loss of
7165 precision. We use U_V to cast to a UV, because some
7166 compilers have issues. Then we try casting it back and see
7167 if it was the same [1]. We only do this if we know we
7168 specifically read an integer. If floatit is true, then we
7169 don't need to do the conversion at all.
7171 [1] Note that this is lossy if our NVs cannot preserve our
7172 UVs. There are metaconfig defines NV_PRESERVES_UV (a boolean)
7173 and NV_PRESERVES_UV_BITS (a number), but in general we really
7174 do hope all such potentially lossy platforms have strtou?ll?
7175 to do a lossless IV/UV conversion.
7177 Maybe could do some tricks with DBL_DIG, LDBL_DIG and
7178 DBL_MANT_DIG and LDBL_MANT_DIG (these are already available
7179 as NV_DIG and NV_MANT_DIG)?
7185 if (!floatit && (NV)uv == nv) {
7187 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
7195 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
7196 (PL_hints & HINT_NEW_INTEGER) )
7197 sv = new_constant(PL_tokenbuf, d - PL_tokenbuf,
7198 (floatit ? "float" : "integer"),
7202 /* if it starts with a v, it could be a v-string */
7208 while (isDIGIT(*pos) || *pos == '_')
7210 if (!isALPHA(*pos)) {
7212 U8 tmpbuf[UTF8_MAXLEN+1];
7214 s++; /* get past 'v' */
7217 sv_setpvn(sv, "", 0);
7220 if (*s == '0' && isDIGIT(s[1]))
7221 yyerror("Octal number in vector unsupported");
7224 /* this is atoi() that tolerates underscores */
7227 while (--end >= s) {
7232 rev += (*end - '0') * mult;
7234 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
7235 Perl_warner(aTHX_ WARN_OVERFLOW,
7236 "Integer overflow in decimal number");
7239 tmpend = uv_to_utf8(tmpbuf, rev);
7242 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
7243 if (*pos == '.' && isDIGIT(pos[1]))
7249 while (isDIGIT(*pos) || *pos == '_')
7258 sv_utf8_downgrade(sv, TRUE);
7265 /* make the op for the constant and return */
7268 lvalp->opval = newSVOP(OP_CONST, 0, sv);
7270 lvalp->opval = Nullop;
7276 S_scan_formline(pTHX_ register char *s)
7280 SV *stuff = newSVpvn("",0);
7281 bool needargs = FALSE;
7284 if (*s == '.' || *s == /*{*/'}') {
7286 #ifdef PERL_STRICT_CR
7287 for (t = s+1;SPACE_OR_TAB(*t); t++) ;
7289 for (t = s+1;SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
7291 if (*t == '\n' || t == PL_bufend)
7294 if (PL_in_eval && !PL_rsfp) {
7295 eol = strchr(s,'\n');
7300 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
7302 for (t = s; t < eol; t++) {
7303 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
7305 goto enough; /* ~~ must be first line in formline */
7307 if (*t == '@' || *t == '^')
7310 sv_catpvn(stuff, s, eol-s);
7311 #ifndef PERL_STRICT_CR
7312 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
7313 char *end = SvPVX(stuff) + SvCUR(stuff);
7322 s = filter_gets(PL_linestr, PL_rsfp, 0);
7323 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
7324 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
7327 yyerror("Format not terminated");
7337 PL_lex_state = LEX_NORMAL;
7338 PL_nextval[PL_nexttoke].ival = 0;
7342 PL_lex_state = LEX_FORMLINE;
7343 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST, 0, stuff);
7345 PL_nextval[PL_nexttoke].ival = OP_FORMLINE;
7349 SvREFCNT_dec(stuff);
7350 PL_lex_formbrack = 0;
7361 PL_cshlen = strlen(PL_cshname);
7366 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
7368 I32 oldsavestack_ix = PL_savestack_ix;
7369 CV* outsidecv = PL_compcv;
7373 assert(SvTYPE(PL_compcv) == SVt_PVCV);
7375 SAVEI32(PL_subline);
7376 save_item(PL_subname);
7379 SAVESPTR(PL_comppad_name);
7380 SAVESPTR(PL_compcv);
7381 SAVEI32(PL_comppad_name_fill);
7382 SAVEI32(PL_min_intro_pending);
7383 SAVEI32(PL_max_intro_pending);
7384 SAVEI32(PL_pad_reset_pending);
7386 PL_compcv = (CV*)NEWSV(1104,0);
7387 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
7388 CvFLAGS(PL_compcv) |= flags;
7390 PL_comppad = newAV();
7391 av_push(PL_comppad, Nullsv);
7392 PL_curpad = AvARRAY(PL_comppad);
7393 PL_comppad_name = newAV();
7394 PL_comppad_name_fill = 0;
7395 PL_min_intro_pending = 0;
7397 PL_subline = CopLINE(PL_curcop);
7399 av_store(PL_comppad_name, 0, newSVpvn("@_", 2));
7400 PL_curpad[0] = (SV*)newAV();
7401 SvPADMY_on(PL_curpad[0]); /* XXX Needed? */
7402 #endif /* USE_THREADS */
7404 comppadlist = newAV();
7405 AvREAL_off(comppadlist);
7406 av_store(comppadlist, 0, (SV*)PL_comppad_name);
7407 av_store(comppadlist, 1, (SV*)PL_comppad);
7409 CvPADLIST(PL_compcv) = comppadlist;
7410 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc(outsidecv);
7412 CvOWNER(PL_compcv) = 0;
7413 New(666, CvMUTEXP(PL_compcv), 1, perl_mutex);
7414 MUTEX_INIT(CvMUTEXP(PL_compcv));
7415 #endif /* USE_THREADS */
7417 return oldsavestack_ix;
7421 Perl_yywarn(pTHX_ char *s)
7423 PL_in_eval |= EVAL_WARNONLY;
7425 PL_in_eval &= ~EVAL_WARNONLY;
7430 Perl_yyerror(pTHX_ char *s)
7433 char *context = NULL;
7437 if (!yychar || (yychar == ';' && !PL_rsfp))
7439 else if (PL_bufptr > PL_oldoldbufptr && PL_bufptr - PL_oldoldbufptr < 200 &&
7440 PL_oldoldbufptr != PL_oldbufptr && PL_oldbufptr != PL_bufptr) {
7441 while (isSPACE(*PL_oldoldbufptr))
7443 context = PL_oldoldbufptr;
7444 contlen = PL_bufptr - PL_oldoldbufptr;
7446 else if (PL_bufptr > PL_oldbufptr && PL_bufptr - PL_oldbufptr < 200 &&
7447 PL_oldbufptr != PL_bufptr) {
7448 while (isSPACE(*PL_oldbufptr))
7450 context = PL_oldbufptr;
7451 contlen = PL_bufptr - PL_oldbufptr;
7453 else if (yychar > 255)
7454 where = "next token ???";
7455 #ifdef USE_PURE_BISON
7456 /* GNU Bison sets the value -2 */
7457 else if (yychar == -2) {
7459 else if ((yychar & 127) == 127) {
7461 if (PL_lex_state == LEX_NORMAL ||
7462 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
7463 where = "at end of line";
7464 else if (PL_lex_inpat)
7465 where = "within pattern";
7467 where = "within string";
7470 SV *where_sv = sv_2mortal(newSVpvn("next char ", 10));
7472 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
7473 else if (isPRINT_LC(yychar))
7474 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
7476 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
7477 where = SvPVX(where_sv);
7479 msg = sv_2mortal(newSVpv(s, 0));
7480 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
7481 CopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
7483 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
7485 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
7486 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
7487 Perl_sv_catpvf(aTHX_ msg,
7488 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
7489 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
7492 if (PL_in_eval & EVAL_WARNONLY)
7493 Perl_warn(aTHX_ "%"SVf, msg);
7496 if (PL_error_count >= 10) {
7497 if (PL_in_eval && SvCUR(ERRSV))
7498 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
7499 ERRSV, CopFILE(PL_curcop));
7501 Perl_croak(aTHX_ "%s has too many errors.\n",
7502 CopFILE(PL_curcop));
7505 PL_in_my_stash = Nullhv;
7510 S_swallow_bom(pTHX_ U8 *s)
7513 slen = SvCUR(PL_linestr);
7517 /* UTF-16 little-endian */
7518 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
7519 Perl_croak(aTHX_ "Unsupported script encoding");
7520 #ifndef PERL_NO_UTF16_FILTER
7521 DEBUG_p(PerlIO_printf(Perl_debug_log, "UTF-LE script encoding\n"));
7523 if (PL_bufend > (char*)s) {
7527 filter_add(utf16rev_textfilter, NULL);
7528 New(898, news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
7529 PL_bufend = (char*)utf16_to_utf8_reversed(s, news,
7530 PL_bufend - (char*)s - 1,
7532 Copy(news, s, newlen, U8);
7533 SvCUR_set(PL_linestr, newlen);
7534 PL_bufend = SvPVX(PL_linestr) + newlen;
7535 news[newlen++] = '\0';
7539 Perl_croak(aTHX_ "Unsupported script encoding");
7544 if (s[1] == 0xFF) { /* UTF-16 big-endian */
7545 #ifndef PERL_NO_UTF16_FILTER
7546 DEBUG_p(PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding\n"));
7548 if (PL_bufend > (char *)s) {
7552 filter_add(utf16_textfilter, NULL);
7553 New(898, news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
7554 PL_bufend = (char*)utf16_to_utf8(s, news,
7555 PL_bufend - (char*)s,
7557 Copy(news, s, newlen, U8);
7558 SvCUR_set(PL_linestr, newlen);
7559 PL_bufend = SvPVX(PL_linestr) + newlen;
7560 news[newlen++] = '\0';
7564 Perl_croak(aTHX_ "Unsupported script encoding");
7569 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
7570 DEBUG_p(PerlIO_printf(Perl_debug_log, "UTF-8 script encoding\n"));
7575 if (slen > 3 && s[1] == 0 && /* UTF-32 big-endian */
7576 s[2] == 0xFE && s[3] == 0xFF)
7578 Perl_croak(aTHX_ "Unsupported script encoding");
7590 * Restore a source filter.
7594 restore_rsfp(pTHXo_ void *f)
7596 PerlIO *fp = (PerlIO*)f;
7598 if (PL_rsfp == PerlIO_stdin())
7599 PerlIO_clearerr(PL_rsfp);
7600 else if (PL_rsfp && (PL_rsfp != fp))
7601 PerlIO_close(PL_rsfp);
7605 #ifndef PERL_NO_UTF16_FILTER
7607 utf16_textfilter(pTHXo_ int idx, SV *sv, int maxlen)
7609 I32 count = FILTER_READ(idx+1, sv, maxlen);
7614 New(898, tmps, SvCUR(sv) * 3 / 2 + 1, U8);
7615 if (!*SvPV_nolen(sv))
7616 /* Game over, but don't feed an odd-length string to utf16_to_utf8 */
7619 tend = utf16_to_utf8((U8*)SvPVX(sv), tmps, SvCUR(sv), &newlen);
7620 sv_usepvn(sv, (char*)tmps, tend - tmps);
7626 utf16rev_textfilter(pTHXo_ int idx, SV *sv, int maxlen)
7628 I32 count = FILTER_READ(idx+1, sv, maxlen);
7633 if (!*SvPV_nolen(sv))
7634 /* Game over, but don't feed an odd-length string to utf16_to_utf8 */
7637 New(898, tmps, SvCUR(sv) * 3 / 2 + 1, U8);
7638 tend = utf16_to_utf8_reversed((U8*)SvPVX(sv), tmps, SvCUR(sv), &newlen);
7639 sv_usepvn(sv, (char*)tmps, tend - tmps);